JingsongLi commented on code in PR #11190: URL: https://github.com/apache/paimon/pull/11190#discussion_r4584356057
########## paimon-spark/paimon-spark-common/src/main/scala/org/apache/paimon/spark/PaimonTableAsSelectWrite.scala: ########## @@ -0,0 +1,87 @@ +/* + * 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.paimon.spark + +import org.apache.paimon.spark.write.PaimonWriteRequirement +import org.apache.paimon.table.FileStoreTable + +import org.apache.spark.sql.SparkSession +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.catalyst.plans.logical.{AppendData, LogicalPlan} +import org.apache.spark.sql.connector.catalog.{Identifier, StagedTable, Table, TableCatalog} +import org.apache.spark.sql.execution.datasources.v2.DataSourceV2Relation +import org.apache.spark.sql.paimon.shims.SparkShimLoader +import org.apache.spark.sql.util.CaseInsensitiveStringMap + +import scala.collection.JavaConverters._ +import scala.util.control.NonFatal + +/** Routes Spark 3.2-3.4 table-as-select writes through the normal write planning rules. */ +object PaimonTableAsSelectWrite { + + def supports(table: Table): Boolean = table match { + case target: SparkTable => target.useV2Write + case staged: RollbackStagedTable => supports(staged.wrappedTable()) + case _ => false + } + + def write( + spark: SparkSession, + catalog: TableCatalog, + table: Table, + ident: Identifier, + query: LogicalPlan, + options: CaseInsensitiveStringMap): Seq[InternalRow] = { + try { + // A staged table delegates its writer but is not a SparkTable relation, so PaimonAnalysis + // cannot add the legacy bucket distribution for it. Prepare that query here instead. + val writeQuery = table match { + case staged: RollbackStagedTable => + staged.wrappedTable() match { + case target: SparkTable if target.useV2Write => + PaimonWriteRequirement.prepareQuery( + target.getTable.asInstanceOf[FileStoreTable], + query) + case _ => query + } + case _ => query + } + val relation = DataSourceV2Relation.create(table, Some(catalog), Some(ident)) + val command = + AppendData.byPosition(relation, writeQuery, options.asCaseSensitiveMap().asScala.toMap) + SparkShimLoader.shim.classicApi.createDataset(spark, command).collect() + table match { + case staged: StagedTable => staged.commitStagedChanges() + case _ => + } + Nil + } catch { + case NonFatal(error) => Review Comment: [P2] Run CTAS failure cleanup when the driver is interrupted Scala NonFatal excludes InterruptedException, so this skips both dropTable and abortStagedChanges when the driver thread is interrupted after table creation. On this exact head with Spark 3.4.4/JDK 8, a real CTAS driver-interruption probe propagates the original InterruptedException for both routes: the V1 control removes the created table, while V2 leaves it registered. The assertion checks the exception before any cancelJobGroup cleanup, so this is not a task failure wrapped in SparkException. A retry then encounters the leftover table. Preserve Spark's native abort/drop behavior on interruption as well as ordinary exceptions, while propagating the original interruption. This helper is used by the new Spark 3.2–3.4 table-as-select path; runtime verification here is on 3.4. ########## paimon-spark/paimon-spark-common/src/main/scala/org/apache/paimon/spark/write/PaimonWriteRequirement.scala: ########## @@ -21,23 +21,61 @@ package org.apache.paimon.spark.write import org.apache.paimon.CoreOptions import org.apache.paimon.CoreOptions.PartitionSinkStrategy import org.apache.paimon.spark.commands.BucketExpression.quote +import org.apache.paimon.spark.commands.FixedBucketExpression import org.apache.paimon.table.{FileStoreTable, FormatTable} import org.apache.paimon.table.BucketMode._ +import org.apache.spark.sql.catalyst.SQLConfHelper +import org.apache.spark.sql.catalyst.expressions.{Expression => CatalystExpression, Literal => CatalystLiteral} +import org.apache.spark.sql.catalyst.plans.logical.{LogicalPlan, RepartitionByExpression} import org.apache.spark.sql.connector.distributions.{ClusteredDistribution, Distribution, Distributions} -import org.apache.spark.sql.connector.expressions.{Expression, Expressions, SortOrder} +import org.apache.spark.sql.connector.expressions.{Expression, Expressions, Literal, NamedReference, SortOrder, Transform} import scala.collection.JavaConverters._ /** Distribution requirements of Spark write. */ case class PaimonWriteRequirement(distribution: Distribution, ordering: Array[SortOrder]) -object PaimonWriteRequirement { +object PaimonWriteRequirement extends SQLConfHelper { private val EMPTY_ORDERING: Array[SortOrder] = Array.empty private val EMPTY: PaimonWriteRequirement = PaimonWriteRequirement(Distributions.unspecified(), EMPTY_ORDERING) + // Spark 3.2/3.3 do not pass the FunctionCatalog when preparing V2 write distributions. + // Bind the same distribution with Paimon's bucket expression before Spark prepares the write. + private[spark] def requiresQueryPreparation: Boolean = org.apache.spark.SPARK_VERSION < "3.4" + + def prepareQuery(table: FileStoreTable, query: LogicalPlan): LogicalPlan = { + if (!requiresQueryPreparation) { + return query + } + + def bind(expression: Expression): CatalystExpression = expression match { + case reference: NamedReference => + query.output.find(a => conf.resolver(a.name, reference.fieldNames().head)).get + case literal: Literal[_] => CatalystLiteral(literal.value(), literal.dataType()) + case transform: Transform if transform.name() == "identity" => + bind(transform.arguments().head) + case transform: Transform if transform.name() == "bucket" => + FixedBucketExpression( + CatalystLiteral(table.coreOptions().bucketFunctionType().toString) +: + transform.arguments().map(bind).toSeq) + case other => + throw new UnsupportedOperationException(s"Unsupported write distribution: $other") + } + + apply(table).distribution match { + case clustered: ClusteredDistribution => + // This Int overload is binary-compatible across Spark 3.2+. + RepartitionByExpression( + clustered.clustering().map(bind).toSeq, + query, + conf.numShufflePartitions) Review Comment: [P2] Keep adaptive coalescing available for legacy V2 writes Passing an explicit partition count creates a REPARTITION_BY_NUM exchange on Spark 3.2/3.3; their CoalesceShufflePartitions rule does not coalesce that origin. The previous V1 repartition(columns) path remains adaptive. On this exact head with Spark 3.2.4/JDK 8, a real one-bucket write of 40 rows from four input partitions, with AQE/coalescing enabled and shuffle.partitions=200, submits stages [4,1] through V1 versus [4,200] through V2. Both read back the same 40 rows and one actual data file. The newly default-enabled V2 route therefore schedules 199 unnecessary write tasks for this small single-bucket input. Construct an unspecified-count repartition through the version-specific adapter rather than freezing conf.numShufflePartitions, and cover actual AQE stage counts. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
