[GitHub] [spark] jaceklaskowski commented on a diff in pull request #40567: [SPARK-42935] [SQL] Add union required distribution push down

["jaceklaskowski (via GitHub)" <gi...@apache.org>]

jaceklaskowski commented on code in PR #40567:
URL: https://github.com/apache/spark/pull/40567#discussion_r1153247899


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/internal/SQLConf.scala:
##########
@@ -4195,6 +4195,15 @@ object SQLConf {
       .booleanConf
       .createWithDefault(false)
 
+  val UNION_REQUIRED_DISTRIBUTION_PUSHDOWN =
+    buildConf("spark.sql.unionRequiredDistributionPushdown.enabled")
+      .internal()
+      .doc("Enable union spark plan required children distribution pushdown. This is useful when " +

Review Comment:
   nit: Enables



##########
sql/core/src/main/scala/org/apache/spark/sql/execution/exchange/UnionDistributionPushdown.scala:
##########
@@ -0,0 +1,146 @@
+/*
+ * 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.execution.exchange
+
+import java.io.{IOException, ObjectOutputStream}
+
+import scala.reflect.ClassTag
+
+import org.apache.spark.{OneToOneDependency, Partition, SparkContext, TaskContext}
+import org.apache.spark.rdd.RDD
+import org.apache.spark.sql.catalyst.expressions.Attribute
+import org.apache.spark.sql.catalyst.plans.physical.{ClusteredDistribution, Distribution}
+import org.apache.spark.sql.catalyst.rules.Rule
+import org.apache.spark.sql.execution.{SparkPlan, UnionExec, UnionZipExec}
+import org.apache.spark.sql.internal.SQLConf
+import org.apache.spark.util.Utils
+
+/**
+ * Pushes parent's required distribution to [[UnionExec]] so that we can
+ * remove unnecessary shuffle exchange if any child output partitioning matches
+ * 1) the operator is UnionExec
+ * 2) the operator's output partitioning can not satisfy parent distribution
+ * 3) the operator's parent requires hashing or clustering distribution
+ * 4) the data types of operator's children output attributes are the same
+ */
+case class UnionDistributionPushdown() extends Rule[SparkPlan] {

Review Comment:
   Why not `object`?



##########
sql/core/src/main/scala/org/apache/spark/sql/execution/exchange/UnionDistributionPushdown.scala:
##########
@@ -0,0 +1,146 @@
+/*
+ * 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.execution.exchange
+
+import java.io.{IOException, ObjectOutputStream}
+
+import scala.reflect.ClassTag
+
+import org.apache.spark.{OneToOneDependency, Partition, SparkContext, TaskContext}
+import org.apache.spark.rdd.RDD
+import org.apache.spark.sql.catalyst.expressions.Attribute
+import org.apache.spark.sql.catalyst.plans.physical.{ClusteredDistribution, Distribution}
+import org.apache.spark.sql.catalyst.rules.Rule
+import org.apache.spark.sql.execution.{SparkPlan, UnionExec, UnionZipExec}
+import org.apache.spark.sql.internal.SQLConf
+import org.apache.spark.util.Utils
+
+/**
+ * Pushes parent's required distribution to [[UnionExec]] so that we can
+ * remove unnecessary shuffle exchange if any child output partitioning matches
+ * 1) the operator is UnionExec
+ * 2) the operator's output partitioning can not satisfy parent distribution
+ * 3) the operator's parent requires hashing or clustering distribution
+ * 4) the data types of operator's children output attributes are the same
+ */
+case class UnionDistributionPushdown() extends Rule[SparkPlan] {
+
+  def allChildrenSameDataType(unionExec: UnionExec): Boolean = {
+    (Seq(unionExec) ++ unionExec.children).map(_.output).transpose.map(attrs => {

