[SPARK-43046] [SS] [Connect] Implemented Python API dropDuplicatesWithinWatermark for Spark Connect

connector/connect/common/src/main/protobuf/spark/connect/relations.proto

@@ -362,6 +362,9 @@ message Deduplicate {
   //
   // This field does not co-use with `column_names`.
   optional bool all_columns_as_keys = 3;
+
+  // (Optional) Deduplicate within the time range of watermark.
+  optional bool within_watermark = 4;
 }
 
 // A relation that does not need to be qualified by name.

connector/connect/server/src/main/scala/org/apache/spark/sql/connect/dsl/package.scala

@@ -596,6 +596,17 @@ package object dsl {
               .addAllColumnNames(colNames.asJava))
           .build()
 
+      def deduplicateWithinWatermark(colNames: Seq[String]): Relation =
+        Relation
+          .newBuilder()
+          .setDeduplicate(
+            Deduplicate
+              .newBuilder()
+              .setInput(logicalPlan)
+              .addAllColumnNames(colNames.asJava)
+              .setWithinWatermark(true))
+          .build()
+
       def distinct(): Relation =
         Relation
           .newBuilder()

connector/connect/server/src/main/scala/org/apache/spark/sql/connect/planner/SparkConnectPlanner.scala

@@ -46,7 +46,7 @@ import org.apache.spark.sql.catalyst.expressions._
 import org.apache.spark.sql.catalyst.parser.{CatalystSqlParser, ParseException, ParserUtils}
 import org.apache.spark.sql.catalyst.plans.{Cross, FullOuter, Inner, JoinType, LeftAnti, LeftOuter, LeftSemi, RightOuter, UsingJoin}
 import org.apache.spark.sql.catalyst.plans.logical
-import org.apache.spark.sql.catalyst.plans.logical.{CollectMetrics, CommandResult, Deduplicate, DeserializeToObject, Except, Intersect, LocalRelation, LogicalPlan, MapPartitions, Project, Sample, SerializeFromObject, Sort, SubqueryAlias, TypedFilter, Union, Unpivot, UnresolvedHint}
+import org.apache.spark.sql.catalyst.plans.logical.{CollectMetrics, CommandResult, Deduplicate, DeduplicateWithinWatermark, DeserializeToObject, Except, Intersect, LocalRelation, LogicalPlan, MapPartitions, Project, Sample, SerializeFromObject, Sort, SubqueryAlias, TypedFilter, Union, Unpivot, UnresolvedHint}
 import org.apache.spark.sql.catalyst.util.{CaseInsensitiveMap, CharVarcharUtils}
 import org.apache.spark.sql.connect.artifact.SparkConnectArtifactManager
 import org.apache.spark.sql.connect.common.{DataTypeProtoConverter, InvalidPlanInput, LiteralValueProtoConverter, StorageLevelProtoConverter, UdfPacket}
@@ -738,7 +738,8 @@ class SparkConnectPlanner(val session: SparkSession) {
     val resolver = session.sessionState.analyzer.resolver
     val allColumns = queryExecution.analyzed.output
     if (rel.getAllColumnsAsKeys) {
-      Deduplicate(allColumns, queryExecution.analyzed)
+      if (rel.getWithinWatermark) DeduplicateWithinWatermark(allColumns, queryExecution.analyzed)
+      else Deduplicate(allColumns, queryExecution.analyzed)
     } else {
       val toGroupColumnNames = rel.getColumnNamesList.asScala.toSeq
       val groupCols = toGroupColumnNames.flatMap { (colName: String) =>
@@ -750,7 +751,8 @@ class SparkConnectPlanner(val session: SparkSession) {
         }
         cols
       }
-      Deduplicate(groupCols, queryExecution.analyzed)
+      if (rel.getWithinWatermark) DeduplicateWithinWatermark(groupCols, queryExecution.analyzed)
+      else Deduplicate(groupCols, queryExecution.analyzed)
     }
   }
 

connector/connect/server/src/test/scala/org/apache/spark/sql/connect/planner/SparkConnectPlannerSuite.scala

@@ -381,6 +381,36 @@ class SparkConnectPlannerSuite extends SparkFunSuite with SparkConnectPlanTest {
     assert(e2.getMessage.contains("either deduplicate on all columns or a subset of columns"))
   }
 
