From 8a4edecc483490da4fce406ac3b401860834ac1e Mon Sep 17 00:00:00 2001 From: zsxwing Date: Tue, 12 May 2015 01:34:33 -0700 Subject: [PATCH 01/27] [SPARK-7534] [CORE] [WEBUI] Fix the Stage table when a stage is missing Just improved the Stage table when a stage is missing. Before: ![screen shot 2015-05-11 at 10 11 51 am](https://cloud.githubusercontent.com/assets/1000778/7570842/2ba37380-f7c8-11e4-9b5f-cf1a6264b2a4.png) After: ![screen shot 2015-05-11 at 10 26 09 am](https://cloud.githubusercontent.com/assets/1000778/7570848/33703152-f7c8-11e4-81a8-d53dd72d7b8d.png) Author: zsxwing Closes #6061 from zsxwing/SPARK-7534 and squashes the following commits: 09fe862 [zsxwing] Leave it blank rather than '-' 6299197 [zsxwing] Fix the Stage table when a stage is missing --- .../org/apache/spark/ui/jobs/StageTable.scala | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) diff --git a/core/src/main/scala/org/apache/spark/ui/jobs/StageTable.scala b/core/src/main/scala/org/apache/spark/ui/jobs/StageTable.scala index a33243d4252bf..82ba561eefb16 100644 --- a/core/src/main/scala/org/apache/spark/ui/jobs/StageTable.scala +++ b/core/src/main/scala/org/apache/spark/ui/jobs/StageTable.scala @@ -121,10 +121,23 @@ private[ui] class StageTableBase(
{stageDesc.getOrElse("")} {killLink} {nameLink} {details}
} + protected def missingStageRow(stageId: Int): Seq[Node] = { + {stageId} ++ + {if (isFairScheduler) {-} else Seq.empty} ++ + No data available for this stage ++ // Description + ++ // Submitted + ++ // Duration + ++ // Tasks: Succeeded/Total + ++ // Input + ++ // Output + ++ // Shuffle Read + // Shuffle Write + } + protected def stageRow(s: StageInfo): Seq[Node] = { val stageDataOption = listener.stageIdToData.get((s.stageId, s.attemptId)) if (stageDataOption.isEmpty) { - return {s.stageId}No data available for this stage + return missingStageRow(s.stageId) } val stageData = stageDataOption.get From 984787526625b4ef8a1635faf7a5ac3cb0b758b7 Mon Sep 17 00:00:00 2001 From: linweizhong Date: Tue, 12 May 2015 01:36:27 -0700 Subject: [PATCH 02/27] [MINOR] [PYSPARK] Set PYTHONPATH to python/lib/pyspark.zip rather than python/pyspark As PR #5580 we have created pyspark.zip on building and set PYTHONPATH to python/lib/pyspark.zip, so to keep consistence update this. Author: linweizhong Closes #6047 from Sephiroth-Lin/pyspark_pythonpath and squashes the following commits: 8cc3d96 [linweizhong] Set PYTHONPATH to python/lib/pyspark.zip rather than python/pyspark as PR#5580 we have create pyspark.zip on build --- .../main/scala/org/apache/spark/api/python/PythonUtils.scala | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/core/src/main/scala/org/apache/spark/api/python/PythonUtils.scala b/core/src/main/scala/org/apache/spark/api/python/PythonUtils.scala index acbaba6791850..9eff0a230cbe2 100644 --- a/core/src/main/scala/org/apache/spark/api/python/PythonUtils.scala +++ b/core/src/main/scala/org/apache/spark/api/python/PythonUtils.scala @@ -31,7 +31,7 @@ private[spark] object PythonUtils { def sparkPythonPath: String = { val pythonPath = new ArrayBuffer[String] for (sparkHome <- sys.env.get("SPARK_HOME")) { - pythonPath += Seq(sparkHome, "python").mkString(File.separator) + pythonPath += Seq(sparkHome, "python", "lib", "pyspark.zip").mkString(File.separator) pythonPath += Seq(sparkHome, "python", "lib", "py4j-0.8.2.1-src.zip").mkString(File.separator) } pythonPath ++= SparkContext.jarOfObject(this) From 82e890fb19d6fbaffa69856eecb4699f2f8a81eb Mon Sep 17 00:00:00 2001 From: Marcelo Vanzin Date: Tue, 12 May 2015 01:39:21 -0700 Subject: [PATCH 03/27] [SPARK-7485] [BUILD] Remove pyspark files from assembly. The sbt part of the build is hacky; it basically tricks sbt into generating the zip by using a generator, but returns an empty list for the generated files so that nothing is actually added to the assembly. Author: Marcelo Vanzin Closes #6022 from vanzin/SPARK-7485 and squashes the following commits: 22c1e04 [Marcelo Vanzin] Remove unneeded code. 4893622 [Marcelo Vanzin] [SPARK-7485] [build] Remove pyspark files from assembly. --- core/pom.xml | 47 ---------------------------------------- mllib/pom.xml | 11 ---------- project/SparkBuild.scala | 44 +++---------------------------------- sql/core/pom.xml | 8 ------- streaming/pom.xml | 8 ------- 5 files changed, 3 insertions(+), 115 deletions(-) diff --git a/core/pom.xml b/core/pom.xml index fc42f48973fe9..262a3320db106 100644 --- a/core/pom.xml +++ b/core/pom.xml @@ -381,35 +381,6 @@ target/scala-${scala.binary.version}/classes target/scala-${scala.binary.version}/test-classes - - - org.apache.maven.plugins - maven-antrun-plugin - - - generate-resources - - run - - - - - - - - - - - maven-clean-plugin - - - - ${basedir}/../python/build - - - true - - org.apache.maven.plugins maven-dependency-plugin @@ -438,24 +409,6 @@ - - - - src/main/resources - - - ../python - - pyspark/*.py - - - - ../python/build - - py4j/*.py - - - diff --git a/mllib/pom.xml b/mllib/pom.xml index a3c57ae26000b..0c07ca1a62fd3 100644 --- a/mllib/pom.xml +++ b/mllib/pom.xml @@ -141,16 +141,5 @@ target/scala-${scala.binary.version}/classes target/scala-${scala.binary.version}/test-classes - - - ../python - - pyspark/mllib/*.py - pyspark/mllib/stat/*.py - pyspark/ml/*.py - pyspark/ml/param/*.py - - - diff --git a/project/SparkBuild.scala b/project/SparkBuild.scala index 186345af0e60e..1b87e4e98bd83 100644 --- a/project/SparkBuild.scala +++ b/project/SparkBuild.scala @@ -168,7 +168,7 @@ object SparkBuild extends PomBuild { /* Enable Assembly for all assembly projects */ assemblyProjects.foreach(enable(Assembly.settings)) - /* Package pyspark artifacts in the main assembly. */ + /* Package pyspark artifacts in a separate zip file for YARN. */ enable(PySparkAssembly.settings)(assembly) /* Enable unidoc only for the root spark project */ @@ -373,22 +373,15 @@ object PySparkAssembly { import java.util.zip.{ZipOutputStream, ZipEntry} lazy val settings = Seq( - unmanagedJars in Compile += { BuildCommons.sparkHome / "python/lib/py4j-0.8.2.1-src.zip" }, // Use a resource generator to copy all .py files from python/pyspark into a managed directory // to be included in the assembly. We can't just add "python/" to the assembly's resource dir // list since that will copy unneeded / unwanted files. resourceGenerators in Compile <+= resourceManaged in Compile map { outDir: File => val src = new File(BuildCommons.sparkHome, "python/pyspark") - val zipFile = new File(BuildCommons.sparkHome , "python/lib/pyspark.zip") zipFile.delete() zipRecursive(src, zipFile) - - val dst = new File(outDir, "pyspark") - if (!dst.isDirectory()) { - require(dst.mkdirs()) - } - copy(src, dst) + Seq[File]() } ) @@ -416,42 +409,11 @@ object PySparkAssembly { output.write(buf, 0, n) } } + output.closeEntry() in.close() } } - private def copy(src: File, dst: File): Seq[File] = { - src.listFiles().flatMap { f => - val child = new File(dst, f.getName()) - if (f.isDirectory()) { - child.mkdir() - copy(f, child) - } else if (f.getName().endsWith(".py")) { - var in: Option[FileInputStream] = None - var out: Option[FileOutputStream] = None - try { - in = Some(new FileInputStream(f)) - out = Some(new FileOutputStream(child)) - - val bytes = new Array[Byte](1024) - var read = 0 - while (read >= 0) { - read = in.get.read(bytes) - if (read > 0) { - out.get.write(bytes, 0, read) - } - } - - Some(child) - } finally { - in.foreach(_.close()) - out.foreach(_.close()) - } - } else { - None - } - } - } } object Unidoc { diff --git a/sql/core/pom.xml b/sql/core/pom.xml index 7d274a73e079f..ffe95bb49188f 100644 --- a/sql/core/pom.xml +++ b/sql/core/pom.xml @@ -103,13 +103,5 @@ target/scala-${scala.binary.version}/classes target/scala-${scala.binary.version}/test-classes - - - ../../python - - pyspark/sql/*.py - - - diff --git a/streaming/pom.xml b/streaming/pom.xml index 5ca55a4f680bb..5ab7f4472c38b 100644 --- a/streaming/pom.xml +++ b/streaming/pom.xml @@ -105,13 +105,5 @@ - - - ../python - - pyspark/streaming/*.py - - - From f3e8e60063ccf0d713d03e671a3231560475f90d Mon Sep 17 00:00:00 2001 From: Andrew Or Date: Tue, 12 May 2015 01:40:55 -0700 Subject: [PATCH 04/27] [SPARK-7467] Dag visualization: treat checkpoint as an RDD operation Such that a checkpoint RDD does not go into random scopes on the UI, e.g. `take`. We've seen this in streaming. Author: Andrew Or Closes #6004 from andrewor14/dag-viz-checkpoint and squashes the following commits: 9217439 [Andrew Or] Fix checkpoints 4ae8806 [Andrew Or] Merge branch 'master' of github.com:apache/spark into dag-viz-checkpoint 19bc07b [Andrew Or] Treat checkpoint as an RDD operation --- .../main/scala/org/apache/spark/rdd/RDD.scala | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) 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 7dad30ecbdd2f..02a94baf372d9 100644 --- a/core/src/main/scala/org/apache/spark/rdd/RDD.scala +++ b/core/src/main/scala/org/apache/spark/rdd/RDD.scala @@ -1523,13 +1523,15 @@ abstract class RDD[T: ClassTag]( * has completed (therefore the RDD has been materialized and potentially stored in memory). * doCheckpoint() is called recursively on the parent RDDs. */ - private[spark] def doCheckpoint() { - if (!doCheckpointCalled) { - doCheckpointCalled = true - if (checkpointData.isDefined) { - checkpointData.get.doCheckpoint() - } else { - dependencies.foreach(_.rdd.doCheckpoint()) + private[spark] def doCheckpoint(): Unit = { + RDDOperationScope.withScope(sc, "checkpoint", false, true) { + if (!doCheckpointCalled) { + doCheckpointCalled = true + if (checkpointData.isDefined) { + checkpointData.get.doCheckpoint() + } else { + dependencies.foreach(_.rdd.doCheckpoint()) + } } } } From ec6f2a9774167014566fb9608ee4394d2ce5fd6a Mon Sep 17 00:00:00 2001 From: Tathagata Das Date: Tue, 12 May 2015 08:48:24 -0700 Subject: [PATCH 05/27] [SPARK-7532] [STREAMING] StreamingContext.start() made to logWarning and not throw exception Author: Tathagata Das Closes #6060 from tdas/SPARK-7532 and squashes the following commits: 6fe2e83 [Tathagata Das] Update docs 7dadfc3 [Tathagata Das] Fixed bug again 99c7678 [Tathagata Das] Added logInfo 65aec20 [Tathagata Das] Fix bug 5bf031b [Tathagata Das] Merge remote-tracking branch 'apache-github/master' into SPARK-7532 1a9a818 [Tathagata Das] Fix scaladoc c584313 [Tathagata Das] StreamingContext.start() made to logWarning and not throw exception --- .../spark/streaming/StreamingContext.scala | 27 +++++++++---------- .../streaming/StreamingContextSuite.scala | 4 +-- 2 files changed, 14 insertions(+), 17 deletions(-) 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 2c5834defa8c8..8461e901208ff 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/StreamingContext.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/StreamingContext.scala @@ -528,28 +528,27 @@ class StreamingContext private[streaming] ( /** * Start the execution of the streams. * - * @throws SparkException if the context has already been started or stopped. + * @throws SparkException if the StreamingContext is already stopped. */ def start(): Unit = synchronized { - import StreamingContext._ state match { case INITIALIZED => - // good to start + validate() + startSite.set(DStream.getCreationSite()) + sparkContext.setCallSite(startSite.get) + StreamingContext.ACTIVATION_LOCK.synchronized { + StreamingContext.assertNoOtherContextIsActive() + scheduler.start() + uiTab.foreach(_.attach()) + state = StreamingContextState.ACTIVE + StreamingContext.setActiveContext(this) + } + logInfo("StreamingContext started") case ACTIVE => - throw new SparkException("StreamingContext has already been started") + logWarning("StreamingContext has already been started") case STOPPED => throw new SparkException("StreamingContext has already been stopped") } - validate() - startSite.set(DStream.getCreationSite()) - sparkContext.setCallSite(startSite.get) - ACTIVATION_LOCK.synchronized { - assertNoOtherContextIsActive() - scheduler.start() - uiTab.foreach(_.attach()) - state = StreamingContextState.ACTIVE - setActiveContext(this) - } } /** 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 b8247db7e8950..47299513de095 100644 --- a/streaming/src/test/scala/org/apache/spark/streaming/StreamingContextSuite.scala +++ b/streaming/src/test/scala/org/apache/spark/streaming/StreamingContextSuite.scala @@ -136,9 +136,7 @@ class StreamingContextSuite extends FunSuite with BeforeAndAfter with Timeouts w addInputStream(ssc).register() ssc.start() assert(ssc.getState() === StreamingContextState.ACTIVE) - intercept[SparkException] { - ssc.start() - } + ssc.start() assert(ssc.getState() === StreamingContextState.ACTIVE) } From d86ce845840a92b4dde7975082738ed94ab8c570 Mon Sep 17 00:00:00 2001 From: Daoyuan Wang Date: Tue, 12 May 2015 10:23:41 -0700 Subject: [PATCH 06/27] [SPARK-6876] [PySpark] [SQL] add DataFrame na.replace in pyspark Author: Daoyuan Wang Closes #6003 from adrian-wang/pynareplace and squashes the following commits: 672efba [Daoyuan Wang] remove py2.7 feature 4a148f7 [Daoyuan Wang] to_replace support dict, value support single value, and add full tests 9e232e7 [Daoyuan Wang] rename scala map af0268a [Daoyuan Wang] remove na 63ac579 [Daoyuan Wang] add na.replace in pyspark --- .../apache/spark/api/python/PythonUtils.scala | 7 ++ python/pyspark/sql/dataframe.py | 85 +++++++++++++++++++ python/pyspark/sql/tests.py | 48 +++++++++++ 3 files changed, 140 insertions(+) diff --git a/core/src/main/scala/org/apache/spark/api/python/PythonUtils.scala b/core/src/main/scala/org/apache/spark/api/python/PythonUtils.scala index 9eff0a230cbe2..efb6b93cfc35d 100644 --- a/core/src/main/scala/org/apache/spark/api/python/PythonUtils.scala +++ b/core/src/main/scala/org/apache/spark/api/python/PythonUtils.scala @@ -53,4 +53,11 @@ private[spark] object PythonUtils { def toSeq[T](cols: JList[T]): Seq[T] = { cols.toList.toSeq } + + /** + * Convert java map of K, V into Map of K, V (for calling API with varargs) + */ + def toScalaMap[K, V](jm: java.util.Map[K, V]): Map[K, V] = { + jm.toMap + } } diff --git a/python/pyspark/sql/dataframe.py b/python/pyspark/sql/dataframe.py index 72180f6d05fbc..078acfdf7e2df 100644 --- a/python/pyspark/sql/dataframe.py +++ b/python/pyspark/sql/dataframe.py @@ -578,6 +578,10 @@ def _jseq(self, cols, converter=None): """Return a JVM Seq of Columns from a list of Column or names""" return _to_seq(self.sql_ctx._sc, cols, converter) + def _jmap(self, jm): + """Return a JVM Scala Map from a dict""" + return _to_scala_map(self.sql_ctx._sc, jm) + def _jcols(self, *cols): """Return a JVM Seq of Columns from a list of Column or column names @@ -924,6 +928,80 @@ def fillna(self, value, subset=None): return DataFrame(self._jdf.na().fill(value, self._jseq(subset)), self.sql_ctx) + def replace(self, to_replace, value, subset=None): + """Returns a new :class:`DataFrame` replacing a value with another value. + + :param to_replace: int, long, float, string, or list. + Value to be replaced. + If the value is a dict, then `value` is ignored and `to_replace` must be a + mapping from column name (string) to replacement value. The value to be + replaced must be an int, long, float, or string. + :param value: int, long, float, string, or list. + Value to use to replace holes. + The replacement value must be an int, long, float, or string. If `value` is a + list or tuple, `value` should be of the same length with `to_replace`. + :param subset: optional list of column names to consider. + Columns specified in subset that do not have matching data type are ignored. + For example, if `value` is a string, and subset contains a non-string column, + then the non-string column is simply ignored. + >>> df4.replace(10, 20).show() + +----+------+-----+ + | age|height| name| + +----+------+-----+ + | 20| 80|Alice| + | 5| null| Bob| + |null| null| Tom| + |null| null| null| + +----+------+-----+ + + >>> df4.replace(['Alice', 'Bob'], ['A', 'B'], 'name').show() + +----+------+----+ + | age|height|name| + +----+------+----+ + | 10| 80| A| + | 5| null| B| + |null| null| Tom| + |null| null|null| + +----+------+----+ + """ + if not isinstance(to_replace, (float, int, long, basestring, list, tuple, dict)): + raise ValueError( + "to_replace should be a float, int, long, string, list, tuple, or dict") + + if not isinstance(value, (float, int, long, basestring, list, tuple)): + raise ValueError("value should be a float, int, long, string, list, or tuple") + + rep_dict = dict() + + if isinstance(to_replace, (float, int, long, basestring)): + to_replace = [to_replace] + + if isinstance(to_replace, tuple): + to_replace = list(to_replace) + + if isinstance(value, tuple): + value = list(value) + + if isinstance(to_replace, list) and isinstance(value, list): + if len(to_replace) != len(value): + raise ValueError("to_replace and value lists should be of the same length") + rep_dict = dict(zip(to_replace, value)) + elif isinstance(to_replace, list) and isinstance(value, (float, int, long, basestring)): + rep_dict = dict([(tr, value) for tr in to_replace]) + elif isinstance(to_replace, dict): + rep_dict = to_replace + + if subset is None: + return DataFrame(self._jdf.na().replace('*', rep_dict), self.sql_ctx) + elif isinstance(subset, basestring): + subset = [subset] + + if not isinstance(subset, (list, tuple)): + raise ValueError("subset should be a list or tuple of column names") + + return DataFrame( + self._jdf.na().replace(self._jseq(subset), self._jmap(rep_dict)), self.sql_ctx) + def corr(self, col1, col2, method=None): """ Calculates the correlation of two columns of a DataFrame as a double value. Currently only @@ -1226,6 +1304,13 @@ def _to_seq(sc, cols, converter=None): return sc._jvm.PythonUtils.toSeq(cols) +def _to_scala_map(sc, jm): + """ + Convert a dict into a JVM Map. + """ + return sc._jvm.PythonUtils.toScalaMap(jm) + + def _unary_op(name, doc="unary operator"): """ Create a method for given unary operator """ def _(self): diff --git a/python/pyspark/sql/tests.py b/python/pyspark/sql/tests.py index 7e63f4d6461f6..1922d03af61da 100644 --- a/python/pyspark/sql/tests.py +++ b/python/pyspark/sql/tests.py @@ -665,6 +665,54 @@ def test_bitwise_operations(self): result = df.select(functions.bitwiseNOT(df.b)).collect()[0].asDict() self.assertEqual(~75, result['~b']) + def test_replace(self): + schema = StructType([ + StructField("name", StringType(), True), + StructField("age", IntegerType(), True), + StructField("height", DoubleType(), True)]) + + # replace with int + row = self.sqlCtx.createDataFrame([(u'Alice', 10, 10.0)], schema).replace(10, 20).first() + self.assertEqual(row.age, 20) + self.assertEqual(row.height, 20.0) + + # replace with double + row = self.sqlCtx.createDataFrame( + [(u'Alice', 80, 80.0)], schema).replace(80.0, 82.1).first() + self.assertEqual(row.age, 82) + self.assertEqual(row.height, 82.1) + + # replace with string + row = self.sqlCtx.createDataFrame( + [(u'Alice', 10, 80.1)], schema).replace(u'Alice', u'Ann').first() + self.assertEqual(row.name, u"Ann") + self.assertEqual(row.age, 10) + + # replace with subset specified by a string of a column name w/ actual change + row = self.sqlCtx.createDataFrame( + [(u'Alice', 10, 80.1)], schema).replace(10, 20, subset='age').first() + self.assertEqual(row.age, 20) + + # replace with subset specified by a string of a column name w/o actual change + row = self.sqlCtx.createDataFrame( + [(u'Alice', 10, 80.1)], schema).replace(10, 20, subset='height').first() + self.assertEqual(row.age, 10) + + # replace with subset specified with one column replaced, another column not in subset + # stays unchanged. + row = self.sqlCtx.createDataFrame( + [(u'Alice', 10, 10.0)], schema).replace(10, 20, subset=['name', 'age']).first() + self.assertEqual(row.name, u'Alice') + self.assertEqual(row.age, 20) + self.assertEqual(row.height, 10.0) + + # replace with subset specified but no column will be replaced + row = self.sqlCtx.createDataFrame( + [(u'Alice', 10, None)], schema).replace(10, 20, subset=['name', 'height']).first() + self.assertEqual(row.name, u'Alice') + self.assertEqual(row.age, 10) + self.assertEqual(row.height, None) + class HiveContextSQLTests(ReusedPySparkTestCase): From 831504cf6bde6b1131005d5552e56a842725c84c Mon Sep 17 00:00:00 2001 From: Wenchen Fan Date: Tue, 12 May 2015 10:28:40 -0700 Subject: [PATCH 07/27] [DataFrame][minor] cleanup unapply methods in DataTypes Author: Wenchen Fan Closes #6079 from cloud-fan/unapply and squashes the following commits: 40da442 [Wenchen Fan] one more 7d90a05 [Wenchen Fan] cleanup unapply in DataTypes --- .../org/apache/spark/sql/types/DataType.scala | 15 +++------------ 1 file changed, 3 insertions(+), 12 deletions(-) diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/types/DataType.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/types/DataType.scala index 0992a7c311ee2..a0b261649f66f 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/types/DataType.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/types/DataType.scala @@ -47,10 +47,7 @@ abstract class DataType { * ... * }}} */ - private[sql] def unapply(a: Expression): Boolean = a match { - case e: Expression if e.dataType == this => true - case _ => false - } + private[sql] def unapply(e: Expression): Boolean = e.dataType == this /** * The default size of a value of this data type, used internally for size estimation. @@ -137,10 +134,7 @@ private[sql] object IntegralType { * ... * }}} */ - def unapply(a: Expression): Boolean = a match { - case e: Expression if e.dataType.isInstanceOf[IntegralType] => true - case _ => false - } + def unapply(e: Expression): Boolean = e.dataType.isInstanceOf[IntegralType] } @@ -157,10 +151,7 @@ private[sql] object FractionalType { * ... * }}} */ - def unapply(a: Expression): Boolean = a match { - case e: Expression if e.dataType.isInstanceOf[FractionalType] => true - case _ => false - } + def unapply(e: Expression): Boolean = e.dataType.isInstanceOf[FractionalType] } From 0595b6de8f1da04baceda082553c2aa1aa2cb006 Mon Sep 17 00:00:00 2001 From: Cheng Lian Date: Wed, 13 May 2015 01:32:28 +0800 Subject: [PATCH 08/27] [SPARK-3928] [SPARK-5182] [SQL] Partitioning support for the data sources API This PR adds partitioning support for the external data sources API. It aims to simplify development of file system based data sources, and provide first class partitioning support for both read path and write path. Existing data sources like JSON and Parquet can be simplified with this work. ## New features provided 1. Hive compatible partition discovery This actually generalizes the partition discovery strategy used in Parquet data source in Spark 1.3.0. 1. Generalized partition pruning optimization Now partition pruning is handled during physical planning phase. Specific data sources don't need to worry about this harness anymore. (This also implies that we can remove `CatalystScan` after migrating the Parquet data source, since now we don't need to pass Catalyst expressions to data source implementations.) 1. Insertion with dynamic partitions When inserting data to a `FSBasedRelation`, data can be partitioned dynamically by specified partition columns. ## New structures provided ### Developer API 1. `FSBasedRelation` Base abstract class for file system based data sources. 1. `OutputWriter` Base abstract class for output row writers, responsible for writing a single row object. 1. `FSBasedRelationProvider` A new relation provider for `FSBasedRelation` subclasses. Note that data sources extending `FSBasedRelation` don't need to extend `RelationProvider` and `SchemaRelationProvider`. ### User API New overloaded versions of 1. `DataFrame.save()` 1. `DataFrame.saveAsTable()` 1. `SQLContext.load()` are provided to allow users to save/load DataFrames with user defined dynamic partition columns. ### Spark SQL query planning 1. `InsertIntoFSBasedRelation` Used to implement write path for `FSBasedRelation`s. 1. New rules for `FSBasedRelation` in `DataSourceStrategy` These are added to hook `FSBasedRelation` into physical query plan in read path, and perform partition pruning. ## TODO - [ ] Use scratch directories when overwriting a table with data selected from itself. Currently, this is not supported, because the table been overwritten is always deleted before writing any data to it. - [ ] When inserting with dynamic partition columns, use external sorter to group the data first. This ensures that we only need to open a single `OutputWriter` at a time. For data sources like Parquet, `OutputWriter`s can be quite memory consuming. One issue is that, this approach breaks the row distribution in the original DataFrame. However, we did't promise to preserve data distribution when writing a DataFrame. - [x] More tests. Specifically, test cases for - [x] Self-join - [x] Loading partitioned relations with a subset of partition columns stored in data files. - [x] `SQLContext.load()` with user defined dynamic partition columns. ## Parquet data source migration Parquet data source migration is covered in PR https://github.com/liancheng/spark/pull/6, which is against this PR branch and for preview only. A formal PR need to be made after this one is merged. Author: Cheng Lian Closes #5526 from liancheng/partitioning-support and squashes the following commits: 5351a1b [Cheng Lian] Fixes compilation error introduced while rebasing 1f9b1a5 [Cheng Lian] Tweaks data schema passed to FSBasedRelations 43ba50e [Cheng Lian] Avoids serializing generated projection code edf49e7 [Cheng Lian] Removed commented stale code block 348a922 [Cheng Lian] Adds projection in FSBasedRelation.buildScan(requiredColumns, inputPaths) ad4d4de [Cheng Lian] Enables HDFS style globbing 8d12e69 [Cheng Lian] Fixes compilation error c71ac6c [Cheng Lian] Addresses comments from @marmbrus 7552168 [Cheng Lian] Fixes typo in MimaExclude.scala 0349e09 [Cheng Lian] Fixes compilation error introduced while rebasing 52b0c9b [Cheng Lian] Adjusts project/MimaExclude.scala c466de6 [Cheng Lian] Addresses comments bc3f9b4 [Cheng Lian] Uses projection to separate partition columns and data columns while inserting rows 795920a [Cheng Lian] Fixes compilation error after rebasing 0b8cd70 [Cheng Lian] Adds Scala/Catalyst row conversion when writing non-partitioned tables fa543f3 [Cheng Lian] Addresses comments 5849dd0 [Cheng Lian] Fixes doc typos. Fixes partition discovery refresh. 51be443 [Cheng Lian] Replaces FSBasedRelation.outputCommitterClass with FSBasedRelation.prepareForWrite c4ed4fe [Cheng Lian] Bug fixes and a new test suite a29e663 [Cheng Lian] Bug fix: should only pass actuall data files to FSBaseRelation.buildScan 5f423d3 [Cheng Lian] Bug fixes. Lets data source to customize OutputCommitter rather than OutputFormat 54c3d7b [Cheng Lian] Enforces that FileOutputFormat must be used be0c268 [Cheng Lian] Uses TaskAttempContext rather than Configuration in OutputWriter.init 0bc6ad1 [Cheng Lian] Resorts to new Hadoop API, and now FSBasedRelation can customize output format class f320766 [Cheng Lian] Adds prepareForWrite() hook, refactored writer containers 422ff4a [Cheng Lian] Fixes style issue ce52353 [Cheng Lian] Adds new SQLContext.load() overload with user defined dynamic partition columns 8d2ff71 [Cheng Lian] Merges partition columns when reading partitioned relations ca1805b [Cheng Lian] Removes duplicated partition discovery code in new Parquet f18dec2 [Cheng Lian] More strict schema checking b746ab5 [Cheng Lian] More tests 9b487bf [Cheng Lian] Fixes compilation errors introduced while rebasing ea6c8dd [Cheng Lian] Removes remote debugging stuff 327bb1d [Cheng Lian] Implements partitioning support for data sources API 3c5073a [Cheng Lian] Fixes SaveModes used in test cases fb5a607 [Cheng Lian] Fixes compilation error 9d17607 [Cheng Lian] Adds the contract that OutputWriter should have zero-arg constructor 5de194a [Cheng Lian] Forgot Apache licence header 95d0b4d [Cheng Lian] Renames PartitionedSchemaRelationProvider to FSBasedRelationProvider 770b5ba [Cheng Lian] Adds tests for FSBasedRelation 3ba9bbf [Cheng Lian] Adds DataFrame.saveAsTable() overrides which support partitioning 1b8231f [Cheng Lian] Renames FSBasedPrunedFilteredScan to FSBasedRelation aa8ba9a [Cheng Lian] Javadoc fix 012ed2d [Cheng Lian] Adds PartitioningOptions 7dd8dd5 [Cheng Lian] Adds new interfaces and stub methods for data sources API partitioning support --- .../apache/spark/deploy/SparkHadoopUtil.scala | 52 +- project/MimaExcludes.scala | 14 +- .../org/apache/spark/sql/DataFrame.scala | 107 +++- .../scala/org/apache/spark/sql/SQLConf.scala | 6 + .../org/apache/spark/sql/SQLContext.scala | 36 +- .../spark/sql/execution/SparkStrategies.scala | 10 +- .../apache/spark/sql/parquet/newParquet.scala | 182 +----- .../sql/sources/DataSourceStrategy.scala | 191 ++++++- .../spark/sql/sources/PartitioningUtils.scala | 207 +++++++ .../apache/spark/sql/sources/commands.scala | 406 +++++++++++++- .../org/apache/spark/sql/sources/ddl.scala | 90 ++- .../apache/spark/sql/sources/interfaces.scala | 283 +++++++++- .../org/apache/spark/sql/sources/rules.scala | 8 +- .../ParquetPartitionDiscoverySuite.scala | 3 +- .../sources/CreateTableAsSelectSuite.scala | 4 +- .../spark/sql/hive/HiveMetastoreCatalog.scala | 13 +- .../spark/sql/hive/HiveStrategies.scala | 6 +- .../spark/sql/hive/execution/commands.scala | 27 +- .../spark/sql/hive/hiveWriterContainers.scala | 2 +- .../sql/sources/FSBasedRelationSuite.scala | 525 ++++++++++++++++++ .../sql/sources/SimpleTextRelation.scala | 125 +++++ 21 files changed, 2042 insertions(+), 255 deletions(-) create mode 100644 sql/core/src/main/scala/org/apache/spark/sql/sources/PartitioningUtils.scala create mode 100644 sql/hive/src/test/scala/org/apache/spark/sql/sources/FSBasedRelationSuite.scala create mode 100644 sql/hive/src/test/scala/org/apache/spark/sql/sources/SimpleTextRelation.scala diff --git a/core/src/main/scala/org/apache/spark/deploy/SparkHadoopUtil.scala b/core/src/main/scala/org/apache/spark/deploy/SparkHadoopUtil.scala index b563034457a91..7fa75ac8c2b54 100644 --- a/core/src/main/scala/org/apache/spark/deploy/SparkHadoopUtil.scala +++ b/core/src/main/scala/org/apache/spark/deploy/SparkHadoopUtil.scala @@ -22,22 +22,22 @@ import java.lang.reflect.Method import java.security.PrivilegedExceptionAction import java.util.{Arrays, Comparator} +import scala.collection.JavaConversions._ +import scala.concurrent.duration._ +import scala.language.postfixOps + import com.google.common.primitives.Longs import org.apache.hadoop.conf.Configuration -import org.apache.hadoop.fs.{FileStatus, FileSystem, Path, PathFilter} import org.apache.hadoop.fs.FileSystem.Statistics +import org.apache.hadoop.fs.{FileStatus, FileSystem, Path, PathFilter} import org.apache.hadoop.hdfs.security.token.delegation.DelegationTokenIdentifier import org.apache.hadoop.mapred.JobConf import org.apache.hadoop.mapreduce.JobContext import org.apache.hadoop.security.{Credentials, UserGroupInformation} -import org.apache.spark.{Logging, SparkConf, SparkException} import org.apache.spark.annotation.DeveloperApi import org.apache.spark.util.Utils - -import scala.collection.JavaConversions._ -import scala.concurrent.duration._ -import scala.language.postfixOps +import org.apache.spark.