RussellSpitzer commented on a change in pull request #2591: URL: https://github.com/apache/iceberg/pull/2591#discussion_r665819822
########## File path: spark3/src/main/java/org/apache/iceberg/spark/actions/Spark3BinPackStrategy.java ########## @@ -0,0 +1,85 @@ +/* + * 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.iceberg.spark.actions; + +import java.util.List; +import java.util.Set; +import java.util.UUID; +import org.apache.iceberg.DataFile; +import org.apache.iceberg.FileScanTask; +import org.apache.iceberg.Table; +import org.apache.iceberg.actions.BinPackStrategy; +import org.apache.iceberg.relocated.com.google.common.collect.ImmutableSet; +import org.apache.iceberg.spark.FileRewriteCoordinator; +import org.apache.iceberg.spark.FileScanTaskSetManager; +import org.apache.iceberg.spark.SparkReadOptions; +import org.apache.iceberg.spark.SparkWriteOptions; +import org.apache.spark.sql.Dataset; +import org.apache.spark.sql.Row; +import org.apache.spark.sql.SparkSession; +import org.apache.spark.sql.internal.SQLConf; + +public class Spark3BinPackStrategy extends BinPackStrategy { + private final Table table; + private final SparkSession spark; + private final FileScanTaskSetManager manager = FileScanTaskSetManager.get(); + private final FileRewriteCoordinator rewriteCoordinator = FileRewriteCoordinator.get(); + + public Spark3BinPackStrategy(Table table, SparkSession spark) { + this.table = table; + this.spark = spark; + } + + @Override + public Table table() { + return table; + } + + @Override + public Set<DataFile> rewriteFiles(List<FileScanTask> filesToRewrite) { + String groupID = UUID.randomUUID().toString(); + try { + manager.stageTasks(table, groupID, filesToRewrite); + + // Disable Adaptive Query Execution as this may change the output partitioning of our write + SparkSession cloneSession = spark.cloneSession(); + cloneSession.conf().set(SQLConf.ADAPTIVE_EXECUTION_ENABLED().key(), false); + + Dataset<Row> scanDF = cloneSession.read().format("iceberg") + .option(SparkReadOptions.FILE_SCAN_TASK_SET_ID, groupID) + .option(SparkReadOptions.SPLIT_SIZE, splitSize(inputFileSize(filesToRewrite))) + .option(SparkReadOptions.FILE_OPEN_COST, "0") Review comment: So normally we are trying to weight the cost of opening a new file vs the cost of starting a new spark task. Opening files is expensive and sometimes it makes more sense to just start another task in parallel rather than waiting to open another file in serial within the same task. Imagine the case of many small files, if we pretend they are slightly larger with this value then we will bias our splits to having less small files in them which helps use make more spark tasks even if they read less data. In this case we aren't optimizing for read performance, instead we are trying to make each Spark task as close to our target size as possible regardless of how many small files it takes. Here our goal is to write files of the correct size so the number of input files doesn't really matter. To do this we set the open cost to 0. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
