EnricoMi commented on a change in pull request #31905: URL: https://github.com/apache/spark/pull/31905#discussion_r654757381
########## File path: sql/core/src/main/scala/org/apache/spark/sql/Observation.scala ########## @@ -0,0 +1,139 @@ +/* + * 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 + +import java.util.UUID +import java.util.concurrent.TimeUnit + +import org.apache.spark.sql.execution.QueryExecution +import org.apache.spark.sql.util.QueryExecutionListener + +/** + * Not thread-safe. + * @param name + * @param sparkSession + */ +class Observation(name: String) { + + private val listener: ObservationListener = ObservationListener(this) + + private var sparkSession: Option[SparkSession] = None + + @volatile private var row: Option[Row] = None + + /** + * Attach this observation to the given Dataset. + * + * @param ds dataset + * @tparam T dataset type + * @return observed dataset + */ + def on[T](ds: Dataset[T])(expr: Column, exprs: Column*): Dataset[T] = { + if (ds.isStreaming) { + throw new IllegalArgumentException("Observation does not support streaming Datasets") + } + register(ds.sparkSession) + ds.observe(name, expr, exprs: _*) + } + + /** + * Wait for the first action on the observed dataset to complete and returns true. + * This method times out after the given amount of time and returns false. + * + * @param time timeout + * @param unit timeout time unit + * @return true if action complete within timeout, false on timeout + */ + def waitCompleted(time: Long, unit: TimeUnit): Boolean = waitCompleted(Some(time), unit) + + /** + * Get the observation results. This waits until the observed dataset finishes its first action. + * If you want to wait for the result and provide a timeout, use waitCompleted. + * Only the result of the first action is available. Subsequent actions do not modify the result. + */ + def get: Row = { + assert(waitCompleted(None, TimeUnit.SECONDS), "waitCompleted without timeout returned false") + row.get + } + + private def waitCompleted(time: Option[Long], unit: TimeUnit): Boolean = { + synchronized { + if (row.isEmpty) { + if (time.isDefined) { + this.wait(unit.toMillis(time.get)) + } else { + this.wait() + } + } + row.isDefined + } + } + + private def register(sparkSession: SparkSession): Unit = { + // makes this class thread-safe: + // only the first thread entering this block can set sparkSession + // all other threads will see the exception, because it is only allowed to do this once + synchronized { + if (this.sparkSession.isDefined) { + throw new IllegalStateException("An Observation can be used with a Dataset only once") + } + this.sparkSession = Some(sparkSession) + } + + sparkSession.listenerManager.register(listener) + } + + private def unregister(): Unit = { + this.sparkSession.foreach(_.listenerManager.unregister(listener)) + } + + private[spark] def onFinish(qe: QueryExecution): Unit = { + synchronized { + this.row = qe.observedMetrics.get(name) + assert(this.row.isDefined, "No metric provided by QueryExecutionListener") + this.notifyAll() + } + unregister() Review comment: > shall we put it in the synchronized block? `` This means we would guard against another thread that is concurrently running `register`. The listener thread can only call into `unregister` after `register` has finished. So yes, two threads are involved in updating and accessing `sparkSession`, but not concurrently. -- 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: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org