{Logging, SparkConf, SparkException} /** * :: DeveloperApi :: @@ -199,13 +199,43 @@ class SparkHadoopUtil extends Logging { * that file. */ def listLeafStatuses(fs: FileSystem, basePath: Path): Seq[FileStatus] = { - def recurse(path: Path): Array[FileStatus] = { - val (directories, leaves) = fs.listStatus(path).partition(_.isDir) - leaves ++ directories.flatMap(f => listLeafStatuses(fs, f.getPath)) + listLeafStatuses(fs, fs.getFileStatus(basePath)) + } + + /** + * Get [[FileStatus]] objects for all leaf children (files) under the given base path. If the + * given path points to a file, return a single-element collection containing [[FileStatus]] of + * that file. + */ + def listLeafStatuses(fs: FileSystem, baseStatus: FileStatus): Seq[FileStatus] = { + def recurse(status: FileStatus): Seq[FileStatus] = { + val (directories, leaves) = fs.listStatus(status.getPath).partition(_.isDir) + leaves ++ directories.flatMap(f => listLeafStatuses(fs, f)) } - val baseStatus = fs.getFileStatus(basePath) - if (baseStatus.isDir) recurse(basePath) else Array(baseStatus) + if (baseStatus.isDir) recurse(baseStatus) else Seq(baseStatus) + } + + def listLeafDirStatuses(fs: FileSystem, basePath: Path): Seq[FileStatus] = { + listLeafDirStatuses(fs, fs.getFileStatus(basePath)) + } + + def listLeafDirStatuses(fs: FileSystem, baseStatus: FileStatus): Seq[FileStatus] = { + def recurse(status: FileStatus): Seq[FileStatus] = { + val (directories, files) = fs.listStatus(status.getPath).partition(_.isDir) + val leaves = if (directories.isEmpty) Seq(status) else Seq.empty[FileStatus] + leaves ++ directories.flatMap(dir => listLeafDirStatuses(fs, dir)) + } + + assert(baseStatus.isDir) + recurse(baseStatus) + } + + def globPath(pattern: Path): Seq[Path] = { + val fs = pattern.getFileSystem(conf) + Option(fs.globStatus(pattern)).map { statuses => + statuses.map(_.getPath.makeQualified(fs.getUri, fs.getWorkingDirectory)).toSeq + }.getOrElse(Seq.empty[Path]) } /** diff --git a/project/MimaExcludes.scala b/project/MimaExcludes.scala index ad3d8426bdedd..a47e29e2ef365 100644 --- a/project/MimaExcludes.scala +++ b/project/MimaExcludes.scala @@ -94,11 +94,23 @@ object MimaExcludes { // This `protected[sql]` method was removed in 1.3.1 ProblemFilters.exclude[MissingMethodProblem]( "org.apache.spark.sql.SQLContext.checkAnalysis"), - // This `private[sql]` class was removed in 1.4.0: + // These `private[sql]` class were removed in 1.4.0: ProblemFilters.exclude[MissingClassProblem]( "org.apache.spark.sql.execution.AddExchange"), ProblemFilters.exclude[MissingClassProblem]( "org.apache.spark.sql.execution.AddExchange$"), + ProblemFilters.exclude[MissingClassProblem]( + "org.apache.spark.sql.parquet.PartitionSpec"), + ProblemFilters.exclude[MissingClassProblem]( + "org.apache.spark.sql.parquet.PartitionSpec$"), + ProblemFilters.exclude[MissingClassProblem]( + "org.apache.spark.sql.parquet.Partition"), + ProblemFilters.exclude[MissingClassProblem]( + "org.apache.spark.sql.parquet.Partition$"), + ProblemFilters.exclude[MissingClassProblem]( + "org.apache.spark.sql.parquet.ParquetRelation2$PartitionValues"), + ProblemFilters.exclude[MissingClassProblem]( + "org.apache.spark.sql.parquet.ParquetRelation2$PartitionValues$"), // These test support classes were moved out of src/main and into src/test: ProblemFilters.exclude[MissingClassProblem]( "org.apache.spark.sql.parquet.ParquetTestData"), 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 265a61592b943..f3107f7b51ad8 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 @@ -27,23 +27,23 @@ import scala.reflect.runtime.universe.TypeTag import scala.util.control.NonFatal import com.fasterxml.jackson.core.JsonFactory - import org.apache.commons.lang3.StringUtils + import org.apache.spark.annotation.{DeveloperApi, Experimental} 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.{CatalystTypeConverters, ScalaReflection, SqlParser} -import org.apache.spark.sql.catalyst.analysis.{UnresolvedAttribute, UnresolvedRelation, ResolvedStar} +import org.apache.spark.sql.catalyst.analysis.{ResolvedStar, UnresolvedAttribute, UnresolvedRelation} import org.apache.spark.sql.catalyst.expressions._ -import org.apache.spark.sql.catalyst.plans.{JoinType, Inner} -import org.apache.spark.sql.catalyst.plans.logical._ +import org.apache.spark.sql.catalyst.plans.logical.{Filter, _} +import org.apache.spark.sql.catalyst.plans.{Inner, JoinType} +import org.apache.spark.sql.catalyst.{expressions, CatalystTypeConverters, ScalaReflection, SqlParser} import org.apache.spark.sql.execution.{EvaluatePython, ExplainCommand, LogicalRDD} import org.apache.spark.sql.jdbc.JDBCWriteDetails import org.apache.spark.sql.json.JacksonGenerator +import org.apache.spark.sql.sources.{CreateTableUsingAsSelect, ResolvedDataSource} import org.apache.spark.sql.types._ -import org.apache.spark.sql.sources.{ResolvedDataSource, CreateTableUsingAsSelect} +import org.apache.spark.storage.StorageLevel import org.apache.spark.util.Utils @@ -400,7 +400,9 @@ class DataFrame private[sql]( joined.left, joined.right, joinType = Inner, - Some(EqualTo(joined.left.resolve(usingColumn), joined.right.resolve(usingColumn)))) + Some(expressions.EqualTo( + joined.left.resolve(usingColumn), + joined.right.resolve(usingColumn)))) ) } @@ -465,8 +467,8 @@ class DataFrame private[sql]( // By the time we get here, since we have already run analysis, all attributes should've been // resolved and become AttributeReference. val cond = plan.condition.map { _.transform { - case EqualTo(a: AttributeReference, b: AttributeReference) if a.sameRef(b) => - EqualTo(plan.left.resolve(a.name), plan.right.resolve(b.name)) + case expressions.EqualTo(a: AttributeReference, b: AttributeReference) if a.sameRef(b) => + expressions.EqualTo(plan.left.resolve(a.name), plan.right.resolve(b.name)) }} plan.copy(condition = cond) } @@ -1324,6 +1326,28 @@ class DataFrame private[sql]( saveAsTable(tableName, source, mode, options.toMap) } + /** + * :: Experimental :: + * Creates a table at the given path from the the contents of this DataFrame + * based on a given data source, [[SaveMode]] specified by mode, a set of options, and a list of + * partition columns. + * + * Note that this currently only works with DataFrames that are created from a HiveContext as + * there is no notion of a persisted catalog in a standard SQL context. Instead you can write + * an RDD out to a parquet file, and then register that file as a table. This "table" can then + * be the target of an `insertInto`. + * @group output + */ + @Experimental + def saveAsTable( + tableName: String, + source: String, + mode: SaveMode, + options: java.util.Map[String, String], + partitionColumns: java.util.List[String]): Unit = { + saveAsTable(tableName, source, mode, options.toMap, partitionColumns) + } + /** * :: Experimental :: * (Scala-specific) @@ -1350,6 +1374,7 @@ class DataFrame private[sql]( tableName, source, temporary = false, + Array.empty[String], mode, options, logicalPlan) @@ -1357,6 +1382,36 @@ class DataFrame private[sql]( sqlContext.executePlan(cmd).toRdd } + /** + * :: Experimental :: + * Creates a table at the given path from the the contents of this DataFrame + * based on a given data source, [[SaveMode]] specified by mode, a set of options, and a list of + * partition columns. + * + * Note that this currently only works with DataFrames that are created from a HiveContext as + * there is no notion of a persisted catalog in a standard SQL context. Instead you can write + * an RDD out to a parquet file, and then register that file as a table. This "table" can then + * be the target of an `insertInto`. + * @group output + */ + @Experimental + def saveAsTable( + tableName: String, + source: String, + mode: SaveMode, + options: Map[String, String], + partitionColumns: Seq[String]): Unit = { + sqlContext.executePlan( + CreateTableUsingAsSelect( + tableName, + source, + temporary = false, + partitionColumns.toArray, + mode, + options, + logicalPlan)).toRdd + } + /** * :: Experimental :: * Saves the contents of this DataFrame to the given path, @@ -1417,6 +1472,21 @@ class DataFrame private[sql]( save(source, mode, options.toMap) } + /** + * :: Experimental :: + * Saves the contents of this DataFrame to the given path based on the given data source, + * [[SaveMode]] specified by mode, and partition columns specified by `partitionColumns`. + * @group output + */ + @Experimental + def save( + source: String, + mode: SaveMode, + options: java.util.Map[String, String], + partitionColumns: java.util.List[String]): Unit = { + save(source, mode, options.toMap, partitionColumns) + } + /** * :: Experimental :: * (Scala-specific) @@ -1429,7 +1499,22 @@ class DataFrame private[sql]( source: String, mode: SaveMode, options: Map[String, String]): Unit = { - ResolvedDataSource(sqlContext, source, mode, options, this) + ResolvedDataSource(sqlContext, source, Array.empty[String], mode, options, this) + } + + /** + * :: Experimental :: + * Saves the contents of this DataFrame to the given path based on the given data source, + * [[SaveMode]] specified by mode, and partition columns specified by `partitionColumns`. + * @group output + */ + @Experimental + def save( + source: String, + mode: SaveMode, + options: Map[String, String], + partitionColumns: Seq[String]): Unit = { + ResolvedDataSource(sqlContext, source, partitionColumns.toArray, mode, options, this) } /** diff --git a/sql/core/src/main/scala/org/apache/spark/sql/SQLConf.scala b/sql/core/src/main/scala/org/apache/spark/sql/SQLConf.scala index dcac97beafb04..f07bb196c11ec 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/SQLConf.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/SQLConf.scala @@ -66,6 +66,9 @@ private[spark] object SQLConf { // to its length exceeds the threshold. val SCHEMA_STRING_LENGTH_THRESHOLD = "spark.sql.sources.schemaStringLengthThreshold" + // Whether to perform partition discovery when loading external data sources. Default to true. + val PARTITION_DISCOVERY_ENABLED = "spark.sql.sources.partitionDiscovery.enabled" + // Whether to perform eager analysis when constructing a dataframe. // Set to false when debugging requires the ability to look at invalid query plans. val DATAFRAME_EAGER_ANALYSIS = "spark.sql.eagerAnalysis" @@ -235,6 +238,9 @@ private[sql] class SQLConf extends Serializable with CatalystConf { private[spark] def defaultDataSourceName: String = getConf(DEFAULT_DATA_SOURCE_NAME, "org.apache.spark.sql.parquet") + private[spark] def partitionDiscoveryEnabled() = + getConf(SQLConf.PARTITION_DISCOVERY_ENABLED, "true").toBoolean + // Do not use a value larger than 4000 as the default value of this property. // See the comments of SCHEMA_STRING_LENGTH_THRESHOLD above for more information. private[spark] def schemaStringLengthThreshold: Int = 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 648021806f3ee..afee09adaae20 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 @@ -762,7 +762,7 @@ class SQLContext(@transient val sparkContext: SparkContext) */ @Experimental def load(source: String, options: Map[String, String]): DataFrame = { - val resolved = ResolvedDataSource(this, None, source, options) + val resolved = ResolvedDataSource(this, None, Array.empty[String], source, options) DataFrame(this, LogicalRelation(resolved.relation)) } @@ -781,6 +781,37 @@ class SQLContext(@transient val sparkContext: SparkContext) load(source, schema, options.toMap) } + /** + * :: Experimental :: + * (Java-specific) Returns the dataset specified by the given data source and + * a set of options as a DataFrame, using the given schema as the schema of the DataFrame. + * + * @group genericdata + */ + @Experimental + def load( + source: String, + schema: StructType, + partitionColumns: Array[String], + options: java.util.Map[String, String]): DataFrame = { + load(source, schema, partitionColumns, options.toMap) + } + + /** + * :: Experimental :: + * (Scala-specific) Returns the dataset specified by the given data source and + * a set of options as a DataFrame, using the given schema as the schema of the DataFrame. + * @group genericdata + */ + @Experimental + def load( + source: String, + schema: StructType, + options: Map[String, String]): DataFrame = { + val resolved = ResolvedDataSource(this, Some(schema), Array.empty[String], source, options) + DataFrame(this, LogicalRelation(resolved.relation)) + } + /** * :: Experimental :: * (Scala-specific) Returns the dataset specified by the given data source and @@ -791,8 +822,9 @@ class SQLContext(@transient val sparkContext: SparkContext) def load( source: String, schema: StructType, + partitionColumns: Array[String], options: Map[String, String]): DataFrame = { - val resolved = ResolvedDataSource(this, Some(schema), source, options) + val resolved = ResolvedDataSource(this, Some(schema), partitionColumns, source, options) DataFrame(this, LogicalRelation(resolved.relation)) } 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 56a4689eb58f0..af0029cb84f9a 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 @@ -343,9 +343,13 @@ private[sql] abstract class SparkStrategies extends QueryPlanner[SparkPlan] { case c: CreateTableUsing if c.temporary && c.allowExisting => sys.error("allowExisting should be set to false when creating a temporary table.") - case CreateTableUsingAsSelect(tableName, provider, true, mode, opts, query) => - val cmd = - CreateTempTableUsingAsSelect(tableName, provider, mode, opts, query) + case CreateTableUsingAsSelect(tableName, provider, true, partitionsCols, mode, opts, query) + if partitionsCols.nonEmpty => + sys.error("Cannot create temporary partitioned table.") + + case CreateTableUsingAsSelect(tableName, provider, true, _, mode, opts, query) => + val cmd = CreateTempTableUsingAsSelect( + tableName, provider, Array.empty[String], mode, opts, query) ExecutedCommand(cmd) :: Nil case c: CreateTableUsingAsSelect if !c.temporary => sys.error("Tables created with SQLContext must be TEMPORARY. Use a HiveContext instead.") 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 85e60733bc57a..ee4b1c72a2148 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 @@ -136,10 +136,6 @@ private[sql] class DefaultSource } } -private[sql] case class Partition(values: Row, path: String) - -private[sql] case class PartitionSpec(partitionColumns: StructType, partitions: Seq[Partition]) - /** * An alternative to [[ParquetRelation]] that plugs in using the data sources API. This class is * intended as a full replacement of the Parquet support in Spark SQL. The old implementation will @@ -307,7 +303,7 @@ private[sql] case class ParquetRelation2( if (partitionDirs.nonEmpty) { // Parses names and values of partition columns, and infer their data types. - ParquetRelation2.parsePartitions(partitionDirs, defaultPartitionName) + PartitioningUtils.parsePartitions(partitionDirs, defaultPartitionName) } else { // No partition directories found, makes an empty specification PartitionSpec(StructType(Seq.empty[StructField]), Seq.empty[Partition]) @@ -805,7 +801,7 @@ private[sql] object ParquetRelation2 extends Logging { val ordinalMap = metastoreSchema.zipWithIndex.map { case (field, index) => field.name.toLowerCase -> index }.toMap - val reorderedParquetSchema = mergedParquetSchema.sortBy(f => + val reorderedParquetSchema = mergedParquetSchema.sortBy(f => ordinalMap.getOrElse(f.name.toLowerCase, metastoreSchema.size + 1)) StructType(metastoreSchema.zip(reorderedParquetSchema).map { @@ -841,178 +837,4 @@ private[sql] object ParquetRelation2 extends Logging { .filter(_.nullable) StructType(parquetSchema ++ missingFields) } - - - // TODO Data source implementations shouldn't touch Catalyst types (`Literal`). - // However, we are already using Catalyst expressions for partition pruning and predicate - // push-down here... - private[parquet] case class PartitionValues(columnNames: Seq[String], literals: Seq[Literal]) { - require(columnNames.size == literals.size) - } - - /** - * Given a group of qualified paths, tries to parse them and returns a partition specification. - * For example, given: - * {{{ - * hdfs://:/path/to/partition/a=1/b=hello/c=3.14 - * hdfs://:/path/to/partition/a=2/b=world/c=6.28 - * }}} - * it returns: - * {{{ - * PartitionSpec( - * partitionColumns = StructType( - * StructField(name = "a", dataType = IntegerType, nullable = true), - * StructField(name = "b", dataType = StringType, nullable = true), - * StructField(name = "c", dataType = DoubleType, nullable = true)), - * partitions = Seq( - * Partition( - * values = Row(1, "hello", 3.14), - * path = "hdfs://:/path/to/partition/a=1/b=hello/c=3.14"), - * Partition( - * values = Row(2, "world", 6.28), - * path = "hdfs://:/path/to/partition/a=2/b=world/c=6.28"))) - * }}} - */ - private[parquet] def parsePartitions( - paths: Seq[Path], - defaultPartitionName: String): PartitionSpec = { - val partitionValues = resolvePartitions(paths.map(parsePartition(_, defaultPartitionName))) - val fields = { - val (PartitionValues(columnNames, literals)) = partitionValues.head - columnNames.zip(literals).map { case (name, Literal(_, dataType)) => - StructField(name, dataType, nullable = true) - } - } - - val partitions = partitionValues.zip(paths).map { - case (PartitionValues(_, literals), path) => - Partition(Row(literals.map(_.value): _*), path.toString) - } - - PartitionSpec(StructType(fields), partitions) - } - - /** - * Parses a single partition, returns column names and values of each partition column. For - * example, given: - * {{{ - * path = hdfs://:/path/to/partition/a=42/b=hello/c=3.14 - * }}} - * it returns: - * {{{ - * PartitionValues( - * Seq("a", "b", "c"), - * Seq( - * Literal.create(42, IntegerType), - * Literal.create("hello", StringType), - * Literal.create(3.14, FloatType))) - * }}} - */ - private[parquet] def parsePartition( - path: Path, - defaultPartitionName: String): PartitionValues = { - val columns = ArrayBuffer.empty[(String, Literal)] - // Old Hadoop versions don't have `Path.isRoot` - var finished = path.getParent == null - var chopped = path - - while (!finished) { - val maybeColumn = parsePartitionColumn(chopped.getName, defaultPartitionName) - maybeColumn.foreach(columns += _) - chopped = chopped.getParent - finished = maybeColumn.isEmpty || chopped.getParent == null - } - - val (columnNames, values) = columns.reverse.unzip - PartitionValues(columnNames, values) - } - - private def parsePartitionColumn( - columnSpec: String, - defaultPartitionName: String): Option[(String, Literal)] = { - val equalSignIndex = columnSpec.indexOf('=') - if (equalSignIndex == -1) { - None - } else { - val columnName = columnSpec.take(equalSignIndex) - assert(columnName.nonEmpty, s"Empty partition column name in '$columnSpec'") - - val rawColumnValue = columnSpec.drop(equalSignIndex + 1) - assert(rawColumnValue.nonEmpty, s"Empty partition column value in '$columnSpec'") - - val literal = inferPartitionColumnValue(rawColumnValue, defaultPartitionName) - Some(columnName -> literal) - } - } - - /** - * Resolves possible type conflicts between partitions by up-casting "lower" types. The up- - * casting order is: - * {{{ - * NullType -> - * IntegerType -> LongType -> - * FloatType -> DoubleType -> DecimalType.Unlimited -> - * StringType - * }}} - */ - private[parquet] def resolvePartitions(values: Seq[PartitionValues]): Seq[PartitionValues] = { - // Column names of all partitions must match - val distinctPartitionsColNames = values.map(_.columnNames).distinct - assert(distinctPartitionsColNames.size == 1, { - val list = distinctPartitionsColNames.mkString("\t", "\n", "") - s"Conflicting partition column names detected:\n$list" - }) - - // Resolves possible type conflicts for each column - val columnCount = values.head.columnNames.size - val resolvedValues = (0 until columnCount).map { i => - resolveTypeConflicts(values.map(_.literals(i))) - } - - // Fills resolved literals back to each partition - values.zipWithIndex.map { case (d, index) => - d.copy(literals = resolvedValues.map(_(index))) - } - } - - /** - * Converts a string to a `Literal` with automatic type inference. Currently only supports - * [[IntegerType]], [[LongType]], [[FloatType]], [[DoubleType]], [[DecimalType.Unlimited]], and - * [[StringType]]. - */ - private[parquet] def inferPartitionColumnValue( - raw: String, - defaultPartitionName: String): Literal = { - // First tries integral types - Try(Literal.create(Integer.parseInt(raw), IntegerType)) - .orElse(Try(Literal.create(JLong.parseLong(raw), LongType))) - // Then falls back to fractional types - .orElse(Try(Literal.create(JFloat.parseFloat(raw), FloatType))) - .orElse(Try(Literal.create(JDouble.parseDouble(raw), DoubleType))) - .orElse(Try(Literal.create(new JBigDecimal(raw), DecimalType.Unlimited))) - // Then falls back to string - .getOrElse { - if (raw == defaultPartitionName) Literal.create(null, NullType) - else Literal.create(raw, StringType) - } - } - - private val upCastingOrder: Seq[DataType] = - Seq(NullType, IntegerType, LongType, FloatType, DoubleType, DecimalType.Unlimited, StringType) - - /** - * Given a collection of [[Literal]]s, resolves possible type conflicts by up-casting "lower" - * types. - */ - private def resolveTypeConflicts(literals: Seq[Literal]): Seq[Literal] = { - val desiredType = { - val topType = literals.map(_.dataType).maxBy(upCastingOrder.indexOf(_)) - // Falls back to string if all values of this column are null or empty string - if (topType == NullType) StringType else topType - } - - literals.map { case l @ Literal(_, dataType) => - Literal.create(Cast(l, desiredType).eval(), desiredType) - } - } } diff --git a/sql/core/src/main/scala/org/apache/spark/sql/sources/DataSourceStrategy.scala b/sql/core/src/main/scala/org/apache/spark/sql/sources/DataSourceStrategy.scala index b3d71f687a60a..a5410cda0fe6b 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/sources/DataSourceStrategy.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/sources/DataSourceStrategy.scala @@ -17,20 +17,25 @@ package org.apache.spark.sql.sources +import org.apache.hadoop.fs.Path + +import org.apache.spark.Logging +import org.apache.spark.deploy.SparkHadoopUtil import org.apache.spark.rdd.RDD +import org.apache.spark.sql.Row import org.apache.spark.sql.catalyst.expressions import org.apache.spark.sql.catalyst.expressions._ import org.apache.spark.sql.catalyst.planning.PhysicalOperation import org.apache.spark.sql.catalyst.plans.logical import org.apache.spark.sql.catalyst.plans.logical.LogicalPlan import org.apache.spark.sql.execution.SparkPlan -import org.apache.spark.sql.types.{UTF8String, StringType} -import org.apache.spark.sql.{Row, Strategy, execution, sources} +import org.apache.spark.sql.types.{StructType, UTF8String, StringType} +import org.apache.spark.sql._ /** * A Strategy for planning scans over data sources defined using the sources API. */ -private[sql] object DataSourceStrategy extends Strategy { +private[sql] object DataSourceStrategy extends Strategy with Logging { def apply(plan: LogicalPlan): Seq[execution.SparkPlan] = plan match { case PhysicalOperation(projectList, filters, l @ LogicalRelation(t: CatalystScan)) => pruneFilterProjectRaw( @@ -53,6 +58,51 @@ private[sql] object DataSourceStrategy extends Strategy { filters, (a, _) => t.buildScan(a)) :: Nil + // Scanning partitioned FSBasedRelation + case PhysicalOperation(projectList, filters, l @ LogicalRelation(t: FSBasedRelation)) + if t.partitionSpec.partitionColumns.nonEmpty => + val selectedPartitions = prunePartitions(filters, t.partitionSpec).toArray + + logInfo { + val total = t.partitionSpec.partitions.length + val selected = selectedPartitions.length + val percentPruned = (1 - total.toDouble / selected.toDouble) * 100 + s"Selected $selected partitions out of $total, pruned $percentPruned% partitions." + } + + // Only pushes down predicates that do not reference partition columns. + val pushedFilters = { + val partitionColumnNames = t.partitionSpec.partitionColumns.map(_.name).toSet + filters.filter { f => + val referencedColumnNames = f.references.map(_.name).toSet + referencedColumnNames.intersect(partitionColumnNames).isEmpty + } + } + + buildPartitionedTableScan( + l, + projectList, + pushedFilters, + t.partitionSpec.partitionColumns, + selectedPartitions) :: Nil + + // Scanning non-partitioned FSBasedRelation + case PhysicalOperation(projectList, filters, l @ LogicalRelation(t: FSBasedRelation)) => + val inputPaths = t.paths.map(new Path(_)).flatMap { path => + val fs = path.getFileSystem(t.sqlContext.sparkContext.hadoopConfiguration) + val qualifiedPath = path.makeQualified(fs.getUri, fs.getWorkingDirectory) + SparkHadoopUtil.get.listLeafStatuses(fs, qualifiedPath).map(_.getPath).filterNot { path => + val name = path.getName + name.startsWith("_") || name.startsWith(".") + }.map(fs.makeQualified(_).toString) + } + + pruneFilterProject( + l, + projectList, + filters, + (a, f) => t.buildScan(a, f, inputPaths)) :: Nil + case l @ LogicalRelation(t: TableScan) => createPhysicalRDD(l.relation, l.output, t.buildScan()) :: Nil @@ -60,9 +110,144 @@ private[sql] object DataSourceStrategy extends Strategy { l @ LogicalRelation(t: InsertableRelation), part, query, overwrite, false) if part.isEmpty => execution.ExecutedCommand(InsertIntoDataSource(l, query, overwrite)) :: Nil + case i @ logical.InsertIntoTable( + l @ LogicalRelation(t: FSBasedRelation), part, query, overwrite, false) if part.isEmpty => + val mode = if (overwrite) SaveMode.Overwrite else SaveMode.Append + execution.ExecutedCommand( + InsertIntoFSBasedRelation(t, query, Array.empty[String], mode)) :: Nil + case _ => Nil } + private def buildPartitionedTableScan( + logicalRelation: LogicalRelation, + projections: Seq[NamedExpression], + filters: Seq[Expression], + partitionColumns: StructType, + partitions: Array[Partition]) = { + val output = projections.map(_.toAttribute) + val relation = logicalRelation.relation.asInstanceOf[FSBasedRelation] + + // Builds RDD[Row]s for each selected partition. + val perPartitionRows = partitions.map { case Partition(partitionValues, dir) => + // Paths to all data files within this partition + val dataFilePaths = { + val dirPath = new Path(dir) + val fs = dirPath.getFileSystem(SparkHadoopUtil.get.conf) + fs.listStatus(dirPath).map(_.getPath).filterNot { path => + val name = path.getName + name.startsWith("_") || name.startsWith(".") + }.map(fs.makeQualified(_).toString) + } + + // The table scan operator (PhysicalRDD) which retrieves required columns from data files. + // Notice that the schema of data files, represented by `relation.dataSchema`, may contain + // some partition column(s). + val scan = + pruneFilterProject( + logicalRelation, + projections, + filters, + (requiredColumns, filters) => { + val partitionColNames = partitionColumns.fieldNames + + // Don't scan any partition columns to save I/O. Here we are being optimistic and + // assuming partition columns data stored in data files are always consistent with those + // partition values encoded in partition directory paths. + val nonPartitionColumns = requiredColumns.filterNot(partitionColNames.contains) + val dataRows = relation.buildScan(nonPartitionColumns, filters, dataFilePaths) + + // Merges data values with partition values. + mergeWithPartitionValues( + relation.schema, + requiredColumns, + partitionColNames, + partitionValues, + dataRows) + }) + + scan.execute() + } + + val unionedRows = perPartitionRows.reduceOption(_ ++ _).getOrElse { + relation.sqlContext.emptyResult + } + + createPhysicalRDD(logicalRelation.relation, output, unionedRows) + } + + private def mergeWithPartitionValues( + schema: StructType, + requiredColumns: Array[String], + partitionColumns: Array[String], + partitionValues: Row, + dataRows: RDD[Row]): RDD[Row] = { + val nonPartitionColumns = requiredColumns.filterNot(partitionColumns.contains) + + // If output columns contain any partition column(s), we need to merge scanned data + // columns and requested partition columns to form the final result. + if (!requiredColumns.sameElements(nonPartitionColumns)) { + val mergers = requiredColumns.zipWithIndex.map { case (name, index) => + // To see whether the `index`-th column is a partition column... + val i = partitionColumns.indexOf(name) + if (i != -1) { + // If yes, gets column value from partition values. + (mutableRow: MutableRow, dataRow: expressions.Row, ordinal: Int) => { + mutableRow(ordinal) = partitionValues(i) + } + } else { + // Otherwise, inherits the value from scanned data. + val i = nonPartitionColumns.indexOf(name) + (mutableRow: MutableRow, dataRow: expressions.Row, ordinal: Int) => { + mutableRow(ordinal) = dataRow(i) + } + } + } + + dataRows.mapPartitions { iterator => + val dataTypes = requiredColumns.map(schema(_).dataType) + val mutableRow = new SpecificMutableRow(dataTypes) + iterator.map { dataRow => + var i = 0 + while (i < mutableRow.length) { + mergers(i)(mutableRow, dataRow, i) + i += 1 + } + mutableRow.asInstanceOf[expressions.Row] + } + } + } else { + dataRows + } + } + + protected def prunePartitions( + predicates: Seq[Expression], + partitionSpec: PartitionSpec): Seq[Partition] = { + val PartitionSpec(partitionColumns, partitions) = partitionSpec + val partitionColumnNames = partitionColumns.map(_.name).toSet + val partitionPruningPredicates = predicates.filter { + _.references.map(_.name).toSet.subsetOf(partitionColumnNames) + } + + if (partitionPruningPredicates.nonEmpty) { + val predicate = + partitionPruningPredicates + .reduceOption(expressions.And) + .getOrElse(Literal(true)) + + val boundPredicate = InterpretedPredicate.create(predicate.transform { + case a: AttributeReference => + val index = partitionColumns.indexWhere(a.name == _.name) + BoundReference(index, partitionColumns(index).dataType, nullable = true) + }) + + partitions.filter { case Partition(values, _) => boundPredicate(values) } + } else { + partitions + } + } + // Based on Public API. protected def pruneFilterProject( relation: LogicalRelation, diff --git a/sql/core/src/main/scala/org/apache/spark/sql/sources/PartitioningUtils.scala b/sql/core/src/main/scala/org/apache/spark/sql/sources/PartitioningUtils.scala new file mode 100644 index 0000000000000..d30f7f65e21c0 --- /dev/null +++ b/sql/core/src/main/scala/org/apache/spark/sql/sources/PartitioningUtils.scala @@ -0,0 +1,207 @@ +/* + * 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.sources + +import java.lang.{Double => JDouble, Float => JFloat, Long => JLong} +import java.math.{BigDecimal => JBigDecimal} + +import scala.collection.mutable.ArrayBuffer +import scala.util.Try + +import com.google.common.cache.{CacheBuilder, Cache} +import org.apache.hadoop.fs.{FileStatus, Path} + +import org.apache.spark.sql.Row +import org.apache.spark.sql.catalyst.expressions.