[ https://issues.apache.org/jira/browse/FLINK-4391?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15633127#comment-15633127 ]
ASF GitHub Bot commented on FLINK-4391: --------------------------------------- Github user tillrohrmann commented on a diff in the pull request: https://github.com/apache/flink/pull/2629#discussion_r86356070 --- Diff: flink-streaming-java/src/main/java/org/apache/flink/streaming/api/operators/async/AsyncCollectorBuffer.java --- @@ -0,0 +1,494 @@ +/* + * 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.flink.streaming.api.operators.async; + +import org.apache.flink.annotation.Internal; +import org.apache.flink.streaming.api.datastream.AsyncDataStream; +import org.apache.flink.streaming.api.operators.Output; +import org.apache.flink.streaming.api.operators.TimestampedCollector; +import org.apache.flink.streaming.api.watermark.Watermark; +import org.apache.flink.streaming.runtime.streamrecord.LatencyMarker; +import org.apache.flink.streaming.runtime.streamrecord.StreamElement; +import org.apache.flink.streaming.runtime.streamrecord.StreamRecord; +import org.apache.flink.util.Preconditions; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.util.ArrayList; +import java.util.HashMap; +import java.util.LinkedList; +import java.util.List; +import java.util.Map; +import java.util.concurrent.locks.Condition; +import java.util.concurrent.locks.Lock; +import java.util.concurrent.locks.ReentrantLock; + +/** + * AsyncCollectorBuffer will hold all {@link AsyncCollector} in its internal buffer, + * and emit results from {@link AsyncCollector} to the next operators following it by + * calling {@link Output#collect(Object)} + */ +@Internal +public class AsyncCollectorBuffer<IN, OUT> { + private static final Logger LOG = LoggerFactory.getLogger(AsyncCollectorBuffer.class); + + /** + * Max number of {@link AsyncCollector} in the buffer. + */ + private final int bufferSize; + + private final AsyncDataStream.OutputMode mode; + + private final AsyncWaitOperator<IN, OUT> operator; + + /** + * {@link AsyncCollector} queue. + */ + private final SimpleLinkedList<AsyncCollector<IN, OUT>> queue = new SimpleLinkedList<>(); + /** + * A hash map keeping {@link AsyncCollector} and their corresponding {@link StreamElement} + */ + private final Map<AsyncCollector<IN, OUT>, StreamElement> collectorToStreamElement = new HashMap<>(); + /** + * A hash map keeping {@link AsyncCollector} and their node references in the #queue. + */ + private final Map<AsyncCollector<IN, OUT>, SimpleLinkedList.Node> collectorToQueue = new HashMap<>(); + + private final LinkedList<AsyncCollector> finishedCollectors = new LinkedList<>(); + + /** + * {@link TimestampedCollector} and {@link Output} to collect results and watermarks. + */ + private TimestampedCollector<OUT> timestampedCollector; + private Output<StreamRecord<OUT>> output; + + /** + * Locks and conditions to synchronize with main thread and emitter thread. + */ + private final Lock lock; + private final Condition notFull; + private final Condition taskDone; + private final Condition isEmpty; + + /** + * Error from user codes. + */ + private volatile Exception error; + + private final Emitter emitter; + private final Thread emitThread; + + private boolean isCheckpointing; + + public AsyncCollectorBuffer(int maxSize, AsyncDataStream.OutputMode mode, AsyncWaitOperator operator) { + Preconditions.checkArgument(maxSize > 0, "Future buffer size should be greater than 0."); + + this.bufferSize = maxSize; + this.mode = mode; + this.operator = operator; + + this.lock = new ReentrantLock(true); + this.notFull = this.lock.newCondition(); + this.taskDone = this.lock.newCondition(); + this.isEmpty = this.lock.newCondition(); + + this.emitter = new Emitter(); + this.emitThread = new Thread(emitter); + } + + /** + * Add an {@link StreamRecord} into the buffer. A new {@link AsyncCollector} will be created and returned + * corresponding to the input StreamRecord. + * <p> + * If buffer is full, caller will wait until new space is available. + * + * @param record StreamRecord + * @return An AsyncCollector + * @throws Exception InterruptedException or exceptions from AsyncCollector. + */ + public AsyncCollector<IN, OUT> add(StreamRecord<IN> record) throws Exception { + try { + lock.lock(); + + notifyCheckpointDone(); + + while (queue.size() >= bufferSize) { + notFull.await(); + } + + // propagate error to the main thread + if (error != null) { + throw error; + } + + AsyncCollector<IN, OUT> collector = new AsyncCollector(this); + + collectorToQueue.put(collector, queue.add(collector)); + collectorToStreamElement.put(collector, record); + + return collector; + } + finally { + lock.unlock(); + } + } + + /** + * Add a {@link Watermark} into queue. A new AsyncCollector will be created and returned. + * <p> + * If queue is full, caller will be blocked here. + * + * @param watermark Watermark + * @return AsyncCollector + * @throws Exception Exceptions from async operation. + */ + public AsyncCollector<IN, OUT> add(Watermark watermark) throws Exception { + return processMark(watermark); + } + + /** + * Add a {@link LatencyMarker} into queue. A new AsyncCollector will be created and returned. + * <p> + * If queue is full, caller will be blocked here. + * + * @param latencyMarker LatencyMarker + * @return AsyncCollector + * @throws Exception Exceptions from async operation. + */ + public AsyncCollector<IN, OUT> add(LatencyMarker latencyMarker) throws Exception { + return processMark(latencyMarker); + } + + private AsyncCollector<IN, OUT> processMark(StreamElement mark) throws Exception { + try { + lock.lock(); + + notifyCheckpointDone(); + + while (queue.size() >= bufferSize) + notFull.await(); + + if (error != null) { + throw error; + } + + AsyncCollector<IN, OUT> collector = new AsyncCollector(this, true); + + collectorToQueue.put(collector, queue.add(collector)); + collectorToStreamElement.put(collector, mark); + + // signal emitter thread that current collector is ready + mark(collector); + + return collector; + } + finally { + lock.unlock(); + } + } + + /** + * Notify the Emitter Thread that an AsyncCollector has completed. + * + * @param collector Completed AsyncCollector + */ + void mark(AsyncCollector<IN, OUT> collector) { + try { + lock.lock(); + + if (mode == AsyncDataStream.OutputMode.UNORDERED) { + finishedCollectors.add(collector); --- End diff -- I think that `OutputMode.UNORDERED` won't work with event time processing. The watermarks can easily overtake elements with a smaller timestamp. Either way, we don't allow the `OutputMode.UNORDERED` when event time was chosen, or we handle watermarks always in order. This means that the watermark is only emitted after all preceding elements have been emitted (potentially in an unordered fashion). > Provide support for asynchronous operations over streams > -------------------------------------------------------- > > Key: FLINK-4391 > URL: https://issues.apache.org/jira/browse/FLINK-4391 > Project: Flink > Issue Type: New Feature > Components: DataStream API > Reporter: Jamie Grier > Assignee: david.wang > > Many Flink users need to do asynchronous processing driven by data from a > DataStream. The classic example would be joining against an external > database in order to enrich a stream with extra information. > It would be nice to add general support for this type of operation in the > Flink API. Ideally this could simply take the form of a new operator that > manages async operations, keeps so many of them in flight, and then emits > results to downstream operators as the async operations complete. -- This message was sent by Atlassian JIRA (v6.3.4#6332)