From 5873c713cc47af311f517ea33a6110993a410377 Mon Sep 17 00:00:00 2001 From: Wenchen Fan Date: Thu, 5 Mar 2015 14:49:01 -0800 Subject: [PATCH 001/122] [SPARK-6145][SQL] fix ORDER BY on nested fields Based on #4904 with style errors fixed. `LogicalPlan#resolve` will not only produce `Attribute`, but also "`GetField` chain". So in `ResolveSortReferences`, after resolve the ordering expressions, we should not just collect the `Attribute` results, but also `Attribute` at the bottom of "`GetField` chain". Author: Wenchen Fan Author: Michael Armbrust Closes #4918 from marmbrus/pr/4904 and squashes the following commits: 997f84e [Michael Armbrust] fix style 3eedbfc [Wenchen Fan] fix 6145 --- .../org/apache/spark/sql/catalyst/SqlParser.scala | 2 +- .../apache/spark/sql/catalyst/analysis/Analyzer.scala | 5 +++-- .../scala/org/apache/spark/sql/SQLQuerySuite.scala | 10 ++++++++++ 3 files changed, 14 insertions(+), 3 deletions(-) diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/SqlParser.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/SqlParser.scala index c363a5efacde8..54ab13ca352d2 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/SqlParser.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/SqlParser.scala @@ -385,7 +385,7 @@ class SqlParser extends AbstractSparkSQLParser { protected lazy val dotExpressionHeader: Parser[Expression] = (ident <~ ".") ~ ident ~ rep("." ~> ident) ^^ { - case i1 ~ i2 ~ rest => UnresolvedAttribute(i1 + "." + i2 + rest.mkString(".", ".", "")) + case i1 ~ i2 ~ rest => UnresolvedAttribute((Seq(i1, i2) ++ rest).mkString(".")) } protected lazy val dataType: Parser[DataType] = diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala index e4e542562f22d..7753331748d7b 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala @@ -310,7 +310,7 @@ class Analyzer(catalog: Catalog, } /** - * In many dialects of SQL is it valid to sort by attributes that are not present in the SELECT + * In many dialects of SQL it is valid to sort by attributes that are not present in the SELECT * clause. This rule detects such queries and adds the required attributes to the original * projection, so that they will be available during sorting. Another projection is added to * remove these attributes after sorting. @@ -321,7 +321,8 @@ class Analyzer(catalog: Catalog, if !s.resolved && p.resolved => val unresolved = ordering.flatMap(_.collect { case UnresolvedAttribute(name) => name }) val resolved = unresolved.flatMap(child.resolve(_, resolver)) - val requiredAttributes = AttributeSet(resolved.collect { case a: Attribute => a }) + val requiredAttributes = + AttributeSet(resolved.flatMap(_.collect { case a: Attribute => a })) val missingInProject = requiredAttributes -- p.output if (missingInProject.nonEmpty) { diff --git a/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala index 097bf0dd23c89..4dedcd365f6cc 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala @@ -1049,4 +1049,14 @@ class SQLQuerySuite extends QueryTest with BeforeAndAfterAll { rdd.toDF().registerTempTable("distinctData") checkAnswer(sql("SELECT COUNT(DISTINCT key,value) FROM distinctData"), Row(2)) } + + test("SPARK-6145: ORDER BY test for nested fields") { + jsonRDD(sparkContext.makeRDD( + """{"a": {"b": 1, "a": {"a": 1}}, "c": [{"d": 1}]}""" :: Nil)).registerTempTable("nestedOrder") + // These should be successfully analyzed + sql("SELECT 1 FROM nestedOrder ORDER BY a.b").queryExecution.analyzed + sql("SELECT a.b FROM nestedOrder ORDER BY a.b").queryExecution.analyzed + sql("SELECT 1 FROM nestedOrder ORDER BY a.a.a").queryExecution.analyzed + sql("SELECT 1 FROM nestedOrder ORDER BY c[0].d").queryExecution.analyzed + } } From 1b4bb25c10d72132d7f4f3835ef9a3b94b2349e0 Mon Sep 17 00:00:00 2001 From: Yin Huai Date: Thu, 5 Mar 2015 14:49:44 -0800 Subject: [PATCH 002/122] [SPARK-6163][SQL] jsonFile should be backed by the data source API jira: https://issues.apache.org/jira/browse/SPARK-6163 Author: Yin Huai Closes #4896 from yhuai/SPARK-6163 and squashes the following commits: 45e023e [Yin Huai] Address @chenghao-intel's comment. 2e8734e [Yin Huai] Use JSON data source for jsonFile. 92a4a33 [Yin Huai] Test. --- .../org/apache/spark/sql/SQLContext.scala | 12 +++----- .../org/apache/spark/sql/json/JsonSuite.scala | 28 +++++++++++++++++++ 2 files changed, 32 insertions(+), 8 deletions(-) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala b/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala index ce800e0754559..9c49e84bf9680 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala @@ -542,20 +542,16 @@ class SQLContext(@transient val sparkContext: SparkContext) * @group specificdata */ @Experimental - def jsonFile(path: String, schema: StructType): DataFrame = { - val json = sparkContext.textFile(path) - jsonRDD(json, schema) - } + def jsonFile(path: String, schema: StructType): DataFrame = + load("json", schema, Map("path" -> path)) /** * :: Experimental :: * @group specificdata */ @Experimental - def jsonFile(path: String, samplingRatio: Double): DataFrame = { - val json = sparkContext.textFile(path) - jsonRDD(json, samplingRatio) - } + def jsonFile(path: String, samplingRatio: Double): DataFrame = + load("json", Map("path" -> path, "samplingRatio" -> samplingRatio.toString)) /** * Loads an RDD[String] storing JSON objects (one object per record), returning the result as a diff --git a/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala index 9d94d3406acfb..0c21f725f0b49 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala @@ -19,6 +19,8 @@ package org.apache.spark.sql.json import java.sql.{Date, Timestamp} +import org.scalactic.Tolerance._ + import org.apache.spark.sql.TestData._ import org.apache.spark.sql.catalyst.util._ import org.apache.spark.sql.functions._ @@ -551,6 +553,32 @@ class JsonSuite extends QueryTest { jsonDF.registerTempTable("jsonTable") } + test("jsonFile should be based on JSONRelation") { + val file = getTempFilePath("json") + val path = file.toString + sparkContext.parallelize(1 to 100).map(i => s"""{"a": 1, "b": "str$i"}""").saveAsTextFile(path) + val jsonDF = jsonFile(path, 0.49) + + val analyzed = jsonDF.queryExecution.analyzed + assert( + analyzed.isInstanceOf[LogicalRelation], + "The DataFrame returned by jsonFile should be based on JSONRelation.") + val relation = analyzed.asInstanceOf[LogicalRelation].relation + assert( + relation.isInstanceOf[JSONRelation], + "The DataFrame returned by jsonFile should be based on JSONRelation.") + assert(relation.asInstanceOf[JSONRelation].path === path) + assert(relation.asInstanceOf[JSONRelation].samplingRatio === (0.49 +- 0.001)) + + val schema = StructType(StructField("a", LongType, true) :: Nil) + val logicalRelation = + jsonFile(path, schema).queryExecution.analyzed.asInstanceOf[LogicalRelation] + val relationWithSchema = logicalRelation.relation.asInstanceOf[JSONRelation] + assert(relationWithSchema.path === path) + assert(relationWithSchema.schema === schema) + assert(relationWithSchema.samplingRatio > 0.99) + } + test("Loading a JSON dataset from a text file") { val file = getTempFilePath("json") val path = file.toString From eb48fd6e9d55fb034c00e61374bb9c2a86a82fb8 Mon Sep 17 00:00:00 2001 From: Michael Armbrust Date: Thu, 5 Mar 2015 14:50:25 -0800 Subject: [PATCH 003/122] [SQL] Make Strategies a public developer API Author: Michael Armbrust Closes #4920 from marmbrus/openStrategies and squashes the following commits: cbc35c0 [Michael Armbrust] [SQL] Make Strategies a public developer API --- sql/core/src/main/scala/org/apache/spark/sql/package.scala | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/package.scala b/sql/core/src/main/scala/org/apache/spark/sql/package.scala index 02e5b015e8ec2..3f97a11ceb97d 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/package.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/package.scala @@ -34,10 +34,13 @@ import org.apache.spark.sql.execution.SparkPlan package object sql { /** - * Converts a logical plan into zero or more SparkPlans. + * Converts a logical plan into zero or more SparkPlans. This API is exposed for experimenting + * with the query planner and is not designed to be stable across spark releases. Developers + * writing libraries should instead consider using the stable APIs provided in + * [[org.apache.spark.sql.sources]] */ @DeveloperApi - protected[sql] type Strategy = org.apache.spark.sql.catalyst.planning.GenericStrategy[SparkPlan] + type Strategy = org.apache.spark.sql.catalyst.planning.GenericStrategy[SparkPlan] /** * Type alias for [[DataFrame]]. Kept here for backward source compatibility for Scala. From d8b3da9ddfe44a2886f3841ceef4ebf9fc318640 Mon Sep 17 00:00:00 2001 From: "Zhang, Liye" Date: Fri, 6 Mar 2015 09:34:07 +0000 Subject: [PATCH 004/122] [CORE, DEPLOY][minor] align arguments order with docs of worker The help message for starting `worker` is `Usage: Worker [options] `. While in `start-slaves.sh`, the format is not align with that, it is confusing for the fist glance. Author: Zhang, Liye Closes #4924 from liyezhang556520/startSlaves and squashes the following commits: 7fd5deb [Zhang, Liye] align arguments order with docs of worker --- sbin/start-slaves.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/sbin/start-slaves.sh b/sbin/start-slaves.sh index ba1a84abc1fef..76316a3067c93 100755 --- a/sbin/start-slaves.sh +++ b/sbin/start-slaves.sh @@ -64,6 +64,6 @@ else SPARK_WORKER_WEBUI_PORT=8081 fi for ((i=0; i<$SPARK_WORKER_INSTANCES; i++)); do - "$sbin/slaves.sh" cd "$SPARK_HOME" \; "$sbin/start-slave.sh" $(( $i + 1 )) "spark://$SPARK_MASTER_IP:$SPARK_MASTER_PORT" --webui-port $(( $SPARK_WORKER_WEBUI_PORT + $i )) + "$sbin/slaves.sh" cd "$SPARK_HOME" \; "$sbin/start-slave.sh" $(( $i + 1 )) --webui-port $(( $SPARK_WORKER_WEBUI_PORT + $i )) "spark://$SPARK_MASTER_IP:$SPARK_MASTER_PORT" done fi From cd7594ca6acf1226bf91f8a783606bf5c116f7df Mon Sep 17 00:00:00 2001 From: Marcelo Vanzin Date: Fri, 6 Mar 2015 09:43:24 +0000 Subject: [PATCH 005/122] [core] [minor] Don't pollute source directory when running UtilsSuite. Author: Marcelo Vanzin Closes #4921 from vanzin/utils-suite and squashes the following commits: 7795dd4 [Marcelo Vanzin] [core] [minor] Don't pollute source directory when running UtilsSuite. --- core/src/test/scala/org/apache/spark/util/UtilsSuite.scala | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/core/src/test/scala/org/apache/spark/util/UtilsSuite.scala b/core/src/test/scala/org/apache/spark/util/UtilsSuite.scala index fd77753c0d362..b91428efadfd0 100644 --- a/core/src/test/scala/org/apache/spark/util/UtilsSuite.scala +++ b/core/src/test/scala/org/apache/spark/util/UtilsSuite.scala @@ -386,10 +386,11 @@ class UtilsSuite extends FunSuite with ResetSystemProperties { } test("fetch hcfs dir") { - val sourceDir = Utils.createTempDir() + val tempDir = Utils.createTempDir() + val sourceDir = new File(tempDir, "source-dir") val innerSourceDir = Utils.createTempDir(root=sourceDir.getPath) val sourceFile = File.createTempFile("someprefix", "somesuffix", innerSourceDir) - val targetDir = new File(Utils.createTempDir(), "target-dir") + val targetDir = new File(tempDir, "target-dir") Files.write("some text", sourceFile, UTF_8) val path = new Path("file://" + sourceDir.getAbsolutePath) @@ -413,7 +414,7 @@ class UtilsSuite extends FunSuite with ResetSystemProperties { assert(destInnerFile.isFile()) val filePath = new Path("file://" + sourceFile.getAbsolutePath) - val testFileDir = new File("test-filename") + val testFileDir = new File(tempDir, "test-filename") val testFileName = "testFName" val testFilefs = Utils.getHadoopFileSystem(filePath.toString, conf) Utils.fetchHcfsFile(filePath, testFileDir, testFilefs, new SparkConf(), From 05cb6b34d8fc25114f3dd3e2bd156386c00eb391 Mon Sep 17 00:00:00 2001 From: GuoQiang Li Date: Fri, 6 Mar 2015 13:20:20 +0000 Subject: [PATCH 006/122] [Minor] Resolve sbt warnings: postfix operator second should be enabled Resolve sbt warnings: ``` [warn] spark/streaming/src/main/scala/org/apache/spark/streaming/util/WriteAheadLogManager.scala:155: postfix operator second should be enabled [warn] by making the implicit value scala.language.postfixOps visible. [warn] This can be achieved by adding the import clause 'import scala.language.postfixOps' [warn] or by setting the compiler option -language:postfixOps. [warn] See the Scala docs for value scala.language.postfixOps for a discussion [warn] why the feature should be explicitly enabled. [warn] Await.ready(f, 1 second) [warn] ^ ``` Author: GuoQiang Li Closes #4908 from witgo/sbt_warnings and squashes the following commits: 0629af4 [GuoQiang Li] Resolve sbt warnings: postfix operator second should be enabled --- .../scala/org/apache/spark/scheduler/local/LocalBackend.scala | 1 + .../org/apache/spark/streaming/util/WriteAheadLogManager.scala | 1 + 2 files changed, 2 insertions(+) diff --git a/core/src/main/scala/org/apache/spark/scheduler/local/LocalBackend.scala b/core/src/main/scala/org/apache/spark/scheduler/local/LocalBackend.scala index 4676b828d3d89..d95426d918e19 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/local/LocalBackend.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/local/LocalBackend.scala @@ -20,6 +20,7 @@ package org.apache.spark.scheduler.local import java.nio.ByteBuffer import scala.concurrent.duration._ +import scala.language.postfixOps import akka.actor.{Actor, ActorRef, Props} diff --git a/streaming/src/main/scala/org/apache/spark/streaming/util/WriteAheadLogManager.scala b/streaming/src/main/scala/org/apache/spark/streaming/util/WriteAheadLogManager.scala index 985ded9111f74..6bdfe45dc7f83 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/util/WriteAheadLogManager.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/util/WriteAheadLogManager.scala @@ -20,6 +20,7 @@ import java.nio.ByteBuffer import scala.collection.mutable.ArrayBuffer import scala.concurrent.{Await, ExecutionContext, Future} +import scala.language.postfixOps import org.apache.hadoop.conf.Configuration import org.apache.hadoop.fs.Path From dba0b2eadb441f41ded0f0b9706b720bcfa84881 Mon Sep 17 00:00:00 2001 From: Vinod K C Date: Fri, 6 Mar 2015 14:43:09 +0000 Subject: [PATCH 007/122] [SPARK-6178][Shuffle] Removed unused imports Author: Vinod K C Author: Vinod K C Closes #4900 from vinodkc/unused_imports and squashes the following commits: 5373456 [Vinod K C] Removed empty lines 9da7438 [Vinod K C] Changed order of import 594d471 [Vinod K C] Removed unused imports --- .../org/apache/spark/network/TransportContext.java | 1 - .../spark/network/protocol/ChunkFetchFailure.java | 1 - .../org/apache/spark/network/protocol/Encoders.java | 1 - .../org/apache/spark/network/protocol/RpcFailure.java | 1 - .../apache/spark/network/server/TransportServer.java | 1 - .../java/org/apache/spark/network/util/JavaUtils.java | 10 ++-------- .../java/org/apache/spark/network/util/NettyUtils.java | 1 - .../org/apache/spark/network/sasl/SaslRpcHandler.java | 3 --- .../spark/network/shuffle/OneForOneBlockFetcher.java | 1 - .../spark/network/shuffle/protocol/OpenBlocks.java | 1 - .../network/shuffle/protocol/RegisterExecutor.java | 1 - .../spark/network/shuffle/protocol/StreamHandle.java | 2 -- .../spark/network/shuffle/protocol/UploadBlock.java | 1 - .../org/apache/spark/network/sasl/SparkSaslSuite.java | 6 +++--- .../network/shuffle/OneForOneBlockFetcherSuite.java | 9 +++++++-- 15 files changed, 12 insertions(+), 28 deletions(-) diff --git a/network/common/src/main/java/org/apache/spark/network/TransportContext.java b/network/common/src/main/java/org/apache/spark/network/TransportContext.java index 5bc6e5a2418a9..f0a89c9d9116c 100644 --- a/network/common/src/main/java/org/apache/spark/network/TransportContext.java +++ b/network/common/src/main/java/org/apache/spark/network/TransportContext.java @@ -35,7 +35,6 @@ import org.apache.spark.network.server.TransportChannelHandler; import org.apache.spark.network.server.TransportRequestHandler; import org.apache.spark.network.server.TransportServer; -import org.apache.spark.network.server.StreamManager; import org.apache.spark.network.util.NettyUtils; import org.apache.spark.network.util.TransportConf; diff --git a/network/common/src/main/java/org/apache/spark/network/protocol/ChunkFetchFailure.java b/network/common/src/main/java/org/apache/spark/network/protocol/ChunkFetchFailure.java index 986957c1509fd..f76bb49e874fc 100644 --- a/network/common/src/main/java/org/apache/spark/network/protocol/ChunkFetchFailure.java +++ b/network/common/src/main/java/org/apache/spark/network/protocol/ChunkFetchFailure.java @@ -17,7 +17,6 @@ package org.apache.spark.network.protocol; -import com.google.common.base.Charsets; import com.google.common.base.Objects; import io.netty.buffer.ByteBuf; diff --git a/network/common/src/main/java/org/apache/spark/network/protocol/Encoders.java b/network/common/src/main/java/org/apache/spark/network/protocol/Encoders.java index 873c694250942..9162d0b977f83 100644 --- a/network/common/src/main/java/org/apache/spark/network/protocol/Encoders.java +++ b/network/common/src/main/java/org/apache/spark/network/protocol/Encoders.java @@ -20,7 +20,6 @@ import com.google.common.base.Charsets; import io.netty.buffer.ByteBuf; -import io.netty.buffer.Unpooled; /** Provides a canonical set of Encoders for simple types. */ public class Encoders { diff --git a/network/common/src/main/java/org/apache/spark/network/protocol/RpcFailure.java b/network/common/src/main/java/org/apache/spark/network/protocol/RpcFailure.java index ebd764eb5eb5f..6b991375fc486 100644 --- a/network/common/src/main/java/org/apache/spark/network/protocol/RpcFailure.java +++ b/network/common/src/main/java/org/apache/spark/network/protocol/RpcFailure.java @@ -17,7 +17,6 @@ package org.apache.spark.network.protocol; -import com.google.common.base.Charsets; import com.google.common.base.Objects; import io.netty.buffer.ByteBuf; diff --git a/network/common/src/main/java/org/apache/spark/network/server/TransportServer.java b/network/common/src/main/java/org/apache/spark/network/server/TransportServer.java index ef209991804b4..b7ce8541e565e 100644 --- a/network/common/src/main/java/org/apache/spark/network/server/TransportServer.java +++ b/network/common/src/main/java/org/apache/spark/network/server/TransportServer.java @@ -28,7 +28,6 @@ import io.netty.channel.ChannelOption; import io.netty.channel.EventLoopGroup; import io.netty.channel.socket.SocketChannel; -import io.netty.util.internal.PlatformDependent; import org.slf4j.Logger; import org.slf4j.LoggerFactory; diff --git a/network/common/src/main/java/org/apache/spark/network/util/JavaUtils.java b/network/common/src/main/java/org/apache/spark/network/util/JavaUtils.java index bf8a1fc42fc6d..73da9b7346f4d 100644 --- a/network/common/src/main/java/org/apache/spark/network/util/JavaUtils.java +++ b/network/common/src/main/java/org/apache/spark/network/util/JavaUtils.java @@ -17,19 +17,13 @@ package org.apache.spark.network.util; -import java.nio.ByteBuffer; - -import java.io.ByteArrayInputStream; -import java.io.ByteArrayOutputStream; import java.io.Closeable; import java.io.File; import java.io.IOException; -import java.io.ObjectInputStream; -import java.io.ObjectOutputStream; +import java.nio.ByteBuffer; -import com.google.common.base.Preconditions; -import com.google.common.io.Closeables; import com.google.common.base.Charsets; +import com.google.common.base.Preconditions; import io.netty.buffer.Unpooled; import org.slf4j.Logger; import org.slf4j.LoggerFactory; diff --git a/network/common/src/main/java/org/apache/spark/network/util/NettyUtils.java b/network/common/src/main/java/org/apache/spark/network/util/NettyUtils.java index 2a4b88b64cdc9..dabd6261d2aa0 100644 --- a/network/common/src/main/java/org/apache/spark/network/util/NettyUtils.java +++ b/network/common/src/main/java/org/apache/spark/network/util/NettyUtils.java @@ -25,7 +25,6 @@ import io.netty.channel.Channel; import io.netty.channel.EventLoopGroup; import io.netty.channel.ServerChannel; -import io.netty.channel.epoll.Epoll; import io.netty.channel.epoll.EpollEventLoopGroup; import io.netty.channel.epoll.EpollServerSocketChannel; import io.netty.channel.epoll.EpollSocketChannel; diff --git a/network/shuffle/src/main/java/org/apache/spark/network/sasl/SaslRpcHandler.java b/network/shuffle/src/main/java/org/apache/spark/network/sasl/SaslRpcHandler.java index 3777a18e33f78..026cbd260d16c 100644 --- a/network/shuffle/src/main/java/org/apache/spark/network/sasl/SaslRpcHandler.java +++ b/network/shuffle/src/main/java/org/apache/spark/network/sasl/SaslRpcHandler.java @@ -19,16 +19,13 @@ import java.util.concurrent.ConcurrentMap; -import com.google.common.base.Charsets; import com.google.common.collect.Maps; -import io.netty.buffer.ByteBuf; import io.netty.buffer.Unpooled; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import org.apache.spark.network.client.RpcResponseCallback; import org.apache.spark.network.client.TransportClient; -import org.apache.spark.network.protocol.Encodable; import org.apache.spark.network.server.RpcHandler; import org.apache.spark.network.server.StreamManager; diff --git a/network/shuffle/src/main/java/org/apache/spark/network/shuffle/OneForOneBlockFetcher.java b/network/shuffle/src/main/java/org/apache/spark/network/shuffle/OneForOneBlockFetcher.java index 8ed2e0b39ad23..e653f5cb147ee 100644 --- a/network/shuffle/src/main/java/org/apache/spark/network/shuffle/OneForOneBlockFetcher.java +++ b/network/shuffle/src/main/java/org/apache/spark/network/shuffle/OneForOneBlockFetcher.java @@ -29,7 +29,6 @@ import org.apache.spark.network.shuffle.protocol.BlockTransferMessage; import org.apache.spark.network.shuffle.protocol.OpenBlocks; import org.apache.spark.network.shuffle.protocol.StreamHandle; -import org.apache.spark.network.util.JavaUtils; /** * Simple wrapper on top of a TransportClient which interprets each chunk as a whole block, and diff --git a/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/OpenBlocks.java b/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/OpenBlocks.java index 62fce9b0d16cd..60485bace643c 100644 --- a/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/OpenBlocks.java +++ b/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/OpenBlocks.java @@ -23,7 +23,6 @@ import io.netty.buffer.ByteBuf; import org.apache.spark.network.protocol.Encoders; -import org.apache.spark.network.shuffle.protocol.BlockTransferMessage.Type; /** Request to read a set of blocks. Returns {@link StreamHandle}. */ public class OpenBlocks extends BlockTransferMessage { diff --git a/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/RegisterExecutor.java b/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/RegisterExecutor.java index 7eb4385044077..38acae3b31d64 100644 --- a/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/RegisterExecutor.java +++ b/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/RegisterExecutor.java @@ -21,7 +21,6 @@ import io.netty.buffer.ByteBuf; import org.apache.spark.network.protocol.Encoders; -import org.apache.spark.network.shuffle.protocol.BlockTransferMessage.Type; /** * Initial registration message between an executor and its local shuffle server. diff --git a/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/StreamHandle.java b/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/StreamHandle.java index bc9daa6158ba3..9a9220211a50c 100644 --- a/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/StreamHandle.java +++ b/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/StreamHandle.java @@ -20,8 +20,6 @@ import com.google.common.base.Objects; import io.netty.buffer.ByteBuf; -import org.apache.spark.network.shuffle.protocol.BlockTransferMessage.Type; - /** * Identifier for a fixed number of chunks to read from a stream created by an "open blocks" * message. This is used by {@link org.apache.spark.network.shuffle.OneForOneBlockFetcher}. diff --git a/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/UploadBlock.java b/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/UploadBlock.java index 0b23e112bd512..2ff9aaa650f92 100644 --- a/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/UploadBlock.java +++ b/network/shuffle/src/main/java/org/apache/spark/network/shuffle/protocol/UploadBlock.java @@ -23,7 +23,6 @@ import io.netty.buffer.ByteBuf; import org.apache.spark.network.protocol.Encoders; -import org.apache.spark.network.shuffle.protocol.BlockTransferMessage.Type; /** Request to upload a block with a certain StorageLevel. Returns nothing (empty byte array). */ diff --git a/network/shuffle/src/test/java/org/apache/spark/network/sasl/SparkSaslSuite.java b/network/shuffle/src/test/java/org/apache/spark/network/sasl/SparkSaslSuite.java index 67a07f38eb5a0..23b4e06f064e1 100644 --- a/network/shuffle/src/test/java/org/apache/spark/network/sasl/SparkSaslSuite.java +++ b/network/shuffle/src/test/java/org/apache/spark/network/sasl/SparkSaslSuite.java @@ -17,12 +17,12 @@ package org.apache.spark.network.sasl; -import java.util.Map; +import static org.junit.Assert.assertFalse; +import static org.junit.Assert.assertTrue; +import static org.junit.Assert.fail; -import com.google.common.collect.ImmutableMap; import org.junit.Test; -import static org.junit.Assert.*; /** * Jointly tests SparkSaslClient and SparkSaslServer, as both are black boxes. diff --git a/network/shuffle/src/test/java/org/apache/spark/network/shuffle/OneForOneBlockFetcherSuite.java b/network/shuffle/src/test/java/org/apache/spark/network/shuffle/OneForOneBlockFetcherSuite.java index 842741e3d354f..b35a6d685dd02 100644 --- a/network/shuffle/src/test/java/org/apache/spark/network/shuffle/OneForOneBlockFetcherSuite.java +++ b/network/shuffle/src/test/java/org/apache/spark/network/shuffle/OneForOneBlockFetcherSuite.java @@ -28,11 +28,16 @@ import org.mockito.invocation.InvocationOnMock; import org.mockito.stubbing.Answer; -import static org.junit.Assert.*; import static org.junit.Assert.assertEquals; +import static org.junit.Assert.fail; import static org.mockito.Matchers.any; +import static org.mockito.Matchers.anyInt; +import static org.mockito.Matchers.anyLong; import static org.mockito.Matchers.eq; -import static org.mockito.Mockito.*; +import static org.mockito.Mockito.doAnswer; +import static org.mockito.Mockito.mock; +import static org.mockito.Mockito.times; +import static org.mockito.Mockito.verify; import org.apache.spark.network.buffer.ManagedBuffer; import org.apache.spark.network.buffer.NettyManagedBuffer; From 48a723c98684c5bb3d185cada4888cae952791bd Mon Sep 17 00:00:00 2001 From: RobertZK Date: Sat, 7 Mar 2015 00:16:50 +0000 Subject: [PATCH 008/122] Fix python typo (+ Scala, Java typos) Author: RobertZK Author: Robert Krzyzanowski Closes #4840 from robertzk/patch-1 and squashes the following commits: d286215 [RobertZK] lambda fix per @laserson 5937989 [Robert Krzyzanowski] Fix python typo --- docs/programming-guide.md | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-) diff --git a/docs/programming-guide.md b/docs/programming-guide.md index 7b0701828878e..b5e04bd0c610d 100644 --- a/docs/programming-guide.md +++ b/docs/programming-guide.md @@ -1336,25 +1336,28 @@ Accumulators do not change the lazy evaluation model of Spark. If they are being
{% highlight scala %} -val acc = sc.accumulator(0) -data.map(x => acc += x; f(x)) -// Here, acc is still 0 because no actions have cause the `map` to be computed. +val accum = sc.accumulator(0) +data.map { x => accum += x; f(x) } +// Here, accum is still 0 because no actions have caused the `map` to be computed. {% endhighlight %}
{% highlight java %} Accumulator accum = sc.accumulator(0); -data.map(x -> accum.add(x); f(x);); -// Here, accum is still 0 because no actions have cause the `map` to be computed. +data.map(x -> { accum.add(x); return f(x); }); +// Here, accum is still 0 because no actions have caused the `map` to be computed. {% endhighlight %}
{% highlight python %} accum = sc.accumulator(0) -data.map(lambda x => acc.add(x); f(x)) -# Here, acc is still 0 because no actions have cause the `map` to be computed. +def g(x): + accum.add(x) + return f(x) +data.map(g) +# Here, accum is still 0 because no actions have caused the `map` to be computed. {% endhighlight %}
From 2646794ffb2970618087e2e964d9f4c953e17e6a Mon Sep 17 00:00:00 2001 From: Nicholas Chammas Date: Sat, 7 Mar 2015 12:33:41 +0000 Subject: [PATCH 009/122] [EC2] Reorder print statements on termination The PR reorders some print statements slightly on cluster termination so that they read better. For example, from this: ``` Are you sure you want to destroy the cluster spark-cluster-test? The following instances will be terminated: Searching for existing cluster spark-cluster-test in region us-west-2... Found 1 master(s), 2 slaves > ... ALL DATA ON ALL NODES WILL BE LOST!! Destroy cluster spark-cluster-test (y/N): ``` To this: ``` Searching for existing cluster spark-cluster-test in region us-west-2... Found 1 master(s), 2 slaves The following instances will be terminated: > ... ALL DATA ON ALL NODES WILL BE LOST!! Are you sure you want to destroy the cluster spark-cluster-test? (y/N) ``` Author: Nicholas Chammas Closes #4932 from nchammas/termination-print-order and squashes the following commits: c23711d [Nicholas Chammas] reorder prints on termination --- ec2/spark_ec2.py | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) diff --git a/ec2/spark_ec2.py b/ec2/spark_ec2.py index c59ab565c6862..dabb9fce40d01 100755 --- a/ec2/spark_ec2.py +++ b/ec2/spark_ec2.py @@ -1126,14 +1126,16 @@ def real_main(): setup_cluster(conn, master_nodes, slave_nodes, opts, True) elif action == "destroy": - print "Are you sure you want to destroy the cluster %s?" % cluster_name - print "The following instances will be terminated:" (master_nodes, slave_nodes) = get_existing_cluster( conn, opts, cluster_name, die_on_error=False) - for inst in master_nodes + slave_nodes: - print "> %s" % inst.public_dns_name - msg = "ALL DATA ON ALL NODES WILL BE LOST!!\nDestroy cluster %s (y/N): " % cluster_name + if any(master_nodes + slave_nodes): + print "The following instances will be terminated:" + for inst in master_nodes + slave_nodes: + print "> %s" % inst.public_dns_name + print "ALL DATA ON ALL NODES WILL BE LOST!!" + + msg = "Are you sure you want to destroy the cluster {c}? (y/N) ".format(c=cluster_name) response = raw_input(msg) if response == "y": print "Terminating master..." @@ -1145,7 +1147,6 @@ def real_main(): # Delete security groups as well if opts.delete_groups: - print "Deleting security groups (this will take some time)..." group_names = [cluster_name + "-master", cluster_name + "-slaves"] wait_for_cluster_state( conn=conn, @@ -1153,6 +1154,7 @@ def real_main(): cluster_instances=(master_nodes + slave_nodes), cluster_state='terminated' ) + print "Deleting security groups (this will take some time)..." attempt = 1 while attempt <= 3: print "Attempt %d" % attempt From 729c05bda87c2383d1c54b31850ed10814617cde Mon Sep 17 00:00:00 2001 From: WangTaoTheTonic Date: Sat, 7 Mar 2015 12:35:26 +0000 Subject: [PATCH 010/122] [Minor]fix the wrong description Found it by accident. I'm not gonna file jira for this as it is a very tiny fix. Author: WangTaoTheTonic Closes #4936 from WangTaoTheTonic/wrongdesc and squashes the following commits: fb8a8ec [WangTaoTheTonic] fix the wrong description aca5596 [WangTaoTheTonic] fix the wrong description --- sbin/stop-all.sh | 4 ++-- sbin/stop-master.sh | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/sbin/stop-all.sh b/sbin/stop-all.sh index 971d5d49da664..1a9abe07db844 100755 --- a/sbin/stop-all.sh +++ b/sbin/stop-all.sh @@ -17,8 +17,8 @@ # limitations under the License. # -# Start all spark daemons. -# Run this on the master nde +# Stop all spark daemons. +# Run this on the master node. sbin="`dirname "$0"`" diff --git a/sbin/stop-master.sh b/sbin/stop-master.sh index b6bdaa4db373c..729702d92191e 100755 --- a/sbin/stop-master.sh +++ b/sbin/stop-master.sh @@ -17,7 +17,7 @@ # limitations under the License. # -# Starts the master on the machine this script is executed on. +# Stops the master on the machine this script is executed on. sbin=`dirname "$0"` sbin=`cd "$sbin"; pwd` From 334c5bd1ae50ac76770e545cab302361673f45de Mon Sep 17 00:00:00 2001 From: Florian Verhein Date: Sat, 7 Mar 2015 12:56:59 +0000 Subject: [PATCH 011/122] [SPARK-5641] [EC2] Allow spark_ec2.py to copy arbitrary files to cluster Give users an easy way to rcp a directory structure to the master's / as part of the cluster launch, at a useful point in the workflow (before setup.sh is called on the master). This is an alternative approach to meeting requirements discussed in https://github.com/apache/spark/pull/4487 Author: Florian Verhein Closes #4583 from florianverhein/master and squashes the following commits: 49dee88 [Florian Verhein] removed addition of trailing / in rsync to give user this option, added documentation in help 7b8e3d8 [Florian Verhein] remove unused args 87d922c [Florian Verhein] [SPARK-5641] [EC2] implement --deploy-root-dir --- ec2/spark_ec2.py | 42 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 42 insertions(+) diff --git a/ec2/spark_ec2.py b/ec2/spark_ec2.py index dabb9fce40d01..b6e7c4c2af39b 100755 --- a/ec2/spark_ec2.py +++ b/ec2/spark_ec2.py @@ -159,6 +159,15 @@ def parse_args(): "--spark-ec2-git-branch", default=DEFAULT_SPARK_EC2_BRANCH, help="Github repo branch of spark-ec2 to use (default: %default)") + parser.add_option( + "--deploy-root-dir", + default=None, + help="A directory to copy into / on the first master. " + + "Must be absolute. Note that a trailing slash is handled as per rsync: " + + "If you omit it, the last directory of the --deploy-root-dir path will be created " + + "in / before copying its contents. If you append the trailing slash, " + + "the directory is not created and its contents are copied directly into /. " + + "(default: %default).") parser.add_option( "--hadoop-major-version", default="1", help="Major version of Hadoop (default: %default)") @@ -694,6 +703,14 @@ def setup_cluster(conn, master_nodes, slave_nodes, opts, deploy_ssh_key): modules=modules ) + if opts.deploy_root_dir is not None: + print "Deploying {s} to master...".format(s=opts.deploy_root_dir) + deploy_user_files( + root_dir=opts.deploy_root_dir, + opts=opts, + master_nodes=master_nodes + ) + print "Running setup on master..." setup_spark_cluster(master, opts) print "Done!" @@ -931,6 +948,23 @@ def deploy_files(conn, root_dir, opts, master_nodes, slave_nodes, modules): shutil.rmtree(tmp_dir) +# Deploy a given local directory to a cluster, WITHOUT parameter substitution. +# Note that unlike deploy_files, this works for binary files. +# Also, it is up to the user to add (or not) the trailing slash in root_dir. +# Files are only deployed to the first master instance in the cluster. +# +# root_dir should be an absolute path. +def deploy_user_files(root_dir, opts, master_nodes): + active_master = master_nodes[0].public_dns_name + command = [ + 'rsync', '-rv', + '-e', stringify_command(ssh_command(opts)), + "%s" % root_dir, + "%s@%s:/" % (opts.user, active_master) + ] + subprocess.check_call(command) + + def stringify_command(parts): if isinstance(parts, str): return parts @@ -1099,6 +1133,14 @@ def real_main(): "Furthermore, we currently only support forks named spark-ec2." sys.exit(1) + if not (opts.deploy_root_dir is None or + (os.path.isabs(opts.deploy_root_dir) and + os.path.isdir(opts.deploy_root_dir) and + os.path.exists(opts.deploy_root_dir))): + print >> stderr, "--deploy-root-dir must be an absolute path to a directory that exists " \ + "on the local file system" + sys.exit(1) + try: conn = ec2.connect_to_region(opts.region) except Exception as e: From 52ed7da12e26c45734ce53a1be19ef148b2b953e Mon Sep 17 00:00:00 2001 From: Nicholas Chammas Date: Sun, 8 Mar 2015 14:01:26 +0000 Subject: [PATCH 012/122] [SPARK-6193] [EC2] Push group filter up to EC2 When looking for a cluster, spark-ec2 currently pulls down [info for all instances](https://github.com/apache/spark/blob/eb48fd6e9d55fb034c00e61374bb9c2a86a82fb8/ec2/spark_ec2.py#L620) and filters locally. When working on an AWS account with hundreds of active instances, this step alone can take over 10 seconds. This PR improves how spark-ec2 searches for clusters by pushing the filter up to EC2. Basically, the problem (and solution) look like this: ```python >>> timeit.timeit('blah = conn.get_all_reservations()', setup='from __main__ import conn', number=10) 116.96390509605408 >>> timeit.timeit('blah = conn.get_all_reservations(filters={"instance.group-name": ["my-cluster-master"]})', setup='from __main__ import conn', number=10) 4.629754066467285 ``` Translated to a user-visible action, this looks like (against an AWS account with ~200 active instances): ```shell # master $ python -m timeit -n 3 --setup 'import subprocess' 'subprocess.call("./spark-ec2 get-master my-cluster --region us-west-2", shell=True)' ... 3 loops, best of 3: 9.83 sec per loop # this PR $ python -m timeit -n 3 --setup 'import subprocess' 'subprocess.call("./spark-ec2 get-master my-cluster --region us-west-2", shell=True)' ... 3 loops, best of 3: 1.47 sec per loop ``` This PR also refactors `get_existing_cluster()` to make it, I hope, simpler. Finally, this PR fixes some minor grammar issues related to printing status to the user. :tophat: :clap: Author: Nicholas Chammas Closes #4922 from nchammas/get-existing-cluster-faster and squashes the following commits: 18802f1 [Nicholas Chammas] ignore shutting-down f2a5b9f [Nicholas Chammas] fix grammar d96a489 [Nicholas Chammas] push group filter up to EC2 --- ec2/spark_ec2.py | 78 +++++++++++++++++++++++++----------------------- 1 file changed, 41 insertions(+), 37 deletions(-) diff --git a/ec2/spark_ec2.py b/ec2/spark_ec2.py index b6e7c4c2af39b..5e636ddd17e94 100755 --- a/ec2/spark_ec2.py +++ b/ec2/spark_ec2.py @@ -22,6 +22,7 @@ from __future__ import with_statement import hashlib +import itertools import logging import os import os.path @@ -299,13 +300,6 @@ def get_validate_spark_version(version, repo): return version -# Check whether a given EC2 instance object is in a state we consider active, -# i.e. not terminating or terminated. We count both stopping and stopped as -# active since we can restart stopped clusters. -def is_active(instance): - return (instance.state in ['pending', 'running', 'stopping', 'stopped']) - - # Source: http://aws.amazon.com/amazon-linux-ami/instance-type-matrix/ # Last Updated: 2014-06-20 # For easy maintainability, please keep this manually-inputted dictionary sorted by key. @@ -573,8 +567,11 @@ def launch_cluster(conn, opts, cluster_name): placement_group=opts.placement_group, user_data=user_data_content) slave_nodes += slave_res.instances - print "Launched %d slaves in %s, regid = %s" % (num_slaves_this_zone, - zone, slave_res.id) + print "Launched {s} slave{plural_s} in {z}, regid = {r}".format( + s=num_slaves_this_zone, + plural_s=('' if num_slaves_this_zone == 1 else 's'), + z=zone, + r=slave_res.id) i += 1 # Launch or resume masters @@ -621,40 +618,47 @@ def launch_cluster(conn, opts, cluster_name): return (master_nodes, slave_nodes) -# Get the EC2 instances in an existing cluster if available. -# Returns a tuple of lists of EC2 instance objects for the masters and slaves def get_existing_cluster(conn, opts, cluster_name, die_on_error=True): - print "Searching for existing cluster " + cluster_name + " in region " \ - + opts.region + "..." - reservations = conn.get_all_reservations() - master_nodes = [] - slave_nodes = [] - for res in reservations: - active = [i for i in res.instances if is_active(i)] - for inst in active: - group_names = [g.name for g in inst.groups] - if (cluster_name + "-master") in group_names: - master_nodes.append(inst) - elif (cluster_name + "-slaves") in group_names: - slave_nodes.append(inst) - if any((master_nodes, slave_nodes)): - print "Found %d master(s), %d slaves" % (len(master_nodes), len(slave_nodes)) - if master_nodes != [] or not die_on_error: - return (master_nodes, slave_nodes) - else: - if master_nodes == [] and slave_nodes != []: - print >> sys.stderr, "ERROR: Could not find master in group " + cluster_name \ - + "-master" + " in region " + opts.region - else: - print >> sys.stderr, "ERROR: Could not find any existing cluster" \ - + " in region " + opts.region + """ + Get the EC2 instances in an existing cluster if available. + Returns a tuple of lists of EC2 instance objects for the masters and slaves. + """ + print "Searching for existing cluster {c} in region {r}...".format( + c=cluster_name, r=opts.region) + + def get_instances(group_names): + """ + Get all non-terminated instances that belong to any of the provided security groups. + + EC2 reservation filters and instance states are documented here: + http://docs.aws.amazon.com/cli/latest/reference/ec2/describe-instances.html#options + """ + reservations = conn.get_all_reservations( + filters={"instance.group-name": group_names}) + instances = itertools.chain.from_iterable(r.instances for r in reservations) + return [i for i in instances if i.state not in ["shutting-down", "terminated"]] + + master_instances = get_instances([cluster_name + "-master"]) + slave_instances = get_instances([cluster_name + "-slaves"]) + + if any((master_instances, slave_instances)): + print "Found {m} master{plural_m}, {s} slave{plural_s}.".format( + m=len(master_instances), + plural_m=('' if len(master_instances) == 1 else 's'), + s=len(slave_instances), + plural_s=('' if len(slave_instances) == 1 else 's')) + + if not master_instances and die_on_error: + print >> sys.stderr, \ + "ERROR: Could not find a master for cluster {c} in region {r}.".format( + c=cluster_name, r=opts.region) sys.exit(1) + return (master_instances, slave_instances) + # Deploy configuration files and run setup scripts on a newly launched # or started EC2 cluster. - - def setup_cluster(conn, master_nodes, slave_nodes, opts, deploy_ssh_key): master = master_nodes[0].public_dns_name if deploy_ssh_key: From f16b7b031feeb13ec9c17608bd99566f56431869 Mon Sep 17 00:00:00 2001 From: Sean Owen Date: Sun, 8 Mar 2015 14:09:40 +0000 Subject: [PATCH 013/122] SPARK-6205 [CORE] UISeleniumSuite fails for Hadoop 2.x test with NoClassDefFoundError Add xml-apis to core test deps to work aroudn UISeleniumSuite classpath issue Author: Sean Owen Closes #4933 from srowen/SPARK-6205 and squashes the following commits: ddd4d32 [Sean Owen] Add xml-apis to core test deps to work aroudn UISeleniumSuite classpath issue --- core/pom.xml | 6 ++++++ pom.xml | 7 +++++++ 2 files changed, 13 insertions(+) diff --git a/core/pom.xml b/core/pom.xml index fab776d142ef7..dc0d07d806635 100644 --- a/core/pom.xml +++ b/core/pom.xml @@ -319,6 +319,12 @@ selenium-java test + + + xml-apis + xml-apis + test + org.mockito mockito-all diff --git a/pom.xml b/pom.xml index f99a83b9994ed..51bef30f9ca8f 100644 --- a/pom.xml +++ b/pom.xml @@ -422,6 +422,13 @@ 2.42.2 test + + + xml-apis + xml-apis + 1.4.01 + test + org.slf4j slf4j-api From 55b1b32dc8b9b25deea8e5864b53fe802bb92741 Mon Sep 17 00:00:00 2001 From: Jacky Li Date: Sun, 8 Mar 2015 19:47:35 +0000 Subject: [PATCH 014/122] [GraphX] Improve LiveJournalPageRank example 1. Removed unnecessary import 2. Modified usage print since user must specify the --numEPart parameter as it is required in Analytics.main Author: Jacky Li Closes #4917 from jackylk/import and squashes the following commits: 6c07682 [Jacky Li] fix comment c0df8f2 [Jacky Li] fix scalastyle b6235e6 [Jacky Li] fix for comment 87be83b [Jacky Li] remove default value description 5caae76 [Jacky Li] remove import and modify usage --- .../spark/examples/graphx/LiveJournalPageRank.scala | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-) diff --git a/examples/src/main/scala/org/apache/spark/examples/graphx/LiveJournalPageRank.scala b/examples/src/main/scala/org/apache/spark/examples/graphx/LiveJournalPageRank.scala index e809a65b79975..f6f8d9f90c275 100644 --- a/examples/src/main/scala/org/apache/spark/examples/graphx/LiveJournalPageRank.scala +++ b/examples/src/main/scala/org/apache/spark/examples/graphx/LiveJournalPageRank.scala @@ -17,11 +17,6 @@ package org.apache.spark.examples.graphx -import org.apache.spark.SparkContext._ -import org.apache.spark._ -import org.apache.spark.graphx._ - - /** * Uses GraphX to run PageRank on a LiveJournal social network graph. Download the dataset from * http://snap.stanford.edu/data/soc-LiveJournal1.html. @@ -31,13 +26,13 @@ object LiveJournalPageRank { if (args.length < 1) { System.err.println( "Usage: LiveJournalPageRank \n" + + " --numEPart=\n" + + " The number of partitions for the graph's edge RDD.\n" + " [--tol=]\n" + " The tolerance allowed at convergence (smaller => more accurate). Default is " + "0.001.\n" + " [--output=]\n" + " If specified, the file to write the ranks to.\n" + - " [--numEPart=]\n" + - " The number of partitions for the graph's edge RDD. Default is 4.\n" + " [--partStrategy=RandomVertexCut | EdgePartition1D | EdgePartition2D | " + "CanonicalRandomVertexCut]\n" + " The way edges are assigned to edge partitions. Default is RandomVertexCut.") From f7c799204358bcc38c5972a29e5994b78b25b515 Mon Sep 17 00:00:00 2001 From: Theodore Vasiloudis Date: Mon, 9 Mar 2015 14:16:07 +0000 Subject: [PATCH 015/122] [EC2] [SPARK-6188] Instance types can be mislabeled when re-starting cluster with default arguments As described in https://issues.apache.org/jira/browse/SPARK-6188 and discovered in https://issues.apache.org/jira/browse/SPARK-5838. When re-starting a cluster, if the user does not provide the instance types, which is the recommended behavior in the docs currently, the instance will be assigned the default type m1.large. This then affects the setup of the machines. This solves this by getting the instance types from the existing instances, and overwriting the default options. EDIT: Further clarification of the issue: In short, while the instances themselves are the same as launched, their setup is done assuming the default instance type, m1.large. This means that the machines are assumed to have 2 disks, and that leads to problems that are described in in issue [5838](https://issues.apache.org/jira/browse/SPARK-5838), where machines that have one disk end up having shuffle spills in the in the small (8GB) snapshot partitions that quickly fills up and results in failing jobs due to "No space left on device" errors. Other instance specific settings that are set in the spark_ec2.py script are likely to be wrong as well. Author: Theodore Vasiloudis Author: Theodore Vasiloudis Closes #4916 from thvasilo/SPARK-6188]-Instance-types-can-be-mislabeled-when-re-starting-cluster-with-default-arguments and squashes the following commits: 6705b98 [Theodore Vasiloudis] Added comment to clarify setting master instance type to the empty string. a3d29fe [Theodore Vasiloudis] More trailing whitespace 7b32429 [Theodore Vasiloudis] Removed trailing whitespace 3ebd52a [Theodore Vasiloudis] Make sure that the instance type is correct when relaunching a cluster. --- ec2/spark_ec2.py | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/ec2/spark_ec2.py b/ec2/spark_ec2.py index 5e636ddd17e94..b50b3816ff890 100755 --- a/ec2/spark_ec2.py +++ b/ec2/spark_ec2.py @@ -1307,6 +1307,17 @@ def real_main(): cluster_instances=(master_nodes + slave_nodes), cluster_state='ssh-ready' ) + + # Determine types of running instances + existing_master_type = master_nodes[0].instance_type + existing_slave_type = slave_nodes[0].instance_type + # Setting opts.master_instance_type to the empty string indicates we + # have the same instance type for the master and the slaves + if existing_master_type == existing_slave_type: + existing_master_type = "" + opts.master_instance_type = existing_master_type + opts.instance_type = existing_slave_type + setup_cluster(conn, master_nodes, slave_nodes, opts, False) else: From 70f88148bb04161a1a4968230d8e3fc7e3f8321a Mon Sep 17 00:00:00 2001 From: Reynold Xin Date: Mon, 9 Mar 2015 13:29:19 -0700 Subject: [PATCH 016/122] [Docs] Replace references to SchemaRDD with DataFrame Author: Reynold Xin Closes #4952 from rxin/schemardd-df-reference and squashes the following commits: b2b1dbe [Reynold Xin] [Docs] Replace references to SchemaRDD with DataFrame --- python/pyspark/ml/pipeline.py | 4 ++-- python/pyspark/ml/wrapper.py | 2 +- .../src/test/scala/org/apache/spark/repl/ReplSuite.scala | 6 +++--- 3 files changed, 6 insertions(+), 6 deletions(-) diff --git a/python/pyspark/ml/pipeline.py b/python/pyspark/ml/pipeline.py index 5233c5801e2e6..83880a5afcd1d 100644 --- a/python/pyspark/ml/pipeline.py +++ b/python/pyspark/ml/pipeline.py @@ -39,7 +39,7 @@ def fit(self, dataset, params={}): Fits a model to the input dataset with optional parameters. :param dataset: input dataset, which is an instance of - :py:class:`pyspark.sql.SchemaRDD` + :py:class:`pyspark.sql.DataFrame` :param params: an optional param map that overwrites embedded params :returns: fitted model @@ -62,7 +62,7 @@ def transform(self, dataset, params={}): Transforms the input dataset with optional parameters. :param dataset: input dataset, which is an instance of - :py:class:`pyspark.sql.SchemaRDD` + :py:class:`pyspark.sql.DataFrame` :param params: an optional param map that overwrites embedded params :returns: transformed dataset diff --git a/python/pyspark/ml/wrapper.py b/python/pyspark/ml/wrapper.py index 4bae96f678388..31a66b3d2f730 100644 --- a/python/pyspark/ml/wrapper.py +++ b/python/pyspark/ml/wrapper.py @@ -102,7 +102,7 @@ def _fit_java(self, dataset, params={}): """ Fits a Java model to the input dataset. :param dataset: input dataset, which is an instance of - :py:class:`pyspark.sql.SchemaRDD` + :py:class:`pyspark.sql.DataFrame` :param params: additional params (overwriting embedded values) :return: fitted Java model """ diff --git a/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala b/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala index f966f25c5a14c..ed9b207a86a0b 100644 --- a/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala +++ b/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala @@ -263,14 +263,14 @@ class ReplSuite extends FunSuite { assertDoesNotContain("Exception", output) } - test("SPARK-2576 importing SQLContext.createSchemaRDD.") { + test("SPARK-2576 importing SQLContext.createDataFrame.") { // We need to use local-cluster to test this case. val output = runInterpreter("local-cluster[1,1,512]", """ |val sqlContext = new org.apache.spark.sql.SQLContext(sc) - |import sqlContext.createSchemaRDD + |import sqlContext.implicits._ |case class TestCaseClass(value: Int) - |sc.parallelize(1 to 10).map(x => TestCaseClass(x)).toSchemaRDD.collect + |sc.parallelize(1 to 10).map(x => TestCaseClass(x)).toDF.collect """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) From 3cac1991a1def0adaf42face2c578d3ab8c27025 Mon Sep 17 00:00:00 2001 From: Reynold Xin Date: Mon, 9 Mar 2015 16:16:16 -0700 Subject: [PATCH 017/122] [SPARK-5310][Doc] Update SQL Programming Guide to include DataFrames. Author: Reynold Xin Closes #4954 from rxin/df-docs and squashes the following commits: c592c70 [Reynold Xin] [SPARK-5310][Doc] Update SQL Programming Guide to include DataFrames. --- docs/_layouts/global.html | 2 +- docs/index.md | 2 +- docs/sql-programming-guide.md | 404 ++++++++++++++++++++++++---------- 3 files changed, 286 insertions(+), 122 deletions(-) diff --git a/docs/_layouts/global.html b/docs/_layouts/global.html index efc4c612937df..2e88b3093652d 100755 --- a/docs/_layouts/global.html +++ b/docs/_layouts/global.html @@ -71,7 +71,7 @@
  • Spark Programming Guide
  • Spark Streaming
    • -
  • Spark SQL
    • +
  • DataFrames and SQL
  • MLlib (Machine Learning)
  • GraphX (Graph Processing)
  • Bagel (Pregel on Spark)
    • diff --git a/docs/index.md b/docs/index.md index 0986398e6f744..b5b016e34795e 100644 --- a/docs/index.md +++ b/docs/index.md @@ -74,7 +74,7 @@ options for deployment: in all supported languages (Scala, Java, Python) * Modules built on Spark: * [Spark Streaming](streaming-programming-guide.html): processing real-time data streams - * [Spark SQL](sql-programming-guide.html): support for structured data and relational queries + * [Spark SQL and DataFrames](sql-programming-guide.html): support for structured data and relational queries * [MLlib](mllib-guide.html): built-in machine learning library * [GraphX](graphx-programming-guide.html): Spark's new API for graph processing * [Bagel (Pregel on Spark)](bagel-programming-guide.html): older, simple graph processing model diff --git a/docs/sql-programming-guide.md b/docs/sql-programming-guide.md index 0146a4ed1b745..4fbdca7397951 100644 --- a/docs/sql-programming-guide.md +++ b/docs/sql-programming-guide.md @@ -1,7 +1,7 @@ --- layout: global -displayTitle: Spark SQL Programming Guide -title: Spark SQL +displayTitle: Spark SQL and DataFrame Guide +title: Spark SQL and DataFrames --- * This will become a table of contents (this text will be scraped). @@ -9,55 +9,24 @@ title: Spark SQL # Overview -
    -
    - -Spark SQL allows relational queries expressed in SQL, HiveQL, or Scala to be executed using -Spark. At the core of this component is a new type of RDD, -[DataFrame](api/scala/index.html#org.apache.spark.sql.DataFrame). DataFrames are composed of -[Row](api/scala/index.html#org.apache.spark.sql.package@Row:org.apache.spark.sql.catalyst.expressions.Row.type) objects, along with -a schema that describes the data types of each column in the row. A DataFrame is similar to a table -in a traditional relational database. A DataFrame can be created from an existing RDD, a [Parquet](http://parquet.io) -file, a JSON dataset, or by running HiveQL against data stored in [Apache Hive](http://hive.apache.org/). - -All of the examples on this page use sample data included in the Spark distribution and can be run in the `spark-shell`. - -
    +Spark SQL is a Spark module for structured data processing. It provides a programming abstraction called DataFrames and can also act as a distributed query engine. -
    -Spark SQL allows relational queries expressed in SQL or HiveQL to be executed using -Spark. At the core of this component is a new type of RDD, -[DataFrame](api/scala/index.html#org.apache.spark.sql.DataFrame). DataFrames are composed of -[Row](api/scala/index.html#org.apache.spark.sql.api.java.Row) objects, along with -a schema that describes the data types of each column in the row. A DataFrame is similar to a table -in a traditional relational database. A DataFrame can be created from an existing RDD, a [Parquet](http://parquet.io) -file, a JSON dataset, or by running HiveQL against data stored in [Apache Hive](http://hive.apache.org/). -
    -
    +# DataFrames -Spark SQL allows relational queries expressed in SQL or HiveQL to be executed using -Spark. At the core of this component is a new type of RDD, -[DataFrame](api/python/pyspark.sql.html#pyspark.sql.DataFrame). DataFrames are composed of -[Row](api/python/pyspark.sql.Row-class.html) objects, along with -a schema that describes the data types of each column in the row. A DataFrame is similar to a table -in a traditional relational database. A DataFrame can be created from an existing RDD, a [Parquet](http://parquet.io) -file, a JSON dataset, or by running HiveQL against data stored in [Apache Hive](http://hive.apache.org/). +A DataFrame is a distributed collection of data organized into named columns. It is conceptually equivalent to a table in a relational database or a data frame in R/Python, but with richer optimizations under the hood. DataFrames can be constructed from a wide array of sources such as: structured data files, tables in Hive, external databases, or existing RDDs. -All of the examples on this page use sample data included in the Spark distribution and can be run in the `pyspark` shell. -
    -
    +The DataFrame API is available in [Scala](api/scala/index.html#org.apache.spark.sql.DataFrame), [Java](api/java/index.html?org/apache/spark/sql/DataFrame.html), and [Python](api/python/pyspark.sql.html#pyspark.sql.DataFrame). -**Spark SQL is currently an alpha component. While we will minimize API changes, some APIs may change in future releases.** +All of the examples on this page use sample data included in the Spark distribution and can be run in the `spark-shell` or the `pyspark` shell. -*************************************************************************************************** -# Getting Started +## Starting Point: SQLContext
    -The entry point into all relational functionality in Spark is the +The entry point into all functionality in Spark SQL is the [SQLContext](api/scala/index.html#org.apache.spark.sql.SQLContext) class, or one of its descendants. To create a basic SQLContext, all you need is a SparkContext. @@ -69,39 +38,19 @@ val sqlContext = new org.apache.spark.sql.SQLContext(sc) import sqlContext.implicits._ {% endhighlight %} -In addition to the basic SQLContext, you can also create a HiveContext, which provides a -superset of the functionality provided by the basic SQLContext. Additional features include -the ability to write queries using the more complete HiveQL parser, access to HiveUDFs, and the -ability to read data from Hive tables. To use a HiveContext, you do not need to have an -existing Hive setup, and all of the data sources available to a SQLContext are still available. -HiveContext is only packaged separately to avoid including all of Hive's dependencies in the default -Spark build. If these dependencies are not a problem for your application then using HiveContext -is recommended for the 1.2 release of Spark. Future releases will focus on bringing SQLContext up to -feature parity with a HiveContext. -
    -The entry point into all relational functionality in Spark is the -[SQLContext](api/scala/index.html#org.apache.spark.sql.api.SQLContext) class, or one -of its descendants. To create a basic SQLContext, all you need is a JavaSparkContext. +The entry point into all functionality in Spark SQL is the +[SQLContext](api/java/index.html#org.apache.spark.sql.SQLContext) class, or one of its +descendants. To create a basic SQLContext, all you need is a SparkContext. {% highlight java %} JavaSparkContext sc = ...; // An existing JavaSparkContext. SQLContext sqlContext = new org.apache.spark.sql.SQLContext(sc); {% endhighlight %} -In addition to the basic SQLContext, you can also create a HiveContext, which provides a strict -super set of the functionality provided by the basic SQLContext. Additional features include -the ability to write queries using the more complete HiveQL parser, access to HiveUDFs, and the -ability to read data from Hive tables. To use a HiveContext, you do not need to have an -existing Hive setup, and all of the data sources available to a SQLContext are still available. -HiveContext is only packaged separately to avoid including all of Hive's dependencies in the default -Spark build. If these dependencies are not a problem for your application then using HiveContext -is recommended for the 1.2 release of Spark. Future releases will focus on bringing SQLContext up to -feature parity with a HiveContext. -
    @@ -115,35 +64,266 @@ from pyspark.sql import SQLContext sqlContext = SQLContext(sc) {% endhighlight %} -In addition to the basic SQLContext, you can also create a HiveContext, which provides a strict -super set of the functionality provided by the basic SQLContext. Additional features include -the ability to write queries using the more complete HiveQL parser, access to HiveUDFs, and the +
    +
    + +In addition to the basic SQLContext, you can also create a HiveContext, which provides a +superset of the functionality provided by the basic SQLContext. Additional features include +the ability to write queries using the more complete HiveQL parser, access to Hive UDFs, and the ability to read data from Hive tables. To use a HiveContext, you do not need to have an existing Hive setup, and all of the data sources available to a SQLContext are still available. HiveContext is only packaged separately to avoid including all of Hive's dependencies in the default Spark build. If these dependencies are not a problem for your application then using HiveContext -is recommended for the 1.2 release of Spark. Future releases will focus on bringing SQLContext up to -feature parity with a HiveContext. - - - - +is recommended for the 1.3 release of Spark. Future releases will focus on bringing SQLContext up +to feature parity with a HiveContext. The specific variant of SQL that is used to parse queries can also be selected using the `spark.sql.dialect` option. This parameter can be changed using either the `setConf` method on a SQLContext or by using a `SET key=value` command in SQL. For a SQLContext, the only dialect available is "sql" which uses a simple SQL parser provided by Spark SQL. In a HiveContext, the default is "hiveql", though "sql" is also available. Since the HiveQL parser is much more complete, - this is recommended for most use cases. +this is recommended for most use cases. -# Data Sources -Spark SQL supports operating on a variety of data sources through the `DataFrame` interface. -A DataFrame can be operated on as normal RDDs and can also be registered as a temporary table. -Registering a DataFrame as a table allows you to run SQL queries over its data. This section -describes the various methods for loading data into a DataFrame. +## Creating DataFrames + +With a `SQLContext`, applications can create `DataFrame`s from an existing `RDD`, from a Hive table, or from data sources. + +As an example, the following creates a `DataFrame` based on the content of a JSON file: + +
    +
    +{% highlight scala %} +val sc: SparkContext // An existing SparkContext. +val sqlContext = new org.apache.spark.sql.SQLContext(sc) + +val df = sqlContext.jsonFile("examples/src/main/resources/people.json") + +// Displays the content of the DataFrame to stdout +df.show() +{% endhighlight %} + +
    + +
    +{% highlight java %} +JavaSparkContext sc = ...; // An existing JavaSparkContext. +SQLContext sqlContext = new org.apache.spark.sql.SQLContext(sc); + +DataFrame df = sqlContext.jsonFile("examples/src/main/resources/people.json"); + +// Displays the content of the DataFrame to stdout +df.show(); +{% endhighlight %} + +
    + +
    +{% highlight python %} +from pyspark.sql import SQLContext +sqlContext = SQLContext(sc) + +df = sqlContext.jsonFile("examples/src/main/resources/people.json") + +# Displays the content of the DataFrame to stdout +df.show() +{% endhighlight %} + +
    +
    + + +## DataFrame Operations + +DataFrames provide a domain-specific language for structured data manipulation in [Scala](api/scala/index.html#org.apache.spark.sql.DataFrame), [Java](api/java/index.html?org/apache/spark/sql/DataFrame.html), and [Python](api/python/pyspark.sql.html#pyspark.sql.DataFrame). + +Here we include some basic examples of structured data processing using DataFrames: + + +
    +
    +{% highlight scala %} +val sc: SparkContext // An existing SparkContext. +val sqlContext = new org.apache.spark.sql.SQLContext(sc) + +// Create the DataFrame +val df = sqlContext.jsonFile("examples/src/main/resources/people.json") + +// Show the content of the DataFrame +df.show() +// age name +// null Michael +// 30 Andy +// 19 Justin + +// Print the schema in a tree format +df.printSchema() +// root +// |-- age: long (nullable = true) +// |-- name: string (nullable = true) + +// Select only the "name" column +df.select("name").show() +// name +// Michael +// Andy +// Justin + +// Select everybody, but increment the age by 1 +df.select("name", df("age") + 1).show() +// name (age + 1) +// Michael null +// Andy 31 +// Justin 20 + +// Select people older than 21 +df.filter(df("name") > 21).show() +// age name +// 30 Andy + +// Count people by age +df.groupBy("age").count().show() +// age count +// null 1 +// 19 1 +// 30 1 +{% endhighlight %} + +
    + +
    +{% highlight java %} +val sc: JavaSparkContext // An existing SparkContext. +val sqlContext = new org.apache.spark.sql.SQLContext(sc) + +// Create the DataFrame +DataFrame df = sqlContext.jsonFile("examples/src/main/resources/people.json"); + +// Show the content of the DataFrame +df.show(); +// age name +// null Michael +// 30 Andy +// 19 Justin + +// Print the schema in a tree format +df.printSchema(); +// root +// |-- age: long (nullable = true) +// |-- name: string (nullable = true) + +// Select only the "name" column +df.select("name").show(); +// name +// Michael +// Andy +// Justin + +// Select everybody, but increment the age by 1 +df.select("name", df.col("age").plus(1)).show(); +// name (age + 1) +// Michael null +// Andy 31 +// Justin 20 + +// Select people older than 21 +df.filter(df("name") > 21).show(); +// age name +// 30 Andy + +// Count people by age +df.groupBy("age").count().show(); +// age count +// null 1 +// 19 1 +// 30 1 +{% endhighlight %} + +
    + +
    +{% highlight python %} +from pyspark.sql import SQLContext +sqlContext = SQLContext(sc) + +# Create the DataFrame +df = sqlContext.jsonFile("examples/src/main/resources/people.json") + +# Show the content of the DataFrame +df.show() +## age name +## null Michael +## 30 Andy +## 19 Justin + +# Print the schema in a tree format +df.printSchema() +## root +## |-- age: long (nullable = true) +## |-- name: string (nullable = true) + +# Select only the "name" column +df.select("name").show() +## name +## Michael +## Andy +## Justin + +# Select everybody, but increment the age by 1 +df.select("name", df.age + 1).show() +## name (age + 1) +## Michael null +## Andy 31 +## Justin 20 + +# Select people older than 21 +df.filter(df.name > 21).show() +## age name +## 30 Andy + +# Count people by age +df.groupBy("age").count().show() +## age count +## null 1 +## 19 1 +## 30 1 + +{% endhighlight %} + +
    +
    + + +## Running SQL Queries Programmatically + +The `sql` function on a `SQLContext` enables applications to run SQL queries programmatically and returns the result as a `DataFrame`. + +
    +
    +{% highlight scala %} +val sqlContext = ... // An existing SQLContext +val df = sqlContext.sql("SELECT * FROM table") +{% endhighlight %} +
    + +
    +{% highlight java %} +val sqlContext = ... // An existing SQLContext +val df = sqlContext.sql("SELECT * FROM table") +{% endhighlight %} +
    + +
    +{% highlight python %} +from pyspark.sql import SQLContext +sqlContext = SQLContext(sc) +df = sqlContext.sql("SELECT * FROM table") +{% endhighlight %} +
    +
    + -## RDDs +## Interoperating with RDDs Spark SQL supports two different methods for converting existing RDDs into DataFrames. The first method uses reflection to infer the schema of an RDD that contains specific types of objects. This @@ -373,12 +553,12 @@ by `SQLContext`. For example: {% highlight java %} // Import factory methods provided by DataType. -import org.apache.spark.sql.api.java.DataType +import org.apache.spark.sql.types.DataType; // Import StructType and StructField -import org.apache.spark.sql.api.java.StructType -import org.apache.spark.sql.api.java.StructField +import org.apache.spark.sql.types.StructType; +import org.apache.spark.sql.types.StructField; // Import Row. -import org.apache.spark.sql.api.java.Row +import org.apache.spark.sql.Row; // sc is an existing JavaSparkContext. SQLContext sqlContext = new org.apache.spark.sql.SQLContext(sc); @@ -472,11 +652,19 @@ for name in names.collect(): print name {% endhighlight %} - + +# Data Sources + +Spark SQL supports operating on a variety of data sources through the `DataFrame` interface. +A DataFrame can be operated on as normal RDDs and can also be registered as a temporary table. +Registering a DataFrame as a table allows you to run SQL queries over its data. This section +describes the various methods for loading data into a DataFrame. + + ## Parquet Files [Parquet](http://parquet.io) is a columnar format that is supported by many other data processing systems. @@ -904,15 +1092,14 @@ that these options will be deprecated in future release as more optimizations ar -# Other SQL Interfaces +# Distributed Query Engine -Spark SQL also supports interfaces for running SQL queries directly without the need to write any -code. +Spark SQL can also act as a distributed query engine using its JDBC/ODBC or command-line interface. In this mode, end-users or applications can interact with Spark SQL directly to run SQL queries, without the need to write any code. ## Running the Thrift JDBC/ODBC server The Thrift JDBC/ODBC server implemented here corresponds to the [`HiveServer2`](https://cwiki.apache.org/confluence/display/Hive/Setting+Up+HiveServer2) -in Hive 0.12. You can test the JDBC server with the beeline script that comes with either Spark or Hive 0.12. +in Hive 0.13. You can test the JDBC server with the beeline script that comes with either Spark or Hive 0.13. To start the JDBC/ODBC server, run the following in the Spark directory: @@ -982,7 +1169,7 @@ Configuration of Hive is done by placing your `hive-site.xml` file in `conf/`. You may run `./bin/spark-sql --help` for a complete list of all available options. -# Compatibility with Other Systems +# Migration Guide ## Migration Guide for Shark User @@ -1139,33 +1326,10 @@ releases of Spark SQL. Hive can optionally merge the small files into fewer large files to avoid overflowing the HDFS metadata. Spark SQL does not support that. -# Writing Language-Integrated Relational Queries - -**Language-Integrated queries are experimental and currently only supported in Scala.** - -Spark SQL also supports a domain specific language for writing queries. Once again, -using the data from the above examples: - -{% highlight scala %} -// sc is an existing SparkContext. -val sqlContext = new org.apache.spark.sql.SQLContext(sc) -// Importing the SQL context gives access to all the public SQL functions and implicit conversions. -import sqlContext._ -val people: RDD[Person] = ... // An RDD of case class objects, from the first example. - -// The following is the same as 'SELECT name FROM people WHERE age >= 10 AND age <= 19' -val teenagers = people.where('age >= 10).where('age <= 19).select('name) -teenagers.map(t => "Name: " + t(0)).collect().foreach(println) -{% endhighlight %} - -The DSL uses Scala symbols to represent columns in the underlying table, which are identifiers -prefixed with a tick (`'`). Implicit conversions turn these symbols into expressions that are -evaluated by the SQL execution engine. A full list of the functions supported can be found in the -[ScalaDoc](api/scala/index.html#org.apache.spark.sql.DataFrame). - +# Data Types -# Spark SQL DataType Reference +Spark SQL and DataFrames support the following data types: * Numeric types - `ByteType`: Represents 1-byte signed integer numbers. @@ -1208,10 +1372,10 @@ evaluated by the SQL execution engine. A full list of the functions supported c
    -All data types of Spark SQL are located in the package `org.apache.spark.sql`. +All data types of Spark SQL are located in the package `org.apache.spark.sql.types`. You can access them by doing {% highlight scala %} -import org.apache.spark.sql._ +import org.apache.spark.sql.types._ {% endhighlight %} @@ -1263,7 +1427,7 @@ import org.apache.spark.sql._ - + @@ -1457,7 +1621,7 @@ please use factory methods provided in - +
    DecimalType scala.math.BigDecimal java.math.BigDecimal DecimalType
    StructType org.apache.spark.sql.api.java.Row org.apache.spark.sql.Row DataTypes.createStructType(fields)
    Note: fields is a List or an array of StructFields. @@ -1478,10 +1642,10 @@ please use factory methods provided in
    -All data types of Spark SQL are located in the package of `pyspark.sql`. +All data types of Spark SQL are located in the package of `pyspark.sql.types`. You can access them by doing {% highlight python %} -from pyspark.sql import * +from pyspark.sql.types import * {% endhighlight %} From 8767565cef01d847f57b7293d8b63b2422009b90 Mon Sep 17 00:00:00 2001 From: Davies Liu Date: Mon, 9 Mar 2015 16:24:06 -0700 Subject: [PATCH 018/122] [SPARK-6194] [SPARK-677] [PySpark] fix memory leak in collect() Because circular reference between JavaObject and JavaMember, an Java object can not be released until Python GC kick in, then it will cause memory leak in collect(), which may consume lots of memory in JVM. This PR change the way we sending collected data back into Python from local file to socket, which could avoid any disk IO during collect, also avoid any referrers of Java object in Python. cc JoshRosen Author: Davies Liu Closes #4923 from davies/fix_collect and squashes the following commits: d730286 [Davies Liu] address comments 24c92a4 [Davies Liu] fix style ba54614 [Davies Liu] use socket to transfer data from JVM 9517c8f [Davies Liu] fix memory leak in collect() --- .../apache/spark/api/python/PythonRDD.scala | 76 ++++++++++++++----- python/pyspark/context.py | 13 ++-- python/pyspark/rdd.py | 30 ++++---- python/pyspark/sql/dataframe.py | 14 +--- 4 files changed, 82 insertions(+), 51 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala b/core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala index b1cec0f6472b0..8d4a53b4ca9b0 100644 --- a/core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala +++ b/core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala @@ -19,26 +19,27 @@ package org.apache.spark.api.python import java.io._ import java.net._ -import java.util.{List => JList, ArrayList => JArrayList, Map => JMap, UUID, Collections} - -import org.apache.spark.input.PortableDataStream +import java.util.{Collections, ArrayList => JArrayList, List => JList, Map => JMap} import scala.collection.JavaConversions._ import scala.collection.mutable import scala.language.existentials import com.google.common.base.Charsets.UTF_8 - import org.apache.hadoop.conf.Configuration import org.apache.hadoop.io.compress.CompressionCodec -import org.apache.hadoop.mapred.{InputFormat, OutputFormat, JobConf} +import org.apache.hadoop.mapred.{InputFormat, JobConf, OutputFormat} import org.apache.hadoop.mapreduce.{InputFormat => NewInputFormat, OutputFormat => NewOutputFormat} + import org.apache.spark._ -import org.apache.spark.api.java.{JavaSparkContext, JavaPairRDD, JavaRDD} +import org.apache.spark.api.java.{JavaPairRDD, JavaRDD, JavaSparkContext} import org.apache.spark.broadcast.Broadcast +import org.apache.spark.input.PortableDataStream import org.apache.spark.rdd.RDD import org.apache.spark.util.Utils +import scala.util.control.NonFatal + private[spark] class PythonRDD( @transient parent: RDD[_], command: Array[Byte], @@ -341,21 +342,33 @@ private[spark] object PythonRDD extends Logging { /** * Adapter for calling SparkContext#runJob from Python. * - * This method will return an iterator of an array that contains all elements in the RDD + * This method will serve an iterator of an array that contains all elements in the RDD * (effectively a collect()), but allows you to run on a certain subset of partitions, * or to enable local execution. + * + * @return the port number of a local socket which serves the data collected from this job. */ def runJob( sc: SparkContext, rdd: JavaRDD[Array[Byte]], partitions: JArrayList[Int], - allowLocal: Boolean): Iterator[Array[Byte]] = { + allowLocal: Boolean): Int = { type ByteArray = Array[Byte] type UnrolledPartition = Array[ByteArray] val allPartitions: Array[UnrolledPartition] = sc.runJob(rdd, (x: Iterator[ByteArray]) => x.toArray, partitions, allowLocal) val flattenedPartition: UnrolledPartition = Array.concat(allPartitions: _*) - flattenedPartition.iterator + serveIterator(flattenedPartition.iterator, + s"serve RDD ${rdd.id} with partitions ${partitions.mkString(",")}") + } + + /** + * A helper function to collect an RDD as an iterator, then serve it via socket. + * + * @return the port number of a local socket which serves the data collected from this job. + */ + def collectAndServe[T](rdd: RDD[T]): Int = { + serveIterator(rdd.collect().iterator, s"serve RDD ${rdd.id}") } def readRDDFromFile(sc: JavaSparkContext, filename: String, parallelism: Int): @@ -575,15 +588,44 @@ private[spark] object PythonRDD extends Logging { dataOut.write(bytes) } - def writeToFile[T](items: java.util.Iterator[T], filename: String) { - import scala.collection.JavaConverters._ - writeToFile(items.asScala, filename) - } + /** + * Create a socket server and a background thread to serve the data in `items`, + * + * The socket server can only accept one connection, or close if no connection + * in 3 seconds. + * + * Once a connection comes in, it tries to serialize all the data in `items` + * and send them into this connection. + * + * The thread will terminate after all the data are sent or any exceptions happen. + */ + private def serveIterator[T](items: Iterator[T], threadName: String): Int = { + val serverSocket = new ServerSocket(0, 1) + serverSocket.setReuseAddress(true) + // Close the socket if no connection in 3 seconds + serverSocket.setSoTimeout(3000) + + new Thread(threadName) { + setDaemon(true) + override def run() { + try { + val sock = serverSocket.accept() + val out = new DataOutputStream(new BufferedOutputStream(sock.getOutputStream)) + try { + writeIteratorToStream(items, out) + } finally { + out.close() + } + } catch { + case NonFatal(e) => + logError(s"Error while sending iterator", e) + } finally { + serverSocket.close() + } + } + }.start() - def writeToFile[T](items: Iterator[T], filename: String) { - val file = new DataOutputStream(new FileOutputStream(filename)) - writeIteratorToStream(items, file) - file.close() + serverSocket.getLocalPort } private def getMergedConf(confAsMap: java.util.HashMap[String, String], diff --git a/python/pyspark/context.py b/python/pyspark/context.py index 6011caf9f1c5a..78dccc40470e3 100644 --- a/python/pyspark/context.py +++ b/python/pyspark/context.py @@ -21,6 +21,8 @@ from threading import Lock from tempfile import NamedTemporaryFile +from py4j.java_collections import ListConverter + from pyspark import accumulators from pyspark.accumulators import Accumulator from pyspark.broadcast import Broadcast @@ -30,13 +32,11 @@ from pyspark.serializers import PickleSerializer, BatchedSerializer, UTF8Deserializer, \ PairDeserializer, AutoBatchedSerializer, NoOpSerializer from pyspark.storagelevel import StorageLevel -from pyspark.rdd import RDD +from pyspark.rdd import RDD, _load_from_socket from pyspark.traceback_utils import CallSite, first_spark_call from pyspark.status import StatusTracker from pyspark.profiler import ProfilerCollector, BasicProfiler -from py4j.java_collections import ListConverter - __all__ = ['SparkContext'] @@ -59,7 +59,6 @@ class SparkContext(object): _gateway = None _jvm = None - _writeToFile = None _next_accum_id = 0 _active_spark_context = None _lock = Lock() @@ -221,7 +220,6 @@ def _ensure_initialized(cls, instance=None, gateway=None): if not SparkContext._gateway: SparkContext._gateway = gateway or launch_gateway() SparkContext._jvm = SparkContext._gateway.jvm - SparkContext._writeToFile = SparkContext._jvm.PythonRDD.writeToFile if instance: if (SparkContext._active_spark_context and @@ -840,8 +838,9 @@ def runJob(self, rdd, partitionFunc, partitions=None, allowLocal=False): # by runJob() in order to avoid having to pass a Python lambda into # SparkContext#runJob. mappedRDD = rdd.mapPartitions(partitionFunc) - it = self._jvm.PythonRDD.runJob(self._jsc.sc(), mappedRDD._jrdd, javaPartitions, allowLocal) - return list(mappedRDD._collect_iterator_through_file(it)) + port = self._jvm.PythonRDD.runJob(self._jsc.sc(), mappedRDD._jrdd, javaPartitions, + allowLocal) + return list(_load_from_socket(port, mappedRDD._jrdd_deserializer)) def show_profiles(self): """ Print the profile stats to stdout """ diff --git a/python/pyspark/rdd.py b/python/pyspark/rdd.py index cb12fed98c53d..bf17f513c0bc3 100644 --- a/python/pyspark/rdd.py +++ b/python/pyspark/rdd.py @@ -19,7 +19,6 @@ from collections import defaultdict from itertools import chain, ifilter, imap import operator -import os import sys import shlex from subprocess import Popen, PIPE @@ -29,6 +28,7 @@ import heapq import bisect import random +import socket from math import sqrt, log, isinf, isnan, pow, ceil from pyspark.serializers import NoOpSerializer, CartesianDeserializer, \ @@ -111,6 +111,17 @@ def _parse_memory(s): return int(float(s[:-1]) * units[s[-1].lower()]) +def _load_from_socket(port, serializer): + sock = socket.socket() + try: + sock.connect(("localhost", port)) + rf = sock.makefile("rb", 65536) + for item in serializer.load_stream(rf): + yield item + finally: + sock.close() + + class Partitioner(object): def __init__(self, numPartitions, partitionFunc): self.numPartitions = numPartitions @@ -698,21 +709,8 @@ def collect(self): Return a list that contains all of the elements in this RDD. """ with SCCallSiteSync(self.context) as css: - bytesInJava = self._jrdd.collect().iterator() - return list(self._collect_iterator_through_file(bytesInJava)) - - def _collect_iterator_through_file(self, iterator): - # Transferring lots of data through Py4J can be slow because - # socket.readline() is inefficient. Instead, we'll dump the data to a - # file and read it back. - tempFile = NamedTemporaryFile(delete=False, dir=self.ctx._temp_dir) - tempFile.close() - self.ctx._writeToFile(iterator, tempFile.name) - # Read the data into Python and deserialize it: - with open(tempFile.name, 'rb') as tempFile: - for item in self._jrdd_deserializer.load_stream(tempFile): - yield item - os.unlink(tempFile.name) + port = self.ctx._jvm.PythonRDD.collectAndServe(self._jrdd.rdd()) + return list(_load_from_socket(port, self._jrdd_deserializer)) def reduce(self, f): """ diff --git a/python/pyspark/sql/dataframe.py b/python/pyspark/sql/dataframe.py index 5c3b7377c33b5..e8ce4547455a5 100644 --- a/python/pyspark/sql/dataframe.py +++ b/python/pyspark/sql/dataframe.py @@ -19,13 +19,11 @@ import itertools import warnings import random -import os -from tempfile import NamedTemporaryFile from py4j.java_collections import ListConverter, MapConverter from pyspark.context import SparkContext -from pyspark.rdd import RDD +from pyspark.rdd import RDD, _load_from_socket from pyspark.serializers import BatchedSerializer, PickleSerializer, UTF8Deserializer from pyspark.storagelevel import StorageLevel from pyspark.traceback_utils import SCCallSiteSync @@ -310,14 +308,8 @@ def collect(self): [Row(age=2, name=u'Alice'), Row(age=5, name=u'Bob')] """ with SCCallSiteSync(self._sc) as css: - bytesInJava = self._jdf.javaToPython().collect().iterator() - tempFile = NamedTemporaryFile(delete=False, dir=self._sc._temp_dir) - tempFile.close() - self._sc._writeToFile(bytesInJava, tempFile.name) - # Read the data into Python and deserialize it: - with open(tempFile.name, 'rb') as tempFile: - rs = list(BatchedSerializer(PickleSerializer()).load_stream(tempFile)) - os.unlink(tempFile.name) + port = self._sc._jvm.PythonRDD.collectAndServe(self._jdf.javaToPython().rdd()) + rs = list(_load_from_socket(port, BatchedSerializer(PickleSerializer()))) cls = _create_cls(self.schema) return [cls(r) for r in rs] From 9a0272fbb322042788f14e9cd99e2db86b456225 Mon Sep 17 00:00:00 2001 From: Yuhao Yang Date: Tue, 10 Mar 2015 10:51:44 +0000 Subject: [PATCH 019/122] [SPARK-6177][MLlib]Add note in LDA example to remind possible coalesce JIRA: https://issues.apache.org/jira/browse/SPARK-6177 Add comment to introduce coalesce to LDA example to avoid the possible massive partitions from `sc.textFile`. sc.textFile will create RDD with one partition for each file, and the possible massive partitions downgrades LDA performance. Author: Yuhao Yang Closes #4899 from hhbyyh/adjustPartition and squashes the following commits: a499630 [Yuhao Yang] update comment 9a2d7b6 [Yuhao Yang] move to comment f7fd5d4 [Yuhao Yang] Merge remote-tracking branch 'upstream/master' into adjustPartition 26a564a [Yuhao Yang] add coalesce to LDAExample --- .../scala/org/apache/spark/examples/mllib/LDAExample.scala | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/examples/src/main/scala/org/apache/spark/examples/mllib/LDAExample.scala b/examples/src/main/scala/org/apache/spark/examples/mllib/LDAExample.scala index 11399a7633638..08a93595a2e17 100644 --- a/examples/src/main/scala/org/apache/spark/examples/mllib/LDAExample.scala +++ b/examples/src/main/scala/org/apache/spark/examples/mllib/LDAExample.scala @@ -173,7 +173,9 @@ object LDAExample { stopwordFile: String): (RDD[(Long, Vector)], Array[String], Long) = { // Get dataset of document texts - // One document per line in each text file. + // One document per line in each text file. If the input consists of many small files, + // this can result in a large number of small partitions, which can degrade performance. + // In this case, consider using coalesce() to create fewer, larger partitions. val textRDD: RDD[String] = sc.textFile(paths.mkString(",")) // Split text into words From c4c4b07bf61cab01d92fde4f902d8c06abdce240 Mon Sep 17 00:00:00 2001 From: Lev Khomich Date: Tue, 10 Mar 2015 10:55:42 +0000 Subject: [PATCH 020/122] [SPARK-6087][CORE] Provide actionable exception if Kryo buffer is not large enough A simple try-catch wrapping KryoException to be more informative. Author: Lev Khomich Closes #4947 from levkhomich/master and squashes the following commits: 0f7a947 [Lev Khomich] [SPARK-6087][CORE] Provide actionable exception if Kryo buffer is not large enough --- .../apache/spark/serializer/KryoSerializer.scala | 8 +++++++- .../spark/serializer/KryoSerializerSuite.scala | 14 ++++++++++++++ 2 files changed, 21 insertions(+), 1 deletion(-) diff --git a/core/src/main/scala/org/apache/spark/serializer/KryoSerializer.scala b/core/src/main/scala/org/apache/spark/serializer/KryoSerializer.scala index 9ce64d41fbc40..dc7aa99738c17 100644 --- a/core/src/main/scala/org/apache/spark/serializer/KryoSerializer.scala +++ b/core/src/main/scala/org/apache/spark/serializer/KryoSerializer.scala @@ -158,7 +158,13 @@ private[spark] class KryoSerializerInstance(ks: KryoSerializer) extends Serializ override def serialize[T: ClassTag](t: T): ByteBuffer = { output.clear() - kryo.writeClassAndObject(output, t) + try { + kryo.writeClassAndObject(output, t) + } catch { + case e: KryoException if e.getMessage.startsWith("Buffer overflow") => + throw new SparkException(s"Kryo serialization failed: ${e.getMessage}. To avoid this, " + + "increase spark.kryoserializer.buffer.max.mb value.") + } ByteBuffer.wrap(output.toBytes) } diff --git a/core/src/test/scala/org/apache/spark/serializer/KryoSerializerSuite.scala b/core/src/test/scala/org/apache/spark/serializer/KryoSerializerSuite.scala index 523d898207447..6198df84fab3d 100644 --- a/core/src/test/scala/org/apache/spark/serializer/KryoSerializerSuite.scala +++ b/core/src/test/scala/org/apache/spark/serializer/KryoSerializerSuite.scala @@ -261,6 +261,20 @@ class KryoSerializerSuite extends FunSuite with SharedSparkContext { ser.serialize(HighlyCompressedMapStatus(BlockManagerId("exec-1", "host", 1234), blockSizes)) } } + + test("serialization buffer overflow reporting") { + import org.apache.spark.SparkException + val kryoBufferMaxProperty = "spark.kryoserializer.buffer.max.mb" + + val largeObject = (1 to 1000000).toArray + + val conf = new SparkConf(false) + conf.set(kryoBufferMaxProperty, "1") + + val ser = new KryoSerializer(conf).newInstance() + val thrown = intercept[SparkException](ser.serialize(largeObject)) + assert(thrown.getMessage.contains(kryoBufferMaxProperty)) + } } From d14df06c05a6228fd6522914c39aa75898eddfc1 Mon Sep 17 00:00:00 2001 From: Nicholas Chammas Date: Tue, 10 Mar 2015 10:58:31 +0000 Subject: [PATCH 021/122] [SPARK-6191] [EC2] Generalize ability to download libs Right now we have a method to specifically download boto. This PR generalizes it so it's easy to download additional libraries if we want. For example, adding new external libraries for spark-ec2 is now as simple as: ```python external_libs = [ { "name": "boto", "version": "2.34.0", "md5": "5556223d2d0cc4d06dd4829e671dcecd" }, { "name": "PyYAML", "version": "3.11", "md5": "f50e08ef0fe55178479d3a618efe21db" }, { "name": "argparse", "version": "1.3.0", "md5": "9bcf7f612190885c8c85e30ba41db3c7" } ] ``` Likely use cases: * Downloading PyYAML to allow spark-ec2 configs to be persisted as a YAML file. ([SPARK-925](https://issues.apache.org/jira/browse/SPARK-925)) * Downloading argparse to clean up / modernize our option parsing. First run output, with PyYAML and argparse added just for demonstration purposes: ```shell $ ./spark-ec2 --version Downloading external libraries that spark-ec2 needs from PyPI to /path/to/spark/ec2/lib... This should be a one-time operation. - Downloading boto... - Finished downloading boto. - Downloading PyYAML... - Finished downloading PyYAML. - Downloading argparse... - Finished downloading argparse. spark-ec2 1.2.1 ``` Output thereafter: ```shell $ ./spark-ec2 --version spark-ec2 1.2.1 ``` Author: Nicholas Chammas Closes #4919 from nchammas/setup-ec2-libs and squashes the following commits: a077955 [Nicholas Chammas] print default region c95fb7d [Nicholas Chammas] to docstring 5448845 [Nicholas Chammas] remove libs added for demo purposes 60d8c23 [Nicholas Chammas] generalize ability to download libs --- ec2/spark_ec2.py | 82 +++++++++++++++++++++++++++++++----------------- 1 file changed, 54 insertions(+), 28 deletions(-) diff --git a/ec2/spark_ec2.py b/ec2/spark_ec2.py index b50b3816ff890..3acb5fea042df 100755 --- a/ec2/spark_ec2.py +++ b/ec2/spark_ec2.py @@ -70,34 +70,60 @@ DEFAULT_SPARK_EC2_BRANCH = "branch-1.3" -def setup_boto(): - # Download Boto if it's not already present in the SPARK_EC2_DIR/lib folder: - version = "boto-2.34.0" - md5 = "5556223d2d0cc4d06dd4829e671dcecd" - url = "https://pypi.python.org/packages/source/b/boto/%s.tar.gz" % version - lib_dir = os.path.join(SPARK_EC2_DIR, "lib") - if not os.path.exists(lib_dir): - os.mkdir(lib_dir) - boto_lib_dir = os.path.join(lib_dir, version) - if not os.path.isdir(boto_lib_dir): - tgz_file_path = os.path.join(lib_dir, "%s.tar.gz" % version) - print "Downloading Boto from PyPi" - download_stream = urllib2.urlopen(url) - with open(tgz_file_path, "wb") as tgz_file: - tgz_file.write(download_stream.read()) - with open(tgz_file_path) as tar: - if hashlib.md5(tar.read()).hexdigest() != md5: - print >> stderr, "ERROR: Got wrong md5sum for Boto" - sys.exit(1) - tar = tarfile.open(tgz_file_path) - tar.extractall(path=lib_dir) - tar.close() - os.remove(tgz_file_path) - print "Finished downloading Boto" - sys.path.insert(0, boto_lib_dir) +def setup_external_libs(libs): + """ + Download external libraries from PyPI to SPARK_EC2_DIR/lib/ and prepend them to our PATH. + """ + PYPI_URL_PREFIX = "https://pypi.python.org/packages/source" + SPARK_EC2_LIB_DIR = os.path.join(SPARK_EC2_DIR, "lib") + + if not os.path.exists(SPARK_EC2_LIB_DIR): + print "Downloading external libraries that spark-ec2 needs from PyPI to {path}...".format( + path=SPARK_EC2_LIB_DIR + ) + print "This should be a one-time operation." + os.mkdir(SPARK_EC2_LIB_DIR) + + for lib in libs: + versioned_lib_name = "{n}-{v}".format(n=lib["name"], v=lib["version"]) + lib_dir = os.path.join(SPARK_EC2_LIB_DIR, versioned_lib_name) + + if not os.path.isdir(lib_dir): + tgz_file_path = os.path.join(SPARK_EC2_LIB_DIR, versioned_lib_name + ".tar.gz") + print " - Downloading {lib}...".format(lib=lib["name"]) + download_stream = urllib2.urlopen( + "{prefix}/{first_letter}/{lib_name}/{lib_name}-{lib_version}.tar.gz".format( + prefix=PYPI_URL_PREFIX, + first_letter=lib["name"][:1], + lib_name=lib["name"], + lib_version=lib["version"] + ) + ) + with open(tgz_file_path, "wb") as tgz_file: + tgz_file.write(download_stream.read()) + with open(tgz_file_path) as tar: + if hashlib.md5(tar.read()).hexdigest() != lib["md5"]: + print >> stderr, "ERROR: Got wrong md5sum for {lib}.".format(lib=lib["name"]) + sys.exit(1) + tar = tarfile.open(tgz_file_path) + tar.extractall(path=SPARK_EC2_LIB_DIR) + tar.close() + os.remove(tgz_file_path) + print " - Finished downloading {lib}.".format(lib=lib["name"]) + sys.path.insert(1, lib_dir) + + +# Only PyPI libraries are supported. +external_libs = [ + { + "name": "boto", + "version": "2.34.0", + "md5": "5556223d2d0cc4d06dd4829e671dcecd" + } +] +setup_external_libs(external_libs) -setup_boto() import boto from boto.ec2.blockdevicemapping import BlockDeviceMapping, BlockDeviceType, EBSBlockDeviceType from boto import ec2 @@ -136,7 +162,7 @@ def parse_args(): help="Master instance type (leave empty for same as instance-type)") parser.add_option( "-r", "--region", default="us-east-1", - help="EC2 region used to launch instances in, or to find them in") + help="EC2 region used to launch instances in, or to find them in (default: %default)") parser.add_option( "-z", "--zone", default="", help="Availability zone to launch instances in, or 'all' to spread " + @@ -230,7 +256,7 @@ def parse_args(): "(e.g -Dspark.worker.timeout=180)") parser.add_option( "--user-data", type="string", default="", - help="Path to a user-data file (most AMI's interpret this as an initialization script)") + help="Path to a user-data file (most AMIs interpret this as an initialization script)") parser.add_option( "--authorized-address", type="string", default="0.0.0.0/0", help="Address to authorize on created security groups (default: %default)") From 7c7d2d5e093b0623edc75dd166ec1179b4e62062 Mon Sep 17 00:00:00 2001 From: cheng chang Date: Tue, 10 Mar 2015 11:02:12 +0000 Subject: [PATCH 022/122] [SPARK-6186] [EC2] Make Tachyon version configurable in EC2 deployment script This PR comes from Tachyon community to solve the issue: https://tachyon.atlassian.net/browse/TACHYON-11 An accompanying PR is in mesos/spark-ec2: https://github.com/mesos/spark-ec2/pull/101 Author: cheng chang Closes #4901 from uronce-cc/master and squashes the following commits: 313aa36 [cheng chang] minor re-wording fd2a48e [cheng chang] Remove Tachyon when deploying through git hash 1d53c5c [cheng chang] add default value to --tachyon-version 6f8887e [cheng chang] make tachyon version configurable --- .../root/spark-ec2/ec2-variables.sh | 3 ++- ec2/spark_ec2.py | 19 +++++++++++++++++++ 2 files changed, 21 insertions(+), 1 deletion(-) diff --git a/ec2/deploy.generic/root/spark-ec2/ec2-variables.sh b/ec2/deploy.generic/root/spark-ec2/ec2-variables.sh index 740c267fd9866..0857657152ec7 100644 --- a/ec2/deploy.generic/root/spark-ec2/ec2-variables.sh +++ b/ec2/deploy.generic/root/spark-ec2/ec2-variables.sh @@ -26,9 +26,10 @@ export SPARK_LOCAL_DIRS="{{spark_local_dirs}}" export MODULES="{{modules}}" export SPARK_VERSION="{{spark_version}}" export SHARK_VERSION="{{shark_version}}" +export TACHYON_VERSION="{{tachyon_version}}" export HADOOP_MAJOR_VERSION="{{hadoop_major_version}}" export SWAP_MB="{{swap}}" export SPARK_WORKER_INSTANCES="{{spark_worker_instances}}" export SPARK_MASTER_OPTS="{{spark_master_opts}}" export AWS_ACCESS_KEY_ID="{{aws_access_key_id}}" -export AWS_SECRET_ACCESS_KEY="{{aws_secret_access_key}}" \ No newline at end of file +export AWS_SECRET_ACCESS_KEY="{{aws_secret_access_key}}" diff --git a/ec2/spark_ec2.py b/ec2/spark_ec2.py index 3acb5fea042df..f848874b0c775 100755 --- a/ec2/spark_ec2.py +++ b/ec2/spark_ec2.py @@ -62,6 +62,16 @@ "1.2.1", ]) +SPARK_TACHYON_MAP = { + "1.0.0": "0.4.1", + "1.0.1": "0.4.1", + "1.0.2": "0.4.1", + "1.1.0": "0.5.0", + "1.1.1": "0.5.0", + "1.2.0": "0.5.0", + "1.2.1": "0.5.0", +} + DEFAULT_SPARK_VERSION = SPARK_EC2_VERSION DEFAULT_SPARK_GITHUB_REPO = "https://github.com/apache/spark" @@ -370,6 +380,10 @@ def get_validate_spark_version(version, repo): } +def get_tachyon_version(spark_version): + return SPARK_TACHYON_MAP.get(spark_version, "") + + # Attempt to resolve an appropriate AMI given the architecture and region of the request. def get_spark_ami(opts): if opts.instance_type in EC2_INSTANCE_TYPES: @@ -919,9 +933,13 @@ def deploy_files(conn, root_dir, opts, master_nodes, slave_nodes, modules): if "." in opts.spark_version: # Pre-built Spark deploy spark_v = get_validate_spark_version(opts.spark_version, opts.spark_git_repo) + tachyon_v = get_tachyon_version(spark_v) else: # Spark-only custom deploy spark_v = "%s|%s" % (opts.spark_git_repo, opts.spark_version) + tachyon_v = "" + print "Deploying Spark via git hash; Tachyon won't be set up" + modules = filter(lambda x: x != "tachyon", modules) template_vars = { "master_list": '\n'.join([i.public_dns_name for i in master_nodes]), @@ -934,6 +952,7 @@ def deploy_files(conn, root_dir, opts, master_nodes, slave_nodes, modules): "swap": str(opts.swap), "modules": '\n'.join(modules), "spark_version": spark_v, + "tachyon_version": tachyon_v, "hadoop_major_version": opts.hadoop_major_version, "spark_worker_instances": "%d" % opts.worker_instances, "spark_master_opts": opts.master_opts From 74fb433702b676225097e1d4d2c2b170915a5d19 Mon Sep 17 00:00:00 2001 From: Reynold Xin Date: Tue, 10 Mar 2015 17:25:04 -0700 Subject: [PATCH 023/122] Minor doc: Remove the extra blank line in data types javadoc. The extra blank line is preventing the first lines from showing up in the package summary page. Author: Reynold Xin Closes #4955 from rxin/datatype-docs and squashes the following commits: 1621114 [Reynold Xin] Minor doc: Remove the extra blank line in data types javadoc. --- .../apache/spark/sql/types/dataTypes.scala | 24 +++++-------------- 1 file changed, 6 insertions(+), 18 deletions(-) diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/types/dataTypes.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/types/dataTypes.scala index 92d322845f5c5..bf39603d13bd5 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/types/dataTypes.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/types/dataTypes.scala @@ -240,7 +240,6 @@ object DataType { /** * :: DeveloperApi :: - * * The base type of all Spark SQL data types. * * @group dataType @@ -282,7 +281,6 @@ abstract class DataType { /** * :: DeveloperApi :: - * * The data type representing `NULL` values. Please use the singleton [[DataTypes.NullType]]. * * @group dataType @@ -342,7 +340,6 @@ protected[sql] abstract class NativeType extends DataType { /** * :: DeveloperApi :: - * * The data type representing `String` values. Please use the singleton [[DataTypes.StringType]]. * * @group dataType @@ -369,7 +366,6 @@ case object StringType extends StringType /** * :: DeveloperApi :: - * * The data type representing `Array[Byte]` values. * Please use the singleton [[DataTypes.BinaryType]]. * @@ -405,7 +401,6 @@ case object BinaryType extends BinaryType /** * :: DeveloperApi :: - * * The data type representing `Boolean` values. Please use the singleton [[DataTypes.BooleanType]]. * *@group dataType @@ -432,7 +427,6 @@ case object BooleanType extends BooleanType /** * :: DeveloperApi :: - * * The data type representing `java.sql.Timestamp` values. * Please use the singleton [[DataTypes.TimestampType]]. * @@ -464,7 +458,6 @@ case object TimestampType extends TimestampType /** * :: DeveloperApi :: - * * The data type representing `java.sql.Date` values. * Please use the singleton [[DataTypes.DateType]]. * @@ -492,6 +485,12 @@ class DateType private() extends NativeType { case object DateType extends DateType +/** + * :: DeveloperApi :: + * Numeric data types. + * + * @group dataType + */ abstract class NumericType extends NativeType with PrimitiveType { // Unfortunately we can't get this implicitly as that breaks Spark Serialization. In order for // implicitly[Numeric[JvmType]] to be valid, we have to change JvmType from a type variable to a @@ -523,7 +522,6 @@ protected[sql] sealed abstract class IntegralType extends NumericType { /** * :: DeveloperApi :: - * * The data type representing `Long` values. Please use the singleton [[DataTypes.LongType]]. * * @group dataType @@ -554,7 +552,6 @@ case object LongType extends LongType /** * :: DeveloperApi :: - * * The data type representing `Int` values. Please use the singleton [[DataTypes.IntegerType]]. * * @group dataType @@ -585,7 +582,6 @@ case object IntegerType extends IntegerType /** * :: DeveloperApi :: - * * The data type representing `Short` values. Please use the singleton [[DataTypes.ShortType]]. * * @group dataType @@ -616,7 +612,6 @@ case object ShortType extends ShortType /** * :: DeveloperApi :: - * * The data type representing `Byte` values. Please use the singleton [[DataTypes.ByteType]]. * * @group dataType @@ -666,7 +661,6 @@ case class PrecisionInfo(precision: Int, scale: Int) /** * :: DeveloperApi :: - * * The data type representing `java.math.BigDecimal` values. * A Decimal that might have fixed precision and scale, or unlimited values for these. * @@ -745,7 +739,6 @@ object DecimalType { /** * :: DeveloperApi :: - * * The data type representing `Double` values. Please use the singleton [[DataTypes.DoubleType]]. * * @group dataType @@ -775,7 +768,6 @@ case object DoubleType extends DoubleType /** * :: DeveloperApi :: - * * The data type representing `Float` values. Please use the singleton [[DataTypes.FloatType]]. * * @group dataType @@ -811,7 +803,6 @@ object ArrayType { /** * :: DeveloperApi :: - * * The data type for collections of multiple values. * Internally these are represented as columns that contain a ``scala.collection.Seq``. * @@ -854,7 +845,6 @@ case class ArrayType(elementType: DataType, containsNull: Boolean) extends DataT /** * A field inside a StructType. - * * @param name The name of this field. * @param dataType The data type of this field. * @param nullable Indicates if values of this field can be `null` values. @@ -949,7 +939,6 @@ object StructType { /** * :: DeveloperApi :: - * * A [[StructType]] object can be constructed by * {{{ * StructType(fields: Seq[StructField]) @@ -1118,7 +1107,6 @@ object MapType { /** * :: DeveloperApi :: - * * The data type for Maps. Keys in a map are not allowed to have `null` values. * * Please use [[DataTypes.createMapType()]] to create a specific instance. From 2672374110d58e45ffae2408e74b96613deddda3 Mon Sep 17 00:00:00 2001 From: Michael Armbrust Date: Tue, 10 Mar 2015 18:13:09 -0700 Subject: [PATCH 024/122] [SPARK-5183][SQL] Update SQL Docs with JDBC and Migration Guide Author: Michael Armbrust Closes #4958 from marmbrus/sqlDocs and squashes the following commits: 9351dbc [Michael Armbrust] fix parquet example 6877e13 [Michael Armbrust] add sql examples d81b7e7 [Michael Armbrust] rxins comments e393528 [Michael Armbrust] fix order 19c2735 [Michael Armbrust] more on data source load/store 00d5914 [Michael Armbrust] Update SQL Docs with JDBC and Migration Guide --- docs/sql-programming-guide.md | 382 ++++++++++++++++++++++++++++++++-- 1 file changed, 370 insertions(+), 12 deletions(-) diff --git a/docs/sql-programming-guide.md b/docs/sql-programming-guide.md index 4fbdca7397951..9c363bc87e890 100644 --- a/docs/sql-programming-guide.md +++ b/docs/sql-programming-guide.md @@ -9,7 +9,7 @@ title: Spark SQL and DataFrames # Overview -Spark SQL is a Spark module for structured data processing. It provides a programming abstraction called DataFrames and can also act as a distributed query engine. +Spark SQL is a Spark module for structured data processing. It provides a programming abstraction called DataFrames and can also act as distributed SQL query engine. # DataFrames @@ -662,8 +662,146 @@ for name in names.collect(): Spark SQL supports operating on a variety of data sources through the `DataFrame` interface. A DataFrame can be operated on as normal RDDs and can also be registered as a temporary table. Registering a DataFrame as a table allows you to run SQL queries over its data. This section -describes the various methods for loading data into a DataFrame. +describes the general methods for loading and saving data using the Spark Data Sources and then +goes into specific options that are available for the built-in data sources. +## Generic Load/Save Functions + +In the simplest form, the default data source (`parquet` unless otherwise configured by +`spark.sql.sources.default`) will be used for all operations. + +
    +
    + +{% highlight scala %} +val df = sqlContext.load("people.parquet") +df.select("name", "age").save("namesAndAges.parquet") +{% endhighlight %} + +
    + +
    + +{% highlight java %} + +DataFrame df = sqlContext.load("people.parquet"); +df.select("name", "age").save("namesAndAges.parquet"); + +{% endhighlight %} + +
    + +
    + +{% highlight python %} + +df = sqlContext.load("people.parquet") +df.select("name", "age").save("namesAndAges.parquet") + +{% endhighlight %} + +
    +
    + +### Manually Specifying Options + +You can also manually specify the data source that will be used along with any extra options +that you would like to pass to the data source. Data sources are specified by their fully qualified +name (i.e., `org.apache.spark.sql.parquet`), but for built-in sources you can also use the shorted +name (`json`, `parquet`, `jdbc`). DataFrames of any type can be converted into other types +using this syntax. + +
    +
    + +{% highlight scala %} +val df = sqlContext.load("people.json", "json") +df.select("name", "age").save("namesAndAges.parquet", "parquet") +{% endhighlight %} + +
    + +
    + +{% highlight java %} + +DataFrame df = sqlContext.load("people.json", "json"); +df.select("name", "age").save("namesAndAges.parquet", "parquet"); + +{% endhighlight %} + +
    + +
    + +{% highlight python %} + +df = sqlContext.load("people.json", "json") +df.select("name", "age").save("namesAndAges.parquet", "parquet") + +{% endhighlight %} + +
    +
    + +### Save Modes + +Save operations can optionally take a `SaveMode`, that specifies how to handle existing data if +present. It is important to realize that these save modes do not utilize any locking and are not +atomic. Thus, it is not safe to have multiple writers attempting to write to the same location. +Additionally, when performing a `Overwrite`, the data will be deleted before writing out the +new data. + +
    + + + + + + + + + + + + + + + + + + + + + +
    Scala/JavaPythonMeaning
    SaveMode.ErrorIfExists (default)"error" (default) + When saving a DataFrame to a data source, if data already exists, + an exception is expected to be thrown. +
    SaveMode.Append"append" + When saving a DataFrame to a data source, if data/table already exists, + contents of the DataFrame are expected to be appended to existing data. +
    SaveMode.Overwrite"overwrite" + Overwrite mode means that when saving a DataFrame to a data source, + if data/table already exists, existing data is expected to be overwritten by the contents of + the DataFrame. +
    SaveMode.Ignore"ignore" + Ignore mode means that when saving a DataFrame to a data source, if data already exists, + the save operation is expected to not save the contents of the DataFrame and to not + change the existing data. This is similar to a `CREATE TABLE IF NOT EXISTS` in SQL. +
    + +### Saving to Persistent Tables + +When working with a `HiveContext`, `DataFrames` can also be saved as persistent tables using the +`saveAsTable` command. Unlike the `registerTempTable` command, `saveAsTable` will materialize the +contents of the dataframe and create a pointer to the data in the HiveMetastore. Persistent tables +will still exist even after your Spark program has restarted, as long as you maintain your connection +to the same metastore. A DataFrame for a persistent table can be created by calling the `table` +method on a SQLContext with the name of the table. + +By default `saveAsTable` will create a "managed table", meaning that the location of the data will +be controlled by the metastore. Managed tables will also have their data deleted automatically +when a table is dropped. ## Parquet Files @@ -751,6 +889,22 @@ for teenName in teenNames.collect():
    +
    + +{% highlight sql %} + +CREATE TEMPORARY TABLE parquetTable +USING org.apache.spark.sql.parquet +OPTIONS ( + path "examples/src/main/resources/people.parquet" +) + +SELECT * FROM parquetTable + +{% endhighlight %} + +
    + ### Configuration @@ -942,6 +1096,22 @@ anotherPeople = sqlContext.jsonRDD(anotherPeopleRDD) {% endhighlight %} +
    + +{% highlight sql %} + +CREATE TEMPORARY TABLE jsonTable +USING org.apache.spark.sql.json +OPTIONS ( + path "examples/src/main/resources/people.json" +) + +SELECT * FROM jsonTable + +{% endhighlight %} + +
    + ## Hive Tables @@ -1022,6 +1192,121 @@ results = sqlContext.sql("FROM src SELECT key, value").collect() +## JDBC To Other Databases + +Spark SQL also includes a data source that can read data from other databases using JDBC. This +functionality should be preferred over using [JdbcRDD](api/scala/index.html#org.apache.spark.rdd.JdbcRDD). +This is because the results are returned +as a DataFrame and they can easily be processed in Spark SQL or joined with other data sources. +The JDBC data source is also easier to use from Java or Python as it does not require the user to +provide a ClassTag. +(Note that this is different than the Spark SQL JDBC server, which allows other applications to +run queries using Spark SQL). + +To get started you will need to include the JDBC driver for you particular database on the +spark classpath. For example, to connect to postgres from the Spark Shell you would run the +following command: + +{% highlight bash %} +SPARK_CLASSPATH=postgresql-9.3-1102-jdbc41.jar bin/spark-shell +{% endhighlight %} + +Tables from the remote database can be loaded as a DataFrame or Spark SQL Temporary table using +the Data Sources API. The following options are supported: + + + + + + + + + + + + + + + + + + + + +
    Property NameMeaning
    url + The JDBC URL to connect to. +
    dbtable + The JDBC table that should be read. Note that anything that is valid in a `FROM` clause of + a SQL query can be used. For example, instead of a full table you could also use a + subquery in parentheses. +
    driver + The class name of the JDBC driver needed to connect to this URL. This class with be loaded + on the master and workers before running an JDBC commands to allow the driver to + register itself with the JDBC subsystem. +
    partitionColumn, lowerBound, upperBound, numPartitions + These options must all be specified if any of them is specified. They describe how to + partition the table when reading in parallel from multiple workers. + partitionColumn must be a numeric column from the table in question. +
    + +
    + +
    + +{% highlight scala %} +val jdbcDF = sqlContext.load("jdbc", Map( + "url" -> "jdbc:postgresql:dbserver", + "dbtable" -> "schema.tablename")) +{% endhighlight %} + +
    + +
    + +{% highlight java %} + +Map options = new HashMap(); +options.put("url", "jdbc:postgresql:dbserver"); +options.put("dbtable", "schema.tablename"); + +DataFrame jdbcDF = sqlContext.load("jdbc", options) +{% endhighlight %} + + +
    + +
    + +{% highlight python %} + +df = sqlContext.load("jdbc", url="jdbc:postgresql:dbserver", dbtable="schema.tablename") + +{% endhighlight %} + +
    + +
    + +{% highlight sql %} + +CREATE TEMPORARY TABLE jdbcTable +USING org.apache.spark.sql.jdbc +OPTIONS ( + url "jdbc:postgresql:dbserver", + dbtable "schema.tablename" +) + +{% endhighlight %} + +
    +
    + +## Troubleshooting + + * The JDBC driver class must be visible to the primordial class loader on the client session and on all executors. This is because Java's DriverManager class does a security check that results in it ignoring all drivers not visible to the primordial class loader when one goes to open a connection. One convenient way to do this is to modify compute_classpath.sh on all worker nodes to include your driver JARs. + * Some databases, such as H2, convert all names to upper case. You'll need to use upper case to refer to those names in Spark SQL. + + # Performance Tuning For some workloads it is possible to improve performance by either caching data in memory, or by @@ -1092,7 +1377,7 @@ that these options will be deprecated in future release as more optimizations ar
    -# Distributed Query Engine +# Distributed SQL Engine Spark SQL can also act as a distributed query engine using its JDBC/ODBC or command-line interface. In this mode, end-users or applications can interact with Spark SQL directly to run SQL queries, without the need to write any code. @@ -1171,6 +1456,87 @@ options. # Migration Guide +## Upgrading from Spark SQL 1.0-1.2 to 1.3 + +In Spark 1.3 we removed the "Alpha" label from Spark SQL and as part of this did a cleanup of the +available APIs. From Spark 1.3 onwards, Spark SQL will provide binary compatibility with other +releases in the 1.X series. This compatibility guarantee excludes APIs that are explicitly marked +as unstable (i.e., DeveloperAPI or Experimental). + +#### Rename of SchemaRDD to DataFrame + +The largest change that users will notice when upgrading to Spark SQL 1.3 is that `SchemaRDD` has +been renamed to `DataFrame`. This is primarily because DataFrames no longer inherit from RDD +directly, but instead provide most of the functionality that RDDs provide though their own +implementation. DataFrames can still be converted to RDDs by calling the `.rdd` method. + +In Scala there is a type alias from `SchemaRDD` to `DataFrame` to provide source compatibility for +some use cases. It is still recommended that users update their code to use `DataFrame` instead. +Java and Python users will need to update their code. + +#### Unification of the Java and Scala APIs + +Prior to Spark 1.3 there were separate Java compatible classes (`JavaSQLContext` and `JavaSchemaRDD`) +that mirrored the Scala API. In Spark 1.3 the Java API and Scala API have been unified. Users +of either language should use `SQLContext` and `DataFrame`. In general theses classes try to +use types that are usable from both languages (i.e. `Array` instead of language specific collections). +In some cases where no common type exists (e.g., for passing in closures or Maps) function overloading +is used instead. + +Additionally the Java specific types API has been removed. Users of both Scala and Java should +use the classes present in `org.apache.spark.sql.types` to describe schema programmatically. + + +#### Isolation of Implicit Conversions and Removal of dsl Package (Scala-only) + +Many of the code examples prior to Spark 1.3 started with `import sqlContext._`, which brought +all of the functions from sqlContext into scope. In Spark 1.3 we have isolated the implicit +conversions for converting `RDD`s into `DataFrame`s into an object inside of the `SQLContext`. +Users should now write `import sqlContext.implicits._`. + +Additionally, the implicit conversions now only augment RDDs that are composed of `Product`s (i.e., +case classes or tuples) with a method `toDF`, instead of applying automatically. + +When using function inside of the DSL (now replaced with the `DataFrame` API) users used to import +`org.apache.spark.sql.catalyst.dsl`. Instead the public dataframe functions API should be used: +`import org.apache.spark.sql.functions._`. + +#### Removal of the type aliases in org.apache.spark.sql for DataType (Scala-only) + +Spark 1.3 removes the type aliases that were present in the base sql package for `DataType`. Users +should instead import the classes in `org.apache.spark.sql.types` + +#### UDF Registration Moved to sqlContext.udf (Java & Scala) + +Functions that are used to register UDFs, either for use in the DataFrame DSL or SQL, have been +moved into the udf object in `SQLContext`. + +
    +
    +{% highlight java %} + +sqlCtx.udf.register("strLen", (s: String) => s.length()) + +{% endhighlight %} +
    + +
    +{% highlight java %} + +sqlCtx.udf().register("strLen", (String s) -> { s.length(); }); + +{% endhighlight %} +
    + +
    + +Python UDF registration is unchanged. + +#### Python DataTypes No Longer Singletons + +When using DataTypes in Python you will need to construct them (i.e. `StringType()`) instead of +referencing a singleton. + ## Migration Guide for Shark User ### Scheduling @@ -1289,15 +1655,10 @@ in Hive deployments. * Tables with buckets: bucket is the hash partitioning within a Hive table partition. Spark SQL doesn't support buckets yet. -**Esoteric Hive Features** -* Tables with partitions using different input formats: In Spark SQL, all table partitions need to - have the same input format. -* Non-equi outer join: For the uncommon use case of using outer joins with non-equi join conditions - (e.g. condition "`key < 10`"), Spark SQL will output wrong result for the `NULL` tuple. +**Esoteric Hive Features** * `UNION` type * Unique join -* Single query multi insert * Column statistics collecting: Spark SQL does not piggyback scans to collect column statistics at the moment and only supports populating the sizeInBytes field of the hive metastore. @@ -1313,9 +1674,6 @@ less important due to Spark SQL's in-memory computational model. Others are slot releases of Spark SQL. * Block level bitmap indexes and virtual columns (used to build indexes) -* Automatically convert a join to map join: For joining a large table with multiple small tables, - Hive automatically converts the join into a map join. We are adding this auto conversion in the - next release. * Automatically determine the number of reducers for joins and groupbys: Currently in Spark SQL, you need to control the degree of parallelism post-shuffle using "`SET spark.sql.shuffle.partitions=[num_tasks];`". * Meta-data only query: For queries that can be answered by using only meta data, Spark SQL still From 2d4e00efe2cf179935ae108a68f28edf6e5a1628 Mon Sep 17 00:00:00 2001 From: Xusen Yin Date: Wed, 11 Mar 2015 00:24:55 -0700 Subject: [PATCH 025/122] [SPARK-5986][MLLib] Add save/load for k-means This PR adds save/load for K-means as described in SPARK-5986. Python version will be added in another PR. Author: Xusen Yin Closes #4951 from yinxusen/SPARK-5986 and squashes the following commits: 6dd74a0 [Xusen Yin] rewrite some functions and classes cd390fd [Xusen Yin] add indexed point b144216 [Xusen Yin] remove invalid comments dce7055 [Xusen Yin] add save/load for k-means for SPARK-5986 --- .../spark/mllib/clustering/KMeansModel.scala | 68 ++++++++++++++++++- .../spark/mllib/clustering/KMeansSuite.scala | 44 +++++++++++- 2 files changed, 108 insertions(+), 4 deletions(-) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeansModel.scala b/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeansModel.scala index 3b95a9e6936e8..707da537d238f 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeansModel.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeansModel.scala @@ -17,15 +17,22 @@ package org.apache.spark.mllib.clustering +import org.json4s._ +import org.json4s.JsonDSL._ +import org.json4s.jackson.JsonMethods._ + import org.apache.spark.api.java.JavaRDD -import org.apache.spark.rdd.RDD -import org.apache.spark.SparkContext._ import org.apache.spark.mllib.linalg.Vector +import org.apache.spark.mllib.util.{Loader, Saveable} +import org.apache.spark.rdd.RDD +import org.apache.spark.SparkContext +import org.apache.spark.sql.SQLContext +import org.apache.spark.sql.Row /** * A clustering model for K-means. Each point belongs to the cluster with the closest center. */ -class KMeansModel (val clusterCenters: Array[Vector]) extends Serializable { +class KMeansModel (val clusterCenters: Array[Vector]) extends Saveable with Serializable { /** Total number of clusters. */ def k: Int = clusterCenters.length @@ -58,4 +65,59 @@ class KMeansModel (val clusterCenters: Array[Vector]) extends Serializable { private def clusterCentersWithNorm: Iterable[VectorWithNorm] = clusterCenters.map(new VectorWithNorm(_)) + + override def save(sc: SparkContext, path: String): Unit = { + KMeansModel.SaveLoadV1_0.save(sc, this, path) + } + + override protected def formatVersion: String = "1.0" +} + +object KMeansModel extends Loader[KMeansModel] { + override def load(sc: SparkContext, path: String): KMeansModel = { + KMeansModel.SaveLoadV1_0.load(sc, path) + } + + private case class Cluster(id: Int, point: Vector) + + private object Cluster { + def apply(r: Row): Cluster = { + Cluster(r.getInt(0), r.getAs[Vector](1)) + } + } + + private[clustering] + object SaveLoadV1_0 { + + private val thisFormatVersion = "1.0" + + private[clustering] + val thisClassName = "org.apache.spark.mllib.clustering.KMeansModel" + + def save(sc: SparkContext, model: KMeansModel, path: String): Unit = { + val sqlContext = new SQLContext(sc) + import sqlContext.implicits._ + val metadata = compact(render( + ("class" -> thisClassName) ~ ("version" -> thisFormatVersion) ~ ("k" -> model.k))) + sc.parallelize(Seq(metadata), 1).saveAsTextFile(Loader.metadataPath(path)) + val dataRDD = sc.parallelize(model.clusterCenters.zipWithIndex).map { case (point, id) => + Cluster(id, point) + }.toDF() + dataRDD.saveAsParquetFile(Loader.dataPath(path)) + } + + def load(sc: SparkContext, path: String): KMeansModel = { + implicit val formats = DefaultFormats + val sqlContext = new SQLContext(sc) + val (className, formatVersion, metadata) = Loader.loadMetadata(sc, path) + assert(className == thisClassName) + assert(formatVersion == thisFormatVersion) + val k = (metadata \ "k").extract[Int] + val centriods = sqlContext.parquetFile(Loader.dataPath(path)) + Loader.checkSchema[Cluster](centriods.schema) + val localCentriods = centriods.map(Cluster.apply).collect() + assert(k == localCentriods.size) + new KMeansModel(localCentriods.sortBy(_.id).map(_.point)) + } + } } diff --git a/mllib/src/test/scala/org/apache/spark/mllib/clustering/KMeansSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/clustering/KMeansSuite.scala index caee5917000aa..7bf250eb5a383 100644 --- a/mllib/src/test/scala/org/apache/spark/mllib/clustering/KMeansSuite.scala +++ b/mllib/src/test/scala/org/apache/spark/mllib/clustering/KMeansSuite.scala @@ -21,9 +21,10 @@ import scala.util.Random import org.scalatest.FunSuite -import org.apache.spark.mllib.linalg.{Vector, Vectors} +import org.apache.spark.mllib.linalg.{DenseVector, SparseVector, Vector, Vectors} import org.apache.spark.mllib.util.{LocalClusterSparkContext, MLlibTestSparkContext} import org.apache.spark.mllib.util.TestingUtils._ +import org.apache.spark.util.Utils class KMeansSuite extends FunSuite with MLlibTestSparkContext { @@ -257,6 +258,47 @@ class KMeansSuite extends FunSuite with MLlibTestSparkContext { assert(predicts(0) != predicts(3)) } } + + test("model save/load") { + val tempDir = Utils.createTempDir() + val path = tempDir.toURI.toString + + Array(true, false).foreach { case selector => + val model = KMeansSuite.createModel(10, 3, selector) + // Save model, load it back, and compare. + try { + model.save(sc, path) + val sameModel = KMeansModel.load(sc, path) + KMeansSuite.checkEqual(model, sameModel) + } finally { + Utils.deleteRecursively(tempDir) + } + } + } +} + +object KMeansSuite extends FunSuite { + def createModel(dim: Int, k: Int, isSparse: Boolean): KMeansModel = { + val singlePoint = isSparse match { + case true => + Vectors.sparse(dim, Array.empty[Int], Array.empty[Double]) + case _ => + Vectors.dense(Array.fill[Double](dim)(0.0)) + } + new KMeansModel(Array.fill[Vector](k)(singlePoint)) + } + + def checkEqual(a: KMeansModel, b: KMeansModel): Unit = { + assert(a.k === b.k) + a.clusterCenters.zip(b.clusterCenters).foreach { + case (ca: SparseVector, cb: SparseVector) => + assert(ca === cb) + case (ca: DenseVector, cb: DenseVector) => + assert(ca === cb) + case _ => + throw new AssertionError("checkEqual failed since the two clusters were not identical.\n") + } + } } class KMeansClusterSuite extends FunSuite with LocalClusterSparkContext { From 517975d89d40a77c7186f488547eed11f79c1e97 Mon Sep 17 00:00:00 2001 From: Marcelo Vanzin Date: Wed, 11 Mar 2015 01:03:01 -0700 Subject: [PATCH 026/122] [SPARK-4924] Add a library for launching Spark jobs programmatically. This change encapsulates all the logic involved in launching a Spark job into a small Java library that can be easily embedded into other applications. The overall goal of this change is twofold, as described in the bug: - Provide a public API for launching Spark processes. This is a common request from users and currently there's no good answer for it. - Remove a lot of the duplicated code and other coupling that exists in the different parts of Spark that deal with launching processes. A lot of the duplication was due to different code needed to build an application's classpath (and the bootstrapper needed to run the driver in certain situations), and also different code needed to parse spark-submit command line options in different contexts. The change centralizes those as much as possible so that all code paths can rely on the library for handling those appropriately. Author: Marcelo Vanzin Closes #3916 from vanzin/SPARK-4924 and squashes the following commits: 18c7e4d [Marcelo Vanzin] Fix make-distribution.sh. 2ce741f [Marcelo Vanzin] Add lots of quotes. 3b28a75 [Marcelo Vanzin] Update new pom. a1b8af1 [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 897141f [Marcelo Vanzin] Review feedback. e2367d2 [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 28cd35e [Marcelo Vanzin] Remove stale comment. b1d86b0 [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 00505f9 [Marcelo Vanzin] Add blurb about new API in the programming guide. 5f4ddcc [Marcelo Vanzin] Better usage messages. 92a9cfb [Marcelo Vanzin] Fix Win32 launcher, usage. 6184c07 [Marcelo Vanzin] Rename field. 4c19196 [Marcelo Vanzin] Update comment. 7e66c18 [Marcelo Vanzin] Fix pyspark tests. 0031a8e [Marcelo Vanzin] Review feedback. c12d84b [Marcelo Vanzin] Review feedback. And fix spark-submit on Windows. e2d4d71 [Marcelo Vanzin] Simplify some code used to launch pyspark. 43008a7 [Marcelo Vanzin] Don't make builder extend SparkLauncher. b4d6912 [Marcelo Vanzin] Use spark-submit script in SparkLauncher. 28b1434 [Marcelo Vanzin] Add a comment. 304333a [Marcelo Vanzin] Fix propagation of properties file arg. bb67b93 [Marcelo Vanzin] Remove unrelated Yarn change (that is also wrong). 8ec0243 [Marcelo Vanzin] Add missing newline. 95ddfa8 [Marcelo Vanzin] Fix handling of --help for spark-class command builder. 72da7ec [Marcelo Vanzin] Rename SparkClassLauncher. 62978e4 [Marcelo Vanzin] Minor cleanup of Windows code path. 9cd5b44 [Marcelo Vanzin] Make all non-public APIs package-private. e4c80b6 [Marcelo Vanzin] Reorganize the code so that only SparkLauncher is public. e50dc5e [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 de81da2 [Marcelo Vanzin] Fix CommandUtils. 86a87bf [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 2061967 [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 46d46da [Marcelo Vanzin] Clean up a test and make it more future-proof. b93692a [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 ad03c48 [Marcelo Vanzin] Revert "Fix a thread-safety issue in "local" mode." 0b509d0 [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 23aa2a9 [Marcelo Vanzin] Read java-opts from conf dir, not spark home. 7cff919 [Marcelo Vanzin] Javadoc updates. eae4d8e [Marcelo Vanzin] Fix new unit tests on Windows. e570fb5 [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 44cd5f7 [Marcelo Vanzin] Add package-info.java, clean up javadocs. f7cacff [Marcelo Vanzin] Remove "launch Spark in new thread" feature. 7ed8859 [Marcelo Vanzin] Some more feedback. 54cd4fd [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 61919df [Marcelo Vanzin] Clean leftover debug statement. aae5897 [Marcelo Vanzin] Use launcher classes instead of jars in non-release mode. e584fc3 [Marcelo Vanzin] Rework command building a little bit. 525ef5b [Marcelo Vanzin] Rework Unix spark-class to handle argument with newlines. 8ac4e92 [Marcelo Vanzin] Minor test cleanup. e946a99 [Marcelo Vanzin] Merge PySparkLauncher into SparkSubmitCliLauncher. c617539 [Marcelo Vanzin] Review feedback round 1. fc6a3e2 [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 f26556b [Marcelo Vanzin] Fix a thread-safety issue in "local" mode. 2f4e8b4 [Marcelo Vanzin] Changes needed to make this work with SPARK-4048. 799fc20 [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 bb5d324 [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 53faef1 [Marcelo Vanzin] Merge branch 'master' into SPARK-4924 a7936ef [Marcelo Vanzin] Fix pyspark tests. 656374e [Marcelo Vanzin] Mima fixes. 4d511e7 [Marcelo Vanzin] Fix tools search code. 7a01e4a [Marcelo Vanzin] Fix pyspark on Yarn. 1b3f6e9 [Marcelo Vanzin] Call SparkSubmit from spark-class launcher for unknown classes. 25c5ae6 [Marcelo Vanzin] Centralize SparkSubmit command line parsing. 27be98a [Marcelo Vanzin] Modify Spark to use launcher lib. 6f70eea [Marcelo Vanzin] [SPARK-4924] Add a library for launching Spark jobs programatically. --- .gitignore | 1 + bin/compute-classpath.cmd | 124 ------ bin/compute-classpath.sh | 161 -------- bin/load-spark-env.sh | 8 +- bin/pyspark | 59 +-- bin/pyspark2.cmd | 57 +-- bin/run-example | 2 +- bin/spark-class | 180 ++------- bin/spark-class2.cmd | 141 ++----- bin/spark-shell | 23 +- bin/spark-shell2.cmd | 27 +- bin/spark-sql | 20 +- bin/spark-submit | 66 +--- bin/spark-submit2.cmd | 71 +--- bin/utils.sh | 60 --- bin/windows-utils.cmd | 60 --- core/pom.xml | 5 + .../spark/deploy/SparkSubmitArguments.scala | 157 ++++---- .../SparkSubmitDriverBootstrapper.scala | 170 -------- .../spark/deploy/worker/CommandUtils.scala | 46 +-- .../org/apache/spark/executor/Executor.scala | 1 - .../launcher/SparkSubmitArgumentsParser.scala | 25 ++ .../spark/launcher/WorkerCommandBuilder.scala | 50 +++ docs/programming-guide.md | 5 + launcher/pom.xml | 83 ++++ .../launcher/AbstractCommandBuilder.java | 362 ++++++++++++++++++ .../spark/launcher/CommandBuilderUtils.java | 296 ++++++++++++++ .../java/org/apache/spark/launcher/Main.java | 173 +++++++++ .../launcher/SparkClassCommandBuilder.java | 108 ++++++ .../apache/spark/launcher/SparkLauncher.java | 279 ++++++++++++++ .../launcher/SparkSubmitCommandBuilder.java | 327 ++++++++++++++++ .../launcher/SparkSubmitOptionParser.java | 224 +++++++++++ .../apache/spark/launcher/package-info.java | 45 +++ .../launcher/CommandBuilderUtilsSuite.java | 101 +++++ .../spark/launcher/SparkLauncherSuite.java | 94 +++++ .../SparkSubmitCommandBuilderSuite.java | 278 ++++++++++++++ .../SparkSubmitOptionParserSuite.java | 108 ++++++ launcher/src/test/resources/log4j.properties | 31 ++ make-distribution.sh | 2 + pom.xml | 3 +- project/SparkBuild.scala | 7 +- python/pyspark/java_gateway.py | 3 +- sbin/spark-daemon.sh | 84 ++-- sbin/start-thriftserver.sh | 2 +- 44 files changed, 2891 insertions(+), 1238 deletions(-) delete mode 100644 bin/compute-classpath.cmd delete mode 100755 bin/compute-classpath.sh delete mode 100755 bin/utils.sh delete mode 100644 bin/windows-utils.cmd delete mode 100644 core/src/main/scala/org/apache/spark/deploy/SparkSubmitDriverBootstrapper.scala create mode 100644 core/src/main/scala/org/apache/spark/launcher/SparkSubmitArgumentsParser.scala create mode 100644 core/src/main/scala/org/apache/spark/launcher/WorkerCommandBuilder.scala create mode 100644 launcher/pom.xml create mode 100644 launcher/src/main/java/org/apache/spark/launcher/AbstractCommandBuilder.java create mode 100644 launcher/src/main/java/org/apache/spark/launcher/CommandBuilderUtils.java create mode 100644 launcher/src/main/java/org/apache/spark/launcher/Main.java create mode 100644 launcher/src/main/java/org/apache/spark/launcher/SparkClassCommandBuilder.java create mode 100644 launcher/src/main/java/org/apache/spark/launcher/SparkLauncher.java create mode 100644 launcher/src/main/java/org/apache/spark/launcher/SparkSubmitCommandBuilder.java create mode 100644 launcher/src/main/java/org/apache/spark/launcher/SparkSubmitOptionParser.java create mode 100644 launcher/src/main/java/org/apache/spark/launcher/package-info.java create mode 100644 launcher/src/test/java/org/apache/spark/launcher/CommandBuilderUtilsSuite.java create mode 100644 launcher/src/test/java/org/apache/spark/launcher/SparkLauncherSuite.java create mode 100644 launcher/src/test/java/org/apache/spark/launcher/SparkSubmitCommandBuilderSuite.java create mode 100644 launcher/src/test/java/org/apache/spark/launcher/SparkSubmitOptionParserSuite.java create mode 100644 launcher/src/test/resources/log4j.properties diff --git a/.gitignore b/.gitignore index 9757054a50f9e..d162fa9cca994 100644 --- a/.gitignore +++ b/.gitignore @@ -6,6 +6,7 @@ *.iml *.iws *.pyc +*.pyo .idea/ .idea_modules/ build/*.jar diff --git a/bin/compute-classpath.cmd b/bin/compute-classpath.cmd deleted file mode 100644 index 088f993954d9e..0000000000000 --- a/bin/compute-classpath.cmd +++ /dev/null @@ -1,124 +0,0 @@ -@echo off - -rem -rem Licensed to the Apache Software Foundation (ASF) under one or more -rem contributor license agreements. See the NOTICE file distributed with -rem this work for additional information regarding copyright ownership. -rem The ASF licenses this file to You under the Apache License, Version 2.0 -rem (the "License"); you may not use this file except in compliance with -rem the License. You may obtain a copy of the License at -rem -rem http://www.apache.org/licenses/LICENSE-2.0 -rem -rem Unless required by applicable law or agreed to in writing, software -rem distributed under the License is distributed on an "AS IS" BASIS, -rem WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -rem See the License for the specific language governing permissions and -rem limitations under the License. -rem - -rem This script computes Spark's classpath and prints it to stdout; it's used by both the "run" -rem script and the ExecutorRunner in standalone cluster mode. - -rem If we're called from spark-class2.cmd, it already set enabledelayedexpansion and setting -rem it here would stop us from affecting its copy of the CLASSPATH variable; otherwise we -rem need to set it here because we use !datanucleus_jars! below. -if "%DONT_PRINT_CLASSPATH%"=="1" goto skip_delayed_expansion -setlocal enabledelayedexpansion -:skip_delayed_expansion - -set SCALA_VERSION=2.10 - -rem Figure out where the Spark framework is installed -set FWDIR=%~dp0..\ - -rem Load environment variables from conf\spark-env.cmd, if it exists -if exist "%FWDIR%conf\spark-env.cmd" call "%FWDIR%conf\spark-env.cmd" - -rem Build up classpath -set CLASSPATH=%SPARK_CLASSPATH%;%SPARK_SUBMIT_CLASSPATH% - -if not "x%SPARK_CONF_DIR%"=="x" ( - set CLASSPATH=%CLASSPATH%;%SPARK_CONF_DIR% -) else ( - set CLASSPATH=%CLASSPATH%;%FWDIR%conf -) - -if exist "%FWDIR%RELEASE" ( - for %%d in ("%FWDIR%lib\spark-assembly*.jar") do ( - set ASSEMBLY_JAR=%%d - ) -) else ( - for %%d in ("%FWDIR%assembly\target\scala-%SCALA_VERSION%\spark-assembly*hadoop*.jar") do ( - set ASSEMBLY_JAR=%%d - ) -) - -set CLASSPATH=%CLASSPATH%;%ASSEMBLY_JAR% - -rem When Hive support is needed, Datanucleus jars must be included on the classpath. -rem Datanucleus jars do not work if only included in the uber jar as plugin.xml metadata is lost. -rem Both sbt and maven will populate "lib_managed/jars/" with the datanucleus jars when Spark is -rem built with Hive, so look for them there. -if exist "%FWDIR%RELEASE" ( - set datanucleus_dir=%FWDIR%lib -) else ( - set datanucleus_dir=%FWDIR%lib_managed\jars -) -set "datanucleus_jars=" -for %%d in ("%datanucleus_dir%\datanucleus-*.jar") do ( - set datanucleus_jars=!datanucleus_jars!;%%d -) -set CLASSPATH=%CLASSPATH%;%datanucleus_jars% - -set SPARK_CLASSES=%FWDIR%core\target\scala-%SCALA_VERSION%\classes -set SPARK_CLASSES=%SPARK_CLASSES%;%FWDIR%repl\target\scala-%SCALA_VERSION%\classes -set SPARK_CLASSES=%SPARK_CLASSES%;%FWDIR%mllib\target\scala-%SCALA_VERSION%\classes -set SPARK_CLASSES=%SPARK_CLASSES%;%FWDIR%bagel\target\scala-%SCALA_VERSION%\classes -set SPARK_CLASSES=%SPARK_CLASSES%;%FWDIR%graphx\target\scala-%SCALA_VERSION%\classes -set SPARK_CLASSES=%SPARK_CLASSES%;%FWDIR%streaming\target\scala-%SCALA_VERSION%\classes -set SPARK_CLASSES=%SPARK_CLASSES%;%FWDIR%tools\target\scala-%SCALA_VERSION%\classes -set SPARK_CLASSES=%SPARK_CLASSES%;%FWDIR%sql\catalyst\target\scala-%SCALA_VERSION%\classes -set SPARK_CLASSES=%SPARK_CLASSES%;%FWDIR%sql\core\target\scala-%SCALA_VERSION%\classes -set SPARK_CLASSES=%SPARK_CLASSES%;%FWDIR%sql\hive\target\scala-%SCALA_VERSION%\classes - -set SPARK_TEST_CLASSES=%FWDIR%core\target\scala-%SCALA_VERSION%\test-classes -set SPARK_TEST_CLASSES=%SPARK_TEST_CLASSES%;%FWDIR%repl\target\scala-%SCALA_VERSION%\test-classes -set SPARK_TEST_CLASSES=%SPARK_TEST_CLASSES%;%FWDIR%mllib\target\scala-%SCALA_VERSION%\test-classes -set SPARK_TEST_CLASSES=%SPARK_TEST_CLASSES%;%FWDIR%bagel\target\scala-%SCALA_VERSION%\test-classes -set SPARK_TEST_CLASSES=%SPARK_TEST_CLASSES%;%FWDIR%graphx\target\scala-%SCALA_VERSION%\test-classes -set SPARK_TEST_CLASSES=%SPARK_TEST_CLASSES%;%FWDIR%streaming\target\scala-%SCALA_VERSION%\test-classes -set SPARK_TEST_CLASSES=%SPARK_TEST_CLASSES%;%FWDIR%sql\catalyst\target\scala-%SCALA_VERSION%\test-classes -set SPARK_TEST_CLASSES=%SPARK_TEST_CLASSES%;%FWDIR%sql\core\target\scala-%SCALA_VERSION%\test-classes -set SPARK_TEST_CLASSES=%SPARK_TEST_CLASSES%;%FWDIR%sql\hive\target\scala-%SCALA_VERSION%\test-classes - -if "x%SPARK_TESTING%"=="x1" ( - rem Add test clases to path - note, add SPARK_CLASSES and SPARK_TEST_CLASSES before CLASSPATH - rem so that local compilation takes precedence over assembled jar - set CLASSPATH=%SPARK_CLASSES%;%SPARK_TEST_CLASSES%;%CLASSPATH% -) - -rem Add hadoop conf dir - else FileSystem.*, etc fail -rem Note, this assumes that there is either a HADOOP_CONF_DIR or YARN_CONF_DIR which hosts -rem the configurtion files. -if "x%HADOOP_CONF_DIR%"=="x" goto no_hadoop_conf_dir - set CLASSPATH=%CLASSPATH%;%HADOOP_CONF_DIR% -:no_hadoop_conf_dir - -if "x%YARN_CONF_DIR%"=="x" goto no_yarn_conf_dir - set CLASSPATH=%CLASSPATH%;%YARN_CONF_DIR% -:no_yarn_conf_dir - -rem To allow for distributions to append needed libraries to the classpath (e.g. when -rem using the "hadoop-provided" profile to build Spark), check SPARK_DIST_CLASSPATH and -rem append it to tbe final classpath. -if not "x%$SPARK_DIST_CLASSPATH%"=="x" ( - set CLASSPATH=%CLASSPATH%;%SPARK_DIST_CLASSPATH% -) - -rem A bit of a hack to allow calling this script within run2.cmd without seeing output -if "%DONT_PRINT_CLASSPATH%"=="1" goto exit - -echo %CLASSPATH% - -:exit diff --git a/bin/compute-classpath.sh b/bin/compute-classpath.sh deleted file mode 100755 index f4f6b7b909490..0000000000000 --- a/bin/compute-classpath.sh +++ /dev/null @@ -1,161 +0,0 @@ -#!/usr/bin/env bash - -# -# Licensed to the Apache Software Foundation (ASF) under one or more -# contributor license agreements. See the NOTICE file distributed with -# this work for additional information regarding copyright ownership. -# The ASF licenses this file to You under the Apache License, Version 2.0 -# (the "License"); you may not use this file except in compliance with -# the License. You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# - -# This script computes Spark's classpath and prints it to stdout; it's used by both the "run" -# script and the ExecutorRunner in standalone cluster mode. - -# Figure out where Spark is installed -FWDIR="$(cd "`dirname "$0"`"/..; pwd)" - -. "$FWDIR"/bin/load-spark-env.sh - -if [ -n "$SPARK_CLASSPATH" ]; then - CLASSPATH="$SPARK_CLASSPATH:$SPARK_SUBMIT_CLASSPATH" -else - CLASSPATH="$SPARK_SUBMIT_CLASSPATH" -fi - -# Build up classpath -if [ -n "$SPARK_CONF_DIR" ]; then - CLASSPATH="$CLASSPATH:$SPARK_CONF_DIR" -else - CLASSPATH="$CLASSPATH:$FWDIR/conf" -fi - -ASSEMBLY_DIR="$FWDIR/assembly/target/scala-$SPARK_SCALA_VERSION" - -if [ -n "$JAVA_HOME" ]; then - JAR_CMD="$JAVA_HOME/bin/jar" -else - JAR_CMD="jar" -fi - -# A developer option to prepend more recently compiled Spark classes -if [ -n "$SPARK_PREPEND_CLASSES" ]; then - echo "NOTE: SPARK_PREPEND_CLASSES is set, placing locally compiled Spark"\ - "classes ahead of assembly." >&2 - # Spark classes - CLASSPATH="$CLASSPATH:$FWDIR/core/target/scala-$SPARK_SCALA_VERSION/classes" - CLASSPATH="$CLASSPATH:$FWDIR/repl/target/scala-$SPARK_SCALA_VERSION/classes" - CLASSPATH="$CLASSPATH:$FWDIR/mllib/target/scala-$SPARK_SCALA_VERSION/classes" - CLASSPATH="$CLASSPATH:$FWDIR/bagel/target/scala-$SPARK_SCALA_VERSION/classes" - CLASSPATH="$CLASSPATH:$FWDIR/graphx/target/scala-$SPARK_SCALA_VERSION/classes" - CLASSPATH="$CLASSPATH:$FWDIR/streaming/target/scala-$SPARK_SCALA_VERSION/classes" - CLASSPATH="$CLASSPATH:$FWDIR/tools/target/scala-$SPARK_SCALA_VERSION/classes" - CLASSPATH="$CLASSPATH:$FWDIR/sql/catalyst/target/scala-$SPARK_SCALA_VERSION/classes" - CLASSPATH="$CLASSPATH:$FWDIR/sql/core/target/scala-$SPARK_SCALA_VERSION/classes" - CLASSPATH="$CLASSPATH:$FWDIR/sql/hive/target/scala-$SPARK_SCALA_VERSION/classes" - CLASSPATH="$CLASSPATH:$FWDIR/sql/hive-thriftserver/target/scala-$SPARK_SCALA_VERSION/classes" - CLASSPATH="$CLASSPATH:$FWDIR/yarn/stable/target/scala-$SPARK_SCALA_VERSION/classes" - # Jars for shaded deps in their original form (copied here during build) - CLASSPATH="$CLASSPATH:$FWDIR/core/target/jars/*" -fi - -# Use spark-assembly jar from either RELEASE or assembly directory -if [ -f "$FWDIR/RELEASE" ]; then - assembly_folder="$FWDIR"/lib -else - assembly_folder="$ASSEMBLY_DIR" -fi - -num_jars=0 - -for f in "${assembly_folder}"/spark-assembly*hadoop*.jar; do - if [[ ! -e "$f" ]]; then - echo "Failed to find Spark assembly in $assembly_folder" 1>&2 - echo "You need to build Spark before running this program." 1>&2 - exit 1 - fi - ASSEMBLY_JAR="$f" - num_jars=$((num_jars+1)) -done - -if [ "$num_jars" -gt "1" ]; then - echo "Found multiple Spark assembly jars in $assembly_folder:" 1>&2 - ls "${assembly_folder}"/spark-assembly*hadoop*.jar 1>&2 - echo "Please remove all but one jar." 1>&2 - exit 1 -fi - -# Verify that versions of java used to build the jars and run Spark are compatible -jar_error_check=$("$JAR_CMD" -tf "$ASSEMBLY_JAR" nonexistent/class/path 2>&1) -if [[ "$jar_error_check" =~ "invalid CEN header" ]]; then - echo "Loading Spark jar with '$JAR_CMD' failed. " 1>&2 - echo "This is likely because Spark was compiled with Java 7 and run " 1>&2 - echo "with Java 6. (see SPARK-1703). Please use Java 7 to run Spark " 1>&2 - echo "or build Spark with Java 6." 1>&2 - exit 1 -fi - -CLASSPATH="$CLASSPATH:$ASSEMBLY_JAR" - -# When Hive support is needed, Datanucleus jars must be included on the classpath. -# Datanucleus jars do not work if only included in the uber jar as plugin.xml metadata is lost. -# Both sbt and maven will populate "lib_managed/jars/" with the datanucleus jars when Spark is -# built with Hive, so first check if the datanucleus jars exist, and then ensure the current Spark -# assembly is built for Hive, before actually populating the CLASSPATH with the jars. -# Note that this check order is faster (by up to half a second) in the case where Hive is not used. -if [ -f "$FWDIR/RELEASE" ]; then - datanucleus_dir="$FWDIR"/lib -else - datanucleus_dir="$FWDIR"/lib_managed/jars -fi - -datanucleus_jars="$(find "$datanucleus_dir" 2>/dev/null | grep "datanucleus-.*\\.jar$")" -datanucleus_jars="$(echo "$datanucleus_jars" | tr "\n" : | sed s/:$//g)" - -if [ -n "$datanucleus_jars" ]; then - hive_files=$("$JAR_CMD" -tf "$ASSEMBLY_JAR" org/apache/hadoop/hive/ql/exec 2>/dev/null) - if [ -n "$hive_files" ]; then - echo "Spark assembly has been built with Hive, including Datanucleus jars on classpath" 1>&2 - CLASSPATH="$CLASSPATH:$datanucleus_jars" - fi -fi - -# Add test classes if we're running from SBT or Maven with SPARK_TESTING set to 1 -if [[ $SPARK_TESTING == 1 ]]; then - CLASSPATH="$CLASSPATH:$FWDIR/core/target/scala-$SPARK_SCALA_VERSION/test-classes" - CLASSPATH="$CLASSPATH:$FWDIR/repl/target/scala-$SPARK_SCALA_VERSION/test-classes" - CLASSPATH="$CLASSPATH:$FWDIR/mllib/target/scala-$SPARK_SCALA_VERSION/test-classes" - CLASSPATH="$CLASSPATH:$FWDIR/bagel/target/scala-$SPARK_SCALA_VERSION/test-classes" - CLASSPATH="$CLASSPATH:$FWDIR/graphx/target/scala-$SPARK_SCALA_VERSION/test-classes" - CLASSPATH="$CLASSPATH:$FWDIR/streaming/target/scala-$SPARK_SCALA_VERSION/test-classes" - CLASSPATH="$CLASSPATH:$FWDIR/sql/catalyst/target/scala-$SPARK_SCALA_VERSION/test-classes" - CLASSPATH="$CLASSPATH:$FWDIR/sql/core/target/scala-$SPARK_SCALA_VERSION/test-classes" - CLASSPATH="$CLASSPATH:$FWDIR/sql/hive/target/scala-$SPARK_SCALA_VERSION/test-classes" -fi - -# Add hadoop conf dir if given -- otherwise FileSystem.*, etc fail ! -# Note, this assumes that there is either a HADOOP_CONF_DIR or YARN_CONF_DIR which hosts -# the configurtion files. -if [ -n "$HADOOP_CONF_DIR" ]; then - CLASSPATH="$CLASSPATH:$HADOOP_CONF_DIR" -fi -if [ -n "$YARN_CONF_DIR" ]; then - CLASSPATH="$CLASSPATH:$YARN_CONF_DIR" -fi - -# To allow for distributions to append needed libraries to the classpath (e.g. when -# using the "hadoop-provided" profile to build Spark), check SPARK_DIST_CLASSPATH and -# append it to tbe final classpath. -if [ -n "$SPARK_DIST_CLASSPATH" ]; then - CLASSPATH="$CLASSPATH:$SPARK_DIST_CLASSPATH" -fi - -echo "$CLASSPATH" diff --git a/bin/load-spark-env.sh b/bin/load-spark-env.sh index 356b3d49b2ffe..2d7070c25d328 100644 --- a/bin/load-spark-env.sh +++ b/bin/load-spark-env.sh @@ -41,9 +41,9 @@ fi if [ -z "$SPARK_SCALA_VERSION" ]; then - ASSEMBLY_DIR2="$FWDIR/assembly/target/scala-2.11" - ASSEMBLY_DIR1="$FWDIR/assembly/target/scala-2.10" - + ASSEMBLY_DIR2="$SPARK_HOME/assembly/target/scala-2.11" + ASSEMBLY_DIR1="$SPARK_HOME/assembly/target/scala-2.10" + if [[ -d "$ASSEMBLY_DIR2" && -d "$ASSEMBLY_DIR1" ]]; then echo -e "Presence of build for both scala versions(SCALA 2.10 and SCALA 2.11) detected." 1>&2 echo -e 'Either clean one of them or, export SPARK_SCALA_VERSION=2.11 in spark-env.sh.' 1>&2 @@ -54,5 +54,5 @@ if [ -z "$SPARK_SCALA_VERSION" ]; then export SPARK_SCALA_VERSION="2.11" else export SPARK_SCALA_VERSION="2.10" - fi + fi fi diff --git a/bin/pyspark b/bin/pyspark index 0b4f695dd06dd..e7f6a1a072c2a 100755 --- a/bin/pyspark +++ b/bin/pyspark @@ -18,36 +18,24 @@ # # Figure out where Spark is installed -FWDIR="$(cd "`dirname "$0"`"/..; pwd)" +export SPARK_HOME="$(cd "`dirname "$0"`"/..; pwd)" -# Export this as SPARK_HOME -export SPARK_HOME="$FWDIR" - -source "$FWDIR/bin/utils.sh" - -source "$FWDIR"/bin/load-spark-env.sh +source "$SPARK_HOME"/bin/load-spark-env.sh function usage() { + if [ -n "$1" ]; then + echo $1 + fi echo "Usage: ./bin/pyspark [options]" 1>&2 - "$FWDIR"/bin/spark-submit --help 2>&1 | grep -v Usage 1>&2 - exit 0 + "$SPARK_HOME"/bin/spark-submit --help 2>&1 | grep -v Usage 1>&2 + exit $2 } +export -f usage if [[ "$@" = *--help ]] || [[ "$@" = *-h ]]; then usage fi -# Exit if the user hasn't compiled Spark -if [ ! -f "$FWDIR/RELEASE" ]; then - # Exit if the user hasn't compiled Spark - ls "$FWDIR"/assembly/target/scala-$SPARK_SCALA_VERSION/spark-assembly*hadoop*.jar >& /dev/null - if [[ $? != 0 ]]; then - echo "Failed to find Spark assembly in $FWDIR/assembly/target" 1>&2 - echo "You need to build Spark before running this program" 1>&2 - exit 1 - fi -fi - # In Spark <= 1.1, setting IPYTHON=1 would cause the driver to be launched using the `ipython` # executable, while the worker would still be launched using PYSPARK_PYTHON. # @@ -95,26 +83,13 @@ export PYTHONPATH="$SPARK_HOME/python/lib/py4j-0.8.2.1-src.zip:$PYTHONPATH" # Load the PySpark shell.py script when ./pyspark is used interactively: export OLD_PYTHONSTARTUP="$PYTHONSTARTUP" -export PYTHONSTARTUP="$FWDIR/python/pyspark/shell.py" - -# Build up arguments list manually to preserve quotes and backslashes. -# We export Spark submit arguments as an environment variable because shell.py must run as a -# PYTHONSTARTUP script, which does not take in arguments. This is required for IPython notebooks. -SUBMIT_USAGE_FUNCTION=usage -gatherSparkSubmitOpts "$@" -PYSPARK_SUBMIT_ARGS="" -whitespace="[[:space:]]" -for i in "${SUBMISSION_OPTS[@]}"; do - if [[ $i =~ \" ]]; then i=$(echo $i | sed 's/\"/\\\"/g'); fi - if [[ $i =~ $whitespace ]]; then i=\"$i\"; fi - PYSPARK_SUBMIT_ARGS="$PYSPARK_SUBMIT_ARGS $i" -done -export PYSPARK_SUBMIT_ARGS +export PYTHONSTARTUP="$SPARK_HOME/python/pyspark/shell.py" # For pyspark tests if [[ -n "$SPARK_TESTING" ]]; then unset YARN_CONF_DIR unset HADOOP_CONF_DIR + export PYSPARK_SUBMIT_ARGS=pyspark-shell if [[ -n "$PYSPARK_DOC_TEST" ]]; then exec "$PYSPARK_DRIVER_PYTHON" -m doctest $1 else @@ -123,14 +98,6 @@ if [[ -n "$SPARK_TESTING" ]]; then exit fi -# If a python file is provided, directly run spark-submit. -if [[ "$1" =~ \.py$ ]]; then - echo -e "\nWARNING: Running python applications through ./bin/pyspark is deprecated as of Spark 1.0." 1>&2 - echo -e "Use ./bin/spark-submit \n" 1>&2 - primary="$1" - shift - gatherSparkSubmitOpts "$@" - exec "$FWDIR"/bin/spark-submit "${SUBMISSION_OPTS[@]}" "$primary" "${APPLICATION_OPTS[@]}" -else - exec "$PYSPARK_DRIVER_PYTHON" $PYSPARK_DRIVER_PYTHON_OPTS -fi +export PYSPARK_DRIVER_PYTHON +export PYSPARK_DRIVER_PYTHON_OPTS +exec "$SPARK_HOME"/bin/spark-submit pyspark-shell-main "$@" diff --git a/bin/pyspark2.cmd b/bin/pyspark2.cmd index a542ec80b49d6..4f5eb5e20614d 100644 --- a/bin/pyspark2.cmd +++ b/bin/pyspark2.cmd @@ -17,59 +17,22 @@ rem See the License for the specific language governing permissions and rem limitations under the License. rem -set SCALA_VERSION=2.10 - rem Figure out where the Spark framework is installed -set FWDIR=%~dp0..\ - -rem Export this as SPARK_HOME -set SPARK_HOME=%FWDIR% - -rem Test whether the user has built Spark -if exist "%FWDIR%RELEASE" goto skip_build_test -set FOUND_JAR=0 -for %%d in ("%FWDIR%assembly\target\scala-%SCALA_VERSION%\spark-assembly*hadoop*.jar") do ( - set FOUND_JAR=1 -) -if [%FOUND_JAR%] == [0] ( - echo Failed to find Spark assembly JAR. - echo You need to build Spark before running this program. - goto exit -) -:skip_build_test +set SPARK_HOME=%~dp0.. rem Load environment variables from conf\spark-env.cmd, if it exists -if exist "%FWDIR%conf\spark-env.cmd" call "%FWDIR%conf\spark-env.cmd" +if exist "%SPARK_HOME%\conf\spark-env.cmd" call "%SPARK_HOME%\conf\spark-env.cmd" rem Figure out which Python to use. -if [%PYSPARK_PYTHON%] == [] set PYSPARK_PYTHON=python +if "x%PYSPARK_DRIVER_PYTHON%"=="x" ( + set PYSPARK_DRIVER_PYTHON=python + if not [%PYSPARK_PYTHON%] == [] set PYSPARK_DRIVER_PYTHON=%PYSPARK_PYTHON% +) -set PYTHONPATH=%FWDIR%python;%PYTHONPATH% -set PYTHONPATH=%FWDIR%python\lib\py4j-0.8.2.1-src.zip;%PYTHONPATH% +set PYTHONPATH=%SPARK_HOME%\python;%PYTHONPATH% +set PYTHONPATH=%SPARK_HOME%\python\lib\py4j-0.8.2.1-src.zip;%PYTHONPATH% set OLD_PYTHONSTARTUP=%PYTHONSTARTUP% -set PYTHONSTARTUP=%FWDIR%python\pyspark\shell.py -set PYSPARK_SUBMIT_ARGS=%* - -echo Running %PYSPARK_PYTHON% with PYTHONPATH=%PYTHONPATH% - -rem Check whether the argument is a file -for /f %%i in ('echo %1^| findstr /R "\.py"') do ( - set PYTHON_FILE=%%i -) - -if [%PYTHON_FILE%] == [] ( - if [%IPYTHON%] == [1] ( - ipython %IPYTHON_OPTS% - ) else ( - %PYSPARK_PYTHON% - ) -) else ( - echo. - echo WARNING: Running python applications through ./bin/pyspark.cmd is deprecated as of Spark 1.0. - echo Use ./bin/spark-submit ^ - echo. - "%FWDIR%\bin\spark-submit.cmd" %PYSPARK_SUBMIT_ARGS% -) +set PYTHONSTARTUP=%SPARK_HOME%\python\pyspark\shell.py -:exit +call %SPARK_HOME%\bin\spark-submit2.cmd pyspark-shell-main %* diff --git a/bin/run-example b/bin/run-example index a106411392e06..798e2caeb88ce 100755 --- a/bin/run-example +++ b/bin/run-example @@ -67,7 +67,7 @@ if [[ ! $EXAMPLE_CLASS == org.apache.spark.examples* ]]; then EXAMPLE_CLASS="org.apache.spark.examples.$EXAMPLE_CLASS" fi -"$FWDIR"/bin/spark-submit \ +exec "$FWDIR"/bin/spark-submit \ --master $EXAMPLE_MASTER \ --class $EXAMPLE_CLASS \ "$SPARK_EXAMPLES_JAR" \ diff --git a/bin/spark-class b/bin/spark-class index 2f0441bb3c1c2..e29b234afaf96 100755 --- a/bin/spark-class +++ b/bin/spark-class @@ -16,89 +16,18 @@ # See the License for the specific language governing permissions and # limitations under the License. # - -# NOTE: Any changes to this file must be reflected in SparkSubmitDriverBootstrapper.scala! - -cygwin=false -case "`uname`" in - CYGWIN*) cygwin=true;; -esac +set -e # Figure out where Spark is installed -FWDIR="$(cd "`dirname "$0"`"/..; pwd)" +export SPARK_HOME="$(cd "`dirname "$0"`"/..; pwd)" -# Export this as SPARK_HOME -export SPARK_HOME="$FWDIR" -export SPARK_CONF_DIR="${SPARK_CONF_DIR:-"$SPARK_HOME/conf"}" - -. "$FWDIR"/bin/load-spark-env.sh +. "$SPARK_HOME"/bin/load-spark-env.sh if [ -z "$1" ]; then echo "Usage: spark-class []" 1>&2 exit 1 fi -if [ -n "$SPARK_MEM" ]; then - echo -e "Warning: SPARK_MEM is deprecated, please use a more specific config option" 1>&2 - echo -e "(e.g., spark.executor.memory or spark.driver.memory)." 1>&2 -fi - -# Use SPARK_MEM or 512m as the default memory, to be overridden by specific options -DEFAULT_MEM=${SPARK_MEM:-512m} - -SPARK_DAEMON_JAVA_OPTS="$SPARK_DAEMON_JAVA_OPTS -Dspark.akka.logLifecycleEvents=true" - -# Add java opts and memory settings for master, worker, history server, executors, and repl. -case "$1" in - # Master, Worker, and HistoryServer use SPARK_DAEMON_JAVA_OPTS (and specific opts) + SPARK_DAEMON_MEMORY. - 'org.apache.spark.deploy.master.Master') - OUR_JAVA_OPTS="$SPARK_DAEMON_JAVA_OPTS $SPARK_MASTER_OPTS" - OUR_JAVA_MEM=${SPARK_DAEMON_MEMORY:-$DEFAULT_MEM} - ;; - 'org.apache.spark.deploy.worker.Worker') - OUR_JAVA_OPTS="$SPARK_DAEMON_JAVA_OPTS $SPARK_WORKER_OPTS" - OUR_JAVA_MEM=${SPARK_DAEMON_MEMORY:-$DEFAULT_MEM} - ;; - 'org.apache.spark.deploy.history.HistoryServer') - OUR_JAVA_OPTS="$SPARK_DAEMON_JAVA_OPTS $SPARK_HISTORY_OPTS" - OUR_JAVA_MEM=${SPARK_DAEMON_MEMORY:-$DEFAULT_MEM} - ;; - - # Executors use SPARK_JAVA_OPTS + SPARK_EXECUTOR_MEMORY. - 'org.apache.spark.executor.CoarseGrainedExecutorBackend') - OUR_JAVA_OPTS="$SPARK_JAVA_OPTS $SPARK_EXECUTOR_OPTS" - OUR_JAVA_MEM=${SPARK_EXECUTOR_MEMORY:-$DEFAULT_MEM} - ;; - 'org.apache.spark.executor.MesosExecutorBackend') - OUR_JAVA_OPTS="$SPARK_JAVA_OPTS $SPARK_EXECUTOR_OPTS" - OUR_JAVA_MEM=${SPARK_EXECUTOR_MEMORY:-$DEFAULT_MEM} - export PYTHONPATH="$FWDIR/python:$PYTHONPATH" - export PYTHONPATH="$FWDIR/python/lib/py4j-0.8.2.1-src.zip:$PYTHONPATH" - ;; - - # Spark submit uses SPARK_JAVA_OPTS + SPARK_SUBMIT_OPTS + - # SPARK_DRIVER_MEMORY + SPARK_SUBMIT_DRIVER_MEMORY. - 'org.apache.spark.deploy.SparkSubmit') - OUR_JAVA_OPTS="$SPARK_JAVA_OPTS $SPARK_SUBMIT_OPTS" - OUR_JAVA_MEM=${SPARK_DRIVER_MEMORY:-$DEFAULT_MEM} - if [ -n "$SPARK_SUBMIT_LIBRARY_PATH" ]; then - if [[ $OSTYPE == darwin* ]]; then - export DYLD_LIBRARY_PATH="$SPARK_SUBMIT_LIBRARY_PATH:$DYLD_LIBRARY_PATH" - else - export LD_LIBRARY_PATH="$SPARK_SUBMIT_LIBRARY_PATH:$LD_LIBRARY_PATH" - fi - fi - if [ -n "$SPARK_SUBMIT_DRIVER_MEMORY" ]; then - OUR_JAVA_MEM="$SPARK_SUBMIT_DRIVER_MEMORY" - fi - ;; - - *) - OUR_JAVA_OPTS="$SPARK_JAVA_OPTS" - OUR_JAVA_MEM=${SPARK_DRIVER_MEMORY:-$DEFAULT_MEM} - ;; -esac - # Find the java binary if [ -n "${JAVA_HOME}" ]; then RUNNER="${JAVA_HOME}/bin/java" @@ -110,83 +39,48 @@ else exit 1 fi fi -JAVA_VERSION=$("$RUNNER" -version 2>&1 | grep 'version' | sed 's/.* version "\(.*\)\.\(.*\)\..*"/\1\2/; 1q') - -# Set JAVA_OPTS to be able to load native libraries and to set heap size -if [ "$JAVA_VERSION" -ge 18 ]; then - JAVA_OPTS="$OUR_JAVA_OPTS" -else - JAVA_OPTS="-XX:MaxPermSize=128m $OUR_JAVA_OPTS" -fi -JAVA_OPTS="$JAVA_OPTS -Xms$OUR_JAVA_MEM -Xmx$OUR_JAVA_MEM" - -# Load extra JAVA_OPTS from conf/java-opts, if it exists -if [ -e "$SPARK_CONF_DIR/java-opts" ] ; then - JAVA_OPTS="$JAVA_OPTS `cat "$SPARK_CONF_DIR"/java-opts`" -fi - -# Attention: when changing the way the JAVA_OPTS are assembled, the change must be reflected in CommandUtils.scala! - -TOOLS_DIR="$FWDIR"/tools -SPARK_TOOLS_JAR="" -if [ -e "$TOOLS_DIR"/target/scala-$SPARK_SCALA_VERSION/spark-tools*[0-9Tg].jar ]; then - # Use the JAR from the SBT build - export SPARK_TOOLS_JAR="`ls "$TOOLS_DIR"/target/scala-$SPARK_SCALA_VERSION/spark-tools*[0-9Tg].jar`" -fi -if [ -e "$TOOLS_DIR"/target/spark-tools*[0-9Tg].jar ]; then - # Use the JAR from the Maven build - # TODO: this also needs to become an assembly! - export SPARK_TOOLS_JAR="`ls "$TOOLS_DIR"/target/spark-tools*[0-9Tg].jar`" -fi -# Compute classpath using external script -classpath_output=$("$FWDIR"/bin/compute-classpath.sh) -if [[ "$?" != "0" ]]; then - echo "$classpath_output" - exit 1 -else - CLASSPATH="$classpath_output" -fi +# Look for the launcher. In non-release mode, add the compiled classes directly to the classpath +# instead of looking for a jar file. +SPARK_LAUNCHER_CP= +if [ -f $SPARK_HOME/RELEASE ]; then + LAUNCHER_DIR="$SPARK_HOME/lib" + num_jars="$(ls -1 "$LAUNCHER_DIR" | grep "^spark-launcher.*\.jar$" | wc -l)" + if [ "$num_jars" -eq "0" -a -z "$SPARK_LAUNCHER_CP" ]; then + echo "Failed to find Spark launcher in $LAUNCHER_DIR." 1>&2 + echo "You need to build Spark before running this program." 1>&2 + exit 1 + fi -if [[ "$1" =~ org.apache.spark.tools.* ]]; then - if test -z "$SPARK_TOOLS_JAR"; then - echo "Failed to find Spark Tools Jar in $FWDIR/tools/target/scala-$SPARK_SCALA_VERSION/" 1>&2 - echo "You need to run \"build/sbt tools/package\" before running $1." 1>&2 + LAUNCHER_JARS="$(ls -1 "$LAUNCHER_DIR" | grep "^spark-launcher.*\.jar$" || true)" + if [ "$num_jars" -gt "1" ]; then + echo "Found multiple Spark launcher jars in $LAUNCHER_DIR:" 1>&2 + echo "$LAUNCHER_JARS" 1>&2 + echo "Please remove all but one jar." 1>&2 exit 1 fi - CLASSPATH="$CLASSPATH:$SPARK_TOOLS_JAR" -fi -if $cygwin; then - CLASSPATH="`cygpath -wp "$CLASSPATH"`" - if [ "$1" == "org.apache.spark.tools.JavaAPICompletenessChecker" ]; then - export SPARK_TOOLS_JAR="`cygpath -w "$SPARK_TOOLS_JAR"`" + SPARK_LAUNCHER_CP="${LAUNCHER_DIR}/${LAUNCHER_JARS}" +else + LAUNCHER_DIR="$SPARK_HOME/launcher/target/scala-$SPARK_SCALA_VERSION" + if [ ! -d "$LAUNCHER_DIR/classes" ]; then + echo "Failed to find Spark launcher classes in $LAUNCHER_DIR." 1>&2 + echo "You need to build Spark before running this program." 1>&2 + exit 1 fi + SPARK_LAUNCHER_CP="$LAUNCHER_DIR/classes" fi -export CLASSPATH -# In Spark submit client mode, the driver is launched in the same JVM as Spark submit itself. -# Here we must parse the properties file for relevant "spark.driver.*" configs before launching -# the driver JVM itself. Instead of handling this complexity in Bash, we launch a separate JVM -# to prepare the launch environment of this driver JVM. +# The launcher library will print arguments separated by a NULL character, to allow arguments with +# characters that would be otherwise interpreted by the shell. Read that in a while loop, populating +# an array that will be used to exec the final command. +CMD=() +while IFS= read -d '' -r ARG; do + CMD+=("$ARG") +done < <("$RUNNER" -cp "$SPARK_LAUNCHER_CP" org.apache.spark.launcher.Main "$@") -if [ -n "$SPARK_SUBMIT_BOOTSTRAP_DRIVER" ]; then - # This is used only if the properties file actually contains these special configs - # Export the environment variables needed by SparkSubmitDriverBootstrapper - export RUNNER - export CLASSPATH - export JAVA_OPTS - export OUR_JAVA_MEM - export SPARK_CLASS=1 - shift # Ignore main class (org.apache.spark.deploy.SparkSubmit) and use our own - exec "$RUNNER" org.apache.spark.deploy.SparkSubmitDriverBootstrapper "$@" +if [ "${CMD[0]}" = "usage" ]; then + "${CMD[@]}" else - # Note: The format of this command is closely echoed in SparkSubmitDriverBootstrapper.scala - if [ -n "$SPARK_PRINT_LAUNCH_COMMAND" ]; then - echo -n "Spark Command: " 1>&2 - echo "$RUNNER" -cp "$CLASSPATH" $JAVA_OPTS "$@" 1>&2 - echo -e "========================================\n" 1>&2 - fi - exec "$RUNNER" -cp "$CLASSPATH" $JAVA_OPTS "$@" + exec "${CMD[@]}" fi - diff --git a/bin/spark-class2.cmd b/bin/spark-class2.cmd index da46543647efd..37d22215a0e7e 100644 --- a/bin/spark-class2.cmd +++ b/bin/spark-class2.cmd @@ -17,135 +17,54 @@ rem See the License for the specific language governing permissions and rem limitations under the License. rem -rem Any changes to this file must be reflected in SparkSubmitDriverBootstrapper.scala! - -setlocal enabledelayedexpansion - -set SCALA_VERSION=2.10 - rem Figure out where the Spark framework is installed -set FWDIR=%~dp0..\ - -rem Export this as SPARK_HOME -set SPARK_HOME=%FWDIR% +set SPARK_HOME=%~dp0.. rem Load environment variables from conf\spark-env.cmd, if it exists -if exist "%FWDIR%conf\spark-env.cmd" call "%FWDIR%conf\spark-env.cmd" +if exist "%SPARK_HOME%\conf\spark-env.cmd" call "%SPARK_HOME%\conf\spark-env.cmd" rem Test that an argument was given -if not "x%1"=="x" goto arg_given +if "x%1"=="x" ( echo Usage: spark-class ^ [^] - goto exit -:arg_given - -if not "x%SPARK_MEM%"=="x" ( - echo Warning: SPARK_MEM is deprecated, please use a more specific config option - echo e.g., spark.executor.memory or spark.driver.memory. + exit /b 1 ) -rem Use SPARK_MEM or 512m as the default memory, to be overridden by specific options -set OUR_JAVA_MEM=%SPARK_MEM% -if "x%OUR_JAVA_MEM%"=="x" set OUR_JAVA_MEM=512m - -set SPARK_DAEMON_JAVA_OPTS=%SPARK_DAEMON_JAVA_OPTS% -Dspark.akka.logLifecycleEvents=true - -rem Add java opts and memory settings for master, worker, history server, executors, and repl. -rem Master, Worker and HistoryServer use SPARK_DAEMON_JAVA_OPTS (and specific opts) + SPARK_DAEMON_MEMORY. -if "%1"=="org.apache.spark.deploy.master.Master" ( - set OUR_JAVA_OPTS=%SPARK_DAEMON_JAVA_OPTS% %SPARK_MASTER_OPTS% - if not "x%SPARK_DAEMON_MEMORY%"=="x" set OUR_JAVA_MEM=%SPARK_DAEMON_MEMORY% -) else if "%1"=="org.apache.spark.deploy.worker.Worker" ( - set OUR_JAVA_OPTS=%SPARK_DAEMON_JAVA_OPTS% %SPARK_WORKER_OPTS% - if not "x%SPARK_DAEMON_MEMORY%"=="x" set OUR_JAVA_MEM=%SPARK_DAEMON_MEMORY% -) else if "%1"=="org.apache.spark.deploy.history.HistoryServer" ( - set OUR_JAVA_OPTS=%SPARK_DAEMON_JAVA_OPTS% %SPARK_HISTORY_OPTS% - if not "x%SPARK_DAEMON_MEMORY%"=="x" set OUR_JAVA_MEM=%SPARK_DAEMON_MEMORY% - -rem Executors use SPARK_JAVA_OPTS + SPARK_EXECUTOR_MEMORY. -) else if "%1"=="org.apache.spark.executor.CoarseGrainedExecutorBackend" ( - set OUR_JAVA_OPTS=%SPARK_JAVA_OPTS% %SPARK_EXECUTOR_OPTS% - if not "x%SPARK_EXECUTOR_MEMORY%"=="x" set OUR_JAVA_MEM=%SPARK_EXECUTOR_MEMORY% -) else if "%1"=="org.apache.spark.executor.MesosExecutorBackend" ( - set OUR_JAVA_OPTS=%SPARK_JAVA_OPTS% %SPARK_EXECUTOR_OPTS% - if not "x%SPARK_EXECUTOR_MEMORY%"=="x" set OUR_JAVA_MEM=%SPARK_EXECUTOR_MEMORY% +set LAUNCHER_CP=0 +if exist %SPARK_HOME%\RELEASE goto find_release_launcher -rem Spark submit uses SPARK_JAVA_OPTS + SPARK_SUBMIT_OPTS + -rem SPARK_DRIVER_MEMORY + SPARK_SUBMIT_DRIVER_MEMORY. -rem The repl also uses SPARK_REPL_OPTS. -) else if "%1"=="org.apache.spark.deploy.SparkSubmit" ( - set OUR_JAVA_OPTS=%SPARK_JAVA_OPTS% %SPARK_SUBMIT_OPTS% %SPARK_REPL_OPTS% - if not "x%SPARK_SUBMIT_LIBRARY_PATH%"=="x" ( - set OUR_JAVA_OPTS=!OUR_JAVA_OPTS! -Djava.library.path=%SPARK_SUBMIT_LIBRARY_PATH% - ) else if not "x%SPARK_LIBRARY_PATH%"=="x" ( - set OUR_JAVA_OPTS=!OUR_JAVA_OPTS! -Djava.library.path=%SPARK_LIBRARY_PATH% - ) - if not "x%SPARK_DRIVER_MEMORY%"=="x" set OUR_JAVA_MEM=%SPARK_DRIVER_MEMORY% - if not "x%SPARK_SUBMIT_DRIVER_MEMORY%"=="x" set OUR_JAVA_MEM=%SPARK_SUBMIT_DRIVER_MEMORY% -) else ( - set OUR_JAVA_OPTS=%SPARK_JAVA_OPTS% - if not "x%SPARK_DRIVER_MEMORY%"=="x" set OUR_JAVA_MEM=%SPARK_DRIVER_MEMORY% +rem Look for the Spark launcher in both Scala build directories. The launcher doesn't use Scala so +rem it doesn't really matter which one is picked up. Add the compiled classes directly to the +rem classpath instead of looking for a jar file, since it's very common for people using sbt to use +rem the "assembly" target instead of "package". +set LAUNCHER_CLASSES=%SPARK_HOME%\launcher\target\scala-2.10\classes +if exist %LAUNCHER_CLASSES% ( + set LAUNCHER_CP=%LAUNCHER_CLASSES% ) - -rem Set JAVA_OPTS to be able to load native libraries and to set heap size -for /f "tokens=3" %%i in ('java -version 2^>^&1 ^| find "version"') do set jversion=%%i -for /f "tokens=1 delims=_" %%i in ("%jversion:~1,-1%") do set jversion=%%i -if "%jversion%" geq "1.8.0" ( - set JAVA_OPTS=%OUR_JAVA_OPTS% -Xms%OUR_JAVA_MEM% -Xmx%OUR_JAVA_MEM% -) else ( - set JAVA_OPTS=-XX:MaxPermSize=128m %OUR_JAVA_OPTS% -Xms%OUR_JAVA_MEM% -Xmx%OUR_JAVA_MEM% +set LAUNCHER_CLASSES=%SPARK_HOME%\launcher\target\scala-2.11\classes +if exist %LAUNCHER_CLASSES% ( + set LAUNCHER_CP=%LAUNCHER_CLASSES% ) -rem Attention: when changing the way the JAVA_OPTS are assembled, the change must be reflected in CommandUtils.scala! +goto check_launcher -rem Test whether the user has built Spark -if exist "%FWDIR%RELEASE" goto skip_build_test -set FOUND_JAR=0 -for %%d in ("%FWDIR%assembly\target\scala-%SCALA_VERSION%\spark-assembly*hadoop*.jar") do ( - set FOUND_JAR=1 -) -if "%FOUND_JAR%"=="0" ( - echo Failed to find Spark assembly JAR. - echo You need to build Spark before running this program. - goto exit +:find_release_launcher +for %%d in (%SPARK_HOME%\lib\spark-launcher*.jar) do ( + set LAUNCHER_CP=%%d ) -:skip_build_test -set TOOLS_DIR=%FWDIR%tools -set SPARK_TOOLS_JAR= -for %%d in ("%TOOLS_DIR%\target\scala-%SCALA_VERSION%\spark-tools*assembly*.jar") do ( - set SPARK_TOOLS_JAR=%%d +:check_launcher +if "%LAUNCHER_CP%"=="0" ( + echo Failed to find Spark launcher JAR. + echo You need to build Spark before running this program. + exit /b 1 ) -rem Compute classpath using external script -set DONT_PRINT_CLASSPATH=1 -call "%FWDIR%bin\compute-classpath.cmd" -set DONT_PRINT_CLASSPATH=0 -set CLASSPATH=%CLASSPATH%;%SPARK_TOOLS_JAR% - rem Figure out where java is. set RUNNER=java if not "x%JAVA_HOME%"=="x" set RUNNER=%JAVA_HOME%\bin\java -rem In Spark submit client mode, the driver is launched in the same JVM as Spark submit itself. -rem Here we must parse the properties file for relevant "spark.driver.*" configs before launching -rem the driver JVM itself. Instead of handling this complexity here, we launch a separate JVM -rem to prepare the launch environment of this driver JVM. - -rem In this case, leave out the main class (org.apache.spark.deploy.SparkSubmit) and use our own. -rem Leaving out the first argument is surprisingly difficult to do in Windows. Note that this must -rem be done here because the Windows "shift" command does not work in a conditional block. -set BOOTSTRAP_ARGS= -shift -:start_parse -if "%~1" == "" goto end_parse -set BOOTSTRAP_ARGS=%BOOTSTRAP_ARGS% %~1 -shift -goto start_parse -:end_parse - -if not [%SPARK_SUBMIT_BOOTSTRAP_DRIVER%] == [] ( - set SPARK_CLASS=1 - "%RUNNER%" org.apache.spark.deploy.SparkSubmitDriverBootstrapper %BOOTSTRAP_ARGS% -) else ( - "%RUNNER%" -cp "%CLASSPATH%" %JAVA_OPTS% %* +rem The launcher library prints the command to be executed in a single line suitable for being +rem executed by the batch interpreter. So read all the output of the launcher into a variable. +for /f "tokens=*" %%i in ('cmd /C ""%RUNNER%" -cp %LAUNCHER_CP% org.apache.spark.launcher.Main %*"') do ( + set SPARK_CMD=%%i ) -:exit +%SPARK_CMD% diff --git a/bin/spark-shell b/bin/spark-shell index cca5aa0676123..b3761b5e1375b 100755 --- a/bin/spark-shell +++ b/bin/spark-shell @@ -28,25 +28,24 @@ esac # Enter posix mode for bash set -o posix -## Global script variables -FWDIR="$(cd "`dirname "$0"`"/..; pwd)" +export FWDIR="$(cd "`dirname "$0"`"/..; pwd)" -function usage() { +usage() { + if [ -n "$1" ]; then + echo "$1" + fi echo "Usage: ./bin/spark-shell [options]" "$FWDIR"/bin/spark-submit --help 2>&1 | grep -v Usage 1>&2 - exit 0 + exit "$2" } +export -f usage if [[ "$@" = *--help ]] || [[ "$@" = *-h ]]; then - usage + usage "" 0 fi -source "$FWDIR"/bin/utils.sh -SUBMIT_USAGE_FUNCTION=usage -gatherSparkSubmitOpts "$@" - # SPARK-4161: scala does not assume use of the java classpath, -# so we need to add the "-Dscala.usejavacp=true" flag mnually. We +# so we need to add the "-Dscala.usejavacp=true" flag manually. We # do this specifically for the Spark shell because the scala REPL # has its own class loader, and any additional classpath specified # through spark.driver.extraClassPath is not automatically propagated. @@ -61,11 +60,11 @@ function main() { # (see https://github.com/sbt/sbt/issues/562). stty -icanon min 1 -echo > /dev/null 2>&1 export SPARK_SUBMIT_OPTS="$SPARK_SUBMIT_OPTS -Djline.terminal=unix" - "$FWDIR"/bin/spark-submit --class org.apache.spark.repl.Main "${SUBMISSION_OPTS[@]}" spark-shell "${APPLICATION_OPTS[@]}" + "$FWDIR"/bin/spark-submit --class org.apache.spark.repl.Main "$@" stty icanon echo > /dev/null 2>&1 else export SPARK_SUBMIT_OPTS - "$FWDIR"/bin/spark-submit --class org.apache.spark.repl.Main "${SUBMISSION_OPTS[@]}" spark-shell "${APPLICATION_OPTS[@]}" + "$FWDIR"/bin/spark-submit --class org.apache.spark.repl.Main "$@" fi } diff --git a/bin/spark-shell2.cmd b/bin/spark-shell2.cmd index 1d1a40da315eb..02f51fe59a911 100644 --- a/bin/spark-shell2.cmd +++ b/bin/spark-shell2.cmd @@ -25,17 +25,28 @@ if %ERRORLEVEL% equ 0 ( exit /b 0 ) -call %SPARK_HOME%\bin\windows-utils.cmd %* -if %ERRORLEVEL% equ 1 ( +rem SPARK-4161: scala does not assume use of the java classpath, +rem so we need to add the "-Dscala.usejavacp=true" flag manually. We +rem do this specifically for the Spark shell because the scala REPL +rem has its own class loader, and any additional classpath specified +rem through spark.driver.extraClassPath is not automatically propagated. +if "x%SPARK_SUBMIT_OPTS%"=="x" ( + set SPARK_SUBMIT_OPTS=-Dscala.usejavacp=true + goto run_shell +) +set SPARK_SUBMIT_OPTS="%SPARK_SUBMIT_OPTS% -Dscala.usejavacp=true" + +:run_shell +call %SPARK_HOME%\bin\spark-submit2.cmd --class org.apache.spark.repl.Main %* +set SPARK_ERROR_LEVEL=%ERRORLEVEL% +if not "x%SPARK_LAUNCHER_USAGE_ERROR%"=="x" ( call :usage exit /b 1 ) - -cmd /V /E /C %SPARK_HOME%\bin\spark-submit.cmd --class org.apache.spark.repl.Main %SUBMISSION_OPTS% spark-shell %APPLICATION_OPTS% - -exit /b 0 +exit /b %SPARK_ERROR_LEVEL% :usage +echo %SPARK_LAUNCHER_USAGE_ERROR% echo "Usage: .\bin\spark-shell.cmd [options]" >&2 -%SPARK_HOME%\bin\spark-submit --help 2>&1 | findstr /V "Usage" 1>&2 -exit /b 0 +call %SPARK_HOME%\bin\spark-submit2.cmd --help 2>&1 | findstr /V "Usage" 1>&2 +goto :eof diff --git a/bin/spark-sql b/bin/spark-sql index 3b6cc420fea81..ca1729f4cfcb4 100755 --- a/bin/spark-sql +++ b/bin/spark-sql @@ -25,12 +25,15 @@ set -o posix # NOTE: This exact class name is matched downstream by SparkSubmit. # Any changes need to be reflected there. -CLASS="org.apache.spark.sql.hive.thriftserver.SparkSQLCLIDriver" +export CLASS="org.apache.spark.sql.hive.thriftserver.SparkSQLCLIDriver" # Figure out where Spark is installed -FWDIR="$(cd "`dirname "$0"`"/..; pwd)" +export FWDIR="$(cd "`dirname "$0"`"/..; pwd)" function usage { + if [ -n "$1" ]; then + echo "$1" + fi echo "Usage: ./bin/spark-sql [options] [cli option]" pattern="usage" pattern+="\|Spark assembly has been built with Hive" @@ -42,16 +45,13 @@ function usage { "$FWDIR"/bin/spark-submit --help 2>&1 | grep -v Usage 1>&2 echo echo "CLI options:" - "$FWDIR"/bin/spark-class $CLASS --help 2>&1 | grep -v "$pattern" 1>&2 + "$FWDIR"/bin/spark-class "$CLASS" --help 2>&1 | grep -v "$pattern" 1>&2 + exit "$2" } +export -f usage if [[ "$@" = *--help ]] || [[ "$@" = *-h ]]; then - usage - exit 0 + usage "" 0 fi -source "$FWDIR"/bin/utils.sh -SUBMIT_USAGE_FUNCTION=usage -gatherSparkSubmitOpts "$@" - -exec "$FWDIR"/bin/spark-submit --class $CLASS "${SUBMISSION_OPTS[@]}" spark-internal "${APPLICATION_OPTS[@]}" +exec "$FWDIR"/bin/spark-submit --class "$CLASS" "$@" diff --git a/bin/spark-submit b/bin/spark-submit index 3e5cbdbb24394..bcff78edd51ca 100755 --- a/bin/spark-submit +++ b/bin/spark-submit @@ -17,58 +17,18 @@ # limitations under the License. # -# NOTE: Any changes in this file must be reflected in SparkSubmitDriverBootstrapper.scala! - -export SPARK_HOME="$(cd "`dirname "$0"`"/..; pwd)" -ORIG_ARGS=("$@") - -# Set COLUMNS for progress bar -export COLUMNS=`tput cols` - -while (($#)); do - if [ "$1" = "--deploy-mode" ]; then - SPARK_SUBMIT_DEPLOY_MODE=$2 - elif [ "$1" = "--properties-file" ]; then - SPARK_SUBMIT_PROPERTIES_FILE=$2 - elif [ "$1" = "--driver-memory" ]; then - export SPARK_SUBMIT_DRIVER_MEMORY=$2 - elif [ "$1" = "--driver-library-path" ]; then - export SPARK_SUBMIT_LIBRARY_PATH=$2 - elif [ "$1" = "--driver-class-path" ]; then - export SPARK_SUBMIT_CLASSPATH=$2 - elif [ "$1" = "--driver-java-options" ]; then - export SPARK_SUBMIT_OPTS=$2 - elif [ "$1" = "--master" ]; then - export MASTER=$2 - fi - shift -done - -if [ -z "$SPARK_CONF_DIR" ]; then - export SPARK_CONF_DIR="$SPARK_HOME/conf" -fi -DEFAULT_PROPERTIES_FILE="$SPARK_CONF_DIR/spark-defaults.conf" -if [ "$MASTER" == "yarn-cluster" ]; then - SPARK_SUBMIT_DEPLOY_MODE=cluster +SPARK_HOME="$(cd "`dirname "$0"`"/..; pwd)" + +# Only define a usage function if an upstream script hasn't done so. +if ! type -t usage >/dev/null 2>&1; then + usage() { + if [ -n "$1" ]; then + echo "$1" + fi + "$SPARK_HOME"/bin/spark-class org.apache.spark.deploy.SparkSubmit --help + exit "$2" + } + export -f usage fi -export SPARK_SUBMIT_DEPLOY_MODE=${SPARK_SUBMIT_DEPLOY_MODE:-"client"} -export SPARK_SUBMIT_PROPERTIES_FILE=${SPARK_SUBMIT_PROPERTIES_FILE:-"$DEFAULT_PROPERTIES_FILE"} - -# For client mode, the driver will be launched in the same JVM that launches -# SparkSubmit, so we may need to read the properties file for any extra class -# paths, library paths, java options and memory early on. Otherwise, it will -# be too late by the time the driver JVM has started. - -if [[ "$SPARK_SUBMIT_DEPLOY_MODE" == "client" && -f "$SPARK_SUBMIT_PROPERTIES_FILE" ]]; then - # Parse the properties file only if the special configs exist - contains_special_configs=$( - grep -e "spark.driver.extra*\|spark.driver.memory" "$SPARK_SUBMIT_PROPERTIES_FILE" | \ - grep -v "^[[:space:]]*#" - ) - if [ -n "$contains_special_configs" ]; then - export SPARK_SUBMIT_BOOTSTRAP_DRIVER=1 - fi -fi - -exec "$SPARK_HOME"/bin/spark-class org.apache.spark.deploy.SparkSubmit "${ORIG_ARGS[@]}" +exec "$SPARK_HOME"/bin/spark-class org.apache.spark.deploy.SparkSubmit "$@" diff --git a/bin/spark-submit2.cmd b/bin/spark-submit2.cmd index 446cbc74b74f9..08ddb185742d2 100644 --- a/bin/spark-submit2.cmd +++ b/bin/spark-submit2.cmd @@ -17,62 +17,19 @@ rem See the License for the specific language governing permissions and rem limitations under the License. rem -rem NOTE: Any changes in this file must be reflected in SparkSubmitDriverBootstrapper.scala! - -set SPARK_HOME=%~dp0.. -set ORIG_ARGS=%* - -rem Reset the values of all variables used -set SPARK_SUBMIT_DEPLOY_MODE=client - -if [%SPARK_CONF_DIR%] == [] ( - set SPARK_CONF_DIR=%SPARK_HOME%\conf -) -set SPARK_SUBMIT_PROPERTIES_FILE=%SPARK_CONF_DIR%\spark-defaults.conf -set SPARK_SUBMIT_DRIVER_MEMORY= -set SPARK_SUBMIT_LIBRARY_PATH= -set SPARK_SUBMIT_CLASSPATH= -set SPARK_SUBMIT_OPTS= -set SPARK_SUBMIT_BOOTSTRAP_DRIVER= - -:loop -if [%1] == [] goto continue - if [%1] == [--deploy-mode] ( - set SPARK_SUBMIT_DEPLOY_MODE=%2 - ) else if [%1] == [--properties-file] ( - set SPARK_SUBMIT_PROPERTIES_FILE=%2 - ) else if [%1] == [--driver-memory] ( - set SPARK_SUBMIT_DRIVER_MEMORY=%2 - ) else if [%1] == [--driver-library-path] ( - set SPARK_SUBMIT_LIBRARY_PATH=%2 - ) else if [%1] == [--driver-class-path] ( - set SPARK_SUBMIT_CLASSPATH=%2 - ) else if [%1] == [--driver-java-options] ( - set SPARK_SUBMIT_OPTS=%2 - ) else if [%1] == [--master] ( - set MASTER=%2 - ) - shift -goto loop -:continue - -if [%MASTER%] == [yarn-cluster] ( - set SPARK_SUBMIT_DEPLOY_MODE=cluster -) - -rem For client mode, the driver will be launched in the same JVM that launches -rem SparkSubmit, so we may need to read the properties file for any extra class -rem paths, library paths, java options and memory early on. Otherwise, it will -rem be too late by the time the driver JVM has started. - -if [%SPARK_SUBMIT_DEPLOY_MODE%] == [client] ( - if exist %SPARK_SUBMIT_PROPERTIES_FILE% ( - rem Parse the properties file only if the special configs exist - for /f %%i in ('findstr /r /c:"^[\t ]*spark.driver.memory" /c:"^[\t ]*spark.driver.extra" ^ - %SPARK_SUBMIT_PROPERTIES_FILE%') do ( - set SPARK_SUBMIT_BOOTSTRAP_DRIVER=1 - ) - ) +rem This is the entry point for running Spark submit. To avoid polluting the +rem environment, it just launches a new cmd to do the real work. + +set CLASS=org.apache.spark.deploy.SparkSubmit +call %~dp0spark-class2.cmd %CLASS% %* +set SPARK_ERROR_LEVEL=%ERRORLEVEL% +if not "x%SPARK_LAUNCHER_USAGE_ERROR%"=="x" ( + call :usage + exit /b 1 ) +exit /b %SPARK_ERROR_LEVEL% -cmd /V /E /C %SPARK_HOME%\bin\spark-class.cmd org.apache.spark.deploy.SparkSubmit %ORIG_ARGS% +:usage +echo %SPARK_LAUNCHER_USAGE_ERROR% +call %SPARK_HOME%\bin\spark-class2.cmd %CLASS% --help +goto :eof diff --git a/bin/utils.sh b/bin/utils.sh deleted file mode 100755 index 748dbe345a74c..0000000000000 --- a/bin/utils.sh +++ /dev/null @@ -1,60 +0,0 @@ -#!/usr/bin/env bash - -# -# Licensed to the Apache Software Foundation (ASF) under one or more -# contributor license agreements. See the NOTICE file distributed with -# this work for additional information regarding copyright ownership. -# The ASF licenses this file to You under the Apache License, Version 2.0 -# (the "License"); you may not use this file except in compliance with -# the License. You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# - -# Gather all spark-submit options into SUBMISSION_OPTS -function gatherSparkSubmitOpts() { - - if [ -z "$SUBMIT_USAGE_FUNCTION" ]; then - echo "Function for printing usage of $0 is not set." 1>&2 - echo "Please set usage function to shell variable 'SUBMIT_USAGE_FUNCTION' in $0" 1>&2 - exit 1 - fi - - # NOTE: If you add or remove spark-submit options, - # modify NOT ONLY this script but also SparkSubmitArgument.scala - SUBMISSION_OPTS=() - APPLICATION_OPTS=() - while (($#)); do - case "$1" in - --master | --deploy-mode | --class | --name | --jars | --packages | --py-files | --files | \ - --conf | --repositories | --properties-file | --driver-memory | --driver-java-options | \ - --driver-library-path | --driver-class-path | --executor-memory | --driver-cores | \ - --total-executor-cores | --executor-cores | --queue | --num-executors | --archives | \ - --proxy-user) - if [[ $# -lt 2 ]]; then - "$SUBMIT_USAGE_FUNCTION" - exit 1; - fi - SUBMISSION_OPTS+=("$1"); shift - SUBMISSION_OPTS+=("$1"); shift - ;; - - --verbose | -v | --supervise) - SUBMISSION_OPTS+=("$1"); shift - ;; - - *) - APPLICATION_OPTS+=("$1"); shift - ;; - esac - done - - export SUBMISSION_OPTS - export APPLICATION_OPTS -} diff --git a/bin/windows-utils.cmd b/bin/windows-utils.cmd deleted file mode 100644 index 0cf9e87ca554b..0000000000000 --- a/bin/windows-utils.cmd +++ /dev/null @@ -1,60 +0,0 @@ -rem -rem Licensed to the Apache Software Foundation (ASF) under one or more -rem contributor license agreements. See the NOTICE file distributed with -rem this work for additional information regarding copyright ownership. -rem The ASF licenses this file to You under the Apache License, Version 2.0 -rem (the "License"); you may not use this file except in compliance with -rem the License. You may obtain a copy of the License at -rem -rem http://www.apache.org/licenses/LICENSE-2.0 -rem -rem Unless required by applicable law or agreed to in writing, software -rem distributed under the License is distributed on an "AS IS" BASIS, -rem WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -rem See the License for the specific language governing permissions and -rem limitations under the License. -rem - -rem Gather all spark-submit options into SUBMISSION_OPTS - -set SUBMISSION_OPTS= -set APPLICATION_OPTS= - -rem NOTE: If you add or remove spark-sumbmit options, -rem modify NOT ONLY this script but also SparkSubmitArgument.scala - -:OptsLoop -if "x%1"=="x" ( - goto :OptsLoopEnd -) - -SET opts="\<--master\> \<--deploy-mode\> \<--class\> \<--name\> \<--jars\> \<--py-files\> \<--files\>" -SET opts="%opts:~1,-1% \<--conf\> \<--properties-file\> \<--driver-memory\> \<--driver-java-options\>" -SET opts="%opts:~1,-1% \<--driver-library-path\> \<--driver-class-path\> \<--executor-memory\>" -SET opts="%opts:~1,-1% \<--driver-cores\> \<--total-executor-cores\> \<--executor-cores\> \<--queue\>" -SET opts="%opts:~1,-1% \<--num-executors\> \<--archives\> \<--packages\> \<--repositories\>" -SET opts="%opts:~1,-1% \<--proxy-user\>" - -echo %1 | findstr %opts% >nul -if %ERRORLEVEL% equ 0 ( - if "x%2"=="x" ( - echo "%1" requires an argument. >&2 - exit /b 1 - ) - set SUBMISSION_OPTS=%SUBMISSION_OPTS% %1 %2 - shift - shift - goto :OptsLoop -) -echo %1 | findstr "\<--verbose\> \<-v\> \<--supervise\>" >nul -if %ERRORLEVEL% equ 0 ( - set SUBMISSION_OPTS=%SUBMISSION_OPTS% %1 - shift - goto :OptsLoop -) -set APPLICATION_OPTS=%APPLICATION_OPTS% %1 -shift -goto :OptsLoop - -:OptsLoopEnd -exit /b 0 diff --git a/core/pom.xml b/core/pom.xml index dc0d07d806635..4164a3a7208d4 100644 --- a/core/pom.xml +++ b/core/pom.xml @@ -76,6 +76,11 @@ + + org.apache.spark + spark-launcher_${scala.binary.version} + ${project.version} + org.apache.spark spark-network-common_${scala.binary.version} diff --git a/core/src/main/scala/org/apache/spark/deploy/SparkSubmitArguments.scala b/core/src/main/scala/org/apache/spark/deploy/SparkSubmitArguments.scala index 82e66a374249c..94e4bdbfb7d7b 100644 --- a/core/src/main/scala/org/apache/spark/deploy/SparkSubmitArguments.scala +++ b/core/src/main/scala/org/apache/spark/deploy/SparkSubmitArguments.scala @@ -18,18 +18,22 @@ package org.apache.spark.deploy import java.net.URI +import java.util.{List => JList} import java.util.jar.JarFile +import scala.collection.JavaConversions._ import scala.collection.mutable.{ArrayBuffer, HashMap} import org.apache.spark.deploy.SparkSubmitAction._ +import org.apache.spark.launcher.SparkSubmitArgumentsParser import org.apache.spark.util.Utils /** * Parses and encapsulates arguments from the spark-submit script. * The env argument is used for testing. */ -private[spark] class SparkSubmitArguments(args: Seq[String], env: Map[String, String] = sys.env) { +private[spark] class SparkSubmitArguments(args: Seq[String], env: Map[String, String] = sys.env) + extends SparkSubmitArgumentsParser { var master: String = null var deployMode: String = null var executorMemory: String = null @@ -84,7 +88,12 @@ private[spark] class SparkSubmitArguments(args: Seq[String], env: Map[String, St } // Set parameters from command line arguments - parseOpts(args.toList) + try { + parse(args.toList) + } catch { + case e: IllegalArgumentException => + SparkSubmit.printErrorAndExit(e.getMessage()) + } // Populate `sparkProperties` map from properties file mergeDefaultSparkProperties() // Use `sparkProperties` map along with env vars to fill in any missing parameters @@ -277,167 +286,139 @@ private[spark] class SparkSubmitArguments(args: Seq[String], env: Map[String, St """.stripMargin } - /** - * Fill in values by parsing user options. - * NOTE: Any changes here must be reflected in YarnClientSchedulerBackend. - */ - private def parseOpts(opts: Seq[String]): Unit = { - val EQ_SEPARATED_OPT="""(--[^=]+)=(.+)""".r - - // Delineates parsing of Spark options from parsing of user options. - parse(opts) - - /** - * NOTE: If you add or remove spark-submit options, - * modify NOT ONLY this file but also utils.sh - */ - def parse(opts: Seq[String]): Unit = opts match { - case ("--name") :: value :: tail => + /** Fill in values by parsing user options. */ + override protected def handle(opt: String, value: String): Boolean = { + opt match { + case NAME => name = value - parse(tail) - case ("--master") :: value :: tail => + case MASTER => master = value - parse(tail) - case ("--class") :: value :: tail => + case CLASS => mainClass = value - parse(tail) - case ("--deploy-mode") :: value :: tail => + case DEPLOY_MODE => if (value != "client" && value != "cluster") { SparkSubmit.printErrorAndExit("--deploy-mode must be either \"client\" or \"cluster\"") } deployMode = value - parse(tail) - case ("--num-executors") :: value :: tail => + case NUM_EXECUTORS => numExecutors = value - parse(tail) - case ("--total-executor-cores") :: value :: tail => + case TOTAL_EXECUTOR_CORES => totalExecutorCores = value - parse(tail) - case ("--executor-cores") :: value :: tail => + case EXECUTOR_CORES => executorCores = value - parse(tail) - case ("--executor-memory") :: value :: tail => + case EXECUTOR_MEMORY => executorMemory = value - parse(tail) - case ("--driver-memory") :: value :: tail => + case DRIVER_MEMORY => driverMemory = value - parse(tail) - case ("--driver-cores") :: value :: tail => + case DRIVER_CORES => driverCores = value - parse(tail) - case ("--driver-class-path") :: value :: tail => + case DRIVER_CLASS_PATH => driverExtraClassPath = value - parse(tail) - case ("--driver-java-options") :: value :: tail => + case DRIVER_JAVA_OPTIONS => driverExtraJavaOptions = value - parse(tail) - case ("--driver-library-path") :: value :: tail => + case DRIVER_LIBRARY_PATH => driverExtraLibraryPath = value - parse(tail) - case ("--properties-file") :: value :: tail => + case PROPERTIES_FILE => propertiesFile = value - parse(tail) - case ("--kill") :: value :: tail => + case KILL_SUBMISSION => submissionToKill = value if (action != null) { SparkSubmit.printErrorAndExit(s"Action cannot be both $action and $KILL.") } action = KILL - parse(tail) - case ("--status") :: value :: tail => + case STATUS => submissionToRequestStatusFor = value if (action != null) { SparkSubmit.printErrorAndExit(s"Action cannot be both $action and $REQUEST_STATUS.") } action = REQUEST_STATUS - parse(tail) - case ("--supervise") :: tail => + case SUPERVISE => supervise = true - parse(tail) - case ("--queue") :: value :: tail => + case QUEUE => queue = value - parse(tail) - case ("--files") :: value :: tail => + case FILES => files = Utils.resolveURIs(value) - parse(tail) - case ("--py-files") :: value :: tail => + case PY_FILES => pyFiles = Utils.resolveURIs(value) - parse(tail) - case ("--archives") :: value :: tail => + case ARCHIVES => archives = Utils.resolveURIs(value) - parse(tail) - case ("--jars") :: value :: tail => + case JARS => jars = Utils.resolveURIs(value) - parse(tail) - case ("--packages") :: value :: tail => + case PACKAGES => packages = value - parse(tail) - case ("--repositories") :: value :: tail => + case REPOSITORIES => repositories = value - parse(tail) - case ("--conf" | "-c") :: value :: tail => + case CONF => value.split("=", 2).toSeq match { case Seq(k, v) => sparkProperties(k) = v case _ => SparkSubmit.printErrorAndExit(s"Spark config without '=': $value") } - parse(tail) - case ("--proxy-user") :: value :: tail => + case PROXY_USER => proxyUser = value - parse(tail) - case ("--help" | "-h") :: tail => + case HELP => printUsageAndExit(0) - case ("--verbose" | "-v") :: tail => + case VERBOSE => verbose = true - parse(tail) - case ("--version") :: tail => + case VERSION => SparkSubmit.printVersionAndExit() - case EQ_SEPARATED_OPT(opt, value) :: tail => - parse(opt :: value :: tail) + case _ => + throw new IllegalArgumentException(s"Unexpected argument '$opt'.") + } + true + } - case value :: tail if value.startsWith("-") => - SparkSubmit.printErrorAndExit(s"Unrecognized option '$value'.") + /** + * Handle unrecognized command line options. + * + * The first unrecognized option is treated as the "primary resource". Everything else is + * treated as application arguments. + */ + override protected def handleUnknown(opt: String): Boolean = { + if (opt.startsWith("-")) { + SparkSubmit.printErrorAndExit(s"Unrecognized option '$opt'.") + } - case value :: tail => - primaryResource = - if (!SparkSubmit.isShell(value) && !SparkSubmit.isInternal(value)) { - Utils.resolveURI(value).toString - } else { - value - } - isPython = SparkSubmit.isPython(value) - childArgs ++= tail + primaryResource = + if (!SparkSubmit.isShell(opt) && !SparkSubmit.isInternal(opt)) { + Utils.resolveURI(opt).toString + } else { + opt + } + isPython = SparkSubmit.isPython(opt) + false + } - case Nil => - } + override protected def handleExtraArgs(extra: JList[String]): Unit = { + childArgs ++= extra } private def printUsageAndExit(exitCode: Int, unknownParam: Any = null): Unit = { diff --git a/core/src/main/scala/org/apache/spark/deploy/SparkSubmitDriverBootstrapper.scala b/core/src/main/scala/org/apache/spark/deploy/SparkSubmitDriverBootstrapper.scala deleted file mode 100644 index 311048cdaa324..0000000000000 --- a/core/src/main/scala/org/apache/spark/deploy/SparkSubmitDriverBootstrapper.scala +++ /dev/null @@ -1,170 +0,0 @@ -/* - * Licensed to the Apache Software Foundation (ASF) under one or more - * contributor license agreements. See the NOTICE file distributed with - * this work for additional information regarding copyright ownership. - * The ASF licenses this file to You under the Apache License, Version 2.0 - * (the "License"); you may not use this file except in compliance with - * the License. You may obtain a copy of the License at - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package org.apache.spark.deploy - -import scala.collection.JavaConversions._ - -import org.apache.spark.util.{RedirectThread, Utils} - -/** - * Launch an application through Spark submit in client mode with the appropriate classpath, - * library paths, java options and memory. These properties of the JVM must be set before the - * driver JVM is launched. The sole purpose of this class is to avoid handling the complexity - * of parsing the properties file for such relevant configs in Bash. - * - * Usage: org.apache.spark.deploy.SparkSubmitDriverBootstrapper - */ -private[spark] object SparkSubmitDriverBootstrapper { - - // Note: This class depends on the behavior of `bin/spark-class` and `bin/spark-submit`. - // Any changes made there must be reflected in this file. - - def main(args: Array[String]): Unit = { - - // This should be called only from `bin/spark-class` - if (!sys.env.contains("SPARK_CLASS")) { - System.err.println("SparkSubmitDriverBootstrapper must be called from `bin/spark-class`!") - System.exit(1) - } - - val submitArgs = args - val runner = sys.env("RUNNER") - val classpath = sys.env("CLASSPATH") - val javaOpts = sys.env("JAVA_OPTS") - val defaultDriverMemory = sys.env("OUR_JAVA_MEM") - - // Spark submit specific environment variables - val deployMode = sys.env("SPARK_SUBMIT_DEPLOY_MODE") - val propertiesFile = sys.env("SPARK_SUBMIT_PROPERTIES_FILE") - val bootstrapDriver = sys.env("SPARK_SUBMIT_BOOTSTRAP_DRIVER") - val submitDriverMemory = sys.env.get("SPARK_SUBMIT_DRIVER_MEMORY") - val submitLibraryPath = sys.env.get("SPARK_SUBMIT_LIBRARY_PATH") - val submitClasspath = sys.env.get("SPARK_SUBMIT_CLASSPATH") - val submitJavaOpts = sys.env.get("SPARK_SUBMIT_OPTS") - - assume(runner != null, "RUNNER must be set") - assume(classpath != null, "CLASSPATH must be set") - assume(javaOpts != null, "JAVA_OPTS must be set") - assume(defaultDriverMemory != null, "OUR_JAVA_MEM must be set") - assume(deployMode == "client", "SPARK_SUBMIT_DEPLOY_MODE must be \"client\"!") - assume(propertiesFile != null, "SPARK_SUBMIT_PROPERTIES_FILE must be set") - assume(bootstrapDriver != null, "SPARK_SUBMIT_BOOTSTRAP_DRIVER must be set") - - // Parse the properties file for the equivalent spark.driver.* configs - val properties = Utils.getPropertiesFromFile(propertiesFile) - val confDriverMemory = properties.get("spark.driver.memory") - val confLibraryPath = properties.get("spark.driver.extraLibraryPath") - val confClasspath = properties.get("spark.driver.extraClassPath") - val confJavaOpts = properties.get("spark.driver.extraJavaOptions") - - // Favor Spark submit arguments over the equivalent configs in the properties file. - // Note that we do not actually use the Spark submit values for library path, classpath, - // and Java opts here, because we have already captured them in Bash. - - val newDriverMemory = submitDriverMemory - .orElse(confDriverMemory) - .getOrElse(defaultDriverMemory) - - val newClasspath = - if (submitClasspath.isDefined) { - classpath - } else { - classpath + confClasspath.map(sys.props("path.separator") + _).getOrElse("") - } - - val newJavaOpts = - if (submitJavaOpts.isDefined) { - // SPARK_SUBMIT_OPTS is already captured in JAVA_OPTS - javaOpts - } else { - javaOpts + confJavaOpts.map(" " + _).getOrElse("") - } - - val filteredJavaOpts = Utils.splitCommandString(newJavaOpts) - .filterNot(_.startsWith("-Xms")) - .filterNot(_.startsWith("-Xmx")) - - // Build up command - val command: Seq[String] = - Seq(runner) ++ - Seq("-cp", newClasspath) ++ - filteredJavaOpts ++ - Seq(s"-Xms$newDriverMemory", s"-Xmx$newDriverMemory") ++ - Seq("org.apache.spark.deploy.SparkSubmit") ++ - submitArgs - - // Print the launch command. This follows closely the format used in `bin/spark-class`. - if (sys.env.contains("SPARK_PRINT_LAUNCH_COMMAND")) { - System.err.print("Spark Command: ") - System.err.println(command.mkString(" ")) - System.err.println("========================================\n") - } - - // Start the driver JVM - val filteredCommand = command.filter(_.nonEmpty) - val builder = new ProcessBuilder(filteredCommand) - val env = builder.environment() - - if (submitLibraryPath.isEmpty && confLibraryPath.nonEmpty) { - val libraryPaths = confLibraryPath ++ sys.env.get(Utils.libraryPathEnvName) - env.put(Utils.libraryPathEnvName, libraryPaths.mkString(sys.props("path.separator"))) - } - - val process = builder.start() - - // If we kill an app while it's running, its sub-process should be killed too. - Runtime.getRuntime().addShutdownHook(new Thread() { - override def run() = { - if (process != null) { - process.destroy() - process.waitFor() - } - } - }) - - // Redirect stdout and stderr from the child JVM - val stdoutThread = new RedirectThread(process.getInputStream, System.out, "redirect stdout") - val stderrThread = new RedirectThread(process.getErrorStream, System.err, "redirect stderr") - stdoutThread.start() - stderrThread.start() - - // Redirect stdin to child JVM only if we're not running Windows. This is because the - // subprocess there already reads directly from our stdin, so we should avoid spawning a - // thread that contends with the subprocess in reading from System.in. - val isWindows = Utils.isWindows - val isSubprocess = sys.env.contains("IS_SUBPROCESS") - if (!isWindows) { - val stdinThread = new RedirectThread(System.in, process.getOutputStream, "redirect stdin", - propagateEof = true) - stdinThread.start() - // Spark submit (JVM) may run as a subprocess, and so this JVM should terminate on - // broken pipe, signaling that the parent process has exited. This is the case if the - // application is launched directly from python, as in the PySpark shell. In Windows, - // the termination logic is handled in java_gateway.py - if (isSubprocess) { - stdinThread.join() - process.destroy() - } - } - val returnCode = process.waitFor() - stdoutThread.join() - stderrThread.join() - sys.exit(returnCode) - } - -} diff --git a/core/src/main/scala/org/apache/spark/deploy/worker/CommandUtils.scala b/core/src/main/scala/org/apache/spark/deploy/worker/CommandUtils.scala index 3e013c32096c5..83f78cf47306c 100644 --- a/core/src/main/scala/org/apache/spark/deploy/worker/CommandUtils.scala +++ b/core/src/main/scala/org/apache/spark/deploy/worker/CommandUtils.scala @@ -20,10 +20,12 @@ package org.apache.spark.deploy.worker import java.io.{File, FileOutputStream, InputStream, IOException} import java.lang.System._ +import scala.collection.JavaConversions._ import scala.collection.Map import org.apache.spark.Logging import org.apache.spark.deploy.Command +import org.apache.spark.launcher.WorkerCommandBuilder import org.apache.spark.util.Utils /** @@ -54,12 +56,10 @@ object CommandUtils extends Logging { } private def buildCommandSeq(command: Command, memory: Int, sparkHome: String): Seq[String] = { - val runner = sys.env.get("JAVA_HOME").map(_ + "/bin/java").getOrElse("java") - // SPARK-698: do not call the run.cmd script, as process.destroy() // fails to kill a process tree on Windows - Seq(runner) ++ buildJavaOpts(command, memory, sparkHome) ++ Seq(command.mainClass) ++ - command.arguments + val cmd = new WorkerCommandBuilder(sparkHome, memory, command).buildCommand() + cmd.toSeq ++ Seq(command.mainClass) ++ command.arguments } /** @@ -92,44 +92,6 @@ object CommandUtils extends Logging { command.javaOpts) } - /** - * Attention: this must always be aligned with the environment variables in the run scripts and - * the way the JAVA_OPTS are assembled there. - */ - private def buildJavaOpts(command: Command, memory: Int, sparkHome: String): Seq[String] = { - val memoryOpts = Seq(s"-Xms${memory}M", s"-Xmx${memory}M") - - // Exists for backwards compatibility with older Spark versions - val workerLocalOpts = Option(getenv("SPARK_JAVA_OPTS")).map(Utils.splitCommandString) - .getOrElse(Nil) - if (workerLocalOpts.length > 0) { - logWarning("SPARK_JAVA_OPTS was set on the worker. It is deprecated in Spark 1.0.") - logWarning("Set SPARK_LOCAL_DIRS for node-specific storage locations.") - } - - // Figure out our classpath with the external compute-classpath script - val ext = if (System.getProperty("os.name").startsWith("Windows")) ".cmd" else ".sh" - val classPath = Utils.executeAndGetOutput( - Seq(sparkHome + "/bin/compute-classpath" + ext), - extraEnvironment = command.environment) - val userClassPath = command.classPathEntries ++ Seq(classPath) - - val javaVersion = System.getProperty("java.version") - - val javaOpts = workerLocalOpts ++ command.javaOpts - - val permGenOpt = - if (!javaVersion.startsWith("1.8") && !javaOpts.exists(_.startsWith("-XX:MaxPermSize="))) { - // do not specify -XX:MaxPermSize if it was already specified by user - Some("-XX:MaxPermSize=128m") - } else { - None - } - - Seq("-cp", userClassPath.filterNot(_.isEmpty).mkString(File.pathSeparator)) ++ - permGenOpt ++ javaOpts ++ memoryOpts - } - /** Spawn a thread that will redirect a given stream to a file */ def redirectStream(in: InputStream, file: File) { val out = new FileOutputStream(file, true) diff --git a/core/src/main/scala/org/apache/spark/executor/Executor.scala b/core/src/main/scala/org/apache/spark/executor/Executor.scala index bed0a08d4d515..a897e532184ac 100644 --- a/core/src/main/scala/org/apache/spark/executor/Executor.scala +++ b/core/src/main/scala/org/apache/spark/executor/Executor.scala @@ -49,7 +49,6 @@ private[spark] class Executor( isLocal: Boolean = false) extends Logging { - logInfo(s"Starting executor ID $executorId on host $executorHostname") // Application dependencies (added through SparkContext) that we've fetched so far on this node. diff --git a/core/src/main/scala/org/apache/spark/launcher/SparkSubmitArgumentsParser.scala b/core/src/main/scala/org/apache/spark/launcher/SparkSubmitArgumentsParser.scala new file mode 100644 index 0000000000000..a835012531052 --- /dev/null +++ b/core/src/main/scala/org/apache/spark/launcher/SparkSubmitArgumentsParser.scala @@ -0,0 +1,25 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher + +/** + * This class makes SparkSubmitOptionParser visible for Spark code outside of the `launcher` + * package, since Java doesn't have a feature similar to `private[spark]`, and we don't want + * that class to be public. + */ +private[spark] abstract class SparkSubmitArgumentsParser extends SparkSubmitOptionParser diff --git a/core/src/main/scala/org/apache/spark/launcher/WorkerCommandBuilder.scala b/core/src/main/scala/org/apache/spark/launcher/WorkerCommandBuilder.scala new file mode 100644 index 0000000000000..9be98723aed14 --- /dev/null +++ b/core/src/main/scala/org/apache/spark/launcher/WorkerCommandBuilder.scala @@ -0,0 +1,50 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher + +import java.io.File +import java.util.{HashMap => JHashMap, List => JList, Map => JMap} + +import scala.collection.JavaConversions._ + +import org.apache.spark.deploy.Command + +/** + * This class is used by CommandUtils. It uses some package-private APIs in SparkLauncher, and since + * Java doesn't have a feature similar to `private[spark]`, and we don't want that class to be + * public, needs to live in the same package as the rest of the library. + */ +private[spark] class WorkerCommandBuilder(sparkHome: String, memoryMb: Int, command: Command) + extends AbstractCommandBuilder { + + childEnv.putAll(command.environment) + childEnv.put(CommandBuilderUtils.ENV_SPARK_HOME, sparkHome) + + override def buildCommand(env: JMap[String, String]): JList[String] = { + val cmd = buildJavaCommand(command.classPathEntries.mkString(File.pathSeparator)) + cmd.add(s"-Xms${memoryMb}M") + cmd.add(s"-Xmx${memoryMb}M") + command.javaOpts.foreach(cmd.add) + addPermGenSizeOpt(cmd) + addOptionString(cmd, getenv("SPARK_JAVA_OPTS")) + cmd + } + + def buildCommand(): JList[String] = buildCommand(new JHashMap[String, String]()) + +} diff --git a/docs/programming-guide.md b/docs/programming-guide.md index b5e04bd0c610d..fa0b4e3705d6e 100644 --- a/docs/programming-guide.md +++ b/docs/programming-guide.md @@ -1369,6 +1369,11 @@ The [application submission guide](submitting-applications.html) describes how t In short, once you package your application into a JAR (for Java/Scala) or a set of `.py` or `.zip` files (for Python), the `bin/spark-submit` script lets you submit it to any supported cluster manager. +# Launching Spark jobs from Java / Scala + +The [org.apache.spark.launcher](api/java/index.html?org/apache/spark/launcher/package-summary.html) +package provides classes for launching Spark jobs as child processes using a simple Java API. + # Unit Testing Spark is friendly to unit testing with any popular unit test framework. diff --git a/launcher/pom.xml b/launcher/pom.xml new file mode 100644 index 0000000000000..ccbd9d0419a98 --- /dev/null +++ b/launcher/pom.xml @@ -0,0 +1,83 @@ + + + + + 4.0.0 + + org.apache.spark + spark-parent_2.10 + 1.3.0-SNAPSHOT + ../pom.xml + + + org.apache.spark + spark-launcher_2.10 + jar + Spark Launcher Project + http://spark.apache.org/ + + launcher + + + + + + log4j + log4j + test + + + junit + junit + test + + + org.mockito + mockito-all + test + + + org.scalatest + scalatest_${scala.binary.version} + test + + + org.slf4j + slf4j-api + test + + + org.slf4j + slf4j-log4j12 + test + + + + + org.apache.hadoop + hadoop-client + test + + + + + target/scala-${scala.binary.version}/classes + target/scala-${scala.binary.version}/test-classes + + diff --git a/launcher/src/main/java/org/apache/spark/launcher/AbstractCommandBuilder.java b/launcher/src/main/java/org/apache/spark/launcher/AbstractCommandBuilder.java new file mode 100644 index 0000000000000..dc90e9e987234 --- /dev/null +++ b/launcher/src/main/java/org/apache/spark/launcher/AbstractCommandBuilder.java @@ -0,0 +1,362 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.io.BufferedReader; +import java.io.File; +import java.io.FileFilter; +import java.io.FileInputStream; +import java.io.InputStreamReader; +import java.io.IOException; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collections; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.Properties; +import java.util.jar.JarFile; +import java.util.regex.Pattern; + +import static org.apache.spark.launcher.CommandBuilderUtils.*; + +/** + * Abstract Spark command builder that defines common functionality. + */ +abstract class AbstractCommandBuilder { + + boolean verbose; + String appName; + String appResource; + String deployMode; + String javaHome; + String mainClass; + String master; + String propertiesFile; + final List appArgs; + final List jars; + final List files; + final List pyFiles; + final Map childEnv; + final Map conf; + + public AbstractCommandBuilder() { + this.appArgs = new ArrayList(); + this.childEnv = new HashMap(); + this.conf = new HashMap(); + this.files = new ArrayList(); + this.jars = new ArrayList(); + this.pyFiles = new ArrayList(); + } + + /** + * Builds the command to execute. + * + * @param env A map containing environment variables for the child process. It may already contain + * entries defined by the user (such as SPARK_HOME, or those defined by the + * SparkLauncher constructor that takes an environment), and may be modified to + * include other variables needed by the process to be executed. + */ + abstract List buildCommand(Map env) throws IOException; + + /** + * Builds a list of arguments to run java. + * + * This method finds the java executable to use and appends JVM-specific options for running a + * class with Spark in the classpath. It also loads options from the "java-opts" file in the + * configuration directory being used. + * + * Callers should still add at least the class to run, as well as any arguments to pass to the + * class. + */ + List buildJavaCommand(String extraClassPath) throws IOException { + List cmd = new ArrayList(); + if (javaHome == null) { + cmd.add(join(File.separator, System.getProperty("java.home"), "bin", "java")); + } else { + cmd.add(join(File.separator, javaHome, "bin", "java")); + } + + // Load extra JAVA_OPTS from conf/java-opts, if it exists. + File javaOpts = new File(join(File.separator, getConfDir(), "java-opts")); + if (javaOpts.isFile()) { + BufferedReader br = new BufferedReader(new InputStreamReader( + new FileInputStream(javaOpts), "UTF-8")); + try { + String line; + while ((line = br.readLine()) != null) { + addOptionString(cmd, line); + } + } finally { + br.close(); + } + } + + cmd.add("-cp"); + cmd.add(join(File.pathSeparator, buildClassPath(extraClassPath))); + return cmd; + } + + /** + * Adds the default perm gen size option for Spark if the VM requires it and the user hasn't + * set it. + */ + void addPermGenSizeOpt(List cmd) { + // Don't set MaxPermSize for Java 8 and later. + String[] version = System.getProperty("java.version").split("\\."); + if (Integer.parseInt(version[0]) > 1 || Integer.parseInt(version[1]) > 7) { + return; + } + + for (String arg : cmd) { + if (arg.startsWith("-XX:MaxPermSize=")) { + return; + } + } + + cmd.add("-XX:MaxPermSize=128m"); + } + + void addOptionString(List cmd, String options) { + if (!isEmpty(options)) { + for (String opt : parseOptionString(options)) { + cmd.add(opt); + } + } + } + + /** + * Builds the classpath for the application. Returns a list with one classpath entry per element; + * each entry is formatted in the way expected by java.net.URLClassLoader (more + * specifically, with trailing slashes for directories). + */ + List buildClassPath(String appClassPath) throws IOException { + String sparkHome = getSparkHome(); + String scala = getScalaVersion(); + + List cp = new ArrayList(); + addToClassPath(cp, getenv("SPARK_CLASSPATH")); + addToClassPath(cp, appClassPath); + + addToClassPath(cp, getConfDir()); + + boolean prependClasses = !isEmpty(getenv("SPARK_PREPEND_CLASSES")); + boolean isTesting = "1".equals(getenv("SPARK_TESTING")); + if (prependClasses || isTesting) { + List projects = Arrays.asList("core", "repl", "mllib", "bagel", "graphx", + "streaming", "tools", "sql/catalyst", "sql/core", "sql/hive", "sql/hive-thriftserver", + "yarn", "launcher"); + if (prependClasses) { + System.err.println( + "NOTE: SPARK_PREPEND_CLASSES is set, placing locally compiled Spark classes ahead of " + + "assembly."); + for (String project : projects) { + addToClassPath(cp, String.format("%s/%s/target/scala-%s/classes", sparkHome, project, + scala)); + } + } + if (isTesting) { + for (String project : projects) { + addToClassPath(cp, String.format("%s/%s/target/scala-%s/test-classes", sparkHome, + project, scala)); + } + } + + // Add this path to include jars that are shaded in the final deliverable created during + // the maven build. These jars are copied to this directory during the build. + addToClassPath(cp, String.format("%s/core/target/jars/*", sparkHome)); + } + + String assembly = findAssembly(scala); + addToClassPath(cp, assembly); + + // When Hive support is needed, Datanucleus jars must be included on the classpath. Datanucleus + // jars do not work if only included in the uber jar as plugin.xml metadata is lost. Both sbt + // and maven will populate "lib_managed/jars/" with the datanucleus jars when Spark is built + // with Hive, so first check if the datanucleus jars exist, and then ensure the current Spark + // assembly is built for Hive, before actually populating the CLASSPATH with the jars. + // + // This block also serves as a check for SPARK-1703, when the assembly jar is built with + // Java 7 and ends up with too many files, causing issues with other JDK versions. + boolean needsDataNucleus = false; + JarFile assemblyJar = null; + try { + assemblyJar = new JarFile(assembly); + needsDataNucleus = assemblyJar.getEntry("org/apache/hadoop/hive/ql/exec/") != null; + } catch (IOException ioe) { + if (ioe.getMessage().indexOf("invalid CEN header") >= 0) { + System.err.println( + "Loading Spark jar failed.\n" + + "This is likely because Spark was compiled with Java 7 and run\n" + + "with Java 6 (see SPARK-1703). Please use Java 7 to run Spark\n" + + "or build Spark with Java 6."); + System.exit(1); + } else { + throw ioe; + } + } finally { + if (assemblyJar != null) { + try { + assemblyJar.close(); + } catch (IOException e) { + // Ignore. + } + } + } + + if (needsDataNucleus) { + System.err.println("Spark assembly has been built with Hive, including Datanucleus jars " + + "in classpath."); + File libdir; + if (new File(sparkHome, "RELEASE").isFile()) { + libdir = new File(sparkHome, "lib"); + } else { + libdir = new File(sparkHome, "lib_managed/jars"); + } + + checkState(libdir.isDirectory(), "Library directory '%s' does not exist.", + libdir.getAbsolutePath()); + for (File jar : libdir.listFiles()) { + if (jar.getName().startsWith("datanucleus-")) { + addToClassPath(cp, jar.getAbsolutePath()); + } + } + } + + addToClassPath(cp, getenv("HADOOP_CONF_DIR")); + addToClassPath(cp, getenv("YARN_CONF_DIR")); + addToClassPath(cp, getenv("SPARK_DIST_CLASSPATH")); + return cp; + } + + /** + * Adds entries to the classpath. + * + * @param cp List to which the new entries are appended. + * @param entries New classpath entries (separated by File.pathSeparator). + */ + private void addToClassPath(List cp, String entries) { + if (isEmpty(entries)) { + return; + } + String[] split = entries.split(Pattern.quote(File.pathSeparator)); + for (String entry : split) { + if (!isEmpty(entry)) { + if (new File(entry).isDirectory() && !entry.endsWith(File.separator)) { + entry += File.separator; + } + cp.add(entry); + } + } + } + + String getScalaVersion() { + String scala = getenv("SPARK_SCALA_VERSION"); + if (scala != null) { + return scala; + } + + String sparkHome = getSparkHome(); + File scala210 = new File(sparkHome, "assembly/target/scala-2.10"); + File scala211 = new File(sparkHome, "assembly/target/scala-2.11"); + checkState(!scala210.isDirectory() || !scala211.isDirectory(), + "Presence of build for both scala versions (2.10 and 2.11) detected.\n" + + "Either clean one of them or set SPARK_SCALA_VERSION in your environment."); + if (scala210.isDirectory()) { + return "2.10"; + } else { + checkState(scala211.isDirectory(), "Cannot find any assembly build directories."); + return "2.11"; + } + } + + String getSparkHome() { + String path = getenv(ENV_SPARK_HOME); + checkState(path != null, + "Spark home not found; set it explicitly or use the SPARK_HOME environment variable."); + return path; + } + + /** + * Loads the configuration file for the application, if it exists. This is either the + * user-specified properties file, or the spark-defaults.conf file under the Spark configuration + * directory. + */ + Properties loadPropertiesFile() throws IOException { + Properties props = new Properties(); + File propsFile; + if (propertiesFile != null) { + propsFile = new File(propertiesFile); + checkArgument(propsFile.isFile(), "Invalid properties file '%s'.", propertiesFile); + } else { + propsFile = new File(getConfDir(), DEFAULT_PROPERTIES_FILE); + } + + if (propsFile.isFile()) { + FileInputStream fd = null; + try { + fd = new FileInputStream(propsFile); + props.load(new InputStreamReader(fd, "UTF-8")); + } finally { + if (fd != null) { + try { + fd.close(); + } catch (IOException e) { + // Ignore. + } + } + } + } + + return props; + } + + String getenv(String key) { + return firstNonEmpty(childEnv.get(key), System.getenv(key)); + } + + private String findAssembly(String scalaVersion) { + String sparkHome = getSparkHome(); + File libdir; + if (new File(sparkHome, "RELEASE").isFile()) { + libdir = new File(sparkHome, "lib"); + checkState(libdir.isDirectory(), "Library directory '%s' does not exist.", + libdir.getAbsolutePath()); + } else { + libdir = new File(sparkHome, String.format("assembly/target/scala-%s", scalaVersion)); + } + + final Pattern re = Pattern.compile("spark-assembly.*hadoop.*\\.jar"); + FileFilter filter = new FileFilter() { + @Override + public boolean accept(File file) { + return file.isFile() && re.matcher(file.getName()).matches(); + } + }; + File[] assemblies = libdir.listFiles(filter); + checkState(assemblies != null && assemblies.length > 0, "No assemblies found in '%s'.", libdir); + checkState(assemblies.length == 1, "Multiple assemblies found in '%s'.", libdir); + return assemblies[0].getAbsolutePath(); + } + + private String getConfDir() { + String confDir = getenv("SPARK_CONF_DIR"); + return confDir != null ? confDir : join(File.separator, getSparkHome(), "conf"); + } + +} diff --git a/launcher/src/main/java/org/apache/spark/launcher/CommandBuilderUtils.java b/launcher/src/main/java/org/apache/spark/launcher/CommandBuilderUtils.java new file mode 100644 index 0000000000000..9b04732afee14 --- /dev/null +++ b/launcher/src/main/java/org/apache/spark/launcher/CommandBuilderUtils.java @@ -0,0 +1,296 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.io.File; +import java.util.ArrayList; +import java.util.List; +import java.util.Map; + +/** + * Helper methods for command builders. + */ +class CommandBuilderUtils { + + static final String DEFAULT_MEM = "512m"; + static final String DEFAULT_PROPERTIES_FILE = "spark-defaults.conf"; + static final String ENV_SPARK_HOME = "SPARK_HOME"; + + /** Returns whether the given string is null or empty. */ + static boolean isEmpty(String s) { + return s == null || s.isEmpty(); + } + + /** Joins a list of strings using the given separator. */ + static String join(String sep, String... elements) { + StringBuilder sb = new StringBuilder(); + for (String e : elements) { + if (e != null) { + if (sb.length() > 0) { + sb.append(sep); + } + sb.append(e); + } + } + return sb.toString(); + } + + /** Joins a list of strings using the given separator. */ + static String join(String sep, Iterable elements) { + StringBuilder sb = new StringBuilder(); + for (String e : elements) { + if (e != null) { + if (sb.length() > 0) { + sb.append(sep); + } + sb.append(e); + } + } + return sb.toString(); + } + + /** + * Returns the first non-empty value mapped to the given key in the given maps, or null otherwise. + */ + static String firstNonEmptyValue(String key, Map... maps) { + for (Map map : maps) { + String value = (String) map.get(key); + if (!isEmpty(value)) { + return value; + } + } + return null; + } + + /** Returns the first non-empty, non-null string in the given list, or null otherwise. */ + static String firstNonEmpty(String... candidates) { + for (String s : candidates) { + if (!isEmpty(s)) { + return s; + } + } + return null; + } + + /** Returns the name of the env variable that holds the native library path. */ + static String getLibPathEnvName() { + if (isWindows()) { + return "PATH"; + } + + String os = System.getProperty("os.name"); + if (os.startsWith("Mac OS X")) { + return "DYLD_LIBRARY_PATH"; + } else { + return "LD_LIBRARY_PATH"; + } + } + + /** Returns whether the OS is Windows. */ + static boolean isWindows() { + String os = System.getProperty("os.name"); + return os.startsWith("Windows"); + } + + /** + * Updates the user environment, appending the given pathList to the existing value of the given + * environment variable (or setting it if it hasn't yet been set). + */ + static void mergeEnvPathList(Map userEnv, String envKey, String pathList) { + if (!isEmpty(pathList)) { + String current = firstNonEmpty(userEnv.get(envKey), System.getenv(envKey)); + userEnv.put(envKey, join(File.pathSeparator, current, pathList)); + } + } + + /** + * Parse a string as if it were a list of arguments, following bash semantics. + * For example: + * + * Input: "\"ab cd\" efgh 'i \" j'" + * Output: [ "ab cd", "efgh", "i \" j" ] + */ + static List parseOptionString(String s) { + List opts = new ArrayList(); + StringBuilder opt = new StringBuilder(); + boolean inOpt = false; + boolean inSingleQuote = false; + boolean inDoubleQuote = false; + boolean escapeNext = false; + + // This is needed to detect when a quoted empty string is used as an argument ("" or ''). + boolean hasData = false; + + for (int i = 0; i < s.length(); i++) { + int c = s.codePointAt(i); + if (escapeNext) { + opt.appendCodePoint(c); + escapeNext = false; + } else if (inOpt) { + switch (c) { + case '\\': + if (inSingleQuote) { + opt.appendCodePoint(c); + } else { + escapeNext = true; + } + break; + case '\'': + if (inDoubleQuote) { + opt.appendCodePoint(c); + } else { + inSingleQuote = !inSingleQuote; + } + break; + case '"': + if (inSingleQuote) { + opt.appendCodePoint(c); + } else { + inDoubleQuote = !inDoubleQuote; + } + break; + default: + if (!Character.isWhitespace(c) || inSingleQuote || inDoubleQuote) { + opt.appendCodePoint(c); + } else { + opts.add(opt.toString()); + opt.setLength(0); + inOpt = false; + hasData = false; + } + } + } else { + switch (c) { + case '\'': + inSingleQuote = true; + inOpt = true; + hasData = true; + break; + case '"': + inDoubleQuote = true; + inOpt = true; + hasData = true; + break; + case '\\': + escapeNext = true; + inOpt = true; + hasData = true; + break; + default: + if (!Character.isWhitespace(c)) { + inOpt = true; + hasData = true; + opt.appendCodePoint(c); + } + } + } + } + + checkArgument(!inSingleQuote && !inDoubleQuote && !escapeNext, "Invalid option string: %s", s); + if (hasData) { + opts.add(opt.toString()); + } + return opts; + } + + /** Throws IllegalArgumentException if the given object is null. */ + static void checkNotNull(Object o, String arg) { + if (o == null) { + throw new IllegalArgumentException(String.format("'%s' must not be null.", arg)); + } + } + + /** Throws IllegalArgumentException with the given message if the check is false. */ + static void checkArgument(boolean check, String msg, Object... args) { + if (!check) { + throw new IllegalArgumentException(String.format(msg, args)); + } + } + + /** Throws IllegalStateException with the given message if the check is false. */ + static void checkState(boolean check, String msg, Object... args) { + if (!check) { + throw new IllegalStateException(String.format(msg, args)); + } + } + + /** + * Quote a command argument for a command to be run by a Windows batch script, if the argument + * needs quoting. Arguments only seem to need quotes in batch scripts if they have certain + * special characters, some of which need extra (and different) escaping. + * + * For example: + * original single argument: ab="cde fgh" + * quoted: "ab^=""cde fgh""" + */ + static String quoteForBatchScript(String arg) { + + boolean needsQuotes = false; + for (int i = 0; i < arg.length(); i++) { + int c = arg.codePointAt(i); + if (Character.isWhitespace(c) || c == '"' || c == '=') { + needsQuotes = true; + break; + } + } + if (!needsQuotes) { + return arg; + } + StringBuilder quoted = new StringBuilder(); + quoted.append("\""); + for (int i = 0; i < arg.length(); i++) { + int cp = arg.codePointAt(i); + switch (cp) { + case '"': + quoted.append('"'); + break; + + case '=': + quoted.append('^'); + break; + + default: + break; + } + quoted.appendCodePoint(cp); + } + quoted.append("\""); + return quoted.toString(); + } + + /** + * Quotes a string so that it can be used in a command string and be parsed back into a single + * argument by python's "shlex.split()" function. + * + * Basically, just add simple escapes. E.g.: + * original single argument : ab "cd" ef + * after: "ab \"cd\" ef" + */ + static String quoteForPython(String s) { + StringBuilder quoted = new StringBuilder().append('"'); + for (int i = 0; i < s.length(); i++) { + int cp = s.codePointAt(i); + if (cp == '"' || cp == '\\') { + quoted.appendCodePoint('\\'); + } + quoted.appendCodePoint(cp); + } + return quoted.append('"').toString(); + } + +} diff --git a/launcher/src/main/java/org/apache/spark/launcher/Main.java b/launcher/src/main/java/org/apache/spark/launcher/Main.java new file mode 100644 index 0000000000000..206acfb514d86 --- /dev/null +++ b/launcher/src/main/java/org/apache/spark/launcher/Main.java @@ -0,0 +1,173 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.util.ArrayList; +import java.util.Arrays; +import java.util.HashMap; +import java.util.List; +import java.util.Map; + +import static org.apache.spark.launcher.CommandBuilderUtils.*; + +/** + * Command line interface for the Spark launcher. Used internally by Spark scripts. + */ +class Main { + + /** + * Usage: Main [class] [class args] + *

    + * This CLI works in two different modes: + *

      + *
    • "spark-submit": if class is "org.apache.spark.deploy.SparkSubmit", the + * {@link SparkLauncher} class is used to launch a Spark application.
      • + *
    • "spark-class": if another class is provided, an internal Spark class is run.
      • + *
    + * + * This class works in tandem with the "bin/spark-class" script on Unix-like systems, and + * "bin/spark-class2.cmd" batch script on Windows to execute the final command. + *

    + * On Unix-like systems, the output is a list of command arguments, separated by the NULL + * character. On Windows, the output is a command line suitable for direct execution from the + * script. + */ + public static void main(String[] argsArray) throws Exception { + checkArgument(argsArray.length > 0, "Not enough arguments: missing class name."); + + List args = new ArrayList(Arrays.asList(argsArray)); + String className = args.remove(0); + + boolean printLaunchCommand; + boolean printUsage; + AbstractCommandBuilder builder; + try { + if (className.equals("org.apache.spark.deploy.SparkSubmit")) { + builder = new SparkSubmitCommandBuilder(args); + } else { + builder = new SparkClassCommandBuilder(className, args); + } + printLaunchCommand = !isEmpty(System.getenv("SPARK_PRINT_LAUNCH_COMMAND")); + printUsage = false; + } catch (IllegalArgumentException e) { + builder = new UsageCommandBuilder(e.getMessage()); + printLaunchCommand = false; + printUsage = true; + } + + Map env = new HashMap(); + List cmd = builder.buildCommand(env); + if (printLaunchCommand) { + System.err.println("Spark Command: " + join(" ", cmd)); + System.err.println("========================================"); + } + + if (isWindows()) { + // When printing the usage message, we can't use "cmd /v" since that prevents the env + // variable from being seen in the caller script. So do not call prepareWindowsCommand(). + if (printUsage) { + System.out.println(join(" ", cmd)); + } else { + System.out.println(prepareWindowsCommand(cmd, env)); + } + } else { + // In bash, use NULL as the arg separator since it cannot be used in an argument. + List bashCmd = prepareBashCommand(cmd, env); + for (String c : bashCmd) { + System.out.print(c); + System.out.print('\0'); + } + } + } + + /** + * Prepare a command line for execution from a Windows batch script. + * + * The method quotes all arguments so that spaces are handled as expected. Quotes within arguments + * are "double quoted" (which is batch for escaping a quote). This page has more details about + * quoting and other batch script fun stuff: http://ss64.com/nt/syntax-esc.html + * + * The command is executed using "cmd /c" and formatted in single line, since that's the + * easiest way to consume this from a batch script (see spark-class2.cmd). + */ + private static String prepareWindowsCommand(List cmd, Map childEnv) { + StringBuilder cmdline = new StringBuilder("cmd /c \""); + for (Map.Entry e : childEnv.entrySet()) { + cmdline.append(String.format("set %s=%s", e.getKey(), e.getValue())); + cmdline.append(" && "); + } + for (String arg : cmd) { + cmdline.append(quoteForBatchScript(arg)); + cmdline.append(" "); + } + cmdline.append("\""); + return cmdline.toString(); + } + + /** + * Prepare the command for execution from a bash script. The final command will have commands to + * set up any needed environment variables needed by the child process. + */ + private static List prepareBashCommand(List cmd, Map childEnv) { + if (childEnv.isEmpty()) { + return cmd; + } + + List newCmd = new ArrayList(); + newCmd.add("env"); + + for (Map.Entry e : childEnv.entrySet()) { + newCmd.add(String.format("%s=%s", e.getKey(), e.getValue())); + } + newCmd.addAll(cmd); + return newCmd; + } + + /** + * Internal builder used when command line parsing fails. This will behave differently depending + * on the platform: + * + * - On Unix-like systems, it will print a call to the "usage" function with two arguments: the + * the error string, and the exit code to use. The function is expected to print the command's + * usage and exit with the provided exit code. The script should use "export -f usage" after + * declaring a function called "usage", so that the function is available to downstream scripts. + * + * - On Windows it will set the variable "SPARK_LAUNCHER_USAGE_ERROR" to the usage error message. + * The batch script should check for this variable and print its usage, since batch scripts + * don't really support the "export -f" functionality used in bash. + */ + private static class UsageCommandBuilder extends AbstractCommandBuilder { + + private final String message; + + UsageCommandBuilder(String message) { + this.message = message; + } + + @Override + public List buildCommand(Map env) { + if (isWindows()) { + return Arrays.asList("set", "SPARK_LAUNCHER_USAGE_ERROR=" + message); + } else { + return Arrays.asList("usage", message, "1"); + } + } + + } + +} diff --git a/launcher/src/main/java/org/apache/spark/launcher/SparkClassCommandBuilder.java b/launcher/src/main/java/org/apache/spark/launcher/SparkClassCommandBuilder.java new file mode 100644 index 0000000000000..e601a0a19f368 --- /dev/null +++ b/launcher/src/main/java/org/apache/spark/launcher/SparkClassCommandBuilder.java @@ -0,0 +1,108 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.io.File; +import java.io.IOException; +import java.util.ArrayList; +import java.util.List; +import java.util.Map; +import java.util.regex.Pattern; + +import static org.apache.spark.launcher.CommandBuilderUtils.*; + +/** + * Command builder for internal Spark classes. + *

    + * This class handles building the command to launch all internal Spark classes except for + * SparkSubmit (which is handled by {@link SparkSubmitCommandBuilder} class. + */ +class SparkClassCommandBuilder extends AbstractCommandBuilder { + + private final String className; + private final List classArgs; + + SparkClassCommandBuilder(String className, List classArgs) { + this.className = className; + this.classArgs = classArgs; + } + + @Override + public List buildCommand(Map env) throws IOException { + List javaOptsKeys = new ArrayList(); + String memKey = null; + String extraClassPath = null; + + // Master, Worker, and HistoryServer use SPARK_DAEMON_JAVA_OPTS (and specific opts) + + // SPARK_DAEMON_MEMORY. + if (className.equals("org.apache.spark.deploy.master.Master")) { + javaOptsKeys.add("SPARK_DAEMON_JAVA_OPTS"); + javaOptsKeys.add("SPARK_MASTER_OPTS"); + memKey = "SPARK_DAEMON_MEMORY"; + } else if (className.equals("org.apache.spark.deploy.worker.Worker")) { + javaOptsKeys.add("SPARK_DAEMON_JAVA_OPTS"); + javaOptsKeys.add("SPARK_WORKER_OPTS"); + memKey = "SPARK_DAEMON_MEMORY"; + } else if (className.equals("org.apache.spark.deploy.history.HistoryServer")) { + javaOptsKeys.add("SPARK_DAEMON_JAVA_OPTS"); + javaOptsKeys.add("SPARK_HISTORY_OPTS"); + memKey = "SPARK_DAEMON_MEMORY"; + } else if (className.equals("org.apache.spark.executor.CoarseGrainedExecutorBackend")) { + javaOptsKeys.add("SPARK_JAVA_OPTS"); + javaOptsKeys.add("SPARK_EXECUTOR_OPTS"); + memKey = "SPARK_EXECUTOR_MEMORY"; + } else if (className.equals("org.apache.spark.executor.MesosExecutorBackend")) { + javaOptsKeys.add("SPARK_EXECUTOR_OPTS"); + memKey = "SPARK_EXECUTOR_MEMORY"; + } else if (className.startsWith("org.apache.spark.tools.")) { + String sparkHome = getSparkHome(); + File toolsDir = new File(join(File.separator, sparkHome, "tools", "target", + "scala-" + getScalaVersion())); + checkState(toolsDir.isDirectory(), "Cannot find tools build directory."); + + Pattern re = Pattern.compile("spark-tools_.*\\.jar"); + for (File f : toolsDir.listFiles()) { + if (re.matcher(f.getName()).matches()) { + extraClassPath = f.getAbsolutePath(); + break; + } + } + + checkState(extraClassPath != null, + "Failed to find Spark Tools Jar in %s.\n" + + "You need to run \"build/sbt tools/package\" before running %s.", + toolsDir.getAbsolutePath(), className); + + javaOptsKeys.add("SPARK_JAVA_OPTS"); + } + + List cmd = buildJavaCommand(extraClassPath); + for (String key : javaOptsKeys) { + addOptionString(cmd, System.getenv(key)); + } + + String mem = firstNonEmpty(memKey != null ? System.getenv(memKey) : null, DEFAULT_MEM); + cmd.add("-Xms" + mem); + cmd.add("-Xmx" + mem); + addPermGenSizeOpt(cmd); + cmd.add(className); + cmd.addAll(classArgs); + return cmd; + } + +} diff --git a/launcher/src/main/java/org/apache/spark/launcher/SparkLauncher.java b/launcher/src/main/java/org/apache/spark/launcher/SparkLauncher.java new file mode 100644 index 0000000000000..b566507ee6061 --- /dev/null +++ b/launcher/src/main/java/org/apache/spark/launcher/SparkLauncher.java @@ -0,0 +1,279 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.io.File; +import java.io.IOException; +import java.util.ArrayList; +import java.util.List; +import java.util.Map; + +import static org.apache.spark.launcher.CommandBuilderUtils.*; + +/** + * Launcher for Spark applications. + *

    + * Use this class to start Spark applications programmatically. The class uses a builder pattern + * to allow clients to configure the Spark application and launch it as a child process. + */ +public class SparkLauncher { + + /** The Spark master. */ + public static final String SPARK_MASTER = "spark.master"; + + /** Configuration key for the driver memory. */ + public static final String DRIVER_MEMORY = "spark.driver.memory"; + /** Configuration key for the driver class path. */ + public static final String DRIVER_EXTRA_CLASSPATH = "spark.driver.extraClassPath"; + /** Configuration key for the driver VM options. */ + public static final String DRIVER_EXTRA_JAVA_OPTIONS = "spark.driver.extraJavaOptions"; + /** Configuration key for the driver native library path. */ + public static final String DRIVER_EXTRA_LIBRARY_PATH = "spark.driver.extraLibraryPath"; + + /** Configuration key for the executor memory. */ + public static final String EXECUTOR_MEMORY = "spark.executor.memory"; + /** Configuration key for the executor class path. */ + public static final String EXECUTOR_EXTRA_CLASSPATH = "spark.executor.extraClassPath"; + /** Configuration key for the executor VM options. */ + public static final String EXECUTOR_EXTRA_JAVA_OPTIONS = "spark.executor.extraJavaOptions"; + /** Configuration key for the executor native library path. */ + public static final String EXECUTOR_EXTRA_LIBRARY_PATH = "spark.executor.extraLibraryOptions"; + /** Configuration key for the number of executor CPU cores. */ + public static final String EXECUTOR_CORES = "spark.executor.cores"; + + private final SparkSubmitCommandBuilder builder; + + public SparkLauncher() { + this(null); + } + + /** + * Creates a launcher that will set the given environment variables in the child. + * + * @param env Environment variables to set. + */ + public SparkLauncher(Map env) { + this.builder = new SparkSubmitCommandBuilder(); + if (env != null) { + this.builder.childEnv.putAll(env); + } + } + + /** + * Set a custom JAVA_HOME for launching the Spark application. + * + * @param javaHome Path to the JAVA_HOME to use. + * @return This launcher. + */ + public SparkLauncher setJavaHome(String javaHome) { + checkNotNull(javaHome, "javaHome"); + builder.javaHome = javaHome; + return this; + } + + /** + * Set a custom Spark installation location for the application. + * + * @param sparkHome Path to the Spark installation to use. + * @return This launcher. + */ + public SparkLauncher setSparkHome(String sparkHome) { + checkNotNull(sparkHome, "sparkHome"); + builder.childEnv.put(ENV_SPARK_HOME, sparkHome); + return this; + } + + /** + * Set a custom properties file with Spark configuration for the application. + * + * @param path Path to custom properties file to use. + * @return This launcher. + */ + public SparkLauncher setPropertiesFile(String path) { + checkNotNull(path, "path"); + builder.propertiesFile = path; + return this; + } + + /** + * Set a single configuration value for the application. + * + * @param key Configuration key. + * @param value The value to use. + * @return This launcher. + */ + public SparkLauncher setConf(String key, String value) { + checkNotNull(key, "key"); + checkNotNull(value, "value"); + checkArgument(key.startsWith("spark."), "'key' must start with 'spark.'"); + builder.conf.put(key, value); + return this; + } + + /** + * Set the application name. + * + * @param appName Application name. + * @return This launcher. + */ + public SparkLauncher setAppName(String appName) { + checkNotNull(appName, "appName"); + builder.appName = appName; + return this; + } + + /** + * Set the Spark master for the application. + * + * @param master Spark master. + * @return This launcher. + */ + public SparkLauncher setMaster(String master) { + checkNotNull(master, "master"); + builder.master = master; + return this; + } + + /** + * Set the deploy mode for the application. + * + * @param mode Deploy mode. + * @return This launcher. + */ + public SparkLauncher setDeployMode(String mode) { + checkNotNull(mode, "mode"); + builder.deployMode = mode; + return this; + } + + /** + * Set the main application resource. This should be the location of a jar file for Scala/Java + * applications, or a python script for PySpark applications. + * + * @param resource Path to the main application resource. + * @return This launcher. + */ + public SparkLauncher setAppResource(String resource) { + checkNotNull(resource, "resource"); + builder.appResource = resource; + return this; + } + + /** + * Sets the application class name for Java/Scala applications. + * + * @param mainClass Application's main class. + * @return This launcher. + */ + public SparkLauncher setMainClass(String mainClass) { + checkNotNull(mainClass, "mainClass"); + builder.mainClass = mainClass; + return this; + } + + /** + * Adds command line arguments for the application. + * + * @param args Arguments to pass to the application's main class. + * @return This launcher. + */ + public SparkLauncher addAppArgs(String... args) { + for (String arg : args) { + checkNotNull(arg, "arg"); + builder.appArgs.add(arg); + } + return this; + } + + /** + * Adds a jar file to be submitted with the application. + * + * @param jar Path to the jar file. + * @return This launcher. + */ + public SparkLauncher addJar(String jar) { + checkNotNull(jar, "jar"); + builder.jars.add(jar); + return this; + } + + /** + * Adds a file to be submitted with the application. + * + * @param file Path to the file. + * @return This launcher. + */ + public SparkLauncher addFile(String file) { + checkNotNull(file, "file"); + builder.files.add(file); + return this; + } + + /** + * Adds a python file / zip / egg to be submitted with the application. + * + * @param file Path to the file. + * @return This launcher. + */ + public SparkLauncher addPyFile(String file) { + checkNotNull(file, "file"); + builder.pyFiles.add(file); + return this; + } + + /** + * Enables verbose reporting for SparkSubmit. + * + * @param verbose Whether to enable verbose output. + * @return This launcher. + */ + public SparkLauncher setVerbose(boolean verbose) { + builder.verbose = verbose; + return this; + } + + /** + * Launches a sub-process that will start the configured Spark application. + * + * @return A process handle for the Spark app. + */ + public Process launch() throws IOException { + List cmd = new ArrayList(); + String script = isWindows() ? "spark-submit.cmd" : "spark-submit"; + cmd.add(join(File.separator, builder.getSparkHome(), "bin", script)); + cmd.addAll(builder.buildSparkSubmitArgs()); + + // Since the child process is a batch script, let's quote things so that special characters are + // preserved, otherwise the batch interpreter will mess up the arguments. Batch scripts are + // weird. + if (isWindows()) { + List winCmd = new ArrayList(); + for (String arg : cmd) { + winCmd.add(quoteForBatchScript(arg)); + } + cmd = winCmd; + } + + ProcessBuilder pb = new ProcessBuilder(cmd.toArray(new String[cmd.size()])); + for (Map.Entry e : builder.childEnv.entrySet()) { + pb.environment().put(e.getKey(), e.getValue()); + } + return pb.start(); + } + +} diff --git a/launcher/src/main/java/org/apache/spark/launcher/SparkSubmitCommandBuilder.java b/launcher/src/main/java/org/apache/spark/launcher/SparkSubmitCommandBuilder.java new file mode 100644 index 0000000000000..6ffdff63d3c78 --- /dev/null +++ b/launcher/src/main/java/org/apache/spark/launcher/SparkSubmitCommandBuilder.java @@ -0,0 +1,327 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collections; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.Properties; + +import static org.apache.spark.launcher.CommandBuilderUtils.*; + +/** + * Special command builder for handling a CLI invocation of SparkSubmit. + *

    + * This builder adds command line parsing compatible with SparkSubmit. It handles setting + * driver-side options and special parsing behavior needed for the special-casing certain internal + * Spark applications. + *

    + * This class has also some special features to aid launching pyspark. + */ +class SparkSubmitCommandBuilder extends AbstractCommandBuilder { + + /** + * Name of the app resource used to identify the PySpark shell. The command line parser expects + * the resource name to be the very first argument to spark-submit in this case. + * + * NOTE: this cannot be "pyspark-shell" since that identifies the PySpark shell to SparkSubmit + * (see java_gateway.py), and can cause this code to enter into an infinite loop. + */ + static final String PYSPARK_SHELL = "pyspark-shell-main"; + + /** + * This is the actual resource name that identifies the PySpark shell to SparkSubmit. + */ + static final String PYSPARK_SHELL_RESOURCE = "pyspark-shell"; + + /** + * This map must match the class names for available special classes, since this modifies the way + * command line parsing works. This maps the class name to the resource to use when calling + * spark-submit. + */ + private static final Map specialClasses = new HashMap(); + static { + specialClasses.put("org.apache.spark.repl.Main", "spark-shell"); + specialClasses.put("org.apache.spark.sql.hive.thriftserver.SparkSQLCLIDriver", + "spark-internal"); + specialClasses.put("org.apache.spark.sql.hive.thriftserver.HiveThriftServer2", + "spark-internal"); + } + + private final List sparkArgs; + + /** + * Controls whether mixing spark-submit arguments with app arguments is allowed. This is needed + * to parse the command lines for things like bin/spark-shell, which allows users to mix and + * match arguments (e.g. "bin/spark-shell SparkShellArg --master foo"). + */ + private boolean allowsMixedArguments; + + SparkSubmitCommandBuilder() { + this.sparkArgs = new ArrayList(); + } + + SparkSubmitCommandBuilder(List args) { + this(); + List submitArgs = args; + if (args.size() > 0 && args.get(0).equals(PYSPARK_SHELL)) { + this.allowsMixedArguments = true; + appResource = PYSPARK_SHELL_RESOURCE; + submitArgs = args.subList(1, args.size()); + } else { + this.allowsMixedArguments = false; + } + + new OptionParser().parse(submitArgs); + } + + @Override + public List buildCommand(Map env) throws IOException { + if (PYSPARK_SHELL_RESOURCE.equals(appResource)) { + return buildPySparkShellCommand(env); + } else { + return buildSparkSubmitCommand(env); + } + } + + List buildSparkSubmitArgs() { + List args = new ArrayList(); + SparkSubmitOptionParser parser = new SparkSubmitOptionParser(); + + if (verbose) { + args.add(parser.VERBOSE); + } + + if (master != null) { + args.add(parser.MASTER); + args.add(master); + } + + if (deployMode != null) { + args.add(parser.DEPLOY_MODE); + args.add(deployMode); + } + + if (appName != null) { + args.add(parser.NAME); + args.add(appName); + } + + for (Map.Entry e : conf.entrySet()) { + args.add(parser.CONF); + args.add(String.format("%s=%s", e.getKey(), e.getValue())); + } + + if (propertiesFile != null) { + args.add(parser.PROPERTIES_FILE); + args.add(propertiesFile); + } + + if (!jars.isEmpty()) { + args.add(parser.JARS); + args.add(join(",", jars)); + } + + if (!files.isEmpty()) { + args.add(parser.FILES); + args.add(join(",", files)); + } + + if (!pyFiles.isEmpty()) { + args.add(parser.PY_FILES); + args.add(join(",", pyFiles)); + } + + if (mainClass != null) { + args.add(parser.CLASS); + args.add(mainClass); + } + + args.addAll(sparkArgs); + if (appResource != null) { + args.add(appResource); + } + args.addAll(appArgs); + + return args; + } + + private List buildSparkSubmitCommand(Map env) throws IOException { + // Load the properties file and check whether spark-submit will be running the app's driver + // or just launching a cluster app. When running the driver, the JVM's argument will be + // modified to cover the driver's configuration. + Properties props = loadPropertiesFile(); + boolean isClientMode = isClientMode(props); + String extraClassPath = isClientMode ? + firstNonEmptyValue(SparkLauncher.DRIVER_EXTRA_CLASSPATH, conf, props) : null; + + List cmd = buildJavaCommand(extraClassPath); + addOptionString(cmd, System.getenv("SPARK_SUBMIT_OPTS")); + addOptionString(cmd, System.getenv("SPARK_JAVA_OPTS")); + + if (isClientMode) { + // Figuring out where the memory value come from is a little tricky due to precedence. + // Precedence is observed in the following order: + // - explicit configuration (setConf()), which also covers --driver-memory cli argument. + // - properties file. + // - SPARK_DRIVER_MEMORY env variable + // - SPARK_MEM env variable + // - default value (512m) + String memory = firstNonEmpty(firstNonEmptyValue(SparkLauncher.DRIVER_MEMORY, conf, props), + System.getenv("SPARK_DRIVER_MEMORY"), System.getenv("SPARK_MEM"), DEFAULT_MEM); + cmd.add("-Xms" + memory); + cmd.add("-Xmx" + memory); + addOptionString(cmd, firstNonEmptyValue(SparkLauncher.DRIVER_EXTRA_JAVA_OPTIONS, conf, props)); + mergeEnvPathList(env, getLibPathEnvName(), + firstNonEmptyValue(SparkLauncher.DRIVER_EXTRA_LIBRARY_PATH, conf, props)); + } + + addPermGenSizeOpt(cmd); + cmd.add("org.apache.spark.deploy.SparkSubmit"); + cmd.addAll(buildSparkSubmitArgs()); + return cmd; + } + + private List buildPySparkShellCommand(Map env) throws IOException { + // For backwards compatibility, if a script is specified in + // the pyspark command line, then run it using spark-submit. + if (!appArgs.isEmpty() && appArgs.get(0).endsWith(".py")) { + System.err.println( + "WARNING: Running python applications through 'pyspark' is deprecated as of Spark 1.0.\n" + + "Use ./bin/spark-submit "); + appResource = appArgs.get(0); + appArgs.remove(0); + return buildCommand(env); + } + + // When launching the pyspark shell, the spark-submit arguments should be stored in the + // PYSPARK_SUBMIT_ARGS env variable. The executable is the PYSPARK_DRIVER_PYTHON env variable + // set by the pyspark script, followed by PYSPARK_DRIVER_PYTHON_OPTS. + checkArgument(appArgs.isEmpty(), "pyspark does not support any application options."); + + Properties props = loadPropertiesFile(); + mergeEnvPathList(env, getLibPathEnvName(), + firstNonEmptyValue(SparkLauncher.DRIVER_EXTRA_LIBRARY_PATH, conf, props)); + + // Store spark-submit arguments in an environment variable, since there's no way to pass + // them to shell.py on the comand line. + StringBuilder submitArgs = new StringBuilder(); + for (String arg : buildSparkSubmitArgs()) { + if (submitArgs.length() > 0) { + submitArgs.append(" "); + } + submitArgs.append(quoteForPython(arg)); + } + env.put("PYSPARK_SUBMIT_ARGS", submitArgs.toString()); + + List pyargs = new ArrayList(); + pyargs.add(firstNonEmpty(System.getenv("PYSPARK_DRIVER_PYTHON"), "python")); + String pyOpts = System.getenv("PYSPARK_DRIVER_PYTHON_OPTS"); + if (!isEmpty(pyOpts)) { + pyargs.addAll(parseOptionString(pyOpts)); + } + + return pyargs; + } + + private boolean isClientMode(Properties userProps) { + String userMaster = firstNonEmpty(master, (String) userProps.get(SparkLauncher.SPARK_MASTER)); + // Default master is "local[*]", so assume client mode in that case. + return userMaster == null || + "client".equals(deployMode) || + (!userMaster.equals("yarn-cluster") && deployMode == null); + } + + private class OptionParser extends SparkSubmitOptionParser { + + private final List driverJvmKeys = Arrays.asList( + SparkLauncher.DRIVER_EXTRA_CLASSPATH, + SparkLauncher.DRIVER_EXTRA_JAVA_OPTIONS, + SparkLauncher.DRIVER_EXTRA_LIBRARY_PATH, + SparkLauncher.DRIVER_MEMORY); + + @Override + protected boolean handle(String opt, String value) { + if (opt.equals(MASTER)) { + master = value; + } else if (opt.equals(DEPLOY_MODE)) { + deployMode = value; + } else if (opt.equals(PROPERTIES_FILE)) { + propertiesFile = value; + } else if (opt.equals(DRIVER_MEMORY)) { + conf.put(SparkLauncher.DRIVER_MEMORY, value); + } else if (opt.equals(DRIVER_JAVA_OPTIONS)) { + conf.put(SparkLauncher.DRIVER_EXTRA_JAVA_OPTIONS, value); + } else if (opt.equals(DRIVER_LIBRARY_PATH)) { + conf.put(SparkLauncher.DRIVER_EXTRA_LIBRARY_PATH, value); + } else if (opt.equals(DRIVER_CLASS_PATH)) { + conf.put(SparkLauncher.DRIVER_EXTRA_CLASSPATH, value); + } else if (opt.equals(CONF)) { + String[] setConf = value.split("=", 2); + checkArgument(setConf.length == 2, "Invalid argument to %s: %s", CONF, value); + if (driverJvmKeys.contains(setConf[0])) { + conf.put(setConf[0], setConf[1]); + } + } else if (opt.equals(CLASS)) { + // The special classes require some special command line handling, since they allow + // mixing spark-submit arguments with arguments that should be propagated to the shell + // itself. Note that for this to work, the "--class" argument must come before any + // non-spark-submit arguments. + mainClass = value; + if (specialClasses.containsKey(value)) { + allowsMixedArguments = true; + appResource = specialClasses.get(value); + } + } else { + sparkArgs.add(opt); + if (value != null) { + sparkArgs.add(value); + } + } + return true; + } + + @Override + protected boolean handleUnknown(String opt) { + // When mixing arguments, add unrecognized parameters directly to the user arguments list. In + // normal mode, any unrecognized parameter triggers the end of command line parsing, and the + // parameter itself will be interpreted by SparkSubmit as the application resource. The + // remaining params will be appended to the list of SparkSubmit arguments. + if (allowsMixedArguments) { + appArgs.add(opt); + return true; + } else { + sparkArgs.add(opt); + return false; + } + } + + @Override + protected void handleExtraArgs(List extra) { + for (String arg : extra) { + sparkArgs.add(arg); + } + } + + } + +} diff --git a/launcher/src/main/java/org/apache/spark/launcher/SparkSubmitOptionParser.java b/launcher/src/main/java/org/apache/spark/launcher/SparkSubmitOptionParser.java new file mode 100644 index 0000000000000..8526d2e7cfa3f --- /dev/null +++ b/launcher/src/main/java/org/apache/spark/launcher/SparkSubmitOptionParser.java @@ -0,0 +1,224 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.util.List; +import java.util.regex.Matcher; +import java.util.regex.Pattern; + +/** + * Parser for spark-submit command line options. + *

    + * This class encapsulates the parsing code for spark-submit command line options, so that there + * is a single list of options that needs to be maintained (well, sort of, but it makes it harder + * to break things). + */ +class SparkSubmitOptionParser { + + // The following constants define the "main" name for the available options. They're defined + // to avoid copy & paste of the raw strings where they're needed. + // + // The fields are not static so that they're exposed to Scala code that uses this class. See + // SparkSubmitArguments.scala. That is also why this class is not abstract - to allow code to + // easily use these constants without having to create dummy implementations of this class. + protected final String CLASS = "--class"; + protected final String CONF = "--conf"; + protected final String DEPLOY_MODE = "--deploy-mode"; + protected final String DRIVER_CLASS_PATH = "--driver-class-path"; + protected final String DRIVER_CORES = "--driver-cores"; + protected final String DRIVER_JAVA_OPTIONS = "--driver-java-options"; + protected final String DRIVER_LIBRARY_PATH = "--driver-library-path"; + protected final String DRIVER_MEMORY = "--driver-memory"; + protected final String EXECUTOR_MEMORY = "--executor-memory"; + protected final String FILES = "--files"; + protected final String JARS = "--jars"; + protected final String KILL_SUBMISSION = "--kill"; + protected final String MASTER = "--master"; + protected final String NAME = "--name"; + protected final String PACKAGES = "--packages"; + protected final String PROPERTIES_FILE = "--properties-file"; + protected final String PROXY_USER = "--proxy-user"; + protected final String PY_FILES = "--py-files"; + protected final String REPOSITORIES = "--repositories"; + protected final String STATUS = "--status"; + protected final String TOTAL_EXECUTOR_CORES = "--total-executor-cores"; + + // Options that do not take arguments. + protected final String HELP = "--help"; + protected final String SUPERVISE = "--supervise"; + protected final String VERBOSE = "--verbose"; + protected final String VERSION = "--version"; + + // Standalone-only options. + + // YARN-only options. + protected final String ARCHIVES = "--archives"; + protected final String EXECUTOR_CORES = "--executor-cores"; + protected final String QUEUE = "--queue"; + protected final String NUM_EXECUTORS = "--num-executors"; + + /** + * This is the canonical list of spark-submit options. Each entry in the array contains the + * different aliases for the same option; the first element of each entry is the "official" + * name of the option, passed to {@link #handle(String, String)}. + *

    + * Options not listed here nor in the "switch" list below will result in a call to + * {@link $#handleUnknown(String)}. + *

    + * These two arrays are visible for tests. + */ + final String[][] opts = { + { ARCHIVES }, + { CLASS }, + { CONF, "-c" }, + { DEPLOY_MODE }, + { DRIVER_CLASS_PATH }, + { DRIVER_CORES }, + { DRIVER_JAVA_OPTIONS }, + { DRIVER_LIBRARY_PATH }, + { DRIVER_MEMORY }, + { EXECUTOR_CORES }, + { EXECUTOR_MEMORY }, + { FILES }, + { JARS }, + { KILL_SUBMISSION }, + { MASTER }, + { NAME }, + { NUM_EXECUTORS }, + { PACKAGES }, + { PROPERTIES_FILE }, + { PROXY_USER }, + { PY_FILES }, + { QUEUE }, + { REPOSITORIES }, + { STATUS }, + { TOTAL_EXECUTOR_CORES }, + }; + + /** + * List of switches (command line options that do not take parameters) recognized by spark-submit. + */ + final String[][] switches = { + { HELP, "-h" }, + { SUPERVISE }, + { VERBOSE, "-v" }, + { VERSION }, + }; + + /** + * Parse a list of spark-submit command line options. + *

    + * See SparkSubmitArguments.scala for a more formal description of available options. + * + * @throws IllegalArgumentException If an error is found during parsing. + */ + protected final void parse(List args) { + Pattern eqSeparatedOpt = Pattern.compile("(--[^=]+)=(.+)"); + + int idx = 0; + for (idx = 0; idx < args.size(); idx++) { + String arg = args.get(idx); + String value = null; + + Matcher m = eqSeparatedOpt.matcher(arg); + if (m.matches()) { + arg = m.group(1); + value = m.group(2); + } + + // Look for options with a value. + String name = findCliOption(arg, opts); + if (name != null) { + if (value == null) { + if (idx == args.size() - 1) { + throw new IllegalArgumentException( + String.format("Missing argument for option '%s'.", arg)); + } + idx++; + value = args.get(idx); + } + if (!handle(name, value)) { + break; + } + continue; + } + + // Look for a switch. + name = findCliOption(arg, switches); + if (name != null) { + if (!handle(name, null)) { + break; + } + continue; + } + + if (!handleUnknown(arg)) { + break; + } + } + + if (idx < args.size()) { + idx++; + } + handleExtraArgs(args.subList(idx, args.size())); + } + + /** + * Callback for when an option with an argument is parsed. + * + * @param opt The long name of the cli option (might differ from actual command line). + * @param value The value. This will be null if the option does not take a value. + * @return Whether to continue parsing the argument list. + */ + protected boolean handle(String opt, String value) { + throw new UnsupportedOperationException(); + } + + /** + * Callback for when an unrecognized option is parsed. + * + * @param opt Unrecognized option from the command line. + * @return Whether to continue parsing the argument list. + */ + protected boolean handleUnknown(String opt) { + throw new UnsupportedOperationException(); + } + + /** + * Callback for remaining command line arguments after either {@link #handle(String, String)} or + * {@link #handleUnknown(String)} return "false". This will be called at the end of parsing even + * when there are no remaining arguments. + * + * @param extra List of remaining arguments. + */ + protected void handleExtraArgs(List extra) { + throw new UnsupportedOperationException(); + } + + private String findCliOption(String name, String[][] available) { + for (String[] candidates : available) { + for (String candidate : candidates) { + if (candidate.equals(name)) { + return candidates[0]; + } + } + } + return null; + } + +} diff --git a/launcher/src/main/java/org/apache/spark/launcher/package-info.java b/launcher/src/main/java/org/apache/spark/launcher/package-info.java new file mode 100644 index 0000000000000..7ed756f4b8591 --- /dev/null +++ b/launcher/src/main/java/org/apache/spark/launcher/package-info.java @@ -0,0 +1,45 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** + * Library for launching Spark applications. + *

    + * This library allows applications to launch Spark programmatically. There's only one entry + * point to the library - the {@link org.apache.spark.launcher.SparkLauncher} class. + *

    + * To launch a Spark application, just instantiate a {@link org.apache.spark.launcher.SparkLauncher} + * and configure the application to run. For example: + * + * 

    + * {@code
+ *   import org.apache.spark.launcher.SparkLauncher;
+ *
+ *   public class MyLauncher {
+ *     public static void main(String[] args) throws Exception {
+ *       Process spark = new SparkLauncher()
+ *         .setAppResource("/my/app.jar")
+ *         .setMainClass("my.spark.app.Main")
+ *         .setMaster("local")
+ *         .setConf(SparkLauncher.DRIVER_MEMORY, "2g")
+ *         .launch();
+ *       spark.waitFor();
+ *     }
+ *   }
+ * }
+ *
    + */ +package org.apache.spark.launcher; diff --git a/launcher/src/test/java/org/apache/spark/launcher/CommandBuilderUtilsSuite.java b/launcher/src/test/java/org/apache/spark/launcher/CommandBuilderUtilsSuite.java new file mode 100644 index 0000000000000..dba0203867372 --- /dev/null +++ b/launcher/src/test/java/org/apache/spark/launcher/CommandBuilderUtilsSuite.java @@ -0,0 +1,101 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.util.ArrayList; +import java.util.Arrays; +import java.util.List; + +import org.junit.Test; +import static org.junit.Assert.*; + +import static org.apache.spark.launcher.CommandBuilderUtils.*; + +public class CommandBuilderUtilsSuite { + + @Test + public void testValidOptionStrings() { + testOpt("a b c d e", Arrays.asList("a", "b", "c", "d", "e")); + testOpt("a 'b c' \"d\" e", Arrays.asList("a", "b c", "d", "e")); + testOpt("a 'b\\\"c' \"'d'\" e", Arrays.asList("a", "b\\\"c", "'d'", "e")); + testOpt("a 'b\"c' \"\\\"d\\\"\" e", Arrays.asList("a", "b\"c", "\"d\"", "e")); + testOpt(" a b c \\\\ ", Arrays.asList("a", "b", "c", "\\")); + + // Following tests ported from UtilsSuite.scala. + testOpt("", new ArrayList()); + testOpt("a", Arrays.asList("a")); + testOpt("aaa", Arrays.asList("aaa")); + testOpt("a b c", Arrays.asList("a", "b", "c")); + testOpt(" a b\t c ", Arrays.asList("a", "b", "c")); + testOpt("a 'b c'", Arrays.asList("a", "b c")); + testOpt("a 'b c' d", Arrays.asList("a", "b c", "d")); + testOpt("'b c'", Arrays.asList("b c")); + testOpt("a \"b c\"", Arrays.asList("a", "b c")); + testOpt("a \"b c\" d", Arrays.asList("a", "b c", "d")); + testOpt("\"b c\"", Arrays.asList("b c")); + testOpt("a 'b\" c' \"d' e\"", Arrays.asList("a", "b\" c", "d' e")); + testOpt("a\t'b\nc'\nd", Arrays.asList("a", "b\nc", "d")); + testOpt("a \"b\\\\c\"", Arrays.asList("a", "b\\c")); + testOpt("a \"b\\\"c\"", Arrays.asList("a", "b\"c")); + testOpt("a 'b\\\"c'", Arrays.asList("a", "b\\\"c")); + testOpt("'a'b", Arrays.asList("ab")); + testOpt("'a''b'", Arrays.asList("ab")); + testOpt("\"a\"b", Arrays.asList("ab")); + testOpt("\"a\"\"b\"", Arrays.asList("ab")); + testOpt("''", Arrays.asList("")); + testOpt("\"\"", Arrays.asList("")); + } + + @Test + public void testInvalidOptionStrings() { + testInvalidOpt("\\"); + testInvalidOpt("\"abcde"); + testInvalidOpt("'abcde"); + } + + @Test + public void testWindowsBatchQuoting() { + assertEquals("abc", quoteForBatchScript("abc")); + assertEquals("\"a b c\"", quoteForBatchScript("a b c")); + assertEquals("\"a \"\"b\"\" c\"", quoteForBatchScript("a \"b\" c")); + assertEquals("\"a\"\"b\"\"c\"", quoteForBatchScript("a\"b\"c")); + assertEquals("\"ab^=\"\"cd\"\"\"", quoteForBatchScript("ab=\"cd\"")); + } + + @Test + public void testPythonArgQuoting() { + assertEquals("\"abc\"", quoteForPython("abc")); + assertEquals("\"a b c\"", quoteForPython("a b c")); + assertEquals("\"a \\\"b\\\" c\"", quoteForPython("a \"b\" c")); + } + + private void testOpt(String opts, List expected) { + assertEquals(String.format("test string failed to parse: [[ %s ]]", opts), + expected, parseOptionString(opts)); + } + + private void testInvalidOpt(String opts) { + try { + parseOptionString(opts); + fail("Expected exception for invalid option string."); + } catch (IllegalArgumentException e) { + // pass. + } + } + +} diff --git a/launcher/src/test/java/org/apache/spark/launcher/SparkLauncherSuite.java b/launcher/src/test/java/org/apache/spark/launcher/SparkLauncherSuite.java new file mode 100644 index 0000000000000..252d5abae1ca3 --- /dev/null +++ b/launcher/src/test/java/org/apache/spark/launcher/SparkLauncherSuite.java @@ -0,0 +1,94 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.io.BufferedReader; +import java.io.InputStream; +import java.io.InputStreamReader; +import java.util.HashMap; +import java.util.Map; + +import org.junit.Test; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; +import static org.junit.Assert.*; + +/** + * These tests require the Spark assembly to be built before they can be run. + */ +public class SparkLauncherSuite { + + private static final Logger LOG = LoggerFactory.getLogger(SparkLauncherSuite.class); + + @Test + public void testChildProcLauncher() throws Exception { + Map env = new HashMap(); + env.put("SPARK_PRINT_LAUNCH_COMMAND", "1"); + + SparkLauncher launcher = new SparkLauncher(env) + .setSparkHome(System.getProperty("spark.test.home")) + .setMaster("local") + .setAppResource("spark-internal") + .setConf(SparkLauncher.DRIVER_EXTRA_JAVA_OPTIONS, + "-Dfoo=bar -Dtest.name=-testChildProcLauncher") + .setConf(SparkLauncher.DRIVER_EXTRA_CLASSPATH, System.getProperty("java.class.path")) + .setMainClass(SparkLauncherTestApp.class.getName()) + .addAppArgs("proc"); + final Process app = launcher.launch(); + new Redirector("stdout", app.getInputStream()).start(); + new Redirector("stderr", app.getErrorStream()).start(); + assertEquals(0, app.waitFor()); + } + + public static class SparkLauncherTestApp { + + public static void main(String[] args) throws Exception { + assertEquals(1, args.length); + assertEquals("proc", args[0]); + assertEquals("bar", System.getProperty("foo")); + assertEquals("local", System.getProperty(SparkLauncher.SPARK_MASTER)); + } + + } + + private static class Redirector extends Thread { + + private final InputStream in; + + Redirector(String name, InputStream in) { + this.in = in; + setName(name); + setDaemon(true); + } + + @Override + public void run() { + try { + BufferedReader reader = new BufferedReader(new InputStreamReader(in, "UTF-8")); + String line; + while ((line = reader.readLine()) != null) { + LOG.warn(line); + } + } catch (Exception e) { + LOG.error("Error reading process output.", e); + } + } + + } + +} diff --git a/launcher/src/test/java/org/apache/spark/launcher/SparkSubmitCommandBuilderSuite.java b/launcher/src/test/java/org/apache/spark/launcher/SparkSubmitCommandBuilderSuite.java new file mode 100644 index 0000000000000..815edc4e4971f --- /dev/null +++ b/launcher/src/test/java/org/apache/spark/launcher/SparkSubmitCommandBuilderSuite.java @@ -0,0 +1,278 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.io.File; +import java.util.Arrays; +import java.util.Collections; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.regex.Pattern; + +import org.junit.AfterClass; +import org.junit.BeforeClass; +import org.junit.Test; +import static org.junit.Assert.*; + +public class SparkSubmitCommandBuilderSuite { + + private static File dummyPropsFile; + private static SparkSubmitOptionParser parser; + + @BeforeClass + public static void setUp() throws Exception { + dummyPropsFile = File.createTempFile("spark", "properties"); + parser = new SparkSubmitOptionParser(); + } + + @AfterClass + public static void cleanUp() throws Exception { + dummyPropsFile.delete(); + } + + @Test + public void testDriverCmdBuilder() throws Exception { + testCmdBuilder(true); + } + + @Test + public void testClusterCmdBuilder() throws Exception { + testCmdBuilder(false); + } + + @Test + public void testCliParser() throws Exception { + List sparkSubmitArgs = Arrays.asList( + parser.MASTER, + "local", + parser.DRIVER_MEMORY, + "42g", + parser.DRIVER_CLASS_PATH, + "/driverCp", + parser.DRIVER_JAVA_OPTIONS, + "extraJavaOpt", + parser.CONF, + SparkLauncher.DRIVER_EXTRA_LIBRARY_PATH + "=/driverLibPath"); + Map env = new HashMap(); + List cmd = buildCommand(sparkSubmitArgs, env); + + assertTrue(findInStringList(env.get(CommandBuilderUtils.getLibPathEnvName()), + File.pathSeparator, "/driverLibPath")); + assertTrue(findInStringList(findArgValue(cmd, "-cp"), File.pathSeparator, "/driverCp")); + assertTrue("Driver -Xms should be configured.", cmd.contains("-Xms42g")); + assertTrue("Driver -Xmx should be configured.", cmd.contains("-Xmx42g")); + } + + @Test + public void testShellCliParser() throws Exception { + List sparkSubmitArgs = Arrays.asList( + parser.CLASS, + "org.apache.spark.repl.Main", + parser.MASTER, + "foo", + "--app-arg", + "bar", + "--app-switch", + parser.FILES, + "baz", + parser.NAME, + "appName"); + + List args = new SparkSubmitCommandBuilder(sparkSubmitArgs).buildSparkSubmitArgs(); + List expected = Arrays.asList("spark-shell", "--app-arg", "bar", "--app-switch"); + assertEquals(expected, args.subList(args.size() - expected.size(), args.size())); + } + + @Test + public void testAlternateSyntaxParsing() throws Exception { + List sparkSubmitArgs = Arrays.asList( + parser.CLASS + "=org.my.Class", + parser.MASTER + "=foo", + parser.DEPLOY_MODE + "=bar"); + + List cmd = new SparkSubmitCommandBuilder(sparkSubmitArgs).buildSparkSubmitArgs(); + assertEquals("org.my.Class", findArgValue(cmd, parser.CLASS)); + assertEquals("foo", findArgValue(cmd, parser.MASTER)); + assertEquals("bar", findArgValue(cmd, parser.DEPLOY_MODE)); + } + + @Test + public void testPySparkLauncher() throws Exception { + List sparkSubmitArgs = Arrays.asList( + SparkSubmitCommandBuilder.PYSPARK_SHELL, + "--master=foo", + "--deploy-mode=bar"); + + Map env = new HashMap(); + List cmd = buildCommand(sparkSubmitArgs, env); + assertEquals("python", cmd.get(cmd.size() - 1)); + assertEquals( + String.format("\"%s\" \"foo\" \"%s\" \"bar\" \"%s\"", + parser.MASTER, parser.DEPLOY_MODE, SparkSubmitCommandBuilder.PYSPARK_SHELL_RESOURCE), + env.get("PYSPARK_SUBMIT_ARGS")); + } + + @Test + public void testPySparkFallback() throws Exception { + List sparkSubmitArgs = Arrays.asList( + "--master=foo", + "--deploy-mode=bar", + "script.py", + "arg1"); + + Map env = new HashMap(); + List cmd = buildCommand(sparkSubmitArgs, env); + + assertEquals("foo", findArgValue(cmd, "--master")); + assertEquals("bar", findArgValue(cmd, "--deploy-mode")); + assertEquals("script.py", cmd.get(cmd.size() - 2)); + assertEquals("arg1", cmd.get(cmd.size() - 1)); + } + + private void testCmdBuilder(boolean isDriver) throws Exception { + String deployMode = isDriver ? "client" : "cluster"; + + SparkSubmitCommandBuilder launcher = + new SparkSubmitCommandBuilder(Collections.emptyList()); + launcher.childEnv.put(CommandBuilderUtils.ENV_SPARK_HOME, + System.getProperty("spark.test.home")); + launcher.master = "yarn"; + launcher.deployMode = deployMode; + launcher.appResource = "/foo"; + launcher.appName = "MyApp"; + launcher.mainClass = "my.Class"; + launcher.propertiesFile = dummyPropsFile.getAbsolutePath(); + launcher.appArgs.add("foo"); + launcher.appArgs.add("bar"); + launcher.conf.put(SparkLauncher.DRIVER_MEMORY, "1g"); + launcher.conf.put(SparkLauncher.DRIVER_EXTRA_CLASSPATH, "/driver"); + launcher.conf.put(SparkLauncher.DRIVER_EXTRA_JAVA_OPTIONS, "-Ddriver -XX:MaxPermSize=256m"); + launcher.conf.put(SparkLauncher.DRIVER_EXTRA_LIBRARY_PATH, "/native"); + launcher.conf.put("spark.foo", "foo"); + + Map env = new HashMap(); + List cmd = launcher.buildCommand(env); + + // Checks below are different for driver and non-driver mode. + + if (isDriver) { + assertTrue("Driver -Xms should be configured.", cmd.contains("-Xms1g")); + assertTrue("Driver -Xmx should be configured.", cmd.contains("-Xmx1g")); + } else { + boolean found = false; + for (String arg : cmd) { + if (arg.startsWith("-Xms") || arg.startsWith("-Xmx")) { + found = true; + break; + } + } + assertFalse("Memory arguments should not be set.", found); + } + + for (String arg : cmd) { + if (arg.startsWith("-XX:MaxPermSize=")) { + if (isDriver) { + assertEquals("-XX:MaxPermSize=256m", arg); + } else { + assertEquals("-XX:MaxPermSize=128m", arg); + } + } + } + + String[] cp = findArgValue(cmd, "-cp").split(Pattern.quote(File.pathSeparator)); + if (isDriver) { + assertTrue("Driver classpath should contain provided entry.", contains("/driver", cp)); + } else { + assertFalse("Driver classpath should not be in command.", contains("/driver", cp)); + } + + String libPath = env.get(CommandBuilderUtils.getLibPathEnvName()); + if (isDriver) { + assertNotNull("Native library path should be set.", libPath); + assertTrue("Native library path should contain provided entry.", + contains("/native", libPath.split(Pattern.quote(File.pathSeparator)))); + } else { + assertNull("Native library should not be set.", libPath); + } + + // Checks below are the same for both driver and non-driver mode. + assertEquals(dummyPropsFile.getAbsolutePath(), findArgValue(cmd, parser.PROPERTIES_FILE)); + assertEquals("yarn", findArgValue(cmd, parser.MASTER)); + assertEquals(deployMode, findArgValue(cmd, parser.DEPLOY_MODE)); + assertEquals("my.Class", findArgValue(cmd, parser.CLASS)); + assertEquals("MyApp", findArgValue(cmd, parser.NAME)); + + boolean appArgsOk = false; + for (int i = 0; i < cmd.size(); i++) { + if (cmd.get(i).equals("/foo")) { + assertEquals("foo", cmd.get(i + 1)); + assertEquals("bar", cmd.get(i + 2)); + assertEquals(cmd.size(), i + 3); + appArgsOk = true; + break; + } + } + assertTrue("App resource and args should be added to command.", appArgsOk); + + Map conf = parseConf(cmd, parser); + assertEquals("foo", conf.get("spark.foo")); + } + + private boolean contains(String needle, String[] haystack) { + for (String entry : haystack) { + if (entry.equals(needle)) { + return true; + } + } + return false; + } + + private Map parseConf(List cmd, SparkSubmitOptionParser parser) { + Map conf = new HashMap(); + for (int i = 0; i < cmd.size(); i++) { + if (cmd.get(i).equals(parser.CONF)) { + String[] val = cmd.get(i + 1).split("=", 2); + conf.put(val[0], val[1]); + i += 1; + } + } + return conf; + } + + private String findArgValue(List cmd, String name) { + for (int i = 0; i < cmd.size(); i++) { + if (cmd.get(i).equals(name)) { + return cmd.get(i + 1); + } + } + fail(String.format("arg '%s' not found", name)); + return null; + } + + private boolean findInStringList(String list, String sep, String needle) { + return contains(needle, list.split(sep)); + } + + private List buildCommand(List args, Map env) throws Exception { + SparkSubmitCommandBuilder builder = new SparkSubmitCommandBuilder(args); + builder.childEnv.put(CommandBuilderUtils.ENV_SPARK_HOME, System.getProperty("spark.test.home")); + return builder.buildCommand(env); + } + +} diff --git a/launcher/src/test/java/org/apache/spark/launcher/SparkSubmitOptionParserSuite.java b/launcher/src/test/java/org/apache/spark/launcher/SparkSubmitOptionParserSuite.java new file mode 100644 index 0000000000000..f3d2109917056 --- /dev/null +++ b/launcher/src/test/java/org/apache/spark/launcher/SparkSubmitOptionParserSuite.java @@ -0,0 +1,108 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.util.Arrays; +import java.util.Collections; +import java.util.List; + +import org.junit.Before; +import org.junit.Test; +import static org.junit.Assert.*; +import static org.mockito.Mockito.*; + +import static org.apache.spark.launcher.SparkSubmitOptionParser.*; + +public class SparkSubmitOptionParserSuite { + + private SparkSubmitOptionParser parser; + + @Before + public void setUp() { + parser = spy(new DummyParser()); + } + + @Test + public void testAllOptions() { + int count = 0; + for (String[] optNames : parser.opts) { + for (String optName : optNames) { + String value = optName + "-value"; + parser.parse(Arrays.asList(optName, value)); + count++; + verify(parser).handle(eq(optNames[0]), eq(value)); + verify(parser, times(count)).handle(anyString(), anyString()); + verify(parser, times(count)).handleExtraArgs(eq(Collections.emptyList())); + } + } + + for (String[] switchNames : parser.switches) { + int switchCount = 0; + for (String name : switchNames) { + parser.parse(Arrays.asList(name)); + count++; + switchCount++; + verify(parser, times(switchCount)).handle(eq(switchNames[0]), same((String) null)); + verify(parser, times(count)).handle(anyString(), any(String.class)); + verify(parser, times(count)).handleExtraArgs(eq(Collections.emptyList())); + } + } + } + + @Test + public void testExtraOptions() { + List args = Arrays.asList(parser.MASTER, parser.MASTER, "foo", "bar"); + parser.parse(args); + verify(parser).handle(eq(parser.MASTER), eq(parser.MASTER)); + verify(parser).handleUnknown(eq("foo")); + verify(parser).handleExtraArgs(eq(Arrays.asList("bar"))); + } + + @Test(expected=IllegalArgumentException.class) + public void testMissingArg() { + parser.parse(Arrays.asList(parser.MASTER)); + } + + @Test + public void testEqualSeparatedOption() { + List args = Arrays.asList(parser.MASTER + "=" + parser.MASTER); + parser.parse(args); + verify(parser).handle(eq(parser.MASTER), eq(parser.MASTER)); + verify(parser).handleExtraArgs(eq(Collections.emptyList())); + } + + private static class DummyParser extends SparkSubmitOptionParser { + + @Override + protected boolean handle(String opt, String value) { + return true; + } + + @Override + protected boolean handleUnknown(String opt) { + return false; + } + + @Override + protected void handleExtraArgs(List extra) { + + } + + } + +} diff --git a/launcher/src/test/resources/log4j.properties b/launcher/src/test/resources/log4j.properties new file mode 100644 index 0000000000000..00c20ad69cd4d --- /dev/null +++ b/launcher/src/test/resources/log4j.properties @@ -0,0 +1,31 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +# Set everything to be logged to the file core/target/unit-tests.log +log4j.rootCategory=INFO, file +log4j.appender.file=org.apache.log4j.FileAppender +log4j.appender.file.append=false + +# Some tests will set "test.name" to avoid overwriting the main log file. +log4j.appender.file.file=target/unit-tests${test.name}.log + +log4j.appender.file.layout=org.apache.log4j.PatternLayout +log4j.appender.file.layout.ConversionPattern=%d{yy/MM/dd HH:mm:ss.SSS} %t %p %c{1}: %m%n + +# Ignore messages below warning level from Jetty, because it's a bit verbose +log4j.logger.org.eclipse.jetty=WARN +org.eclipse.jetty.LEVEL=WARN diff --git a/make-distribution.sh b/make-distribution.sh index dd990d4b96e46..82d33408cd5e5 100755 --- a/make-distribution.sh +++ b/make-distribution.sh @@ -127,6 +127,7 @@ if [ ! $(command -v "$MVN") ] ; then fi VERSION=$("$MVN" help:evaluate -Dexpression=project.version 2>/dev/null | grep -v "INFO" | tail -n 1) +SCALA_VERSION=$("$MVN" help:evaluate -Dexpression=scala.binary.version 2>/dev/null | grep -v "INFO" | tail -n 1) SPARK_HADOOP_VERSION=$("$MVN" help:evaluate -Dexpression=hadoop.version $@ 2>/dev/null\ | grep -v "INFO"\ | tail -n 1) @@ -196,6 +197,7 @@ echo "Build flags: $@" >> "$DISTDIR/RELEASE" # Copy jars cp "$SPARK_HOME"/assembly/target/scala*/*assembly*hadoop*.jar "$DISTDIR/lib/" cp "$SPARK_HOME"/examples/target/scala*/spark-examples*.jar "$DISTDIR/lib/" +cp "$SPARK_HOME"/launcher/target/spark-launcher_$SCALA_VERSION-$VERSION.jar "$DISTDIR/lib/" # This will fail if the -Pyarn profile is not provided # In this case, silence the error and ignore the return code of this command cp "$SPARK_HOME"/network/yarn/target/scala*/spark-*-yarn-shuffle.jar "$DISTDIR/lib/" &> /dev/null || : diff --git a/pom.xml b/pom.xml index 51bef30f9ca8f..a19da73cf45b3 100644 --- a/pom.xml +++ b/pom.xml @@ -105,6 +105,7 @@ external/zeromq examples repl + launcher @@ -1195,7 +1196,7 @@ true - ${session.executionRootDirectory} + ${spark.test.home} 1 false false diff --git a/project/SparkBuild.scala b/project/SparkBuild.scala index 4f17df59f4c1f..35e748f26bbaa 100644 --- a/project/SparkBuild.scala +++ b/project/SparkBuild.scala @@ -34,11 +34,11 @@ object BuildCommons { val allProjects@Seq(bagel, catalyst, core, graphx, hive, hiveThriftServer, mllib, repl, sql, networkCommon, networkShuffle, streaming, streamingFlumeSink, streamingFlume, streamingKafka, - streamingMqtt, streamingTwitter, streamingZeromq) = + streamingMqtt, streamingTwitter, streamingZeromq, launcher) = Seq("bagel", "catalyst", "core", "graphx", "hive", "hive-thriftserver", "mllib", "repl", "sql", "network-common", "network-shuffle", "streaming", "streaming-flume-sink", "streaming-flume", "streaming-kafka", "streaming-mqtt", "streaming-twitter", - "streaming-zeromq").map(ProjectRef(buildLocation, _)) + "streaming-zeromq", "launcher").map(ProjectRef(buildLocation, _)) val optionallyEnabledProjects@Seq(yarn, yarnStable, java8Tests, sparkGangliaLgpl, sparkKinesisAsl) = Seq("yarn", "yarn-stable", "java8-tests", "ganglia-lgpl", @@ -155,8 +155,9 @@ object SparkBuild extends PomBuild { (allProjects ++ optionallyEnabledProjects).foreach(enable(TestSettings.settings)) // TODO: Add Sql to mima checks + // TODO: remove launcher from this list after 1.3. allProjects.filterNot(x => Seq(spark, sql, hive, hiveThriftServer, catalyst, repl, - networkCommon, networkShuffle, networkYarn).contains(x)).foreach { + networkCommon, networkShuffle, networkYarn, launcher).contains(x)).foreach { x => enable(MimaBuild.mimaSettings(sparkHome, x))(x) } diff --git a/python/pyspark/java_gateway.py b/python/pyspark/java_gateway.py index 936857e75c7e9..43d2cf5171880 100644 --- a/python/pyspark/java_gateway.py +++ b/python/pyspark/java_gateway.py @@ -41,7 +41,7 @@ def launch_gateway(): submit_args = os.environ.get("PYSPARK_SUBMIT_ARGS") submit_args = submit_args if submit_args is not None else "" submit_args = shlex.split(submit_args) - command = [os.path.join(SPARK_HOME, script)] + submit_args + ["pyspark-shell"] + command = [os.path.join(SPARK_HOME, script)] + submit_args # Start a socket that will be used by PythonGatewayServer to communicate its port to us callback_socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM) @@ -58,7 +58,6 @@ def launch_gateway(): # Don't send ctrl-c / SIGINT to the Java gateway: def preexec_func(): signal.signal(signal.SIGINT, signal.SIG_IGN) - env["IS_SUBPROCESS"] = "1" # tell JVM to exit after python exits proc = Popen(command, stdin=PIPE, preexec_fn=preexec_func, env=env) else: # preexec_fn not supported on Windows diff --git a/sbin/spark-daemon.sh b/sbin/spark-daemon.sh index 5e812a1d91c6b..92e76a3fe6ca2 100755 --- a/sbin/spark-daemon.sh +++ b/sbin/spark-daemon.sh @@ -121,45 +121,63 @@ if [ "$SPARK_NICENESS" = "" ]; then export SPARK_NICENESS=0 fi +run_command() { + mode="$1" + shift -case $option in + mkdir -p "$SPARK_PID_DIR" - (start|spark-submit) + if [ -f "$pid" ]; then + TARGET_ID="$(cat "$pid")" + if [[ $(ps -p "$TARGET_ID" -o args=) =~ $command ]]; then + echo "$command running as process $TARGET_ID. Stop it first." + exit 1 + fi + fi - mkdir -p "$SPARK_PID_DIR" + if [ "$SPARK_MASTER" != "" ]; then + echo rsync from "$SPARK_MASTER" + rsync -a -e ssh --delete --exclude=.svn --exclude='logs/*' --exclude='contrib/hod/logs/*' "$SPARK_MASTER/" "$SPARK_HOME" + fi - if [ -f $pid ]; then - TARGET_ID="$(cat "$pid")" - if [[ $(ps -p "$TARGET_ID" -o args=) =~ $command ]]; then - echo "$command running as process $TARGET_ID. Stop it first." - exit 1 - fi - fi + spark_rotate_log "$log" + echo "starting $command, logging to $log" + + case "$mode" in + (class) + nohup nice -n "$SPARK_NICENESS" "$SPARK_PREFIX"/bin/spark-class $command "$@" >> "$log" 2>&1 < /dev/null & + newpid="$!" + ;; + + (submit) + nohup nice -n "$SPARK_NICENESS" "$SPARK_PREFIX"/bin/spark-submit --class $command "$@" >> "$log" 2>&1 < /dev/null & + newpid="$!" + ;; + + (*) + echo "unknown mode: $mode" + exit 1 + ;; + esac + + echo "$newpid" > "$pid" + sleep 2 + # Check if the process has died; in that case we'll tail the log so the user can see + if [[ ! $(ps -p "$newpid" -o args=) =~ $command ]]; then + echo "failed to launch $command:" + tail -2 "$log" | sed 's/^/ /' + echo "full log in $log" + fi +} - if [ "$SPARK_MASTER" != "" ]; then - echo rsync from "$SPARK_MASTER" - rsync -a -e ssh --delete --exclude=.svn --exclude='logs/*' --exclude='contrib/hod/logs/*' $SPARK_MASTER/ "$SPARK_HOME" - fi +case $option in - spark_rotate_log "$log" - echo "starting $command, logging to $log" - if [ $option == spark-submit ]; then - source "$SPARK_HOME"/bin/utils.sh - gatherSparkSubmitOpts "$@" - nohup nice -n $SPARK_NICENESS "$SPARK_PREFIX"/bin/spark-submit --class $command \ - "${SUBMISSION_OPTS[@]}" spark-internal "${APPLICATION_OPTS[@]}" >> "$log" 2>&1 < /dev/null & - else - nohup nice -n $SPARK_NICENESS "$SPARK_PREFIX"/bin/spark-class $command "$@" >> "$log" 2>&1 < /dev/null & - fi - newpid=$! - echo $newpid > $pid - sleep 2 - # Check if the process has died; in that case we'll tail the log so the user can see - if [[ ! $(ps -p "$newpid" -o args=) =~ $command ]]; then - echo "failed to launch $command:" - tail -2 "$log" | sed 's/^/ /' - echo "full log in $log" - fi + (submit) + run_command submit "$@" + ;; + + (start) + run_command class "$@" ;; (stop) diff --git a/sbin/start-thriftserver.sh b/sbin/start-thriftserver.sh index 070cc7a87e6f2..5b0aeb177fff3 100755 --- a/sbin/start-thriftserver.sh +++ b/sbin/start-thriftserver.sh @@ -52,4 +52,4 @@ fi export SUBMIT_USAGE_FUNCTION=usage -exec "$FWDIR"/sbin/spark-daemon.sh spark-submit $CLASS 1 "$@" +exec "$FWDIR"/sbin/spark-daemon.sh submit $CLASS 1 "$@" From 35b25640a4debddd5a4498455888f6241caf6223 Mon Sep 17 00:00:00 2001 From: Sean Owen Date: Wed, 11 Mar 2015 12:16:32 +0000 Subject: [PATCH 027/122] [MINOR] [DOCS] Fix map -> mapToPair in Streaming Java example Fix map -> mapToPair in Java example. (And zap some unneeded "throws Exception" while here) Author: Sean Owen Closes #4967 from srowen/MapToPairFix and squashes the following commits: ded2bc0 [Sean Owen] Fix map -> mapToPair in Java example. (And zap some unneeded "throws Exception" while here) --- docs/streaming-programming-guide.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/docs/streaming-programming-guide.md b/docs/streaming-programming-guide.md index 815c98713b738..062ac2648db30 100644 --- a/docs/streaming-programming-guide.md +++ b/docs/streaming-programming-guide.md @@ -189,15 +189,15 @@ Next, we want to count these words. {% highlight java %} // Count each word in each batch -JavaPairDStream pairs = words.map( +JavaPairDStream pairs = words.mapToPair( new PairFunction() { - @Override public Tuple2 call(String s) throws Exception { + @Override public Tuple2 call(String s) { return new Tuple2(s, 1); } }); JavaPairDStream wordCounts = pairs.reduceByKey( new Function2() { - @Override public Integer call(Integer i1, Integer i2) throws Exception { + @Override public Integer call(Integer i1, Integer i2) { return i1 + i2; } }); @@ -1041,7 +1041,7 @@ val windowedWordCounts = pairs.reduceByKeyAndWindow((a:Int,b:Int) => (a + b), Se {% highlight java %} // Reduce function adding two integers, defined separately for clarity Function2 reduceFunc = new Function2() { - @Override public Integer call(Integer i1, Integer i2) throws Exception { + @Override public Integer call(Integer i1, Integer i2) { return i1 + i2; } }; From 40f49795e2624326dd4e38eedbf63d35860ea00e Mon Sep 17 00:00:00 2001 From: Hongbo Liu Date: Wed, 11 Mar 2015 12:18:24 +0000 Subject: [PATCH 028/122] [SQL][Minor] fix typo in comments Removed an repeated "from" in the comments. Author: Hongbo Liu Closes #4976 from liuhb86/mine and squashes the following commits: e280e7c [Hongbo Liu] [SQL][Minor] fix typo in comments --- .../scala/org/apache/spark/sql/catalyst/trees/TreeNode.scala | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/trees/TreeNode.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/trees/TreeNode.scala index 109671bdca361..7e191ad0315a5 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/trees/TreeNode.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/trees/TreeNode.scala @@ -63,7 +63,7 @@ abstract class TreeNode[BaseType <: TreeNode[BaseType]] { /** * Faster version of equality which short-circuits when two treeNodes are the same instance. - * We don't just override Object.Equals, as doing so prevents the scala compiler from from + * We don't just override Object.equals, as doing so prevents the scala compiler from * generating case class `equals` methods */ def fastEquals(other: TreeNode[_]): Boolean = { From ec30c17822329e6d2b8c85625b31ba8bd8679fcf Mon Sep 17 00:00:00 2001 From: zzcclp Date: Wed, 11 Mar 2015 12:22:24 +0000 Subject: [PATCH 029/122] [SPARK-6279][Streaming]In KafkaRDD.scala, Miss expressions flag "s" at logging string In KafkaRDD.scala, Miss expressions flag "s" at logging string In logging file, it print `Beginning offset $ {part.fromOffset} is the same as ending offset ` but not `Beginning offset 111 is the same as ending offset `. Author: zzcclp Closes #4979 from zzcclp/SPARK-6279 and squashes the following commits: 768f88e [zzcclp] Miss expressions flag "s" --- .../main/scala/org/apache/spark/streaming/kafka/KafkaRDD.scala | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/external/kafka/src/main/scala/org/apache/spark/streaming/kafka/KafkaRDD.scala b/external/kafka/src/main/scala/org/apache/spark/streaming/kafka/KafkaRDD.scala index d56cc01be9514..6d465bcb6bfc0 100644 --- a/external/kafka/src/main/scala/org/apache/spark/streaming/kafka/KafkaRDD.scala +++ b/external/kafka/src/main/scala/org/apache/spark/streaming/kafka/KafkaRDD.scala @@ -86,7 +86,7 @@ class KafkaRDD[ val part = thePart.asInstanceOf[KafkaRDDPartition] assert(part.fromOffset <= part.untilOffset, errBeginAfterEnd(part)) if (part.fromOffset == part.untilOffset) { - log.warn("Beginning offset ${part.fromOffset} is the same as ending offset " + + log.warn(s"Beginning offset ${part.fromOffset} is the same as ending offset " + s"skipping ${part.topic} ${part.partition}") Iterator.empty } else { From 6e94c4eadf443ac3d34eaae4c334c8386fdec960 Mon Sep 17 00:00:00 2001 From: Sean Owen Date: Wed, 11 Mar 2015 13:15:19 +0000 Subject: [PATCH 030/122] SPARK-6225 [CORE] [SQL] [STREAMING] Resolve most build warnings, 1.3.0 edition Resolve javac, scalac warnings of various types -- deprecations, Scala lang, unchecked cast, etc. Author: Sean Owen Closes #4950 from srowen/SPARK-6225 and squashes the following commits: 3080972 [Sean Owen] Ordered imports: Java, Scala, 3rd party, Spark c67985b [Sean Owen] Resolve javac, scalac warnings of various types -- deprecations, Scala lang, unchecked cast, etc. --- .../scala/org/apache/spark/SparkContext.scala | 2 +- .../scheduler/EventLoggingListener.scala | 2 +- .../spark/util/MutableURLClassLoader.scala | 2 -- .../OutputCommitCoordinatorSuite.scala | 2 +- .../util/MutableURLClassLoaderSuite.scala | 3 +-- .../JavaStatefulNetworkWordCount.java | 1 + .../org/apache/spark/examples/HBaseTest.scala | 4 ++-- .../kafka/JavaDirectKafkaStreamSuite.java | 21 +++++++------------ .../streaming/kafka/JavaKafkaRDDSuite.java | 21 +++++++------------ .../streaming/kafka/JavaKafkaStreamSuite.java | 14 ++++++------- .../MatrixFactorizationModel.scala | 8 +++---- .../org/apache/spark/sql/sources/ddl.scala | 1 + .../sql/ScalaReflectionRelationSuite.scala | 2 +- .../spark/sql/hive/HiveInspectorSuite.scala | 2 +- .../apache/spark/streaming/JavaAPISuite.java | 4 ++++ 15 files changed, 40 insertions(+), 49 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/SparkContext.scala b/core/src/main/scala/org/apache/spark/SparkContext.scala index 1a0bee4e3aea9..8121aab3b0b34 100644 --- a/core/src/main/scala/org/apache/spark/SparkContext.scala +++ b/core/src/main/scala/org/apache/spark/SparkContext.scala @@ -1104,7 +1104,7 @@ class SparkContext(config: SparkConf) extends Logging with ExecutorAllocationCli if (!fs.exists(hadoopPath)) { throw new FileNotFoundException(s"Added file $hadoopPath does not exist.") } - val isDir = fs.isDirectory(hadoopPath) + val isDir = fs.getFileStatus(hadoopPath).isDir if (!isLocal && scheme == "file" && isDir) { throw new SparkException(s"addFile does not support local directories when not running " + "local mode.") diff --git a/core/src/main/scala/org/apache/spark/scheduler/EventLoggingListener.scala b/core/src/main/scala/org/apache/spark/scheduler/EventLoggingListener.scala index 2091a9fe8d0d3..34fa6d27c3a45 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/EventLoggingListener.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/EventLoggingListener.scala @@ -95,7 +95,7 @@ private[spark] class EventLoggingListener( * Creates the log file in the configured log directory. */ def start() { - if (!fileSystem.isDirectory(new Path(logBaseDir))) { + if (!fileSystem.getFileStatus(new Path(logBaseDir)).isDir) { throw new IllegalArgumentException(s"Log directory $logBaseDir does not exist.") } diff --git a/core/src/main/scala/org/apache/spark/util/MutableURLClassLoader.scala b/core/src/main/scala/org/apache/spark/util/MutableURLClassLoader.scala index d9c7103b2f3bf..1e0ba5c28754a 100644 --- a/core/src/main/scala/org/apache/spark/util/MutableURLClassLoader.scala +++ b/core/src/main/scala/org/apache/spark/util/MutableURLClassLoader.scala @@ -23,8 +23,6 @@ import java.util.concurrent.ConcurrentHashMap import scala.collection.JavaConversions._ -import org.apache.spark.util.ParentClassLoader - /** * URL class loader that exposes the `addURL` and `getURLs` methods in URLClassLoader. */ diff --git a/core/src/test/scala/org/apache/spark/scheduler/OutputCommitCoordinatorSuite.scala b/core/src/test/scala/org/apache/spark/scheduler/OutputCommitCoordinatorSuite.scala index 3cc860caa1d9b..c8c957856247a 100644 --- a/core/src/test/scala/org/apache/spark/scheduler/OutputCommitCoordinatorSuite.scala +++ b/core/src/test/scala/org/apache/spark/scheduler/OutputCommitCoordinatorSuite.scala @@ -153,7 +153,7 @@ class OutputCommitCoordinatorSuite extends FunSuite with BeforeAndAfter { def resultHandler(x: Int, y: Unit): Unit = {} val futureAction: SimpleFutureAction[Unit] = sc.submitJob[Int, Unit, Unit](rdd, OutputCommitFunctions(tempDir.getAbsolutePath).commitSuccessfully, - 0 until rdd.partitions.size, resultHandler, 0) + 0 until rdd.partitions.size, resultHandler, () => Unit) // It's an error if the job completes successfully even though no committer was authorized, // so throw an exception if the job was allowed to complete. intercept[TimeoutException] { diff --git a/core/src/test/scala/org/apache/spark/util/MutableURLClassLoaderSuite.scala b/core/src/test/scala/org/apache/spark/util/MutableURLClassLoaderSuite.scala index 31e3b7e7bb71b..87de90bb0dfb0 100644 --- a/core/src/test/scala/org/apache/spark/util/MutableURLClassLoaderSuite.scala +++ b/core/src/test/scala/org/apache/spark/util/MutableURLClassLoaderSuite.scala @@ -21,8 +21,7 @@ import java.net.URLClassLoader import org.scalatest.FunSuite -import org.apache.spark.{LocalSparkContext, SparkContext, SparkException, TestUtils} -import org.apache.spark.util.Utils +import org.apache.spark.{SparkContext, SparkException, TestUtils} class MutableURLClassLoaderSuite extends FunSuite { diff --git a/examples/src/main/java/org/apache/spark/examples/streaming/JavaStatefulNetworkWordCount.java b/examples/src/main/java/org/apache/spark/examples/streaming/JavaStatefulNetworkWordCount.java index d46c7107c7a21..dbf2ef02d7b76 100644 --- a/examples/src/main/java/org/apache/spark/examples/streaming/JavaStatefulNetworkWordCount.java +++ b/examples/src/main/java/org/apache/spark/examples/streaming/JavaStatefulNetworkWordCount.java @@ -82,6 +82,7 @@ public Optional call(List values, Optional state) { ssc.checkpoint("."); // Initial RDD input to updateStateByKey + @SuppressWarnings("unchecked") List> tuples = Arrays.asList(new Tuple2("hello", 1), new Tuple2("world", 1)); JavaPairRDD initialRDD = ssc.sc().parallelizePairs(tuples); diff --git a/examples/src/main/scala/org/apache/spark/examples/HBaseTest.scala b/examples/src/main/scala/org/apache/spark/examples/HBaseTest.scala index 822673347bdce..f4684b42b5d41 100644 --- a/examples/src/main/scala/org/apache/spark/examples/HBaseTest.scala +++ b/examples/src/main/scala/org/apache/spark/examples/HBaseTest.scala @@ -18,7 +18,7 @@ package org.apache.spark.examples import org.apache.hadoop.hbase.client.HBaseAdmin -import org.apache.hadoop.hbase.{HBaseConfiguration, HTableDescriptor} +import org.apache.hadoop.hbase.{HBaseConfiguration, HTableDescriptor, TableName} import org.apache.hadoop.hbase.mapreduce.TableInputFormat import org.apache.spark._ @@ -36,7 +36,7 @@ object HBaseTest { // Initialize hBase table if necessary val admin = new HBaseAdmin(conf) if (!admin.isTableAvailable(args(0))) { - val tableDesc = new HTableDescriptor(args(0)) + val tableDesc = new HTableDescriptor(TableName.valueOf(args(0))) admin.createTable(tableDesc) } diff --git a/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaDirectKafkaStreamSuite.java b/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaDirectKafkaStreamSuite.java index 1334cc8fd1b57..d6ca6d58b5665 100644 --- a/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaDirectKafkaStreamSuite.java +++ b/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaDirectKafkaStreamSuite.java @@ -20,32 +20,27 @@ import java.io.Serializable; import java.util.HashMap; import java.util.HashSet; -import java.util.Random; import java.util.Arrays; -import org.apache.spark.SparkConf; - import scala.Tuple2; -import junit.framework.Assert; - import kafka.common.TopicAndPartition; import kafka.message.MessageAndMetadata; import kafka.serializer.StringDecoder; +import org.junit.After; +import org.junit.Assert; +import org.junit.Before; +import org.junit.Test; +import org.apache.spark.SparkConf; import org.apache.spark.api.java.JavaRDD; import org.apache.spark.api.java.function.Function; -import org.apache.spark.streaming.api.java.JavaDStream; import org.apache.spark.streaming.Durations; +import org.apache.spark.streaming.api.java.JavaDStream; import org.apache.spark.streaming.api.java.JavaStreamingContext; -import org.junit.Test; -import org.junit.After; -import org.junit.Before; - public class JavaDirectKafkaStreamSuite implements Serializable { private transient JavaStreamingContext ssc = null; - private transient Random random = new Random(); private transient KafkaStreamSuiteBase suiteBase = null; @Before @@ -93,7 +88,7 @@ public void testKafkaStream() throws InterruptedException { ).map( new Function, String>() { @Override - public String call(scala.Tuple2 kv) throws Exception { + public String call(Tuple2 kv) throws Exception { return kv._2(); } } @@ -121,7 +116,7 @@ public String call(MessageAndMetadata msgAndMd) throws Exception unifiedStream.foreachRDD( new Function, Void>() { @Override - public Void call(org.apache.spark.api.java.JavaRDD rdd) throws Exception { + public Void call(JavaRDD rdd) throws Exception { result.addAll(rdd.collect()); return null; } diff --git a/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaKafkaRDDSuite.java b/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaKafkaRDDSuite.java index 9d2e1705c6c73..4477b81827c70 100644 --- a/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaKafkaRDDSuite.java +++ b/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaKafkaRDDSuite.java @@ -19,27 +19,22 @@ import java.io.Serializable; import java.util.HashMap; -import java.util.HashSet; -import java.util.Arrays; - -import org.apache.spark.SparkConf; import scala.Tuple2; -import junit.framework.Assert; - import kafka.common.TopicAndPartition; import kafka.message.MessageAndMetadata; import kafka.serializer.StringDecoder; +import org.junit.After; +import org.junit.Assert; +import org.junit.Before; +import org.junit.Test; +import org.apache.spark.SparkConf; import org.apache.spark.api.java.JavaRDD; import org.apache.spark.api.java.JavaSparkContext; import org.apache.spark.api.java.function.Function; -import org.junit.Test; -import org.junit.After; -import org.junit.Before; - public class JavaKafkaRDDSuite implements Serializable { private transient JavaSparkContext sc = null; private transient KafkaStreamSuiteBase suiteBase = null; @@ -78,8 +73,8 @@ public void testKafkaRDD() throws InterruptedException { OffsetRange.create(topic2, 0, 0, 1) }; - HashMap emptyLeaders = new HashMap(); - HashMap leaders = new HashMap(); + HashMap emptyLeaders = new HashMap(); + HashMap leaders = new HashMap(); String[] hostAndPort = suiteBase.brokerAddress().split(":"); Broker broker = Broker.create(hostAndPort[0], Integer.parseInt(hostAndPort[1])); leaders.put(new TopicAndPartition(topic1, 0), broker); @@ -96,7 +91,7 @@ public void testKafkaRDD() throws InterruptedException { ).map( new Function, String>() { @Override - public String call(scala.Tuple2 kv) throws Exception { + public String call(Tuple2 kv) throws Exception { return kv._2(); } } diff --git a/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaKafkaStreamSuite.java b/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaKafkaStreamSuite.java index 208cc51b29876..bad0a93eb2e84 100644 --- a/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaKafkaStreamSuite.java +++ b/external/kafka/src/test/java/org/apache/spark/streaming/kafka/JavaKafkaStreamSuite.java @@ -22,27 +22,25 @@ import java.util.List; import java.util.Random; -import org.apache.spark.SparkConf; -import org.apache.spark.streaming.Duration; import scala.Predef; import scala.Tuple2; import scala.collection.JavaConverters; -import junit.framework.Assert; - import kafka.serializer.StringDecoder; +import org.junit.After; +import org.junit.Assert; +import org.junit.Before; +import org.junit.Test; +import org.apache.spark.SparkConf; import org.apache.spark.api.java.JavaPairRDD; import org.apache.spark.api.java.function.Function; import org.apache.spark.storage.StorageLevel; +import org.apache.spark.streaming.Duration; import org.apache.spark.streaming.api.java.JavaDStream; import org.apache.spark.streaming.api.java.JavaPairDStream; import org.apache.spark.streaming.api.java.JavaStreamingContext; -import org.junit.Test; -import org.junit.After; -import org.junit.Before; - public class JavaKafkaStreamSuite implements Serializable { private transient JavaStreamingContext ssc = null; private transient Random random = new Random(); diff --git a/mllib/src/main/scala/org/apache/spark/mllib/recommendation/MatrixFactorizationModel.scala b/mllib/src/main/scala/org/apache/spark/mllib/recommendation/MatrixFactorizationModel.scala index c399496568bfb..5f5a996a87b81 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/recommendation/MatrixFactorizationModel.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/recommendation/MatrixFactorizationModel.scala @@ -199,12 +199,12 @@ object MatrixFactorizationModel extends Loader[MatrixFactorizationModel] { assert(formatVersion == thisFormatVersion) val rank = (metadata \ "rank").extract[Int] val userFeatures = sqlContext.parquetFile(userPath(path)) - .map { case Row(id: Int, features: Seq[Double]) => - (id, features.toArray) + .map { case Row(id: Int, features: Seq[_]) => + (id, features.asInstanceOf[Seq[Double]].toArray) } val productFeatures = sqlContext.parquetFile(productPath(path)) - .map { case Row(id: Int, features: Seq[Double]) => - (id, features.toArray) + .map { case Row(id: Int, features: Seq[_]) => + (id, features.asInstanceOf[Seq[Double]].toArray) } new MatrixFactorizationModel(rank, userFeatures, productFeatures) } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/sources/ddl.scala b/sql/core/src/main/scala/org/apache/spark/sql/sources/ddl.scala index 5020689f7a105..76754a6ce4617 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/sources/ddl.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/sources/ddl.scala @@ -17,6 +17,7 @@ package org.apache.spark.sql.sources +import scala.language.existentials import scala.language.implicitConversions import org.apache.spark.Logging diff --git a/sql/core/src/test/scala/org/apache/spark/sql/ScalaReflectionRelationSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/ScalaReflectionRelationSuite.scala index 23df6e7eac043..17e923ca48502 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/ScalaReflectionRelationSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/ScalaReflectionRelationSuite.scala @@ -86,7 +86,7 @@ class ScalaReflectionRelationSuite extends FunSuite { assert(sql("SELECT * FROM reflectData").collect().head === Row("a", 1, 1L, 1.toFloat, 1.toDouble, 1.toShort, 1.toByte, true, - new java.math.BigDecimal(1), new Date(70, 0, 1), // This is 1970-01-01 + new java.math.BigDecimal(1), Date.valueOf("1970-01-01"), new Timestamp(12345), Seq(1,2,3))) } diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/hive/HiveInspectorSuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/hive/HiveInspectorSuite.scala index 09bbd5c867e4e..3181cfe40016c 100644 --- a/sql/hive/src/test/scala/org/apache/spark/sql/hive/HiveInspectorSuite.scala +++ b/sql/hive/src/test/scala/org/apache/spark/sql/hive/HiveInspectorSuite.scala @@ -75,7 +75,7 @@ class HiveInspectorSuite extends FunSuite with HiveInspectors { Literal(0.asInstanceOf[Float]) :: Literal(0.asInstanceOf[Double]) :: Literal("0") :: - Literal(new java.sql.Date(114, 8, 23)) :: + Literal(java.sql.Date.valueOf("2014-09-23")) :: Literal(Decimal(BigDecimal(123.123))) :: Literal(new java.sql.Timestamp(123123)) :: Literal(Array[Byte](1,2,3)) :: diff --git a/streaming/src/test/java/org/apache/spark/streaming/JavaAPISuite.java b/streaming/src/test/java/org/apache/spark/streaming/JavaAPISuite.java index 57302ff407183..90340753a4eed 100644 --- a/streaming/src/test/java/org/apache/spark/streaming/JavaAPISuite.java +++ b/streaming/src/test/java/org/apache/spark/streaming/JavaAPISuite.java @@ -316,6 +316,7 @@ public void testReduceByWindowWithoutInverse() { testReduceByWindow(false); } + @SuppressWarnings("unchecked") private void testReduceByWindow(boolean withInverse) { List> inputData = Arrays.asList( Arrays.asList(1,2,3), @@ -684,6 +685,7 @@ public void testStreamingContextTransform(){ JavaDStream transformed1 = ssc.transform( listOfDStreams1, new Function2>, Time, JavaRDD>() { + @Override public JavaRDD call(List> listOfRDDs, Time time) { Assert.assertEquals(2, listOfRDDs.size()); return null; @@ -697,6 +699,7 @@ public JavaRDD call(List> listOfRDDs, Time time) { JavaPairDStream> transformed2 = ssc.transformToPair( listOfDStreams2, new Function2>, Time, JavaPairRDD>>() { + @Override public JavaPairRDD> call(List> listOfRDDs, Time time) { Assert.assertEquals(3, listOfRDDs.size()); JavaRDD rdd1 = (JavaRDD)listOfRDDs.get(0); @@ -1829,6 +1832,7 @@ private List> fileTestPrepare(File testDir) throws IOException { return expected; } + @SuppressWarnings("unchecked") // SPARK-5795: no logic assertions, just testing that intended API invocations compile private void compileSaveAsJavaAPI(JavaPairDStream pds) { pds.saveAsNewAPIHadoopFiles( From 5b335bdda3efb7c6a5b18b4eeff189064c11e6c3 Mon Sep 17 00:00:00 2001 From: Marcelo Vanzin Date: Wed, 11 Mar 2015 13:16:22 +0000 Subject: [PATCH 031/122] [SPARK-6228] [network] Move SASL classes from network/shuffle to network... .../common. No code changes. Left the shuffle-related files in the shuffle module. Author: Marcelo Vanzin Closes #4953 from vanzin/SPARK-6228 and squashes the following commits: 664ef30 [Marcelo Vanzin] [SPARK-6228] [network] Move SASL classes from network/shuffle to network/common. --- .../java/org/apache/spark/network/sasl/SaslClientBootstrap.java | 0 .../src/main/java/org/apache/spark/network/sasl/SaslMessage.java | 0 .../main/java/org/apache/spark/network/sasl/SaslRpcHandler.java | 0 .../main/java/org/apache/spark/network/sasl/SecretKeyHolder.java | 0 .../main/java/org/apache/spark/network/sasl/SparkSaslClient.java | 0 .../main/java/org/apache/spark/network/sasl/SparkSaslServer.java | 0 .../test/java/org/apache/spark/network/sasl/SparkSaslSuite.java | 0 7 files changed, 0 insertions(+), 0 deletions(-) rename network/{shuffle => common}/src/main/java/org/apache/spark/network/sasl/SaslClientBootstrap.java (100%) rename network/{shuffle => common}/src/main/java/org/apache/spark/network/sasl/SaslMessage.java (100%) rename network/{shuffle => common}/src/main/java/org/apache/spark/network/sasl/SaslRpcHandler.java (100%) rename network/{shuffle => common}/src/main/java/org/apache/spark/network/sasl/SecretKeyHolder.java (100%) rename network/{shuffle => common}/src/main/java/org/apache/spark/network/sasl/SparkSaslClient.java (100%) rename network/{shuffle => common}/src/main/java/org/apache/spark/network/sasl/SparkSaslServer.java (100%) rename network/{shuffle => common}/src/test/java/org/apache/spark/network/sasl/SparkSaslSuite.java (100%) diff --git a/network/shuffle/src/main/java/org/apache/spark/network/sasl/SaslClientBootstrap.java b/network/common/src/main/java/org/apache/spark/network/sasl/SaslClientBootstrap.java similarity index 100% rename from network/shuffle/src/main/java/org/apache/spark/network/sasl/SaslClientBootstrap.java rename to network/common/src/main/java/org/apache/spark/network/sasl/SaslClientBootstrap.java diff --git a/network/shuffle/src/main/java/org/apache/spark/network/sasl/SaslMessage.java b/network/common/src/main/java/org/apache/spark/network/sasl/SaslMessage.java similarity index 100% rename from network/shuffle/src/main/java/org/apache/spark/network/sasl/SaslMessage.java rename to network/common/src/main/java/org/apache/spark/network/sasl/SaslMessage.java diff --git a/network/shuffle/src/main/java/org/apache/spark/network/sasl/SaslRpcHandler.java b/network/common/src/main/java/org/apache/spark/network/sasl/SaslRpcHandler.java similarity index 100% rename from network/shuffle/src/main/java/org/apache/spark/network/sasl/SaslRpcHandler.java rename to network/common/src/main/java/org/apache/spark/network/sasl/SaslRpcHandler.java diff --git a/network/shuffle/src/main/java/org/apache/spark/network/sasl/SecretKeyHolder.java b/network/common/src/main/java/org/apache/spark/network/sasl/SecretKeyHolder.java similarity index 100% rename from network/shuffle/src/main/java/org/apache/spark/network/sasl/SecretKeyHolder.java rename to network/common/src/main/java/org/apache/spark/network/sasl/SecretKeyHolder.java diff --git a/network/shuffle/src/main/java/org/apache/spark/network/sasl/SparkSaslClient.java b/network/common/src/main/java/org/apache/spark/network/sasl/SparkSaslClient.java similarity index 100% rename from network/shuffle/src/main/java/org/apache/spark/network/sasl/SparkSaslClient.java rename to network/common/src/main/java/org/apache/spark/network/sasl/SparkSaslClient.java diff --git a/network/shuffle/src/main/java/org/apache/spark/network/sasl/SparkSaslServer.java b/network/common/src/main/java/org/apache/spark/network/sasl/SparkSaslServer.java similarity index 100% rename from network/shuffle/src/main/java/org/apache/spark/network/sasl/SparkSaslServer.java rename to network/common/src/main/java/org/apache/spark/network/sasl/SparkSaslServer.java diff --git a/network/shuffle/src/test/java/org/apache/spark/network/sasl/SparkSaslSuite.java b/network/common/src/test/java/org/apache/spark/network/sasl/SparkSaslSuite.java similarity index 100% rename from network/shuffle/src/test/java/org/apache/spark/network/sasl/SparkSaslSuite.java rename to network/common/src/test/java/org/apache/spark/network/sasl/SparkSaslSuite.java From 548643a9e4690b69e2a496cdcd0a426b6de8d8b5 Mon Sep 17 00:00:00 2001 From: Ilya Ganelin Date: Wed, 11 Mar 2015 13:20:15 +0000 Subject: [PATCH 032/122] [SPARK-4423] Improve foreach() documentation to avoid confusion between local- and cluster-mode behavior Hi all - I've added a writeup on how closures work within Spark to help clarify the general case for this problem and similar problems. I hope this addresses the issue and would love any feedback. Author: Ilya Ganelin Closes #4696 from ilganeli/SPARK-4423 and squashes the following commits: c5dc498 [Ilya Ganelin] Fixed typo 07b78e8 [Ilya Ganelin] Updated to fix capitalization 48c1983 [Ilya Ganelin] Updated to fix capitalization and clarify wording 2fd2a07 [Ilya Ganelin] Incoporated a few more minor fixes. Fixed a bug in python code. Added semicolons for java 4772f99 [Ilya Ganelin] Incorporated latest feedback 448bd79 [Ilya Ganelin] Updated some verbage and added section links 5dbbda5 [Ilya Ganelin] Improved some wording d374d3a [Ilya Ganelin] Merge remote-tracking branch 'upstream/master' into SPARK-4423 2600668 [Ilya Ganelin] Minor edits c768ab2 [Ilya Ganelin] Updated documentation to add a section on closures. This helps understand confusing behavior of foreach and map functions when attempting to modify variables outside of the scope of an RDD action or transformation --- docs/programming-guide.md | 72 ++++++++++++++++++++++++++++++++++++--- 1 file changed, 68 insertions(+), 4 deletions(-) diff --git a/docs/programming-guide.md b/docs/programming-guide.md index fa0b4e3705d6e..c011a8404f7c9 100644 --- a/docs/programming-guide.md +++ b/docs/programming-guide.md @@ -725,7 +725,7 @@ class MyClass(object): def __init__(self): self.field = "Hello" def doStuff(self, rdd): - return rdd.map(lambda s: self.field + x) + return rdd.map(lambda s: self.field + s) {% endhighlight %} To avoid this issue, the simplest way is to copy `field` into a local variable instead @@ -734,13 +734,76 @@ of accessing it externally: {% highlight python %} def doStuff(self, rdd): field = self.field - return rdd.map(lambda s: field + x) + return rdd.map(lambda s: field + s) {% endhighlight %}