{Cast, Literal} +import org.apache.spark.sql.types._ + +private[sql] case class Partition(values: Row, path: String) + +private[sql] case class PartitionSpec(partitionColumns: StructType, partitions: Seq[Partition]) + +private[sql] object PartitioningUtils { + private[sql] case class PartitionValues(columnNames: Seq[String], literals: Seq[Literal]) { + require(columnNames.size == literals.size) + } + + /** + * Given a group of qualified paths, tries to parse them and returns a partition specification. + * For example, given: + * {{{ + * hdfs://:/path/to/partition/a=1/b=hello/c=3.14 + * hdfs://:/path/to/partition/a=2/b=world/c=6.28 + * }}} + * it returns: + * {{{ + * PartitionSpec( + * partitionColumns = StructType( + * StructField(name = "a", dataType = IntegerType, nullable = true), + * StructField(name = "b", dataType = StringType, nullable = true), + * StructField(name = "c", dataType = DoubleType, nullable = true)), + * partitions = Seq( + * Partition( + * values = Row(1, "hello", 3.14), + * path = "hdfs://:/path/to/partition/a=1/b=hello/c=3.14"), + * Partition( + * values = Row(2, "world", 6.28), + * path = "hdfs://:/path/to/partition/a=2/b=world/c=6.28"))) + * }}} + */ + private[sql] def parsePartitions( + paths: Seq[Path], + defaultPartitionName: String): PartitionSpec = { + val partitionValues = resolvePartitions(paths.map(parsePartition(_, defaultPartitionName))) + val fields = { + val (PartitionValues(columnNames, literals)) = partitionValues.head + columnNames.zip(literals).map { case (name, Literal(_, dataType)) => + StructField(name, dataType, nullable = true) + } + } + + val partitions = partitionValues.zip(paths).map { + case (PartitionValues(_, literals), path) => + Partition(Row(literals.map(_.value): _*), path.toString) + } + + PartitionSpec(StructType(fields), partitions) + } + + /** + * Parses a single partition, returns column names and values of each partition column. For + * example, given: + * {{{ + * path = hdfs://:/path/to/partition/a=42/b=hello/c=3.14 + * }}} + * it returns: + * {{{ + * PartitionValues( + * Seq("a", "b", "c"), + * Seq( + * Literal.create(42, IntegerType), + * Literal.create("hello", StringType), + * Literal.create(3.14, FloatType))) + * }}} + */ + private[sql] def parsePartition( + path: Path, + defaultPartitionName: String): PartitionValues = { + val columns = ArrayBuffer.empty[(String, Literal)] + // Old Hadoop versions don't have `Path.isRoot` + var finished = path.getParent == null + var chopped = path + + while (!finished) { + val maybeColumn = parsePartitionColumn(chopped.getName, defaultPartitionName) + maybeColumn.foreach(columns += _) + chopped = chopped.getParent + finished = maybeColumn.isEmpty || chopped.getParent == null + } + + val (columnNames, values) = columns.reverse.unzip + PartitionValues(columnNames, values) + } + + private def parsePartitionColumn( + columnSpec: String, + defaultPartitionName: String): Option[(String, Literal)] = { + val equalSignIndex = columnSpec.indexOf('=') + if (equalSignIndex == -1) { + None + } else { + val columnName = columnSpec.take(equalSignIndex) + assert(columnName.nonEmpty, s"Empty partition column name in '$columnSpec'") + + val rawColumnValue = columnSpec.drop(equalSignIndex + 1) + assert(rawColumnValue.nonEmpty, s"Empty partition column value in '$columnSpec'") + + val literal = inferPartitionColumnValue(rawColumnValue, defaultPartitionName) + Some(columnName -> literal) + } + } + + /** + * Resolves possible type conflicts between partitions by up-casting "lower" types. The up- + * casting order is: + * {{{ + * NullType -> + * IntegerType -> LongType -> + * FloatType -> DoubleType -> DecimalType.Unlimited -> + * StringType + * }}} + */ + private[sql] def resolvePartitions(values: Seq[PartitionValues]): Seq[PartitionValues] = { + // Column names of all partitions must match + val distinctPartitionsColNames = values.map(_.columnNames).distinct + assert(distinctPartitionsColNames.size == 1, { + val list = distinctPartitionsColNames.mkString("\t", "\n", "") + s"Conflicting partition column names detected:\n$list" + }) + + // Resolves possible type conflicts for each column + val columnCount = values.head.columnNames.size + val resolvedValues = (0 until columnCount).map { i => + resolveTypeConflicts(values.map(_.literals(i))) + } + + // Fills resolved literals back to each partition + values.zipWithIndex.map { case (d, index) => + d.copy(literals = resolvedValues.map(_(index))) + } + } + + /** + * Converts a string to a `Literal` with automatic type inference. Currently only supports + * [[IntegerType]], [[LongType]], [[FloatType]], [[DoubleType]], [[DecimalType.Unlimited]], and + * [[StringType]]. + */ + private[sql] def inferPartitionColumnValue( + raw: String, + defaultPartitionName: String): Literal = { + // First tries integral types + Try(Literal.create(Integer.parseInt(raw), IntegerType)) + .orElse(Try(Literal.create(JLong.parseLong(raw), LongType))) + // Then falls back to fractional types + .orElse(Try(Literal.create(JFloat.parseFloat(raw), FloatType))) + .orElse(Try(Literal.create(JDouble.parseDouble(raw), DoubleType))) + .orElse(Try(Literal.create(new JBigDecimal(raw), DecimalType.Unlimited))) + // Then falls back to string + .getOrElse { + if (raw == defaultPartitionName) Literal.create(null, NullType) + else Literal.create(raw, StringType) + } + } + + private val upCastingOrder: Seq[DataType] = + Seq(NullType, IntegerType, LongType, FloatType, DoubleType, DecimalType.Unlimited, StringType) + + /** + * Given a collection of [[Literal]]s, resolves possible type conflicts by up-casting "lower" + * types. + */ + private def resolveTypeConflicts(literals: Seq[Literal]): Seq[Literal] = { + val desiredType = { + val topType = literals.map(_.dataType).maxBy(upCastingOrder.indexOf(_)) + // Falls back to string if all values of this column are null or empty string + if (topType == NullType) StringType else topType + } + + literals.map { case l @ Literal(_, dataType) => + Literal.create(Cast(l, desiredType).eval(), desiredType) + } + } +} 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 dbdb0d39c26a1..127133bfaf662 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 @@ -14,12 +14,28 @@ * See the License for the specific language governing permissions and * limitations under the License. */ + package org.apache.spark.sql.sources -import org.apache.spark.sql.{DataFrame, SQLContext} +import java.util.Date + +import scala.collection.mutable + +import org.apache.hadoop.fs.Path +import org.apache.hadoop.mapreduce._ +import org.apache.hadoop.mapreduce.lib.output.{FileOutputCommitter, FileOutputFormat} +import org.apache.hadoop.util.Shell +import parquet.hadoop.util.ContextUtil + +import org.apache.spark._ +import org.apache.spark.mapred.SparkHadoopMapRedUtil +import org.apache.spark.mapreduce.SparkHadoopMapReduceUtil +import org.apache.spark.sql.catalyst.CatalystTypeConverters import org.apache.spark.sql.catalyst.expressions._ +import org.apache.spark.sql.catalyst.expressions.codegen.GenerateProjection import org.apache.spark.sql.catalyst.plans.logical.LogicalPlan import org.apache.spark.sql.execution.RunnableCommand +import org.apache.spark.sql.{DataFrame, SQLContext, SaveMode} private[sql] case class InsertIntoDataSource( logicalRelation: LogicalRelation, @@ -41,3 +57,391 @@ private[sql] case class InsertIntoDataSource( Seq.empty[Row] } } + +private[sql] case class InsertIntoFSBasedRelation( + @transient relation: FSBasedRelation, + @transient query: LogicalPlan, + partitionColumns: Array[String], + mode: SaveMode) + extends RunnableCommand { + + override def run(sqlContext: SQLContext): Seq[Row] = { + require( + relation.paths.length == 1, + s"Cannot write to multiple destinations: ${relation.paths.mkString(",")}") + + val hadoopConf = sqlContext.sparkContext.hadoopConfiguration + val outputPath = new Path(relation.paths.head) + val fs = outputPath.getFileSystem(hadoopConf) + val qualifiedOutputPath = outputPath.makeQualified(fs.getUri, fs.getWorkingDirectory) + + val doInsertion = (mode, fs.exists(qualifiedOutputPath)) match { + case (SaveMode.ErrorIfExists, true) => + sys.error(s"path $qualifiedOutputPath already exists.") + case (SaveMode.Overwrite, true) => + fs.delete(qualifiedOutputPath, true) + true + case (SaveMode.Append, _) | (SaveMode.Overwrite, _) | (SaveMode.ErrorIfExists, false) => + true + case (SaveMode.Ignore, exists) => + !exists + } + + if (doInsertion) { + val job = Job.getInstance(hadoopConf) + job.setOutputKeyClass(classOf[Void]) + job.setOutputValueClass(classOf[Row]) + FileOutputFormat.setOutputPath(job, qualifiedOutputPath) + + val df = sqlContext.createDataFrame( + DataFrame(sqlContext, query).queryExecution.toRdd, + relation.schema, + needsConversion = false) + + if (partitionColumns.isEmpty) { + insert(new DefaultWriterContainer(relation, job), df) + } else { + val writerContainer = new DynamicPartitionWriterContainer( + relation, job, partitionColumns, "__HIVE_DEFAULT_PARTITION__") + insertWithDynamicPartitions(sqlContext, writerContainer, df, partitionColumns) + } + } + + Seq.empty[Row] + } + + private def insert(writerContainer: BaseWriterContainer, df: DataFrame): Unit = { + // Uses local vals for serialization + val needsConversion = relation.needConversion + val dataSchema = relation.dataSchema + + try { + writerContainer.driverSideSetup() + df.sqlContext.sparkContext.runJob(df.queryExecution.executedPlan.execute(), writeRows _) + writerContainer.commitJob() + relation.refresh() + } catch { case cause: Throwable => + writerContainer.abortJob() + throw new SparkException("Job aborted.", cause) + } + + def writeRows(taskContext: TaskContext, iterator: Iterator[Row]): Unit = { + writerContainer.executorSideSetup(taskContext) + + try { + if (needsConversion) { + val converter = CatalystTypeConverters.createToScalaConverter(dataSchema) + while (iterator.hasNext) { + val row = converter(iterator.next()).asInstanceOf[Row] + writerContainer.outputWriterForRow(row).write(row) + } + } else { + while (iterator.hasNext) { + val row = iterator.next() + writerContainer.outputWriterForRow(row).write(row) + } + } + writerContainer.commitTask() + } catch { case cause: Throwable => + writerContainer.abortTask() + throw new SparkException("Task failed while writing rows.", cause) + } + } + } + + private def insertWithDynamicPartitions( + sqlContext: SQLContext, + writerContainer: BaseWriterContainer, + df: DataFrame, + partitionColumns: Array[String]): Unit = { + // Uses a local val for serialization + val needsConversion = relation.needConversion + val dataSchema = relation.dataSchema + + require( + df.schema == relation.schema, + s"""DataFrame must have the same schema as the relation to which is inserted. + |DataFrame schema: ${df.schema} + |Relation schema: ${relation.schema} + """.stripMargin) + + val partitionColumnsInSpec = relation.partitionColumns.fieldNames + require( + partitionColumnsInSpec.sameElements(partitionColumns), + s"""Partition columns mismatch. + |Expected: ${partitionColumnsInSpec.mkString(", ")} + |Actual: ${partitionColumns.mkString(", ")} + """.stripMargin) + + val output = df.queryExecution.executedPlan.output + val (partitionOutput, dataOutput) = output.partition(a => partitionColumns.contains(a.name)) + val codegenEnabled = df.sqlContext.conf.codegenEnabled + + try { + writerContainer.driverSideSetup() + df.sqlContext.sparkContext.runJob(df.queryExecution.executedPlan.execute(), writeRows _) + writerContainer.commitJob() + relation.refresh() + } catch { case cause: Throwable => + logError("Aborting job.", cause) + writerContainer.abortJob() + throw new SparkException("Job aborted.", cause) + } + + def writeRows(taskContext: TaskContext, iterator: Iterator[Row]): Unit = { + writerContainer.executorSideSetup(taskContext) + + val partitionProj = newProjection(codegenEnabled, partitionOutput, output) + val dataProj = newProjection(codegenEnabled, dataOutput, output) + + if (needsConversion) { + val converter = CatalystTypeConverters.createToScalaConverter(dataSchema) + while (iterator.hasNext) { + val row = iterator.next() + val partitionPart = partitionProj(row) + val dataPart = dataProj(row) + val convertedDataPart = converter(dataPart).asInstanceOf[Row] + writerContainer.outputWriterForRow(partitionPart).write(convertedDataPart) + } + } else { + while (iterator.hasNext) { + val row = iterator.next() + val partitionPart = partitionProj(row) + val dataPart = dataProj(row) + writerContainer.outputWriterForRow(partitionPart).write(dataPart) + } + } + + writerContainer.commitTask() + } + } + + // This is copied from SparkPlan, probably should move this to a more general place. + private def newProjection( + codegenEnabled: Boolean, + expressions: Seq[Expression], + inputSchema: Seq[Attribute]): Projection = { + log.debug( + s"Creating Projection: $expressions, inputSchema: $inputSchema, codegen:$codegenEnabled") + if (codegenEnabled) { + GenerateProjection.generate(expressions, inputSchema) + } else { + new InterpretedProjection(expressions, inputSchema) + } + } +} + +private[sql] abstract class BaseWriterContainer( + @transient val relation: FSBasedRelation, + @transient job: Job) + extends SparkHadoopMapReduceUtil + with Logging + with Serializable { + + protected val serializableConf = new SerializableWritable(ContextUtil.getConfiguration(job)) + + // This is only used on driver side. + @transient private val jobContext: JobContext = job + + // The following fields are initialized and used on both driver and executor side. + @transient protected var outputCommitter: FileOutputCommitter = _ + @transient private var jobId: JobID = _ + @transient private var taskId: TaskID = _ + @transient private var taskAttemptId: TaskAttemptID = _ + @transient protected var taskAttemptContext: TaskAttemptContext = _ + + protected val outputPath: String = { + assert( + relation.paths.length == 1, + s"Cannot write to multiple destinations: ${relation.paths.mkString(",")}") + relation.paths.head + } + + protected val dataSchema = relation.dataSchema + + protected val outputWriterClass: Class[_ <: OutputWriter] = relation.outputWriterClass + + private var outputFormatClass: Class[_ <: OutputFormat[_, _]] = _ + + def driverSideSetup(): Unit = { + setupIDs(0, 0, 0) + setupConf() + taskAttemptContext = newTaskAttemptContext(serializableConf.value, taskAttemptId) + relation.prepareForWrite(job) + outputFormatClass = job.getOutputFormatClass + outputCommitter = newOutputCommitter(taskAttemptContext) + outputCommitter.setupJob(jobContext) + } + + def executorSideSetup(taskContext: TaskContext): Unit = { + setupIDs(taskContext.stageId(), taskContext.partitionId(), taskContext.attemptNumber()) + setupConf() + taskAttemptContext = newTaskAttemptContext(serializableConf.value, taskAttemptId) + outputCommitter = newOutputCommitter(taskAttemptContext) + outputCommitter.setupTask(taskAttemptContext) + initWriters() + } + + private def newOutputCommitter(context: TaskAttemptContext): FileOutputCommitter = { + outputFormatClass.newInstance().getOutputCommitter(context) match { + case f: FileOutputCommitter => f + case f => sys.error( + s"FileOutputCommitter or its subclass is expected, but got a ${f.getClass.getName}.") + } + } + + private def setupIDs(jobId: Int, splitId: Int, attemptId: Int): Unit = { + this.jobId = SparkHadoopWriter.createJobID(new Date, jobId) + this.taskId = new TaskID(this.jobId, true, splitId) + this.taskAttemptId = new TaskAttemptID(taskId, attemptId) + } + + private def setupConf(): Unit = { + serializableConf.value.set("mapred.job.id", jobId.toString) + serializableConf.value.set("mapred.tip.id", taskAttemptId.getTaskID.toString) + serializableConf.value.set("mapred.task.id", taskAttemptId.toString) + serializableConf.value.setBoolean("mapred.task.is.map", true) + serializableConf.value.setInt("mapred.task.partition", 0) + } + + // Called on executor side when writing rows + def outputWriterForRow(row: Row): OutputWriter + + protected def initWriters(): Unit + + def commitTask(): Unit = { + SparkHadoopMapRedUtil.commitTask( + outputCommitter, taskAttemptContext, jobId.getId, taskId.getId, taskAttemptId.getId) + } + + def abortTask(): Unit = { + outputCommitter.abortTask(taskAttemptContext) + logError(s"Task attempt $taskAttemptId aborted.") + } + + def commitJob(): Unit = { + outputCommitter.commitJob(jobContext) + logInfo(s"Job $jobId committed.") + } + + def abortJob(): Unit = { + outputCommitter.abortJob(jobContext, JobStatus.State.FAILED) + logError(s"Job $jobId aborted.") + } +} + +private[sql] class DefaultWriterContainer( + @transient relation: FSBasedRelation, + @transient job: Job) + extends BaseWriterContainer(relation, job) { + + @transient private var writer: OutputWriter = _ + + override protected def initWriters(): Unit = { + writer = outputWriterClass.newInstance() + writer.init(outputCommitter.getWorkPath.toString, dataSchema, taskAttemptContext) + } + + override def outputWriterForRow(row: Row): OutputWriter = writer + + override def commitTask(): Unit = { + writer.close() + super.commitTask() + } + + override def abortTask(): Unit = { + writer.close() + super.abortTask() + } +} + +private[sql] class DynamicPartitionWriterContainer( + @transient relation: FSBasedRelation, + @transient job: Job, + partitionColumns: Array[String], + defaultPartitionName: String) + extends BaseWriterContainer(relation, job) { + + // All output writers are created on executor side. + @transient protected var outputWriters: mutable.Map[String, OutputWriter] = _ + + override protected def initWriters(): Unit = { + outputWriters = mutable.Map.empty[String, OutputWriter] + } + + override def outputWriterForRow(row: Row): OutputWriter = { + val partitionPath = partitionColumns.zip(row.toSeq).map { case (col, rawValue) => + val string = if (rawValue == null) null else String.valueOf(rawValue) + val valueString = if (string == null || string.isEmpty) { + defaultPartitionName + } else { + DynamicPartitionWriterContainer.escapePathName(string) + } + s"/$col=$valueString" + }.mkString + + outputWriters.getOrElseUpdate(partitionPath, { + val path = new Path(outputCommitter.getWorkPath, partitionPath.stripPrefix(Path.SEPARATOR)) + val writer = outputWriterClass.newInstance() + writer.init(path.toString, dataSchema, taskAttemptContext) + writer + }) + } + + override def commitTask(): Unit = { + outputWriters.values.foreach(_.close()) + super.commitTask() + } + + override def abortTask(): Unit = { + outputWriters.values.foreach(_.close()) + super.abortTask() + } +} + +private[sql] object DynamicPartitionWriterContainer { + ////////////////////////////////////////////////////////////////////////////////////////////////// + // The following string escaping code is mainly copied from Hive (o.a.h.h.common.FileUtils). + ////////////////////////////////////////////////////////////////////////////////////////////////// + + val charToEscape = { + val bitSet = new java.util.BitSet(128) + + /** + * ASCII 01-1F are HTTP control characters that need to be escaped. + * \u000A and \u000D are \n and \r, respectively. + */ + val clist = Array( + '\u0001', '\u0002', '\u0003', '\u0004', '\u0005', '\u0006', '\u0007', '\u0008', '\u0009', + '\n', '\u000B', '\u000C', '\r', '\u000E', '\u000F', '\u0010', '\u0011', '\u0012', '\u0013', + '\u0014', '\u0015', '\u0016', '\u0017', '\u0018', '\u0019', '\u001A', '\u001B', '\u001C', + '\u001D', '\u001E', '\u001F', '"', '#', '%', '\'', '*', '/', ':', '=', '?', '\\', '\u007F', + '{', '[', ']', '^') + + clist.foreach(bitSet.set(_)) + + if (Shell.WINDOWS) { + Array(' ', '<', '>', '|').foreach(bitSet.set(_)) + } + + bitSet + } + + def needsEscaping(c: Char): Boolean = { + c >= 0 && c < charToEscape.size() && charToEscape.get(c) + } + + def escapePathName(path: String): String = { + val builder = new StringBuilder() + path.foreach { c => + if (DynamicPartitionWriterContainer.needsEscaping(c)) { + builder.append('%') + builder.append(f"${c.asInstanceOf[Int]}%02x") + } else { + builder.append(c) + } + } + + builder.toString() + } +} 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 06c64f2bdd59e..595c5eb40e295 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 @@ -17,18 +17,20 @@ package org.apache.spark.sql.sources -import scala.language.existentials +import scala.language.{existentials, implicitConversions} import scala.util.matching.Regex -import scala.language.implicitConversions + +import org.apache.hadoop.fs.Path import org.apache.spark.Logging -import org.apache.spark.sql.{AnalysisException, SaveMode, DataFrame, SQLContext} -import org.apache.spark.sql.catalyst.plans.logical._ +import org.apache.spark.deploy.SparkHadoopUtil import org.apache.spark.sql.catalyst.AbstractSparkSQLParser import org.apache.spark.sql.catalyst.analysis.UnresolvedRelation import org.apache.spark.sql.catalyst.expressions.{Attribute, AttributeReference, Row} +import org.apache.spark.sql.catalyst.plans.logical._ import org.apache.spark.sql.execution.RunnableCommand import org.apache.spark.sql.types._ +import org.apache.spark.sql.{AnalysisException, DataFrame, SQLContext, SaveMode} import org.apache.spark.util.Utils /** @@ -111,6 +113,7 @@ private[sql] class DDLParser( CreateTableUsingAsSelect(tableName, provider, temp.isDefined, + Array.empty[String], mode, options, queryPlan) @@ -157,7 +160,7 @@ private[sql] class DDLParser( protected lazy val className: Parser[String] = repsep(ident, ".") ^^ { case s => s.mkString(".")} override implicit def regexToParser(regex: Regex): Parser[String] = acceptMatch( - s"identifier matching regex ${regex}", { + s"identifier matching regex $regex", { case lexical.Identifier(str) if regex.unapplySeq(str).isDefined => str case lexical.Keyword(str) if regex.unapplySeq(str).isDefined => str } @@ -214,6 +217,7 @@ private[sql] object ResolvedDataSource { def apply( sqlContext: SQLContext, userSpecifiedSchema: Option[StructType], + partitionColumns: Array[String], provider: String, options: Map[String, String]): ResolvedDataSource = { val clazz: Class[_] = lookupDataSource(provider) @@ -222,6 +226,27 @@ private[sql] object ResolvedDataSource { case Some(schema: StructType) => clazz.newInstance() match { case dataSource: SchemaRelationProvider => dataSource.createRelation(sqlContext, new CaseInsensitiveMap(options), schema) + case dataSource: FSBasedRelationProvider => + val maybePartitionsSchema = if (partitionColumns.isEmpty) { + None + } else { + Some(partitionColumnsSchema(schema, partitionColumns)) + } + + val caseInsensitiveOptions= new CaseInsensitiveMap(options) + val paths = { + val patternPath = new Path(caseInsensitiveOptions("path")) + SparkHadoopUtil.get.globPath(patternPath).map(_.toString).toArray + } + + val dataSchema = StructType(schema.filterNot(f => partitionColumns.contains(f.name))) + + dataSource.createRelation( + sqlContext, + paths, + Some(schema), + maybePartitionsSchema, + caseInsensitiveOptions) case dataSource: org.apache.spark.sql.sources.RelationProvider => throw new AnalysisException(s"$className does not allow user-specified schemas.") case _ => @@ -231,20 +256,39 @@ private[sql] object ResolvedDataSource { case None => clazz.newInstance() match { case dataSource: RelationProvider => dataSource.createRelation(sqlContext, new CaseInsensitiveMap(options)) + case dataSource: FSBasedRelationProvider => + val caseInsensitiveOptions = new CaseInsensitiveMap(options) + val paths = { + val patternPath = new Path(caseInsensitiveOptions("path")) + SparkHadoopUtil.get.globPath(patternPath).map(_.toString).toArray + } + dataSource.createRelation(sqlContext, paths, None, None, caseInsensitiveOptions) case dataSource: org.apache.spark.sql.sources.SchemaRelationProvider => throw new AnalysisException( s"A schema needs to be specified when using $className.") case _ => - throw new AnalysisException(s"$className is not a RelationProvider.") + throw new AnalysisException( + s"$className is neither a RelationProvider nor a FSBasedRelationProvider.") } } new ResolvedDataSource(clazz, relation) } + private def partitionColumnsSchema( + schema: StructType, + partitionColumns: Array[String]): StructType = { + StructType(partitionColumns.map { col => + schema.find(_.name == col).getOrElse { + throw new RuntimeException(s"Partition column $col not found in schema $schema") + } + }).asNullable + } + /** Create a [[ResolvedDataSource]] for saving the content of the given [[DataFrame]]. */ def apply( sqlContext: SQLContext, provider: String, + partitionColumns: Array[String], mode: SaveMode, options: Map[String, String], data: DataFrame): ResolvedDataSource = { @@ -252,6 +296,31 @@ private[sql] object ResolvedDataSource { val relation = clazz.newInstance() match { case dataSource: CreatableRelationProvider => dataSource.createRelation(sqlContext, mode, options, data) + case dataSource: FSBasedRelationProvider => + // Don't glob path for the write path. The contracts here are: + // 1. Only one output path can be specified on the write path; + // 2. Output path must be a legal HDFS style file system path; + // 3. It's OK that the output path doesn't exist yet; + val caseInsensitiveOptions = new CaseInsensitiveMap(options) + val outputPath = { + val path = new Path(caseInsensitiveOptions("path")) + val fs = path.getFileSystem(sqlContext.sparkContext.hadoopConfiguration) + path.makeQualified(fs.getUri, fs.getWorkingDirectory) + } + val dataSchema = StructType(data.schema.filterNot(f => partitionColumns.contains(f.name))) + val r = dataSource.createRelation( + sqlContext, + Array(outputPath.toString), + Some(dataSchema.asNullable), + Some(partitionColumnsSchema(data.schema, partitionColumns)), + caseInsensitiveOptions) + sqlContext.executePlan( + InsertIntoFSBasedRelation( + r, + data.logicalPlan, + partitionColumns.toArray, + mode)).toRdd + r case _ => sys.error(s"${clazz.getCanonicalName} does not allow create table as select.") } @@ -310,6 +379,7 @@ private[sql] case class CreateTableUsingAsSelect( tableName: String, provider: String, temporary: Boolean, + partitionColumns: Array[String], mode: SaveMode, options: Map[String, String], child: LogicalPlan) extends UnaryNode { @@ -324,8 +394,9 @@ private[sql] case class CreateTempTableUsing( provider: String, options: Map[String, String]) extends RunnableCommand { - override def run(sqlContext: SQLContext): Seq[Row] = { - val resolved = ResolvedDataSource(sqlContext, userSpecifiedSchema, provider, options) + def run(sqlContext: SQLContext): Seq[Row] = { + val resolved = ResolvedDataSource( + sqlContext, userSpecifiedSchema, Array.empty[String], provider, options) sqlContext.registerDataFrameAsTable( DataFrame(sqlContext, LogicalRelation(resolved.relation)), tableName) Seq.empty @@ -335,13 +406,14 @@ private[sql] case class CreateTempTableUsing( private[sql] case class CreateTempTableUsingAsSelect( tableName: String, provider: String, + partitionColumns: Array[String], mode: SaveMode, options: Map[String, String], query: LogicalPlan) extends RunnableCommand { override def run(sqlContext: SQLContext): Seq[Row] = { val df = DataFrame(sqlContext, query) - val resolved = ResolvedDataSource(sqlContext, provider, mode, options, df) + val resolved = ResolvedDataSource(sqlContext, provider, partitionColumns, mode, options, df) sqlContext.registerDataFrameAsTable( DataFrame(sqlContext, LogicalRelation(resolved.relation)), tableName) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/sources/interfaces.scala b/sql/core/src/main/scala/org/apache/spark/sql/sources/interfaces.scala index ca53dcdb92c52..5e010d21120f6 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/sources/interfaces.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/sources/interfaces.scala @@ -17,11 +17,19 @@ package org.apache.spark.sql.sources -import org.apache.spark.annotation.{Experimental, DeveloperApi} +import scala.util.Try + +import org.apache.hadoop.conf.Configuration +import org.apache.hadoop.fs.{FileStatus, Path} +import org.apache.hadoop.mapreduce.{Job, TaskAttemptContext} + +import org.apache.spark.annotation.{DeveloperApi, Experimental} +import org.apache.spark.deploy.SparkHadoopUtil import org.apache.spark.rdd.RDD -import org.apache.spark.sql.{SaveMode, DataFrame, Row, SQLContext} -import org.apache.spark.sql.catalyst.expressions.{Expression, Attribute} -import org.apache.spark.sql.types.StructType +import org.apache.spark.sql.{Row, _} +import org.apache.spark.sql.catalyst.expressions._ +import org.apache.spark.sql.catalyst.expressions.codegen.GenerateMutableProjection +import org.apache.spark.sql.types.