Java版Flink使用指南—

Java版Flink使用指南——合流

大纲

新建工程
无界流
- 奇数Long型无界流
- 偶数Long型无界流
- 奇数String型无界流
合流
- Union
- Connect
测试
工程代码

在《Java版Flink使用指南——分流导出》中，我们通过addSink进行了输出分流。本文我们将介绍几种通过多个无界流输入合并成一个流来进行处理的方案。

新建工程

我们新建一个名字叫MultiSource的工程。
Archetype：org.apache.flink:flink-quickstart-java
版本：1.19.1
在这里插入图片描述

无界流

我们使用《Java版Flink使用指南——自定义无界流生成器》中的方法，我们定义3个无界流。其中两个是Long类型，一个是String类型。

奇数Long型无界流

src/main/java/org/example/generator/UnBoundedOddStreamGenerator.java
这个类每隔1秒钟产生一个Long型奇数。

package org.example.generator;import org.apache.flink.streaming.api.functions.source.RichSourceFunction;public class UnBoundedOddStreamGenerator extends RichSourceFunction<Long> {private volatile boolean isRunning = true;@Overridepublic void run(SourceContext<Long> ctx) throws Exception {long count = 1L;while (isRunning) {Thread.sleep(1000); // Simulate delayctx.collect(count); // Emit datacount = count + 2;}}@Overridepublic void cancel() {isRunning = false;System.out.println("UnBoundedStreamGenerator canceled");}
}

偶数Long型无界流

src/main/java/org/example/generator/UnBoundedEvenStreamGenerator.java
这个类每隔1秒钟产生一个Long型偶数。

package org.example.generator;import org.apache.flink.streaming.api.functions.source.RichSourceFunction;public class UnBoundedEvenStreamGenerator extends RichSourceFunction<Long> {private volatile boolean isRunning = true;@Overridepublic void run(SourceContext<Long> ctx) throws Exception {long count = 0L;while (isRunning) {Thread.sleep(1000); // Simulate delayctx.collect(count); // Emit datacount = count + 2;}}@Overridepublic void cancel() {isRunning = false;System.out.println("UnBoundedStreamGenerator canceled");}
}

奇数String型无界流

src/main/java/org/example/generator/UnBoundedOddStringStreamGenerator.java
这个类每隔1秒钟产生一个String型奇数。

package org.example.generator;import org.apache.flink.streaming.api.functions.source.RichSourceFunction;public class UnBoundedOddStringStreamGenerator extends RichSourceFunction<String> {private volatile boolean isRunning = true;@Overridepublic void run(SourceContext<String> ctx) throws Exception {long count = 1L;while (isRunning) {Thread.sleep(1000); // Simulate delayctx.collect(String.valueOf(count)); // Emit datacount = count + 2;}}@Overridepublic void cancel() {isRunning = false;System.out.println("UnBoundedStreamGenerator canceled");}
}

合流

Union

Union是最简单的算子。它可以把两个数据类型相同的流合并。
上面奇数和偶数Long型流就可以使用Union去做合并。

DataStreamSource<Long> evenLongDataStreamSource = env.addSource(new UnBoundedEvenStreamGenerator());
DataStreamSource<Long> oddLongDataStreamSource = env.addSource(new UnBoundedOddStreamGenerator());evenLongDataStreamSource.union(oddLongDataStreamSource).addSink(new SinkFunction<Long>() {@Overridepublic void invoke(Long value, Context context) throws Exception {System.out.println("sink union value: " + value);}}
).name("union stream");

Connect

Connect可以用于连接两个不同类型的流。这就意味着它需要提供针对不同类型的处理方法。
上面这个例子，如果使用Connect实现，则如下

evenLongDataStreamSource.connect(oddLongDataStreamSource).map(new CoMapFunction<Long, Long, Long>() {@Overridepublic Long map1(Long value) throws Exception {return value;}@Overridepublic Long map2(Long value) throws Exception {return value;}}).addSink(new SinkFunction<Long>() {@Overridepublic void invoke(Long value, Context context) throws Exception {System.out.println("sink connect value: " + value);}}).name("connect stream");

map方法中的CoMapFunction接口类中的map1和map2就是将两个不同类型的流归一化处理的中间方法。
IN1是Connect方法调用者的流数据类型；IN2是Connect参数的流数据类型；R是它们归一化后的类型。

@Public
public interface CoMapFunction<IN1, IN2, OUT> extends Function, Serializable {/*** This method is called for each element in the first of the connected streams.** @param value The stream element* @return The resulting element* @throws Exception The function may throw exceptions which cause the streaming program to fail*     and go into recovery.*/OUT map1(IN1 value) throws Exception;/*** This method is called for each element in the second of the connected streams.** @param value The stream element* @return The resulting element* @throws Exception The function may throw exceptions which cause the streaming program to fail*     and go into recovery.*/OUT map2(IN2 value) throws Exception;
}

假如我们将Long型偶数流和String型奇数流合并，并生成一个Double类型的流，则可以如下

evenLongDataStreamSource.connect(oddStringDataStreamSource).map(new CoMapFunction<Long, String, Double>() {@Overridepublic Double map1(Long value) throws Exception {return Double.valueOf(value);}@Overridepublic Double map2(String value) throws Exception {return Double.valueOf(value);}}).addSink(new SinkFunction<Double>() {@Overridepublic void invoke(Double value, Context context) throws Exception {System.out.println("sink union connect value: " + value);}}
).name("union connect stream");