Flink旁路输出OutputTag

---

前言

除了由 DataStream 操作产生的主要流之外，还可以产生任意数量的旁路输出结果流。结果流中的数据类型不必与主要流中的数据类型相匹配，并且不同旁路输出的类型也可以不同。当你需要拆分数据流时，通常必须复制该数据流，然后从每个流中过滤掉不需要的数据。

使用旁路输出时，首先需要定义用于标识旁路输出流的 OutputTag：

//需要使用匿名内部类，其中T是泛型
OutputTag<T> outputTag = new OutputTag<T>("side-output") {};

可以通过以下方法将数据发送到旁路输出：

• ProcessFunction
• KeyedProcessFunction
• CoProcessFunction
• KeyedCoProcessFunction
• ProcessWindowFunction
• ProcessAllWindowFunction

代码示例

1.流复制

将流复制两份 发到测输出流stream1 和stream2,代码如下（示例）：

import com.alibaba.fastjson.JSONObject;
import org.apache.flink.api.java.utils.ParameterTool;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.functions.ProcessFunction;
import org.apache.flink.util.Collector;
import org.apache.flink.util.OutputTag;

public class SideOutputTest {
    public static final String TYPE = "type";
    public static void main(String[] args) throws Exception {
        //获取执行环境
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        final ParameterTool params = ParameterTool.fromArgs(args);
        String hostName = params.get("hostname", "10.68.8.59");
        int port = params.getInt("port", 9999);
     // nc -l 9999
        DataStream<String> sourceStream = env.socketTextStream(hostName, port, "\n");
        SingleOutputStreamOperator<JSONObject> jsonObjectStream = sourceStream.map(s -> JSONObject.parseObject(s));
        //定义OutputTag
        OutputTag<JSONObject> outputTag1 = new OutputTag<JSONObject>("stream1") {
        };
        OutputTag<JSONObject> outputTag2 = new OutputTag<JSONObject>("stream2") {
        };

        //将流复制两份 发到测输出流stream1 和stream2
        SingleOutputStreamOperator<JSONObject> outputStream = jsonObjectStream.process(new ProcessFunction<JSONObject, JSONObject>() {

            @Override
            public void processElement(JSONObject jsonObject, Context context, Collector<JSONObject> collector)
                    throws Exception {
                context.output(outputTag1, jsonObject);
                context.output(outputTag2, jsonObject);
            }
        });
        DataStream<JSONObject> stream1 = outputStream.getSideOutput(outputTag1);
        DataStream<JSONObject> stream2 = outputStream.getSideOutput(outputTag2);
        //数据去向
        //stream1
        stream1.map(e -> {
            e.put("stream", "stream1");
            return e;
        }).print();
        //stream2
        stream2.map(e -> {
            e.put("stream", "stream2");
            return e;
        }).print();
        env.execute("SocketStreamTest");
    }
}

2.条件分流

可以根据自定义条件将数据分流。

public class SplitDemo {
    public static final OutputTag<Integer> evenTag = new OutputTag<Integer>("even"){};
    public static final OutputTag<Integer> oddTag = new OutputTag<Integer>("odd"){};

    public static void main(String[] args) throws Exception {
        StreamExecutionEnvironment executionEnvironment = StreamExecutionEnvironment.getExecutionEnvironment();

        DataStreamSource<Integer> source = executionEnvironment.fromElements(1, 2, 3, 4, 5);

        SingleOutputStreamOperator<Integer> process = source.process(new ProcessFunction<Integer, Integer>() {
            @Override
            public void processElement(Integer value, ProcessFunction<Integer, Integer>.Context ctx, Collector<Integer> out) throws Exception {
                if (value % 2 == 0) {
                    // 这里不使用out.collect，而是使用ctx.output
                    // 这个方法多了一个参数，可以指定output tag，从而实现数据分流
                    ctx.output(evenTag, value);
                } else {
                    ctx.output(oddTag, value);
                }
            }
        });

        // 依赖OutputTag获取对应的旁路输出
        DataStream<Integer> evenStream = process.getSideOutput(evenTag);
        DataStream<Integer> oddStream = process.getSideOutput(oddTag);