{StructField, StructType} /** * ::DeveloperApi:: @@ -78,6 +86,41 @@ trait SchemaRelationProvider { schema: StructType): BaseRelation } +/** + * ::DeveloperApi:: + * Implemented by objects that produce relations for a specific kind of data source + * with a given schema and partitioned columns. When Spark SQL is given a DDL operation with a + * USING clause specified (to specify the implemented [[FSBasedRelationProvider]]), a user defined + * schema, and an optional list of partition columns, this interface is used to pass in the + * parameters specified by a user. + * + * Users may specify the fully qualified class name of a given data source. When that class is + * not found Spark SQL will append the class name `DefaultSource` to the path, allowing for + * less verbose invocation. For example, 'org.apache.spark.sql.json' would resolve to the + * data source 'org.apache.spark.sql.json.DefaultSource' + * + * A new instance of this class with be instantiated each time a DDL call is made. + * + * The difference between a [[RelationProvider]] and a [[FSBasedRelationProvider]] is + * that users need to provide a schema and a (possibly empty) list of partition columns when + * using a SchemaRelationProvider. A relation provider can inherits both [[RelationProvider]], + * and [[FSBasedRelationProvider]] if it can support schema inference, user-specified + * schemas, and accessing partitioned relations. + */ +trait FSBasedRelationProvider { + /** + * Returns a new base relation with the given parameters, a user defined schema, and a list of + * partition columns. Note: the parameters' keywords are case insensitive and this insensitivity + * is enforced by the Map that is passed to the function. + */ + def createRelation( + sqlContext: SQLContext, + paths: Array[String], + schema: Option[StructType], + partitionColumns: Option[StructType], + parameters: Map[String, String]): FSBasedRelation +} + @DeveloperApi trait CreatableRelationProvider { /** @@ -207,3 +250,235 @@ trait InsertableRelation { trait CatalystScan { def buildScan(requiredColumns: Seq[Attribute], filters: Seq[Expression]): RDD[Row] } + +/** + * ::Experimental:: + * [[OutputWriter]] is used together with [[FSBasedRelation]] for persisting rows to the + * underlying file system. Subclasses of [[OutputWriter]] must provide a zero-argument constructor. + * An [[OutputWriter]] instance is created and initialized when a new output file is opened on + * executor side. This instance is used to persist rows to this single output file. + */ +@Experimental +abstract class OutputWriter { + /** + * Initializes this [[OutputWriter]] before any rows are persisted. + * + * @param path Path of the file to which this [[OutputWriter]] is supposed to write. Note that + * this may not point to the final output file. For example, `FileOutputFormat` writes to + * temporary directories and then merge written files back to the final destination. In + * this case, `path` points to a temporary output file under the temporary directory. + * @param dataSchema Schema of the rows to be written. Partition columns are not included in the + * schema if the corresponding relation is partitioned. + * @param context The Hadoop MapReduce task context. + */ + def init( + path: String, + dataSchema: StructType, + context: TaskAttemptContext): Unit = () + + /** + * Persists a single row. Invoked on the executor side. When writing to dynamically partitioned + * tables, dynamic partition columns are not included in rows to be written. + */ + def write(row: Row): Unit + + /** + * Closes the [[OutputWriter]]. Invoked on the executor side after all rows are persisted, before + * the task output is committed. + */ + def close(): Unit +} + +/** + * ::Experimental:: + * A [[BaseRelation]] that provides much of the common code required for formats that store their + * data to an HDFS compatible filesystem. + * + * For the read path, similar to [[PrunedFilteredScan]], it can eliminate unneeded columns and + * filter using selected predicates before producing an RDD containing all matching tuples as + * [[Row]] objects. In addition, when reading from Hive style partitioned tables stored in file + * systems, it's able to discover partitioning information from the paths of input directories, and + * perform partition pruning before start reading the data. Subclasses of [[FSBasedRelation()]] must + * override one of the three `buildScan` methods to implement the read path. + * + * For the write path, it provides the ability to write to both non-partitioned and partitioned + * tables. Directory layout of the partitioned tables is compatible with Hive. + * + * @constructor This constructor is for internal uses only. The [[PartitionSpec]] argument is for + * implementing metastore table conversion. + * @param paths Base paths of this relation. For partitioned relations, it should be the root + * directories of all partition directories. + * @param maybePartitionSpec An [[FSBasedRelation]] can be created with an optional + * [[PartitionSpec]], so that partition discovery can be skipped. + */ +@Experimental +abstract class FSBasedRelation private[sql]( + val paths: Array[String], + maybePartitionSpec: Option[PartitionSpec]) + extends BaseRelation { + + /** + * Constructs an [[FSBasedRelation]]. + * + * @param paths Base paths of this relation. For partitioned relations, it should be either root + * directories of all partition directories. + * @param partitionColumns Partition columns of this relation. + */ + def this(paths: Array[String], partitionColumns: StructType) = + this(paths, { + if (partitionColumns.isEmpty) None + else Some(PartitionSpec(partitionColumns, Array.empty[Partition])) + }) + + /** + * Constructs an [[FSBasedRelation]]. + * + * @param paths Base paths of this relation. For partitioned relations, it should be root + * directories of all partition directories. + */ + def this(paths: Array[String]) = this(paths, None) + + private val hadoopConf = new Configuration(sqlContext.sparkContext.hadoopConfiguration) + + private val codegenEnabled = sqlContext.conf.codegenEnabled + + private var _partitionSpec: PartitionSpec = maybePartitionSpec.map { spec => + spec.copy(partitionColumns = spec.partitionColumns.asNullable) + }.getOrElse { + if (sqlContext.conf.partitionDiscoveryEnabled()) { + discoverPartitions() + } else { + PartitionSpec(StructType(Nil), Array.empty[Partition]) + } + } + + private[sql] def partitionSpec: PartitionSpec = _partitionSpec + + /** + * Partition columns. Note that they are always nullable. + */ + def partitionColumns: StructType = partitionSpec.partitionColumns + + private[sql] def refresh(): Unit = { + if (sqlContext.conf.partitionDiscoveryEnabled()) { + _partitionSpec = discoverPartitions() + } + } + + private def discoverPartitions(): PartitionSpec = { + val basePaths = paths.map(new Path(_)) + val leafDirs = basePaths.flatMap { path => + val fs = path.getFileSystem(hadoopConf) + Try(fs.getFileStatus(path.makeQualified(fs.getUri, fs.getWorkingDirectory))) + .filter(_.isDir) + .map(SparkHadoopUtil.get.listLeafDirStatuses(fs, _)) + .getOrElse(Seq.empty[FileStatus]) + }.map(_.getPath) + + if (leafDirs.nonEmpty) { + PartitioningUtils.parsePartitions(leafDirs, "__HIVE_DEFAULT_PARTITION__") + } else { + PartitionSpec(StructType(Array.empty[StructField]), Array.empty[Partition]) + } + } + + /** + * Schema of this relation. It consists of columns appearing in [[dataSchema]] and all partition + * columns not appearing in [[dataSchema]]. + */ + override lazy val schema: StructType = { + val dataSchemaColumnNames = dataSchema.map(_.name.toLowerCase).toSet + StructType(dataSchema ++ partitionSpec.partitionColumns.filterNot { column => + dataSchemaColumnNames.contains(column.name.toLowerCase) + }) + } + + /** + * Specifies schema of actual data files. For partitioned relations, if one or more partitioned + * columns are contained in the data files, they should also appear in `dataSchema`. + */ + def dataSchema: StructType + + /** + * For a non-partitioned relation, this method builds an `RDD[Row]` containing all rows within + * this relation. For partitioned relations, this method is called for each selected partition, + * and builds an `RDD[Row]` containing all rows within that single partition. + * + * @param inputPaths For a non-partitioned relation, it contains paths of all data files in the + * relation. For a partitioned relation, it contains paths of all data files in a single + * selected partition. + */ + def buildScan(inputPaths: Array[String]): RDD[Row] = { + throw new RuntimeException( + "At least one buildScan() method should be overridden to read the relation.") + } + + /** + * For a non-partitioned relation, this method builds an `RDD[Row]` containing all rows within + * this relation. For partitioned relations, this method is called for each selected partition, + * and builds an `RDD[Row]` containing all rows within that single partition. + * + * @param requiredColumns Required columns. + * @param inputPaths For a non-partitioned relation, it contains paths of all data files in the + * relation. For a partitioned relation, it contains paths of all data files in a single + * selected partition. + */ + def buildScan(requiredColumns: Array[String], inputPaths: Array[String]): RDD[Row] = { + // Yeah, to workaround serialization... + val dataSchema = this.dataSchema + val codegenEnabled = this.codegenEnabled + + val requiredOutput = requiredColumns.map { col => + val field = dataSchema(col) + BoundReference(dataSchema.fieldIndex(col), field.dataType, field.nullable) + }.toSeq + + buildScan(inputPaths).mapPartitions { rows => + val buildProjection = if (codegenEnabled) { + GenerateMutableProjection.generate(requiredOutput, dataSchema.toAttributes) + } else { + () => new InterpretedMutableProjection(requiredOutput, dataSchema.toAttributes) + } + + val mutableProjection = buildProjection() + rows.map(mutableProjection) + } + } + + /** + * For a non-partitioned relation, this method builds an `RDD[Row]` containing all rows within + * this relation. For partitioned relations, this method is called for each selected partition, + * and builds an `RDD[Row]` containing all rows within that single partition. + * + * @param requiredColumns Required columns. + * @param filters Candidate filters to be pushed down. The actual filter should be the conjunction + * of all `filters`. The pushed down filters are currently purely an optimization as they + * will all be evaluated again. This means it is safe to use them with methods that produce + * false positives such as filtering partitions based on a bloom filter. + * @param inputPaths For a non-partitioned relation, it contains paths of all data files in the + * relation. For a partitioned relation, it contains paths of all data files in a single + * selected partition. + */ + def buildScan( + requiredColumns: Array[String], + filters: Array[Filter], + inputPaths: Array[String]): RDD[Row] = { + buildScan(requiredColumns, inputPaths) + } + + /** + * Client side preparation for data writing can be put here. For example, user defined output + * committer can be configured here. + * + * Note that the only side effect expected here is mutating `job` via its setters. Especially, + * Spark SQL caches [[BaseRelation]] instances for performance, mutating relation internal states + * may cause unexpected behaviors. + */ + def prepareForWrite(job: Job): Unit = () + + /** + * This method is responsible for producing a new [[OutputWriter]] for each newly opened output + * file on the executor side. + */ + def outputWriterClass: Class[_ <: OutputWriter] +} 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 6ed68d179edc9..aad1d248d0a28 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 @@ -101,13 +101,13 @@ private[sql] case class PreWriteCheck(catalog: Catalog) extends (LogicalPlan => } } - case i @ logical.InsertIntoTable( - l: LogicalRelation, partition, query, overwrite, ifNotExists) - if !l.isInstanceOf[InsertableRelation] => + case logical.InsertIntoTable(LogicalRelation(_: InsertableRelation), _, _, _, _) => // OK + case logical.InsertIntoTable(LogicalRelation(_: FSBasedRelation), _, _, _, _) => // OK + case logical.InsertIntoTable(l: LogicalRelation, _, _, _, _) => // The relation in l is not an InsertableRelation. failAnalysis(s"$l does not allow insertion.") - case CreateTableUsingAsSelect(tableName, _, _, SaveMode.Overwrite, _, query) => + case CreateTableUsingAsSelect(tableName, _, _, _, SaveMode.Overwrite, _, query) => // When the SaveMode is Overwrite, we need to check if the table is an input table of // the query. If so, we will throw an AnalysisException to let users know it is not allowed. if (catalog.tableExists(Seq(tableName))) { diff --git a/sql/core/src/test/scala/org/apache/spark/sql/parquet/ParquetPartitionDiscoverySuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/parquet/ParquetPartitionDiscoverySuite.scala index b7561ce7298cb..bea568ed40049 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/parquet/ParquetPartitionDiscoverySuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/parquet/ParquetPartitionDiscoverySuite.scala @@ -21,7 +21,8 @@ import scala.collection.mutable.ArrayBuffer import org.apache.hadoop.fs.Path import org.apache.spark.sql.catalyst.expressions.Literal -import org.apache.spark.sql.parquet.ParquetRelation2._ +import org.apache.spark.sql.sources.PartitioningUtils._ +import org.apache.spark.sql.sources.{Partition, PartitionSpec} import org.apache.spark.sql.test.TestSQLContext import org.apache.spark.sql.types._ import org.apache.spark.sql.{QueryTest, Row, SQLContext} 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 54f2f3cdec298..4e54b2eb8df7a 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 @@ -17,11 +17,11 @@ package org.apache.spark.sql.sources -import java.io.{IOException, File} +import java.io.{File, IOException} -import org.apache.spark.sql.AnalysisException import org.scalatest.BeforeAndAfterAll +import org.apache.spark.sql.AnalysisException import org.apache.spark.util.Utils class CreateTableAsSelectSuite extends DataSourceTest with BeforeAndAfterAll { diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveMetastoreCatalog.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveMetastoreCatalog.scala index bbf48efb24440..d754c8e3a8aa1 100644 --- a/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveMetastoreCatalog.scala +++ b/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveMetastoreCatalog.scala @@ -19,25 +19,24 @@ package org.apache.spark.sql.hive import com.google.common.base.Objects import com.google.common.cache.{CacheBuilder, CacheLoader, LoadingCache} - import org.apache.hadoop.fs.Path -import org.apache.hadoop.hive.metastore.api.FieldSchema import org.apache.hadoop.hive.metastore.Warehouse +import org.apache.hadoop.hive.metastore.api.FieldSchema import org.apache.hadoop.hive.ql.metadata._ import org.apache.hadoop.hive.serde2.Deserializer import org.apache.spark.Logging -import org.apache.spark.sql.{SaveMode, AnalysisException, SQLContext} -import org.apache.spark.sql.catalyst.analysis.{MultiInstanceRelation, Catalog, OverrideCatalog} +import org.apache.spark.sql.catalyst.analysis.{Catalog, MultiInstanceRelation, OverrideCatalog} import org.apache.spark.sql.catalyst.expressions._ import org.apache.spark.sql.catalyst.planning.PhysicalOperation import org.apache.spark.sql.catalyst.plans.logical import org.apache.spark.sql.catalyst.plans.logical._ import org.apache.spark.sql.catalyst.rules._ import org.apache.spark.sql.hive.client._ -import org.apache.spark.sql.parquet.{ParquetRelation2, Partition => ParquetPartition, PartitionSpec} -import org.apache.spark.sql.sources.{CreateTableUsingAsSelect, LogicalRelation, ResolvedDataSource} +import org.apache.spark.sql.parquet.ParquetRelation2 +import org.apache.spark.sql.sources.{CreateTableUsingAsSelect, LogicalRelation, Partition => ParquetPartition, PartitionSpec, ResolvedDataSource} import org.apache.spark.sql.types._ +import org.apache.spark.sql.{AnalysisException, SQLContext, SaveMode} import org.apache.spark.util.Utils /* Implicit conversions */ @@ -98,6 +97,7 @@ private[hive] class HiveMetastoreCatalog(val client: ClientInterface, hive: Hive ResolvedDataSource( hive, userSpecifiedSchema, + Array.empty[String], table.properties("spark.sql.sources.provider"), options) @@ -438,6 +438,7 @@ private[hive] class HiveMetastoreCatalog(val client: ClientInterface, hive: Hive desc.name, hive.conf.defaultDataSourceName, temporary = false, + Array.empty[String], mode, options = Map.empty[String, String], child diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveStrategies.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveStrategies.scala index be9249a8b1f44..d46a127d47d31 100644 --- a/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveStrategies.scala +++ b/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveStrategies.scala @@ -221,14 +221,14 @@ private[hive] trait HiveStrategies { object HiveDDLStrategy extends Strategy { def apply(plan: LogicalPlan): Seq[SparkPlan] = plan match { case CreateTableUsing( - tableName, userSpecifiedSchema, provider, false, opts, allowExisting, managedIfNoPath) => + tableName, userSpecifiedSchema, provider, false, opts, allowExisting, managedIfNoPath) => ExecutedCommand( CreateMetastoreDataSource( tableName, userSpecifiedSchema, provider, opts, allowExisting, managedIfNoPath)) :: Nil - case CreateTableUsingAsSelect(tableName, provider, false, mode, opts, query) => + case CreateTableUsingAsSelect(tableName, provider, false, partitionCols, mode, opts, query) => val cmd = - CreateMetastoreDataSourceAsSelect(tableName, provider, mode, opts, query) + CreateMetastoreDataSourceAsSelect(tableName, provider, partitionCols, mode, opts, query) ExecutedCommand(cmd) :: Nil case _ => Nil diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/execution/commands.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/execution/commands.scala index abab1a223a43a..8e405e080489f 100644 --- a/sql/hive/src/main/scala/org/apache/spark/sql/hive/execution/commands.scala +++ b/sql/hive/src/main/scala/org/apache/spark/sql/hive/execution/commands.scala @@ -158,6 +158,7 @@ private[hive] case class CreateMetastoreDataSourceAsSelect( tableName: String, provider: String, + partitionColumns: Array[String], mode: SaveMode, options: Map[String, String], query: LogicalPlan) extends RunnableCommand { @@ -189,12 +190,12 @@ case class CreateMetastoreDataSourceAsSelect( return Seq.empty[Row] case SaveMode.Append => // Check if the specified data source match the data source of the existing table. - val resolved = - ResolvedDataSource(sqlContext, Some(query.schema), provider, optionsWithPath) + val resolved = ResolvedDataSource( + sqlContext, Some(query.schema.asNullable), partitionColumns, provider, optionsWithPath) val createdRelation = LogicalRelation(resolved.relation) EliminateSubQueries(sqlContext.table(tableName).logicalPlan) match { - case l @ LogicalRelation(i: InsertableRelation) => - if (i != createdRelation.relation) { + case l @ LogicalRelation(_: InsertableRelation | _: FSBasedRelation) => + if (l.relation != createdRelation.relation) { val errorDescription = s"Cannot append to table $tableName because the resolved relation does not " + s"match the existing relation of $tableName. " + @@ -202,14 +203,13 @@ case class CreateMetastoreDataSourceAsSelect( s"table $tableName and using its data source and options." val errorMessage = s""" - |$errorDescription - |== Relations == - |${sideBySide( - s"== Expected Relation ==" :: - l.toString :: Nil, - s"== Actual Relation ==" :: - createdRelation.toString :: Nil).mkString("\n")} - """.stripMargin + |$errorDescription + |== Relations == + |${sideBySide( + s"== Expected Relation ==" :: l.toString :: Nil, + s"== Actual Relation ==" :: createdRelation.toString :: Nil + ).mkString("\n")} + """.stripMargin throw new AnalysisException(errorMessage) } existingSchema = Some(l.schema) @@ -234,7 +234,8 @@ case class CreateMetastoreDataSourceAsSelect( } // Create the relation based on the data of df. - val resolved = ResolvedDataSource(sqlContext, provider, mode, optionsWithPath, df) + val resolved = + ResolvedDataSource(sqlContext, provider, partitionColumns, mode, optionsWithPath, df) if (createMetastoreTable) { // We will use the schema of resolved.relation as the schema of the table (instead of diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveWriterContainers.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveWriterContainers.scala index 8398da268174d..cbc381cc81b59 100644 --- a/sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveWriterContainers.scala +++ b/sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveWriterContainers.scala @@ -204,7 +204,7 @@ private[spark] class SparkHiveDynamicPartitionWriterContainer( if (string == null || string.isEmpty) { defaultPartName } else { - FileUtils.escapePathName(string) + FileUtils.escapePathName(string, defaultPartName) } s"/$col=$colString" }.mkString diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/sources/FSBasedRelationSuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/sources/FSBasedRelationSuite.scala new file mode 100644 index 0000000000000..415b1cd168848 --- /dev/null +++ b/sql/hive/src/test/scala/org/apache/spark/sql/sources/FSBasedRelationSuite.scala @@ -0,0 +1,525 @@ +/* + * 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.sources + +import org.apache.hadoop.fs.Path + +import org.apache.spark.deploy.SparkHadoopUtil +import org.apache.spark.sql._ +import org.apache.spark.sql.hive.test.TestHive +import org.apache.spark.sql.parquet.ParquetTest +import org.apache.spark.sql.types._ + +// TODO Don't extend ParquetTest +// This test suite extends ParquetTest for some convenient utility methods. These methods should be +// moved to some more general places, maybe QueryTest. +class FSBasedRelationSuite extends QueryTest with ParquetTest { + override val sqlContext: SQLContext = TestHive + + import sqlContext._ + import sqlContext.implicits._ + + val dataSchema = + StructType( + Seq( + StructField("a", IntegerType, nullable = false), + StructField("b", StringType, nullable = false))) + + val testDF = (1 to 3).map(i => (i, s"val_$i")).toDF("a", "b") + + val partitionedTestDF1 = (for { + i <- 1 to 3 + p2 <- Seq("foo", "bar") + } yield (i, s"val_$i", 1, p2)).toDF("a", "b", "p1", "p2") + + val partitionedTestDF2 = (for { + i <- 1 to 3 + p2 <- Seq("foo", "bar") + } yield (i, s"val_$i", 2, p2)).toDF("a", "b", "p1", "p2") + + val partitionedTestDF = partitionedTestDF1.unionAll(partitionedTestDF2) + + def checkQueries(df: DataFrame): Unit = { + // Selects everything + checkAnswer( + df, + for (i <- 1 to 3; p1 <- 1 to 2; p2 <- Seq("foo", "bar")) yield Row(i, s"val_$i", p1, p2)) + + // Simple filtering and partition pruning + checkAnswer( + df.filter('a > 1 && 'p1 === 2), + for (i <- 2 to 3; p2 <- Seq("foo", "bar")) yield Row(i, s"val_$i", 2, p2)) + + // Simple projection and filtering + checkAnswer( + df.filter('a > 1).select('b, 'a + 1), + for (i <- 2 to 3; _ <- 1 to 2; _ <- Seq("foo", "bar")) yield Row(s"val_$i", i + 1)) + + // Simple projection and partition pruning + checkAnswer( + df.filter('a > 1 && 'p1 < 2).select('b, 'p1), + for (i <- 2 to 3; _ <- Seq("foo", "bar")) yield Row(s"val_$i", 1)) + + // Self-join + df.registerTempTable("t") + withTempTable("t") { + checkAnswer( + sql( + """SELECT l.a, r.b, l.p1, r.p2 + |FROM t l JOIN t r + |ON l.a = r.a AND l.p1 = r.p1 AND l.p2 = r.p2 + """.stripMargin), + for (i <- 1 to 3; p1 <- 1 to 2; p2 <- Seq("foo", "bar")) yield Row(i, s"val_$i", p1, p2)) + } + } + + test("save()/load() - non-partitioned table - Overwrite") { + withTempPath { file => + testDF.save( + path = file.getCanonicalPath, + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite) + + testDF.save( + path = file.getCanonicalPath, + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite) + + checkAnswer( + load( + source = classOf[SimpleTextSource].getCanonicalName, + options = Map( + "path" -> file.getCanonicalPath, + "dataSchema" -> dataSchema.json)), + testDF.collect()) + } + } + + test("save()/load() - non-partitioned table - Append") { + withTempPath { file => + testDF.save( + path = file.getCanonicalPath, + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite) + + testDF.save( + path = file.getCanonicalPath, + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Append) + + checkAnswer( + load( + source = classOf[SimpleTextSource].getCanonicalName, + options = Map( + "path" -> file.getCanonicalPath, + "dataSchema" -> dataSchema.json)).orderBy("a"), + testDF.unionAll(testDF).orderBy("a").collect()) + } + } + + test("save()/load() - non-partitioned table - ErrorIfExists") { + withTempDir { file => + intercept[RuntimeException] { + testDF.save( + path = file.getCanonicalPath, + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.ErrorIfExists) + } + } + } + + test("save()/load() - non-partitioned table - Ignore") { + withTempDir { file => + testDF.save( + path = file.getCanonicalPath, + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Ignore) + + val path = new Path(file.getCanonicalPath) + val fs = path.getFileSystem(sqlContext.sparkContext.hadoopConfiguration) + assert(fs.listStatus(path).isEmpty) + } + } + + test("save()/load() - partitioned table - simple queries") { + withTempPath { file => + partitionedTestDF.save( + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.ErrorIfExists, + options = Map("path" -> file.getCanonicalPath), + partitionColumns = Seq("p1", "p2")) + + checkQueries( + load( + source = classOf[SimpleTextSource].getCanonicalName, + options = Map( + "path" -> file.getCanonicalPath, + "dataSchema" -> dataSchema.json))) + } + } + + test("save()/load() - partitioned table - simple queries - partition columns in data") { + withTempDir { file => + val basePath = new Path(file.getCanonicalPath) + val fs = basePath.getFileSystem(SparkHadoopUtil.get.conf) + val qualifiedBasePath = fs.makeQualified(basePath) + + for (p1 <- 1 to 2; p2 <- Seq("foo", "bar")) { + val partitionDir = new Path(qualifiedBasePath, s"p1=$p1/p2=$p2") + sparkContext + .parallelize(for (i <- 1 to 3) yield s"$i,val_$i,$p1") + .saveAsTextFile(partitionDir.toString) + } + + val dataSchemaWithPartition = + StructType(dataSchema.fields :+ StructField("p1", IntegerType, nullable = true)) + + checkQueries( + load( + source = classOf[SimpleTextSource].getCanonicalName, + options = Map( + "path" -> file.getCanonicalPath, + "dataSchema" -> dataSchemaWithPartition.json))) + } + } + + test("save()/load() - partitioned table - Overwrite") { + withTempPath { file => + partitionedTestDF.save( + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + options = Map("path" -> file.getCanonicalPath), + partitionColumns = Seq("p1", "p2")) + + partitionedTestDF.save( + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + options = Map("path" -> file.getCanonicalPath), + partitionColumns = Seq("p1", "p2")) + + checkAnswer( + load( + source = classOf[SimpleTextSource].getCanonicalName, + options = Map( + "path" -> file.getCanonicalPath, + "dataSchema" -> dataSchema.json)), + partitionedTestDF.collect()) + } + } + + test("save()/load() - partitioned table - Append") { + withTempPath { file => + partitionedTestDF.save( + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + options = Map("path" -> file.getCanonicalPath), + partitionColumns = Seq("p1", "p2")) + + partitionedTestDF.save( + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Append, + options = Map("path" -> file.getCanonicalPath), + partitionColumns = Seq("p1", "p2")) + + checkAnswer( + load( + source = classOf[SimpleTextSource].getCanonicalName, + options = Map( + "path" -> file.getCanonicalPath, + "dataSchema" -> dataSchema.json)), + partitionedTestDF.unionAll(partitionedTestDF).collect()) + } + } + + test("save()/load() - partitioned table - Append - new partition values") { + withTempPath { file => + partitionedTestDF1.save( + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + options = Map("path" -> file.getCanonicalPath), + partitionColumns = Seq("p1", "p2")) + + partitionedTestDF2.save( + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Append, + options = Map("path" -> file.getCanonicalPath), + partitionColumns = Seq("p1", "p2")) + + checkAnswer( + load( + source = classOf[SimpleTextSource].getCanonicalName, + options = Map( + "path" -> file.getCanonicalPath, + "dataSchema" -> dataSchema.json)), + partitionedTestDF.collect()) + } + } + + test("save()/load() - partitioned table - ErrorIfExists") { + withTempDir { file => + intercept[RuntimeException] { + partitionedTestDF.save( + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.ErrorIfExists, + options = Map("path" -> file.getCanonicalPath), + partitionColumns = Seq("p1", "p2")) + } + } + } + + test("save()/load() - partitioned table - Ignore") { + withTempDir { file => + partitionedTestDF.save( + path = file.getCanonicalPath, + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Ignore) + + val path = new Path(file.getCanonicalPath) + val fs = path.getFileSystem(SparkHadoopUtil.get.conf) + assert(fs.listStatus(path).isEmpty) + } + } + + def withTable(tableName: String)(f: => Unit): Unit = { + try f finally sql(s"DROP TABLE $tableName") + } + + test("saveAsTable()/load() - non-partitioned table - Overwrite") { + testDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + Map("dataSchema" -> dataSchema.json)) + + withTable("t") { + checkAnswer(table("t"), testDF.collect()) + } + } + + test("saveAsTable()/load() - non-partitioned table - Append") { + testDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite) + + testDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Append) + + withTable("t") { + checkAnswer(table("t"), testDF.unionAll(testDF).orderBy("a").collect()) + } + } + + test("saveAsTable()/load() - non-partitioned table - ErrorIfExists") { + Seq.empty[(Int, String)].toDF().registerTempTable("t") + + withTempTable("t") { + intercept[AnalysisException] { + testDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.ErrorIfExists) + } + } + } + + test("saveAsTable()/load() - non-partitioned table - Ignore") { + Seq.empty[(Int, String)].toDF().registerTempTable("t") + + withTempTable("t") { + testDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Ignore) + + assert(table("t").collect().isEmpty) + } + } + + test("saveAsTable()/load() - partitioned table - simple queries") { + partitionedTestDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + Map("dataSchema" -> dataSchema.json)) + + withTable("t") { + checkQueries(table("t")) + } + } + + test("saveAsTable()/load() - partitioned table - Overwrite") { + partitionedTestDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + options = Map("dataSchema" -> dataSchema.json), + partitionColumns = Seq("p1", "p2")) + + partitionedTestDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + options = Map("dataSchema" -> dataSchema.json), + partitionColumns = Seq("p1", "p2")) + + withTable("t") { + checkAnswer(table("t"), partitionedTestDF.collect()) + } + } + + test("saveAsTable()/load() - partitioned table - Append") { + partitionedTestDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + options = Map("dataSchema" -> dataSchema.json), + partitionColumns = Seq("p1", "p2")) + + partitionedTestDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Append, + options = Map("dataSchema" -> dataSchema.json), + partitionColumns = Seq("p1", "p2")) + + withTable("t") { + checkAnswer(table("t"), partitionedTestDF.unionAll(partitionedTestDF).collect()) + } + } + + test("saveAsTable()/load() - partitioned table - Append - new partition values") { + partitionedTestDF1.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + options = Map("dataSchema" -> dataSchema.json), + partitionColumns = Seq("p1", "p2")) + + partitionedTestDF2.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Append, + options = Map("dataSchema" -> dataSchema.json), + partitionColumns = Seq("p1", "p2")) + + withTable("t") { + checkAnswer(table("t"), partitionedTestDF.collect()) + } + } + + test("saveAsTable()/load() - partitioned table - Append - mismatched partition columns") { + partitionedTestDF1.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + options = Map("dataSchema" -> dataSchema.json), + partitionColumns = Seq("p1", "p2")) + + // Using only a subset of all partition columns + intercept[Throwable] { + partitionedTestDF2.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Append, + options = Map("dataSchema" -> dataSchema.json), + partitionColumns = Seq("p1")) + } + + // Using different order of partition columns + intercept[Throwable] { + partitionedTestDF2.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Append, + options = Map("dataSchema" -> dataSchema.json), + partitionColumns = Seq("p2", "p1")) + } + } + + test("saveAsTable()/load() - partitioned table - ErrorIfExists") { + Seq.empty[(Int, String)].toDF().registerTempTable("t") + + withTempTable("t") { + intercept[AnalysisException] { + partitionedTestDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.ErrorIfExists, + options = Map("dataSchema" -> dataSchema.json), + partitionColumns = Seq("p1", "p2")) + } + } + } + + test("saveAsTable()/load() - partitioned table - Ignore") { + Seq.empty[(Int, String)].toDF().registerTempTable("t") + + withTempTable("t") { + partitionedTestDF.saveAsTable( + tableName = "t", + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Ignore, + options = Map("dataSchema" -> dataSchema.json), + partitionColumns = Seq("p1", "p2")) + + assert(table("t").collect().isEmpty) + } + } + + test("Hadoop style globbing") { + withTempPath { file => + partitionedTestDF.save( + source = classOf[SimpleTextSource].getCanonicalName, + mode = SaveMode.Overwrite, + options = Map("path" -> file.getCanonicalPath), + partitionColumns = Seq("p1", "p2")) + + val df = load( + source = classOf[SimpleTextSource].getCanonicalName, + options = Map( + "path" -> s"${file.getCanonicalPath}/p1=*/p2=???", + "dataSchema" -> dataSchema.json)) + + val expectedPaths = Set( + s"${file.getCanonicalFile}/p1=1/p2=foo", + s"${file.getCanonicalFile}/p1=2/p2=foo", + s"${file.getCanonicalFile}/p1=1/p2=bar", + s"${file.getCanonicalFile}/p1=2/p2=bar" + ).map { p => + val path = new Path(p) + val fs = path.getFileSystem(sqlContext.sparkContext.hadoopConfiguration) + path.makeQualified(fs.getUri, fs.getWorkingDirectory).toString + } + + println(df.queryExecution) + + val actualPaths = df.queryExecution.analyzed.collectFirst { + case LogicalRelation(relation: FSBasedRelation) => + relation.paths.toSet + }.getOrElse { + fail("Expect an FSBasedRelation, but none could be found") + } + + assert(actualPaths === expectedPaths) + checkAnswer(df, partitionedTestDF.collect()) + } + } +} diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/sources/SimpleTextRelation.scala b/sql/hive/src/test/scala/org/apache/spark/sql/sources/SimpleTextRelation.scala new file mode 100644 index 0000000000000..8801aba2f64c3 --- /dev/null +++ b/sql/hive/src/test/scala/org/apache/spark/sql/sources/SimpleTextRelation.scala @@ -0,0 +1,125 @@ +/* + * 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.sources + +import java.text.NumberFormat +import java.util.UUID + +import com.google.common.base.Objects +import org.apache.hadoop.fs.Path +import org.apache.hadoop.io.{NullWritable, Text} +import org.apache.hadoop.mapreduce.lib.output.{FileOutputFormat, TextOutputFormat} +import org.apache.hadoop.mapreduce.{RecordWriter, TaskAttemptContext} + +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.catalyst.expressions.{Cast, Literal} +import org.apache.spark.sql.types.{DataType, StructField, StructType} +import org.apache.spark.sql.