    +### Understanding closures +One of the harder things about Spark is understanding the scope and life cycle of variables and methods when executing code across a cluster. RDD operations that modify variables outside of their scope can be a frequent source of confusion. In the example below we'll look at code that uses `foreach()` to increment a counter, but similar issues can occur for other operations as well. + +#### Example + +Consider the naive RDD element sum below, which behaves completely differently depending on whether execution is happening within the same JVM. A common example of this is when running Spark in `local` mode (`--master = local[n]`) versus deploying a Spark application to a cluster (e.g. via spark-submit to YARN): + +
    + +
    +{% highlight scala %} +var counter = 0 +var rdd = sc.parallelize(data) + +// Wrong: Don't do this!! +rdd.foreach(x => counter += x) + +println("Counter value: " + counter) +{% endhighlight %} +
    + +
    +{% highlight java %} +int counter = 0; +JavaRDD rdd = sc.parallelize(data); + +// Wrong: Don't do this!! +rdd.foreach(x -> counter += x); + +println("Counter value: " + counter); +{% endhighlight %} +
    + +
    +{% highlight python %} +counter = 0 +rdd = sc.parallelize(data) + +# Wrong: Don't do this!! +rdd.foreach(lambda x: counter += x) + +print("Counter value: " + counter) + +{% endhighlight %} +
    + +
    + +#### Local vs. cluster modes + +The primary challenge is that the behavior of the above code is undefined. In local mode with a single JVM, the above code will sum the values within the RDD and store it in **counter**. This is because both the RDD and the variable **counter** are in the same memory space on the driver node. + +However, in `cluster` mode, what happens is more complicated, and the above may not work as intended. To execute jobs, Spark breaks up the processing of RDD operations into tasks - each of which is operated on by an executor. Prior to execution, Spark computes the **closure**. The closure is those variables and methods which must be visible for the executor to perform its computations on the RDD (in this case `foreach()`). This closure is serialized and sent to each executor. In `local` mode, there is only the one executors so everything shares the same closure. In other modes however, this is not the case and the executors running on seperate worker nodes each have their own copy of the closure. + +What is happening here is that the variables within the closure sent to each executor are now copies and thus, when **counter** is referenced within the `foreach` function, it's no longer the **counter** on the driver node. There is still a **counter** in the memory of the driver node but this is no longer visible to the executors! The executors only sees the copy from the serialized closure. Thus, the final value of **counter** will still be zero since all operations on **counter** were referencing the value within the serialized closure. + +To ensure well-defined behavior in these sorts of scenarios one should use an [`Accumulator`](#AccumLink). Accumulators in Spark are used specifically to provide a mechanism for safely updating a variable when execution is split up across worker nodes in a cluster. The Accumulators section of this guide discusses these in more detail. + +In general, closures - constructs like loops or locally defined methods, should not be used to mutate some global state. Spark does not define or guarantee the behavior of mutations to objects referenced from outside of closures. Some code that does this may work in local mode, but that's just by accident and such code will not behave as expected in distributed mode. Use an Accumulator instead if some global aggregation is needed. + +#### Printing elements of an RDD +Another common idiom is attempting to print out the elements of an RDD using `rdd.foreach(println)` or `rdd.map(println)`. On a single machine, this will generate the expected output and print all the RDD's elements. However, in `cluster` mode, the output to `stdout` being called by the executors is now writing to the executor's `stdout` instead, not the one on the driver, so `stdout` on the driver won't show these! To print all elements on the driver, one can use the `collect()` method to first bring the RDD to the driver node thus: `rdd.collect().foreach(println)`. This can cause the driver to run out of memory, though, because `collect()` fetches the entire RDD to a single machine; if you only need to print a few elements of the RDD, a safer approach is to use the `take()`: `rdd.take(100).foreach(println)`. + ### Working with Key-Value Pairs
    @@ -1018,7 +1081,8 @@ for details. foreach(func) - Run a function func on each element of the dataset. This is usually done for side effects such as updating an accumulator variable (see below) or interacting with external storage systems. + Run a function func on each element of the dataset. This is usually done for side effects such as updating an Accumulator or interacting with external storage systems. +
    Note: modifying variables other than Accumulators outside of the foreach() may result in undefined behavior. See Understanding closures for more details. @@ -1191,7 +1255,7 @@ run on the cluster so that `v` is not shipped to the nodes more than once. In ad `v` should not be modified after it is broadcast in order to ensure that all nodes get the same value of the broadcast variable (e.g. if the variable is shipped to a new node later). -## Accumulators +## Accumulators Accumulators are variables that are only "added" to through an associative operation and can therefore be efficiently supported in parallel. They can be used to implement counters (as in From 2d87a415f20c85487537d6791a73827ff537f2c0 Mon Sep 17 00:00:00 2001 From: Sandy Ryza Date: Wed, 11 Mar 2015 13:22:05 +0000 Subject: [PATCH 033/122] SPARK-3642. Document the nuances of shared variables. Author: Sandy Ryza Closes #2490 from sryza/sandy-spark-3642 and squashes the following commits: aae3340 [Sandy Ryza] SPARK-3642. Document the nuances of broadcast variables --- docs/programming-guide.md | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/docs/programming-guide.md b/docs/programming-guide.md index c011a8404f7c9..eda3a95426182 100644 --- a/docs/programming-guide.md +++ b/docs/programming-guide.md @@ -1207,6 +1207,12 @@ than shipping a copy of it with tasks. They can be used, for example, to give ev large input dataset in an efficient manner. Spark also attempts to distribute broadcast variables using efficient broadcast algorithms to reduce communication cost. +Spark actions are executed through a set of stages, separated by distributed "shuffle" operations. +Spark automatically broadcasts the common data needed by tasks within each stage. The data +broadcasted this way is cached in serialized form and deserialized before running each task. This +means that explicitly creating broadcast variables is only useful when tasks across multiple stages +need the same data or when caching the data in deserialized form is important. + Broadcast variables are created from a variable `v` by calling `SparkContext.broadcast(v)`. The broadcast variable is a wrapper around `v`, and its value can be accessed by calling the `value` method. The code below shows this: From 55c4831d68c8326380086b5540244f984ea9ec27 Mon Sep 17 00:00:00 2001 From: Sean Owen Date: Wed, 11 Mar 2015 14:09:09 +0000 Subject: [PATCH 034/122] SPARK-6245 [SQL] jsonRDD() of empty RDD results in exception Avoid `UnsupportedOperationException` from JsonRDD.inferSchema on empty RDD. Not sure if this is supposed to be an error (but a better one), but it seems like this case can come up if the input is down-sampled so much that nothing is sampled. Now stuff like this: ``` sqlContext.jsonRDD(sc.parallelize(List[String]())) ``` just results in ``` org.apache.spark.sql.DataFrame = [] ``` Author: Sean Owen Closes #4971 from srowen/SPARK-6245 and squashes the following commits: 3699964 [Sean Owen] Set() -> Set.empty 3c619e1 [Sean Owen] Avoid UnsupportedOperationException from JsonRDD.inferSchema on empty RDD --- .../src/main/scala/org/apache/spark/sql/json/JsonRDD.scala | 6 +++++- .../test/scala/org/apache/spark/sql/json/JsonSuite.scala | 7 +++++++ .../scala/org/apache/spark/sql/json/TestJsonData.scala | 3 +++ 3 files changed, 15 insertions(+), 1 deletion(-) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/json/JsonRDD.scala b/sql/core/src/main/scala/org/apache/spark/sql/json/JsonRDD.scala index e54a2a3679272..2b0358c4e2a1e 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/json/JsonRDD.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/json/JsonRDD.scala @@ -48,7 +48,11 @@ private[sql] object JsonRDD extends Logging { require(samplingRatio > 0, s"samplingRatio ($samplingRatio) should be greater than 0") val schemaData = if (samplingRatio > 0.99) json else json.sample(false, samplingRatio, 1) val allKeys = - parseJson(schemaData, columnNameOfCorruptRecords).map(allKeysWithValueTypes).reduce(_ ++ _) + if (schemaData.isEmpty()) { + Set.empty[(String,DataType)] + } else { + parseJson(schemaData, columnNameOfCorruptRecords).map(allKeysWithValueTypes).reduce(_ ++ _) + } createSchema(allKeys) } diff --git a/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala index 0c21f725f0b49..320b80d80e997 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala @@ -1033,4 +1033,11 @@ class JsonSuite extends QueryTest { assert(!logicalRelation2.sameResult(logicalRelation3), s"$logicalRelation2 and $logicalRelation3 should be considered not having the same result.") } + + test("SPARK-6245 JsonRDD.inferSchema on empty RDD") { + // This is really a test that it doesn't throw an exception + val emptySchema = JsonRDD.inferSchema(empty, 1.0, "") + assert(StructType(Seq()) === emptySchema) + } + } diff --git a/sql/core/src/test/scala/org/apache/spark/sql/json/TestJsonData.scala b/sql/core/src/test/scala/org/apache/spark/sql/json/TestJsonData.scala index 15698f61e0837..47a97a49daabb 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/json/TestJsonData.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/json/TestJsonData.scala @@ -185,4 +185,7 @@ object TestJsonData { """{"a":{, b:3}""" :: """{"b":"str_b_4", "a":"str_a_4", "c":"str_c_4"}""" :: """]""" :: Nil) + + val empty = + TestSQLContext.sparkContext.parallelize(Seq[String]()) } From 51a79a770a8356bd0ed244af5ca7f1c44c9437d2 Mon Sep 17 00:00:00 2001 From: Tathagata Das Date: Wed, 11 Mar 2015 11:19:51 -0700 Subject: [PATCH 035/122] [SPARK-6274][Streaming][Examples] Added examples streaming + sql examples. Added Scala, Java and Python streaming examples showing DataFrame and SQL operations within streaming. Author: Tathagata Das Closes #4975 from tdas/streaming-sql-examples and squashes the following commits: 705cba1 [Tathagata Das] Fixed python lint error 75a3fad [Tathagata Das] Fixed python lint error 5fbf789 [Tathagata Das] Removed empty lines at the end 874b943 [Tathagata Das] Added examples streaming + sql examples. --- .../spark/examples/streaming/JavaRecord.java | 31 +++++ .../streaming/JavaSqlNetworkWordCount.java | 122 ++++++++++++++++++ .../python/streaming/sql_network_wordcount.py | 82 ++++++++++++ .../streaming/SqlNetworkWordCount.scala | 101 +++++++++++++++ 4 files changed, 336 insertions(+) create mode 100644 examples/src/main/java/org/apache/spark/examples/streaming/JavaRecord.java create mode 100644 examples/src/main/java/org/apache/spark/examples/streaming/JavaSqlNetworkWordCount.java create mode 100644 examples/src/main/python/streaming/sql_network_wordcount.py create mode 100644 examples/src/main/scala/org/apache/spark/examples/streaming/SqlNetworkWordCount.scala diff --git a/examples/src/main/java/org/apache/spark/examples/streaming/JavaRecord.java b/examples/src/main/java/org/apache/spark/examples/streaming/JavaRecord.java new file mode 100644 index 0000000000000..e63697a79f23a --- /dev/null +++ b/examples/src/main/java/org/apache/spark/examples/streaming/JavaRecord.java @@ -0,0 +1,31 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.examples.streaming; + +/** Java Bean class to be used with the example JavaSqlNetworkWordCount. */ +public class JavaRecord implements java.io.Serializable { + private String word; + + public String getWord() { + return word; + } + + public void setWord(String word) { + this.word = word; + } +} diff --git a/examples/src/main/java/org/apache/spark/examples/streaming/JavaSqlNetworkWordCount.java b/examples/src/main/java/org/apache/spark/examples/streaming/JavaSqlNetworkWordCount.java new file mode 100644 index 0000000000000..46562ddbbcb57 --- /dev/null +++ b/examples/src/main/java/org/apache/spark/examples/streaming/JavaSqlNetworkWordCount.java @@ -0,0 +1,122 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.examples.streaming; + +import java.util.regex.Pattern; + +import com.google.common.collect.Lists; + +import org.apache.spark.SparkConf; +import org.apache.spark.SparkContext; +import org.apache.spark.api.java.JavaRDD; +import org.apache.spark.api.java.function.FlatMapFunction; +import org.apache.spark.api.java.function.Function; +import org.apache.spark.api.java.function.Function2; +import org.apache.spark.sql.SQLContext; +import org.apache.spark.sql.DataFrame; +import org.apache.spark.api.java.StorageLevels; +import org.apache.spark.streaming.Durations; +import org.apache.spark.streaming.Time; +import org.apache.spark.streaming.api.java.JavaDStream; +import org.apache.spark.streaming.api.java.JavaReceiverInputDStream; +import org.apache.spark.streaming.api.java.JavaStreamingContext; + +/** + * Use DataFrames and SQL to count words in UTF8 encoded, '\n' delimited text received from the + * network every second. + * + * Usage: JavaSqlNetworkWordCount + * and describe the TCP server that Spark Streaming would connect to receive data. + * + * To run this on your local machine, you need to first run a Netcat server + * `$ nc -lk 9999` + * and then run the example + * `$ bin/run-example org.apache.spark.examples.streaming.JavaSqlNetworkWordCount localhost 9999` + */ + +public final class JavaSqlNetworkWordCount { + private static final Pattern SPACE = Pattern.compile(" "); + + public static void main(String[] args) { + if (args.length < 2) { + System.err.println("Usage: JavaNetworkWordCount "); + System.exit(1); + } + + StreamingExamples.setStreamingLogLevels(); + + // Create the context with a 1 second batch size + SparkConf sparkConf = new SparkConf().setAppName("JavaSqlNetworkWordCount"); + JavaStreamingContext ssc = new JavaStreamingContext(sparkConf, Durations.seconds(1)); + + // Create a JavaReceiverInputDStream on target ip:port and count the + // words in input stream of \n delimited text (eg. generated by 'nc') + // Note that no duplication in storage level only for running locally. + // Replication necessary in distributed scenario for fault tolerance. + JavaReceiverInputDStream lines = ssc.socketTextStream( + args[0], Integer.parseInt(args[1]), StorageLevels.MEMORY_AND_DISK_SER); + JavaDStream words = lines.flatMap(new FlatMapFunction() { + @Override + public Iterable call(String x) { + return Lists.newArrayList(SPACE.split(x)); + } + }); + + // Convert RDDs of the words DStream to DataFrame and run SQL query + words.foreachRDD(new Function2, Time, Void>() { + @Override + public Void call(JavaRDD rdd, Time time) { + SQLContext sqlContext = JavaSQLContextSingleton.getInstance(rdd.context()); + + // Convert JavaRDD[String] to JavaRDD[bean class] to DataFrame + JavaRDD rowRDD = rdd.map(new Function() { + public JavaRecord call(String word) { + JavaRecord record = new JavaRecord(); + record.setWord(word); + return record; + } + }); + DataFrame wordsDataFrame = sqlContext.createDataFrame(rowRDD, JavaRecord.class); + + // Register as table + wordsDataFrame.registerTempTable("words"); + + // Do word count on table using SQL and print it + DataFrame wordCountsDataFrame = + sqlContext.sql("select word, count(*) as total from words group by word"); + System.out.println("========= " + time + "========="); + wordCountsDataFrame.show(); + return null; + } + }); + + ssc.start(); + ssc.awaitTermination(); + } +} + +/** Lazily instantiated singleton instance of SQLContext */ +class JavaSQLContextSingleton { + static private transient SQLContext instance = null; + static public SQLContext getInstance(SparkContext sparkContext) { + if (instance == null) { + instance = new SQLContext(sparkContext); + } + return instance; + } +} diff --git a/examples/src/main/python/streaming/sql_network_wordcount.py b/examples/src/main/python/streaming/sql_network_wordcount.py new file mode 100644 index 0000000000000..f89bc562d856b --- /dev/null +++ b/examples/src/main/python/streaming/sql_network_wordcount.py @@ -0,0 +1,82 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +""" + Use DataFrames and SQL to count words in UTF8 encoded, '\n' delimited text received from the + network every second. + + Usage: sql_network_wordcount.py + and describe the TCP server that Spark Streaming would connect to receive data. + + To run this on your local machine, you need to first run a Netcat server + `$ nc -lk 9999` + and then run the example + `$ bin/spark-submit examples/src/main/python/streaming/sql_network_wordcount.py localhost 9999` +""" + +import os +import sys + +from pyspark import SparkContext +from pyspark.streaming import StreamingContext +from pyspark.sql import SQLContext, Row + + +def getSqlContextInstance(sparkContext): + if ('sqlContextSingletonInstance' not in globals()): + globals()['sqlContextSingletonInstance'] = SQLContext(sparkContext) + return globals()['sqlContextSingletonInstance'] + + +if __name__ == "__main__": + if len(sys.argv) != 3: + print >> sys.stderr, "Usage: sql_network_wordcount.py " + exit(-1) + host, port = sys.argv[1:] + sc = SparkContext(appName="PythonSqlNetworkWordCount") + ssc = StreamingContext(sc, 1) + + # Create a socket stream on target ip:port and count the + # words in input stream of \n delimited text (eg. generated by 'nc') + lines = ssc.socketTextStream(host, int(port)) + words = lines.flatMap(lambda line: line.split(" ")) + + # Convert RDDs of the words DStream to DataFrame and run SQL query + def process(time, rdd): + print "========= %s =========" % str(time) + + try: + # Get the singleton instance of SQLContext + sqlContext = getSqlContextInstance(rdd.context) + + # Convert RDD[String] to RDD[Row] to DataFrame + rowRdd = rdd.map(lambda w: Row(word=w)) + wordsDataFrame = sqlContext.createDataFrame(rowRdd) + + # Register as table + wordsDataFrame.registerTempTable("words") + + # Do word count on table using SQL and print it + wordCountsDataFrame = \ + sqlContext.sql("select word, count(*) as total from words group by word") + wordCountsDataFrame.show() + except: + pass + + words.foreachRDD(process) + ssc.start() + ssc.awaitTermination() diff --git a/examples/src/main/scala/org/apache/spark/examples/streaming/SqlNetworkWordCount.scala b/examples/src/main/scala/org/apache/spark/examples/streaming/SqlNetworkWordCount.scala new file mode 100644 index 0000000000000..5a6b9216a3fbc --- /dev/null +++ b/examples/src/main/scala/org/apache/spark/examples/streaming/SqlNetworkWordCount.scala @@ -0,0 +1,101 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.examples.streaming + +import org.apache.spark.SparkConf +import org.apache.spark.SparkContext +import org.apache.spark.rdd.RDD +import org.apache.spark.streaming.{Time, Seconds, StreamingContext} +import org.apache.spark.util.IntParam +import org.apache.spark.sql.SQLContext +import org.apache.spark.storage.StorageLevel + +/** + * Use DataFrames and SQL to count words in UTF8 encoded, '\n' delimited text received from the + * network every second. + * + * Usage: SqlNetworkWordCount + * and describe the TCP server that Spark Streaming would connect to receive data. + * + * To run this on your local machine, you need to first run a Netcat server + * `$ nc -lk 9999` + * and then run the example + * `$ bin/run-example org.apache.spark.examples.streaming.SqlNetworkWordCount localhost 9999` + */ + +object SqlNetworkWordCount { + def main(args: Array[String]) { + if (args.length < 2) { + System.err.println("Usage: NetworkWordCount ") + System.exit(1) + } + + StreamingExamples.setStreamingLogLevels() + + // Create the context with a 2 second batch size + val sparkConf = new SparkConf().setAppName("SqlNetworkWordCount") + val ssc = new StreamingContext(sparkConf, Seconds(2)) + + // Create a socket stream on target ip:port and count the + // words in input stream of \n delimited text (eg. generated by 'nc') + // Note that no duplication in storage level only for running locally. + // Replication necessary in distributed scenario for fault tolerance. + val lines = ssc.socketTextStream(args(0), args(1).toInt, StorageLevel.MEMORY_AND_DISK_SER) + val words = lines.flatMap(_.split(" ")) + + // Convert RDDs of the words DStream to DataFrame and run SQL query + words.foreachRDD((rdd: RDD[String], time: Time) => { + // Get the singleton instance of SQLContext + val sqlContext = SQLContextSingleton.getInstance(rdd.sparkContext) + import sqlContext.implicits._ + + // Convert RDD[String] to RDD[case class] to DataFrame + val wordsDataFrame = rdd.map(w => Record(w)).toDF() + + // Register as table + wordsDataFrame.registerTempTable("words") + + // Do word count on table using SQL and print it + val wordCountsDataFrame = + sqlContext.sql("select word, count(*) as total from words group by word") + println(s"========= $time =========") + wordCountsDataFrame.show() + }) + + ssc.start() + ssc.awaitTermination() + } +} + + +/** Case class for converting RDD to DataFrame */ +case class Record(word: String) + + +/** Lazily instantiated singleton instance of SQLContext */ +object SQLContextSingleton { + + @transient private var instance: SQLContext = _ + + def getInstance(sparkContext: SparkContext): SQLContext = { + if (instance == null) { + instance = new SQLContext(sparkContext) + } + instance + } +} From cd3b68d93a01f11bd3d5a441b341cb33d227e900 Mon Sep 17 00:00:00 2001 From: Tathagata Das Date: Wed, 11 Mar 2015 18:48:21 -0700 Subject: [PATCH 036/122] [SPARK-6128][Streaming][Documentation] Updates to Spark Streaming Programming Guide Updates to the documentation are as follows: - Added information on Kafka Direct API and Kafka Python API - Added joins to the main streaming guide - Improved details on the fault-tolerance semantics Generated docs located here http://people.apache.org/~tdas/spark-1.3.0-temp-docs/streaming-programming-guide.html#fault-tolerance-semantics More things to add: - Configuration for Kafka receive rate - May be add concurrentJobs Author: Tathagata Das Closes #4956 from tdas/streaming-guide-update-1.3 and squashes the following commits: 819408c [Tathagata Das] Minor fixes. debe484 [Tathagata Das] Added DataFrames and MLlib 380cf8d [Tathagata Das] Fix link 04167a6 [Tathagata Das] Merge remote-tracking branch 'apache-github/master' into streaming-guide-update-1.3 0b77486 [Tathagata Das] Updates based on Josh's comments. 86c4c2a [Tathagata Das] Updated streaming guides 82de92a [Tathagata Das] Add Kafka to Python api docs --- docs/configuration.md | 14 +- docs/streaming-flume-integration.md | 2 + docs/streaming-kafka-integration.md | 151 +++++++-- docs/streaming-programming-guide.md | 470 +++++++++++++++++++++++----- 4 files changed, 528 insertions(+), 109 deletions(-) diff --git a/docs/configuration.md b/docs/configuration.md index ae90fe1f8f6b9..a7116fbece9bb 100644 --- a/docs/configuration.md +++ b/docs/configuration.md @@ -1345,9 +1345,9 @@ Apart from these, the following properties are also available, and may be useful spark.streaming.receiver.maxRate - infinite + not set - Maximum number records per second at which each receiver will receive data. + Maximum rate (number of records per second) at which each receiver will receive data. Effectively, each stream will consume at most this number of records per second. Setting this configuration to 0 or a negative number will put no limit on the rate. See the deployment guide @@ -1375,6 +1375,16 @@ Apart from these, the following properties are also available, and may be useful higher memory usage in Spark. + + spark.streaming.kafka.maxRatePerPartition + not set + + Maximum rate (number of records per second) at which data will be read from each Kafka + partition when using the new Kafka direct stream API. See the + Kafka Integration guide + for more details. + + #### Cluster Managers diff --git a/docs/streaming-flume-integration.md b/docs/streaming-flume-integration.md index 40e17246fea83..c8ab146bcae0a 100644 --- a/docs/streaming-flume-integration.md +++ b/docs/streaming-flume-integration.md @@ -5,6 +5,8 @@ title: Spark Streaming + Flume Integration Guide [Apache Flume](https://flume.apache.org/) is a distributed, reliable, and available service for efficiently collecting, aggregating, and moving large amounts of log data. Here we explain how to configure Flume and Spark Streaming to receive data from Flume. There are two approaches to this. +Python API Flume is not yet available in the Python API. + ## Approach 1: Flume-style Push-based Approach Flume is designed to push data between Flume agents. In this approach, Spark Streaming essentially sets up a receiver that acts an Avro agent for Flume, to which Flume can push the data. Here are the configuration steps. diff --git a/docs/streaming-kafka-integration.md b/docs/streaming-kafka-integration.md index 77c0abbbacbd0..64714f0b799fc 100644 --- a/docs/streaming-kafka-integration.md +++ b/docs/streaming-kafka-integration.md @@ -2,58 +2,155 @@ layout: global title: Spark Streaming + Kafka Integration Guide --- -[Apache Kafka](http://kafka.apache.org/) is publish-subscribe messaging rethought as a distributed, partitioned, replicated commit log service. Here we explain how to configure Spark Streaming to receive data from Kafka. +[Apache Kafka](http://kafka.apache.org/) is publish-subscribe messaging rethought as a distributed, partitioned, replicated commit log service. Here we explain how to configure Spark Streaming to receive data from Kafka. There are two approaches to this - the old approach using Receivers and Kafka's high-level API, and a new experimental approach (introduced in Spark 1.3) without using Receivers. They have different programming models, performance characteristics, and semantics guarantees, so read on for more details. -1. **Linking:** In your SBT/Maven project definition, link your streaming application against the following artifact (see [Linking section](streaming-programming-guide.html#linking) in the main programming guide for further information). +## Approach 1: Receiver-based Approach +This approach uses a Receiver to receive the data. The Received is implemented using the Kafka high-level consumer API. As with all receivers, the data received from Kafka through a Receiver is stored in Spark executors, and then jobs launched by Spark Streaming processes the data. + +However, under default configuration, this approach can lose data under failures (see [receiver reliability](streaming-programming-guide.html#receiver-reliability). To ensure zero-data loss, you have to additionally enable Write Ahead Logs in Spark Streaming. To ensure zero data loss, enable the Write Ahead Logs (introduced in Spark 1.2). This synchronously saves all the received Kafka data into write ahead logs on a distributed file system (e.g HDFS), so that all the data can be recovered on failure. See [Deploying section](streaming-programming-guide.html#deploying-applications) in the streaming programming guide for more details on Write Ahead Logs. + +Next, we discuss how to use this approach in your streaming application. + +1. **Linking:** For Scala/Java applications using SBT/Maven project definitions, link your streaming application with the following artifact (see [Linking section](streaming-programming-guide.html#linking) in the main programming guide for further information). groupId = org.apache.spark artifactId = spark-streaming-kafka_{{site.SCALA_BINARY_VERSION}} version = {{site.SPARK_VERSION_SHORT}} -2. **Programming:** In the streaming application code, import `KafkaUtils` and create input DStream as follows. + For Python applications, you will have to add this above library and its dependencies when deploying your application. See the *Deploying* subsection below. + +2. **Programming:** In the streaming application code, import `KafkaUtils` and create an input DStream as follows.
    import org.apache.spark.streaming.kafka._ - val kafkaStream = KafkaUtils.createStream( - streamingContext, [zookeeperQuorum], [group id of the consumer], [per-topic number of Kafka partitions to consume]) + val kafkaStream = KafkaUtils.createStream(streamingContext, + [ZK quorum], [consumer group id], [per-topic number of Kafka partitions to consume]) - See the [API docs](api/scala/index.html#org.apache.spark.streaming.kafka.KafkaUtils$) + You can also specify the key and value classes and their corresponding decoder classes using variations of `createStream`. See the [API docs](api/scala/index.html#org.apache.spark.streaming.kafka.KafkaUtils$) and the [example]({{site.SPARK_GITHUB_URL}}/blob/master/examples/scala-2.10/src/main/scala/org/apache/spark/examples/streaming/KafkaWordCount.scala).
    import org.apache.spark.streaming.kafka.*; - JavaPairReceiverInputDStream kafkaStream = KafkaUtils.createStream( - streamingContext, [zookeeperQuorum], [group id of the consumer], [per-topic number of Kafka partitions to consume]); + JavaPairReceiverInputDStream kafkaStream = + KafkaUtils.createStream(streamingContext, + [ZK quorum], [consumer group id], [per-topic number of Kafka partitions to consume]); - See the [API docs](api/java/index.html?org/apache/spark/streaming/kafka/KafkaUtils.html) + You can also specify the key and value classes and their corresponding decoder classes using variations of `createStream`. See the [API docs](api/java/index.html?org/apache/spark/streaming/kafka/KafkaUtils.html) and the [example]({{site.SPARK_GITHUB_URL}}/blob/master/examples/scala-2.10/src/main/java/org/apache/spark/examples/streaming/JavaKafkaWordCount.java). + +
    +
    + from pyspark.streaming.kafka import KafkaUtils + + kafkaStream = KafkaUtils.createStream(streamingContext, \ + [ZK quorum], [consumer group id], [per-topic number of Kafka partitions to consume]) + + By default, the Python API will decode Kafka data as UTF8 encoded strings. You can specify your custom decoding function to decode the byte arrays in Kafka records to any arbitrary data type. See the [API docs](api/python/pyspark.streaming.html#pyspark.streaming.kafka.KafkaUtils) + and the [example]({{site.SPARK_GITHUB_URL}}/blob/master/examples/src/main/python/streaming/kafka_wordcount.py).
    - *Points to remember:* + **Points to remember:** - Topic partitions in Kafka does not correlate to partitions of RDDs generated in Spark Streaming. So increasing the number of topic-specific partitions in the `KafkaUtils.createStream()` only increases the number of threads using which topics that are consumed within a single receiver. It does not increase the parallelism of Spark in processing the data. Refer to the main document for more information on that. - Multiple Kafka input DStreams can be created with different groups and topics for parallel receiving of data using multiple receivers. -3. **Deploying:** Package `spark-streaming-kafka_{{site.SCALA_BINARY_VERSION}}` and its dependencies (except `spark-core_{{site.SCALA_BINARY_VERSION}}` and `spark-streaming_{{site.SCALA_BINARY_VERSION}}` which are provided by `spark-submit`) into the application JAR. Then use `spark-submit` to launch your application (see [Deploying section](streaming-programming-guide.html#deploying-applications) in the main programming guide). - -Note that the Kafka receiver used by default is an -[*unreliable* receiver](streaming-programming-guide.html#receiver-reliability) section in the -programming guide). In Spark 1.2, we have added an experimental *reliable* Kafka receiver that -provides stronger -[fault-tolerance guarantees](streaming-programming-guide.html#fault-tolerance-semantics) of zero -data loss on failures. This receiver is automatically used when the write ahead log -(also introduced in Spark 1.2) is enabled -(see [Deployment](#deploying-applications.html) section in the programming guide). This -may reduce the receiving throughput of individual Kafka receivers compared to the unreliable -receivers, but this can be corrected by running -[more receivers in parallel](streaming-programming-guide.html#level-of-parallelism-in-data-receiving) -to increase aggregate throughput. Additionally, it is recommended that the replication of the -received data within Spark be disabled when the write ahead log is enabled as the log is already stored -in a replicated storage system. This can be done by setting the storage level for the input -stream to `StorageLevel.MEMORY_AND_DISK_SER` (that is, use + - If you have enabled Write Ahead Logs with a replicated file system like HDFS, the received data is already being replicated in the log. Hence, the storage level in storage level for the input stream to `StorageLevel.MEMORY_AND_DISK_SER` (that is, use `KafkaUtils.createStream(..., StorageLevel.MEMORY_AND_DISK_SER)`). + +3. **Deploying:** As with any Spark applications, `spark-submit` is used to launch your application. However, the details are slightly different for Scala/Java applications and Python applications. + + For Scala and Java applications, if you are using SBT or Maven for project management, then package `spark-streaming-kafka_{{site.SCALA_BINARY_VERSION}}` and its dependencies into the application JAR. Make sure `spark-core_{{site.SCALA_BINARY_VERSION}}` and `spark-streaming_{{site.SCALA_BINARY_VERSION}}` are marked as `provided` dependencies as those are already present in a Spark installation. Then use `spark-submit` to launch your application (see [Deploying section](streaming-programming-guide.html#deploying-applications) in the main programming guide). + + For Python applications which lack SBT/Maven project management, `spark-streaming-kafka_{{site.SCALA_BINARY_VERSION}}` and its dependencies can be directly added to `spark-submit` using `--packages` (see [Application Submission Guide](submitting-applications.html)). That is, + + ./bin/spark-submit --packages org.apache.spark:spark-streaming-kafka_{{site.SCALA_BINARY_VERSION}}:{{site.SPARK_VERSION_SHORT}} ... + + Alternatively, you can also download the JAR of the Maven artifact `spark-streaming-kafka-assembly` from the + [Maven repository](http://search.maven.org/#search|ga|1|a%3A%22spark-streaming-kafka-assembly_2.10%22%20AND%20v%3A%22{{site.SPARK_VERSION_SHORT}}%22) and add it to `spark-submit` with `--jars`. + +## Approach 2: Direct Approach (No Receivers) +This is a new receiver-less "direct" approach has been introduced in Spark 1.3 to ensure stronger end-to-end guarantees. Instead of using receivers to receive data, this approach periodically queries Kafka for the latest offsets in each topic+partition, and accordingly defines the offset ranges to process in each batch. When the jobs to process the data are launched, Kafka's simple consumer API is used to read the defined ranges of offsets from Kafka (similar to read files from a file system). Note that this is an experimental feature in Spark 1.3 and is only available in the Scala and Java API. + +This approach has the following advantages over the received-based approach (i.e. Approach 1). + +- *Simplified Parallelism:* No need to create multiple input Kafka streams and union-ing them. With `directStream`, Spark Streaming will create as many RDD partitions as there is Kafka partitions to consume, which will all read data from Kafka in parallel. So there is one-to-one mapping between Kafka and RDD partitions, which is easier to understand and tune. + +- *Efficiency:* Achieving zero-data loss in the first approach required the data to be stored in a Write Ahead Log, which further replicated the data. This is actually inefficient as the data effectively gets replicated twice - once by Kafka, and a second time by the Write Ahead Log. This second approach eliminate the problem as there is no receiver, and hence no need for Write Ahead Logs. + +- *Exactly-once semantics:* The first approach uses Kafka's high level API to store consumed offsets in Zookeeper. This is traditionally the way to consume data from Kafka. While this approach (in combination with write ahead logs) can ensure zero data loss (i.e. at-least once semantics), there is a small chance some records may get consumed twice under some failures. This occurs because of inconsistencies between data reliably received by Spark Streaming and offsets tracked by Zookeeper. Hence, in this second approach, we use simple Kafka API that does not use Zookeeper and offsets tracked only by Spark Streaming within its checkpoints. This eliminates inconsistencies between Spark Streaming and Zookeeper/Kafka, and so each record is received by Spark Streaming effectively exactly once despite failures. + +Note that one disadvantage of this approach is that it does not update offsets in Zookeeper, hence Zookeeper-based Kafka monitoring tools will not show progress. However, you can access the offsets processed by this approach in each batch and update Zookeeper yourself (see below). + +Next, we discuss how to use this approach in your streaming application. + +1. **Linking:** This approach is supported only in Scala/Java application. Link your SBT/Maven project with the following artifact (see [Linking section](streaming-programming-guide.html#linking) in the main programming guide for further information). + + groupId = org.apache.spark + artifactId = spark-streaming-kafka_{{site.SCALA_BINARY_VERSION}} + version = {{site.SPARK_VERSION_SHORT}} + +2. **Programming:** In the streaming application code, import `KafkaUtils` and create an input DStream as follows. + +
    +
    + import org.apache.spark.streaming.kafka._ + + val directKafkaStream = KafkaUtils.createDirectStream[ + [key class], [value class], [key decoder class], [value decoder class] ]( + streamingContext, [map of Kafka parameters], [set of topics to consume]) + + See the [API docs](api/scala/index.html#org.apache.spark.streaming.kafka.KafkaUtils$) + and the [example]({{site.SPARK_GITHUB_URL}}/blob/master/examples/scala-2.10/src/main/scala/org/apache/spark/examples/streaming/DirectKafkaWordCount.scala). +
    +
    + import org.apache.spark.streaming.kafka.*; + + JavaPairReceiverInputDStream directKafkaStream = + KafkaUtils.createDirectStream(streamingContext, + [key class], [value class], [key decoder class], [value decoder class], + [map of Kafka parameters], [set of topics to consume]); + + See the [API docs](api/java/index.html?org/apache/spark/streaming/kafka/KafkaUtils.html) + and the [example]({{site.SPARK_GITHUB_URL}}/blob/master/examples/scala-2.10/src/main/java/org/apache/spark/examples/streaming/JavaDirectKafkaWordCount.java). + +
    +
    + + In the Kafka parameters, you must specify either `metadata.broker.list` or `bootstrap.servers`. + By default, it will start consuming from the latest offset of each Kafka partition. If you set configuration `auto.offset.reset` in Kafka parameters to `smallest`, then it will start consuming from the smallest offset. + + You can also start consuming from any arbitrary offset using other variations of `KafkaUtils.createDirectStream`. Furthermore, if you want to access the Kafka offsets consumed in each batch, you can do the following. + +
    +
    + directKafkaStream.foreachRDD { rdd => + val offsetRanges = rdd.asInstanceOf[HasOffsetRanges] + // offsetRanges.length = # of Kafka partitions being consumed + ... + } +
    +
    + directKafkaStream.foreachRDD( + new Function, Void>() { + @Override + public Void call(JavaPairRDD rdd) throws IOException { + OffsetRange[] offsetRanges = ((HasOffsetRanges)rdd).offsetRanges + // offsetRanges.length = # of Kafka partitions being consumed + ... + return null; + } + } + ); +
    +
    + + You can use this to update Zookeeper yourself if you want Zookeeper-based Kafka monitoring tools to show progress of the streaming application. + + Another thing to note is that since this approach does not use Receivers, the standard receiver-related (that is, [configurations](configuration.html) of the form `spark.streaming.receiver.*` ) will not apply to the input DStreams created by this approach (will apply to other input DStreams though). Instead, use the [configurations](configuration.html) `spark.streaming.kafka.*`. An important one is `spark.streaming.kafka.maxRatePerPartition` which is the maximum rate at which each Kafka partition will be read by this direct API. + +3. **Deploying:** Similar to the first approach, you can package `spark-streaming-kafka_{{site.SCALA_BINARY_VERSION}}` and its dependencies into the application JAR and the launch the application using `spark-submit`. Make sure `spark-core_{{site.SCALA_BINARY_VERSION}}` and `spark-streaming_{{site.SCALA_BINARY_VERSION}}` are marked as `provided` dependencies as those are already present in a Spark installation. \ No newline at end of file diff --git a/docs/streaming-programming-guide.md b/docs/streaming-programming-guide.md index 062ac2648db30..6d6229625f3f9 100644 --- a/docs/streaming-programming-guide.md +++ b/docs/streaming-programming-guide.md @@ -432,7 +432,7 @@ some of the common ones are as follows. For an up-to-date list, please refer to the -[Apache repository](http://search.maven.org/#search%7Cga%7C1%7Cg%3A%22org.apache.spark%22%20AND%20v%3A%22{{site.SPARK_VERSION_SHORT}}%22) +[Maven repository](http://search.maven.org/#search%7Cga%7C1%7Cg%3A%22org.apache.spark%22%20AND%20v%3A%22{{site.SPARK_VERSION_SHORT}}%22) for the full list of supported sources and artifacts. *** @@ -662,8 +662,7 @@ methods for creating DStreams from files and Akka actors as input sources. For simple text files, there is an easier method `streamingContext.textFileStream(dataDirectory)`. And file streams do not require running a receiver, hence does not require allocating cores. - Python API As of Spark 1.2, - `fileStream` is not available in the Python API, only `textFileStream` is available. + Python API `fileStream` is not available in the Python API, only `textFileStream` is available. - **Streams based on Custom Actors:** DStreams can be created with data streams received through Akka actors by using `streamingContext.actorStream(actorProps, actor-name)`. See the [Custom Receiver @@ -682,8 +681,9 @@ for Java, and [StreamingContext](api/python/pyspark.streaming.html#pyspark.strea ### Advanced Sources {:.no_toc} -Python API As of Spark 1.2, -these sources are not available in the Python API. + +Python API As of Spark 1.3, +out of these sources, *only* Kafka is available in the Python API. We will add more advanced sources in the Python API in future. This category of sources require interfacing with external non-Spark libraries, some of them with complex dependencies (e.g., Kafka and Flume). Hence, to minimize issues related to version conflicts @@ -723,6 +723,12 @@ and it in the classpath. Some of these advanced sources are as follows. +- **Kafka:** Spark Streaming {{site.SPARK_VERSION_SHORT}} is compatible with Kafka 0.8.1.1. See the [Kafka Integration Guide](streaming-kafka-integration.html) for more details. + +- **Flume:** Spark Streaming {{site.SPARK_VERSION_SHORT}} is compatible with Flume 1.4.0. See the [Flume Integration Guide](streaming-flume-integration.html) for more details. + +- **Kinesis:** See the [Kinesis Integration Guide](streaming-kinesis-integration.html) for more details. + - **Twitter:** Spark Streaming's TwitterUtils uses Twitter4j 3.0.3 to get the public stream of tweets using [Twitter's Streaming API](https://dev.twitter.com/docs/streaming-apis). Authentication information can be provided by any of the [methods](http://twitter4j.org/en/configuration.html) supported by @@ -732,17 +738,10 @@ Some of these advanced sources are as follows. ([TwitterPopularTags]({{site.SPARK_GITHUB_URL}}/blob/master/examples/src/main/scala/org/apache/spark/examples/streaming/TwitterPopularTags.scala) and [TwitterAlgebirdCMS]({{site.SPARK_GITHUB_URL}}/blob/master/examples/src/main/scala/org/apache/spark/examples/streaming/TwitterAlgebirdCMS.scala)). -- **Flume:** Spark Streaming {{site.SPARK_VERSION_SHORT}} can received data from Flume 1.4.0. See the [Flume Integration Guide](streaming-flume-integration.html) for more details. - -- **Kafka:** Spark Streaming {{site.SPARK_VERSION_SHORT}} can receive data from Kafka 0.8.0. See the [Kafka Integration Guide](streaming-kafka-integration.html) for more details. - -- **Kinesis:** See the [Kinesis Integration Guide](streaming-kinesis-integration.html) for more details. - ### Custom Sources {:.no_toc} -Python API As of Spark 1.2, -these sources are not available in the Python API. +Python API This is not yet supported in Python. Input DStreams can also be created out of custom data sources. All you have to do is implement an user-defined **receiver** (see next section to understand what that is) that can receive data from @@ -846,7 +845,7 @@ Some of the common ones are as follows. -The last two transformations are worth highlighting again. +A few of these transformations are worth discussing in more detail. #### UpdateStateByKey Operation {:.no_toc} @@ -997,7 +996,7 @@ In fact, you can also use [machine learning](mllib-guide.html) and #### Window Operations {:.no_toc} -Finally, Spark Streaming also provides *windowed computations*, which allow you to apply +Spark Streaming also provides *windowed computations*, which allow you to apply transformations over a sliding window of data. This following figure illustrates this sliding window. @@ -1120,6 +1119,100 @@ said two parameters - windowLength and slideInterval. +#### Join Operations +{:.no_toc} +Finally, its worth highlighting how easily you can perform different kinds of joins in Spark Streaming. + + +##### Stream-stream joins +{:.no_toc} +Streams can be very easily joined with other streams. + +
    +
    +{% highlight scala %} +val stream1: DStream[String, String] = ... +val stream2: DStream[String, String] = ... +val joinedStream = stream1.join(stream2) +{% endhighlight %} +
    +
    +{% highlight java %} +JavaPairDStream stream1 = ... +JavaPairDStream stream2 = ... +JavaPairDStream joinedStream = stream1.join(stream2); +{% endhighlight %} +
    +
    +{% highlight python %} +stream1 = ... +stream2 = ... +joinedStream = stream1.join(stream2) +{% endhighlight %} +
    +
    +Here, in each batch interval, the RDD generated by `stream1` will be joined with the RDD generated by `stream2`. You can also do `leftOuterJoin`, `rightOuterJoin`, `fullOuterJoin`. Furthermore, it is often very useful to do joins over windows of the streams. That is pretty easy as well. + +
    +
    +{% highlight scala %} +val windowedStream1 = stream1.window(Seconds(20)) +val windowedStream2 = stream2.window(Minutes(1)) +val joinedStream = windowedStream1.join(windowedStream2) +{% endhighlight %} +
    +
    +{% highlight java %} +JavaPairDStream windowedStream1 = stream1.window(Durations.seconds(20)); +JavaPairDStream windowedStream2 = stream2.window(Durations.minutes(1)); +JavaPairDStream joinedStream = windowedStream1.join(windowedStream2); +{% endhighlight %} +
    +
    +{% highlight python %} +windowedStream1 = stream1.window(20) +windowedStream2 = stream2.window(60) +joinedStream = windowedStream1.join(windowedStream2) +{% endhighlight %} +
    +
    + +##### Stream-dataset joins +{:.no_toc} +This has already been shown earlier while explain `DStream.transform` operation. Here is yet another example of joining a windowed stream with a dataset. + +
    +
    +{% highlight scala %} +val dataset: RDD[String, String] = ... +val windowedStream = stream.window(Seconds(20))... +val joinedStream = windowedStream.transform { rdd => rdd.join(dataset) } +{% endhighlight %} +
    +
    +{% highlight java %} +JavaPairRDD dataset = ... +JavaPairDStream windowedStream = stream.window(Durations.seconds(20)); +JavaPairDStream joinedStream = windowedStream.transform( + new Function>, JavaRDD>>() { + @Override + public JavaRDD> call(JavaRDD> rdd) { + return rdd.join(dataset); + } + } +); +{% endhighlight %} +
    +
    +{% highlight python %} +dataset = ... # some RDD +windowedStream = stream.window(20) +joinedStream = windowedStream.transform(lambda rdd: rdd.join(dataset)) +{% endhighlight %} +
    +
    + +In fact, you can also dynamically change the dataset you want to join against. The function provided to `transform` is evaluated every batch interval and therefore will use the current dataset that `dataset` reference points to. The complete list of DStream transformations is available in the API documentation. For the Scala API, see [DStream](api/scala/index.html#org.apache.spark.streaming.dstream.DStream) @@ -1327,6 +1420,178 @@ Note that the connections in the pool should be lazily created on demand and tim *** +## DataFrame and SQL Operations +You can easily use [DataFrames and SQL](sql-programming-guide.html) operations on streaming data. You have to create a SQLContext using the SparkContext that the StreamingContext is using. Furthermore this has to done such that it can be restarted on driver failures. This is done by creating a lazily instantiated singleton instance of SQLContext. This is shown in the following example. It modifies the earlier [word count example](#a-quick-example) to generate word counts using DataFrames and SQL. Each RDD is converted to a DataFrame, registered as a temporary table and then queried using SQL. + +
    +
    +{% highlight scala %} + +/** Lazily instantiated singleton instance of SQLContext */ +object SQLContextSingleton { + @transient private var instance: SQLContext = null + + // Instantiate SQLContext on demand + def getInstance(sparkContext: SparkContext): SQLContext = synchronized { + if (instance == null) { + instance = new SQLContext(sparkContext) + } + instance + } +} + +... + +/** Case class for converting RDD to DataFrame */ +case class Row(word: String) + +... + +/** DataFrame operations inside your streaming program */ + +val words: DStream[String] = ... + +words.foreachRDD { rdd => + + // Get the singleton instance of SQLContext + val sqlContext = SQLContextSingleton.getInstance(rdd.sparkContext) + import sqlContext.implicits._ + + // Convert RDD[String] to RDD[case class] to DataFrame + val wordsDataFrame = rdd.map(w => Row(w)).toDF() + + // Register as table + wordsDataFrame.registerTempTable("words") + + // Do word count on DataFrame using SQL and print it + val wordCountsDataFrame = + sqlContext.sql("select word, count(*) as total from words group by word") + wordCountsDataFrame.show() +} + +{% endhighlight %} + +See the full [source code]({{site.SPARK_GITHUB_URL}}/blob/master/examples/src/main/scala/org/apache/spark/examples/streaming/SqlNetworkWordCount.scala). +
    +
    +{% highlight java %} + +/** Lazily instantiated singleton instance of SQLContext */ +class JavaSQLContextSingleton { + static private transient SQLContext instance = null; + static public SQLContext getInstance(SparkContext sparkContext) { + if (instance == null) { + instance = new SQLContext(sparkContext); + } + return instance; + } +} + +... + +/** Java Bean class for converting RDD to DataFrame */ +public class JavaRow implements java.io.Serializable { + private String word; + + public String getWord() { + return word; + } + + public void setWord(String word) { + this.word = word; + } +} + +... + +/** DataFrame operations inside your streaming program */ + +JavaDStream words = ... + +words.foreachRDD( + new Function2, Time, Void>() { + @Override + public Void call(JavaRDD rdd, Time time) { + SQLContext sqlContext = JavaSQLContextSingleton.getInstance(rdd.context()); + + // Convert RDD[String] to RDD[case class] to DataFrame + JavaRDD rowRDD = rdd.map(new Function() { + public JavaRow call(String word) { + JavaRow record = new JavaRow(); + record.setWord(word); + return record; + } + }); + DataFrame wordsDataFrame = sqlContext.createDataFrame(rowRDD, JavaRow.class); + + // Register as table + wordsDataFrame.registerTempTable("words"); + + // Do word count on table using SQL and print it + DataFrame wordCountsDataFrame = + sqlContext.sql("select word, count(*) as total from words group by word"); + wordCountsDataFrame.show(); + return null; + } + } +); +{% endhighlight %} + +See the full [source code]({{site.SPARK_GITHUB_URL}}/blob/master/examples/src/main/java/org/apache/spark/examples/streaming/JavaSqlNetworkWordCount.java). +
    +
    +{% highlight python %} + +# Lazily instantiated global instance of SQLContext +def getSqlContextInstance(sparkContext): + if ('sqlContextSingletonInstance' not in globals()): + globals()['sqlContextSingletonInstance'] = SQLContext(sparkContext) + return globals()['sqlContextSingletonInstance'] + +... + +# DataFrame operations inside your streaming program + +words = ... # DStream of strings + +def process(time, rdd): + print "========= %s =========" % str(time) + try: + # Get the singleton instance of SQLContext + sqlContext = getSqlContextInstance(rdd.context) + + # Convert RDD[String] to RDD[Row] to DataFrame + rowRdd = rdd.map(lambda w: Row(word=w)) + wordsDataFrame = sqlContext.createDataFrame(rowRdd) + + # Register as table + wordsDataFrame.registerTempTable("words") + + # Do word count on table using SQL and print it + wordCountsDataFrame = sqlContext.sql("select word, count(*) as total from words group by word") + wordCountsDataFrame.show() + except: + pass + +words.foreachRDD(process) +{% endhighlight %} + +See the full [source code]({{site.SPARK_GITHUB_URL}}/blob/master/examples/src/main/python/streaming/sql_network_wordcount.py). + +
    +
    + +You can also run SQL queries on tables defined on streaming data from a different thread (that is, asynchronous to the running StreamingContext). Just make sure that you set the StreamingContext to remember sufficient amount of streaming data such that query can run. Otherwise the StreamingContext, which is unaware of the any asynchronous SQL queries, will delete off old streaming data before the query can complete. For example, if you want to query the last batch, but your query can take 5 minutes to run, then call `streamingContext.remember(Minutes(5))` (in Scala, or equivalent in other languages). + +See the [DataFrames and SQL](sql-programming-guide.html) guide to learn more about DataFrames. + +*** + +## MLlib Operations +You can also easily use machine learning algorithms provided by [MLlib](mllib-guide.html). First of all, there are streaming machine learning algorithms (e.g. (Streaming Linear Regression](mllib-linear-methods.html#streaming-linear-regression), [Streaming KMeans](mllib-clustering.html#streaming-k-means), etc.) which can simultaneously learn from the streaming data as well as apply the model on the streaming data. Beyond these, for a much larger class of machine learning algorithms, you can learn a learning model offline (i.e. using historical data) and then apply the model online on streaming data. See the [MLlib](mllib-guide.html) guide for more details. + +*** + ## Caching / Persistence Similar to RDDs, DStreams also allow developers to persist the stream's data in memory. That is, using `persist()` method on a DStream will automatically persist every RDD of that DStream in @@ -1580,9 +1845,8 @@ To run a Spark Streaming applications, you need to have the following. + *Mesos* - [Marathon](https://github.com/mesosphere/marathon) has been used to achieve this with Mesos. - -- *[Experimental in Spark 1.2] Configuring write ahead logs* - In Spark 1.2, - we have introduced a new experimental feature of write ahead logs for achieving strong +- *[Since Spark 1.2] Configuring write ahead logs* - Since Spark 1.2, + we have introduced _write ahead logs_ for achieving strong fault-tolerance guarantees. If enabled, all the data received from a receiver gets written into a write ahead log in the configuration checkpoint directory. This prevents data loss on driver recovery, thus ensuring zero data loss (discussed in detail in the @@ -1668,7 +1932,7 @@ improve the performance of you application. At a high level, you need to conside 2. Setting the right batch size such that the batches of data can be processed as fast as they are received (that is, data processing keeps up with the data ingestion). -## Reducing the Processing Time of each Batch +## Reducing the Batch Processing Times There are a number of optimizations that can be done in Spark to minimize the processing time of each batch. These have been discussed in detail in [Tuning Guide](tuning.html). This section highlights some of the most important ones. @@ -1740,16 +2004,15 @@ documentation), or set the `spark.default.parallelism` ### Data Serialization {:.no_toc} -The overhead of data serialization can be significant, especially when sub-second batch sizes are - to be achieved. There are two aspects to it. +The overheads of data serialization can be reduce by tuning the serialization formats. In case of streaming, there are two types of data that are being serialized. + +* **Input data**: By default, the input data received through Receivers is stored in the executors' memory with [StorageLevel.MEMORY_AND_DISK_SER_2](api/scala/index.html#org.apache.spark.storage.StorageLevel$). That is, the data is serialized into bytes to reduce GC overheads, and replicated for tolerating executor failures. Also, the data is kept first in memory, and spilled over to disk only if the memory is unsufficient to hold all the input data necessary for the streaming computation. This serialization obviously has overheads -- the receiver must deserialize the received data and re-serialize it using Spark's serialization format. -* **Serialization of RDD data in Spark**: Please refer to the detailed discussion on data - serialization in the [Tuning Guide](tuning.html). However, note that unlike Spark, by default - RDDs are persisted as serialized byte arrays to minimize pauses related to GC. +* **Persisted RDDs generated by Streaming Operations**: RDDs generated by streaming computations may be persisted in memory. For example, window operation persist data in memory as they would be processed multiple times. However, unlike Spark, by default RDDs are persisted with [StorageLevel.MEMORY_ONLY_SER](api/scala/index.html#org.apache.spark.storage.StorageLevel$) (i.e. serialized) to minimize GC overheads. -* **Serialization of input data**: To ingest external data into Spark, data received as bytes - (say, from the network) needs to deserialized from bytes and re-serialized into Spark's - serialization format. Hence, the deserialization overhead of input data may be a bottleneck. +In both cases, using Kryo serialization can reduce both CPU and memory overheads. See the [Spark Tuning Guide](tuning.html#data-serialization)) for more details. Consider registering custom classes, and disabling object reference tracking for Kryo (see Kryo-related configurations in the [Configuration Guide](configuration.html#compression-and-serialization)). + +In specific cases where the amount of data that needs to be retained for the streaming application is not large, it may be feasible to persist data (both types) as deserialized objects without incurring excessive GC overheads. For example, if you are using batch intervals of few seconds and no window operations, then you can try disabling serialization in persisted data by explicitly setting the storage level accordingly. This would reduce the CPU overheads due to serialization, potentially improving performance without too much GC overheads. ### Task Launching Overheads {:.no_toc} @@ -1769,7 +2032,7 @@ thus allowing sub-second batch size to be viable. *** -## Setting the Right Batch Size +## Setting the Right Batch Interval For a Spark Streaming application running on a cluster to be stable, the system should be able to process data as fast as it is being received. In other words, batches of data should be processed as fast as they are being generated. Whether this is true for an application can be found by @@ -1801,40 +2064,40 @@ temporary data rate increases maybe fine as long as the delay reduces back to a ## Memory Tuning Tuning the memory usage and GC behavior of Spark applications have been discussed in great detail -in the [Tuning Guide](tuning.html). It is recommended that you read that. In this section, -we highlight a few customizations that are strongly recommended to minimize GC related pauses -in Spark Streaming applications and achieving more consistent batch processing times. - -* **Default persistence level of DStreams**: Unlike RDDs, the default persistence level of DStreams -serializes the data in memory (that is, -[StorageLevel.MEMORY_ONLY_SER](api/scala/index.html#org.apache.spark.storage.StorageLevel$) for -DStream compared to -[StorageLevel.MEMORY_ONLY](api/scala/index.html#org.apache.spark.storage.StorageLevel$) for RDDs). -Even though keeping the data serialized incurs higher serialization/deserialization overheads, -it significantly reduces GC pauses. - -* **Clearing persistent RDDs**: By default, all persistent RDDs generated by Spark Streaming will - be cleared from memory based on Spark's built-in policy (LRU). If `spark.cleaner.ttl` is set, - then persistent RDDs that are older than that value are periodically cleared. As mentioned - [earlier](#operation), this needs to be careful set based on operations used in the Spark - Streaming program. However, a smarter unpersisting of RDDs can be enabled by setting the - [configuration property](configuration.html#spark-properties) `spark.streaming.unpersist` to - `true`. This makes the system to figure out which RDDs are not necessary to be kept around and - unpersists them. This is likely to reduce - the RDD memory usage of Spark, potentially improving GC behavior as well. - -* **Concurrent garbage collector**: Using the concurrent mark-and-sweep GC further -minimizes the variability of GC pauses. Even though concurrent GC is known to reduce the +in the [Tuning Guide](tuning.html#memory-tuning). It is strongly recommended that you read that. In this section, we discuss a few tuning parameters specifically in the context of Spark Streaming applications. + +The amount of cluster memory required by a Spark Streaming application depends heavily on the type of transformations used. For example, if you want to use a window operation on last 10 minutes of data, then your cluster should have sufficient memory to hold 10 minutes of worth of data in memory. Or if you want to use `updateStateByKey` with a large number of keys, then the necessary memory will be high. On the contrary, if you want to do a simple map-filter-store operation, then necessary memory will be low. + +In general, since the data received through receivers are stored with StorageLevel.MEMORY_AND_DISK_SER_2, the data that does not fit in memory will spill over to the disk. This may reduce the performance of the streaming application, and hence it is advised to provide sufficient memory as required by your streaming application. Its best to try and see the memory usage on a small scale and estimate accordingly. + +Another aspect of memory tuning is garbage collection. For a streaming application that require low latency, it is undesirable to have large pauses caused by JVM Garbage Collection. + +There are a few parameters that can help you tune the memory usage and GC overheads. + +* **Persistence Level of DStreams**: As mentioned earlier in the [Data Serialization](#data-serialization) section, the input data and RDDs are by default persisted as serialized bytes. This reduces both, the memory usage and GC overheads, compared to deserialized persistence. Enabling Kryo serialization further reduces serialized sizes and memory usage. Further reduction in memory usage can be achieved with compression (see the Spark configuration `spark.rdd.compress`), at the cost of CPU time. + +* **Clearing old data**: By default, all input data and persisted RDDs generated by DStream transformations are automatically cleared. Spark Streaming decides when to clear the data based on the transformations that are used. For example, if you are using window operation of 10 minutes, then Spark Streaming will keep around last 10 minutes of data, and actively throw away older data. +Data can be retained for longer duration (e.g. interactively querying older data) by setting `streamingContext.remember`. + +* **CMS Garbage Collector**: Use of the concurrent mark-and-sweep GC is strongly recommended for keeping GC-related pauses consistently low. Even though concurrent GC is known to reduce the overall processing throughput of the system, its use is still recommended to achieve more -consistent batch processing times. +consistent batch processing times. Make sure you set the CMS GC on both the driver (using `--driver-java-options` in `spark-submit`) and the executors (using [Spark configuration](configuration.html#runtime-environment) `spark.executor.extraJavaOptions`). + +* **Other tips**: To further reduce GC overheads, here are some more tips to try. + - Use Tachyon for off-heap storage of persisted RDDs. See more detail in the [Spark Programming Guide](programming-guide.html#rdd-persistence). + - Use more executors with smaller heap sizes. This will reduce the GC pressure within each JVM heap. + *************************************************************************************************** *************************************************************************************************** # Fault-tolerance Semantics In this section, we will discuss the behavior of Spark Streaming applications in the event -of node failures. To understand this, let us remember the basic fault-tolerance semantics of -Spark's RDDs. +of failures. + +## Background +{:.no_toc} +To understand the semantics provided by Spark Streaming, let us remember the basic fault-tolerance semantics of Spark's RDDs. 1. An RDD is an immutable, deterministically re-computable, distributed dataset. Each RDD remembers the lineage of deterministic operations that were used on a fault-tolerant input @@ -1868,13 +2131,43 @@ Furthermore, there are two kinds of failures that we should be concerned about: With this basic knowledge, let us understand the fault-tolerance semantics of Spark Streaming. -## Semantics with files as input source +## Definitions +{:.no_toc} +The semantics of streaming systems are often captured in terms of how many times each record can be processed by the system. There are three types of guarantees that a system can provide under all possible operating conditions (despite failures, etc.) + +1. *At most once*: Each record will be either processed once or not processed at all. +2. *At least once*: Each record will be processed one or more times. This is stronger than *at-most once* as it ensure that no data will be lost. But there may be duplicates. +3. *Exactly once*: Each record will be processed exactly once - no data will be lost and no data will be processed multiple times. This is obviously the strongest guarantee of the three. + +## Basic Semantics +{:.no_toc} +In any stream processing system, broadly speaking, there are three steps in processing the data. + +1. *Receiving the data*: The data is received from sources using Receivers or otherwise. + +1. *Transforming the data*: The data received data is transformed using DStream and RDD transformations. + +1. *Pushing out the data*: The final transformed data is pushed out to external systems like file systems, databases, dashboards, etc. + +If a streaming application has to achieve end-to-end exactly-once guarantees, then each step has to provide exactly-once guarantee. That is, each record must be received exactly once, transformed exactly once, and pushed to downstream systems exactly once. Let's understand the semantics of these steps in the context of Spark Streaming. + +1. *Receiving the data*: Different input sources provided different guarantees. This is discussed in detail in the next subsection. + +1. *Transforming the data*: All data that has been received will be processed _exactly once_, thanks to the guarantees that RDDs provide. Even if there are failures, as long as the received input data is accessible, the final transformed RDDs will always have the same contents. + +1. *Pushing out the data*: Output operations by default ensure _at-least once_ semantics because it depends on the type of output operation (idempotent, or not) and the semantics of the downstream system (supports transactions or not). But users can implement their own transaction mechanisms to achieve _exactly-once_ semantics. This is discussed in more details later in the section. + +## Semantics of Received Data +{:.no_toc} +Different input sources provide different guarantees, ranging from _at-least once_ to _exactly once_. Read for more details. + +### With Files {:.no_toc} If all of the input data is already present in a fault-tolerant files system like HDFS, Spark Streaming can always recover from any failure and process all the data. This gives *exactly-once* semantics, that all the data will be processed exactly once no matter what fails. -## Semantics with input sources based on receivers +### With Receiver-based Sources {:.no_toc} For input sources based on receivers, the fault-tolerance semantics depend on both the failure scenario and the type of receiver. @@ -1893,10 +2186,9 @@ receivers, data received but not replicated can get lost. If the driver node fai then besides these losses, all the past data that was received and replicated in memory will be lost. This will affect the results of the stateful transformations. -To avoid this loss of past received data, Spark 1.2 introduces an experimental feature of _write +To avoid this loss of past received data, Spark 1.2 introduced _write ahead logs_ which saves the received data to fault-tolerant storage. With the [write ahead logs -enabled](#deploying-applications) and reliable receivers, there is zero data loss and -exactly-once semantics. +enabled](#deploying-applications) and reliable receivers, there is zero data loss. In terms of semantics, it provides at-least once guarantee. The following table summarizes the semantics under failures: @@ -1908,23 +2200,30 @@ The following table summarizes the semantics under failures: - Spark 1.1 or earlier, or
    - Spark 1.2 without write ahead log + Spark 1.1 or earlier, OR
    + Spark 1.2 or later without write ahead logs Buffered data lost with unreliable receivers
    - Zero data loss with reliable receivers and files
    + Zero data loss with reliable receivers
    + At-least once semantics Buffered data lost with unreliable receivers
    Past data lost with all receivers
    - Zero data loss with files - + Undefined semantics + - Spark 1.2 with write ahead log - Zero data loss with reliable receivers and files - Zero data loss with reliable receivers and files + Spark 1.2 or later with write ahead logs + + Zero data loss with reliable receivers
    + At-least once semantics + + + Zero data loss with reliable receivers and files
    + At-least once semantics + @@ -1933,17 +2232,24 @@ The following table summarizes the semantics under failures: +### With Kafka Direct API +{:.no_toc} +In Spark 1.3, we have introduced a new Kafka Direct API, which can ensure that all the Kafka data is received by Spark Streaming exactly once. Along with this, if you implement exactly-once output operation, you can achieve end-to-end exactly-once guarantees. This approach (experimental as of Spark 1.3) is further discussed in the [Kafka Integration Guide](streaming-kafka-integration.html). + ## Semantics of output operations {:.no_toc} -Since all data is modeled as RDDs with their lineage of deterministic operations, any recomputation - always leads to the same result. As a result, all DStream transformations are guaranteed to have - _exactly-once_ semantics. That is, the final transformed result will be same even if there were - was a worker node failure. However, output operations (like `foreachRDD`) have _at-least once_ - semantics, that is, the transformed data may get written to an external entity more than once in - the event of a worker failure. While this is acceptable for saving to HDFS using the - `saveAs***Files` operations (as the file will simply get over-written by the same data), - additional transactions-like mechanisms may be necessary to achieve exactly-once semantics - for output operations. +Output operations (like `foreachRDD`) have _at-least once_ semantics, that is, +the transformed data may get written to an external entity more than once in +the event of a worker failure. While this is acceptable for saving to file systems using the +`saveAs***Files` operations (as the file will simply get overwritten with the same data), +additional effort may be necessary to achieve exactly-once semantics. There are two approaches. + +- *Idempotent updates*: Multiple attempts always write the same data. For example, `saveAs***Files` always writes the same data to the generated files. + +- *Transactional updates*: All updates are made transactionally so that updates are made exactly once atomically. One way to do this would be the following. + + - Use the batch time (available in `foreachRDD`) and the partition index of the transformed RDD to create an identifier. This identifier uniquely identifies a blob data in the streaming application. + - Update external system with this blob transactionally (that is, exactly once, atomically) using the identifier. That is, if the identifier is not already committed, commit the partition data and the identifier atomically. Else if this was already committed, skip the update. *************************************************************************************************** @@ -2001,7 +2307,11 @@ package and renamed for better clarity. *************************************************************************************************** # Where to Go from Here - +* Additional guides + - [Kafka Integration Guide](streaming-kafka-integration.html) + - [Flume Integration Guide](streaming-flume-integration.html) + - [Kinesis Integration Guide](streaming-kinesis-integration.html) + - [Custom Receiver Guide](streaming-custom-receivers.html) * API documentation - Scala docs * [StreamingContext](api/scala/index.html#org.apache.spark.streaming.StreamingContext) and @@ -2023,8 +2333,8 @@ package and renamed for better clarity. [ZeroMQUtils](api/java/index.html?org/apache/spark/streaming/zeromq/ZeroMQUtils.html), and [MQTTUtils](api/java/index.html?org/apache/spark/streaming/mqtt/MQTTUtils.html) - Python docs - * [StreamingContext](api/python/pyspark.streaming.html#pyspark.streaming.StreamingContext) - * [DStream](api/python/pyspark.streaming.html#pyspark.streaming.DStream) + * [StreamingContext](api/python/pyspark.streaming.html#pyspark.streaming.StreamingContext) and [DStream](api/python/pyspark.streaming.html#pyspark.streaming.DStream) + * [KafkaUtils](api/python/pyspark.streaming.html#pyspark.streaming.kafka.KafkaUtils) * More examples in [Scala]({{site.SPARK_GITHUB_URL}}/tree/master/examples/src/main/scala/org/apache/spark/examples/streaming) and [Java]({{site.SPARK_GITHUB_URL}}/tree/master/examples/src/main/java/org/apache/spark/examples/streaming) From e921a665c56950c03155f4b47500627265a4ba8e Mon Sep 17 00:00:00 2001 From: Patrick Wendell Date: Wed, 11 Mar 2015 22:24:08 -0700 Subject: [PATCH 037/122] BUILD: Adding more known contributor names --- dev/create-release/known_translations | 34 +++++++++++++++++++++++++++ 1 file changed, 34 insertions(+) diff --git a/dev/create-release/known_translations b/dev/create-release/known_translations index b74e4ee8a330b..0a599b5a65549 100644 --- a/dev/create-release/known_translations +++ b/dev/create-release/known_translations @@ -57,3 +57,37 @@ watermen - Yadong Qi witgo - Guoqiang Li xinyunh - Xinyun Huang zsxwing - Shixiong Zhu +Bilna - Bilna P +DoingDone9 - Doing Done +Earne - Ernest +FlytxtRnD - Meethu Mathew +GenTang - Gen TANG +JoshRosen - Josh Rosen +MechCoder - Manoj Kumar +OopsOutOfMemory - Sheng Li +Peishen-Jia - Peishen Jia +SaintBacchus - Huang Zhaowei +azagrebin - Andrey Zagrebin +bzz - Alexander Bezzubov +fjiang6 - Fan Jiang +gasparms - Gaspar Munoz +guowei2 - Guo Wei +hhbyyh - Yuhao Yang +hseagle - Peng Xu +javadba - Stephen Boesch +jbencook - Ben Cook +kul - Kuldeep +ligangty - Gang Li +marsishandsome - Liangliang Gu +medale - Markus Dale +nemccarthy - Nathan McCarthy +nxwhite-str - Nate Crosswhite +seayi - Xiaohua Yi +tianyi - Yi Tian +uncleGen - Uncle Gen +viper-kun - Xu Kun +x1- - Yuri Saito +zapletal-martin - Martin Zapletal +zuxqoj - Shekhar Bansal +mingyukim - Mingyu Kim +sigmoidanalytics - Mayur Rustagi From 25b71d8c15572f0f2b951c827c169f8c65f726ad Mon Sep 17 00:00:00 2001 From: Volodymyr Lyubinets Date: Thu, 12 Mar 2015 00:55:26 -0700 Subject: [PATCH 038/122] [SPARK-6296] [SQL] Added equals to Column Author: Volodymyr Lyubinets Closes #4988 from vlyubin/columncomp and squashes the following commits: 92d7c8f [Volodymyr Lyubinets] Added equals to Column --- sql/core/src/main/scala/org/apache/spark/sql/Column.scala | 7 +++++++ .../scala/org/apache/spark/sql/ColumnExpressionSuite.scala | 5 +++++ 2 files changed, 12 insertions(+) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/Column.scala b/sql/core/src/main/scala/org/apache/spark/sql/Column.scala index a2cc9a9b93eb8..908c78a4d3f10 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/Column.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/Column.scala @@ -59,6 +59,13 @@ class Column(protected[sql] val expr: Expression) { override def toString: String = expr.prettyString + override def equals(that: Any) = that match { + case that: Column => that.expr.equals(this.expr) + case _ => false + } + + override def hashCode: Int = this.expr.hashCode + /** * Unary minus, i.e. negate the expression. * {{{ diff --git a/sql/core/src/test/scala/org/apache/spark/sql/ColumnExpressionSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/ColumnExpressionSuite.scala index 37c02aaa5460b..3036fbc05d021 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/ColumnExpressionSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/ColumnExpressionSuite.scala @@ -313,4 +313,9 @@ class ColumnExpressionSuite extends QueryTest { test("lift alias out of cast") { assert(col("1234").as("name").cast("int").expr === col("1234").cast("int").as("name").expr) } + + test("columns can be compared") { + assert('key.desc == 'key.desc) + assert('key.desc != 'key.asc) + } } From 712679a7b447346a365b38574d7a86d56a93f767 Mon Sep 17 00:00:00 2001 From: Davies Liu Date: Thu, 12 Mar 2015 01:34:38 -0700 Subject: [PATCH 039/122] [SPARK-6294] fix hang when call take() in JVM on PythonRDD The Thread.interrupt() can not terminate the thread in some cases, so we should not wait for the writerThread of PythonRDD. This PR also ignore some exception during clean up. cc JoshRosen mengxr Author: Davies Liu Closes #4987 from davies/fix_take and squashes the following commits: 4488f1a [Davies Liu] fix hang when call take() in JVM on PythonRDD --- .../scala/org/apache/spark/api/python/PythonRDD.scala | 9 ++++++--- python/pyspark/daemon.py | 5 ++++- python/pyspark/tests.py | 5 +++++ 3 files changed, 15 insertions(+), 4 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala b/core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala index 8d4a53b4ca9b0..4c71b69069eb3 100644 --- a/core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala +++ b/core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala @@ -76,7 +76,6 @@ private[spark] class PythonRDD( context.addTaskCompletionListener { context => writerThread.shutdownOnTaskCompletion() - writerThread.join() if (!reuse_worker || !released) { try { worker.close() @@ -248,13 +247,17 @@ private[spark] class PythonRDD( } catch { case e: Exception if context.isCompleted || context.isInterrupted => logDebug("Exception thrown after task completion (likely due to cleanup)", e) - Utils.tryLog(worker.shutdownOutput()) + if (!worker.isClosed) { + Utils.tryLog(worker.shutdownOutput()) + } case e: Exception => // We must avoid throwing exceptions here, because the thread uncaught exception handler // will kill the whole executor (see org.apache.spark.executor.Executor). _exception = e - Utils.tryLog(worker.shutdownOutput()) + if (!worker.isClosed) { + Utils.tryLog(worker.shutdownOutput()) + } } finally { // Release memory used by this thread for shuffles env.shuffleMemoryManager.releaseMemoryForThisThread() diff --git a/python/pyspark/daemon.py b/python/pyspark/daemon.py index f09587f211708..93885985fe377 100644 --- a/python/pyspark/daemon.py +++ b/python/pyspark/daemon.py @@ -61,7 +61,10 @@ def worker(sock): except SystemExit as exc: exit_code = compute_real_exit_code(exc.code) finally: - outfile.flush() + try: + outfile.flush() + except Exception: + pass return exit_code diff --git a/python/pyspark/tests.py b/python/pyspark/tests.py index 06ba2b461d53e..dd8d3b1c53733 100644 --- a/python/pyspark/tests.py +++ b/python/pyspark/tests.py @@ -782,6 +782,11 @@ def test_narrow_dependency_in_join(self): jobId = tracker.getJobIdsForGroup("test4")[0] self.assertEqual(3, len(tracker.getJobInfo(jobId).stageIds)) + # Regression test for SPARK-6294 + def test_take_on_jrdd(self): + rdd = self.sc.parallelize(range(1 << 20)).map(lambda x: str(x)) + rdd._jrdd.first() + class ProfilerTests(PySparkTestCase): From 0cba802adf15f5ab8da24dd1e8a5e7214cc4e148 Mon Sep 17 00:00:00 2001 From: Xiangrui Meng Date: Thu, 12 Mar 2015 01:39:04 -0700 Subject: [PATCH 040/122] [SPARK-5814][MLLIB][GRAPHX] Remove JBLAS from runtime The issue is discussed in https://issues.apache.org/jira/browse/SPARK-5669. Replacing all JBLAS usage by netlib-java gives us a simpler dependency tree and less license issues to worry about. I didn't touch the test scope in this PR. The user guide is not modified to avoid merge conflicts with branch-1.3. srowen ankurdave pwendell Author: Xiangrui Meng Closes #4699 from mengxr/SPARK-5814 and squashes the following commits: 48635c6 [Xiangrui Meng] move netlib-java version to parent pom ca21c74 [Xiangrui Meng] remove jblas from ml-guide 5f7767a [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-5814 c5c4183 [Xiangrui Meng] merge master 0f20cad [Xiangrui Meng] add mima excludes e53e9f4 [Xiangrui Meng] remove jblas from mllib runtime ceaa14d [Xiangrui Meng] replace jblas by netlib-java in graphx fa7c2ca [Xiangrui Meng] move jblas to test scope --- assembly/pom.xml | 10 -- docs/mllib-guide.md | 5 - graphx/pom.xml | 11 ++- .../apache/spark/graphx/lib/SVDPlusPlus.scala | 96 ++++++++++++------- .../spark/graphx/lib/SVDPlusPlusSuite.scala | 6 +- mllib/pom.xml | 3 +- .../apache/spark/ml/recommendation/ALS.scala | 14 ++- .../spark/mllib/optimization/NNLS.scala | 86 +++++++++-------- .../MatrixFactorizationModel.scala | 15 ++- .../mllib/util/LinearDataGenerator.scala | 9 +- .../spark/mllib/util/MFDataGenerator.scala | 26 ++--- .../spark/mllib/util/SVMDataGenerator.scala | 7 +- .../spark/mllib/optimization/NNLSSuite.scala | 6 +- .../spark/mllib/stat/KernelDensitySuite.scala | 4 +- pom.xml | 1 + project/MimaExcludes.scala | 28 ++++++ 16 files changed, 183 insertions(+), 144 deletions(-) diff --git a/assembly/pom.xml b/assembly/pom.xml index cbf5b6c4aa8df..d3bb4bde0c412 100644 --- a/assembly/pom.xml +++ b/assembly/pom.xml @@ -114,16 +114,6 @@ META-INF/*.RSA - - - org.jblas:jblas - - - lib/static/Linux/i386/** - lib/static/Mac OS X/** - lib/static/Windows/** - - diff --git a/docs/mllib-guide.md b/docs/mllib-guide.md index 4c7a7d9115ca1..598374f66df5e 100644 --- a/docs/mllib-guide.md +++ b/docs/mllib-guide.md @@ -80,11 +80,6 @@ include `netlib-java`'s native proxies by default. To configure [netlib-java](https://github.com/fommil/netlib-java) documentation for your platform's additional installation instructions. -MLlib also uses [jblas](https://github.com/mikiobraun/jblas) which -will require you to install the -[gfortran runtime library](https://github.com/mikiobraun/jblas/wiki/Missing-Libraries) -if it is not already present on your nodes. - To use MLlib in Python, you will need [NumPy](http://www.numpy.org) version 1.4 or newer. diff --git a/graphx/pom.xml b/graphx/pom.xml index 57e338c03ecf9..c0d534e185d7f 100644 --- a/graphx/pom.xml +++ b/graphx/pom.xml @@ -45,9 +45,14 @@ guava - org.jblas - jblas - ${jblas.version} + com.github.fommil.netlib + core + ${netlib.java.version} + + + net.sourceforge.f2j + arpack_combined_all + 0.1 org.scalacheck diff --git a/graphx/src/main/scala/org/apache/spark/graphx/lib/SVDPlusPlus.scala b/graphx/src/main/scala/org/apache/spark/graphx/lib/SVDPlusPlus.scala index 3e4157a63fd1c..1a7178b82e3af 100644 --- a/graphx/src/main/scala/org/apache/spark/graphx/lib/SVDPlusPlus.scala +++ b/graphx/src/main/scala/org/apache/spark/graphx/lib/SVDPlusPlus.scala @@ -18,7 +18,9 @@ package org.apache.spark.graphx.lib import scala.util.Random -import org.jblas.DoubleMatrix + +import com.github.fommil.netlib.BLAS.{getInstance => blas} + import org.apache.spark.rdd._ import org.apache.spark.graphx._ @@ -53,7 +55,7 @@ object SVDPlusPlus { * a Multifaceted Collaborative Filtering Model", * available at [[http://public.research.att.com/~volinsky/netflix/kdd08koren.pdf]]. * - * The prediction rule is rui = u + bu + bi + qi*(pu + |N(u)|^(-0.5)*sum(y)), + * The prediction rule is rui = u + bu + bi + qi*(pu + |N(u)|^^-0.5^^*sum(y)), * see the details on page 6. * * @param edges edges for constructing the graph @@ -66,13 +68,10 @@ object SVDPlusPlus { : (Graph[(Array[Double], Array[Double], Double, Double), Double], Double) = { // Generate default vertex attribute - def defaultF(rank: Int): (DoubleMatrix, DoubleMatrix, Double, Double) = { - val v1 = new DoubleMatrix(rank) - val v2 = new DoubleMatrix(rank) - for (i <- 0 until rank) { - v1.put(i, Random.nextDouble()) - v2.put(i, Random.nextDouble()) - } + def defaultF(rank: Int): (Array[Double], Array[Double], Double, Double) = { + // TODO: use a fixed random seed + val v1 = Array.fill(rank)(Random.nextDouble()) + val v2 = Array.fill(rank)(Random.nextDouble()) (v1, v2, 0.0, 0.0) } @@ -92,7 +91,7 @@ object SVDPlusPlus { (g1, g2) => (g1._1 + g2._1, g1._2 + g2._2)) val gJoinT0 = g.outerJoinVertices(t0) { - (vid: VertexId, vd: (DoubleMatrix, DoubleMatrix, Double, Double), + (vid: VertexId, vd: (Array[Double], Array[Double], Double, Double), msg: Option[(Long, Double)]) => (vd._1, vd._2, msg.get._2 / msg.get._1, 1.0 / scala.math.sqrt(msg.get._1)) }.cache() @@ -102,24 +101,28 @@ object SVDPlusPlus { def sendMsgTrainF(conf: Conf, u: Double) (ctx: EdgeContext[ - (DoubleMatrix, DoubleMatrix, Double, Double), + (Array[Double], Array[Double], Double, Double), Double, - (DoubleMatrix, DoubleMatrix, Double)]) { + (Array[Double], Array[Double], Double)]) { val (usr, itm) = (ctx.srcAttr, ctx.dstAttr) val (p, q) = (usr._1, itm._1) - var pred = u + usr._3 + itm._3 + q.dot(usr._2) + val rank = p.length + var pred = u + usr._3 + itm._3 + blas.ddot(rank, q, 1, usr._2, 1) pred = math.max(pred, conf.minVal) pred = math.min(pred, conf.maxVal) val err = ctx.attr - pred - val updateP = q.mul(err) - .subColumnVector(p.mul(conf.gamma7)) - .mul(conf.gamma2) - val updateQ = usr._2.mul(err) - .subColumnVector(q.mul(conf.gamma7)) - .mul(conf.gamma2) - val updateY = q.mul(err * usr._4) - .subColumnVector(itm._2.mul(conf.gamma7)) - .mul(conf.gamma2) + // updateP = (err * q - conf.gamma7 * p) * conf.gamma2 + val updateP = q.clone() + blas.dscal(rank, err * conf.gamma2, updateP, 1) + blas.daxpy(rank, -conf.gamma7 * conf.gamma2, p, 1, updateP, 1) + // updateQ = (err * usr._2 - conf.gamma7 * q) * conf.gamma2 + val updateQ = usr._2.clone() + blas.dscal(rank, err * conf.gamma2, updateQ, 1) + blas.daxpy(rank, -conf.gamma7 * conf.gamma2, q, 1, updateQ, 1) + // updateY = (err * usr._4 * q - conf.gamma7 * itm._2) * conf.gamma2 + val updateY = q.clone() + blas.dscal(rank, err * usr._4 * conf.gamma2, updateY, 1) + blas.daxpy(rank, -conf.gamma7 * conf.gamma2, itm._2, 1, updateY, 1) ctx.sendToSrc((updateP, updateY, (err - conf.gamma6 * usr._3) * conf.gamma1)) ctx.sendToDst((updateQ, updateY, (err - conf.gamma6 * itm._3) * conf.gamma1)) } @@ -127,14 +130,23 @@ object SVDPlusPlus { for (i <- 0 until conf.maxIters) { // Phase 1, calculate pu + |N(u)|^(-0.5)*sum(y) for user nodes g.cache() - val t1 = g.aggregateMessages[DoubleMatrix]( + val t1 = g.aggregateMessages[Array[Double]]( ctx => ctx.sendToSrc(ctx.dstAttr._2), - (g1, g2) => g1.addColumnVector(g2)) + (g1, g2) => { + val out = g1.clone() + blas.daxpy(out.length, 1.0, g2, 1, out, 1) + out + }) val gJoinT1 = g.outerJoinVertices(t1) { - (vid: VertexId, vd: (DoubleMatrix, DoubleMatrix, Double, Double), - msg: Option[DoubleMatrix]) => - if (msg.isDefined) (vd._1, vd._1 - .addColumnVector(msg.get.mul(vd._4)), vd._3, vd._4) else vd + (vid: VertexId, vd: (Array[Double], Array[Double], Double, Double), + msg: Option[Array[Double]]) => + if (msg.isDefined) { + val out = vd._1.clone() + blas.daxpy(out.length, vd._4, msg.get, 1, out, 1) + (vd._1, out, vd._3, vd._4) + } else { + vd + } }.cache() materialize(gJoinT1) g.unpersist() @@ -144,14 +156,24 @@ object SVDPlusPlus { g.cache() val t2 = g.aggregateMessages( sendMsgTrainF(conf, u), - (g1: (DoubleMatrix, DoubleMatrix, Double), g2: (DoubleMatrix, DoubleMatrix, Double)) => - (g1._1.addColumnVector(g2._1), g1._2.addColumnVector(g2._2), g1._3 + g2._3)) + (g1: (Array[Double], Array[Double], Double), g2: (Array[Double], Array[Double], Double)) => + { + val out1 = g1._1.clone() + blas.daxpy(out1.length, 1.0, g2._1, 1, out1, 1) + val out2 = g2._2.clone() + blas.daxpy(out2.length, 1.0, g2._2, 1, out2, 1) + (out1, out2, g1._3 + g2._3) + }) val gJoinT2 = g.outerJoinVertices(t2) { (vid: VertexId, - vd: (DoubleMatrix, DoubleMatrix, Double, Double), - msg: Option[(DoubleMatrix, DoubleMatrix, Double)]) => - (vd._1.addColumnVector(msg.get._1), vd._2.addColumnVector(msg.get._2), - vd._3 + msg.get._3, vd._4) + vd: (Array[Double], Array[Double], Double, Double), + msg: Option[(Array[Double], Array[Double], Double)]) => { + val out1 = vd._1.clone() + blas.daxpy(out1.length, 1.0, msg.get._1, 1, out1, 1) + val out2 = vd._2.clone() + blas.daxpy(out2.length, 1.0, msg.get._2, 1, out2, 1) + (out1, out2, vd._3 + msg.get._3, vd._4) + } }.cache() materialize(gJoinT2) g.unpersist() @@ -160,10 +182,10 @@ object SVDPlusPlus { // calculate error on training set def sendMsgTestF(conf: Conf, u: Double) - (ctx: EdgeContext[(DoubleMatrix, DoubleMatrix, Double, Double), Double, Double]) { + (ctx: EdgeContext[(Array[Double], Array[Double], Double, Double), Double, Double]) { val (usr, itm) = (ctx.srcAttr, ctx.dstAttr) val (p, q) = (usr._1, itm._1) - var pred = u + usr._3 + itm._3 + q.dot(usr._2) + var pred = u + usr._3 + itm._3 + blas.ddot(q.length, q, 1, usr._2, 1) pred = math.max(pred, conf.minVal) pred = math.min(pred, conf.maxVal) val err = (ctx.attr - pred) * (ctx.attr - pred) @@ -173,7 +195,7 @@ object SVDPlusPlus { g.cache() val t3 = g.aggregateMessages[Double](sendMsgTestF(conf, u), _ + _) val gJoinT3 = g.outerJoinVertices(t3) { - (vid: VertexId, vd: (DoubleMatrix, DoubleMatrix, Double, Double), msg: Option[Double]) => + (vid: VertexId, vd: (Array[Double], Array[Double], Double, Double), msg: Option[Double]) => if (msg.isDefined) (vd._1, vd._2, vd._3, msg.get) else vd }.cache() materialize(gJoinT3) diff --git a/graphx/src/test/scala/org/apache/spark/graphx/lib/SVDPlusPlusSuite.scala b/graphx/src/test/scala/org/apache/spark/graphx/lib/SVDPlusPlusSuite.scala index 9987a4b1a3c25..7bd6b7f3c4ab2 100644 --- a/graphx/src/test/scala/org/apache/spark/graphx/lib/SVDPlusPlusSuite.scala +++ b/graphx/src/test/scala/org/apache/spark/graphx/lib/SVDPlusPlusSuite.scala @@ -32,11 +32,11 @@ class SVDPlusPlusSuite extends FunSuite with LocalSparkContext { Edge(fields(0).toLong * 2, fields(1).toLong * 2 + 1, fields(2).toDouble) } val conf = new SVDPlusPlus.Conf(10, 2, 0.0, 5.0, 0.007, 0.007, 0.005, 0.015) // 2 iterations - var (graph, u) = SVDPlusPlus.runSVDPlusPlus(edges, conf) + val (graph, _) = SVDPlusPlus.run(edges, conf) graph.cache() - val err = graph.vertices.collect().map{ case (vid, vd) => + val err = graph.vertices.map { case (vid, vd) => if (vid % 2 == 1) vd._4 else 0.0 - }.reduce(_ + _) / graph.triplets.collect().size + }.reduce(_ + _) / graph.numEdges assert(err <= svdppErr) } } diff --git a/mllib/pom.xml b/mllib/pom.xml index b5c949e155cfd..a76704a8c2c59 100644 --- a/mllib/pom.xml +++ b/mllib/pom.xml @@ -59,6 +59,7 @@ org.jblas jblas ${jblas.version} + test org.scalanlp @@ -116,7 +117,7 @@ com.github.fommil.netlib all - 1.1.2 + ${netlib.java.version} pom diff --git a/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala b/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala index 7bb69df65362b..e3515ee81af3d 100644 --- a/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala +++ b/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala @@ -26,7 +26,6 @@ import scala.util.hashing.byteswap64 import com.github.fommil.netlib.BLAS.{getInstance => blas} import com.github.fommil.netlib.LAPACK.{getInstance => lapack} -import org.jblas.DoubleMatrix import org.netlib.util.intW import org.apache.spark.{Logging, Partitioner} @@ -361,14 +360,14 @@ object ALS extends Logging { private[recommendation] class NNLSSolver extends LeastSquaresNESolver { private var rank: Int = -1 private var workspace: NNLS.Workspace = _ - private var ata: DoubleMatrix = _ + private var ata: Array[Double] = _ private var initialized: Boolean = false private def initialize(rank: Int): Unit = { if (!initialized) { this.rank = rank workspace = NNLS.createWorkspace(rank) - ata = new DoubleMatrix(rank, rank) + ata = new Array[Double](rank * rank) initialized = true } else { require(this.rank == rank) @@ -385,7 +384,7 @@ object ALS extends Logging { val rank = ne.k initialize(rank) fillAtA(ne.ata, lambda * ne.n) - val x = NNLS.solve(ata, new DoubleMatrix(rank, 1, ne.atb: _*), workspace) + val x = NNLS.solve(ata, ne.atb, workspace) ne.reset() x.map(x => x.toFloat) } @@ -398,17 +397,16 @@ object ALS extends Logging { var i = 0 var pos = 0 var a = 0.0 - val data = ata.data while (i < rank) { var j = 0 while (j <= i) { a = triAtA(pos) - data(i * rank + j) = a - data(j * rank + i) = a + ata(i * rank + j) = a + ata(j * rank + i) = a pos += 1 j += 1 } - data(i * rank + i) += lambda + ata(i * rank + i) += lambda i += 1 } } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/optimization/NNLS.scala b/mllib/src/main/scala/org/apache/spark/mllib/optimization/NNLS.scala index ccd93b318bc23..4766f7708295d 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/optimization/NNLS.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/optimization/NNLS.scala @@ -17,7 +17,9 @@ package org.apache.spark.mllib.optimization -import org.jblas.{DoubleMatrix, SimpleBlas} +import java.{util => ju} + +import com.github.fommil.netlib.BLAS.{getInstance => blas} /** * Object used to solve nonnegative least squares problems using a modified @@ -25,20 +27,20 @@ import org.jblas.{DoubleMatrix, SimpleBlas} */ private[spark] object NNLS { class Workspace(val n: Int) { - val scratch = new DoubleMatrix(n, 1) - val grad = new DoubleMatrix(n, 1) - val x = new DoubleMatrix(n, 1) - val dir = new DoubleMatrix(n, 1) - val lastDir = new DoubleMatrix(n, 1) - val res = new DoubleMatrix(n, 1) - - def wipe() { - scratch.fill(0.0) - grad.fill(0.0) - x.fill(0.0) - dir.fill(0.0) - lastDir.fill(0.0) - res.fill(0.0) + val scratch = new Array[Double](n) + val grad = new Array[Double](n) + val x = new Array[Double](n) + val dir = new Array[Double](n) + val lastDir = new Array[Double](n) + val res = new Array[Double](n) + + def wipe(): Unit = { + ju.Arrays.fill(scratch, 0.0) + ju.Arrays.fill(grad, 0.0) + ju.Arrays.fill(x, 0.0) + ju.Arrays.fill(dir, 0.0) + ju.Arrays.fill(lastDir, 0.0) + ju.Arrays.fill(res, 0.0) } } @@ -60,18 +62,18 @@ private[spark] object NNLS { * direction, however, while this method only uses a conjugate gradient direction if the last * iteration did not cause a previously-inactive constraint to become active. */ - def solve(ata: DoubleMatrix, atb: DoubleMatrix, ws: Workspace): Array[Double] = { + def solve(ata: Array[Double], atb: Array[Double], ws: Workspace): Array[Double] = { ws.wipe() - val n = atb.rows + val n = atb.length val scratch = ws.scratch // find the optimal unconstrained step - def steplen(dir: DoubleMatrix, res: DoubleMatrix): Double = { - val top = SimpleBlas.dot(dir, res) - SimpleBlas.gemv(1.0, ata, dir, 0.0, scratch) + def steplen(dir: Array[Double], res: Array[Double]): Double = { + val top = blas.ddot(n, dir, 1, res, 1) + blas.dgemv("N", n, n, 1.0, ata, n, dir, 1, 0.0, scratch, 1) // Push the denominator upward very slightly to avoid infinities and silliness - top / (SimpleBlas.dot(scratch, dir) + 1e-20) + top / (blas.ddot(n, scratch, 1, dir, 1) + 1e-20) } // stopping condition @@ -96,52 +98,52 @@ private[spark] object NNLS { var i = 0 while (iterno < iterMax) { // find the residual - SimpleBlas.gemv(1.0, ata, x, 0.0, res) - SimpleBlas.axpy(-1.0, atb, res) - SimpleBlas.copy(res, grad) + blas.dgemv("N", n, n, 1.0, ata, n, x, 1, 0.0, res, 1) + blas.daxpy(n, -1.0, atb, 1, res, 1) + blas.dcopy(n, res, 1, grad, 1) // project the gradient i = 0 while (i < n) { - if (grad.data(i) > 0.0 && x.data(i) == 0.0) { - grad.data(i) = 0.0 + if (grad(i) > 0.0 && x(i) == 0.0) { + grad(i) = 0.0 } i = i + 1 } - val ngrad = SimpleBlas.dot(grad, grad) + val ngrad = blas.ddot(n, grad, 1, grad, 1) - SimpleBlas.copy(grad, dir) + blas.dcopy(n, grad, 1, dir, 1) // use a CG direction under certain conditions var step = steplen(grad, res) var ndir = 0.0 - val nx = SimpleBlas.dot(x, x) + val nx = blas.ddot(n, x, 1, x, 1) if (iterno > lastWall + 1) { val alpha = ngrad / lastNorm - SimpleBlas.axpy(alpha, lastDir, dir) + blas.daxpy(n, alpha, lastDir, 1, dir, 1) val dstep = steplen(dir, res) - ndir = SimpleBlas.dot(dir, dir) + ndir = blas.ddot(n, dir, 1, dir, 1) if (stop(dstep, ndir, nx)) { // reject the CG step if it could lead to premature termination - SimpleBlas.copy(grad, dir) - ndir = SimpleBlas.dot(dir, dir) + blas.dcopy(n, grad, 1, dir, 1) + ndir = blas.ddot(n, dir, 1, dir, 1) } else { step = dstep } } else { - ndir = SimpleBlas.dot(dir, dir) + ndir = blas.ddot(n, dir, 1, dir, 1) } // terminate? if (stop(step, ndir, nx)) { - return x.data.clone + return x.clone } // don't run through the walls i = 0 while (i < n) { - if (step * dir.data(i) > x.data(i)) { - step = x.data(i) / dir.data(i) + if (step * dir(i) > x(i)) { + step = x(i) / dir(i) } i = i + 1 } @@ -149,19 +151,19 @@ private[spark] object NNLS { // take the step i = 0 while (i < n) { - if (step * dir.data(i) > x.data(i) * (1 - 1e-14)) { - x.data(i) = 0 + if (step * dir(i) > x(i) * (1 - 1e-14)) { + x(i) = 0 lastWall = iterno } else { - x.data(i) -= step * dir.data(i) + x(i) -= step * dir(i) } i = i + 1 } iterno = iterno + 1 - SimpleBlas.copy(dir, lastDir) + blas.dcopy(n, dir, 1, lastDir, 1) lastNorm = ngrad } - x.data.clone + x.clone } } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/recommendation/MatrixFactorizationModel.scala b/mllib/src/main/scala/org/apache/spark/mllib/recommendation/MatrixFactorizationModel.scala index 5f5a996a87b81..36cbf060d9998 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/recommendation/MatrixFactorizationModel.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/recommendation/MatrixFactorizationModel.scala @@ -21,10 +21,10 @@ import java.io.IOException import java.lang.{Integer => JavaInteger} import org.apache.hadoop.fs.Path -import org.jblas.DoubleMatrix import org.json4s._ import org.json4s.JsonDSL._ import org.json4s.jackson.JsonMethods._ +import com.github.fommil.netlib.BLAS.{getInstance => blas} import org.apache.spark.{Logging, SparkContext} import org.apache.spark.api.java.{JavaPairRDD, JavaRDD} @@ -70,9 +70,9 @@ class MatrixFactorizationModel( /** Predict the rating of one user for one product. */ def predict(user: Int, product: Int): Double = { - val userVector = new DoubleMatrix(userFeatures.lookup(user).head) - val productVector = new DoubleMatrix(productFeatures.lookup(product).head) - userVector.dot(productVector) + val userVector = userFeatures.lookup(user).head + val productVector = productFeatures.lookup(product).head + blas.ddot(userVector.length, userVector, 1, productVector, 1) } /** @@ -89,9 +89,7 @@ class MatrixFactorizationModel( } users.join(productFeatures).map { case (product, ((user, uFeatures), pFeatures)) => - val userVector = new DoubleMatrix(uFeatures) - val productVector = new DoubleMatrix(pFeatures) - Rating(user, product, userVector.dot(productVector)) + Rating(user, product, blas.ddot(uFeatures.length, uFeatures, 1, pFeatures, 1)) } } @@ -143,9 +141,8 @@ class MatrixFactorizationModel( recommendToFeatures: Array[Double], recommendableFeatures: RDD[(Int, Array[Double])], num: Int): Array[(Int, Double)] = { - val recommendToVector = new DoubleMatrix(recommendToFeatures) val scored = recommendableFeatures.map { case (id,features) => - (id, recommendToVector.dot(new DoubleMatrix(features))) + (id, blas.ddot(features.length, recommendToFeatures, 1, features, 1)) } scored.top(num)(Ordering.by(_._2)) } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/util/LinearDataGenerator.scala b/mllib/src/main/scala/org/apache/spark/mllib/util/LinearDataGenerator.scala index 97f54aa62d31c..c9d33787b0bb5 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/util/LinearDataGenerator.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/util/LinearDataGenerator.scala @@ -20,7 +20,7 @@ package org.apache.spark.mllib.util import scala.collection.JavaConversions._ import scala.util.Random -import org.jblas.DoubleMatrix +import com.github.fommil.netlib.BLAS.{getInstance => blas} import org.apache.spark.annotation.DeveloperApi import org.apache.spark.SparkContext @@ -72,11 +72,10 @@ object LinearDataGenerator { eps: Double = 0.1): Seq[LabeledPoint] = { val rnd = new Random(seed) - val weightsMat = new DoubleMatrix(1, weights.length, weights:_*) val x = Array.fill[Array[Double]](nPoints)( Array.fill[Double](weights.length)(2 * rnd.nextDouble - 1.0)) val y = x.map { xi => - new DoubleMatrix(1, xi.length, xi: _*).dot(weightsMat) + intercept + eps * rnd.nextGaussian() + blas.ddot(weights.length, xi, 1, weights, 1) + intercept + eps * rnd.nextGaussian() } y.zip(x).map(p => LabeledPoint(p._1, Vectors.dense(p._2))) } @@ -100,9 +99,9 @@ object LinearDataGenerator { eps: Double, nparts: Int = 2, intercept: Double = 0.0) : RDD[LabeledPoint] = { - org.jblas.util.Random.seed(42) + val random = new Random(42) // Random values distributed uniformly in [-0.5, 0.5] - val w = DoubleMatrix.rand(nfeatures, 1).subi(0.5) + val w = Array.fill(nfeatures)(random.nextDouble() - 0.5) val data: RDD[LabeledPoint] = sc.parallelize(0 until nparts, nparts).flatMap { p => val seed = 42 + p diff --git a/mllib/src/main/scala/org/apache/spark/mllib/util/MFDataGenerator.scala b/mllib/src/main/scala/org/apache/spark/mllib/util/MFDataGenerator.scala index b76fbe89c3681..0c5b4f9d04a74 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/util/MFDataGenerator.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/util/MFDataGenerator.scala @@ -17,13 +17,14 @@ package org.apache.spark.mllib.util +import java.{util => ju} + import scala.language.postfixOps import scala.util.Random -import org.jblas.DoubleMatrix - -import org.apache.spark.annotation.DeveloperApi import org.apache.spark.SparkContext +import org.apache.spark.annotation.DeveloperApi +import org.apache.spark.mllib.linalg.{BLAS, DenseMatrix} import org.apache.spark.rdd.RDD /** @@ -72,24 +73,25 @@ object MFDataGenerator { val sc = new SparkContext(sparkMaster, "MFDataGenerator") - val A = DoubleMatrix.randn(m, rank) - val B = DoubleMatrix.randn(rank, n) - val z = 1 / scala.math.sqrt(scala.math.sqrt(rank)) - A.mmuli(z) - B.mmuli(z) - val fullData = A.mmul(B) + val random = new ju.Random(42L) + + val A = DenseMatrix.randn(m, rank, random) + val B = DenseMatrix.randn(rank, n, random) + val z = 1 / math.sqrt(rank) + val fullData = DenseMatrix.zeros(m, n) + BLAS.gemm(z, A, B, 1.0, fullData) val df = rank * (m + n - rank) val sampSize = scala.math.min(scala.math.round(trainSampFact * df), scala.math.round(.99 * m * n)).toInt val rand = new Random() val mn = m * n - val shuffled = rand.shuffle(1 to mn toList) + val shuffled = rand.shuffle((0 until mn).toList) val omega = shuffled.slice(0, sampSize) val ordered = omega.sortWith(_ < _).toArray val trainData: RDD[(Int, Int, Double)] = sc.parallelize(ordered) - .map(x => (fullData.indexRows(x - 1), fullData.indexColumns(x - 1), fullData.get(x - 1))) + .map(x => (x % m, x / m, fullData.values(x))) // optionally add gaussian noise if (noise) { @@ -105,7 +107,7 @@ object MFDataGenerator { val testOmega = shuffled.slice(sampSize, sampSize + testSampSize) val testOrdered = testOmega.sortWith(_ < _).toArray val testData: RDD[(Int, Int, Double)] = sc.parallelize(testOrdered) - .map(x => (fullData.indexRows(x - 1), fullData.indexColumns(x - 1), fullData.get(x - 1))) + .map(x => (x % m, x / m, fullData.values(x))) testData.map(x => x._1 + "," + x._2 + "," + x._3).saveAsTextFile(outputPath) } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/util/SVMDataGenerator.scala b/mllib/src/main/scala/org/apache/spark/mllib/util/SVMDataGenerator.scala index 7db97e6bac688..a8e30cc9d730c 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/util/SVMDataGenerator.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/util/SVMDataGenerator.scala @@ -19,7 +19,7 @@ package org.apache.spark.mllib.util import scala.util.Random -import org.jblas.DoubleMatrix +import com.github.fommil.netlib.BLAS.{getInstance => blas} import org.apache.spark.annotation.DeveloperApi import org.apache.spark.SparkContext @@ -51,8 +51,7 @@ object SVMDataGenerator { val sc = new SparkContext(sparkMaster, "SVMGenerator") val globalRnd = new Random(94720) - val trueWeights = new DoubleMatrix(1, nfeatures + 1, - Array.fill[Double](nfeatures + 1)(globalRnd.nextGaussian()):_*) + val trueWeights = Array.fill[Double](nfeatures + 1)(globalRnd.nextGaussian()) val data: RDD[LabeledPoint] = sc.parallelize(0 until nexamples, parts).map { idx => val rnd = new Random(42 + idx) @@ -60,7 +59,7 @@ object SVMDataGenerator { val x = Array.fill[Double](nfeatures) { rnd.nextDouble() * 2.0 - 1.0 } - val yD = new DoubleMatrix(1, x.length, x: _*).dot(trueWeights) + rnd.nextGaussian() * 0.1 + val yD = blas.ddot(trueWeights.length, x, 1, trueWeights, 1) + rnd.nextGaussian() * 0.1 val y = if (yD < 0) 0.0 else 1.0 LabeledPoint(y, Vectors.dense(x)) } diff --git a/mllib/src/test/scala/org/apache/spark/mllib/optimization/NNLSSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/optimization/NNLSSuite.scala index 82c327bd49fcd..22855e4e8f247 100644 --- a/mllib/src/test/scala/org/apache/spark/mllib/optimization/NNLSSuite.scala +++ b/mllib/src/test/scala/org/apache/spark/mllib/optimization/NNLSSuite.scala @@ -55,7 +55,7 @@ class NNLSSuite extends FunSuite { for (k <- 0 until 100) { val (ata, atb) = genOnesData(n, rand) - val x = new DoubleMatrix(NNLS.solve(ata, atb, ws)) + val x = new DoubleMatrix(NNLS.solve(ata.data, atb.data, ws)) assert(x.length === n) val answer = DoubleMatrix.ones(n, 1) SimpleBlas.axpy(-1.0, answer, x) @@ -79,7 +79,7 @@ class NNLSSuite extends FunSuite { val goodx = Array(0.13025, 0.54506, 0.2874, 0.0, 0.028628) val ws = NNLS.createWorkspace(n) - val x = NNLS.solve(ata, atb, ws) + val x = NNLS.solve(ata.data, atb.data, ws) for (i <- 0 until n) { assert(x(i) ~== goodx(i) absTol 1E-3) assert(x(i) >= 0) @@ -104,7 +104,7 @@ class NNLSSuite extends FunSuite { val ws = NNLS.createWorkspace(n) - val x = new DoubleMatrix(NNLS.solve(ata, atb, ws)) + val x = new DoubleMatrix(NNLS.solve(ata.data, atb.data, ws)) val obj = computeObjectiveValue(ata, atb, x) assert(obj < refObj + 1E-5) diff --git a/mllib/src/test/scala/org/apache/spark/mllib/stat/KernelDensitySuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/stat/KernelDensitySuite.scala index f6a1e19f50296..16ecae23dd9d4 100644 --- a/mllib/src/test/scala/org/apache/spark/mllib/stat/KernelDensitySuite.scala +++ b/mllib/src/test/scala/org/apache/spark/mllib/stat/KernelDensitySuite.scala @@ -21,9 +21,9 @@ import org.scalatest.FunSuite import org.apache.commons.math3.distribution.NormalDistribution -import org.apache.spark.mllib.util.LocalClusterSparkContext +import org.apache.spark.mllib.util.MLlibTestSparkContext -class KernelDensitySuite extends FunSuite with LocalClusterSparkContext { +class KernelDensitySuite extends FunSuite with MLlibTestSparkContext { test("kernel density single sample") { val rdd = sc.parallelize(Array(5.0)) val evaluationPoints = Array(5.0, 6.0) diff --git a/pom.xml b/pom.xml index a19da73cf45b3..6fc56a86d44ac 100644 --- a/pom.xml +++ b/pom.xml @@ -157,6 +157,7 @@ 1.8.8 2.4.4 1.1.1.6 + 1.1.2 [Review on Reviewable](https://reviewable.io/reviews/apache/spark/5001) Author: Cheng Lian Closes #5001 from liancheng/parquet-doc and squashes the following commits: 89ad3db [Cheng Lian] Addresses @rxin's comments 7eb6955 [Cheng Lian] Docs for the new Parquet data source 415eefb [Cheng Lian] Some minor formatting improvements --- docs/sql-programming-guide.md | 237 ++++++++++++++++++++++++++-------- 1 file changed, 180 insertions(+), 57 deletions(-) diff --git a/docs/sql-programming-guide.md b/docs/sql-programming-guide.md index 76aa1a533d56e..11c29e20632ae 100644 --- a/docs/sql-programming-guide.md +++ b/docs/sql-programming-guide.md @@ -21,14 +21,14 @@ The DataFrame API is available in [Scala](api/scala/index.html#org.apache.spark. All of the examples on this page use sample data included in the Spark distribution and can be run in the `spark-shell` or the `pyspark` shell. -## Starting Point: SQLContext +## Starting Point: `SQLContext`
    The entry point into all functionality in Spark SQL is the -[SQLContext](api/scala/index.html#org.apache.spark.sql.SQLContext) class, or one of its -descendants. To create a basic SQLContext, all you need is a SparkContext. +[`SQLContext`](api/scala/index.html#org.apache.spark.sql.`SQLContext`) class, or one of its +descendants. To create a basic `SQLContext`, all you need is a SparkContext. {% highlight scala %} val sc: SparkContext // An existing SparkContext. @@ -43,8 +43,8 @@ import sqlContext.implicits._
    The entry point into all functionality in Spark SQL is the -[SQLContext](api/java/index.html#org.apache.spark.sql.SQLContext) class, or one of its -descendants. To create a basic SQLContext, all you need is a SparkContext. +[`SQLContext`](api/java/index.html#org.apache.spark.sql.SQLContext) class, or one of its +descendants. To create a basic `SQLContext`, all you need is a SparkContext. {% highlight java %} JavaSparkContext sc = ...; // An existing JavaSparkContext. @@ -56,8 +56,8 @@ SQLContext sqlContext = new org.apache.spark.sql.SQLContext(sc);
    The entry point into all relational functionality in Spark is the -[SQLContext](api/python/pyspark.sql.SQLContext-class.html) class, or one -of its decedents. To create a basic SQLContext, all you need is a SparkContext. +[`SQLContext`](api/python/pyspark.sql.SQLContext-class.html) class, or one +of its decedents. To create a basic `SQLContext`, all you need is a SparkContext. {% highlight python %} from pyspark.sql import SQLContext @@ -67,20 +67,20 @@ sqlContext = SQLContext(sc)
    -In addition to the basic SQLContext, you can also create a HiveContext, which provides a -superset of the functionality provided by the basic SQLContext. Additional features include +In addition to the basic `SQLContext`, you can also create a `HiveContext`, which provides a +superset of the functionality provided by the basic `SQLContext`. Additional features include the ability to write queries using the more complete HiveQL parser, access to Hive UDFs, and the -ability to read data from Hive tables. To use a HiveContext, you do not need to have an -existing Hive setup, and all of the data sources available to a SQLContext are still available. -HiveContext is only packaged separately to avoid including all of Hive's dependencies in the default -Spark build. If these dependencies are not a problem for your application then using HiveContext -is recommended for the 1.3 release of Spark. Future releases will focus on bringing SQLContext up -to feature parity with a HiveContext. +ability to read data from Hive tables. To use a `HiveContext`, you do not need to have an +existing Hive setup, and all of the data sources available to a `SQLContext` are still available. +`HiveContext` is only packaged separately to avoid including all of Hive's dependencies in the default +Spark build. If these dependencies are not a problem for your application then using `HiveContext` +is recommended for the 1.3 release of Spark. Future releases will focus on bringing `SQLContext` up +to feature parity with a `HiveContext`. The specific variant of SQL that is used to parse queries can also be selected using the `spark.sql.dialect` option. This parameter can be changed using either the `setConf` method on -a SQLContext or by using a `SET key=value` command in SQL. For a SQLContext, the only dialect -available is "sql" which uses a simple SQL parser provided by Spark SQL. In a HiveContext, the +a `SQLContext` or by using a `SET key=value` command in SQL. For a `SQLContext`, the only dialect +available is "sql" which uses a simple SQL parser provided by Spark SQL. In a `HiveContext`, the default is "hiveql", though "sql" is also available. Since the HiveQL parser is much more complete, this is recommended for most use cases. @@ -100,7 +100,7 @@ val sqlContext = new org.apache.spark.sql.SQLContext(sc) val df = sqlContext.jsonFile("examples/src/main/resources/people.json") // Displays the content of the DataFrame to stdout -df.show() +df.show() {% endhighlight %}
    @@ -151,10 +151,10 @@ val df = sqlContext.jsonFile("examples/src/main/resources/people.json") // Show the content of the DataFrame df.show() -// age name +// age name // null Michael -// 30 Andy -// 19 Justin +// 30 Andy +// 19 Justin // Print the schema in a tree format df.printSchema() @@ -164,17 +164,17 @@ df.printSchema() // Select only the "name" column df.select("name").show() -// name +// name // Michael -// Andy -// Justin +// Andy +// Justin // Select everybody, but increment the age by 1 df.select("name", df("age") + 1).show() // name (age + 1) -// Michael null -// Andy 31 -// Justin 20 +// Michael null +// Andy 31 +// Justin 20 // Select people older than 21 df.filter(df("name") > 21).show() @@ -201,10 +201,10 @@ DataFrame df = sqlContext.jsonFile("examples/src/main/resources/people.json"); // Show the content of the DataFrame df.show(); -// age name +// age name // null Michael -// 30 Andy -// 19 Justin +// 30 Andy +// 19 Justin // Print the schema in a tree format df.printSchema(); @@ -214,17 +214,17 @@ df.printSchema(); // Select only the "name" column df.select("name").show(); -// name +// name // Michael -// Andy -// Justin +// Andy +// Justin // Select everybody, but increment the age by 1 df.select("name", df.col("age").plus(1)).show(); // name (age + 1) -// Michael null -// Andy 31 -// Justin 20 +// Michael null +// Andy 31 +// Justin 20 // Select people older than 21 df.filter(df("name") > 21).show(); @@ -251,10 +251,10 @@ df = sqlContext.jsonFile("examples/src/main/resources/people.json") # Show the content of the DataFrame df.show() -## age name +## age name ## null Michael -## 30 Andy -## 19 Justin +## 30 Andy +## 19 Justin # Print the schema in a tree format df.printSchema() @@ -264,17 +264,17 @@ df.printSchema() # Select only the "name" column df.select("name").show() -## name +## name ## Michael -## Andy -## Justin +## Andy +## Justin # Select everybody, but increment the age by 1 df.select("name", df.age + 1).show() ## name (age + 1) -## Michael null -## Andy 31 -## Justin 20 +## Michael null +## Andy 31 +## Justin 20 # Select people older than 21 df.filter(df.name > 21).show() @@ -797,7 +797,7 @@ When working with a `HiveContext`, `DataFrames` can also be saved as persistent contents of the dataframe and create a pointer to the data in the HiveMetastore. Persistent tables will still exist even after your Spark program has restarted, as long as you maintain your connection to the same metastore. A DataFrame for a persistent table can be created by calling the `table` -method on a SQLContext with the name of the table. +method on a `SQLContext` with the name of the table. By default `saveAsTable` will create a "managed table", meaning that the location of the data will be controlled by the metastore. Managed tables will also have their data deleted automatically @@ -907,9 +907,132 @@ SELECT * FROM parquetTable
    +### Partition discovery + +Table partitioning is a common optimization approach used in systems like Hive. In a partitioned +table, data are usually stored in different directories, with partitioning column values encoded in +the path of each partition directory. The Parquet data source is now able to discover and infer +partitioning information automatically. For exmaple, we can store all our previously used +population data into a partitioned table using the following directory structure, with two extra +columns, `gender` and `country` as partitioning columns: + +{% highlight text %} + +path +└── to + └── table + ├── gender=male + │   ├── ... + │   │ + │   ├── country=US + │   │   └── data.parquet + │   ├── country=CN + │   │   └── data.parquet + │   └── ... + └── gender=female +    ├── ... +    │ +    ├── country=US +    │   └── data.parquet +    ├── country=CN +    │   └── data.parquet +    └── ... + +{% endhighlight %} + +By passing `path/to/table` to either `SQLContext.parquetFile` or `SQLContext.load`, Spark SQL will +automatically extract the partitioning information from the paths. Now the schema of the returned +DataFrame becomes: + +{% highlight text %} + +root +|-- name: string (nullable = true) +|-- age: long (nullable = true) +|-- gender: string (nullable = true) +|-- country: string (nullable = true) + +{% endhighlight %} + +Notice that the data types of the partitioning columns are automatically inferred. Currently, +numeric data types and string type are supported. + +### Schema merging + +Like ProtocolBuffer, Avro, and Thrift, Parquet also supports schema evolution. Users can start with +a simple schema, and gradually add more columns to the schema as needed. In this way, users may end +up with multiple Parquet files with different but mutually compatible schemas. The Parquet data +source is now able to automatically detect this case and merge schemas of all these files. + +
    + +
    + +{% highlight scala %} +// sqlContext from the previous example is used in this example. +// This is used to implicitly convert an RDD to a DataFrame. +import sqlContext.implicits._ + +// Create a simple DataFrame, stored into a partition directory +val df1 = sparkContext.makeRDD(1 to 5).map(i => (i, i * 2)).toDF("single", "double") +df1.saveAsParquetFile("data/test_table/key=1") + +// Create another DataFrame in a new partition directory, +// adding a new column and dropping an existing column +val df2 = sparkContext.makeRDD(6 to 10).map(i => (i, i * 3)).toDF("single", "triple") +df2.saveAsParquetFile("data/test_table/key=2") + +// Read the partitioned table +val df3 = sqlContext.parquetFile("data/test_table") +df3.printSchema() + +// The final schema consists of all 3 columns in the Parquet files together +// with the partiioning column appeared in the partition directory paths. +// root +// |-- single: int (nullable = true) +// |-- double: int (nullable = true) +// |-- triple: int (nullable = true) +// |-- key : int (nullable = true) +{% endhighlight %} + +
    + +
    + +{% highlight python %} +# sqlContext from the previous example is used in this example. + +# Create a simple DataFrame, stored into a partition directory +df1 = sqlContext.createDataFrame(sc.parallelize(range(1, 6))\ + .map(lambda i: Row(single=i, double=i * 2))) +df1.save("data/test_table/key=1", "parquet") + +# Create another DataFrame in a new partition directory, +# adding a new column and dropping an existing column +df2 = sqlContext.createDataFrame(sc.parallelize(range(6, 11)) + .map(lambda i: Row(single=i, triple=i * 3))) +df2.save("data/test_table/key=2", "parquet") + +# Read the partitioned table +df3 = sqlContext.parquetFile("data/test_table") +df3.printSchema() + +# The final schema consists of all 3 columns in the Parquet files together +# with the partiioning column appeared in the partition directory paths. +# root +# |-- single: int (nullable = true) +# |-- double: int (nullable = true) +# |-- triple: int (nullable = true) +# |-- key : int (nullable = true) +{% endhighlight %} + +
    + +
    + ### Configuration -Configuration of Parquet can be done using the `setConf` method on SQLContext or by running +Configuration of Parquet can be done using the `setConf` method on `SQLContext` or by running `SET key=value` commands using SQL. @@ -972,7 +1095,7 @@ Configuration of Parquet can be done using the `setConf` method on SQLContext or
    Spark SQL can automatically infer the schema of a JSON dataset and load it as a DataFrame. -This conversion can be done using one of two methods in a SQLContext: +This conversion can be done using one of two methods in a `SQLContext`: * `jsonFile` - loads data from a directory of JSON files where each line of the files is a JSON object. * `jsonRDD` - loads data from an existing RDD where each element of the RDD is a string containing a JSON object. @@ -1014,7 +1137,7 @@ val anotherPeople = sqlContext.jsonRDD(anotherPeopleRDD)
    Spark SQL can automatically infer the schema of a JSON dataset and load it as a DataFrame. -This conversion can be done using one of two methods in a SQLContext : +This conversion can be done using one of two methods in a `SQLContext` : * `jsonFile` - loads data from a directory of JSON files where each line of the files is a JSON object. * `jsonRDD` - loads data from an existing RDD where each element of the RDD is a string containing a JSON object. @@ -1056,7 +1179,7 @@ DataFrame anotherPeople = sqlContext.jsonRDD(anotherPeopleRDD);
    Spark SQL can automatically infer the schema of a JSON dataset and load it as a DataFrame. -This conversion can be done using one of two methods in a SQLContext: +This conversion can be done using one of two methods in a `SQLContext`: * `jsonFile` - loads data from a directory of JSON files where each line of the files is a JSON object. * `jsonRDD` - loads data from an existing RDD where each element of the RDD is a string containing a JSON object. @@ -1085,7 +1208,7 @@ people.printSchema() # Register this DataFrame as a table. people.registerTempTable("people") -# SQL statements can be run by using the sql methods provided by sqlContext. +# SQL statements can be run by using the sql methods provided by `sqlContext`. teenagers = sqlContext.sql("SELECT name FROM people WHERE age >= 13 AND age <= 19") # Alternatively, a DataFrame can be created for a JSON dataset represented by @@ -1131,7 +1254,7 @@ Configuration of Hive is done by placing your `hive-site.xml` file in `conf/`. When working with Hive one must construct a `HiveContext`, which inherits from `SQLContext`, and adds support for finding tables in the MetaStore and writing queries using HiveQL. Users who do -not have an existing Hive deployment can still create a HiveContext. When not configured by the +not have an existing Hive deployment can still create a `HiveContext`. When not configured by the hive-site.xml, the context automatically creates `metastore_db` and `warehouse` in the current directory. @@ -1318,7 +1441,7 @@ Spark SQL can cache tables using an in-memory columnar format by calling `sqlCon Then Spark SQL will scan only required columns and will automatically tune compression to minimize memory usage and GC pressure. You can call `sqlContext.uncacheTable("tableName")` to remove the table from memory. -Configuration of in-memory caching can be done using the `setConf` method on SQLContext or by running +Configuration of in-memory caching can be done using the `setConf` method on `SQLContext` or by running `SET key=value` commands using SQL.