        // 分别打印两个旁路输出流中的数据
        evenStream.process(new ProcessFunction<Integer, String>() {
            @Override
            public void processElement(Integer value, ProcessFunction<Integer, String>.Context ctx, Collector<String> out) throws Exception {
                out.collect("Even: " + value);
            }
        }).print();

        oddStream.process(new ProcessFunction<Integer, String>() {
            @Override
            public void processElement(Integer value, ProcessFunction<Integer, String>.Context ctx, Collector<String> out) throws Exception {
                out.collect("Odd: " + value);
            }
        }).print();

        executionEnvironment.execute();
    }
}

3.迟到数据分流

public class OutOfOrderDemo {
    // 创建tag
    public static final OutputTag<Tuple2<String, Integer>> lateTag = new OutputTag<Tuple2<String, Integer>>("late"){};

    public static void main(String[] args) throws Exception {
        StreamExecutionEnvironment executionEnvironment = StreamExecutionEnvironment.getExecutionEnvironment();
        // 示例数据，其中D乱序，I来迟（H到来的时候认为15000ms之前的数据已经到齐）
        SingleOutputStreamOperator<Tuple2<String, Integer>> source = executionEnvironment.fromElements(
                new Tuple2<>("A", 0),
                new Tuple2<>("B", 1000),
                new Tuple2<>("C", 2000),
                new Tuple2<>("D", 7000),
                new Tuple2<>("E", 3000),
                new Tuple2<>("F", 4000),
                new Tuple2<>("G", 5000),
                new Tuple2<>("H", 20000),
                new Tuple2<>("I", 8000)
        ).assignTimestampsAndWatermarks(WatermarkStrategy.<Tuple2<String, Integer>>forGenerator(new WatermarkGeneratorSupplier<Tuple2<String, Integer>>() {
            // 这里自定义WatermarkGenerator的原因是Flink按照运行时间周期发送watermark，但我们的例子是单次执行的，可以认为数据是一瞬间到来
            // 因此我们改写为每到来一条数据发送一次watermark，watermark的时间戳为数据的事件事件减去5000毫秒，意思是最多容忍数据来迟5000毫秒
            @Override
            public WatermarkGenerator<Tuple2<String, Integer>> createWatermarkGenerator(Context context) {
                return new WatermarkGenerator<Tuple2<String, Integer>>() {
                    @Override
                    public void onEvent(Tuple2<String, Integer> event, long eventTimestamp, WatermarkOutput output) {
                        long watermark = eventTimestamp - 5000L < 0 ? 0L : eventTimestamp - 5000L;
                        output.emitWatermark(new Watermark(watermark));
                    }

                    @Override
                    public void onPeriodicEmit(WatermarkOutput output) {

                    }
                };
            }
        // 取第二个字段为watermark
        }).withTimestampAssigner((element, timestamp) -> element.f1));

        // 窗口大小5秒，允许延迟5秒
        // watermark和allowedLateness的区别是，watermark决定了什么时候窗口数据触发计算，allowedLateness决定什么数据被认为是lateElement，从而发送到sideOutput
        // 设置side output tag
        source.windowAll(TumblingEventTimeWindows.of(Time.seconds(5))).allowedLateness(Time.seconds(5)).sideOutputLateData(lateTag).process(new ProcessAllWindowFunction<Tuple2<String, Integer>, Object, TimeWindow>() {
            @Override
            public void process(ProcessAllWindowFunction<Tuple2<String, Integer>, Object, TimeWindow>.Context context, Iterable<Tuple2<String, Integer>> elements, Collector<Object> out) throws Exception {
                Iterator<Tuple2<String, Integer>> iterator = elements.iterator();
                System.out.println("--------------------");
                while(iterator.hasNext()) {
                    System.out.println(iterator.next());
                }
            }
        // 打印sideoutput流内容
        }).getSideOutput(lateTag).process(new ProcessFunction<Tuple2<String, Integer>, Object>() {
            @Override
            public void processElement(Tuple2<String, Integer> value, ProcessFunction<Tuple2<String, Integer>, Object>.Context ctx, Collector<Object> out) throws Exception {
                System.out.println("Late element: " + value);
            }
        });

        executionEnvironment.execute();
    }
}

---