attilapiros commented on a change in pull request #28618: URL: https://github.com/apache/spark/pull/28618#discussion_r446384655
########## File path: core/src/main/scala/org/apache/spark/shuffle/MemoizingShuffleDataIO.scala ########## @@ -0,0 +1,51 @@ +/* + * 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.shuffle + +import scala.collection.JavaConverters._ + +import org.apache.spark.shuffle.api.{ShuffleDataIO, ShuffleDriverComponents, ShuffleExecutorComponents} + +/** + * Thin wrapper around {@link ShuffleDataIO} that ensures the given components are + * only initialized once and providing the same instance each time. + * <p> + * Used to ensure the SparkEnv only instantiates the given components once lazily + * and then reuses them throughout the lifetime of the SparkEnv. + */ +class MemoizingShuffleDataIO(delegate: ShuffleDataIO) { + private var _driver: ShuffleDriverComponents = _ + private var _executor: ShuffleExecutorComponents = _ + + def getOrCreateDriverComponents(): ShuffleDriverComponents = synchronized { + if (_driver == null) { + _driver = delegate.initializeShuffleDriverComponents() + } + _driver + } + + def getOrCreateExecutorComponents( + appId: String, + execId: String, + extraConfigs: Map[String, String]): ShuffleExecutorComponents = synchronized { + if (_executor == null) { + _executor = delegate.initializeShuffleExecutorComponents(appId, execId, extraConfigs.asJava) + } + _executor + } Review comment: I think we can improve the methods of this class. Especially the `getOrCreateExecutorComponents` as reading that I have a feeling it is easy to use it badly. For example by calling it twice with two different values (for any of its params) as always the object generated for the first call will be returned. But first `getOrCreateDriverComponents` that is the easier here as it has no input params at all. So this could be replaced with a simple `lazy val` (it is thread-safe for a long time ago: https://github.com/scala/bug/issues/3007). And the same is true for `getOrCreateExecutorComponents ` as all its parameters are basically coming from a SparkEnv instance. So my idea is: ```suggestion lazy val driverComponents = delegate.initializeShuffleDriverComponents() lazy val executorComponents = { val env = SparkEnv.get delegate.initializeShuffleExecutorComponents( env.conf.getAppId, env.executorId, env.conf.getAllWithPrefix(ShuffleDataIOUtils.SHUFFLE_SPARK_CONF_PREFIX).toMap.asJava) } ``` I still have to test it. What is your opinion @mccheah ? ---------------------------------------------------------------- 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]