    @@ -1429,10 +1552,10 @@ Configuration of Hive is done by placing your `hive-site.xml` file in `conf/`. You may also use the beeline script that comes with Hive. -Thrift JDBC server also supports sending thrift RPC messages over HTTP transport. -Use the following setting to enable HTTP mode as system property or in `hive-site.xml` file in `conf/`: +Thrift JDBC server also supports sending thrift RPC messages over HTTP transport. +Use the following setting to enable HTTP mode as system property or in `hive-site.xml` file in `conf/`: - hive.server2.transport.mode - Set this to value: http + hive.server2.transport.mode - Set this to value: http hive.server2.thrift.http.port - HTTP port number fo listen on; default is 10001 hive.server2.http.endpoint - HTTP endpoint; default is cliservice @@ -1506,7 +1629,7 @@ When using function inside of the DSL (now replaced with the `DataFrame` API) us Spark 1.3 removes the type aliases that were present in the base sql package for `DataType`. Users should instead import the classes in `org.apache.spark.sql.types` -#### UDF Registration Moved to sqlContext.udf (Java & Scala) +#### UDF Registration Moved to `sqlContext.udf` (Java & Scala) Functions that are used to register UDFs, either for use in the DataFrame DSL or SQL, have been moved into the udf object in `SQLContext`. From 9048e8102e3f564842fa0dc6e82edce70b7dd3d7 Mon Sep 17 00:00:00 2001 From: "Zhang, Liye" Date: Fri, 13 Mar 2015 13:59:54 +0000 Subject: [PATCH 050/122] [SPARK-6197][CORE] handle json exception when hisotry file not finished writing For details, please refer to [SPARK-6197](https://issues.apache.org/jira/browse/SPARK-6197) Author: Zhang, Liye Closes #4927 from liyezhang556520/jsonParseError and squashes the following commits: 5cbdc82 [Zhang, Liye] without unnecessary wrap 2b48831 [Zhang, Liye] small changes with sean owen's comments 2973024 [Zhang, Liye] handle json exception when file not finished writing --- .../apache/spark/deploy/master/Master.scala | 3 ++- .../spark/scheduler/ReplayListenerBus.scala | 25 ++++++++++++++++--- 2 files changed, 23 insertions(+), 5 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/deploy/master/Master.scala b/core/src/main/scala/org/apache/spark/deploy/master/Master.scala index 15814293227ab..22935c9b1d394 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/Master.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/Master.scala @@ -764,8 +764,9 @@ private[spark] class Master( val replayBus = new ReplayListenerBus() val ui = SparkUI.createHistoryUI(new SparkConf, replayBus, new SecurityManager(conf), appName + status, HistoryServer.UI_PATH_PREFIX + s"/${app.id}") + val maybeTruncated = eventLogFile.endsWith(EventLoggingListener.IN_PROGRESS) try { - replayBus.replay(logInput, eventLogFile) + replayBus.replay(logInput, eventLogFile, maybeTruncated) } finally { logInput.close() } diff --git a/core/src/main/scala/org/apache/spark/scheduler/ReplayListenerBus.scala b/core/src/main/scala/org/apache/spark/scheduler/ReplayListenerBus.scala index 95273c716b3e2..86f357abb8723 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/ReplayListenerBus.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/ReplayListenerBus.scala @@ -21,6 +21,7 @@ import java.io.{InputStream, IOException} import scala.io.Source +import com.fasterxml.jackson.core.JsonParseException import org.json4s.jackson.JsonMethods._ import org.apache.spark.Logging @@ -40,15 +41,31 @@ private[spark] class ReplayListenerBus extends SparkListenerBus with Logging { * * @param logData Stream containing event log data. * @param sourceName Filename (or other source identifier) from whence @logData is being read + * @param maybeTruncated Indicate whether log file might be truncated (some abnormal situations + * encountered, log file might not finished writing) or not */ - def replay(logData: InputStream, sourceName: String): Unit = { + def replay( + logData: InputStream, + sourceName: String, + maybeTruncated: Boolean = false): Unit = { var currentLine: String = null var lineNumber: Int = 1 try { val lines = Source.fromInputStream(logData).getLines() - lines.foreach { line => - currentLine = line - postToAll(JsonProtocol.sparkEventFromJson(parse(line))) + while (lines.hasNext) { + currentLine = lines.next() + try { + postToAll(JsonProtocol.sparkEventFromJson(parse(currentLine))) + } catch { + case jpe: JsonParseException => + // We can only ignore exception from last line of the file that might be truncated + if (!maybeTruncated || lines.hasNext) { + throw jpe + } else { + logWarning(s"Got JsonParseException from log file $sourceName" + + s" at line $lineNumber, the file might not have finished writing cleanly.") + } + } lineNumber += 1 } } catch { From ea3d2eed9b0a94b34543d9a9df87dc63a279deb1 Mon Sep 17 00:00:00 2001 From: Wenchen Fan Date: Fri, 13 Mar 2015 14:08:56 +0000 Subject: [PATCH 051/122] [CORE][minor] remove unnecessary ClassTag in `DAGScheduler` This existed at the very beginning, but became unnecessary after [this commit](https://github.com/apache/spark/commit/37d8f37a8ec110416fba0d51d8ba70370ac380c1#diff-6a9ff7fb74fd490a50462d45db2d5e11L272). I think we should remove it if we don't plan to use it in the future. Author: Wenchen Fan Closes #4992 from cloud-fan/small and squashes the following commits: e857f2e [Wenchen Fan] remove unnecessary ClassTag --- .../main/scala/org/apache/spark/scheduler/DAGScheduler.scala | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala b/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala index bc84e2351ad74..e4170a55b7981 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala @@ -26,7 +26,6 @@ import scala.collection.mutable.{ArrayBuffer, HashMap, HashSet, Map, Stack} import scala.concurrent.Await import scala.concurrent.duration._ import scala.language.postfixOps -import scala.reflect.ClassTag import scala.util.control.NonFatal import akka.pattern.ask @@ -497,7 +496,7 @@ class DAGScheduler( waiter } - def runJob[T, U: ClassTag]( + def runJob[T, U]( rdd: RDD[T], func: (TaskContext, Iterator[T]) => U, partitions: Seq[Int], From dc4abd4dc40deacab39bfa9572b06bf0ea6daa6d Mon Sep 17 00:00:00 2001 From: "Joseph K. Bradley" Date: Fri, 13 Mar 2015 10:26:09 -0700 Subject: [PATCH 052/122] [SPARK-6252] [mllib] Added getLambda to Scala NaiveBayes Note: not relevant for Python API since it only has a static train method Author: Joseph K. Bradley Author: Joseph K. Bradley Closes #4969 from jkbradley/SPARK-6252 and squashes the following commits: a471d90 [Joseph K. Bradley] small edits from review 63eff48 [Joseph K. Bradley] Added getLambda to Scala NaiveBayes --- .../apache/spark/mllib/classification/NaiveBayes.scala | 3 +++ .../spark/mllib/classification/NaiveBayesSuite.scala | 8 ++++++++ 2 files changed, 11 insertions(+) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala b/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala index b11fd4f128c56..2ebc7fa5d4234 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala @@ -166,6 +166,9 @@ class NaiveBayes private (private var lambda: Double) extends Serializable with this } + /** Get the smoothing parameter. Default: 1.0. */ + def getLambda: Double = lambda + /** * Run the algorithm with the configured parameters on an input RDD of LabeledPoint entries. * diff --git a/mllib/src/test/scala/org/apache/spark/mllib/classification/NaiveBayesSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/classification/NaiveBayesSuite.scala index 64dcc0fb9f82c..5a27c7d2309c5 100644 --- a/mllib/src/test/scala/org/apache/spark/mllib/classification/NaiveBayesSuite.scala +++ b/mllib/src/test/scala/org/apache/spark/mllib/classification/NaiveBayesSuite.scala @@ -85,6 +85,14 @@ class NaiveBayesSuite extends FunSuite with MLlibTestSparkContext { assert(numOfPredictions < input.length / 5) } + test("get, set params") { + val nb = new NaiveBayes() + nb.setLambda(2.0) + assert(nb.getLambda === 2.0) + nb.setLambda(3.0) + assert(nb.getLambda === 3.0) + } + test("Naive Bayes") { val nPoints = 10000 From 7f13434a5c52b815c584ec773ab0e5df1a35ea86 Mon Sep 17 00:00:00 2001 From: Xiangrui Meng Date: Fri, 13 Mar 2015 10:27:28 -0700 Subject: [PATCH 053/122] [SPARK-6278][MLLIB] Mention the change of objective in linear regression As discussed in the RC3 vote thread, we should mention the change of objective in linear regression in the migration guide. srowen Author: Xiangrui Meng Closes #4978 from mengxr/SPARK-6278 and squashes the following commits: fb3bbe6 [Xiangrui Meng] mention regularization parameter bfd6cff [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-6278 375fd09 [Xiangrui Meng] address Sean's comments f87ae71 [Xiangrui Meng] mention step size change --- docs/mllib-guide.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/docs/mllib-guide.md b/docs/mllib-guide.md index 598374f66df5e..f8e879496c135 100644 --- a/docs/mllib-guide.md +++ b/docs/mllib-guide.md @@ -102,6 +102,8 @@ In the `spark.mllib` package, there were several breaking changes. The first ch * In `DecisionTree`, the deprecated class method `train` has been removed. (The object/static `train` methods remain.) * In `Strategy`, the `checkpointDir` parameter has been removed. Checkpointing is still supported, but the checkpoint directory must be set before calling tree and tree ensemble training. * `PythonMLlibAPI` (the interface between Scala/Java and Python for MLlib) was a public API but is now private, declared `private[python]`. This was never meant for external use. +* In linear regression (including Lasso and ridge regression), the squared loss is now divided by 2. + So in order to produce the same result as in 1.2, the regularization parameter needs to be divided by 2 and the step size needs to be multiplied by 2. ## Previous Spark Versions From b943f5d907df0607ecffb729f2bccfa436438d7e Mon Sep 17 00:00:00 2001 From: Brennon York Date: Fri, 13 Mar 2015 18:48:31 +0000 Subject: [PATCH 054/122] [SPARK-4600][GraphX]: org.apache.spark.graphx.VertexRDD.diff does not work Turns out, per the [convo on the JIRA](https://issues.apache.org/jira/browse/SPARK-4600), `diff` is acting exactly as should. It became a large misconception as I thought it meant set difference, when in fact it does not. To that extent I merely updated the `diff` documentation to, hopefully, better reflect its true intentions moving forward. Author: Brennon York Closes #5015 from brennonyork/SPARK-4600 and squashes the following commits: 1e1d1e5 [Brennon York] reverted internal diff docs 92288f7 [Brennon York] reverted both the test suite and the diff function back to its origin functionality f428623 [Brennon York] updated diff documentation to better represent its function cc16d65 [Brennon York] Merge remote-tracking branch 'upstream/master' into SPARK-4600 66818b9 [Brennon York] added small secondary diff test 99ad412 [Brennon York] Merge remote-tracking branch 'upstream/master' into SPARK-4600 74b8c95 [Brennon York] corrected method by leveraging bitmask operations to correctly return only the portions of that are different from the calling VertexRDD 9717120 [Brennon York] updated diff impl to cause fewer objects to be created 710a21c [Brennon York] working diff given test case aa57f83 [Brennon York] updated to set ShortestPaths to run 'forward' rather than 'backward' --- .../src/main/scala/org/apache/spark/graphx/VertexRDD.scala | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/graphx/src/main/scala/org/apache/spark/graphx/VertexRDD.scala b/graphx/src/main/scala/org/apache/spark/graphx/VertexRDD.scala index 09ae3f9f6c09b..40ecff7107109 100644 --- a/graphx/src/main/scala/org/apache/spark/graphx/VertexRDD.scala +++ b/graphx/src/main/scala/org/apache/spark/graphx/VertexRDD.scala @@ -122,8 +122,11 @@ abstract class VertexRDD[VD]( def mapValues[VD2: ClassTag](f: (VertexId, VD) => VD2): VertexRDD[VD2] /** - * Hides vertices that are the same between `this` and `other`; for vertices that are different, - * keeps the values from `other`. + * For each vertex present in both `this` and `other`, `diff` returns only those vertices with + * differing values; for values that are different, keeps the values from `other`. This is + * only guaranteed to work if the VertexRDDs share a common ancestor. + * + * @param other the other VertexRDD with which to diff against. */ def diff(other: VertexRDD[VD]): VertexRDD[VD] From cdc34ed9108688fea32ad170b1ba344fe047716b Mon Sep 17 00:00:00 2001 From: Cheng Lian Date: Sat, 14 Mar 2015 07:09:53 +0800 Subject: [PATCH 055/122] [SPARK-6285] [SQL] Removes unused ParquetTestData and duplicated TestGroupWriteSupport All the contents in this file are not referenced anywhere and should have been removed in #4116 when I tried to get rid of the old Parquet test suites. [Review on Reviewable](https://reviewable.io/reviews/apache/spark/5010) Author: Cheng Lian Closes #5010 from liancheng/spark-6285 and squashes the following commits: 06ed057 [Cheng Lian] Removes unused ParquetTestData and duplicated TestGroupWriteSupport --- .../spark/sql/parquet/ParquetTestData.scala | 466 ------------------ 1 file changed, 466 deletions(-) delete mode 100644 sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTestData.scala diff --git a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTestData.scala b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTestData.scala deleted file mode 100644 index e4a10aa2ae6c3..0000000000000 --- a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTestData.scala +++ /dev/null @@ -1,466 +0,0 @@ -/* - * Licensed to the Apache Software Foundation (ASF) under one or more - * contributor license agreements. See the NOTICE file distributed with - * this work for additional information regarding copyright ownership. - * The ASF licenses this file to You under the Apache License, Version 2.0 - * (the "License"); you may not use this file except in compliance with - * the License. You may obtain a copy of the License at - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package org.apache.spark.sql.parquet - -import java.io.File - -import org.apache.hadoop.conf.Configuration -import org.apache.hadoop.fs.{FileStatus, FileSystem, Path} -import org.apache.hadoop.mapreduce.Job -import org.apache.spark.sql.test.TestSQLContext - -import parquet.example.data.{GroupWriter, Group} -import parquet.example.data.simple.{NanoTime, SimpleGroup} -import parquet.hadoop.{ParquetReader, ParquetFileReader, ParquetWriter} -import parquet.hadoop.api.WriteSupport -import parquet.hadoop.api.WriteSupport.WriteContext -import parquet.hadoop.example.GroupReadSupport -import parquet.hadoop.util.ContextUtil -import parquet.io.api.RecordConsumer -import parquet.schema.{MessageType, MessageTypeParser} - -import org.apache.spark.util.Utils - -// Write support class for nested groups: ParquetWriter initializes GroupWriteSupport -// with an empty configuration (it is after all not intended to be used in this way?) -// and members are private so we need to make our own in order to pass the schema -// to the writer. -private class TestGroupWriteSupport(schema: MessageType) extends WriteSupport[Group] { - var groupWriter: GroupWriter = null - override def prepareForWrite(recordConsumer: RecordConsumer): Unit = { - groupWriter = new GroupWriter(recordConsumer, schema) - } - override def init(configuration: Configuration): WriteContext = { - new WriteContext(schema, new java.util.HashMap[String, String]()) - } - override def write(record: Group) { - groupWriter.write(record) - } -} - -private[sql] object ParquetTestData { - - val testSchema = - """message myrecord { - optional boolean myboolean; - optional int32 myint; - optional binary mystring (UTF8); - optional int64 mylong; - optional float myfloat; - optional double mydouble; - optional int96 mytimestamp; - }""" - - // field names for test assertion error messages - val testSchemaFieldNames = Seq( - "myboolean:Boolean", - "myint:Int", - "mystring:String", - "mylong:Long", - "myfloat:Float", - "mydouble:Double", - "mytimestamp:Timestamp" - ) - - val subTestSchema = - """ - message myrecord { - optional boolean myboolean; - optional int64 mylong; - } - """ - - val testFilterSchema = - """ - message myrecord { - required boolean myboolean; - required int32 myint; - required binary mystring (UTF8); - required int64 mylong; - required float myfloat; - required double mydouble; - optional boolean myoptboolean; - optional int32 myoptint; - optional binary myoptstring (UTF8); - optional int64 myoptlong; - optional float myoptfloat; - optional double myoptdouble; - optional int96 mytimestamp; - } - """ - - // field names for test assertion error messages - val subTestSchemaFieldNames = Seq( - "myboolean:Boolean", - "mylong:Long" - ) - - val testDir = Utils.createTempDir() - val testFilterDir = Utils.createTempDir() - - lazy val testData = new ParquetRelation(testDir.toURI.toString, None, TestSQLContext) - - val testNestedSchema1 = - // based on blogpost example, source: - // https://blog.twitter.com/2013/dremel-made-simple-with-parquet - // note: instead of string we have to use binary (?) otherwise - // Parquet gives us: - // IllegalArgumentException: expected one of [INT64, INT32, BOOLEAN, - // BINARY, FLOAT, DOUBLE, INT96, FIXED_LEN_BYTE_ARRAY] - // Also repeated primitives seem tricky to convert (AvroParquet - // only uses them in arrays?) so only use at most one in each group - // and nothing else in that group (-> is mapped to array)! - // The "values" inside ownerPhoneNumbers is a keyword currently - // so that array types can be translated correctly. - """ - message AddressBook { - required binary owner (UTF8); - optional group ownerPhoneNumbers { - repeated binary array (UTF8); - } - optional group contacts { - repeated group array { - required binary name (UTF8); - optional binary phoneNumber (UTF8); - } - } - } - """ - - - val testNestedSchema2 = - """ - message TestNested2 { - required int32 firstInt; - optional int32 secondInt; - optional group longs { - repeated int64 array; - } - required group entries { - repeated group array { - required double value; - optional boolean truth; - } - } - optional group outerouter { - repeated group array { - repeated group array { - repeated int32 array; - } - } - } - } - """ - - val testNestedSchema3 = - """ - message TestNested3 { - required int32 x; - optional group booleanNumberPairs { - repeated group array { - required int32 key; - optional group value { - repeated group array { - required double nestedValue; - optional boolean truth; - } - } - } - } - } - """ - - val testNestedSchema4 = - """ - message TestNested4 { - required int32 x; - optional group data1 { - repeated group map { - required binary key (UTF8); - required int32 value; - } - } - required group data2 { - repeated group map { - required binary key (UTF8); - required group value { - required int64 payload1; - optional binary payload2 (UTF8); - } - } - } - } - """ - - val testNestedDir1 = Utils.createTempDir() - val testNestedDir2 = Utils.createTempDir() - val testNestedDir3 = Utils.createTempDir() - val testNestedDir4 = Utils.createTempDir() - - lazy val testNestedData1 = - new ParquetRelation(testNestedDir1.toURI.toString, None, TestSQLContext) - lazy val testNestedData2 = - new ParquetRelation(testNestedDir2.toURI.toString, None, TestSQLContext) - - def writeFile() = { - testDir.delete() - val path: Path = new Path(new Path(testDir.toURI), new Path("part-r-0.parquet")) - val job = new Job() - val schema: MessageType = MessageTypeParser.parseMessageType(testSchema) - val writeSupport = new TestGroupWriteSupport(schema) - val writer = new ParquetWriter[Group](path, writeSupport) - - for(i <- 0 until 15) { - val record = new SimpleGroup(schema) - if (i % 3 == 0) { - record.add(0, true) - } else { - record.add(0, false) - } - if (i % 5 == 0) { - record.add(1, 5) - } - record.add(2, "abc") - record.add(3, i.toLong << 33) - record.add(4, 2.5F) - record.add(5, 4.5D) - record.add(6, new NanoTime(1,2)) - writer.write(record) - } - writer.close() - } - - def writeFilterFile(records: Int = 200) = { - // for microbenchmark use: records = 300000000 - testFilterDir.delete - val path: Path = new Path(new Path(testFilterDir.toURI), new Path("part-r-0.parquet")) - val schema: MessageType = MessageTypeParser.parseMessageType(testFilterSchema) - val writeSupport = new TestGroupWriteSupport(schema) - val writer = new ParquetWriter[Group](path, writeSupport) - - for(i <- 0 to records) { - val record = new SimpleGroup(schema) - if (i % 4 == 0) { - record.add(0, true) - } else { - record.add(0, false) - } - record.add(1, i) - record.add(2, i.toString) - record.add(3, i.toLong) - record.add(4, i.toFloat + 0.5f) - record.add(5, i.toDouble + 0.5d) - if (i % 2 == 0) { - if (i % 3 == 0) { - record.add(6, true) - } else { - record.add(6, false) - } - record.add(7, i) - record.add(8, i.toString) - record.add(9, i.toLong) - record.add(10, i.toFloat + 0.5f) - record.add(11, i.toDouble + 0.5d) - } - - writer.write(record) - } - writer.close() - } - - def writeNestedFile1() { - // example data from https://blog.twitter.com/2013/dremel-made-simple-with-parquet - testNestedDir1.delete() - val path: Path = new Path(new Path(testNestedDir1.toURI), new Path("part-r-0.parquet")) - val schema: MessageType = MessageTypeParser.parseMessageType(testNestedSchema1) - - val r1 = new SimpleGroup(schema) - r1.add(0, "Julien Le Dem") - r1.addGroup(1) - .append(CatalystConverter.ARRAY_ELEMENTS_SCHEMA_NAME, "555 123 4567") - .append(CatalystConverter.ARRAY_ELEMENTS_SCHEMA_NAME, "555 666 1337") - .append(CatalystConverter.ARRAY_ELEMENTS_SCHEMA_NAME, "XXX XXX XXXX") - val contacts = r1.addGroup(2) - contacts.addGroup(0) - .append("name", "Dmitriy Ryaboy") - .append("phoneNumber", "555 987 6543") - contacts.addGroup(0) - .append("name", "Chris Aniszczyk") - - val r2 = new SimpleGroup(schema) - r2.add(0, "A. Nonymous") - - val writeSupport = new TestGroupWriteSupport(schema) - val writer = new ParquetWriter[Group](path, writeSupport) - writer.write(r1) - writer.write(r2) - writer.close() - } - - def writeNestedFile2() { - testNestedDir2.delete() - val path: Path = new Path(new Path(testNestedDir2.toURI), new Path("part-r-0.parquet")) - val schema: MessageType = MessageTypeParser.parseMessageType(testNestedSchema2) - - val r1 = new SimpleGroup(schema) - r1.add(0, 1) - r1.add(1, 7) - val longs = r1.addGroup(2) - longs.add(CatalystConverter.ARRAY_ELEMENTS_SCHEMA_NAME , 1.toLong << 32) - longs.add(CatalystConverter.ARRAY_ELEMENTS_SCHEMA_NAME, 1.toLong << 33) - longs.add(CatalystConverter.ARRAY_ELEMENTS_SCHEMA_NAME, 1.toLong << 34) - val booleanNumberPair = r1.addGroup(3).addGroup(0) - booleanNumberPair.add("value", 2.5) - booleanNumberPair.add("truth", false) - val top_level = r1.addGroup(4) - val second_level_a = top_level.addGroup(0) - val second_level_b = top_level.addGroup(0) - val third_level_aa = second_level_a.addGroup(0) - val third_level_ab = second_level_a.addGroup(0) - val third_level_c = second_level_b.addGroup(0) - third_level_aa.add( - CatalystConverter.ARRAY_ELEMENTS_SCHEMA_NAME, - 7) - third_level_ab.add( - CatalystConverter.ARRAY_ELEMENTS_SCHEMA_NAME, - 8) - third_level_c.add( - CatalystConverter.ARRAY_ELEMENTS_SCHEMA_NAME, - 9) - - val writeSupport = new TestGroupWriteSupport(schema) - val writer = new ParquetWriter[Group](path, writeSupport) - writer.write(r1) - writer.close() - } - - def writeNestedFile3() { - testNestedDir3.delete() - val path: Path = new Path(new Path(testNestedDir3.toURI), new Path("part-r-0.parquet")) - val schema: MessageType = MessageTypeParser.parseMessageType(testNestedSchema3) - - val r1 = new SimpleGroup(schema) - r1.add(0, 1) - val booleanNumberPairs = r1.addGroup(1) - val g1 = booleanNumberPairs.addGroup(0) - g1.add(0, 1) - val nested1 = g1.addGroup(1) - val ng1 = nested1.addGroup(0) - ng1.add(0, 1.5) - ng1.add(1, false) - val ng2 = nested1.addGroup(0) - ng2.add(0, 2.5) - ng2.add(1, true) - val g2 = booleanNumberPairs.addGroup(0) - g2.add(0, 2) - val ng3 = g2.addGroup(1) - .addGroup(0) - ng3.add(0, 3.5) - ng3.add(1, false) - - val writeSupport = new TestGroupWriteSupport(schema) - val writer = new ParquetWriter[Group](path, writeSupport) - writer.write(r1) - writer.close() - } - - def writeNestedFile4() { - testNestedDir4.delete() - val path: Path = new Path(new Path(testNestedDir4.toURI), new Path("part-r-0.parquet")) - val schema: MessageType = MessageTypeParser.parseMessageType(testNestedSchema4) - - val r1 = new SimpleGroup(schema) - r1.add(0, 7) - val map1 = r1.addGroup(1) - val keyValue1 = map1.addGroup(0) - keyValue1.add(0, "key1") - keyValue1.add(1, 1) - val keyValue2 = map1.addGroup(0) - keyValue2.add(0, "key2") - keyValue2.add(1, 2) - val map2 = r1.addGroup(2) - val keyValue3 = map2.addGroup(0) - // TODO: currently only string key type supported - keyValue3.add(0, "seven") - val valueGroup1 = keyValue3.addGroup(1) - valueGroup1.add(0, 42.toLong) - valueGroup1.add(1, "the answer") - val keyValue4 = map2.addGroup(0) - // TODO: currently only string key type supported - keyValue4.add(0, "eight") - val valueGroup2 = keyValue4.addGroup(1) - valueGroup2.add(0, 49.toLong) - - val writeSupport = new TestGroupWriteSupport(schema) - val writer = new ParquetWriter[Group](path, writeSupport) - writer.write(r1) - writer.close() - } - - // TODO: this is not actually used anywhere but useful for debugging - /* def readNestedFile(file: File, schemaString: String): Unit = { - val configuration = new Configuration() - val path = new Path(new Path(file.toURI), new Path("part-r-0.parquet")) - val fs: FileSystem = path.getFileSystem(configuration) - val schema: MessageType = MessageTypeParser.parseMessageType(schemaString) - assert(schema != null) - val outputStatus: FileStatus = fs.getFileStatus(new Path(path.toString)) - val footers = ParquetFileReader.readFooter(configuration, outputStatus) - assert(footers != null) - val reader = new ParquetReader(new Path(path.toString), new GroupReadSupport()) - val first = reader.read() - assert(first != null) - } */ - - // to test golb pattern (wild card pattern matching for parquetFile input - val testGlobDir = Utils.createTempDir() - val testGlobSubDir1 = Utils.createTempDir(testGlobDir.getPath) - val testGlobSubDir2 = Utils.createTempDir(testGlobDir.getPath) - val testGlobSubDir3 = Utils.createTempDir(testGlobDir.getPath) - - def writeGlobFiles() = { - val subDirs = Array(testGlobSubDir1, testGlobSubDir2, testGlobSubDir3) - - subDirs.foreach { dir => - val path: Path = new Path(new Path(dir.toURI), new Path("part-r-0.parquet")) - val job = new Job() - val schema: MessageType = MessageTypeParser.parseMessageType(testSchema) - val writeSupport = new TestGroupWriteSupport(schema) - val writer = new ParquetWriter[Group](path, writeSupport) - - for(i <- 0 until 15) { - val record = new SimpleGroup(schema) - if(i % 3 == 0) { - record.add(0, true) - } else { - record.add(0, false) - } - if(i % 5 == 0) { - record.add(1, 5) - } - record.add(2, "abc") - record.add(3, i.toLong << 33) - record.add(4, 2.5F) - record.add(5, 4.5D) - writer.write(record) - } - writer.close() - } - } -} - From e360d5e4adf287444c10e72f8e4d57548839bf6e Mon Sep 17 00:00:00 2001 From: vinodkc Date: Sat, 14 Mar 2015 07:17:54 +0800 Subject: [PATCH 056/122] [SPARK-6317][SQL]Fixed HIVE console startup issue Author: vinodkc Author: Vinod K C Closes #5011 from vinodkc/HIVE_console_startupError and squashes the following commits: b43925f [vinodkc] Changed order of import b4f5453 [Vinod K C] Fixed HIVE console startup issue --- project/SparkBuild.scala | 4 ++-- sql/README.md | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/project/SparkBuild.scala b/project/SparkBuild.scala index 4a06b9821bb98..f4c74c4051014 100644 --- a/project/SparkBuild.scala +++ b/project/SparkBuild.scala @@ -269,8 +269,8 @@ object SQL { |import org.apache.spark.sql.catalyst.plans.logical._ |import org.apache.spark.sql.catalyst.rules._ |import org.apache.spark.sql.catalyst.util._ - |import org.apache.spark.sql.Dsl._ |import org.apache.spark.sql.execution + |import org.apache.spark.sql.functions._ |import org.apache.spark.sql.test.TestSQLContext._ |import org.apache.spark.sql.types._ |import org.apache.spark.sql.parquet.ParquetTestData""".stripMargin, @@ -300,8 +300,8 @@ object Hive { |import org.apache.spark.sql.catalyst.plans.logical._ |import org.apache.spark.sql.catalyst.rules._ |import org.apache.spark.sql.catalyst.util._ - |import org.apache.spark.sql.Dsl._ |import org.apache.spark.sql.execution + |import org.apache.spark.sql.functions._ |import org.apache.spark.sql.hive._ |import org.apache.spark.sql.hive.test.TestHive._ |import org.apache.spark.sql.types._ diff --git a/sql/README.md b/sql/README.md index a79249965ee67..48f83340e37b3 100644 --- a/sql/README.md +++ b/sql/README.md @@ -36,8 +36,8 @@ import org.apache.spark.sql.catalyst.expressions._ import org.apache.spark.sql.catalyst.plans.logical._ import org.apache.spark.sql.catalyst.rules._ import org.apache.spark.sql.catalyst.util._ -import org.apache.spark.sql.Dsl._ import org.apache.spark.sql.execution +import org.apache.spark.sql.functions._ import org.apache.spark.sql.hive._ import org.apache.spark.sql.hive.test.TestHive._ import org.apache.spark.sql.types._ From b38e073fee794188d5267f1812b095e51874839e Mon Sep 17 00:00:00 2001 From: Davies Liu Date: Sat, 14 Mar 2015 00:43:33 -0700 Subject: [PATCH 057/122] [SPARK-6210] [SQL] use prettyString as column name in agg() use prettyString instead of toString() (which include id of expression) as column name in agg() Author: Davies Liu Closes #5006 from davies/prettystring and squashes the following commits: cb1fdcf [Davies Liu] use prettyString as column name in agg() --- python/pyspark/sql/dataframe.py | 32 +++++++++---------- .../spark/sql/catalyst/trees/TreeNode.scala | 2 +- .../org/apache/spark/sql/GroupedData.scala | 8 ++--- 3 files changed, 21 insertions(+), 21 deletions(-) diff --git a/python/pyspark/sql/dataframe.py b/python/pyspark/sql/dataframe.py index e8ce4547455a5..94001aec3774b 100644 --- a/python/pyspark/sql/dataframe.py +++ b/python/pyspark/sql/dataframe.py @@ -631,11 +631,11 @@ def groupBy(self, *cols): for all the available aggregate functions. >>> df.groupBy().avg().collect() - [Row(AVG(age#0)=3.5)] + [Row(AVG(age)=3.5)] >>> df.groupBy('name').agg({'age': 'mean'}).collect() - [Row(name=u'Bob', AVG(age#0)=5.0), Row(name=u'Alice', AVG(age#0)=2.0)] + [Row(name=u'Bob', AVG(age)=5.0), Row(name=u'Alice', AVG(age)=2.0)] >>> df.groupBy(df.name).avg().collect() - [Row(name=u'Bob', AVG(age#0)=5.0), Row(name=u'Alice', AVG(age#0)=2.0)] + [Row(name=u'Bob', AVG(age)=5.0), Row(name=u'Alice', AVG(age)=2.0)] """ jcols = ListConverter().convert([_to_java_column(c) for c in cols], self._sc._gateway._gateway_client) @@ -647,10 +647,10 @@ def agg(self, *exprs): (shorthand for df.groupBy.agg()). >>> df.agg({"age": "max"}).collect() - [Row(MAX(age#0)=5)] + [Row(MAX(age)=5)] >>> from pyspark.sql import functions as F >>> df.agg(F.min(df.age)).collect() - [Row(MIN(age#0)=2)] + [Row(MIN(age)=2)] """ return self.groupBy().agg(*exprs) @@ -766,7 +766,7 @@ def agg(self, *exprs): >>> from pyspark.sql import functions as F >>> gdf.agg(F.min(df.age)).collect() - [Row(MIN(age#0)=5), Row(MIN(age#0)=2)] + [Row(MIN(age)=5), Row(MIN(age)=2)] """ assert exprs, "exprs should not be empty" if len(exprs) == 1 and isinstance(exprs[0], dict): @@ -795,9 +795,9 @@ def mean(self, *cols): for each group. This is an alias for `avg`. >>> df.groupBy().mean('age').collect() - [Row(AVG(age#0)=3.5)] + [Row(AVG(age)=3.5)] >>> df3.groupBy().mean('age', 'height').collect() - [Row(AVG(age#4L)=3.5, AVG(height#5L)=82.5)] + [Row(AVG(age)=3.5, AVG(height)=82.5)] """ @df_varargs_api @@ -806,9 +806,9 @@ def avg(self, *cols): for each group. >>> df.groupBy().avg('age').collect() - [Row(AVG(age#0)=3.5)] + [Row(AVG(age)=3.5)] >>> df3.groupBy().avg('age', 'height').collect() - [Row(AVG(age#4L)=3.5, AVG(height#5L)=82.5)] + [Row(AVG(age)=3.5, AVG(height)=82.5)] """ @df_varargs_api @@ -817,9 +817,9 @@ def max(self, *cols): each group. >>> df.groupBy().max('age').collect() - [Row(MAX(age#0)=5)] + [Row(MAX(age)=5)] >>> df3.groupBy().max('age', 'height').collect() - [Row(MAX(age#4L)=5, MAX(height#5L)=85)] + [Row(MAX(age)=5, MAX(height)=85)] """ @df_varargs_api @@ -828,9 +828,9 @@ def min(self, *cols): each group. >>> df.groupBy().min('age').collect() - [Row(MIN(age#0)=2)] + [Row(MIN(age)=2)] >>> df3.groupBy().min('age', 'height').collect() - [Row(MIN(age#4L)=2, MIN(height#5L)=80)] + [Row(MIN(age)=2, MIN(height)=80)] """ @df_varargs_api @@ -839,9 +839,9 @@ def sum(self, *cols): group. >>> df.groupBy().sum('age').collect() - [Row(SUM(age#0)=7)] + [Row(SUM(age)=7)] >>> df3.groupBy().sum('age', 'height').collect() - [Row(SUM(age#4L)=7, SUM(height#5L)=165)] + [Row(SUM(age)=7, SUM(height)=165)] """ diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/trees/TreeNode.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/trees/TreeNode.scala index 7e191ad0315a5..f84ffe4e176cc 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/trees/TreeNode.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/trees/TreeNode.scala @@ -343,7 +343,7 @@ abstract class TreeNode[BaseType <: TreeNode[BaseType]] { }.mkString(", ") /** String representation of this node without any children */ - def simpleString = s"$nodeName $argString" + def simpleString = s"$nodeName $argString".trim override def toString: String = treeString diff --git a/sql/core/src/main/scala/org/apache/spark/sql/GroupedData.scala b/sql/core/src/main/scala/org/apache/spark/sql/GroupedData.scala index d00175265924c..45a63ae26ed71 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/GroupedData.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/GroupedData.scala @@ -37,7 +37,7 @@ class GroupedData protected[sql](df: DataFrame, groupingExprs: Seq[Expression]) private[this] implicit def toDF(aggExprs: Seq[NamedExpression]): DataFrame = { val namedGroupingExprs = groupingExprs.map { case expr: NamedExpression => expr - case expr: Expression => Alias(expr, expr.toString)() + case expr: Expression => Alias(expr, expr.prettyString)() } DataFrame( df.sqlContext, Aggregate(groupingExprs, namedGroupingExprs ++ aggExprs, df.logicalPlan)) @@ -63,7 +63,7 @@ class GroupedData protected[sql](df: DataFrame, groupingExprs: Seq[Expression]) } columnExprs.map { c => val a = f(c) - Alias(a, a.toString)() + Alias(a, a.prettyString)() } } @@ -115,7 +115,7 @@ class GroupedData protected[sql](df: DataFrame, groupingExprs: Seq[Expression]) def agg(exprs: Map[String, String]): DataFrame = { exprs.map { case (colName, expr) => val a = strToExpr(expr)(df(colName).expr) - Alias(a, a.toString)() + Alias(a, a.prettyString)() }.toSeq } @@ -159,7 +159,7 @@ class GroupedData protected[sql](df: DataFrame, groupingExprs: Seq[Expression]) def agg(expr: Column, exprs: Column*): DataFrame = { val aggExprs = (expr +: exprs).map(_.expr).map { case expr: NamedExpression => expr - case expr: Expression => Alias(expr, expr.toString)() + case expr: Expression => Alias(expr, expr.prettyString)() } DataFrame(df.sqlContext, Aggregate(groupingExprs, aggExprs, df.logicalPlan)) } From ee15404a2b0009fc70119ac7af69137b54890d48 Mon Sep 17 00:00:00 2001 From: ArcherShao Date: Sat, 14 Mar 2015 08:27:18 +0000 Subject: [PATCH 058/122] [SQL]Delete some dupliate code in HiveThriftServer2 Author: ArcherShao Author: ArcherShao Closes #5007 from ArcherShao/20150313 and squashes the following commits: ae422ae [ArcherShao] Updated 459efbd [ArcherShao] [SQL]Delete some dupliate code in HiveThriftServer2 --- .../sql/hive/thriftserver/HiveThriftServer2.scala | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-) diff --git a/sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2.scala b/sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2.scala index 6e07df18b0e15..c3a3f8c0f41df 100644 --- a/sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2.scala +++ b/sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2.scala @@ -98,16 +98,14 @@ private[hive] class HiveThriftServer2(hiveContext: HiveContext) setSuperField(this, "cliService", sparkSqlCliService) addService(sparkSqlCliService) - if (isHTTPTransportMode(hiveConf)) { - val thriftCliService = new ThriftHttpCLIService(sparkSqlCliService) - setSuperField(this, "thriftCLIService", thriftCliService) - addService(thriftCliService) + val thriftCliService = if (isHTTPTransportMode(hiveConf)) { + new ThriftHttpCLIService(sparkSqlCliService) } else { - val thriftCliService = new ThriftBinaryCLIService(sparkSqlCliService) - setSuperField(this, "thriftCLIService", thriftCliService) - addService(thriftCliService) + new ThriftBinaryCLIService(sparkSqlCliService) } + setSuperField(this, "thriftCLIService", thriftCliService) + addService(thriftCliService) initCompositeService(hiveConf) } From 5be6b0e4f48aca12fcd47c1b77c4675ad651c332 Mon Sep 17 00:00:00 2001 From: Cheng Lian Date: Sat, 14 Mar 2015 19:53:54 +0800 Subject: [PATCH 059/122] [SPARK-6195] [SQL] Adds in-memory column type for fixed-precision decimals This PR adds a specialized in-memory column type for fixed-precision decimals. For all other column types, a single integer column type ID is enough to determine which column type to use. However, this doesn't apply to fixed-precision decimal types with different precision and scale parameters. Moreover, according to the previous design, there seems no trivial way to encode precision and scale information into the columnar byte buffer. On the other hand, considering we always know the data type of the column to be built / scanned ahead of time. This PR no longer use column type ID to construct `ColumnBuilder`s and `ColumnAccessor`s, but resorts to the actual column data type. In this way, we can pass precision / scale information along the way. The column type ID is now not used anymore and can be removed in a future PR. ### Micro benchmark result The following micro benchmark builds a simple table with 2 million decimals (precision = 10, scale = 0), cache it in memory, then count all the rows. Code (simply paste it into Spark shell): ```scala import sc._ import sqlContext._ import sqlContext.implicits._ import org.apache.spark.sql.types._ import com.google.common.base.Stopwatch def benchmark(n: Int)(f: => Long) { val stopwatch = new Stopwatch() def run() = { stopwatch.reset() stopwatch.start() f stopwatch.stop() stopwatch.elapsedMillis() } val records = (0 until n).map(_ => run()) (0 until n).foreach(i => println(s"Round $i: ${records(i)} ms")) println(s"Average: ${records.sum / n.toDouble} ms") } // Explicit casting is required because ScalaReflection can't inspect decimal precision parallelize(1 to 2000000) .map(i => Tuple1(Decimal(i, 10, 0))) .toDF("dec") .select($"dec" cast DecimalType(10, 0)) .registerTempTable("dec") sql("CACHE TABLE dec") val df = table("dec") // Warm up df.count() df.count() benchmark(5) { df.count() } ``` With `FIXED_DECIMAL` column type: - Round 0: 75 ms - Round 1: 97 ms - Round 2: 75 ms - Round 3: 70 ms - Round 4: 72 ms - Average: 77.8 ms Without `FIXED_DECIMAL` column type: - Round 0: 1233 ms - Round 1: 1170 ms - Round 2: 1171 ms - Round 3: 1141 ms - Round 4: 1141 ms - Average: 1171.2 ms [Review on Reviewable](https://reviewable.io/reviews/apache/spark/4938) Author: Cheng Lian Closes #4938 from liancheng/decimal-column-type and squashes the following commits: fef5338 [Cheng Lian] Updates fixed decimal column type related test cases e08ab5b [Cheng Lian] Only resorts to FIXED_DECIMAL when the value can be held in a long 4db713d [Cheng Lian] Adds in-memory column type for fixed-precision decimals --- .../spark/sql/columnar/ColumnAccessor.scala | 43 +++++++++------ .../spark/sql/columnar/ColumnBuilder.scala | 39 +++++++------ .../spark/sql/columnar/ColumnStats.scala | 17 ++++++ .../spark/sql/columnar/ColumnType.scala | 55 ++++++++++++++----- .../columnar/InMemoryColumnarTableScan.scala | 8 ++- .../spark/sql/columnar/ColumnStatsSuite.scala | 1 + .../spark/sql/columnar/ColumnTypeSuite.scala | 46 ++++++++++++---- .../sql/columnar/ColumnarTestUtils.scala | 23 ++++---- .../columnar/InMemoryColumnarQuerySuite.scala | 17 +++++- .../NullableColumnAccessorSuite.scala | 3 +- .../columnar/NullableColumnBuilderSuite.scala | 3 +- 11 files changed, 179 insertions(+), 76 deletions(-) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnAccessor.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnAccessor.scala index 91c4c105b14e6..b615eaa0dca0d 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnAccessor.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnAccessor.scala @@ -21,7 +21,7 @@ import java.nio.{ByteBuffer, ByteOrder} import org.apache.spark.sql.catalyst.expressions.MutableRow import org.apache.spark.sql.columnar.compression.CompressibleColumnAccessor -import org.apache.spark.sql.types.{BinaryType, DataType, NativeType} +import org.apache.spark.sql.types._ /** * An `Iterator` like trait used to extract values from columnar byte buffer. When a value is @@ -89,6 +89,9 @@ private[sql] class DoubleColumnAccessor(buffer: ByteBuffer) private[sql] class FloatColumnAccessor(buffer: ByteBuffer) extends NativeColumnAccessor(buffer, FLOAT) +private[sql] class FixedDecimalColumnAccessor(buffer: ByteBuffer, precision: Int, scale: Int) + extends NativeColumnAccessor(buffer, FIXED_DECIMAL(precision, scale)) + private[sql] class StringColumnAccessor(buffer: ByteBuffer) extends NativeColumnAccessor(buffer, STRING) @@ -107,24 +110,28 @@ private[sql] class GenericColumnAccessor(buffer: ByteBuffer) with NullableColumnAccessor private[sql] object ColumnAccessor { - def apply(buffer: ByteBuffer): ColumnAccessor = { + def apply(dataType: DataType, buffer: ByteBuffer): ColumnAccessor = { val dup = buffer.duplicate().order(ByteOrder.nativeOrder) - // The first 4 bytes in the buffer indicate the column type. - val columnTypeId = dup.getInt() - - columnTypeId match { - case INT.typeId => new IntColumnAccessor(dup) - case LONG.typeId => new LongColumnAccessor(dup) - case FLOAT.typeId => new FloatColumnAccessor(dup) - case DOUBLE.typeId => new DoubleColumnAccessor(dup) - case BOOLEAN.typeId => new BooleanColumnAccessor(dup) - case BYTE.typeId => new ByteColumnAccessor(dup) - case SHORT.typeId => new ShortColumnAccessor(dup) - case STRING.typeId => new StringColumnAccessor(dup) - case DATE.typeId => new DateColumnAccessor(dup) - case TIMESTAMP.typeId => new TimestampColumnAccessor(dup) - case BINARY.typeId => new BinaryColumnAccessor(dup) - case GENERIC.typeId => new GenericColumnAccessor(dup) + + // The first 4 bytes in the buffer indicate the column type. This field is not used now, + // because we always know the data type of the column ahead of time. + dup.getInt() + + dataType match { + case IntegerType => new IntColumnAccessor(dup) + case LongType => new LongColumnAccessor(dup) + case FloatType => new FloatColumnAccessor(dup) + case DoubleType => new DoubleColumnAccessor(dup) + case BooleanType => new BooleanColumnAccessor(dup) + case ByteType => new ByteColumnAccessor(dup) + case ShortType => new ShortColumnAccessor(dup) + case StringType => new StringColumnAccessor(dup) + case BinaryType => new BinaryColumnAccessor(dup) + case DateType => new DateColumnAccessor(dup) + case TimestampType => new TimestampColumnAccessor(dup) + case DecimalType.Fixed(precision, scale) if precision < 19 => + new FixedDecimalColumnAccessor(dup, precision, scale) + case _ => new GenericColumnAccessor(dup) } } } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnBuilder.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnBuilder.scala index 3a4977b836af7..d8d24a577347c 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnBuilder.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnBuilder.scala @@ -106,6 +106,13 @@ private[sql] class DoubleColumnBuilder extends NativeColumnBuilder(new DoubleCol private[sql] class FloatColumnBuilder extends NativeColumnBuilder(new FloatColumnStats, FLOAT) +private[sql] class FixedDecimalColumnBuilder( + precision: Int, + scale: Int) + extends NativeColumnBuilder( + new FixedDecimalColumnStats, + FIXED_DECIMAL(precision, scale)) + private[sql] class StringColumnBuilder extends NativeColumnBuilder(new StringColumnStats, STRING) private[sql] class DateColumnBuilder extends NativeColumnBuilder(new DateColumnStats, DATE) @@ -139,25 +146,25 @@ private[sql] object ColumnBuilder { } def apply( - typeId: Int, + dataType: DataType, initialSize: Int = 0, columnName: String = "", useCompression: Boolean = false): ColumnBuilder = { - - val builder = (typeId match { - case INT.typeId => new IntColumnBuilder - case LONG.typeId => new LongColumnBuilder - case FLOAT.typeId => new FloatColumnBuilder - case DOUBLE.typeId => new DoubleColumnBuilder - case BOOLEAN.typeId => new BooleanColumnBuilder - case BYTE.typeId => new ByteColumnBuilder - case SHORT.typeId => new ShortColumnBuilder - case STRING.typeId => new StringColumnBuilder - case BINARY.typeId => new BinaryColumnBuilder - case GENERIC.typeId => new GenericColumnBuilder - case DATE.typeId => new DateColumnBuilder - case TIMESTAMP.typeId => new TimestampColumnBuilder - }).asInstanceOf[ColumnBuilder] + val builder: ColumnBuilder = dataType match { + case IntegerType => new IntColumnBuilder + case LongType => new LongColumnBuilder + case DoubleType => new DoubleColumnBuilder + case BooleanType => new BooleanColumnBuilder + case ByteType => new ByteColumnBuilder + case ShortType => new ShortColumnBuilder + case StringType => new StringColumnBuilder + case BinaryType => new BinaryColumnBuilder + case DateType => new DateColumnBuilder + case TimestampType => new TimestampColumnBuilder + case DecimalType.Fixed(precision, scale) if precision < 19 => + new FixedDecimalColumnBuilder(precision, scale) + case _ => new GenericColumnBuilder + } builder.initialize(initialSize, columnName, useCompression) builder diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnStats.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnStats.scala index cad0667b46435..04047b9c062be 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnStats.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnStats.scala @@ -181,6 +181,23 @@ private[sql] class FloatColumnStats extends ColumnStats { def collectedStatistics = Row(lower, upper, nullCount, count, sizeInBytes) } +private[sql] class FixedDecimalColumnStats extends ColumnStats { + protected var upper: Decimal = null + protected var lower: Decimal = null + + override def gatherStats(row: Row, ordinal: Int): Unit = { + super.gatherStats(row, ordinal) + if (!row.isNullAt(ordinal)) { + val value = row(ordinal).asInstanceOf[Decimal] + if (upper == null || value.compareTo(upper) > 0) upper = value + if (lower == null || value.compareTo(lower) < 0) lower = value + sizeInBytes += FIXED_DECIMAL.defaultSize + } + } + + override def collectedStatistics: Row = Row(lower, upper, nullCount, count, sizeInBytes) +} + private[sql] class IntColumnStats extends ColumnStats { protected var upper = Int.MinValue protected var lower = Int.MaxValue diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnType.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnType.scala index db5bc0de363c7..36ea1c77e0470 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnType.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnType.scala @@ -373,6 +373,33 @@ private[sql] object TIMESTAMP extends NativeColumnType(TimestampType, 9, 12) { } } +private[sql] case class FIXED_DECIMAL(precision: Int, scale: Int) + extends NativeColumnType( + DecimalType(Some(PrecisionInfo(precision, scale))), + 10, + FIXED_DECIMAL.defaultSize) { + + override def extract(buffer: ByteBuffer): Decimal = { + Decimal(buffer.getLong(), precision, scale) + } + + override def append(v: Decimal, buffer: ByteBuffer): Unit = { + buffer.putLong(v.toUnscaledLong) + } + + override def getField(row: Row, ordinal: Int): Decimal = { + row(ordinal).asInstanceOf[Decimal] + } + + override def setField(row: MutableRow, ordinal: Int, value: Decimal): Unit = { + row(ordinal) = value + } +} + +private[sql] object FIXED_DECIMAL { + val defaultSize = 8 +} + private[sql] sealed abstract class ByteArrayColumnType[T <: DataType]( typeId: Int, defaultSize: Int) @@ -394,7 +421,7 @@ private[sql] sealed abstract class ByteArrayColumnType[T <: DataType]( } } -private[sql] object BINARY extends ByteArrayColumnType[BinaryType.type](10, 16) { +private[sql] object BINARY extends ByteArrayColumnType[BinaryType.type](11, 16) { override def setField(row: MutableRow, ordinal: Int, value: Array[Byte]): Unit = { row(ordinal) = value } @@ -405,7 +432,7 @@ private[sql] object BINARY extends ByteArrayColumnType[BinaryType.type](10, 16) // Used to process generic objects (all types other than those listed above). Objects should be // serialized first before appending to the column `ByteBuffer`, and is also extracted as serialized // byte array. -private[sql] object GENERIC extends ByteArrayColumnType[DataType](11, 16) { +private[sql] object GENERIC extends ByteArrayColumnType[DataType](12, 16) { override def setField(row: MutableRow, ordinal: Int, value: Array[Byte]): Unit = { row(ordinal) = SparkSqlSerializer.deserialize[Any](value) } @@ -416,18 +443,20 @@ private[sql] object GENERIC extends ByteArrayColumnType[DataType](11, 16) { private[sql] object ColumnType { def apply(dataType: DataType): ColumnType[_, _] = { dataType match { - case IntegerType => INT - case LongType => LONG - case FloatType => FLOAT - case DoubleType => DOUBLE - case BooleanType => BOOLEAN - case ByteType => BYTE - case ShortType => SHORT - case StringType => STRING - case BinaryType => BINARY - case DateType => DATE + case IntegerType => INT + case LongType => LONG + case FloatType => FLOAT + case DoubleType => DOUBLE + case BooleanType => BOOLEAN + case ByteType => BYTE + case ShortType => SHORT + case StringType => STRING + case BinaryType => BINARY + case DateType => DATE case TimestampType => TIMESTAMP - case _ => GENERIC + case DecimalType.Fixed(precision, scale) if precision < 19 => + FIXED_DECIMAL(precision, scale) + case _ => GENERIC } } } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/InMemoryColumnarTableScan.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/InMemoryColumnarTableScan.scala index 8944a32bc3887..387faee12b3cd 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/InMemoryColumnarTableScan.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/InMemoryColumnarTableScan.scala @@ -113,7 +113,7 @@ private[sql] case class InMemoryRelation( val columnBuilders = output.map { attribute => val columnType = ColumnType(attribute.dataType) val initialBufferSize = columnType.defaultSize * batchSize - ColumnBuilder(columnType.typeId, initialBufferSize, attribute.name, useCompression) + ColumnBuilder(attribute.dataType, initialBufferSize, attribute.name, useCompression) }.toArray var rowCount = 0 @@ -274,8 +274,10 @@ private[sql] case class InMemoryColumnarTableScan( def cachedBatchesToRows(cacheBatches: Iterator[CachedBatch]) = { val rows = cacheBatches.flatMap { cachedBatch => // Build column accessors - val columnAccessors = requestedColumnIndices.map { batch => - ColumnAccessor(ByteBuffer.wrap(cachedBatch.buffers(batch))) + val columnAccessors = requestedColumnIndices.map { batchColumnIndex => + ColumnAccessor( + relation.output(batchColumnIndex).dataType, + ByteBuffer.wrap(cachedBatch.buffers(batchColumnIndex))) } // Extract rows via column accessors diff --git a/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnStatsSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnStatsSuite.scala index 581fccf8ee613..fec487f1d2c82 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnStatsSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnStatsSuite.scala @@ -29,6 +29,7 @@ class ColumnStatsSuite extends FunSuite { testColumnStats(classOf[LongColumnStats], LONG, Row(Long.MaxValue, Long.MinValue, 0)) testColumnStats(classOf[FloatColumnStats], FLOAT, Row(Float.MaxValue, Float.MinValue, 0)) testColumnStats(classOf[DoubleColumnStats], DOUBLE, Row(Double.MaxValue, Double.MinValue, 0)) + testColumnStats(classOf[FixedDecimalColumnStats], FIXED_DECIMAL(15, 10), Row(null, null, 0)) testColumnStats(classOf[StringColumnStats], STRING, Row(null, null, 0)) testColumnStats(classOf[DateColumnStats], DATE, Row(Int.MaxValue, Int.MinValue, 0)) testColumnStats(classOf[TimestampColumnStats], TIMESTAMP, Row(null, null, 0)) diff --git a/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnTypeSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnTypeSuite.scala index 9ce845912f1c7..5f08834f73c6b 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnTypeSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnTypeSuite.scala @@ -33,8 +33,9 @@ class ColumnTypeSuite extends FunSuite with Logging { test("defaultSize") { val checks = Map( - INT -> 4, SHORT -> 2, LONG -> 8, BYTE -> 1, DOUBLE -> 8, FLOAT -> 4, BOOLEAN -> 1, - STRING -> 8, DATE -> 4, TIMESTAMP -> 12, BINARY -> 16, GENERIC -> 16) + INT -> 4, SHORT -> 2, LONG -> 8, BYTE -> 1, DOUBLE -> 8, FLOAT -> 4, + FIXED_DECIMAL(15, 10) -> 8, BOOLEAN -> 1, STRING -> 8, DATE -> 4, TIMESTAMP -> 12, + BINARY -> 16, GENERIC -> 16) checks.foreach { case (columnType, expectedSize) => assertResult(expectedSize, s"Wrong defaultSize for $columnType") { @@ -56,15 +57,16 @@ class ColumnTypeSuite extends FunSuite with Logging { } } - checkActualSize(INT, Int.MaxValue, 4) - checkActualSize(SHORT, Short.MaxValue, 2) - checkActualSize(LONG, Long.MaxValue, 8) - checkActualSize(BYTE, Byte.MaxValue, 1) - checkActualSize(DOUBLE, Double.MaxValue, 8) - checkActualSize(FLOAT, Float.MaxValue, 4) - checkActualSize(BOOLEAN, true, 1) - checkActualSize(STRING, "hello", 4 + "hello".getBytes("utf-8").length) - checkActualSize(DATE, 0, 4) + checkActualSize(INT, Int.MaxValue, 4) + checkActualSize(SHORT, Short.MaxValue, 2) + checkActualSize(LONG, Long.MaxValue, 8) + checkActualSize(BYTE, Byte.MaxValue, 1) + checkActualSize(DOUBLE, Double.MaxValue, 8) + checkActualSize(FLOAT, Float.MaxValue, 4) + checkActualSize(FIXED_DECIMAL(15, 10), Decimal(0, 15, 10), 8) + checkActualSize(BOOLEAN, true, 1) + checkActualSize(STRING, "hello", 4 + "hello".getBytes("utf-8").length) + checkActualSize(DATE, 0, 4) checkActualSize(TIMESTAMP, new Timestamp(0L), 12) val binary = Array.fill[Byte](4)(0: Byte) @@ -93,12 +95,20 @@ class ColumnTypeSuite extends FunSuite with Logging { testNativeColumnType[DoubleType.type](DOUBLE, _.putDouble(_), _.getDouble) + testNativeColumnType[DecimalType]( + FIXED_DECIMAL(15, 10), + (buffer: ByteBuffer, decimal: Decimal) => { + buffer.putLong(decimal.toUnscaledLong) + }, + (buffer: ByteBuffer) => { + Decimal(buffer.getLong(), 15, 10) + }) + testNativeColumnType[FloatType.type](FLOAT, _.putFloat(_), _.getFloat) testNativeColumnType[StringType.type]( STRING, (buffer: ByteBuffer, string: String) => { - val bytes = string.getBytes("utf-8") buffer.putInt(bytes.length) buffer.put(bytes) @@ -206,4 +216,16 @@ class ColumnTypeSuite extends FunSuite with Logging { if (sb.nonEmpty) sb.setLength(sb.length - 1) sb.toString() } + + test("column type for decimal types with different precision") { + (1 to 18).foreach { i => + assertResult(FIXED_DECIMAL(i, 0)) { + ColumnType(DecimalType(i, 0)) + } + } + + assertResult(GENERIC) { + ColumnType(DecimalType(19, 0)) + } + } } diff --git a/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnarTestUtils.scala b/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnarTestUtils.scala index 60ed28cc97bf1..c7a40845db16c 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnarTestUtils.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/columnar/ColumnarTestUtils.scala @@ -24,7 +24,7 @@ import scala.util.Random import org.apache.spark.sql.Row import org.apache.spark.sql.catalyst.expressions.GenericMutableRow -import org.apache.spark.sql.types.{DataType, NativeType} +import org.apache.spark.sql.types.{Decimal, DataType, NativeType} object ColumnarTestUtils { def makeNullRow(length: Int) = { @@ -41,16 +41,17 @@ object ColumnarTestUtils { } (columnType match { - case BYTE => (Random.nextInt(Byte.MaxValue * 2) - Byte.MaxValue).toByte - case SHORT => (Random.nextInt(Short.MaxValue * 2) - Short.MaxValue).toShort - case INT => Random.nextInt() - case LONG => Random.nextLong() - case FLOAT => Random.nextFloat() - case DOUBLE => Random.nextDouble() - case STRING => Random.nextString(Random.nextInt(32)) - case BOOLEAN => Random.nextBoolean() - case BINARY => randomBytes(Random.nextInt(32)) - case DATE => Random.nextInt() + case BYTE => (Random.nextInt(Byte.MaxValue * 2) - Byte.MaxValue).toByte + case SHORT => (Random.nextInt(Short.MaxValue * 2) - Short.MaxValue).toShort + case INT => Random.nextInt() + case LONG => Random.nextLong() + case FLOAT => Random.nextFloat() + case DOUBLE => Random.nextDouble() + case FIXED_DECIMAL(precision, scale) => Decimal(Random.nextLong() % 100, precision, scale) + case STRING => Random.nextString(Random.nextInt(32)) + case BOOLEAN => Random.nextBoolean() + case BINARY => randomBytes(Random.nextInt(32)) + case DATE => Random.nextInt() case TIMESTAMP => val timestamp = new Timestamp(Random.nextLong()) timestamp.setNanos(Random.nextInt(999999999)) diff --git a/sql/core/src/test/scala/org/apache/spark/sql/columnar/InMemoryColumnarQuerySuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/columnar/InMemoryColumnarQuerySuite.scala index 38b0f666ab90b..27dfabca90217 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/columnar/InMemoryColumnarQuerySuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/columnar/InMemoryColumnarQuerySuite.scala @@ -17,11 +17,11 @@ package org.apache.spark.sql.columnar -import org.apache.spark.sql.functions._ import org.apache.spark.sql.TestData._ import org.apache.spark.sql.catalyst.expressions.Row import org.apache.spark.sql.test.TestSQLContext._ import org.apache.spark.sql.test.TestSQLContext.implicits._ +import org.apache.spark.sql.types.{DecimalType, Decimal} import org.apache.spark.sql.{QueryTest, TestData} import org.apache.spark.storage.StorageLevel.MEMORY_ONLY @@ -117,4 +117,19 @@ class InMemoryColumnarQuerySuite extends QueryTest { complexData.count() complexData.unpersist() } + + test("decimal type") { + // Casting is required here because ScalaReflection can't capture decimal precision information. + val df = (1 to 10) + .map(i => Tuple1(Decimal(i, 15, 10))) + .toDF("dec") + .select($"dec" cast DecimalType(15, 10)) + + assert(df.schema.head.dataType === DecimalType(15, 10)) + + df.cache().registerTempTable("test_fixed_decimal") + checkAnswer( + sql("SELECT * FROM test_fixed_decimal"), + (1 to 10).map(i => Row(Decimal(i, 15, 10).toJavaBigDecimal))) + } } diff --git a/sql/core/src/test/scala/org/apache/spark/sql/columnar/NullableColumnAccessorSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/columnar/NullableColumnAccessorSuite.scala index f95c895587f3f..bb305355276bf 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/columnar/NullableColumnAccessorSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/columnar/NullableColumnAccessorSuite.scala @@ -42,7 +42,8 @@ class NullableColumnAccessorSuite extends FunSuite { import ColumnarTestUtils._ Seq( - INT, LONG, SHORT, BOOLEAN, BYTE, STRING, DOUBLE, FLOAT, BINARY, GENERIC, DATE, TIMESTAMP + INT, LONG, SHORT, BOOLEAN, BYTE, STRING, DOUBLE, FLOAT, FIXED_DECIMAL(15, 10), BINARY, GENERIC, + DATE, TIMESTAMP ).foreach { testNullableColumnAccessor(_) } diff --git a/sql/core/src/test/scala/org/apache/spark/sql/columnar/NullableColumnBuilderSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/columnar/NullableColumnBuilderSuite.scala index 80bd5c94570cb..75a47498683f4 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/columnar/NullableColumnBuilderSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/columnar/NullableColumnBuilderSuite.scala @@ -38,7 +38,8 @@ class NullableColumnBuilderSuite extends FunSuite { import ColumnarTestUtils._ Seq( - INT, LONG, SHORT, BOOLEAN, BYTE, STRING, DOUBLE, FLOAT, BINARY, GENERIC, DATE, TIMESTAMP + INT, LONG, SHORT, BOOLEAN, BYTE, STRING, DOUBLE, FLOAT, FIXED_DECIMAL(15, 10), BINARY, GENERIC, + DATE, TIMESTAMP ).foreach { testNullableColumnBuilder(_) } From 127268bc3999201ec1c0a040a29c7fa9ac25476b Mon Sep 17 00:00:00 2001 From: Brennon York Date: Sat, 14 Mar 2015 17:28:13 +0000 Subject: [PATCH 060/122] [SPARK-6329][Docs]: Minor doc changes for Mesos and TOC Updated the configuration docs from the minor items that Reynold had left over from SPARK-1182; specifically I updated the `running-on-mesos` link to point directly to `running-on-mesos#configuration` and upgraded the `yarn`, `mesos`, etc. bullets to `
    ` tags in hopes that they'll get pushed into the TOC. Author: Brennon York Closes #5022 from brennonyork/SPARK-6329 and squashes the following commits: 42a10a9 [Brennon York] minor doc fixes --- docs/configuration.md | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/docs/configuration.md b/docs/configuration.md index a7116fbece9bb..63fc99e7d3e29 100644 --- a/docs/configuration.md +++ b/docs/configuration.md @@ -1391,9 +1391,11 @@ Apart from these, the following properties are also available, and may be useful Each cluster manager in Spark has additional configuration options. Configurations can be found on the pages for each mode: - * [YARN](running-on-yarn.html#configuration) - * [Mesos](running-on-mesos.html) - * [Standalone Mode](spark-standalone.html#cluster-launch-scripts) +##### [YARN](running-on-yarn.html#configuration) + +##### [Mesos](running-on-mesos.html#configuration) + +##### [Standalone Mode](spark-standalone.html#cluster-launch-scripts) # Environment Variables From c49d156624624a719c0d1262a58933ea3e346963 Mon Sep 17 00:00:00 2001 From: Brennon York Date: Sat, 14 Mar 2015 17:38:12 +0000 Subject: [PATCH 061/122] [SPARK-5790][GraphX]: VertexRDD's won't zip properly for `diff` capability (added tests) Added tests that maropu [created](https://github.com/maropu/spark/blob/1f64794b2ce33e64f340e383d4e8a60639a7eb4b/graphx/src/test/scala/org/apache/spark/graphx/VertexRDDSuite.scala) for vertices with differing partition counts. Wanted to make sure his work got captured /merged as its not in the master branch and I don't believe there's a PR out already for it. Author: Brennon York Closes #5023 from brennonyork/SPARK-5790 and squashes the following commits: 83bbd29 [Brennon York] added maropu's tests for vertices with differing partition counts --- .../apache/spark/graphx/VertexRDDSuite.scala | 38 ++++++++++++++++++- 1 file changed, 37 insertions(+), 1 deletion(-) diff --git a/graphx/src/test/scala/org/apache/spark/graphx/VertexRDDSuite.scala b/graphx/src/test/scala/org/apache/spark/graphx/VertexRDDSuite.scala index 131959cea3ef7..97533dd3aa6ce 100644 --- a/graphx/src/test/scala/org/apache/spark/graphx/VertexRDDSuite.scala +++ b/graphx/src/test/scala/org/apache/spark/graphx/VertexRDDSuite.scala @@ -19,7 +19,7 @@ package org.apache.spark.graphx import org.scalatest.FunSuite -import org.apache.spark.SparkContext +import org.apache.spark.{HashPartitioner, SparkContext} import org.apache.spark.storage.StorageLevel class VertexRDDSuite extends FunSuite with LocalSparkContext { @@ -58,6 +58,16 @@ class VertexRDDSuite extends FunSuite with LocalSparkContext { } } + test("diff vertices with the non-equal number of partitions") { + withSpark { sc => + val vertexA = VertexRDD(sc.parallelize(0 until 24, 3).map(i => (i.toLong, 0))) + val vertexB = VertexRDD(sc.parallelize(8 until 16, 2).map(i => (i.toLong, 1))) + assert(vertexA.partitions.size != vertexB.partitions.size) + val vertexC = vertexA.diff(vertexB) + assert(vertexC.map(_._1).collect.toSet === (8 until 16).toSet) + } + } + test("leftJoin") { withSpark { sc => val n = 100 @@ -73,6 +83,19 @@ class VertexRDDSuite extends FunSuite with LocalSparkContext { } } + test("leftJoin vertices with the non-equal number of partitions") { + withSpark { sc => + val vertexA = VertexRDD(sc.parallelize(0 until 100, 2).map(i => (i.toLong, 1))) + val vertexB = VertexRDD( + vertexA.filter(v => v._1 % 2 == 0).partitionBy(new HashPartitioner(3))) + assert(vertexA.partitions.size != vertexB.partitions.size) + val vertexC = vertexA.leftJoin(vertexB) { (vid, old, newOpt) => + old - newOpt.getOrElse(0) + } + assert(vertexC.filter(v => v._2 != 0).map(_._1).collect.toSet == (1 to 99 by 2).toSet) + } + } + test("innerJoin") { withSpark { sc => val n = 100 @@ -87,6 +110,19 @@ class VertexRDDSuite extends FunSuite with LocalSparkContext { (0 to n by 2).map(x => (x.toLong, 0)).toSet) } } + test("innerJoin vertices with the non-equal number of partitions") { + withSpark { sc => + val vertexA = VertexRDD(sc.parallelize(0 until 100, 2).map(i => (i.toLong, 1))) + val vertexB = VertexRDD( + vertexA.filter(v => v._1 % 2 == 0).partitionBy(new HashPartitioner(3))) + assert(vertexA.partitions.size != vertexB.partitions.size) + val vertexC = vertexA.innerJoin(vertexB) { (vid, old, newVal) => + old - newVal + } + assert(vertexC.filter(v => v._2 == 0).map(_._1).collect.toSet == (0 to 98 by 2).toSet) + } + } + test("aggregateUsingIndex") { withSpark { sc => val n = 100 From 62ede5383f64b69570a66d46939638f4bf38d1b1 Mon Sep 17 00:00:00 2001 From: OopsOutOfMemory Date: Sun, 15 Mar 2015 20:44:45 +0800 Subject: [PATCH 062/122] [SPARK-6285][SQL]Remove ParquetTestData in SparkBuild.scala and in README.md This is a following clean up PR for #5010 This will resolve issues when launching `hive/console` like below: ``` :20: error: object ParquetTestData is not a member of package org.apache.spark.sql.parquet import org.apache.spark.sql.parquet.ParquetTestData ``` Author: OopsOutOfMemory Closes #5032 from OopsOutOfMemory/SPARK-6285 and squashes the following commits: 2996aeb [OopsOutOfMemory] remove ParquetTestData --- project/SparkBuild.scala | 6 ++---- sql/README.md | 1 - 2 files changed, 2 insertions(+), 5 deletions(-) diff --git a/project/SparkBuild.scala b/project/SparkBuild.scala index f4c74c4051014..ac37c605de4b6 100644 --- a/project/SparkBuild.scala +++ b/project/SparkBuild.scala @@ -272,8 +272,7 @@ object SQL { |import org.apache.spark.sql.execution |import org.apache.spark.sql.functions._ |import org.apache.spark.sql.test.TestSQLContext._ - |import org.apache.spark.sql.types._ - |import org.apache.spark.sql.parquet.ParquetTestData""".stripMargin, + |import org.apache.spark.sql.types._""".stripMargin, cleanupCommands in console := "sparkContext.stop()" ) } @@ -304,8 +303,7 @@ object Hive { |import org.apache.spark.sql.functions._ |import org.apache.spark.sql.hive._ |import org.apache.spark.sql.hive.test.TestHive._ - |import org.apache.spark.sql.types._ - |import org.apache.spark.sql.parquet.ParquetTestData""".stripMargin, + |import org.apache.spark.sql.types._""".stripMargin, cleanupCommands in console := "sparkContext.stop()", // Some of our log4j jars make it impossible to submit jobs from this JVM to Hive Map/Reduce // in order to generate golden files. This is only required for developers who are adding new diff --git a/sql/README.md b/sql/README.md index 48f83340e37b3..fbb3200a3a4b4 100644 --- a/sql/README.md +++ b/sql/README.md @@ -41,7 +41,6 @@ import org.apache.spark.sql.functions._ import org.apache.spark.sql.hive._ import org.apache.spark.sql.hive.test.TestHive._ import org.apache.spark.sql.types._ -import org.apache.spark.sql.parquet.ParquetTestData Type in expressions to have them evaluated. Type :help for more information. From aa6536fa3c2ed1cac47abc79fc22e273f0814858 Mon Sep 17 00:00:00 2001 From: Jongyoul Lee Date: Sun, 15 Mar 2015 15:46:55 +0000 Subject: [PATCH 063/122] [SPARK-3619] Part 2. Upgrade to Mesos 0.21 to work around MESOS-1688 - MESOS_NATIVE_LIBRARY become deprecated - Chagned MESOS_NATIVE_LIBRARY to MESOS_NATIVE_JAVA_LIBRARY Author: Jongyoul Lee Closes #4361 from jongyoul/SPARK-3619-1 and squashes the following commits: f1ea91f [Jongyoul Lee] Merge branch 'SPARK-3619-1' of https://github.com/jongyoul/spark into SPARK-3619-1 a6a00c2 [Jongyoul Lee] [SPARK-3619] Upgrade to Mesos 0.21 to work around MESOS-1688 - Removed 'Known issues' section 2e15a21 [Jongyoul Lee] [SPARK-3619] Upgrade to Mesos 0.21 to work around MESOS-1688 - MESOS_NATIVE_LIBRARY become deprecated - Chagned MESOS_NATIVE_LIBRARY to MESOS_NATIVE_JAVA_LIBRARY 0dace7b [Jongyoul Lee] [SPARK-3619] Upgrade to Mesos 0.21 to work around MESOS-1688 - MESOS_NATIVE_LIBRARY become deprecated - Chagned MESOS_NATIVE_LIBRARY to MESOS_NATIVE_JAVA_LIBRARY --- conf/spark-env.sh.template | 2 +- docs/running-on-mesos.md | 5 +---- .../src/test/scala/org/apache/spark/repl/ReplSuite.scala | 2 +- .../src/test/scala/org/apache/spark/repl/ReplSuite.scala | 2 +- 4 files changed, 4 insertions(+), 7 deletions(-) diff --git a/conf/spark-env.sh.template b/conf/spark-env.sh.template index 0886b0276fb90..67f81d33361e1 100755 --- a/conf/spark-env.sh.template +++ b/conf/spark-env.sh.template @@ -15,7 +15,7 @@ # - SPARK_PUBLIC_DNS, to set the public DNS name of the driver program # - SPARK_CLASSPATH, default classpath entries to append # - SPARK_LOCAL_DIRS, storage directories to use on this node for shuffle and RDD data -# - MESOS_NATIVE_LIBRARY, to point to your libmesos.so if you use Mesos +# - MESOS_NATIVE_JAVA_LIBRARY, to point to your libmesos.so if you use Mesos # Options read in YARN client mode # - HADOOP_CONF_DIR, to point Spark towards Hadoop configuration files diff --git a/docs/running-on-mesos.md b/docs/running-on-mesos.md index e509e4bf37396..59a3e9d25baf1 100644 --- a/docs/running-on-mesos.md +++ b/docs/running-on-mesos.md @@ -110,7 +110,7 @@ cluster, or `mesos://zk://host:2181` for a multi-master Mesos cluster using ZooK The driver also needs some configuration in `spark-env.sh` to interact properly with Mesos: 1. In `spark-env.sh` set some environment variables: - * `export MESOS_NATIVE_LIBRARY=`. This path is typically + * `export MESOS_NATIVE_JAVA_LIBRARY=`. This path is typically `/lib/libmesos.so` where the prefix is `/usr/local` by default. See Mesos installation instructions above. On Mac OS X, the library is called `libmesos.dylib` instead of `libmesos.so`. @@ -167,9 +167,6 @@ acquire. By default, it will acquire *all* cores in the cluster (that get offere only makes sense if you run just one application at a time. You can cap the maximum number of cores using `conf.set("spark.cores.max", "10")` (for example). -# Known issues -- When using the "fine-grained" mode, make sure that your executors always leave 32 MB free on the slaves. Otherwise it can happen that your Spark job does not proceed anymore. Currently, Apache Mesos only offers resources if there are at least 32 MB memory allocatable. But as Spark allocates memory only for the executor and cpu only for tasks, it can happen on high slave memory usage that no new tasks will be started anymore. More details can be found in [MESOS-1688](https://issues.apache.org/jira/browse/MESOS-1688). Alternatively use the "coarse-gained" mode, which is not affected by this issue. - # Running Alongside Hadoop You can run Spark and Mesos alongside your existing Hadoop cluster by just launching them as a diff --git a/repl/scala-2.10/src/test/scala/org/apache/spark/repl/ReplSuite.scala b/repl/scala-2.10/src/test/scala/org/apache/spark/repl/ReplSuite.scala index 529914a2b6141..249f438459300 100644 --- a/repl/scala-2.10/src/test/scala/org/apache/spark/repl/ReplSuite.scala +++ b/repl/scala-2.10/src/test/scala/org/apache/spark/repl/ReplSuite.scala @@ -281,7 +281,7 @@ class ReplSuite extends FunSuite { assertDoesNotContain("Exception", output) } - if (System.getenv("MESOS_NATIVE_LIBRARY") != null) { + if (System.getenv("MESOS_NATIVE_JAVA_LIBRARY") != null) { test("running on Mesos") { val output = runInterpreter("localquiet", """ diff --git a/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala b/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala index ed9b207a86a0b..b3bd135548124 100644 --- a/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala +++ b/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala @@ -289,7 +289,7 @@ class ReplSuite extends FunSuite { assertDoesNotContain("Exception", output) } - if (System.getenv("MESOS_NATIVE_LIBRARY") != null) { + if (System.getenv("MESOS_NATIVE_JAVA_LIBRARY") != null) { test("running on Mesos") { val output = runInterpreter("localquiet", """ From 45f4c66122c57011e74c694a424756812ab77d99 Mon Sep 17 00:00:00 2001 From: Brennon York Date: Mon, 16 Mar 2015 01:06:26 -0700 Subject: [PATCH 064/122] [SPARK-5922][GraphX]: Add diff(other: RDD[VertexId, VD]) in VertexRDD Changed method invocation of 'diff' to match that of 'innerJoin' and 'leftJoin' from VertexRDD[VD] to RDD[(VertexId, VD)]. This change maintains backwards compatibility and better unifies the VertexRDD methods to match each other. Author: Brennon York Closes #4733 from brennonyork/SPARK-5922 and squashes the following commits: e800f08 [Brennon York] fixed merge conflicts b9274af [Brennon York] fixed merge conflicts f86375c [Brennon York] fixed minor include line 398ddb4 [Brennon York] fixed merge conflicts aac1810 [Brennon York] updated to aggregateUsingIndex and added test to ensure that method works properly 2af0b88 [Brennon York] removed deprecation line 753c963 [Brennon York] fixed merge conflicts and set preference to use the diff(other: VertexRDD[VD]) method 2c678c6 [Brennon York] added mima exclude to exclude new public diff method from VertexRDD 93186f3 [Brennon York] added back the original diff method to sustain binary compatibility f18356e [Brennon York] changed method invocation of 'diff' to match that of 'innerJoin' and 'leftJoin' from VertexRDD[VD] to RDD[(VertexId, VD)] --- .../scala/org/apache/spark/graphx/VertexRDD.scala | 9 +++++++++ .../apache/spark/graphx/impl/VertexRDDImpl.scala | 4 ++++ .../org/apache/spark/graphx/VertexRDDSuite.scala | 13 +++++++++++++ project/MimaExcludes.scala | 3 +++ 4 files changed, 29 insertions(+) diff --git a/graphx/src/main/scala/org/apache/spark/graphx/VertexRDD.scala b/graphx/src/main/scala/org/apache/spark/graphx/VertexRDD.scala index 40ecff7107109..ad4bfe077293a 100644 --- a/graphx/src/main/scala/org/apache/spark/graphx/VertexRDD.scala +++ b/graphx/src/main/scala/org/apache/spark/graphx/VertexRDD.scala @@ -121,6 +121,15 @@ abstract class VertexRDD[VD]( */ def mapValues[VD2: ClassTag](f: (VertexId, VD) => VD2): VertexRDD[VD2] + /** + * For each vertex present in both `this` and `other`, `diff` returns only those vertices with + * differing values; for values that are different, keeps the values from `other`. This is + * only guaranteed to work if the VertexRDDs share a common ancestor. + * + * @param other the other RDD[(VertexId, VD)] with which to diff against. + */ + def diff(other: RDD[(VertexId, VD)]): VertexRDD[VD] + /** * For each vertex present in both `this` and `other`, `diff` returns only those vertices with * differing values; for values that are different, keeps the values from `other`. This is diff --git a/graphx/src/main/scala/org/apache/spark/graphx/impl/VertexRDDImpl.scala b/graphx/src/main/scala/org/apache/spark/graphx/impl/VertexRDDImpl.scala index 904be213147dc..125692ddaad83 100644 --- a/graphx/src/main/scala/org/apache/spark/graphx/impl/VertexRDDImpl.scala +++ b/graphx/src/main/scala/org/apache/spark/graphx/impl/VertexRDDImpl.scala @@ -103,6 +103,10 @@ class VertexRDDImpl[VD] private[graphx] ( override def mapValues[VD2: ClassTag](f: (VertexId, VD) => VD2): VertexRDD[VD2] = this.mapVertexPartitions(_.map(f)) + override def diff(other: RDD[(VertexId, VD)]): VertexRDD[VD] = { + diff(this.aggregateUsingIndex(other, (a: VD, b: VD) => a)) + } + override def diff(other: VertexRDD[VD]): VertexRDD[VD] = { val otherPartition = other match { case other: VertexRDD[_] if this.partitioner == other.partitioner => diff --git a/graphx/src/test/scala/org/apache/spark/graphx/VertexRDDSuite.scala b/graphx/src/test/scala/org/apache/spark/graphx/VertexRDDSuite.scala index 97533dd3aa6ce..4f7a442ab503d 100644 --- a/graphx/src/test/scala/org/apache/spark/graphx/VertexRDDSuite.scala +++ b/graphx/src/test/scala/org/apache/spark/graphx/VertexRDDSuite.scala @@ -20,6 +20,7 @@ package org.apache.spark.graphx import org.scalatest.FunSuite import org.apache.spark.{HashPartitioner, SparkContext} +import org.apache.spark.rdd.RDD import org.apache.spark.storage.StorageLevel class VertexRDDSuite extends FunSuite with LocalSparkContext { @@ -58,6 +59,18 @@ class VertexRDDSuite extends FunSuite with LocalSparkContext { } } + test("diff with RDD[(VertexId, VD)]") { + withSpark { sc => + val n = 100 + val verts = vertices(sc, n).cache() + val flipEvens: RDD[(VertexId, Int)] = + sc.parallelize(0L to 100L) + .map(id => if (id % 2 == 0) (id, -id.toInt) else (id, id.toInt)).cache() + // diff should keep only the changed vertices + assert(verts.diff(flipEvens).map(_._2).collect().toSet === (2 to n by 2).map(-_).toSet) + } + } + test("diff vertices with the non-equal number of partitions") { withSpark { sc => val vertexA = VertexRDD(sc.parallelize(0 until 24, 3).map(i => (i.toLong, 0))) diff --git a/project/MimaExcludes.scala b/project/MimaExcludes.scala index 627b2cea4d020..a6b07fa7cddec 100644 --- a/project/MimaExcludes.scala +++ b/project/MimaExcludes.scala @@ -181,6 +181,9 @@ object MimaExcludes { ProblemFilters.exclude[MissingClassProblem]("org.apache.spark.RealClock"), ProblemFilters.exclude[MissingClassProblem]("org.apache.spark.Clock"), ProblemFilters.exclude[MissingClassProblem]("org.apache.spark.TestClock") + ) ++ Seq( + // SPARK-5922 Adding a generalized diff(other: RDD[(VertexId, VD)]) to VertexRDD + ProblemFilters.exclude[MissingMethodProblem]("org.apache.spark.graphx.VertexRDD.diff") ) case v if v.startsWith("1.2") => From 00e730b94cba1202a73af1e2476ff5a44af4b6b2 Mon Sep 17 00:00:00 2001 From: DoingDone9 <799203320@qq.com> Date: Mon, 16 Mar 2015 12:27:15 +0000 Subject: [PATCH 065/122] [SPARK-6300][Spark Core] sc.addFile(path) does not support the relative path. when i run cmd like that sc.addFile("../test.txt"), it did not work and throwed an exception: java.lang.IllegalArgumentException: java.net.URISyntaxException: Relative path in absolute URI: file:../test.txt at org.apache.hadoop.fs.Path.initialize(Path.java:206) at org.apache.hadoop.fs.Path.(Path.java:172) ........ ....... Caused by: java.net.URISyntaxException: Relative path in absolute URI: file:../test.txt at java.net.URI.checkPath(URI.java:1804) at java.net.URI.(URI.java:752) at org.apache.hadoop.fs.Path.initialize(Path.java:203) Author: DoingDone9 <799203320@qq.com> Closes #4993 from DoingDone9/relativePath and squashes the following commits: ee375cd [DoingDone9] Update SparkContextSuite.scala d594e16 [DoingDone9] Update SparkContext.scala 0ff3fa8 [DoingDone9] test for add file dced8eb [DoingDone9] Update SparkContext.scala e4a13fe [DoingDone9] getCanonicalPath 161cae3 [DoingDone9] Merge pull request #4 from apache/master c87e8b6 [DoingDone9] Merge pull request #3 from apache/master cb1852d [DoingDone9] Merge pull request #2 from apache/master c3f046f [DoingDone9] Merge pull request #1 from apache/master --- .../scala/org/apache/spark/SparkContext.scala | 2 +- .../org/apache/spark/SparkContextSuite.scala | 51 ++++++++++++++----- 2 files changed, 38 insertions(+), 15 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/SparkContext.scala b/core/src/main/scala/org/apache/spark/SparkContext.scala index 8121aab3b0b34..4457f40286fda 100644 --- a/core/src/main/scala/org/apache/spark/SparkContext.scala +++ b/core/src/main/scala/org/apache/spark/SparkContext.scala @@ -1093,7 +1093,7 @@ class SparkContext(config: SparkConf) extends Logging with ExecutorAllocationCli def addFile(path: String, recursive: Boolean): Unit = { val uri = new URI(path) val schemeCorrectedPath = uri.getScheme match { - case null | "local" => "file:" + uri.getPath + case null | "local" => new File(path).getCanonicalFile.toURI.toString case _ => path } diff --git a/core/src/test/scala/org/apache/spark/SparkContextSuite.scala b/core/src/test/scala/org/apache/spark/SparkContextSuite.scala index 50f347f1954de..b8e3e83b5a47b 100644 --- a/core/src/test/scala/org/apache/spark/SparkContextSuite.scala +++ b/core/src/test/scala/org/apache/spark/SparkContextSuite.scala @@ -79,26 +79,49 @@ class SparkContextSuite extends FunSuite with LocalSparkContext { val byteArray2 = converter.convert(bytesWritable) assert(byteArray2.length === 0) } - + test("addFile works") { - val file = File.createTempFile("someprefix", "somesuffix") - val absolutePath = file.getAbsolutePath + val file1 = File.createTempFile("someprefix1", "somesuffix1") + val absolutePath1 = file1.getAbsolutePath + + val pluto = Utils.createTempDir() + val file2 = File.createTempFile("someprefix2", "somesuffix2", pluto) + val relativePath = file2.getParent + "/../" + file2.getParentFile.getName + "/" + file2.getName + val absolutePath2 = file2.getAbsolutePath + try { - Files.write("somewords", file, UTF_8) - val length = file.length() + Files.write("somewords1", file1, UTF_8) + Files.write("somewords2", file2, UTF_8) + val length1 = file1.length() + val length2 = file2.length() + sc = new SparkContext(new SparkConf().setAppName("test").setMaster("local")) - sc.addFile(file.getAbsolutePath) + sc.addFile(file1.getAbsolutePath) + sc.addFile(relativePath) sc.parallelize(Array(1), 1).map(x => { - val gotten = new File(SparkFiles.get(file.getName)) - if (!gotten.exists()) { - throw new SparkException("file doesn't exist") + val gotten1 = new File(SparkFiles.get(file1.getName)) + val gotten2 = new File(SparkFiles.get(file2.getName)) + if (!gotten1.exists()) { + throw new SparkException("file doesn't exist : " + absolutePath1) + } + if (!gotten2.exists()) { + throw new SparkException("file doesn't exist : " + absolutePath2) } - if (length != gotten.length()) { + + if (length1 != gotten1.length()) { + throw new SparkException( + s"file has different length $length1 than added file ${gotten1.length()} : " + absolutePath1) + } + if (length2 != gotten2.length()) { throw new SparkException( - s"file has different length $length than added file ${gotten.length()}") + s"file has different length $length2 than added file ${gotten2.length()} : " + absolutePath2) } - if (absolutePath == gotten.getAbsolutePath) { - throw new SparkException("file should have been copied") + + if (absolutePath1 == gotten1.getAbsolutePath) { + throw new SparkException("file should have been copied :" + absolutePath1) + } + if (absolutePath2 == gotten2.getAbsolutePath) { + throw new SparkException("file should have been copied : " + absolutePath2) } x }).count() @@ -106,7 +129,7 @@ class SparkContextSuite extends FunSuite with LocalSparkContext { sc.stop() } } - + test("addFile recursive works") { val pluto = Utils.createTempDir() val neptune = Utils.createTempDir(pluto.getAbsolutePath) From 12a345adcbaee359199ddfed4f41bf0e19d66d48 Mon Sep 17 00:00:00 2001 From: Cheng Hao Date: Tue, 17 Mar 2015 01:09:27 +0800 Subject: [PATCH 066/122] [SPARK-2087] [SQL] Multiple thriftserver sessions with single HiveContext instance Still, we keep only a single HiveContext within ThriftServer, and we also create a object called `SQLSession` for isolating the different user states. Developers can obtain/release a new user session via `openSession` and `closeSession`, and `SQLContext` and `HiveContext` will also provide a default session if no `openSession` called, for backward-compatibility. Author: Cheng Hao Closes #4885 from chenghao-intel/multisessions_singlecontext and squashes the following commits: 1c47b2a [Cheng Hao] rename the tss => tlSession 815b27a [Cheng Hao] code style issue 57e3fa0 [Cheng Hao] openSession is not compatible between Hive0.12 & 0.13.1 4665b0d [Cheng Hao] thriftservice with single context --- .../org/apache/spark/sql/SQLContext.scala | 43 ++++- .../spark/sql/test/TestSQLContext.scala | 17 +- .../thriftserver/SparkSQLSessionManager.scala | 56 ------ .../HiveThriftServer2Suites.scala | 161 +++++++++++++++++- .../spark/sql/hive/thriftserver/Shim12.scala | 48 +++++- .../spark/sql/hive/thriftserver/Shim13.scala | 49 +++++- .../apache/spark/sql/hive/HiveContext.scala | 70 ++++---- .../apache/spark/sql/hive/test/TestHive.scala | 14 +- 8 files changed, 353 insertions(+), 105 deletions(-) delete mode 100644 sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/SparkSQLSessionManager.scala diff --git a/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala b/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala index 9c49e84bf9680..297d0d644a423 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala @@ -63,8 +63,10 @@ class SQLContext(@transient val sparkContext: SparkContext) def this(sparkContext: JavaSparkContext) = this(sparkContext.sc) - // Note that this is a lazy val so we can override the default value in subclasses. - protected[sql] lazy val conf: SQLConf = new SQLConf + /** + * @return Spark SQL configuration + */ + protected[sql] def conf = tlSession.get().conf /** * Set Spark SQL configuration properties. @@ -103,9 +105,11 @@ class SQLContext(@transient val sparkContext: SparkContext) */ def getAllConfs: immutable.Map[String, String] = conf.getAllConfs + // TODO how to handle the temp table per user session? @transient protected[sql] lazy val catalog: Catalog = new SimpleCatalog(true) + // TODO how to handle the temp function per user session? @transient protected[sql] lazy val functionRegistry: FunctionRegistry = new SimpleFunctionRegistry(true) @@ -138,6 +142,14 @@ class SQLContext(@transient val sparkContext: SparkContext) protected[sql] def executePlan(plan: LogicalPlan) = new this.QueryExecution(plan) + @transient + protected[sql] val tlSession = new ThreadLocal[SQLSession]() { + override def initialValue = defaultSession + } + + @transient + protected[sql] val defaultSession = createSession() + sparkContext.getConf.getAll.foreach { case (key, value) if key.startsWith("spark.sql") => setConf(key, value) case _ => @@ -194,6 +206,7 @@ class SQLContext(@transient val sparkContext: SparkContext) * }}} * * @group basic + * TODO move to SQLSession? */ @transient val udf: UDFRegistration = new UDFRegistration(this) @@ -1059,6 +1072,32 @@ class SQLContext(@transient val sparkContext: SparkContext) ) } + + protected[sql] def openSession(): SQLSession = { + detachSession() + val session = createSession() + tlSession.set(session) + + session + } + + protected[sql] def currentSession(): SQLSession = { + tlSession.get() + } + + protected[sql] def createSession(): SQLSession = { + new this.SQLSession() + } + + protected[sql] def detachSession(): Unit = { + tlSession.remove() + } + + protected[sql] class SQLSession { + // Note that this is a lazy val so we can override the default value in subclasses. + protected[sql] lazy val conf: SQLConf = new SQLConf + } + /** * :: DeveloperApi :: * The primary workflow for executing relational queries using Spark. Designed to allow easy diff --git a/sql/core/src/main/scala/org/apache/spark/sql/test/TestSQLContext.scala b/sql/core/src/main/scala/org/apache/spark/sql/test/TestSQLContext.scala index 4e1ec38bd0158..356a6100d2cf5 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/test/TestSQLContext.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/test/TestSQLContext.scala @@ -24,16 +24,22 @@ import org.apache.spark.sql.{DataFrame, SQLConf, SQLContext} import org.apache.spark.sql.catalyst.plans.logical.LogicalPlan /** A SQLContext that can be used for local testing. */ -object TestSQLContext +class LocalSQLContext extends SQLContext( new SparkContext( "local[2]", "TestSQLContext", new SparkConf().set("spark.sql.testkey", "true"))) { - /** Fewer partitions to speed up testing. */ - protected[sql] override lazy val conf: SQLConf = new SQLConf { - override def numShufflePartitions: Int = this.getConf(SQLConf.SHUFFLE_PARTITIONS, "5").toInt + override protected[sql] def createSession(): SQLSession = { + new this.SQLSession() + } + + protected[sql] class SQLSession extends super.SQLSession { + protected[sql] override lazy val conf: SQLConf = new SQLConf { + /** Fewer partitions to speed up testing. */ + override def numShufflePartitions: Int = this.getConf(SQLConf.SHUFFLE_PARTITIONS, "5").toInt + } } /** @@ -45,3 +51,6 @@ object TestSQLContext } } + +object TestSQLContext extends LocalSQLContext + diff --git a/sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/SparkSQLSessionManager.scala b/sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/SparkSQLSessionManager.scala deleted file mode 100644 index 89e9ede7261c9..0000000000000 --- a/sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/SparkSQLSessionManager.scala +++ /dev/null @@ -1,56 +0,0 @@ -/* - * Licensed to the Apache Software Foundation (ASF) under one or more - * contributor license agreements. See the NOTICE file distributed with - * this work for additional information regarding copyright ownership. - * The ASF licenses this file to You under the Apache License, Version 2.0 - * (the "License"); you may not use this file except in compliance with - * the License. You may obtain a copy of the License at - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package org.apache.spark.sql.hive.thriftserver - -import java.util.concurrent.Executors - -import org.apache.commons.logging.Log -import org.apache.hadoop.hive.conf.HiveConf -import org.apache.hadoop.hive.conf.HiveConf.ConfVars -import org.apache.hive.service.cli.session.SessionManager - -import org.apache.spark.sql.hive.HiveContext -import org.apache.spark.sql.hive.thriftserver.ReflectionUtils._ -import org.apache.spark.sql.hive.thriftserver.server.SparkSQLOperationManager -import org.apache.hive.service.cli.SessionHandle - -private[hive] class SparkSQLSessionManager(hiveContext: HiveContext) - extends SessionManager - with ReflectedCompositeService { - - private lazy val sparkSqlOperationManager = new SparkSQLOperationManager(hiveContext) - - override def init(hiveConf: HiveConf) { - setSuperField(this, "hiveConf", hiveConf) - - val backgroundPoolSize = hiveConf.getIntVar(ConfVars.HIVE_SERVER2_ASYNC_EXEC_THREADS) - setSuperField(this, "backgroundOperationPool", Executors.newFixedThreadPool(backgroundPoolSize)) - getAncestorField[Log](this, 3, "LOG").info( - s"HiveServer2: Async execution pool size $backgroundPoolSize") - - setSuperField(this, "operationManager", sparkSqlOperationManager) - addService(sparkSqlOperationManager) - - initCompositeService(hiveConf) - } - - override def closeSession(sessionHandle: SessionHandle) { - super.closeSession(sessionHandle) - sparkSqlOperationManager.sessionToActivePool -= sessionHandle - } -} diff --git a/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2Suites.scala b/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2Suites.scala index d783d487b5c60..aff96e21a5373 100644 --- a/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2Suites.scala +++ b/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2Suites.scala @@ -195,6 +195,146 @@ class HiveThriftBinaryServerSuite extends HiveThriftJdbcTest { } } } + + test("test multiple session") { + import org.apache.spark.sql.SQLConf + var defaultV1: String = null + var defaultV2: String = null + + withMultipleConnectionJdbcStatement( + // create table + { statement => + + val queries = Seq( + "DROP TABLE IF EXISTS test_map", + "CREATE TABLE test_map(key INT, value STRING)", + s"LOAD DATA LOCAL INPATH '${TestData.smallKv}' OVERWRITE INTO TABLE test_map", + "CACHE TABLE test_table AS SELECT key FROM test_map ORDER BY key DESC") + + queries.foreach(statement.execute) + + val rs1 = statement.executeQuery("SELECT key FROM test_table ORDER BY KEY DESC") + val buf1 = new collection.mutable.ArrayBuffer[Int]() + while (rs1.next()) { + buf1 += rs1.getInt(1) + } + rs1.close() + + val rs2 = statement.executeQuery("SELECT key FROM test_map ORDER BY KEY DESC") + val buf2 = new collection.mutable.ArrayBuffer[Int]() + while (rs2.next()) { + buf2 += rs2.getInt(1) + } + rs2.close() + + assert(buf1 === buf2) + }, + + // first session, we get the default value of the session status + { statement => + + val rs1 = statement.executeQuery(s"SET ${SQLConf.SHUFFLE_PARTITIONS}") + rs1.next() + defaultV1 = rs1.getString(1) + assert(defaultV1 != "200") + rs1.close() + + val rs2 = statement.executeQuery("SET hive.cli.print.header") + rs2.next() + + defaultV2 = rs2.getString(1) + assert(defaultV1 != "true") + rs2.close() + }, + + // second session, we update the session status + { statement => + + val queries = Seq( + s"SET ${SQLConf.SHUFFLE_PARTITIONS}=291", + "SET hive.cli.print.header=true" + ) + + queries.map(statement.execute) + val rs1 = statement.executeQuery(s"SET ${SQLConf.SHUFFLE_PARTITIONS}") + rs1.next() + assert("spark.sql.shuffle.partitions=291" === rs1.getString(1)) + rs1.close() + + val rs2 = statement.executeQuery("SET hive.cli.print.header") + rs2.next() + assert("hive.cli.print.header=true" === rs2.getString(1)) + rs2.close() + }, + + // third session, we get the latest session status, supposed to be the + // default value + { statement => + + val rs1 = statement.executeQuery(s"SET ${SQLConf.SHUFFLE_PARTITIONS}") + rs1.next() + assert(defaultV1 === rs1.getString(1)) + rs1.close() + + val rs2 = statement.executeQuery("SET hive.cli.print.header") + rs2.next() + assert(defaultV2 === rs2.getString(1)) + rs2.close() + }, + + // accessing the cached data in another session + { statement => + + val rs1 = statement.executeQuery("SELECT key FROM test_table ORDER BY KEY DESC") + val buf1 = new collection.mutable.ArrayBuffer[Int]() + while (rs1.next()) { + buf1 += rs1.getInt(1) + } + rs1.close() + + val rs2 = statement.executeQuery("SELECT key FROM test_map ORDER BY KEY DESC") + val buf2 = new collection.mutable.ArrayBuffer[Int]() + while (rs2.next()) { + buf2 += rs2.getInt(1) + } + rs2.close() + + assert(buf1 === buf2) + statement.executeQuery("UNCACHE TABLE test_table") + + // TODO need to figure out how to determine if the data loaded from cache + val rs3 = statement.executeQuery("SELECT key FROM test_map ORDER BY KEY DESC") + val buf3 = new collection.mutable.ArrayBuffer[Int]() + while (rs3.next()) { + buf3 += rs3.getInt(1) + } + rs3.close() + + assert(buf1 === buf3) + }, + + // accessing the uncached table + { statement => + + // TODO need to figure out how to determine if the data loaded from cache + val rs1 = statement.executeQuery("SELECT key FROM test_table ORDER BY KEY DESC") + val buf1 = new collection.mutable.ArrayBuffer[Int]() + while (rs1.next()) { + buf1 += rs1.getInt(1) + } + rs1.close() + + val rs2 = statement.executeQuery("SELECT key FROM test_map ORDER BY KEY DESC") + val buf2 = new collection.mutable.ArrayBuffer[Int]() + while (rs2.next()) { + buf2 += rs2.getInt(1) + } + rs2.close() + + assert(buf1 === buf2) + } + ) + } } class HiveThriftHttpServerSuite extends HiveThriftJdbcTest { @@ -245,15 +385,22 @@ abstract class HiveThriftJdbcTest extends HiveThriftServer2Test { s"jdbc:hive2://localhost:$serverPort/" } - protected def withJdbcStatement(f: Statement => Unit): Unit = { - val connection = DriverManager.getConnection(jdbcUri, user, "") - val statement = connection.createStatement() - - try f(statement) finally { - statement.close() - connection.close() + def withMultipleConnectionJdbcStatement(fs: (Statement => Unit)*) { + val user = System.getProperty("user.name") + val connections = fs.map { _ => DriverManager.getConnection(jdbcUri, user, "") } + val statements = connections.map(_.createStatement()) + + try { + statements.zip(fs).map { case (s, f) => f(s) } + } finally { + statements.map(_.close()) + connections.map(_.close()) } } + + def withJdbcStatement(f: Statement => Unit) { + withMultipleConnectionJdbcStatement(f) + } } abstract class HiveThriftServer2Test extends FunSuite with BeforeAndAfterAll with Logging { diff --git a/sql/hive-thriftserver/v0.12.0/src/main/scala/org/apache/spark/sql/hive/thriftserver/Shim12.scala b/sql/hive-thriftserver/v0.12.0/src/main/scala/org/apache/spark/sql/hive/thriftserver/Shim12.scala index 13116b40bb259..95a6e86d0546d 100644 --- a/sql/hive-thriftserver/v0.12.0/src/main/scala/org/apache/spark/sql/hive/thriftserver/Shim12.scala +++ b/sql/hive-thriftserver/v0.12.0/src/main/scala/org/apache/spark/sql/hive/thriftserver/Shim12.scala @@ -18,8 +18,15 @@ package org.apache.spark.sql.hive.thriftserver import java.sql.{Date, Timestamp} +import java.util.concurrent.Executors import java.util.{ArrayList => JArrayList, Map => JMap} +import org.apache.commons.logging.Log +import org.apache.hadoop.hive.conf.HiveConf +import org.apache.hadoop.hive.conf.HiveConf.ConfVars +import org.apache.hive.service.cli.thrift.TProtocolVersion +import org.apache.spark.sql.hive.thriftserver.server.SparkSQLOperationManager + import scala.collection.JavaConversions._ import scala.collection.mutable.{ArrayBuffer, Map => SMap} @@ -29,7 +36,7 @@ import org.apache.hadoop.hive.shims.ShimLoader import org.apache.hadoop.security.UserGroupInformation import org.apache.hive.service.cli._ import org.apache.hive.service.cli.operation.ExecuteStatementOperation -import org.apache.hive.service.cli.session.HiveSession +import org.apache.hive.service.cli.session.{SessionManager, HiveSession} import org.apache.spark.Logging import org.apache.spark.sql.{DataFrame, SQLConf, Row => SparkRow} @@ -220,3 +227,42 @@ private[hive] class SparkExecuteStatementOperation( setState(OperationState.FINISHED) } } + +private[hive] class SparkSQLSessionManager(hiveContext: HiveContext) + extends SessionManager + with ReflectedCompositeService { + + private lazy val sparkSqlOperationManager = new SparkSQLOperationManager(hiveContext) + + override def init(hiveConf: HiveConf) { + setSuperField(this, "hiveConf", hiveConf) + + val backgroundPoolSize = hiveConf.getIntVar(ConfVars.HIVE_SERVER2_ASYNC_EXEC_THREADS) + setSuperField(this, "backgroundOperationPool", Executors.newFixedThreadPool(backgroundPoolSize)) + getAncestorField[Log](this, 3, "LOG").info( + s"HiveServer2: Async execution pool size $backgroundPoolSize") + + setSuperField(this, "operationManager", sparkSqlOperationManager) + addService(sparkSqlOperationManager) + + initCompositeService(hiveConf) + } + + override def openSession( + username: String, + passwd: String, + sessionConf: java.util.Map[String, String], + withImpersonation: Boolean, + delegationToken: String): SessionHandle = { + hiveContext.openSession() + + super.openSession(username, passwd, sessionConf, withImpersonation, delegationToken) + } + + override def closeSession(sessionHandle: SessionHandle) { + super.closeSession(sessionHandle) + sparkSqlOperationManager.sessionToActivePool -= sessionHandle + + hiveContext.detachSession() + } +} diff --git a/sql/hive-thriftserver/v0.13.1/src/main/scala/org/apache/spark/sql/hive/thriftserver/Shim13.scala b/sql/hive-thriftserver/v0.13.1/src/main/scala/org/apache/spark/sql/hive/thriftserver/Shim13.scala index 9b8faeff94eab..178eb1af7cdcd 100644 --- a/sql/hive-thriftserver/v0.13.1/src/main/scala/org/apache/spark/sql/hive/thriftserver/Shim13.scala +++ b/sql/hive-thriftserver/v0.13.1/src/main/scala/org/apache/spark/sql/hive/thriftserver/Shim13.scala @@ -18,8 +18,15 @@ package org.apache.spark.sql.hive.thriftserver import java.sql.{Date, Timestamp} +import java.util.concurrent.Executors import java.util.{ArrayList => JArrayList, List => JList, Map => JMap} +import org.apache.commons.logging.Log +import org.apache.hadoop.hive.conf.HiveConf +import org.apache.hadoop.hive.conf.HiveConf.ConfVars +import org.apache.hive.service.cli.thrift.TProtocolVersion +import org.apache.spark.sql.hive.thriftserver.server.SparkSQLOperationManager + import scala.collection.JavaConversions._ import scala.collection.mutable.{ArrayBuffer, Map => SMap} @@ -27,7 +34,7 @@ import org.apache.hadoop.hive.metastore.api.FieldSchema import org.apache.hadoop.security.UserGroupInformation import org.apache.hive.service.cli._ import org.apache.hive.service.cli.operation.ExecuteStatementOperation -import org.apache.hive.service.cli.session.HiveSession +import org.apache.hive.service.cli.session.{SessionManager, HiveSession} import org.apache.spark.Logging import org.apache.spark.sql.{DataFrame, Row => SparkRow, SQLConf} @@ -191,3 +198,43 @@ private[hive] class SparkExecuteStatementOperation( setState(OperationState.FINISHED) } } + +private[hive] class SparkSQLSessionManager(hiveContext: HiveContext) + extends SessionManager + with ReflectedCompositeService { + + private lazy val sparkSqlOperationManager = new SparkSQLOperationManager(hiveContext) + + override def init(hiveConf: HiveConf) { + setSuperField(this, "hiveConf", hiveConf) + + val backgroundPoolSize = hiveConf.getIntVar(ConfVars.HIVE_SERVER2_ASYNC_EXEC_THREADS) + setSuperField(this, "backgroundOperationPool", Executors.newFixedThreadPool(backgroundPoolSize)) + getAncestorField[Log](this, 3, "LOG").info( + s"HiveServer2: Async execution pool size $backgroundPoolSize") + + setSuperField(this, "operationManager", sparkSqlOperationManager) + addService(sparkSqlOperationManager) + + initCompositeService(hiveConf) + } + + override def openSession( + protocol: TProtocolVersion, + username: String, + passwd: String, + sessionConf: java.util.Map[String, String], + withImpersonation: Boolean, + delegationToken: String): SessionHandle = { + hiveContext.openSession() + + super.openSession(protocol, username, passwd, sessionConf, withImpersonation, delegationToken) + } + + override def closeSession(sessionHandle: SessionHandle) { + super.closeSession(sessionHandle) + sparkSqlOperationManager.sessionToActivePool -= sessionHandle + + hiveContext.detachSession() + } +} diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveContext.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveContext.scala index c439dfe0a71f8..a5c435fdfa778 100644 --- a/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveContext.scala +++ b/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveContext.scala @@ -49,10 +49,6 @@ import org.apache.spark.sql.types._ class HiveContext(sc: SparkContext) extends SQLContext(sc) { self => - protected[sql] override lazy val conf: SQLConf = new SQLConf { - override def dialect: String = getConf(SQLConf.DIALECT, "hiveql") - } - /** * When true, enables an experimental feature where metastore tables that use the parquet SerDe * are automatically converted to use the Spark SQL parquet table scan, instead of the Hive @@ -214,33 +210,9 @@ class HiveContext(sc: SparkContext) extends SQLContext(sc) { } } - /** - * SQLConf and HiveConf contracts: - * - * 1. reuse existing started SessionState if any - * 2. when the Hive session is first initialized, params in HiveConf will get picked up by the - * SQLConf. Additionally, any properties set by set() or a SET command inside sql() will be - * set in the SQLConf *as well as* in the HiveConf. - */ - @transient protected[hive] lazy val sessionState: SessionState = { - var state = SessionState.get() - if (state == null) { - state = new SessionState(new HiveConf(classOf[SessionState])) - SessionState.start(state) - } - if (state.out == null) { - state.out = new PrintStream(outputBuffer, true, "UTF-8") - } - if (state.err == null) { - state.err = new PrintStream(outputBuffer, true, "UTF-8") - } - state - } + protected[hive] def sessionState = tlSession.get().asInstanceOf[this.SQLSession].sessionState - @transient protected[hive] lazy val hiveconf: HiveConf = { - setConf(sessionState.getConf.getAllProperties) - sessionState.getConf - } + protected[hive] def hiveconf = tlSession.get().asInstanceOf[this.SQLSession].hiveconf override def setConf(key: String, value: String): Unit = { super.setConf(key, value) @@ -272,6 +244,44 @@ class HiveContext(sc: SparkContext) extends SQLContext(sc) { Nil } + override protected[sql] def createSession(): SQLSession = { + new this.SQLSession() + } + + protected[hive] class SQLSession extends super.SQLSession { + protected[sql] override lazy val conf: SQLConf = new SQLConf { + override def dialect: String = getConf(SQLConf.DIALECT, "hiveql") + } + + protected[hive] lazy val hiveconf: HiveConf = { + setConf(sessionState.getConf.getAllProperties) + sessionState.getConf + } + + /** + * SQLConf and HiveConf contracts: + * + * 1. reuse existing started SessionState if any + * 2. when the Hive session is first initialized, params in HiveConf will get picked up by the + * SQLConf. Additionally, any properties set by set() or a SET command inside sql() will be + * set in the SQLConf *as well as* in the HiveConf. + */ + protected[hive] lazy val sessionState: SessionState = { + var state = SessionState.get() + if (state == null) { + state = new SessionState(new HiveConf(classOf[SessionState])) + SessionState.start(state) + } + if (state.out == null) { + state.out = new PrintStream(outputBuffer, true, "UTF-8") + } + if (state.err == null) { + state.err = new PrintStream(outputBuffer, true, "UTF-8") + } + state + } + } + /** * Runs the specified SQL query using Hive. */ diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/test/TestHive.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/test/TestHive.scala index a2d99f1f4b28d..4859991e2351a 100644 --- a/sql/hive/src/main/scala/org/apache/spark/sql/hive/test/TestHive.scala +++ b/sql/hive/src/main/scala/org/apache/spark/sql/hive/test/TestHive.scala @@ -102,10 +102,16 @@ class TestHiveContext(sc: SparkContext) extends HiveContext(sc) { override def executePlan(plan: LogicalPlan): this.QueryExecution = new this.QueryExecution(plan) - /** Fewer partitions to speed up testing. */ - protected[sql] override lazy val conf: SQLConf = new SQLConf { - override def numShufflePartitions: Int = getConf(SQLConf.SHUFFLE_PARTITIONS, "5").toInt - override def dialect: String = getConf(SQLConf.DIALECT, "hiveql") + override protected[sql] def createSession(): SQLSession = { + new this.SQLSession() + } + + protected[hive] class SQLSession extends super.SQLSession { + /** Fewer partitions to speed up testing. */ + protected[sql] override lazy val conf: SQLConf = new SQLConf { + override def numShufflePartitions: Int = getConf(SQLConf.SHUFFLE_PARTITIONS, "5").toInt + override def dialect: String = getConf(SQLConf.DIALECT, "hiveql") + } } /** From d19efeddc0cb710c9496af11e447d39e1ad61b31 Mon Sep 17 00:00:00 2001 From: Volodymyr Lyubinets Date: Mon, 16 Mar 2015 12:13:18 -0700 Subject: [PATCH 067/122] [SPARK-6330] Fix filesystem bug in newParquet relation If I'm running this locally and my path points to S3, this would currently error out because of incorrect FS. I tested this in a scenario that previously didn't work, this change seemed to fix the issue. Author: Volodymyr Lyubinets Closes #5020 from vlyubin/parquertbug and squashes the following commits: a645ad5 [Volodymyr Lyubinets] Fix filesystem bug in newParquet relation --- .../main/scala/org/apache/spark/sql/parquet/newParquet.scala | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/parquet/newParquet.scala b/sql/core/src/main/scala/org/apache/spark/sql/parquet/newParquet.scala index 234e6bb8443af..c38b6e8c61d8a 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/parquet/newParquet.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/parquet/newParquet.scala @@ -19,6 +19,7 @@ package org.apache.spark.sql.parquet import java.io.IOException import java.lang.{Double => JDouble, Float => JFloat, Long => JLong} import java.math.{BigDecimal => JBigDecimal} +import java.net.URI import java.text.SimpleDateFormat import java.util.{Date, List => JList} @@ -244,11 +245,10 @@ private[sql] case class ParquetRelation2( * Refreshes `FileStatus`es, footers, partition spec, and table schema. */ def refresh(): Unit = { - val fs = FileSystem.get(sparkContext.hadoopConfiguration) - // Support either reading a collection of raw Parquet part-files, or a collection of folders // containing Parquet files (e.g. partitioned Parquet table). val baseStatuses = paths.distinct.map { p => + val fs = FileSystem.get(URI.create(p), sparkContext.hadoopConfiguration) val qualified = fs.makeQualified(new Path(p)) if (!fs.exists(qualified) && maybeSchema.isDefined) { @@ -262,6 +262,7 @@ private[sql] case class ParquetRelation2( // Lists `FileStatus`es of all leaf nodes (files) under all base directories. val leaves = baseStatuses.flatMap { f => + val fs = FileSystem.get(f.getPath.toUri, sparkContext.hadoopConfiguration) SparkHadoopUtil.get.listLeafStatuses(fs, f.getPath).filter { f => isSummaryFile(f.getPath) || !(f.getPath.getName.startsWith("_") || f.getPath.getName.startsWith(".")) From f149b8b5e542af44650923d0156f037121b45a20 Mon Sep 17 00:00:00 2001 From: lisurprise Date: Mon, 16 Mar 2015 13:10:32 -0700 Subject: [PATCH 068/122] [SPARK-6077] Remove streaming tab while stopping StreamingContext Currently we would create a new streaming tab for each streamingContext even if there's already one on the same sparkContext which would cause duplicate StreamingTab created and none of them is taking effect. snapshot: https://www.dropbox.com/s/t4gd6hqyqo0nivz/bad%20multiple%20streamings.png?dl=0 How to reproduce: 1) import org.apache.spark.SparkConf import org.apache.spark.streaming. {Seconds, StreamingContext} import org.apache.spark.storage.StorageLevel val ssc = new StreamingContext(sc, Seconds(1)) val lines = ssc.socketTextStream("localhost", 9999, StorageLevel.MEMORY_AND_DISK_SER) val words = lines.flatMap(_.split(" ")) val wordCounts = words.map(x => (x, 1)).reduceByKey(_ + _) wordCounts.print() ssc.start() ..... 2) ssc.stop(false) val ssc = new StreamingContext(sc, Seconds(1)) val lines = ssc.socketTextStream("localhost", 9999, StorageLevel.MEMORY_AND_DISK_SER) val words = lines.flatMap(_.split(" ")) val wordCounts = words.map(x => (x, 1)).reduceByKey(_ + _) wordCounts.print() ssc.start() Author: lisurprise Closes #4828 from zhichao-li/master and squashes the following commits: c329806 [lisurprise] add test for attaching/detaching streaming tab 51e6c7f [lisurprise] move detach method into StreamingTab 31a44fa [lisurprise] add unit test for attaching and detaching new tab db25ed2 [lisurprise] clean code 8281bcb [lisurprise] clean code 193c542 [lisurprise] remove streaming tab while closing streaming context --- .../scala/org/apache/spark/ui/WebUI.scala | 28 +++++- .../org/apache/spark/ui/UISeleniumSuite.scala | 50 +++++++++- .../scala/org/apache/spark/ui/UISuite.scala | 38 +------- streaming/pom.xml | 5 + .../spark/streaming/StreamingContext.scala | 1 + .../spark/streaming/ui/StreamingPage.scala | 4 +- .../spark/streaming/ui/StreamingTab.scala | 4 + .../spark/streaming/UISeleniumSuite.scala | 95 +++++++++++++++++++ .../org/apache/spark/streaming/UISuite.scala | 55 ----------- 9 files changed, 179 insertions(+), 101 deletions(-) create mode 100644 streaming/src/test/scala/org/apache/spark/streaming/UISeleniumSuite.scala delete mode 100644 streaming/src/test/scala/org/apache/spark/streaming/UISuite.scala diff --git a/core/src/main/scala/org/apache/spark/ui/WebUI.scala b/core/src/main/scala/org/apache/spark/ui/WebUI.scala index ec68837a1516c..ea548f23120d9 100644 --- a/core/src/main/scala/org/apache/spark/ui/WebUI.scala +++ b/core/src/main/scala/org/apache/spark/ui/WebUI.scala @@ -20,14 +20,15 @@ package org.apache.spark.ui import javax.servlet.http.HttpServletRequest import scala.collection.mutable.ArrayBuffer +import scala.collection.mutable.HashMap import scala.xml.Node import org.eclipse.jetty.servlet.ServletContextHandler import org.json4s.JsonAST.{JNothing, JValue} -import org.apache.spark.{Logging, SecurityManager, SparkConf} import org.apache.spark.ui.JettyUtils._ import org.apache.spark.util.Utils +import org.apache.spark.{Logging, SecurityManager, SparkConf} /** * The top level component of the UI hierarchy that contains the server. @@ -45,6 +46,7 @@ private[spark] abstract class WebUI( protected val tabs = ArrayBuffer[WebUITab]() protected val handlers = ArrayBuffer[ServletContextHandler]() + protected val pageToHandlers = new HashMap[WebUIPage, ArrayBuffer[ServletContextHandler]] protected var serverInfo: Option[ServerInfo] = None protected val localHostName = Utils.localHostName() protected val publicHostName = Option(conf.getenv("SPARK_PUBLIC_DNS")).getOrElse(localHostName) @@ -60,14 +62,30 @@ private[spark] abstract class WebUI( tab.pages.foreach(attachPage) tabs += tab } + + def detachTab(tab: WebUITab) { + tab.pages.foreach(detachPage) + tabs -= tab + } + + def detachPage(page: WebUIPage) { + pageToHandlers.remove(page).foreach(_.foreach(detachHandler)) + } /** Attach a page to this UI. */ def attachPage(page: WebUIPage) { val pagePath = "/" + page.prefix - attachHandler(createServletHandler(pagePath, - (request: HttpServletRequest) => page.render(request), securityManager, basePath)) - attachHandler(createServletHandler(pagePath.stripSuffix("/") + "/json", - (request: HttpServletRequest) => page.renderJson(request), securityManager, basePath)) + val renderHandler = createServletHandler(pagePath, + (request: HttpServletRequest) => page.render(request), securityManager, basePath) + val renderJsonHandler = createServletHandler(pagePath.stripSuffix("/") + "/json", + (request: HttpServletRequest) => page.renderJson(request), securityManager, basePath) + attachHandler(renderHandler) + attachHandler(renderJsonHandler) + pageToHandlers.getOrElseUpdate(page, ArrayBuffer[ServletContextHandler]()) + .append(renderHandler) + pageToHandlers.getOrElseUpdate(page, ArrayBuffer[ServletContextHandler]()) + .append(renderJsonHandler) + } /** Attach a handler to this UI. */ diff --git a/core/src/test/scala/org/apache/spark/ui/UISeleniumSuite.scala b/core/src/test/scala/org/apache/spark/ui/UISeleniumSuite.scala index 6a972381faf14..0d155982a8c54 100644 --- a/core/src/test/scala/org/apache/spark/ui/UISeleniumSuite.scala +++ b/core/src/test/scala/org/apache/spark/ui/UISeleniumSuite.scala @@ -17,20 +17,24 @@ package org.apache.spark.ui +import javax.servlet.http.HttpServletRequest + import scala.collection.JavaConversions._ +import scala.xml.Node -import org.openqa.selenium.{By, WebDriver} import org.openqa.selenium.htmlunit.HtmlUnitDriver +import org.openqa.selenium.{By, WebDriver} import org.scalatest._ import org.scalatest.concurrent.Eventually._ import org.scalatest.selenium.WebBrowser import org.scalatest.time.SpanSugar._ -import org.apache.spark._ import org.apache.spark.LocalSparkContext._ +import org.apache.spark._ import org.apache.spark.api.java.StorageLevels import org.apache.spark.shuffle.FetchFailedException + /** * Selenium tests for the Spark Web UI. */ @@ -310,4 +314,46 @@ class UISeleniumSuite extends FunSuite with WebBrowser with Matchers with Before } } } + + test("attaching and detaching a new tab") { + withSpark(newSparkContext()) { sc => + val sparkUI = sc.ui.get + + val newTab = new WebUITab(sparkUI, "foo") { + attachPage(new WebUIPage("") { + def render(request: HttpServletRequest): Seq[Node] = { + "html magic" + } + }) + } + sparkUI.attachTab(newTab) + eventually(timeout(10 seconds), interval(50 milliseconds)) { + go to (sc.ui.get.appUIAddress.stripSuffix("/")) + find(cssSelector("""ul li a[href*="jobs"]""")) should not be(None) + find(cssSelector("""ul li a[href*="stages"]""")) should not be(None) + find(cssSelector("""ul li a[href*="storage"]""")) should not be(None) + find(cssSelector("""ul li a[href*="environment"]""")) should not be(None) + find(cssSelector("""ul li a[href*="foo"]""")) should not be(None) + } + eventually(timeout(10 seconds), interval(50 milliseconds)) { + // check whether new page exists + go to (sc.ui.get.appUIAddress.stripSuffix("/") + "/foo") + find(cssSelector("b")).get.text should include ("html magic") + } + sparkUI.detachTab(newTab) + eventually(timeout(10 seconds), interval(50 milliseconds)) { + go to (sc.ui.get.appUIAddress.stripSuffix("/")) + find(cssSelector("""ul li a[href*="jobs"]""")) should not be(None) + find(cssSelector("""ul li a[href*="stages"]""")) should not be(None) + find(cssSelector("""ul li a[href*="storage"]""")) should not be(None) + find(cssSelector("""ul li a[href*="environment"]""")) should not be(None) + find(cssSelector("""ul li a[href*="foo"]""")) should be(None) + } + eventually(timeout(10 seconds), interval(50 milliseconds)) { + // check new page not exist + go to (sc.ui.get.appUIAddress.stripSuffix("/") + "/foo") + find(cssSelector("b")) should be(None) + } + } + } } diff --git a/core/src/test/scala/org/apache/spark/ui/UISuite.scala b/core/src/test/scala/org/apache/spark/ui/UISuite.scala index 92a21f82f3c21..77a038dc1720d 100644 --- a/core/src/test/scala/org/apache/spark/ui/UISuite.scala +++ b/core/src/test/scala/org/apache/spark/ui/UISuite.scala @@ -18,7 +18,6 @@ package org.apache.spark.ui import java.net.ServerSocket -import javax.servlet.http.HttpServletRequest import scala.io.Source import scala.util.{Failure, Success, Try} @@ -28,9 +27,8 @@ import org.scalatest.FunSuite import org.scalatest.concurrent.Eventually._ import org.scalatest.time.SpanSugar._ -import org.apache.spark.{SparkContext, SparkConf} import org.apache.spark.LocalSparkContext._ -import scala.xml.Node +import org.apache.spark.{SparkConf, SparkContext} class UISuite extends FunSuite { @@ -72,40 +70,6 @@ class UISuite extends FunSuite { } } - ignore("attaching a new tab") { - withSpark(newSparkContext()) { sc => - val sparkUI = sc.ui.get - - val newTab = new WebUITab(sparkUI, "foo") { - attachPage(new WebUIPage("") { - def render(request: HttpServletRequest): Seq[Node] = { - "html magic" - } - }) - } - sparkUI.attachTab(newTab) - eventually(timeout(10 seconds), interval(50 milliseconds)) { - val html = Source.fromURL(sparkUI.appUIAddress).mkString - assert(!html.contains("random data that should not be present")) - - // check whether new page exists - assert(html.toLowerCase.contains("foo")) - - // check whether other pages still exist - assert(html.toLowerCase.contains("stages")) - assert(html.toLowerCase.contains("storage")) - assert(html.toLowerCase.contains("environment")) - assert(html.toLowerCase.contains("executors")) - } - - eventually(timeout(10 seconds), interval(50 milliseconds)) { - val html = Source.fromURL(sparkUI.appUIAddress.stripSuffix("/") + "/foo").mkString - // check whether new page exists - assert(html.contains("magic")) - } - } - } - test("jetty selects different port under contention") { val server = new ServerSocket(0) val startPort = server.getLocalPort diff --git a/streaming/pom.xml b/streaming/pom.xml index 0370b0e9e1aa3..96508d83f4049 100644 --- a/streaming/pom.xml +++ b/streaming/pom.xml @@ -82,6 +82,11 @@ junit test + + org.seleniumhq.selenium + selenium-java + test + com.novocode junit-interface diff --git a/streaming/src/main/scala/org/apache/spark/streaming/StreamingContext.scala b/streaming/src/main/scala/org/apache/spark/streaming/StreamingContext.scala index ba3f23434f24c..b5b6770a8a150 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/StreamingContext.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/StreamingContext.scala @@ -578,6 +578,7 @@ class StreamingContext private[streaming] ( // Even if we have already stopped, we still need to attempt to stop the SparkContext because // a user might stop(stopSparkContext = false) and then call stop(stopSparkContext = true). if (stopSparkContext) sc.stop() + uiTab.foreach(_.detach()) // The state should always be Stopped after calling `stop()`, even if we haven't started yet: state = Stopped } diff --git a/streaming/src/main/scala/org/apache/spark/streaming/ui/StreamingPage.scala b/streaming/src/main/scala/org/apache/spark/streaming/ui/StreamingPage.scala index 98e9a2e639e25..bfe8086fcf8fe 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/ui/StreamingPage.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/ui/StreamingPage.scala @@ -32,7 +32,7 @@ private[ui] class StreamingPage(parent: StreamingTab) extends WebUIPage("") with Logging { private val listener = parent.listener - private val startTime = Calendar.getInstance().getTime() + private val startTime = System.currentTimeMillis() private val emptyCell = "-" /** Render the page */ @@ -47,7 +47,7 @@ private[ui] class StreamingPage(parent: StreamingTab) /** Generate basic stats of the streaming program */ private def generateBasicStats(): Seq[Node] = { - val timeSinceStart = System.currentTimeMillis() - startTime.getTime + val timeSinceStart = System.currentTimeMillis() - startTime
