imback82 commented on a change in pull request #28123: URL: https://github.com/apache/spark/pull/28123#discussion_r439706623
########## File path: sql/core/src/main/scala/org/apache/spark/sql/execution/bucketing/CoalesceBucketsInSortMergeJoin.scala ########## @@ -0,0 +1,112 @@ +/* + * 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.bucketing + +import org.apache.spark.sql.catalyst.catalog.BucketSpec +import org.apache.spark.sql.catalyst.rules.Rule +import org.apache.spark.sql.execution.{FileSourceScanExec, FilterExec, ProjectExec, SparkPlan} +import org.apache.spark.sql.execution.joins.SortMergeJoinExec +import org.apache.spark.sql.internal.SQLConf + +/** + * This rule coalesces one side of the `SortMergeJoin` if the following conditions are met: + * - Two bucketed tables are joined. + * - The larger bucket number is divisible by the smaller bucket number. + * - COALESCE_BUCKETS_IN_SORT_MERGE_JOIN_ENABLED is set to true. + * - The ratio of the number of buckets is less than the value set in + * COALESCE_BUCKETS_IN_SORT_MERGE_JOIN_MAX_BUCKET_RATIO. + */ +case class CoalesceBucketsInSortMergeJoin(conf: SQLConf) extends Rule[SparkPlan] { + private def mayCoalesce(numBuckets1: Int, numBuckets2: Int, conf: SQLConf): Option[Int] = { + assert(numBuckets1 != numBuckets2) + val (small, large) = (math.min(numBuckets1, numBuckets2), math.max(numBuckets1, numBuckets2)) + // A bucket can be coalesced only if the bigger number of buckets is divisible by the smaller + // number of buckets because bucket id is calculated by modding the total number of buckets. + if (large % small == 0 && + large / small <= conf.getConf(SQLConf.COALESCE_BUCKETS_IN_SORT_MERGE_JOIN_MAX_BUCKET_RATIO)) { + Some(small) + } else { + None + } + } + + private def updateNumCoalescedBuckets(plan: SparkPlan, numCoalescedBuckets: Int): SparkPlan = { + plan.transformUp { + case f: FileSourceScanExec => + f.copy(optionalNumCoalescedBuckets = Some(numCoalescedBuckets)) + } + } + + def apply(plan: SparkPlan): SparkPlan = { + if (!conf.getConf(SQLConf.COALESCE_BUCKETS_IN_SORT_MERGE_JOIN_ENABLED)) { + return plan + } + + plan transform { + case ExtractSortMergeJoinWithBuckets(smj, numLeftBuckets, numRightBuckets) + if numLeftBuckets != numRightBuckets => + mayCoalesce(numLeftBuckets, numRightBuckets, conf).map { numCoalescedBuckets => + if (numCoalescedBuckets != numLeftBuckets) { + smj.copy(left = updateNumCoalescedBuckets(smj.left, numCoalescedBuckets)) + } else { + smj.copy(right = updateNumCoalescedBuckets(smj.right, numCoalescedBuckets)) + } + }.getOrElse(smj) + case other => other + } + } +} + +/** + * An extractor that extracts `SortMergeJoinExec` where both sides of the join have the bucketed + * tables and are consisted of only the scan operation. + */ +object ExtractSortMergeJoinWithBuckets { + private def isScanOperation(plan: SparkPlan): Boolean = plan match { + case f: FilterExec => isScanOperation(f.child) + case p: ProjectExec => isScanOperation(p.child) Review comment: Would this be a cleaner approach? Or handling buckets in one place (`FileSourceScanExec`) would be cleaner? What do you think @cloud-fan / @maropu? ---------------------------------------------------------------- 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 --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org
