HeartSaVioR commented on a change in pull request #31083: URL: https://github.com/apache/spark/pull/31083#discussion_r556051127
########## 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: My concern is mostly the "static partitions" as I provided as an example like state data source. That's not a matter of whether it's ill-pattern or not, because for the case the ability of restricting the number of partitions is not optional but "required" - the data should be partitioned exactly the same with Spark partitions the rows for hash shuffle, and a partition shouldn't be written concurrently. I don't think end users should do the repartition manually in their queries to not break a thing. That is easily achievable in DSv1 (I have an implementation based on DSv1 and want to migrate to DSv2) as Spark provides DataFrame to the data source on write. While I don't expect such flexibility for DSv2 (the behavior seems too open), I'm not sure the case is something we'd like to define as "not supported on DSv2 and have to live with DSv1". ---------------------------------------------------------------- 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]