Review Comment:
   nit: Replace `map(... => {...})` with `map { ... => }`



##########
sql/core/src/main/scala/org/apache/spark/sql/execution/exchange/UnionDistributionPushdown.scala:
##########
@@ -0,0 +1,146 @@
+/*
+ * 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.execution.exchange
+
+import java.io.{IOException, ObjectOutputStream}
+
+import scala.reflect.ClassTag
+
+import org.apache.spark.{OneToOneDependency, Partition, SparkContext, TaskContext}
+import org.apache.spark.rdd.RDD
+import org.apache.spark.sql.catalyst.expressions.Attribute
+import org.apache.spark.sql.catalyst.plans.physical.{ClusteredDistribution, Distribution}
+import org.apache.spark.sql.catalyst.rules.Rule
+import org.apache.spark.sql.execution.{SparkPlan, UnionExec, UnionZipExec}
+import org.apache.spark.sql.internal.SQLConf
+import org.apache.spark.util.Utils
+
+/**
+ * Pushes parent's required distribution to [[UnionExec]] so that we can
+ * remove unnecessary shuffle exchange if any child output partitioning matches
+ * 1) the operator is UnionExec
+ * 2) the operator's output partitioning can not satisfy parent distribution
+ * 3) the operator's parent requires hashing or clustering distribution
+ * 4) the data types of operator's children output attributes are the same
+ */
+case class UnionDistributionPushdown() extends Rule[SparkPlan] {
+
+  def allChildrenSameDataType(unionExec: UnionExec): Boolean = {
+    (Seq(unionExec) ++ unionExec.children).map(_.output).transpose.map(attrs => {
+      val firstAttr = attrs.head
+      attrs.forall(_.dataType == firstAttr.dataType)
+    }).forall(_ == true)
+  }
+
+  def createNewChildrenDistribution(unionExec: UnionExec,
+                                    unionRequired: Distribution): Seq[Distribution] = {
+    unionExec.children.map(unionChild => {
+      // create output attributes map (union output exprId -> child output)
+      val attributes = unionExec.output.map(_.exprId).zip(unionChild.output).toMap
+      unionRequired match {
+        case distribution: ClusteredDistribution =>
+          val newExpressions = distribution.clustering.map(exp => exp.transformUp {
+            case attr: Attribute if attributes.contains(attr.exprId) => attributes(attr.exprId)
+          })
+          ClusteredDistribution(newExpressions, distribution.requireAllClusterKeys,
+            distribution.requiredNumPartitions)
+
+        case distribution => distribution
+      }
+    })
+  }
+
+  override def apply(plan: SparkPlan): SparkPlan = plan.transformUp {
+    case operator: SparkPlan if conf.getConf(SQLConf.UNION_REQUIRED_DISTRIBUTION_PUSHDOWN) &&
+      operator.children.exists(_.isInstanceOf[UnionExec]) =>
+      val newChildren = operator.children.zipWithIndex.map {
+        case (unionExec: UnionExec, childIndex: Int) if !unionExec.outputPartitioning.
+          satisfies(operator.requiredChildDistribution(childIndex)) =>
+          val unionRequired = operator.requiredChildDistribution(childIndex)
+          unionRequired match {
+            case _: ClusteredDistribution if allChildrenSameDataType(unionExec) =>
+              val requiredChildDistribution: Seq[Distribution] =
+                createNewChildrenDistribution(unionExec, unionRequired)
+              val numPartitions = unionRequired.requiredNumPartitions
+                .getOrElse(conf.numShufflePartitions)
+              val outputPartitioning = unionRequired.createPartitioning(numPartitions)
+              UnionZipExec(unionExec.children, requiredChildDistribution,
+                outputPartitioning)
+
+            case _: Distribution => unionExec
+          }
+
+        case (child, _) => child
+      }
+      operator.withNewChildren(newChildren)
+  }
+}
+
+/**
+ * Class representing partitions of UnionZipRDD, which maintains the list of
+ * corresponding partitions of parent RDDs.
+ */
+private[spark]
+class UnionZipRDDPartition(@transient val rdds: Seq[RDD[_]],
+                           override val index: Int) extends Partition {
+  var parents: Array[Partition] = rdds.map(_.partitions(index)).toArray
+
+  override def hashCode(): Int = index
+
+  override def equals(other: Any): Boolean = super.equals(other)
+
+  @throws(classOf[IOException])
+  private def writeObject(oos: ObjectOutputStream): Unit = Utils.tryOrIOException {
+    // Update the reference to parent partition at the time of task serialization
+    parents = rdds.map(_.partitions(index)).toArray
+    oos.defaultWriteObject()
+  }
+}
+
+/**
+ * Class representing an RDD that can take multiple RDDs with same partition size and
+ * unify them into a single RDD while preserving the partition. So m RDDs with p partitions
+ * will be unified to a single RDD with p partitions, which the ith partition consists of the
+ * ith partition in each parent RDD.
+ */
+private[spark]
+class UnionZipRDD[T: ClassTag](sc: SparkContext, var rdds: Seq[RDD[T]])

Review Comment:
   nit: Can you move arguments to their own lines?