{Row, SQLContext} + +/** + * A simple example [[FSBasedRelationProvider]]. + */ +class SimpleTextSource extends FSBasedRelationProvider { + override def createRelation( + sqlContext: SQLContext, + paths: Array[String], + schema: Option[StructType], + partitionColumns: Option[StructType], + parameters: Map[String, String]): FSBasedRelation = { + val partitionsSchema = partitionColumns.getOrElse(StructType(Array.empty[StructField])) + new SimpleTextRelation(paths, schema, partitionsSchema, parameters)(sqlContext) + } +} + +class AppendingTextOutputFormat(outputFile: Path) extends TextOutputFormat[NullWritable, Text] { + val numberFormat = NumberFormat.getInstance() + + numberFormat.setMinimumIntegerDigits(5) + numberFormat.setGroupingUsed(false) + + override def getDefaultWorkFile(context: TaskAttemptContext, extension: String): Path = { + val split = context.getTaskAttemptID.getTaskID.getId + val name = FileOutputFormat.getOutputName(context) + new Path(outputFile, s"$name-${numberFormat.format(split)}-${UUID.randomUUID()}") + } +} + +class SimpleTextOutputWriter extends OutputWriter { + private var recordWriter: RecordWriter[NullWritable, Text] = _ + private var taskAttemptContext: TaskAttemptContext = _ + + override def init( + path: String, + dataSchema: StructType, + context: TaskAttemptContext): Unit = { + recordWriter = new AppendingTextOutputFormat(new Path(path)).getRecordWriter(context) + taskAttemptContext = context + } + + override def write(row: Row): Unit = { + val serialized = row.toSeq.map(_.toString).mkString(",") + recordWriter.write(null, new Text(serialized)) + } + + override def close(): Unit = recordWriter.close(taskAttemptContext) +} + +/** + * A simple example [[FSBasedRelation]], used for testing purposes. Data are stored as comma + * separated string lines. When scanning data, schema must be explicitly provided via data source + * option `"dataSchema"`. + */ +class SimpleTextRelation( + paths: Array[String], + val maybeDataSchema: Option[StructType], + partitionsSchema: StructType, + parameters: Map[String, String])( + @transient val sqlContext: SQLContext) + extends FSBasedRelation(paths, partitionsSchema) { + + import sqlContext.sparkContext + + override val dataSchema: StructType = + maybeDataSchema.getOrElse(DataType.fromJson(parameters("dataSchema")).asInstanceOf[StructType]) + + override def equals(other: Any): Boolean = other match { + case that: SimpleTextRelation => + this.paths.sameElements(that.paths) && + this.maybeDataSchema == that.maybeDataSchema && + this.dataSchema == that.dataSchema && + this.partitionColumns == that.partitionColumns + + case _ => false + } + + override def hashCode(): Int = + Objects.hashCode(paths, maybeDataSchema, dataSchema) + + override def outputWriterClass: Class[_ <: OutputWriter] = + classOf[SimpleTextOutputWriter] + + override def buildScan(inputPaths: Array[String]): RDD[Row] = { + val fields = dataSchema.map(_.dataType) + + sparkContext.textFile(inputPaths.mkString(",")).map { record => + Row(record.split(",").zip(fields).map { case (value, dataType) => + Cast(Literal(value), dataType).eval() + }: _*) + } + } +} From bfcaf8adcdc20dec203e2e9d5a72b52dd6f226a9 Mon Sep 17 00:00:00 2001 From: Wenchen Fan Date: Tue, 12 May 2015 10:37:57 -0700 Subject: [PATCH 09/27] [DataFrame][minor] support column in field accessor Minor improvement, now we can use `Column` as extraction expression. Author: Wenchen Fan Closes #6080 from cloud-fan/tmp and squashes the following commits: 0fdefb7 [Wenchen Fan] support column in field accessor --- sql/core/src/main/scala/org/apache/spark/sql/Column.scala | 2 +- .../src/test/scala/org/apache/spark/sql/DataFrameSuite.scala | 1 + 2 files changed, 2 insertions(+), 1 deletion(-) 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 e6e475bb82f82..4d50821620f5e 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 @@ -78,7 +78,7 @@ class Column(protected[sql] val expr: Expression) extends Logging { * * @group expr_ops */ - def apply(field: Any): Column = UnresolvedExtractValue(expr, Literal(field)) + def apply(extraction: Any): Column = UnresolvedExtractValue(expr, lit(extraction).expr) /** * Unary minus, i.e. negate the expression. diff --git a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala index 2ade955864b71..d58438e5d129c 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala @@ -456,6 +456,7 @@ class DataFrameSuite extends QueryTest { assert(complexData.filter(complexData("a")(0) === 2).count() == 1) assert(complexData.filter(complexData("m")("1") === 1).count() == 1) assert(complexData.filter(complexData("s")("key") === 1).count() == 1) + assert(complexData.filter(complexData("m")(complexData("s")("value")) === 1).count() == 1) } test("SPARK-7324 dropDuplicates") { From 65697bbeafe507dda066e2dc14ca5183f278dfe9 Mon Sep 17 00:00:00 2001 From: Andrew Or Date: Tue, 12 May 2015 11:17:59 -0700 Subject: [PATCH 10/27] [SPARK-7500] DAG visualization: move cluster labeling to dagre-d3 This fixes the label bleeding issue described in the JIRA and pictured in the screenshots below. I also took the opportunity to move some code to the places that they belong more to. In particular: (1) Drawing cluster labels is now implemented in my branch of dagre-d3 instead of in Spark (2) All graph styling is now moved from Scala to JS Note that these changes are related because our existing mechanism of "tacking on cluster labels" afterwards isn't flexible enough for us to fix issues like this one easily. For the other half of the changes, visit http://github.com/andrewor14/dagre-d3. ------------------- **Before.** ------------------- **After.** Author: Andrew Or Closes #6076 from andrewor14/dag-viz-bleed and squashes the following commits: 5858d7a [Andrew Or] Merge branch 'master' of github.com:apache/spark into dag-viz-bleed c686dc4 [Andrew Or] Fix tooltip placement d908c36 [Andrew Or] Add link to dagre-d3 changes (minor) 4a4fb58 [Andrew Or] Fix bleeding + move all styling to JS --- .../apache/spark/ui/static/dagre-d3.min.js | 14 ++-- .../apache/spark/ui/static/spark-dag-viz.js | 79 ++++++++----------- .../scala/org/apache/spark/ui/UIUtils.scala | 2 +- .../spark/ui/scope/RDDOperationGraph.scala | 40 +++------- 4 files changed, 48 insertions(+), 87 deletions(-) diff --git a/core/src/main/resources/org/apache/spark/ui/static/dagre-d3.min.js b/core/src/main/resources/org/apache/spark/ui/static/dagre-d3.min.js index 6beb7db855ca0..acf2d93b718b2 100644 --- a/core/src/main/resources/org/apache/spark/ui/static/dagre-d3.min.js +++ b/core/src/main/resources/org/apache/spark/ui/static/dagre-d3.min.js @@ -1,4 +1,4 @@ -!function(e){if("object"==typeof exports&&"undefined"!=typeof module)module.exports=e();else if("function"==typeof define&&define.amd)define([],e);else{var f;"undefined"!=typeof window?f=window:"undefined"!=typeof global?f=global:"undefined"!=typeof self&&(f=self),f.dagreD3=e()}}(function(){var define,module,exports;return function e(t,n,r){function s(o,u){if(!n[o]){if(!t[o]){var a=typeof require=="function"&&require;if(!u&&a)return a(o,!0);if(i)return i(o,!0);var f=new Error("Cannot find module '"+o+"'");throw f.code="MODULE_NOT_FOUND",f}var l=n[o]={exports:{}};t[o][0].call(l.exports,function(e){var n=t[o][1][e];return s(n?n:e)},l,l.exports,e,t,n,r)}return n[o].exports}var i=typeof require=="function"&&require;for(var o=0;o0}},{}],14:[function(require,module,exports){module.exports=intersectNode;function intersectNode(node,point){return node.intersect(point)}},{}],15:[function(require,module,exports){var intersectLine=require("./intersect-line");module.exports=intersectPolygon;function intersectPolygon(node,polyPoints,point){var x1=node.x;var y1=node.y;var intersections=[];var minX=Number.POSITIVE_INFINITY,minY=Number.POSITIVE_INFINITY;polyPoints.forEach(function(entry){minX=Math.min(minX,entry.x);minY=Math.min(minY,entry.y)});var left=x1-node.width/2-minX;var top=y1-node.height/2-minY;for(var i=0;i1){intersections.sort(function(p,q){var pdx=p.x-point.x,pdy=p.y-point.y,distp=Math.sqrt(pdx*pdx+pdy*pdy),qdx=q.x-point.x,qdy=q.y-point.y,distq=Math.sqrt(qdx*qdx+qdy*qdy);return distpMath.abs(dx)*h){if(dy<0){h=-h}sx=dy===0?0:h*dx/dy;sy=h}else{if(dx<0){w=-w}sx=w;sy=dx===0?0:w*dy/dx}return{x:x+sx,y:y+sy}}},{}],17:[function(require,module,exports){var util=require("../util");module.exports=addHtmlLabel;function addHtmlLabel(root,node){var fo=root.append("foreignObject").attr("width","100000");var div=fo.append("xhtml:div");var label=node.label;switch(typeof label){case"function":div.insert(label);break;case"object":div.insert(function(){return label});break;default:div.html(label)}util.applyStyle(div,node.labelStyle);div.style("display","inline-block");div.style("white-space","nowrap");var w,h;div.each(function(){w=this.clientWidth;h=this.clientHeight});fo.attr("width",w).attr("height",h);return fo}},{"../util":25}],18:[function(require,module,exports){var addTextLabel=require("./add-text-label"),addHtmlLabel=require("./add-html-label");module.exports=addLabel;function addLabel(root,node){var label=node.label;var labelSvg=root.append("g");if(typeof label!=="string"||node.labelType==="html"){addHtmlLabel(labelSvg,node)}else{addTextLabel(labelSvg,node)}var labelBBox=labelSvg.node().getBBox();labelSvg.attr("transform","translate("+-labelBBox.width/2+","+-labelBBox.height/2+")");return labelSvg}},{"./add-html-label":17,"./add-text-label":19}],19:[function(require,module,exports){var util=require("../util");module.exports=addTextLabel;function addTextLabel(root,node){var domNode=root.append("text");var lines=processEscapeSequences(node.label).split("\n");for(var i=0;i0;--i){entry=buckets[i].dequeue();if(entry){results=results.concat(removeNode(g,buckets,zeroIdx,entry,true));break}}}}return results}function removeNode(g,buckets,zeroIdx,entry,collectPredecessors){var results=collectPredecessors?[]:undefined;_.each(g.inEdges(entry.v),function(edge){var weight=g.edge(edge),uEntry=g.node(edge.v);if(collectPredecessors){results.push({v:edge.v,w:edge.w})}uEntry.out-=weight;assignBucket(buckets,zeroIdx,uEntry)});_.each(g.outEdges(entry.v),function(edge){var weight=g.edge(edge),w=edge.w,wEntry=g.node(w);wEntry["in"]-=weight;assignBucket(buckets,zeroIdx,wEntry)});g.removeNode(entry.v);return results}function buildState(g,weightFn){var fasGraph=new Graph,maxIn=0,maxOut=0;_.each(g.nodes(),function(v){fasGraph.setNode(v,{v:v,"in":0,out:0})});_.each(g.edges(),function(e){var prevWeight=fasGraph.edge(e.v,e.w)||0,weight=weightFn(e),edgeWeight=prevWeight+weight;fasGraph.setEdge(e.v,e.w,edgeWeight);maxOut=Math.max(maxOut,fasGraph.node(e.v).out+=weight);maxIn=Math.max(maxIn,fasGraph.node(e.w)["in"]+=weight)});var buckets=_.range(maxOut+maxIn+3).map(function(){return new List});var zeroIdx=maxIn+1;_.each(fasGraph.nodes(),function(v){assignBucket(buckets,zeroIdx,fasGraph.node(v))});return{graph:fasGraph,buckets:buckets,zeroIdx:zeroIdx}}function assignBucket(buckets,zeroIdx,entry){if(!entry.out){buckets[0].enqueue(entry)}else if(!entry["in"]){buckets[buckets.length-1].enqueue(entry)}else{buckets[entry.out-entry["in"]+zeroIdx].enqueue(entry)}}},{"./data/list":31,"./graphlib":33,"./lodash":36}],35:[function(require,module,exports){"use strict";var _=require("./lodash"),acyclic=require("./acyclic"),normalize=require("./normalize"),rank=require("./rank"),normalizeRanks=require("./util").normalizeRanks,parentDummyChains=require("./parent-dummy-chains"),removeEmptyRanks=require("./util").removeEmptyRanks,nestingGraph=require("./nesting-graph"),addBorderSegments=require("./add-border-segments"),coordinateSystem=require("./coordinate-system"),order=require("./order"),position=require("./position"),util=require("./util"),Graph=require("./graphlib").Graph;module.exports=layout;function layout(g,opts){var time=opts&&opts.debugTiming?util.time:util.notime;time("layout",function(){var layoutGraph=time(" buildLayoutGraph",function(){return buildLayoutGraph(g)});time(" runLayout",function(){runLayout(layoutGraph,time)});time(" updateInputGraph",function(){updateInputGraph(g,layoutGraph)})})}function runLayout(g,time){time(" makeSpaceForEdgeLabels",function(){makeSpaceForEdgeLabels(g)});time(" removeSelfEdges",function(){removeSelfEdges(g)});time(" acyclic",function(){acyclic.run(g)});time(" nestingGraph.run",function(){nestingGraph.run(g)});time(" rank",function(){rank(util.asNonCompoundGraph(g))});time(" injectEdgeLabelProxies",function(){injectEdgeLabelProxies(g)});time(" removeEmptyRanks",function(){removeEmptyRanks(g)});time(" nestingGraph.cleanup",function(){nestingGraph.cleanup(g)});time(" normalizeRanks",function(){normalizeRanks(g)});time(" assignRankMinMax",function(){assignRankMinMax(g)});time(" removeEdgeLabelProxies",function(){removeEdgeLabelProxies(g)});time(" normalize.run",function(){normalize.run(g)});time(" parentDummyChains",function(){parentDummyChains(g)});time(" addBorderSegments",function(){addBorderSegments(g)});time(" order",function(){order(g)});time(" insertSelfEdges",function(){insertSelfEdges(g)});time(" adjustCoordinateSystem",function(){coordinateSystem.adjust(g)});time(" position",function(){position(g)});time(" positionSelfEdges",function(){positionSelfEdges(g)});time(" removeBorderNodes",function(){removeBorderNodes(g)});time(" normalize.undo",function(){normalize.undo(g)});time(" fixupEdgeLabelCoords",function(){fixupEdgeLabelCoords(g)});time(" undoCoordinateSystem",function(){coordinateSystem.undo(g)});time(" translateGraph",function(){translateGraph(g)});time(" assignNodeIntersects",function(){assignNodeIntersects(g)});time(" reversePoints",function(){reversePointsForReversedEdges(g)});time(" acyclic.undo",function(){acyclic.undo(g)})}function updateInputGraph(inputGraph,layoutGraph){_.each(inputGraph.nodes(),function(v){var inputLabel=inputGraph.node(v),layoutLabel=layoutGraph.node(v);if(inputLabel){inputLabel.x=layoutLabel.x;inputLabel.y=layoutLabel.y;if(layoutGraph.children(v).length){inputLabel.width=layoutLabel.width;inputLabel.height=layoutLabel.height}}});_.each(inputGraph.edges(),function(e){var inputLabel=inputGraph.edge(e),layoutLabel=layoutGraph.edge(e);inputLabel.points=layoutLabel.points;if(_.has(layoutLabel,"x")){inputLabel.x=layoutLabel.x;inputLabel.y=layoutLabel.y}});inputGraph.graph().width=layoutGraph.graph().width;inputGraph.graph().height=layoutGraph.graph().height}var graphNumAttrs=["nodesep","edgesep","ranksep","marginx","marginy"],graphDefaults={ranksep:50,edgesep:20,nodesep:50,rankdir:"tb"},graphAttrs=["acyclicer","ranker","rankdir","align"],nodeNumAttrs=["width","height"],nodeDefaults={width:0,height:0},edgeNumAttrs=["minlen","weight","width","height","labeloffset"],edgeDefaults={minlen:1,weight:1,width:0,height:0,labeloffset:10,labelpos:"r"},edgeAttrs=["labelpos"];function buildLayoutGraph(inputGraph){var g=new Graph({multigraph:true,compound:true}),graph=canonicalize(inputGraph.graph());g.setGraph(_.merge({},graphDefaults,selectNumberAttrs(graph,graphNumAttrs),_.pick(graph,graphAttrs))); +module.exports={graphlib:require("./lib/graphlib"),dagre:require("./lib/dagre"),intersect:require("./lib/intersect"),render:require("./lib/render"),util:require("./lib/util"),version:require("./lib/version")}},{"./lib/dagre":8,"./lib/graphlib":9,"./lib/intersect":10,"./lib/render":23,"./lib/util":25,"./lib/version":26}],2:[function(require,module,exports){var util=require("./util");module.exports={"default":normal,normal:normal,vee:vee,undirected:undirected};function normal(parent,id,edge,type){var marker=parent.append("marker").attr("id",id).attr("viewBox","0 0 10 10").attr("refX",9).attr("refY",5).attr("markerUnits","strokeWidth").attr("markerWidth",8).attr("markerHeight",6).attr("orient","auto");var path=marker.append("path").attr("d","M 0 0 L 10 5 L 0 10 z").style("stroke-width",1).style("stroke-dasharray","1,0");util.applyStyle(path,edge[type+"Style"])}function vee(parent,id,edge,type){var marker=parent.append("marker").attr("id",id).attr("viewBox","0 0 10 10").attr("refX",9).attr("refY",5).attr("markerUnits","strokeWidth").attr("markerWidth",8).attr("markerHeight",6).attr("orient","auto");var path=marker.append("path").attr("d","M 0 0 L 10 5 L 0 10 L 4 5 z").style("stroke-width",1).style("stroke-dasharray","1,0");util.applyStyle(path,edge[type+"Style"])}function undirected(parent,id,edge,type){var marker=parent.append("marker").attr("id",id).attr("viewBox","0 0 10 10").attr("refX",9).attr("refY",5).attr("markerUnits","strokeWidth").attr("markerWidth",8).attr("markerHeight",6).attr("orient","auto");var path=marker.append("path").attr("d","M 0 5 L 10 5").style("stroke-width",1).style("stroke-dasharray","1,0");util.applyStyle(path,edge[type+"Style"])}},{"./util":25}],3:[function(require,module,exports){var _=require("./lodash"),addLabel=require("./label/add-label"),util=require("./util");module.exports=createClusters;function createClusters(selection,g){var clusters=g.nodes().filter(function(v){return util.isSubgraph(g,v)}),svgClusters=selection.selectAll("g.cluster").data(clusters,function(v){return v});var makeClusterIdentifier=function(v){return"cluster_"+v.replace(/^cluster/,"")};svgClusters.enter().append("g").attr("id",makeClusterIdentifier).attr("name",function(v){return g.node(v).label}).classed("cluster",true).style("opacity",0).append("rect");var sortedClusters=util.orderByRank(g,svgClusters.data());for(var i=0;i0}},{}],14:[function(require,module,exports){module.exports=intersectNode;function intersectNode(node,point){return node.intersect(point)}},{}],15:[function(require,module,exports){var intersectLine=require("./intersect-line");module.exports=intersectPolygon;function intersectPolygon(node,polyPoints,point){var x1=node.x;var y1=node.y;var intersections=[];var minX=Number.POSITIVE_INFINITY,minY=Number.POSITIVE_INFINITY;polyPoints.forEach(function(entry){minX=Math.min(minX,entry.x);minY=Math.min(minY,entry.y)});var left=x1-node.width/2-minX;var top=y1-node.height/2-minY;for(var i=0;i1){intersections.sort(function(p,q){var pdx=p.x-point.x,pdy=p.y-point.y,distp=Math.sqrt(pdx*pdx+pdy*pdy),qdx=q.x-point.x,qdy=q.y-point.y,distq=Math.sqrt(qdx*qdx+qdy*qdy);return distpMath.abs(dx)*h){if(dy<0){h=-h}sx=dy===0?0:h*dx/dy;sy=h}else{if(dx<0){w=-w}sx=w;sy=dx===0?0:w*dy/dx}return{x:x+sx,y:y+sy}}},{}],17:[function(require,module,exports){var util=require("../util");module.exports=addHtmlLabel;function addHtmlLabel(root,node){var fo=root.append("foreignObject").attr("width","100000");var div=fo.append("xhtml:div");var label=node.label;switch(typeof label){case"function":div.insert(label);break;case"object":div.insert(function(){return label});break;default:div.html(label)}util.applyStyle(div,node.labelStyle);div.style("display","inline-block");div.style("white-space","nowrap");var w,h;div.each(function(){w=this.clientWidth;h=this.clientHeight});fo.attr("width",w).attr("height",h);return fo}},{"../util":25}],18:[function(require,module,exports){var addTextLabel=require("./add-text-label"),addHtmlLabel=require("./add-html-label");module.exports=addLabel;function addLabel(root,node){var label=node.label;var labelSvg=root.append("g");if(typeof label!=="string"||node.labelType==="html"){addHtmlLabel(labelSvg,node)}else{addTextLabel(labelSvg,node)}var labelBBox=labelSvg.node().getBBox();labelSvg.attr("transform","translate("+-labelBBox.width/2+","+-labelBBox.height/2+")");return labelSvg}},{"./add-html-label":17,"./add-text-label":19}],19:[function(require,module,exports){var util=require("../util");module.exports=addTextLabel;function addTextLabel(root,node){var domNode=root.append("text");var lines=processEscapeSequences(node.label).split("\n");for(var i=0;imaxPadding){maxPadding=child.paddingTop}}return maxPadding}function getRank(g,v){var maxRank=0;var children=g.children(v);for(var i=0;imaxRank){maxRank=thisRank}}return maxRank}function orderByRank(g,nodes){return nodes.sort(function(x,y){return getRank(g,x)-getRank(g,y)})}function edgeToId(e){return escapeId(e.v)+":"+escapeId(e.w)+":"+escapeId(e.name)}var ID_DELIM=/:/g;function escapeId(str){return str?String(str).replace(ID_DELIM,"\\:"):""}function applyStyle(dom,styleFn){if(styleFn){dom.attr("style",styleFn)}}function applyClass(dom,classFn,otherClasses){if(classFn){dom.attr("class",classFn).attr("class",otherClasses+" "+dom.attr("class"))}}function applyTransition(selection,g){var graph=g.graph();if(_.isPlainObject(graph)){var transition=graph.transition;if(_.isFunction(transition)){return transition(selection)}}return selection}},{"./lodash":20}],26:[function(require,module,exports){module.exports="0.4.4-pre"},{}],27:[function(require,module,exports){module.exports={graphlib:require("./lib/graphlib"),layout:require("./lib/layout"),debug:require("./lib/debug"),util:{time:require("./lib/util").time,notime:require("./lib/util").notime},version:require("./lib/version")}},{"./lib/debug":32,"./lib/graphlib":33,"./lib/layout":35,"./lib/util":55,"./lib/version":56}],28:[function(require,module,exports){"use strict";var _=require("./lodash"),greedyFAS=require("./greedy-fas");module.exports={run:run,undo:undo};function run(g){var fas=g.graph().acyclicer==="greedy"?greedyFAS(g,weightFn(g)):dfsFAS(g);_.each(fas,function(e){var label=g.edge(e);g.removeEdge(e);label.forwardName=e.name;label.reversed=true;g.setEdge(e.w,e.v,label,_.uniqueId("rev"))});function weightFn(g){return function(e){return g.edge(e).weight}}}function dfsFAS(g){var fas=[],stack={},visited={};function dfs(v){if(_.has(visited,v)){return}visited[v]=true;stack[v]=true;_.each(g.outEdges(v),function(e){if(_.has(stack,e.w)){fas.push(e)}else{dfs(e.w)}});delete stack[v]}_.each(g.nodes(),dfs);return fas}function undo(g){_.each(g.edges(),function(e){var label=g.edge(e);if(label.reversed){g.removeEdge(e);var forwardName=label.forwardName;delete label.reversed;delete label.forwardName;g.setEdge(e.w,e.v,label,forwardName)}})}},{"./greedy-fas":34,"./lodash":36}],29:[function(require,module,exports){var _=require("./lodash"),util=require("./util");module.exports=addBorderSegments;function addBorderSegments(g){function dfs(v){var children=g.children(v),node=g.node(v);if(children.length){_.each(children,dfs)}if(_.has(node,"minRank")){node.borderLeft=[];node.borderRight=[];for(var rank=node.minRank,maxRank=node.maxRank+1;rank0;--i){entry=buckets[i].dequeue();if(entry){results=results.concat(removeNode(g,buckets,zeroIdx,entry,true));break}}}}return results}function removeNode(g,buckets,zeroIdx,entry,collectPredecessors){var results=collectPredecessors?[]:undefined;_.each(g.inEdges(entry.v),function(edge){var weight=g.edge(edge),uEntry=g.node(edge.v);if(collectPredecessors){results.push({v:edge.v,w:edge.w})}uEntry.out-=weight;assignBucket(buckets,zeroIdx,uEntry)});_.each(g.outEdges(entry.v),function(edge){var weight=g.edge(edge),w=edge.w,wEntry=g.node(w);wEntry["in"]-=weight;assignBucket(buckets,zeroIdx,wEntry)});g.removeNode(entry.v);return results}function buildState(g,weightFn){var fasGraph=new Graph,maxIn=0,maxOut=0;_.each(g.nodes(),function(v){fasGraph.setNode(v,{v:v,"in":0,out:0})});_.each(g.edges(),function(e){var prevWeight=fasGraph.edge(e.v,e.w)||0,weight=weightFn(e),edgeWeight=prevWeight+weight;fasGraph.setEdge(e.v,e.w,edgeWeight);maxOut=Math.max(maxOut,fasGraph.node(e.v).out+=weight);maxIn=Math.max(maxIn,fasGraph.node(e.w)["in"]+=weight)});var buckets=_.range(maxOut+maxIn+3).map(function(){return new List});var zeroIdx=maxIn+1;_.each(fasGraph.nodes(),function(v){assignBucket(buckets,zeroIdx,fasGraph.node(v))});return{graph:fasGraph,buckets:buckets,zeroIdx:zeroIdx}}function assignBucket(buckets,zeroIdx,entry){if(!entry.out){buckets[0].enqueue(entry)}else if(!entry["in"]){buckets[buckets.length-1].enqueue(entry)}else{buckets[entry.out-entry["in"]+zeroIdx].enqueue(entry)}}},{"./data/list":31,"./graphlib":33,"./lodash":36}],35:[function(require,module,exports){"use strict";var _=require("./lodash"),acyclic=require("./acyclic"),normalize=require("./normalize"),rank=require("./rank"),normalizeRanks=require("./util").normalizeRanks,parentDummyChains=require("./parent-dummy-chains"),removeEmptyRanks=require("./util").removeEmptyRanks,nestingGraph=require("./nesting-graph"),addBorderSegments=require("./add-border-segments"),coordinateSystem=require("./coordinate-system"),order=require("./order"),position=require("./position"),util=require("./util"),Graph=require("./graphlib").Graph;module.exports=layout;function layout(g,opts){var time=opts&&opts.debugTiming?util.time:util.notime;time("layout",function(){var layoutGraph=time(" buildLayoutGraph",function(){return buildLayoutGraph(g)});time(" runLayout",function(){runLayout(layoutGraph,time)});time(" updateInputGraph",function(){updateInputGraph(g,layoutGraph)})})}function runLayout(g,time){time(" makeSpaceForEdgeLabels",function(){makeSpaceForEdgeLabels(g)});time(" removeSelfEdges",function(){removeSelfEdges(g)});time(" acyclic",function(){acyclic.run(g)});time(" nestingGraph.run",function(){nestingGraph.run(g)});time(" rank",function(){rank(util.asNonCompoundGraph(g))});time(" injectEdgeLabelProxies",function(){injectEdgeLabelProxies(g)});time(" removeEmptyRanks",function(){removeEmptyRanks(g)});time(" nestingGraph.cleanup",function(){nestingGraph.cleanup(g)});time(" normalizeRanks",function(){normalizeRanks(g)});time(" assignRankMinMax",function(){assignRankMinMax(g)});time(" removeEdgeLabelProxies",function(){removeEdgeLabelProxies(g)});time(" normalize.run",function(){normalize.run(g)});time(" parentDummyChains",function(){ +parentDummyChains(g)});time(" addBorderSegments",function(){addBorderSegments(g)});time(" order",function(){order(g)});time(" insertSelfEdges",function(){insertSelfEdges(g)});time(" adjustCoordinateSystem",function(){coordinateSystem.adjust(g)});time(" position",function(){position(g)});time(" positionSelfEdges",function(){positionSelfEdges(g)});time(" removeBorderNodes",function(){removeBorderNodes(g)});time(" normalize.undo",function(){normalize.undo(g)});time(" fixupEdgeLabelCoords",function(){fixupEdgeLabelCoords(g)});time(" undoCoordinateSystem",function(){coordinateSystem.undo(g)});time(" translateGraph",function(){translateGraph(g)});time(" assignNodeIntersects",function(){assignNodeIntersects(g)});time(" reversePoints",function(){reversePointsForReversedEdges(g)});time(" acyclic.undo",function(){acyclic.undo(g)})}function updateInputGraph(inputGraph,layoutGraph){_.each(inputGraph.nodes(),function(v){var inputLabel=inputGraph.node(v),layoutLabel=layoutGraph.node(v);if(inputLabel){inputLabel.x=layoutLabel.x;inputLabel.y=layoutLabel.y;if(layoutGraph.children(v).length){inputLabel.width=layoutLabel.width;inputLabel.height=layoutLabel.height}}});_.each(inputGraph.edges(),function(e){var inputLabel=inputGraph.edge(e),layoutLabel=layoutGraph.edge(e);inputLabel.points=layoutLabel.points;if(_.has(layoutLabel,"x")){inputLabel.x=layoutLabel.x;inputLabel.y=layoutLabel.y}});inputGraph.graph().width=layoutGraph.graph().width;inputGraph.graph().height=layoutGraph.graph().height}var graphNumAttrs=["nodesep","edgesep","ranksep","marginx","marginy"],graphDefaults={ranksep:50,edgesep:20,nodesep:50,rankdir:"tb"},graphAttrs=["acyclicer","ranker","rankdir","align"],nodeNumAttrs=["width","height"],nodeDefaults={width:0,height:0},edgeNumAttrs=["minlen","weight","width","height","labeloffset"],edgeDefaults={minlen:1,weight:1,width:0,height:0,labeloffset:10,labelpos:"r"},edgeAttrs=["labelpos"];function buildLayoutGraph(inputGraph){var g=new Graph({multigraph:true,compound:true}),graph=canonicalize(inputGraph.graph());g.setGraph(_.merge({},graphDefaults,selectNumberAttrs(graph,graphNumAttrs),_.pick(graph,graphAttrs)));_.each(inputGraph.nodes(),function(v){var node=canonicalize(inputGraph.node(v));g.setNode(v,_.defaults(selectNumberAttrs(node,nodeNumAttrs),nodeDefaults));g.setParent(v,inputGraph.parent(v))});_.each(inputGraph.edges(),function(e){var edge=canonicalize(inputGraph.edge(e));g.setEdge(e,_.merge({},edgeDefaults,selectNumberAttrs(edge,edgeNumAttrs),_.pick(edge,edgeAttrs)))});return g}function makeSpaceForEdgeLabels(g){var graph=g.graph();graph.ranksep/=2;_.each(g.edges(),function(e){var edge=g.edge(e);edge.minlen*=2;if(edge.labelpos.toLowerCase()!=="c"){if(graph.rankdir==="TB"||graph.rankdir==="BT"){edge.width+=edge.labeloffset}else{edge.height+=edge.labeloffset}}})}function injectEdgeLabelProxies(g){_.each(g.edges(),function(e){var edge=g.edge(e);if(edge.width&&edge.height){var v=g.node(e.v),w=g.node(e.w),label={rank:(w.rank-v.rank)/2+v.rank,e:e};util.addDummyNode(g,"edge-proxy",label,"_ep")}})}function assignRankMinMax(g){var maxRank=0;_.each(g.nodes(),function(v){var node=g.node(v);if(node.borderTop){node.minRank=g.node(node.borderTop).rank;node.maxRank=g.node(node.borderBottom).rank;maxRank=_.max(maxRank,node.maxRank)}});g.graph().maxRank=maxRank}function removeEdgeLabelProxies(g){_.each(g.nodes(),function(v){var node=g.node(v);if(node.dummy==="edge-proxy"){g.edge(node.e).labelRank=node.rank;g.removeNode(v)}})}function translateGraph(g){var minX=Number.POSITIVE_INFINITY,maxX=0,minY=Number.POSITIVE_INFINITY,maxY=0,graphLabel=g.graph(),marginX=graphLabel.marginx||0,marginY=graphLabel.marginy||0;function getExtremes(attrs){var x=attrs.x,y=attrs.y,w=attrs.width,h=attrs.height;minX=Math.min(minX,x-w/2);maxX=Math.max(maxX,x+w/2);minY=Math.min(minY,y-h/2);maxY=Math.max(maxY,y+h/2)}_.each(g.nodes(),function(v){getExtremes(g.node(v))});_.each(g.edges(),function(e){var edge=g.edge(e);if(_.has(edge,"x")){getExtremes(edge)}});minX-=marginX;minY-=marginY;_.each(g.nodes(),function(v){var node=g.node(v);node.x-=minX;node.y-=minY});_.each(g.edges(),function(e){var edge=g.edge(e);_.each(edge.points,function(p){p.x-=minX;p.y-=minY});if(_.has(edge,"x")){edge.x-=minX}if(_.has(edge,"y")){edge.y-=minY}});graphLabel.width=maxX-minX+marginX;graphLabel.height=maxY-minY+marginY}function assignNodeIntersects(g){_.each(g.edges(),function(e){var edge=g.edge(e),nodeV=g.node(e.v),nodeW=g.node(e.w),p1,p2;if(!edge.points){edge.points=[];p1=nodeW;p2=nodeV}else{p1=edge.points[0];p2=edge.points[edge.points.length-1]}edge.points.unshift(util.intersectRect(nodeV,p1));edge.points.push(util.intersectRect(nodeW,p2))})}function fixupEdgeLabelCoords(g){_.each(g.edges(),function(e){var edge=g.edge(e);if(_.has(edge,"x")){if(edge.labelpos==="l"||edge.labelpos==="r"){edge.width-=edge.labeloffset}switch(edge.labelpos){case"l":edge.x-=edge.width/2+edge.labeloffset;break;case"r":edge.x+=edge.width/2+edge.labeloffset;break}}})}function reversePointsForReversedEdges(g){_.each(g.edges(),function(e){var edge=g.edge(e);if(edge.reversed){edge.points.reverse()}})}function removeBorderNodes(g){_.each(g.nodes(),function(v){if(g.children(v).length){var node=g.node(v),t=g.node(node.borderTop),b=g.node(node.borderBottom),l=g.node(_.last(node.borderLeft)),r=g.node(_.last(node.borderRight));node.width=Math.abs(r.x-l.x);node.height=Math.abs(b.y-t.y);node.x=l.x+node.width/2;node.y=t.y+node.height/2}});_.each(g.nodes(),function(v){if(g.node(v).dummy==="border"){g.removeNode(v)}})}function removeSelfEdges(g){_.each(g.edges(),function(e){if(e.v===e.w){var node=g.node(e.v);if(!node.selfEdges){node.selfEdges=[]}node.selfEdges.push({e:e,label:g.edge(e)});g.removeEdge(e)}})}function insertSelfEdges(g){var layers=util.buildLayerMatrix(g);_.each(layers,function(layer){var orderShift=0;_.each(layer,function(v,i){var node=g.node(v);node.order=i+orderShift;_.each(node.selfEdges,function(selfEdge){util.addDummyNode(g,"selfedge",{width:selfEdge.label.width,height:selfEdge.label.height,rank:node.rank,order:i+ ++orderShift,e:selfEdge.e,label:selfEdge.label},"_se")});delete node.selfEdges})})}function positionSelfEdges(g){_.each(g.nodes(),function(v){var node=g.node(v);if(node.dummy==="selfedge"){var selfNode=g.node(node.e.v),x=selfNode.x+selfNode.width/2,y=selfNode.y,dx=node.x-x,dy=selfNode.height/2;g.setEdge(node.e,node.label);g.removeNode(v);node.label.points=[{x:x+2*dx/3,y:y-dy},{x:x+5*dx/6,y:y-dy},{x:x+dx,y:y},{x:x+5*dx/6,y:y+dy},{x:x+2*dx/3,y:y+dy}];node.label.x=node.x;node.label.y=node.y}})}function selectNumberAttrs(obj,attrs){return _.mapValues(_.pick(obj,attrs),Number)}function canonicalize(attrs){var newAttrs={};_.each(attrs,function(v,k){newAttrs[k.toLowerCase()]=v});return newAttrs}},{"./acyclic":28,"./add-border-segments":29,"./coordinate-system":30,"./graphlib":33,"./lodash":36,"./nesting-graph":37,"./normalize":38,"./order":43,"./parent-dummy-chains":48,"./position":50,"./rank":52,"./util":55}],36:[function(require,module,exports){module.exports=require(20)},{"/Users/andrew/Documents/dev/dagre-d3/lib/lodash.js":20,lodash:77}],37:[function(require,module,exports){var _=require("./lodash"),util=require("./util");module.exports={run:run,cleanup:cleanup};function run(g){var root=util.addDummyNode(g,"root",{},"_root"),depths=treeDepths(g),height=_.max(depths)-1,nodeSep=2*height+1;g.graph().nestingRoot=root;_.each(g.edges(),function(e){g.edge(e).minlen*=nodeSep});var weight=sumWeights(g)+1;_.each(g.children(),function(child){dfs(g,root,nodeSep,weight,height,depths,child)});g.graph().nodeRankFactor=nodeSep}function dfs(g,root,nodeSep,weight,height,depths,v){var children=g.children(v);if(!children.length){if(v!