aokolnychyi commented on a change in pull request #35395: URL: https://github.com/apache/spark/pull/35395#discussion_r824191664
########## File path: sql/catalyst/src/test/scala/org/apache/spark/sql/connector/catalog/InMemoryRowLevelOperationTable.scala ########## @@ -0,0 +1,96 @@ +/* + * 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.connector.catalog + +import java.util + +import org.apache.spark.sql.connector.distributions.{Distribution, Distributions} +import org.apache.spark.sql.connector.expressions.{FieldReference, LogicalExpressions, NamedReference, SortDirection, SortOrder, Transform} +import org.apache.spark.sql.connector.read.{Scan, ScanBuilder} +import org.apache.spark.sql.connector.write.{BatchWrite, LogicalWriteInfo, RequiresDistributionAndOrdering, RowLevelOperation, RowLevelOperationBuilder, RowLevelOperationInfo, Write, WriteBuilder, WriterCommitMessage} +import org.apache.spark.sql.connector.write.RowLevelOperation.Command +import org.apache.spark.sql.types.StructType +import org.apache.spark.sql.util.CaseInsensitiveStringMap + +class InMemoryRowLevelOperationTable( + name: String, + schema: StructType, + partitioning: Array[Transform], + properties: util.Map[String, String]) + extends InMemoryTable(name, schema, partitioning, properties) with SupportsRowLevelOperations { + + override def newRowLevelOperationBuilder( + info: RowLevelOperationInfo): RowLevelOperationBuilder = { + () => PartitionBasedOperation(info.command) + } + + case class PartitionBasedOperation(command: Command) extends RowLevelOperation { + private final val PARTITION_COLUMN_REF = FieldReference(PartitionKeyColumn.name) + + var configuredScan: InMemoryBatchScan = _ + + override def requiredMetadataAttributes(): Array[NamedReference] = { + Array(PARTITION_COLUMN_REF) + } + + override def newScanBuilder(options: CaseInsensitiveStringMap): ScanBuilder = { + new InMemoryScanBuilder(schema) { + override def build: Scan = { + val scan = super.build() + configuredScan = scan.asInstanceOf[InMemoryBatchScan] + scan + } + } + } + + override def newWriteBuilder(info: LogicalWriteInfo): WriteBuilder = new WriteBuilder { + + override def build(): Write = new Write with RequiresDistributionAndOrdering { + override def requiredDistribution(): Distribution = { + Distributions.clustered(Array(PARTITION_COLUMN_REF)) + } + + override def requiredOrdering(): Array[SortOrder] = { + Array[SortOrder]( + LogicalExpressions.sort( + PARTITION_COLUMN_REF, + SortDirection.ASCENDING, + SortDirection.ASCENDING.defaultNullOrdering()) + ) + } + + override def toBatch: BatchWrite = PartitionBasedReplaceData(configuredScan) + + override def description(): String = "InMemoryWrite" + } + } + + override def description(): String = "InMemoryPartitionReplaceOperation" + } + + private case class PartitionBasedReplaceData(scan: InMemoryBatchScan) extends TestBatchWrite { + + override def commit(messages: Array[WriterCommitMessage]): Unit = dataMap.synchronized { + val newData = messages.map(_.asInstanceOf[BufferedRows]) + val readRows = scan.data.flatMap(_.asInstanceOf[BufferedRows].rows) Review comment: I don't think the result will be discarded. Let me explain how I see this: - The data source plans input splits for groups that may have matches using parts of the condition that can be converted into data source filters. - The data source scan can cache group IDs or input splits that may have matches and expose filter attributes for runtime filtering. In case of Delta, the scan needs to remember a set of files that potentially have rows to update/delete. - If the data source supports runtime filtering, Spark will assign a filtering subquery that will be executed at runtime based on the SQL command. The same scan object will be used in the filtering subquery and in the row-level operation. In case of Delta, the runtime filtering will be done using the `_file_name` metadata column. - Spark executes the filtering subquery via the existing runtime filtering mechanism, collects unique values for the filtering attributes and passes them into the scan. In case of Delta, the scan will receive a set of files that have matches. - The data source uses the passed unique values to filter out cached group IDs or tasks and reports back input splits for groups that definitely have matches. In case of Delta, it can filter out previously cached files and include tasks only for files that have matches. - Writes have access to scans, so each scan can report a list of affected group IDs (which was cached). It can be in any format: strings, case classes, etc. Whatever a particular data source needs. -- 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: reviews-unsubscr...@spark.apache.org 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
