cloud-fan commented on a change in pull request #32550:
URL: https://github.com/apache/spark/pull/32550#discussion_r633577681
##########
File path:
sql/catalyst/src/main/scala/org/apache/spark/sql/internal/SQLConf.scala
##########
@@ -598,6 +598,16 @@ object SQLConf {
.bytesConf(ByteUnit.BYTE)
.createOptional
+ val ADAPTIVE_SHUFFLE_HASH_JOIN_LOCAL_MAP_THRESHOLD =
+ buildConf("spark.sql.adaptive.shuffledHashJoinLocalMapThreshold")
+ .doc("Configures the maximum size in bytes for per partition that can be
allowed to build " +
+ "local hash map. If all the partition size not larger than this
threshold, join " +
+ "selection may use shuffled hash join instead of sort merge join when
" +
Review comment:
`when` -> `even if`
##########
File path:
sql/catalyst/src/main/scala/org/apache/spark/sql/internal/SQLConf.scala
##########
@@ -598,6 +598,16 @@ object SQLConf {
.bytesConf(ByteUnit.BYTE)
.createOptional
+ val ADAPTIVE_SHUFFLE_HASH_JOIN_LOCAL_MAP_THRESHOLD =
+ buildConf("spark.sql.adaptive.shuffledHashJoinLocalMapThreshold")
+ .doc("Configures the maximum size in bytes for per partition that can be
allowed to build " +
+ "local hash map. If all the partition size not larger than this
threshold, join " +
+ "selection may use shuffled hash join instead of sort merge join when
" +
+ s"${PREFER_SORTMERGEJOIN.key} is false.")
Review comment:
really? Shall we use shuffle hash join regardless of this
`PREFER_SORTMERGEJOIN` config?
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/DynamicJoinSelection.scala
##########
@@ -0,0 +1,81 @@
+/*
+ * 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.adaptive
+
+import org.apache.spark.sql.catalyst.plans.logical.{HintInfo, Join,
JoinStrategyHint, JoinStrategyHintCollection, LogicalPlan, NO_BROADCAST_HASH,
PREFER_SHUFFLE_HASH}
+import org.apache.spark.sql.catalyst.rules.Rule
+
+/**
+ * This optimization rule incluede two join selection:
+ * 1. detects a join child that has a high ratio of empty partitions and
adds a
+ * no_broadcast_hash_join hint to avoid it being broadcast.
+ * 2. detects a join child that every partition size less than local map
threshold and adds a
+ * prefer_shuffled_hash hint to encourage being shuffle hash join instead
of sort merge join.
+ */
+object DynamicJoinSelection extends Rule[LogicalPlan] {
+
+ private def shouldDemoteBroadcastHashJoin(stage: ShuffleQueryStageExec):
Boolean = {
+ val mapStats = stage.mapStats.get
+ val partitionCnt = mapStats.bytesByPartitionId.length
+ val nonZeroCnt = mapStats.bytesByPartitionId.count(_ > 0)
+ partitionCnt > 0 && nonZeroCnt > 0 &&
+ (nonZeroCnt * 1.0 / partitionCnt) <
conf.nonEmptyPartitionRatioForBroadcastJoin
+ }
+
+ private def preferShuffledHashJoin(stage: ShuffleQueryStageExec): Boolean = {
+ val localMapThreshold = conf.shuffleHashJoinLocalMapThreshold
+ stage.mapStats.get.bytesByPartitionId.forall(_ <= localMapThreshold)
+ }
+
+ private def selectJoinStrategy(plan: LogicalPlan): Option[JoinStrategyHint]
= plan match {
+ case LogicalQueryStage(_, stage: ShuffleQueryStageExec) if
stage.resultOption.get().isDefined
+ && stage.mapStats.isDefined =>
+ var strategies = Seq[JoinStrategyHint]()
+ if (shouldDemoteBroadcastHashJoin(stage)) {
+ strategies = strategies ++ Seq(NO_BROADCAST_HASH)
+ }
+ if (preferShuffledHashJoin(stage)) {
Review comment:
are these two join hints exclusive? shall we use `else if` here?
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/DynamicJoinSelection.scala
##########
@@ -0,0 +1,81 @@
+/*
+ * 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.adaptive
+
+import org.apache.spark.sql.catalyst.plans.logical.{HintInfo, Join,
JoinStrategyHint, JoinStrategyHintCollection, LogicalPlan, NO_BROADCAST_HASH,
PREFER_SHUFFLE_HASH}
+import org.apache.spark.sql.catalyst.rules.Rule
+
+/**
+ * This optimization rule incluede two join selection:
+ * 1. detects a join child that has a high ratio of empty partitions and
adds a
+ * no_broadcast_hash_join hint to avoid it being broadcast.
+ * 2. detects a join child that every partition size less than local map
threshold and adds a
+ * prefer_shuffled_hash hint to encourage being shuffle hash join instead
of sort merge join.
+ */
+object DynamicJoinSelection extends Rule[LogicalPlan] {
+
+ private def shouldDemoteBroadcastHashJoin(stage: ShuffleQueryStageExec):
Boolean = {
+ val mapStats = stage.mapStats.get
+ val partitionCnt = mapStats.bytesByPartitionId.length
+ val nonZeroCnt = mapStats.bytesByPartitionId.count(_ > 0)
+ partitionCnt > 0 && nonZeroCnt > 0 &&
+ (nonZeroCnt * 1.0 / partitionCnt) <
conf.nonEmptyPartitionRatioForBroadcastJoin
+ }
+
+ private def preferShuffledHashJoin(stage: ShuffleQueryStageExec): Boolean = {
+ val localMapThreshold = conf.shuffleHashJoinLocalMapThreshold
+ stage.mapStats.get.bytesByPartitionId.forall(_ <= localMapThreshold)
+ }
+
+ private def selectJoinStrategy(plan: LogicalPlan): Option[JoinStrategyHint]
= plan match {
+ case LogicalQueryStage(_, stage: ShuffleQueryStageExec) if
stage.resultOption.get().isDefined
+ && stage.mapStats.isDefined =>
+ var strategies = Seq[JoinStrategyHint]()
+ if (shouldDemoteBroadcastHashJoin(stage)) {
+ strategies = strategies ++ Seq(NO_BROADCAST_HASH)
+ }
+ if (preferShuffledHashJoin(stage)) {
Review comment:
I think `PREFER_SHUFFLE_HASH` means we shouldn't do broadcast join.
--
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.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
