JkSelf commented on a change in pull request #25295: [SPARK-28560][SQL] Optimize shuffle reader to local shuffle reader when smj converted to bhj in adaptive execution URL: https://github.com/apache/spark/pull/25295#discussion_r323548060
########## File path: sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/LocalShuffledRowRDD.scala ########## @@ -0,0 +1,118 @@ +/* + * 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._ +import org.apache.spark.rdd.{RDD, ShuffledRDDPartition} +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.execution.metric.{SQLMetric, SQLShuffleReadMetricsReporter} + +/** + * This is a specialized version of [[org.apache.spark.sql.execution.ShuffledRowRDD]]. This is used + * in Spark SQL adaptive execution when a shuffle join is converted to broadcast join at runtime + * because the map output of one input table is small enough for broadcast. This RDD represents the + * data of another input table of the join that reads from shuffle. Each partition of the RDD reads + * the whole data from just one mapper output locally. So actually there is no data transferred + * from the network. + + * This RDD takes a [[ShuffleDependency]] (`dependency`). + * + * The `dependency` has the parent RDD of this RDD, which represents the dataset before shuffle + * (i.e. map output). Elements of this RDD are (partitionId, Row) pairs. + * Partition ids should be in the range [0, numPartitions - 1]. + * `dependency.partitioner.numPartitions` is the number of pre-shuffle partitions. (i.e. the number + * of partitions of the map output). The post-shuffle partition number is the same to the parent + * RDD's partition number. + */ +class LocalShuffledRowRDD( + var dependency: ShuffleDependency[Int, InternalRow, InternalRow], + metrics: Map[String, SQLMetric], + specifiedPartitionStartIndices: Option[Array[Int]] = None, + specifiedPartitionEndIndices: Option[Array[Int]] = None) Review comment: @viirya Currently not. We may need the `specifiedPartitionEndIndices `variable to skip the partitions with 0 size in the following optimization. And I will retain and use it when create `LocalShuffledRowRDD `later. ---------------------------------------------------------------- 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: us...@infra.apache.org With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org
