flink如何正确分流

flink如何正确分流

分流方式

1. filter分流
2. split分流 (只能一次分流，分流后的流不能继续分流)
3. side output分流 (推荐使用)

场景

 

输入数据：

 

{"key":"001","type":"1","data":"data1"}
{"key":"001","type":"11","data":"data11"}
{"key":"001","type":"12","data":"data12"}
{"key":"002","type":"2","data":"data2"}
{"key":"002","type":"21","data":"data21"}
{"key":"002","type":"22","data":"data22"}

Operator中处理：转为jsonObject处理,并在jsonObject对象中添加流标志；

pom中需要引入fastjson (fastjson使用教程);

 

<dependency>
  <groupId>com.alibaba</groupId>
  <artifactId>fastjson</artifactId>
  <version>1.2.60</version>
</dependency>

输出结果为：

11> {"data":"data1","stream":"stream1","type":"1","key":"001"}
12> {"data":"data11","stream":"stream1","type":"11","key":"001"}
12> {"data":"data11","stream":"stream11","type":"11","key":"001"}
1> {"data":"data12","stream":"stream1","type":"12","key":"001"}
1> {"data":"data12","stream":"stream12","type":"12","key":"001"}
2> {"data":"data2","stream":"stream2","type":"2","key":"002"}
3> {"data":"data21","stream":"stream2","type":"21","key":"002"}
3> {"data":"data21","stream":"stream21","type":"21","key":"002"}
4> {"data":"data22","stream":"stream2","type":"22","key":"002"}
4> {"data":"data22","stream":"stream22","type":"22","key":"002"}

发送数据客户端

nc程序发送(nc程序实现)

Fliter分流

代码实现：

 

public class FliterStreamTest {

    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", "localhost");
        int port = params.getInt("port", 9000);

        DataStream<String> sourceStream = env.socketTextStream(hostName, port, "\n");

        SingleOutputStreamOperator<JSONObject> jsonObjectStream = sourceStream.map(s -> JSONObject.parseObject(s));

        DataStream<JSONObject> stream1 = jsonObjectStream
                .filter(e -> e.get(TYPE).equals("1") || e.get(TYPE).equals("11") || e.get(TYPE).equals("12"));
        DataStream<JSONObject> stream11 = stream1.filter(e -> e.get(TYPE).equals("11"));
        DataStream<JSONObject> stream12 = stream1.filter(e -> e.get(TYPE).equals("12"));

        DataStream<JSONObject> stream2 = jsonObjectStream
                .filter(e -> e.get(TYPE).equals("2") || e.get(TYPE).equals("21") || e.get(TYPE).equals("22"));
        DataStream<JSONObject> stream21 = stream2.filter(e -> e.get(TYPE).equals("21"));
        DataStream<JSONObject> stream22 = stream2.filter(e -> e.get(TYPE).equals("22"));

        //数据去向
        stream1.map(e -> {e.put("stream", "stream1");return e;}).print();
        stream11.map(e -> {e.put("stream", "stream11");return e;}).print();
        stream12.map(e -> {e.put("stream", "stream12");return e;}).print();
        stream2.map(e -> {e.put("stream", "stream2");return e;}).print();
        stream21.map(e -> {e.put("stream", "stream21");return e;}).print();
        stream22.map(e -> {e.put("stream", "stream22");return e;}).print();

        env.execute("SocketStreamTest");
    }
}

输出结果(与正确结果一直)：

11> {"data":"data1","stream":"stream1","type":"1","key":"001"}
12> {"data":"data11","stream":"stream1","type":"11","key":"001"}
12> {"data":"data11","stream":"stream11","type":"11","key":"001"}
1> {"data":"data12","stream":"stream1","type":"12","key":"001"}
1> {"data":"data12","stream":"stream12","type":"12","key":"001"}
2> {"data":"data2","stream":"stream2","type":"2","key":"002"}
3> {"data":"data21","stream":"stream2","type":"21","key":"002"}
3> {"data":"data21","stream":"stream21","type":"21","key":"002"}
4> {"data":"data22","stream":"stream2","type":"22","key":"002"}
4> {"data":"data22","stream":"stream22","type":"22","key":"002"}