+  test("Test invalid deduplicateWithinWatermark") {
+    val deduplicateWithinWatermark = proto.Deduplicate
+      .newBuilder()
+      .setInput(readRel)
+      .setAllColumnsAsKeys(true)
+      .addColumnNames("test")
+      .setWithinWatermark(true)
+
+    val e = intercept[InvalidPlanInput] {
+      transform(
+        proto.Relation.newBuilder
+          .setDeduplicate(deduplicateWithinWatermark)
+          .build())
+    }
+    assert(
+      e.getMessage.contains("Cannot deduplicate on both all columns and a subset of columns"))
+
+    val deduplicateWithinWatermark2 = proto.Deduplicate
+      .newBuilder()
+      .setInput(readRel)
+      .setWithinWatermark(true)
+    val e2 = intercept[InvalidPlanInput] {
+      transform(
+        proto.Relation.newBuilder
+          .setDeduplicate(deduplicateWithinWatermark2)
+          .build())
+    }
+    assert(e2.getMessage.contains("either deduplicate on all columns or a subset of columns"))
+  }
+
   test("Test invalid intersect, except") {
     // Except with union_by_name=true
     val except = proto.SetOperation

connector/connect/server/src/test/scala/org/apache/spark/sql/connect/planner/SparkConnectProtoSuite.scala

@@ -348,6 +348,16 @@ class SparkConnectProtoSuite extends PlanTest with SparkConnectPlanTest {
     comparePlans(connectPlan2, sparkPlan2)
   }
 
+  test("Test basic deduplicateWithinWatermark") {
+    val connectPlan = connectTestRelation.distinct()
+    val sparkPlan = sparkTestRelation.distinct()
+    comparePlans(connectPlan, sparkPlan)
+
+    val connectPlan2 = connectTestRelation.deduplicateWithinWatermark(Seq("id", "name"))
+    val sparkPlan2 = sparkTestRelation.dropDuplicatesWithinWatermark(Seq("id", "name"))
+    comparePlans(connectPlan2, sparkPlan2)
+  }
+
   test("Test union, except, intersect") {
     val connectPlan1 = connectTestRelation.except(connectTestRelation, isAll = false)
     val sparkPlan1 = sparkTestRelation.except(sparkTestRelation)

python/pyspark/sql/connect/dataframe.py

@@ -379,7 +379,24 @@ def dropDuplicates(self, subset: Optional[List[str]] = None) -> "DataFrame":
     drop_duplicates = dropDuplicates
 
     def dropDuplicatesWithinWatermark(self, subset: Optional[List[str]] = None) -> "DataFrame":
-        raise NotImplementedError("dropDuplicatesWithinWatermark() is not implemented.")
+        if subset is not None and not isinstance(subset, (list, tuple)):
+            raise PySparkTypeError(
+                error_class="NOT_LIST_OR_TUPLE",
+                message_parameters={"arg_name": "subset", "arg_type": type(subset).__name__},
+            )
+
+        if subset is None:
+            return DataFrame.withPlan(
+                plan.DeduplicateWithinWatermark(child=self._plan, all_columns_as_keys=True), session=self._session
+            )
+        else:
+            return DataFrame.withPlan(
+                plan.DeduplicateWithinWatermark(child=self._plan, column_names=subset), session=self._session
+            )
+
+    dropDuplicatesWithinWatermark.__doc__ = PySparkDataFrame.dropDuplicatesWithinWatermark.__doc__
+
+    drop_duplicates_within_watermark = dropDuplicatesWithinWatermark
 
     def distinct(self) -> "DataFrame":
         return DataFrame.withPlan(

python/pyspark/sql/connect/plan.py

@@ -623,6 +623,30 @@ def plan(self, session: "SparkConnectClient") -> proto.Relation:
         return plan
 
 
+class DeduplicateWithinWatermark(LogicalPlan):
+    def __init__(
+            self,
+            child: Optional["LogicalPlan"],
+            all_columns_as_keys: bool = False,
+            column_names: Optional[List[str]] = None,
+            within_watermark: bool = True,
+    ) -> None:
+        super().__init__(child)
+        self.all_columns_as_keys = all_columns_as_keys
+        self.column_names = column_names
+        self.within_watermark = within_watermark
+
+    def plan(self, session: "SparkConnectClient") -> proto.Relation:
+        assert self._child is not None
+        plan = self._create_proto_relation()
+        plan.deduplicate.input.CopyFrom(self._child.plan(session))
+        plan.deduplicate.all_columns_as_keys = self.all_columns_as_keys
+        plan.deduplicate.within_watermark = self.within_watermark
+        if self.column_names is not None:
+            plan.deduplicate.column_names.extend(self.column_names)
+        return plan
+
+
 class Sort(LogicalPlan):
     def __init__(
         self,

python/pyspark/sql/connect/proto/base_pb2.py

@@ -699,7 +699,6 @@
 
 _SPARKCONNECTSERVICE = DESCRIPTOR.services_by_name["SparkConnectService"]
 if _descriptor._USE_C_DESCRIPTORS == False:
-
     DESCRIPTOR._options = None
     DESCRIPTOR._serialized_options = b"\n\036org.apache.spark.connect.protoP\001"
     _EXECUTEPLANRESPONSE_METRICS_METRICOBJECT_EXECUTIONMETRICSENTRY._options = None