haiyangsun-db commented on code in PR #55611: URL: https://github.com/apache/spark/pull/55611#discussion_r3192223030
########## sql/core/src/main/scala/org/apache/spark/sql/execution/externalUDF/MapPartitionExternalUDFExec.scala: ########## @@ -0,0 +1,68 @@ +/* + * 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.externalUDF + +import org.apache.spark.TaskContext +import org.apache.spark.annotation.Experimental +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.catalyst.expressions.{Attribute, + ExternalUserDefinedFunction} +import org.apache.spark.sql.execution.SparkPlan +import org.apache.spark.udf.worker.UDFWorkerSpecification +import org.apache.spark.udf.worker.core.InitMessage + +/** + * :: Experimental :: + * Physical plan node that executes a mapPartitions-style UDF in an + * external worker process. + * + * @param workerSpec Specification describing the UDF worker. + * @param functionExpr The UDF to invoke. + * @param resultAttributes Output attributes produced by the UDF. + * @param child Child plan providing input partitions. + */ +@Experimental +case class MapPartitionExternalUDFExec( + workerSpec: UDFWorkerSpecification, + functionExpr: ExternalUserDefinedFunction, + resultAttributes: Seq[Attribute], + child: SparkPlan) + extends ExternalUDFExec { + + override protected def doExecute(): RDD[InternalRow] = { + child.execute().mapPartitionsInternal { rows => + val taskContext = TaskContext.get() + withUDFWorkerSession(taskContext, securityScope = None) { + session => + session.init(InitMessage( + functionPayload = functionExpr.payload, + inputSchema = Array.empty, + outputSchema = Array.empty)) + session.close() Review Comment: we may need a placeholder for session.process - we can't simply call session.close() after session.process as it is an asynchronous operator that returns an operator. We need to properly call session.cancel(), session.close() in the lifecyle of a task here: 1. in most graceful case, we should trigger session.close() when the result iterator has been exhausted. 2. in case of task failure, we should be able to trigger session.cancel() in the failure listener, then call session.close() 3. in case of graceful task completion before result iterator is exhausted (e.g., due to limit), we should trigger a cancel and then a close in task completion listener. -- 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]