Split分流

代码实现：

 

public class SplitStreamTest {

    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", "localhost");
        int port = params.getInt("port", 9000);

        DataStream<String> sourceStream = env.socketTextStream(hostName, port, "\n");

        SingleOutputStreamOperator<JSONObject> jsonObjectStream = sourceStream.map(s -> JSONObject.parseObject(s));

        //flink split流已被标记为过期
        SplitStream<JSONObject> splitStream = jsonObjectStream.split(new OutputSelector<JSONObject>() {
            @Override
            public Iterable<String> select(JSONObject jsonObject) {
                List<String> tags = new ArrayList<>();
                String type = jsonObject.get("type").toString();
                if (type.equals("1") || type.equals("11") || type.equals("12")) {
                    tags.add("stream1");
                } else if (type.equals("2") || type.equals("21") || type.equals("22")) {
                    tags.add("stream2");
                }
                return tags;
            }
        });

        DataStream<JSONObject> stream1 = splitStream.select("stream1");

        DataStream<JSONObject> stream2 = splitStream.select("stream2");

        SplitStream<JSONObject> splitStream1 = stream1.split(new OutputSelector<JSONObject>() {
            @Override
            public Iterable<String> select(JSONObject jsonObject) {
                List<String> tags = new ArrayList<>();
                String type = jsonObject.get("type").toString();
                if (type.equals("11")) {
                    tags.add("stream11");
                } else if (type.equals("12")) {
                    tags.add("stream12");
                }
                return tags;
            }
        });

        SplitStream<JSONObject> splitStream2 = stream1.split(new OutputSelector<JSONObject>() {
            @Override
            public Iterable<String> select(JSONObject jsonObject) {
                List<String> tags = new ArrayList<>();
                String type = jsonObject.get("type").toString();
                if (type.equals("21")) {
                    tags.add("stream21");
                } else if (type.equals("22")) {
                    tags.add("stream22");
                }
                return tags;
            }
        });

        DataStream<JSONObject> stream11 = splitStream1.select("stream11");
        DataStream<JSONObject> stream12 = splitStream1.select("stream12");
        DataStream<JSONObject> stream21 = splitStream2.select("stream21");
        DataStream<JSONObject> stream22 = splitStream2.select("stream22");

        //数据去向
        stream1.map(e -> {e.put("stream", "stream1");return e;}).print();
        stream11.map(e -> {e.put("stream", "stream11");return e;}).print();
        stream12.map(e -> {e.put("stream", "stream12");return e;}).print();
        stream2.map(e -> {e.put("stream", "stream2");return e;}).print();
        stream21.map(e -> {e.put("stream", "stream21");return e;}).print();
        stream22.map(e -> {e.put("stream", "stream22");return e;}).print();

        env.execute("SocketStreamTest");
    }

}

输出结果：

Flink 1.9以前可以输出，但是第二次split将会失效；Flink 1.9 如果二次split，直接报错并提示使用side-outputs

Exception in thread "main" java.lang.IllegalStateException: Consecutive multiple splits are not supported. Splits are deprecated. Please use side-outputs.

