Flink--广播状态

/**
 *
 * 广播状态的使用场景
 * 1.少量的数据可以使用广播状态(广播状态的数据存储在内存中,如果数据量非常大,不能使用广播状态)
 * 2.广播状态的数据可以动态修改
 *
 * 需要先将维度数据读取过去才能进行关联
 *
 * 如果要广播的数据量非常大,或者根本无法获得数据
 * 为了高效的关联维度数据,可以使用异步IO
 */
public class BroadcastStateDemo {

    public static void main(String[] args) throws Exception {

        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        //维度流
        //c001,图书,INSERT
        //c002,电脑,INSERT
        //c002,文具,UPDATE
        //c002,文具,DELETE
        DataStreamSource<String> lines1 = env.socketTextStream("linux01", 7777);
        //事实流
        //o001,1000,c001
        //o002,2000,c002
        DataStreamSource<String> lines2 = env.socketTextStream("linux01", 8888);
        //期望得到的数据
        //o001,1000,c001,图书

        //对维度流进行处理
        SingleOutputStreamOperator<Tuple3<String, String, String>> tpStream1 = lines1.map(new MapFunction<String, Tuple3<String, String, String>>() {
            @Override
            public Tuple3<String, String, String> map(String input) throws Exception {
                String[] fields = input.split(",");
                return Tuple3.of(fields[0], fields[1], fields[2]);
            }
        });

        //将维度流数据以MapState的形式广播到下游
        MapStateDescriptor<String, String> stateDescriptor = new MapStateDescriptor<>("broadcast-state", Types.STRING, Types.STRING);
        BroadcastStream<Tuple3<String, String, String>> broadcastStream = tpStream1.broadcast(stateDescriptor);

        //对事实流进行处理
        SingleOutputStreamOperator<Tuple3<String, Double, String>> tpStream2 = lines2.map(new MapFunction<String, Tuple3<String, Double, String>>() {
            @Override
            public Tuple3<String, Double, String> map(String input) throws Exception {
                String[] fields = input.split(",");
                return Tuple3.of(fields[0], Double.parseDouble(fields[1]), fields[2]);
            }
        });

        //将两个流connect 可以共享状态
        SingleOutputStreamOperator<Tuple4<String, Double, String, String>> result = tpStream2.connect(broadcastStream).process(new BroadcastProcessFunction<Tuple3<String, Double, String>, Tuple3<String, String, String>, Tuple4<String, Double, String, String>>() {
            //处理事实流
            @Override
            public void processElement(Tuple3<String, Double, String> input, ReadOnlyContext ctx, Collector<Tuple4<String, Double, String, String>> out) throws Exception {
                //获取状态
                ReadOnlyBroadcastState<String, String> broadcastState = ctx.getBroadcastState(stateDescriptor);
                out.collect(Tuple4.of(input.f0, input.f1, input.f2, broadcastState.get(input.f2)));
            }

            //处理广播流
            @Override
            public void processBroadcastElement(Tuple3<String, String, String> input, Context ctx, Collector<Tuple4<String, Double, String, String>> out) throws Exception {
                //获取状态
                BroadcastState<String, String> broadcastState = ctx.getBroadcastState(stateDescriptor);
                String cid = input.f0;
                String cname = input.f1;
                String type = input.f2;
                if ("DELETE".equals(type)) {
                    //删除状态中对应的数据
                    broadcastState.remove(cid);
                } else {
                    //添加,更新状态中的数据
                    broadcastState.put(cid, cname);
                }
            }
        });

        result.print();

        env.execute();


    }
}