    • Started at: {startTime.toString} diff --git a/streaming/src/main/scala/org/apache/spark/streaming/ui/StreamingTab.scala b/streaming/src/main/scala/org/apache/spark/streaming/ui/StreamingTab.scala index d9d04cd706a04..9a860ea4a6c68 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/ui/StreamingTab.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/ui/StreamingTab.scala @@ -36,6 +36,10 @@ private[spark] class StreamingTab(ssc: StreamingContext) ssc.addStreamingListener(listener) attachPage(new StreamingPage(this)) parent.attachTab(this) + + def detach() { + getSparkUI(ssc).detachTab(this) + } } private object StreamingTab { diff --git a/streaming/src/test/scala/org/apache/spark/streaming/UISeleniumSuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/UISeleniumSuite.scala new file mode 100644 index 0000000000000..87a0395efbf2a --- /dev/null +++ b/streaming/src/test/scala/org/apache/spark/streaming/UISeleniumSuite.scala @@ -0,0 +1,95 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.streaming + +import org.openqa.selenium.WebDriver +import org.openqa.selenium.htmlunit.HtmlUnitDriver +import org.scalatest._ +import org.scalatest.concurrent.Eventually._ +import org.scalatest.selenium.WebBrowser +import org.scalatest.time.SpanSugar._ + +import org.apache.spark._ + + + + +/** + * Selenium tests for the Spark Web UI. + */ +class UISeleniumSuite extends FunSuite with WebBrowser with Matchers with BeforeAndAfterAll with TestSuiteBase { + + implicit var webDriver: WebDriver = _ + + override def beforeAll(): Unit = { + webDriver = new HtmlUnitDriver + } + + override def afterAll(): Unit = { + if (webDriver != null) { + webDriver.quit() + } + } + + /** + * Create a test SparkStreamingContext with the SparkUI enabled. + */ + private def newSparkStreamingContext(): StreamingContext = { + val conf = new SparkConf() + .setMaster("local") + .setAppName("test") + .set("spark.ui.enabled", "true") + val ssc = new StreamingContext(conf, Seconds(1)) + assert(ssc.sc.ui.isDefined, "Spark UI is not started!") + ssc + } + + test("attaching and detaching a Streaming tab") { + withStreamingContext(newSparkStreamingContext()) { ssc => + val sparkUI = ssc.sparkContext.ui.get + + eventually(timeout(10 seconds), interval(50 milliseconds)) { + go to (sparkUI.appUIAddress.stripSuffix("/")) + find(cssSelector( """ul li a[href*="streaming"]""")) should not be (None) + } + + eventually(timeout(10 seconds), interval(50 milliseconds)) { + // check whether streaming page exists + go to (sparkUI.appUIAddress.stripSuffix("/") + "/streaming") + val statisticText = findAll(cssSelector("li strong")).map(_.text).toSeq + statisticText should contain("Network receivers:") + statisticText should contain("Batch interval:") + } + + ssc.stop(false) + + eventually(timeout(10 seconds), interval(50 milliseconds)) { + go to (sparkUI.appUIAddress.stripSuffix("/")) + find(cssSelector( """ul li a[href*="streaming"]""")) should be(None) + } + + eventually(timeout(10 seconds), interval(50 milliseconds)) { + go to (sparkUI.appUIAddress.stripSuffix("/") + "/streaming") + val statisticText = findAll(cssSelector("li strong")).map(_.text).toSeq + statisticText should not contain ("Network receivers:") + statisticText should not contain ("Batch interval:") + } + } + } +} + diff --git a/streaming/src/test/scala/org/apache/spark/streaming/UISuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/UISuite.scala deleted file mode 100644 index 8e30118266855..0000000000000 --- a/streaming/src/test/scala/org/apache/spark/streaming/UISuite.scala +++ /dev/null @@ -1,55 +0,0 @@ -/* - * Licensed to the Apache Software Foundation (ASF) under one or more - * contributor license agreements. See the NOTICE file distributed with - * this work for additional information regarding copyright ownership. - * The ASF licenses this file to You under the Apache License, Version 2.0 - * (the "License"); you may not use this file except in compliance with - * the License. You may obtain a copy of the License at - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - * See the License for the specific language governing permissions and - * limitations under the License. - */ - -package org.apache.spark.streaming - -import scala.io.Source - -import org.scalatest.FunSuite -import org.scalatest.concurrent.Eventually._ -import org.scalatest.time.SpanSugar._ - -import org.apache.spark.SparkConf - -class UISuite extends FunSuite { - - // Ignored: See SPARK-1530 - ignore("streaming tab in spark UI") { - val conf = new SparkConf() - .setMaster("local") - .setAppName("test") - .set("spark.ui.enabled", "true") - val ssc = new StreamingContext(conf, Seconds(1)) - assert(ssc.sc.ui.isDefined, "Spark UI is not started!") - val ui = ssc.sc.ui.get - - eventually(timeout(10 seconds), interval(50 milliseconds)) { - val html = Source.fromURL(ui.appUIAddress).mkString - assert(!html.contains("random data that should not be present")) - // test if streaming tab exist - assert(html.toLowerCase.contains("streaming")) - // test if other Spark tabs still exist - assert(html.toLowerCase.contains("stages")) - } - - eventually(timeout(10 seconds), interval(50 milliseconds)) { - val html = Source.fromURL(ui.appUIAddress.stripSuffix("/") + "/streaming").mkString - assert(html.toLowerCase.contains("batch")) - assert(html.toLowerCase.contains("network")) - } - } -} From e3f315ac358dfe4f5b9705c3eac76e8b1e24f82a Mon Sep 17 00:00:00 2001 From: Davies Liu Date: Mon, 16 Mar 2015 16:26:55 -0700 Subject: [PATCH 069/122] [SPARK-6327] [PySpark] fix launch spark-submit from python SparkSubmit should be launched without setting PYSPARK_SUBMIT_ARGS cc JoshRosen , this mode is actually used by python unit test, so I will not add more test for it. Author: Davies Liu Closes #5019 from davies/fix_submit and squashes the following commits: 2c20b0c [Davies Liu] fix launch spark-submit from python --- bin/pyspark | 1 - python/pyspark/java_gateway.py | 6 ++---- 2 files changed, 2 insertions(+), 5 deletions(-) diff --git a/bin/pyspark b/bin/pyspark index e7f6a1a072c2a..776b28dc41099 100755 --- a/bin/pyspark +++ b/bin/pyspark @@ -89,7 +89,6 @@ export PYTHONSTARTUP="$SPARK_HOME/python/pyspark/shell.py" if [[ -n "$SPARK_TESTING" ]]; then unset YARN_CONF_DIR unset HADOOP_CONF_DIR - export PYSPARK_SUBMIT_ARGS=pyspark-shell if [[ -n "$PYSPARK_DOC_TEST" ]]; then exec "$PYSPARK_DRIVER_PYTHON" -m doctest $1 else diff --git a/python/pyspark/java_gateway.py b/python/pyspark/java_gateway.py index 43d2cf5171880..0a16cbd8bff62 100644 --- a/python/pyspark/java_gateway.py +++ b/python/pyspark/java_gateway.py @@ -38,10 +38,8 @@ def launch_gateway(): # proper classpath and settings from spark-env.sh on_windows = platform.system() == "Windows" script = "./bin/spark-submit.cmd" if on_windows else "./bin/spark-submit" - submit_args = os.environ.get("PYSPARK_SUBMIT_ARGS") - submit_args = submit_args if submit_args is not None else "" - submit_args = shlex.split(submit_args) - command = [os.path.join(SPARK_HOME, script)] + submit_args + submit_args = os.environ.get("PYSPARK_SUBMIT_ARGS", "pyspark-shell") + command = [os.path.join(SPARK_HOME, script)] + shlex.split(submit_args) # Start a socket that will be used by PythonGatewayServer to communicate its port to us callback_socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM) From 9667b9f9c3239f814a0b1120355d9e7bd7a89158 Mon Sep 17 00:00:00 2001 From: Daoyuan Wang Date: Tue, 17 Mar 2015 12:29:15 +0800 Subject: [PATCH 070/122] [SPARK-5712] [SQL] fix comment with semicolon at end ---- comment; Author: Daoyuan Wang Closes #4500 from adrian-wang/semicolon and squashes the following commits: 70b8abb [Daoyuan Wang] use mkstring instead of reduce 2d49738 [Daoyuan Wang] remove outdated golden file 317346e [Daoyuan Wang] only skip comment with semicolon at end of line, to avoid golden file outdated d3ae01e [Daoyuan Wang] fix error a11602d [Daoyuan Wang] fix comment with semicolon at end --- .../hive/thriftserver/SparkSQLCLIDriver.scala | 25 ++++++++++--------- .../execution/HiveCompatibilitySuite.scala | 1 + ...micolon-0-f104632770dc96b81f00ccdac51fe5a8 | 1 + .../hive/execution/HiveComparisonTest.scala | 5 +++- 4 files changed, 19 insertions(+), 13 deletions(-) create mode 100644 sql/hive/src/test/resources/golden/semicolon-0-f104632770dc96b81f00ccdac51fe5a8 diff --git a/sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/SparkSQLCLIDriver.scala b/sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/SparkSQLCLIDriver.scala index 401e97b162dea..895688ab2ec2e 100644 --- a/sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/SparkSQLCLIDriver.scala +++ b/sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/SparkSQLCLIDriver.scala @@ -202,20 +202,21 @@ private[hive] object SparkSQLCLIDriver { var line = reader.readLine(currentPrompt + "> ") while (line != null) { - if (prefix.nonEmpty) { - prefix += '\n' - } + if (!line.startsWith("--")) { + if (prefix.nonEmpty) { + prefix += '\n' + } - if (line.trim().endsWith(";") && !line.trim().endsWith("\\;")) { - line = prefix + line - ret = cli.processLine(line, true) - prefix = "" - currentPrompt = promptWithCurrentDB - } else { - prefix = prefix + line - currentPrompt = continuedPromptWithDBSpaces + if (line.trim().endsWith(";") && !line.trim().endsWith("\\;")) { + line = prefix + line + ret = cli.processLine(line, true) + prefix = "" + currentPrompt = promptWithCurrentDB + } else { + prefix = prefix + line + currentPrompt = continuedPromptWithDBSpaces + } } - line = reader.readLine(currentPrompt + "> ") } diff --git a/sql/hive/compatibility/src/test/scala/org/apache/spark/sql/hive/execution/HiveCompatibilitySuite.scala b/sql/hive/compatibility/src/test/scala/org/apache/spark/sql/hive/execution/HiveCompatibilitySuite.scala index 6126ce7130426..68cb34d698ef3 100644 --- a/sql/hive/compatibility/src/test/scala/org/apache/spark/sql/hive/execution/HiveCompatibilitySuite.scala +++ b/sql/hive/compatibility/src/test/scala/org/apache/spark/sql/hive/execution/HiveCompatibilitySuite.scala @@ -726,6 +726,7 @@ class HiveCompatibilitySuite extends HiveQueryFileTest with BeforeAndAfter { "select_unquote_and", "select_unquote_not", "select_unquote_or", + "semicolon", "semijoin", "serde_regex", "serde_reported_schema", diff --git a/sql/hive/src/test/resources/golden/semicolon-0-f104632770dc96b81f00ccdac51fe5a8 b/sql/hive/src/test/resources/golden/semicolon-0-f104632770dc96b81f00ccdac51fe5a8 new file mode 100644 index 0000000000000..1b79f38e25b24 --- /dev/null +++ b/sql/hive/src/test/resources/golden/semicolon-0-f104632770dc96b81f00ccdac51fe5a8 @@ -0,0 +1 @@ +500 diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveComparisonTest.scala b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveComparisonTest.scala index a90bd1e257ade..8f3285242091c 100644 --- a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveComparisonTest.scala +++ b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveComparisonTest.scala @@ -241,7 +241,10 @@ abstract class HiveComparisonTest // Clear old output for this testcase. outputDirectories.map(new File(_, testCaseName)).filter(_.exists()).foreach(_.delete()) - val allQueries = sql.split("(?<=[^\\\\]);").map(_.trim).filterNot(q => q == "").toSeq + val sqlWithoutComment = + sql.split("\n").filterNot(l => l.matches("--.*(?<=[^\\\\]);")).mkString("\n") + val allQueries = + sqlWithoutComment.split("(?<=[^\\\\]);").map(_.trim).filterNot(q => q == "").toSeq // TODO: DOCUMENT UNSUPPORTED val queryList = From f0edeae7f9ab7eae02c227be9162ec69d22c92bd Mon Sep 17 00:00:00 2001 From: "Kevin (Sangwoo) Kim" Date: Mon, 16 Mar 2015 23:49:23 -0700 Subject: [PATCH 071/122] [SPARK-6299][CORE] ClassNotFoundException in standalone mode when running groupByKey with class defined in REPL ``` case class ClassA(value: String) val rdd = sc.parallelize(List(("k1", ClassA("v1")), ("k1", ClassA("v2")) )) rdd.groupByKey.collect ``` This code used to be throw exception in spark-shell, because while shuffling ```JavaSerializer```uses ```defaultClassLoader``` which was defined like ```env.serializer.setDefaultClassLoader(urlClassLoader)```. It should be ```env.serializer.setDefaultClassLoader(replClassLoader)```, like ``` override def run() { val deserializeStartTime = System.currentTimeMillis() Thread.currentThread.setContextClassLoader(replClassLoader) ``` in TaskRunner. When ```replClassLoader``` cannot be defined, it's identical with ```urlClassLoader``` Author: Kevin (Sangwoo) Kim Closes #5046 from swkimme/master and squashes the following commits: fa2b9ee [Kevin (Sangwoo) Kim] stylish test codes ( collect -> collect() ) 6e9620b [Kevin (Sangwoo) Kim] stylish test codes ( collect -> collect() ) d23e4e2 [Kevin (Sangwoo) Kim] stylish test codes ( collect -> collect() ) a4a3c8a [Kevin (Sangwoo) Kim] add 'class defined in repl - shuffle' test to ReplSuite bd00da5 [Kevin (Sangwoo) Kim] add 'class defined in repl - shuffle' test to ReplSuite c1b1fc7 [Kevin (Sangwoo) Kim] use REPL class loader for executor's serializer --- .../org/apache/spark/executor/Executor.scala | 2 +- .../org/apache/spark/repl/ReplSuite.scala | 50 ++++++++++++------- .../org/apache/spark/repl/ReplSuite.scala | 50 ++++++++++++------- 3 files changed, 63 insertions(+), 39 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/executor/Executor.scala b/core/src/main/scala/org/apache/spark/executor/Executor.scala index a897e532184ac..6196f7b165049 100644 --- a/core/src/main/scala/org/apache/spark/executor/Executor.scala +++ b/core/src/main/scala/org/apache/spark/executor/Executor.scala @@ -103,7 +103,7 @@ private[spark] class Executor( private val replClassLoader = addReplClassLoaderIfNeeded(urlClassLoader) // Set the classloader for serializer - env.serializer.setDefaultClassLoader(urlClassLoader) + env.serializer.setDefaultClassLoader(replClassLoader) // Akka's message frame size. If task result is bigger than this, we use the block manager // to send the result back. diff --git a/repl/scala-2.10/src/test/scala/org/apache/spark/repl/ReplSuite.scala b/repl/scala-2.10/src/test/scala/org/apache/spark/repl/ReplSuite.scala index 249f438459300..934daaeaafca1 100644 --- a/repl/scala-2.10/src/test/scala/org/apache/spark/repl/ReplSuite.scala +++ b/repl/scala-2.10/src/test/scala/org/apache/spark/repl/ReplSuite.scala @@ -121,9 +121,9 @@ class ReplSuite extends FunSuite { val output = runInterpreter("local", """ |var v = 7 - |sc.parallelize(1 to 10).map(x => v).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => v).collect().reduceLeft(_+_) |v = 10 - |sc.parallelize(1 to 10).map(x => v).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => v).collect().reduceLeft(_+_) """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -137,7 +137,7 @@ class ReplSuite extends FunSuite { |class C { |def foo = 5 |} - |sc.parallelize(1 to 10).map(x => (new C).foo).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => (new C).foo).collect().reduceLeft(_+_) """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -148,7 +148,7 @@ class ReplSuite extends FunSuite { val output = runInterpreter("local", """ |def double(x: Int) = x + x - |sc.parallelize(1 to 10).map(x => double(x)).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => double(x)).collect().reduceLeft(_+_) """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -160,9 +160,9 @@ class ReplSuite extends FunSuite { """ |var v = 7 |def getV() = v - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) |v = 10 - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -178,9 +178,9 @@ class ReplSuite extends FunSuite { """ |var array = new Array[Int](5) |val broadcastArray = sc.broadcast(array) - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() |array(0) = 5 - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -216,14 +216,14 @@ class ReplSuite extends FunSuite { """ |var v = 7 |def getV() = v - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) |v = 10 - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) |var array = new Array[Int](5) |val broadcastArray = sc.broadcast(array) - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() |array(0) = 5 - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -262,7 +262,7 @@ class ReplSuite extends FunSuite { |val sqlContext = new org.apache.spark.sql.SQLContext(sc) |import sqlContext.implicits._ |case class TestCaseClass(value: Int) - |sc.parallelize(1 to 10).map(x => TestCaseClass(x)).toDF.collect() + |sc.parallelize(1 to 10).map(x => TestCaseClass(x)).toDF().collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -275,7 +275,7 @@ class ReplSuite extends FunSuite { |val t = new TestClass |import t.testMethod |case class TestCaseClass(value: Int) - |sc.parallelize(1 to 10).map(x => TestCaseClass(x)).collect + |sc.parallelize(1 to 10).map(x => TestCaseClass(x)).collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -287,14 +287,14 @@ class ReplSuite extends FunSuite { """ |var v = 7 |def getV() = v - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) |v = 10 - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) |var array = new Array[Int](5) |val broadcastArray = sc.broadcast(array) - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() |array(0) = 5 - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -309,10 +309,22 @@ class ReplSuite extends FunSuite { val output = runInterpreter("local[2]", """ |case class Foo(i: Int) - |val ret = sc.parallelize((1 to 100).map(Foo), 10).collect + |val ret = sc.parallelize((1 to 100).map(Foo), 10).collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) assertContains("ret: Array[Foo] = Array(Foo(1),", output) } + + test("collecting objects of class defined in repl - shuffling") { + val output = runInterpreter("local-cluster[1,1,512]", + """ + |case class Foo(i: Int) + |val list = List((1, Foo(1)), (1, Foo(2))) + |val ret = sc.parallelize(list).groupByKey().collect() + """.stripMargin) + assertDoesNotContain("error:", output) + assertDoesNotContain("Exception", output) + assertContains("ret: Array[(Int, Iterable[Foo])] = Array((1,", output) + } } diff --git a/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala b/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala index b3bd135548124..fbef5b25ba688 100644 --- a/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala +++ b/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala @@ -128,9 +128,9 @@ class ReplSuite extends FunSuite { val output = runInterpreter("local", """ |var v = 7 - |sc.parallelize(1 to 10).map(x => v).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => v).collect().reduceLeft(_+_) |v = 10 - |sc.parallelize(1 to 10).map(x => v).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => v).collect().reduceLeft(_+_) """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -144,7 +144,7 @@ class ReplSuite extends FunSuite { |class C { |def foo = 5 |} - |sc.parallelize(1 to 10).map(x => (new C).foo).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => (new C).foo).collect().reduceLeft(_+_) """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -155,7 +155,7 @@ class ReplSuite extends FunSuite { val output = runInterpreter("local", """ |def double(x: Int) = x + x - |sc.parallelize(1 to 10).map(x => double(x)).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => double(x)).collect().reduceLeft(_+_) """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -167,9 +167,9 @@ class ReplSuite extends FunSuite { """ |var v = 7 |def getV() = v - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) |v = 10 - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -185,9 +185,9 @@ class ReplSuite extends FunSuite { """ |var array = new Array[Int](5) |val broadcastArray = sc.broadcast(array) - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() |array(0) = 5 - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -224,14 +224,14 @@ class ReplSuite extends FunSuite { """ |var v = 7 |def getV() = v - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) |v = 10 - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) |var array = new Array[Int](5) |val broadcastArray = sc.broadcast(array) - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() |array(0) = 5 - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -270,7 +270,7 @@ class ReplSuite extends FunSuite { |val sqlContext = new org.apache.spark.sql.SQLContext(sc) |import sqlContext.implicits._ |case class TestCaseClass(value: Int) - |sc.parallelize(1 to 10).map(x => TestCaseClass(x)).toDF.collect + |sc.parallelize(1 to 10).map(x => TestCaseClass(x)).toDF().collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -283,7 +283,7 @@ class ReplSuite extends FunSuite { |val t = new TestClass |import t.testMethod |case class TestCaseClass(value: Int) - |sc.parallelize(1 to 10).map(x => TestCaseClass(x)).collect + |sc.parallelize(1 to 10).map(x => TestCaseClass(x)).collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -295,14 +295,14 @@ class ReplSuite extends FunSuite { """ |var v = 7 |def getV() = v - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) |v = 10 - |sc.parallelize(1 to 10).map(x => getV()).collect.reduceLeft(_+_) + |sc.parallelize(1 to 10).map(x => getV()).collect().reduceLeft(_+_) |var array = new Array[Int](5) |val broadcastArray = sc.broadcast(array) - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() |array(0) = 5 - |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect + |sc.parallelize(0 to 4).map(x => broadcastArray.value(x)).collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) @@ -317,10 +317,22 @@ class ReplSuite extends FunSuite { val output = runInterpreter("local[2]", """ |case class Foo(i: Int) - |val ret = sc.parallelize((1 to 100).map(Foo), 10).collect + |val ret = sc.parallelize((1 to 100).map(Foo), 10).collect() """.stripMargin) assertDoesNotContain("error:", output) assertDoesNotContain("Exception", output) assertContains("ret: Array[Foo] = Array(Foo(1),", output) } + + test("collecting objects of class defined in repl - shuffling") { + val output = runInterpreter("local-cluster[1,1,512]", + """ + |case class Foo(i: Int) + |val list = List((1, Foo(1)), (1, Foo(2))) + |val ret = sc.parallelize(list).groupByKey().collect() + """.stripMargin) + assertDoesNotContain("error:", output) + assertDoesNotContain("Exception", output) + assertContains("ret: Array[(Int, Iterable[Foo])] = Array((1,", output) + } } From 68707225f1a4081aadbf0fd7e6221293a190529b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Lomig=20Me=CC=81gard?= Date: Mon, 16 Mar 2015 23:52:42 -0700 Subject: [PATCH 072/122] [SQL][docs][minor] Fixed sample code in SQLContext scaladoc MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Error in the code sample of the `implicits` object in `SQLContext`. Author: Lomig Mégard Closes #5051 from tarfaa/simple and squashes the following commits: 5a88acc [Lomig Mégard] [docs][minor] Fixed sample code in SQLContext scaladoc --- sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala b/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala index 297d0d644a423..6de46a50db20e 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala @@ -242,8 +242,8 @@ class SQLContext(@transient val sparkContext: SparkContext) * common Scala objects into [[DataFrame]]s. * * {{{ - * val sqlContext = new SQLContext - * import sqlContext._ + * val sqlContext = new SQLContext(sc) + * import sqlContext.implicits._ * }}} * * @group basic From b3e6eca81f79ba3c9205211797fa825b199bac83 Mon Sep 17 00:00:00 2001 From: Takeshi YAMAMURO Date: Mon, 16 Mar 2015 23:54:54 -0700 Subject: [PATCH 073/122] [SPARK-6357][GraphX] Add unapply in EdgeContext This extractor is mainly used for Graph#aggregateMessages*. Author: Takeshi YAMAMURO Closes #5047 from maropu/AddUnapplyInEdgeContext and squashes the following commits: 87e04df [Takeshi YAMAMURO] Add unapply in EdgeContext --- .../org/apache/spark/graphx/EdgeContext.scala | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) diff --git a/graphx/src/main/scala/org/apache/spark/graphx/EdgeContext.scala b/graphx/src/main/scala/org/apache/spark/graphx/EdgeContext.scala index f70715fca6eea..d8be02e2023d5 100644 --- a/graphx/src/main/scala/org/apache/spark/graphx/EdgeContext.scala +++ b/graphx/src/main/scala/org/apache/spark/graphx/EdgeContext.scala @@ -49,3 +49,20 @@ abstract class EdgeContext[VD, ED, A] { et } } + +object EdgeContext { + + /** + * Extractor mainly used for Graph#aggregateMessages*. + * Example: + * {{{ + * val messages = graph.aggregateMessages( + * case ctx @ EdgeContext(_, _, _, _, attr) => + * ctx.sendToDst(attr) + * , _ + _) + * }}} + */ + def unapply[VD, ED, A](edge: EdgeContext[VD, ED, A]) = + Some(edge.srcId, edge.dstId, edge.srcAttr, edge.dstAttr, edge.attr) +} + From b2d8c02224892192b1aa314b4265fe50845932f9 Mon Sep 17 00:00:00 2001 From: Sean Owen Date: Mon, 16 Mar 2015 23:58:52 -0700 Subject: [PATCH 074/122] SPARK-6044 [CORE] RDD.aggregate() should not use the closure serializer on the zero value Use configured serializer in RDD.aggregate, to match PairRDDFunctions.aggregateByKey, instead of closure serializer. Compare with https://github.com/apache/spark/blob/e60ad2f4c47b011be7a3198689ac2b82ee317d96/core/src/main/scala/org/apache/spark/rdd/PairRDDFunctions.scala#L127 Author: Sean Owen Closes #5028 from srowen/SPARK-6044 and squashes the following commits: a4040a7 [Sean Owen] Use configured serializer in RDD.aggregate, to match PairRDDFunctions.aggregateByKey, instead of closure serializer --- core/src/main/scala/org/apache/spark/rdd/RDD.scala | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/core/src/main/scala/org/apache/spark/rdd/RDD.scala b/core/src/main/scala/org/apache/spark/rdd/RDD.scala index cf0433010aa03..a139780d967e9 100644 --- a/core/src/main/scala/org/apache/spark/rdd/RDD.scala +++ b/core/src/main/scala/org/apache/spark/rdd/RDD.scala @@ -960,7 +960,7 @@ abstract class RDD[T: ClassTag]( */ def aggregate[U: ClassTag](zeroValue: U)(seqOp: (U, T) => U, combOp: (U, U) => U): U = { // Clone the zero value since we will also be serializing it as part of tasks - var jobResult = Utils.clone(zeroValue, sc.env.closureSerializer.newInstance()) + var jobResult = Utils.clone(zeroValue, sc.env.serializer.newInstance()) val cleanSeqOp = sc.clean(seqOp) val cleanCombOp = sc.clean(combOp) val aggregatePartition = (it: Iterator[T]) => it.aggregate(zeroValue)(cleanSeqOp, cleanCombOp) From 25f35806e307c9635e63b8b12698446a14bdd29d Mon Sep 17 00:00:00 2001 From: CodingCat Date: Tue, 17 Mar 2015 11:18:27 +0000 Subject: [PATCH 075/122] [SPARK-4011] tighten the visibility of the members in Master/Worker class https://issues.apache.org/jira/browse/SPARK-4011 Currently, most of the members in Master/Worker are with public accessibility. We might wish to tighten the accessibility of them a bit more discussion is here: https://github.com/apache/spark/pull/2828 Author: CodingCat Closes #4844 from CodingCat/SPARK-4011 and squashes the following commits: 1a64175 [CodingCat] fix compilation issue e7fd375 [CodingCat] Sean is right.... f5034a4 [CodingCat] fix rebase mistake 8d5b0c0 [CodingCat] loose more fields 0072f96 [CodingCat] lose some restrictions based on the possible design intention de77286 [CodingCat] tighten accessibility of deploy package 12b4fd3 [CodingCat] tighten accessibility of deploy.worker 1243bc7 [CodingCat] tighten accessibility of deploy.rest c5f622c [CodingCat] tighten the accessibility of deploy.history d441e20 [CodingCat] tighten accessibility of deploy.client 4e0ce4a [CodingCat] tighten the accessibility of the members of classes in master 23cddbb [CodingCat] stylistic fix 9a3a340 [CodingCat] tighten the access of worker class 67a0559 [CodingCat] tighten the access permission in Master --- .../apache/spark/deploy/ClientArguments.scala | 12 +- .../spark/deploy/DriverDescription.scala | 2 +- .../spark/deploy/ExecutorDescription.scala | 2 +- .../apache/spark/deploy/ExecutorState.scala | 2 +- .../spark/deploy/FaultToleranceTest.scala | 60 +++++----- .../apache/spark/deploy/JsonProtocol.scala | 2 +- .../org/apache/spark/deploy/SparkSubmit.scala | 38 +++--- .../spark/deploy/SparkSubmitArguments.scala | 2 +- .../spark/deploy/client/AppClient.scala | 18 +-- .../spark/deploy/client/TestClient.scala | 2 +- .../history/ApplicationHistoryProvider.scala | 4 +- .../spark/deploy/history/HistoryPage.scala | 2 +- .../history/HistoryServerArguments.scala | 3 +- .../spark/deploy/master/ApplicationInfo.scala | 19 +-- .../deploy/master/ApplicationSource.scala | 2 +- .../deploy/master/ApplicationState.scala | 2 +- .../spark/deploy/master/DriverInfo.scala | 2 +- .../spark/deploy/master/DriverState.scala | 2 +- .../spark/deploy/master/ExecutorDesc.scala | 2 +- .../master/FileSystemPersistenceEngine.scala | 2 +- .../apache/spark/deploy/master/Master.scala | 110 +++++++++--------- .../spark/deploy/master/MasterArguments.scala | 6 +- .../deploy/master/PersistenceEngine.scala | 2 +- .../deploy/master/RecoveryModeFactory.scala | 4 +- .../spark/deploy/master/RecoveryState.scala | 2 +- .../deploy/master/SparkCuratorUtil.scala | 10 +- .../spark/deploy/master/WorkerState.scala | 2 +- .../master/ZooKeeperLeaderElectionAgent.scala | 6 +- .../master/ZooKeeperPersistenceEngine.scala | 12 +- .../deploy/master/ui/ApplicationPage.scala | 2 +- .../master/ui/HistoryNotFoundPage.scala | 2 +- .../spark/deploy/master/ui/MasterPage.scala | 2 +- .../spark/deploy/master/ui/MasterWebUI.scala | 6 +- .../deploy/rest/StandaloneRestClient.scala | 14 +-- .../deploy/rest/StandaloneRestServer.scala | 2 +- .../rest/SubmitRestProtocolException.scala | 6 +- .../rest/SubmitRestProtocolMessage.scala | 2 +- .../rest/SubmitRestProtocolRequest.scala | 4 +- .../rest/SubmitRestProtocolResponse.scala | 10 +- .../spark/deploy/worker/CommandUtils.scala | 2 +- .../spark/deploy/worker/DriverRunner.scala | 24 ++-- .../spark/deploy/worker/ExecutorRunner.scala | 24 ++-- .../apache/spark/deploy/worker/Worker.scala | 85 +++++++------- .../spark/deploy/worker/WorkerArguments.scala | 4 +- .../spark/deploy/worker/WorkerSource.scala | 2 +- .../spark/deploy/worker/WorkerWatcher.scala | 2 +- .../spark/deploy/worker/ui/LogPage.scala | 2 +- .../spark/deploy/worker/ui/WorkerPage.scala | 7 +- .../spark/deploy/worker/ui/WorkerWebUI.scala | 6 +- 49 files changed, 277 insertions(+), 265 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/deploy/ClientArguments.scala b/core/src/main/scala/org/apache/spark/deploy/ClientArguments.scala index 415bd50591692..53bc62aff7395 100644 --- a/core/src/main/scala/org/apache/spark/deploy/ClientArguments.scala +++ b/core/src/main/scala/org/apache/spark/deploy/ClientArguments.scala @@ -28,7 +28,7 @@ import org.apache.spark.util.{IntParam, MemoryParam} /** * Command-line parser for the driver client. */ -private[spark] class ClientArguments(args: Array[String]) { +private[deploy] class ClientArguments(args: Array[String]) { import ClientArguments._ var cmd: String = "" // 'launch' or 'kill' @@ -96,7 +96,7 @@ private[spark] class ClientArguments(args: Array[String]) { /** * Print usage and exit JVM with the given exit code. */ - def printUsageAndExit(exitCode: Int) { + private def printUsageAndExit(exitCode: Int) { // TODO: It wouldn't be too hard to allow users to submit their app and dependency jars // separately similar to in the YARN client. val usage = @@ -116,10 +116,10 @@ private[spark] class ClientArguments(args: Array[String]) { } } -object ClientArguments { - private[spark] val DEFAULT_CORES = 1 - private[spark] val DEFAULT_MEMORY = 512 // MB - private[spark] val DEFAULT_SUPERVISE = false +private[deploy] object ClientArguments { + val DEFAULT_CORES = 1 + val DEFAULT_MEMORY = 512 // MB + val DEFAULT_SUPERVISE = false def isValidJarUrl(s: String): Boolean = { try { diff --git a/core/src/main/scala/org/apache/spark/deploy/DriverDescription.scala b/core/src/main/scala/org/apache/spark/deploy/DriverDescription.scala index b056a19ce6598..659fb434a80f5 100644 --- a/core/src/main/scala/org/apache/spark/deploy/DriverDescription.scala +++ b/core/src/main/scala/org/apache/spark/deploy/DriverDescription.scala @@ -17,7 +17,7 @@ package org.apache.spark.deploy -private[spark] class DriverDescription( +private[deploy] class DriverDescription( val jarUrl: String, val mem: Int, val cores: Int, diff --git a/core/src/main/scala/org/apache/spark/deploy/ExecutorDescription.scala b/core/src/main/scala/org/apache/spark/deploy/ExecutorDescription.scala index 2abf0b69dddb3..ec23371b52f93 100644 --- a/core/src/main/scala/org/apache/spark/deploy/ExecutorDescription.scala +++ b/core/src/main/scala/org/apache/spark/deploy/ExecutorDescription.scala @@ -22,7 +22,7 @@ package org.apache.spark.deploy * This state is sufficient for the Master to reconstruct its internal data structures during * failover. */ -private[spark] class ExecutorDescription( +private[deploy] class ExecutorDescription( val appId: String, val execId: Int, val cores: Int, diff --git a/core/src/main/scala/org/apache/spark/deploy/ExecutorState.scala b/core/src/main/scala/org/apache/spark/deploy/ExecutorState.scala index 9f34d01e6db48..efa88c62e1f5d 100644 --- a/core/src/main/scala/org/apache/spark/deploy/ExecutorState.scala +++ b/core/src/main/scala/org/apache/spark/deploy/ExecutorState.scala @@ -17,7 +17,7 @@ package org.apache.spark.deploy -private[spark] object ExecutorState extends Enumeration { +private[deploy] object ExecutorState extends Enumeration { val LAUNCHING, LOADING, RUNNING, KILLED, FAILED, LOST, EXITED = Value diff --git a/core/src/main/scala/org/apache/spark/deploy/FaultToleranceTest.scala b/core/src/main/scala/org/apache/spark/deploy/FaultToleranceTest.scala index 47dbcd87c35b5..4e58aa0ed4c7e 100644 --- a/core/src/main/scala/org/apache/spark/deploy/FaultToleranceTest.scala +++ b/core/src/main/scala/org/apache/spark/deploy/FaultToleranceTest.scala @@ -55,29 +55,29 @@ import org.apache.spark.deploy.master.{RecoveryState, SparkCuratorUtil} * - The docker images tagged spark-test-master and spark-test-worker are built from the * docker/ directory. Run 'docker/spark-test/build' to generate these. */ -private[spark] object FaultToleranceTest extends App with Logging { +private object FaultToleranceTest extends App with Logging { - val conf = new SparkConf() - val ZK_DIR = conf.get("spark.deploy.zookeeper.dir", "/spark") + private val conf = new SparkConf() + private val ZK_DIR = conf.get("spark.deploy.zookeeper.dir", "/spark") - val masters = ListBuffer[TestMasterInfo]() - val workers = ListBuffer[TestWorkerInfo]() - var sc: SparkContext = _ + private val masters = ListBuffer[TestMasterInfo]() + private val workers = ListBuffer[TestWorkerInfo]() + private var sc: SparkContext = _ - val zk = SparkCuratorUtil.newClient(conf) + private val zk = SparkCuratorUtil.newClient(conf) - var numPassed = 0 - var numFailed = 0 + private var numPassed = 0 + private var numFailed = 0 - val sparkHome = System.getenv("SPARK_HOME") + private val sparkHome = System.getenv("SPARK_HOME") assertTrue(sparkHome != null, "Run with a valid SPARK_HOME") - val containerSparkHome = "/opt/spark" - val dockerMountDir = "%s:%s".format(sparkHome, containerSparkHome) + private val containerSparkHome = "/opt/spark" + private val dockerMountDir = "%s:%s".format(sparkHome, containerSparkHome) System.setProperty("spark.driver.host", "172.17.42.1") // default docker host ip - def afterEach() { + private def afterEach() { if (sc != null) { sc.stop() sc = null @@ -179,7 +179,7 @@ private[spark] object FaultToleranceTest extends App with Logging { } } - def test(name: String)(fn: => Unit) { + private def test(name: String)(fn: => Unit) { try { fn numPassed += 1 @@ -197,19 +197,19 @@ private[spark] object FaultToleranceTest extends App with Logging { afterEach() } - def addMasters(num: Int) { + private def addMasters(num: Int) { logInfo(s">>>>> ADD MASTERS $num <<<<<") (1 to num).foreach { _ => masters += SparkDocker.startMaster(dockerMountDir) } } - def addWorkers(num: Int) { + private def addWorkers(num: Int) { logInfo(s">>>>> ADD WORKERS $num <<<<<") val masterUrls = getMasterUrls(masters) (1 to num).foreach { _ => workers += SparkDocker.startWorker(dockerMountDir, masterUrls) } } /** Creates a SparkContext, which constructs a Client to interact with our cluster. */ - def createClient() = { + private def createClient() = { logInfo(">>>>> CREATE CLIENT <<<<<") if (sc != null) { sc.stop() } // Counter-hack: Because of a hack in SparkEnv#create() that changes this @@ -218,17 +218,17 @@ private[spark] object FaultToleranceTest extends App with Logging { sc = new SparkContext(getMasterUrls(masters), "fault-tolerance", containerSparkHome) } - def getMasterUrls(masters: Seq[TestMasterInfo]): String = { + private def getMasterUrls(masters: Seq[TestMasterInfo]): String = { "spark://" + masters.map(master => master.ip + ":7077").mkString(",") } - def getLeader: TestMasterInfo = { + private def getLeader: TestMasterInfo = { val leaders = masters.filter(_.state == RecoveryState.ALIVE) assertTrue(leaders.size == 1) leaders(0) } - def killLeader(): Unit = { + private def killLeader(): Unit = { logInfo(">>>>> KILL LEADER <<<<<") masters.foreach(_.readState()) val leader = getLeader @@ -236,9 +236,9 @@ private[spark] object FaultToleranceTest extends App with Logging { leader.kill() } - def delay(secs: Duration = 5.seconds) = Thread.sleep(secs.toMillis) + private def delay(secs: Duration = 5.seconds) = Thread.sleep(secs.toMillis) - def terminateCluster() { + private def terminateCluster() { logInfo(">>>>> TERMINATE CLUSTER <<<<<") masters.foreach(_.kill()) workers.foreach(_.kill()) @@ -247,7 +247,7 @@ private[spark] object FaultToleranceTest extends App with Logging { } /** This includes Client retry logic, so it may take a while if the cluster is recovering. */ - def assertUsable() = { + private def assertUsable() = { val f = future { try { val res = sc.parallelize(0 until 10).collect() @@ -269,7 +269,7 @@ private[spark] object FaultToleranceTest extends App with Logging { * Asserts that the cluster is usable and that the expected masters and workers * are all alive in a proper configuration (e.g., only one leader). */ - def assertValidClusterState() = { + private def assertValidClusterState() = { logInfo(">>>>> ASSERT VALID CLUSTER STATE <<<<<") assertUsable() var numAlive = 0 @@ -325,7 +325,7 @@ private[spark] object FaultToleranceTest extends App with Logging { } } - def assertTrue(bool: Boolean, message: String = "") { + private def assertTrue(bool: Boolean, message: String = "") { if (!bool) { throw new IllegalStateException("Assertion failed: " + message) } @@ -335,7 +335,7 @@ private[spark] object FaultToleranceTest extends App with Logging { numFailed)) } -private[spark] class TestMasterInfo(val ip: String, val dockerId: DockerId, val logFile: File) +private class TestMasterInfo(val ip: String, val dockerId: DockerId, val logFile: File) extends Logging { implicit val formats = org.json4s.DefaultFormats @@ -377,7 +377,7 @@ private[spark] class TestMasterInfo(val ip: String, val dockerId: DockerId, val format(ip, dockerId.id, logFile.getAbsolutePath, state) } -private[spark] class TestWorkerInfo(val ip: String, val dockerId: DockerId, val logFile: File) +private class TestWorkerInfo(val ip: String, val dockerId: DockerId, val logFile: File) extends Logging { implicit val formats = org.json4s.DefaultFormats @@ -390,7 +390,7 @@ private[spark] class TestWorkerInfo(val ip: String, val dockerId: DockerId, val "[ip=%s, id=%s, logFile=%s]".format(ip, dockerId, logFile.getAbsolutePath) } -private[spark] object SparkDocker { +private object SparkDocker { def startMaster(mountDir: String): TestMasterInfo = { val cmd = Docker.makeRunCmd("spark-test-master", mountDir = mountDir) val (ip, id, outFile) = startNode(cmd) @@ -425,11 +425,11 @@ private[spark] object SparkDocker { } } -private[spark] class DockerId(val id: String) { +private class DockerId(val id: String) { override def toString = id } -private[spark] object Docker extends Logging { +private object Docker extends Logging { def makeRunCmd(imageTag: String, args: String = "", mountDir: String = ""): ProcessBuilder = { val mountCmd = if (mountDir != "") { " -v " + mountDir } else "" diff --git a/core/src/main/scala/org/apache/spark/deploy/JsonProtocol.scala b/core/src/main/scala/org/apache/spark/deploy/JsonProtocol.scala index 696f32a6f5730..458a7c3a455de 100644 --- a/core/src/main/scala/org/apache/spark/deploy/JsonProtocol.scala +++ b/core/src/main/scala/org/apache/spark/deploy/JsonProtocol.scala @@ -23,7 +23,7 @@ import org.apache.spark.deploy.DeployMessages.{MasterStateResponse, WorkerStateR import org.apache.spark.deploy.master.{ApplicationInfo, DriverInfo, WorkerInfo} import org.apache.spark.deploy.worker.ExecutorRunner -private[spark] object JsonProtocol { +private[deploy] object JsonProtocol { def writeWorkerInfo(obj: WorkerInfo) = { ("id" -> obj.id) ~ ("host" -> obj.host) ~ diff --git a/core/src/main/scala/org/apache/spark/deploy/SparkSubmit.scala b/core/src/main/scala/org/apache/spark/deploy/SparkSubmit.scala index 4a74641f4e1fa..4f506be63fe59 100644 --- a/core/src/main/scala/org/apache/spark/deploy/SparkSubmit.scala +++ b/core/src/main/scala/org/apache/spark/deploy/SparkSubmit.scala @@ -45,7 +45,7 @@ import org.apache.spark.util.{ChildFirstURLClassLoader, MutableURLClassLoader, U * Whether to submit, kill, or request the status of an application. * The latter two operations are currently supported only for standalone cluster mode. */ -private[spark] object SparkSubmitAction extends Enumeration { +private[deploy] object SparkSubmitAction extends Enumeration { type SparkSubmitAction = Value val SUBMIT, KILL, REQUEST_STATUS = Value } @@ -137,7 +137,7 @@ object SparkSubmit { * Second, we use this launch environment to invoke the main method of the child * main class. */ - private[spark] def submit(args: SparkSubmitArguments): Unit = { + private def submit(args: SparkSubmitArguments): Unit = { val (childArgs, childClasspath, sysProps, childMainClass) = prepareSubmitEnvironment(args) def doRunMain(): Unit = { @@ -199,7 +199,7 @@ object SparkSubmit { * (4) the main class for the child * Exposed for testing. */ - private[spark] def prepareSubmitEnvironment(args: SparkSubmitArguments) + private[deploy] def prepareSubmitEnvironment(args: SparkSubmitArguments) : (Seq[String], Seq[String], Map[String, String], String) = { // Return values val childArgs = new ArrayBuffer[String]() @@ -598,32 +598,32 @@ object SparkSubmit { /** * Return whether the given primary resource represents a shell. */ - private[spark] def isShell(primaryResource: String): Boolean = { + private[deploy] def isShell(primaryResource: String): Boolean = { primaryResource == SPARK_SHELL || primaryResource == PYSPARK_SHELL } /** * Return whether the given main class represents a sql shell. */ - private[spark] def isSqlShell(mainClass: String): Boolean = { + private def isSqlShell(mainClass: String): Boolean = { mainClass == "org.apache.spark.sql.hive.thriftserver.SparkSQLCLIDriver" } /** * Return whether the given main class represents a thrift server. */ - private[spark] def isThriftServer(mainClass: String): Boolean = { + private def isThriftServer(mainClass: String): Boolean = { mainClass == "org.apache.spark.sql.hive.thriftserver.HiveThriftServer2" } /** * Return whether the given primary resource requires running python. */ - private[spark] def isPython(primaryResource: String): Boolean = { + private[deploy] def isPython(primaryResource: String): Boolean = { primaryResource.endsWith(".py") || primaryResource == PYSPARK_SHELL } - private[spark] def isInternal(primaryResource: String): Boolean = { + private[deploy] def isInternal(primaryResource: String): Boolean = { primaryResource == SPARK_INTERNAL } @@ -631,7 +631,7 @@ object SparkSubmit { * Merge a sequence of comma-separated file lists, some of which may be null to indicate * no files, into a single comma-separated string. */ - private[spark] def mergeFileLists(lists: String*): String = { + private def mergeFileLists(lists: String*): String = { val merged = lists.filter(_ != null) .flatMap(_.split(",")) .mkString(",") @@ -640,10 +640,10 @@ object SparkSubmit { } /** Provides utility functions to be used inside SparkSubmit. */ -private[spark] object SparkSubmitUtils { +private[deploy] object SparkSubmitUtils { // Exposed for testing - private[spark] var printStream = SparkSubmit.printStream + var printStream = SparkSubmit.printStream /** * Represents a Maven Coordinate @@ -651,7 +651,7 @@ private[spark] object SparkSubmitUtils { * @param artifactId the artifactId of the coordinate * @param version the version of the coordinate */ - private[spark] case class MavenCoordinate(groupId: String, artifactId: String, version: String) + private[deploy] case class MavenCoordinate(groupId: String, artifactId: String, version: String) /** * Extracts maven coordinates from a comma-delimited string. Coordinates should be provided @@ -659,7 +659,7 @@ private[spark] object SparkSubmitUtils { * @param coordinates Comma-delimited string of maven coordinates * @return Sequence of Maven coordinates */ - private[spark] def extractMavenCoordinates(coordinates: String): Seq[MavenCoordinate] = { + def extractMavenCoordinates(coordinates: String): Seq[MavenCoordinate] = { coordinates.split(",").map { p => val splits = p.replace("/", ":").split(":") require(splits.length == 3, s"Provided Maven Coordinates must be in the form " + @@ -679,7 +679,7 @@ private[spark] object SparkSubmitUtils { * @param remoteRepos Comma-delimited string of remote repositories * @return A ChainResolver used by Ivy to search for and resolve dependencies. */ - private[spark] def createRepoResolvers(remoteRepos: Option[String]): ChainResolver = { + def createRepoResolvers(remoteRepos: Option[String]): ChainResolver = { // We need a chain resolver if we want to check multiple repositories val cr = new ChainResolver cr.setName("list") @@ -722,7 +722,7 @@ private[spark] object SparkSubmitUtils { * @param cacheDirectory directory where jars are cached * @return a comma-delimited list of paths for the dependencies */ - private[spark] def resolveDependencyPaths( + def resolveDependencyPaths( artifacts: Array[AnyRef], cacheDirectory: File): String = { artifacts.map { artifactInfo => @@ -734,7 +734,7 @@ private[spark] object SparkSubmitUtils { } /** Adds the given maven coordinates to Ivy's module descriptor. */ - private[spark] def addDependenciesToIvy( + def addDependenciesToIvy( md: DefaultModuleDescriptor, artifacts: Seq[MavenCoordinate], ivyConfName: String): Unit = { @@ -748,7 +748,7 @@ private[spark] object SparkSubmitUtils { } /** Add exclusion rules for dependencies already included in the spark-assembly */ - private[spark] def addExclusionRules( + def addExclusionRules( ivySettings: IvySettings, ivyConfName: String, md: DefaultModuleDescriptor): Unit = { @@ -777,7 +777,7 @@ private[spark] object SparkSubmitUtils { } /** A nice function to use in tests as well. Values are dummy strings. */ - private[spark] def getModuleDescriptor = DefaultModuleDescriptor.newDefaultInstance( + def getModuleDescriptor = DefaultModuleDescriptor.newDefaultInstance( ModuleRevisionId.newInstance("org.apache.spark", "spark-submit-parent", "1.0")) /** @@ -788,7 +788,7 @@ private[spark] object SparkSubmitUtils { * @return The comma-delimited path to the jars of the given maven artifacts including their * transitive dependencies */ - private[spark] def resolveMavenCoordinates( + def resolveMavenCoordinates( coordinates: String, remoteRepos: Option[String], ivyPath: Option[String], diff --git a/core/src/main/scala/org/apache/spark/deploy/SparkSubmitArguments.scala b/core/src/main/scala/org/apache/spark/deploy/SparkSubmitArguments.scala index 94e4bdbfb7d7b..2250d5a28e4ef 100644 --- a/core/src/main/scala/org/apache/spark/deploy/SparkSubmitArguments.scala +++ b/core/src/main/scala/org/apache/spark/deploy/SparkSubmitArguments.scala @@ -32,7 +32,7 @@ import org.apache.spark.util.Utils * Parses and encapsulates arguments from the spark-submit script. * The env argument is used for testing. */ -private[spark] class SparkSubmitArguments(args: Seq[String], env: Map[String, String] = sys.env) +private[deploy] class SparkSubmitArguments(args: Seq[String], env: Map[String, String] = sys.env) extends SparkSubmitArgumentsParser { var master: String = null var deployMode: String = null diff --git a/core/src/main/scala/org/apache/spark/deploy/client/AppClient.scala b/core/src/main/scala/org/apache/spark/deploy/client/AppClient.scala index ffe940fbda2fb..2d24083a77b73 100644 --- a/core/src/main/scala/org/apache/spark/deploy/client/AppClient.scala +++ b/core/src/main/scala/org/apache/spark/deploy/client/AppClient.scala @@ -47,18 +47,18 @@ private[spark] class AppClient( conf: SparkConf) extends Logging { - val masterAkkaUrls = masterUrls.map(Master.toAkkaUrl(_, AkkaUtils.protocol(actorSystem))) + private val masterAkkaUrls = masterUrls.map(Master.toAkkaUrl(_, AkkaUtils.protocol(actorSystem))) - val REGISTRATION_TIMEOUT = 20.seconds - val REGISTRATION_RETRIES = 3 + private val REGISTRATION_TIMEOUT = 20.seconds + private val REGISTRATION_RETRIES = 3 - var masterAddress: Address = null - var actor: ActorRef = null - var appId: String = null - var registered = false - var activeMasterUrl: String = null + private var masterAddress: Address = null + private var actor: ActorRef = null + private var appId: String = null + private var registered = false + private var activeMasterUrl: String = null - class ClientActor extends Actor with ActorLogReceive with Logging { + private class ClientActor extends Actor with ActorLogReceive with Logging { var master: ActorSelection = null var alreadyDisconnected = false // To avoid calling listener.disconnected() multiple times var alreadyDead = false // To avoid calling listener.dead() multiple times diff --git a/core/src/main/scala/org/apache/spark/deploy/client/TestClient.scala b/core/src/main/scala/org/apache/spark/deploy/client/TestClient.scala index 88a0862b96afe..c1c4812f17fbe 100644 --- a/core/src/main/scala/org/apache/spark/deploy/client/TestClient.scala +++ b/core/src/main/scala/org/apache/spark/deploy/client/TestClient.scala @@ -23,7 +23,7 @@ import org.apache.spark.util.{AkkaUtils, Utils} private[spark] object TestClient { - class TestListener extends AppClientListener with Logging { + private class TestListener extends AppClientListener with Logging { def connected(id: String) { logInfo("Connected to master, got app ID " + id) } diff --git a/core/src/main/scala/org/apache/spark/deploy/history/ApplicationHistoryProvider.scala b/core/src/main/scala/org/apache/spark/deploy/history/ApplicationHistoryProvider.scala index 553bf3cb945ab..ea6c85ee511d5 100644 --- a/core/src/main/scala/org/apache/spark/deploy/history/ApplicationHistoryProvider.scala +++ b/core/src/main/scala/org/apache/spark/deploy/history/ApplicationHistoryProvider.scala @@ -19,7 +19,7 @@ package org.apache.spark.deploy.history import org.apache.spark.ui.SparkUI -private[spark] case class ApplicationHistoryInfo( +private[history] case class ApplicationHistoryInfo( id: String, name: String, startTime: Long, @@ -28,7 +28,7 @@ private[spark] case class ApplicationHistoryInfo( sparkUser: String, completed: Boolean = false) -private[spark] abstract class ApplicationHistoryProvider { +private[history] abstract class ApplicationHistoryProvider { /** * Returns a list of applications available for the history server to show. diff --git a/core/src/main/scala/org/apache/spark/deploy/history/HistoryPage.scala b/core/src/main/scala/org/apache/spark/deploy/history/HistoryPage.scala index 26ebc75971c66..6e432d63c6b5a 100644 --- a/core/src/main/scala/org/apache/spark/deploy/history/HistoryPage.scala +++ b/core/src/main/scala/org/apache/spark/deploy/history/HistoryPage.scala @@ -23,7 +23,7 @@ import scala.xml.Node import org.apache.spark.ui.{WebUIPage, UIUtils} -private[spark] class HistoryPage(parent: HistoryServer) extends WebUIPage("") { +private[history] class HistoryPage(parent: HistoryServer) extends WebUIPage("") { private val pageSize = 20 private val plusOrMinus = 2 diff --git a/core/src/main/scala/org/apache/spark/deploy/history/HistoryServerArguments.scala b/core/src/main/scala/org/apache/spark/deploy/history/HistoryServerArguments.scala index b1270ade9f750..a2a97a7877ce7 100644 --- a/core/src/main/scala/org/apache/spark/deploy/history/HistoryServerArguments.scala +++ b/core/src/main/scala/org/apache/spark/deploy/history/HistoryServerArguments.scala @@ -23,7 +23,8 @@ import org.apache.spark.util.Utils /** * Command-line parser for the master. */ -private[spark] class HistoryServerArguments(conf: SparkConf, args: Array[String]) extends Logging { +private[history] class HistoryServerArguments(conf: SparkConf, args: Array[String]) + extends Logging { private var propertiesFile: String = null parse(args.toList) diff --git a/core/src/main/scala/org/apache/spark/deploy/master/ApplicationInfo.scala b/core/src/main/scala/org/apache/spark/deploy/master/ApplicationInfo.scala index a962dc4af2f6c..536aedb6f9fe9 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/ApplicationInfo.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/ApplicationInfo.scala @@ -28,7 +28,7 @@ import org.apache.spark.annotation.DeveloperApi import org.apache.spark.deploy.ApplicationDescription import org.apache.spark.util.Utils -private[spark] class ApplicationInfo( +private[deploy] class ApplicationInfo( val startTime: Long, val id: String, val desc: ApplicationDescription, @@ -75,14 +75,15 @@ private[spark] class ApplicationInfo( } } - def addExecutor(worker: WorkerInfo, cores: Int, useID: Option[Int] = None): ExecutorDesc = { + private[master] def addExecutor(worker: WorkerInfo, cores: Int, useID: Option[Int] = None): + ExecutorDesc = { val exec = new ExecutorDesc(newExecutorId(useID), this, worker, cores, desc.memoryPerSlave) executors(exec.id) = exec coresGranted += cores exec } - def removeExecutor(exec: ExecutorDesc) { + private[master] def removeExecutor(exec: ExecutorDesc) { if (executors.contains(exec.id)) { removedExecutors += executors(exec.id) executors -= exec.id @@ -90,22 +91,22 @@ private[spark] class ApplicationInfo( } } - val requestedCores = desc.maxCores.getOrElse(defaultCores) + private[master] val requestedCores = desc.maxCores.getOrElse(defaultCores) - def coresLeft: Int = requestedCores - coresGranted + private[master] def coresLeft: Int = requestedCores - coresGranted private var _retryCount = 0 - def retryCount = _retryCount + private[master] def retryCount = _retryCount - def incrementRetryCount() = { + private[master] def incrementRetryCount() = { _retryCount += 1 _retryCount } - def resetRetryCount() = _retryCount = 0 + private[master] def resetRetryCount() = _retryCount = 0 - def markFinished(endState: ApplicationState.Value) { + private[master] def markFinished(endState: ApplicationState.Value) { state = endState endTime = System.currentTimeMillis() } diff --git a/core/src/main/scala/org/apache/spark/deploy/master/ApplicationSource.scala b/core/src/main/scala/org/apache/spark/deploy/master/ApplicationSource.scala index 38db02cd2421b..017e8b55cbe7f 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/ApplicationSource.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/ApplicationSource.scala @@ -21,7 +21,7 @@ import com.codahale.metrics.{Gauge, MetricRegistry} import org.apache.spark.metrics.source.Source -class ApplicationSource(val application: ApplicationInfo) extends Source { +private[master] class ApplicationSource(val application: ApplicationInfo) extends Source { override val metricRegistry = new MetricRegistry() override val sourceName = "%s.%s.%s".format("application", application.desc.name, System.currentTimeMillis()) diff --git a/core/src/main/scala/org/apache/spark/deploy/master/ApplicationState.scala b/core/src/main/scala/org/apache/spark/deploy/master/ApplicationState.scala index f5b946329ae9b..37bfcdfdf4777 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/ApplicationState.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/ApplicationState.scala @@ -17,7 +17,7 @@ package org.apache.spark.deploy.master -private[spark] object ApplicationState extends Enumeration { +private[master] object ApplicationState extends Enumeration { type ApplicationState = Value diff --git a/core/src/main/scala/org/apache/spark/deploy/master/DriverInfo.scala b/core/src/main/scala/org/apache/spark/deploy/master/DriverInfo.scala index 9d3d7938c6ccb..b197dbcbfe294 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/DriverInfo.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/DriverInfo.scala @@ -23,7 +23,7 @@ import org.apache.spark.annotation.DeveloperApi import org.apache.spark.deploy.DriverDescription import org.apache.spark.util.Utils -private[spark] class DriverInfo( +private[deploy] class DriverInfo( val startTime: Long, val id: String, val desc: DriverDescription, diff --git a/core/src/main/scala/org/apache/spark/deploy/master/DriverState.scala b/core/src/main/scala/org/apache/spark/deploy/master/DriverState.scala index 26a68bade3c60..35ff33a61653c 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/DriverState.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/DriverState.scala @@ -17,7 +17,7 @@ package org.apache.spark.deploy.master -private[spark] object DriverState extends Enumeration { +private[deploy] object DriverState extends Enumeration { type DriverState = Value diff --git a/core/src/main/scala/org/apache/spark/deploy/master/ExecutorDesc.scala b/core/src/main/scala/org/apache/spark/deploy/master/ExecutorDesc.scala index 5d620dfcabad5..fc62b094def67 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/ExecutorDesc.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/ExecutorDesc.scala @@ -19,7 +19,7 @@ package org.apache.spark.deploy.master import org.apache.spark.deploy.{ExecutorDescription, ExecutorState} -private[spark] class ExecutorDesc( +private[master] class ExecutorDesc( val id: Int, val application: ApplicationInfo, val worker: WorkerInfo, diff --git a/core/src/main/scala/org/apache/spark/deploy/master/FileSystemPersistenceEngine.scala b/core/src/main/scala/org/apache/spark/deploy/master/FileSystemPersistenceEngine.scala index 36a2e2c6a6349..d2d30bfd7fcba 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/FileSystemPersistenceEngine.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/FileSystemPersistenceEngine.scala @@ -33,7 +33,7 @@ import org.apache.spark.Logging * @param dir Directory to store files. Created if non-existent (but not recursively). * @param serialization Used to serialize our objects. */ -private[spark] class FileSystemPersistenceEngine( +private[master] class FileSystemPersistenceEngine( val dir: String, val serialization: Serialization) extends PersistenceEngine with Logging { diff --git a/core/src/main/scala/org/apache/spark/deploy/master/Master.scala b/core/src/main/scala/org/apache/spark/deploy/master/Master.scala index 22935c9b1d394..1b42121c8db05 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/Master.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/Master.scala @@ -49,7 +49,7 @@ import org.apache.spark.scheduler.{EventLoggingListener, ReplayListenerBus} import org.apache.spark.ui.SparkUI import org.apache.spark.util.{ActorLogReceive, AkkaUtils, SignalLogger, Utils} -private[spark] class Master( +private[master] class Master( host: String, port: Int, webUiPort: Int, @@ -59,65 +59,68 @@ private[spark] class Master( import context.dispatcher // to use Akka's scheduler.schedule() - val hadoopConf = SparkHadoopUtil.get.newConfiguration(conf) + private val hadoopConf = SparkHadoopUtil.get.newConfiguration(conf) - def createDateFormat = new SimpleDateFormat("yyyyMMddHHmmss") // For application IDs - val WORKER_TIMEOUT = conf.getLong("spark.worker.timeout", 60) * 1000 - val RETAINED_APPLICATIONS = conf.getInt("spark.deploy.retainedApplications", 200) - val RETAINED_DRIVERS = conf.getInt("spark.deploy.retainedDrivers", 200) - val REAPER_ITERATIONS = conf.getInt("spark.dead.worker.persistence", 15) - val RECOVERY_MODE = conf.get("spark.deploy.recoveryMode", "NONE") + private def createDateFormat = new SimpleDateFormat("yyyyMMddHHmmss") // For application IDs + + private val WORKER_TIMEOUT = conf.getLong("spark.worker.timeout", 60) * 1000 + private val RETAINED_APPLICATIONS = conf.getInt("spark.deploy.retainedApplications", 200) + private val RETAINED_DRIVERS = conf.getInt("spark.deploy.retainedDrivers", 200) + private val REAPER_ITERATIONS = conf.getInt("spark.dead.worker.persistence", 15) + private val RECOVERY_MODE = conf.get("spark.deploy.recoveryMode", "NONE") val workers = new HashSet[WorkerInfo] - val idToWorker = new HashMap[String, WorkerInfo] - val addressToWorker = new HashMap[Address, WorkerInfo] - - val apps = new HashSet[ApplicationInfo] val idToApp = new HashMap[String, ApplicationInfo] - val actorToApp = new HashMap[ActorRef, ApplicationInfo] - val addressToApp = new HashMap[Address, ApplicationInfo] val waitingApps = new ArrayBuffer[ApplicationInfo] - val completedApps = new ArrayBuffer[ApplicationInfo] - var nextAppNumber = 0 - val appIdToUI = new HashMap[String, SparkUI] + val apps = new HashSet[ApplicationInfo] + + private val idToWorker = new HashMap[String, WorkerInfo] + private val addressToWorker = new HashMap[Address, WorkerInfo] + + private val actorToApp = new HashMap[ActorRef, ApplicationInfo] + private val addressToApp = new HashMap[Address, ApplicationInfo] + private val completedApps = new ArrayBuffer[ApplicationInfo] + private var nextAppNumber = 0 + private val appIdToUI = new HashMap[String, SparkUI] - val drivers = new HashSet[DriverInfo] - val completedDrivers = new ArrayBuffer[DriverInfo] - val waitingDrivers = new ArrayBuffer[DriverInfo] // Drivers currently spooled for scheduling - var nextDriverNumber = 0 + private val drivers = new HashSet[DriverInfo] + private val completedDrivers = new ArrayBuffer[DriverInfo] + // Drivers currently spooled for scheduling + private val waitingDrivers = new ArrayBuffer[DriverInfo] + private var nextDriverNumber = 0 Utils.checkHost(host, "Expected hostname") - val masterMetricsSystem = MetricsSystem.createMetricsSystem("master", conf, securityMgr) - val applicationMetricsSystem = MetricsSystem.createMetricsSystem("applications", conf, + private val masterMetricsSystem = MetricsSystem.createMetricsSystem("master", conf, securityMgr) + private val applicationMetricsSystem = MetricsSystem.createMetricsSystem("applications", conf, securityMgr) - val masterSource = new MasterSource(this) + private val masterSource = new MasterSource(this) - val webUi = new MasterWebUI(this, webUiPort) + private val webUi = new MasterWebUI(this, webUiPort) - val masterPublicAddress = { + private val masterPublicAddress = { val envVar = conf.getenv("SPARK_PUBLIC_DNS") if (envVar != null) envVar else host } - val masterUrl = "spark://" + host + ":" + port - var masterWebUiUrl: String = _ + private val masterUrl = "spark://" + host + ":" + port + private var masterWebUiUrl: String = _ - var state = RecoveryState.STANDBY + private var state = RecoveryState.STANDBY - var persistenceEngine: PersistenceEngine = _ + private var persistenceEngine: PersistenceEngine = _ - var leaderElectionAgent: LeaderElectionAgent = _ + private var leaderElectionAgent: LeaderElectionAgent = _ private var recoveryCompletionTask: Cancellable = _ // As a temporary workaround before better ways of configuring memory, we allow users to set // a flag that will perform round-robin scheduling across the nodes (spreading out each app // among all the nodes) instead of trying to consolidate each app onto a small # of nodes. - val spreadOutApps = conf.getBoolean("spark.deploy.spreadOut", true) + private val spreadOutApps = conf.getBoolean("spark.deploy.spreadOut", true) // Default maxCores for applications that don't specify it (i.e. pass Int.MaxValue) - val defaultCores = conf.getInt("spark.deploy.defaultCores", Int.MaxValue) + private val defaultCores = conf.getInt("spark.deploy.defaultCores", Int.MaxValue) if (defaultCores < 1) { throw new SparkException("spark.deploy.defaultCores must be positive") } @@ -449,11 +452,11 @@ private[spark] class Master( } } - def canCompleteRecovery = + private def canCompleteRecovery = workers.count(_.state == WorkerState.UNKNOWN) == 0 && apps.count(_.state == ApplicationState.UNKNOWN) == 0 - def beginRecovery(storedApps: Seq[ApplicationInfo], storedDrivers: Seq[DriverInfo], + private def beginRecovery(storedApps: Seq[ApplicationInfo], storedDrivers: Seq[DriverInfo], storedWorkers: Seq[WorkerInfo]) { for (app <- storedApps) { logInfo("Trying to recover app: " + app.id) @@ -484,7 +487,7 @@ private[spark] class Master( } } - def completeRecovery() { + private def completeRecovery() { // Ensure "only-once" recovery semantics using a short synchronization period. synchronized { if (state != RecoveryState.RECOVERING) { return } @@ -517,7 +520,7 @@ private[spark] class Master( * launched an executor for the app on it (right now the standalone backend doesn't like having * two executors on the same worker). */ - def canUse(app: ApplicationInfo, worker: WorkerInfo): Boolean = { + private def canUse(app: ApplicationInfo, worker: WorkerInfo): Boolean = { worker.memoryFree >= app.desc.memoryPerSlave && !worker.hasExecutor(app) } @@ -596,7 +599,7 @@ private[spark] class Master( } } - def launchExecutor(worker: WorkerInfo, exec: ExecutorDesc) { + private def launchExecutor(worker: WorkerInfo, exec: ExecutorDesc) { logInfo("Launching executor " + exec.fullId + " on worker " + worker.id) worker.addExecutor(exec) worker.actor ! LaunchExecutor(masterUrl, @@ -605,7 +608,7 @@ private[spark] class Master( exec.id, worker.id, worker.hostPort, exec.cores, exec.memory) } - def registerWorker(worker: WorkerInfo): Boolean = { + private def registerWorker(worker: WorkerInfo): Boolean = { // There may be one or more refs to dead workers on this same node (w/ different ID's), // remove them. workers.filter { w => @@ -633,7 +636,7 @@ private[spark] class Master( true } - def removeWorker(worker: WorkerInfo) { + private def removeWorker(worker: WorkerInfo) { logInfo("Removing worker " + worker.id + " on " + worker.host + ":" + worker.port) worker.setState(WorkerState.DEAD) idToWorker -= worker.id @@ -656,20 +659,20 @@ private[spark] class Master( persistenceEngine.removeWorker(worker) } - def relaunchDriver(driver: DriverInfo) { + private def relaunchDriver(driver: DriverInfo) { driver.worker = None driver.state = DriverState.RELAUNCHING waitingDrivers += driver schedule() } - def createApplication(desc: ApplicationDescription, driver: ActorRef): ApplicationInfo = { + private def createApplication(desc: ApplicationDescription, driver: ActorRef): ApplicationInfo = { val now = System.currentTimeMillis() val date = new Date(now) new ApplicationInfo(now, newApplicationId(date), desc, date, driver, defaultCores) } - def registerApplication(app: ApplicationInfo): Unit = { + private def registerApplication(app: ApplicationInfo): Unit = { val appAddress = app.driver.path.address if (addressToApp.contains(appAddress)) { logInfo("Attempted to re-register application at same address: " + appAddress) @@ -684,7 +687,7 @@ private[spark] class Master( waitingApps += app } - def finishApplication(app: ApplicationInfo) { + private def finishApplication(app: ApplicationInfo) { removeApplication(app, ApplicationState.FINISHED) } @@ -732,7 +735,7 @@ private[spark] class Master( * Rebuild a new SparkUI from the given application's event logs. * Return whether this is successful. */ - def rebuildSparkUI(app: ApplicationInfo): Boolean = { + private def rebuildSparkUI(app: ApplicationInfo): Boolean = { val appName = app.desc.name val notFoundBasePath = HistoryServer.UI_PATH_PREFIX + "/not-found" try { @@ -798,14 +801,14 @@ private[spark] class Master( } /** Generate a new app ID given a app's submission date */ - def newApplicationId(submitDate: Date): String = { + private def newApplicationId(submitDate: Date): String = { val appId = "app-%s-%04d".format(createDateFormat.format(submitDate), nextAppNumber) nextAppNumber += 1 appId } /** Check for, and remove, any timed-out workers */ - def timeOutDeadWorkers() { + private def timeOutDeadWorkers() { // Copy the workers into an array so we don't modify the hashset while iterating through it val currentTime = System.currentTimeMillis() val toRemove = workers.filter(_.lastHeartbeat < currentTime - WORKER_TIMEOUT).toArray @@ -822,19 +825,19 @@ private[spark] class Master( } } - def newDriverId(submitDate: Date): String = { + private def newDriverId(submitDate: Date): String = { val appId = "driver-%s-%04d".format(createDateFormat.format(submitDate), nextDriverNumber) nextDriverNumber += 1 appId } - def createDriver(desc: DriverDescription): DriverInfo = { + private def createDriver(desc: DriverDescription): DriverInfo = { val now = System.currentTimeMillis() val date = new Date(now) new DriverInfo(now, newDriverId(date), desc, date) } - def launchDriver(worker: WorkerInfo, driver: DriverInfo) { + private def launchDriver(worker: WorkerInfo, driver: DriverInfo) { logInfo("Launching driver " + driver.id + " on worker " + worker.id) worker.addDriver(driver) driver.worker = Some(worker) @@ -842,7 +845,10 @@ private[spark] class Master( driver.state = DriverState.RUNNING } - def removeDriver(driverId: String, finalState: DriverState, exception: Option[Exception]) { + private def removeDriver( + driverId: String, + finalState: DriverState, + exception: Option[Exception]) { drivers.find(d => d.id == driverId) match { case Some(driver) => logInfo(s"Removing driver: $driverId") @@ -863,7 +869,7 @@ private[spark] class Master( } } -private[spark] object Master extends Logging { +private[deploy] object Master extends Logging { val systemName = "sparkMaster" private val actorName = "Master" diff --git a/core/src/main/scala/org/apache/spark/deploy/master/MasterArguments.scala b/core/src/main/scala/org/apache/spark/deploy/master/MasterArguments.scala index e34bee7854292..435b9b12f83b8 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/MasterArguments.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/MasterArguments.scala @@ -23,7 +23,7 @@ import org.apache.spark.util.{IntParam, Utils} /** * Command-line parser for the master. */ -private[spark] class MasterArguments(args: Array[String], conf: SparkConf) { +private[master] class MasterArguments(args: Array[String], conf: SparkConf) { var host = Utils.localHostName() var port = 7077 var webUiPort = 8080 @@ -49,7 +49,7 @@ private[spark] class MasterArguments(args: Array[String], conf: SparkConf) { webUiPort = conf.get("spark.master.ui.port").toInt } - def parse(args: List[String]): Unit = args match { + private def parse(args: List[String]): Unit = args match { case ("--ip" | "-i") :: value :: tail => Utils.checkHost(value, "ip no longer supported, please use hostname " + value) host = value @@ -84,7 +84,7 @@ private[spark] class MasterArguments(args: Array[String], conf: SparkConf) { /** * Print usage and exit JVM with the given exit code. */ - def printUsageAndExit(exitCode: Int) { + private def printUsageAndExit(exitCode: Int) { System.err.println( "Usage: Master [options]\n" + "\n" + diff --git a/core/src/main/scala/org/apache/spark/deploy/master/PersistenceEngine.scala b/core/src/main/scala/org/apache/spark/deploy/master/PersistenceEngine.scala index 2e0e1e7036ac8..da5060778edeb 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/PersistenceEngine.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/PersistenceEngine.scala @@ -87,7 +87,7 @@ trait PersistenceEngine { def close() {} } -private[spark] class BlackHolePersistenceEngine extends PersistenceEngine { +private[master] class BlackHolePersistenceEngine extends PersistenceEngine { override def persist(name: String, obj: Object): Unit = {} diff --git a/core/src/main/scala/org/apache/spark/deploy/master/RecoveryModeFactory.scala b/core/src/main/scala/org/apache/spark/deploy/master/RecoveryModeFactory.scala index 1096eb0368357..1583bf1f60032 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/RecoveryModeFactory.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/RecoveryModeFactory.scala @@ -49,7 +49,7 @@ abstract class StandaloneRecoveryModeFactory(conf: SparkConf, serializer: Serial * LeaderAgent in this case is a no-op. Since leader is forever leader as the actual * recovery is made by restoring from filesystem. */ -private[spark] class FileSystemRecoveryModeFactory(conf: SparkConf, serializer: Serialization) +private[master] class FileSystemRecoveryModeFactory(conf: SparkConf, serializer: Serialization) extends StandaloneRecoveryModeFactory(conf, serializer) with Logging { val RECOVERY_DIR = conf.get("spark.deploy.recoveryDirectory", "") @@ -61,7 +61,7 @@ private[spark] class FileSystemRecoveryModeFactory(conf: SparkConf, serializer: def createLeaderElectionAgent(master: LeaderElectable) = new MonarchyLeaderAgent(master) } -private[spark] class ZooKeeperRecoveryModeFactory(conf: SparkConf, serializer: Serialization) +private[master] class ZooKeeperRecoveryModeFactory(conf: SparkConf, serializer: Serialization) extends StandaloneRecoveryModeFactory(conf, serializer) { def createPersistenceEngine() = new ZooKeeperPersistenceEngine(conf, serializer) diff --git a/core/src/main/scala/org/apache/spark/deploy/master/RecoveryState.scala b/core/src/main/scala/org/apache/spark/deploy/master/RecoveryState.scala index 256a5a7c28e47..aa0f02fa625cc 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/RecoveryState.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/RecoveryState.scala @@ -17,7 +17,7 @@ package org.apache.spark.deploy.master -private[spark] object RecoveryState extends Enumeration { +private[deploy] object RecoveryState extends Enumeration { type MasterState = Value val STANDBY, ALIVE, RECOVERING, COMPLETING_RECOVERY = Value diff --git a/core/src/main/scala/org/apache/spark/deploy/master/SparkCuratorUtil.scala b/core/src/main/scala/org/apache/spark/deploy/master/SparkCuratorUtil.scala index 4781a80d470e1..5b22481ea8c5f 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/SparkCuratorUtil.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/SparkCuratorUtil.scala @@ -25,12 +25,12 @@ import org.apache.zookeeper.KeeperException import org.apache.spark.{Logging, SparkConf} -object SparkCuratorUtil extends Logging { +private[deploy] object SparkCuratorUtil extends Logging { - val ZK_CONNECTION_TIMEOUT_MILLIS = 15000 - val ZK_SESSION_TIMEOUT_MILLIS = 60000 - val RETRY_WAIT_MILLIS = 5000 - val MAX_RECONNECT_ATTEMPTS = 3 + private val ZK_CONNECTION_TIMEOUT_MILLIS = 15000 + private val ZK_SESSION_TIMEOUT_MILLIS = 60000 + private val RETRY_WAIT_MILLIS = 5000 + private val MAX_RECONNECT_ATTEMPTS = 3 def newClient(conf: SparkConf): CuratorFramework = { val ZK_URL = conf.get("spark.deploy.zookeeper.url") diff --git a/core/src/main/scala/org/apache/spark/deploy/master/WorkerState.scala b/core/src/main/scala/org/apache/spark/deploy/master/WorkerState.scala index 0b36ef60051fc..b60baaadfb4bc 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/WorkerState.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/WorkerState.scala @@ -17,7 +17,7 @@ package org.apache.spark.deploy.master -private[spark] object WorkerState extends Enumeration { +private[master] object WorkerState extends Enumeration { type WorkerState = Value val ALIVE, DEAD, DECOMMISSIONED, UNKNOWN = Value diff --git a/core/src/main/scala/org/apache/spark/deploy/master/ZooKeeperLeaderElectionAgent.scala b/core/src/main/scala/org/apache/spark/deploy/master/ZooKeeperLeaderElectionAgent.scala index 8eaa0ad948519..4823fd7cac0cb 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/ZooKeeperLeaderElectionAgent.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/ZooKeeperLeaderElectionAgent.scala @@ -24,7 +24,7 @@ import org.apache.spark.deploy.master.MasterMessages._ import org.apache.curator.framework.CuratorFramework import org.apache.curator.framework.recipes.leader.{LeaderLatchListener, LeaderLatch} -private[spark] class ZooKeeperLeaderElectionAgent(val masterActor: LeaderElectable, +private[master] class ZooKeeperLeaderElectionAgent(val masterActor: LeaderElectable, conf: SparkConf) extends LeaderLatchListener with LeaderElectionAgent with Logging { val WORKING_DIR = conf.get("spark.deploy.zookeeper.dir", "/spark") + "/leader_election" @@ -35,7 +35,7 @@ private[spark] class ZooKeeperLeaderElectionAgent(val masterActor: LeaderElectab start() - def start() { + private def start() { logInfo("Starting ZooKeeper LeaderElection agent") zk = SparkCuratorUtil.newClient(conf) leaderLatch = new LeaderLatch(zk, WORKING_DIR) @@ -72,7 +72,7 @@ private[spark] class ZooKeeperLeaderElectionAgent(val masterActor: LeaderElectab } } - def updateLeadershipStatus(isLeader: Boolean) { + private def updateLeadershipStatus(isLeader: Boolean) { if (isLeader && status == LeadershipStatus.NOT_LEADER) { status = LeadershipStatus.LEADER masterActor.electedLeader() diff --git a/core/src/main/scala/org/apache/spark/deploy/master/ZooKeeperPersistenceEngine.scala b/core/src/main/scala/org/apache/spark/deploy/master/ZooKeeperPersistenceEngine.scala index e11ac031fb9c6..1ac6677ad2b6d 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/ZooKeeperPersistenceEngine.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/ZooKeeperPersistenceEngine.scala @@ -28,12 +28,12 @@ import org.apache.zookeeper.CreateMode import org.apache.spark.{Logging, SparkConf} -private[spark] class ZooKeeperPersistenceEngine(conf: SparkConf, val serialization: Serialization) +private[master] class ZooKeeperPersistenceEngine(conf: SparkConf, val serialization: Serialization) extends PersistenceEngine - with Logging -{ - val WORKING_DIR = conf.get("spark.deploy.zookeeper.dir", "/spark") + "/master_status" - val zk: CuratorFramework = SparkCuratorUtil.newClient(conf) + with Logging { + + private val WORKING_DIR = conf.get("spark.deploy.zookeeper.dir", "/spark") + "/master_status" + private val zk: CuratorFramework = SparkCuratorUtil.newClient(conf) SparkCuratorUtil.mkdir(zk, WORKING_DIR) @@ -61,7 +61,7 @@ private[spark] class ZooKeeperPersistenceEngine(conf: SparkConf, val serializati zk.create().withMode(CreateMode.PERSISTENT).forPath(path, serialized) } - def deserializeFromFile[T](filename: String)(implicit m: ClassTag[T]): Option[T] = { + private def deserializeFromFile[T](filename: String)(implicit m: ClassTag[T]): Option[T] = { val fileData = zk.getData().forPath(WORKING_DIR + "/" + filename) val clazz = m.runtimeClass.asInstanceOf[Class[T]] val serializer = serialization.serializerFor(clazz) diff --git a/core/src/main/scala/org/apache/spark/deploy/master/ui/ApplicationPage.scala b/core/src/main/scala/org/apache/spark/deploy/master/ui/ApplicationPage.scala index 76fc40e17d9a8..761aa8f7b1ef6 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/ui/ApplicationPage.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/ui/ApplicationPage.scala @@ -32,7 +32,7 @@ import org.apache.spark.deploy.master.ExecutorDesc import org.apache.spark.ui.{UIUtils, WebUIPage} import org.apache.spark.util.Utils -private[spark] class ApplicationPage(parent: MasterWebUI) extends WebUIPage("app") { +private[ui] class ApplicationPage(parent: MasterWebUI) extends WebUIPage("app") { private val master = parent.masterActorRef private val timeout = parent.timeout diff --git a/core/src/main/scala/org/apache/spark/deploy/master/ui/HistoryNotFoundPage.scala b/core/src/main/scala/org/apache/spark/deploy/master/ui/HistoryNotFoundPage.scala index d8daff3e7fb9c..e021f1eef794f 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/ui/HistoryNotFoundPage.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/ui/HistoryNotFoundPage.scala @@ -24,7 +24,7 @@ import scala.xml.Node import org.apache.spark.ui.{UIUtils, WebUIPage} -private[spark] class HistoryNotFoundPage(parent: MasterWebUI) +private[ui] class HistoryNotFoundPage(parent: MasterWebUI) extends WebUIPage("history/not-found") { /** diff --git a/core/src/main/scala/org/apache/spark/deploy/master/ui/MasterPage.scala b/core/src/main/scala/org/apache/spark/deploy/master/ui/MasterPage.scala index c086cadca2c7d..dee2e4a447c6e 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/ui/MasterPage.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/ui/MasterPage.scala @@ -31,7 +31,7 @@ import org.apache.spark.deploy.master._ import org.apache.spark.ui.{WebUIPage, UIUtils} import org.apache.spark.util.Utils -private[spark] class MasterPage(parent: MasterWebUI) extends WebUIPage("") { +private[ui] class MasterPage(parent: MasterWebUI) extends WebUIPage("") { private val master = parent.masterActorRef private val timeout = parent.timeout diff --git a/core/src/main/scala/org/apache/spark/deploy/master/ui/MasterWebUI.scala b/core/src/main/scala/org/apache/spark/deploy/master/ui/MasterWebUI.scala index 170f90a00ad2a..1b670418ab1ff 100644 --- a/core/src/main/scala/org/apache/spark/deploy/master/ui/MasterWebUI.scala +++ b/core/src/main/scala/org/apache/spark/deploy/master/ui/MasterWebUI.scala @@ -26,7 +26,7 @@ import org.apache.spark.util.AkkaUtils /** * Web UI server for the standalone master. */ -private[spark] +private[master] class MasterWebUI(val master: Master, requestedPort: Int) extends WebUI(master.securityMgr, requestedPort, master.conf, name = "MasterUI") with Logging { @@ -62,6 +62,6 @@ class MasterWebUI(val master: Master, requestedPort: Int) } } -private[spark] object MasterWebUI { - val STATIC_RESOURCE_DIR = SparkUI.STATIC_RESOURCE_DIR +private[master] object MasterWebUI { + private val STATIC_RESOURCE_DIR = SparkUI.STATIC_RESOURCE_DIR } diff --git a/core/src/main/scala/org/apache/spark/deploy/rest/StandaloneRestClient.scala b/core/src/main/scala/org/apache/spark/deploy/rest/StandaloneRestClient.scala index c4be1f19e8e9f..420442f7564cc 100644 --- a/core/src/main/scala/org/apache/spark/deploy/rest/StandaloneRestClient.scala +++ b/core/src/main/scala/org/apache/spark/deploy/rest/StandaloneRestClient.scala @@ -52,7 +52,7 @@ import org.apache.spark.{Logging, SparkConf, SPARK_VERSION => sparkVersion} * implementation of this client can use that information to retry using the version specified * by the server. */ -private[spark] class StandaloneRestClient extends Logging { +private[deploy] class StandaloneRestClient extends Logging { import StandaloneRestClient._ /** @@ -61,7 +61,7 @@ private[spark] class StandaloneRestClient extends Logging { * If the submission was successful, poll the status of the submission and report * it to the user. Otherwise, report the error message provided by the server. */ - def createSubmission( + private[rest] def createSubmission( master: String, request: CreateSubmissionRequest): SubmitRestProtocolResponse = { logInfo(s"Submitting a request to launch an application in $master.") @@ -106,7 +106,7 @@ private[spark] class StandaloneRestClient extends Logging { } /** Construct a message that captures the specified parameters for submitting an application. */ - def constructSubmitRequest( + private[rest] def constructSubmitRequest( appResource: String, mainClass: String, appArgs: Array[String], @@ -291,16 +291,16 @@ private[spark] class StandaloneRestClient extends Logging { } } -private[spark] object StandaloneRestClient { - val REPORT_DRIVER_STATUS_INTERVAL = 1000 - val REPORT_DRIVER_STATUS_MAX_TRIES = 10 +private[rest] object StandaloneRestClient { + private val REPORT_DRIVER_STATUS_INTERVAL = 1000 + private val REPORT_DRIVER_STATUS_MAX_TRIES = 10 val PROTOCOL_VERSION = "v1" /** * Submit an application, assuming Spark parameters are specified through the given config. * This is abstracted to its own method for testing purposes. */ - private[rest] def run( + def run( appResource: String, mainClass: String, appArgs: Array[String], diff --git a/core/src/main/scala/org/apache/spark/deploy/rest/StandaloneRestServer.scala b/core/src/main/scala/org/apache/spark/deploy/rest/StandaloneRestServer.scala index f9e0478e4f874..4f19af59f409f 100644 --- a/core/src/main/scala/org/apache/spark/deploy/rest/StandaloneRestServer.scala +++ b/core/src/main/scala/org/apache/spark/deploy/rest/StandaloneRestServer.scala @@ -58,7 +58,7 @@ import org.apache.spark.deploy.ClientArguments._ * @param masterUrl the URL of the Master new drivers will attempt to connect to * @param masterConf the conf used by the Master */ -private[spark] class StandaloneRestServer( +private[deploy] class StandaloneRestServer( host: String, requestedPort: Int, masterActor: ActorRef, diff --git a/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolException.scala b/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolException.scala index d7a0bdbe10778..b97921ec934a0 100644 --- a/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolException.scala +++ b/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolException.scala @@ -20,17 +20,17 @@ package org.apache.spark.deploy.rest /** * An exception thrown in the REST application submission protocol. */ -private[spark] class SubmitRestProtocolException(message: String, cause: Throwable = null) +private[rest] class SubmitRestProtocolException(message: String, cause: Throwable = null) extends Exception(message, cause) /** * An exception thrown if a field is missing from a [[SubmitRestProtocolMessage]]. */ -private[spark] class SubmitRestMissingFieldException(message: String) +private[rest] class SubmitRestMissingFieldException(message: String) extends SubmitRestProtocolException(message) /** * An exception thrown if the REST client cannot reach the REST server. */ -private[spark] class SubmitRestConnectionException(message: String, cause: Throwable) +private[deploy] class SubmitRestConnectionException(message: String, cause: Throwable) extends SubmitRestProtocolException(message, cause) diff --git a/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolMessage.scala b/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolMessage.scala index 8f36635674a28..e6615a3174ce1 100644 --- a/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolMessage.scala +++ b/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolMessage.scala @@ -39,7 +39,7 @@ import org.apache.spark.util.Utils @JsonInclude(Include.NON_NULL) @JsonAutoDetect(getterVisibility = Visibility.ANY, setterVisibility = Visibility.ANY) @JsonPropertyOrder(alphabetic = true) -private[spark] abstract class SubmitRestProtocolMessage { +private[rest] abstract class SubmitRestProtocolMessage { @JsonIgnore val messageType = Utils.getFormattedClassName(this) diff --git a/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolRequest.scala b/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolRequest.scala index 9e1fd8c40cabd..d80abdf15fb34 100644 --- a/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolRequest.scala +++ b/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolRequest.scala @@ -24,7 +24,7 @@ import org.apache.spark.util.Utils /** * An abstract request sent from the client in the REST application submission protocol. */ -private[spark] abstract class SubmitRestProtocolRequest extends SubmitRestProtocolMessage { +private[rest] abstract class SubmitRestProtocolRequest extends SubmitRestProtocolMessage { var clientSparkVersion: String = null protected override def doValidate(): Unit = { super.doValidate() @@ -35,7 +35,7 @@ private[spark] abstract class SubmitRestProtocolRequest extends SubmitRestProtoc /** * A request to launch a new application in the REST application submission protocol. */ -private[spark] class CreateSubmissionRequest extends SubmitRestProtocolRequest { +private[rest] class CreateSubmissionRequest extends SubmitRestProtocolRequest { var appResource: String = null var mainClass: String = null var appArgs: Array[String] = null diff --git a/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolResponse.scala b/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolResponse.scala index 16dfe041d4bea..8fde8c142a4c1 100644 --- a/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolResponse.scala +++ b/core/src/main/scala/org/apache/spark/deploy/rest/SubmitRestProtocolResponse.scala @@ -22,7 +22,7 @@ import java.lang.Boolean /** * An abstract response sent from the server in the REST application submission protocol. */ -private[spark] abstract class SubmitRestProtocolResponse extends SubmitRestProtocolMessage { +private[rest] abstract class SubmitRestProtocolResponse extends SubmitRestProtocolMessage { var serverSparkVersion: String = null var success: Boolean = null var unknownFields: Array[String] = null @@ -35,7 +35,7 @@ private[spark] abstract class SubmitRestProtocolResponse extends SubmitRestProto /** * A response to a [[CreateSubmissionRequest]] in the REST application submission protocol. */ -private[spark] class CreateSubmissionResponse extends SubmitRestProtocolResponse { +private[rest] class CreateSubmissionResponse extends SubmitRestProtocolResponse { var submissionId: String = null protected override def doValidate(): Unit = { super.doValidate() @@ -46,7 +46,7 @@ private[spark] class CreateSubmissionResponse extends SubmitRestProtocolResponse /** * A response to a kill request in the REST application submission protocol. */ -private[spark] class KillSubmissionResponse extends SubmitRestProtocolResponse { +private[rest] class KillSubmissionResponse extends SubmitRestProtocolResponse { var submissionId: String = null protected override def doValidate(): Unit = { super.doValidate() @@ -58,7 +58,7 @@ private[spark] class KillSubmissionResponse extends SubmitRestProtocolResponse { /** * A response to a status request in the REST application submission protocol. */ -private[spark] class SubmissionStatusResponse extends SubmitRestProtocolResponse { +private[rest] class SubmissionStatusResponse extends SubmitRestProtocolResponse { var submissionId: String = null var driverState: String = null var workerId: String = null @@ -74,7 +74,7 @@ private[spark] class SubmissionStatusResponse extends SubmitRestProtocolResponse /** * An error response message used in the REST application submission protocol. */ -private[spark] class ErrorResponse extends SubmitRestProtocolResponse { +private[rest] class ErrorResponse extends SubmitRestProtocolResponse { // The highest protocol version that the server knows about // This is set when the client specifies an unknown version var highestProtocolVersion: String = null diff --git a/core/src/main/scala/org/apache/spark/deploy/worker/CommandUtils.scala b/core/src/main/scala/org/apache/spark/deploy/worker/CommandUtils.scala index 83f78cf47306c..0a1d60f58bc58 100644 --- a/core/src/main/scala/org/apache/spark/deploy/worker/CommandUtils.scala +++ b/core/src/main/scala/org/apache/spark/deploy/worker/CommandUtils.scala @@ -31,7 +31,7 @@ import org.apache.spark.util.Utils /** ** Utilities for running commands with the spark classpath. */ -private[spark] +private[deploy] object CommandUtils extends Logging { /** diff --git a/core/src/main/scala/org/apache/spark/deploy/worker/DriverRunner.scala b/core/src/main/scala/org/apache/spark/deploy/worker/DriverRunner.scala index e16bccb24d2c4..27a9eabb1ede7 100644 --- a/core/src/main/scala/org/apache/spark/deploy/worker/DriverRunner.scala +++ b/core/src/main/scala/org/apache/spark/deploy/worker/DriverRunner.scala @@ -37,8 +37,8 @@ import org.apache.spark.util.{Clock, SystemClock} * Manages the execution of one driver, including automatically restarting the driver on failure. * This is currently only used in standalone cluster deploy mode. */ -private[spark] class DriverRunner( - val conf: SparkConf, +private[deploy] class DriverRunner( + conf: SparkConf, val driverId: String, val workDir: File, val sparkHome: File, @@ -47,24 +47,24 @@ private[spark] class DriverRunner( val workerUrl: String) extends Logging { - @volatile var process: Option[Process] = None - @volatile var killed = false + @volatile private var process: Option[Process] = None + @volatile private var killed = false // Populated once finished - var finalState: Option[DriverState] = None - var finalException: Option[Exception] = None - var finalExitCode: Option[Int] = None + private[worker] var finalState: Option[DriverState] = None + private[worker] var finalException: Option[Exception] = None + private var finalExitCode: Option[Int] = None // Decoupled for testing - private[deploy] def setClock(_clock: Clock) = clock = _clock - private[deploy] def setSleeper(_sleeper: Sleeper) = sleeper = _sleeper + def setClock(_clock: Clock) = clock = _clock + def setSleeper(_sleeper: Sleeper) = sleeper = _sleeper private var clock: Clock = new SystemClock() private var sleeper = new Sleeper { def sleep(seconds: Int): Unit = (0 until seconds).takeWhile(f => {Thread.sleep(1000); !killed}) } /** Starts a thread to run and manage the driver. */ - def start() = { + private[worker] def start() = { new Thread("DriverRunner for " + driverId) { override def run() { try { @@ -106,7 +106,7 @@ private[spark] class DriverRunner( } /** Terminate this driver (or prevent it from ever starting if not yet started) */ - def kill() { + private[worker] def kill() { synchronized { process.foreach(p => p.destroy()) killed = true @@ -169,7 +169,7 @@ private[spark] class DriverRunner( runCommandWithRetry(ProcessBuilderLike(builder), initialize, supervise) } - private[deploy] def runCommandWithRetry(command: ProcessBuilderLike, initialize: Process => Unit, + def runCommandWithRetry(command: ProcessBuilderLike, initialize: Process => Unit, supervise: Boolean) { // Time to wait between submission retries. var waitSeconds = 1 diff --git a/core/src/main/scala/org/apache/spark/deploy/worker/ExecutorRunner.scala b/core/src/main/scala/org/apache/spark/deploy/worker/ExecutorRunner.scala index 023f3c6269062..83e24a7a1f80c 100644 --- a/core/src/main/scala/org/apache/spark/deploy/worker/ExecutorRunner.scala +++ b/core/src/main/scala/org/apache/spark/deploy/worker/ExecutorRunner.scala @@ -34,7 +34,7 @@ import org.apache.spark.util.logging.FileAppender * Manages the execution of one executor process. * This is currently only used in standalone mode. */ -private[spark] class ExecutorRunner( +private[deploy] class ExecutorRunner( val appId: String, val execId: Int, val appDesc: ApplicationDescription, @@ -48,22 +48,22 @@ private[spark] class ExecutorRunner( val sparkHome: File, val executorDir: File, val workerUrl: String, - val conf: SparkConf, + conf: SparkConf, val appLocalDirs: Seq[String], var state: ExecutorState.Value) extends Logging { - val fullId = appId + "/" + execId - var workerThread: Thread = null - var process: Process = null - var stdoutAppender: FileAppender = null - var stderrAppender: FileAppender = null + private val fullId = appId + "/" + execId + private var workerThread: Thread = null + private var process: Process = null + private var stdoutAppender: FileAppender = null + private var stderrAppender: FileAppender = null // NOTE: This is now redundant with the automated shut-down enforced by the Executor. It might // make sense to remove this in the future. - var shutdownHook: Thread = null + private var shutdownHook: Thread = null - def start() { + private[worker] def start() { workerThread = new Thread("ExecutorRunner for " + fullId) { override def run() { fetchAndRunExecutor() } } @@ -99,7 +99,7 @@ private[spark] class ExecutorRunner( } /** Stop this executor runner, including killing the process it launched */ - def kill() { + private[worker] def kill() { if (workerThread != null) { // the workerThread will kill the child process when interrupted workerThread.interrupt() @@ -114,7 +114,7 @@ private[spark] class ExecutorRunner( } /** Replace variables such as {{EXECUTOR_ID}} and {{CORES}} in a command argument passed to us */ - def substituteVariables(argument: String): String = argument match { + private[worker] def substituteVariables(argument: String): String = argument match { case "{{WORKER_URL}}" => workerUrl case "{{EXECUTOR_ID}}" => execId.toString case "{{HOSTNAME}}" => host @@ -126,7 +126,7 @@ private[spark] class ExecutorRunner( /** * Download and run the executor described in our ApplicationDescription */ - def fetchAndRunExecutor() { + private def fetchAndRunExecutor() { try { // Launch the process val builder = CommandUtils.buildProcessBuilder(appDesc.command, memory, diff --git a/core/src/main/scala/org/apache/spark/deploy/worker/Worker.scala b/core/src/main/scala/org/apache/spark/deploy/worker/Worker.scala index f2e7418f4bf15..c1b0a295f9f74 100755 --- a/core/src/main/scala/org/apache/spark/deploy/worker/Worker.scala +++ b/core/src/main/scala/org/apache/spark/deploy/worker/Worker.scala @@ -42,7 +42,7 @@ import org.apache.spark.util.{ActorLogReceive, AkkaUtils, SignalLogger, Utils} /** * @param masterAkkaUrls Each url should be a valid akka url. */ -private[spark] class Worker( +private[worker] class Worker( host: String, port: Int, webUiPort: Int, @@ -60,85 +60,90 @@ private[spark] class Worker( Utils.checkHost(host, "Expected hostname") assert (port > 0) - def createDateFormat = new SimpleDateFormat("yyyyMMddHHmmss") // For worker and executor IDs + // For worker and executor IDs + private def createDateFormat = new SimpleDateFormat("yyyyMMddHHmmss") // Send a heartbeat every (heartbeat timeout) / 4 milliseconds - val HEARTBEAT_MILLIS = conf.getLong("spark.worker.timeout", 60) * 1000 / 4 + private val HEARTBEAT_MILLIS = conf.getLong("spark.worker.timeout", 60) * 1000 / 4 // Model retries to connect to the master, after Hadoop's model. // The first six attempts to reconnect are in shorter intervals (between 5 and 15 seconds) // Afterwards, the next 10 attempts are between 30 and 90 seconds. // A bit of randomness is introduced so that not all of the workers attempt to reconnect at // the same time. - val INITIAL_REGISTRATION_RETRIES = 6 - val TOTAL_REGISTRATION_RETRIES = INITIAL_REGISTRATION_RETRIES + 10 - val FUZZ_MULTIPLIER_INTERVAL_LOWER_BOUND = 0.500 - val REGISTRATION_RETRY_FUZZ_MULTIPLIER = { + private val INITIAL_REGISTRATION_RETRIES = 6 + private val TOTAL_REGISTRATION_RETRIES = INITIAL_REGISTRATION_RETRIES + 10 + private val FUZZ_MULTIPLIER_INTERVAL_LOWER_BOUND = 0.500 + private val REGISTRATION_RETRY_FUZZ_MULTIPLIER = { val randomNumberGenerator = new Random(UUID.randomUUID.getMostSignificantBits) randomNumberGenerator.nextDouble + FUZZ_MULTIPLIER_INTERVAL_LOWER_BOUND } - val INITIAL_REGISTRATION_RETRY_INTERVAL = (math.round(10 * + private val INITIAL_REGISTRATION_RETRY_INTERVAL = (math.round(10 * REGISTRATION_RETRY_FUZZ_MULTIPLIER)).seconds - val PROLONGED_REGISTRATION_RETRY_INTERVAL = (math.round(60 + private val PROLONGED_REGISTRATION_RETRY_INTERVAL = (math.round(60 * REGISTRATION_RETRY_FUZZ_MULTIPLIER)).seconds - val CLEANUP_ENABLED = conf.getBoolean("spark.worker.cleanup.enabled", false) + private val CLEANUP_ENABLED = conf.getBoolean("spark.worker.cleanup.enabled", false) // How often worker will clean up old app folders - val CLEANUP_INTERVAL_MILLIS = conf.getLong("spark.worker.cleanup.interval", 60 * 30) * 1000 + private val CLEANUP_INTERVAL_MILLIS = + conf.getLong("spark.worker.cleanup.interval", 60 * 30) * 1000 // TTL for app folders/data; after TTL expires it will be cleaned up - val APP_DATA_RETENTION_SECS = conf.getLong("spark.worker.cleanup.appDataTtl", 7 * 24 * 3600) - - val testing: Boolean = sys.props.contains("spark.testing") - var master: ActorSelection = null - var masterAddress: Address = null - var activeMasterUrl: String = "" - var activeMasterWebUiUrl : String = "" - val akkaUrl = AkkaUtils.address( + private val APP_DATA_RETENTION_SECS = + conf.getLong("spark.worker.cleanup.appDataTtl", 7 * 24 * 3600) + + private val testing: Boolean = sys.props.contains("spark.testing") + private var master: ActorSelection = null + private var masterAddress: Address = null + private var activeMasterUrl: String = "" + private[worker] var activeMasterWebUiUrl : String = "" + private val akkaUrl = AkkaUtils.address( AkkaUtils.protocol(context.system), actorSystemName, host, port, actorName) - @volatile var registered = false - @volatile var connected = false - val workerId = generateWorkerId() - val sparkHome = + @volatile private var registered = false + @volatile private var connected = false + private val workerId = generateWorkerId() + private val sparkHome = if (testing) { assert(sys.props.contains("spark.test.home"), "spark.test.home is not set!") new File(sys.props("spark.test.home")) } else { new File(sys.env.get("SPARK_HOME").getOrElse(".")) } + var workDir: File = null - val executors = new HashMap[String, ExecutorRunner] val finishedExecutors = new HashMap[String, ExecutorRunner] val drivers = new HashMap[String, DriverRunner] + val executors = new HashMap[String, ExecutorRunner] val finishedDrivers = new HashMap[String, DriverRunner] val appDirectories = new HashMap[String, Seq[String]] val finishedApps = new HashSet[String] // The shuffle service is not actually started unless configured. - val shuffleService = new StandaloneWorkerShuffleService(conf, securityMgr) + private val shuffleService = new StandaloneWorkerShuffleService(conf, securityMgr) - val publicAddress = { + private val publicAddress = { val envVar = conf.getenv("SPARK_PUBLIC_DNS") if (envVar != null) envVar else host } - var webUi: WorkerWebUI = null + private var webUi: WorkerWebUI = null - var coresUsed = 0 - var memoryUsed = 0 - var connectionAttemptCount = 0 + private var connectionAttemptCount = 0 - val metricsSystem = MetricsSystem.createMetricsSystem("worker", conf, securityMgr) - val workerSource = new WorkerSource(this) + private val metricsSystem = MetricsSystem.createMetricsSystem("worker", conf, securityMgr) + private val workerSource = new WorkerSource(this) + + private var registrationRetryTimer: Option[Cancellable] = None - var registrationRetryTimer: Option[Cancellable] = None + var coresUsed = 0 + var memoryUsed = 0 def coresFree: Int = cores - coresUsed def memoryFree: Int = memory - memoryUsed - def createWorkDir() { + private def createWorkDir() { workDir = Option(workDirPath).map(new File(_)).getOrElse(new File(sparkHome, "work")) try { // This sporadically fails - not sure why ... !workDir.exists() && !workDir.mkdirs() @@ -175,7 +180,7 @@ private[spark] class Worker( metricsSystem.getServletHandlers.foreach(webUi.attachHandler) } - def changeMaster(url: String, uiUrl: String) { + private def changeMaster(url: String, uiUrl: String) { // activeMasterUrl it's a valid Spark url since we receive it from master. activeMasterUrl = url activeMasterWebUiUrl = uiUrl @@ -252,7 +257,7 @@ private[spark] class Worker( } } - def registerWithMaster() { + private def registerWithMaster() { // DisassociatedEvent may be triggered multiple times, so don't attempt registration // if there are outstanding registration attempts scheduled. registrationRetryTimer match { @@ -506,7 +511,7 @@ private[spark] class Worker( } } - def generateWorkerId(): String = { + private def generateWorkerId(): String = { "worker-%s-%s-%d".format(createDateFormat.format(new Date), host, port) } @@ -521,7 +526,7 @@ private[spark] class Worker( } } -private[spark] object Worker extends Logging { +private[deploy] object Worker extends Logging { def main(argStrings: Array[String]) { SignalLogger.register(log) val conf = new SparkConf @@ -554,7 +559,7 @@ private[spark] object Worker extends Logging { (actorSystem, boundPort) } - private[spark] def isUseLocalNodeSSLConfig(cmd: Command): Boolean = { + def isUseLocalNodeSSLConfig(cmd: Command): Boolean = { val pattern = """\-Dspark\.ssl\.useNodeLocalConf\=(.+)""".r val result = cmd.javaOpts.collectFirst { case pattern(_result) => _result.toBoolean @@ -562,7 +567,7 @@ private[spark] object Worker extends Logging { result.getOrElse(false) } - private[spark] def maybeUpdateSSLSettings(cmd: Command, conf: SparkConf): Command = { + def maybeUpdateSSLSettings(cmd: Command, conf: SparkConf): Command = { val prefix = "spark.ssl." val useNLC = "spark.ssl.useNodeLocalConf" if (isUseLocalNodeSSLConfig(cmd)) { diff --git a/core/src/main/scala/org/apache/spark/deploy/worker/WorkerArguments.scala b/core/src/main/scala/org/apache/spark/deploy/worker/WorkerArguments.scala index 019cd70f2a229..88f9d880ac209 100644 --- a/core/src/main/scala/org/apache/spark/deploy/worker/WorkerArguments.scala +++ b/core/src/main/scala/org/apache/spark/deploy/worker/WorkerArguments.scala @@ -25,7 +25,7 @@ import org.apache.spark.SparkConf /** * Command-line parser for the worker. */ -private[spark] class WorkerArguments(args: Array[String], conf: SparkConf) { +private[worker] class WorkerArguments(args: Array[String], conf: SparkConf) { var host = Utils.localHostName() var port = 0 var webUiPort = 8081 @@ -63,7 +63,7 @@ private[spark] class WorkerArguments(args: Array[String], conf: SparkConf) { checkWorkerMemory() - def parse(args: List[String]): Unit = args match { + private def parse(args: List[String]): Unit = args match { case ("--ip" | "-i") :: value :: tail => Utils.checkHost(value, "ip no longer supported, please use hostname " + value) host = value diff --git a/core/src/main/scala/org/apache/spark/deploy/worker/WorkerSource.scala b/core/src/main/scala/org/apache/spark/deploy/worker/WorkerSource.scala index df1e01b23b932..b36023bc40c3d 100644 --- a/core/src/main/scala/org/apache/spark/deploy/worker/WorkerSource.scala +++ b/core/src/main/scala/org/apache/spark/deploy/worker/WorkerSource.scala @@ -21,7 +21,7 @@ import com.codahale.metrics.{Gauge, MetricRegistry} import org.apache.spark.metrics.source.Source -private[spark] class WorkerSource(val worker: Worker) extends Source { +private[worker] class WorkerSource(val worker: Worker) extends Source { override val sourceName = "worker" override val metricRegistry = new MetricRegistry() diff --git a/core/src/main/scala/org/apache/spark/deploy/worker/WorkerWatcher.scala b/core/src/main/scala/org/apache/spark/deploy/worker/WorkerWatcher.scala index 63a8ac817b618..09d866fb0cd90 100644 --- a/core/src/main/scala/org/apache/spark/deploy/worker/WorkerWatcher.scala +++ b/core/src/main/scala/org/apache/spark/deploy/worker/WorkerWatcher.scala @@ -48,7 +48,7 @@ private[spark] class WorkerWatcher(workerUrl: String) private val expectedHostPort = AddressFromURIString(workerUrl).hostPort private def isWorker(address: Address) = address.hostPort == expectedHostPort - def exitNonZero() = if (isTesting) isShutDown = true else System.exit(-1) + private def exitNonZero() = if (isTesting) isShutDown = true else System.exit(-1) override def receiveWithLogging = { case AssociatedEvent(localAddress, remoteAddress, inbound) if isWorker(remoteAddress) => diff --git a/core/src/main/scala/org/apache/spark/deploy/worker/ui/LogPage.scala b/core/src/main/scala/org/apache/spark/deploy/worker/ui/LogPage.scala index ecb358c399819..88170d4df3053 100644 --- a/core/src/main/scala/org/apache/spark/deploy/worker/ui/LogPage.scala +++ b/core/src/main/scala/org/apache/spark/deploy/worker/ui/LogPage.scala @@ -26,7 +26,7 @@ import org.apache.spark.util.Utils import org.apache.spark.Logging import org.apache.spark.util.logging.RollingFileAppender -private[spark] class LogPage(parent: WorkerWebUI) extends WebUIPage("logPage") with Logging { +private[ui] class LogPage(parent: WorkerWebUI) extends WebUIPage("logPage") with Logging { private val worker = parent.worker private val workDir = parent.workDir diff --git a/core/src/main/scala/org/apache/spark/deploy/worker/ui/WorkerPage.scala b/core/src/main/scala/org/apache/spark/deploy/worker/ui/WorkerPage.scala index 720f13bfa829b..9f9f27d71e1ae 100644 --- a/core/src/main/scala/org/apache/spark/deploy/worker/ui/WorkerPage.scala +++ b/core/src/main/scala/org/apache/spark/deploy/worker/ui/WorkerPage.scala @@ -31,10 +31,9 @@ import org.apache.spark.deploy.worker.{DriverRunner, ExecutorRunner} import org.apache.spark.ui.{WebUIPage, UIUtils} import org.apache.spark.util.Utils -private[spark] class WorkerPage(parent: WorkerWebUI) extends WebUIPage("") { - val workerActor = parent.worker.self - val worker = parent.worker - val timeout = parent.timeout +private[ui] class WorkerPage(parent: WorkerWebUI) extends WebUIPage("") { + private val workerActor = parent.worker.self + private val timeout = parent.timeout override def renderJson(request: HttpServletRequest): JValue = { val stateFuture = (workerActor ? RequestWorkerState)(timeout).mapTo[WorkerStateResponse] diff --git a/core/src/main/scala/org/apache/spark/deploy/worker/ui/WorkerWebUI.scala b/core/src/main/scala/org/apache/spark/deploy/worker/ui/WorkerWebUI.scala index 7ac81a2d87efd..de6423beb543e 100644 --- a/core/src/main/scala/org/apache/spark/deploy/worker/ui/WorkerWebUI.scala +++ b/core/src/main/scala/org/apache/spark/deploy/worker/ui/WorkerWebUI.scala @@ -30,7 +30,7 @@ import org.apache.spark.util.AkkaUtils /** * Web UI server for the standalone worker. */ -private[spark] +private[worker] class WorkerWebUI( val worker: Worker, val workDir: File, @@ -38,7 +38,7 @@ class WorkerWebUI( extends WebUI(worker.securityMgr, requestedPort, worker.conf, name = "WorkerUI") with Logging { - val timeout = AkkaUtils.askTimeout(worker.conf) + private[ui] val timeout = AkkaUtils.askTimeout(worker.conf) initialize() @@ -53,6 +53,6 @@ class WorkerWebUI( } } -private[spark] object WorkerWebUI { +private[ui] object WorkerWebUI { val STATIC_RESOURCE_BASE = SparkUI.STATIC_RESOURCE_DIR } From 005d1c5f290decc606a0be59fb191136dafc0c9d Mon Sep 17 00:00:00 2001 From: mcheah Date: Tue, 17 Mar 2015 11:20:20 +0000 Subject: [PATCH 076/122] [SPARK-6269] [CORE] Use ScalaRunTime's array methods instead of java.lang.reflect.Array in size estimation This patch switches the usage of java.lang.reflect.Array in Size estimation to using scala's RunTime array-getter methods. The notes on https://bugs.openjdk.java.net/browse/JDK-8051447 tipped me off to the fact that using java.lang.reflect.Array was not ideal. At first, I used the code from that ticket, but it turns out that ScalaRunTime's array-related methods avoid the bottleneck of invoking native code anyways, so that was sufficient to boost performance in size estimation. The idea is to use pure Java code in implementing the methods there, as opposed to relying on native C code which ends up being ill-performing. This improves the performance of estimating the size of arrays when we are checking for spilling in Spark. Here's the benchmark discussion from the ticket: I did two tests. The first, less convincing, take-with-a-block-of-salt test I did was do a simple groupByKey operation to collect objects in a 4.0 GB text file RDD into 30,000 buckets. I ran 1 Master and 4 Spark Worker JVMs on my mac, fetching the RDD from a text file simply stored on disk, and saving it out to another file located on local disk. The wall clock times I got back before and after the change were: Before: 352.195s, 343.871s, 359.080s After (using code directly from the JDK ticket, not the scala code in this PR): 342.929583s, 329.456623s, 326.151481s So, there is a bit of an improvement after the change. I also did some YourKit profiling of the executors to get an idea of how much time was spent in size estimation before and after the change. I roughly saw that size estimation took up less of the time after my change, but YourKit's profiling can be inconsistent and who knows if I was profiling the executors that had the same data between runs? The more convincing test I did was to run the size-estimation logic itself in an isolated unit test. I ran the following code: ``` val bigArray = Array.fill(1000)(Array.fill(1000)(java.util.UUID.randomUUID().toString())) test("String arrays only perf testing") { val startTime = System.currentTimeMillis() for (i <- 1 to 50000) { SizeEstimator.estimate(bigArray) } println("Runtime: " + (System.currentTimeMillis() - startTime) / 1000.0000) } ``` I wanted to use a 2D array specifically because I wanted to measure the performance of repeatedly calling Array.getLength. I used UUID-Strings to ensure that the strings were randomized (so String object re-use doesn't happen), but that they would all be the same size. The results were as follows: Before PR: 222.681 s, 218.34 s, 211.739s After latest change: 170.715 s, 176.775 s, 180.298 s . Author: mcheah Author: Justin Uang Closes #4972 from mccheah/feature/spark-6269-reflect-array and squashes the following commits: 8527852 [mcheah] Respect CamelCase for numElementsDrawn 18d4b50 [mcheah] Addressing style comments - while loops instead of for loops 16ce534 [mcheah] Organizing imports properly db890ea [mcheah] Removing CastedArray and just using ScalaRunTime. cb67ce2 [mcheah] Fixing a scalastyle error - line too long 5d53c4c [mcheah] Removing unused parameter in visitArray. 6467759 [mcheah] Including primitive size information inside CastedArray. 93f4b05 [mcheah] Using Scala instead of Java for the array-reflection implementation. a557ab8 [mcheah] Using a wrapper around arrays to do casting only once ca063fc [mcheah] Fixing a compiler error made while refactoring style 1fe09de [Justin Uang] [SPARK-6269] Use a different implementation of java.lang.reflect.Array --- .../org/apache/spark/util/SizeEstimator.scala | 28 ++++++++++--------- 1 file changed, 15 insertions(+), 13 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/util/SizeEstimator.scala b/core/src/main/scala/org/apache/spark/util/SizeEstimator.scala index bce3b3afe9aba..26ffbf9350388 100644 --- a/core/src/main/scala/org/apache/spark/util/SizeEstimator.scala +++ b/core/src/main/scala/org/apache/spark/util/SizeEstimator.scala @@ -18,18 +18,16 @@ package org.apache.spark.util import java.lang.management.ManagementFactory -import java.lang.reflect.{Array => JArray} -import java.lang.reflect.Field -import java.lang.reflect.Modifier -import java.util.IdentityHashMap -import java.util.Random +import java.lang.reflect.{Field, Modifier} +import java.util.{IdentityHashMap, Random} import java.util.concurrent.ConcurrentHashMap - import scala.collection.mutable.ArrayBuffer +import scala.runtime.ScalaRunTime import org.apache.spark.Logging import org.apache.spark.util.collection.OpenHashSet + /** * Estimates the sizes of Java objects (number of bytes of memory they occupy), for use in * memory-aware caches. @@ -184,9 +182,9 @@ private[spark] object SizeEstimator extends Logging { private val ARRAY_SIZE_FOR_SAMPLING = 200 private val ARRAY_SAMPLE_SIZE = 100 // should be lower than ARRAY_SIZE_FOR_SAMPLING - private def visitArray(array: AnyRef, cls: Class[_], state: SearchState) { - val length = JArray.getLength(array) - val elementClass = cls.getComponentType + private def visitArray(array: AnyRef, arrayClass: Class[_], state: SearchState) { + val length = ScalaRunTime.array_length(array) + val elementClass = arrayClass.getComponentType() // Arrays have object header and length field which is an integer var arrSize: Long = alignSize(objectSize + INT_SIZE) @@ -199,22 +197,26 @@ private[spark] object SizeEstimator extends Logging { state.size += arrSize if (length <= ARRAY_SIZE_FOR_SAMPLING) { - for (i <- 0 until length) { - state.enqueue(JArray.get(array, i)) + var arrayIndex = 0 + while (arrayIndex < length) { + state.enqueue(ScalaRunTime.array_apply(array, arrayIndex).asInstanceOf[AnyRef]) + arrayIndex += 1 } } else { // Estimate the size of a large array by sampling elements without replacement. var size = 0.0 val rand = new Random(42) val drawn = new OpenHashSet[Int](ARRAY_SAMPLE_SIZE) - for (i <- 0 until ARRAY_SAMPLE_SIZE) { + var numElementsDrawn = 0 + while (numElementsDrawn < ARRAY_SAMPLE_SIZE) { var index = 0 do { index = rand.nextInt(length) } while (drawn.contains(index)) drawn.add(index) - val elem = JArray.get(array, index) + val elem = ScalaRunTime.array_apply(array, index).asInstanceOf[AnyRef] size += SizeEstimator.estimate(elem, state.visited) + numElementsDrawn += 1 } state.size += ((length / (ARRAY_SAMPLE_SIZE * 1.0)) * size).toLong } From e26db9be47b295c4ec9e651f4cce321ba1fddfcd Mon Sep 17 00:00:00 2001 From: Theodore Vasiloudis Date: Tue, 17 Mar 2015 11:25:01 +0000 Subject: [PATCH 077/122] [docs] [SPARK-4820] Spark build encounters "File name too long" on some encrypted filesystems Added a note instructing users how to build Spark in an encrypted file system. Author: Theodore Vasiloudis Closes #5041 from thvasilo/patch-2 and squashes the following commits: 09d890b [Theodore Vasiloudis] Workaroung for buiding in an encrypted filesystem --- docs/building-spark.md | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/docs/building-spark.md b/docs/building-spark.md index 57d0ca834f460..ea79c5bc276d3 100644 --- a/docs/building-spark.md +++ b/docs/building-spark.md @@ -23,6 +23,18 @@ build/mvn -Pyarn -Phadoop-2.4 -Dhadoop.version=2.4.0 -DskipTests clean package Other build examples can be found below. +**Note:** When building on an encrypted filesystem (if your home directory is encrypted, for example), then the Spark build might fail with a "Filename too long" error. As a workaround, add the following in the configuration args of the `scala-maven-plugin` in the project `pom.xml`: + + -Xmax-classfile-name + 128 + +and in `project/SparkBuild.scala` add: + + scalacOptions in Compile ++= Seq("-Xmax-classfile-name", "128"), + +to the `sharedSettings` val. See also [this PR](https://github.com/apache/spark/pull/2883/files) if you are unsure of where to add these lines. + + # Setting up Maven's Memory Usage You'll need to configure Maven to use more memory than usual by setting `MAVEN_OPTS`. We recommend the following settings: From c928796ade54f68e26bc55734a9867a046d2e3fe Mon Sep 17 00:00:00 2001 From: Tathagata Das Date: Tue, 17 Mar 2015 05:31:27 -0700 Subject: [PATCH 078/122] [SPARK-6331] Load new master URL if present when recovering streaming context from checkpoint In streaming driver recovery, when the SparkConf is reconstructed based on the checkpointed configuration, it recovers the old master URL. This okay if the cluster on which the streaming application is relaunched is the same cluster as it was running before. But if that cluster changes, there is no way to inject the new master URL of the new cluster. As a result, the restarted app tries to connect to the non-existent old cluster and fails. The solution is to check whether a master URL is set in the System properties (by Spark submit) before recreating the SparkConf. If a new master url is set in the properties, then use it as that is obviously the most relevant one. Otherwise load the old one (to maintain existing behavior). Author: Tathagata Das Closes #5024 from tdas/SPARK-6331 and squashes the following commits: 392fd44 [Tathagata Das] Fixed naming issue. c7c0b99 [Tathagata Das] Addressed comments. 6a0857c [Tathagata Das] Updated testsuites. 222485d [Tathagata Das] Load new master URL if present when recovering streaming context from checkpoint --- .../apache/spark/streaming/Checkpoint.scala | 7 +++++-- .../spark/streaming/StreamingContext.scala | 2 +- .../spark/streaming/CheckpointSuite.scala | 21 ++++++++++++++++--- .../streaming/StreamingContextSuite.scala | 2 +- 4 files changed, 25 insertions(+), 7 deletions(-) diff --git a/streaming/src/main/scala/org/apache/spark/streaming/Checkpoint.scala b/streaming/src/main/scala/org/apache/spark/streaming/Checkpoint.scala index f88a8a0151550..cb4c94fb9d5a6 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/Checkpoint.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/Checkpoint.scala @@ -43,10 +43,13 @@ class Checkpoint(@transient ssc: StreamingContext, val checkpointTime: Time) val delaySeconds = MetadataCleaner.getDelaySeconds(ssc.conf) val sparkConfPairs = ssc.conf.getAll - def sparkConf = { - new SparkConf(false).setAll(sparkConfPairs) + def createSparkConf(): SparkConf = { + val newSparkConf = new SparkConf(loadDefaults = false).setAll(sparkConfPairs) .remove("spark.driver.host") .remove("spark.driver.port") + val newMasterOption = new SparkConf(loadDefaults = true).getOption("spark.master") + newMasterOption.foreach { newMaster => newSparkConf.setMaster(newMaster) } + newSparkConf } def validate() { diff --git a/streaming/src/main/scala/org/apache/spark/streaming/StreamingContext.scala b/streaming/src/main/scala/org/apache/spark/streaming/StreamingContext.scala index b5b6770a8a150..543224d4b07bc 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/StreamingContext.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/StreamingContext.scala @@ -116,7 +116,7 @@ class StreamingContext private[streaming] ( private[streaming] val sc: SparkContext = { if (isCheckpointPresent) { - new SparkContext(cp_.sparkConf) + new SparkContext(cp_.createSparkConf()) } else { sc_ } diff --git a/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala index 03c448f1df5f1..8ea91eca683cf 100644 --- a/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala +++ b/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala @@ -146,7 +146,7 @@ class CheckpointSuite extends TestSuiteBase { // This tests whether spark conf persists through checkpoints, and certain // configs gets scrubbed - test("persistence of conf through checkpoints") { + test("recovery of conf through checkpoints") { val key = "spark.mykey" val value = "myvalue" System.setProperty(key, value) @@ -154,7 +154,7 @@ class CheckpointSuite extends TestSuiteBase { val originalConf = ssc.conf val cp = new Checkpoint(ssc, Time(1000)) - val cpConf = cp.sparkConf + val cpConf = cp.createSparkConf() assert(cpConf.get("spark.master") === originalConf.get("spark.master")) assert(cpConf.get("spark.app.name") === originalConf.get("spark.app.name")) assert(cpConf.get(key) === value) @@ -163,7 +163,8 @@ class CheckpointSuite extends TestSuiteBase { // Serialize/deserialize to simulate write to storage and reading it back val newCp = Utils.deserialize[Checkpoint](Utils.serialize(cp)) - val newCpConf = newCp.sparkConf + // Verify new SparkConf has all the previous properties + val newCpConf = newCp.createSparkConf() assert(newCpConf.get("spark.master") === originalConf.get("spark.master")) assert(newCpConf.get("spark.app.name") === originalConf.get("spark.app.name")) assert(newCpConf.get(key) === value) @@ -174,6 +175,20 @@ class CheckpointSuite extends TestSuiteBase { ssc = new StreamingContext(null, newCp, null) val restoredConf = ssc.conf assert(restoredConf.get(key) === value) + ssc.stop() + + // Verify new SparkConf picks up new master url if it is set in the properties. See SPARK-6331. + try { + val newMaster = "local[100]" + System.setProperty("spark.master", newMaster) + val newCpConf = newCp.createSparkConf() + assert(newCpConf.get("spark.master") === newMaster) + assert(newCpConf.get("spark.app.name") === originalConf.get("spark.app.name")) + ssc = new StreamingContext(null, newCp, null) + assert(ssc.sparkContext.master === newMaster) + } finally { + System.clearProperty("spark.master") + } } diff --git a/streaming/src/test/scala/org/apache/spark/streaming/StreamingContextSuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/StreamingContextSuite.scala index 6a7cd97aa3222..2e5005ef6ff14 100644 --- a/streaming/src/test/scala/org/apache/spark/streaming/StreamingContextSuite.scala +++ b/streaming/src/test/scala/org/apache/spark/streaming/StreamingContextSuite.scala @@ -100,7 +100,7 @@ class StreamingContextSuite extends FunSuite with BeforeAndAfter with Timeouts w assert(cp.sparkConfPairs.toMap.getOrElse("spark.cleaner.ttl", "-1") === "10") ssc1.stop() val newCp = Utils.deserialize[Checkpoint](Utils.serialize(cp)) - assert(newCp.sparkConf.getInt("spark.cleaner.ttl", -1) === 10) + assert(newCp.createSparkConf().getInt("spark.cleaner.ttl", -1) === 10) ssc = new StreamingContext(null, newCp, null) assert(ssc.conf.getInt("spark.cleaner.ttl", -1) === 10) } From e9f22c61290348c58af54c0ae3c6226105126a8d Mon Sep 17 00:00:00 2001 From: Imran Rashid Date: Tue, 17 Mar 2015 09:41:06 -0500 Subject: [PATCH 079/122] [SPARK-6365] jetty-security also needed for SPARK_PREPEND_CLASSES to work https://issues.apache.org/jira/browse/SPARK-6365 thanks vanzin for helping me figure this out Author: Imran Rashid Closes #5052 from squito/fix_prepend_classes and squashes the following commits: 09d334c [Imran Rashid] jetty-security also needed for SPARK_PREPEND_CLASSES to work --- core/pom.xml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/core/pom.xml b/core/pom.xml index 4164a3a7208d4..81f8cba711df6 100644 --- a/core/pom.xml +++ b/core/pom.xml @@ -414,7 +414,7 @@ true true - guava,jetty-io,jetty-servlet,jetty-continuation,jetty-http,jetty-plus,jetty-util,jetty-server + guava,jetty-io,jetty-servlet,jetty-continuation,jetty-http,jetty-plus,jetty-util,jetty-server,jetty-security true From 0f673c21f68ee3d5df3c01ae405709d3c1f4909b Mon Sep 17 00:00:00 2001 From: Josh Rosen Date: Tue, 17 Mar 2015 09:18:57 -0700 Subject: [PATCH 080/122] [SPARK-3266] Use intermediate abstract classes to fix type erasure issues in Java APIs This PR addresses a Scala compiler bug ([SI-8905](https://issues.scala-lang.org/browse/SI-8905)) that was breaking some of the Spark Java APIs. In a nutshell, it seems that methods whose implementations are inherited from generic traits sometimes have their type parameters erased to Object. This was causing methods like `DoubleRDD.min()` to throw confusing NoSuchMethodErrors at runtime. The fix implemented here is to introduce an intermediate layer of abstract classes and inherit from those instead of directly extends the `Java*Like` traits. This should not break binary compatibility. I also improved the test coverage of the Java API, adding several new tests for methods that failed at runtime due to this bug. Author: Josh Rosen Closes #5050 from JoshRosen/javardd-si-8905-fix and squashes the following commits: 2feb068 [Josh Rosen] Use intermediate abstract classes to work around SPARK-3266 d5f3e5d [Josh Rosen] Add failing regression tests for SPARK-3266 --- .../apache/spark/api/java/JavaDoubleRDD.scala | 3 +- .../apache/spark/api/java/JavaPairRDD.scala | 2 +- .../org/apache/spark/api/java/JavaRDD.scala | 2 +- .../apache/spark/api/java/JavaRDDLike.scala | 8 ++ .../java/org/apache/spark/JavaAPISuite.java | 129 ++++++++++++++++++ .../streaming/api/java/JavaDStream.scala | 2 +- .../streaming/api/java/JavaDStreamLike.scala | 9 ++ .../streaming/api/java/JavaPairDStream.scala | 2 +- 8 files changed, 152 insertions(+), 5 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/api/java/JavaDoubleRDD.scala b/core/src/main/scala/org/apache/spark/api/java/JavaDoubleRDD.scala index 8e8f7f6c4fda2..79e4ebf2db578 100644 --- a/core/src/main/scala/org/apache/spark/api/java/JavaDoubleRDD.scala +++ b/core/src/main/scala/org/apache/spark/api/java/JavaDoubleRDD.scala @@ -32,7 +32,8 @@ import org.apache.spark.storage.StorageLevel import org.apache.spark.util.StatCounter import org.apache.spark.util.Utils -class JavaDoubleRDD(val srdd: RDD[scala.Double]) extends JavaRDDLike[JDouble, JavaDoubleRDD] { +class JavaDoubleRDD(val srdd: RDD[scala.Double]) + extends AbstractJavaRDDLike[JDouble, JavaDoubleRDD] { override val classTag: ClassTag[JDouble] = implicitly[ClassTag[JDouble]] diff --git a/core/src/main/scala/org/apache/spark/api/java/JavaPairRDD.scala b/core/src/main/scala/org/apache/spark/api/java/JavaPairRDD.scala index 7af3538262fd6..4eadc9a85613e 100644 --- a/core/src/main/scala/org/apache/spark/api/java/JavaPairRDD.scala +++ b/core/src/main/scala/org/apache/spark/api/java/JavaPairRDD.scala @@ -44,7 +44,7 @@ import org.apache.spark.util.Utils class JavaPairRDD[K, V](val rdd: RDD[(K, V)]) (implicit val kClassTag: ClassTag[K], implicit val vClassTag: ClassTag[V]) - extends JavaRDDLike[(K, V), JavaPairRDD[K, V]] { + extends AbstractJavaRDDLike[(K, V), JavaPairRDD[K, V]] { override def wrapRDD(rdd: RDD[(K, V)]): JavaPairRDD[K, V] = JavaPairRDD.fromRDD(rdd) diff --git a/core/src/main/scala/org/apache/spark/api/java/JavaRDD.scala b/core/src/main/scala/org/apache/spark/api/java/JavaRDD.scala index 86fb374bef1e3..645dc3bfb6b06 100644 --- a/core/src/main/scala/org/apache/spark/api/java/JavaRDD.scala +++ b/core/src/main/scala/org/apache/spark/api/java/JavaRDD.scala @@ -30,7 +30,7 @@ import org.apache.spark.storage.StorageLevel import org.apache.spark.util.Utils class JavaRDD[T](val rdd: RDD[T])(implicit val classTag: ClassTag[T]) - extends JavaRDDLike[T, JavaRDD[T]] { + extends AbstractJavaRDDLike[T, JavaRDD[T]] { override def wrapRDD(rdd: RDD[T]): JavaRDD[T] = JavaRDD.fromRDD(rdd) diff --git a/core/src/main/scala/org/apache/spark/api/java/JavaRDDLike.scala b/core/src/main/scala/org/apache/spark/api/java/JavaRDDLike.scala index 0f91c942ecd50..8da42934a7d96 100644 --- a/core/src/main/scala/org/apache/spark/api/java/JavaRDDLike.scala +++ b/core/src/main/scala/org/apache/spark/api/java/JavaRDDLike.scala @@ -38,6 +38,14 @@ import org.apache.spark.rdd.RDD import org.apache.spark.storage.StorageLevel import org.apache.spark.util.Utils +/** + * As a workaround for https://issues.scala-lang.org/browse/SI-8905, implementations + * of JavaRDDLike should extend this dummy abstract class instead of directly inheriting + * from the trait. See SPARK-3266 for additional details. + */ +private[spark] abstract class AbstractJavaRDDLike[T, This <: JavaRDDLike[T, This]] + extends JavaRDDLike[T, This] + /** * Defines operations common to several Java RDD implementations. * Note that this trait is not intended to be implemented by user code. diff --git a/core/src/test/java/org/apache/spark/JavaAPISuite.java b/core/src/test/java/org/apache/spark/JavaAPISuite.java index 74e88c767ee07..8ec54360ca42a 100644 --- a/core/src/test/java/org/apache/spark/JavaAPISuite.java +++ b/core/src/test/java/org/apache/spark/JavaAPISuite.java @@ -267,6 +267,22 @@ public void call(String s) throws IOException { Assert.assertEquals(2, accum.value().intValue()); } + @Test + public void foreachPartition() { + final Accumulator accum = sc.accumulator(0); + JavaRDD rdd = sc.parallelize(Arrays.asList("Hello", "World")); + rdd.foreachPartition(new VoidFunction>() { + @Override + public void call(Iterator iter) throws IOException { + while (iter.hasNext()) { + iter.next(); + accum.add(1); + } + } + }); + Assert.assertEquals(2, accum.value().intValue()); + } + @Test public void toLocalIterator() { List correct = Arrays.asList(1, 2, 3, 4); @@ -657,6 +673,13 @@ public Boolean call(Integer i) { }).isEmpty()); } + @Test + public void toArray() { + JavaRDD rdd = sc.parallelize(Arrays.asList(1, 2, 3)); + List list = rdd.toArray(); + Assert.assertEquals(Arrays.asList(1, 2, 3), list); + } + @Test public void cartesian() { JavaDoubleRDD doubleRDD = sc.parallelizeDoubles(Arrays.asList(1.0, 1.0, 2.0, 3.0, 5.0, 8.0)); @@ -714,6 +737,80 @@ public void javaDoubleRDDHistoGram() { sc.parallelizeDoubles(new ArrayList(0), 1).histogram(new double[]{0.0, 1.0})); } + private static class DoubleComparator implements Comparator, Serializable { + public int compare(Double o1, Double o2) { + return o1.compareTo(o2); + } + } + + @Test + public void max() { + JavaDoubleRDD rdd = sc.parallelizeDoubles(Arrays.asList(1.0, 2.0, 3.0, 4.0)); + double max = rdd.max(new DoubleComparator()); + Assert.assertEquals(4.0, max, 0.001); + } + + @Test + public void min() { + JavaDoubleRDD rdd = sc.parallelizeDoubles(Arrays.asList(1.0, 2.0, 3.0, 4.0)); + double max = rdd.min(new DoubleComparator()); + Assert.assertEquals(1.0, max, 0.001); + } + + @Test + public void takeOrdered() { + JavaDoubleRDD rdd = sc.parallelizeDoubles(Arrays.asList(1.0, 2.0, 3.0, 4.0)); + Assert.assertEquals(Arrays.asList(1.0, 2.0), rdd.takeOrdered(2, new DoubleComparator())); + Assert.assertEquals(Arrays.asList(1.0, 2.0), rdd.takeOrdered(2)); + } + + @Test + public void top() { + JavaRDD rdd = sc.parallelize(Arrays.asList(1, 2, 3, 4)); + List top2 = rdd.top(2); + Assert.assertEquals(Arrays.asList(4, 3), top2); + } + + private static class AddInts implements Function2 { + @Override + public Integer call(Integer a, Integer b) { + return a + b; + } + } + + @Test + public void reduce() { + JavaRDD rdd = sc.parallelize(Arrays.asList(1, 2, 3, 4)); + int sum = rdd.reduce(new AddInts()); + Assert.assertEquals(10, sum); + } + + @Test + public void reduceOnJavaDoubleRDD() { + JavaDoubleRDD rdd = sc.parallelizeDoubles(Arrays.asList(1.0, 2.0, 3.0, 4.0)); + double sum = rdd.reduce(new Function2() { + @Override + public Double call(Double v1, Double v2) throws Exception { + return v1 + v2; + } + }); + Assert.assertEquals(10.0, sum, 0.001); + } + + @Test + public void fold() { + JavaRDD rdd = sc.parallelize(Arrays.asList(1, 2, 3, 4)); + int sum = rdd.fold(0, new AddInts()); + Assert.assertEquals(10, sum); + } + + @Test + public void aggregate() { + JavaRDD rdd = sc.parallelize(Arrays.asList(1, 2, 3, 4)); + int sum = rdd.aggregate(0, new AddInts(), new AddInts()); + Assert.assertEquals(10, sum); + } + @Test public void map() { JavaRDD rdd = sc.parallelize(Arrays.asList(1, 2, 3, 4, 5)); @@ -830,6 +927,25 @@ public Iterable call(Iterator iter) { Assert.assertEquals("[3, 7]", partitionSums.collect().toString()); } + + @Test + public void mapPartitionsWithIndex() { + JavaRDD rdd = sc.parallelize(Arrays.asList(1, 2, 3, 4), 2); + JavaRDD partitionSums = rdd.mapPartitionsWithIndex( + new Function2, Iterator>() { + @Override + public Iterator call(Integer index, Iterator iter) throws Exception { + int sum = 0; + while (iter.hasNext()) { + sum += iter.next(); + } + return Collections.singletonList(sum).iterator(); + } + }, false); + Assert.assertEquals("[3, 7]", partitionSums.collect().toString()); + } + + @Test public void repartition() { // Shrinking number of partitions @@ -1516,6 +1632,19 @@ public void collectAsync() throws Exception { Assert.assertEquals(1, future.jobIds().size()); } + @Test + public void takeAsync() throws Exception { + List data = Arrays.asList(1, 2, 3, 4, 5); + JavaRDD rdd = sc.parallelize(data, 1); + JavaFutureAction> future = rdd.takeAsync(1); + List result = future.get(); + Assert.assertEquals(1, result.size()); + Assert.assertEquals((Integer) 1, result.get(0)); + Assert.assertFalse(future.isCancelled()); + Assert.assertTrue(future.isDone()); + Assert.assertEquals(1, future.jobIds().size()); + } + @Test public void foreachAsync() throws Exception { List data = Arrays.asList(1, 2, 3, 4, 5); diff --git a/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStream.scala b/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStream.scala index 505e4431e4350..01cdcb0574040 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStream.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStream.scala @@ -36,7 +36,7 @@ import org.apache.spark.streaming.dstream.DStream * [[org.apache.spark.streaming.api.java.JavaPairDStream]]. */ class JavaDStream[T](val dstream: DStream[T])(implicit val classTag: ClassTag[T]) - extends JavaDStreamLike[T, JavaDStream[T], JavaRDD[T]] { + extends AbstractJavaDStreamLike[T, JavaDStream[T], JavaRDD[T]] { override def wrapRDD(rdd: RDD[T]): JavaRDD[T] = JavaRDD.fromRDD(rdd) diff --git a/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStreamLike.scala b/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStreamLike.scala index c382a12f4d099..2eabdd9387913 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStreamLike.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStreamLike.scala @@ -34,6 +34,15 @@ import org.apache.spark.streaming._ import org.apache.spark.streaming.api.java.JavaDStream._ import org.apache.spark.streaming.dstream.DStream +/** + * As a workaround for https://issues.scala-lang.org/browse/SI-8905, implementations + * of JavaDStreamLike should extend this dummy abstract class instead of directly inheriting + * from the trait. See SPARK-3266 for additional details. + */ +private[streaming] +abstract class AbstractJavaDStreamLike[T, This <: JavaDStreamLike[T, This, R], + R <: JavaRDDLike[T, R]] extends JavaDStreamLike[T, This, R] + trait JavaDStreamLike[T, This <: JavaDStreamLike[T, This, R], R <: JavaRDDLike[T, R]] extends Serializable { implicit val classTag: ClassTag[T] diff --git a/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaPairDStream.scala b/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaPairDStream.scala index bd01789b611a4..f94f2d0e8bd31 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaPairDStream.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaPairDStream.scala @@ -45,7 +45,7 @@ import org.apache.spark.streaming.dstream.DStream class JavaPairDStream[K, V](val dstream: DStream[(K, V)])( implicit val kManifest: ClassTag[K], implicit val vManifest: ClassTag[V]) - extends JavaDStreamLike[(K, V), JavaPairDStream[K, V], JavaPairRDD[K, V]] { + extends AbstractJavaDStreamLike[(K, V), JavaPairDStream[K, V], JavaPairRDD[K, V]] { override def wrapRDD(rdd: RDD[(K, V)]): JavaPairRDD[K, V] = JavaPairRDD.fromRDD(rdd) From 4cca3917dc30ee907e6cbd6a569b6ac58af963f7 Mon Sep 17 00:00:00 2001 From: nemccarthy Date: Tue, 17 Mar 2015 09:33:11 -0700 Subject: [PATCH 081/122] [SPARK-6313] Add config option to disable file locks/fetchFile cache to ... ...support NFS mounts. This is a work around for now with the goal to find a more permanent solution. https://issues.apache.org/jira/browse/SPARK-6313 Author: nemccarthy Closes #5036 from nemccarthy/master and squashes the following commits: 2eaaf42 [nemccarthy] [SPARK-6313] Update config wording doc for spark.files.useFetchCache 5de7eb4 [nemccarthy] [SPARK-6313] Add config option to disable file locks/fetchFile cache to support NFS mounts --- .../src/main/scala/org/apache/spark/util/Utils.scala | 3 ++- docs/configuration.md | 12 ++++++++++++ 2 files changed, 14 insertions(+), 1 deletion(-) diff --git a/core/src/main/scala/org/apache/spark/util/Utils.scala b/core/src/main/scala/org/apache/spark/util/Utils.scala index d3dc1d09cb7b4..af8a24553a461 100644 --- a/core/src/main/scala/org/apache/spark/util/Utils.scala +++ b/core/src/main/scala/org/apache/spark/util/Utils.scala @@ -403,7 +403,8 @@ private[spark] object Utils extends Logging { useCache: Boolean) { val fileName = url.split("/").last val targetFile = new File(targetDir, fileName) - if (useCache) { + val fetchCacheEnabled = conf.getBoolean("spark.files.useFetchCache", defaultValue = true) + if (useCache && fetchCacheEnabled) { val cachedFileName = s"${url.hashCode}${timestamp}_cache" val lockFileName = s"${url.hashCode}${timestamp}_lock" val localDir = new File(getLocalDir(conf)) diff --git a/docs/configuration.md b/docs/configuration.md index 63fc99e7d3e29..7fe11475212b3 100644 --- a/docs/configuration.md +++ b/docs/configuration.md @@ -745,6 +745,18 @@ Apart from these, the following properties are also available, and may be useful the driver, in seconds. +