Side-outputs分流(推荐使用)

步骤：

1. 定义 OutputTag
2. process 算子中收集数据
   1. ProcessFunction
   2. KeyedProcessFunction
   3. CoProcessFunction
   4. ProcessWindowFunction
   5. ProcessAllWindowFunction

代码实现：

 

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", "localhost");
        int port = params.getInt("port", 9000);

        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") {};

        SingleOutputStreamOperator<JSONObject> outputStream = jsonObjectStream.process(new ProcessFunction<JSONObject, JSONObject>() {

            @Override
            public void processElement(JSONObject jsonObject, Context context, Collector<JSONObject> collector)
                    throws Exception {
                String type = jsonObject.getString(TYPE);
                if (type.equals("1") || type.equals("11") || type.equals("12")) {
                    context.output(outputTag1, jsonObject);
                } else if (type.equals("2") || type.equals("21") || type.equals("22")) {
                    context.output(outputTag2, jsonObject);
                }
            }
        });

        DataStream<JSONObject> stream1 = outputStream.getSideOutput(outputTag1);
        DataStream<JSONObject> stream2 = outputStream.getSideOutput(outputTag2);

        //定义OutputTag
        OutputTag<JSONObject> outputTag11 = new OutputTag<JSONObject>("stream11") {};
        OutputTag<JSONObject> outputTag12 = new OutputTag<JSONObject>("stream12") {};

        SingleOutputStreamOperator<JSONObject> outputStream1 = stream1.process(new ProcessFunction<JSONObject, JSONObject>() {

            @Override
            public void processElement(JSONObject jsonObject, Context context, Collector<JSONObject> collector)
                    throws Exception {
                String type = jsonObject.getString(TYPE);
                if (type.equals("11")) {
                    context.output(outputTag11, jsonObject);
                } else if (type.equals("12")) {
                    context.output(outputTag12, jsonObject);
                }
            }
        });

        DataStream<JSONObject> stream11 = outputStream1.getSideOutput(outputTag11);
        DataStream<JSONObject> stream12 = outputStream1.getSideOutput(outputTag12);

        //定义OutputTag
        OutputTag<JSONObject> outputTag21 = new OutputTag<JSONObject>("stream21") {};
        OutputTag<JSONObject> outputTag22 = new OutputTag<JSONObject>("stream22") {};

        SingleOutputStreamOperator<JSONObject> outputStream2 = stream2.process(new ProcessFunction<JSONObject, JSONObject>() {

            @Override
            public void processElement(JSONObject jsonObject, Context context, Collector<JSONObject> collector)
                    throws Exception {
                String type = jsonObject.getString(TYPE);
                if (type.equals("21")) {
                    context.output(outputTag21, jsonObject);
                } else if (type.equals("22")) {
                    context.output(outputTag22, jsonObject);
                }
            }
        });

        DataStream<JSONObject> stream21 = outputStream2.getSideOutput(outputTag21);
        DataStream<JSONObject> stream22 = outputStream2.getSideOutput(outputTag22);

        //数据去向
        stream1.map(e -> {e.put("stream", "stream1");return e;}).print();
        stream11.map(e -> {e.put("stream", "stream11");return e;}).print();
        stream12.map(e -> {e.put("stream", "stream12");return e;}).print();
        stream2.map(e -> {e.put("stream", "stream2");return e;}).print();
        stream21.map(e -> {e.put("stream", "stream21");return e;}).print();
        stream22.map(e -> {e.put("stream", "stream22");return e;}).print();

        env.execute("SocketStreamTest");
    }
}

输出结果(与结果一致)：

3> {"data":"data1","stream":"stream1","type":"1","key":"001"}
4> {"data":"data11","stream":"stream11","type":"11","key":"001"}
4> {"data":"data11","stream":"stream1","type":"11","key":"001"}
5> {"data":"data12","stream":"stream12","type":"12","key":"001"}
5> {"data":"data12","stream":"stream1","type":"12","key":"001"}
6> {"data":"data2","stream":"stream2","type":"2","key":"002"}
7> {"data":"data21","stream":"stream21","type":"21","key":"002"}
7> {"data":"data21","stream":"stream2","type":"21","key":"002"}
8> {"data":"data22","stream":"stream22","type":"22","key":"002"}
8> {"data":"data22","stream":"stream2","type":"22","key":"002"}

作者：蜗牛写java
链接：https://www.jianshu.com/p/274d0b78d378
来源：简书
著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。