##########
sql/core/src/main/scala/org/apache/spark/sql/execution/exchange/UnionDistributionPushdown.scala:
##########
@@ -0,0 +1,146 @@
+/*
+ * 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.execution.exchange
+
+import java.io.{IOException, ObjectOutputStream}
+
+import scala.reflect.ClassTag
+
+import org.apache.spark.{OneToOneDependency, Partition, SparkContext, TaskContext}
+import org.apache.spark.rdd.RDD
+import org.apache.spark.sql.catalyst.expressions.Attribute
+import org.apache.spark.sql.catalyst.plans.physical.{ClusteredDistribution, Distribution}
+import org.apache.spark.sql.catalyst.rules.Rule
+import org.apache.spark.sql.execution.{SparkPlan, UnionExec, UnionZipExec}
+import org.apache.spark.sql.internal.SQLConf
+import org.apache.spark.util.Utils
+
+/**
+ * Pushes parent's required distribution to [[UnionExec]] so that we can
+ * remove unnecessary shuffle exchange if any child output partitioning matches
+ * 1) the operator is UnionExec
+ * 2) the operator's output partitioning can not satisfy parent distribution
+ * 3) the operator's parent requires hashing or clustering distribution
+ * 4) the data types of operator's children output attributes are the same
+ */
+case class UnionDistributionPushdown() extends Rule[SparkPlan] {
+
+  def allChildrenSameDataType(unionExec: UnionExec): Boolean = {
+    (Seq(unionExec) ++ unionExec.children).map(_.output).transpose.map(attrs => {
+      val firstAttr = attrs.head
+      attrs.forall(_.dataType == firstAttr.dataType)
+    }).forall(_ == true)
+  }
+
+  def createNewChildrenDistribution(unionExec: UnionExec,
+                                    unionRequired: Distribution): Seq[Distribution] = {
+    unionExec.children.map(unionChild => {
+      // create output attributes map (union output exprId -> child output)
+      val attributes = unionExec.output.map(_.exprId).zip(unionChild.output).toMap
+      unionRequired match {
+        case distribution: ClusteredDistribution =>
+          val newExpressions = distribution.clustering.map(exp => exp.transformUp {
+            case attr: Attribute if attributes.contains(attr.exprId) => attributes(attr.exprId)
+          })
+          ClusteredDistribution(newExpressions, distribution.requireAllClusterKeys,
+            distribution.requiredNumPartitions)
+
+        case distribution => distribution
+      }
+    })
+  }
+
+  override def apply(plan: SparkPlan): SparkPlan = plan.transformUp {
+    case operator: SparkPlan if conf.getConf(SQLConf.UNION_REQUIRED_DISTRIBUTION_PUSHDOWN) &&
+      operator.children.exists(_.isInstanceOf[UnionExec]) =>
+      val newChildren = operator.children.zipWithIndex.map {
+        case (unionExec: UnionExec, childIndex: Int) if !unionExec.outputPartitioning.
+          satisfies(operator.requiredChildDistribution(childIndex)) =>
+          val unionRequired = operator.requiredChildDistribution(childIndex)
+          unionRequired match {
+            case _: ClusteredDistribution if allChildrenSameDataType(unionExec) =>
+              val requiredChildDistribution: Seq[Distribution] =
+                createNewChildrenDistribution(unionExec, unionRequired)
+              val numPartitions = unionRequired.requiredNumPartitions
+                .getOrElse(conf.numShufflePartitions)
+              val outputPartitioning = unionRequired.createPartitioning(numPartitions)
+              UnionZipExec(unionExec.children, requiredChildDistribution,
+                outputPartitioning)
+
+            case _: Distribution => unionExec
+          }
+
+        case (child, _) => child
+      }
+      operator.withNewChildren(newChildren)
+  }
+}
+
+/**
+ * Class representing partitions of UnionZipRDD, which maintains the list of
+ * corresponding partitions of parent RDDs.
+ */
+private[spark]
+class UnionZipRDDPartition(@transient val rdds: Seq[RDD[_]],
+                           override val index: Int) extends Partition {
+  var parents: Array[Partition] = rdds.map(_.partitions(index)).toArray
+
+  override def hashCode(): Int = index
+
+  override def equals(other: Any): Boolean = super.equals(other)

Review Comment:
   Is `equals` required? Doesn't do much.



##########
sql/core/src/main/scala/org/apache/spark/sql/execution/exchange/UnionDistributionPushdown.scala:
##########
@@ -0,0 +1,146 @@
+/*
+ * 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.execution.exchange
+
+import java.io.{IOException, ObjectOutputStream}
+
+import scala.reflect.ClassTag
+
+import org.apache.spark.{OneToOneDependency, Partition, SparkContext, TaskContext}
+import org.apache.spark.rdd.RDD
+import org.apache.spark.sql.catalyst.expressions.Attribute
+import org.apache.spark.sql.catalyst.plans.physical.{ClusteredDistribution, Distribution}
+import org.apache.spark.sql.catalyst.rules.Rule
+import org.apache.spark.sql.execution.{SparkPlan, UnionExec, UnionZipExec}
+import org.apache.spark.sql.internal.SQLConf
+import org.apache.spark.util.Utils
+
+/**
+ * Pushes parent's required distribution to [[UnionExec]] so that we can
+ * remove unnecessary shuffle exchange if any child output partitioning matches
+ * 1) the operator is UnionExec
+ * 2) the operator's output partitioning can not satisfy parent distribution
+ * 3) the operator's parent requires hashing or clustering distribution
+ * 4) the data types of operator's children output attributes are the same
+ */
+case class UnionDistributionPushdown() extends Rule[SparkPlan] {
+
+  def allChildrenSameDataType(unionExec: UnionExec): Boolean = {
+    (Seq(unionExec) ++ unionExec.children).map(_.output).transpose.map(attrs => {
+      val firstAttr = attrs.head
+      attrs.forall(_.dataType == firstAttr.dataType)
+    }).forall(_ == true)
+  }
+
+  def createNewChildrenDistribution(unionExec: UnionExec,
+                                    unionRequired: Distribution): Seq[Distribution] = {
+    unionExec.children.map(unionChild => {

Review Comment:
   same as above



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org