Question

我正在开始使用flink并研究one of the official tutorials。

据我所知，本练习的目的是在time属性上加入两个流。

任务：

此练习的结果是Tuple2记录的数据流，每个不同的rideId一个。您应该忽略   END事件，并且只能在每次骑行的START时加入该事件   其相应的票价数据。

结果流应打印为标准输出。

问题：EnrichmentFunction如何又可以将两个流合并在一起。它怎么知道参加哪个游乐项目的公平？我希望它可以缓冲多个博览会/竞赛，直到传入的博览会/竞赛有一个匹配的伙伴。

以我的理解，它只是保存了它看到的每一个乘车/展览，并将其与下一个最佳乘车/展览结合在一起。为什么这是适当的联接？

提供的解决方案：

/*
 * Copyright 2017 data Artisans GmbH
 *
 * Licensed 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 com.dataartisans.flinktraining.solutions.datastream_java.state;

import com.dataartisans.flinktraining.exercises.datastream_java.datatypes.TaxiFare;
import com.dataartisans.flinktraining.exercises.datastream_java.datatypes.TaxiRide;
import com.dataartisans.flinktraining.exercises.datastream_java.sources.TaxiFareSource;
import com.dataartisans.flinktraining.exercises.datastream_java.sources.TaxiRideSource;
import com.dataartisans.flinktraining.exercises.datastream_java.utils.ExerciseBase;
import org.apache.flink.api.common.state.ValueState;
import org.apache.flink.api.common.state.ValueStateDescriptor;
import org.apache.flink.api.java.tuple.Tuple2;
import org.apache.flink.api.java.utils.ParameterTool;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.streaming.api.TimeCharacteristic;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.functions.co.RichCoFlatMapFunction;
import org.apache.flink.util.Collector;

/**
 * Java reference implementation for the "Stateful Enrichment" exercise of the Flink training
 * (http://training.data-artisans.com).
 *
 * The goal for this exercise is to enrich TaxiRides with fare information.
 *
 * Parameters:
 * -rides path-to-input-file
 * -fares path-to-input-file
 *
 */
public class RidesAndFaresSolution extends ExerciseBase {
    public static void main(String[] args) throws Exception {

        ParameterTool params = ParameterTool.fromArgs(args);
        final String ridesFile = params.get("rides", pathToRideData);
        final String faresFile = params.get("fares", pathToFareData);

        final int delay = 60;                   // at most 60 seconds of delay
        final int servingSpeedFactor = 1800;    // 30 minutes worth of events are served every second

        // set up streaming execution environment
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        env.setStreamTimeCharacteristic(TimeCharacteristic.EventTime);
        env.setParallelism(ExerciseBase.parallelism);

        DataStream<TaxiRide> rides = env
                .addSource(rideSourceOrTest(new TaxiRideSource(ridesFile, delay, servingSpeedFactor)))
                .filter((TaxiRide ride) -> ride.isStart)
                .keyBy("rideId");

        DataStream<TaxiFare> fares = env
                .addSource(fareSourceOrTest(new TaxiFareSource(faresFile, delay, servingSpeedFactor)))
                .keyBy("rideId");

        DataStream<Tuple2<TaxiRide, TaxiFare>> enrichedRides = rides
                .connect(fares)
                .flatMap(new EnrichmentFunction());

        printOrTest(enrichedRides);

        env.execute("Join Rides with Fares (java RichCoFlatMap)");
    }

    public static class EnrichmentFunction extends RichCoFlatMapFunction<TaxiRide, TaxiFare, Tuple2<TaxiRide, TaxiFare>> {
        // keyed, managed state
        private ValueState<TaxiRide> rideState;
        private ValueState<TaxiFare> fareState;

        @Override
        public void open(Configuration config) {
            rideState = getRuntimeContext().getState(new ValueStateDescriptor<>("saved ride", TaxiRide.class));
            fareState = getRuntimeContext().getState(new ValueStateDescriptor<>("saved fare", TaxiFare.class));
        }

        @Override
        public void flatMap1(TaxiRide ride, Collector<Tuple2<TaxiRide, TaxiFare>> out) throws Exception {
            TaxiFare fare = fareState.value();
            if (fare != null) {
                fareState.clear();
                out.collect(new Tuple2(ride, fare));
            } else {
                rideState.update(ride);
            }
        }

        @Override
        public void flatMap2(TaxiFare fare, Collector<Tuple2<TaxiRide, TaxiFare>> out) throws Exception {
            TaxiRide ride = rideState.value();
            if (ride != null) {
                rideState.clear();
                out.collect(new Tuple2(ride, fare));
            } else {
                fareState.update(fare);
            }
        }
    }
}

Answer 1

在此特定的训练练习中，每个rideId值都有三个事件-TaxiRide开始事件，TaxiRide结束事件和TaxiFare。本练习的目的是将每个TaxiRide启动事件与具有相同rideId的一个TaxiFare事件相关联-换句话说，在rideId上加入乘车流和票价流，同时知道每个将只有一个。

此练习演示了Flink中的键控状态如何工作。键控状态实际上是分片键值存储。当我们有一个ValueState项，例如ValueState<TaxiRide> rideState时，Flink将在状态后端为密钥的每个不同值（rideId）存储一个单独的记录。

每次调用flatMap1和flatMap2时，都会在上下文中隐含一个键（一个rideId），当我们调用rideState.update(ride)或rideState.value()时，不是访问单个变量，而是使用rideId作为键在键值存储中设置和获取条目。

在此练习中，两个流都由rideId键控，因此每个不同的rideState可能都有fareState的一个元素和rideId的一个元素。因此，提供的解决方案是缓冲大量的游乐设施和票价，但是每个rideId仅缓冲一个（这足以满足条件，因为游乐设施和票价在此数据集中是完美配对的。）

所以，你问：

EnrichmentFunction如何又可以将两个流加入。它怎么知道哪种票价和哪种搭车？

答案是

它加入具有相同rideId的票价。

您询问过的这个特定练习显示了如何实现简单的扩充连接，以了解键控状态和连接的流的思想。但是使用Flink当然可以进行更复杂的连接。有关更多信息，请参见the docs on joining，joins with Flink's Table API，joins with Flink SQL和exercise on time-based enrichment joins。

如何在Apache Flink中加入两个流？

1 个答案: