flink 广播变量

     使用过spark的人都知道广播变量这个概念。广播变量相当于一个共享变量，将一个小数据集复制分发到每个task，task直接从本地读取。flink中有两种广播变量，一种静态的广播变量，一种实时动态的广播变量。

    静态广播变量示例:

      使用场景如: 黑名单判断，将黑名单广播出去进行数据匹配。

public class FlinkBroadcast2 {
    public static void main(String[] args) throws Exception {
        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
        DataSet ds1 = env.fromElements(1, 2, 3, 4);
        DataSource ds2 = env.fromElements(7, 8, 9, 10);
        ds2.map(new RichMapFunction() {
            List list = new ArrayList();
            public void open(Configuration parameters) throws Exception {
               list = getRuntimeContext().getBroadcastVariable("bs");
            }

            @Override
            public String map(Integer integer) throws Exception {

                return integer.intValue()+":"+list;
            }
        }).withBroadcastSet(ds1,"bs").print();

//       env.execute();
    }

 动态广播变量示例:

   使用场景： 数据依赖某些动态变化的处理规则

   广播流一般都是从kafka或其他数据源获取，这里演示直接固定了。从kafka获取流，修改数据后，下游也会更新广播流。

   key streaming 使用KeyedBroadcastProcessFunction.

   非key streaming 使用 BroadcastProcessFunction.

public class FlinkBroadcast {
    public static void main(String[] args) throws Exception {
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        env.setParallelism(1);
        DataStream ds1 = env.fromElements(2);
        DataStreamSource ds2 = env.fromElements(7,8,9,10);
        MapStateDescriptor ruleStateDescriptor = new MapStateDescriptor<>(
                "BroadcastState",
                BasicTypeInfo.STRING_TYPE_INFO,
                TypeInformation.of(Integer.TYPE));

        BroadcastStream ruleBroadcastStream = ds1
                .broadcast(ruleStateDescriptor);
         ds2.connect(ruleBroadcastStream)
                .process(


                        new BroadcastProcessFunction() {

                            @Override
                            public void processElement(Integer integer, ReadOnlyContext readOnlyContext, Collector collector) throws Exception {
                                ReadOnlyBroadcastState state = readOnlyContext.getBroadcastState(ruleStateDescriptor);
                                Integer integer1 = state.get("test");
                                collector.collect(integer+"="+integer1);
                            }

                            @Override
                            public void processBroadcastElement(Integer integer, Context context, Collector collector) throws Exception {
                                context.getBroadcastState(ruleStateDescriptor).put("test",integer);
                            }
                        }
                ).print();
      env.execute();
    }
}