==root){g.setEdge(root,v,{weight:0,minlen:nodeSep})}return}var top=util.addBorderNode(g,"_bt"),bottom=util.addBorderNode(g,"_bb"),label=g.node(v);g.setParent(top,v);label.borderTop=top;g.setParent(bottom,v);label.borderBottom=bottom;_.each(children,function(child){dfs(g,root,nodeSep,weight,height,depths,child);var childNode=g.node(child),childTop=childNode.borderTop?childNode.borderTop:child,childBottom=childNode.borderBottom?childNode.borderBottom:child,thisWeight=childNode.borderTop?weight:2*weight,minlen=childTop!==childBottom?1:height-depths[v]+1;g.setEdge(top,childTop,{weight:thisWeight,minlen:minlen,nestingEdge:true});g.setEdge(childBottom,bottom,{weight:thisWeight,minlen:minlen,nestingEdge:true})});if(!g.parent(v)){g.setEdge(root,top,{weight:0,minlen:height+depths[v]})}}function treeDepths(g){var depths={};function dfs(v,depth){var children=g.children(v);if(children&&children.length){_.each(children,function(child){dfs(child,depth+1)})}depths[v]=depth}_.each(g.children(),function(v){dfs(v,1)});return depths}function sumWeights(g){return _.reduce(g.edges(),function(acc,e){return acc+g.edge(e).weight},0)}function cleanup(g){var graphLabel=g.graph();g.removeNode(graphLabel.nestingRoot);delete graphLabel.nestingRoot;_.each(g.edges(),function(e){var edge=g.edge(e);if(edge.nestingEdge){g.removeEdge(e)}})}},{"./lodash":36,"./util":55}],38:[function(require,module,exports){"use strict";var _=require("./lodash"),util=require("./util");module.exports={run:run,undo:undo};function run(g){g.graph().dummyChains=[];_.each(g.edges(),function(edge){normalizeEdge(g,edge)})}function normalizeEdge(g,e){var v=e.v,vRank=g.node(v).rank,w=e.w,wRank=g.node(w).rank,name=e.name,edgeLabel=g.edge(e),labelRank=edgeLabel.labelRank;if(wRank===vRank+1)return;g.removeEdge(e);var dummy,attrs,i;for(i=0,++vRank;vRank0){if(index%2){weightSum+=tree[index+1]}index=index-1>>1;tree[index]+=entry.weight}cc+=entry.weight*weightSum}));return cc}},{"../lodash":36}],43:[function(require,module,exports){"use strict";var _=require("../lodash"),initOrder=require("./init-order"),crossCount=require("./cross-count"),sortSubgraph=require("./sort-subgraph"),buildLayerGraph=require("./build-layer-graph"),addSubgraphConstraints=require("./add-subgraph-constraints"),Graph=require("../graphlib").Graph,util=require("../util");module.exports=order;function order(g){var maxRank=util.maxRank(g),downLayerGraphs=buildLayerGraphs(g,_.range(1,maxRank+1),"inEdges"),upLayerGraphs=buildLayerGraphs(g,_.range(maxRank-1,-1,-1),"outEdges");var layering=initOrder(g);assignOrder(g,layering);var bestCC=Number.POSITIVE_INFINITY,best;for(var i=0,lastBest=0;lastBest<4;++i,++lastBest){sweepLayerGraphs(i%2?downLayerGraphs:upLayerGraphs,i%4>=2);layering=util.buildLayerMatrix(g);var cc=crossCount(g,layering);if(cc=vEntry.barycenter){mergeEntries(vEntry,uEntry)}}}function handleOut(vEntry){return function(wEntry){wEntry["in"].push(vEntry);if(--wEntry.indegree===0){sourceSet.push(wEntry)}}}while(sourceSet.length){var entry=sourceSet.pop();entries.push(entry);_.each(entry["in"].reverse(),handleIn(entry));_.each(entry.out,handleOut(entry))}return _.chain(entries).filter(function(entry){return!entry.merged}).map(function(entry){return _.pick(entry,["vs","i","barycenter","weight"])}).value()}function mergeEntries(target,source){var sum=0,weight=0;if(target.weight){sum+=target.barycenter*target.weight;weight+=target.weight}if(source.weight){sum+=source.barycenter*source.weight;weight+=source.weight}target.vs=source.vs.concat(target.vs);target.barycenter=sum/weight;target.weight=weight;target.i=Math.min(source.i,target.i);source.merged=true}},{"../lodash":36}],46:[function(require,module,exports){var _=require("../lodash"),barycenter=require("./barycenter"),resolveConflicts=require("./resolve-conflicts"),sort=require("./sort");module.exports=sortSubgraph;function sortSubgraph(g,v,cg,biasRight){var movable=g.children(v),node=g.node(v),bl=node?node.borderLeft:undefined,br=node?node.borderRight:undefined,subgraphs={};if(bl){movable=_.filter(movable,function(w){return w!==bl&&w!==br})}var barycenters=barycenter(g,movable);_.each(barycenters,function(entry){if(g.children(entry.v).length){var subgraphResult=sortSubgraph(g,entry.v,cg,biasRight);subgraphs[entry.v]=subgraphResult;if(_.has(subgraphResult,"barycenter")){mergeBarycenters(entry,subgraphResult)}}});var entries=resolveConflicts(barycenters,cg);expandSubgraphs(entries,subgraphs);var result=sort(entries,biasRight);if(bl){result.vs=_.flatten([bl,result.vs,br],true);if(g.predecessors(bl).length){var blPred=g.node(g.predecessors(bl)[0]),brPred=g.node(g.predecessors(br)[0]);if(!_.has(result,"barycenter")){result.barycenter=0;result.weight=0}result.barycenter=(result.barycenter*result.weight+blPred.order+brPred.order)/(result.weight+2);result.weight+=2}}return result}function expandSubgraphs(entries,subgraphs){_.each(entries,function(entry){entry.vs=_.flatten(entry.vs.map(function(v){if(subgraphs[v]){return subgraphs[v].vs}return v}),true)})}function mergeBarycenters(target,other){if(!_.isUndefined(target.barycenter)){target.barycenter=(target.barycenter*target.weight+other.barycenter*other.weight)/(target.weight+other.weight);target.weight+=other.weight}else{target.barycenter=other.barycenter;target.weight=other.weight}}},{"../lodash":36,"./barycenter":40,"./resolve-conflicts":45,"./sort":47}],47:[function(require,module,exports){var _=require("../lodash"),util=require("../util");module.exports=sort;function sort(entries,biasRight){var parts=util.partition(entries,function(entry){return _.has(entry,"barycenter")});var sortable=parts.lhs,unsortable=_.sortBy(parts.rhs,function(entry){return-entry.i}),vs=[],sum=0,weight=0,vsIndex=0;sortable.sort(compareWithBias(!!biasRight));vsIndex=consumeUnsortable(vs,unsortable,vsIndex);_.each(sortable,function(entry){vsIndex+=entry.vs.length;vs.push(entry.vs);sum+=entry.barycenter*entry.weight;weight+=entry.weight;vsIndex=consumeUnsortable(vs,unsortable,vsIndex)});var result={vs:_.flatten(vs,true)};if(weight){result.barycenter=sum/weight;result.weight=weight}return result}function consumeUnsortable(vs,unsortable,index){var last;while(unsortable.length&&(last=_.last(unsortable)).i<=index){unsortable.pop();vs.push(last.vs);index++}return index}function compareWithBias(bias){return function(entryV,entryW){if(entryV.barycenterentryW.barycenter){return 1}return!bias?entryV.i-entryW.i:entryW.i-entryV.i}}},{"../lodash":36,"../util":55}],48:[function(require,module,exports){var _=require("./lodash");module.exports=parentDummyChains;function parentDummyChains(g){var postorderNums=postorder(g);_.each(g.graph().dummyChains,function(v){var node=g.node(v),edgeObj=node.edgeObj,pathData=findPath(g,postorderNums,edgeObj.v,edgeObj.w),path=pathData.path,lca=pathData.lca,pathIdx=0,pathV=path[pathIdx],ascending=true;while(v!==edgeObj.w){node=g.node(v);if(ascending){while((pathV=path[pathIdx])!==lca&&g.node(pathV).maxRanklow||lim>postorderNums[parent].lim));lca=parent;parent=w;while((parent=g.parent(parent))!==lca){wPath.push(parent)}return{path:vPath.concat(wPath.reverse()),lca:lca}}function postorder(g){var result={},lim=0;function dfs(v){var low=lim;_.each(g.children(v),dfs);result[v]={low:low,lim:lim++}}_.each(g.children(),dfs);return result}},{"./lodash":36}],49:[function(require,module,exports){"use strict";var _=require("../lodash"),Graph=require("../graphlib").Graph,util=require("../util");module.exports={positionX:positionX,findType1Conflicts:findType1Conflicts,findType2Conflicts:findType2Conflicts,addConflict:addConflict,hasConflict:hasConflict,verticalAlignment:verticalAlignment,horizontalCompaction:horizontalCompaction,alignCoordinates:alignCoordinates,findSmallestWidthAlignment:findSmallestWidthAlignment,balance:balance};function findType1Conflicts(g,layering){var conflicts={};function visitLayer(prevLayer,layer){var k0=0,scanPos=0,prevLayerLength=prevLayer.length,lastNode=_.last(layer);_.each(layer,function(v,i){var w=findOtherInnerSegmentNode(g,v),k1=w?g.node(w).order:prevLayerLength;if(w||v===lastNode){_.each(layer.slice(scanPos,i+1),function(scanNode){_.each(g.predecessors(scanNode),function(u){var uLabel=g.node(u),uPos=uLabel.order;if((uPosnextNorthBorder)){addConflict(conflicts,u,v)}})}})}function visitLayer(north,south){var prevNorthPos=-1,nextNorthPos,southPos=0;_.each(south,function(v,southLookahead){if(g.node(v).dummy==="border"){var predecessors=g.predecessors(v);if(predecessors.length){nextNorthPos=g.node(predecessors[0]).order;scan(south,southPos,southLookahead,prevNorthPos,nextNorthPos);southPos=southLookahead;prevNorthPos=nextNorthPos}}scan(south,southPos,south.length,nextNorthPos,north.length)});return south}_.reduce(layering,visitLayer);return conflicts}function findOtherInnerSegmentNode(g,v){if(g.node(v).dummy){return _.find(g.predecessors(v),function(u){return g.node(u).dummy})}}function addConflict(conflicts,v,w){if(v>w){var tmp=v;v=w;w=tmp}var conflictsV=conflicts[v];if(!conflictsV){conflicts[v]=conflictsV={}}conflictsV[w]=true}function hasConflict(conflicts,v,w){if(v>w){var tmp=v;v=w;w=tmp}return _.has(conflicts[v],w)}function verticalAlignment(g,layering,conflicts,neighborFn){var root={},align={},pos={};_.each(layering,function(layer){_.each(layer,function(v,order){root[v]=v;align[v]=v;pos[v]=order})});_.each(layering,function(layer){var prevIdx=-1;_.each(layer,function(v){var ws=neighborFn(v);if(ws.length){ws=_.sortBy(ws,function(w){return pos[w]});var mp=(ws.length-1)/2;for(var i=Math.floor(mp),il=Math.ceil(mp);i<=il;++i){var w=ws[i];if(align[v]===v&&prevIdx0){if(index%2){weightSum+=tree[index+1]}index=index-1>>1;tree[index]+=entry.weight}cc+=entry.weight*weightSum}));return cc}},{"../lodash":36}],43:[function(require,module,exports){"use strict";var _=require("../lodash"),initOrder=require("./init-order"),crossCount=require("./cross-count"),sortSubgraph=require("./sort-subgraph"),buildLayerGraph=require("./build-layer-graph"),addSubgraphConstraints=require("./add-subgraph-constraints"),Graph=require("../graphlib").Graph,util=require("../util");module.exports=order;function order(g){var maxRank=util.maxRank(g),downLayerGraphs=buildLayerGraphs(g,_.range(1,maxRank+1),"inEdges"),upLayerGraphs=buildLayerGraphs(g,_.range(maxRank-1,-1,-1),"outEdges");var layering=initOrder(g);assignOrder(g,layering);var bestCC=Number.POSITIVE_INFINITY,best;for(var i=0,lastBest=0;lastBest<4;++i,++lastBest){sweepLayerGraphs(i%2?downLayerGraphs:upLayerGraphs,i%4>=2);layering=util.buildLayerMatrix(g);var cc=crossCount(g,layering);if(cc=vEntry.barycenter){mergeEntries(vEntry,uEntry)}}}function handleOut(vEntry){return function(wEntry){wEntry["in"].push(vEntry);if(--wEntry.indegree===0){sourceSet.push(wEntry)}}}while(sourceSet.length){var entry=sourceSet.pop();entries.push(entry);_.each(entry["in"].reverse(),handleIn(entry));_.each(entry.out,handleOut(entry))}return _.chain(entries).filter(function(entry){return!entry.merged}).map(function(entry){return _.pick(entry,["vs","i","barycenter","weight"])}).value()}function mergeEntries(target,source){var sum=0,weight=0;if(target.weight){sum+=target.barycenter*target.weight;weight+=target.weight}if(source.weight){sum+=source.barycenter*source.weight;weight+=source.weight}target.vs=source.vs.concat(target.vs);target.barycenter=sum/weight;target.weight=weight;target.i=Math.min(source.i,target.i);source.merged=true}},{"../lodash":36}],46:[function(require,module,exports){var _=require("../lodash"),barycenter=require("./barycenter"),resolveConflicts=require("./resolve-conflicts"),sort=require("./sort");module.exports=sortSubgraph;function sortSubgraph(g,v,cg,biasRight){var movable=g.children(v),node=g.node(v),bl=node?node.borderLeft:undefined,br=node?node.borderRight:undefined,subgraphs={};if(bl){movable=_.filter(movable,function(w){return w!==bl&&w!==br})}var barycenters=barycenter(g,movable);_.each(barycenters,function(entry){if(g.children(entry.v).length){var subgraphResult=sortSubgraph(g,entry.v,cg,biasRight);subgraphs[entry.v]=subgraphResult;if(_.has(subgraphResult,"barycenter")){mergeBarycenters(entry,subgraphResult)}}});var entries=resolveConflicts(barycenters,cg);expandSubgraphs(entries,subgraphs);var result=sort(entries,biasRight);if(bl){result.vs=_.flatten([bl,result.vs,br],true);if(g.predecessors(bl).length){var blPred=g.node(g.predecessors(bl)[0]),brPred=g.node(g.predecessors(br)[0]);if(!_.has(result,"barycenter")){result.barycenter=0;result.weight=0}result.barycenter=(result.barycenter*result.weight+blPred.order+brPred.order)/(result.weight+2);result.weight+=2}}return result}function expandSubgraphs(entries,subgraphs){_.each(entries,function(entry){entry.vs=_.flatten(entry.vs.map(function(v){if(subgraphs[v]){return subgraphs[v].vs}return v}),true)})}function mergeBarycenters(target,other){if(!_.isUndefined(target.barycenter)){target.barycenter=(target.barycenter*target.weight+other.barycenter*other.weight)/(target.weight+other.weight);target.weight+=other.weight}else{target.barycenter=other.barycenter;target.weight=other.weight}}},{"../lodash":36,"./barycenter":40,"./resolve-conflicts":45,"./sort":47}],47:[function(require,module,exports){var _=require("../lodash"),util=require("../util");module.exports=sort;function sort(entries,biasRight){var parts=util.partition(entries,function(entry){return _.has(entry,"barycenter")});var sortable=parts.lhs,unsortable=_.sortBy(parts.rhs,function(entry){return-entry.i}),vs=[],sum=0,weight=0,vsIndex=0;sortable.sort(compareWithBias(!!biasRight));vsIndex=consumeUnsortable(vs,unsortable,vsIndex);_.each(sortable,function(entry){vsIndex+=entry.vs.length;vs.push(entry.vs);sum+=entry.barycenter*entry.weight;weight+=entry.weight;vsIndex=consumeUnsortable(vs,unsortable,vsIndex)});var result={vs:_.flatten(vs,true)};if(weight){result.barycenter=sum/weight;result.weight=weight}return result}function consumeUnsortable(vs,unsortable,index){var last;while(unsortable.length&&(last=_.last(unsortable)).i<=index){unsortable.pop();vs.push(last.vs);index++}return index}function compareWithBias(bias){return function(entryV,entryW){if(entryV.barycenterentryW.barycenter){return 1}return!bias?entryV.i-entryW.i:entryW.i-entryV.i}}},{"../lodash":36,"../util":55}],48:[function(require,module,exports){var _=require("./lodash");module.exports=parentDummyChains;function parentDummyChains(g){var postorderNums=postorder(g);_.each(g.graph().dummyChains,function(v){var node=g.node(v),edgeObj=node.edgeObj,pathData=findPath(g,postorderNums,edgeObj.v,edgeObj.w),path=pathData.path,lca=pathData.lca,pathIdx=0,pathV=path[pathIdx],ascending=true;while(v!==edgeObj.w){node=g.node(v);if(ascending){while((pathV=path[pathIdx])!==lca&&g.node(pathV).maxRanklow||lim>postorderNums[parent].lim));lca=parent;parent=w;while((parent=g.parent(parent))!==lca){wPath.push(parent)}return{path:vPath.concat(wPath.reverse()),lca:lca}}function postorder(g){var result={},lim=0;function dfs(v){var low=lim;_.each(g.children(v),dfs);result[v]={low:low,lim:lim++}}_.each(g.children(),dfs);return result}},{"./lodash":36}],49:[function(require,module,exports){"use strict";var _=require("../lodash"),Graph=require("../graphlib").Graph,util=require("../util");module.exports={positionX:positionX,findType1Conflicts:findType1Conflicts,findType2Conflicts:findType2Conflicts,addConflict:addConflict,hasConflict:hasConflict,verticalAlignment:verticalAlignment,horizontalCompaction:horizontalCompaction,alignCoordinates:alignCoordinates,findSmallestWidthAlignment:findSmallestWidthAlignment,balance:balance};function findType1Conflicts(g,layering){var conflicts={};function visitLayer(prevLayer,layer){var k0=0,scanPos=0,prevLayerLength=prevLayer.length,lastNode=_.last(layer);_.each(layer,function(v,i){var w=findOtherInnerSegmentNode(g,v),k1=w?g.node(w).order:prevLayerLength;if(w||v===lastNode){_.each(layer.slice(scanPos,i+1),function(scanNode){_.each(g.predecessors(scanNode),function(u){var uLabel=g.node(u),uPos=uLabel.order;if((uPosnextNorthBorder)){addConflict(conflicts,u,v)}})}})}function visitLayer(north,south){var prevNorthPos=-1,nextNorthPos,southPos=0;_.each(south,function(v,southLookahead){if(g.node(v).dummy==="border"){var predecessors=g.predecessors(v);if(predecessors.length){nextNorthPos=g.node(predecessors[0]).order;scan(south,southPos,southLookahead,prevNorthPos,nextNorthPos);southPos=southLookahead;prevNorthPos=nextNorthPos}}scan(south,southPos,south.length,nextNorthPos,north.length)});return south}_.reduce(layering,visitLayer);return conflicts}function findOtherInnerSegmentNode(g,v){if(g.node(v).dummy){return _.find(g.predecessors(v),function(u){return g.node(u).dummy})}}function addConflict(conflicts,v,w){if(v>w){var tmp=v;v=w;w=tmp}var conflictsV=conflicts[v];if(!conflictsV){conflicts[v]=conflictsV={}}conflictsV[w]=true}function hasConflict(conflicts,v,w){if(v>w){var tmp=v;v=w;w=tmp}return _.has(conflicts[v],w)}function verticalAlignment(g,layering,conflicts,neighborFn){var root={},align={},pos={};_.each(layering,function(layer){_.each(layer,function(v,order){root[v]=v;align[v]=v;pos[v]=order})});_.each(layering,function(layer){var prevIdx=-1;_.each(layer,function(v){var ws=neighborFn(v);if(ws.length){ws=_.sortBy(ws,function(w){return pos[w]});var mp=(ws.length-1)/2;for(var i=Math.floor(mp),il=Math.ceil(mp);i<=il;++i){var w=ws[i];if(align[v]===v&&prevIdxwLabel.lim){tailLabel=wLabel;flip=true}var candidates=_.filter(g.edges(),function(edge){return flip===isDescendant(t,t.node(edge.v),tailLabel)&&flip!==isDescendant(t,t.node(edge.w),tailLabel)});return _.min(candidates,function(edge){return slack(g,edge)})}function exchangeEdges(t,g,e,f){var v=e.v,w=e.w;t.removeEdge(v,w);t.setEdge(f.v,f.w,{});initLowLimValues(t);initCutValues(t,g);updateRanks(t,g)}function updateRanks(t,g){var root=_.find(t.nodes(),function(v){return!g.node(v).parent}),vs=preorder(t,root);vs=vs.slice(1);_.each(vs,function(v){var parent=t.node(v).parent,edge=g.edge(v,parent),flipped=false;if(!edge){edge=g.edge(parent,v);flipped=true}g.node(v).rank=g.node(parent).rank+(flipped?edge.minlen:-edge.minlen)})}function isTreeEdge(tree,u,v){return tree.hasEdge(u,v)}function isDescendant(tree,vLabel,rootLabel){return rootLabel.low<=vLabel.lim&&vLabel.lim<=rootLabel.lim}},{"../graphlib":33,"../lodash":36,"../util":55,"./feasible-tree":51,"./util":54}],54:[function(require,module,exports){"use strict";var _=require("../lodash");module.exports={longestPath:longestPath,slack:slack};function longestPath(g){var visited={};function dfs(v){var label=g.node(v);if(_.has(visited,v)){return label.rank}visited[v]=true;var rank=_.min(_.map(g.outEdges(v),function(e){return dfs(e.w)-g.edge(e).minlen}));if(rank===Number.POSITIVE_INFINITY){rank=0}return label.rank=rank}_.each(g.sources(),dfs)}function slack(g,e){return g.node(e.w).rank-g.node(e.v).rank-g.edge(e).minlen}},{"../lodash":36}],55:[function(require,module,exports){"use strict";var _=require("./lodash"),Graph=require("./graphlib").Graph;module.exports={addDummyNode:addDummyNode,simplify:simplify,asNonCompoundGraph:asNonCompoundGraph,successorWeights:successorWeights,predecessorWeights:predecessorWeights,intersectRect:intersectRect,buildLayerMatrix:buildLayerMatrix,normalizeRanks:normalizeRanks,removeEmptyRanks:removeEmptyRanks,addBorderNode:addBorderNode,maxRank:maxRank,partition:partition,time:time,notime:notime};function addDummyNode(g,type,attrs,name){var v; +if(!g.hasEdge(v,w)){v=edge.w;w=edge.v}var vLabel=t.node(v),wLabel=t.node(w),tailLabel=vLabel,flip=false;if(vLabel.lim>wLabel.lim){tailLabel=wLabel;flip=true}var candidates=_.filter(g.edges(),function(edge){return flip===isDescendant(t,t.node(edge.v),tailLabel)&&flip!==isDescendant(t,t.node(edge.w),tailLabel)});return _.min(candidates,function(edge){return slack(g,edge)})}function exchangeEdges(t,g,e,f){var v=e.v,w=e.w;t.removeEdge(v,w);t.setEdge(f.v,f.w,{});initLowLimValues(t);initCutValues(t,g);updateRanks(t,g)}function updateRanks(t,g){var root=_.find(t.nodes(),function(v){return!g.node(v).parent}),vs=preorder(t,root);vs=vs.slice(1);_.each(vs,function(v){var parent=t.node(v).parent,edge=g.edge(v,parent),flipped=false;if(!edge){edge=g.edge(parent,v);flipped=true}g.node(v).rank=g.node(parent).rank+(flipped?edge.minlen:-edge.minlen)})}function isTreeEdge(tree,u,v){return tree.hasEdge(u,v)}function isDescendant(tree,vLabel,rootLabel){return rootLabel.low<=vLabel.lim&&vLabel.lim<=rootLabel.lim}},{"../graphlib":33,"../lodash":36,"../util":55,"./feasible-tree":51,"./util":54}],54:[function(require,module,exports){"use strict";var _=require("../lodash");module.exports={longestPath:longestPath,slack:slack};function longestPath(g){var visited={};function dfs(v){var label=g.node(v);if(_.has(visited,v)){return label.rank}visited[v]=true;var rank=_.min(_.map(g.outEdges(v),function(e){return dfs(e.w)-g.edge(e).minlen}));if(rank===Number.POSITIVE_INFINITY){rank=0}return label.rank=rank}_.each(g.sources(),dfs)}function slack(g,e){return g.node(e.w).rank-g.node(e.v).rank-g.edge(e).minlen}},{"../lodash":36}],55:[function(require,module,exports){"use strict";var _=require("./lodash"),Graph=require("./graphlib").Graph;module.exports={addDummyNode:addDummyNode,simplify:simplify,asNonCompoundGraph:asNonCompoundGraph,successorWeights:successorWeights,predecessorWeights:predecessorWeights,intersectRect:intersectRect,buildLayerMatrix:buildLayerMatrix,normalizeRanks:normalizeRanks,removeEmptyRanks:removeEmptyRanks,addBorderNode:addBorderNode,maxRank:maxRank,partition:partition,time:time,notime:notime};function addDummyNode(g,type,attrs,name){var v;do{v=_.uniqueId(name)}while(g.hasNode(v));attrs.dummy=type;g.setNode(v,attrs);return v}function simplify(g){var simplified=(new Graph).setGraph(g.graph());_.each(g.nodes(),function(v){simplified.setNode(v,g.node(v))});_.each(g.edges(),function(e){var simpleLabel=simplified.edge(e.v,e.w)||{weight:0,minlen:1},label=g.edge(e);simplified.setEdge(e.v,e.w,{weight:simpleLabel.weight+label.weight,minlen:Math.max(simpleLabel.minlen,label.minlen)})});return simplified}function asNonCompoundGraph(g){var simplified=new Graph({multigraph:g.isMultigraph()}).setGraph(g.graph());_.each(g.nodes(),function(v){if(!g.children(v).length){simplified.setNode(v,g.node(v))}});_.each(g.edges(),function(e){simplified.setEdge(e,g.edge(e))});return simplified}function successorWeights(g){var weightMap=_.map(g.nodes(),function(v){var sucs={};_.each(g.outEdges(v),function(e){sucs[e.w]=(sucs[e.w]||0)+g.edge(e).weight});return sucs});return _.zipObject(g.nodes(),weightMap)}function predecessorWeights(g){var weightMap=_.map(g.nodes(),function(v){var preds={};_.each(g.inEdges(v),function(e){preds[e.v]=(preds[e.v]||0)+g.edge(e).weight});return preds});return _.zipObject(g.nodes(),weightMap)}function intersectRect(rect,point){var x=rect.x;var y=rect.y;var dx=point.x-x;var dy=point.y-y;var w=rect.width/2;var h=rect.height/2;if(!dx&&!dy){throw new Error("Not possible to find intersection inside of the rectangle")}var sx,sy;if(Math.abs(dy)*w>Math.abs(dx)*h){if(dy<0){h=-h}sx=h*dx/dy;sy=h}else{if(dx<0){w=-w}sx=w;sy=w*dy/dx}return{x:x+sx,y:y+sy}}function buildLayerMatrix(g){var layering=_.map(_.range(maxRank(g)+1),function(){return[]});_.each(g.nodes(),function(v){var node=g.node(v),rank=node.rank;if(!_.isUndefined(rank)){layering[rank][node.order]=v}});return layering}function normalizeRanks(g){var min=_.min(_.map(g.nodes(),function(v){return g.node(v).rank}));_.each(g.nodes(),function(v){var node=g.node(v);if(_.has(node,"rank")){node.rank-=min}})}function removeEmptyRanks(g){var offset=_.min(_.map(g.nodes(),function(v){return g.node(v).rank}));var layers=[];_.each(g.nodes(),function(v){var rank=g.node(v).rank-offset;if(!_.has(layers,rank)){layers[rank]=[]}layers[rank].push(v)});var delta=0,nodeRankFactor=g.graph().nodeRankFactor;_.each(layers,function(vs,i){if(_.isUndefined(vs)&&i%nodeRankFactor!==0){--delta}else if(delta){_.each(vs,function(v){g.node(v).rank+=delta})}})}function addBorderNode(g,prefix,rank,order){var node={width:0,height:0};if(arguments.length>=4){node.rank=rank;node.order=order}return addDummyNode(g,"border",node,prefix)}function maxRank(g){return _.max(_.map(g.nodes(),function(v){var rank=g.node(v).rank;if(!_.isUndefined(rank)){return rank}}))}function partition(collection,fn){var result={lhs:[],rhs:[]};_.each(collection,function(value){if(fn(value)){result.lhs.push(value)}else{result.rhs.push(value)}});return result}function time(name,fn){var start=_.now();try{return fn()}finally{console.log(name+" time: "+(_.now()-start)+"ms")}}function notime(name,fn){return fn()}},{"./graphlib":33,"./lodash":36}],56:[function(require,module,exports){module.exports="0.7.1"},{}],57:[function(require,module,exports){var lib=require("./lib");module.exports={Graph:lib.Graph,json:require("./lib/json"),alg:require("./lib/alg"),version:lib.version}},{"./lib":73,"./lib/alg":64,"./lib/json":74}],58:[function(require,module,exports){var _=require("../lodash");module.exports=components;function components(g){var visited={},cmpts=[],cmpt;function dfs(v){if(_.has(visited,v))return;visited[v]=true;cmpt.push(v);_.each(g.successors(v),dfs);_.each(g.predecessors(v),dfs)}_.each(g.nodes(),function(v){cmpt=[];dfs(v);if(cmpt.length){cmpts.push(cmpt)}});return cmpts}},{"../lodash":75}],59:[function(require,module,exports){var _=require("../lodash");module.exports=dfs;function dfs(g,vs,order){if(!_.isArray(vs)){vs=[vs]}var acc=[],visited={};_.each(vs,function(v){if(!g.hasNode(v)){throw new Error("Graph does not have node: "+v)}doDfs(g,v,order==="post",visited,acc)});return acc}function doDfs(g,v,postorder,visited,acc){if(!_.has(visited,v)){visited[v]=true;if(!postorder){acc.push(v)}_.each(g.neighbors(v),function(w){doDfs(g,w,postorder,visited,acc)});if(postorder){acc.push(v)}}}},{"../lodash":75}],60:[function(require,module,exports){var dijkstra=require("./dijkstra"),_=require("../lodash");module.exports=dijkstraAll;function dijkstraAll(g,weightFunc,edgeFunc){return _.transform(g.nodes(),function(acc,v){acc[v]=dijkstra(g,v,weightFunc,edgeFunc)},{})}},{"../lodash":75,"./dijkstra":61}],61:[function(require,module,exports){var _=require("../lodash"),PriorityQueue=require("../data/priority-queue");module.exports=dijkstra;var DEFAULT_WEIGHT_FUNC=_.constant(1);function dijkstra(g,source,weightFn,edgeFn){return runDijkstra(g,String(source),weightFn||DEFAULT_WEIGHT_FUNC,edgeFn||function(v){return g.outEdges(v)})}function runDijkstra(g,source,weightFn,edgeFn){var results={},pq=new PriorityQueue,v,vEntry;var updateNeighbors=function(edge){var w=edge.v!==v?edge.v:edge.w,wEntry=results[w],weight=weightFn(edge),distance=vEntry.distance+weight;if(weight<0){throw new Error("dijkstra does not allow negative edge weights. "+"Bad edge: "+edge+" Weight: "+weight)}if(distance0){v=pq.removeMin();vEntry=results[v];if(vEntry.distance===Number.POSITIVE_INFINITY){break}edgeFn(v).forEach(updateNeighbors)}return results}},{"../data/priority-queue":71,"../lodash":75}],62:[function(require,module,exports){var _=require("../lodash"),tarjan=require("./tarjan");module.exports=findCycles;function findCycles(g){return _.filter(tarjan(g),function(cmpt){return cmpt.length>1})}},{"../lodash":75,"./tarjan":69}],63:[function(require,module,exports){var _=require("../lodash");module.exports=floydWarshall;var DEFAULT_WEIGHT_FUNC=_.constant(1);function floydWarshall(g,weightFn,edgeFn){return runFloydWarshall(g,weightFn||DEFAULT_WEIGHT_FUNC,edgeFn||function(v){return g.outEdges(v)})}function runFloydWarshall(g,weightFn,edgeFn){var results={},nodes=g.nodes();nodes.forEach(function(v){results[v]={};results[v][v]={distance:0};nodes.forEach(function(w){if(v!==w){results[v][w]={distance:Number.POSITIVE_INFINITY}}});edgeFn(v).forEach(function(edge){var w=edge.v===v?edge.w:edge.v,d=weightFn(edge);results[v][w]={distance:d,predecessor:v}})});nodes.forEach(function(k){var rowK=results[k];nodes.forEach(function(i){var rowI=results[i];nodes.forEach(function(j){var ik=rowI[k];var kj=rowK[j];var ij=rowI[j];var altDistance=ik.distance+kj.distance;if(altDistance0){v=pq.removeMin();if(_.has(parents,v)){result.setEdge(v,parents[v])}else if(init){throw new Error("Input graph is not connected: "+g)}else{init=true}g.nodeEdges(v).forEach(updateNeighbors)}return result}},{"../data/priority-queue":71,"../graph":72,"../lodash":75}],69:[function(require,module,exports){var _=require("../lodash");module.exports=tarjan;function tarjan(g){var index=0,stack=[],visited={},results=[];function dfs(v){var entry=visited[v]={onStack:true,lowlink:index,index:index++};stack.push(v);g.successors(v).forEach(function(w){if(!_.has(visited,w)){dfs(w);entry.lowlink=Math.min(entry.lowlink,visited[w].lowlink)}else if(visited[w].onStack){entry.lowlink=Math.min(entry.lowlink,visited[w].index)}});if(entry.lowlink===entry.index){var cmpt=[],w;do{w=stack.pop();visited[w].onStack=false;cmpt.push(w)}while(v!==w);results.push(cmpt)}}g.nodes().forEach(function(v){if(!_.has(visited,v)){dfs(v)}});return results}},{"../lodash":75}],70:[function(require,module,exports){var _=require("../lodash");module.exports=topsort;topsort.CycleException=CycleException;function topsort(g){var visited={},stack={},results=[];function visit(node){if(_.has(stack,node)){throw new CycleException}if(!_.has(visited,node)){stack[node]=true;visited[node]=true;_.each(g.predecessors(node),visit);delete stack[node];results.push(node)}}_.each(g.sinks(),visit);if(_.size(visited)!==g.nodeCount()){throw new CycleException}return results}function CycleException(){}},{"../lodash":75}],71:[function(require,module,exports){var _=require("../lodash");module.exports=PriorityQueue;function PriorityQueue(){this._arr=[];this._keyIndices={}}PriorityQueue.prototype.size=function(){return this._arr.length};PriorityQueue.prototype.keys=function(){return this._arr.map(function(x){return x.key})};PriorityQueue.prototype.has=function(key){return _.has(this._keyIndices,key)};PriorityQueue.prototype.priority=function(key){var index=this._keyIndices[key];if(index!==undefined){return this._arr[index].priority}};PriorityQueue.prototype.min=function(){if(this.size()===0){throw new Error("Queue underflow")}return this._arr[0].key};PriorityQueue.prototype.add=function(key,priority){var keyIndices=this._keyIndices;key=String(key);if(!