aokolnychyi commented on a change in pull request #31083: URL: https://github.com/apache/spark/pull/31083#discussion_r555708684
########## File path: sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/v2/DistributionAndOrderingUtils.scala ########## @@ -0,0 +1,110 @@ +/* + * 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.datasources.v2 + +import org.apache.spark.sql.{catalyst, AnalysisException} +import org.apache.spark.sql.catalyst.analysis.Resolver +import org.apache.spark.sql.catalyst.expressions.{NamedExpression, SortOrder} +import org.apache.spark.sql.catalyst.plans.logical.{LogicalPlan, RepartitionByExpression, Sort} +import org.apache.spark.sql.connector.distributions.{ClusteredDistribution, OrderedDistribution, UnspecifiedDistribution} +import org.apache.spark.sql.connector.expressions.{Expression, FieldReference, IdentityTransform, NullOrdering, SortDirection, SortValue} +import org.apache.spark.sql.connector.write.{RequiresDistributionAndOrdering, Write} +import org.apache.spark.sql.internal.SQLConf + +object DistributionAndOrderingUtils { + + def prepareQuery(write: Write, query: LogicalPlan, conf: SQLConf): LogicalPlan = write match { + case write: RequiresDistributionAndOrdering => + val resolver = conf.resolver + + val distribution = write.requiredDistribution match { + case d: OrderedDistribution => + d.ordering.map(e => toCatalyst(e, query, resolver)) + case d: ClusteredDistribution => + d.clustering.map(e => toCatalyst(e, query, resolver)) + case _: UnspecifiedDistribution => + Array.empty[catalyst.expressions.Expression] + } + + val queryWithDistribution = if (distribution.nonEmpty) { + val numShufflePartitions = conf.numShufflePartitions Review comment: @HeartSaVioR, I think it is going to be useful for some data sources. I did not want to cover this in first PRs as there was no consensus on the [dev list](https://lists.apache.org/thread.html/d8bb72fc9b4be8acc3f49367bfc99cbf029194a58333eba69df49717@%3Cdev.spark.apache.org%3E) around controlling the number of tasks. That's why I added this topic to non-goals of the design doc. I think one point to think about is who should control the parallelism. I guess the parallelism should depend on incoming data volume in most cases (except when the number of requested partitions is static, like probably in the case mentioned above). Without having statistics about the number of incoming records or their shape, it will be hard for a data source to determine the right number of partitions. That being said, I think making that number optional like in your change can be a reasonable starting point. However, I'd like us to think about how this will look like in the future. Should Spark report stats about the incoming batch so that data sources can make a better estimate? How will that API look like? ---------------------------------------------------------------- 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]