     + + + + From d9f3e01688ad0a8d5fc2419a948a682ad7d957c9 Mon Sep 17 00:00:00 2001 From: lewuathe Date: Tue, 17 Mar 2015 12:11:57 -0700 Subject: [PATCH 082/122] [SPARK-6336] LBFGS should document what convergenceTol means LBFGS uses convergence tolerance. This value should be written in document as an argument. Author: lewuathe Closes #5033 from Lewuathe/SPARK-6336 and squashes the following commits: e738b33 [lewuathe] Modify text to be more natural ac03c3a [lewuathe] Modify documentations 6ccb304 [lewuathe] [SPARK-6336] LBFGS should document what convergenceTol means --- docs/mllib-optimization.md | 4 ++++ .../scala/org/apache/spark/mllib/optimization/LBFGS.scala | 6 +++++- 2 files changed, 9 insertions(+), 1 deletion(-) diff --git a/docs/mllib-optimization.md b/docs/mllib-optimization.md index 4d101afca2c97..6cabc1610a151 100644 --- a/docs/mllib-optimization.md +++ b/docs/mllib-optimization.md @@ -203,6 +203,10 @@ regularization, as well as L2 regularizer. recommended. * `maxNumIterations` is the maximal number of iterations that L-BFGS can be run. * `regParam` is the regularization parameter when using regularization. +* `convergenceTol` controls how much relative change is still allowed when L-BFGS +is considered to converge. This must be nonnegative. Lower values are less tolerant and +therefore generally cause more iterations to be run. This value looks at both average +improvement and the norm of gradient inside [Breeze LBFGS](https://github.com/scalanlp/breeze/blob/master/math/src/main/scala/breeze/optimize/LBFGS.scala). The `return` is a tuple containing two elements. The first element is a column matrix containing weights for every feature, and the second element is an array containing diff --git a/mllib/src/main/scala/org/apache/spark/mllib/optimization/LBFGS.scala b/mllib/src/main/scala/org/apache/spark/mllib/optimization/LBFGS.scala index d5e4f4ccbff10..ef6eccd90711a 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/optimization/LBFGS.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/optimization/LBFGS.scala @@ -60,6 +60,8 @@ class LBFGS(private var gradient: Gradient, private var updater: Updater) /** * Set the convergence tolerance of iterations for L-BFGS. Default 1E-4. * Smaller value will lead to higher accuracy with the cost of more iterations. + * This value must be nonnegative. Lower convergence values are less tolerant + * and therefore generally cause more iterations to be run. */ def setConvergenceTol(tolerance: Double): this.type = { this.convergenceTol = tolerance @@ -142,7 +144,9 @@ object LBFGS extends Logging { * one single data example) * @param updater - Updater function to actually perform a gradient step in a given direction. * @param numCorrections - The number of corrections used in the L-BFGS update. - * @param convergenceTol - The convergence tolerance of iterations for L-BFGS + * @param convergenceTol - The convergence tolerance of iterations for L-BFGS which is must be + * nonnegative. Lower values are less tolerant and therefore generally + * cause more iterations to be run. * @param maxNumIterations - Maximal number of iterations that L-BFGS can be run. * @param regParam - Regularization parameter * From c94d0626471e209ab7ebfc588f9a2992946b7ed5 Mon Sep 17 00:00:00 2001 From: Xiangrui Meng Date: Tue, 17 Mar 2015 12:14:40 -0700 Subject: [PATCH 083/122] [SPARK-6226][MLLIB] add save/load in PySpark's KMeansModel Use `_py2java` and `_java2py` to convert Python model to/from Java model. yinxusen Author: Xiangrui Meng Closes #5049 from mengxr/SPARK-6226-mengxr and squashes the following commits: 570ba81 [Xiangrui Meng] fix python style b10b911 [Xiangrui Meng] add save/load in PySpark's KMeansModel --- .../spark/mllib/clustering/KMeansModel.scala | 5 ++++ python/pyspark/mllib/clustering.py | 28 +++++++++++++++++-- python/pyspark/mllib/common.py | 4 +-- 3 files changed, 32 insertions(+), 5 deletions(-) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeansModel.scala b/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeansModel.scala index 707da537d238f..e4e411a3c8b42 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeansModel.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeansModel.scala @@ -17,6 +17,8 @@ package org.apache.spark.mllib.clustering +import scala.collection.JavaConverters._ + import org.json4s._ import org.json4s.JsonDSL._ import org.json4s.jackson.JsonMethods._ @@ -34,6 +36,9 @@ import org.apache.spark.sql.Row */ class KMeansModel (val clusterCenters: Array[Vector]) extends Saveable with Serializable { + /** A Java-friendly constructor that takes an Iterable of Vectors. */ + def this(centers: java.lang.Iterable[Vector]) = this(centers.asScala.toArray) + /** Total number of clusters. */ def k: Int = clusterCenters.length diff --git a/python/pyspark/mllib/clustering.py b/python/pyspark/mllib/clustering.py index 949db5705abd7..464f49aeee3cd 100644 --- a/python/pyspark/mllib/clustering.py +++ b/python/pyspark/mllib/clustering.py @@ -19,14 +19,16 @@ from pyspark import RDD from pyspark import SparkContext -from pyspark.mllib.common import callMLlibFunc, callJavaFunc -from pyspark.mllib.linalg import DenseVector, SparseVector, _convert_to_vector +from pyspark.mllib.common import callMLlibFunc, callJavaFunc, _py2java, _java2py +from pyspark.mllib.linalg import SparseVector, _convert_to_vector from pyspark.mllib.stat.distribution import MultivariateGaussian +from pyspark.mllib.util import Saveable, Loader, inherit_doc __all__ = ['KMeansModel', 'KMeans', 'GaussianMixtureModel', 'GaussianMixture'] -class KMeansModel(object): +@inherit_doc +class KMeansModel(Saveable, Loader): """A clustering model derived from the k-means method. @@ -55,6 +57,16 @@ class KMeansModel(object): True >>> type(model.clusterCenters) + >>> import os, tempfile + >>> path = tempfile.mkdtemp() + >>> model.save(sc, path) + >>> sameModel = KMeansModel.load(sc, path) + >>> sameModel.predict(sparse_data[0]) == model.predict(sparse_data[0]) + True + >>> try: + ... os.removedirs(path) + ... except OSError: + ... pass """ def __init__(self, centers): @@ -77,6 +89,16 @@ def predict(self, x): best_distance = distance return best + def save(self, sc, path): + java_centers = _py2java(sc, map(_convert_to_vector, self.centers)) + java_model = sc._jvm.org.apache.spark.mllib.clustering.KMeansModel(java_centers) + java_model.save(sc._jsc.sc(), path) + + @classmethod + def load(cls, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.clustering.KMeansModel.load(sc._jsc.sc(), path) + return KMeansModel(_java2py(sc, java_model.clusterCenters())) + class KMeans(object): diff --git a/python/pyspark/mllib/common.py b/python/pyspark/mllib/common.py index 621591c26b77f..a539d2f2846f9 100644 --- a/python/pyspark/mllib/common.py +++ b/python/pyspark/mllib/common.py @@ -70,8 +70,8 @@ def _py2java(sc, obj): obj = _to_java_object_rdd(obj) elif isinstance(obj, SparkContext): obj = obj._jsc - elif isinstance(obj, list) and (obj or isinstance(obj[0], JavaObject)): - obj = ListConverter().convert(obj, sc._gateway._gateway_client) + elif isinstance(obj, list): + obj = ListConverter().convert([_py2java(sc, x) for x in obj], sc._gateway._gateway_client) elif isinstance(obj, JavaObject): pass elif isinstance(obj, (int, long, float, bool, basestring)): From 4633a87b86a6ef01fa724d31763dcb97cb5bc746 Mon Sep 17 00:00:00 2001 From: Pei-Lun Lee Date: Wed, 18 Mar 2015 08:34:46 +0800 Subject: [PATCH 084/122] [SPARK-6330] [SQL] Add a test case for SPARK-6330 When getting file statuses, create file system from each path instead of a single one from hadoop configuration. Author: Pei-Lun Lee Closes #5039 from ypcat/spark-6351 and squashes the following commits: a19a3fe [Pei-Lun Lee] [SPARK-6330] [SQL] fix test 506f5a0 [Pei-Lun Lee] [SPARK-6351] [SQL] fix test fa2290e [Pei-Lun Lee] [SPARK-6330] [SQL] Rename test case and add comment 606c967 [Pei-Lun Lee] Merge branch 'master' of https://github.com/apache/spark into spark-6351 896e80a [Pei-Lun Lee] [SPARK-6351] [SQL] Add test case 2ae0916 [Pei-Lun Lee] [SPARK-6351] [SQL] ParquetRelation2 supporting multiple file systems --- .../apache/spark/sql/parquet/ParquetIOSuite.scala | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/sql/core/src/test/scala/org/apache/spark/sql/parquet/ParquetIOSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/parquet/ParquetIOSuite.scala index 36f3406a7825f..a70b3c7ce48d3 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/parquet/ParquetIOSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/parquet/ParquetIOSuite.scala @@ -329,6 +329,7 @@ class ParquetIOSuiteBase extends QueryTest with ParquetTest { checkAnswer(parquetFile(file), (data ++ newData).map(Row.fromTuple)) } } + } class ParquetDataSourceOnIOSuite extends ParquetIOSuiteBase with BeforeAndAfterAll { @@ -341,6 +342,18 @@ class ParquetDataSourceOnIOSuite extends ParquetIOSuiteBase with BeforeAndAfterA override protected def afterAll(): Unit = { sqlContext.setConf(SQLConf.PARQUET_USE_DATA_SOURCE_API, originalConf.toString) } + + test("SPARK-6330 regression test") { + // In 1.3.0, save to fs other than file: without configuring core-site.xml would get: + // IllegalArgumentException: Wrong FS: hdfs://..., expected: file:/// + intercept[java.io.FileNotFoundException] { + sqlContext.parquetFile("file:///nonexistent") + } + val errorMessage = intercept[Throwable] { + sqlContext.parquetFile("hdfs://nonexistent") + }.toString + assert(errorMessage.contains("UnknownHostException")) + } } class ParquetDataSourceOffIOSuite extends ParquetIOSuiteBase with BeforeAndAfterAll { From dc9c9196d63aa465e86ac52f0e86e10c12472100 Mon Sep 17 00:00:00 2001 From: Yin Huai Date: Wed, 18 Mar 2015 09:41:06 +0800 Subject: [PATCH 085/122] [SPARK-6366][SQL] In Python API, the default save mode for save and saveAsTable should be "error" instead of "append". https://issues.apache.org/jira/browse/SPARK-6366 Author: Yin Huai Closes #5053 from yhuai/SPARK-6366 and squashes the following commits: fc81897 [Yin Huai] Use error as the default save mode for save/saveAsTable. --- python/pyspark/sql/dataframe.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/python/pyspark/sql/dataframe.py b/python/pyspark/sql/dataframe.py index 94001aec3774b..5cb89da7a8ed5 100644 --- a/python/pyspark/sql/dataframe.py +++ b/python/pyspark/sql/dataframe.py @@ -162,7 +162,7 @@ def _java_save_mode(self, mode): "Only 'append', 'overwrite', 'ignore', and 'error' are acceptable save mode.") return jmode - def saveAsTable(self, tableName, source=None, mode="append", **options): + def saveAsTable(self, tableName, source=None, mode="error", **options): """Saves the contents of the :class:`DataFrame` to a data source as a table. The data source is specified by the `source` and a set of `options`. @@ -188,7 +188,7 @@ def saveAsTable(self, tableName, source=None, mode="append", **options): self.sql_ctx._sc._gateway._gateway_client) self._jdf.saveAsTable(tableName, source, jmode, joptions) - def save(self, path=None, source=None, mode="append", **options): + def save(self, path=None, source=None, mode="error", **options): """Saves the contents of the :class:`DataFrame` to a data source. The data source is specified by the `source` and a set of `options`. From a012e08635dc2d643715e11680fd6a3fb3afe44a Mon Sep 17 00:00:00 2001 From: Tijo Thomas Date: Tue, 17 Mar 2015 18:50:19 -0700 Subject: [PATCH 086/122] [SPARK-6383][SQL]Fixed compiler and errors in Dataframe examples Author: Tijo Thomas Closes #5068 from tijoparacka/fix_sql_dataframe_example and squashes the following commits: 6953ac1 [Tijo Thomas] Handled Java and Python example sections 0751a74 [Tijo Thomas] Fixed compiler and errors in Dataframe examples --- docs/sql-programming-guide.md | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/docs/sql-programming-guide.md b/docs/sql-programming-guide.md index 11c29e20632ae..2cbb4c967eb81 100644 --- a/docs/sql-programming-guide.md +++ b/docs/sql-programming-guide.md @@ -170,14 +170,14 @@ df.select("name").show() // Justin // Select everybody, but increment the age by 1 -df.select("name", df("age") + 1).show() +df.select(df("name"), df("age") + 1).show() // name (age + 1) // Michael null // Andy 31 // Justin 20 // Select people older than 21 -df.filter(df("name") > 21).show() +df.filter(df("age") > 21).show() // age name // 30 Andy @@ -220,14 +220,14 @@ df.select("name").show(); // Justin // Select everybody, but increment the age by 1 -df.select("name", df.col("age").plus(1)).show(); +df.select(df.col("name"), df.col("age").plus(1)).show(); // name (age + 1) // Michael null // Andy 31 // Justin 20 // Select people older than 21 -df.filter(df("name") > 21).show(); +df.filter(df.col("age").gt(21)).show(); // age name // 30 Andy @@ -270,14 +270,14 @@ df.select("name").show() ## Justin # Select everybody, but increment the age by 1 -df.select("name", df.age + 1).show() +df.select(df.name, df.age + 1).show() ## name (age + 1) ## Michael null ## Andy 31 ## Justin 20 # Select people older than 21 -df.filter(df.name > 21).show() +df.filter(df.age > 21).show() ## age name ## 30 Andy From 5c80643d137008ce8a0ac7467b31d8d52383c105 Mon Sep 17 00:00:00 2001 From: Liang-Chi Hsieh Date: Tue, 17 Mar 2015 18:58:52 -0700 Subject: [PATCH 087/122] [SPARK-5908][SQL] Resolve UdtfsAlias when only single Alias is used `ResolveUdtfsAlias` in `hiveUdfs` only considers the `HiveGenericUdtf` with multiple alias. When only single alias is used with `HiveGenericUdtf`, the alias is not working. Author: Liang-Chi Hsieh Closes #4692 from viirya/udft_alias and squashes the following commits: 8a3bae4 [Liang-Chi Hsieh] No need to test selected column from DataFrame since DataFrame API is updated. 160a379 [Liang-Chi Hsieh] Merge remote-tracking branch 'upstream/master' into udft_alias e6531cc [Liang-Chi Hsieh] Selected column from DataFrame should not re-analyze logical plan. a45cc2a [Liang-Chi Hsieh] Resolve UdtfsAlias when only single Alias is used. --- .../main/scala/org/apache/spark/sql/hive/hiveUdfs.scala | 2 ++ .../apache/spark/sql/hive/execution/SQLQuerySuite.scala | 7 +++++++ 2 files changed, 9 insertions(+) diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveUdfs.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveUdfs.scala index 34c21c11761ae..4a702d96563d5 100644 --- a/sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveUdfs.scala +++ b/sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveUdfs.scala @@ -333,6 +333,8 @@ private[spark] object ResolveUdtfsAlias extends Rule[LogicalPlan] { if projectList.exists(_.isInstanceOf[MultiAlias]) && projectList.size != 1 => throw new TreeNodeException(p, "only single Generator supported for SELECT clause") + case Project(Seq(Alias(udtf @ HiveGenericUdtf(_, _, _), name)), child) => + Generate(udtf.copy(aliasNames = Seq(name)), join = false, outer = false, None, child) case Project(Seq(MultiAlias(udtf @ HiveGenericUdtf(_, _, _), names)), child) => Generate(udtf.copy(aliasNames = names), join = false, outer = false, None, child) } diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/SQLQuerySuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/SQLQuerySuite.scala index 22ea19bd82f86..1187228f4c3db 100644 --- a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/SQLQuerySuite.scala +++ b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/SQLQuerySuite.scala @@ -397,6 +397,13 @@ class SQLQuerySuite extends QueryTest { dropTempTable("data") } + test("resolve udtf with single alias") { + val rdd = sparkContext.makeRDD((1 to 5).map(i => s"""{"a":[$i, ${i+1}]}""")) + jsonRDD(rdd).registerTempTable("data") + val df = sql("SELECT explode(a) AS val FROM data") + val col = df("val") + } + test("logical.Project should not be resolved if it contains aggregates or generators") { // This test is used to test the fix of SPARK-5875. // The original issue was that Project's resolved will be true when it contains From 78cb08a5db7b3e1b61ffb28bc95d0b23e8db5c40 Mon Sep 17 00:00:00 2001 From: Cheng Hao Date: Tue, 17 Mar 2015 19:32:38 -0700 Subject: [PATCH 088/122] [SPARK-5404] [SQL] Update the default statistic number By default, the statistic for logical plan with multiple children is quite aggressive, and those statistic are quite critical for the join optimization, hence we need to estimate the statistics as accurate as possible. For `Union`, which has 2 children, and overwrite the default implementation by `adding` its children `byteInSize` instead of `multiplying`. For `Expand`, which only has a single child, but it will grows the size, and we need to multiply its inflating factor. Author: Cheng Hao Closes #4914 from chenghao-intel/statistic and squashes the following commits: d466bbc [Cheng Hao] Update the default statistic --- .../sql/catalyst/plans/logical/basicOperators.scala | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/basicOperators.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/basicOperators.scala index 20cc8e90a71a3..624912dab4652 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/basicOperators.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/basicOperators.scala @@ -81,6 +81,11 @@ case class Union(left: LogicalPlan, right: LogicalPlan) extends BinaryNode { override lazy val resolved = childrenResolved && !left.output.zip(right.output).exists { case (l,r) => l.dataType != r.dataType } + + override def statistics: Statistics = { + val sizeInBytes = left.statistics.sizeInBytes + right.statistics.sizeInBytes + Statistics(sizeInBytes = sizeInBytes) + } } case class Join( @@ -174,7 +179,12 @@ case class Aggregate( case class Expand( projections: Seq[GroupExpression], output: Seq[Attribute], - child: LogicalPlan) extends UnaryNode + child: LogicalPlan) extends UnaryNode { + override def statistics: Statistics = { + val sizeInBytes = child.statistics.sizeInBytes * projections.length + Statistics(sizeInBytes = sizeInBytes) + } +} trait GroupingAnalytics extends UnaryNode { self: Product => From a6ee2f7940b9a64a81667615586ae597da837974 Mon Sep 17 00:00:00 2001 From: watermen Date: Tue, 17 Mar 2015 19:35:18 -0700 Subject: [PATCH 089/122] [SPARK-5651][SQL] Add input64 in blacklist and add test suit for create table within backticks Now spark version is only support ```create table table_in_database_creation.test1 as select * from src limit 1;``` in HiveContext. This patch is used to support ```create table `table_in_database_creation.test2` as select * from src limit 1;``` in HiveContext. Author: watermen Author: q00251598 Closes #4427 from watermen/SPARK-5651 and squashes the following commits: c5c8ed1 [watermen] add the generated golden files 1f0e42e [q00251598] add input64 in blacklist and add test suit --- .../execution/HiveCompatibilitySuite.scala | 6 +- ...ckticks-0-a253b1ed35dbf503d1b8902dacbe23ac | 0 ...ckticks-1-61dc640dfeaff471f3d2b730f9cbf959 | 0 ...ckticks-2-ce780d068b8d24786e639e361101a0c7 | 500 ++++++++++++++++++ ...ckticks-3-afd6e46b6a289c3c24a8eec75a94043c | 0 .../sql/hive/execution/HiveQuerySuite.scala | 8 + 6 files changed, 513 insertions(+), 1 deletion(-) create mode 100644 sql/hive/src/test/resources/golden/create table as with db name within backticks-0-a253b1ed35dbf503d1b8902dacbe23ac create mode 100644 sql/hive/src/test/resources/golden/create table as with db name within backticks-1-61dc640dfeaff471f3d2b730f9cbf959 create mode 100644 sql/hive/src/test/resources/golden/create table as with db name within backticks-2-ce780d068b8d24786e639e361101a0c7 create mode 100644 sql/hive/src/test/resources/golden/create table as with db name within backticks-3-afd6e46b6a289c3c24a8eec75a94043c diff --git a/sql/hive/compatibility/src/test/scala/org/apache/spark/sql/hive/execution/HiveCompatibilitySuite.scala b/sql/hive/compatibility/src/test/scala/org/apache/spark/sql/hive/execution/HiveCompatibilitySuite.scala index 68cb34d698ef3..5180a7f09d80f 100644 --- a/sql/hive/compatibility/src/test/scala/org/apache/spark/sql/hive/execution/HiveCompatibilitySuite.scala +++ b/sql/hive/compatibility/src/test/scala/org/apache/spark/sql/hive/execution/HiveCompatibilitySuite.scala @@ -236,7 +236,11 @@ class HiveCompatibilitySuite extends HiveQueryFileTest with BeforeAndAfter { // timestamp in array, the output format of Hive contains double quotes, while // Spark SQL doesn't - "udf_sort_array" + "udf_sort_array", + + // It has a bug and it has been fixed by + // https://issues.apache.org/jira/browse/HIVE-7673 (in Hive 0.14 and trunk). + "input46" ) ++ HiveShim.compatibilityBlackList /** diff --git a/sql/hive/src/test/resources/golden/create table as with db name within backticks-0-a253b1ed35dbf503d1b8902dacbe23ac b/sql/hive/src/test/resources/golden/create table as with db name within backticks-0-a253b1ed35dbf503d1b8902dacbe23ac new file mode 100644 index 0000000000000..e69de29bb2d1d diff --git a/sql/hive/src/test/resources/golden/create table as with db name within backticks-1-61dc640dfeaff471f3d2b730f9cbf959 b/sql/hive/src/test/resources/golden/create table as with db name within backticks-1-61dc640dfeaff471f3d2b730f9cbf959 new file mode 100644 index 0000000000000..e69de29bb2d1d diff --git a/sql/hive/src/test/resources/golden/create table as with db name within backticks-2-ce780d068b8d24786e639e361101a0c7 b/sql/hive/src/test/resources/golden/create table as with db name within backticks-2-ce780d068b8d24786e639e361101a0c7 new file mode 100644 index 0000000000000..7aae61e5eb82f --- /dev/null +++ b/sql/hive/src/test/resources/golden/create table as with db name within backticks-2-ce780d068b8d24786e639e361101a0c7 @@ -0,0 +1,500 @@ +238 val_238 +86 val_86 +311 val_311 +27 val_27 +165 val_165 +409 val_409 +255 val_255 +278 val_278 +98 val_98 +484 val_484 +265 val_265 +193 val_193 +401 val_401 +150 val_150 +273 val_273 +224 val_224 +369 val_369 +66 val_66 +128 val_128 +213 val_213 +146 val_146 +406 val_406 +429 val_429 +374 val_374 +152 val_152 +469 val_469 +145 val_145 +495 val_495 +37 val_37 +327 val_327 +281 val_281 +277 val_277 +209 val_209 +15 val_15 +82 val_82 +403 val_403 +166 val_166 +417 val_417 +430 val_430 +252 val_252 +292 val_292 +219 val_219 +287 val_287 +153 val_153 +193 val_193 +338 val_338 +446 val_446 +459 val_459 +394 val_394 +237 val_237 +482 val_482 +174 val_174 +413 val_413 +494 val_494 +207 val_207 +199 val_199 +466 val_466 +208 val_208 +174 val_174 +399 val_399 +396 val_396 +247 val_247 +417 val_417 +489 val_489 +162 val_162 +377 val_377 +397 val_397 +309 val_309 +365 val_365 +266 val_266 +439 val_439 +342 val_342 +367 val_367 +325 val_325 +167 val_167 +195 val_195 +475 val_475 +17 val_17 +113 val_113 +155 val_155 +203 val_203 +339 val_339 +0 val_0 +455 val_455 +128 val_128 +311 val_311 +316 val_316 +57 val_57 +302 val_302 +205 val_205 +149 val_149 +438 val_438 +345 val_345 +129 val_129 +170 val_170 +20 val_20 +489 val_489 +157 val_157 +378 val_378 +221 val_221 +92 val_92 +111 val_111 +47 val_47 +72 val_72 +4 val_4 +280 val_280 +35 val_35 +427 val_427 +277 val_277 +208 val_208 +356 val_356 +399 val_399 +169 val_169 +382 val_382 +498 val_498 +125 val_125 +386 val_386 +437 val_437 +469 val_469 +192 val_192 +286 val_286 +187 val_187 +176 val_176 +54 val_54 +459 val_459 +51 val_51 +138 val_138 +103 val_103 +239 val_239 +213 val_213 +216 val_216 +430 val_430 +278 val_278 +176 val_176 +289 val_289 +221 val_221 +65 val_65 +318 val_318 +332 val_332 +311 val_311 +275 val_275 +137 val_137 +241 val_241 +83 val_83 +333 val_333 +180 val_180 +284 val_284 +12 val_12 +230 val_230 +181 val_181 +67 val_67 +260 val_260 +404 val_404 +384 val_384 +489 val_489 +353 val_353 +373 val_373 +272 val_272 +138 val_138 +217 val_217 +84 val_84 +348 val_348 +466 val_466 +58 val_58 +8 val_8 +411 val_411 +230 val_230 +208 val_208 +348 val_348 +24 val_24 +463 val_463 +431 val_431 +179 val_179 +172 val_172 +42 val_42 +129 val_129 +158 val_158 +119 val_119 +496 val_496 +0 val_0 +322 val_322 +197 val_197 +468 val_468 +393 val_393 +454 val_454 +100 val_100 +298 val_298 +199 val_199 +191 val_191 +418 val_418 +96 val_96 +26 val_26 +165 val_165 +327 val_327 +230 val_230 +205 val_205 +120 val_120 +131 val_131 +51 val_51 +404 val_404 +43 val_43 +436 val_436 +156 val_156 +469 val_469 +468 val_468 +308 val_308 +95 val_95 +196 val_196 +288 val_288 +481 val_481 +457 val_457 +98 val_98 +282 val_282 +197 val_197 +187 val_187 +318 val_318 +318 val_318 +409 val_409 +470 val_470 +137 val_137 +369 val_369 +316 val_316 +169 val_169 +413 val_413 +85 val_85 +77 val_77 +0 val_0 +490 val_490 +87 val_87 +364 val_364 +179 val_179 +118 val_118 +134 val_134 +395 val_395 +282 val_282 +138 val_138 +238 val_238 +419 val_419 +15 val_15 +118 val_118 +72 val_72 +90 val_90 +307 val_307 +19 val_19 +435 val_435 +10 val_10 +277 val_277 +273 val_273 +306 val_306 +224 val_224 +309 val_309 +389 val_389 +327 val_327 +242 val_242 +369 val_369 +392 val_392 +272 val_272 +331 val_331 +401 val_401 +242 val_242 +452 val_452 +177 val_177 +226 val_226 +5 val_5 +497 val_497 +402 val_402 +396 val_396 +317 val_317 +395 val_395 +58 val_58 +35 val_35 +336 val_336 +95 val_95 +11 val_11 +168 val_168 +34 val_34 +229 val_229 +233 val_233 +143 val_143 +472 val_472 +322 val_322 +498 val_498 +160 val_160 +195 val_195 +42 val_42 +321 val_321 +430 val_430 +119 val_119 +489 val_489 +458 val_458 +78 val_78 +76 val_76 +41 val_41 +223 val_223 +492 val_492 +149 val_149 +449 val_449 +218 val_218 +228 val_228 +138 val_138 +453 val_453 +30 val_30 +209 val_209 +64 val_64 +468 val_468 +76 val_76 +74 val_74 +342 val_342 +69 val_69 +230 val_230 +33 val_33 +368 val_368 +103 val_103 +296 val_296 +113 val_113 +216 val_216 +367 val_367 +344 val_344 +167 val_167 +274 val_274 +219 val_219 +239 val_239 +485 val_485 +116 val_116 +223 val_223 +256 val_256 +263 val_263 +70 val_70 +487 val_487 +480 val_480 +401 val_401 +288 val_288 +191 val_191 +5 val_5 +244 val_244 +438 val_438 +128 val_128 +467 val_467 +432 val_432 +202 val_202 +316 val_316 +229 val_229 +469 val_469 +463 val_463 +280 val_280 +2 val_2 +35 val_35 +283 val_283 +331 val_331 +235 val_235 +80 val_80 +44 val_44 +193 val_193 +321 val_321 +335 val_335 +104 val_104 +466 val_466 +366 val_366 +175 val_175 +403 val_403 +483 val_483 +53 val_53 +105 val_105 +257 val_257 +406 val_406 +409 val_409 +190 val_190 +406 val_406 +401 val_401 +114 val_114 +258 val_258 +90 val_90 +203 val_203 +262 val_262 +348 val_348 +424 val_424 +12 val_12 +396 val_396 +201 val_201 +217 val_217 +164 val_164 +431 val_431 +454 val_454 +478 val_478 +298 val_298 +125 val_125 +431 val_431 +164 val_164 +424 val_424 +187 val_187 +382 val_382 +5 val_5 +70 val_70 +397 val_397 +480 val_480 +291 val_291 +24 val_24 +351 val_351 +255 val_255 +104 val_104 +70 val_70 +163 val_163 +438 val_438 +119 val_119 +414 val_414 +200 val_200 +491 val_491 +237 val_237 +439 val_439 +360 val_360 +248 val_248 +479 val_479 +305 val_305 +417 val_417 +199 val_199 +444 val_444 +120 val_120 +429 val_429 +169 val_169 +443 val_443 +323 val_323 +325 val_325 +277 val_277 +230 val_230 +478 val_478 +178 val_178 +468 val_468 +310 val_310 +317 val_317 +333 val_333 +493 val_493 +460 val_460 +207 val_207 +249 val_249 +265 val_265 +480 val_480 +83 val_83 +136 val_136 +353 val_353 +172 val_172 +214 val_214 +462 val_462 +233 val_233 +406 val_406 +133 val_133 +175 val_175 +189 val_189 +454 val_454 +375 val_375 +401 val_401 +421 val_421 +407 val_407 +384 val_384 +256 val_256 +26 val_26 +134 val_134 +67 val_67 +384 val_384 +379 val_379 +18 val_18 +462 val_462 +492 val_492 +100 val_100 +298 val_298 +9 val_9 +341 val_341 +498 val_498 +146 val_146 +458 val_458 +362 val_362 +186 val_186 +285 val_285 +348 val_348 +167 val_167 +18 val_18 +273 val_273 +183 val_183 +281 val_281 +344 val_344 +97 val_97 +469 val_469 +315 val_315 +84 val_84 +28 val_28 +37 val_37 +448 val_448 +152 val_152 +348 val_348 +307 val_307 +194 val_194 +414 val_414 +477 val_477 +222 val_222 +126 val_126 +90 val_90 +169 val_169 +403 val_403 +400 val_400 +200 val_200 +97 val_97 diff --git a/sql/hive/src/test/resources/golden/create table as with db name within backticks-3-afd6e46b6a289c3c24a8eec75a94043c b/sql/hive/src/test/resources/golden/create table as with db name within backticks-3-afd6e46b6a289c3c24a8eec75a94043c new file mode 100644 index 0000000000000..e69de29bb2d1d diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveQuerySuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveQuerySuite.scala index c0d21bc9a89da..de140fc72a2c3 100644 --- a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveQuerySuite.scala +++ b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveQuerySuite.scala @@ -334,6 +334,14 @@ class HiveQuerySuite extends HiveComparisonTest with BeforeAndAfter { |DROP DATABASE IF EXISTS testdb CASCADE """.stripMargin) + createQueryTest("create table as with db name within backticks", + """ + |CREATE DATABASE IF NOT EXISTS testdb; + |CREATE TABLE `testdb`.`createdtable` AS SELECT * FROM default.src; + |SELECT * FROM testdb.createdtable; + |DROP DATABASE IF EXISTS testdb CASCADE + """.stripMargin) + createQueryTest("insert table with db name", """ |CREATE DATABASE IF NOT EXISTS testdb; From 3579003115fa3217cff6aa400729d96b0c7b257b Mon Sep 17 00:00:00 2001 From: Michael Armbrust Date: Tue, 17 Mar 2015 19:47:51 -0700 Subject: [PATCH 090/122] [SPARK-6247][SQL] Fix resolution of ambiguous joins caused by new aliases We need to handle ambiguous `exprId`s that are produced by new aliases as well as those caused by leaf nodes (`MultiInstanceRelation`). Attempting to fix this revealed a bug in `equals` for `Alias` as these objects were comparing equal even when the expression ids did not match. Additionally, `LocalRelation` did not correctly provide statistics, and some tests in `catalyst` and `hive` were not using the helper functions for comparing plans. Based on #4991 by chenghao-intel Author: Michael Armbrust Closes #5062 from marmbrus/selfJoins and squashes the following commits: 8e9b84b [Michael Armbrust] check qualifier too 8038a36 [Michael Armbrust] handle aggs too 0b9c687 [Michael Armbrust] fix more tests c3c574b [Michael Armbrust] revert change. 725f1ab [Michael Armbrust] add statistics a925d08 [Michael Armbrust] check for conflicting attributes in join resolution b022ef7 [Michael Armbrust] Handle project aliases. d8caa40 [Michael Armbrust] test case: SPARK-6247 f9c67c2 [Michael Armbrust] Check for duplicate attributes in join resolution. 898af73 [Michael Armbrust] Fix Alias equality. --- .../sql/catalyst/analysis/Analyzer.scala | 30 +++++++++++++++--- .../expressions/namedExpressions.scala | 6 ++++ .../plans/logical/LocalRelation.scala | 3 ++ .../plans/logical/basicOperators.scala | 7 +++++ .../analysis/HiveTypeCoercionSuite.scala | 10 +++--- .../spark/sql/catalyst/plans/PlanTest.scala | 11 +++++-- .../spark/sql/ColumnExpressionSuite.scala | 6 +++- .../org/apache/spark/sql/SQLQuerySuite.scala | 31 +++++++++++++++++++ .../spark/sql/catalyst/plans/PlanTest.scala | 4 ++- 9 files changed, 96 insertions(+), 12 deletions(-) diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala index 7753331748d7b..92d3db077c5e1 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala @@ -237,22 +237,33 @@ class Analyzer(catalog: Catalog, // Special handling for cases when self-join introduce duplicate expression ids. case j @ Join(left, right, _, _) if left.outputSet.intersect(right.outputSet).nonEmpty => val conflictingAttributes = left.outputSet.intersect(right.outputSet) + logDebug(s"Conflicting attributes ${conflictingAttributes.mkString(",")} in $j") - val (oldRelation, newRelation, attributeRewrites) = right.collect { + val (oldRelation, newRelation) = right.collect { + // Handle base relations that might appear more than once. case oldVersion: MultiInstanceRelation if oldVersion.outputSet.intersect(conflictingAttributes).nonEmpty => val newVersion = oldVersion.newInstance() - val newAttributes = AttributeMap(oldVersion.output.zip(newVersion.output)) - (oldVersion, newVersion, newAttributes) + (oldVersion, newVersion) + + // Handle projects that create conflicting aliases. + case oldVersion @ Project(projectList, _) + if findAliases(projectList).intersect(conflictingAttributes).nonEmpty => + (oldVersion, oldVersion.copy(projectList = newAliases(projectList))) + + case oldVersion @ Aggregate(_, aggregateExpressions, _) + if findAliases(aggregateExpressions).intersect(conflictingAttributes).nonEmpty => + (oldVersion, oldVersion.copy(aggregateExpressions = newAliases(aggregateExpressions))) }.head // Only handle first case found, others will be fixed on the next pass. + val attributeRewrites = AttributeMap(oldRelation.output.zip(newRelation.output)) val newRight = right transformUp { case r if r == oldRelation => newRelation + } transformUp { case other => other transformExpressions { case a: Attribute => attributeRewrites.get(a).getOrElse(a) } } - j.copy(right = newRight) case q: LogicalPlan => @@ -272,6 +283,17 @@ class Analyzer(catalog: Catalog, } } + def newAliases(expressions: Seq[NamedExpression]): Seq[NamedExpression] = { + expressions.map { + case a: Alias => Alias(a.child, a.name)() + case other => other + } + } + + def findAliases(projectList: Seq[NamedExpression]): AttributeSet = { + AttributeSet(projectList.collect { case a: Alias => a.toAttribute }) + } + /** * Returns true if `exprs` contains a [[Star]]. */ diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/namedExpressions.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/namedExpressions.scala index 62c062be6d820..17f7f9fe51376 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/namedExpressions.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/namedExpressions.scala @@ -124,6 +124,12 @@ case class Alias(child: Expression, name: String) override def toString: String = s"$child AS $name#${exprId.id}$typeSuffix" override protected final def otherCopyArgs = exprId :: qualifiers :: Nil + + override def equals(other: Any): Boolean = other match { + case a: Alias => + name == a.name && exprId == a.exprId && child == a.child && qualifiers == a.qualifiers + case _ => false + } } /** diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/LocalRelation.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/LocalRelation.scala index 92bd057c6f4b6..bb79dc340553b 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/LocalRelation.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/LocalRelation.scala @@ -54,4 +54,7 @@ case class LocalRelation(output: Seq[Attribute], data: Seq[Row] = Nil) otherOutput.map(_.dataType) == output.map(_.dataType) && otherData == data case _ => false } + + override lazy val statistics = + Statistics(sizeInBytes = output.map(_.dataType.defaultSize).sum * data.length) } diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/basicOperators.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/basicOperators.scala index 624912dab4652..1e7b449d75b80 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/basicOperators.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/basicOperators.scala @@ -108,6 +108,13 @@ case class Join( left.output ++ right.output } } + + def selfJoinResolved = left.outputSet.intersect(right.outputSet).isEmpty + + // Joins are only resolved if they don't introduce ambiguious expression ids. + override lazy val resolved: Boolean = { + childrenResolved && !expressions.exists(!_.resolved) && selfJoinResolved + } } case class Except(left: LogicalPlan, right: LogicalPlan) extends BinaryNode { diff --git a/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/analysis/HiveTypeCoercionSuite.scala b/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/analysis/HiveTypeCoercionSuite.scala index 85798d0871fda..ecbb54218d457 100644 --- a/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/analysis/HiveTypeCoercionSuite.scala +++ b/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/analysis/HiveTypeCoercionSuite.scala @@ -17,13 +17,13 @@ package org.apache.spark.sql.catalyst.analysis -import org.scalatest.FunSuite +import org.apache.spark.sql.catalyst.plans.PlanTest import org.apache.spark.sql.catalyst.expressions._ import org.apache.spark.sql.catalyst.plans.logical.{LocalRelation, Project} import org.apache.spark.sql.types._ -class HiveTypeCoercionSuite extends FunSuite { +class HiveTypeCoercionSuite extends PlanTest { test("tightest common bound for types") { def widenTest(t1: DataType, t2: DataType, tightestCommon: Option[DataType]) { @@ -106,7 +106,8 @@ class HiveTypeCoercionSuite extends FunSuite { val booleanCasts = new HiveTypeCoercion { }.BooleanCasts def ruleTest(initial: Expression, transformed: Expression) { val testRelation = LocalRelation(AttributeReference("a", IntegerType)()) - assert(booleanCasts(Project(Seq(Alias(initial, "a")()), testRelation)) == + comparePlans( + booleanCasts(Project(Seq(Alias(initial, "a")()), testRelation)), Project(Seq(Alias(transformed, "a")()), testRelation)) } // Remove superflous boolean -> boolean casts. @@ -119,7 +120,8 @@ class HiveTypeCoercionSuite extends FunSuite { val fac = new HiveTypeCoercion { }.FunctionArgumentConversion def ruleTest(initial: Expression, transformed: Expression) { val testRelation = LocalRelation(AttributeReference("a", IntegerType)()) - assert(fac(Project(Seq(Alias(initial, "a")()), testRelation)) == + comparePlans( + fac(Project(Seq(Alias(initial, "a")()), testRelation)), Project(Seq(Alias(transformed, "a")()), testRelation)) } ruleTest( diff --git a/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/plans/PlanTest.scala b/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/plans/PlanTest.scala index 7d609b91389c6..48884040bfce7 100644 --- a/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/plans/PlanTest.scala +++ b/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/plans/PlanTest.scala @@ -19,8 +19,8 @@ package org.apache.spark.sql.catalyst.plans import org.scalatest.FunSuite -import org.apache.spark.sql.catalyst.expressions.{ExprId, AttributeReference} -import org.apache.spark.sql.catalyst.plans.logical.LogicalPlan +import org.apache.spark.sql.catalyst.expressions._ +import org.apache.spark.sql.catalyst.plans.logical.{NoRelation, Filter, LogicalPlan} import org.apache.spark.sql.catalyst.util._ /** @@ -36,6 +36,8 @@ class PlanTest extends FunSuite { plan transformAllExpressions { case a: AttributeReference => AttributeReference(a.name, a.dataType, a.nullable)(exprId = ExprId(0)) + case a: Alias => + Alias(a.child, a.name)(exprId = ExprId(0)) } } @@ -50,4 +52,9 @@ class PlanTest extends FunSuite { |${sideBySide(normalized1.treeString, normalized2.treeString).mkString("\n")} """.stripMargin) } + + /** Fails the test if the two expressions do not match */ + protected def compareExpressions(e1: Expression, e2: Expression): Unit = { + comparePlans(Filter(e1, NoRelation), Filter(e2, NoRelation)) + } } diff --git a/sql/core/src/test/scala/org/apache/spark/sql/ColumnExpressionSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/ColumnExpressionSuite.scala index 3036fbc05d021..a53ae97d6243a 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/ColumnExpressionSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/ColumnExpressionSuite.scala @@ -17,6 +17,8 @@ package org.apache.spark.sql +import org.apache.spark.sql.catalyst.expressions.NamedExpression +import org.apache.spark.sql.catalyst.plans.logical.{Project, NoRelation} import org.apache.spark.sql.functions._ import org.apache.spark.sql.test.TestSQLContext import org.apache.spark.sql.test.TestSQLContext.implicits._ @@ -311,7 +313,9 @@ class ColumnExpressionSuite extends QueryTest { } test("lift alias out of cast") { - assert(col("1234").as("name").cast("int").expr === col("1234").cast("int").as("name").expr) + compareExpressions( + col("1234").as("name").cast("int").expr, + col("1234").cast("int").as("name").expr) } test("columns can be compared") { diff --git a/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala index 4dedcd365f6cc..a3c0076e16d6c 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala @@ -36,6 +36,37 @@ class SQLQuerySuite extends QueryTest with BeforeAndAfterAll { import org.apache.spark.sql.test.TestSQLContext.implicits._ val sqlCtx = TestSQLContext + test("self join with aliases") { + Seq(1,2,3).map(i => (i, i.toString)).toDF("int", "str").registerTempTable("df") + + checkAnswer( + sql( + """ + |SELECT x.str, COUNT(*) + |FROM df x JOIN df y ON x.str = y.str + |GROUP BY x.str + """.stripMargin), + Row("1", 1) :: Row("2", 1) :: Row("3", 1) :: Nil) + } + + test("self join with alias in agg") { + Seq(1,2,3) + .map(i => (i, i.toString)) + .toDF("int", "str") + .groupBy("str") + .agg($"str", count("str").as("strCount")) + .registerTempTable("df") + + checkAnswer( + sql( + """ + |SELECT x.str, SUM(x.strCount) + |FROM df x JOIN df y ON x.str = y.str + |GROUP BY x.str + """.stripMargin), + Row("1", 1) :: Row("2", 1) :: Row("3", 1) :: Nil) + } + test("SPARK-4625 support SORT BY in SimpleSQLParser & DSL") { checkAnswer( sql("SELECT a FROM testData2 SORT BY a"), diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/catalyst/plans/PlanTest.scala b/sql/hive/src/test/scala/org/apache/spark/sql/catalyst/plans/PlanTest.scala index 44ee5ab5975fb..98f1c0e69e29d 100644 --- a/sql/hive/src/test/scala/org/apache/spark/sql/catalyst/plans/PlanTest.scala +++ b/sql/hive/src/test/scala/org/apache/spark/sql/catalyst/plans/PlanTest.scala @@ -17,7 +17,7 @@ package org.apache.spark.sql.catalyst.plans -import org.apache.spark.sql.catalyst.expressions.{AttributeReference, ExprId} +import org.apache.spark.sql.catalyst.expressions.{Alias, AttributeReference, ExprId} import org.apache.spark.sql.catalyst.plans.logical.LogicalPlan import org.apache.spark.sql.catalyst.util._ import org.scalatest.FunSuite @@ -38,6 +38,8 @@ class PlanTest extends FunSuite { plan transformAllExpressions { case a: AttributeReference => AttributeReference(a.name, a.dataType, a.nullable)(exprId = ExprId(0)) + case a: Alias => + Alias(a.child, a.name)(exprId = ExprId(0)) } } From 6205a255aea0652dddadf953771f5405065b5bec Mon Sep 17 00:00:00 2001 From: Marcelo Vanzin Date: Wed, 18 Mar 2015 09:06:57 -0400 Subject: [PATCH 091/122] [SPARK-6372] [core] Propagate --conf to child processes. And add unit test. Author: Marcelo Vanzin Closes #5057 from vanzin/SPARK-6372 and squashes the following commits: b33728b [Marcelo Vanzin] [SPARK-6372] [core] Propagate --conf to child processes. --- .../spark/launcher/SparkSubmitCommandBuilder.java | 10 +--------- .../spark/launcher/SparkSubmitCommandBuilderSuite.java | 4 ++++ 2 files changed, 5 insertions(+), 9 deletions(-) diff --git a/launcher/src/main/java/org/apache/spark/launcher/SparkSubmitCommandBuilder.java b/launcher/src/main/java/org/apache/spark/launcher/SparkSubmitCommandBuilder.java index 6ffdff63d3c78..91dcf70f105db 100644 --- a/launcher/src/main/java/org/apache/spark/launcher/SparkSubmitCommandBuilder.java +++ b/launcher/src/main/java/org/apache/spark/launcher/SparkSubmitCommandBuilder.java @@ -253,12 +253,6 @@ private boolean isClientMode(Properties userProps) { private class OptionParser extends SparkSubmitOptionParser { - private final List driverJvmKeys = Arrays.asList( - SparkLauncher.DRIVER_EXTRA_CLASSPATH, - SparkLauncher.DRIVER_EXTRA_JAVA_OPTIONS, - SparkLauncher.DRIVER_EXTRA_LIBRARY_PATH, - SparkLauncher.DRIVER_MEMORY); - @Override protected boolean handle(String opt, String value) { if (opt.equals(MASTER)) { @@ -278,9 +272,7 @@ protected boolean handle(String opt, String value) { } else if (opt.equals(CONF)) { String[] setConf = value.split("=", 2); checkArgument(setConf.length == 2, "Invalid argument to %s: %s", CONF, value); - if (driverJvmKeys.contains(setConf[0])) { - conf.put(setConf[0], setConf[1]); - } + conf.put(setConf[0], setConf[1]); } else if (opt.equals(CLASS)) { // The special classes require some special command line handling, since they allow // mixing spark-submit arguments with arguments that should be propagated to the shell diff --git a/launcher/src/test/java/org/apache/spark/launcher/SparkSubmitCommandBuilderSuite.java b/launcher/src/test/java/org/apache/spark/launcher/SparkSubmitCommandBuilderSuite.java index 815edc4e4971f..626116a9e7477 100644 --- a/launcher/src/test/java/org/apache/spark/launcher/SparkSubmitCommandBuilderSuite.java +++ b/launcher/src/test/java/org/apache/spark/launcher/SparkSubmitCommandBuilderSuite.java @@ -68,6 +68,8 @@ public void testCliParser() throws Exception { parser.DRIVER_JAVA_OPTIONS, "extraJavaOpt", parser.CONF, + "spark.randomOption=foo", + parser.CONF, SparkLauncher.DRIVER_EXTRA_LIBRARY_PATH + "=/driverLibPath"); Map env = new HashMap(); List cmd = buildCommand(sparkSubmitArgs, env); @@ -77,6 +79,8 @@ public void testCliParser() throws Exception { assertTrue(findInStringList(findArgValue(cmd, "-cp"), File.pathSeparator, "/driverCp")); assertTrue("Driver -Xms should be configured.", cmd.contains("-Xms42g")); assertTrue("Driver -Xmx should be configured.", cmd.contains("-Xmx42g")); + assertTrue("Command should contain user-defined conf.", + Collections.indexOfSubList(cmd, Arrays.asList(parser.CONF, "spark.randomOption=foo")) > 0); } @Test From e09c852d6b83b9b112685d113f2792daec8785a3 Mon Sep 17 00:00:00 2001 From: Steve Loughran Date: Wed, 18 Mar 2015 09:09:32 -0400 Subject: [PATCH 092/122] SPARK-6389 YARN app diagnostics report doesn't report NPEs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Trivial patch to implicitly call `Exception.toString()` over `Exception.getMessage()` —this defaults to including the exception class & any non-null message; some subclasses include more. No test. Author: Steve Loughran Closes #5070 from steveloughran/stevel/patches/SPARK-6389-NPE-reporting and squashes the following commits: 8239d85 [Steve Loughran] SPARK-6389 cull use of getMessage over toString in the container launcher 6fbaf6a [Steve Loughran] SPARK-6389 YARN app diagnostics report doesn't report NPEs --- .../org/apache/spark/deploy/yarn/ApplicationMaster.scala | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala b/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala index e966bfba7bb7d..056b8c0257cfe 100644 --- a/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala +++ b/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala @@ -151,7 +151,7 @@ private[spark] class ApplicationMaster( logError("Uncaught exception: ", e) finish(FinalApplicationStatus.FAILED, ApplicationMaster.EXIT_UNCAUGHT_EXCEPTION, - "Uncaught exception: " + e.getMessage()) + "Uncaught exception: " + e) } exitCode } @@ -486,10 +486,10 @@ private[spark] class ApplicationMaster( case _: InterruptedException => // Reporter thread can interrupt to stop user class case cause: Throwable => - logError("User class threw exception: " + cause.getMessage, cause) + logError("User class threw exception: " + cause, cause) finish(FinalApplicationStatus.FAILED, ApplicationMaster.EXIT_EXCEPTION_USER_CLASS, - "User class threw exception: " + cause.getMessage) + "User class threw exception: " + cause) } } } From 9d112a958ee2facad179344dd367a6d1ccbc9614 Mon Sep 17 00:00:00 2001 From: Iulian Dragos Date: Wed, 18 Mar 2015 09:15:33 -0400 Subject: [PATCH 093/122] [SPARK-6286][minor] Handle missing Mesos case TASK_ERROR. Author: Iulian Dragos Closes #5000 from dragos/issue/task-error-case and squashes the following commits: e063627 [Iulian Dragos] Handle TASK_ERROR in Mesos scheduler backends. ac17cf0 [Iulian Dragos] Handle missing Mesos case TASK_ERROR. --- core/src/main/scala/org/apache/spark/TaskState.scala | 1 + .../cluster/mesos/CoarseMesosSchedulerBackend.scala | 12 ++---------- .../cluster/mesos/MesosSchedulerBackend.scala | 10 +--------- 3 files changed, 4 insertions(+), 19 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/TaskState.scala b/core/src/main/scala/org/apache/spark/TaskState.scala index 0bf1e4a5e2ccd..d85a6d683427d 100644 --- a/core/src/main/scala/org/apache/spark/TaskState.scala +++ b/core/src/main/scala/org/apache/spark/TaskState.scala @@ -46,5 +46,6 @@ private[spark] object TaskState extends Enumeration { case MesosTaskState.TASK_FAILED => FAILED case MesosTaskState.TASK_KILLED => KILLED case MesosTaskState.TASK_LOST => LOST + case MesosTaskState.TASK_ERROR => LOST } } diff --git a/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/CoarseMesosSchedulerBackend.scala b/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/CoarseMesosSchedulerBackend.scala index 90dfe14352a8e..fc92b9c35c3a3 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/CoarseMesosSchedulerBackend.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/CoarseMesosSchedulerBackend.scala @@ -28,7 +28,7 @@ import org.apache.mesos.{Scheduler => MScheduler} import org.apache.mesos._ import org.apache.mesos.Protos.{TaskInfo => MesosTaskInfo, TaskState => MesosTaskState, _} -import org.apache.spark.{Logging, SparkContext, SparkEnv, SparkException} +import org.apache.spark.{Logging, SparkContext, SparkEnv, SparkException, TaskState} import org.apache.spark.scheduler.TaskSchedulerImpl import org.apache.spark.scheduler.cluster.CoarseGrainedSchedulerBackend import org.apache.spark.util.{Utils, AkkaUtils} @@ -262,20 +262,12 @@ private[spark] class CoarseMesosSchedulerBackend( .build() } - /** Check whether a Mesos task state represents a finished task */ - private def isFinished(state: MesosTaskState) = { - state == MesosTaskState.TASK_FINISHED || - state == MesosTaskState.TASK_FAILED || - state == MesosTaskState.TASK_KILLED || - state == MesosTaskState.TASK_LOST - } - override def statusUpdate(d: SchedulerDriver, status: TaskStatus) { val taskId = status.getTaskId.getValue.toInt val state = status.getState logInfo("Mesos task " + taskId + " is now " + state) synchronized { - if (isFinished(state)) { + if (TaskState.isFinished(TaskState.fromMesos(state))) { val slaveId = taskIdToSlaveId(taskId) slaveIdsWithExecutors -= slaveId taskIdToSlaveId -= taskId diff --git a/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MesosSchedulerBackend.scala b/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MesosSchedulerBackend.scala index cfb6592e14aa8..df8f4306b88a8 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MesosSchedulerBackend.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MesosSchedulerBackend.scala @@ -313,14 +313,6 @@ private[spark] class MesosSchedulerBackend( .build() } - /** Check whether a Mesos task state represents a finished task */ - def isFinished(state: MesosTaskState) = { - state == MesosTaskState.TASK_FINISHED || - state == MesosTaskState.TASK_FAILED || - state == MesosTaskState.TASK_KILLED || - state == MesosTaskState.TASK_LOST - } - override def statusUpdate(d: SchedulerDriver, status: TaskStatus) { inClassLoader() { val tid = status.getTaskId.getValue.toLong @@ -330,7 +322,7 @@ private[spark] class MesosSchedulerBackend( // We lost the executor on this slave, so remember that it's gone removeExecutor(taskIdToSlaveId(tid), "Lost executor") } - if (isFinished(status.getState)) { + if (TaskState.isFinished(state)) { taskIdToSlaveId.remove(tid) } } From 981fbafa2a878e86abeefe1d77cca01fd848f9f6 Mon Sep 17 00:00:00 2001 From: Marcelo Vanzin Date: Wed, 18 Mar 2015 09:18:28 -0400 Subject: [PATCH 094/122] [SPARK-6325] [core,yarn] Do not change target executor count when killing executors. The dynamic execution code has two ways to reduce the number of executors: one where it reduces the total number of executors it wants, by asking for an absolute number of executors that is lower than the previous one. The second is by explicitly killing idle executors. YarnAllocator was mixing those up and lowering the target number of executors when a kill was issued. Instead, trust the frontend knows what it's doing, and kill executors without messing with other accounting. That means that if the frontend kills an executor without lowering the target, it will get a new executor shortly. The one situation where both actions (lower the target and kill executor) need to happen together is when user code explicitly calls `SparkContext.killExecutors`. In that case, issue two calls to the backend to achieve the goal. I also did some minor cleanup in related code: - avoid sending a request for executors when target is unchanged, to avoid log spam in the AM - avoid printing misleading log messages in the AM when there are no requests to cancel - fix a slow memory leak plus misleading error message on the driver caused by failing to completely unregister the executor. Author: Marcelo Vanzin Closes #5018 from vanzin/SPARK-6325 and squashes the following commits: 2e782a3 [Marcelo Vanzin] Avoid redundant logging on the AM side. a3567cd [Marcelo Vanzin] Add parentheses. a363926 [Marcelo Vanzin] Update logic. a158101 [Marcelo Vanzin] [SPARK-6325] [core,yarn] Disallow reducing executor count past running count. --- .../CoarseGrainedSchedulerBackend.scala | 7 ++++++ .../spark/deploy/yarn/ApplicationMaster.scala | 1 - .../spark/deploy/yarn/YarnAllocator.scala | 13 ++++++----- .../deploy/yarn/YarnAllocatorSuite.scala | 22 +++++++++++++++++++ 4 files changed, 37 insertions(+), 6 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/scheduler/cluster/CoarseGrainedSchedulerBackend.scala b/core/src/main/scala/org/apache/spark/scheduler/cluster/CoarseGrainedSchedulerBackend.scala index 6f77fa32ce37b..87ebf31139ce9 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/cluster/CoarseGrainedSchedulerBackend.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/cluster/CoarseGrainedSchedulerBackend.scala @@ -211,6 +211,7 @@ class CoarseGrainedSchedulerBackend(scheduler: TaskSchedulerImpl, val actorSyste // This must be synchronized because variables mutated // in this block are read when requesting executors CoarseGrainedSchedulerBackend.this.synchronized { + addressToExecutorId -= executorInfo.executorAddress executorDataMap -= executorId executorsPendingToRemove -= executorId } @@ -371,6 +372,12 @@ class CoarseGrainedSchedulerBackend(scheduler: TaskSchedulerImpl, val actorSyste logWarning(s"Executor to kill $id does not exist!") } } + // Killing executors means effectively that we want less executors than before, so also update + // the target number of executors to avoid having the backend allocate new ones. + val newTotal = (numExistingExecutors + numPendingExecutors - executorsPendingToRemove.size + - filteredExecutorIds.size) + doRequestTotalExecutors(newTotal) + executorsPendingToRemove ++= filteredExecutorIds doKillExecutors(filteredExecutorIds) } diff --git a/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala b/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala index 056b8c0257cfe..3d18690cd9cbf 100644 --- a/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala +++ b/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala @@ -534,7 +534,6 @@ private[spark] class ApplicationMaster( driver ! x case RequestExecutors(requestedTotal) => - logInfo(s"Driver requested a total number of $requestedTotal executor(s).") Option(allocator) match { case Some(a) => a.requestTotalExecutors(requestedTotal) case None => logWarning("Container allocator is not ready to request executors yet.") diff --git a/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala b/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala index 55bfbcd9cb84b..c98763e15b58f 100644 --- a/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala +++ b/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala @@ -86,7 +86,8 @@ private[yarn] class YarnAllocator( @volatile private var targetNumExecutors = args.numExecutors // Keep track of which container is running which executor to remove the executors later - private val executorIdToContainer = new HashMap[String, Container] + // Visible for testing. + private[yarn] val executorIdToContainer = new HashMap[String, Container] // Executor memory in MB. protected val executorMemory = args.executorMemory @@ -137,7 +138,10 @@ private[yarn] class YarnAllocator( * be killed. */ def requestTotalExecutors(requestedTotal: Int): Unit = synchronized { - targetNumExecutors = requestedTotal + if (requestedTotal != targetNumExecutors) { + logInfo(s"Driver requested a total number of $requestedTotal executor(s).") + targetNumExecutors = requestedTotal + } } /** @@ -148,8 +152,6 @@ private[yarn] class YarnAllocator( val container = executorIdToContainer.remove(executorId).get internalReleaseContainer(container) numExecutorsRunning -= 1 - targetNumExecutors -= 1 - assert(targetNumExecutors >= 0, "Allocator killed more executors than are allocated!") } else { logWarning(s"Attempted to kill unknown executor $executorId!") } @@ -351,7 +353,8 @@ private[yarn] class YarnAllocator( } } - private def processCompletedContainers(completedContainers: Seq[ContainerStatus]): Unit = { + // Visible for testing. + private[yarn] def processCompletedContainers(completedContainers: Seq[ContainerStatus]): Unit = { for (completedContainer <- completedContainers) { val containerId = completedContainer.getContainerId diff --git a/yarn/src/test/scala/org/apache/spark/deploy/yarn/YarnAllocatorSuite.scala b/yarn/src/test/scala/org/apache/spark/deploy/yarn/YarnAllocatorSuite.scala index 3c224f148802e..c09b01bafce37 100644 --- a/yarn/src/test/scala/org/apache/spark/deploy/yarn/YarnAllocatorSuite.scala +++ b/yarn/src/test/scala/org/apache/spark/deploy/yarn/YarnAllocatorSuite.scala @@ -206,6 +206,28 @@ class YarnAllocatorSuite extends FunSuite with Matchers with BeforeAndAfterEach handler.getNumExecutorsRunning should be (2) } + test("kill executors") { + val handler = createAllocator(4) + handler.updateResourceRequests() + handler.getNumExecutorsRunning should be (0) + handler.getNumPendingAllocate should be (4) + + val container1 = createContainer("host1") + val container2 = createContainer("host2") + handler.handleAllocatedContainers(Array(container1, container2)) + + handler.requestTotalExecutors(1) + handler.executorIdToContainer.keys.foreach { id => handler.killExecutor(id ) } + + val statuses = Seq(container1, container2).map { c => + ContainerStatus.newInstance(c.getId(), ContainerState.COMPLETE, "Finished", 0) + } + handler.updateResourceRequests() + handler.processCompletedContainers(statuses.toSeq) + handler.getNumExecutorsRunning should be (0) + handler.getNumPendingAllocate should be (1) + } + test("memory exceeded diagnostic regexes") { val diagnostics = "Container [pid=12465,containerID=container_1412887393566_0003_01_000002] is running " + From a95ee242b0a9644c912fc54ed68b4301c9558bc9 Mon Sep 17 00:00:00 2001 From: Yuhao Yang Date: Wed, 18 Mar 2015 13:44:37 -0400 Subject: [PATCH 095/122] [SPARK-6374] [MLlib] add get for GeneralizedLinearAlgo I find it's better to have getter for NumFeatures and addIntercept within GeneralizedLinearAlgorithm during actual usage, otherwise I 'll have to get the value through debug. Author: Yuhao Yang Closes #5058 from hhbyyh/addGetLinear and squashes the following commits: 9dc90e8 [Yuhao Yang] add get for GeneralizedLinearAlgo --- .../mllib/regression/GeneralizedLinearAlgorithm.scala | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala b/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala index 7c66e8cdebdbe..b262bec904525 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala @@ -123,6 +123,11 @@ abstract class GeneralizedLinearAlgorithm[M <: GeneralizedLinearModel] */ private var useFeatureScaling = false + /** + * The dimension of training features. + */ + def getNumFeatures: Int = this.numFeatures + /** * The dimension of training features. */ @@ -141,6 +146,11 @@ abstract class GeneralizedLinearAlgorithm[M <: GeneralizedLinearModel] */ protected def createModel(weights: Vector, intercept: Double): M + /** + * Get if the algorithm uses addIntercept + */ + def isAddIntercept: Boolean = this.addIntercept + /** * Set if the algorithm should add an intercept. Default false. * We set the default to false because adding the intercept will cause memory allocation. From 3db13874250ded267d7455898e4048a70a47fdcd Mon Sep 17 00:00:00 2001 From: Jongyoul Lee Date: Wed, 18 Mar 2015 20:54:22 -0400 Subject: [PATCH 096/122] SPARK-6085 Part. 2 Increase default value for memory overhead - fixed a description of spark.mesos.executor.memoryOverhead from 7% to 10% - This is a second part of SPARK-6085 Author: Jongyoul Lee Closes #5065 from jongyoul/SPARK-6085-1 and squashes the following commits: c5af84c [Jongyoul Lee] SPARK-6085 Part. 2 Increase default value for memory overhead - Changed "MiB" to "MB" dbac1c0 [Jongyoul Lee] SPARK-6085 Part. 2 Increase default value for memory overhead - fixed a description of spark.mesos.executor.memoryOverhead from 7% to 10% --- docs/running-on-mesos.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/docs/running-on-mesos.md b/docs/running-on-mesos.md index 59a3e9d25baf1..6a9d304501dc0 100644 --- a/docs/running-on-mesos.md +++ b/docs/running-on-mesos.md @@ -224,11 +224,11 @@ See the [configuration page](configuration.html) for information on Spark config
    spark.files.useFetchCachetrue + If set to true (default), file fetching will use a local cache that is shared by executors + that belong to the same application, which can improve task launching performance when + running many executors on the same host. If set to false, these caching optimizations will + be disabled and all executors will fetch their own copies of files. This optimization may be + disabled in order to use Spark local directories that reside on NFS filesystems (see + SPARK-6313 for more details). +
    spark.files.overwrite falsespark.mesos.executor.memoryOverhead executor memory * 0.10, with minimum of 384 - This value is an additive for spark.executor.memory, specified in MiB, + This value is an additive for spark.executor.memory, specified in MB, which is used to calculate the total Mesos task memory. A value of 384 - implies a 384MiB overhead. Additionally, there is a hard-coded 7% minimum + implies a 384MB overhead. Additionally, there is a hard-coded 10% minimum overhead. The final overhead will be the larger of either - `spark.mesos.executor.memoryOverhead` or 7% of `spark.executor.memory`. + `spark.mesos.executor.memoryOverhead` or 10% of `spark.executor.memory`.
    From 540b2a4eabe0bad2455f5140c4ad6a315e37cc3f Mon Sep 17 00:00:00 2001 From: Wenchen Fan Date: Wed, 18 Mar 2015 19:43:04 -0700 Subject: [PATCH 097/122] [SPARK-6394][Core] cleanup BlockManager companion object and improve the getCacheLocs method in DAGScheduler The current implementation include searching a HashMap many times, we can avoid this. Actually if you look into `BlockManager.blockIdsToBlockManagers`, the core function call is [this](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/storage/BlockManager.scala#L1258), so we can call `blockManagerMaster.getLocations` directly and avoid building a HashMap. Author: Wenchen Fan Closes #5043 from cloud-fan/small and squashes the following commits: e959d12 [Wenchen Fan] fix style 203c493 [Wenchen Fan] some cleanup in BlockManager companion object d409099 [Wenchen Fan] address rxin's comment faec999 [Wenchen Fan] add regression test 2fb57aa [Wenchen Fan] imporve the getCacheLocs method --- .../apache/spark/scheduler/DAGScheduler.scala | 11 +++++----- .../apache/spark/storage/BlockManager.scala | 22 ++++--------------- .../spark/scheduler/DAGSchedulerSuite.scala | 12 ++++++++++ 3 files changed, 22 insertions(+), 23 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala b/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala index e4170a55b7981..1021172e6afb4 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala @@ -104,7 +104,7 @@ class DAGScheduler( * * All accesses to this map should be guarded by synchronizing on it (see SPARK-4454). */ - private val cacheLocs = new HashMap[Int, Array[Seq[TaskLocation]]] + private val cacheLocs = new HashMap[Int, Seq[Seq[TaskLocation]]] // For tracking failed nodes, we use the MapOutputTracker's epoch number, which is sent with // every task. When we detect a node failing, we note the current epoch number and failed @@ -188,14 +188,15 @@ class DAGScheduler( eventProcessLoop.post(TaskSetFailed(taskSet, reason)) } - private def getCacheLocs(rdd: RDD[_]): Array[Seq[TaskLocation]] = cacheLocs.synchronized { + private[scheduler] + def getCacheLocs(rdd: RDD[_]): Seq[Seq[TaskLocation]] = cacheLocs.synchronized { // Note: this doesn't use `getOrElse()` because this method is called O(num tasks) times if (!cacheLocs.contains(rdd.id)) { val blockIds = rdd.partitions.indices.map(index => RDDBlockId(rdd.id, index)).toArray[BlockId] - val locs = BlockManager.blockIdsToBlockManagers(blockIds, env, blockManagerMaster) - cacheLocs(rdd.id) = blockIds.map { id => - locs.getOrElse(id, Nil).map(bm => TaskLocation(bm.host, bm.executorId)) + val locs: Seq[Seq[TaskLocation]] = blockManagerMaster.getLocations(blockIds).map { bms => + bms.map(bm => TaskLocation(bm.host, bm.executorId)) } + cacheLocs(rdd.id) = locs } cacheLocs(rdd.id) } diff --git a/core/src/main/scala/org/apache/spark/storage/BlockManager.scala b/core/src/main/scala/org/apache/spark/storage/BlockManager.scala index c8b7763f03fb7..80d66e59132da 100644 --- a/core/src/main/scala/org/apache/spark/storage/BlockManager.scala +++ b/core/src/main/scala/org/apache/spark/storage/BlockManager.scala @@ -1245,10 +1245,10 @@ private[spark] object BlockManager extends Logging { } } - def blockIdsToBlockManagers( + def blockIdsToHosts( blockIds: Array[BlockId], env: SparkEnv, - blockManagerMaster: BlockManagerMaster = null): Map[BlockId, Seq[BlockManagerId]] = { + blockManagerMaster: BlockManagerMaster = null): Map[BlockId, Seq[String]] = { // blockManagerMaster != null is used in tests assert(env != null || blockManagerMaster != null) @@ -1258,24 +1258,10 @@ private[spark] object BlockManager extends Logging { blockManagerMaster.getLocations(blockIds) } - val blockManagers = new HashMap[BlockId, Seq[BlockManagerId]] + val blockManagers = new HashMap[BlockId, Seq[String]] for (i <- 0 until blockIds.length) { - blockManagers(blockIds(i)) = blockLocations(i) + blockManagers(blockIds(i)) = blockLocations(i).map(_.host) } blockManagers.toMap } - - def blockIdsToExecutorIds( - blockIds: Array[BlockId], - env: SparkEnv, - blockManagerMaster: BlockManagerMaster = null): Map[BlockId, Seq[String]] = { - blockIdsToBlockManagers(blockIds, env, blockManagerMaster).mapValues(s => s.map(_.executorId)) - } - - def blockIdsToHosts( - blockIds: Array[BlockId], - env: SparkEnv, - blockManagerMaster: BlockManagerMaster = null): Map[BlockId, Seq[String]] = { - blockIdsToBlockManagers(blockIds, env, blockManagerMaster).mapValues(s => s.map(_.host)) - } } diff --git a/core/src/test/scala/org/apache/spark/scheduler/DAGSchedulerSuite.scala b/core/src/test/scala/org/apache/spark/scheduler/DAGSchedulerSuite.scala index 30119ce5d4eec..63360a0f189a3 100644 --- a/core/src/test/scala/org/apache/spark/scheduler/DAGSchedulerSuite.scala +++ b/core/src/test/scala/org/apache/spark/scheduler/DAGSchedulerSuite.scala @@ -322,6 +322,18 @@ class DAGSchedulerSuite extends FunSuiteLike with BeforeAndAfter with LocalSpar assertDataStructuresEmpty } + test("regression test for getCacheLocs") { + val rdd = new MyRDD(sc, 3, Nil) + cacheLocations(rdd.id -> 0) = + Seq(makeBlockManagerId("hostA"), makeBlockManagerId("hostB")) + cacheLocations(rdd.id -> 1) = + Seq(makeBlockManagerId("hostB"), makeBlockManagerId("hostC")) + cacheLocations(rdd.id -> 2) = + Seq(makeBlockManagerId("hostC"), makeBlockManagerId("hostD")) + val locs = scheduler.getCacheLocs(rdd).map(_.map(_.host)) + assert(locs === Seq(Seq("hostA", "hostB"), Seq("hostB", "hostC"), Seq("hostC", "hostD"))) + } + test("avoid exponential blowup when getting preferred locs list") { // Build up a complex dependency graph with repeated zip operations, without preferred locations. var rdd: RDD[_] = new MyRDD(sc, 1, Nil) From 645cf3fcc21987417b2946bdeeeb60af3edf667e Mon Sep 17 00:00:00 2001 From: Tathagata Das Date: Thu, 19 Mar 2015 02:15:50 -0400 Subject: [PATCH 098/122] [SPARK-6222][Streaming] Dont delete checkpoint data when doing pre-batch-start checkpoint This is another alternative approach to https://github.com/apache/spark/pull/4964/ I think this is a simpler fix that can be backported easily to other branches (1.2 and 1.3). All it does it introduce a flag so that the pre-batch-start checkpoint does not call clear checkpoint. There is not unit test yet. I will add it when this approach is commented upon. Not sure if this is testable easily. Author: Tathagata Das Closes #5008 from tdas/SPARK-6222 and squashes the following commits: 7315bc2 [Tathagata Das] Removed empty line. c438de4 [Tathagata Das] Revert unnecessary change. 5e98374 [Tathagata Das] Added unit test 50cb60b [Tathagata Das] Fixed style issue 295ca5c [Tathagata Das] Fixing SPARK-6222 --- .../apache/spark/streaming/Checkpoint.scala | 12 +- .../streaming/scheduler/JobGenerator.scala | 20 +-- .../scheduler/JobGeneratorSuite.scala | 133 ++++++++++++++++++ 3 files changed, 153 insertions(+), 12 deletions(-) create mode 100644 streaming/src/test/scala/org/apache/spark/streaming/scheduler/JobGeneratorSuite.scala diff --git a/streaming/src/main/scala/org/apache/spark/streaming/Checkpoint.scala b/streaming/src/main/scala/org/apache/spark/streaming/Checkpoint.scala index cb4c94fb9d5a6..db64e11e16304 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/Checkpoint.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/Checkpoint.scala @@ -119,7 +119,10 @@ class CheckpointWriter( private var stopped = false private var fs_ : FileSystem = _ - class CheckpointWriteHandler(checkpointTime: Time, bytes: Array[Byte]) extends Runnable { + class CheckpointWriteHandler( + checkpointTime: Time, + bytes: Array[Byte], + clearCheckpointDataLater: Boolean) extends Runnable { def run() { var attempts = 0 val startTime = System.currentTimeMillis() @@ -166,7 +169,7 @@ class CheckpointWriter( val finishTime = System.currentTimeMillis() logInfo("Checkpoint for time " + checkpointTime + " saved to file '" + checkpointFile + "', took " + bytes.length + " bytes and " + (finishTime - startTime) + " ms") - jobGenerator.onCheckpointCompletion(checkpointTime) + jobGenerator.onCheckpointCompletion(checkpointTime, clearCheckpointDataLater) return } catch { case ioe: IOException => @@ -180,7 +183,7 @@ class CheckpointWriter( } } - def write(checkpoint: Checkpoint) { + def write(checkpoint: Checkpoint, clearCheckpointDataLater: Boolean) { val bos = new ByteArrayOutputStream() val zos = compressionCodec.compressedOutputStream(bos) val oos = new ObjectOutputStream(zos) @@ -188,7 +191,8 @@ class CheckpointWriter( oos.close() bos.close() try { - executor.execute(new CheckpointWriteHandler(checkpoint.checkpointTime, bos.toByteArray)) + executor.execute(new CheckpointWriteHandler( + checkpoint.checkpointTime, bos.toByteArray, clearCheckpointDataLater)) logDebug("Submitted checkpoint of time " + checkpoint.checkpointTime + " writer queue") } catch { case rej: RejectedExecutionException => diff --git a/streaming/src/main/scala/org/apache/spark/streaming/scheduler/JobGenerator.scala b/streaming/src/main/scala/org/apache/spark/streaming/scheduler/JobGenerator.scala index ac92774a38273..59488dfb0f8c6 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/scheduler/JobGenerator.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/scheduler/JobGenerator.scala @@ -30,7 +30,8 @@ import org.apache.spark.util.{Clock, ManualClock} private[scheduler] sealed trait JobGeneratorEvent private[scheduler] case class GenerateJobs(time: Time) extends JobGeneratorEvent private[scheduler] case class ClearMetadata(time: Time) extends JobGeneratorEvent -private[scheduler] case class DoCheckpoint(time: Time) extends JobGeneratorEvent +private[scheduler] case class DoCheckpoint( + time: Time, clearCheckpointDataLater: Boolean) extends JobGeneratorEvent private[scheduler] case class ClearCheckpointData(time: Time) extends JobGeneratorEvent /** @@ -163,8 +164,10 @@ class JobGenerator(jobScheduler: JobScheduler) extends Logging { /** * Callback called when the checkpoint of a batch has been written. */ - def onCheckpointCompletion(time: Time) { - eventActor ! ClearCheckpointData(time) + def onCheckpointCompletion(time: Time, clearCheckpointDataLater: Boolean) { + if (clearCheckpointDataLater) { + eventActor ! ClearCheckpointData(time) + } } /** Processes all events */ @@ -173,7 +176,8 @@ class JobGenerator(jobScheduler: JobScheduler) extends Logging { event match { case GenerateJobs(time) => generateJobs(time) case ClearMetadata(time) => clearMetadata(time) - case DoCheckpoint(time) => doCheckpoint(time) + case DoCheckpoint(time, clearCheckpointDataLater) => + doCheckpoint(time, clearCheckpointDataLater) case ClearCheckpointData(time) => clearCheckpointData(time) } } @@ -245,7 +249,7 @@ class JobGenerator(jobScheduler: JobScheduler) extends Logging { case Failure(e) => jobScheduler.reportError("Error generating jobs for time " + time, e) } - eventActor ! DoCheckpoint(time) + eventActor ! DoCheckpoint(time, clearCheckpointDataLater = false) } /** Clear DStream metadata for the given `time`. */ @@ -255,7 +259,7 @@ class JobGenerator(jobScheduler: JobScheduler) extends Logging { // If checkpointing is enabled, then checkpoint, // else mark batch to be fully processed if (shouldCheckpoint) { - eventActor ! DoCheckpoint(time) + eventActor ! DoCheckpoint(time, clearCheckpointDataLater = true) } else { // If checkpointing is not enabled, then delete metadata information about // received blocks (block data not saved in any case). Otherwise, wait for @@ -278,11 +282,11 @@ class JobGenerator(jobScheduler: JobScheduler) extends Logging { } /** Perform checkpoint for the give `time`. */ - private def doCheckpoint(time: Time) { + private def doCheckpoint(time: Time, clearCheckpointDataLater: Boolean) { if (shouldCheckpoint && (time - graph.zeroTime).isMultipleOf(ssc.checkpointDuration)) { logInfo("Checkpointing graph for time " + time) ssc.graph.updateCheckpointData(time) - checkpointWriter.write(new Checkpoint(ssc, time)) + checkpointWriter.write(new Checkpoint(ssc, time), clearCheckpointDataLater) } } diff --git a/streaming/src/test/scala/org/apache/spark/streaming/scheduler/JobGeneratorSuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/scheduler/JobGeneratorSuite.scala new file mode 100644 index 0000000000000..4150b60635ed6 --- /dev/null +++ b/streaming/src/test/scala/org/apache/spark/streaming/scheduler/JobGeneratorSuite.scala @@ -0,0 +1,133 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.streaming.scheduler + +import java.util.concurrent.CountDownLatch + +import scala.concurrent.duration._ +import scala.language.postfixOps + +import org.scalatest.concurrent.Eventually._ + +import org.apache.spark.rdd.RDD +import org.apache.spark.streaming._ +import org.apache.spark.util.{ManualClock, Utils} + +class JobGeneratorSuite extends TestSuiteBase { + + // SPARK-6222 is a tricky regression bug which causes received block metadata + // to be deleted before the corresponding batch has completed. This occurs when + // the following conditions are met. + // 1. streaming checkpointing is enabled by setting streamingContext.checkpoint(dir) + // 2. input data is received through a receiver as blocks + // 3. a batch processing a set of blocks takes a long time, such that a few subsequent + // batches have been generated and submitted for processing. + // + // The JobGenerator (as of Mar 16, 2015) checkpoints twice per batch, once after generation + // of a batch, and another time after the completion of a batch. The cleanup of + // checkpoint data (including block metadata, etc.) from DStream must be done only after the + // 2nd checkpoint has completed, that is, after the batch has been completely processed. + // However, the issue is that the checkpoint data and along with it received block data is + // cleaned even in the case of the 1st checkpoint, causing pre-mature deletion of received block + // data. For example, if the 3rd batch is still being process, the 7th batch may get generated, + // and the corresponding "1st checkpoint" will delete received block metadata of batch older + // than 6th batch. That, is 3rd batch's block metadata gets deleted even before 3rd batch has + // been completely processed. + // + // This test tries to create that scenario by the following. + // 1. enable checkpointing + // 2. generate batches with received blocks + // 3. make the 3rd batch never complete + // 4. allow subsequent batches to be generated (to allow premature deletion of 3rd batch metadata) + // 5. verify whether 3rd batch's block metadata still exists + // + test("SPARK-6222: Do not clear received block data too soon") { + import JobGeneratorSuite._ + val checkpointDir = Utils.createTempDir() + val testConf = conf + testConf.set("spark.streaming.clock", "org.apache.spark.streaming.util.ManualClock") + testConf.set("spark.streaming.receiver.writeAheadLog.rollingInterval", "1") + + withStreamingContext(new StreamingContext(testConf, batchDuration)) { ssc => + val clock = ssc.scheduler.clock.asInstanceOf[ManualClock] + val numBatches = 10 + val longBatchNumber = 3 // 3rd batch will take a long time + val longBatchTime = longBatchNumber * batchDuration.milliseconds + + val testTimeout = timeout(10 seconds) + val inputStream = ssc.receiverStream(new TestReceiver) + + inputStream.foreachRDD((rdd: RDD[Int], time: Time) => { + if (time.milliseconds == longBatchTime) { + while (waitLatch.getCount() > 0) { + waitLatch.await() + println("Await over") + } + } + }) + + val batchCounter = new BatchCounter(ssc) + ssc.checkpoint(checkpointDir.getAbsolutePath) + ssc.start() + + // Make sure the only 1 batch of information is to be remembered + assert(inputStream.rememberDuration === batchDuration) + val receiverTracker = ssc.scheduler.receiverTracker + + // Get the blocks belonging to a batch + def getBlocksOfBatch(batchTime: Long) = { + receiverTracker.getBlocksOfBatchAndStream(Time(batchTime), inputStream.id) + } + + // Wait for new blocks to be received + def waitForNewReceivedBlocks() { + eventually(testTimeout) { + assert(receiverTracker.hasUnallocatedBlocks) + } + } + + // Wait for received blocks to be allocated to a batch + def waitForBlocksToBeAllocatedToBatch(batchTime: Long) { + eventually(testTimeout) { + assert(getBlocksOfBatch(batchTime).nonEmpty) + } + } + + // Generate a large number of batches with blocks in them + for (batchNum <- 1 to numBatches) { + waitForNewReceivedBlocks() + clock.advance(batchDuration.milliseconds) + waitForBlocksToBeAllocatedToBatch(clock.getTimeMillis()) + } + + // Wait for 3rd batch to start + eventually(testTimeout) { + ssc.scheduler.getPendingTimes().contains(Time(numBatches * batchDuration.milliseconds)) + } + + // Verify that the 3rd batch's block data is still present while the 3rd batch is incomplete + assert(getBlocksOfBatch(longBatchTime).nonEmpty, "blocks of incomplete batch already deleted") + assert(batchCounter.getNumCompletedBatches < longBatchNumber) + waitLatch.countDown() + } + } +} + +object JobGeneratorSuite { + val waitLatch = new CountDownLatch(1) +} From 2c3f83c34bb8d2c1bf13b33633d8c5a8089545d1 Mon Sep 17 00:00:00 2001 From: CodingCat Date: Wed, 18 Mar 2015 23:48:45 -0700 Subject: [PATCH 099/122] [SPARK-4012] stop SparkContext when the exception is thrown from an infinite loop https://issues.apache.org/jira/browse/SPARK-4012 This patch is a resubmission for https://github.com/apache/spark/pull/2864 What I am proposing in this patch is that ***when the exception is thrown from an infinite loop, we should stop the SparkContext, instead of let JVM throws exception forever*** So, in the infinite loops where we originally wrapped with a ` logUncaughtExceptions`, I changed to `tryOrStopSparkContext`, so that the Spark component is stopped Early stopped JVM process is helpful for HA scheme design, for example, The user has a script checking the existence of the pid of the Spark Streaming driver for monitoring the availability; with the code before this patch, the JVM process is still available but not functional when the exceptions are thrown andrewor14, srowen , mind taking further consideration about the change? Author: CodingCat Closes #5004 from CodingCat/SPARK-4012-1 and squashes the following commits: 589276a [CodingCat] throw fatal error again 3c72cd8 [CodingCat] address the comments 6087864 [CodingCat] revise comments 6ad3eb0 [CodingCat] stop SparkContext instead of quit the JVM process 6322959 [CodingCat] exit JVM process when the exception is thrown from an infinite loop --- .../org/apache/spark/ContextCleaner.scala | 2 +- .../scala/org/apache/spark/SparkContext.scala | 2 +- .../deploy/history/FsHistoryProvider.scala | 2 +- .../spark/scheduler/TaskSchedulerImpl.scala | 2 +- .../spark/util/AsynchronousListenerBus.scala | 10 +++++-- .../scala/org/apache/spark/util/Utils.scala | 28 +++++++++++++++++++ .../scheduler/EventLoggingListenerSuite.scala | 9 +++--- .../spark/scheduler/SparkListenerSuite.scala | 10 +++---- .../streaming/scheduler/JobScheduler.scala | 2 +- 9 files changed, 51 insertions(+), 16 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/ContextCleaner.scala b/core/src/main/scala/org/apache/spark/ContextCleaner.scala index 0c59a61e81393..9b05c9623b704 100644 --- a/core/src/main/scala/org/apache/spark/ContextCleaner.scala +++ b/core/src/main/scala/org/apache/spark/ContextCleaner.scala @@ -145,7 +145,7 @@ private[spark] class ContextCleaner(sc: SparkContext) extends Logging { } /** Keep cleaning RDD, shuffle, and broadcast state. */ - private def keepCleaning(): Unit = Utils.logUncaughtExceptions { + private def keepCleaning(): Unit = Utils.tryOrStopSparkContext(sc) { while (!stopped) { try { val reference = Option(referenceQueue.remove(ContextCleaner.REF_QUEUE_POLL_TIMEOUT)) diff --git a/core/src/main/scala/org/apache/spark/SparkContext.scala b/core/src/main/scala/org/apache/spark/SparkContext.scala index 4457f40286fda..228ff715fe7cb 100644 --- a/core/src/main/scala/org/apache/spark/SparkContext.scala +++ b/core/src/main/scala/org/apache/spark/SparkContext.scala @@ -1736,7 +1736,7 @@ class SparkContext(config: SparkConf) extends Logging with ExecutorAllocationCli } } - listenerBus.start() + listenerBus.start(this) } /** Post the application start event */ diff --git a/core/src/main/scala/org/apache/spark/deploy/history/FsHistoryProvider.scala b/core/src/main/scala/org/apache/spark/deploy/history/FsHistoryProvider.scala index 16d88c17d1a76..7fde02040927d 100644 --- a/core/src/main/scala/org/apache/spark/deploy/history/FsHistoryProvider.scala +++ b/core/src/main/scala/org/apache/spark/deploy/history/FsHistoryProvider.scala @@ -93,7 +93,7 @@ private[history] class FsHistoryProvider(conf: SparkConf) extends ApplicationHis */ private def getRunner(operateFun: () => Unit): Runnable = { new Runnable() { - override def run() = Utils.logUncaughtExceptions { + override def run() = Utils.tryOrExit { operateFun() } } diff --git a/core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala b/core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala index 7a9cf1c2e7f30..f33fd4450b2a6 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala @@ -145,7 +145,7 @@ private[spark] class TaskSchedulerImpl( import sc.env.actorSystem.dispatcher sc.env.actorSystem.scheduler.schedule(SPECULATION_INTERVAL milliseconds, SPECULATION_INTERVAL milliseconds) { - Utils.tryOrExit { checkSpeculatableTasks() } + Utils.tryOrStopSparkContext(sc) { checkSpeculatableTasks() } } } } diff --git a/core/src/main/scala/org/apache/spark/util/AsynchronousListenerBus.scala b/core/src/main/scala/org/apache/spark/util/AsynchronousListenerBus.scala index 18c627e8c7a15..ce7887b76ff96 100644 --- a/core/src/main/scala/org/apache/spark/util/AsynchronousListenerBus.scala +++ b/core/src/main/scala/org/apache/spark/util/AsynchronousListenerBus.scala @@ -21,6 +21,7 @@ import java.util.concurrent._ import java.util.concurrent.atomic.AtomicBoolean import com.google.common.annotations.VisibleForTesting +import org.apache.spark.SparkContext /** * Asynchronously passes events to registered listeners. @@ -38,6 +39,8 @@ private[spark] abstract class AsynchronousListenerBus[L <: AnyRef, E](name: Stri self => + private var sparkContext: SparkContext = null + /* Cap the capacity of the event queue so we get an explicit error (rather than * an OOM exception) if it's perpetually being added to more quickly than it's being drained. */ private val EVENT_QUEUE_CAPACITY = 10000 @@ -57,7 +60,7 @@ private[spark] abstract class AsynchronousListenerBus[L <: AnyRef, E](name: Stri private val listenerThread = new Thread(name) { setDaemon(true) - override def run(): Unit = Utils.logUncaughtExceptions { + override def run(): Unit = Utils.tryOrStopSparkContext(sparkContext) { while (true) { eventLock.acquire() self.synchronized { @@ -89,9 +92,12 @@ private[spark] abstract class AsynchronousListenerBus[L <: AnyRef, E](name: Stri * This first sends out all buffered events posted before this listener bus has started, then * listens for any additional events asynchronously while the listener bus is still running. * This should only be called once. + * + * @param sc Used to stop the SparkContext in case the listener thread dies. */ - def start() { + def start(sc: SparkContext) { if (started.compareAndSet(false, true)) { + sparkContext = sc listenerThread.start() } else { throw new IllegalStateException(s"$name already started!") diff --git a/core/src/main/scala/org/apache/spark/util/Utils.scala b/core/src/main/scala/org/apache/spark/util/Utils.scala index af8a24553a461..91aa70870ab20 100644 --- a/core/src/main/scala/org/apache/spark/util/Utils.scala +++ b/core/src/main/scala/org/apache/spark/util/Utils.scala @@ -1146,6 +1146,8 @@ private[spark] object Utils extends Logging { /** * Execute a block of code that evaluates to Unit, forwarding any uncaught exceptions to the * default UncaughtExceptionHandler + * + * NOTE: This method is to be called by the spark-started JVM process. */ def tryOrExit(block: => Unit) { try { @@ -1156,6 +1158,32 @@ private[spark] object Utils extends Logging { } } + /** + * Execute a block of code that evaluates to Unit, stop SparkContext is there is any uncaught + * exception + * + * NOTE: This method is to be called by the driver-side components to avoid stopping the + * user-started JVM process completely; in contrast, tryOrExit is to be called in the + * spark-started JVM process . + */ + def tryOrStopSparkContext(sc: SparkContext)(block: => Unit) { + try { + block + } catch { + case e: ControlThrowable => throw e + case t: Throwable => + val currentThreadName = Thread.currentThread().getName + if (sc != null) { + logError(s"uncaught error in thread $currentThreadName, stopping SparkContext", t) + sc.stop() + } + if (!NonFatal(t)) { + logError(s"throw uncaught fatal error in thread $currentThreadName", t) + throw t + } + } + } + /** * Execute a block of code that evaluates to Unit, re-throwing any non-fatal uncaught * exceptions as IOException. This is used when implementing Externalizable and Serializable's diff --git a/core/src/test/scala/org/apache/spark/scheduler/EventLoggingListenerSuite.scala b/core/src/test/scala/org/apache/spark/scheduler/EventLoggingListenerSuite.scala index 992dde66f982f..448258a754153 100644 --- a/core/src/test/scala/org/apache/spark/scheduler/EventLoggingListenerSuite.scala +++ b/core/src/test/scala/org/apache/spark/scheduler/EventLoggingListenerSuite.scala @@ -25,9 +25,9 @@ import scala.io.Source import org.apache.hadoop.fs.Path import org.json4s.jackson.JsonMethods._ -import org.scalatest.{BeforeAndAfter, FunSuite} +import org.scalatest.{FunSuiteLike, BeforeAndAfter, FunSuite} -import org.apache.spark.{Logging, SparkConf, SparkContext, SPARK_VERSION} +import org.apache.spark._ import org.apache.spark.deploy.SparkHadoopUtil import org.apache.spark.io._ import org.apache.spark.util.{JsonProtocol, Utils} @@ -39,7 +39,8 @@ import org.apache.spark.util.{JsonProtocol, Utils} * logging events, whether the parsing of the file names is correct, and whether the logged events * can be read and deserialized into actual SparkListenerEvents. */ -class EventLoggingListenerSuite extends FunSuite with BeforeAndAfter with Logging { +class EventLoggingListenerSuite extends FunSuite with LocalSparkContext with BeforeAndAfter + with Logging { import EventLoggingListenerSuite._ private val fileSystem = Utils.getHadoopFileSystem("/", @@ -144,7 +145,7 @@ class EventLoggingListenerSuite extends FunSuite with BeforeAndAfter with Loggin // A comprehensive test on JSON de/serialization of all events is in JsonProtocolSuite eventLogger.start() - listenerBus.start() + listenerBus.start(sc) listenerBus.addListener(eventLogger) listenerBus.postToAll(applicationStart) listenerBus.postToAll(applicationEnd) diff --git a/core/src/test/scala/org/apache/spark/scheduler/SparkListenerSuite.scala b/core/src/test/scala/org/apache/spark/scheduler/SparkListenerSuite.scala index 3a41ee8d4ae0c..627c9a4ddfffc 100644 --- a/core/src/test/scala/org/apache/spark/scheduler/SparkListenerSuite.scala +++ b/core/src/test/scala/org/apache/spark/scheduler/SparkListenerSuite.scala @@ -46,7 +46,7 @@ class SparkListenerSuite extends FunSuite with LocalSparkContext with Matchers assert(counter.count === 0) // Starting listener bus should flush all buffered events - bus.start() + bus.start(sc) assert(bus.waitUntilEmpty(WAIT_TIMEOUT_MILLIS)) assert(counter.count === 5) @@ -58,8 +58,8 @@ class SparkListenerSuite extends FunSuite with LocalSparkContext with Matchers // Listener bus must not be started twice intercept[IllegalStateException] { val bus = new LiveListenerBus - bus.start() - bus.start() + bus.start(sc) + bus.start(sc) } // ... or stopped before starting @@ -96,7 +96,7 @@ class SparkListenerSuite extends FunSuite with LocalSparkContext with Matchers val blockingListener = new BlockingListener bus.addListener(blockingListener) - bus.start() + bus.start(sc) bus.post(SparkListenerJobEnd(0, jobCompletionTime, JobSucceeded)) listenerStarted.acquire() @@ -347,7 +347,7 @@ class SparkListenerSuite extends FunSuite with LocalSparkContext with Matchers bus.addListener(badListener) bus.addListener(jobCounter1) bus.addListener(jobCounter2) - bus.start() + bus.start(sc) // Post events to all listeners, and wait until the queue is drained (1 to 5).foreach { _ => bus.post(SparkListenerJobEnd(0, jobCompletionTime, JobSucceeded)) } diff --git a/streaming/src/main/scala/org/apache/spark/streaming/scheduler/JobScheduler.scala b/streaming/src/main/scala/org/apache/spark/streaming/scheduler/JobScheduler.scala index b3ffc71904c76..60bc099b27a4c 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/scheduler/JobScheduler.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/scheduler/JobScheduler.scala @@ -61,7 +61,7 @@ class JobScheduler(val ssc: StreamingContext) extends Logging { } }), "JobScheduler") - listenerBus.start() + listenerBus.start(ssc.sparkContext) receiverTracker = new ReceiverTracker(ssc) receiverTracker.start() jobGenerator.start() From 797f8a000773d848fa52c7fe2eb1b5e5e7f6c55a Mon Sep 17 00:00:00 2001 From: Pierre Borckmans Date: Thu, 19 Mar 2015 08:02:06 -0400 Subject: [PATCH 100/122] [SPARK-6402][DOC] - Remove some refererences to shark in docs and ec2 EC2 script and job scheduling documentation still refered to Shark. I removed these references. I also removed a remaining `SHARK_VERSION` variable from `ec2-variables.sh`. Author: Pierre Borckmans Closes #5083 from pierre-borckmans/remove_refererences_to_shark_in_docs and squashes the following commits: 4e90ffc [Pierre Borckmans] Removed deprecated SHARK_VERSION caea407 [Pierre Borckmans] Remove shark reference from ec2 script doc 196c744 [Pierre Borckmans] Removed references to Shark --- docs/ec2-scripts.md | 2 +- docs/job-scheduling.md | 6 ++---- ec2/deploy.generic/root/spark-ec2/ec2-variables.sh | 1 - 3 files changed, 3 insertions(+), 6 deletions(-) diff --git a/docs/ec2-scripts.md b/docs/ec2-scripts.md index 8c9a1e1262d8f..7f60f82b966fe 100644 --- a/docs/ec2-scripts.md +++ b/docs/ec2-scripts.md @@ -5,7 +5,7 @@ title: Running Spark on EC2 The `spark-ec2` script, located in Spark's `ec2` directory, allows you to launch, manage and shut down Spark clusters on Amazon EC2. It automatically -sets up Spark, Shark and HDFS on the cluster for you. This guide describes +sets up Spark and HDFS on the cluster for you. This guide describes how to use `spark-ec2` to launch clusters, how to run jobs on them, and how to shut them down. It assumes you've already signed up for an EC2 account on the [Amazon Web Services site](http://aws.amazon.com/). diff --git a/docs/job-scheduling.md b/docs/job-scheduling.md index 5295e351dd711..963e88a3e1d8f 100644 --- a/docs/job-scheduling.md +++ b/docs/job-scheduling.md @@ -14,8 +14,7 @@ runs an independent set of executor processes. The cluster managers that Spark r facilities for [scheduling across applications](#scheduling-across-applications). Second, _within_ each Spark application, multiple "jobs" (Spark actions) may be running concurrently if they were submitted by different threads. This is common if your application is serving requests -over the network; for example, the [Shark](http://shark.cs.berkeley.edu) server works this way. Spark -includes a [fair scheduler](#scheduling-within-an-application) to schedule resources within each SparkContext. +over the network. Spark includes a [fair scheduler](#scheduling-within-an-application) to schedule resources within each SparkContext. # Scheduling Across Applications @@ -52,8 +51,7 @@ an application to gain back cores on one node when it has work to do. To use thi Note that none of the modes currently provide memory sharing across applications. If you would like to share data this way, we recommend running a single server application that can serve multiple requests by querying -the same RDDs. For example, the [Shark](http://shark.cs.berkeley.edu) JDBC server works this way for SQL -queries. In future releases, in-memory storage systems such as [Tachyon](http://tachyon-project.org) will +the same RDDs. In future releases, in-memory storage systems such as [Tachyon](http://tachyon-project.org) will provide another approach to share RDDs. ## Dynamic Resource Allocation diff --git a/ec2/deploy.generic/root/spark-ec2/ec2-variables.sh b/ec2/deploy.generic/root/spark-ec2/ec2-variables.sh index 0857657152ec7..4f3e8da809f7f 100644 --- a/ec2/deploy.generic/root/spark-ec2/ec2-variables.sh +++ b/ec2/deploy.generic/root/spark-ec2/ec2-variables.sh @@ -25,7 +25,6 @@ export MAPRED_LOCAL_DIRS="{{mapred_local_dirs}}" export SPARK_LOCAL_DIRS="{{spark_local_dirs}}" export MODULES="{{modules}}" export SPARK_VERSION="{{spark_version}}" -export SHARK_VERSION="{{shark_version}}" export TACHYON_VERSION="{{tachyon_version}}" export HADOOP_MAJOR_VERSION="{{hadoop_major_version}}" export SWAP_MB="{{swap}}" From 3c4e486b9c8d3f256e801db7c176ab650c976135 Mon Sep 17 00:00:00 2001 From: mcheah Date: Thu, 19 Mar 2015 08:51:49 -0400 Subject: [PATCH 101/122] [SPARK-5843] [API] Allowing map-side combine to be specified in Java. Specifically, when calling JavaPairRDD.combineByKey(), there is a new six-parameter method that exposes the map-side-combine boolean as the fifth parameter and the serializer as the sixth parameter. Author: mcheah Closes #4634 from mccheah/pair-rdd-map-side-combine and squashes the following commits: 5c58319 [mcheah] Fixing compiler errors. 3ce7deb [mcheah] Addressing style and documentation comments. 7455c7a [mcheah] Allowing Java combineByKey to specify Serializer as well. 6ddd729 [mcheah] [SPARK-5843] Allowing map-side combine to be specified in Java. --- .../apache/spark/api/java/JavaPairRDD.scala | 46 ++++++++++++---- .../java/org/apache/spark/JavaAPISuite.java | 53 +++++++++++++++++-- 2 files changed, 87 insertions(+), 12 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/api/java/JavaPairRDD.scala b/core/src/main/scala/org/apache/spark/api/java/JavaPairRDD.scala index 4eadc9a85613e..a023712be1166 100644 --- a/core/src/main/scala/org/apache/spark/api/java/JavaPairRDD.scala +++ b/core/src/main/scala/org/apache/spark/api/java/JavaPairRDD.scala @@ -39,6 +39,7 @@ import org.apache.spark.api.java.function.{Function => JFunction, Function2 => J import org.apache.spark.partial.{BoundedDouble, PartialResult} import org.apache.spark.rdd.{OrderedRDDFunctions, RDD} import org.apache.spark.rdd.RDD.rddToPairRDDFunctions +import org.apache.spark.serializer.Serializer import org.apache.spark.storage.StorageLevel import org.apache.spark.util.Utils @@ -227,24 +228,51 @@ class JavaPairRDD[K, V](val rdd: RDD[(K, V)]) * - `mergeValue`, to merge a V into a C (e.g., adds it to the end of a list) * - `mergeCombiners`, to combine two C's into a single one. * - * In addition, users can control the partitioning of the output RDD, and whether to perform - * map-side aggregation (if a mapper can produce multiple items with the same key). + * In addition, users can control the partitioning of the output RDD, the serializer that is use + * for the shuffle, and whether to perform map-side aggregation (if a mapper can produce multiple + * items with the same key). */ def combineByKey[C](createCombiner: JFunction[V, C], - mergeValue: JFunction2[C, V, C], - mergeCombiners: JFunction2[C, C, C], - partitioner: Partitioner): JavaPairRDD[K, C] = { - implicit val ctag: ClassTag[C] = fakeClassTag + mergeValue: JFunction2[C, V, C], + mergeCombiners: JFunction2[C, C, C], + partitioner: Partitioner, + mapSideCombine: Boolean, + serializer: Serializer): JavaPairRDD[K, C] = { + implicit val ctag: ClassTag[C] = fakeClassTag fromRDD(rdd.combineByKey( createCombiner, mergeValue, mergeCombiners, - partitioner + partitioner, + mapSideCombine, + serializer )) } /** - * Simplified version of combineByKey that hash-partitions the output RDD. + * Generic function to combine the elements for each key using a custom set of aggregation + * functions. Turns a JavaPairRDD[(K, V)] into a result of type JavaPairRDD[(K, C)], for a + * "combined type" C * Note that V and C can be different -- for example, one might group an + * RDD of type (Int, Int) into an RDD of type (Int, List[Int]). Users provide three + * functions: + * + * - `createCombiner`, which turns a V into a C (e.g., creates a one-element list) + * - `mergeValue`, to merge a V into a C (e.g., adds it to the end of a list) + * - `mergeCombiners`, to combine two C's into a single one. + * + * In addition, users can control the partitioning of the output RDD. This method automatically + * uses map-side aggregation in shuffling the RDD. + */ + def combineByKey[C](createCombiner: JFunction[V, C], + mergeValue: JFunction2[C, V, C], + mergeCombiners: JFunction2[C, C, C], + partitioner: Partitioner): JavaPairRDD[K, C] = { + combineByKey(createCombiner, mergeValue, mergeCombiners, partitioner, true, null) + } + + /** + * Simplified version of combineByKey that hash-partitions the output RDD and uses map-side + * aggregation. */ def combineByKey[C](createCombiner: JFunction[V, C], mergeValue: JFunction2[C, V, C], @@ -488,7 +516,7 @@ class JavaPairRDD[K, V](val rdd: RDD[(K, V)]) /** * Simplified version of combineByKey that hash-partitions the resulting RDD using the existing - * partitioner/parallelism level. + * partitioner/parallelism level and using map-side aggregation. */ def combineByKey[C](createCombiner: JFunction[V, C], mergeValue: JFunction2[C, V, C], diff --git a/core/src/test/java/org/apache/spark/JavaAPISuite.java b/core/src/test/java/org/apache/spark/JavaAPISuite.java index 8ec54360ca42a..d4b5bb519157c 100644 --- a/core/src/test/java/org/apache/spark/JavaAPISuite.java +++ b/core/src/test/java/org/apache/spark/JavaAPISuite.java @@ -24,11 +24,12 @@ import java.util.*; import java.util.concurrent.*; -import org.apache.spark.input.PortableDataStream; +import scala.collection.JavaConversions; import scala.Tuple2; import scala.Tuple3; import scala.Tuple4; +import com.google.common.collect.ImmutableMap; import com.google.common.collect.Iterables; import com.google.common.collect.Iterators; import com.google.common.collect.Lists; @@ -51,8 +52,11 @@ import org.apache.spark.api.java.*; import org.apache.spark.api.java.function.*; import org.apache.spark.executor.TaskMetrics; +import org.apache.spark.input.PortableDataStream; import org.apache.spark.partial.BoundedDouble; import org.apache.spark.partial.PartialResult; +import org.apache.spark.rdd.RDD; +import org.apache.spark.serializer.KryoSerializer; import org.apache.spark.storage.StorageLevel; import org.apache.spark.util.StatCounter; @@ -726,8 +730,8 @@ public void javaDoubleRDDHistoGram() { Tuple2 results = rdd.histogram(2); double[] expected_buckets = {1.0, 2.5, 4.0}; long[] expected_counts = {2, 2}; - Assert.assertArrayEquals(expected_buckets, results._1, 0.1); - Assert.assertArrayEquals(expected_counts, results._2); + Assert.assertArrayEquals(expected_buckets, results._1(), 0.1); + Assert.assertArrayEquals(expected_counts, results._2()); // Test with provided buckets long[] histogram = rdd.histogram(expected_buckets); Assert.assertArrayEquals(expected_counts, histogram); @@ -1424,6 +1428,49 @@ public void checkpointAndRestore() { Assert.assertEquals(Arrays.asList(1, 2, 3, 4, 5), recovered.collect()); } + @Test + public void combineByKey() { + JavaRDD originalRDD = sc.parallelize(Arrays.asList(1, 2, 3, 4, 5, 6)); + Function keyFunction = new Function() { + @Override + public Integer call(Integer v1) throws Exception { + return v1 % 3; + } + }; + Function createCombinerFunction = new Function() { + @Override + public Integer call(Integer v1) throws Exception { + return v1; + } + }; + + Function2 mergeValueFunction = new Function2() { + @Override + public Integer call(Integer v1, Integer v2) throws Exception { + return v1 + v2; + } + }; + + JavaPairRDD combinedRDD = originalRDD.keyBy(keyFunction) + .combineByKey(createCombinerFunction, mergeValueFunction, mergeValueFunction); + Map results = combinedRDD.collectAsMap(); + ImmutableMap expected = ImmutableMap.of(0, 9, 1, 5, 2, 7); + Assert.assertEquals(expected, results); + + Partitioner defaultPartitioner = Partitioner.defaultPartitioner( + combinedRDD.rdd(), JavaConversions.asScalaBuffer(Lists.>newArrayList())); + combinedRDD = originalRDD.keyBy(keyFunction) + .combineByKey( + createCombinerFunction, + mergeValueFunction, + mergeValueFunction, + defaultPartitioner, + false, + new KryoSerializer(new SparkConf())); + results = combinedRDD.collectAsMap(); + Assert.assertEquals(expected, results); + } + @SuppressWarnings("unchecked") @Test public void mapOnPairRDD() { From dda4dedca0459fc7c00eb1d9cb07e14af1621e0f Mon Sep 17 00:00:00 2001 From: Yanbo Liang Date: Thu, 19 Mar 2015 11:10:20 -0400 Subject: [PATCH 102/122] [SPARK-6291] [MLLIB] GLM toString & toDebugString GLM toString prints out intercept, numFeatures. For LogisticRegression and SVM model, toString also prints out numClasses, threshold. GLM toDebugString prints out the whole weights, intercept. Author: Yanbo Liang Closes #5038 from yanboliang/spark-6291 and squashes the following commits: 2f578b0 [Yanbo Liang] code format 78b33f2 [Yanbo Liang] fix typos 1e8a023 [Yanbo Liang] GLM toString & toDebugString --- .../spark/mllib/classification/LogisticRegression.scala | 4 ++++ .../scala/org/apache/spark/mllib/classification/SVM.scala | 4 ++++ .../mllib/regression/GeneralizedLinearAlgorithm.scala | 7 ++++++- 3 files changed, 14 insertions(+), 1 deletion(-) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/classification/LogisticRegression.scala b/mllib/src/main/scala/org/apache/spark/mllib/classification/LogisticRegression.scala index b787667b018e6..e7c3599ff619c 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/classification/LogisticRegression.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/classification/LogisticRegression.scala @@ -163,6 +163,10 @@ class LogisticRegressionModel ( } override protected def formatVersion: String = "1.0" + + override def toString: String = { + s"${super.toString}, numClasses = ${numClasses}, threshold = ${threshold.get}" + } } object LogisticRegressionModel extends Loader[LogisticRegressionModel] { diff --git a/mllib/src/main/scala/org/apache/spark/mllib/classification/SVM.scala b/mllib/src/main/scala/org/apache/spark/mllib/classification/SVM.scala index cfc7f868a02f0..52fb62dcff1b4 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/classification/SVM.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/classification/SVM.scala @@ -86,6 +86,10 @@ class SVMModel ( } override protected def formatVersion: String = "1.0" + + override def toString: String = { + s"${super.toString}, numClasses = 2, threshold = ${threshold.get}" + } } object SVMModel extends Loader[SVMModel] { diff --git a/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala b/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala index b262bec904525..45b9ebb4cc0d6 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala @@ -76,7 +76,12 @@ abstract class GeneralizedLinearModel(val weights: Vector, val intercept: Double predictPoint(testData, weights, intercept) } - override def toString() = "(weights=%s, intercept=%s)".format(weights, intercept) + /** + * Print a summary of the model. + */ + override def toString: String = { + s"${this.getClass.getName}: intercept = ${intercept}, numFeatures = ${weights.size}" + } } /** From 8cb23a1f9a3ed08e57865bcb6cc1cc7902881073 Mon Sep 17 00:00:00 2001 From: Brennon York Date: Thu, 19 Mar 2015 11:18:24 -0400 Subject: [PATCH 103/122] [SPARK-5313][Project Infra]: Create simple framework for highlighting changes introduced in a PR Built a simple framework with a `dev/tests` directory to house all pull request related tests. I've moved the two original tests (`pr_merge_ability` and `pr_public_classes`) into the new `dev/tests` directory and tested to the best of my ability. At this point I need to test against Jenkins actually running the new `run-tests-jenkins` script to ensure things aren't broken down the path. Author: Brennon York Closes #5072 from brennonyork/SPARK-5313 and squashes the following commits: 8ae990c [Brennon York] added dev/run-tests back, removed echo 5db4ed4 [Brennon York] removed the git checkout 1b50050 [Brennon York] adding echos to see what jenkins is seeing b823959 [Brennon York] removed run-tests to further test the public_classes pr test 2b9ce12 [Brennon York] added the dev/run-tests call back in ffd49c0 [Brennon York] remove -c from bash as that was removing the trailing args 735d615 [Brennon York] removed the actual dev/run-tests command to further test jenkins d579662 [Brennon York] Merge remote-tracking branch 'upstream/master' into SPARK-5313 aa48029 [Brennon York] removed echo lines for testing jenkins 24cd965 [Brennon York] added test output to check within jenkins to verify 3a38e73 [Brennon York] removed the temporary read 9c881ff [Brennon York] updated test suite 183b7ee [Brennon York] added documentation on how to create tests 0bc2efe [Brennon York] ensure each test starts on the current pr branch 1743378 [Brennon York] added tests in test suite abd7430 [Brennon York] updated to include test suite --- dev/run-tests-jenkins | 75 ++++++++++++++-------------------- dev/tests/pr_merge_ability.sh | 39 ++++++++++++++++++ dev/tests/pr_public_classes.sh | 65 +++++++++++++++++++++++++++++ 3 files changed, 135 insertions(+), 44 deletions(-) create mode 100755 dev/tests/pr_merge_ability.sh create mode 100755 dev/tests/pr_public_classes.sh diff --git a/dev/run-tests-jenkins b/dev/run-tests-jenkins index 6a849e4f77207..5f4000e83925c 100755 --- a/dev/run-tests-jenkins +++ b/dev/run-tests-jenkins @@ -49,6 +49,21 @@ SHORT_COMMIT_HASH="${ghprbActualCommit:0:7}" TESTS_TIMEOUT="120m" # format: http://linux.die.net/man/1/timeout +# Array to capture all tests to run on the pull request. These tests are held under the +#+ dev/tests/ directory. +# +# To write a PR test: +#+ * the file must reside within the dev/tests directory +#+ * be an executable bash script +#+ * accept two arguments on the command line, the first being the Github PR long commit +#+ hash and the second the Github SHA1 hash +#+ * and, lastly, return string output to be included in the pr message output that will +#+ be posted to Github +PR_TESTS=( + "pr_merge_ability" + "pr_public_classes" +) + function post_message () { local message=$1 local data="{\"body\": \"$message\"}" @@ -131,48 +146,22 @@ function send_archived_logs () { fi } - -# We diff master...$ghprbActualCommit because that gets us changes introduced in the PR -#+ and not anything else added to master since the PR was branched. - -# check PR merge-ability and check for new public classes -{ - if [ "$sha1" == "$ghprbActualCommit" ]; then - merge_note=" * This patch **does not merge cleanly**." - else - merge_note=" * This patch merges cleanly." +# Environment variable to capture PR test output +pr_message="" + +# Run pull request tests +for t in "${PR_TESTS[@]}"; do + this_test="${FWDIR}/dev/tests/${t}.sh" + # Ensure the test is a file and is executable + if [ -x "$this_test" ]; then + echo "ghprb: $ghprbActualCommit sha1: $sha1" + this_mssg="`bash \"${this_test}\" \"${ghprbActualCommit}\" \"${sha1}\" 2>/dev/null`" + # Check if this is the merge test as we submit that note *before* and *after* + # the tests run + [ "$t" == "pr_merge_ability" ] && merge_note="${this_mssg}" + pr_message="${pr_message}\n${this_mssg}" fi - - source_files=$( - git diff master...$ghprbActualCommit --name-only `# diff patch against master from branch point` \ - | grep -v -e "\/test" `# ignore files in test directories` \ - | grep -e "\.py$" -e "\.java$" -e "\.scala$" `# include only code files` \ - | tr "\n" " " - ) - new_public_classes=$( - git diff master...$ghprbActualCommit ${source_files} `# diff patch against master from branch point` \ - | grep "^\+" `# filter in only added lines` \ - | sed -r -e "s/^\+//g" `# remove the leading +` \ - | grep -e "trait " -e "class " `# filter in lines with these key words` \ - | grep -e "{" -e "(" `# filter in lines with these key words, too` \ - | grep -v -e "\@\@" -e "private" `# exclude lines with these words` \ - | grep -v -e "^// " -e "^/\*" -e "^ \* " `# exclude comment lines` \ - | sed -r -e "s/\{.*//g" `# remove from the { onwards` \ - | sed -r -e "s/\}//g" `# just in case, remove }; they mess the JSON` \ - | sed -r -e "s/\"/\\\\\"/g" `# escape double quotes; they mess the JSON` \ - | sed -r -e "s/^(.*)$/\`\1\`/g" `# surround with backticks for style` \ - | sed -r -e "s/^/ \* /g" `# prepend ' *' to start of line` \ - | sed -r -e "s/$/\\\n/g" `# append newline to end of line` \ - | tr -d "\n" `# remove actual LF characters` - ) - - if [ -z "$new_public_classes" ]; then - public_classes_note=" * This patch adds no public classes." - else - public_classes_note=" * This patch adds the following public classes _(experimental)_:" - public_classes_note="${public_classes_note}\n${new_public_classes}" - fi -} +done # post start message { @@ -181,7 +170,6 @@ function send_archived_logs () { PR $ghprbPullId at commit [\`${SHORT_COMMIT_HASH}\`](${COMMIT_URL})." start_message="${start_message}\n${merge_note}" - # start_message="${start_message}\n${public_classes_note}" post_message "$start_message" } @@ -234,8 +222,7 @@ function send_archived_logs () { PR $ghprbPullId at commit [\`${SHORT_COMMIT_HASH}\`](${COMMIT_URL})." result_message="${result_message}\n${test_result_note}" - result_message="${result_message}\n${merge_note}" - result_message="${result_message}\n${public_classes_note}" + result_message="${result_message}\n${pr_message}" post_message "$result_message" } diff --git a/dev/tests/pr_merge_ability.sh b/dev/tests/pr_merge_ability.sh new file mode 100755 index 0000000000000..d9a347fe24a8c --- /dev/null +++ b/dev/tests/pr_merge_ability.sh @@ -0,0 +1,39 @@ +#!/usr/bin/env bash + +# +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +# +# This script follows the base format for testing pull requests against +# another branch and returning results to be published. More details can be +# found at dev/run-tests-jenkins. +# +# Arg1: The Github Pull Request Actual Commit +#+ known as `ghprbActualCommit` in `run-tests-jenkins` +# Arg2: The SHA1 hash +#+ known as `sha1` in `run-tests-jenkins` +# + +ghprbActualCommit="$1" +sha1="$2" + +# check PR merge-ability +if [ "${sha1}" == "${ghprbActualCommit}" ]; then + echo " * This patch **does not merge cleanly**." +else + echo " * This patch merges cleanly." +fi diff --git a/dev/tests/pr_public_classes.sh b/dev/tests/pr_public_classes.sh new file mode 100755 index 0000000000000..927295b88c963 --- /dev/null +++ b/dev/tests/pr_public_classes.sh @@ -0,0 +1,65 @@ +#!/usr/bin/env bash + +# +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +# +# This script follows the base format for testing pull requests against +# another branch and returning results to be published. More details can be +# found at dev/run-tests-jenkins. +# +# Arg1: The Github Pull Request Actual Commit +#+ known as `ghprbActualCommit` in `run-tests-jenkins` +# Arg2: The SHA1 hash +#+ known as `sha1` in `run-tests-jenkins` +# + +# We diff master...$ghprbActualCommit because that gets us changes introduced in the PR +#+ and not anything else added to master since the PR was branched. + +ghprbActualCommit="$1" +sha1="$2" + +source_files=$( + git diff master...$ghprbActualCommit --name-only `# diff patch against master from branch point` \ + | grep -v -e "\/test" `# ignore files in test directories` \ + | grep -e "\.py$" -e "\.java$" -e "\.scala$" `# include only code files` \ + | tr "\n" " " +) +new_public_classes=$( + git diff master...$ghprbActualCommit ${source_files} `# diff patch against master from branch point` \ + | grep "^\+" `# filter in only added lines` \ + | sed -r -e "s/^\+//g" `# remove the leading +` \ + | grep -e "trait " -e "class " `# filter in lines with these key words` \ + | grep -e "{" -e "(" `# filter in lines with these key words, too` \ + | grep -v -e "\@\@" -e "private" `# exclude lines with these words` \ + | grep -v -e "^// " -e "^/\*" -e "^ \* " `# exclude comment lines` \ + | sed -r -e "s/\{.*//g" `# remove from the { onwards` \ + | sed -r -e "s/\}//g" `# just in case, remove }; they mess the JSON` \ + | sed -r -e "s/\"/\\\\\"/g" `# escape double quotes; they mess the JSON` \ + | sed -r -e "s/^(.*)$/\`\1\`/g" `# surround with backticks for style` \ + | sed -r -e "s/^/ \* /g" `# prepend ' *' to start of line` \ + | sed -r -e "s/$/\\\n/g" `# append newline to end of line` \ + | tr -d "\n" `# remove actual LF characters` +) + +if [ -z "$new_public_classes" ]; then + echo " * This patch adds no public classes." +else + public_classes_note=" * This patch adds the following public classes _(experimental)_:" + echo "${public_classes_note}\n${new_public_classes}" +fi From 3b5aaa6a5fe0d838a8570c9d500ebca5f63764f8 Mon Sep 17 00:00:00 2001 From: Wenchen Fan Date: Thu, 19 Mar 2015 15:25:32 -0400 Subject: [PATCH 104/122] [Core][minor] remove unused `visitedStages` in `DAGScheduler.stageDependsOn` We define and update `visitedStages` in `DAGScheduler.stageDependsOn`, but never read it. So we can safely remove it. Author: Wenchen Fan Closes #5086 from cloud-fan/minor and squashes the following commits: 24663ea [Wenchen Fan] remove un-used variable --- .../main/scala/org/apache/spark/scheduler/DAGScheduler.scala | 2 -- 1 file changed, 2 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala b/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala index 1021172e6afb4..8feac6cb6b7a1 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala @@ -1262,7 +1262,6 @@ class DAGScheduler( return true } val visitedRdds = new HashSet[RDD[_]] - val visitedStages = new HashSet[Stage] // We are manually maintaining a stack here to prevent StackOverflowError // caused by recursively visiting val waitingForVisit = new Stack[RDD[_]] @@ -1274,7 +1273,6 @@ class DAGScheduler( case shufDep: ShuffleDependency[_, _, _] => val mapStage = getShuffleMapStage(shufDep, stage.jobId) if (!mapStage.isAvailable) { - visitedStages += mapStage waitingForVisit.push(mapStage.rdd) } // Otherwise there's no need to follow the dependency back case narrowDep: NarrowDependency[_] => From f17d43b033d928dbc46aef8e367aa08902e698ad Mon Sep 17 00:00:00 2001 From: Nicholas Chammas Date: Thu, 19 Mar 2015 12:46:10 -0700 Subject: [PATCH 105/122] [SPARK-6219] [Build] Check that Python code compiles This PR expands the Python lint checks so that they check for obvious compilation errors in our Python code. For example: ``` $ ./dev/lint-python Python lint checks failed. Compiling ./ec2/spark_ec2.py ... File "./ec2/spark_ec2.py", line 618 return (master_nodes,, slave_nodes) ^ SyntaxError: invalid syntax ./ec2/spark_ec2.py:618:25: E231 missing whitespace after ',' ./ec2/spark_ec2.py:1117:101: E501 line too long (102 > 100 characters) ``` This PR also bumps up the version of `pep8`. It ignores new types of checks introduced by that version bump while fixing problems missed by the older version of `pep8` we were using. Author: Nicholas Chammas Closes #4941 from nchammas/compile-spark-ec2 and squashes the following commits: 75e31d8 [Nicholas Chammas] upgrade pep8 + check compile b33651c [Nicholas Chammas] PEP8 line length --- dev/lint-python | 44 +++++++++++++++++++++++++++----------------- ec2/spark_ec2.py | 4 ++-- 2 files changed, 29 insertions(+), 19 deletions(-) diff --git a/dev/lint-python b/dev/lint-python index 772f856154ae0..fded654893a7c 100755 --- a/dev/lint-python +++ b/dev/lint-python @@ -19,43 +19,53 @@ SCRIPT_DIR="$( cd "$( dirname "$0" )" && pwd )" SPARK_ROOT_DIR="$(dirname "$SCRIPT_DIR")" -PEP8_REPORT_PATH="$SPARK_ROOT_DIR/dev/pep8-report.txt" +PATHS_TO_CHECK="./python/pyspark/ ./ec2/spark_ec2.py ./examples/src/main/python/" +PYTHON_LINT_REPORT_PATH="$SPARK_ROOT_DIR/dev/python-lint-report.txt" cd "$SPARK_ROOT_DIR" +# compileall: https://docs.python.org/2/library/compileall.html +python -B -m compileall -q -l $PATHS_TO_CHECK > "$PYTHON_LINT_REPORT_PATH" +compile_status="${PIPESTATUS[0]}" + # Get pep8 at runtime so that we don't rely on it being installed on the build server. #+ See: https://github.com/apache/spark/pull/1744#issuecomment-50982162 #+ TODOs: -#+ - Dynamically determine latest release version of pep8 and use that. -#+ - Download this from a more reliable source. (GitHub raw can be flaky, apparently. (?)) +#+ - Download pep8 from PyPI. It's more "official". PEP8_SCRIPT_PATH="$SPARK_ROOT_DIR/dev/pep8.py" -PEP8_SCRIPT_REMOTE_PATH="https://raw.githubusercontent.com/jcrocholl/pep8/1.5.7/pep8.py" -PEP8_PATHS_TO_CHECK="./python/pyspark/ ./ec2/spark_ec2.py ./examples/src/main/python/" +PEP8_SCRIPT_REMOTE_PATH="https://raw.githubusercontent.com/jcrocholl/pep8/1.6.2/pep8.py" +# if [ ! -e "$PEP8_SCRIPT_PATH" ]; then curl --silent -o "$PEP8_SCRIPT_PATH" "$PEP8_SCRIPT_REMOTE_PATH" -curl_status=$? +curl_status="$?" -if [ $curl_status -ne 0 ]; then +if [ "$curl_status" -ne 0 ]; then echo "Failed to download pep8.py from \"$PEP8_SCRIPT_REMOTE_PATH\"." - exit $curl_status + exit "$curl_status" fi - +# fi # There is no need to write this output to a file #+ first, but we do so so that the check status can #+ be output before the report, like with the #+ scalastyle and RAT checks. -python "$PEP8_SCRIPT_PATH" $PEP8_PATHS_TO_CHECK > "$PEP8_REPORT_PATH" -pep8_status=${PIPESTATUS[0]} #$? +python "$PEP8_SCRIPT_PATH" --ignore=E402,E731,E241,W503,E226 $PATHS_TO_CHECK >> "$PYTHON_LINT_REPORT_PATH" +pep8_status="${PIPESTATUS[0]}" + +if [ "$compile_status" -eq 0 -a "$pep8_status" -eq 0 ]; then + lint_status=0 +else + lint_status=1 +fi -if [ $pep8_status -ne 0 ]; then - echo "PEP 8 checks failed." - cat "$PEP8_REPORT_PATH" +if [ "$lint_status" -ne 0 ]; then + echo "Python lint checks failed." + cat "$PYTHON_LINT_REPORT_PATH" else - echo "PEP 8 checks passed." + echo "Python lint checks passed." fi -rm "$PEP8_REPORT_PATH" rm "$PEP8_SCRIPT_PATH" +rm "$PYTHON_LINT_REPORT_PATH" -exit $pep8_status +exit "$lint_status" diff --git a/ec2/spark_ec2.py b/ec2/spark_ec2.py index f848874b0c775..c467cd08ed742 100755 --- a/ec2/spark_ec2.py +++ b/ec2/spark_ec2.py @@ -1159,8 +1159,8 @@ def real_main(): if EC2_INSTANCE_TYPES[opts.instance_type] != \ EC2_INSTANCE_TYPES[opts.master_instance_type]: print >> stderr, \ - "Error: spark-ec2 currently does not support having a master and slaves with " + \ - "different AMI virtualization types." + "Error: spark-ec2 currently does not support having a master and slaves " + \ + "with different AMI virtualization types." print >> stderr, "master instance virtualization type: {t}".format( t=EC2_INSTANCE_TYPES[opts.master_instance_type]) print >> stderr, "slave instance virtualization type: {t}".format( From 0745a305fac622a6eeb8aa4a7401205a14252939 Mon Sep 17 00:00:00 2001 From: Reynold Xin Date: Thu, 19 Mar 2015 22:12:01 -0400 Subject: [PATCH 106/122] Tighten up field/method visibility in Executor and made some code more clear to read. I was reading Executor just now and found that some latest changes introduced some weird code path with too much monadic chaining and unnecessary fields. I cleaned it up a bit, and also tightened up the visibility of various fields/methods. Also added some inline documentation to help understand this code better. Author: Reynold Xin Closes #4850 from rxin/executor and squashes the following commits: 866fc60 [Reynold Xin] Code review feedback. 020efbb [Reynold Xin] Tighten up field/method visibility in Executor and made some code more clear to read. --- .../org/apache/spark/TaskEndReason.scala | 6 +- .../executor/CommitDeniedException.scala | 6 +- .../org/apache/spark/executor/Executor.scala | 196 ++++++++++-------- .../spark/executor/ExecutorSource.scala | 16 +- .../org/apache/spark/scheduler/Task.scala | 2 +- 5 files changed, 120 insertions(+), 106 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/TaskEndReason.scala b/core/src/main/scala/org/apache/spark/TaskEndReason.scala index 29a5cd5fdac76..48fd3e7e23d52 100644 --- a/core/src/main/scala/org/apache/spark/TaskEndReason.scala +++ b/core/src/main/scala/org/apache/spark/TaskEndReason.scala @@ -151,11 +151,7 @@ case object TaskKilled extends TaskFailedReason { * Task requested the driver to commit, but was denied. */ @DeveloperApi -case class TaskCommitDenied( - jobID: Int, - partitionID: Int, - attemptID: Int) - extends TaskFailedReason { +case class TaskCommitDenied(jobID: Int, partitionID: Int, attemptID: Int) extends TaskFailedReason { override def toErrorString: String = s"TaskCommitDenied (Driver denied task commit)" + s" for job: $jobID, partition: $partitionID, attempt: $attemptID" } diff --git a/core/src/main/scala/org/apache/spark/executor/CommitDeniedException.scala b/core/src/main/scala/org/apache/spark/executor/CommitDeniedException.scala index f7604a321f007..f47d7ef511da1 100644 --- a/core/src/main/scala/org/apache/spark/executor/CommitDeniedException.scala +++ b/core/src/main/scala/org/apache/spark/executor/CommitDeniedException.scala @@ -22,14 +22,12 @@ import org.apache.spark.{TaskCommitDenied, TaskEndReason} /** * Exception thrown when a task attempts to commit output to HDFS but is denied by the driver. */ -class CommitDeniedException( +private[spark] class CommitDeniedException( msg: String, jobID: Int, splitID: Int, attemptID: Int) extends Exception(msg) { - def toTaskEndReason: TaskEndReason = new TaskCommitDenied(jobID, splitID, attemptID) - + def toTaskEndReason: TaskEndReason = TaskCommitDenied(jobID, splitID, attemptID) } - diff --git a/core/src/main/scala/org/apache/spark/executor/Executor.scala b/core/src/main/scala/org/apache/spark/executor/Executor.scala index 6196f7b165049..bf3135ef081c1 100644 --- a/core/src/main/scala/org/apache/spark/executor/Executor.scala +++ b/core/src/main/scala/org/apache/spark/executor/Executor.scala @@ -21,7 +21,7 @@ import java.io.File import java.lang.management.ManagementFactory import java.net.URL import java.nio.ByteBuffer -import java.util.concurrent._ +import java.util.concurrent.ConcurrentHashMap import scala.collection.JavaConversions._ import scala.collection.mutable.{ArrayBuffer, HashMap} @@ -31,15 +31,17 @@ import akka.actor.Props import org.apache.spark._ import org.apache.spark.deploy.SparkHadoopUtil -import org.apache.spark.scheduler._ +import org.apache.spark.scheduler.{DirectTaskResult, IndirectTaskResult, Task} import org.apache.spark.shuffle.FetchFailedException import org.apache.spark.storage.{StorageLevel, TaskResultBlockId} -import org.apache.spark.util.{ChildFirstURLClassLoader, MutableURLClassLoader, - SparkUncaughtExceptionHandler, AkkaUtils, Utils} +import org.apache.spark.util._ /** - * Spark executor used with Mesos, YARN, and the standalone scheduler. - * In coarse-grained mode, an existing actor system is provided. + * Spark executor, backed by a threadpool to run tasks. + * + * This can be used with Mesos, YARN, and the standalone scheduler. + * An internal RPC interface (at the moment Akka) is used for communication with the driver, + * except in the case of Mesos fine-grained mode. */ private[spark] class Executor( executorId: String, @@ -47,8 +49,8 @@ private[spark] class Executor( env: SparkEnv, userClassPath: Seq[URL] = Nil, isLocal: Boolean = false) - extends Logging -{ + extends Logging { + logInfo(s"Starting executor ID $executorId on host $executorHostname") // Application dependencies (added through SparkContext) that we've fetched so far on this node. @@ -78,9 +80,8 @@ private[spark] class Executor( } // Start worker thread pool - val threadPool = Utils.newDaemonCachedThreadPool("Executor task launch worker") - - val executorSource = new ExecutorSource(this, executorId) + private val threadPool = Utils.newDaemonCachedThreadPool("Executor task launch worker") + private val executorSource = new ExecutorSource(threadPool, executorId) if (!isLocal) { env.metricsSystem.registerSource(executorSource) @@ -122,21 +123,21 @@ private[spark] class Executor( taskId: Long, attemptNumber: Int, taskName: String, - serializedTask: ByteBuffer) { + serializedTask: ByteBuffer): Unit = { val tr = new TaskRunner(context, taskId = taskId, attemptNumber = attemptNumber, taskName, serializedTask) runningTasks.put(taskId, tr) threadPool.execute(tr) } - def killTask(taskId: Long, interruptThread: Boolean) { + def killTask(taskId: Long, interruptThread: Boolean): Unit = { val tr = runningTasks.get(taskId) if (tr != null) { tr.kill(interruptThread) } } - def stop() { + def stop(): Unit = { env.metricsSystem.report() env.actorSystem.stop(executorActor) isStopped = true @@ -146,7 +147,10 @@ private[spark] class Executor( } } - private def gcTime = ManagementFactory.getGarbageCollectorMXBeans.map(_.getCollectionTime).sum + /** Returns the total amount of time this JVM process has spent in garbage collection. */ + private def computeTotalGcTime(): Long = { + ManagementFactory.getGarbageCollectorMXBeans.map(_.getCollectionTime).sum + } class TaskRunner( execBackend: ExecutorBackend, @@ -156,12 +160,19 @@ private[spark] class Executor( serializedTask: ByteBuffer) extends Runnable { + /** Whether this task has been killed. */ @volatile private var killed = false - @volatile var task: Task[Any] = _ - @volatile var attemptedTask: Option[Task[Any]] = None + + /** How much the JVM process has spent in GC when the task starts to run. */ @volatile var startGCTime: Long = _ - def kill(interruptThread: Boolean) { + /** + * The task to run. This will be set in run() by deserializing the task binary coming + * from the driver. Once it is set, it will never be changed. + */ + @volatile var task: Task[Any] = _ + + def kill(interruptThread: Boolean): Unit = { logInfo(s"Executor is trying to kill $taskName (TID $taskId)") killed = true if (task != null) { @@ -169,14 +180,14 @@ private[spark] class Executor( } } - override def run() { + override def run(): Unit = { val deserializeStartTime = System.currentTimeMillis() Thread.currentThread.setContextClassLoader(replClassLoader) val ser = env.closureSerializer.newInstance() logInfo(s"Running $taskName (TID $taskId)") execBackend.statusUpdate(taskId, TaskState.RUNNING, EMPTY_BYTE_BUFFER) var taskStart: Long = 0 - startGCTime = gcTime + startGCTime = computeTotalGcTime() try { val (taskFiles, taskJars, taskBytes) = Task.deserializeWithDependencies(serializedTask) @@ -193,7 +204,6 @@ private[spark] class Executor( throw new TaskKilledException } - attemptedTask = Some(task) logDebug("Task " + taskId + "'s epoch is " + task.epoch) env.mapOutputTracker.updateEpoch(task.epoch) @@ -215,18 +225,17 @@ private[spark] class Executor( for (m <- task.metrics) { m.setExecutorDeserializeTime(taskStart - deserializeStartTime) m.setExecutorRunTime(taskFinish - taskStart) - m.setJvmGCTime(gcTime - startGCTime) + m.setJvmGCTime(computeTotalGcTime() - startGCTime) m.setResultSerializationTime(afterSerialization - beforeSerialization) } val accumUpdates = Accumulators.values - val directResult = new DirectTaskResult(valueBytes, accumUpdates, task.metrics.orNull) val serializedDirectResult = ser.serialize(directResult) val resultSize = serializedDirectResult.limit // directSend = sending directly back to the driver - val serializedResult = { + val serializedResult: ByteBuffer = { if (maxResultSize > 0 && resultSize > maxResultSize) { logWarning(s"Finished $taskName (TID $taskId). Result is larger than maxResultSize " + s"(${Utils.bytesToString(resultSize)} > ${Utils.bytesToString(maxResultSize)}), " + @@ -248,42 +257,40 @@ private[spark] class Executor( execBackend.statusUpdate(taskId, TaskState.FINISHED, serializedResult) } catch { - case ffe: FetchFailedException => { + case ffe: FetchFailedException => val reason = ffe.toTaskEndReason execBackend.statusUpdate(taskId, TaskState.FAILED, ser.serialize(reason)) - } - case _: TaskKilledException | _: InterruptedException if task.killed => { + case _: TaskKilledException | _: InterruptedException if task.killed => logInfo(s"Executor killed $taskName (TID $taskId)") execBackend.statusUpdate(taskId, TaskState.KILLED, ser.serialize(TaskKilled)) - } - case cDE: CommitDeniedException => { + case cDE: CommitDeniedException => val reason = cDE.toTaskEndReason execBackend.statusUpdate(taskId, TaskState.FAILED, ser.serialize(reason)) - } - case t: Throwable => { + case t: Throwable => // Attempt to exit cleanly by informing the driver of our failure. // If anything goes wrong (or this was a fatal exception), we will delegate to // the default uncaught exception handler, which will terminate the Executor. logError(s"Exception in $taskName (TID $taskId)", t) - val serviceTime = System.currentTimeMillis() - taskStart - val metrics = attemptedTask.flatMap(t => t.metrics) - for (m <- metrics) { - m.setExecutorRunTime(serviceTime) - m.setJvmGCTime(gcTime - startGCTime) + val metrics: Option[TaskMetrics] = Option(task).flatMap { task => + task.metrics.map { m => + m.setExecutorRunTime(System.currentTimeMillis() - taskStart) + m.setJvmGCTime(computeTotalGcTime() - startGCTime) + m + } } - val reason = new ExceptionFailure(t, metrics) - execBackend.statusUpdate(taskId, TaskState.FAILED, ser.serialize(reason)) + val taskEndReason = new ExceptionFailure(t, metrics) + execBackend.statusUpdate(taskId, TaskState.FAILED, ser.serialize(taskEndReason)) // Don't forcibly exit unless the exception was inherently fatal, to avoid // stopping other tasks unnecessarily. if (Utils.isFatalError(t)) { SparkUncaughtExceptionHandler.uncaughtException(t) } - } + } finally { // Release memory used by this thread for shuffles env.shuffleMemoryManager.releaseMemoryForThisThread() @@ -358,7 +365,7 @@ private[spark] class Executor( for ((name, timestamp) <- newFiles if currentFiles.getOrElse(name, -1L) < timestamp) { logInfo("Fetching " + name + " with timestamp " + timestamp) // Fetch file with useCache mode, close cache for local mode. - Utils.fetchFile(name, new File(SparkFiles.getRootDirectory), conf, + Utils.fetchFile(name, new File(SparkFiles.getRootDirectory()), conf, env.securityManager, hadoopConf, timestamp, useCache = !isLocal) currentFiles(name) = timestamp } @@ -370,12 +377,12 @@ private[spark] class Executor( if (currentTimeStamp < timestamp) { logInfo("Fetching " + name + " with timestamp " + timestamp) // Fetch file with useCache mode, close cache for local mode. - Utils.fetchFile(name, new File(SparkFiles.getRootDirectory), conf, + Utils.fetchFile(name, new File(SparkFiles.getRootDirectory()), conf, env.securityManager, hadoopConf, timestamp, useCache = !isLocal) currentJars(name) = timestamp // Add it to our class loader - val url = new File(SparkFiles.getRootDirectory, localName).toURI.toURL - if (!urlClassLoader.getURLs.contains(url)) { + val url = new File(SparkFiles.getRootDirectory(), localName).toURI.toURL + if (!urlClassLoader.getURLs().contains(url)) { logInfo("Adding " + url + " to class loader") urlClassLoader.addURL(url) } @@ -384,61 +391,70 @@ private[spark] class Executor( } } - def startDriverHeartbeater() { - val interval = conf.getInt("spark.executor.heartbeatInterval", 10000) - val timeout = AkkaUtils.lookupTimeout(conf) - val retryAttempts = AkkaUtils.numRetries(conf) - val retryIntervalMs = AkkaUtils.retryWaitMs(conf) - val heartbeatReceiverRef = AkkaUtils.makeDriverRef("HeartbeatReceiver", conf, env.actorSystem) + private val timeout = AkkaUtils.lookupTimeout(conf) + private val retryAttempts = AkkaUtils.numRetries(conf) + private val retryIntervalMs = AkkaUtils.retryWaitMs(conf) + private val heartbeatReceiverRef = + AkkaUtils.makeDriverRef("HeartbeatReceiver", conf, env.actorSystem) + + /** Reports heartbeat and metrics for active tasks to the driver. */ + private def reportHeartBeat(): Unit = { + // list of (task id, metrics) to send back to the driver + val tasksMetrics = new ArrayBuffer[(Long, TaskMetrics)]() + val curGCTime = computeTotalGcTime() + + for (taskRunner <- runningTasks.values()) { + if (taskRunner.task != null) { + taskRunner.task.metrics.foreach { metrics => + metrics.updateShuffleReadMetrics() + metrics.updateInputMetrics() + metrics.setJvmGCTime(curGCTime - taskRunner.startGCTime) + + if (isLocal) { + // JobProgressListener will hold an reference of it during + // onExecutorMetricsUpdate(), then JobProgressListener can not see + // the changes of metrics any more, so make a deep copy of it + val copiedMetrics = Utils.deserialize[TaskMetrics](Utils.serialize(metrics)) + tasksMetrics += ((taskRunner.taskId, copiedMetrics)) + } else { + // It will be copied by serialization + tasksMetrics += ((taskRunner.taskId, metrics)) + } + } + } + } - val t = new Thread() { + val message = Heartbeat(executorId, tasksMetrics.toArray, env.blockManager.blockManagerId) + try { + val response = AkkaUtils.askWithReply[HeartbeatResponse](message, heartbeatReceiverRef, + retryAttempts, retryIntervalMs, timeout) + if (response.reregisterBlockManager) { + logWarning("Told to re-register on heartbeat") + env.blockManager.reregister() + } + } catch { + case NonFatal(e) => logWarning("Issue communicating with driver in heartbeater", e) + } + } + + /** + * Starts a thread to report heartbeat and partial metrics for active tasks to driver. + * This thread stops running when the executor is stopped. + */ + private def startDriverHeartbeater(): Unit = { + val interval = conf.getInt("spark.executor.heartbeatInterval", 10000) + val thread = new Thread() { override def run() { // Sleep a random interval so the heartbeats don't end up in sync Thread.sleep(interval + (math.random * interval).asInstanceOf[Int]) - while (!isStopped) { - val tasksMetrics = new ArrayBuffer[(Long, TaskMetrics)]() - val curGCTime = gcTime - - for (taskRunner <- runningTasks.values()) { - if (taskRunner.attemptedTask.nonEmpty) { - Option(taskRunner.task).flatMap(_.metrics).foreach { metrics => - metrics.updateShuffleReadMetrics() - metrics.updateInputMetrics() - metrics.setJvmGCTime(curGCTime - taskRunner.startGCTime) - - if (isLocal) { - // JobProgressListener will hold an reference of it during - // onExecutorMetricsUpdate(), then JobProgressListener can not see - // the changes of metrics any more, so make a deep copy of it - val copiedMetrics = Utils.deserialize[TaskMetrics](Utils.serialize(metrics)) - tasksMetrics += ((taskRunner.taskId, copiedMetrics)) - } else { - // It will be copied by serialization - tasksMetrics += ((taskRunner.taskId, metrics)) - } - } - } - } - - val message = Heartbeat(executorId, tasksMetrics.toArray, env.blockManager.blockManagerId) - try { - val response = AkkaUtils.askWithReply[HeartbeatResponse](message, heartbeatReceiverRef, - retryAttempts, retryIntervalMs, timeout) - if (response.reregisterBlockManager) { - logWarning("Told to re-register on heartbeat") - env.blockManager.reregister() - } - } catch { - case NonFatal(t) => logWarning("Issue communicating with driver in heartbeater", t) - } - + reportHeartBeat() Thread.sleep(interval) } } } - t.setDaemon(true) - t.setName("Driver Heartbeater") - t.start() + thread.setDaemon(true) + thread.setName("driver-heartbeater") + thread.start() } } diff --git a/core/src/main/scala/org/apache/spark/executor/ExecutorSource.scala b/core/src/main/scala/org/apache/spark/executor/ExecutorSource.scala index c4d73622c4727..293c512f8b70c 100644 --- a/core/src/main/scala/org/apache/spark/executor/ExecutorSource.scala +++ b/core/src/main/scala/org/apache/spark/executor/ExecutorSource.scala @@ -17,6 +17,8 @@ package org.apache.spark.executor +import java.util.concurrent.ThreadPoolExecutor + import scala.collection.JavaConversions._ import com.codahale.metrics.