_.has(keyIndices,key)){var arr=this._arr;var index=arr.length;keyIndices[key]=index;arr.push({key:key,priority:priority});this._decrease(index);return true}return false};PriorityQueue.prototype.removeMin=function(){this._swap(0,this._arr.length-1);var min=this._arr.pop();delete this._keyIndices[min.key];this._heapify(0);return min.key};PriorityQueue.prototype.decrease=function(key,priority){var index=this._keyIndices[key];if(priority>this._arr[index].priority){throw new Error("New priority is greater than current priority. "+"Key: "+key+" Old: "+this._arr[index].priority+" New: "+priority)}this._arr[index].priority=priority;this._decrease(index)};PriorityQueue.prototype._heapify=function(i){var arr=this._arr;var l=2*i,r=l+1,largest=i;if(l>1;if(arr[parent].priority1){this.setNode(v,value)}else{this.setNode(v)}},this);return this};Graph.prototype.setNode=function(v,value){if(_.has(this._nodes,v)){if(arguments.length>1){this._nodes[v]=value}return this}this._nodes[v]=arguments.length>1?value:this._defaultNodeLabelFn(v);if(this._isCompound){this._parent[v]=GRAPH_NODE;this._children[v]={};this._children[GRAPH_NODE][v]=true}this._in[v]={};this._preds[v]={};this._out[v]={};this._sucs[v]={};++this._nodeCount;return this};Graph.prototype.node=function(v){return this._nodes[v]};Graph.prototype.hasNode=function(v){return _.has(this._nodes,v)};Graph.prototype.removeNode=function(v){var self=this;if(_.has(this._nodes,v)){var removeEdge=function(e){self.removeEdge(self._edgeObjs[e])};delete this._nodes[v];if(this._isCompound){this._removeFromParentsChildList(v);delete this._parent[v];_.each(this.children(v),function(child){this.setParent(child)},this);delete this._children[v]}_.each(_.keys(this._in[v]),removeEdge);delete this._in[v];delete this._preds[v];_.each(_.keys(this._out[v]),removeEdge);delete this._out[v];delete this._sucs[v];--this._nodeCount}return this};Graph.prototype.setParent=function(v,parent){if(!this._isCompound){throw new Error("Cannot set parent in a non-compound graph")}if(_.isUndefined(parent)){parent=GRAPH_NODE}else{for(var ancestor=parent;!_.isUndefined(ancestor);ancestor=this.parent(ancestor)){if(ancestor===v){throw new Error("Setting "+parent+" as parent of "+v+" would create create a cycle")}}this.setNode(parent)}this.setNode(v);this._removeFromParentsChildList(v);this._parent[v]=parent;this._children[parent][v]=true;return this};Graph.prototype._removeFromParentsChildList=function(v){delete this._children[this._parent[v]][v]};Graph.prototype.parent=function(v){if(this._isCompound){var parent=this._parent[v];if(parent!==GRAPH_NODE){return parent}}};Graph.prototype.children=function(v){if(_.isUndefined(v)){v=GRAPH_NODE}if(this._isCompound){var children=this._children[v];if(children){return _.keys(children)}}else if(v===GRAPH_NODE){return this.nodes()}else if(this.hasNode(v)){return[]}};Graph.prototype.predecessors=function(v){var predsV=this._preds[v];if(predsV){return _.keys(predsV)}};Graph.prototype.successors=function(v){var sucsV=this._sucs[v];if(sucsV){return _.keys(sucsV)}};Graph.prototype.neighbors=function(v){var preds=this.predecessors(v);if(preds){return _.union(preds,this.successors(v))}};Graph.prototype.setDefaultEdgeLabel=function(newDefault){if(!_.isFunction(newDefault)){newDefault=_.constant(newDefault)}this._defaultEdgeLabelFn=newDefault;return this};Graph.prototype.edgeCount=function(){return this._edgeCount};Graph.prototype.edges=function(){return _.values(this._edgeObjs)};Graph.prototype.setPath=function(vs,value){var self=this,args=arguments;_.reduce(vs,function(v,w){if(args.length>1){self.setEdge(v,w,value)}else{self.setEdge(v,w)}return w});return this};Graph.prototype.setEdge=function(){var v,w,name,value,valueSpecified=false;if(_.isPlainObject(arguments[0])){v=arguments[0].v;w=arguments[0].w;name=arguments[0].name;if(arguments.length===2){value=arguments[1];valueSpecified=true}}else{v=arguments[0];w=arguments[1];name=arguments[3];if(arguments.length>2){value=arguments[2];valueSpecified=true}}v=""+v;w=""+w;if(!_.isUndefined(name)){name=""+name}var e=edgeArgsToId(this._isDirected,v,w,name);if(_.has(this._edgeLabels,e)){if(valueSpecified){this._edgeLabels[e]=value}return this}if(!_.isUndefined(name)&&!this._isMultigraph){throw new Error("Cannot set a named edge when isMultigraph = false")}this.setNode(v);this.setNode(w);this._edgeLabels[e]=valueSpecified?value:this._defaultEdgeLabelFn(v,w,name);var edgeObj=edgeArgsToObj(this._isDirected,v,w,name);v=edgeObj.v;w=edgeObj.w;Object.freeze(edgeObj);this._edgeObjs[e]=edgeObj;incrementOrInitEntry(this._preds[w],v);incrementOrInitEntry(this._sucs[v],w);this._in[w][e]=edgeObj;this._out[v][e]=edgeObj;this._edgeCount++;return this};Graph.prototype.edge=function(v,w,name){var e=arguments.length===1?edgeObjToId(this._isDirected,arguments[0]):edgeArgsToId(this._isDirected,v,w,name);return this._edgeLabels[e]};Graph.prototype.hasEdge=function(v,w,name){var e=arguments.length===1?edgeObjToId(this._isDirected,arguments[0]):edgeArgsToId(this._isDirected,v,w,name);return _.has(this._edgeLabels,e)};Graph.prototype.removeEdge=function(v,w,name){var e=arguments.length===1?edgeObjToId(this._isDirected,arguments[0]):edgeArgsToId(this._isDirected,v,w,name),edge=this._edgeObjs[e];if(edge){v=edge.v;w=edge.w;delete this._edgeLabels[e];delete this._edgeObjs[e];decrementOrRemoveEntry(this._preds[w],v);decrementOrRemoveEntry(this._sucs[v],w);delete this._in[w][e];delete this._out[v][e];this._edgeCount--}return this};Graph.prototype.inEdges=function(v,u){var inV=this._in[v];if(inV){var edges=_.values(inV);if(!u){return edges}return _.filter(edges,function(edge){return edge.v===u})}};Graph.prototype.outEdges=function(v,w){var outV=this._out[v];if(outV){var edges=_.values(outV);if(!w){return edges}return _.filter(edges,function(edge){return edge.w===w})}};Graph.prototype.nodeEdges=function(v,w){var inEdges=this.inEdges(v,w);if(inEdges){return inEdges.concat(this.outEdges(v,w))}};function incrementOrInitEntry(map,k){if(_.has(map,k)){map[k]++}else{map[k]=1}}function decrementOrRemoveEntry(map,k){if(!--map[k]){delete map[k]}}function edgeArgsToId(isDirected,v,w,name){if(!isDirected&&v>w){var tmp=v;v=w;w=tmp}return v+EDGE_KEY_DELIM+w+EDGE_KEY_DELIM+(_.isUndefined(name)?DEFAULT_EDGE_NAME:name)}function edgeArgsToObj(isDirected,v,w,name){if(!isDirected&&v>w){var tmp=v;v=w;w=tmp}var edgeObj={v:v,w:w};if(name){edgeObj.name=name}return edgeObj}function edgeObjToId(isDirected,edgeObj){return edgeArgsToId(isDirected,edgeObj.v,edgeObj.w,edgeObj.name)}},{"./lodash":75}],73:[function(require,module,exports){module.exports={Graph:require("./graph"),version:require("./version")}},{"./graph":72,"./version":76}],74:[function(require,module,exports){var _=require("./lodash"),Graph=require("./graph");module.exports={write:write,read:read};function write(g){var json={options:{directed:g.isDirected(),multigraph:g.isMultigraph(),compound:g.isCompound()},nodes:writeNodes(g),edges:writeEdges(g)};if(!_.isUndefined(g.graph())){json.value=_.clone(g.graph())}return json}function writeNodes(g){return _.map(g.nodes(),function(v){var nodeValue=g.node(v),parent=g.parent(v),node={v:v};if(!_.isUndefined(nodeValue)){node.value=nodeValue}if(!_.isUndefined(parent)){node.parent=parent}return node})}function writeEdges(g){return _.map(g.edges(),function(e){var edgeValue=g.edge(e),edge={v:e.v,w:e.w};if(!_.isUndefined(e.name)){edge.name=e.name}if(!_.isUndefined(edgeValue)){edge.value=edgeValue}return edge})}function read(json){var g=new Graph(json.options).setGraph(json.value);_.each(json.nodes,function(entry){g.setNode(entry.v,entry.value);if(entry.parent){g.setParent(entry.v,entry.parent)}});_.each(json.edges,function(entry){g.setEdge({v:entry.v,w:entry.w,name:entry.name},entry.value)});return g}},{"./graph":72,"./lodash":75}],75:[function(require,module,exports){module.exports=require(20)},{"/Users/andrew/Documents/dev/dagre-d3/lib/lodash.js":20,lodash:77}],76:[function(require,module,exports){module.exports="1.0.1"},{}],77:[function(require,module,exports){(function(global){(function(){var undefined;var arrayPool=[],objectPool=[];var idCounter=0;var keyPrefix=+new Date+"";var largeArraySize=75;var maxPoolSize=40;var whitespace=" \f \ufeff"+"\n\r\u2028\u2029"+" ᠎             　";var reEmptyStringLeading=/\b__p \+= '';/g,reEmptyStringMiddle=/\b(__p \+=) '' \+/g,reEmptyStringTrailing=/(__e\(.*?\)|\b__t\)) \+\n'';/g;var reEsTemplate=/\$\{([^\\}]*(?:\\.[^\\}]*)*)\}/g;var reFlags=/\w*$/;var reFuncName=/^\s*function[ \n\r\t]+\w/;var reInterpolate=/<%=([\s\S]+?)%>/g;var reLeadingSpacesAndZeros=RegExp("^["+whitespace+"]*0+(?=.$)");var reNoMatch=/($^)/;var reThis=/\bthis\b/;var reUnescapedString=/['\n\r\t\u2028\u2029\\]/g;var contextProps=["Array","Boolean","Date","Function","Math","Number","Object","RegExp","String","_","attachEvent","clearTimeout","isFinite","isNaN","parseInt","setTimeout"];var templateCounter=0;var argsClass="[object Arguments]",arrayClass="[object Array]",boolClass="[object Boolean]",dateClass="[object Date]",funcClass="[object Function]",numberClass="[object Number]",objectClass="[object Object]",regexpClass="[object RegExp]",stringClass="[object String]";var cloneableClasses={};cloneableClasses[funcClass]=false;cloneableClasses[argsClass]=cloneableClasses[arrayClass]=cloneableClasses[boolClass]=cloneableClasses[dateClass]=cloneableClasses[numberClass]=cloneableClasses[objectClass]=cloneableClasses[regexpClass]=cloneableClasses[stringClass]=true;var debounceOptions={leading:false,maxWait:0,trailing:false};var descriptor={configurable:false,enumerable:false,value:null,writable:false};var objectTypes={"boolean":false,"function":true,object:true,number:false,string:false,undefined:false};var stringEscapes={"\\":"\\","'":"'","\n":"n","\r":"r"," ":"t","\u2028":"u2028","\u2029":"u2029"};var root=objectTypes[typeof window]&&window||this;var freeExports=objectTypes[typeof exports]&&exports&&!exports.nodeType&&exports;var freeModule=objectTypes[typeof module]&&module&&!module.nodeType&&module;var moduleExports=freeModule&&freeModule.exports===freeExports&&freeExports;var freeGlobal=objectTypes[typeof global]&&global;if(freeGlobal&&(freeGlobal.global===freeGlobal||freeGlobal.window===freeGlobal)){root=freeGlobal}function baseIndexOf(array,value,fromIndex){var index=(fromIndex||0)-1,length=array?array.length:0;while(++index-1?0:-1:cache?0:-1}function cachePush(value){var cache=this.cache,type=typeof value;if(type=="boolean"||value==null){cache[value]=true}else{if(type!="number"&&type!="string"){type="object"}var key=type=="number"?value:keyPrefix+value,typeCache=cache[type]||(cache[type]={});if(type=="object"){(typeCache[key]||(typeCache[key]=[])).push(value)}else{typeCache[key]=true}}}function charAtCallback(value){return value.charCodeAt(0)}function compareAscending(a,b){var ac=a.criteria,bc=b.criteria,index=-1,length=ac.length;while(++indexother||typeof value=="undefined"){return 1}if(value/g,evaluate:/<%([\s\S]+?)%>/g,interpolate:reInterpolate,variable:"",imports:{_:lodash}};function baseBind(bindData){var func=bindData[0],partialArgs=bindData[2],thisArg=bindData[4];function bound(){if(partialArgs){var args=slice(partialArgs);push.apply(args,arguments)}if(this instanceof bound){var thisBinding=baseCreate(func.prototype),result=func.apply(thisBinding,args||arguments);return isObject(result)?result:thisBinding}return func.apply(thisArg,args||arguments)}setBindData(bound,bindData);return bound}function baseClone(value,isDeep,callback,stackA,stackB){if(callback){var result=callback(value);if(typeof result!="undefined"){return result}}var isObj=isObject(value);if(isObj){var className=toString.call(value);if(!cloneableClasses[className]){return value}var ctor=ctorByClass[className];switch(className){case boolClass:case dateClass:return new ctor(+value);case numberClass:case stringClass:return new ctor(value);case regexpClass:result=ctor(value.source,reFlags.exec(value));result.lastIndex=value.lastIndex;return result}}else{return value}var isArr=isArray(value);if(isDeep){var initedStack=!stackA;stackA||(stackA=getArray());stackB||(stackB=getArray());var length=stackA.length;while(length--){if(stackA[length]==value){return stackB[length]}}result=isArr?ctor(value.length):{}}else{result=isArr?slice(value):assign({},value)}if(isArr){if(hasOwnProperty.call(value,"index")){result.index=value.index}if(hasOwnProperty.call(value,"input")){result.input=value.input}}if(!isDeep){return result}stackA.push(value);stackB.push(result);(isArr?forEach:forOwn)(value,function(objValue,key){result[key]=baseClone(objValue,isDeep,callback,stackA,stackB)});if(initedStack){releaseArray(stackA);releaseArray(stackB)}return result}function baseCreate(prototype,properties){return isObject(prototype)?nativeCreate(prototype):{}; -do{v=_.uniqueId(name)}while(g.hasNode(v));attrs.dummy=type;g.setNode(v,attrs);return v}function simplify(g){var simplified=(new Graph).setGraph(g.graph());_.each(g.nodes(),function(v){simplified.setNode(v,g.node(v))});_.each(g.edges(),function(e){var simpleLabel=simplified.edge(e.v,e.w)||{weight:0,minlen:1},label=g.edge(e);simplified.setEdge(e.v,e.w,{weight:simpleLabel.weight+label.weight,minlen:Math.max(simpleLabel.minlen,label.minlen)})});return simplified}function asNonCompoundGraph(g){var simplified=new Graph({multigraph:g.isMultigraph()}).setGraph(g.graph());_.each(g.nodes(),function(v){if(!g.children(v).length){simplified.setNode(v,g.node(v))}});_.each(g.edges(),function(e){simplified.setEdge(e,g.edge(e))});return simplified}function successorWeights(g){var weightMap=_.map(g.nodes(),function(v){var sucs={};_.each(g.outEdges(v),function(e){sucs[e.w]=(sucs[e.w]||0)+g.edge(e).weight});return sucs});return _.zipObject(g.nodes(),weightMap)}function predecessorWeights(g){var weightMap=_.map(g.nodes(),function(v){var preds={};_.each(g.inEdges(v),function(e){preds[e.v]=(preds[e.v]||0)+g.edge(e).weight});return preds});return _.zipObject(g.nodes(),weightMap)}function intersectRect(rect,point){var x=rect.x;var y=rect.y;var dx=point.x-x;var dy=point.y-y;var w=rect.width/2;var h=rect.height/2;if(!dx&&!dy){throw new Error("Not possible to find intersection inside of the rectangle")}var sx,sy;if(Math.abs(dy)*w>Math.abs(dx)*h){if(dy<0){h=-h}sx=h*dx/dy;sy=h}else{if(dx<0){w=-w}sx=w;sy=w*dy/dx}return{x:x+sx,y:y+sy}}function buildLayerMatrix(g){var layering=_.map(_.range(maxRank(g)+1),function(){return[]});_.each(g.nodes(),function(v){var node=g.node(v),rank=node.rank;if(!_.isUndefined(rank)){layering[rank][node.order]=v}});return layering}function normalizeRanks(g){var min=_.min(_.map(g.nodes(),function(v){return g.node(v).rank}));_.each(g.nodes(),function(v){var node=g.node(v);if(_.has(node,"rank")){node.rank-=min}})}function removeEmptyRanks(g){var offset=_.min(_.map(g.nodes(),function(v){return g.node(v).rank}));var layers=[];_.each(g.nodes(),function(v){var rank=g.node(v).rank-offset;if(!_.has(layers,rank)){layers[rank]=[]}layers[rank].push(v)});var delta=0,nodeRankFactor=g.graph().nodeRankFactor;_.each(layers,function(vs,i){if(_.isUndefined(vs)&&i%nodeRankFactor!==0){--delta}else if(delta){_.each(vs,function(v){g.node(v).rank+=delta})}})}function addBorderNode(g,prefix,rank,order){var node={width:0,height:0};if(arguments.length>=4){node.rank=rank;node.order=order}return addDummyNode(g,"border",node,prefix)}function maxRank(g){return _.max(_.map(g.nodes(),function(v){var rank=g.node(v).rank;if(!_.isUndefined(rank)){return rank}}))}function partition(collection,fn){var result={lhs:[],rhs:[]};_.each(collection,function(value){if(fn(value)){result.lhs.push(value)}else{result.rhs.push(value)}});return result}function time(name,fn){var start=_.now();try{return fn()}finally{console.log(name+" time: "+(_.now()-start)+"ms")}}function notime(name,fn){return fn()}},{"./graphlib":33,"./lodash":36}],56:[function(require,module,exports){module.exports="0.7.1"},{}],57:[function(require,module,exports){var lib=require("./lib");module.exports={Graph:lib.Graph,json:require("./lib/json"),alg:require("./lib/alg"),version:lib.version}},{"./lib":73,"./lib/alg":64,"./lib/json":74}],58:[function(require,module,exports){var _=require("../lodash");module.exports=components;function components(g){var visited={},cmpts=[],cmpt;function dfs(v){if(_.has(visited,v))return;visited[v]=true;cmpt.push(v);_.each(g.successors(v),dfs);_.each(g.predecessors(v),dfs)}_.each(g.nodes(),function(v){cmpt=[];dfs(v);if(cmpt.length){cmpts.push(cmpt)}});return cmpts}},{"../lodash":75}],59:[function(require,module,exports){var _=require("../lodash");module.exports=dfs;function dfs(g,vs,order){if(!_.isArray(vs)){vs=[vs]}var acc=[],visited={};_.each(vs,function(v){if(!g.hasNode(v)){throw new Error("Graph does not have node: "+v)}doDfs(g,v,order==="post",visited,acc)});return acc}function doDfs(g,v,postorder,visited,acc){if(!_.has(visited,v)){visited[v]=true;if(!postorder){acc.push(v)}_.each(g.neighbors(v),function(w){doDfs(g,w,postorder,visited,acc)});if(postorder){acc.push(v)}}}},{"../lodash":75}],60:[function(require,module,exports){var dijkstra=require("./dijkstra"),_=require("../lodash");module.exports=dijkstraAll;function dijkstraAll(g,weightFunc,edgeFunc){return _.transform(g.nodes(),function(acc,v){acc[v]=dijkstra(g,v,weightFunc,edgeFunc)},{})}},{"../lodash":75,"./dijkstra":61}],61:[function(require,module,exports){var _=require("../lodash"),PriorityQueue=require("../data/priority-queue");module.exports=dijkstra;var DEFAULT_WEIGHT_FUNC=_.constant(1);function dijkstra(g,source,weightFn,edgeFn){return runDijkstra(g,String(source),weightFn||DEFAULT_WEIGHT_FUNC,edgeFn||function(v){return g.outEdges(v)})}function runDijkstra(g,source,weightFn,edgeFn){var results={},pq=new PriorityQueue,v,vEntry;var updateNeighbors=function(edge){var w=edge.v!==v?edge.v:edge.w,wEntry=results[w],weight=weightFn(edge),distance=vEntry.distance+weight;if(weight<0){throw new Error("dijkstra does not allow negative edge weights. "+"Bad edge: "+edge+" Weight: "+weight)}if(distance0){v=pq.removeMin();vEntry=results[v];if(vEntry.distance===Number.POSITIVE_INFINITY){break}edgeFn(v).forEach(updateNeighbors)}return results}},{"../data/priority-queue":71,"../lodash":75}],62:[function(require,module,exports){var _=require("../lodash"),tarjan=require("./tarjan");module.exports=findCycles;function findCycles(g){return _.filter(tarjan(g),function(cmpt){return cmpt.length>1})}},{"../lodash":75,"./tarjan":69}],63:[function(require,module,exports){var _=require("../lodash");module.exports=floydWarshall;var DEFAULT_WEIGHT_FUNC=_.constant(1);function floydWarshall(g,weightFn,edgeFn){return runFloydWarshall(g,weightFn||DEFAULT_WEIGHT_FUNC,edgeFn||function(v){return g.outEdges(v)})}function runFloydWarshall(g,weightFn,edgeFn){var results={},nodes=g.nodes();nodes.forEach(function(v){results[v]={};results[v][v]={distance:0};nodes.forEach(function(w){if(v!==w){results[v][w]={distance:Number.POSITIVE_INFINITY}}});edgeFn(v).forEach(function(edge){var w=edge.v===v?edge.w:edge.v,d=weightFn(edge);results[v][w]={distance:d,predecessor:v}})});nodes.forEach(function(k){var rowK=results[k];nodes.forEach(function(i){var rowI=results[i];nodes.forEach(function(j){var ik=rowI[k];var kj=rowK[j];var ij=rowI[j];var altDistance=ik.distance+kj.distance;if(altDistance0){v=pq.removeMin();if(_.has(parents,v)){result.setEdge(v,parents[v])}else if(init){throw new Error("Input graph is not connected: "+g)}else{init=true}g.nodeEdges(v).forEach(updateNeighbors)}return result}},{"../data/priority-queue":71,"../graph":72,"../lodash":75}],69:[function(require,module,exports){var _=require("../lodash");module.exports=tarjan;function tarjan(g){var index=0,stack=[],visited={},results=[];function dfs(v){var entry=visited[v]={onStack:true,lowlink:index,index:index++};stack.push(v);g.successors(v).forEach(function(w){if(!_.has(visited,w)){dfs(w);entry.lowlink=Math.min(entry.lowlink,visited[w].lowlink)}else if(visited[w].onStack){entry.lowlink=Math.min(entry.lowlink,visited[w].index)}});if(entry.lowlink===entry.index){var cmpt=[],w;do{w=stack.pop();visited[w].onStack=false;cmpt.push(w)}while(v!==w);results.push(cmpt)}}g.nodes().forEach(function(v){if(!_.has(visited,v)){dfs(v)}});return results}},{"../lodash":75}],70:[function(require,module,exports){var _=require("../lodash");module.exports=topsort;topsort.CycleException=CycleException;function topsort(g){var visited={},stack={},results=[];function visit(node){if(_.has(stack,node)){throw new CycleException}if(!_.has(visited,node)){stack[node]=true;visited[node]=true;_.each(g.predecessors(node),visit);delete stack[node];results.push(node)}}_.each(g.sinks(),visit);if(_.size(visited)!==g.nodeCount()){throw new CycleException}return results}function CycleException(){}},{"../lodash":75}],71:[function(require,module,exports){var _=require("../lodash");module.exports=PriorityQueue;function PriorityQueue(){this._arr=[];this._keyIndices={}}PriorityQueue.prototype.size=function(){return this._arr.length};PriorityQueue.prototype.keys=function(){return this._arr.map(function(x){return x.key})};PriorityQueue.prototype.has=function(key){return _.has(this._keyIndices,key)};PriorityQueue.prototype.priority=function(key){var index=this._keyIndices[key];if(index!==undefined){return this._arr[index].priority}};PriorityQueue.prototype.min=function(){if(this.size()===0){throw new Error("Queue underflow")}return this._arr[0].key};PriorityQueue.prototype.add=function(key,priority){var keyIndices=this._keyIndices;key=String(key);if(!_.has(keyIndices,key)){var arr=this._arr;var index=arr.length;keyIndices[key]=index;arr.push({key:key,priority:priority});this._decrease(index);return true}return false};PriorityQueue.prototype.removeMin=function(){this._swap(0,this._arr.length-1);var min=this._arr.pop();delete this._keyIndices[min.key];this._heapify(0);return min.key};PriorityQueue.prototype.decrease=function(key,priority){var index=this._keyIndices[key];if(priority>this._arr[index].priority){throw new Error("New priority is greater than current priority. "+"Key: "+key+" Old: "+this._arr[index].priority+" New: "+priority)}this._arr[index].priority=priority;this._decrease(index)};PriorityQueue.prototype._heapify=function(i){var arr=this._arr;var l=2*i,r=l+1,largest=i;if(l>1;if(arr[parent].priority1){this.setNode(v,value)}else{this.setNode(v)}},this);return this};Graph.prototype.setNode=function(v,value){if(_.has(this._nodes,v)){if(arguments.length>1){this._nodes[v]=value}return this}this._nodes[v]=arguments.length>1?value:this._defaultNodeLabelFn(v);if(this._isCompound){this._parent[v]=GRAPH_NODE;this._children[v]={};this._children[GRAPH_NODE][v]=true}this._in[v]={};this._preds[v]={};this._out[v]={};this._sucs[v]={};++this._nodeCount;return this};Graph.prototype.node=function(v){return this._nodes[v]};Graph.prototype.hasNode=function(v){return _.has(this._nodes,v)};Graph.prototype.removeNode=function(v){var self=this;if(_.has(this._nodes,v)){var removeEdge=function(e){self.removeEdge(self._edgeObjs[e])};delete this._nodes[v];if(this._isCompound){this._removeFromParentsChildList(v);delete this._parent[v];_.each(this.children(v),function(child){this.setParent(child)},this);delete this._children[v]}_.each(_.keys(this._in[v]),removeEdge);delete this._in[v];delete this._preds[v];_.each(_.keys(this._out[v]),removeEdge);delete this._out[v];delete this._sucs[v];--this._nodeCount}return this};Graph.prototype.setParent=function(v,parent){if(!this._isCompound){throw new Error("Cannot set parent in a non-compound graph")}if(_.isUndefined(parent)){parent=GRAPH_NODE}else{for(var ancestor=parent;!_.isUndefined(ancestor);ancestor=this.parent(ancestor)){if(ancestor===v){throw new Error("Setting "+parent+" as parent of "+v+" would create create a cycle")}}this.setNode(parent)}this.setNode(v);this._removeFromParentsChildList(v);this._parent[v]=parent;this._children[parent][v]=true;return this};Graph.prototype._removeFromParentsChildList=function(v){delete this._children[this._parent[v]][v]};Graph.prototype.parent=function(v){if(this._isCompound){var parent=this._parent[v];if(parent!==GRAPH_NODE){return parent}}};Graph.prototype.children=function(v){if(_.isUndefined(v)){v=GRAPH_NODE}if(this._isCompound){var children=this._children[v];if(children){return _.keys(children)}}else if(v===GRAPH_NODE){return this.nodes()}else if(this.hasNode(v)){return[]}};Graph.prototype.predecessors=function(v){var predsV=this._preds[v];if(predsV){return _.keys(predsV)}};Graph.prototype.successors=function(v){var sucsV=this._sucs[v];if(sucsV){return _.keys(sucsV)}};Graph.prototype.neighbors=function(v){var preds=this.predecessors(v);if(preds){return _.union(preds,this.successors(v))}};Graph.prototype.setDefaultEdgeLabel=function(newDefault){if(!_.isFunction(newDefault)){newDefault=_.constant(newDefault)}this._defaultEdgeLabelFn=newDefault;return this};Graph.prototype.edgeCount=function(){return this._edgeCount};Graph.prototype.edges=function(){return _.values(this._edgeObjs)};Graph.prototype.setPath=function(vs,value){var self=this,args=arguments;_.reduce(vs,function(v,w){if(args.length>1){self.setEdge(v,w,value)}else{self.setEdge(v,w)}return w});return this};Graph.prototype.setEdge=function(){var v,w,name,value,valueSpecified=false;if(_.isPlainObject(arguments[0])){v=arguments[0].v;w=arguments[0].w;name=arguments[0].name;if(arguments.length===2){value=arguments[1];valueSpecified=true}}else{v=arguments[0];w=arguments[1];name=arguments[3];if(arguments.length>2){value=arguments[2];valueSpecified=true}}v=""+v;w=""+w;if(!_.isUndefined(name)){name=""+name}var e=edgeArgsToId(this._isDirected,v,w,name);if(_.has(this._edgeLabels,e)){if(valueSpecified){this._edgeLabels[e]=value}return this}if(!_.isUndefined(name)&&!this._isMultigraph){throw new Error("Cannot set a named edge when isMultigraph = false")}this.setNode(v);this.setNode(w);this._edgeLabels[e]=valueSpecified?value:this._defaultEdgeLabelFn(v,w,name);var edgeObj=edgeArgsToObj(this._isDirected,v,w,name);v=edgeObj.v;w=edgeObj.w;Object.freeze(edgeObj);this._edgeObjs[e]=edgeObj;incrementOrInitEntry(this._preds[w],v);incrementOrInitEntry(this._sucs[v],w);this._in[w][e]=edgeObj;this._out[v][e]=edgeObj;this._edgeCount++;return this};Graph.prototype.edge=function(v,w,name){var e=arguments.length===1?edgeObjToId(this._isDirected,arguments[0]):edgeArgsToId(this._isDirected,v,w,name);return this._edgeLabels[e]};Graph.prototype.hasEdge=function(v,w,name){var e=arguments.length===1?edgeObjToId(this._isDirected,arguments[0]):edgeArgsToId(this._isDirected,v,w,name);return _.has(this._edgeLabels,e)};Graph.prototype.removeEdge=function(v,w,name){var e=arguments.length===1?edgeObjToId(this._isDirected,arguments[0]):edgeArgsToId(this._isDirected,v,w,name),edge=this._edgeObjs[e];if(edge){v=edge.v;w=edge.w;delete this._edgeLabels[e];delete this._edgeObjs[e];decrementOrRemoveEntry(this._preds[w],v);decrementOrRemoveEntry(this._sucs[v],w);delete this._in[w][e];delete this._out[v][e];this._edgeCount--}return this};Graph.prototype.inEdges=function(v,u){var inV=this._in[v];if(inV){var edges=_.values(inV);if(!u){return edges}return _.filter(edges,function(edge){return edge.v===u})}};Graph.prototype.outEdges=function(v,w){var outV=this._out[v];if(outV){var edges=_.values(outV);if(!w){return edges}return _.filter(edges,function(edge){return edge.w===w})}};Graph.prototype.nodeEdges=function(v,w){var inEdges=this.inEdges(v,w);if(inEdges){return inEdges.concat(this.outEdges(v,w))}};function incrementOrInitEntry(map,k){if(_.has(map,k)){map[k]++}else{map[k]=1}}function decrementOrRemoveEntry(map,k){if(!--map[k]){delete map[k]}}function edgeArgsToId(isDirected,v,w,name){if(!isDirected&&v>w){var tmp=v;v=w;w=tmp}return v+EDGE_KEY_DELIM+w+EDGE_KEY_DELIM+(_.isUndefined(name)?DEFAULT_EDGE_NAME:name)}function edgeArgsToObj(isDirected,v,w,name){if(!isDirected&&v>w){var tmp=v;v=w;w=tmp}var edgeObj={v:v,w:w};if(name){edgeObj.name=name}return edgeObj}function edgeObjToId(isDirected,edgeObj){return edgeArgsToId(isDirected,edgeObj.v,edgeObj.w,edgeObj.name)}},{"./lodash":75}],73:[function(require,module,exports){module.exports={Graph:require("./graph"),version:require("./version")}},{"./graph":72,"./version":76}],74:[function(require,module,exports){var _=require("./lodash"),Graph=require("./graph");module.exports={write:write,read:read};function write(g){var json={options:{directed:g.isDirected(),multigraph:g.isMultigraph(),compound:g.isCompound()},nodes:writeNodes(g),edges:writeEdges(g)};if(!_.isUndefined(g.graph())){json.value=_.clone(g.graph())}return json}function writeNodes(g){return _.map(g.nodes(),function(v){var nodeValue=g.node(v),parent=g.parent(v),node={v:v};if(!_.isUndefined(nodeValue)){node.value=nodeValue}if(!_.isUndefined(parent)){node.parent=parent}return node})}function writeEdges(g){return _.map(g.edges(),function(e){var edgeValue=g.edge(e),edge={v:e.v,w:e.w};if(!_.isUndefined(e.name)){edge.name=e.name}if(!_.isUndefined(edgeValue)){edge.value=edgeValue}return edge})}function read(json){var g=new Graph(json.options).setGraph(json.value);_.each(json.nodes,function(entry){g.setNode(entry.v,entry.value);if(entry.parent){g.setParent(entry.v,entry.parent)}});_.each(json.edges,function(entry){g.setEdge({v:entry.v,w:entry.w,name:entry.name},entry.value)});return g}},{"./graph":72,"./lodash":75}],75:[function(require,module,exports){module.exports=require(20)},{"/Users/andrew/Documents/dev/dagre-d3/lib/lodash.js":20,lodash:77}],76:[function(require,module,exports){module.exports="1.0.1"},{}],77:[function(require,module,exports){(function(global){(function(){var undefined;var arrayPool=[],objectPool=[];var idCounter=0;var keyPrefix=+new Date+"";var largeArraySize=75;var maxPoolSize=40;var whitespace=" \f \ufeff"+"\n\r\u2028\u2029"+" ᠎             　";var reEmptyStringLeading=/\b__p \+= '';/g,reEmptyStringMiddle=/\b(__p \+=) '' \+/g,reEmptyStringTrailing=/(__e\(.*?\)|\b__t\)) \+\n'';/g;var reEsTemplate=/\$\{([^\\}]*(?:\\.[^\\}]*)*)\}/g;var reFlags=/\w*$/;var reFuncName=/^\s*function[ \n\r\t]+\w/;var reInterpolate=/<%=([\s\S]+?)%>/g;var reLeadingSpacesAndZeros=RegExp("^["+whitespace+"]*0+(?=.$)");var reNoMatch=/($^)/;var reThis=/\bthis\b/;var reUnescapedString=/['\n\r\t\u2028\u2029\\]/g;var contextProps=["Array","Boolean","Date","Function","Math","Number","Object","RegExp","String","_","attachEvent","clearTimeout","isFinite","isNaN","parseInt","setTimeout"];var templateCounter=0;var argsClass="[object Arguments]",arrayClass="[object Array]",boolClass="[object Boolean]",dateClass="[object Date]",funcClass="[object Function]",numberClass="[object Number]",objectClass="[object Object]",regexpClass="[object RegExp]",stringClass="[object String]";var cloneableClasses={};cloneableClasses[funcClass]=false;cloneableClasses[argsClass]=cloneableClasses[arrayClass]=cloneableClasses[boolClass]=cloneableClasses[dateClass]=cloneableClasses[numberClass]=cloneableClasses[objectClass]=cloneableClasses[regexpClass]=cloneableClasses[stringClass]=true;var debounceOptions={leading:false,maxWait:0,trailing:false};var descriptor={configurable:false,enumerable:false,value:null,writable:false};var objectTypes={"boolean":false,"function":true,object:true,number:false,string:false,undefined:false};var stringEscapes={"\\":"\\","'":"'","\n":"n","\r":"r"," ":"t","\u2028":"u2028","\u2029":"u2029"};var root=objectTypes[typeof window]&&window||this;var freeExports=objectTypes[typeof exports]&&exports&&!exports.nodeType&&exports;var freeModule=objectTypes[typeof module]&&module&&!module.nodeType&&module;var moduleExports=freeModule&&freeModule.exports===freeExports&&freeExports;var freeGlobal=objectTypes[typeof global]&&global;if(freeGlobal&&(freeGlobal.global===freeGlobal||freeGlobal.window===freeGlobal)){root=freeGlobal}function baseIndexOf(array,value,fromIndex){var index=(fromIndex||0)-1,length=array?array.length:0;while(++index-1?0:-1:cache?0:-1}function cachePush(value){var cache=this.cache,type=typeof value;if(type=="boolean"||value==null){cache[value]=true}else{if(type!="number"&&type!="string"){type="object"}var key=type=="number"?value:keyPrefix+value,typeCache=cache[type]||(cache[type]={});if(type=="object"){(typeCache[key]||(typeCache[key]=[])).push(value)}else{typeCache[key]=true}}}function charAtCallback(value){return value.charCodeAt(0)}function compareAscending(a,b){var ac=a.criteria,bc=b.criteria,index=-1,length=ac.length;while(++indexother||typeof value=="undefined"){return 1}if(value/g,evaluate:/<%([\s\S]+?)%>/g,interpolate:reInterpolate,variable:"",imports:{_:lodash}};function baseBind(bindData){var func=bindData[0],partialArgs=bindData[2],thisArg=bindData[4];function bound(){if(partialArgs){var args=slice(partialArgs);push.apply(args,arguments)}if(this instanceof bound){var thisBinding=baseCreate(func.prototype),result=func.apply(thisBinding,args||arguments);return isObject(result)?result:thisBinding}return func.apply(thisArg,args||arguments)}setBindData(bound,bindData);return bound}function baseClone(value,isDeep,callback,stackA,stackB){if(callback){var result=callback(value);if(typeof result!="undefined"){return result}}var isObj=isObject(value);if(isObj){var className=toString.call(value);if(!cloneableClasses[className]){return value}var ctor=ctorByClass[className];switch(className){case boolClass:case dateClass:return new ctor(+value);case numberClass:case stringClass:return new ctor(value);case regexpClass:result=ctor(value.source,reFlags.exec(value));result.lastIndex=value.lastIndex;return result}}else{return value}var isArr=isArray(value);if(isDeep){var initedStack=!stackA;stackA||(stackA=getArray());stackB||(stackB=getArray());var length=stackA.length;while(length--){if(stackA[length]==value){return stackB[length]}}result=isArr?ctor(value.length):{}}else{result=isArr?slice(value):assign({},value)}if(isArr){if(hasOwnProperty.call(value,"index")){result.index=value.index}if(hasOwnProperty.call(value,"input")){result.input=value.input}}if(!isDeep){return result}stackA.push(value);stackB.push(result);(isArr?forEach:forOwn)(value,function(objValue,key){result[key]=baseClone(objValue,isDeep,callback,stackA,stackB)});if(initedStack){releaseArray(stackA);releaseArray(stackB)}return result}function baseCreate(prototype,properties){return isObject(prototype)?nativeCreate(prototype):{}}if(!nativeCreate){baseCreate=function(){function Object(){}return function(prototype){if(isObject(prototype)){Object.prototype=prototype;var result=new Object;Object.prototype=null}return result||context.Object()}}()}function baseCreateCallback(func,thisArg,argCount){if(typeof func!="function"){return identity}if(typeof thisArg=="undefined"||!("prototype"in func)){return func}var bindData=func.__bindData__;if(typeof bindData=="undefined"){if(support.funcNames){bindData=!func.name}bindData=bindData||!support.funcDecomp;if(!bindData){var source=fnToString.call(func);if(!support.funcNames){bindData=!reFuncName.test(source)}if(!bindData){bindData=reThis.test(source);setBindData(func,bindData)}}}if(bindData===false||bindData!==true&&bindData[1]&1){return func}switch(argCount){case 1:return function(value){return func.call(thisArg,value)};case 2:return function(a,b){return func.call(thisArg,a,b)};case 3:return function(value,index,collection){return func.call(thisArg,value,index,collection)};case 4:return function(accumulator,value,index,collection){return func.call(thisArg,accumulator,value,index,collection)}}return bind(func,thisArg)}function baseCreateWrapper(bindData){var func=bindData[0],bitmask=bindData[1],partialArgs=bindData[2],partialRightArgs=bindData[3],thisArg=bindData[4],arity=bindData[5];var isBind=bitmask&1,isBindKey=bitmask&2,isCurry=bitmask&4,isCurryBound=bitmask&8,key=func;function bound(){var thisBinding=isBind?thisArg:this;if(partialArgs){var args=slice(partialArgs);push.apply(args,arguments)}if(partialRightArgs||isCurry){args||(args=slice(arguments));if(partialRightArgs){push.apply(args,partialRightArgs)}if(isCurry&&args.length=largeArraySize&&indexOf===baseIndexOf,result=[]; - -if(isLarge){var cache=createCache(values);if(cache){indexOf=cacheIndexOf;values=cache}else{isLarge=false}}while(++index-1}})}}stackA.pop();stackB.pop();if(initedStack){releaseArray(stackA);releaseArray(stackB)}return result}function baseMerge(object,source,callback,stackA,stackB){(isArray(source)?forEach:forOwn)(source,function(source,key){var found,isArr,result=source,value=object[key];if(source&&((isArr=isArray(source))||isPlainObject(source))){var stackLength=stackA.length;while(stackLength--){if(found=stackA[stackLength]==source){value=stackB[stackLength];break}}if(!found){var isShallow;if(callback){result=callback(value,source);if(isShallow=typeof result!="undefined"){value=result}}if(!isShallow){value=isArr?isArray(value)?value:[]:isPlainObject(value)?value:{}}stackA.push(source);stackB.push(value);if(!isShallow){baseMerge(value,source,callback,stackA,stackB)}}}else{if(callback){result=callback(value,source);if(typeof result=="undefined"){result=source}}if(typeof result!="undefined"){value=result}}object[key]=value})}function baseRandom(min,max){return min+floor(nativeRandom()*(max-min+1))}function baseUniq(array,isSorted,callback){var index=-1,indexOf=getIndexOf(),length=array?array.length:0,result=[];var isLarge=!isSorted&&length>=largeArraySize&&indexOf===baseIndexOf,seen=callback||isLarge?getArray():result;if(isLarge){var cache=createCache(seen);indexOf=cacheIndexOf;seen=cache}while(++index":">",'"':""","'":"'"};var htmlUnescapes=invert(htmlEscapes);var reEscapedHtml=RegExp("("+keys(htmlUnescapes).join("|")+")","g"),reUnescapedHtml=RegExp("["+keys(htmlEscapes).join("")+"]","g");var assign=function(object,source,guard){var index,iterable=object,result=iterable;if(!iterable)return result;var args=arguments,argsIndex=0,argsLength=typeof guard=="number"?2:args.length;if(argsLength>3&&typeof args[argsLength-2]=="function"){var callback=baseCreateCallback(args[--argsLength-1],args[argsLength--],2)}else if(argsLength>2&&typeof args[argsLength-1]=="function"){callback=args[--argsLength]}while(++argsIndex3&&typeof args[length-2]=="function"){var callback=baseCreateCallback(args[--length-1],args[length--],2)}else if(length>2&&typeof args[length-1]=="function"){callback=args[--length]}var sources=slice(arguments,1,length),index=-1,stackA=getArray(),stackB=getArray();while(++index-1}else if(typeof length=="number"){result=(isString(collection)?collection.indexOf(target,fromIndex):indexOf(collection,target,fromIndex))>-1}else{forOwn(collection,function(value){if(++index>=fromIndex){return!(result=value===target)}})}return result}var countBy=createAggregator(function(result,value,key){hasOwnProperty.call(result,key)?result[key]++:result[key]=1});function every(collection,callback,thisArg){var result=true;callback=lodash.createCallback(callback,thisArg,3);var index=-1,length=collection?collection.length:0;if(typeof length=="number"){while(++indexresult){result=value}}}else{callback=callback==null&&isString(collection)?charAtCallback:lodash.createCallback(callback,thisArg,3);forEach(collection,function(value,index,collection){var current=callback(value,index,collection);if(current>computed){computed=current;result=value}})}return result}function min(collection,callback,thisArg){var computed=Infinity,result=computed;if(typeof callback!="function"&&thisArg&&thisArg[callback]===collection){callback=null}if(callback==null&&isArray(collection)){var index=-1,length=collection.length;while(++index=largeArraySize&&createCache(argsIndex?args[argsIndex]:seen))}}var array=args[0],index=-1,length=array?array.length:0,result=[];outer:while(++index>>1;callback(array[mid])1?arguments:arguments[0],index=-1,length=array?max(pluck(array,"length")):0,result=Array(length<0?0:length);while(++index2?createWrapper(func,17,slice(arguments,2),null,thisArg):createWrapper(func,1,null,null,thisArg)}function bindAll(object){var funcs=arguments.length>1?baseFlatten(arguments,true,false,1):functions(object),index=-1,length=funcs.length;while(++index2?createWrapper(key,19,slice(arguments,2),null,object):createWrapper(key,3,null,null,object)}function compose(){var funcs=arguments,length=funcs.length;while(length--){if(!isFunction(funcs[length])){throw new TypeError}}return function(){var args=arguments,length=funcs.length;while(length--){args=[funcs[length].apply(this,args)]}return args[0]}}function curry(func,arity){arity=typeof arity=="number"?arity:+arity||func.length;return createWrapper(func,4,null,null,null,arity)}function debounce(func,wait,options){var args,maxTimeoutId,result,stamp,thisArg,timeoutId,trailingCall,lastCalled=0,maxWait=false,trailing=true;if(!isFunction(func)){throw new TypeError}wait=nativeMax(0,wait)||0;if(options===true){var leading=true;trailing=false}else if(isObject(options)){leading=options.leading;maxWait="maxWait"in options&&(nativeMax(wait,options.maxWait)||0);trailing="trailing"in options?options.trailing:trailing}var delayed=function(){var remaining=wait-(now()-stamp);if(remaining<=0){if(maxTimeoutId){clearTimeout(maxTimeoutId); - -}var isCalled=trailingCall;maxTimeoutId=timeoutId=trailingCall=undefined;if(isCalled){lastCalled=now();result=func.apply(thisArg,args);if(!timeoutId&&!maxTimeoutId){args=thisArg=null}}}else{timeoutId=setTimeout(delayed,remaining)}};var maxDelayed=function(){if(timeoutId){clearTimeout(timeoutId)}maxTimeoutId=timeoutId=trailingCall=undefined;if(trailing||maxWait!==wait){lastCalled=now();result=func.apply(thisArg,args);if(!timeoutId&&!maxTimeoutId){args=thisArg=null}}};return function(){args=arguments;stamp=now();thisArg=this;trailingCall=trailing&&(timeoutId||!leading);if(maxWait===false){var leadingCall=leading&&!timeoutId}else{if(!maxTimeoutId&&!leading){lastCalled=stamp}var remaining=maxWait-(stamp-lastCalled),isCalled=remaining<=0;if(isCalled){if(maxTimeoutId){maxTimeoutId=clearTimeout(maxTimeoutId)}lastCalled=stamp;result=func.apply(thisArg,args)}else if(!maxTimeoutId){maxTimeoutId=setTimeout(maxDelayed,remaining)}}if(isCalled&&timeoutId){timeoutId=clearTimeout(timeoutId)}else if(!timeoutId&&wait!==maxWait){timeoutId=setTimeout(delayed,wait)}if(leadingCall){isCalled=true;result=func.apply(thisArg,args)}if(isCalled&&!timeoutId&&!maxTimeoutId){args=thisArg=null}return result}}function defer(func){if(!isFunction(func)){throw new TypeError}var args=slice(arguments,1);return setTimeout(function(){func.apply(undefined,args)},1)}function delay(func,wait){if(!isFunction(func)){throw new TypeError}var args=slice(arguments,2);return setTimeout(function(){func.apply(undefined,args)},wait)}function memoize(func,resolver){if(!isFunction(func)){throw new TypeError}var memoized=function(){var cache=memoized.cache,key=resolver?resolver.apply(this,arguments):keyPrefix+arguments[0];return hasOwnProperty.call(cache,key)?cache[key]:cache[key]=func.apply(this,arguments)};memoized.cache={};return memoized}function once(func){var ran,result;if(!isFunction(func)){throw new TypeError}return function(){if(ran){return result}ran=true;result=func.apply(this,arguments);func=null;return result}}function partial(func){return createWrapper(func,16,slice(arguments,1))}function partialRight(func){return createWrapper(func,32,null,slice(arguments,1))}function throttle(func,wait,options){var leading=true,trailing=true;if(!isFunction(func)){throw new TypeError}if(options===false){leading=false}else if(isObject(options)){leading="leading"in options?options.leading:leading;trailing="trailing"in options?options.trailing:trailing}debounceOptions.leading=leading;debounceOptions.maxWait=wait;debounceOptions.trailing=trailing;return debounce(func,wait,debounceOptions)}function wrap(value,wrapper){return createWrapper(wrapper,16,[value])}function constant(value){return function(){return value}}function createCallback(func,thisArg,argCount){var type=typeof func;if(func==null||type=="function"){return baseCreateCallback(func,thisArg,argCount)}if(type!="object"){return property(func)}var props=keys(func),key=props[0],a=func[key];if(props.length==1&&a===a&&!isObject(a)){return function(object){var b=object[key];return a===b&&(a!==0||1/a==1/b)}}return function(object){var length=props.length,result=false;while(length--){if(!(result=baseIsEqual(object[props[length]],func[props[length]],null,true))){break}}return result}}function escape(string){return string==null?"":String(string).replace(reUnescapedHtml,escapeHtmlChar)}function identity(value){return value}function mixin(object,source,options){var chain=true,methodNames=source&&functions(source);if(!source||!options&&!methodNames.length){if(options==null){options=source}ctor=lodashWrapper;source=object;object=lodash;methodNames=functions(source)}if(options===false){chain=false}else if(isObject(options)&&"chain"in options){chain=options.chain}var ctor=object,isFunc=isFunction(ctor);forEach(methodNames,function(methodName){var func=object[methodName]=source[methodName];if(isFunc){ctor.prototype[methodName]=function(){var chainAll=this.__chain__,value=this.__wrapped__,args=[value];push.apply(args,arguments);var result=func.apply(object,args);if(chain||chainAll){if(value===result&&isObject(result)){return this}result=new ctor(result);result.__chain__=chainAll}return result}}})}function noConflict(){context._=oldDash;return this}function noop(){}var now=isNative(now=Date.now)&&now||function(){return(new Date).getTime()};var parseInt=nativeParseInt(whitespace+"08")==8?nativeParseInt:function(value,radix){return nativeParseInt(isString(value)?value.replace(reLeadingSpacesAndZeros,""):value,radix||0)};function property(key){return function(object){return object[key]}}function random(min,max,floating){var noMin=min==null,noMax=max==null;if(floating==null){if(typeof min=="boolean"&&noMax){floating=min;min=1}else if(!noMax&&typeof max=="boolean"){floating=max;noMax=true}}if(noMin&&noMax){max=1}min=+min||0;if(noMax){max=min;min=0}else{max=+max||0}if(floating||min%1||max%1){var rand=nativeRandom();return nativeMin(min+rand*(max-min+parseFloat("1e-"+((rand+"").length-1))),max)}return baseRandom(min,max)}function result(object,key){if(object){var value=object[key];return isFunction(value)?object[key]():value}}function template(text,data,options){var settings=lodash.templateSettings;text=String(text||"");options=defaults({},options,settings);var imports=defaults({},options.imports,settings.imports),importsKeys=keys(imports),importsValues=values(imports);var isEvaluating,index=0,interpolate=options.interpolate||reNoMatch,source="__p += '";var reDelimiters=RegExp((options.escape||reNoMatch).source+"|"+interpolate.source+"|"+(interpolate===reInterpolate?reEsTemplate:reNoMatch).source+"|"+(options.evaluate||reNoMatch).source+"|$","g");text.replace(reDelimiters,function(match,escapeValue,interpolateValue,esTemplateValue,evaluateValue,offset){interpolateValue||(interpolateValue=esTemplateValue);source+=text.slice(index,offset).replace(reUnescapedString,escapeStringChar);if(escapeValue){source+="' +\n__e("+escapeValue+") +\n'"}if(evaluateValue){isEvaluating=true;source+="';\n"+evaluateValue+";\n__p += '"}if(interpolateValue){source+="' +\n((__t = ("+interpolateValue+")) == null ? '' : __t) +\n'"}index=offset+match.length;return match});source+="';\n";var variable=options.variable,hasVariable=variable;if(!hasVariable){variable="obj";source="with ("+variable+") {\n"+source+"\n}\n"}source=(isEvaluating?source.replace(reEmptyStringLeading,""):source).replace(reEmptyStringMiddle,"$1").replace(reEmptyStringTrailing,"$1;");source="function("+variable+") {\n"+(hasVariable?"":variable+" || ("+variable+" = {});\n")+"var __t, __p = '', __e = _.escape"+(isEvaluating?", __j = Array.prototype.join;\n"+"function print() { __p += __j.call(arguments, '') }\n":";\n")+source+"return __p\n}";var sourceURL="\n/*\n//# sourceURL="+(options.sourceURL||"/lodash/template/source["+templateCounter++ +"]")+"\n*/";try{var result=Function(importsKeys,"return "+source+sourceURL).apply(undefined,importsValues)}catch(e){e.source=source;throw e}if(data){return result(data)}result.source=source;return result}function times(n,callback,thisArg){n=(n=+n)>-1?n:0;var index=-1,result=Array(n);callback=baseCreateCallback(callback,thisArg,1);while(++index=largeArraySize&&indexOf===baseIndexOf,result=[];if(isLarge){var cache=createCache(values);if(cache){indexOf=cacheIndexOf;values=cache}else{isLarge=false}}while(++index-1}})}}stackA.pop();stackB.pop();if(initedStack){releaseArray(stackA);releaseArray(stackB)}return result}function baseMerge(object,source,callback,stackA,stackB){(isArray(source)?forEach:forOwn)(source,function(source,key){var found,isArr,result=source,value=object[key];if(source&&((isArr=isArray(source))||isPlainObject(source))){var stackLength=stackA.length;while(stackLength--){if(found=stackA[stackLength]==source){value=stackB[stackLength];break}}if(!found){var isShallow;if(callback){result=callback(value,source);if(isShallow=typeof result!="undefined"){value=result}}if(!isShallow){value=isArr?isArray(value)?value:[]:isPlainObject(value)?value:{}}stackA.push(source);stackB.push(value);if(!isShallow){baseMerge(value,source,callback,stackA,stackB)}}}else{if(callback){result=callback(value,source);if(typeof result=="undefined"){result=source}}if(typeof result!="undefined"){value=result}}object[key]=value})}function baseRandom(min,max){return min+floor(nativeRandom()*(max-min+1))}function baseUniq(array,isSorted,callback){var index=-1,indexOf=getIndexOf(),length=array?array.length:0,result=[];var isLarge=!isSorted&&length>=largeArraySize&&indexOf===baseIndexOf,seen=callback||isLarge?getArray():result;if(isLarge){var cache=createCache(seen);indexOf=cacheIndexOf;seen=cache}while(++index":">",'"':""","'":"'"};var htmlUnescapes=invert(htmlEscapes);var reEscapedHtml=RegExp("("+keys(htmlUnescapes).join("|")+")","g"),reUnescapedHtml=RegExp("["+keys(htmlEscapes).join("")+"]","g");var assign=function(object,source,guard){var index,iterable=object,result=iterable;if(!iterable)return result;var args=arguments,argsIndex=0,argsLength=typeof guard=="number"?2:args.length;if(argsLength>3&&typeof args[argsLength-2]=="function"){var callback=baseCreateCallback(args[--argsLength-1],args[argsLength--],2)}else if(argsLength>2&&typeof args[argsLength-1]=="function"){callback=args[--argsLength]}while(++argsIndex3&&typeof args[length-2]=="function"){var callback=baseCreateCallback(args[--length-1],args[length--],2)}else if(length>2&&typeof args[length-1]=="function"){callback=args[--length]}var sources=slice(arguments,1,length),index=-1,stackA=getArray(),stackB=getArray();while(++index-1}else if(typeof length=="number"){result=(isString(collection)?collection.indexOf(target,fromIndex):indexOf(collection,target,fromIndex))>-1}else{forOwn(collection,function(value){if(++index>=fromIndex){return!(result=value===target)}})}return result}var countBy=createAggregator(function(result,value,key){hasOwnProperty.call(result,key)?result[key]++:result[key]=1});function every(collection,callback,thisArg){var result=true;callback=lodash.createCallback(callback,thisArg,3);var index=-1,length=collection?collection.length:0;if(typeof length=="number"){while(++indexresult){result=value}}}else{callback=callback==null&&isString(collection)?charAtCallback:lodash.createCallback(callback,thisArg,3);forEach(collection,function(value,index,collection){var current=callback(value,index,collection);if(current>computed){computed=current;result=value}})}return result}function min(collection,callback,thisArg){var computed=Infinity,result=computed;if(typeof callback!="function"&&thisArg&&thisArg[callback]===collection){callback=null}if(callback==null&&isArray(collection)){var index=-1,length=collection.length;while(++index=largeArraySize&&createCache(argsIndex?args[argsIndex]:seen))}}var array=args[0],index=-1,length=array?array.length:0,result=[];outer:while(++index>>1;callback(array[mid])1?arguments:arguments[0],index=-1,length=array?max(pluck(array,"length")):0,result=Array(length<0?0:length);while(++index2?createWrapper(func,17,slice(arguments,2),null,thisArg):createWrapper(func,1,null,null,thisArg)}function bindAll(object){var funcs=arguments.length>1?baseFlatten(arguments,true,false,1):functions(object),index=-1,length=funcs.length;while(++index2?createWrapper(key,19,slice(arguments,2),null,object):createWrapper(key,3,null,null,object)}function compose(){var funcs=arguments,length=funcs.length;while(length--){if(!isFunction(funcs[length])){throw new TypeError}}return function(){var args=arguments,length=funcs.length;while(length--){args=[funcs[length].apply(this,args)]}return args[0]}}function curry(func,arity){arity=typeof arity=="number"?arity:+arity||func.length;return createWrapper(func,4,null,null,null,arity)}function debounce(func,wait,options){var args,maxTimeoutId,result,stamp,thisArg,timeoutId,trailingCall,lastCalled=0,maxWait=false,trailing=true;if(!isFunction(func)){throw new TypeError}wait=nativeMax(0,wait)||0;if(options===true){var leading=true;trailing=false}else if(isObject(options)){leading=options.leading;maxWait="maxWait"in options&&(nativeMax(wait,options.maxWait)||0);trailing="trailing"in options?options.trailing:trailing}var delayed=function(){var remaining=wait-(now()-stamp);if(remaining<=0){if(maxTimeoutId){clearTimeout(maxTimeoutId)}var isCalled=trailingCall;maxTimeoutId=timeoutId=trailingCall=undefined;if(isCalled){lastCalled=now();result=func.apply(thisArg,args);if(!timeoutId&&!maxTimeoutId){args=thisArg=null}}}else{timeoutId=setTimeout(delayed,remaining)}};var maxDelayed=function(){if(timeoutId){clearTimeout(timeoutId)}maxTimeoutId=timeoutId=trailingCall=undefined;if(trailing||maxWait!==wait){lastCalled=now();result=func.apply(thisArg,args);if(!timeoutId&&!maxTimeoutId){args=thisArg=null}}};return function(){args=arguments;stamp=now();thisArg=this;trailingCall=trailing&&(timeoutId||!leading);if(maxWait===false){var leadingCall=leading&&!timeoutId}else{if(!maxTimeoutId&&!leading){lastCalled=stamp}var remaining=maxWait-(stamp-lastCalled),isCalled=remaining<=0;if(isCalled){if(maxTimeoutId){maxTimeoutId=clearTimeout(maxTimeoutId)}lastCalled=stamp;result=func.apply(thisArg,args)}else if(!maxTimeoutId){maxTimeoutId=setTimeout(maxDelayed,remaining)}}if(isCalled&&timeoutId){timeoutId=clearTimeout(timeoutId)}else if(!timeoutId&&wait!==maxWait){timeoutId=setTimeout(delayed,wait)}if(leadingCall){isCalled=true;result=func.apply(thisArg,args)}if(isCalled&&!timeoutId&&!maxTimeoutId){args=thisArg=null}return result}}function defer(func){if(!isFunction(func)){throw new TypeError}var args=slice(arguments,1);return setTimeout(function(){func.apply(undefined,args)},1)}function delay(func,wait){if(!isFunction(func)){throw new TypeError}var args=slice(arguments,2);return setTimeout(function(){func.apply(undefined,args)},wait)}function memoize(func,resolver){if(!isFunction(func)){throw new TypeError}var memoized=function(){var cache=memoized.cache,key=resolver?resolver.apply(this,arguments):keyPrefix+arguments[0];return hasOwnProperty.call(cache,key)?cache[key]:cache[key]=func.apply(this,arguments)};memoized.cache={};return memoized}function once(func){var ran,result;if(!isFunction(func)){throw new TypeError}return function(){if(ran){return result}ran=true;result=func.apply(this,arguments);func=null;return result}}function partial(func){return createWrapper(func,16,slice(arguments,1))}function partialRight(func){return createWrapper(func,32,null,slice(arguments,1))}function throttle(func,wait,options){var leading=true,trailing=true;if(!isFunction(func)){throw new TypeError}if(options===false){leading=false}else if(isObject(options)){leading="leading"in options?options.leading:leading;trailing="trailing"in options?options.trailing:trailing}debounceOptions.leading=leading;debounceOptions.maxWait=wait;debounceOptions.trailing=trailing;return debounce(func,wait,debounceOptions)}function wrap(value,wrapper){return createWrapper(wrapper,16,[value])}function constant(value){return function(){return value}}function createCallback(func,thisArg,argCount){var type=typeof func;if(func==null||type=="function"){return baseCreateCallback(func,thisArg,argCount)}if(type!="object"){return property(func)}var props=keys(func),key=props[0],a=func[key];if(props.length==1&&a===a&&!isObject(a)){return function(object){var b=object[key];return a===b&&(a!==0||1/a==1/b)}}return function(object){var length=props.length,result=false;while(length--){if(!(result=baseIsEqual(object[props[length]],func[props[length]],null,true))){break}}return result}}function escape(string){return string==null?"":String(string).replace(reUnescapedHtml,escapeHtmlChar)}function identity(value){return value}function mixin(object,source,options){var chain=true,methodNames=source&&functions(source);if(!source||!options&&!methodNames.length){if(options==null){options=source}ctor=lodashWrapper;source=object;object=lodash;methodNames=functions(source)}if(options===false){chain=false}else if(isObject(options)&&"chain"in options){chain=options.chain}var ctor=object,isFunc=isFunction(ctor);forEach(methodNames,function(methodName){var func=object[methodName]=source[methodName];if(isFunc){ctor.prototype[methodName]=function(){var chainAll=this.__chain__,value=this.__wrapped__,args=[value];push.apply(args,arguments);var result=func.apply(object,args);if(chain||chainAll){if(value===result&&isObject(result)){return this}result=new ctor(result);result.__chain__=chainAll}return result}}})}function noConflict(){context._=oldDash;return this}function noop(){}var now=isNative(now=Date.now)&&now||function(){return(new Date).getTime()};var parseInt=nativeParseInt(whitespace+"08")==8?nativeParseInt:function(value,radix){return nativeParseInt(isString(value)?value.replace(reLeadingSpacesAndZeros,""):value,radix||0)};function property(key){return function(object){return object[key]}}function random(min,max,floating){var noMin=min==null,noMax=max==null;if(floating==null){if(typeof min=="boolean"&&noMax){floating=min;min=1}else if(!noMax&&typeof max=="boolean"){floating=max;noMax=true}}if(noMin&&noMax){max=1}min=+min||0;if(noMax){max=min;min=0}else{max=+max||0}if(floating||min%1||max%1){var rand=nativeRandom();return nativeMin(min+rand*(max-min+parseFloat("1e-"+((rand+"").length-1))),max)}return baseRandom(min,max)}function result(object,key){if(object){var value=object[key];return isFunction(value)?object[key]():value}}function template(text,data,options){var settings=lodash.templateSettings;text=String(text||"");options=defaults({},options,settings);var imports=defaults({},options.imports,settings.imports),importsKeys=keys(imports),importsValues=values(imports);var isEvaluating,index=0,interpolate=options.interpolate||reNoMatch,source="__p += '";var reDelimiters=RegExp((options.escape||reNoMatch).source+"|"+interpolate.source+"|"+(interpolate===reInterpolate?reEsTemplate:reNoMatch).source+"|"+(options.evaluate||reNoMatch).source+"|$","g");text.replace(reDelimiters,function(match,escapeValue,interpolateValue,esTemplateValue,evaluateValue,offset){interpolateValue||(interpolateValue=esTemplateValue);source+=text.slice(index,offset).replace(reUnescapedString,escapeStringChar);if(escapeValue){source+="' +\n__e("+escapeValue+") +\n'"}if(evaluateValue){isEvaluating=true;source+="';\n"+evaluateValue+";\n__p += '"}if(interpolateValue){source+="' +\n((__t = ("+interpolateValue+")) == null ? '' : __t) +\n'"}index=offset+match.length;return match});source+="';\n";var variable=options.variable,hasVariable=variable;if(!hasVariable){variable="obj";source="with ("+variable+") {\n"+source+"\n}\n"}source=(isEvaluating?source.replace(reEmptyStringLeading,""):source).replace(reEmptyStringMiddle,"$1").replace(reEmptyStringTrailing,"$1;");source="function("+variable+") {\n"+(hasVariable?"":variable+" || ("+variable+" = {});\n")+"var __t, __p = '', __e = _.escape"+(isEvaluating?", __j = Array.prototype.join;\n"+"function print() { __p += __j.call(arguments, '') }\n":";\n")+source+"return __p\n}";var sourceURL="\n/*\n//# sourceURL="+(options.sourceURL||"/lodash/template/source["+templateCounter++ +"]")+"\n*/";try{var result=Function(importsKeys,"return "+source+sourceURL).apply(undefined,importsValues)}catch(e){e.source=source;throw e}if(data){return result(data)}result.source=source;return result}function times(n,callback,thisArg){n=(n=+n)>-1?n:0;var index=-1,result=Array(n);callback=baseCreateCallback(callback,thisArg,1);while(++index") .replace(/"/g, "\""); var g = graphlibDot.read(escaped_dot); - g.graph().rankSep = rankSep; var renderer = new dagreD3.render(); + preprocessGraphLayout(g, forJob); renderer(container, g); } @@ -251,50 +249,37 @@ function graphContainer() { return d3.select("#dag-viz-graph"); } function metadataContainer() { return d3.select("#dag-viz-metadata"); } /* - * Helper function to create draw a label for each cluster. - * - * We need to do this manually because dagre-d3 does not support labeling clusters. - * In general, the clustering support for dagre-d3 is quite limited at this point. + * Helper function to pre-process the graph layout. + * This step is necessary for certain styles that affect the positioning + * and sizes of graph elements, e.g. padding, font style, shape. */ -function drawClusterLabels(svgContainer, forJob) { - var clusterLabelSize, stageClusterLabelSize; +function preprocessGraphLayout(g, forJob) { + var nodes = g.nodes(); + for (var i = 0; i < nodes.length; i++) { + var isCluster = g.children(nodes[i]).length > 0; + if (!isCluster) { + var node = g.node(nodes[i]); + if (forJob) { + // Do not display RDD name on job page + node.shape = "circle"; + node.labelStyle = "font-size: 0px"; + } else { + node.labelStyle = "font-size: 12px"; + } + node.padding = "5"; + } + } + // Curve the edges + var edges = g.edges(); + for (var j = 0; j < edges.length; j++) { + edges[j].lineInterpolate = "basis"; + } + // Adjust vertical separation between nodes if (forJob) { - clusterLabelSize = JobPageVizConstants.clusterLabelSize; - stageClusterLabelSize = JobPageVizConstants.stageClusterLabelSize; + g.graph().rankSep = JobPageVizConstants.rankSep; } else { - clusterLabelSize = StagePageVizConstants.clusterLabelSize; - stageClusterLabelSize = StagePageVizConstants.stageClusterLabelSize; + g.graph().rankSep = StagePageVizConstants.rankSep; } - svgContainer.selectAll("g.cluster").each(function() { - var cluster = d3.select(this); - var isStage = cluster.attr("id").indexOf(VizConstants.stageClusterPrefix) > -1; - var labelSize = isStage ? stageClusterLabelSize : clusterLabelSize; - drawClusterLabel(cluster, labelSize); - }); -} - -/* - * Helper function to draw a label for the given cluster element based on its name. - * - * In the process, we need to expand the bounding box to make room for the label. - * We need to do this because dagre-d3 did not take this into account when it first - * rendered the bounding boxes. Note that this means we need to adjust the view box - * of the SVG afterwards since we shifted a few boxes around. - */ -function drawClusterLabel(d3cluster, fontSize) { - var cluster = d3cluster; - var rect = d3cluster.select("rect"); - rect.attr("y", toFloat(rect.attr("y")) - fontSize); - rect.attr("height", toFloat(rect.attr("height")) + fontSize); - var labelX = toFloat(rect.attr("x")) + toFloat(rect.attr("width")) - fontSize / 2; - var labelY = toFloat(rect.attr("y")) + fontSize * 1.5; - var labelText = cluster.attr("name").replace(VizConstants.clusterPrefix, ""); - cluster.append("text") - .attr("x", labelX) - .attr("y", labelY) - .attr("text-anchor", "end") - .style("font-size", fontSize + "px") - .text(labelText); } /* @@ -444,7 +429,7 @@ function addTooltipsForRDDs(svgContainer) { if (tooltipText) { node.select("circle") .attr("data-toggle", "tooltip") - .attr("data-placement", "right") + .attr("data-placement", "bottom") .attr("title", tooltipText) } }); diff --git a/core/src/main/scala/org/apache/spark/ui/UIUtils.scala b/core/src/main/scala/org/apache/spark/ui/UIUtils.scala index 6a0f5c5d16daa..441c97d6d3e2a 100644 --- a/core/src/main/scala/org/apache/spark/ui/UIUtils.scala +++ b/core/src/main/scala/org/apache/spark/ui/UIUtils.scala @@ -360,7 +360,7 @@ private[spark] object UIUtils extends Logging { { graphs.map { g => } + def tooltip(text: String, position: String): Seq[Node] = { + + (?) + + } + /** Return a script element that automatically expands the DAG visualization on page load. */ def expandDagVizOnLoad(forJob: Boolean): Seq[Node] = {