{Gauge, MetricRegistry} @@ -24,9 +26,11 @@ import org.apache.hadoop.fs.FileSystem import org.apache.spark.metrics.source.Source -private[spark] class ExecutorSource(val executor: Executor, executorId: String) extends Source { +private[spark] +class ExecutorSource(threadPool: ThreadPoolExecutor, executorId: String) extends Source { + private def fileStats(scheme: String) : Option[FileSystem.Statistics] = - FileSystem.getAllStatistics().filter(s => s.getScheme.equals(scheme)).headOption + FileSystem.getAllStatistics().find(s => s.getScheme.equals(scheme)) private def registerFileSystemStat[T]( scheme: String, name: String, f: FileSystem.Statistics => T, defaultValue: T) = { @@ -41,23 +45,23 @@ private[spark] class ExecutorSource(val executor: Executor, executorId: String) // Gauge for executor thread pool's actively executing task counts metricRegistry.register(MetricRegistry.name("threadpool", "activeTasks"), new Gauge[Int] { - override def getValue: Int = executor.threadPool.getActiveCount() + override def getValue: Int = threadPool.getActiveCount() }) // Gauge for executor thread pool's approximate total number of tasks that have been completed metricRegistry.register(MetricRegistry.name("threadpool", "completeTasks"), new Gauge[Long] { - override def getValue: Long = executor.threadPool.getCompletedTaskCount() + override def getValue: Long = threadPool.getCompletedTaskCount() }) // Gauge for executor thread pool's current number of threads metricRegistry.register(MetricRegistry.name("threadpool", "currentPool_size"), new Gauge[Int] { - override def getValue: Int = executor.threadPool.getPoolSize() + override def getValue: Int = threadPool.getPoolSize() }) // Gauge got executor thread pool's largest number of threads that have ever simultaneously // been in th pool metricRegistry.register(MetricRegistry.name("threadpool", "maxPool_size"), new Gauge[Int] { - override def getValue: Int = executor.threadPool.getMaximumPoolSize() + override def getValue: Int = threadPool.getMaximumPoolSize() }) // Gauge for file system stats of this executor diff --git a/core/src/main/scala/org/apache/spark/scheduler/Task.scala b/core/src/main/scala/org/apache/spark/scheduler/Task.scala index 847a4912eec13..4d9f940813b8e 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/Task.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/Task.scala @@ -45,7 +45,7 @@ import org.apache.spark.util.Utils private[spark] abstract class Task[T](val stageId: Int, var partitionId: Int) extends Serializable { /** - * Called by Executor to run this task. + * Called by [[Executor]] to run this task. * * @param taskAttemptId an identifier for this task attempt that is unique within a SparkContext. * @param attemptNumber how many times this task has been attempted (0 for the first attempt) From 116c553fd6f6d2adcbbf000cd80b5c46d4516e87 Mon Sep 17 00:00:00 2001 From: Jongyoul Lee Date: Fri, 20 Mar 2015 12:24:34 +0000 Subject: [PATCH 107/122] [SPARK-6286][Mesos][minor] Handle missing Mesos case TASK_ERROR - Made TaskState.isFailed for handling TASK_LOST and TASK_ERROR and synchronizing CoarseMesosSchedulerBackend and MesosSchedulerBackend - This is related #5000 Author: Jongyoul Lee Closes #5088 from jongyoul/SPARK-6286-1 and squashes the following commits: 4f2362f [Jongyoul Lee] [SPARK-6286][Mesos][minor] Handle missing Mesos case TASK_ERROR - Fixed scalastyle ac4336a [Jongyoul Lee] [SPARK-6286][Mesos][minor] Handle missing Mesos case TASK_ERROR - Made TaskState.isFailed for handling TASK_LOST and TASK_ERROR and synchronizing CoarseMesosSchedulerBackend and MesosSchedulerBackend --- core/src/main/scala/org/apache/spark/TaskState.scala | 2 ++ .../scheduler/cluster/mesos/CoarseMesosSchedulerBackend.scala | 2 +- .../spark/scheduler/cluster/mesos/MesosSchedulerBackend.scala | 3 ++- 3 files changed, 5 insertions(+), 2 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/TaskState.scala b/core/src/main/scala/org/apache/spark/TaskState.scala index d85a6d683427d..c415fe99b105e 100644 --- a/core/src/main/scala/org/apache/spark/TaskState.scala +++ b/core/src/main/scala/org/apache/spark/TaskState.scala @@ -27,6 +27,8 @@ private[spark] object TaskState extends Enumeration { type TaskState = Value + def isFailed(state: TaskState) = (LOST == state) || (FAILED == state) + def isFinished(state: TaskState) = FINISHED_STATES.contains(state) def toMesos(state: TaskState): MesosTaskState = state match { diff --git a/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/CoarseMesosSchedulerBackend.scala b/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/CoarseMesosSchedulerBackend.scala index fc92b9c35c3a3..e13de0f46ef89 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/CoarseMesosSchedulerBackend.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/CoarseMesosSchedulerBackend.scala @@ -277,7 +277,7 @@ private[spark] class CoarseMesosSchedulerBackend( coresByTaskId -= taskId } // If it was a failure, mark the slave as failed for blacklisting purposes - if (state == MesosTaskState.TASK_FAILED || state == MesosTaskState.TASK_LOST) { + if (TaskState.isFailed(TaskState.fromMesos(state))) { failuresBySlaveId(slaveId) = failuresBySlaveId.getOrElse(slaveId, 0) + 1 if (failuresBySlaveId(slaveId) >= MAX_SLAVE_FAILURES) { logInfo("Blacklisting Mesos slave " + slaveId + " due to too many failures; " + diff --git a/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MesosSchedulerBackend.scala b/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MesosSchedulerBackend.scala index df8f4306b88a8..06bb527522141 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MesosSchedulerBackend.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MesosSchedulerBackend.scala @@ -318,7 +318,8 @@ private[spark] class MesosSchedulerBackend( val tid = status.getTaskId.getValue.toLong val state = TaskState.fromMesos(status.getState) synchronized { - if (status.getState == MesosTaskState.TASK_LOST && taskIdToSlaveId.contains(tid)) { + if (TaskState.isFailed(TaskState.fromMesos(status.getState)) + && taskIdToSlaveId.contains(tid)) { // We lost the executor on this slave, so remember that it's gone removeExecutor(taskIdToSlaveId(tid), "Lost executor") } From d08e3eb3dc455970b685a7b8b7e00c537c89a8e4 Mon Sep 17 00:00:00 2001 From: Sean Owen Date: Fri, 20 Mar 2015 14:14:53 +0000 Subject: [PATCH 108/122] SPARK-5134 [BUILD] Bump default Hadoop version to 2+ Bump default Hadoop version to 2.2.0. (This is already the dependency version reported by published Maven artifacts.) See JIRA for further discussion. Author: Sean Owen Closes #5027 from srowen/SPARK-5134 and squashes the following commits: acbee14 [Sean Owen] Bump default Hadoop version to 2.2.0. (This is already the dependency version reported by published Maven artifacts.) --- pom.xml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/pom.xml b/pom.xml index 6fc56a86d44ac..efb9f172f4751 100644 --- a/pom.xml +++ b/pom.xml @@ -120,7 +120,7 @@ shaded-protobuf 1.7.10 1.2.17 - 1.0.4 + 2.2.0 2.4.1 ${hadoop.version} 0.98.7-hadoop1 From 6f80c3e8880340597f161f87e64697bec86cc586 Mon Sep 17 00:00:00 2001 From: Sean Owen Date: Fri, 20 Mar 2015 14:16:21 +0000 Subject: [PATCH 109/122] SPARK-6338 [CORE] Use standard temp dir mechanisms in tests to avoid orphaned temp files Use `Utils.createTempDir()` to replace other temp file mechanisms used in some tests, to further ensure they are cleaned up, and simplify Author: Sean Owen Closes #5029 from srowen/SPARK-6338 and squashes the following commits: 27b740a [Sean Owen] Fix hive-thriftserver tests that don't expect an existing dir 4a212fa [Sean Owen] Standardize a bit more temp dir management 9004081 [Sean Owen] Revert some added recursive-delete calls 57609e4 [Sean Owen] Use Utils.createTempDir() to replace other temp file mechanisms used in some tests, to further ensure they are cleaned up, and simplify --- .../spark/deploy/FaultToleranceTest.scala | 4 ++-- .../scala/org/apache/spark/util/Utils.scala | 2 +- .../org/apache/spark/CheckpointSuite.scala | 3 +-- .../apache/spark/SecurityManagerSuite.scala | 5 +++-- .../org/apache/spark/SparkContextSuite.scala | 11 +++++----- .../spark/deploy/SparkSubmitSuite.scala | 8 ++++--- .../org/apache/spark/rdd/PipedRDDSuite.scala | 6 +++-- .../storage/BlockObjectWriterSuite.scala | 16 +++++++------- .../apache/spark/util/FileAppenderSuite.scala | 2 +- .../org/apache/spark/util/UtilsSuite.scala | 7 +++--- .../kafka/ReliableKafkaStreamSuite.scala | 10 +++------ .../org/apache/spark/graphx/GraphSuite.scala | 6 ++--- .../org/apache/spark/repl/ReplSuite.scala | 5 +---- .../expressions/codegen/package.scala | 4 ++-- .../spark/sql/catalyst/util/package.scala | 15 ++----------- .../spark/sql/parquet/ParquetTest.scala | 6 ++--- .../spark/sql/UserDefinedTypeSuite.scala | 6 +++-- .../org/apache/spark/sql/json/JsonSuite.scala | 22 +++++++++++-------- .../sources/CreateTableAsSelectSuite.scala | 5 ++--- .../spark/sql/sources/InsertSuite.scala | 5 ++--- .../spark/sql/sources/SaveLoadSuite.scala | 6 ++--- .../sql/hive/thriftserver/CliSuite.scala | 8 ++++--- .../HiveThriftServer2Suites.scala | 7 +++--- .../apache/spark/sql/hive/test/TestHive.scala | 16 +++++--------- .../sql/hive/InsertIntoHiveTableSuite.scala | 5 ++--- .../sql/hive/MetastoreDataSourcesSuite.scala | 14 ++++++------ .../apache/spark/sql/hive/parquetSuites.scala | 22 +++++-------------- .../spark/streaming/CheckpointSuite.scala | 6 ++--- .../apache/spark/streaming/FailureSuite.scala | 10 ++++----- .../streaming/ReceivedBlockHandlerSuite.scala | 11 +++------- .../streaming/ReceivedBlockTrackerSuite.scala | 9 ++------ .../spark/streaming/ReceiverSuite.scala | 5 ++--- .../yarn/YarnSparkHadoopUtilSuite.scala | 2 +- 33 files changed, 116 insertions(+), 153 deletions(-) diff --git a/core/src/main/scala/org/apache/spark/deploy/FaultToleranceTest.scala b/core/src/main/scala/org/apache/spark/deploy/FaultToleranceTest.scala index 4e58aa0ed4c7e..5668b53fc6f4f 100644 --- a/core/src/main/scala/org/apache/spark/deploy/FaultToleranceTest.scala +++ b/core/src/main/scala/org/apache/spark/deploy/FaultToleranceTest.scala @@ -33,6 +33,7 @@ import org.json4s.jackson.JsonMethods import org.apache.spark.{Logging, SparkConf, SparkContext} import org.apache.spark.deploy.master.{RecoveryState, SparkCuratorUtil} +import org.apache.spark.util.Utils /** * This suite tests the fault tolerance of the Spark standalone scheduler, mainly the Master. @@ -405,8 +406,7 @@ private object SparkDocker { private def startNode(dockerCmd: ProcessBuilder) : (String, DockerId, File) = { val ipPromise = promise[String]() - val outFile = File.createTempFile("fault-tolerance-test", "") - outFile.deleteOnExit() + val outFile = File.createTempFile("fault-tolerance-test", "", Utils.createTempDir()) val outStream: FileWriter = new FileWriter(outFile) def findIpAndLog(line: String): Unit = { if (line.startsWith("CONTAINER_IP=")) { diff --git a/core/src/main/scala/org/apache/spark/util/Utils.scala b/core/src/main/scala/org/apache/spark/util/Utils.scala index 91aa70870ab20..fa56bb09e4e5c 100644 --- a/core/src/main/scala/org/apache/spark/util/Utils.scala +++ b/core/src/main/scala/org/apache/spark/util/Utils.scala @@ -288,7 +288,7 @@ private[spark] object Utils extends Logging { } catch { case e: SecurityException => dir = null; } } - dir + dir.getCanonicalFile } /** diff --git a/core/src/test/scala/org/apache/spark/CheckpointSuite.scala b/core/src/test/scala/org/apache/spark/CheckpointSuite.scala index 3b10b3a042317..32abc65385267 100644 --- a/core/src/test/scala/org/apache/spark/CheckpointSuite.scala +++ b/core/src/test/scala/org/apache/spark/CheckpointSuite.scala @@ -33,8 +33,7 @@ class CheckpointSuite extends FunSuite with LocalSparkContext with Logging { override def beforeEach() { super.beforeEach() - checkpointDir = File.createTempFile("temp", "") - checkpointDir.deleteOnExit() + checkpointDir = File.createTempFile("temp", "", Utils.createTempDir()) checkpointDir.delete() sc = new SparkContext("local", "test") sc.setCheckpointDir(checkpointDir.toString) diff --git a/core/src/test/scala/org/apache/spark/SecurityManagerSuite.scala b/core/src/test/scala/org/apache/spark/SecurityManagerSuite.scala index 43fbd3ff3f756..62cb7649c0284 100644 --- a/core/src/test/scala/org/apache/spark/SecurityManagerSuite.scala +++ b/core/src/test/scala/org/apache/spark/SecurityManagerSuite.scala @@ -21,6 +21,8 @@ import java.io.File import org.scalatest.FunSuite +import org.apache.spark.util.Utils + class SecurityManagerSuite extends FunSuite { test("set security with conf") { @@ -160,8 +162,7 @@ class SecurityManagerSuite extends FunSuite { } test("ssl off setup") { - val file = File.createTempFile("SSLOptionsSuite", "conf") - file.deleteOnExit() + val file = File.createTempFile("SSLOptionsSuite", "conf", Utils.createTempDir()) System.setProperty("spark.ssl.configFile", file.getAbsolutePath) val conf = new SparkConf() diff --git a/core/src/test/scala/org/apache/spark/SparkContextSuite.scala b/core/src/test/scala/org/apache/spark/SparkContextSuite.scala index b8e3e83b5a47b..b07c4d93db4e6 100644 --- a/core/src/test/scala/org/apache/spark/SparkContextSuite.scala +++ b/core/src/test/scala/org/apache/spark/SparkContextSuite.scala @@ -79,13 +79,14 @@ class SparkContextSuite extends FunSuite with LocalSparkContext { val byteArray2 = converter.convert(bytesWritable) assert(byteArray2.length === 0) } - + test("addFile works") { - val file1 = File.createTempFile("someprefix1", "somesuffix1") + val dir = Utils.createTempDir() + + val file1 = File.createTempFile("someprefix1", "somesuffix1", dir) val absolutePath1 = file1.getAbsolutePath - val pluto = Utils.createTempDir() - val file2 = File.createTempFile("someprefix2", "somesuffix2", pluto) + val file2 = File.createTempFile("someprefix2", "somesuffix2", dir) val relativePath = file2.getParent + "/../" + file2.getParentFile.getName + "/" + file2.getName val absolutePath2 = file2.getAbsolutePath @@ -129,7 +130,7 @@ class SparkContextSuite extends FunSuite with LocalSparkContext { sc.stop() } } - + test("addFile recursive works") { val pluto = Utils.createTempDir() val neptune = Utils.createTempDir(pluto.getAbsolutePath) diff --git a/core/src/test/scala/org/apache/spark/deploy/SparkSubmitSuite.scala b/core/src/test/scala/org/apache/spark/deploy/SparkSubmitSuite.scala index 46d745c4ecbfa..4561e5b8e9663 100644 --- a/core/src/test/scala/org/apache/spark/deploy/SparkSubmitSuite.scala +++ b/core/src/test/scala/org/apache/spark/deploy/SparkSubmitSuite.scala @@ -402,8 +402,10 @@ class SparkSubmitSuite extends FunSuite with Matchers with ResetSystemProperties val archives = "file:/archive1,archive2" // spark.yarn.dist.archives val pyFiles = "py-file1,py-file2" // spark.submit.pyFiles + val tmpDir = Utils.createTempDir() + // Test jars and files - val f1 = File.createTempFile("test-submit-jars-files", "") + val f1 = File.createTempFile("test-submit-jars-files", "", tmpDir) val writer1 = new PrintWriter(f1) writer1.println("spark.jars " + jars) writer1.println("spark.files " + files) @@ -420,7 +422,7 @@ class SparkSubmitSuite extends FunSuite with Matchers with ResetSystemProperties sysProps("spark.files") should be(Utils.resolveURIs(files)) // Test files and archives (Yarn) - val f2 = File.createTempFile("test-submit-files-archives", "") + val f2 = File.createTempFile("test-submit-files-archives", "", tmpDir) val writer2 = new PrintWriter(f2) writer2.println("spark.yarn.dist.files " + files) writer2.println("spark.yarn.dist.archives " + archives) @@ -437,7 +439,7 @@ class SparkSubmitSuite extends FunSuite with Matchers with ResetSystemProperties sysProps2("spark.yarn.dist.archives") should be(Utils.resolveURIs(archives)) // Test python files - val f3 = File.createTempFile("test-submit-python-files", "") + val f3 = File.createTempFile("test-submit-python-files", "", tmpDir) val writer3 = new PrintWriter(f3) writer3.println("spark.submit.pyFiles " + pyFiles) writer3.close() diff --git a/core/src/test/scala/org/apache/spark/rdd/PipedRDDSuite.scala b/core/src/test/scala/org/apache/spark/rdd/PipedRDDSuite.scala index 1a9a0e857e546..aea76c1adcc09 100644 --- a/core/src/test/scala/org/apache/spark/rdd/PipedRDDSuite.scala +++ b/core/src/test/scala/org/apache/spark/rdd/PipedRDDSuite.scala @@ -22,7 +22,6 @@ import java.io.File import org.apache.hadoop.fs.Path import org.apache.hadoop.io.{LongWritable, Text} import org.apache.hadoop.mapred.{FileSplit, JobConf, TextInputFormat} -import org.apache.spark._ import org.scalatest.FunSuite import scala.collection.Map @@ -30,6 +29,9 @@ import scala.language.postfixOps import scala.sys.process._ import scala.util.Try +import org.apache.spark._ +import org.apache.spark.util.Utils + class PipedRDDSuite extends FunSuite with SharedSparkContext { test("basic pipe") { @@ -141,7 +143,7 @@ class PipedRDDSuite extends FunSuite with SharedSparkContext { // make sure symlinks were created assert(pipedLs.length > 0) // clean up top level tasks directory - new File("tasks").delete() + Utils.deleteRecursively(new File("tasks")) } else { assert(true) } diff --git a/core/src/test/scala/org/apache/spark/storage/BlockObjectWriterSuite.scala b/core/src/test/scala/org/apache/spark/storage/BlockObjectWriterSuite.scala index c21c92b63ad13..78bbc4ec2c620 100644 --- a/core/src/test/scala/org/apache/spark/storage/BlockObjectWriterSuite.scala +++ b/core/src/test/scala/org/apache/spark/storage/BlockObjectWriterSuite.scala @@ -16,16 +16,18 @@ */ package org.apache.spark.storage -import org.scalatest.FunSuite import java.io.File + +import org.scalatest.FunSuite + +import org.apache.spark.SparkConf import org.apache.spark.executor.ShuffleWriteMetrics import org.apache.spark.serializer.JavaSerializer -import org.apache.spark.SparkConf +import org.apache.spark.util.Utils class BlockObjectWriterSuite extends FunSuite { test("verify write metrics") { - val file = new File("somefile") - file.deleteOnExit() + val file = new File(Utils.createTempDir(), "somefile") val writeMetrics = new ShuffleWriteMetrics() val writer = new DiskBlockObjectWriter(new TestBlockId("0"), file, new JavaSerializer(new SparkConf()), 1024, os => os, true, writeMetrics) @@ -47,8 +49,7 @@ class BlockObjectWriterSuite extends FunSuite { } test("verify write metrics on revert") { - val file = new File("somefile") - file.deleteOnExit() + val file = new File(Utils.createTempDir(), "somefile") val writeMetrics = new ShuffleWriteMetrics() val writer = new DiskBlockObjectWriter(new TestBlockId("0"), file, new JavaSerializer(new SparkConf()), 1024, os => os, true, writeMetrics) @@ -71,8 +72,7 @@ class BlockObjectWriterSuite extends FunSuite { } test("Reopening a closed block writer") { - val file = new File("somefile") - file.deleteOnExit() + val file = new File(Utils.createTempDir(), "somefile") val writeMetrics = new ShuffleWriteMetrics() val writer = new DiskBlockObjectWriter(new TestBlockId("0"), file, new JavaSerializer(new SparkConf()), 1024, os => os, true, writeMetrics) diff --git a/core/src/test/scala/org/apache/spark/util/FileAppenderSuite.scala b/core/src/test/scala/org/apache/spark/util/FileAppenderSuite.scala index 4dc5b6103db74..43b6a405cb68c 100644 --- a/core/src/test/scala/org/apache/spark/util/FileAppenderSuite.scala +++ b/core/src/test/scala/org/apache/spark/util/FileAppenderSuite.scala @@ -32,7 +32,7 @@ import org.apache.spark.util.logging.{RollingFileAppender, SizeBasedRollingPolic class FileAppenderSuite extends FunSuite with BeforeAndAfter with Logging { - val testFile = new File("FileAppenderSuite-test-" + System.currentTimeMillis).getAbsoluteFile + val testFile = new File(Utils.createTempDir(), "FileAppenderSuite-test").getAbsoluteFile before { cleanup() diff --git a/core/src/test/scala/org/apache/spark/util/UtilsSuite.scala b/core/src/test/scala/org/apache/spark/util/UtilsSuite.scala index b91428efadfd0..5d93086082189 100644 --- a/core/src/test/scala/org/apache/spark/util/UtilsSuite.scala +++ b/core/src/test/scala/org/apache/spark/util/UtilsSuite.scala @@ -122,7 +122,6 @@ class UtilsSuite extends FunSuite with ResetSystemProperties { test("reading offset bytes of a file") { val tmpDir2 = Utils.createTempDir() - tmpDir2.deleteOnExit() val f1Path = tmpDir2 + "/f1" val f1 = new FileOutputStream(f1Path) f1.write("1\n2\n3\n4\n5\n6\n7\n8\n9\n".getBytes(UTF_8)) @@ -151,7 +150,6 @@ class UtilsSuite extends FunSuite with ResetSystemProperties { test("reading offset bytes across multiple files") { val tmpDir = Utils.createTempDir() - tmpDir.deleteOnExit() val files = (1 to 3).map(i => new File(tmpDir, i.toString)) Files.write("0123456789", files(0), UTF_8) Files.write("abcdefghij", files(1), UTF_8) @@ -357,7 +355,8 @@ class UtilsSuite extends FunSuite with ResetSystemProperties { } test("loading properties from file") { - val outFile = File.createTempFile("test-load-spark-properties", "test") + val tmpDir = Utils.createTempDir() + val outFile = File.createTempFile("test-load-spark-properties", "test", tmpDir) try { System.setProperty("spark.test.fileNameLoadB", "2") Files.write("spark.test.fileNameLoadA true\n" + @@ -370,7 +369,7 @@ class UtilsSuite extends FunSuite with ResetSystemProperties { assert(sparkConf.getBoolean("spark.test.fileNameLoadA", false) === true) assert(sparkConf.getInt("spark.test.fileNameLoadB", 1) === 2) } finally { - outFile.delete() + Utils.deleteRecursively(tmpDir) } } diff --git a/external/kafka/src/test/scala/org/apache/spark/streaming/kafka/ReliableKafkaStreamSuite.scala b/external/kafka/src/test/scala/org/apache/spark/streaming/kafka/ReliableKafkaStreamSuite.scala index fc53c23abda85..3cd960d1fd1d4 100644 --- a/external/kafka/src/test/scala/org/apache/spark/streaming/kafka/ReliableKafkaStreamSuite.scala +++ b/external/kafka/src/test/scala/org/apache/spark/streaming/kafka/ReliableKafkaStreamSuite.scala @@ -25,16 +25,15 @@ import scala.concurrent.duration._ import scala.language.postfixOps import scala.util.Random -import com.google.common.io.Files import kafka.serializer.StringDecoder import kafka.utils.{ZKGroupTopicDirs, ZkUtils} -import org.apache.commons.io.FileUtils import org.scalatest.BeforeAndAfter import org.scalatest.concurrent.Eventually import org.apache.spark.SparkConf import org.apache.spark.storage.StorageLevel import org.apache.spark.streaming.{Milliseconds, StreamingContext} +import org.apache.spark.util.Utils class ReliableKafkaStreamSuite extends KafkaStreamSuiteBase with BeforeAndAfter with Eventually { @@ -60,7 +59,7 @@ class ReliableKafkaStreamSuite extends KafkaStreamSuiteBase with BeforeAndAfter ) ssc = new StreamingContext(sparkConf, Milliseconds(500)) - tempDirectory = Files.createTempDir() + tempDirectory = Utils.createTempDir() ssc.checkpoint(tempDirectory.getAbsolutePath) } @@ -68,10 +67,7 @@ class ReliableKafkaStreamSuite extends KafkaStreamSuiteBase with BeforeAndAfter if (ssc != null) { ssc.stop() } - if (tempDirectory != null && tempDirectory.exists()) { - FileUtils.deleteDirectory(tempDirectory) - tempDirectory = null - } + Utils.deleteRecursively(tempDirectory) tearDownKafka() } diff --git a/graphx/src/test/scala/org/apache/spark/graphx/GraphSuite.scala b/graphx/src/test/scala/org/apache/spark/graphx/GraphSuite.scala index b61d9f0fbe5e4..8d15150458d26 100644 --- a/graphx/src/test/scala/org/apache/spark/graphx/GraphSuite.scala +++ b/graphx/src/test/scala/org/apache/spark/graphx/GraphSuite.scala @@ -19,13 +19,12 @@ package org.apache.spark.graphx import org.scalatest.FunSuite -import com.google.common.io.Files - import org.apache.spark.SparkContext import org.apache.spark.graphx.Graph._ import org.apache.spark.graphx.PartitionStrategy._ import org.apache.spark.rdd._ import org.apache.spark.storage.StorageLevel +import org.apache.spark.util.Utils class GraphSuite extends FunSuite with LocalSparkContext { @@ -369,8 +368,7 @@ class GraphSuite extends FunSuite with LocalSparkContext { } test("checkpoint") { - val checkpointDir = Files.createTempDir() - checkpointDir.deleteOnExit() + val checkpointDir = Utils.createTempDir() withSpark { sc => sc.setCheckpointDir(checkpointDir.getAbsolutePath) val ring = (0L to 100L).zip((1L to 99L) :+ 0L).map { case (a, b) => Edge(a, b, 1)} diff --git a/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala b/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala index fbef5b25ba688..14f5e9ed4f25e 100644 --- a/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala +++ b/repl/scala-2.11/src/test/scala/org/apache/spark/repl/ReplSuite.scala @@ -21,11 +21,9 @@ import java.io._ import java.net.URLClassLoader import scala.collection.mutable.ArrayBuffer -import scala.concurrent.Await import scala.concurrent.duration._ import scala.tools.nsc.interpreter.SparkILoop -import com.google.common.io.Files import org.scalatest.FunSuite import org.apache.commons.lang3.StringEscapeUtils import org.apache.spark.SparkContext @@ -196,8 +194,7 @@ class ReplSuite extends FunSuite { } test("interacting with files") { - val tempDir = Files.createTempDir() - tempDir.deleteOnExit() + val tempDir = Utils.createTempDir() val out = new FileWriter(tempDir + "/input") out.write("Hello world!\n") out.write("What's up?\n") diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/package.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/package.scala index 80c7dfd376c96..528e38a50a740 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/package.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/codegen/package.scala @@ -19,7 +19,7 @@ package org.apache.spark.sql.catalyst.expressions import org.apache.spark.annotation.DeveloperApi import org.apache.spark.sql.catalyst.rules -import org.apache.spark.sql.catalyst.util +import org.apache.spark.util.Utils /** * A collection of generators that build custom bytecode at runtime for performing the evaluation @@ -52,7 +52,7 @@ package object codegen { @DeveloperApi object DumpByteCode { import scala.sys.process._ - val dumpDirectory = util.getTempFilePath("sparkSqlByteCode") + val dumpDirectory = Utils.createTempDir() dumpDirectory.mkdir() def apply(obj: Any): Unit = { diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/util/package.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/util/package.scala index d8da45ae70c4b..feed50f9a2a2d 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/util/package.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/util/package.scala @@ -19,20 +19,9 @@ package org.apache.spark.sql.catalyst import java.io.{PrintWriter, ByteArrayOutputStream, FileInputStream, File} -import org.apache.spark.util.{Utils => SparkUtils} +import org.apache.spark.util.Utils package object util { - /** - * Returns a path to a temporary file that probably does not exist. - * Note, there is always the race condition that someone created this - * file since the last time we checked. Thus, this shouldn't be used - * for anything security conscious. - */ - def getTempFilePath(prefix: String, suffix: String = ""): File = { - val tempFile = File.createTempFile(prefix, suffix) - tempFile.delete() - tempFile - } def fileToString(file: File, encoding: String = "UTF-8") = { val inStream = new FileInputStream(file) @@ -56,7 +45,7 @@ package object util { def resourceToString( resource:String, encoding: String = "UTF-8", - classLoader: ClassLoader = SparkUtils.getSparkClassLoader) = { + classLoader: ClassLoader = Utils.getSparkClassLoader) = { val inStream = classLoader.getResourceAsStream(resource) val outStream = new ByteArrayOutputStream try { diff --git a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTest.scala b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTest.scala index d6ea6679c5966..9d17516e0ef7d 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTest.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTest.scala @@ -23,7 +23,6 @@ import scala.reflect.ClassTag import scala.reflect.runtime.universe.TypeTag import scala.util.Try -import org.apache.spark.sql.catalyst.util import org.apache.spark.sql.{DataFrame, SQLContext, SaveMode} import org.apache.spark.util.Utils @@ -67,8 +66,9 @@ private[sql] trait ParquetTest { * @todo Probably this method should be moved to a more general place */ protected def withTempPath(f: File => Unit): Unit = { - val file = util.getTempFilePath("parquetTest").getCanonicalFile - try f(file) finally if (file.exists()) Utils.deleteRecursively(file) + val path = Utils.createTempDir() + path.delete() + try f(path) finally Utils.deleteRecursively(path) } /** diff --git a/sql/core/src/test/scala/org/apache/spark/sql/UserDefinedTypeSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/UserDefinedTypeSuite.scala index 23f424c0bfc7c..fe618e0e8e767 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/UserDefinedTypeSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/UserDefinedTypeSuite.scala @@ -19,6 +19,8 @@ package org.apache.spark.sql import java.io.File +import org.apache.spark.util.Utils + import scala.beans.{BeanInfo, BeanProperty} import org.apache.spark.rdd.RDD @@ -98,13 +100,13 @@ class UserDefinedTypeSuite extends QueryTest { test("UDTs with Parquet") { - val tempDir = File.createTempFile("parquet", "test") + val tempDir = Utils.createTempDir() tempDir.delete() pointsRDD.saveAsParquetFile(tempDir.getCanonicalPath) } test("Repartition UDTs with Parquet") { - val tempDir = File.createTempFile("parquet", "test") + val tempDir = Utils.createTempDir() tempDir.delete() pointsRDD.repartition(1).saveAsParquetFile(tempDir.getCanonicalPath) } diff --git a/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala index 320b80d80e997..706c966ee05f5 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/json/JsonSuite.scala @@ -22,7 +22,6 @@ import java.sql.{Date, Timestamp} import org.scalactic.Tolerance._ import org.apache.spark.sql.TestData._ -import org.apache.spark.sql.catalyst.util._ import org.apache.spark.sql.functions._ import org.apache.spark.sql.json.JsonRDD.{compatibleType, enforceCorrectType} import org.apache.spark.sql.sources.LogicalRelation @@ -31,6 +30,7 @@ import org.apache.spark.sql.test.TestSQLContext._ import org.apache.spark.sql.test.TestSQLContext.implicits._ import org.apache.spark.sql.types._ import org.apache.spark.sql.{QueryTest, Row, SQLConf} +import org.apache.spark.util.Utils class JsonSuite extends QueryTest { import org.apache.spark.sql.json.TestJsonData._ @@ -554,8 +554,9 @@ class JsonSuite extends QueryTest { } test("jsonFile should be based on JSONRelation") { - val file = getTempFilePath("json") - val path = file.toString + val dir = Utils.createTempDir() + dir.delete() + val path = dir.getCanonicalPath sparkContext.parallelize(1 to 100).map(i => s"""{"a": 1, "b": "str$i"}""").saveAsTextFile(path) val jsonDF = jsonFile(path, 0.49) @@ -580,8 +581,9 @@ class JsonSuite extends QueryTest { } test("Loading a JSON dataset from a text file") { - val file = getTempFilePath("json") - val path = file.toString + val dir = Utils.createTempDir() + dir.delete() + val path = dir.getCanonicalPath primitiveFieldAndType.map(record => record.replaceAll("\n", " ")).saveAsTextFile(path) val jsonDF = jsonFile(path) @@ -611,8 +613,9 @@ class JsonSuite extends QueryTest { } test("Loading a JSON dataset from a text file with SQL") { - val file = getTempFilePath("json") - val path = file.toString + val dir = Utils.createTempDir() + dir.delete() + val path = dir.getCanonicalPath primitiveFieldAndType.map(record => record.replaceAll("\n", " ")).saveAsTextFile(path) sql( @@ -637,8 +640,9 @@ class JsonSuite extends QueryTest { } test("Applying schemas") { - val file = getTempFilePath("json") - val path = file.toString + val dir = Utils.createTempDir() + dir.delete() + val path = dir.getCanonicalPath primitiveFieldAndType.map(record => record.replaceAll("\n", " ")).saveAsTextFile(path) val schema = StructType( diff --git a/sql/core/src/test/scala/org/apache/spark/sql/sources/CreateTableAsSelectSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/sources/CreateTableAsSelectSuite.scala index 60355414a40fa..2975a7fee4c96 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/sources/CreateTableAsSelectSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/sources/CreateTableAsSelectSuite.scala @@ -22,7 +22,6 @@ import java.io.File import org.apache.spark.sql.AnalysisException import org.scalatest.BeforeAndAfterAll -import org.apache.spark.sql.catalyst.util import org.apache.spark.util.Utils class CreateTableAsSelectSuite extends DataSourceTest with BeforeAndAfterAll { @@ -32,7 +31,7 @@ class CreateTableAsSelectSuite extends DataSourceTest with BeforeAndAfterAll { var path: File = null override def beforeAll(): Unit = { - path = util.getTempFilePath("jsonCTAS").getCanonicalFile + path = Utils.createTempDir() val rdd = sparkContext.parallelize((1 to 10).map(i => s"""{"a":$i, "b":"str${i}"}""")) jsonRDD(rdd).registerTempTable("jt") } @@ -42,7 +41,7 @@ class CreateTableAsSelectSuite extends DataSourceTest with BeforeAndAfterAll { } after { - if (path.exists()) Utils.deleteRecursively(path) + Utils.deleteRecursively(path) } test("CREATE TEMPORARY TABLE AS SELECT") { diff --git a/sql/core/src/test/scala/org/apache/spark/sql/sources/InsertSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/sources/InsertSuite.scala index b5b16f9546691..80efe9728fbc2 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/sources/InsertSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/sources/InsertSuite.scala @@ -22,7 +22,6 @@ import java.io.File import org.scalatest.BeforeAndAfterAll import org.apache.spark.sql.{AnalysisException, Row} -import org.apache.spark.sql.catalyst.util import org.apache.spark.util.Utils class InsertSuite extends DataSourceTest with BeforeAndAfterAll { @@ -32,7 +31,7 @@ class InsertSuite extends DataSourceTest with BeforeAndAfterAll { var path: File = null override def beforeAll: Unit = { - path = util.getTempFilePath("jsonCTAS").getCanonicalFile + path = Utils.createTempDir() val rdd = sparkContext.parallelize((1 to 10).map(i => s"""{"a":$i, "b":"str${i}"}""")) jsonRDD(rdd).registerTempTable("jt") sql( @@ -48,7 +47,7 @@ class InsertSuite extends DataSourceTest with BeforeAndAfterAll { override def afterAll: Unit = { dropTempTable("jsonTable") dropTempTable("jt") - if (path.exists()) Utils.deleteRecursively(path) + Utils.deleteRecursively(path) } test("Simple INSERT OVERWRITE a JSONRelation") { diff --git a/sql/core/src/test/scala/org/apache/spark/sql/sources/SaveLoadSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/sources/SaveLoadSuite.scala index 607488ccfdd6a..43bc8eb2d11a7 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/sources/SaveLoadSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/sources/SaveLoadSuite.scala @@ -21,7 +21,6 @@ import java.io.File import org.scalatest.BeforeAndAfterAll -import org.apache.spark.sql.catalyst.util import org.apache.spark.sql.{SaveMode, SQLConf, DataFrame} import org.apache.spark.sql.types._ import org.apache.spark.util.Utils @@ -39,7 +38,8 @@ class SaveLoadSuite extends DataSourceTest with BeforeAndAfterAll { override def beforeAll(): Unit = { originalDefaultSource = conf.defaultDataSourceName - path = util.getTempFilePath("datasource").getCanonicalFile + path = Utils.createTempDir() + path.delete() val rdd = sparkContext.parallelize((1 to 10).map(i => s"""{"a":$i, "b":"str${i}"}""")) df = jsonRDD(rdd) @@ -52,7 +52,7 @@ class SaveLoadSuite extends DataSourceTest with BeforeAndAfterAll { after { conf.setConf(SQLConf.DEFAULT_DATA_SOURCE_NAME, originalDefaultSource) - if (path.exists()) Utils.deleteRecursively(path) + Utils.deleteRecursively(path) } def checkLoad(): Unit = { diff --git a/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/CliSuite.scala b/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/CliSuite.scala index 8bca4b33b3ad1..75738fa22b572 100644 --- a/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/CliSuite.scala +++ b/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/CliSuite.scala @@ -29,7 +29,7 @@ import org.apache.hadoop.hive.conf.HiveConf.ConfVars import org.scalatest.{BeforeAndAfterAll, FunSuite} import org.apache.spark.Logging -import org.apache.spark.sql.catalyst.util.getTempFilePath +import org.apache.spark.util.Utils class CliSuite extends FunSuite with BeforeAndAfterAll with Logging { def runCliWithin( @@ -38,8 +38,10 @@ class CliSuite extends FunSuite with BeforeAndAfterAll with Logging { queriesAndExpectedAnswers: (String, String)*) { val (queries, expectedAnswers) = queriesAndExpectedAnswers.unzip - val warehousePath = getTempFilePath("warehouse") - val metastorePath = getTempFilePath("metastore") + val warehousePath = Utils.createTempDir() + warehousePath.delete() + val metastorePath = Utils.createTempDir() + metastorePath.delete() val cliScript = "../../bin/spark-sql".split("/").mkString(File.separator) val command = { diff --git a/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2Suites.scala b/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2Suites.scala index aff96e21a5373..bf20acecb1f32 100644 --- a/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2Suites.scala +++ b/sql/hive-thriftserver/src/test/scala/org/apache/spark/sql/hive/thriftserver/HiveThriftServer2Suites.scala @@ -37,7 +37,6 @@ import org.apache.thrift.transport.TSocket import org.scalatest.{BeforeAndAfterAll, FunSuite} import org.apache.spark.Logging -import org.apache.spark.sql.catalyst.util import org.apache.spark.sql.hive.HiveShim import org.apache.spark.util.Utils @@ -447,8 +446,10 @@ abstract class HiveThriftServer2Test extends FunSuite with BeforeAndAfterAll wit } private def startThriftServer(port: Int, attempt: Int) = { - warehousePath = util.getTempFilePath("warehouse") - metastorePath = util.getTempFilePath("metastore") + warehousePath = Utils.createTempDir() + warehousePath.delete() + metastorePath = Utils.createTempDir() + metastorePath.delete() logPath = null logTailingProcess = null diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/test/TestHive.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/test/TestHive.scala index 4859991e2351a..b4aee78046383 100644 --- a/sql/hive/src/main/scala/org/apache/spark/sql/hive/test/TestHive.scala +++ b/sql/hive/src/main/scala/org/apache/spark/sql/hive/test/TestHive.scala @@ -30,7 +30,6 @@ import org.apache.hadoop.hive.serde2.avro.AvroSerDe import org.apache.spark.sql.SQLConf import org.apache.spark.sql.catalyst.analysis._ import org.apache.spark.sql.catalyst.plans.logical.LogicalPlan -import org.apache.spark.sql.catalyst.util._ import org.apache.spark.sql.execution.CacheTableCommand import org.apache.spark.sql.hive._ import org.apache.spark.sql.hive.execution.HiveNativeCommand @@ -69,22 +68,19 @@ class TestHiveContext(sc: SparkContext) extends HiveContext(sc) { hiveconf.set("hive.plan.serialization.format", "javaXML") - lazy val warehousePath = getTempFilePath("sparkHiveWarehouse").getCanonicalPath - lazy val metastorePath = getTempFilePath("sparkHiveMetastore").getCanonicalPath + lazy val warehousePath = Utils.createTempDir() + lazy val metastorePath = Utils.createTempDir() /** Sets up the system initially or after a RESET command */ protected def configure(): Unit = { + warehousePath.delete() + metastorePath.delete() setConf("javax.jdo.option.ConnectionURL", s"jdbc:derby:;databaseName=$metastorePath;create=true") - setConf("hive.metastore.warehouse.dir", warehousePath) - Utils.registerShutdownDeleteDir(new File(warehousePath)) - Utils.registerShutdownDeleteDir(new File(metastorePath)) + setConf("hive.metastore.warehouse.dir", warehousePath.toString) } - val testTempDir = File.createTempFile("testTempFiles", "spark.hive.tmp") - testTempDir.delete() - testTempDir.mkdir() - Utils.registerShutdownDeleteDir(testTempDir) + val testTempDir = Utils.createTempDir() // For some hive test case which contain ${system:test.tmp.dir} System.setProperty("test.tmp.dir", testTempDir.getCanonicalPath) diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/hive/InsertIntoHiveTableSuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/hive/InsertIntoHiveTableSuite.scala index d4b175fa443a4..381cd2a29123e 100644 --- a/sql/hive/src/test/scala/org/apache/spark/sql/hive/InsertIntoHiveTableSuite.scala +++ b/sql/hive/src/test/scala/org/apache/spark/sql/hive/InsertIntoHiveTableSuite.scala @@ -21,12 +21,11 @@ import java.io.File import org.scalatest.BeforeAndAfter -import com.google.common.io.Files - import org.apache.spark.sql.execution.QueryExecutionException import org.apache.spark.sql.{QueryTest, _} import org.apache.spark.sql.hive.test.TestHive import org.apache.spark.sql.types._ +import org.apache.spark.util.Utils /* Implicits */ import org.apache.spark.sql.hive.test.TestHive._ @@ -112,7 +111,7 @@ class InsertIntoHiveTableSuite extends QueryTest with BeforeAndAfter { test("SPARK-4203:random partition directory order") { sql("CREATE TABLE tmp_table (key int, value string)") - val tmpDir = Files.createTempDir() + val tmpDir = Utils.createTempDir() sql(s"CREATE TABLE table_with_partition(c1 string) PARTITIONED by (p1 string,p2 string,p3 string,p4 string,p5 string) location '${tmpDir.toURI.toString}' ") sql("INSERT OVERWRITE TABLE table_with_partition partition (p1='a',p2='b',p3='c',p4='c',p5='1') SELECT 'blarr' FROM tmp_table") sql("INSERT OVERWRITE TABLE table_with_partition partition (p1='a',p2='b',p3='c',p4='c',p5='2') SELECT 'blarr' FROM tmp_table") diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/hive/MetastoreDataSourcesSuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/hive/MetastoreDataSourcesSuite.scala index 5d6a6f3b64f03..ff2e6ea9ea51d 100644 --- a/sql/hive/src/test/scala/org/apache/spark/sql/hive/MetastoreDataSourcesSuite.scala +++ b/sql/hive/src/test/scala/org/apache/spark/sql/hive/MetastoreDataSourcesSuite.scala @@ -19,13 +19,14 @@ package org.apache.spark.sql.hive import java.io.File +import scala.collection.mutable.ArrayBuffer + import org.scalatest.BeforeAndAfterEach import org.apache.commons.io.FileUtils import org.apache.hadoop.fs.Path import org.apache.hadoop.mapred.InvalidInputException -import org.apache.spark.sql.catalyst.util import org.apache.spark.sql._ import org.apache.spark.util.Utils import org.apache.spark.sql.types._ @@ -34,8 +35,6 @@ import org.apache.spark.sql.hive.test.TestHive.implicits._ import org.apache.spark.sql.parquet.ParquetRelation2 import org.apache.spark.sql.sources.LogicalRelation -import scala.collection.mutable.ArrayBuffer - /** * Tests for persisting tables created though the data sources API into the metastore. */ @@ -43,11 +42,12 @@ class MetastoreDataSourcesSuite extends QueryTest with BeforeAndAfterEach { override def afterEach(): Unit = { reset() - if (tempPath.exists()) Utils.deleteRecursively(tempPath) + Utils.deleteRecursively(tempPath) } val filePath = Utils.getSparkClassLoader.getResource("sample.json").getFile - var tempPath: File = util.getTempFilePath("jsonCTAS").getCanonicalFile + var tempPath: File = Utils.createTempDir() + tempPath.delete() test ("persistent JSON table") { sql( @@ -154,7 +154,7 @@ class MetastoreDataSourcesSuite extends QueryTest with BeforeAndAfterEach { } test("check change without refresh") { - val tempDir = File.createTempFile("sparksql", "json") + val tempDir = File.createTempFile("sparksql", "json", Utils.createTempDir()) tempDir.delete() sparkContext.parallelize(("a", "b") :: Nil).toDF() .toJSON.saveAsTextFile(tempDir.getCanonicalPath) @@ -192,7 +192,7 @@ class MetastoreDataSourcesSuite extends QueryTest with BeforeAndAfterEach { } test("drop, change, recreate") { - val tempDir = File.createTempFile("sparksql", "json") + val tempDir = File.createTempFile("sparksql", "json", Utils.createTempDir()) tempDir.delete() sparkContext.parallelize(("a", "b") :: Nil).toDF() .toJSON.saveAsTextFile(tempDir.getCanonicalPath) diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/hive/parquetSuites.scala b/sql/hive/src/test/scala/org/apache/spark/sql/hive/parquetSuites.scala index 1904f5faef3a0..d891c4e8903d9 100644 --- a/sql/hive/src/test/scala/org/apache/spark/sql/hive/parquetSuites.scala +++ b/sql/hive/src/test/scala/org/apache/spark/sql/hive/parquetSuites.scala @@ -32,6 +32,7 @@ import org.apache.spark.sql.sources.{InsertIntoDataSource, LogicalRelation} import org.apache.spark.sql.parquet.{ParquetRelation2, ParquetTableScan} import org.apache.spark.sql.SaveMode import org.apache.spark.sql.types._ +import org.apache.spark.util.Utils // The data where the partitioning key exists only in the directory structure. case class ParquetData(intField: Int, stringField: String) @@ -579,13 +580,8 @@ abstract class ParquetPartitioningTest extends QueryTest with BeforeAndAfterAll var partitionedTableDirWithKeyAndComplexTypes: File = null override def beforeAll(): Unit = { - partitionedTableDir = File.createTempFile("parquettests", "sparksql") - partitionedTableDir.delete() - partitionedTableDir.mkdir() - - normalTableDir = File.createTempFile("parquettests", "sparksql") - normalTableDir.delete() - normalTableDir.mkdir() + partitionedTableDir = Utils.createTempDir() + normalTableDir = Utils.createTempDir() (1 to 10).foreach { p => val partDir = new File(partitionedTableDir, s"p=$p") @@ -601,9 +597,7 @@ abstract class ParquetPartitioningTest extends QueryTest with BeforeAndAfterAll .toDF() .saveAsParquetFile(new File(normalTableDir, "normal").getCanonicalPath) - partitionedTableDirWithKey = File.createTempFile("parquettests", "sparksql") - partitionedTableDirWithKey.delete() - partitionedTableDirWithKey.mkdir() + partitionedTableDirWithKey = Utils.createTempDir() (1 to 10).foreach { p => val partDir = new File(partitionedTableDirWithKey, s"p=$p") @@ -613,9 +607,7 @@ abstract class ParquetPartitioningTest extends QueryTest with BeforeAndAfterAll .saveAsParquetFile(partDir.getCanonicalPath) } - partitionedTableDirWithKeyAndComplexTypes = File.createTempFile("parquettests", "sparksql") - partitionedTableDirWithKeyAndComplexTypes.delete() - partitionedTableDirWithKeyAndComplexTypes.mkdir() + partitionedTableDirWithKeyAndComplexTypes = Utils.createTempDir() (1 to 10).foreach { p => val partDir = new File(partitionedTableDirWithKeyAndComplexTypes, s"p=$p") @@ -625,9 +617,7 @@ abstract class ParquetPartitioningTest extends QueryTest with BeforeAndAfterAll }.toDF().saveAsParquetFile(partDir.getCanonicalPath) } - partitionedTableDirWithComplexTypes = File.createTempFile("parquettests", "sparksql") - partitionedTableDirWithComplexTypes.delete() - partitionedTableDirWithComplexTypes.mkdir() + partitionedTableDirWithComplexTypes = Utils.createTempDir() (1 to 10).foreach { p => val partDir = new File(partitionedTableDirWithComplexTypes, s"p=$p") diff --git a/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala index 8ea91eca683cf..91a2b2bba461d 100644 --- a/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala +++ b/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala @@ -222,7 +222,7 @@ class CheckpointSuite extends TestSuiteBase { } test("recovery with saveAsHadoopFiles operation") { - val tempDir = Files.createTempDir() + val tempDir = Utils.createTempDir() try { testCheckpointedOperation( Seq(Seq("a", "a", "b"), Seq("", ""), Seq(), Seq("a", "a", "b"), Seq("", ""), Seq()), @@ -245,7 +245,7 @@ class CheckpointSuite extends TestSuiteBase { } test("recovery with saveAsNewAPIHadoopFiles operation") { - val tempDir = Files.createTempDir() + val tempDir = Utils.createTempDir() try { testCheckpointedOperation( Seq(Seq("a", "a", "b"), Seq("", ""), Seq(), Seq("a", "a", "b"), Seq("", ""), Seq()), @@ -283,7 +283,7 @@ class CheckpointSuite extends TestSuiteBase { // // After SPARK-5079 is addressed, should be able to remove this test since a strengthened // version of the other saveAsHadoopFile* tests would prevent regressions for this issue. - val tempDir = Files.createTempDir() + val tempDir = Utils.createTempDir() try { testCheckpointedOperation( Seq(Seq("a", "a", "b"), Seq("", ""), Seq(), Seq("a", "a", "b"), Seq("", ""), Seq()), diff --git a/streaming/src/test/scala/org/apache/spark/streaming/FailureSuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/FailureSuite.scala index 6500608bba87c..26435d8515815 100644 --- a/streaming/src/test/scala/org/apache/spark/streaming/FailureSuite.scala +++ b/streaming/src/test/scala/org/apache/spark/streaming/FailureSuite.scala @@ -20,15 +20,13 @@ package org.apache.spark.streaming import org.apache.spark.Logging import org.apache.spark.util.Utils -import java.io.File - /** * This testsuite tests master failures at random times while the stream is running using * the real clock. */ class FailureSuite extends TestSuiteBase with Logging { - val directory = Utils.createTempDir().getAbsolutePath + val directory = Utils.createTempDir() val numBatches = 30 override def batchDuration = Milliseconds(1000) @@ -36,16 +34,16 @@ class FailureSuite extends TestSuiteBase with Logging { override def useManualClock = false override def afterFunction() { - Utils.deleteRecursively(new File(directory)) + Utils.deleteRecursively(directory) super.afterFunction() } test("multiple failures with map") { - MasterFailureTest.testMap(directory, numBatches, batchDuration) + MasterFailureTest.testMap(directory.getAbsolutePath, numBatches, batchDuration) } test("multiple failures with updateStateByKey") { - MasterFailureTest.testUpdateStateByKey(directory, numBatches, batchDuration) + MasterFailureTest.testUpdateStateByKey(directory.getAbsolutePath, numBatches, batchDuration) } } diff --git a/streaming/src/test/scala/org/apache/spark/streaming/ReceivedBlockHandlerSuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/ReceivedBlockHandlerSuite.scala index 818f551dbe996..18a477f92094d 100644 --- a/streaming/src/test/scala/org/apache/spark/streaming/ReceivedBlockHandlerSuite.scala +++ b/streaming/src/test/scala/org/apache/spark/streaming/ReceivedBlockHandlerSuite.scala @@ -25,8 +25,6 @@ import scala.concurrent.duration._ import scala.language.postfixOps import akka.actor.{ActorSystem, Props} -import com.google.common.io.Files -import org.apache.commons.io.FileUtils import org.apache.hadoop.conf.Configuration import org.scalatest.{BeforeAndAfter, FunSuite, Matchers} import org.scalatest.concurrent.Eventually._ @@ -39,7 +37,7 @@ import org.apache.spark.shuffle.hash.HashShuffleManager import org.apache.spark.storage._ import org.apache.spark.streaming.receiver._ import org.apache.spark.streaming.util._ -import org.apache.spark.util.{AkkaUtils, ManualClock} +import org.apache.spark.util.{AkkaUtils, ManualClock, Utils} import WriteAheadLogBasedBlockHandler._ import WriteAheadLogSuite._ @@ -76,7 +74,7 @@ class ReceivedBlockHandlerSuite extends FunSuite with BeforeAndAfter with Matche new NioBlockTransferService(conf, securityMgr), securityMgr, 0) blockManager.initialize("app-id") - tempDirectory = Files.createTempDir() + tempDirectory = Utils.createTempDir() manualClock.setTime(0) } @@ -93,10 +91,7 @@ class ReceivedBlockHandlerSuite extends FunSuite with BeforeAndAfter with Matche actorSystem.awaitTermination() actorSystem = null - if (tempDirectory != null && tempDirectory.exists()) { - FileUtils.deleteDirectory(tempDirectory) - tempDirectory = null - } + Utils.deleteRecursively(tempDirectory) } test("BlockManagerBasedBlockHandler - store blocks") { diff --git a/streaming/src/test/scala/org/apache/spark/streaming/ReceivedBlockTrackerSuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/ReceivedBlockTrackerSuite.scala index a3a0fd5187403..42fad769f0c1a 100644 --- a/streaming/src/test/scala/org/apache/spark/streaming/ReceivedBlockTrackerSuite.scala +++ b/streaming/src/test/scala/org/apache/spark/streaming/ReceivedBlockTrackerSuite.scala @@ -24,8 +24,6 @@ import scala.concurrent.duration._ import scala.language.{implicitConversions, postfixOps} import scala.util.Random -import com.google.common.io.Files -import org.apache.commons.io.FileUtils import org.apache.hadoop.conf.Configuration import org.scalatest.{BeforeAndAfter, FunSuite, Matchers} import org.scalatest.concurrent.Eventually._ @@ -51,15 +49,12 @@ class ReceivedBlockTrackerSuite before { conf = new SparkConf().setMaster("local[2]").setAppName("ReceivedBlockTrackerSuite") - checkpointDirectory = Files.createTempDir() + checkpointDirectory = Utils.createTempDir() } after { allReceivedBlockTrackers.foreach { _.stop() } - if (checkpointDirectory != null && checkpointDirectory.exists()) { - FileUtils.deleteDirectory(checkpointDirectory) - checkpointDirectory = null - } + Utils.deleteRecursively(checkpointDirectory) } test("block addition, and block to batch allocation") { diff --git a/streaming/src/test/scala/org/apache/spark/streaming/ReceiverSuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/ReceiverSuite.scala index e8c34a9ee40b9..aa20ad0b5374e 100644 --- a/streaming/src/test/scala/org/apache/spark/streaming/ReceiverSuite.scala +++ b/streaming/src/test/scala/org/apache/spark/streaming/ReceiverSuite.scala @@ -24,7 +24,6 @@ import java.util.concurrent.Semaphore import scala.collection.mutable import scala.collection.mutable.ArrayBuffer -import com.google.common.io.Files import org.scalatest.concurrent.Timeouts import org.scalatest.concurrent.Eventually._ import org.scalatest.time.SpanSugar._ @@ -34,6 +33,7 @@ import org.apache.spark.storage.StorageLevel import org.apache.spark.storage.StreamBlockId import org.apache.spark.streaming.receiver._ import org.apache.spark.streaming.receiver.WriteAheadLogBasedBlockHandler._ +import org.apache.spark.util.Utils /** Testsuite for testing the network receiver behavior */ class ReceiverSuite extends TestSuiteBase with Timeouts with Serializable { @@ -222,7 +222,7 @@ class ReceiverSuite extends TestSuiteBase with Timeouts with Serializable { .set("spark.streaming.receiver.writeAheadLog.enable", "true") .set("spark.streaming.receiver.writeAheadLog.rollingInterval", "1") val batchDuration = Milliseconds(500) - val tempDirectory = Files.createTempDir() + val tempDirectory = Utils.createTempDir() val logDirectory1 = new File(checkpointDirToLogDir(tempDirectory.getAbsolutePath, 0)) val logDirectory2 = new File(checkpointDirToLogDir(tempDirectory.getAbsolutePath, 1)) val allLogFiles1 = new mutable.HashSet[String]() @@ -251,7 +251,6 @@ class ReceiverSuite extends TestSuiteBase with Timeouts with Serializable { } withStreamingContext(new StreamingContext(sparkConf, batchDuration)) { ssc => - tempDirectory.deleteOnExit() val receiver1 = ssc.sparkContext.clean(new FakeReceiver(sendData = true)) val receiver2 = ssc.sparkContext.clean(new FakeReceiver(sendData = true)) val receiverStream1 = ssc.receiverStream(receiver1) diff --git a/yarn/src/test/scala/org/apache/spark/deploy/yarn/YarnSparkHadoopUtilSuite.scala b/yarn/src/test/scala/org/apache/spark/deploy/yarn/YarnSparkHadoopUtilSuite.scala index b5a2db8f6225c..4194f36499e66 100644 --- a/yarn/src/test/scala/org/apache/spark/deploy/yarn/YarnSparkHadoopUtilSuite.scala +++ b/yarn/src/test/scala/org/apache/spark/deploy/yarn/YarnSparkHadoopUtilSuite.scala @@ -50,7 +50,7 @@ class YarnSparkHadoopUtilSuite extends FunSuite with Matchers with Logging { if (hasBash) test(name)(fn) else ignore(name)(fn) bashTest("shell script escaping") { - val scriptFile = File.createTempFile("script.", ".sh") + val scriptFile = File.createTempFile("script.", ".sh", Utils.createTempDir()) val args = Array("arg1", "${arg.2}", "\"arg3\"", "'arg4'", "$arg5", "\\arg6") try { val argLine = args.map(a => YarnSparkHadoopUtil.escapeForShell(a)).mkString(" ") From db4d317ccfdd9bd1dc7e8beac54ebcc35966b7d5 Mon Sep 17 00:00:00 2001 From: Reynold Xin Date: Fri, 20 Mar 2015 14:13:02 -0400 Subject: [PATCH 110/122] [SPARK-6428][MLlib] Added explicit type for public methods and implemented hashCode when equals is defined. I want to add a checker to turn public type checking on, since future pull requests can accidentally expose a non-public type. This is the first cleanup task. Author: Reynold Xin Closes #5102 from rxin/mllib-hashcode-publicmethodtypes and squashes the following commits: 617f19e [Reynold Xin] Fixed Scala compilation error. 52bc2d5 [Reynold Xin] [MLlib] Added explicit type for public methods and implemented hashCode when equals is defined. --- .../spark/examples/mllib/MovieLensALS.scala | 3 +- .../PowerIterationClusteringExample.scala | 4 +-- .../apache/spark/ml/feature/HashingTF.scala | 2 +- .../mllib/api/python/PythonMLLibAPI.scala | 18 ++++++---- .../mllib/classification/NaiveBayes.scala | 6 ++-- .../impl/GLMClassificationModel.scala | 2 +- .../spark/mllib/clustering/KMeans.scala | 2 +- .../mllib/evaluation/MultilabelMetrics.scala | 18 +++++----- .../apache/spark/mllib/linalg/Matrices.scala | 12 +++++-- .../apache/spark/mllib/linalg/Vectors.scala | 4 ++- .../linalg/distributed/BlockMatrix.scala | 10 +++++- .../mllib/random/RandomDataGenerator.scala | 4 +-- .../regression/impl/GLMRegressionModel.scala | 2 +- .../mllib/tree/configuration/Strategy.scala | 9 +++-- .../spark/mllib/tree/impurity/Entropy.scala | 2 +- .../spark/mllib/tree/impurity/Gini.scala | 2 +- .../spark/mllib/tree/impurity/Variance.scala | 2 +- .../mllib/tree/model/DecisionTreeModel.scala | 4 +-- .../tree/model/InformationGainStats.scala | 35 ++++++++++++------- .../apache/spark/mllib/tree/model/Node.scala | 6 ++-- .../spark/mllib/tree/model/Predict.scala | 6 +++- .../apache/spark/mllib/tree/model/Split.scala | 3 +- .../mllib/tree/model/treeEnsembleModels.scala | 6 ++-- 23 files changed, 101 insertions(+), 61 deletions(-) diff --git a/examples/src/main/scala/org/apache/spark/examples/mllib/MovieLensALS.scala b/examples/src/main/scala/org/apache/spark/examples/mllib/MovieLensALS.scala index 91a0a860d6c71..1f4ca4fbe7778 100644 --- a/examples/src/main/scala/org/apache/spark/examples/mllib/MovieLensALS.scala +++ b/examples/src/main/scala/org/apache/spark/examples/mllib/MovieLensALS.scala @@ -175,7 +175,8 @@ object MovieLensALS { } /** Compute RMSE (Root Mean Squared Error). */ - def computeRmse(model: MatrixFactorizationModel, data: RDD[Rating], implicitPrefs: Boolean) = { + def computeRmse(model: MatrixFactorizationModel, data: RDD[Rating], implicitPrefs: Boolean) + : Double = { def mapPredictedRating(r: Double) = if (implicitPrefs) math.max(math.min(r, 1.0), 0.0) else r diff --git a/examples/src/main/scala/org/apache/spark/examples/mllib/PowerIterationClusteringExample.scala b/examples/src/main/scala/org/apache/spark/examples/mllib/PowerIterationClusteringExample.scala index 91c9772744f18..9f22d40c15f3f 100644 --- a/examples/src/main/scala/org/apache/spark/examples/mllib/PowerIterationClusteringExample.scala +++ b/examples/src/main/scala/org/apache/spark/examples/mllib/PowerIterationClusteringExample.scala @@ -116,7 +116,7 @@ object PowerIterationClusteringExample { sc.stop() } - def generateCircle(radius: Double, n: Int) = { + def generateCircle(radius: Double, n: Int): Seq[(Double, Double)] = { Seq.tabulate(n) { i => val theta = 2.0 * math.Pi * i / n (radius * math.cos(theta), radius * math.sin(theta)) @@ -147,7 +147,7 @@ object PowerIterationClusteringExample { /** * Gaussian Similarity: http://en.wikipedia.org/wiki/Radial_basis_function_kernel */ - def gaussianSimilarity(p1: (Double, Double), p2: (Double, Double), sigma: Double) = { + def gaussianSimilarity(p1: (Double, Double), p2: (Double, Double), sigma: Double): Double = { val coeff = 1.0 / (math.sqrt(2.0 * math.Pi) * sigma) val expCoeff = -1.0 / 2.0 * math.pow(sigma, 2.0) val ssquares = (p1._1 - p2._1) * (p1._1 - p2._1) + (p1._2 - p2._2) * (p1._2 - p2._2) diff --git a/mllib/src/main/scala/org/apache/spark/ml/feature/HashingTF.scala b/mllib/src/main/scala/org/apache/spark/ml/feature/HashingTF.scala index 6131ba8832691..fc4e12773c46d 100644 --- a/mllib/src/main/scala/org/apache/spark/ml/feature/HashingTF.scala +++ b/mllib/src/main/scala/org/apache/spark/ml/feature/HashingTF.scala @@ -41,7 +41,7 @@ class HashingTF extends UnaryTransformer[Iterable[_], Vector, HashingTF] { def getNumFeatures: Int = get(numFeatures) /** @group setParam */ - def setNumFeatures(value: Int) = set(numFeatures, value) + def setNumFeatures(value: Int): this.type = set(numFeatures, value) override protected def createTransformFunc(paramMap: ParamMap): Iterable[_] => Vector = { val hashingTF = new feature.HashingTF(paramMap(numFeatures)) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala b/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala index cbd87ea8aeb37..15ca2547d56a8 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala @@ -345,9 +345,13 @@ private[python] class PythonMLLibAPI extends Serializable { def predict(userAndProducts: JavaRDD[Array[Any]]): RDD[Rating] = predict(SerDe.asTupleRDD(userAndProducts.rdd)) - def getUserFeatures = SerDe.fromTuple2RDD(userFeatures.asInstanceOf[RDD[(Any, Any)]]) + def getUserFeatures: RDD[Array[Any]] = { + SerDe.fromTuple2RDD(userFeatures.asInstanceOf[RDD[(Any, Any)]]) + } - def getProductFeatures = SerDe.fromTuple2RDD(productFeatures.asInstanceOf[RDD[(Any, Any)]]) + def getProductFeatures: RDD[Array[Any]] = { + SerDe.fromTuple2RDD(productFeatures.asInstanceOf[RDD[(Any, Any)]]) + } } @@ -909,7 +913,7 @@ private[spark] object SerDe extends Serializable { // Pickler for DenseVector private[python] class DenseVectorPickler extends BasePickler[DenseVector] { - def saveState(obj: Object, out: OutputStream, pickler: Pickler) = { + def saveState(obj: Object, out: OutputStream, pickler: Pickler): Unit = { val vector: DenseVector = obj.asInstanceOf[DenseVector] val bytes = new Array[Byte](8 * vector.size) val bb = ByteBuffer.wrap(bytes) @@ -941,7 +945,7 @@ private[spark] object SerDe extends Serializable { // Pickler for DenseMatrix private[python] class DenseMatrixPickler extends BasePickler[DenseMatrix] { - def saveState(obj: Object, out: OutputStream, pickler: Pickler) = { + def saveState(obj: Object, out: OutputStream, pickler: Pickler): Unit = { val m: DenseMatrix = obj.asInstanceOf[DenseMatrix] val bytes = new Array[Byte](8 * m.values.size) val order = ByteOrder.nativeOrder() @@ -973,7 +977,7 @@ private[spark] object SerDe extends Serializable { // Pickler for SparseVector private[python] class SparseVectorPickler extends BasePickler[SparseVector] { - def saveState(obj: Object, out: OutputStream, pickler: Pickler) = { + def saveState(obj: Object, out: OutputStream, pickler: Pickler): Unit = { val v: SparseVector = obj.asInstanceOf[SparseVector] val n = v.indices.size val indiceBytes = new Array[Byte](4 * n) @@ -1015,7 +1019,7 @@ private[spark] object SerDe extends Serializable { // Pickler for LabeledPoint private[python] class LabeledPointPickler extends BasePickler[LabeledPoint] { - def saveState(obj: Object, out: OutputStream, pickler: Pickler) = { + def saveState(obj: Object, out: OutputStream, pickler: Pickler): Unit = { val point: LabeledPoint = obj.asInstanceOf[LabeledPoint] saveObjects(out, pickler, point.label, point.features) } @@ -1031,7 +1035,7 @@ private[spark] object SerDe extends Serializable { // Pickler for Rating private[python] class RatingPickler extends BasePickler[Rating] { - def saveState(obj: Object, out: OutputStream, pickler: Pickler) = { + def saveState(obj: Object, out: OutputStream, pickler: Pickler): Unit = { val rating: Rating = obj.asInstanceOf[Rating] saveObjects(out, pickler, rating.user, rating.product, rating.rating) } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala b/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala index 2ebc7fa5d4234..068449aa1d346 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala @@ -83,10 +83,10 @@ object NaiveBayesModel extends Loader[NaiveBayesModel] { private object SaveLoadV1_0 { - def thisFormatVersion = "1.0" + def thisFormatVersion: String = "1.0" /** Hard-code class name string in case it changes in the future */ - def thisClassName = "org.apache.spark.mllib.classification.NaiveBayesModel" + def thisClassName: String = "org.apache.spark.mllib.classification.NaiveBayesModel" /** Model data for model import/export */ case class Data(labels: Array[Double], pi: Array[Double], theta: Array[Array[Double]]) @@ -174,7 +174,7 @@ class NaiveBayes private (private var lambda: Double) extends Serializable with * * @param data RDD of [[org.apache.spark.mllib.regression.LabeledPoint]]. */ - def run(data: RDD[LabeledPoint]) = { + def run(data: RDD[LabeledPoint]): NaiveBayesModel = { val requireNonnegativeValues: Vector => Unit = (v: Vector) => { val values = v match { case SparseVector(size, indices, values) => diff --git a/mllib/src/main/scala/org/apache/spark/mllib/classification/impl/GLMClassificationModel.scala b/mllib/src/main/scala/org/apache/spark/mllib/classification/impl/GLMClassificationModel.scala index 8956189ff1158..3b6790cce47c6 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/classification/impl/GLMClassificationModel.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/classification/impl/GLMClassificationModel.scala @@ -32,7 +32,7 @@ private[classification] object GLMClassificationModel { object SaveLoadV1_0 { - def thisFormatVersion = "1.0" + def thisFormatVersion: String = "1.0" /** Model data for import/export */ case class Data(weights: Vector, intercept: Double, threshold: Option[Double]) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeans.scala b/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeans.scala index e41f941fd2c2c..0f8d6a399682d 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeans.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/clustering/KMeans.scala @@ -536,5 +536,5 @@ class VectorWithNorm(val vector: Vector, val norm: Double) extends Serializable def this(array: Array[Double]) = this(Vectors.dense(array)) /** Converts the vector to a dense vector. */ - def toDense = new VectorWithNorm(Vectors.dense(vector.toArray), norm) + def toDense: VectorWithNorm = new VectorWithNorm(Vectors.dense(vector.toArray), norm) } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/evaluation/MultilabelMetrics.scala b/mllib/src/main/scala/org/apache/spark/mllib/evaluation/MultilabelMetrics.scala index ea10bde5fa252..a8378a76d20ae 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/evaluation/MultilabelMetrics.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/evaluation/MultilabelMetrics.scala @@ -96,30 +96,30 @@ class MultilabelMetrics(predictionAndLabels: RDD[(Array[Double], Array[Double])] * Returns precision for a given label (category) * @param label the label. */ - def precision(label: Double) = { + def precision(label: Double): Double = { val tp = tpPerClass(label) val fp = fpPerClass.getOrElse(label, 0L) - if (tp + fp == 0) 0 else tp.toDouble / (tp + fp) + if (tp + fp == 0) 0.0 else tp.toDouble / (tp + fp) } /** * Returns recall for a given label (category) * @param label the label. */ - def recall(label: Double) = { + def recall(label: Double): Double = { val tp = tpPerClass(label) val fn = fnPerClass.getOrElse(label, 0L) - if (tp + fn == 0) 0 else tp.toDouble / (tp + fn) + if (tp + fn == 0) 0.0 else tp.toDouble / (tp + fn) } /** * Returns f1-measure for a given label (category) * @param label the label. */ - def f1Measure(label: Double) = { + def f1Measure(label: Double): Double = { val p = precision(label) val r = recall(label) - if((p + r) == 0) 0 else 2 * p * r / (p + r) + if((p + r) == 0) 0.0 else 2 * p * r / (p + r) } private lazy val sumTp = tpPerClass.foldLeft(0L) { case (sum, (_, tp)) => sum + tp } @@ -130,7 +130,7 @@ class MultilabelMetrics(predictionAndLabels: RDD[(Array[Double], Array[Double])] * Returns micro-averaged label-based precision * (equals to micro-averaged document-based precision) */ - lazy val microPrecision = { + lazy val microPrecision: Double = { val sumFp = fpPerClass.foldLeft(0L){ case(cum, (_, fp)) => cum + fp} sumTp.toDouble / (sumTp + sumFp) } @@ -139,7 +139,7 @@ class MultilabelMetrics(predictionAndLabels: RDD[(Array[Double], Array[Double])] * Returns micro-averaged label-based recall * (equals to micro-averaged document-based recall) */ - lazy val microRecall = { + lazy val microRecall: Double = { val sumFn = fnPerClass.foldLeft(0.0){ case(cum, (_, fn)) => cum + fn} sumTp.toDouble / (sumTp + sumFn) } @@ -148,7 +148,7 @@ class MultilabelMetrics(predictionAndLabels: RDD[(Array[Double], Array[Double])] * Returns micro-averaged label-based f1-measure * (equals to micro-averaged document-based f1-measure) */ - lazy val microF1Measure = 2.0 * sumTp / (2 * sumTp + sumFnClass + sumFpClass) + lazy val microF1Measure: Double = 2.0 * sumTp / (2 * sumTp + sumFnClass + sumFpClass) /** * Returns the sequence of labels in ascending order diff --git a/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala b/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala index 0e4a4d0085895..fdd8848189f19 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala @@ -146,12 +146,16 @@ class DenseMatrix( def this(numRows: Int, numCols: Int, values: Array[Double]) = this(numRows, numCols, values, false) - override def equals(o: Any) = o match { + override def equals(o: Any): Boolean = o match { case m: DenseMatrix => m.numRows == numRows && m.numCols == numCols && Arrays.equals(toArray, m.toArray) case _ => false } + override def hashCode: Int = { + com.google.common.base.Objects.hashCode(numRows : Integer, numCols: Integer, toArray) + } + private[mllib] def toBreeze: BM[Double] = { if (!isTransposed) { new BDM[Double](numRows, numCols, values) @@ -173,7 +177,7 @@ class DenseMatrix( values(index(i, j)) = v } - override def copy = new DenseMatrix(numRows, numCols, values.clone()) + override def copy: DenseMatrix = new DenseMatrix(numRows, numCols, values.clone()) private[mllib] def map(f: Double => Double) = new DenseMatrix(numRows, numCols, values.map(f)) @@ -431,7 +435,9 @@ class SparseMatrix( } } - override def copy = new SparseMatrix(numRows, numCols, colPtrs, rowIndices, values.clone()) + override def copy: SparseMatrix = { + new SparseMatrix(numRows, numCols, colPtrs, rowIndices, values.clone()) + } private[mllib] def map(f: Double => Double) = new SparseMatrix(numRows, numCols, colPtrs, rowIndices, values.map(f)) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/linalg/Vectors.scala b/mllib/src/main/scala/org/apache/spark/mllib/linalg/Vectors.scala index e9d25dcb7e778..2cda9b252ee06 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/linalg/Vectors.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/linalg/Vectors.scala @@ -183,6 +183,8 @@ private[spark] class VectorUDT extends UserDefinedType[Vector] { } } + override def hashCode: Int = 7919 + private[spark] override def asNullable: VectorUDT = this } @@ -478,7 +480,7 @@ class DenseVector(val values: Array[Double]) extends Vector { private[mllib] override def toBreeze: BV[Double] = new BDV[Double](values) - override def apply(i: Int) = values(i) + override def apply(i: Int): Double = values(i) override def copy: DenseVector = { new DenseVector(values.clone()) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/BlockMatrix.scala b/mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/BlockMatrix.scala index 1d253963130f1..3323ae7b1fba0 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/BlockMatrix.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/BlockMatrix.scala @@ -49,7 +49,7 @@ private[mllib] class GridPartitioner( private val rowPartitions = math.ceil(rows * 1.0 / rowsPerPart).toInt private val colPartitions = math.ceil(cols * 1.0 / colsPerPart).toInt - override val numPartitions = rowPartitions * colPartitions + override val numPartitions: Int = rowPartitions * colPartitions /** * Returns the index of the partition the input coordinate belongs to. @@ -85,6 +85,14 @@ private[mllib] class GridPartitioner( false } } + + override def hashCode: Int = { + com.google.common.base.Objects.hashCode( + rows: java.lang.Integer, + cols: java.lang.Integer, + rowsPerPart: java.lang.Integer, + colsPerPart: java.lang.Integer) + } } private[mllib] object GridPartitioner { diff --git a/mllib/src/main/scala/org/apache/spark/mllib/random/RandomDataGenerator.scala b/mllib/src/main/scala/org/apache/spark/mllib/random/RandomDataGenerator.scala index 405bae62ee8b6..9349ecaa13f56 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/random/RandomDataGenerator.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/random/RandomDataGenerator.scala @@ -56,7 +56,7 @@ class UniformGenerator extends RandomDataGenerator[Double] { random.nextDouble() } - override def setSeed(seed: Long) = random.setSeed(seed) + override def setSeed(seed: Long): Unit = random.setSeed(seed) override def copy(): UniformGenerator = new UniformGenerator() } @@ -75,7 +75,7 @@ class StandardNormalGenerator extends RandomDataGenerator[Double] { random.nextGaussian() } - override def setSeed(seed: Long) = random.setSeed(seed) + override def setSeed(seed: Long): Unit = random.setSeed(seed) override def copy(): StandardNormalGenerator = new StandardNormalGenerator() } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/regression/impl/GLMRegressionModel.scala b/mllib/src/main/scala/org/apache/spark/mllib/regression/impl/GLMRegressionModel.scala index bd7e340ca2d8e..b55944f74f623 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/regression/impl/GLMRegressionModel.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/regression/impl/GLMRegressionModel.scala @@ -32,7 +32,7 @@ private[regression] object GLMRegressionModel { object SaveLoadV1_0 { - def thisFormatVersion = "1.0" + def thisFormatVersion: String = "1.0" /** Model data for model import/export */ case class Data(weights: Vector, intercept: Double) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/configuration/Strategy.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/configuration/Strategy.scala index 8d5c36da32bdb..ada227c200a79 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/configuration/Strategy.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/configuration/Strategy.scala @@ -83,10 +83,13 @@ class Strategy ( @BeanProperty var useNodeIdCache: Boolean = false, @BeanProperty var checkpointInterval: Int = 10) extends Serializable { - def isMulticlassClassification = + def isMulticlassClassification: Boolean = { algo == Classification && numClasses > 2 - def isMulticlassWithCategoricalFeatures - = isMulticlassClassification && (categoricalFeaturesInfo.size > 0) + } + + def isMulticlassWithCategoricalFeatures: Boolean = { + isMulticlassClassification && (categoricalFeaturesInfo.size > 0) + } /** * Java-friendly constructor for [[org.apache.spark.mllib.tree.configuration.Strategy]] diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Entropy.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Entropy.scala index b7950e00786ab..5ac10f3fd32dd 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Entropy.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Entropy.scala @@ -71,7 +71,7 @@ object Entropy extends Impurity { * Get this impurity instance. * This is useful for passing impurity parameters to a Strategy in Java. */ - def instance = this + def instance: this.type = this } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Gini.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Gini.scala index c946db9c0d1c8..19d318203c344 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Gini.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Gini.scala @@ -67,7 +67,7 @@ object Gini extends Impurity { * Get this impurity instance. * This is useful for passing impurity parameters to a Strategy in Java. */ - def instance = this + def instance: this.type = this } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Variance.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Variance.scala index df9eafa5da16a..7104a7fa4dd4c 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Variance.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/impurity/Variance.scala @@ -58,7 +58,7 @@ object Variance extends Impurity { * Get this impurity instance. * This is useful for passing impurity parameters to a Strategy in Java. */ - def instance = this + def instance: this.type = this } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/DecisionTreeModel.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/DecisionTreeModel.scala index 8a57ebc387d01..c9bafd60fba4d 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/DecisionTreeModel.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/DecisionTreeModel.scala @@ -120,10 +120,10 @@ object DecisionTreeModel extends Loader[DecisionTreeModel] with Logging { private[tree] object SaveLoadV1_0 { - def thisFormatVersion = "1.0" + def thisFormatVersion: String = "1.0" // Hard-code class name string in case it changes in the future - def thisClassName = "org.apache.spark.mllib.tree.DecisionTreeModel" + def thisClassName: String = "org.apache.spark.mllib.tree.DecisionTreeModel" case class PredictData(predict: Double, prob: Double) { def toPredict: Predict = new Predict(predict, prob) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/InformationGainStats.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/InformationGainStats.scala index 80990aa9a603f..f209fdafd3653 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/InformationGainStats.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/InformationGainStats.scala @@ -38,23 +38,32 @@ class InformationGainStats( val leftPredict: Predict, val rightPredict: Predict) extends Serializable { - override def toString = { + override def toString: String = { "gain = %f, impurity = %f, left impurity = %f, right impurity = %f" .format(gain, impurity, leftImpurity, rightImpurity) } - override def equals(o: Any) = - o match { - case other: InformationGainStats => { - gain == other.gain && - impurity == other.impurity && - leftImpurity == other.leftImpurity && - rightImpurity == other.rightImpurity && - leftPredict == other.leftPredict && - rightPredict == other.rightPredict - } - case _ => false - } + override def equals(o: Any): Boolean = o match { + case other: InformationGainStats => + gain == other.gain && + impurity == other.impurity && + leftImpurity == other.leftImpurity && + rightImpurity == other.rightImpurity && + leftPredict == other.leftPredict && + rightPredict == other.rightPredict + + case _ => false + } + + override def hashCode: Int = { + com.google.common.base.Objects.hashCode( + gain: java.lang.Double, + impurity: java.lang.Double, + leftImpurity: java.lang.Double, + rightImpurity: java.lang.Double, + leftPredict, + rightPredict) + } } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Node.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Node.scala index d961081d185e9..4f72bb8014cc0 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Node.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Node.scala @@ -50,8 +50,10 @@ class Node ( var rightNode: Option[Node], var stats: Option[InformationGainStats]) extends Serializable with Logging { - override def toString = "id = " + id + ", isLeaf = " + isLeaf + ", predict = " + predict + ", " + - "impurity = " + impurity + "split = " + split + ", stats = " + stats + override def toString: String = { + "id = " + id + ", isLeaf = " + isLeaf + ", predict = " + predict + ", " + + "impurity = " + impurity + "split = " + split + ", stats = " + stats + } /** * build the left node and right nodes if not leaf diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Predict.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Predict.scala index ad4c0dbbfb3e5..25990af7c6cf7 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Predict.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Predict.scala @@ -29,7 +29,7 @@ class Predict( val predict: Double, val prob: Double = 0.0) extends Serializable { - override def toString = { + override def toString: String = { "predict = %f, prob = %f".format(predict, prob) } @@ -39,4 +39,8 @@ class Predict( case _ => false } } + + override def hashCode: Int = { + com.google.common.base.Objects.hashCode(predict: java.lang.Double, prob: java.lang.Double) + } } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Split.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Split.scala index b7a85f58544a3..fb35e70a8d077 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Split.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/Split.scala @@ -38,9 +38,10 @@ case class Split( featureType: FeatureType, categories: List[Double]) { - override def toString = + override def toString: String = { "Feature = " + feature + ", threshold = " + threshold + ", featureType = " + featureType + ", categories = " + categories + } } /** diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala index 30a8f7ca301af..f160852c69c77 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala @@ -79,7 +79,7 @@ object RandomForestModel extends Loader[RandomForestModel] { private object SaveLoadV1_0 { // Hard-code class name string in case it changes in the future - def thisClassName = "org.apache.spark.mllib.tree.model.RandomForestModel" + def thisClassName: String = "org.apache.spark.mllib.tree.model.RandomForestModel" } } @@ -130,7 +130,7 @@ object GradientBoostedTreesModel extends Loader[GradientBoostedTreesModel] { private object SaveLoadV1_0 { // Hard-code class name string in case it changes in the future - def thisClassName = "org.apache.spark.mllib.tree.model.GradientBoostedTreesModel" + def thisClassName: String = "org.apache.spark.mllib.tree.model.GradientBoostedTreesModel" } } @@ -257,7 +257,7 @@ private[tree] object TreeEnsembleModel extends Logging { import org.apache.spark.mllib.tree.model.DecisionTreeModel.SaveLoadV1_0.{NodeData, constructTrees} - def thisFormatVersion = "1.0" + def thisFormatVersion: String = "1.0" case class Metadata( algo: String, From 28bcb9e9e86a4b643fbf96b2b7e03928ebcfc060 Mon Sep 17 00:00:00 2001 From: mbonaci Date: Fri, 20 Mar 2015 18:30:45 +0000 Subject: [PATCH 111/122] [SPARK-6370][core] Documentation: Improve all 3 docs for RDD.sample The docs for the `sample` method were insufficient, now less so. Author: mbonaci Closes #5097 from mbonaci/master and squashes the following commits: a6a9d97 [mbonaci] [SPARK-6370][core] Documentation: Improve all 3 docs for RDD.sample method --- .../scala/org/apache/spark/api/java/JavaRDD.scala | 11 +++++++++++ core/src/main/scala/org/apache/spark/rdd/RDD.scala | 6 ++++++ python/pyspark/rdd.py | 6 ++++++ 3 files changed, 23 insertions(+) diff --git a/core/src/main/scala/org/apache/spark/api/java/JavaRDD.scala b/core/src/main/scala/org/apache/spark/api/java/JavaRDD.scala index 645dc3bfb6b06..3e9beb670f7ad 100644 --- a/core/src/main/scala/org/apache/spark/api/java/JavaRDD.scala +++ b/core/src/main/scala/org/apache/spark/api/java/JavaRDD.scala @@ -101,12 +101,23 @@ class JavaRDD[T](val rdd: RDD[T])(implicit val classTag: ClassTag[T]) /** * Return a sampled subset of this RDD. + * + * @param withReplacement can elements be sampled multiple times (replaced when sampled out) + * @param fraction expected size of the sample as a fraction of this RDD's size + * without replacement: probability that each element is chosen; fraction must be [0, 1] + * with replacement: expected number of times each element is chosen; fraction must be >= 0 */ def sample(withReplacement: Boolean, fraction: Double): JavaRDD[T] = sample(withReplacement, fraction, Utils.random.nextLong) /** * Return a sampled subset of this RDD. + * + * @param withReplacement can elements be sampled multiple times (replaced when sampled out) + * @param fraction expected size of the sample as a fraction of this RDD's size + * without replacement: probability that each element is chosen; fraction must be [0, 1] + * with replacement: expected number of times each element is chosen; fraction must be >= 0 + * @param seed seed for the random number generator */ def sample(withReplacement: Boolean, fraction: Double, seed: Long): JavaRDD[T] = wrapRDD(rdd.sample(withReplacement, fraction, seed)) diff --git a/core/src/main/scala/org/apache/spark/rdd/RDD.scala b/core/src/main/scala/org/apache/spark/rdd/RDD.scala index a139780d967e9..a4c74ed03e330 100644 --- a/core/src/main/scala/org/apache/spark/rdd/RDD.scala +++ b/core/src/main/scala/org/apache/spark/rdd/RDD.scala @@ -377,6 +377,12 @@ abstract class RDD[T: ClassTag]( /** * Return a sampled subset of this RDD. + * + * @param withReplacement can elements be sampled multiple times (replaced when sampled out) + * @param fraction expected size of the sample as a fraction of this RDD's size + * without replacement: probability that each element is chosen; fraction must be [0, 1] + * with replacement: expected number of times each element is chosen; fraction must be >= 0 + * @param seed seed for the random number generator */ def sample(withReplacement: Boolean, fraction: Double, diff --git a/python/pyspark/rdd.py b/python/pyspark/rdd.py index bf17f513c0bc3..c337a43c8a7fc 100644 --- a/python/pyspark/rdd.py +++ b/python/pyspark/rdd.py @@ -346,6 +346,12 @@ def sample(self, withReplacement, fraction, seed=None): """ Return a sampled subset of this RDD. + :param withReplacement: can elements be sampled multiple times (replaced when sampled out) + :param fraction: expected size of the sample as a fraction of this RDD's size + without replacement: probability that each element is chosen; fraction must be [0, 1] + with replacement: expected number of times each element is chosen; fraction must be >= 0 + :param seed: seed for the random number generator + >>> rdd = sc.parallelize(range(100), 4) >>> rdd.sample(False, 0.1, 81).count() 10 From 385b2ff10d9ef5083df49233f77c8e873561dc16 Mon Sep 17 00:00:00 2001 From: WangTaoTheTonic Date: Fri, 20 Mar 2015 18:42:18 +0000 Subject: [PATCH 112/122] [SPARK-6426][Doc]User could also point the yarn cluster config directory via YARN_CONF_DI... ...R https://issues.apache.org/jira/browse/SPARK-6426 Author: WangTaoTheTonic Closes #5103 from WangTaoTheTonic/SPARK-6426 and squashes the following commits: e6dd78d [WangTaoTheTonic] User could also point the yarn cluster config directory via YARN_CONF_DIR --- docs/submitting-applications.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/submitting-applications.md b/docs/submitting-applications.md index 57b074778f2b0..3ecbf2308cd44 100644 --- a/docs/submitting-applications.md +++ b/docs/submitting-applications.md @@ -133,10 +133,10 @@ The master URL passed to Spark can be in one of the following formats: Or, for a Mesos cluster using ZooKeeper, use mesos://zk://.... yarn-client Connect to a YARN cluster in -client mode. The cluster location will be found based on the HADOOP_CONF_DIR variable. +client mode. The cluster location will be found based on the HADOOP_CONF_DIR or YARN_CONF_DIR variable. yarn-cluster Connect to a YARN cluster in -cluster mode. The cluster location will be found based on HADOOP_CONF_DIR. +cluster mode. The cluster location will be found based on the HADOOP_CONF_DIR or YARN_CONF_DIR variable. From a74564591f1c824f9eed516ae79e079b355fd32b Mon Sep 17 00:00:00 2001 From: Marcelo Vanzin Date: Fri, 20 Mar 2015 18:43:57 +0000 Subject: [PATCH 113/122] [SPARK-6371] [build] Update version to 1.4.0-SNAPSHOT. Author: Marcelo Vanzin Closes #5056 from vanzin/SPARK-6371 and squashes the following commits: 63220df [Marcelo Vanzin] Merge branch 'master' into SPARK-6371 6506f75 [Marcelo Vanzin] Use more fine-grained exclusion. 178ba71 [Marcelo Vanzin] Oops. 75b2375 [Marcelo Vanzin] Exclude VertexRDD in MiMA. a45a62c [Marcelo Vanzin] Work around MIMA warning. 1d8a670 [Marcelo Vanzin] Re-group jetty exclusion. 0e8e909 [Marcelo Vanzin] Ignore ml, don't ignore graphx. cef4603 [Marcelo Vanzin] Indentation. 296cf82 [Marcelo Vanzin] [SPARK-6371] [build] Update version to 1.4.0-SNAPSHOT. --- assembly/pom.xml | 2 +- bagel/pom.xml | 2 +- core/pom.xml | 2 +- core/src/main/scala/org/apache/spark/package.scala | 2 +- docs/_config.yml | 4 ++-- examples/pom.xml | 2 +- external/flume-sink/pom.xml | 2 +- external/flume/pom.xml | 2 +- external/kafka-assembly/pom.xml | 2 +- external/kafka/pom.xml | 2 +- external/mqtt/pom.xml | 2 +- external/twitter/pom.xml | 2 +- external/zeromq/pom.xml | 2 +- extras/java8-tests/pom.xml | 2 +- extras/kinesis-asl/pom.xml | 2 +- extras/spark-ganglia-lgpl/pom.xml | 2 +- graphx/pom.xml | 2 +- launcher/pom.xml | 2 +- mllib/pom.xml | 2 +- network/common/pom.xml | 2 +- network/shuffle/pom.xml | 2 +- network/yarn/pom.xml | 2 +- pom.xml | 2 +- project/MimaBuild.scala | 2 +- project/MimaExcludes.scala | 14 ++++++++++++++ repl/pom.xml | 2 +- sql/catalyst/pom.xml | 2 +- sql/core/pom.xml | 2 +- sql/hive-thriftserver/pom.xml | 2 +- sql/hive/pom.xml | 2 +- streaming/pom.xml | 2 +- tools/pom.xml | 2 +- yarn/pom.xml | 2 +- 33 files changed, 47 insertions(+), 33 deletions(-) diff --git a/assembly/pom.xml b/assembly/pom.xml index d3bb4bde0c412..f1f8b0d3682e2 100644 --- a/assembly/pom.xml +++ b/assembly/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../pom.xml diff --git a/bagel/pom.xml b/bagel/pom.xml index 1fe61062b4606..1f3dec91314f2 100644 --- a/bagel/pom.xml +++ b/bagel/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../pom.xml diff --git a/core/pom.xml b/core/pom.xml index 81f8cba711df6..6cd1965ec37c2 100644 --- a/core/pom.xml +++ b/core/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../pom.xml diff --git a/core/src/main/scala/org/apache/spark/package.scala b/core/src/main/scala/org/apache/spark/package.scala index b6249b492150a..2ab41ba488ff6 100644 --- a/core/src/main/scala/org/apache/spark/package.scala +++ b/core/src/main/scala/org/apache/spark/package.scala @@ -43,5 +43,5 @@ package org.apache package object spark { // For package docs only - val SPARK_VERSION = "1.3.0-SNAPSHOT" + val SPARK_VERSION = "1.4.0-SNAPSHOT" } diff --git a/docs/_config.yml b/docs/_config.yml index 0652927a8ce9b..b22b627f09007 100644 --- a/docs/_config.yml +++ b/docs/_config.yml @@ -14,8 +14,8 @@ include: # These allow the documentation to be updated with newer releases # of Spark, Scala, and Mesos. -SPARK_VERSION: 1.3.0-SNAPSHOT -SPARK_VERSION_SHORT: 1.3.0 +SPARK_VERSION: 1.4.0-SNAPSHOT +SPARK_VERSION_SHORT: 1.4.0 SCALA_BINARY_VERSION: "2.10" SCALA_VERSION: "2.10.4" MESOS_VERSION: 0.21.0 diff --git a/examples/pom.xml b/examples/pom.xml index 994071d94d0ad..7e93f0eec0b91 100644 --- a/examples/pom.xml +++ b/examples/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../pom.xml diff --git a/external/flume-sink/pom.xml b/external/flume-sink/pom.xml index 96c2787e35cd0..67907bbfb6d1b 100644 --- a/external/flume-sink/pom.xml +++ b/external/flume-sink/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/external/flume/pom.xml b/external/flume/pom.xml index 172d447b77cda..8df7edbdcad33 100644 --- a/external/flume/pom.xml +++ b/external/flume/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/external/kafka-assembly/pom.xml b/external/kafka-assembly/pom.xml index 5109b8ed87524..0b79f47647f6b 100644 --- a/external/kafka-assembly/pom.xml +++ b/external/kafka-assembly/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/external/kafka/pom.xml b/external/kafka/pom.xml index 369856187a244..f695cff410a18 100644 --- a/external/kafka/pom.xml +++ b/external/kafka/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/external/mqtt/pom.xml b/external/mqtt/pom.xml index a344f000c5002..98f95a9a64fa0 100644 --- a/external/mqtt/pom.xml +++ b/external/mqtt/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/external/twitter/pom.xml b/external/twitter/pom.xml index e95853f005ce2..8b6a8959ac4cf 100644 --- a/external/twitter/pom.xml +++ b/external/twitter/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/external/zeromq/pom.xml b/external/zeromq/pom.xml index 9b3475d7c3dc2..a50d378b34335 100644 --- a/external/zeromq/pom.xml +++ b/external/zeromq/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/extras/java8-tests/pom.xml b/extras/java8-tests/pom.xml index bc2f8be10c9ce..4351a8a12fe21 100644 --- a/extras/java8-tests/pom.xml +++ b/extras/java8-tests/pom.xml @@ -20,7 +20,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/extras/kinesis-asl/pom.xml b/extras/kinesis-asl/pom.xml index 7e49a71907336..25847a1b33d9c 100644 --- a/extras/kinesis-asl/pom.xml +++ b/extras/kinesis-asl/pom.xml @@ -20,7 +20,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/extras/spark-ganglia-lgpl/pom.xml b/extras/spark-ganglia-lgpl/pom.xml index 6eb29af03f833..e14bbae4a9b6e 100644 --- a/extras/spark-ganglia-lgpl/pom.xml +++ b/extras/spark-ganglia-lgpl/pom.xml @@ -20,7 +20,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/graphx/pom.xml b/graphx/pom.xml index c0d534e185d7f..d38a3aa8256b7 100644 --- a/graphx/pom.xml +++ b/graphx/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../pom.xml diff --git a/launcher/pom.xml b/launcher/pom.xml index ccbd9d0419a98..0fe2814135d88 100644 --- a/launcher/pom.xml +++ b/launcher/pom.xml @@ -22,7 +22,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../pom.xml diff --git a/mllib/pom.xml b/mllib/pom.xml index a76704a8c2c59..4c183543e3fa8 100644 --- a/mllib/pom.xml +++ b/mllib/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../pom.xml diff --git a/network/common/pom.xml b/network/common/pom.xml index 74437f37c47e4..7b51845206f4a 100644 --- a/network/common/pom.xml +++ b/network/common/pom.xml @@ -22,7 +22,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/network/shuffle/pom.xml b/network/shuffle/pom.xml index a2bcca26d8344..7dc7c65825e34 100644 --- a/network/shuffle/pom.xml +++ b/network/shuffle/pom.xml @@ -22,7 +22,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/network/yarn/pom.xml b/network/yarn/pom.xml index cea7a20c223e2..1e2e9c80af6cc 100644 --- a/network/yarn/pom.xml +++ b/network/yarn/pom.xml @@ -22,7 +22,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/pom.xml b/pom.xml index efb9f172f4751..23bb16130b504 100644 --- a/pom.xml +++ b/pom.xml @@ -26,7 +26,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT pom Spark Project Parent POM http://spark.apache.org/ diff --git a/project/MimaBuild.scala b/project/MimaBuild.scala index f0cbf4e57b8c5..dde92949fa175 100644 --- a/project/MimaBuild.scala +++ b/project/MimaBuild.scala @@ -91,7 +91,7 @@ object MimaBuild { def mimaSettings(sparkHome: File, projectRef: ProjectRef) = { val organization = "org.apache.spark" - val previousSparkVersion = "1.2.0" + val previousSparkVersion = "1.3.0" val fullId = "spark-" + projectRef.project + "_2.10" mimaDefaultSettings ++ Seq(previousArtifact := Some(organization % fullId % previousSparkVersion), diff --git a/project/MimaExcludes.scala b/project/MimaExcludes.scala index a6b07fa7cddec..328d59485a731 100644 --- a/project/MimaExcludes.scala +++ b/project/MimaExcludes.scala @@ -16,6 +16,7 @@ */ import com.typesafe.tools.mima.core._ +import com.typesafe.tools.mima.core.ProblemFilters._ /** * Additional excludes for checking of Spark's binary compatibility. @@ -33,6 +34,19 @@ import com.typesafe.tools.mima.core._ object MimaExcludes { def excludes(version: String) = version match { + case v if v.startsWith("1.4") => + Seq( + MimaBuild.excludeSparkPackage("deploy"), + MimaBuild.excludeSparkPackage("ml"), + // SPARK-5922 Adding a generalized diff(other: RDD[(VertexId, VD)]) to VertexRDD + ProblemFilters.exclude[MissingMethodProblem]("org.apache.spark.graphx.VertexRDD.diff"), + // These are needed if checking against the sbt build, since they are part of + // the maven-generated artifacts in 1.3. + excludePackage("org.spark-project.jetty"), + MimaBuild.excludeSparkPackage("unused"), + ProblemFilters.exclude[MissingClassProblem]("com.google.common.base.Optional") + ) + case v if v.startsWith("1.3") => Seq( MimaBuild.excludeSparkPackage("deploy"), diff --git a/repl/pom.xml b/repl/pom.xml index 295f88ea3ecf9..edfa1c7f2c29c 100644 --- a/repl/pom.xml +++ b/repl/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../pom.xml diff --git a/sql/catalyst/pom.xml b/sql/catalyst/pom.xml index 8ad026dbdf8ff..3dea2ee76542f 100644 --- a/sql/catalyst/pom.xml +++ b/sql/catalyst/pom.xml @@ -22,7 +22,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/sql/core/pom.xml b/sql/core/pom.xml index 3640104e497d4..e3a6b1fe72435 100644 --- a/sql/core/pom.xml +++ b/sql/core/pom.xml @@ -22,7 +22,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/sql/hive-thriftserver/pom.xml b/sql/hive-thriftserver/pom.xml index f466a3c0b5dc2..a96b1ffc26966 100644 --- a/sql/hive-thriftserver/pom.xml +++ b/sql/hive-thriftserver/pom.xml @@ -22,7 +22,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/sql/hive/pom.xml b/sql/hive/pom.xml index 0e3f4eb98cbf7..a9816f6c38cd2 100644 --- a/sql/hive/pom.xml +++ b/sql/hive/pom.xml @@ -22,7 +22,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../../pom.xml diff --git a/streaming/pom.xml b/streaming/pom.xml index 96508d83f4049..23a8358d45c2a 100644 --- a/streaming/pom.xml +++ b/streaming/pom.xml @@ -21,7 +21,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../pom.xml diff --git a/tools/pom.xml b/tools/pom.xml index 181236d1bcbf6..1c6f3e83a1819 100644 --- a/tools/pom.xml +++ b/tools/pom.xml @@ -20,7 +20,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../pom.xml diff --git a/yarn/pom.xml b/yarn/pom.xml index c13534f0410a1..7c8c3613e7a05 100644 --- a/yarn/pom.xml +++ b/yarn/pom.xml @@ -20,7 +20,7 @@ org.apache.spark spark-parent_2.10 - 1.3.0-SNAPSHOT + 1.4.0-SNAPSHOT ../pom.xml From 48866f789712b0cdbaf76054d1014c6df032fff1 Mon Sep 17 00:00:00 2001 From: Yanbo Liang Date: Fri, 20 Mar 2015 14:44:21 -0400 Subject: [PATCH 114/122] [SPARK-6095] [MLLIB] Support model save/load in Python's linear models For Python's linear models, weights and intercept are stored in Python. This PR implements Python's linear models sava/load functions which do the same thing as scala. It can also make model import/export cross languages. Author: Yanbo Liang Closes #5016 from yanboliang/spark-6095 and squashes the following commits: d9bb824 [Yanbo Liang] fix python style b3813ca [Yanbo Liang] linear model save/load for Python reuse the Scala implementation --- python/pyspark/mllib/classification.py | 58 +++++++++++++++++- python/pyspark/mllib/regression.py | 84 +++++++++++++++++++++++++- python/pyspark/mllib/util.py | 6 +- 3 files changed, 145 insertions(+), 3 deletions(-) diff --git a/python/pyspark/mllib/classification.py b/python/pyspark/mllib/classification.py index e4765173709e8..b66159c5bfb66 100644 --- a/python/pyspark/mllib/classification.py +++ b/python/pyspark/mllib/classification.py @@ -21,7 +21,7 @@ from numpy import array from pyspark import RDD -from pyspark.mllib.common import callMLlibFunc +from pyspark.mllib.common import callMLlibFunc, _py2java, _java2py from pyspark.mllib.linalg import SparseVector, _convert_to_vector from pyspark.mllib.regression import LabeledPoint, LinearModel, _regression_train_wrapper @@ -99,6 +99,18 @@ class LogisticRegressionModel(LinearBinaryClassificationModel): 1 >>> lrm.predict(SparseVector(2, {0: 1.0})) 0 + >>> import os, tempfile + >>> path = tempfile.mkdtemp() + >>> lrm.save(sc, path) + >>> sameModel = LogisticRegressionModel.load(sc, path) + >>> sameModel.predict(array([0.0, 1.0])) + 1 + >>> sameModel.predict(SparseVector(2, {0: 1.0})) + 0 + >>> try: + ... os.removedirs(path) + ... except: + ... pass """ def __init__(self, weights, intercept): super(LogisticRegressionModel, self).__init__(weights, intercept) @@ -124,6 +136,22 @@ def predict(self, x): else: return 1 if prob > self._threshold else 0 + def save(self, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.classification.LogisticRegressionModel( + _py2java(sc, self._coeff), self.intercept) + java_model.save(sc._jsc.sc(), path) + + @classmethod + def load(cls, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.classification.LogisticRegressionModel.load( + sc._jsc.sc(), path) + weights = _java2py(sc, java_model.weights()) + intercept = java_model.intercept() + threshold = java_model.getThreshold().get() + model = LogisticRegressionModel(weights, intercept) + model.setThreshold(threshold) + return model + class LogisticRegressionWithSGD(object): @@ -243,6 +271,18 @@ class SVMModel(LinearBinaryClassificationModel): 1 >>> svm.predict(SparseVector(2, {0: -1.0})) 0 + >>> import os, tempfile + >>> path = tempfile.mkdtemp() + >>> svm.save(sc, path) + >>> sameModel = SVMModel.load(sc, path) + >>> sameModel.predict(SparseVector(2, {1: 1.0})) + 1 + >>> sameModel.predict(SparseVector(2, {0: -1.0})) + 0 + >>> try: + ... os.removedirs(path) + ... except: + ... pass """ def __init__(self, weights, intercept): super(SVMModel, self).__init__(weights, intercept) @@ -263,6 +303,22 @@ def predict(self, x): else: return 1 if margin > self._threshold else 0 + def save(self, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.classification.SVMModel( + _py2java(sc, self._coeff), self.intercept) + java_model.save(sc._jsc.sc(), path) + + @classmethod + def load(cls, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.classification.SVMModel.load( + sc._jsc.sc(), path) + weights = _java2py(sc, java_model.weights()) + intercept = java_model.intercept() + threshold = java_model.getThreshold().get() + model = SVMModel(weights, intercept) + model.setThreshold(threshold) + return model + class SVMWithSGD(object): diff --git a/python/pyspark/mllib/regression.py b/python/pyspark/mllib/regression.py index 0c21ad578793f..015a7860116c9 100644 --- a/python/pyspark/mllib/regression.py +++ b/python/pyspark/mllib/regression.py @@ -18,8 +18,9 @@ import numpy as np from numpy import array -from pyspark.mllib.common import callMLlibFunc, inherit_doc +from pyspark.mllib.common import callMLlibFunc, _py2java, _java2py, inherit_doc from pyspark.mllib.linalg import SparseVector, _convert_to_vector +from pyspark.mllib.util import Saveable, Loader __all__ = ['LabeledPoint', 'LinearModel', 'LinearRegressionModel', 'LinearRegressionWithSGD', @@ -114,6 +115,20 @@ class LinearRegressionModel(LinearRegressionModelBase): True >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 True + >>> import os, tempfile + >>> path = tempfile.mkdtemp() + >>> lrm.save(sc, path) + >>> sameModel = LinearRegressionModel.load(sc, path) + >>> abs(sameModel.predict(np.array([0.0])) - 0) < 0.5 + True + >>> abs(sameModel.predict(np.array([1.0])) - 1) < 0.5 + True + >>> abs(sameModel.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 + True + >>> try: + ... os.removedirs(path) + ... except: + ... pass >>> data = [ ... LabeledPoint(0.0, SparseVector(1, {0: 0.0})), ... LabeledPoint(1.0, SparseVector(1, {0: 1.0})), @@ -126,6 +141,19 @@ class LinearRegressionModel(LinearRegressionModelBase): >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 True """ + def save(self, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.regression.LinearRegressionModel( + _py2java(sc, self._coeff), self.intercept) + java_model.save(sc._jsc.sc(), path) + + @classmethod + def load(cls, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.regression.LinearRegressionModel.load( + sc._jsc.sc(), path) + weights = _java2py(sc, java_model.weights()) + intercept = java_model.intercept() + model = LinearRegressionModel(weights, intercept) + return model # train_func should take two parameters, namely data and initial_weights, and @@ -199,6 +227,20 @@ class LassoModel(LinearRegressionModelBase): True >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 True + >>> import os, tempfile + >>> path = tempfile.mkdtemp() + >>> lrm.save(sc, path) + >>> sameModel = LassoModel.load(sc, path) + >>> abs(sameModel.predict(np.array([0.0])) - 0) < 0.5 + True + >>> abs(sameModel.predict(np.array([1.0])) - 1) < 0.5 + True + >>> abs(sameModel.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 + True + >>> try: + ... os.removedirs(path) + ... except: + ... pass >>> data = [ ... LabeledPoint(0.0, SparseVector(1, {0: 0.0})), ... LabeledPoint(1.0, SparseVector(1, {0: 1.0})), @@ -211,6 +253,19 @@ class LassoModel(LinearRegressionModelBase): >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 True """ + def save(self, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.regression.LassoModel( + _py2java(sc, self._coeff), self.intercept) + java_model.save(sc._jsc.sc(), path) + + @classmethod + def load(cls, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.regression.LassoModel.load( + sc._jsc.sc(), path) + weights = _java2py(sc, java_model.weights()) + intercept = java_model.intercept() + model = LassoModel(weights, intercept) + return model class LassoWithSGD(object): @@ -246,6 +301,20 @@ class RidgeRegressionModel(LinearRegressionModelBase): True >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 True + >>> import os, tempfile + >>> path = tempfile.mkdtemp() + >>> lrm.save(sc, path) + >>> sameModel = RidgeRegressionModel.load(sc, path) + >>> abs(sameModel.predict(np.array([0.0])) - 0) < 0.5 + True + >>> abs(sameModel.predict(np.array([1.0])) - 1) < 0.5 + True + >>> abs(sameModel.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 + True + >>> try: + ... os.removedirs(path) + ... except: + ... pass >>> data = [ ... LabeledPoint(0.0, SparseVector(1, {0: 0.0})), ... LabeledPoint(1.0, SparseVector(1, {0: 1.0})), @@ -258,6 +327,19 @@ class RidgeRegressionModel(LinearRegressionModelBase): >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 True """ + def save(self, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.regression.RidgeRegressionModel( + _py2java(sc, self._coeff), self.intercept) + java_model.save(sc._jsc.sc(), path) + + @classmethod + def load(cls, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.regression.RidgeRegressionModel.load( + sc._jsc.sc(), path) + weights = _java2py(sc, java_model.weights()) + intercept = java_model.intercept() + model = RidgeRegressionModel(weights, intercept) + return model class RidgeRegressionWithSGD(object): diff --git a/python/pyspark/mllib/util.py b/python/pyspark/mllib/util.py index e877c720ac77a..c5c3468eb95e9 100644 --- a/python/pyspark/mllib/util.py +++ b/python/pyspark/mllib/util.py @@ -20,7 +20,6 @@ from pyspark.mllib.common import callMLlibFunc, JavaModelWrapper, inherit_doc from pyspark.mllib.linalg import Vectors, SparseVector, _convert_to_vector -from pyspark.mllib.regression import LabeledPoint class MLUtils(object): @@ -50,6 +49,7 @@ def _parse_libsvm_line(line, multiclass=None): @staticmethod def _convert_labeled_point_to_libsvm(p): """Converts a LabeledPoint to a string in LIBSVM format.""" + from pyspark.mllib.regression import LabeledPoint assert isinstance(p, LabeledPoint) items = [str(p.label)] v = _convert_to_vector(p.features) @@ -92,6 +92,7 @@ def loadLibSVMFile(sc, path, numFeatures=-1, minPartitions=None, multiclass=None >>> from tempfile import NamedTemporaryFile >>> from pyspark.mllib.util import MLUtils + >>> from pyspark.mllib.regression import LabeledPoint >>> tempFile = NamedTemporaryFile(delete=True) >>> tempFile.write("+1 1:1.0 3:2.0 5:3.0\\n-1\\n-1 2:4.0 4:5.0 6:6.0") >>> tempFile.flush() @@ -110,6 +111,7 @@ def loadLibSVMFile(sc, path, numFeatures=-1, minPartitions=None, multiclass=None >>> print examples[2] (-1.0,(6,[1,3,5],[4.0,5.0,6.0])) """ + from pyspark.mllib.regression import LabeledPoint if multiclass is not None: warnings.warn("deprecated", DeprecationWarning) @@ -130,6 +132,7 @@ def saveAsLibSVMFile(data, dir): >>> from tempfile import NamedTemporaryFile >>> from fileinput import input + >>> from pyspark.mllib.regression import LabeledPoint >>> from glob import glob >>> from pyspark.mllib.util import MLUtils >>> examples = [LabeledPoint(1.1, Vectors.sparse(3, [(0, 1.23), (2, 4.56)])), \ @@ -156,6 +159,7 @@ def loadLabeledPoints(sc, path, minPartitions=None): >>> from tempfile import NamedTemporaryFile >>> from pyspark.mllib.util import MLUtils + >>> from pyspark.mllib.regression import LabeledPoint >>> examples = [LabeledPoint(1.1, Vectors.sparse(3, [(0, -1.23), (2, 4.56e-7)])), \ LabeledPoint(0.0, Vectors.dense([1.01, 2.02, 3.03]))] >>> tempFile = NamedTemporaryFile(delete=True) From 5e6ad24ff645a9b0f63d9c0f17193550963aa0a7 Mon Sep 17 00:00:00 2001 From: Shuo Xiang Date: Fri, 20 Mar 2015 14:45:44 -0400 Subject: [PATCH 115/122] [MLlib] SPARK-5954: Top by key This PR implements two functions - `topByKey(num: Int): RDD[(K, Array[V])]` finds the top-k values for each key in a pair RDD. This can be used, e.g., in computing top recommendations. - `takeOrderedByKey(num: Int): RDD[(K, Array[V])] ` does the opposite of `topByKey` The `sorted` is used here as the `toArray` method of the PriorityQueue does not return a necessarily sorted array. Author: Shuo Xiang Closes #5075 from coderxiang/topByKey and squashes the following commits: 1611c37 [Shuo Xiang] code clean up 6f565c0 [Shuo Xiang] naming a80e0ec [Shuo Xiang] typo and warning 82dded9 [Shuo Xiang] Merge remote-tracking branch 'upstream/master' into topByKey d202745 [Shuo Xiang] move to MLPairRDDFunctions 901b0af [Shuo Xiang] style check 70c6e35 [Shuo Xiang] remove takeOrderedByKey, update doc and test 0895c17 [Shuo Xiang] Merge remote-tracking branch 'upstream/master' into topByKey b10e325 [Shuo Xiang] Merge remote-tracking branch 'upstream/master' into topByKey debccad [Shuo Xiang] topByKey --- .../spark/mllib/rdd/MLPairRDDFunctions.scala | 60 +++++++++++++++++++ .../mllib/rdd/MLPairRDDFunctionsSuite.scala | 36 +++++++++++ 2 files changed, 96 insertions(+) create mode 100644 mllib/src/main/scala/org/apache/spark/mllib/rdd/MLPairRDDFunctions.scala create mode 100644 mllib/src/test/scala/org/apache/spark/mllib/rdd/MLPairRDDFunctionsSuite.scala diff --git a/mllib/src/main/scala/org/apache/spark/mllib/rdd/MLPairRDDFunctions.scala b/mllib/src/main/scala/org/apache/spark/mllib/rdd/MLPairRDDFunctions.scala new file mode 100644 index 0000000000000..9213fd3f595c3 --- /dev/null +++ b/mllib/src/main/scala/org/apache/spark/mllib/rdd/MLPairRDDFunctions.scala @@ -0,0 +1,60 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.mllib.rdd + +import scala.language.implicitConversions +import scala.reflect.ClassTag + +import org.apache.spark.annotation.DeveloperApi +import org.apache.spark.rdd.RDD +import org.apache.spark.util.BoundedPriorityQueue + +/** + * Machine learning specific Pair RDD functions. + */ +@DeveloperApi +class MLPairRDDFunctions[K: ClassTag, V: ClassTag](self: RDD[(K, V)]) extends Serializable { + /** + * Returns the top k (largest) elements for each key from this RDD as defined by the specified + * implicit Ordering[T]. + * If the number of elements for a certain key is less than k, all of them will be returned. + * + * @param num k, the number of top elements to return + * @param ord the implicit ordering for T + * @return an RDD that contains the top k values for each key + */ + def topByKey(num: Int)(implicit ord: Ordering[V]): RDD[(K, Array[V])] = { + self.aggregateByKey(new BoundedPriorityQueue[V](num)(ord))( + seqOp = (queue, item) => { + queue += item + queue + }, + combOp = (queue1, queue2) => { + queue1 ++= queue2 + queue1 + } + ).mapValues(_.toArray.sorted(ord.reverse)) + } +} + +@DeveloperApi +object MLPairRDDFunctions { + /** Implicit conversion from a pair RDD to MLPairRDDFunctions. */ + implicit def fromPairRDD[K: ClassTag, V: ClassTag](rdd: RDD[(K, V)]): MLPairRDDFunctions[K, V] = + new MLPairRDDFunctions[K, V](rdd) +} diff --git a/mllib/src/test/scala/org/apache/spark/mllib/rdd/MLPairRDDFunctionsSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/rdd/MLPairRDDFunctionsSuite.scala new file mode 100644 index 0000000000000..1ac7c12c4e8e6 --- /dev/null +++ b/mllib/src/test/scala/org/apache/spark/mllib/rdd/MLPairRDDFunctionsSuite.scala @@ -0,0 +1,36 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.mllib.rdd + +import org.scalatest.FunSuite + +import org.apache.spark.mllib.util.MLlibTestSparkContext +import org.apache.spark.mllib.rdd.MLPairRDDFunctions._ + +class MLPairRDDFunctionsSuite extends FunSuite with MLlibTestSparkContext { + test("topByKey") { + val topMap = sc.parallelize(Array((1, 1), (1, 2), (3, 2), (3, 7), (3, 5), (5, 1), (5, 3)), 2) + .topByKey(2) + .collectAsMap() + + assert(topMap.size === 3) + assert(topMap(1) === Array(2, 1)) + assert(topMap(3) === Array(7, 5)) + assert(topMap(5) === Array(3, 1)) + } +} From 25636d9867c6bc901463b6b227cb444d701cfdd1 Mon Sep 17 00:00:00 2001 From: Xusen Yin Date: Fri, 20 Mar 2015 14:53:59 -0400 Subject: [PATCH 116/122] [Spark 6096][MLlib] Add Naive Bayes load save methods in Python See [SPARK-6096](https://issues.apache.org/jira/browse/SPARK-6096). Author: Xusen Yin Closes #5090 from yinxusen/SPARK-6096 and squashes the following commits: bd0fea5 [Xusen Yin] fix style problem, etc. 3fd41f2 [Xusen Yin] use hanging indent in Python style e83803d [Xusen Yin] fix Python style d6dbde5 [Xusen Yin] fix python call java error a054bb3 [Xusen Yin] add save load for NaiveBayes python --- .../mllib/classification/NaiveBayes.scala | 11 +++++++ python/pyspark/mllib/classification.py | 31 ++++++++++++++++++- 2 files changed, 41 insertions(+), 1 deletion(-) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala b/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala index 068449aa1d346..d60e82c410979 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/classification/NaiveBayes.scala @@ -17,6 +17,10 @@ package org.apache.spark.mllib.classification +import java.lang.{Iterable => JIterable} + +import scala.collection.JavaConverters._ + import breeze.linalg.{DenseMatrix => BDM, DenseVector => BDV, argmax => brzArgmax, sum => brzSum} import org.json4s.JsonDSL._ import org.json4s.jackson.JsonMethods._ @@ -41,6 +45,13 @@ class NaiveBayesModel private[mllib] ( val pi: Array[Double], val theta: Array[Array[Double]]) extends ClassificationModel with Serializable with Saveable { + /** A Java-friendly constructor that takes three Iterable parameters. */ + private[mllib] def this( + labels: JIterable[Double], + pi: JIterable[Double], + theta: JIterable[JIterable[Double]]) = + this(labels.asScala.toArray, pi.asScala.toArray, theta.asScala.toArray.map(_.asScala.toArray)) + private val brzPi = new BDV[Double](pi) private val brzTheta = new BDM[Double](theta.length, theta(0).length) diff --git a/python/pyspark/mllib/classification.py b/python/pyspark/mllib/classification.py index b66159c5bfb66..6766f3ebb8894 100644 --- a/python/pyspark/mllib/classification.py +++ b/python/pyspark/mllib/classification.py @@ -24,6 +24,7 @@ from pyspark.mllib.common import callMLlibFunc, _py2java, _java2py from pyspark.mllib.linalg import SparseVector, _convert_to_vector from pyspark.mllib.regression import LabeledPoint, LinearModel, _regression_train_wrapper +from pyspark.mllib.util import Saveable, Loader, inherit_doc __all__ = ['LogisticRegressionModel', 'LogisticRegressionWithSGD', 'LogisticRegressionWithLBFGS', @@ -359,7 +360,8 @@ def train(rdd, i): return _regression_train_wrapper(train, SVMModel, data, initialWeights) -class NaiveBayesModel(object): +@inherit_doc +class NaiveBayesModel(Saveable, Loader): """ Model for Naive Bayes classifiers. @@ -390,6 +392,16 @@ class NaiveBayesModel(object): 0.0 >>> model.predict(SparseVector(2, {0: 1.0})) 1.0 + >>> import os, tempfile + >>> path = tempfile.mkdtemp() + >>> model.save(sc, path) + >>> sameModel = NaiveBayesModel.load(sc, path) + >>> sameModel.predict(SparseVector(2, {0: 1.0})) == model.predict(SparseVector(2, {0: 1.0})) + True + >>> try: + ... os.removedirs(path) + ... except OSError: + ... pass """ def __init__(self, labels, pi, theta): @@ -404,6 +416,23 @@ def predict(self, x): x = _convert_to_vector(x) return self.labels[numpy.argmax(self.pi + x.dot(self.theta.transpose()))] + def save(self, sc, path): + java_labels = _py2java(sc, self.labels.tolist()) + java_pi = _py2java(sc, self.pi.tolist()) + java_theta = _py2java(sc, self.theta.tolist()) + java_model = sc._jvm.org.apache.spark.mllib.classification.NaiveBayesModel( + java_labels, java_pi, java_theta) + java_model.save(sc._jsc.sc(), path) + + @classmethod + def load(cls, sc, path): + java_model = sc._jvm.org.apache.spark.mllib.classification.NaiveBayesModel.load( + sc._jsc.sc(), path) + py_labels = _java2py(sc, java_model.labels()) + py_pi = _java2py(sc, java_model.pi()) + py_theta = _java2py(sc, java_model.theta()) + return NaiveBayesModel(py_labels, py_pi, numpy.array(py_theta)) + class NaiveBayes(object): From 6b36470c66bd6140c45e45d3f1d51b0082c3fd97 Mon Sep 17 00:00:00 2001 From: Xiangrui Meng Date: Fri, 20 Mar 2015 15:02:57 -0400 Subject: [PATCH 117/122] [SPARK-5955][MLLIB] add checkpointInterval to ALS Add checkpiontInterval to ALS to prevent: 1. StackOverflow exceptions caused by long lineage, 2. large shuffle files generated during iterations, 3. slow recovery when some node fail. srowen coderxiang Author: Xiangrui Meng Closes #5076 from mengxr/SPARK-5955 and squashes the following commits: df56791 [Xiangrui Meng] update impl to reuse code 29affcb [Xiangrui Meng] do not materialize factors in implicit 20d3f7f [Xiangrui Meng] add checkpointInterval to ALS --- .../apache/spark/ml/param/sharedParams.scala | 11 +++++ .../apache/spark/ml/recommendation/ALS.scala | 42 ++++++++++++++++--- .../spark/mllib/recommendation/ALS.scala | 17 ++++++++ .../spark/ml/recommendation/ALSSuite.scala | 17 ++++++++ 4 files changed, 82 insertions(+), 5 deletions(-) diff --git a/mllib/src/main/scala/org/apache/spark/ml/param/sharedParams.scala b/mllib/src/main/scala/org/apache/spark/ml/param/sharedParams.scala index 1a70322b4cace..5d660d1e151a7 100644 --- a/mllib/src/main/scala/org/apache/spark/ml/param/sharedParams.scala +++ b/mllib/src/main/scala/org/apache/spark/ml/param/sharedParams.scala @@ -138,3 +138,14 @@ private[ml] trait HasOutputCol extends Params { /** @group getParam */ def getOutputCol: String = get(outputCol) } + +private[ml] trait HasCheckpointInterval extends Params { + /** + * param for checkpoint interval + * @group param + */ + val checkpointInterval: IntParam = new IntParam(this, "checkpointInterval", "checkpoint interval") + + /** @group getParam */ + def getCheckpointInterval: Int = get(checkpointInterval) +} diff --git a/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala b/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala index e3515ee81af3d..514b4ef98dc5b 100644 --- a/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala +++ b/mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala @@ -18,6 +18,7 @@ package org.apache.spark.ml.recommendation import java.{util => ju} +import java.io.IOException import scala.collection.mutable import scala.reflect.ClassTag @@ -26,6 +27,7 @@ import scala.util.hashing.byteswap64 import com.github.fommil.netlib.BLAS.{getInstance => blas} import com.github.fommil.netlib.LAPACK.{getInstance => lapack} +import org.apache.hadoop.fs.{FileSystem, Path} import org.netlib.util.intW import org.apache.spark.{Logging, Partitioner} @@ -46,7 +48,7 @@ import org.apache.spark.util.random.XORShiftRandom * Common params for ALS. */ private[recommendation] trait ALSParams extends Params with HasMaxIter with HasRegParam - with HasPredictionCol { + with HasPredictionCol with HasCheckpointInterval { /** * Param for rank of the matrix factorization. @@ -164,6 +166,7 @@ class ALSModel private[ml] ( itemFactors: RDD[(Int, Array[Float])]) extends Model[ALSModel] with ALSParams { + /** @group setParam */ def setPredictionCol(value: String): this.type = set(predictionCol, value) override def transform(dataset: DataFrame, paramMap: ParamMap): DataFrame = { @@ -262,6 +265,9 @@ class ALS extends Estimator[ALSModel] with ALSParams { /** @group setParam */ def setNonnegative(value: Boolean): this.type = set(nonnegative, value) + /** @group setParam */ + def setCheckpointInterval(value: Int): this.type = set(checkpointInterval, value) + /** * Sets both numUserBlocks and numItemBlocks to the specific value. * @group setParam @@ -274,6 +280,7 @@ class ALS extends Estimator[ALSModel] with ALSParams { setMaxIter(20) setRegParam(1.0) + setCheckpointInterval(10) override def fit(dataset: DataFrame, paramMap: ParamMap): ALSModel = { val map = this.paramMap ++ paramMap @@ -285,7 +292,8 @@ class ALS extends Estimator[ALSModel] with ALSParams { val (userFactors, itemFactors) = ALS.train(ratings, rank = map(rank), numUserBlocks = map(numUserBlocks), numItemBlocks = map(numItemBlocks), maxIter = map(maxIter), regParam = map(regParam), implicitPrefs = map(implicitPrefs), - alpha = map(alpha), nonnegative = map(nonnegative)) + alpha = map(alpha), nonnegative = map(nonnegative), + checkpointInterval = map(checkpointInterval)) val model = new ALSModel(this, map, map(rank), userFactors, itemFactors) Params.inheritValues(map, this, model) model @@ -494,6 +502,7 @@ object ALS extends Logging { nonnegative: Boolean = false, intermediateRDDStorageLevel: StorageLevel = StorageLevel.MEMORY_AND_DISK, finalRDDStorageLevel: StorageLevel = StorageLevel.MEMORY_AND_DISK, + checkpointInterval: Int = 10, seed: Long = 0L)( implicit ord: Ordering[ID]): (RDD[(ID, Array[Float])], RDD[(ID, Array[Float])]) = { require(intermediateRDDStorageLevel != StorageLevel.NONE, @@ -521,6 +530,18 @@ object ALS extends Logging { val seedGen = new XORShiftRandom(seed) var userFactors = initialize(userInBlocks, rank, seedGen.nextLong()) var itemFactors = initialize(itemInBlocks, rank, seedGen.nextLong()) + var previousCheckpointFile: Option[String] = None + val shouldCheckpoint: Int => Boolean = (iter) => + sc.checkpointDir.isDefined && (iter % checkpointInterval == 0) + val deletePreviousCheckpointFile: () => Unit = () => + previousCheckpointFile.foreach { file => + try { + FileSystem.get(sc.hadoopConfiguration).delete(new Path(file), true) + } catch { + case e: IOException => + logWarning(s"Cannot delete checkpoint file $file:", e) + } + } if (implicitPrefs) { for (iter <- 1 to maxIter) { userFactors.setName(s"userFactors-$iter").persist(intermediateRDDStorageLevel) @@ -528,19 +549,30 @@ object ALS extends Logging { itemFactors = computeFactors(userFactors, userOutBlocks, itemInBlocks, rank, regParam, userLocalIndexEncoder, implicitPrefs, alpha, solver) previousItemFactors.unpersist() - if (sc.checkpointDir.isDefined && (iter % 3 == 0)) { - itemFactors.checkpoint() - } itemFactors.setName(s"itemFactors-$iter").persist(intermediateRDDStorageLevel) + // TODO: Generalize PeriodicGraphCheckpointer and use it here. + if (shouldCheckpoint(iter)) { + itemFactors.checkpoint() // itemFactors gets materialized in computeFactors. + } val previousUserFactors = userFactors userFactors = computeFactors(itemFactors, itemOutBlocks, userInBlocks, rank, regParam, itemLocalIndexEncoder, implicitPrefs, alpha, solver) + if (shouldCheckpoint(iter)) { + deletePreviousCheckpointFile() + previousCheckpointFile = itemFactors.getCheckpointFile + } previousUserFactors.unpersist() } } else { for (iter <- 0 until maxIter) { itemFactors = computeFactors(userFactors, userOutBlocks, itemInBlocks, rank, regParam, userLocalIndexEncoder, solver = solver) + if (shouldCheckpoint(iter)) { + itemFactors.checkpoint() + itemFactors.count() // checkpoint item factors and cut lineage + deletePreviousCheckpointFile() + previousCheckpointFile = itemFactors.getCheckpointFile + } userFactors = computeFactors(itemFactors, itemOutBlocks, userInBlocks, rank, regParam, itemLocalIndexEncoder, solver = solver) } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/recommendation/ALS.scala b/mllib/src/main/scala/org/apache/spark/mllib/recommendation/ALS.scala index caacab943030b..dddefe1944e9d 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/recommendation/ALS.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/recommendation/ALS.scala @@ -82,6 +82,9 @@ class ALS private ( private var intermediateRDDStorageLevel: StorageLevel = StorageLevel.MEMORY_AND_DISK private var finalRDDStorageLevel: StorageLevel = StorageLevel.MEMORY_AND_DISK + /** checkpoint interval */ + private var checkpointInterval: Int = 10 + /** * Set the number of blocks for both user blocks and product blocks to parallelize the computation * into; pass -1 for an auto-configured number of blocks. Default: -1. @@ -182,6 +185,19 @@ class ALS private ( this } + /** + * Set period (in iterations) between checkpoints (default = 10). Checkpointing helps with + * recovery (when nodes fail) and StackOverflow exceptions caused by long lineage. It also helps + * with eliminating temporary shuffle files on disk, which can be important when there are many + * ALS iterations. If the checkpoint directory is not set in [[org.apache.spark.SparkContext]], + * this setting is ignored. + */ + @DeveloperApi + def setCheckpointInterval(checkpointInterval: Int): this.type = { + this.checkpointInterval = checkpointInterval + this + } + /** * Run ALS with the configured parameters on an input RDD of (user, product, rating) triples. * Returns a MatrixFactorizationModel with feature vectors for each user and product. @@ -212,6 +228,7 @@ class ALS private ( nonnegative = nonnegative, intermediateRDDStorageLevel = intermediateRDDStorageLevel, finalRDDStorageLevel = StorageLevel.NONE, + checkpointInterval = checkpointInterval, seed = seed) val userFactors = floatUserFactors diff --git a/mllib/src/test/scala/org/apache/spark/ml/recommendation/ALSSuite.scala b/mllib/src/test/scala/org/apache/spark/ml/recommendation/ALSSuite.scala index bb86bafc0eb0a..0bb06e9e8ac9c 100644 --- a/mllib/src/test/scala/org/apache/spark/ml/recommendation/ALSSuite.scala +++ b/mllib/src/test/scala/org/apache/spark/ml/recommendation/ALSSuite.scala @@ -17,6 +17,7 @@ package org.apache.spark.ml.recommendation +import java.io.File import java.util.Random import scala.collection.mutable @@ -32,16 +33,25 @@ import org.apache.spark.mllib.util.MLlibTestSparkContext import org.apache.spark.mllib.util.TestingUtils._ import org.apache.spark.rdd.RDD import org.apache.spark.sql.{Row, SQLContext} +import org.apache.spark.util.Utils class ALSSuite extends FunSuite with MLlibTestSparkContext with Logging { private var sqlContext: SQLContext = _ + private var tempDir: File = _ override def beforeAll(): Unit = { super.beforeAll() + tempDir = Utils.createTempDir() + sc.setCheckpointDir(tempDir.getAbsolutePath) sqlContext = new SQLContext(sc) } + override def afterAll(): Unit = { + Utils.deleteRecursively(tempDir) + super.afterAll() + } + test("LocalIndexEncoder") { val random = new Random for (numBlocks <- Seq(1, 2, 5, 10, 20, 50, 100)) { @@ -485,4 +495,11 @@ class ALSSuite extends FunSuite with MLlibTestSparkContext with Logging { }.count() } } + + test("als with large number of iterations") { + val (ratings, _) = genExplicitTestData(numUsers = 4, numItems = 4, rank = 1) + ALS.train(ratings, rank = 1, maxIter = 50, numUserBlocks = 2, numItemBlocks = 2) + ALS.train( + ratings, rank = 1, maxIter = 50, numUserBlocks = 2, numItemBlocks = 2, implicitPrefs = true) + } } From 49a01c7ea2c48feee7ab4551c4fa03fd1cdb1a32 Mon Sep 17 00:00:00 2001 From: Jongyoul Lee Date: Fri, 20 Mar 2015 19:14:35 +0000 Subject: [PATCH 118/122] [SPARK-6423][Mesos] MemoryUtils should use memoryOverhead if it's set - Fixed calculateTotalMemory to use spark.mesos.executor.memoryOverhead - Added testCase Author: Jongyoul Lee Closes #5099 from jongyoul/SPARK-6423 and squashes the following commits: 6747fce [Jongyoul Lee] [SPARK-6423][Mesos] MemoryUtils should use memoryOverhead if it's set - Changed a description of spark.mesos.executor.memoryOverhead 475a7c8 [Jongyoul Lee] [SPARK-6423][Mesos] MemoryUtils should use memoryOverhead if it's set - Fit the import rules 453c5a2 [Jongyoul Lee] [SPARK-6423][Mesos] MemoryUtils should use memoryOverhead if it's set - Fixed calculateTotalMemory to use spark.mesos.executor.memoryOverhead - Added testCase --- .../scheduler/cluster/mesos/MemoryUtils.scala | 10 ++-- .../cluster/mesos/MemoryUtilsSuite.scala | 47 +++++++++++++++++++ docs/running-on-mesos.md | 8 ++-- 3 files changed, 53 insertions(+), 12 deletions(-) create mode 100644 core/src/test/scala/org/apache/spark/scheduler/cluster/mesos/MemoryUtilsSuite.scala diff --git a/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MemoryUtils.scala b/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MemoryUtils.scala index 705116cb13f54..aa3ec0f8cfb9c 100644 --- a/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MemoryUtils.scala +++ b/core/src/main/scala/org/apache/spark/scheduler/cluster/mesos/MemoryUtils.scala @@ -21,15 +21,11 @@ import org.apache.spark.SparkContext private[spark] object MemoryUtils { // These defaults copied from YARN - val OVERHEAD_FRACTION = 1.10 + val OVERHEAD_FRACTION = 0.10 val OVERHEAD_MINIMUM = 384 def calculateTotalMemory(sc: SparkContext) = { - math.max( - sc.conf.getOption("spark.mesos.executor.memoryOverhead") - .getOrElse(OVERHEAD_MINIMUM.toString) - .toInt + sc.executorMemory, - OVERHEAD_FRACTION * sc.executorMemory - ) + sc.conf.getInt("spark.mesos.executor.memoryOverhead", + math.max(OVERHEAD_FRACTION * sc.executorMemory, OVERHEAD_MINIMUM).toInt) + sc.executorMemory } } diff --git a/core/src/test/scala/org/apache/spark/scheduler/cluster/mesos/MemoryUtilsSuite.scala b/core/src/test/scala/org/apache/spark/scheduler/cluster/mesos/MemoryUtilsSuite.scala new file mode 100644 index 0000000000000..3fa0115e68259 --- /dev/null +++ b/core/src/test/scala/org/apache/spark/scheduler/cluster/mesos/MemoryUtilsSuite.scala @@ -0,0 +1,47 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.scheduler.cluster.mesos + +import org.mockito.Mockito._ +import org.scalatest.FunSuite +import org.scalatest.mock.MockitoSugar + +import org.apache.spark.{SparkConf, SparkContext} + +class MemoryUtilsSuite extends FunSuite with MockitoSugar { + test("MesosMemoryUtils should always override memoryOverhead when it's set") { + val sparkConf = new SparkConf + + val sc = mock[SparkContext] + when(sc.conf).thenReturn(sparkConf) + + // 384 > sc.executorMemory * 0.1 => 512 + 384 = 896 + when(sc.executorMemory).thenReturn(512) + assert(MemoryUtils.calculateTotalMemory(sc) === 896) + + // 384 < sc.executorMemory * 0.1 => 4096 + (4096 * 0.1) = 4505.6 + when(sc.executorMemory).thenReturn(4096) + assert(MemoryUtils.calculateTotalMemory(sc) === 4505) + + // set memoryOverhead + sparkConf.set("spark.mesos.executor.memoryOverhead", "100") + assert(MemoryUtils.calculateTotalMemory(sc) === 4196) + sparkConf.set("spark.mesos.executor.memoryOverhead", "400") + assert(MemoryUtils.calculateTotalMemory(sc) === 4496) + } +} diff --git a/docs/running-on-mesos.md b/docs/running-on-mesos.md index 6a9d304501dc0..c984639bd34cf 100644 --- a/docs/running-on-mesos.md +++ b/docs/running-on-mesos.md @@ -224,11 +224,9 @@ See the [configuration page](configuration.html) for information on Spark config spark.mesos.executor.memoryOverhead executor memory * 0.10, with minimum of 384 - This value is an additive for spark.executor.memory, specified in MB, - which is used to calculate the total Mesos task memory. A value of 384 - implies a 384MB overhead. Additionally, there is a hard-coded 10% minimum - overhead. The final overhead will be the larger of either - `spark.mesos.executor.memoryOverhead` or 10% of `spark.executor.memory`. + The amount of additional memory, specified in MB, to be allocated per executor. By default, + the overhead will be larger of either 384 or 10% of `spark.executor.memory`. If it's set, + the final overhead will be this value. From 11e025956be3818c00effef0d650734f8feeb436 Mon Sep 17 00:00:00 2001 From: MechCoder Date: Fri, 20 Mar 2015 17:13:18 -0400 Subject: [PATCH 119/122] [SPARK-6309] [SQL] [MLlib] Implement MatrixUDT Utilities to serialize and deserialize Matrices in MLlib Author: MechCoder Closes #5048 from MechCoder/spark-6309 and squashes the following commits: 05dc6f2 [MechCoder] Hashcode and organize imports 16d5d47 [MechCoder] Test some more 6e67020 [MechCoder] TST: Test using Array conversion instead of equals 7fa7a2c [MechCoder] [SPARK-6309] [SQL] [MLlib] Implement MatrixUDT --- .../apache/spark/mllib/linalg/Matrices.scala | 90 +++++++++++++++++++ .../spark/mllib/linalg/MatricesSuite.scala | 13 +++ 2 files changed, 103 insertions(+) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala b/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala index fdd8848189f19..849f44295f089 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/linalg/Matrices.scala @@ -23,9 +23,15 @@ import scala.collection.mutable.{ArrayBuilder => MArrayBuilder, HashSet => MHash import breeze.linalg.{CSCMatrix => BSM, DenseMatrix => BDM, Matrix => BM} +import org.apache.spark.annotation.DeveloperApi +import org.apache.spark.sql.Row +import org.apache.spark.sql.types._ +import org.apache.spark.sql.catalyst.expressions.GenericMutableRow + /** * Trait for a local matrix. */ +@SQLUserDefinedType(udt = classOf[MatrixUDT]) sealed trait Matrix extends Serializable { /** Number of rows. */ @@ -102,6 +108,88 @@ sealed trait Matrix extends Serializable { private[spark] def foreachActive(f: (Int, Int, Double) => Unit) } +@DeveloperApi +private[spark] class MatrixUDT extends UserDefinedType[Matrix] { + + override def sqlType: StructType = { + // type: 0 = sparse, 1 = dense + // the dense matrix is built by numRows, numCols, values and isTransposed, all of which are + // set as not nullable, except values since in the future, support for binary matrices might + // be added for which values are not needed. + // the sparse matrix needs colPtrs and rowIndices, which are set as + // null, while building the dense matrix. + StructType(Seq( + StructField("type", ByteType, nullable = false), + StructField("numRows", IntegerType, nullable = false), + StructField("numCols", IntegerType, nullable = false), + StructField("colPtrs", ArrayType(IntegerType, containsNull = false), nullable = true), + StructField("rowIndices", ArrayType(IntegerType, containsNull = false), nullable = true), + StructField("values", ArrayType(DoubleType, containsNull = false), nullable = true), + StructField("isTransposed", BooleanType, nullable = false) + )) + } + + override def serialize(obj: Any): Row = { + val row = new GenericMutableRow(7) + obj match { + case sm: SparseMatrix => + row.setByte(0, 0) + row.setInt(1, sm.numRows) + row.setInt(2, sm.numCols) + row.update(3, sm.colPtrs.toSeq) + row.update(4, sm.rowIndices.toSeq) + row.update(5, sm.values.toSeq) + row.setBoolean(6, sm.isTransposed) + + case dm: DenseMatrix => + row.setByte(0, 1) + row.setInt(1, dm.numRows) + row.setInt(2, dm.numCols) + row.setNullAt(3) + row.setNullAt(4) + row.update(5, dm.values.toSeq) + row.setBoolean(6, dm.isTransposed) + } + row + } + + override def deserialize(datum: Any): Matrix = { + datum match { + // TODO: something wrong with UDT serialization, should never happen. + case m: Matrix => m + case row: Row => + require(row.length == 7, + s"MatrixUDT.deserialize given row with length ${row.length} but requires length == 7") + val tpe = row.getByte(0) + val numRows = row.getInt(1) + val numCols = row.getInt(2) + val values = row.getAs[Iterable[Double]](5).toArray + val isTransposed = row.getBoolean(6) + tpe match { + case 0 => + val colPtrs = row.getAs[Iterable[Int]](3).toArray + val rowIndices = row.getAs[Iterable[Int]](4).toArray + new SparseMatrix(numRows, numCols, colPtrs, rowIndices, values, isTransposed) + case 1 => + new DenseMatrix(numRows, numCols, values, isTransposed) + } + } + } + + override def userClass: Class[Matrix] = classOf[Matrix] + + override def equals(o: Any): Boolean = { + o match { + case v: MatrixUDT => true + case _ => false + } + } + + override def hashCode(): Int = 1994 + + private[spark] override def asNullable: MatrixUDT = this +} + /** * Column-major dense matrix. * The entry values are stored in a single array of doubles with columns listed in sequence. @@ -119,6 +207,7 @@ sealed trait Matrix extends Serializable { * @param isTransposed whether the matrix is transposed. If true, `values` stores the matrix in * row major. */ +@SQLUserDefinedType(udt = classOf[MatrixUDT]) class DenseMatrix( val numRows: Int, val numCols: Int, @@ -360,6 +449,7 @@ object DenseMatrix { * Compressed Sparse Row (CSR) format, where `colPtrs` behaves as rowPtrs, * and `rowIndices` behave as colIndices, and `values` are stored in row major. */ +@SQLUserDefinedType(udt = classOf[MatrixUDT]) class SparseMatrix( val numRows: Int, val numCols: Int, diff --git a/mllib/src/test/scala/org/apache/spark/mllib/linalg/MatricesSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/linalg/MatricesSuite.scala index c098b5458fe6b..96f677db3f377 100644 --- a/mllib/src/test/scala/org/apache/spark/mllib/linalg/MatricesSuite.scala +++ b/mllib/src/test/scala/org/apache/spark/mllib/linalg/MatricesSuite.scala @@ -424,4 +424,17 @@ class MatricesSuite extends FunSuite { assert(mat.rowIndices.toSeq === Seq(3, 0, 2, 1)) assert(mat.values.toSeq === Seq(1.0, 2.0, 3.0, 4.0)) } + + test("MatrixUDT") { + val dm1 = new DenseMatrix(2, 2, Array(0.9, 1.2, 2.3, 9.8)) + val dm2 = new DenseMatrix(3, 2, Array(0.0, 1.21, 2.3, 9.8, 9.0, 0.0)) + val dm3 = new DenseMatrix(0, 0, Array()) + val sm1 = dm1.toSparse + val sm2 = dm2.toSparse + val sm3 = dm3.toSparse + val mUDT = new MatrixUDT() + Seq(dm1, dm2, dm3, sm1, sm2, sm3).foreach { + mat => assert(mat.toArray === mUDT.deserialize(mUDT.serialize(mat)).toArray) + } + } } From 257cde7c363efb3317bfb5c13975cca9154894e2 Mon Sep 17 00:00:00 2001 From: lewuathe Date: Fri, 20 Mar 2015 17:18:18 -0400 Subject: [PATCH 120/122] [SPARK-6421][MLLIB] _regression_train_wrapper does not test initialWeights correctly Weight parameters must be initialized correctly even when numpy array is passed as initial weights. Author: lewuathe Closes #5101 from Lewuathe/SPARK-6421 and squashes the following commits: 7795201 [lewuathe] Fix lint-python errors 21d4fe3 [lewuathe] Fix init logic of weights --- python/pyspark/mllib/regression.py | 3 ++- python/pyspark/mllib/tests.py | 7 +++++++ 2 files changed, 9 insertions(+), 1 deletion(-) diff --git a/python/pyspark/mllib/regression.py b/python/pyspark/mllib/regression.py index 015a7860116c9..414a0ada80787 100644 --- a/python/pyspark/mllib/regression.py +++ b/python/pyspark/mllib/regression.py @@ -163,7 +163,8 @@ def _regression_train_wrapper(train_func, modelClass, data, initial_weights): first = data.first() if not isinstance(first, LabeledPoint): raise ValueError("data should be an RDD of LabeledPoint, but got %s" % first) - initial_weights = initial_weights or [0.0] * len(data.first().features) + if initial_weights is None: + initial_weights = [0.0] * len(data.first().features) weights, intercept = train_func(data, _convert_to_vector(initial_weights)) return modelClass(weights, intercept) diff --git a/python/pyspark/mllib/tests.py b/python/pyspark/mllib/tests.py index 5328d99b69684..155019638f806 100644 --- a/python/pyspark/mllib/tests.py +++ b/python/pyspark/mllib/tests.py @@ -323,6 +323,13 @@ def test_regression(self): self.assertTrue(gbt_model.predict(features[2]) <= 0) self.assertTrue(gbt_model.predict(features[3]) > 0) + try: + LinearRegressionWithSGD.train(rdd, initialWeights=array([1.0, 1.0])) + LassoWithSGD.train(rdd, initialWeights=array([1.0, 1.0])) + RidgeRegressionWithSGD.train(rdd, initialWeights=array([1.0, 1.0])) + except ValueError: + self.fail() + class StatTests(PySparkTestCase): # SPARK-4023 From a95043b1780bfde556db2dcc01511e40a12498dd Mon Sep 17 00:00:00 2001 From: Reynold Xin Date: Fri, 20 Mar 2015 15:47:07 -0700 Subject: [PATCH 121/122] [SPARK-6428][SQL] Added explicit type for all public methods in sql/core Also implemented equals/hashCode when they are missing. This is done in order to enable automatic public method type checking. Author: Reynold Xin Closes #5104 from rxin/sql-hashcode-explicittype and squashes the following commits: ffce6f3 [Reynold Xin] Code review feedback. 8b36733 [Reynold Xin] [SPARK-6428][SQL] Added explicit type for all public methods. --- .../catalyst/expressions/AttributeSet.scala | 3 +- .../spark/sql/catalyst/expressions/rows.scala | 21 +++++ .../org/apache/spark/sql/types/Decimal.scala | 2 +- .../apache/spark/sql/types/dataTypes.scala | 20 ++--- .../scala/org/apache/spark/sql/Column.scala | 2 +- .../org/apache/spark/sql/DataFrame.scala | 6 +- .../org/apache/spark/sql/SQLContext.scala | 8 +- .../apache/spark/sql/UDFRegistration.scala | 2 +- .../spark/sql/columnar/ColumnAccessor.scala | 4 +- .../spark/sql/columnar/ColumnBuilder.scala | 4 +- .../spark/sql/columnar/ColumnStats.scala | 24 +++--- .../spark/sql/columnar/ColumnType.scala | 56 +++++++------ .../columnar/InMemoryColumnarTableScan.scala | 56 +++++++------ .../sql/columnar/NullableColumnAccessor.scala | 2 +- .../CompressibleColumnAccessor.scala | 4 +- .../CompressibleColumnBuilder.scala | 2 +- .../compression/compressionSchemes.scala | 80 ++++++++++--------- .../spark/sql/execution/Aggregate.scala | 8 +- .../apache/spark/sql/execution/Exchange.scala | 19 ++--- .../spark/sql/execution/ExistingRDD.scala | 20 ++--- .../apache/spark/sql/execution/Expand.scala | 5 +- .../apache/spark/sql/execution/Generate.scala | 3 +- .../sql/execution/GeneratedAggregate.scala | 11 +-- .../spark/sql/execution/LocalTableScan.scala | 7 +- .../spark/sql/execution/SparkPlan.scala | 1 + .../sql/execution/SparkSqlSerializer.scala | 2 +- .../spark/sql/execution/SparkStrategies.scala | 6 +- .../spark/sql/execution/basicOperators.scala | 73 +++++++++-------- .../apache/spark/sql/execution/commands.scala | 33 ++++---- .../spark/sql/execution/debug/package.scala | 30 +++---- .../execution/joins/BroadcastHashJoin.scala | 10 ++- .../joins/BroadcastLeftSemiJoinHash.scala | 10 +-- .../joins/BroadcastNestedLoopJoin.scala | 5 +- .../execution/joins/CartesianProduct.scala | 8 +- .../spark/sql/execution/joins/HashJoin.scala | 4 +- .../sql/execution/joins/HashOuterJoin.scala | 53 ++++++------ .../sql/execution/joins/HashedRelation.scala | 4 +- .../sql/execution/joins/LeftSemiJoinBNL.scala | 10 ++- .../execution/joins/LeftSemiJoinHash.scala | 11 +-- .../execution/joins/ShuffledHashJoin.scala | 6 +- .../spark/sql/execution/pythonUdfs.scala | 21 +++-- .../org/apache/spark/sql/jdbc/JDBCRDD.scala | 3 +- .../apache/spark/sql/jdbc/JDBCRelation.scala | 8 +- .../apache/spark/sql/json/JSONRelation.scala | 6 +- .../spark/sql/parquet/ParquetConverter.scala | 2 +- .../spark/sql/parquet/ParquetRelation.scala | 10 ++- .../sql/parquet/ParquetTableOperations.scala | 12 +-- .../apache/spark/sql/parquet/newParquet.scala | 37 ++++++--- .../sql/parquet/timestamp/NanoTime.scala | 6 +- .../spark/sql/sources/LogicalRelation.scala | 16 ++-- .../apache/spark/sql/sources/commands.scala | 2 +- .../org/apache/spark/sql/sources/ddl.scala | 6 +- .../org/apache/spark/sql/sources/rules.scala | 4 +- 53 files changed, 438 insertions(+), 330 deletions(-) diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/AttributeSet.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/AttributeSet.scala index a9ba0be596349..adaeab0b5c027 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/AttributeSet.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/AttributeSet.scala @@ -17,7 +17,6 @@ package org.apache.spark.sql.catalyst.expressions -import org.apache.spark.sql.catalyst.analysis.Star protected class AttributeEquals(val a: Attribute) { override def hashCode() = a match { @@ -115,7 +114,7 @@ class AttributeSet private (val baseSet: Set[AttributeEquals]) // sorts of things in its closure. override def toSeq: Seq[Attribute] = baseSet.map(_.a).toArray.toSeq - override def toString = "{" + baseSet.map(_.a).mkString(", ") + "}" + override def toString: String = "{" + baseSet.map(_.a).mkString(", ") + "}" override def isEmpty: Boolean = baseSet.isEmpty } diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/rows.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/rows.scala index faa366771824b..f03d6f71a9fae 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/rows.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/rows.scala @@ -146,6 +146,27 @@ class GenericRow(protected[sql] val values: Array[Any]) extends Row { result } + override def equals(o: Any): Boolean = o match { + case other: Row => + if (values.length != other.length) { + return false + } + + var i = 0 + while (i < values.length) { + if (isNullAt(i) != other.isNullAt(i)) { + return false + } + if (apply(i) != other.apply(i)) { + return false + } + i += 1 + } + true + + case _ => false + } + def copy() = this } diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/types/Decimal.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/types/Decimal.scala index 21cc6cea4bf54..994c5202c15dc 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/types/Decimal.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/types/Decimal.scala @@ -246,7 +246,7 @@ final class Decimal extends Ordered[Decimal] with Serializable { } } - override def equals(other: Any) = other match { + override def equals(other: Any): Boolean = other match { case d: Decimal => compare(d) == 0 case _ => diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/types/dataTypes.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/types/dataTypes.scala index bf39603d13bd5..d973144de3468 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/types/dataTypes.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/types/dataTypes.scala @@ -307,7 +307,7 @@ protected[sql] object NativeType { protected[sql] trait PrimitiveType extends DataType { - override def isPrimitive = true + override def isPrimitive: Boolean = true } @@ -442,7 +442,7 @@ class TimestampType private() extends NativeType { @transient private[sql] lazy val tag = ScalaReflectionLock.synchronized { typeTag[JvmType] } private[sql] val ordering = new Ordering[JvmType] { - def compare(x: Timestamp, y: Timestamp) = x.compareTo(y) + def compare(x: Timestamp, y: Timestamp): Int = x.compareTo(y) } /** @@ -542,7 +542,7 @@ class LongType private() extends IntegralType { */ override def defaultSize: Int = 8 - override def simpleString = "bigint" + override def simpleString: String = "bigint" private[spark] override def asNullable: LongType = this } @@ -572,7 +572,7 @@ class IntegerType private() extends IntegralType { */ override def defaultSize: Int = 4 - override def simpleString = "int" + override def simpleString: String = "int" private[spark] override def asNullable: IntegerType = this } @@ -602,7 +602,7 @@ class ShortType private() extends IntegralType { */ override def defaultSize: Int = 2 - override def simpleString = "smallint" + override def simpleString: String = "smallint" private[spark] override def asNullable: ShortType = this } @@ -632,7 +632,7 @@ class ByteType private() extends IntegralType { */ override def defaultSize: Int = 1 - override def simpleString = "tinyint" + override def simpleString: String = "tinyint" private[spark] override def asNullable: ByteType = this } @@ -696,7 +696,7 @@ case class DecimalType(precisionInfo: Option[PrecisionInfo]) extends FractionalT */ override def defaultSize: Int = 4096 - override def simpleString = precisionInfo match { + override def simpleString: String = precisionInfo match { case Some(PrecisionInfo(precision, scale)) => s"decimal($precision,$scale)" case None => "decimal(10,0)" } @@ -836,7 +836,7 @@ case class ArrayType(elementType: DataType, containsNull: Boolean) extends DataT */ override def defaultSize: Int = 100 * elementType.defaultSize - override def simpleString = s"array<${elementType.simpleString}>" + override def simpleString: String = s"array<${elementType.simpleString}>" private[spark] override def asNullable: ArrayType = ArrayType(elementType.asNullable, containsNull = true) @@ -1065,7 +1065,7 @@ case class StructType(fields: Array[StructField]) extends DataType with Seq[Stru */ override def defaultSize: Int = fields.map(_.dataType.defaultSize).sum - override def simpleString = { + override def simpleString: String = { val fieldTypes = fields.map(field => s"${field.name}:${field.dataType.simpleString}") s"struct<${fieldTypes.mkString(",")}>" } @@ -1142,7 +1142,7 @@ case class MapType( */ override def defaultSize: Int = 100 * (keyType.defaultSize + valueType.defaultSize) - override def simpleString = s"map<${keyType.simpleString},${valueType.simpleString}>" + override def simpleString: String = s"map<${keyType.simpleString},${valueType.simpleString}>" private[spark] override def asNullable: MapType = MapType(keyType.asNullable, valueType.asNullable, valueContainsNull = true) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/Column.scala b/sql/core/src/main/scala/org/apache/spark/sql/Column.scala index 908c78a4d3f10..b7a13a1b26802 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/Column.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/Column.scala @@ -59,7 +59,7 @@ class Column(protected[sql] val expr: Expression) { override def toString: String = expr.prettyString - override def equals(that: Any) = that match { + override def equals(that: Any): Boolean = that match { case that: Column => that.expr.equals(this.expr) case _ => false } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/DataFrame.scala b/sql/core/src/main/scala/org/apache/spark/sql/DataFrame.scala index 46f50708a9184..8b8f86c4127e0 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/DataFrame.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/DataFrame.scala @@ -33,7 +33,7 @@ import org.apache.spark.api.java.JavaRDD import org.apache.spark.api.python.SerDeUtil import org.apache.spark.rdd.RDD import org.apache.spark.storage.StorageLevel -import org.apache.spark.sql.catalyst.{ScalaReflection, SqlParser} +import org.apache.spark.sql.catalyst.{expressions, ScalaReflection, SqlParser} import org.apache.spark.sql.catalyst.analysis.{UnresolvedRelation, ResolvedStar} import org.apache.spark.sql.catalyst.expressions._ import org.apache.spark.sql.catalyst.plans.{JoinType, Inner} @@ -722,7 +722,7 @@ class DataFrame private[sql]( : DataFrame = { val dataType = ScalaReflection.schemaFor[B].dataType val attributes = AttributeReference(outputColumn, dataType)() :: Nil - def rowFunction(row: Row) = { + def rowFunction(row: Row): TraversableOnce[Row] = { f(row(0).asInstanceOf[A]).map(o => Row(ScalaReflection.convertToCatalyst(o, dataType))) } val generator = UserDefinedGenerator(attributes, rowFunction, apply(inputColumn).expr :: Nil) @@ -1155,7 +1155,7 @@ class DataFrame private[sql]( val gen = new JsonFactory().createGenerator(writer).setRootValueSeparator(null) new Iterator[String] { - override def hasNext = iter.hasNext + override def hasNext: Boolean = iter.hasNext override def next(): String = { JsonRDD.rowToJSON(rowSchema, gen)(iter.next()) gen.flush() diff --git a/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala b/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala index 6de46a50db20e..dc9912b52dcab 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala @@ -144,7 +144,7 @@ class SQLContext(@transient val sparkContext: SparkContext) @transient protected[sql] val tlSession = new ThreadLocal[SQLSession]() { - override def initialValue = defaultSession + override def initialValue: SQLSession = defaultSession } @transient @@ -988,9 +988,9 @@ class SQLContext(@transient val sparkContext: SparkContext) val sqlContext: SQLContext = self - def codegenEnabled = self.conf.codegenEnabled + def codegenEnabled: Boolean = self.conf.codegenEnabled - def numPartitions = self.conf.numShufflePartitions + def numPartitions: Int = self.conf.numShufflePartitions def strategies: Seq[Strategy] = experimental.extraStrategies ++ ( @@ -1109,7 +1109,7 @@ class SQLContext(@transient val sparkContext: SparkContext) lazy val analyzed: LogicalPlan = analyzer(logical) lazy val withCachedData: LogicalPlan = { - assertAnalyzed + assertAnalyzed() cacheManager.useCachedData(analyzed) } lazy val optimizedPlan: LogicalPlan = optimizer(withCachedData) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/UDFRegistration.scala b/sql/core/src/main/scala/org/apache/spark/sql/UDFRegistration.scala index 8051df299252c..b97aaf73529a3 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/UDFRegistration.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/UDFRegistration.scala @@ -61,7 +61,7 @@ class UDFRegistration private[sql] (sqlContext: SQLContext) extends Logging { val dataType = sqlContext.parseDataType(stringDataType) - def builder(e: Seq[Expression]) = + def builder(e: Seq[Expression]): PythonUDF = PythonUDF( name, command, diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnAccessor.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnAccessor.scala index b615eaa0dca0d..f615fb33a7c35 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnAccessor.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnAccessor.scala @@ -48,9 +48,9 @@ private[sql] abstract class BasicColumnAccessor[T <: DataType, JvmType]( protected def initialize() {} - def hasNext = buffer.hasRemaining + override def hasNext: Boolean = buffer.hasRemaining - def extractTo(row: MutableRow, ordinal: Int): Unit = { + override def extractTo(row: MutableRow, ordinal: Int): Unit = { extractSingle(row, ordinal) } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnBuilder.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnBuilder.scala index d8d24a577347c..c881747751520 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnBuilder.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnBuilder.scala @@ -58,7 +58,7 @@ private[sql] class BasicColumnBuilder[T <: DataType, JvmType]( override def initialize( initialSize: Int, columnName: String = "", - useCompression: Boolean = false) = { + useCompression: Boolean = false): Unit = { val size = if (initialSize == 0) DEFAULT_INITIAL_BUFFER_SIZE else initialSize this.columnName = columnName @@ -73,7 +73,7 @@ private[sql] class BasicColumnBuilder[T <: DataType, JvmType]( columnType.append(row, ordinal, buffer) } - override def build() = { + override def build(): ByteBuffer = { buffer.flip().asInstanceOf[ByteBuffer] } } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnStats.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnStats.scala index 04047b9c062be..87a6631da8300 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnStats.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnStats.scala @@ -76,7 +76,7 @@ private[sql] sealed trait ColumnStats extends Serializable { private[sql] class NoopColumnStats extends ColumnStats { override def gatherStats(row: Row, ordinal: Int): Unit = super.gatherStats(row, ordinal) - def collectedStatistics = Row(null, null, nullCount, count, 0L) + override def collectedStatistics: Row = Row(null, null, nullCount, count, 0L) } private[sql] class BooleanColumnStats extends ColumnStats { @@ -93,7 +93,7 @@ private[sql] class BooleanColumnStats extends ColumnStats { } } - def collectedStatistics = Row(lower, upper, nullCount, count, sizeInBytes) + override def collectedStatistics: Row = Row(lower, upper, nullCount, count, sizeInBytes) } private[sql] class ByteColumnStats extends ColumnStats { @@ -110,7 +110,7 @@ private[sql] class ByteColumnStats extends ColumnStats { } } - def collectedStatistics = Row(lower, upper, nullCount, count, sizeInBytes) + override def collectedStatistics: Row = Row(lower, upper, nullCount, count, sizeInBytes) } private[sql] class ShortColumnStats extends ColumnStats { @@ -127,7 +127,7 @@ private[sql] class ShortColumnStats extends ColumnStats { } } - def collectedStatistics = Row(lower, upper, nullCount, count, sizeInBytes) + override def collectedStatistics: Row = Row(lower, upper, nullCount, count, sizeInBytes) } private[sql] class LongColumnStats extends ColumnStats { @@ -144,7 +144,7 @@ private[sql] class LongColumnStats extends ColumnStats { } } - def collectedStatistics = Row(lower, upper, nullCount, count, sizeInBytes) + override def collectedStatistics: Row = Row(lower, upper, nullCount, count, sizeInBytes) } private[sql] class DoubleColumnStats extends ColumnStats { @@ -161,7 +161,7 @@ private[sql] class DoubleColumnStats extends ColumnStats { } } - def collectedStatistics = Row(lower, upper, nullCount, count, sizeInBytes) + override def collectedStatistics: Row = Row(lower, upper, nullCount, count, sizeInBytes) } private[sql] class FloatColumnStats extends ColumnStats { @@ -178,7 +178,7 @@ private[sql] class FloatColumnStats extends ColumnStats { } } - def collectedStatistics = Row(lower, upper, nullCount, count, sizeInBytes) + override def collectedStatistics: Row = Row(lower, upper, nullCount, count, sizeInBytes) } private[sql] class FixedDecimalColumnStats extends ColumnStats { @@ -212,7 +212,7 @@ private[sql] class IntColumnStats extends ColumnStats { } } - def collectedStatistics = Row(lower, upper, nullCount, count, sizeInBytes) + override def collectedStatistics: Row = Row(lower, upper, nullCount, count, sizeInBytes) } private[sql] class StringColumnStats extends ColumnStats { @@ -229,7 +229,7 @@ private[sql] class StringColumnStats extends ColumnStats { } } - def collectedStatistics = Row(lower, upper, nullCount, count, sizeInBytes) + override def collectedStatistics: Row = Row(lower, upper, nullCount, count, sizeInBytes) } private[sql] class DateColumnStats extends IntColumnStats @@ -248,7 +248,7 @@ private[sql] class TimestampColumnStats extends ColumnStats { } } - def collectedStatistics = Row(lower, upper, nullCount, count, sizeInBytes) + override def collectedStatistics: Row = Row(lower, upper, nullCount, count, sizeInBytes) } private[sql] class BinaryColumnStats extends ColumnStats { @@ -259,7 +259,7 @@ private[sql] class BinaryColumnStats extends ColumnStats { } } - def collectedStatistics = Row(null, null, nullCount, count, sizeInBytes) + override def collectedStatistics: Row = Row(null, null, nullCount, count, sizeInBytes) } private[sql] class GenericColumnStats extends ColumnStats { @@ -270,5 +270,5 @@ private[sql] class GenericColumnStats extends ColumnStats { } } - def collectedStatistics = Row(null, null, nullCount, count, sizeInBytes) + override def collectedStatistics: Row = Row(null, null, nullCount, count, sizeInBytes) } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnType.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnType.scala index 36ea1c77e0470..c47497e0662d9 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnType.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/ColumnType.scala @@ -98,7 +98,7 @@ private[sql] sealed abstract class ColumnType[T <: DataType, JvmType]( */ def clone(v: JvmType): JvmType = v - override def toString = getClass.getSimpleName.stripSuffix("$") + override def toString: String = getClass.getSimpleName.stripSuffix("$") } private[sql] abstract class NativeColumnType[T <: NativeType]( @@ -114,7 +114,7 @@ private[sql] abstract class NativeColumnType[T <: NativeType]( } private[sql] object INT extends NativeColumnType(IntegerType, 0, 4) { - def append(v: Int, buffer: ByteBuffer): Unit = { + override def append(v: Int, buffer: ByteBuffer): Unit = { buffer.putInt(v) } @@ -122,7 +122,7 @@ private[sql] object INT extends NativeColumnType(IntegerType, 0, 4) { buffer.putInt(row.getInt(ordinal)) } - def extract(buffer: ByteBuffer) = { + override def extract(buffer: ByteBuffer): Int = { buffer.getInt() } @@ -134,7 +134,7 @@ private[sql] object INT extends NativeColumnType(IntegerType, 0, 4) { row.setInt(ordinal, value) } - override def getField(row: Row, ordinal: Int) = row.getInt(ordinal) + override def getField(row: Row, ordinal: Int): Int = row.getInt(ordinal) override def copyField(from: Row, fromOrdinal: Int, to: MutableRow, toOrdinal: Int): Unit = { to.setInt(toOrdinal, from.getInt(fromOrdinal)) @@ -150,7 +150,7 @@ private[sql] object LONG extends NativeColumnType(LongType, 1, 8) { buffer.putLong(row.getLong(ordinal)) } - override def extract(buffer: ByteBuffer) = { + override def extract(buffer: ByteBuffer): Long = { buffer.getLong() } @@ -162,7 +162,7 @@ private[sql] object LONG extends NativeColumnType(LongType, 1, 8) { row.setLong(ordinal, value) } - override def getField(row: Row, ordinal: Int) = row.getLong(ordinal) + override def getField(row: Row, ordinal: Int): Long = row.getLong(ordinal) override def copyField(from: Row, fromOrdinal: Int, to: MutableRow, toOrdinal: Int): Unit = { to.setLong(toOrdinal, from.getLong(fromOrdinal)) @@ -178,7 +178,7 @@ private[sql] object FLOAT extends NativeColumnType(FloatType, 2, 4) { buffer.putFloat(row.getFloat(ordinal)) } - override def extract(buffer: ByteBuffer) = { + override def extract(buffer: ByteBuffer): Float = { buffer.getFloat() } @@ -190,7 +190,7 @@ private[sql] object FLOAT extends NativeColumnType(FloatType, 2, 4) { row.setFloat(ordinal, value) } - override def getField(row: Row, ordinal: Int) = row.getFloat(ordinal) + override def getField(row: Row, ordinal: Int): Float = row.getFloat(ordinal) override def copyField(from: Row, fromOrdinal: Int, to: MutableRow, toOrdinal: Int): Unit = { to.setFloat(toOrdinal, from.getFloat(fromOrdinal)) @@ -206,7 +206,7 @@ private[sql] object DOUBLE extends NativeColumnType(DoubleType, 3, 8) { buffer.putDouble(row.getDouble(ordinal)) } - override def extract(buffer: ByteBuffer) = { + override def extract(buffer: ByteBuffer): Double = { buffer.getDouble() } @@ -218,7 +218,7 @@ private[sql] object DOUBLE extends NativeColumnType(DoubleType, 3, 8) { row.setDouble(ordinal, value) } - override def getField(row: Row, ordinal: Int) = row.getDouble(ordinal) + override def getField(row: Row, ordinal: Int): Double = row.getDouble(ordinal) override def copyField(from: Row, fromOrdinal: Int, to: MutableRow, toOrdinal: Int): Unit = { to.setDouble(toOrdinal, from.getDouble(fromOrdinal)) @@ -234,7 +234,7 @@ private[sql] object BOOLEAN extends NativeColumnType(BooleanType, 4, 1) { buffer.put(if (row.getBoolean(ordinal)) 1: Byte else 0: Byte) } - override def extract(buffer: ByteBuffer) = buffer.get() == 1 + override def extract(buffer: ByteBuffer): Boolean = buffer.get() == 1 override def extract(buffer: ByteBuffer, row: MutableRow, ordinal: Int): Unit = { row.setBoolean(ordinal, buffer.get() == 1) @@ -244,7 +244,7 @@ private[sql] object BOOLEAN extends NativeColumnType(BooleanType, 4, 1) { row.setBoolean(ordinal, value) } - override def getField(row: Row, ordinal: Int) = row.getBoolean(ordinal) + override def getField(row: Row, ordinal: Int): Boolean = row.getBoolean(ordinal) override def copyField(from: Row, fromOrdinal: Int, to: MutableRow, toOrdinal: Int): Unit = { to.setBoolean(toOrdinal, from.getBoolean(fromOrdinal)) @@ -260,7 +260,7 @@ private[sql] object BYTE extends NativeColumnType(ByteType, 5, 1) { buffer.put(row.getByte(ordinal)) } - override def extract(buffer: ByteBuffer) = { + override def extract(buffer: ByteBuffer): Byte = { buffer.get() } @@ -272,7 +272,7 @@ private[sql] object BYTE extends NativeColumnType(ByteType, 5, 1) { row.setByte(ordinal, value) } - override def getField(row: Row, ordinal: Int) = row.getByte(ordinal) + override def getField(row: Row, ordinal: Int): Byte = row.getByte(ordinal) override def copyField(from: Row, fromOrdinal: Int, to: MutableRow, toOrdinal: Int): Unit = { to.setByte(toOrdinal, from.getByte(fromOrdinal)) @@ -288,7 +288,7 @@ private[sql] object SHORT extends NativeColumnType(ShortType, 6, 2) { buffer.putShort(row.getShort(ordinal)) } - override def extract(buffer: ByteBuffer) = { + override def extract(buffer: ByteBuffer): Short = { buffer.getShort() } @@ -300,7 +300,7 @@ private[sql] object SHORT extends NativeColumnType(ShortType, 6, 2) { row.setShort(ordinal, value) } - override def getField(row: Row, ordinal: Int) = row.getShort(ordinal) + override def getField(row: Row, ordinal: Int): Short = row.getShort(ordinal) override def copyField(from: Row, fromOrdinal: Int, to: MutableRow, toOrdinal: Int): Unit = { to.setShort(toOrdinal, from.getShort(fromOrdinal)) @@ -317,7 +317,7 @@ private[sql] object STRING extends NativeColumnType(StringType, 7, 8) { buffer.putInt(stringBytes.length).put(stringBytes, 0, stringBytes.length) } - override def extract(buffer: ByteBuffer) = { + override def extract(buffer: ByteBuffer): String = { val length = buffer.getInt() val stringBytes = new Array[Byte](length) buffer.get(stringBytes, 0, length) @@ -328,7 +328,7 @@ private[sql] object STRING extends NativeColumnType(StringType, 7, 8) { row.setString(ordinal, value) } - override def getField(row: Row, ordinal: Int) = row.getString(ordinal) + override def getField(row: Row, ordinal: Int): String = row.getString(ordinal) override def copyField(from: Row, fromOrdinal: Int, to: MutableRow, toOrdinal: Int): Unit = { to.setString(toOrdinal, from.getString(fromOrdinal)) @@ -336,7 +336,7 @@ private[sql] object STRING extends NativeColumnType(StringType, 7, 8) { } private[sql] object DATE extends NativeColumnType(DateType, 8, 4) { - override def extract(buffer: ByteBuffer) = { + override def extract(buffer: ByteBuffer): Int = { buffer.getInt } @@ -344,7 +344,7 @@ private[sql] object DATE extends NativeColumnType(DateType, 8, 4) { buffer.putInt(v) } - override def getField(row: Row, ordinal: Int) = { + override def getField(row: Row, ordinal: Int): Int = { row(ordinal).asInstanceOf[Int] } @@ -354,7 +354,7 @@ private[sql] object DATE extends NativeColumnType(DateType, 8, 4) { } private[sql] object TIMESTAMP extends NativeColumnType(TimestampType, 9, 12) { - override def extract(buffer: ByteBuffer) = { + override def extract(buffer: ByteBuffer): Timestamp = { val timestamp = new Timestamp(buffer.getLong()) timestamp.setNanos(buffer.getInt()) timestamp @@ -364,7 +364,7 @@ private[sql] object TIMESTAMP extends NativeColumnType(TimestampType, 9, 12) { buffer.putLong(v.getTime).putInt(v.getNanos) } - override def getField(row: Row, ordinal: Int) = { + override def getField(row: Row, ordinal: Int): Timestamp = { row(ordinal).asInstanceOf[Timestamp] } @@ -405,7 +405,7 @@ private[sql] sealed abstract class ByteArrayColumnType[T <: DataType]( defaultSize: Int) extends ColumnType[T, Array[Byte]](typeId, defaultSize) { - override def actualSize(row: Row, ordinal: Int) = { + override def actualSize(row: Row, ordinal: Int): Int = { getField(row, ordinal).length + 4 } @@ -413,7 +413,7 @@ private[sql] sealed abstract class ByteArrayColumnType[T <: DataType]( buffer.putInt(v.length).put(v, 0, v.length) } - override def extract(buffer: ByteBuffer) = { + override def extract(buffer: ByteBuffer): Array[Byte] = { val length = buffer.getInt() val bytes = new Array[Byte](length) buffer.get(bytes, 0, length) @@ -426,7 +426,9 @@ private[sql] object BINARY extends ByteArrayColumnType[BinaryType.type](11, 16) row(ordinal) = value } - override def getField(row: Row, ordinal: Int) = row(ordinal).asInstanceOf[Array[Byte]] + override def getField(row: Row, ordinal: Int): Array[Byte] = { + row(ordinal).asInstanceOf[Array[Byte]] + } } // Used to process generic objects (all types other than those listed above). Objects should be @@ -437,7 +439,9 @@ private[sql] object GENERIC extends ByteArrayColumnType[DataType](12, 16) { row(ordinal) = SparkSqlSerializer.deserialize[Any](value) } - override def getField(row: Row, ordinal: Int) = SparkSqlSerializer.serialize(row(ordinal)) + override def getField(row: Row, ordinal: Int): Array[Byte] = { + SparkSqlSerializer.serialize(row(ordinal)) + } } private[sql] object ColumnType { diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/InMemoryColumnarTableScan.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/InMemoryColumnarTableScan.scala index 387faee12b3cd..6eee0c86d6a1c 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/InMemoryColumnarTableScan.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/InMemoryColumnarTableScan.scala @@ -19,6 +19,9 @@ package org.apache.spark.sql.columnar import java.nio.ByteBuffer +import org.apache.spark.Accumulator +import org.apache.spark.sql.catalyst.expressions + import scala.collection.mutable.ArrayBuffer import org.apache.spark.rdd.RDD @@ -77,20 +80,23 @@ private[sql] case class InMemoryRelation( _statistics } - override def statistics = if (_statistics == null) { - if (batchStats.value.isEmpty) { - // Underlying columnar RDD hasn't been materialized, no useful statistics information - // available, return the default statistics. - Statistics(sizeInBytes = child.sqlContext.conf.defaultSizeInBytes) + override def statistics: Statistics = { + if (_statistics == null) { + if (batchStats.value.isEmpty) { + // Underlying columnar RDD hasn't been materialized, no useful statistics information + // available, return the default statistics. + Statistics(sizeInBytes = child.sqlContext.conf.defaultSizeInBytes) + } else { + // Underlying columnar RDD has been materialized, required information has also been + // collected via the `batchStats` accumulator, compute the final statistics, + // and update `_statistics`. + _statistics = Statistics(sizeInBytes = computeSizeInBytes) + _statistics + } } else { - // Underlying columnar RDD has been materialized, required information has also been collected - // via the `batchStats` accumulator, compute the final statistics, and update `_statistics`. - _statistics = Statistics(sizeInBytes = computeSizeInBytes) + // Pre-computed statistics _statistics } - } else { - // Pre-computed statistics - _statistics } // If the cached column buffers were not passed in, we calculate them in the constructor. @@ -99,7 +105,7 @@ private[sql] case class InMemoryRelation( buildBuffers() } - def recache() = { + def recache(): Unit = { _cachedColumnBuffers.unpersist() _cachedColumnBuffers = null buildBuffers() @@ -109,7 +115,7 @@ private[sql] case class InMemoryRelation( val output = child.output val cached = child.execute().mapPartitions { rowIterator => new Iterator[CachedBatch] { - def next() = { + def next(): CachedBatch = { val columnBuilders = output.map { attribute => val columnType = ColumnType(attribute.dataType) val initialBufferSize = columnType.defaultSize * batchSize @@ -144,7 +150,7 @@ private[sql] case class InMemoryRelation( CachedBatch(columnBuilders.map(_.build().array()), stats) } - def hasNext = rowIterator.hasNext + def hasNext: Boolean = rowIterator.hasNext } }.persist(storageLevel) @@ -158,9 +164,9 @@ private[sql] case class InMemoryRelation( _cachedColumnBuffers, statisticsToBePropagated) } - override def children = Seq.empty + override def children: Seq[LogicalPlan] = Seq.empty - override def newInstance() = { + override def newInstance(): this.type = { new InMemoryRelation( output.map(_.newInstance()), useCompression, @@ -172,7 +178,7 @@ private[sql] case class InMemoryRelation( statisticsToBePropagated).asInstanceOf[this.type] } - def cachedColumnBuffers = _cachedColumnBuffers + def cachedColumnBuffers: RDD[CachedBatch] = _cachedColumnBuffers override protected def otherCopyArgs: Seq[AnyRef] = Seq(_cachedColumnBuffers, statisticsToBePropagated) @@ -220,7 +226,7 @@ private[sql] case class InMemoryColumnarTableScan( case IsNotNull(a: Attribute) => statsFor(a).count - statsFor(a).nullCount > 0 } - val partitionFilters = { + val partitionFilters: Seq[Expression] = { predicates.flatMap { p => val filter = buildFilter.lift(p) val boundFilter = @@ -239,12 +245,12 @@ private[sql] case class InMemoryColumnarTableScan( } // Accumulators used for testing purposes - val readPartitions = sparkContext.accumulator(0) - val readBatches = sparkContext.accumulator(0) + val readPartitions: Accumulator[Int] = sparkContext.accumulator(0) + val readBatches: Accumulator[Int] = sparkContext.accumulator(0) private val inMemoryPartitionPruningEnabled = sqlContext.conf.inMemoryPartitionPruning - override def execute() = { + override def execute(): RDD[Row] = { readPartitions.setValue(0) readBatches.setValue(0) @@ -271,7 +277,7 @@ private[sql] case class InMemoryColumnarTableScan( val nextRow = new SpecificMutableRow(requestedColumnDataTypes) - def cachedBatchesToRows(cacheBatches: Iterator[CachedBatch]) = { + def cachedBatchesToRows(cacheBatches: Iterator[CachedBatch]): Iterator[Row] = { val rows = cacheBatches.flatMap { cachedBatch => // Build column accessors val columnAccessors = requestedColumnIndices.map { batchColumnIndex => @@ -283,7 +289,7 @@ private[sql] case class InMemoryColumnarTableScan( // Extract rows via column accessors new Iterator[Row] { private[this] val rowLen = nextRow.length - override def next() = { + override def next(): Row = { var i = 0 while (i < rowLen) { columnAccessors(i).extractTo(nextRow, i) @@ -292,7 +298,7 @@ private[sql] case class InMemoryColumnarTableScan( nextRow } - override def hasNext = columnAccessors(0).hasNext + override def hasNext: Boolean = columnAccessors(0).hasNext } } @@ -308,7 +314,7 @@ private[sql] case class InMemoryColumnarTableScan( if (inMemoryPartitionPruningEnabled) { cachedBatchIterator.filter { cachedBatch => if (!partitionFilter(cachedBatch.stats)) { - def statsString = relation.partitionStatistics.schema + def statsString: String = relation.partitionStatistics.schema .zip(cachedBatch.stats.toSeq) .map { case (a, s) => s"${a.name}: $s" } .mkString(", ") diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/NullableColumnAccessor.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/NullableColumnAccessor.scala index 965782a40031b..4d35650d4b1eb 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/NullableColumnAccessor.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/NullableColumnAccessor.scala @@ -55,5 +55,5 @@ private[sql] trait NullableColumnAccessor extends ColumnAccessor { pos += 1 } - abstract override def hasNext = seenNulls < nullCount || super.hasNext + abstract override def hasNext: Boolean = seenNulls < nullCount || super.hasNext } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/CompressibleColumnAccessor.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/CompressibleColumnAccessor.scala index 7dff9deac8dc0..d0b602a834dfe 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/CompressibleColumnAccessor.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/CompressibleColumnAccessor.scala @@ -26,12 +26,12 @@ private[sql] trait CompressibleColumnAccessor[T <: NativeType] extends ColumnAcc private var decoder: Decoder[T] = _ - abstract override protected def initialize() = { + abstract override protected def initialize(): Unit = { super.initialize() decoder = CompressionScheme(underlyingBuffer.getInt()).decoder(buffer, columnType) } - abstract override def hasNext = super.hasNext || decoder.hasNext + abstract override def hasNext: Boolean = super.hasNext || decoder.hasNext override def extractSingle(row: MutableRow, ordinal: Int): Unit = { decoder.next(row, ordinal) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/CompressibleColumnBuilder.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/CompressibleColumnBuilder.scala index aead768ecdf0a..b9cfc5df550d1 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/CompressibleColumnBuilder.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/CompressibleColumnBuilder.scala @@ -81,7 +81,7 @@ private[sql] trait CompressibleColumnBuilder[T <: NativeType] } } - override def build() = { + override def build(): ByteBuffer = { val nonNullBuffer = buildNonNulls() val typeId = nonNullBuffer.getInt() val encoder: Encoder[T] = { diff --git a/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/compressionSchemes.scala b/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/compressionSchemes.scala index 68a5b1de7691b..8727d71c48bb7 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/compressionSchemes.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/columnar/compression/compressionSchemes.scala @@ -33,22 +33,23 @@ import org.apache.spark.util.Utils private[sql] case object PassThrough extends CompressionScheme { override val typeId = 0 - override def supports(columnType: ColumnType[_, _]) = true + override def supports(columnType: ColumnType[_, _]): Boolean = true - override def encoder[T <: NativeType](columnType: NativeColumnType[T]) = { + override def encoder[T <: NativeType](columnType: NativeColumnType[T]): Encoder[T] = { new this.Encoder[T](columnType) } - override def decoder[T <: NativeType](buffer: ByteBuffer, columnType: NativeColumnType[T]) = { + override def decoder[T <: NativeType]( + buffer: ByteBuffer, columnType: NativeColumnType[T]): Decoder[T] = { new this.Decoder(buffer, columnType) } class Encoder[T <: NativeType](columnType: NativeColumnType[T]) extends compression.Encoder[T] { - override def uncompressedSize = 0 + override def uncompressedSize: Int = 0 - override def compressedSize = 0 + override def compressedSize: Int = 0 - override def compress(from: ByteBuffer, to: ByteBuffer) = { + override def compress(from: ByteBuffer, to: ByteBuffer): ByteBuffer = { // Writes compression type ID and copies raw contents to.putInt(PassThrough.typeId).put(from).rewind() to @@ -62,22 +63,23 @@ private[sql] case object PassThrough extends CompressionScheme { columnType.extract(buffer, row, ordinal) } - override def hasNext = buffer.hasRemaining + override def hasNext: Boolean = buffer.hasRemaining } } private[sql] case object RunLengthEncoding extends CompressionScheme { override val typeId = 1 - override def encoder[T <: NativeType](columnType: NativeColumnType[T]) = { + override def encoder[T <: NativeType](columnType: NativeColumnType[T]): Encoder[T] = { new this.Encoder[T](columnType) } - override def decoder[T <: NativeType](buffer: ByteBuffer, columnType: NativeColumnType[T]) = { + override def decoder[T <: NativeType]( + buffer: ByteBuffer, columnType: NativeColumnType[T]): Decoder[T] = { new this.Decoder(buffer, columnType) } - override def supports(columnType: ColumnType[_, _]) = columnType match { + override def supports(columnType: ColumnType[_, _]): Boolean = columnType match { case INT | LONG | SHORT | BYTE | STRING | BOOLEAN => true case _ => false } @@ -90,9 +92,9 @@ private[sql] case object RunLengthEncoding extends CompressionScheme { private val lastValue = new SpecificMutableRow(Seq(columnType.dataType)) private var lastRun = 0 - override def uncompressedSize = _uncompressedSize + override def uncompressedSize: Int = _uncompressedSize - override def compressedSize = _compressedSize + override def compressedSize: Int = _compressedSize override def gatherCompressibilityStats(row: Row, ordinal: Int): Unit = { val value = columnType.getField(row, ordinal) @@ -114,7 +116,7 @@ private[sql] case object RunLengthEncoding extends CompressionScheme { } } - override def compress(from: ByteBuffer, to: ByteBuffer) = { + override def compress(from: ByteBuffer, to: ByteBuffer): ByteBuffer = { to.putInt(RunLengthEncoding.typeId) if (from.hasRemaining) { @@ -169,7 +171,7 @@ private[sql] case object RunLengthEncoding extends CompressionScheme { columnType.setField(row, ordinal, currentValue) } - override def hasNext = valueCount < run || buffer.hasRemaining + override def hasNext: Boolean = valueCount < run || buffer.hasRemaining } } @@ -179,15 +181,16 @@ private[sql] case object DictionaryEncoding extends CompressionScheme { // 32K unique values allowed val MAX_DICT_SIZE = Short.MaxValue - override def decoder[T <: NativeType](buffer: ByteBuffer, columnType: NativeColumnType[T]) = { + override def decoder[T <: NativeType](buffer: ByteBuffer, columnType: NativeColumnType[T]) + : Decoder[T] = { new this.Decoder(buffer, columnType) } - override def encoder[T <: NativeType](columnType: NativeColumnType[T]) = { + override def encoder[T <: NativeType](columnType: NativeColumnType[T]): Encoder[T] = { new this.Encoder[T](columnType) } - override def supports(columnType: ColumnType[_, _]) = columnType match { + override def supports(columnType: ColumnType[_, _]): Boolean = columnType match { case INT | LONG | STRING => true case _ => false } @@ -237,7 +240,7 @@ private[sql] case object DictionaryEncoding extends CompressionScheme { } } - override def compress(from: ByteBuffer, to: ByteBuffer) = { + override def compress(from: ByteBuffer, to: ByteBuffer): ByteBuffer = { if (overflow) { throw new IllegalStateException( "Dictionary encoding should not be used because of dictionary overflow.") @@ -260,9 +263,9 @@ private[sql] case object DictionaryEncoding extends CompressionScheme { to } - override def uncompressedSize = _uncompressedSize + override def uncompressedSize: Int = _uncompressedSize - override def compressedSize = if (overflow) Int.MaxValue else dictionarySize + count * 2 + override def compressedSize: Int = if (overflow) Int.MaxValue else dictionarySize + count * 2 } class Decoder[T <: NativeType](buffer: ByteBuffer, columnType: NativeColumnType[T]) @@ -284,7 +287,7 @@ private[sql] case object DictionaryEncoding extends CompressionScheme { columnType.setField(row, ordinal, dictionary(buffer.getShort())) } - override def hasNext = buffer.hasRemaining + override def hasNext: Boolean = buffer.hasRemaining } } @@ -293,15 +296,16 @@ private[sql] case object BooleanBitSet extends CompressionScheme { val BITS_PER_LONG = 64 - override def decoder[T <: NativeType](buffer: ByteBuffer, columnType: NativeColumnType[T]) = { + override def decoder[T <: NativeType](buffer: ByteBuffer, columnType: NativeColumnType[T]) + : compression.Decoder[T] = { new this.Decoder(buffer).asInstanceOf[compression.Decoder[T]] } - override def encoder[T <: NativeType](columnType: NativeColumnType[T]) = { + override def encoder[T <: NativeType](columnType: NativeColumnType[T]): compression.Encoder[T] = { (new this.Encoder).asInstanceOf[compression.Encoder[T]] } - override def supports(columnType: ColumnType[_, _]) = columnType == BOOLEAN + override def supports(columnType: ColumnType[_, _]): Boolean = columnType == BOOLEAN class Encoder extends compression.Encoder[BooleanType.type] { private var _uncompressedSize = 0 @@ -310,7 +314,7 @@ private[sql] case object BooleanBitSet extends CompressionScheme { _uncompressedSize += BOOLEAN.defaultSize } - override def compress(from: ByteBuffer, to: ByteBuffer) = { + override def compress(from: ByteBuffer, to: ByteBuffer): ByteBuffer = { to.putInt(BooleanBitSet.typeId) // Total element count (1 byte per Boolean value) .putInt(from.remaining) @@ -347,9 +351,9 @@ private[sql] case object BooleanBitSet extends CompressionScheme { to } - override def uncompressedSize = _uncompressedSize + override def uncompressedSize: Int = _uncompressedSize - override def compressedSize = { + override def compressedSize: Int = { val extra = if (_uncompressedSize % BITS_PER_LONG == 0) 0 else 1 (_uncompressedSize / BITS_PER_LONG + extra) * 8 + 4 } @@ -380,22 +384,23 @@ private[sql] case object BooleanBitSet extends CompressionScheme { private[sql] case object IntDelta extends CompressionScheme { override def typeId: Int = 4 - override def decoder[T <: NativeType](buffer: ByteBuffer, columnType: NativeColumnType[T]) = { + override def decoder[T <: NativeType](buffer: ByteBuffer, columnType: NativeColumnType[T]) + : compression.Decoder[T] = { new Decoder(buffer, INT).asInstanceOf[compression.Decoder[T]] } - override def encoder[T <: NativeType](columnType: NativeColumnType[T]) = { + override def encoder[T <: NativeType](columnType: NativeColumnType[T]): compression.Encoder[T] = { (new Encoder).asInstanceOf[compression.Encoder[T]] } - override def supports(columnType: ColumnType[_, _]) = columnType == INT + override def supports(columnType: ColumnType[_, _]): Boolean = columnType == INT class Encoder extends compression.Encoder[IntegerType.type] { protected var _compressedSize: Int = 0 protected var _uncompressedSize: Int = 0 - override def compressedSize = _compressedSize - override def uncompressedSize = _uncompressedSize + override def compressedSize: Int = _compressedSize + override def uncompressedSize: Int = _uncompressedSize private var prevValue: Int = _ @@ -459,22 +464,23 @@ private[sql] case object IntDelta extends CompressionScheme { private[sql] case object LongDelta extends CompressionScheme { override def typeId: Int = 5 - override def decoder[T <: NativeType](buffer: ByteBuffer, columnType: NativeColumnType[T]) = { + override def decoder[T <: NativeType](buffer: ByteBuffer, columnType: NativeColumnType[T]) + : compression.Decoder[T] = { new Decoder(buffer, LONG).asInstanceOf[compression.Decoder[T]] } - override def encoder[T <: NativeType](columnType: NativeColumnType[T]) = { + override def encoder[T <: NativeType](columnType: NativeColumnType[T]): compression.Encoder[T] = { (new Encoder).asInstanceOf[compression.Encoder[T]] } - override def supports(columnType: ColumnType[_, _]) = columnType == LONG + override def supports(columnType: ColumnType[_, _]): Boolean = columnType == LONG class Encoder extends compression.Encoder[LongType.type] { protected var _compressedSize: Int = 0 protected var _uncompressedSize: Int = 0 - override def compressedSize = _compressedSize - override def uncompressedSize = _uncompressedSize + override def compressedSize: Int = _compressedSize + override def uncompressedSize: Int = _uncompressedSize private var prevValue: Long = _ diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/Aggregate.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/Aggregate.scala index ad44a01d0e164..18b1ba4c5c4b9 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/Aggregate.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/Aggregate.scala @@ -21,6 +21,7 @@ import java.util.HashMap import org.apache.spark.annotation.DeveloperApi import org.apache.spark.SparkContext +import org.apache.spark.rdd.RDD import org.apache.spark.sql.catalyst.errors._ import org.apache.spark.sql.catalyst.expressions._ import org.apache.spark.sql.catalyst.plans.physical._ @@ -45,7 +46,7 @@ case class Aggregate( child: SparkPlan) extends UnaryNode { - override def requiredChildDistribution = + override def requiredChildDistribution: List[Distribution] = { if (partial) { UnspecifiedDistribution :: Nil } else { @@ -55,8 +56,9 @@ case class Aggregate( ClusteredDistribution(groupingExpressions) :: Nil } } + } - override def output = aggregateExpressions.map(_.toAttribute) + override def output: Seq[Attribute] = aggregateExpressions.map(_.toAttribute) /** * An aggregate that needs to be computed for each row in a group. @@ -119,7 +121,7 @@ case class Aggregate( } } - override def execute() = attachTree(this, "execute") { + override def execute(): RDD[Row] = attachTree(this, "execute") { if (groupingExpressions.isEmpty) { child.execute().mapPartitions { iter => val buffer = newAggregateBuffer() diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/Exchange.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/Exchange.scala index 7c0b72aab448e..437408d30bfd2 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/Exchange.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/Exchange.scala @@ -19,11 +19,12 @@ package org.apache.spark.sql.execution import org.apache.spark.annotation.DeveloperApi import org.apache.spark.shuffle.sort.SortShuffleManager +import org.apache.spark.sql.catalyst.expressions import org.apache.spark.{SparkEnv, HashPartitioner, RangePartitioner, SparkConf} -import org.apache.spark.rdd.ShuffledRDD +import org.apache.spark.rdd.{RDD, ShuffledRDD} import org.apache.spark.sql.{SQLContext, Row} import org.apache.spark.sql.catalyst.errors.attachTree -import org.apache.spark.sql.catalyst.expressions.RowOrdering +import org.apache.spark.sql.catalyst.expressions.{Attribute, RowOrdering} import org.apache.spark.sql.catalyst.plans.physical._ import org.apache.spark.sql.catalyst.rules.Rule import org.apache.spark.util.MutablePair @@ -34,9 +35,9 @@ import org.apache.spark.util.MutablePair @DeveloperApi case class Exchange(newPartitioning: Partitioning, child: SparkPlan) extends UnaryNode { - override def outputPartitioning = newPartitioning + override def outputPartitioning: Partitioning = newPartitioning - override def output = child.output + override def output: Seq[Attribute] = child.output /** We must copy rows when sort based shuffle is on */ protected def sortBasedShuffleOn = SparkEnv.get.shuffleManager.isInstanceOf[SortShuffleManager] @@ -44,7 +45,7 @@ case class Exchange(newPartitioning: Partitioning, child: SparkPlan) extends Una private val bypassMergeThreshold = child.sqlContext.sparkContext.conf.getInt("spark.shuffle.sort.bypassMergeThreshold", 200) - override def execute() = attachTree(this , "execute") { + override def execute(): RDD[Row] = attachTree(this , "execute") { newPartitioning match { case HashPartitioning(expressions, numPartitions) => // TODO: Eliminate redundant expressions in grouping key and value. @@ -123,13 +124,13 @@ case class Exchange(newPartitioning: Partitioning, child: SparkPlan) extends Una */ private[sql] case class AddExchange(sqlContext: SQLContext) extends Rule[SparkPlan] { // TODO: Determine the number of partitions. - def numPartitions = sqlContext.conf.numShufflePartitions + def numPartitions: Int = sqlContext.conf.numShufflePartitions def apply(plan: SparkPlan): SparkPlan = plan.transformUp { case operator: SparkPlan => // Check if every child's outputPartitioning satisfies the corresponding // required data distribution. - def meetsRequirements = + def meetsRequirements: Boolean = !operator.requiredChildDistribution.zip(operator.children).map { case (required, child) => val valid = child.outputPartitioning.satisfies(required) @@ -147,7 +148,7 @@ private[sql] case class AddExchange(sqlContext: SQLContext) extends Rule[SparkPl // datasets are both clustered by "a", but these two outputPartitionings are not // compatible. // TODO: ASSUMES TRANSITIVITY? - def compatible = + def compatible: Boolean = !operator.children .map(_.outputPartitioning) .sliding(2) @@ -158,7 +159,7 @@ private[sql] case class AddExchange(sqlContext: SQLContext) extends Rule[SparkPl // Check if the partitioning we want to ensure is the same as the child's output // partitioning. If so, we do not need to add the Exchange operator. - def addExchangeIfNecessary(partitioning: Partitioning, child: SparkPlan) = + def addExchangeIfNecessary(partitioning: Partitioning, child: SparkPlan): SparkPlan = if (child.outputPartitioning != partitioning) Exchange(partitioning, child) else child if (meetsRequirements && compatible) { diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/ExistingRDD.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/ExistingRDD.scala index 248dc1512b4d3..d8955725e59b1 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/ExistingRDD.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/ExistingRDD.scala @@ -26,6 +26,8 @@ import org.apache.spark.sql.catalyst.expressions.{Attribute, GenericMutableRow} import org.apache.spark.sql.catalyst.plans.logical.{LogicalPlan, Statistics} import org.apache.spark.sql.types.StructType +import scala.collection.immutable + /** * :: DeveloperApi :: */ @@ -58,17 +60,17 @@ object RDDConversions { case class LogicalRDD(output: Seq[Attribute], rdd: RDD[Row])(sqlContext: SQLContext) extends LogicalPlan with MultiInstanceRelation { - override def children = Nil + override def children: Seq[LogicalPlan] = Nil - override def newInstance() = + override def newInstance(): LogicalRDD.this.type = LogicalRDD(output.map(_.newInstance()), rdd)(sqlContext).asInstanceOf[this.type] - override def sameResult(plan: LogicalPlan) = plan match { + override def sameResult(plan: LogicalPlan): Boolean = plan match { case LogicalRDD(_, otherRDD) => rdd.id == otherRDD.id case _ => false } - @transient override lazy val statistics = Statistics( + @transient override lazy val statistics: Statistics = Statistics( // TODO: Instead of returning a default value here, find a way to return a meaningful size // estimate for RDDs. See PR 1238 for more discussions. sizeInBytes = BigInt(sqlContext.conf.defaultSizeInBytes) @@ -77,24 +79,24 @@ case class LogicalRDD(output: Seq[Attribute], rdd: RDD[Row])(sqlContext: SQLCont /** Physical plan node for scanning data from an RDD. */ case class PhysicalRDD(output: Seq[Attribute], rdd: RDD[Row]) extends LeafNode { - override def execute() = rdd + override def execute(): RDD[Row] = rdd } /** Logical plan node for scanning data from a local collection. */ case class LogicalLocalTable(output: Seq[Attribute], rows: Seq[Row])(sqlContext: SQLContext) extends LogicalPlan with MultiInstanceRelation { - override def children = Nil + override def children: Seq[LogicalPlan] = Nil - override def newInstance() = + override def newInstance(): this.type = LogicalLocalTable(output.map(_.newInstance()), rows)(sqlContext).asInstanceOf[this.type] - override def sameResult(plan: LogicalPlan) = plan match { + override def sameResult(plan: LogicalPlan): Boolean = plan match { case LogicalRDD(_, otherRDD) => rows == rows case _ => false } - @transient override lazy val statistics = Statistics( + @transient override lazy val statistics: Statistics = Statistics( // TODO: Improve the statistics estimation. // This is made small enough so it can be broadcasted. sizeInBytes = sqlContext.conf.autoBroadcastJoinThreshold - 1 diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/Expand.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/Expand.scala index 95172420608f9..575849481faad 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/Expand.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/Expand.scala @@ -18,6 +18,7 @@ package org.apache.spark.sql.execution import org.apache.spark.annotation.DeveloperApi +import org.apache.spark.rdd.RDD import org.apache.spark.sql.SQLContext import org.apache.spark.sql.catalyst.errors._ import org.apache.spark.sql.catalyst.expressions._ @@ -42,7 +43,7 @@ case class Expand( // as UNKNOWN partitioning override def outputPartitioning: Partitioning = UnknownPartitioning(0) - override def execute() = attachTree(this, "execute") { + override def execute(): RDD[Row] = attachTree(this, "execute") { child.execute().mapPartitions { iter => // TODO Move out projection objects creation and transfer to // workers via closure. However we can't assume the Projection @@ -55,7 +56,7 @@ case class Expand( private[this] var idx = -1 // -1 means the initial state private[this] var input: Row = _ - override final def hasNext = (-1 < idx && idx < groups.length) || iter.hasNext + override final def hasNext: Boolean = (-1 < idx && idx < groups.length) || iter.hasNext override final def next(): Row = { if (idx <= 0) { diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/Generate.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/Generate.scala index 38877c28de3a8..12271048bb39c 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/Generate.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/Generate.scala @@ -18,6 +18,7 @@ package org.apache.spark.sql.execution import org.apache.spark.annotation.DeveloperApi +import org.apache.spark.rdd.RDD import org.apache.spark.sql.catalyst.expressions._ /** @@ -54,7 +55,7 @@ case class Generate( val boundGenerator = BindReferences.bindReference(generator, child.output) - override def execute() = { + override def execute(): RDD[Row] = { if (join) { child.execute().mapPartitions { iter => val nullValues = Seq.fill(generator.output.size)(Literal(null)) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/GeneratedAggregate.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/GeneratedAggregate.scala index 4abe26fe4afc6..89682d25ca7dc 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/GeneratedAggregate.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/GeneratedAggregate.scala @@ -18,6 +18,7 @@ package org.apache.spark.sql.execution import org.apache.spark.annotation.DeveloperApi +import org.apache.spark.rdd.RDD import org.apache.spark.sql.catalyst.trees._ import org.apache.spark.sql.catalyst.expressions._ import org.apache.spark.sql.catalyst.plans.physical._ @@ -49,7 +50,7 @@ case class GeneratedAggregate( child: SparkPlan) extends UnaryNode { - override def requiredChildDistribution = + override def requiredChildDistribution: Seq[Distribution] = if (partial) { UnspecifiedDistribution :: Nil } else { @@ -60,9 +61,9 @@ case class GeneratedAggregate( } } - override def output = aggregateExpressions.map(_.toAttribute) + override def output: Seq[Attribute] = aggregateExpressions.map(_.toAttribute) - override def execute() = { + override def execute(): RDD[Row] = { val aggregatesToCompute = aggregateExpressions.flatMap { a => a.collect { case agg: AggregateExpression => agg} } @@ -271,9 +272,9 @@ case class GeneratedAggregate( private[this] val resultIterator = buffers.entrySet.iterator() private[this] val resultProjection = resultProjectionBuilder() - def hasNext = resultIterator.hasNext + def hasNext: Boolean = resultIterator.hasNext - def next() = { + def next(): Row = { val currentGroup = resultIterator.next() resultProjection(joinedRow(currentGroup.getKey, currentGroup.getValue)) } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/LocalTableScan.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/LocalTableScan.scala index d3a18b37d52b9..5bd699a2fa949 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/LocalTableScan.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/LocalTableScan.scala @@ -17,6 +17,7 @@ package org.apache.spark.sql.execution +import org.apache.spark.rdd.RDD import org.apache.spark.sql.Row import org.apache.spark.sql.catalyst.ScalaReflection import org.apache.spark.sql.catalyst.expressions.Attribute @@ -29,11 +30,11 @@ case class LocalTableScan(output: Seq[Attribute], rows: Seq[Row]) extends LeafNo private lazy val rdd = sqlContext.sparkContext.parallelize(rows) - override def execute() = rdd + override def execute(): RDD[Row] = rdd - override def executeCollect() = + override def executeCollect(): Array[Row] = rows.map(ScalaReflection.convertRowToScala(_, schema)).toArray - override def executeTake(limit: Int) = + override def executeTake(limit: Int): Array[Row] = rows.map(ScalaReflection.convertRowToScala(_, schema)).take(limit).toArray } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkPlan.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkPlan.scala index 052766c20abc2..d239637cd4b4e 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkPlan.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkPlan.scala @@ -67,6 +67,7 @@ abstract class SparkPlan extends QueryPlan[SparkPlan] with Logging with Serializ // TODO: Move to `DistributedPlan` /** Specifies how data is partitioned across different nodes in the cluster. */ def outputPartitioning: Partitioning = UnknownPartitioning(0) // TODO: WRONG WIDTH! + /** Specifies any partition requirements on the input data for this operator. */ def requiredChildDistribution: Seq[Distribution] = Seq.fill(children.size)(UnspecifiedDistribution) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkSqlSerializer.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkSqlSerializer.scala index 30564e14fa896..c4534fd5f67e4 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkSqlSerializer.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkSqlSerializer.scala @@ -74,7 +74,7 @@ private[execution] class KryoResourcePool(size: Int) new KryoSerializer(sparkConf) } - def newInstance() = ser.newInstance() + def newInstance(): SerializerInstance = ser.newInstance() } private[sql] object SparkSqlSerializer { diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkStrategies.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkStrategies.scala index 5281c7502556a..2b581152e5f77 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkStrategies.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/SparkStrategies.scala @@ -154,7 +154,7 @@ private[sql] abstract class SparkStrategies extends QueryPlanner[SparkPlan] { case _ => Nil } - def canBeCodeGened(aggs: Seq[AggregateExpression]) = !aggs.exists { + def canBeCodeGened(aggs: Seq[AggregateExpression]): Boolean = !aggs.exists { case _: Sum | _: Count | _: Max | _: CombineSetsAndCount => false // The generated set implementation is pretty limited ATM. case CollectHashSet(exprs) if exprs.size == 1 && @@ -162,7 +162,7 @@ private[sql] abstract class SparkStrategies extends QueryPlanner[SparkPlan] { case _ => true } - def allAggregates(exprs: Seq[Expression]) = + def allAggregates(exprs: Seq[Expression]): Seq[AggregateExpression] = exprs.flatMap(_.collect { case a: AggregateExpression => a }) } @@ -257,7 +257,7 @@ private[sql] abstract class SparkStrategies extends QueryPlanner[SparkPlan] { // Can we automate these 'pass through' operations? object BasicOperators extends Strategy { - def numPartitions = self.numPartitions + def numPartitions: Int = self.numPartitions def apply(plan: LogicalPlan): Seq[SparkPlan] = plan match { case r: RunnableCommand => ExecutedCommand(r) :: Nil diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/basicOperators.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/basicOperators.scala index 710268584cff1..20c9bc3e75542 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/basicOperators.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/basicOperators.scala @@ -24,7 +24,7 @@ import org.apache.spark.shuffle.sort.SortShuffleManager import org.apache.spark.sql.catalyst.ScalaReflection import org.apache.spark.sql.catalyst.errors._ import org.apache.spark.sql.catalyst.expressions._ -import org.apache.spark.sql.catalyst.plans.physical.{ClusteredDistribution, OrderedDistribution, SinglePartition, UnspecifiedDistribution} +import org.apache.spark.sql.catalyst.plans.physical._ import org.apache.spark.util.MutablePair import org.apache.spark.util.collection.ExternalSorter @@ -33,11 +33,11 @@ import org.apache.spark.util.collection.ExternalSorter */ @DeveloperApi case class Project(projectList: Seq[NamedExpression], child: SparkPlan) extends UnaryNode { - override def output = projectList.map(_.toAttribute) + override def output: Seq[Attribute] = projectList.map(_.toAttribute) @transient lazy val buildProjection = newMutableProjection(projectList, child.output) - override def execute() = child.execute().mapPartitions { iter => + override def execute(): RDD[Row] = child.execute().mapPartitions { iter => val resuableProjection = buildProjection() iter.map(resuableProjection) } @@ -48,11 +48,11 @@ case class Project(projectList: Seq[NamedExpression], child: SparkPlan) extends */ @DeveloperApi case class Filter(condition: Expression, child: SparkPlan) extends UnaryNode { - override def output = child.output + override def output: Seq[Attribute] = child.output - @transient lazy val conditionEvaluator = newPredicate(condition, child.output) + @transient lazy val conditionEvaluator: (Row) => Boolean = newPredicate(condition, child.output) - override def execute() = child.execute().mapPartitions { iter => + override def execute(): RDD[Row] = child.execute().mapPartitions { iter => iter.filter(conditionEvaluator) } } @@ -64,10 +64,12 @@ case class Filter(condition: Expression, child: SparkPlan) extends UnaryNode { case class Sample(fraction: Double, withReplacement: Boolean, seed: Long, child: SparkPlan) extends UnaryNode { - override def output = child.output + override def output: Seq[Attribute] = child.output // TODO: How to pick seed? - override def execute() = child.execute().map(_.copy()).sample(withReplacement, fraction, seed) + override def execute(): RDD[Row] = { + child.execute().map(_.copy()).sample(withReplacement, fraction, seed) + } } /** @@ -76,8 +78,8 @@ case class Sample(fraction: Double, withReplacement: Boolean, seed: Long, child: @DeveloperApi case class Union(children: Seq[SparkPlan]) extends SparkPlan { // TODO: attributes output by union should be distinct for nullability purposes - override def output = children.head.output - override def execute() = sparkContext.union(children.map(_.execute())) + override def output: Seq[Attribute] = children.head.output + override def execute(): RDD[Row] = sparkContext.union(children.map(_.execute())) } /** @@ -97,12 +99,12 @@ case class Limit(limit: Int, child: SparkPlan) /** We must copy rows when sort based shuffle is on */ private def sortBasedShuffleOn = SparkEnv.get.shuffleManager.isInstanceOf[SortShuffleManager] - override def output = child.output - override def outputPartitioning = SinglePartition + override def output: Seq[Attribute] = child.output + override def outputPartitioning: Partitioning = SinglePartition override def executeCollect(): Array[Row] = child.executeTake(limit) - override def execute() = { + override def execute(): RDD[Row] = { val rdd: RDD[_ <: Product2[Boolean, Row]] = if (sortBasedShuffleOn) { child.execute().mapPartitions { iter => iter.take(limit).map(row => (false, row.copy())) @@ -129,20 +131,21 @@ case class Limit(limit: Int, child: SparkPlan) @DeveloperApi case class TakeOrdered(limit: Int, sortOrder: Seq[SortOrder], child: SparkPlan) extends UnaryNode { - override def output = child.output - override def outputPartitioning = SinglePartition + override def output: Seq[Attribute] = child.output + + override def outputPartitioning: Partitioning = SinglePartition - val ord = new RowOrdering(sortOrder, child.output) + private val ord: RowOrdering = new RowOrdering(sortOrder, child.output) - private def collectData() = child.execute().map(_.copy()).takeOrdered(limit)(ord) + private def collectData(): Array[Row] = child.execute().map(_.copy()).takeOrdered(limit)(ord) // TODO: Is this copying for no reason? - override def executeCollect() = + override def executeCollect(): Array[Row] = collectData().map(ScalaReflection.convertRowToScala(_, this.schema)) // TODO: Terminal split should be implemented differently from non-terminal split. // TODO: Pick num splits based on |limit|. - override def execute() = sparkContext.makeRDD(collectData(), 1) + override def execute(): RDD[Row] = sparkContext.makeRDD(collectData(), 1) } /** @@ -157,17 +160,17 @@ case class Sort( global: Boolean, child: SparkPlan) extends UnaryNode { - override def requiredChildDistribution = + override def requiredChildDistribution: Seq[Distribution] = if (global) OrderedDistribution(sortOrder) :: Nil else UnspecifiedDistribution :: Nil - override def execute() = attachTree(this, "sort") { + override def execute(): RDD[Row] = attachTree(this, "sort") { child.execute().mapPartitions( { iterator => val ordering = newOrdering(sortOrder, child.output) iterator.map(_.copy()).toArray.sorted(ordering).iterator }, preservesPartitioning = true) } - override def output = child.output + override def output: Seq[Attribute] = child.output } /** @@ -182,10 +185,11 @@ case class ExternalSort( global: Boolean, child: SparkPlan) extends UnaryNode { - override def requiredChildDistribution = + + override def requiredChildDistribution: Seq[Distribution] = if (global) OrderedDistribution(sortOrder) :: Nil else UnspecifiedDistribution :: Nil - override def execute() = attachTree(this, "sort") { + override def execute(): RDD[Row] = attachTree(this, "sort") { child.execute().mapPartitions( { iterator => val ordering = newOrdering(sortOrder, child.output) val sorter = new ExternalSorter[Row, Null, Row](ordering = Some(ordering)) @@ -194,7 +198,7 @@ case class ExternalSort( }, preservesPartitioning = true) } - override def output = child.output + override def output: Seq[Attribute] = child.output } /** @@ -206,12 +210,12 @@ case class ExternalSort( */ @DeveloperApi case class Distinct(partial: Boolean, child: SparkPlan) extends UnaryNode { - override def output = child.output + override def output: Seq[Attribute] = child.output - override def requiredChildDistribution = + override def requiredChildDistribution: Seq[Distribution] = if (partial) UnspecifiedDistribution :: Nil else ClusteredDistribution(child.output) :: Nil - override def execute() = { + override def execute(): RDD[Row] = { child.execute().mapPartitions { iter => val hashSet = new scala.collection.mutable.HashSet[Row]() @@ -236,9 +240,9 @@ case class Distinct(partial: Boolean, child: SparkPlan) extends UnaryNode { */ @DeveloperApi case class Except(left: SparkPlan, right: SparkPlan) extends BinaryNode { - override def output = left.output + override def output: Seq[Attribute] = left.output - override def execute() = { + override def execute(): RDD[Row] = { left.execute().map(_.copy()).subtract(right.execute().map(_.copy())) } } @@ -250,9 +254,9 @@ case class Except(left: SparkPlan, right: SparkPlan) extends BinaryNode { */ @DeveloperApi case class Intersect(left: SparkPlan, right: SparkPlan) extends BinaryNode { - override def output = children.head.output + override def output: Seq[Attribute] = children.head.output - override def execute() = { + override def execute(): RDD[Row] = { left.execute().map(_.copy()).intersection(right.execute().map(_.copy())) } } @@ -265,6 +269,7 @@ case class Intersect(left: SparkPlan, right: SparkPlan) extends BinaryNode { */ @DeveloperApi case class OutputFaker(output: Seq[Attribute], child: SparkPlan) extends SparkPlan { - def children = child :: Nil - def execute() = child.execute() + def children: Seq[SparkPlan] = child :: Nil + + def execute(): RDD[Row] = child.execute() } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/commands.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/commands.scala index a11232142d0fb..fad7a281dc1e2 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/commands.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/commands.scala @@ -26,7 +26,6 @@ import org.apache.spark.sql.catalyst.errors.TreeNodeException import org.apache.spark.sql.catalyst.expressions.{AttributeReference, Row, Attribute} import org.apache.spark.sql.catalyst.plans.logical import org.apache.spark.sql.catalyst.plans.logical.LogicalPlan -import scala.collection.mutable.ArrayBuffer /** * A logical command that is executed for its side-effects. `RunnableCommand`s are @@ -54,9 +53,9 @@ case class ExecutedCommand(cmd: RunnableCommand) extends SparkPlan { */ protected[sql] lazy val sideEffectResult: Seq[Row] = cmd.run(sqlContext) - override def output = cmd.output + override def output: Seq[Attribute] = cmd.output - override def children = Nil + override def children: Seq[SparkPlan] = Nil override def executeCollect(): Array[Row] = sideEffectResult.toArray @@ -71,9 +70,10 @@ case class ExecutedCommand(cmd: RunnableCommand) extends SparkPlan { @DeveloperApi case class SetCommand( kv: Option[(String, Option[String])], - override val output: Seq[Attribute]) extends RunnableCommand with Logging { + override val output: Seq[Attribute]) + extends RunnableCommand with Logging { - override def run(sqlContext: SQLContext) = kv match { + override def run(sqlContext: SQLContext): Seq[Row] = kv match { // Configures the deprecated "mapred.reduce.tasks" property. case Some((SQLConf.Deprecated.MAPRED_REDUCE_TASKS, Some(value))) => logWarning( @@ -119,10 +119,11 @@ case class ExplainCommand( logicalPlan: LogicalPlan, override val output: Seq[Attribute] = Seq(AttributeReference("plan", StringType, nullable = false)()), - extended: Boolean = false) extends RunnableCommand { + extended: Boolean = false) + extends RunnableCommand { // Run through the optimizer to generate the physical plan. - override def run(sqlContext: SQLContext) = try { + override def run(sqlContext: SQLContext): Seq[Row] = try { // TODO in Hive, the "extended" ExplainCommand prints the AST as well, and detailed properties. val queryExecution = sqlContext.executePlan(logicalPlan) val outputString = if (extended) queryExecution.toString else queryExecution.simpleString @@ -140,9 +141,10 @@ case class ExplainCommand( case class CacheTableCommand( tableName: String, plan: Option[LogicalPlan], - isLazy: Boolean) extends RunnableCommand { + isLazy: Boolean) + extends RunnableCommand { - override def run(sqlContext: SQLContext) = { + override def run(sqlContext: SQLContext): Seq[Row] = { plan.foreach { logicalPlan => sqlContext.registerDataFrameAsTable(DataFrame(sqlContext, logicalPlan), tableName) } @@ -166,7 +168,7 @@ case class CacheTableCommand( @DeveloperApi case class UncacheTableCommand(tableName: String) extends RunnableCommand { - override def run(sqlContext: SQLContext) = { + override def run(sqlContext: SQLContext): Seq[Row] = { sqlContext.table(tableName).unpersist(blocking = false) Seq.empty[Row] } @@ -181,7 +183,7 @@ case class UncacheTableCommand(tableName: String) extends RunnableCommand { @DeveloperApi case object ClearCacheCommand extends RunnableCommand { - override def run(sqlContext: SQLContext) = { + override def run(sqlContext: SQLContext): Seq[Row] = { sqlContext.clearCache() Seq.empty[Row] } @@ -196,9 +198,10 @@ case object ClearCacheCommand extends RunnableCommand { case class DescribeCommand( child: SparkPlan, override val output: Seq[Attribute], - isExtended: Boolean) extends RunnableCommand { + isExtended: Boolean) + extends RunnableCommand { - override def run(sqlContext: SQLContext) = { + override def run(sqlContext: SQLContext): Seq[Row] = { child.schema.fields.map { field => val cmtKey = "comment" val comment = if (field.metadata.contains(cmtKey)) field.metadata.getString(cmtKey) else "" @@ -220,7 +223,7 @@ case class DescribeCommand( case class ShowTablesCommand(databaseName: Option[String]) extends RunnableCommand { // The result of SHOW TABLES has two columns, tableName and isTemporary. - override val output = { + override val output: Seq[Attribute] = { val schema = StructType( StructField("tableName", StringType, false) :: StructField("isTemporary", BooleanType, false) :: Nil) @@ -228,7 +231,7 @@ case class ShowTablesCommand(databaseName: Option[String]) extends RunnableComma schema.toAttributes } - override def run(sqlContext: SQLContext) = { + override def run(sqlContext: SQLContext): Seq[Row] = { // Since we need to return a Seq of rows, we will call getTables directly // instead of calling tables in sqlContext. val rows = sqlContext.catalog.getTables(databaseName).map { diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/debug/package.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/debug/package.scala index ffe388cfa9532..e916e68e58b5d 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/debug/package.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/debug/package.scala @@ -17,11 +17,13 @@ package org.apache.spark.sql.execution +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.catalyst.expressions.Attribute + import scala.collection.mutable.HashSet -import org.apache.spark.{AccumulatorParam, Accumulator, SparkContext} +import org.apache.spark.{AccumulatorParam, Accumulator} import org.apache.spark.annotation.DeveloperApi -import org.apache.spark.SparkContext._ import org.apache.spark.sql.{SQLConf, SQLContext, DataFrame, Row} import org.apache.spark.sql.catalyst.trees.TreeNodeRef import org.apache.spark.sql.types._ @@ -43,7 +45,7 @@ package object debug { * Augments [[SQLContext]] with debug methods. */ implicit class DebugSQLContext(sqlContext: SQLContext) { - def debug() = { + def debug(): Unit = { sqlContext.setConf(SQLConf.DATAFRAME_EAGER_ANALYSIS, "false") } } @@ -88,7 +90,7 @@ package object debug { } private[sql] case class DebugNode(child: SparkPlan) extends UnaryNode { - def output = child.output + def output: Seq[Attribute] = child.output implicit object SetAccumulatorParam extends AccumulatorParam[HashSet[String]] { def zero(initialValue: HashSet[String]): HashSet[String] = { @@ -109,10 +111,10 @@ package object debug { */ case class ColumnMetrics( elementTypes: Accumulator[HashSet[String]] = sparkContext.accumulator(HashSet.empty)) - val tupleCount = sparkContext.accumulator[Int](0) + val tupleCount: Accumulator[Int] = sparkContext.accumulator[Int](0) - val numColumns = child.output.size - val columnStats = Array.fill(child.output.size)(new ColumnMetrics()) + val numColumns: Int = child.output.size + val columnStats: Array[ColumnMetrics] = Array.fill(child.output.size)(new ColumnMetrics()) def dumpStats(): Unit = { println(s"== ${child.simpleString} ==") @@ -123,11 +125,11 @@ package object debug { } } - def execute() = { + def execute(): RDD[Row] = { child.execute().mapPartitions { iter => new Iterator[Row] { - def hasNext = iter.hasNext - def next() = { + def hasNext: Boolean = iter.hasNext + def next(): Row = { val currentRow = iter.next() tupleCount += 1 var i = 0 @@ -180,18 +182,18 @@ package object debug { private[sql] case class TypeCheck(child: SparkPlan) extends SparkPlan { import TypeCheck._ - override def nodeName = "" + override def nodeName: String = "" /* Only required when defining this class in a REPL. override def makeCopy(args: Array[Object]): this.type = TypeCheck(args(0).asInstanceOf[SparkPlan]).asInstanceOf[this.type] */ - def output = child.output + def output: Seq[Attribute] = child.output - def children = child :: Nil + def children: List[SparkPlan] = child :: Nil - def execute() = { + def execute(): RDD[Row] = { child.execute().map { row => try typeCheck(row, child.schema) catch { case e: Exception => diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastHashJoin.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastHashJoin.scala index 2dd22c020ef12..926f5e6c137ee 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastHashJoin.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastHashJoin.scala @@ -17,13 +17,15 @@ package org.apache.spark.sql.execution.joins +import org.apache.spark.rdd.RDD + import scala.concurrent._ import scala.concurrent.duration._ import scala.concurrent.ExecutionContext.Implicits.global import org.apache.spark.annotation.DeveloperApi import org.apache.spark.sql.catalyst.expressions.{Row, Expression} -import org.apache.spark.sql.catalyst.plans.physical.{Partitioning, UnspecifiedDistribution} +import org.apache.spark.sql.catalyst.plans.physical.{Distribution, Partitioning, UnspecifiedDistribution} import org.apache.spark.sql.execution.{BinaryNode, SparkPlan} /** @@ -42,7 +44,7 @@ case class BroadcastHashJoin( right: SparkPlan) extends BinaryNode with HashJoin { - val timeout = { + val timeout: Duration = { val timeoutValue = sqlContext.conf.broadcastTimeout if (timeoutValue < 0) { Duration.Inf @@ -53,7 +55,7 @@ case class BroadcastHashJoin( override def outputPartitioning: Partitioning = streamedPlan.outputPartitioning - override def requiredChildDistribution = + override def requiredChildDistribution: Seq[Distribution] = UnspecifiedDistribution :: UnspecifiedDistribution :: Nil @transient @@ -64,7 +66,7 @@ case class BroadcastHashJoin( sparkContext.broadcast(hashed) } - override def execute() = { + override def execute(): RDD[Row] = { val broadcastRelation = Await.result(broadcastFuture, timeout) streamedPlan.execute().mapPartitions { streamedIter => diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastLeftSemiJoinHash.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastLeftSemiJoinHash.scala index 2ab064fd0151e..3ef1e0d7fbdd4 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastLeftSemiJoinHash.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastLeftSemiJoinHash.scala @@ -18,8 +18,8 @@ package org.apache.spark.sql.execution.joins import org.apache.spark.annotation.DeveloperApi -import org.apache.spark.sql.catalyst.expressions.{Expression, Row} -import org.apache.spark.sql.catalyst.plans.physical.ClusteredDistribution +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.catalyst.expressions.{Attribute, Expression, Row} import org.apache.spark.sql.execution.{BinaryNode, SparkPlan} /** @@ -34,11 +34,11 @@ case class BroadcastLeftSemiJoinHash( left: SparkPlan, right: SparkPlan) extends BinaryNode with HashJoin { - override val buildSide = BuildRight + override val buildSide: BuildSide = BuildRight - override def output = left.output + override def output: Seq[Attribute] = left.output - override def execute() = { + override def execute(): RDD[Row] = { val buildIter= buildPlan.execute().map(_.copy()).collect().toIterator val hashSet = new java.util.HashSet[Row]() var currentRow: Row = null diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastNestedLoopJoin.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastNestedLoopJoin.scala index 36aad13778bd2..83b1a83765153 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastNestedLoopJoin.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/BroadcastNestedLoopJoin.scala @@ -18,6 +18,7 @@ package org.apache.spark.sql.execution.joins import org.apache.spark.annotation.DeveloperApi +import org.apache.spark.rdd.RDD import org.apache.spark.sql.catalyst.expressions._ import org.apache.spark.sql.catalyst.plans.physical.Partitioning import org.apache.spark.sql.catalyst.plans.{FullOuter, JoinType, LeftOuter, RightOuter} @@ -44,7 +45,7 @@ case class BroadcastNestedLoopJoin( override def outputPartitioning: Partitioning = streamed.outputPartitioning - override def output = { + override def output: Seq[Attribute] = { joinType match { case LeftOuter => left.output ++ right.output.map(_.withNullability(true)) @@ -63,7 +64,7 @@ case class BroadcastNestedLoopJoin( .map(c => BindReferences.bindReference(c, left.output ++ right.output)) .getOrElse(Literal(true))) - override def execute() = { + override def execute(): RDD[Row] = { val broadcastedRelation = sparkContext.broadcast(broadcast.execute().map(_.copy()).collect().toIndexedSeq) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/CartesianProduct.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/CartesianProduct.scala index 76c14c02aab34..1cbc98354d673 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/CartesianProduct.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/CartesianProduct.scala @@ -18,7 +18,9 @@ package org.apache.spark.sql.execution.joins import org.apache.spark.annotation.DeveloperApi -import org.apache.spark.sql.catalyst.expressions.JoinedRow +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.Row +import org.apache.spark.sql.catalyst.expressions.{Attribute, JoinedRow} import org.apache.spark.sql.execution.{BinaryNode, SparkPlan} /** @@ -26,9 +28,9 @@ import org.apache.spark.sql.execution.{BinaryNode, SparkPlan} */ @DeveloperApi case class CartesianProduct(left: SparkPlan, right: SparkPlan) extends BinaryNode { - override def output = left.output ++ right.output + override def output: Seq[Attribute] = left.output ++ right.output - override def execute() = { + override def execute(): RDD[Row] = { val leftResults = left.execute().map(_.copy()) val rightResults = right.execute().map(_.copy()) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashJoin.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashJoin.scala index 4012d757d5f9a..851de1685509a 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashJoin.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashJoin.scala @@ -41,7 +41,7 @@ trait HashJoin { case BuildRight => (rightKeys, leftKeys) } - override def output = left.output ++ right.output + override def output: Seq[Attribute] = left.output ++ right.output @transient protected lazy val buildSideKeyGenerator: Projection = newProjection(buildKeys, buildPlan.output) @@ -65,7 +65,7 @@ trait HashJoin { (currentMatchPosition != -1 && currentMatchPosition < currentHashMatches.size) || (streamIter.hasNext && fetchNext()) - override final def next() = { + override final def next(): Row = { val ret = buildSide match { case BuildRight => joinRow(currentStreamedRow, currentHashMatches(currentMatchPosition)) case BuildLeft => joinRow(currentHashMatches(currentMatchPosition), currentStreamedRow) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashOuterJoin.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashOuterJoin.scala index 59ef904272545..a396c0f5d56ee 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashOuterJoin.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashOuterJoin.scala @@ -19,6 +19,8 @@ package org.apache.spark.sql.execution.joins import java.util.{HashMap => JavaHashMap} +import org.apache.spark.rdd.RDD + import scala.collection.JavaConversions._ import org.apache.spark.annotation.DeveloperApi @@ -49,10 +51,10 @@ case class HashOuterJoin( case x => throw new Exception(s"HashOuterJoin should not take $x as the JoinType") } - override def requiredChildDistribution = + override def requiredChildDistribution: Seq[ClusteredDistribution] = ClusteredDistribution(leftKeys) :: ClusteredDistribution(rightKeys) :: Nil - override def output = { + override def output: Seq[Attribute] = { joinType match { case LeftOuter => left.output ++ right.output.map(_.withNullability(true)) @@ -78,12 +80,12 @@ case class HashOuterJoin( private[this] def leftOuterIterator( key: Row, joinedRow: JoinedRow, rightIter: Iterable[Row]): Iterator[Row] = { - val ret: Iterable[Row] = ( + val ret: Iterable[Row] = { if (!key.anyNull) { val temp = rightIter.collect { - case r if (boundCondition(joinedRow.withRight(r))) => joinedRow.copy + case r if boundCondition(joinedRow.withRight(r)) => joinedRow.copy() } - if (temp.size == 0) { + if (temp.size == 0) { joinedRow.withRight(rightNullRow).copy :: Nil } else { temp @@ -91,19 +93,19 @@ case class HashOuterJoin( } else { joinedRow.withRight(rightNullRow).copy :: Nil } - ) + } ret.iterator } private[this] def rightOuterIterator( key: Row, leftIter: Iterable[Row], joinedRow: JoinedRow): Iterator[Row] = { - val ret: Iterable[Row] = ( + val ret: Iterable[Row] = { if (!key.anyNull) { val temp = leftIter.collect { - case l if (boundCondition(joinedRow.withLeft(l))) => joinedRow.copy + case l if boundCondition(joinedRow.withLeft(l)) => joinedRow.copy } - if (temp.size == 0) { + if (temp.size == 0) { joinedRow.withLeft(leftNullRow).copy :: Nil } else { temp @@ -111,7 +113,7 @@ case class HashOuterJoin( } else { joinedRow.withLeft(leftNullRow).copy :: Nil } - ) + } ret.iterator } @@ -130,12 +132,12 @@ case class HashOuterJoin( // 1. For those matched (satisfy the join condition) records with both sides filled, // append them directly - case (r, idx) if (boundCondition(joinedRow.withRight(r)))=> { + case (r, idx) if boundCondition(joinedRow.withRight(r)) => matched = true // if the row satisfy the join condition, add its index into the matched set rightMatchedSet.add(idx) - joinedRow.copy - } + joinedRow.copy() + } ++ DUMMY_LIST.filter(_ => !matched).map( _ => { // 2. For those unmatched records in left, append additional records with empty right. @@ -143,22 +145,21 @@ case class HashOuterJoin( // as we don't know whether we need to append it until finish iterating all // of the records in right side. // If we didn't get any proper row, then append a single row with empty right. - joinedRow.withRight(rightNullRow).copy + joinedRow.withRight(rightNullRow).copy() }) } ++ rightIter.zipWithIndex.collect { // 3. For those unmatched records in right, append additional records with empty left. // Re-visiting the records in right, and append additional row with empty left, if its not // in the matched set. - case (r, idx) if (!rightMatchedSet.contains(idx)) => { - joinedRow(leftNullRow, r).copy - } + case (r, idx) if !rightMatchedSet.contains(idx) => + joinedRow(leftNullRow, r).copy() } } else { leftIter.iterator.map[Row] { l => - joinedRow(l, rightNullRow).copy + joinedRow(l, rightNullRow).copy() } ++ rightIter.iterator.map[Row] { r => - joinedRow(leftNullRow, r).copy + joinedRow(leftNullRow, r).copy() } } } @@ -182,13 +183,13 @@ case class HashOuterJoin( hashTable } - override def execute() = { + override def execute(): RDD[Row] = { val joinedRow = new JoinedRow() left.execute().zipPartitions(right.execute()) { (leftIter, rightIter) => // TODO this probably can be replaced by external sort (sort merged join?) joinType match { - case LeftOuter => { + case LeftOuter => val rightHashTable = buildHashTable(rightIter, newProjection(rightKeys, right.output)) val keyGenerator = newProjection(leftKeys, left.output) leftIter.flatMap( currentRow => { @@ -196,8 +197,8 @@ case class HashOuterJoin( joinedRow.withLeft(currentRow) leftOuterIterator(rowKey, joinedRow, rightHashTable.getOrElse(rowKey, EMPTY_LIST)) }) - } - case RightOuter => { + + case RightOuter => val leftHashTable = buildHashTable(leftIter, newProjection(leftKeys, left.output)) val keyGenerator = newProjection(rightKeys, right.output) rightIter.flatMap ( currentRow => { @@ -205,8 +206,8 @@ case class HashOuterJoin( joinedRow.withRight(currentRow) rightOuterIterator(rowKey, leftHashTable.getOrElse(rowKey, EMPTY_LIST), joinedRow) }) - } - case FullOuter => { + + case FullOuter => val leftHashTable = buildHashTable(leftIter, newProjection(leftKeys, left.output)) val rightHashTable = buildHashTable(rightIter, newProjection(rightKeys, right.output)) (leftHashTable.keySet ++ rightHashTable.keySet).iterator.flatMap { key => @@ -214,7 +215,7 @@ case class HashOuterJoin( leftHashTable.getOrElse(key, EMPTY_LIST), rightHashTable.getOrElse(key, EMPTY_LIST), joinedRow) } - } + case x => throw new Exception(s"HashOuterJoin should not take $x as the JoinType") } } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashedRelation.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashedRelation.scala index 38b8993b03f82..2fa1cf5add3b5 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashedRelation.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashedRelation.scala @@ -38,7 +38,7 @@ private[joins] sealed trait HashedRelation { private[joins] final class GeneralHashedRelation(hashTable: JavaHashMap[Row, CompactBuffer[Row]]) extends HashedRelation with Serializable { - override def get(key: Row) = hashTable.get(key) + override def get(key: Row): CompactBuffer[Row] = hashTable.get(key) } @@ -49,7 +49,7 @@ private[joins] final class GeneralHashedRelation(hashTable: JavaHashMap[Row, Com private[joins] final class UniqueKeyHashedRelation(hashTable: JavaHashMap[Row, Row]) extends HashedRelation with Serializable { - override def get(key: Row) = { + override def get(key: Row): CompactBuffer[Row] = { val v = hashTable.get(key) if (v eq null) null else CompactBuffer(v) } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/LeftSemiJoinBNL.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/LeftSemiJoinBNL.scala index 60003d1900d85..1fa7e7bd0406c 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/LeftSemiJoinBNL.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/LeftSemiJoinBNL.scala @@ -18,6 +18,7 @@ package org.apache.spark.sql.execution.joins import org.apache.spark.annotation.DeveloperApi +import org.apache.spark.rdd.RDD import org.apache.spark.sql.catalyst.expressions._ import org.apache.spark.sql.catalyst.plans.physical.Partitioning import org.apache.spark.sql.execution.{BinaryNode, SparkPlan} @@ -35,12 +36,13 @@ case class LeftSemiJoinBNL( override def outputPartitioning: Partitioning = streamed.outputPartitioning - override def output = left.output + override def output: Seq[Attribute] = left.output /** The Streamed Relation */ - override def left = streamed + override def left: SparkPlan = streamed + /** The Broadcast relation */ - override def right = broadcast + override def right: SparkPlan = broadcast @transient private lazy val boundCondition = InterpretedPredicate( @@ -48,7 +50,7 @@ case class LeftSemiJoinBNL( .map(c => BindReferences.bindReference(c, left.output ++ right.output)) .getOrElse(Literal(true))) - override def execute() = { + override def execute(): RDD[Row] = { val broadcastedRelation = sparkContext.broadcast(broadcast.execute().map(_.copy()).collect().toIndexedSeq) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/LeftSemiJoinHash.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/LeftSemiJoinHash.scala index ea7babf3be948..a04f2a63b5a55 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/LeftSemiJoinHash.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/LeftSemiJoinHash.scala @@ -18,7 +18,8 @@ package org.apache.spark.sql.execution.joins import org.apache.spark.annotation.DeveloperApi -import org.apache.spark.sql.catalyst.expressions.{Expression, Row} +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.catalyst.expressions.{Attribute, Expression, Row} import org.apache.spark.sql.catalyst.plans.physical.ClusteredDistribution import org.apache.spark.sql.execution.{BinaryNode, SparkPlan} @@ -34,14 +35,14 @@ case class LeftSemiJoinHash( left: SparkPlan, right: SparkPlan) extends BinaryNode with HashJoin { - override val buildSide = BuildRight + override val buildSide: BuildSide = BuildRight - override def requiredChildDistribution = + override def requiredChildDistribution: Seq[ClusteredDistribution] = ClusteredDistribution(leftKeys) :: ClusteredDistribution(rightKeys) :: Nil - override def output = left.output + override def output: Seq[Attribute] = left.output - override def execute() = { + override def execute(): RDD[Row] = { buildPlan.execute().zipPartitions(streamedPlan.execute()) { (buildIter, streamIter) => val hashSet = new java.util.HashSet[Row]() var currentRow: Row = null diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/ShuffledHashJoin.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/ShuffledHashJoin.scala index 418c1c23e5546..a6cd8337c1c3e 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/ShuffledHashJoin.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/ShuffledHashJoin.scala @@ -18,6 +18,8 @@ package org.apache.spark.sql.execution.joins import org.apache.spark.annotation.DeveloperApi +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.Row import org.apache.spark.sql.catalyst.expressions.Expression import org.apache.spark.sql.catalyst.plans.physical.{ClusteredDistribution, Partitioning} import org.apache.spark.sql.execution.{BinaryNode, SparkPlan} @@ -38,10 +40,10 @@ case class ShuffledHashJoin( override def outputPartitioning: Partitioning = left.outputPartitioning - override def requiredChildDistribution = + override def requiredChildDistribution: Seq[ClusteredDistribution] = ClusteredDistribution(leftKeys) :: ClusteredDistribution(rightKeys) :: Nil - override def execute() = { + override def execute(): RDD[Row] = { buildPlan.execute().zipPartitions(streamedPlan.execute()) { (buildIter, streamIter) => val hashed = HashedRelation(buildIter, buildSideKeyGenerator) hashJoin(streamIter, hashed) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/pythonUdfs.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/pythonUdfs.scala index 33632b8e82ff9..5b308d88d4cdf 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/pythonUdfs.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/pythonUdfs.scala @@ -19,6 +19,8 @@ package org.apache.spark.sql.execution import java.util.{List => JList, Map => JMap} +import org.apache.spark.rdd.RDD + import scala.collection.JavaConversions._ import scala.collection.JavaConverters._ @@ -48,11 +50,13 @@ private[spark] case class PythonUDF( dataType: DataType, children: Seq[Expression]) extends Expression with SparkLogging { - override def toString = s"PythonUDF#$name(${children.mkString(",")})" + override def toString: String = s"PythonUDF#$name(${children.mkString(",")})" def nullable: Boolean = true - override def eval(input: Row) = sys.error("PythonUDFs can not be directly evaluated.") + override def eval(input: Row): PythonUDF.this.EvaluatedType = { + sys.error("PythonUDFs can not be directly evaluated.") + } } /** @@ -63,7 +67,7 @@ private[spark] case class PythonUDF( * multiple child operators. */ private[spark] object ExtractPythonUdfs extends Rule[LogicalPlan] { - def apply(plan: LogicalPlan) = plan transform { + def apply(plan: LogicalPlan): LogicalPlan = plan transform { // Skip EvaluatePython nodes. case p: EvaluatePython => p @@ -107,7 +111,7 @@ private[spark] object ExtractPythonUdfs extends Rule[LogicalPlan] { } object EvaluatePython { - def apply(udf: PythonUDF, child: LogicalPlan) = + def apply(udf: PythonUDF, child: LogicalPlan): EvaluatePython = new EvaluatePython(udf, child, AttributeReference("pythonUDF", udf.dataType)()) /** @@ -205,10 +209,10 @@ case class EvaluatePython( resultAttribute: AttributeReference) extends logical.UnaryNode { - def output = child.output :+ resultAttribute + def output: Seq[Attribute] = child.output :+ resultAttribute // References should not include the produced attribute. - override def references = udf.references + override def references: AttributeSet = udf.references } /** @@ -219,9 +223,10 @@ case class EvaluatePython( @DeveloperApi case class BatchPythonEvaluation(udf: PythonUDF, output: Seq[Attribute], child: SparkPlan) extends SparkPlan { - def children = child :: Nil - def execute() = { + def children: Seq[SparkPlan] = child :: Nil + + def execute(): RDD[Row] = { // TODO: Clean up after ourselves? val childResults = child.execute().map(_.copy()).cache() diff --git a/sql/core/src/main/scala/org/apache/spark/sql/jdbc/JDBCRDD.scala b/sql/core/src/main/scala/org/apache/spark/sql/jdbc/JDBCRDD.scala index 87304ce2496b4..3266b972128ea 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/jdbc/JDBCRDD.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/jdbc/JDBCRDD.scala @@ -306,7 +306,8 @@ private[sql] class JDBCRDD( /** * Runs the SQL query against the JDBC driver. */ - override def compute(thePart: Partition, context: TaskContext) = new Iterator[Row] { + override def compute(thePart: Partition, context: TaskContext): Iterator[Row] = new Iterator[Row] + { var closed = false var finished = false var gotNext = false diff --git a/sql/core/src/main/scala/org/apache/spark/sql/jdbc/JDBCRelation.scala b/sql/core/src/main/scala/org/apache/spark/sql/jdbc/JDBCRelation.scala index 1778d39c42e2b..df687e6da9bea 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/jdbc/JDBCRelation.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/jdbc/JDBCRelation.scala @@ -17,6 +17,10 @@ package org.apache.spark.sql.jdbc +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.catalyst.expressions.Row +import org.apache.spark.sql.types.StructType + import scala.collection.mutable.ArrayBuffer import java.sql.DriverManager @@ -122,9 +126,9 @@ private[sql] case class JDBCRelation( extends BaseRelation with PrunedFilteredScan { - override val schema = JDBCRDD.resolveTable(url, table) + override val schema: StructType = JDBCRDD.resolveTable(url, table) - override def buildScan(requiredColumns: Array[String], filters: Array[Filter]) = { + override def buildScan(requiredColumns: Array[String], filters: Array[Filter]): RDD[Row] = { val driver: String = DriverManager.getDriver(url).getClass.getCanonicalName JDBCRDD.scanTable( sqlContext.sparkContext, diff --git a/sql/core/src/main/scala/org/apache/spark/sql/json/JSONRelation.scala b/sql/core/src/main/scala/org/apache/spark/sql/json/JSONRelation.scala index b645199ded18c..b1e363d02edfe 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/json/JSONRelation.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/json/JSONRelation.scala @@ -20,6 +20,8 @@ package org.apache.spark.sql.json import java.io.IOException import org.apache.hadoop.fs.Path +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.catalyst.expressions.Row import org.apache.spark.sql.{SaveMode, DataFrame, SQLContext} import org.apache.spark.sql.sources._ @@ -104,10 +106,10 @@ private[sql] case class JSONRelation( samplingRatio, sqlContext.conf.columnNameOfCorruptRecord))) - override def buildScan() = + override def buildScan(): RDD[Row] = JsonRDD.jsonStringToRow(baseRDD, schema, sqlContext.conf.columnNameOfCorruptRecord) - override def insert(data: DataFrame, overwrite: Boolean) = { + override def insert(data: DataFrame, overwrite: Boolean): Unit = { val filesystemPath = new Path(path) val fs = filesystemPath.getFileSystem(sqlContext.sparkContext.hadoopConfiguration) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetConverter.scala b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetConverter.scala index 7d62f3728f036..f898e4b37a56b 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetConverter.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetConverter.scala @@ -488,7 +488,7 @@ private[parquet] object CatalystTimestampConverter { // Also we use NanoTime and Int96Values from parquet-examples. // We utilize jodd to convert between NanoTime and Timestamp val parquetTsCalendar = new ThreadLocal[Calendar] - def getCalendar = { + def getCalendar: Calendar = { // this is a cache for the calendar instance. if (parquetTsCalendar.get == null) { parquetTsCalendar.set(Calendar.getInstance(TimeZone.getTimeZone("GMT"))) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetRelation.scala b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetRelation.scala index fd161bae128ad..fcb9513ab66f6 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetRelation.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetRelation.scala @@ -71,16 +71,22 @@ private[sql] case class ParquetRelation( sqlContext.conf.isParquetINT96AsTimestamp) lazy val attributeMap = AttributeMap(output.map(o => o -> o)) - override def newInstance() = ParquetRelation(path, conf, sqlContext).asInstanceOf[this.type] + override def newInstance(): this.type = { + ParquetRelation(path, conf, sqlContext).asInstanceOf[this.type] + } // Equals must also take into account the output attributes so that we can distinguish between // different instances of the same relation, - override def equals(other: Any) = other match { + override def equals(other: Any): Boolean = other match { case p: ParquetRelation => p.path == path && p.output == output case _ => false } + override def hashCode: Int = { + com.google.common.base.Objects.hashCode(path, output) + } + // TODO: Use data from the footers. override lazy val statistics = Statistics(sizeInBytes = sqlContext.conf.defaultSizeInBytes) } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTableOperations.scala b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTableOperations.scala index 62813a981e685..5130d8ad5e003 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTableOperations.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTableOperations.scala @@ -152,8 +152,8 @@ private[sql] case class ParquetTableScan( if (primitiveRow) { new Iterator[Row] { - def hasNext = iter.hasNext - def next() = { + def hasNext: Boolean = iter.hasNext + def next(): Row = { // We are using CatalystPrimitiveRowConverter and it returns a SpecificMutableRow. val row = iter.next()._2.asInstanceOf[SpecificMutableRow] @@ -171,8 +171,8 @@ private[sql] case class ParquetTableScan( // Create a mutable row since we need to fill in values from partition columns. val mutableRow = new GenericMutableRow(outputSize) new Iterator[Row] { - def hasNext = iter.hasNext - def next() = { + def hasNext: Boolean = iter.hasNext + def next(): Row = { // We are using CatalystGroupConverter and it returns a GenericRow. // Since GenericRow is not mutable, we just cast it to a Row. val row = iter.next()._2.asInstanceOf[Row] @@ -255,7 +255,7 @@ private[sql] case class InsertIntoParquetTable( /** * Inserts all rows into the Parquet file. */ - override def execute() = { + override def execute(): RDD[Row] = { // TODO: currently we do not check whether the "schema"s are compatible // That means if one first creates a table and then INSERTs data with // and incompatible schema the execution will fail. It would be nice @@ -302,7 +302,7 @@ private[sql] case class InsertIntoParquetTable( childRdd } - override def output = child.output + override def output: Seq[Attribute] = child.output /** * Stores the given Row RDD as a Hadoop file. diff --git a/sql/core/src/main/scala/org/apache/spark/sql/parquet/newParquet.scala b/sql/core/src/main/scala/org/apache/spark/sql/parquet/newParquet.scala index c38b6e8c61d8a..10b8876c1d31c 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/parquet/newParquet.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/parquet/newParquet.scala @@ -181,7 +181,7 @@ private[sql] case class ParquetRelation2( private val defaultPartitionName = parameters.getOrElse( ParquetRelation2.DEFAULT_PARTITION_NAME, "__HIVE_DEFAULT_PARTITION__") - override def equals(other: Any) = other match { + override def equals(other: Any): Boolean = other match { case relation: ParquetRelation2 => // If schema merging is required, we don't compare the actual schemas since they may evolve. val schemaEquality = if (shouldMergeSchemas) { @@ -198,6 +198,23 @@ private[sql] case class ParquetRelation2( case _ => false } + override def hashCode(): Int = { + if (shouldMergeSchemas) { + com.google.common.base.Objects.hashCode( + shouldMergeSchemas: java.lang.Boolean, + paths.toSet, + maybeMetastoreSchema, + maybePartitionSpec) + } else { + com.google.common.base.Objects.hashCode( + shouldMergeSchemas: java.lang.Boolean, + schema, + paths.toSet, + maybeMetastoreSchema, + maybePartitionSpec) + } + } + private[sql] def sparkContext = sqlContext.sparkContext private class MetadataCache { @@ -370,19 +387,19 @@ private[sql] case class ParquetRelation2( @transient private val metadataCache = new MetadataCache metadataCache.refresh() - def partitionSpec = metadataCache.partitionSpec + def partitionSpec: PartitionSpec = metadataCache.partitionSpec - def partitionColumns = metadataCache.partitionSpec.partitionColumns + def partitionColumns: StructType = metadataCache.partitionSpec.partitionColumns - def partitions = metadataCache.partitionSpec.partitions + def partitions: Seq[Partition] = metadataCache.partitionSpec.partitions - def isPartitioned = partitionColumns.nonEmpty + def isPartitioned: Boolean = partitionColumns.nonEmpty private def partitionKeysIncludedInDataSchema = metadataCache.partitionKeysIncludedInParquetSchema private def parquetSchema = metadataCache.parquetSchema - override def schema = metadataCache.schema + override def schema: StructType = metadataCache.schema private def isSummaryFile(file: Path): Boolean = { file.getName == ParquetFileWriter.PARQUET_COMMON_METADATA_FILE || @@ -425,8 +442,10 @@ private[sql] case class ParquetRelation2( .foreach(ParquetInputFormat.setFilterPredicate(jobConf, _)) if (isPartitioned) { - def percentRead = selectedPartitions.size.toDouble / partitions.size.toDouble * 100 - logInfo(s"Reading $percentRead% of partitions") + logInfo { + val percentRead = selectedPartitions.size.toDouble / partitions.size.toDouble * 100 + s"Reading $percentRead% of partitions" + } } val requiredColumns = output.map(_.name) @@ -703,7 +722,7 @@ private[sql] object ParquetRelation2 { private[parquet] def mergeMetastoreParquetSchema( metastoreSchema: StructType, parquetSchema: StructType): StructType = { - def schemaConflictMessage = + def schemaConflictMessage: String = s"""Converting Hive Metastore Parquet, but detected conflicting schemas. Metastore schema: |${metastoreSchema.prettyJson} | diff --git a/sql/core/src/main/scala/org/apache/spark/sql/parquet/timestamp/NanoTime.scala b/sql/core/src/main/scala/org/apache/spark/sql/parquet/timestamp/NanoTime.scala index e24475292ceaf..70bcca7526aae 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/parquet/timestamp/NanoTime.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/parquet/timestamp/NanoTime.scala @@ -26,7 +26,7 @@ private[parquet] class NanoTime extends Serializable { private var julianDay = 0 private var timeOfDayNanos = 0L - def set(julianDay: Int, timeOfDayNanos: Long) = { + def set(julianDay: Int, timeOfDayNanos: Long): this.type = { this.julianDay = julianDay this.timeOfDayNanos = timeOfDayNanos this @@ -45,11 +45,11 @@ private[parquet] class NanoTime extends Serializable { Binary.fromByteBuffer(buf) } - def writeValue(recordConsumer: RecordConsumer) { + def writeValue(recordConsumer: RecordConsumer): Unit = { recordConsumer.addBinary(toBinary) } - override def toString = + override def toString: String = "NanoTime{julianDay=" + julianDay + ", timeOfDayNanos=" + timeOfDayNanos + "}" } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/sources/LogicalRelation.scala b/sql/core/src/main/scala/org/apache/spark/sql/sources/LogicalRelation.scala index 12b59ba20bb10..f374abffdd505 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/sources/LogicalRelation.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/sources/LogicalRelation.scala @@ -30,24 +30,28 @@ private[sql] case class LogicalRelation(relation: BaseRelation) override val output: Seq[AttributeReference] = relation.schema.toAttributes // Logical Relations are distinct if they have different output for the sake of transformations. - override def equals(other: Any) = other match { + override def equals(other: Any): Boolean = other match { case l @ LogicalRelation(otherRelation) => relation == otherRelation && output == l.output case _ => false } - override def sameResult(otherPlan: LogicalPlan) = otherPlan match { + override def hashCode: Int = { + com.google.common.base.Objects.hashCode(relation, output) + } + + override def sameResult(otherPlan: LogicalPlan): Boolean = otherPlan match { case LogicalRelation(otherRelation) => relation == otherRelation case _ => false } - @transient override lazy val statistics = Statistics( + @transient override lazy val statistics: Statistics = Statistics( sizeInBytes = BigInt(relation.sizeInBytes) ) /** Used to lookup original attribute capitalization */ - val attributeMap = AttributeMap(output.map(o => (o, o))) + val attributeMap: AttributeMap[AttributeReference] = AttributeMap(output.map(o => (o, o))) - def newInstance() = LogicalRelation(relation).asInstanceOf[this.type] + def newInstance(): this.type = LogicalRelation(relation).asInstanceOf[this.type] - override def simpleString = s"Relation[${output.mkString(",")}] $relation" + override def simpleString: String = s"Relation[${output.mkString(",")}] $relation" } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/sources/commands.scala b/sql/core/src/main/scala/org/apache/spark/sql/sources/commands.scala index 0e540dad81283..9bbe06e59ba30 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/sources/commands.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/sources/commands.scala @@ -27,7 +27,7 @@ private[sql] case class InsertIntoDataSource( overwrite: Boolean) extends RunnableCommand { - override def run(sqlContext: SQLContext) = { + override def run(sqlContext: SQLContext): Seq[Row] = { val relation = logicalRelation.relation.asInstanceOf[InsertableRelation] val data = DataFrame(sqlContext, query) // Apply the schema of the existing table to the new data. diff --git a/sql/core/src/main/scala/org/apache/spark/sql/sources/ddl.scala b/sql/core/src/main/scala/org/apache/spark/sql/sources/ddl.scala index 76754a6ce4617..d57406645eefa 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/sources/ddl.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/sources/ddl.scala @@ -362,7 +362,7 @@ private[sql] case class CreateTableUsingAsSelect( mode: SaveMode, options: Map[String, String], child: LogicalPlan) extends UnaryNode { - override def output = Seq.empty[Attribute] + override def output: Seq[Attribute] = Seq.empty[Attribute] // TODO: Override resolved after we support databaseName. // override lazy val resolved = databaseName != None && childrenResolved } @@ -373,7 +373,7 @@ private[sql] case class CreateTempTableUsing( provider: String, options: Map[String, String]) extends RunnableCommand { - def run(sqlContext: SQLContext) = { + def run(sqlContext: SQLContext): Seq[Row] = { val resolved = ResolvedDataSource(sqlContext, userSpecifiedSchema, provider, options) sqlContext.registerDataFrameAsTable( DataFrame(sqlContext, LogicalRelation(resolved.relation)), tableName) @@ -388,7 +388,7 @@ private[sql] case class CreateTempTableUsingAsSelect( options: Map[String, String], query: LogicalPlan) extends RunnableCommand { - def run(sqlContext: SQLContext) = { + def run(sqlContext: SQLContext): Seq[Row] = { val df = DataFrame(sqlContext, query) val resolved = ResolvedDataSource(sqlContext, provider, mode, options, df) sqlContext.registerDataFrameAsTable( diff --git a/sql/core/src/main/scala/org/apache/spark/sql/sources/rules.scala b/sql/core/src/main/scala/org/apache/spark/sql/sources/rules.scala index cfa58f1442218..5a78001117d1b 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/sources/rules.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/sources/rules.scala @@ -53,7 +53,7 @@ private[sql] object PreInsertCastAndRename extends Rule[LogicalPlan] { def castAndRenameChildOutput( insertInto: InsertIntoTable, expectedOutput: Seq[Attribute], - child: LogicalPlan) = { + child: LogicalPlan): InsertIntoTable = { val newChildOutput = expectedOutput.zip(child.output).map { case (expected, actual) => val needCast = !expected.dataType.sameType(actual.dataType) @@ -79,7 +79,7 @@ private[sql] object PreInsertCastAndRename extends Rule[LogicalPlan] { * A rule to do various checks before inserting into or writing to a data source table. */ private[sql] case class PreWriteCheck(catalog: Catalog) extends (LogicalPlan => Unit) { - def failAnalysis(msg: String) = { throw new AnalysisException(msg) } + def failAnalysis(msg: String): Unit = { throw new AnalysisException(msg) } def apply(plan: LogicalPlan): Unit = { plan.foreach { From 25e271d9fbb3394931d23822a1b2020e9d9b46b3 Mon Sep 17 00:00:00 2001 From: MechCoder Date: Fri, 20 Mar 2015 17:14:09 -0700 Subject: [PATCH 122/122] [SPARK-6025] [MLlib] Add helper method evaluateEachIteration to extract learning curve Added evaluateEachIteration to allow the user to manually extract the error for each iteration of GradientBoosting. The internal optimisation can be dealt with later. Author: MechCoder Closes #4906 from MechCoder/spark-6025 and squashes the following commits: 67146ab [MechCoder] Minor 352001f [MechCoder] Minor 6e8aa10 [MechCoder] Made the following changes Used mapPartition instead of map Refactored computeError and unpersisted broadcast variables bc99ac6 [MechCoder] Refactor the method and stuff dbda033 [MechCoder] [SPARK-6025] Add helper method evaluateEachIteration to extract learning curve --- docs/mllib-ensembles.md | 4 +- .../spark/mllib/tree/loss/AbsoluteError.scala | 17 ++---- .../spark/mllib/tree/loss/LogLoss.scala | 20 ++----- .../apache/spark/mllib/tree/loss/Loss.scala | 14 ++++- .../spark/mllib/tree/loss/SquaredError.scala | 17 ++---- .../mllib/tree/model/treeEnsembleModels.scala | 54 +++++++++++++++++++ .../tree/GradientBoostedTreesSuite.scala | 16 +++++- 7 files changed, 96 insertions(+), 46 deletions(-) diff --git a/docs/mllib-ensembles.md b/docs/mllib-ensembles.md index cbfb682609af3..7521fb14a7bd6 100644 --- a/docs/mllib-ensembles.md +++ b/docs/mllib-ensembles.md @@ -464,8 +464,8 @@ first one being the training dataset and the second being the validation dataset The training is stopped when the improvement in the validation error is not more than a certain tolerance (supplied by the `validationTol` argument in `BoostingStrategy`). In practice, the validation error decreases initially and later increases. There might be cases in which the validation error does not change monotonically, -and the user is advised to set a large enough negative tolerance and examine the validation curve to to tune the number of -iterations. +and the user is advised to set a large enough negative tolerance and examine the validation curve using `evaluateEachIteration` +(which gives the error or loss per iteration) to tune the number of iterations. ### Examples diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/AbsoluteError.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/AbsoluteError.scala index d1bde15e6b150..793dd664c5d5a 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/AbsoluteError.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/AbsoluteError.scala @@ -47,18 +47,9 @@ object AbsoluteError extends Loss { if ((point.label - model.predict(point.features)) < 0) 1.0 else -1.0 } - /** - * Method to calculate loss of the base learner for the gradient boosting calculation. - * Note: This method is not used by the gradient boosting algorithm but is useful for debugging - * purposes. - * @param model Ensemble model - * @param data Training dataset: RDD of [[org.apache.spark.mllib.regression.LabeledPoint]]. - * @return Mean absolute error of model on data - */ - override def computeError(model: TreeEnsembleModel, data: RDD[LabeledPoint]): Double = { - data.map { y => - val err = model.predict(y.features) - y.label - math.abs(err) - }.mean() + override def computeError(prediction: Double, label: Double): Double = { + val err = label - prediction + math.abs(err) } + } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/LogLoss.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/LogLoss.scala index 55213e695638c..51b1aed167b66 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/LogLoss.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/LogLoss.scala @@ -50,20 +50,10 @@ object LogLoss extends Loss { - 4.0 * point.label / (1.0 + math.exp(2.0 * point.label * prediction)) } - /** - * Method to calculate loss of the base learner for the gradient boosting calculation. - * Note: This method is not used by the gradient boosting algorithm but is useful for debugging - * purposes. - * @param model Ensemble model - * @param data Training dataset: RDD of [[org.apache.spark.mllib.regression.LabeledPoint]]. - * @return Mean log loss of model on data - */ - override def computeError(model: TreeEnsembleModel, data: RDD[LabeledPoint]): Double = { - data.map { case point => - val prediction = model.predict(point.features) - val margin = 2.0 * point.label * prediction - // The following is equivalent to 2.0 * log(1 + exp(-margin)) but more numerically stable. - 2.0 * MLUtils.log1pExp(-margin) - }.mean() + override def computeError(prediction: Double, label: Double): Double = { + val margin = 2.0 * label * prediction + // The following is equivalent to 2.0 * log(1 + exp(-margin)) but more numerically stable. + 2.0 * MLUtils.log1pExp(-margin) } + } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/Loss.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/Loss.scala index e1169d9f66ea4..357869ff6b333 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/Loss.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/Loss.scala @@ -47,6 +47,18 @@ trait Loss extends Serializable { * @param data Training dataset: RDD of [[org.apache.spark.mllib.regression.LabeledPoint]]. * @return Measure of model error on data */ - def computeError(model: TreeEnsembleModel, data: RDD[LabeledPoint]): Double + def computeError(model: TreeEnsembleModel, data: RDD[LabeledPoint]): Double = { + data.map(point => computeError(model.predict(point.features), point.label)).mean() + } + + /** + * Method to calculate loss when the predictions are already known. + * Note: This method is used in the method evaluateEachIteration to avoid recomputing the + * predicted values from previously fit trees. + * @param prediction Predicted label. + * @param label True label. + * @return Measure of model error on datapoint. + */ + def computeError(prediction: Double, label: Double): Double } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/SquaredError.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/SquaredError.scala index 50ecaa2f86f35..b990707ca4525 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/SquaredError.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/loss/SquaredError.scala @@ -47,18 +47,9 @@ object SquaredError extends Loss { 2.0 * (model.predict(point.features) - point.label) } - /** - * Method to calculate loss of the base learner for the gradient boosting calculation. - * Note: This method is not used by the gradient boosting algorithm but is useful for debugging - * purposes. - * @param model Ensemble model - * @param data Training dataset: RDD of [[org.apache.spark.mllib.regression.LabeledPoint]]. - * @return Mean squared error of model on data - */ - override def computeError(model: TreeEnsembleModel, data: RDD[LabeledPoint]): Double = { - data.map { y => - val err = model.predict(y.features) - y.label - err * err - }.mean() + override def computeError(prediction: Double, label: Double): Double = { + val err = prediction - label + err * err } + } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala index f160852c69c77..1950254b2aa6d 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala @@ -28,9 +28,11 @@ import org.apache.spark.{Logging, SparkContext} import org.apache.spark.annotation.Experimental import org.apache.spark.api.java.JavaRDD import org.apache.spark.mllib.linalg.Vector +import org.apache.spark.mllib.regression.LabeledPoint import org.apache.spark.mllib.tree.configuration.Algo import org.apache.spark.mllib.tree.configuration.Algo._ import org.apache.spark.mllib.tree.configuration.EnsembleCombiningStrategy._ +import org.apache.spark.mllib.tree.loss.Loss import org.apache.spark.mllib.util.{Loader, Saveable} import org.apache.spark.rdd.RDD import org.apache.spark.sql.SQLContext @@ -108,6 +110,58 @@ class GradientBoostedTreesModel( } override protected def formatVersion: String = TreeEnsembleModel.SaveLoadV1_0.thisFormatVersion + + /** + * Method to compute error or loss for every iteration of gradient boosting. + * @param data RDD of [[org.apache.spark.mllib.regression.LabeledPoint]] + * @param loss evaluation metric. + * @return an array with index i having the losses or errors for the ensemble + * containing the first i+1 trees + */ + def evaluateEachIteration( + data: RDD[LabeledPoint], + loss: Loss): Array[Double] = { + + val sc = data.sparkContext + val remappedData = algo match { + case Classification => data.map(x => new LabeledPoint((x.label * 2) - 1, x.features)) + case _ => data + } + + val numIterations = trees.length + val evaluationArray = Array.fill(numIterations)(0.0) + + var predictionAndError: RDD[(Double, Double)] = remappedData.map { i => + val pred = treeWeights(0) * trees(0).predict(i.features) + val error = loss.computeError(pred, i.label) + (pred, error) + } + evaluationArray(0) = predictionAndError.values.mean() + + // Avoid the model being copied across numIterations. + val broadcastTrees = sc.broadcast(trees) + val broadcastWeights = sc.broadcast(treeWeights) + + (1 until numIterations).map { nTree => + predictionAndError = remappedData.zip(predictionAndError).mapPartitions { iter => + val currentTree = broadcastTrees.value(nTree) + val currentTreeWeight = broadcastWeights.value(nTree) + iter.map { + case (point, (pred, error)) => { + val newPred = pred + currentTree.predict(point.features) * currentTreeWeight + val newError = loss.computeError(newPred, point.label) + (newPred, newError) + } + } + } + evaluationArray(nTree) = predictionAndError.values.mean() + } + + broadcastTrees.unpersist() + broadcastWeights.unpersist() + evaluationArray + } + } object GradientBoostedTreesModel extends Loader[GradientBoostedTreesModel] { diff --git a/mllib/src/test/scala/org/apache/spark/mllib/tree/GradientBoostedTreesSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/tree/GradientBoostedTreesSuite.scala index b437aeaaf0547..55b0bac7d49fe 100644 --- a/mllib/src/test/scala/org/apache/spark/mllib/tree/GradientBoostedTreesSuite.scala +++ b/mllib/src/test/scala/org/apache/spark/mllib/tree/GradientBoostedTreesSuite.scala @@ -175,10 +175,11 @@ class GradientBoostedTreesSuite extends FunSuite with MLlibTestSparkContext { new BoostingStrategy(treeStrategy, loss, numIterations, validationTol = 0.0) val gbtValidate = new GradientBoostedTrees(boostingStrategy) .runWithValidation(trainRdd, validateRdd) - assert(gbtValidate.numTrees !== numIterations) + val numTrees = gbtValidate.numTrees + assert(numTrees !== numIterations) // Test that it performs better on the validation dataset. - val gbt = GradientBoostedTrees.train(trainRdd, boostingStrategy) + val gbt = new GradientBoostedTrees(boostingStrategy).run(trainRdd) val (errorWithoutValidation, errorWithValidation) = { if (algo == Classification) { val remappedRdd = validateRdd.map(x => new LabeledPoint(2 * x.label - 1, x.features)) @@ -188,6 +189,17 @@ class GradientBoostedTreesSuite extends FunSuite with MLlibTestSparkContext { } } assert(errorWithValidation <= errorWithoutValidation) + + // Test that results from evaluateEachIteration comply with runWithValidation. + // Note that convergenceTol is set to 0.0 + val evaluationArray = gbt.evaluateEachIteration(validateRdd, loss) + assert(evaluationArray.length === numIterations) + assert(evaluationArray(numTrees) > evaluationArray(numTrees - 1)) + var i = 1 + while (i < numTrees) { + assert(evaluationArray(i) <= evaluationArray(i - 1)) + i += 1 + } } } }