Flink广播变量

• 实时更新配置，例如：任务在统计3个页面的uv，又要统计另外三个页面的uv，那我是不是可以通过配置的方式，快速实现类似需求
• 实时加载维表，例如：kafka里用户购买的订单信息的binlog，但不知道商品id对应的商品，商品也在不断增加，这个时候我需要加载一个维表来做mapping，此时我不想查寻数据库，那怎么做mapping呢？
• 部分流join场景

简单来说就是流合并，一个流作为广播流，一个流作为数据流，即一个大流，一个小流，广播流会将配置或者维表不定时的广播发布，数据流收到广播流的数据后进行相应的操作。
两种用法，一种有key的DataStream(指的是进行过group操作的)，一种无key的DataStream

重写processBroadcastElement将配置进行广播
val state = ctx.getBroadcastState(configStateDescriptor)
state.put(table, config)

重写processElement获取广播数据进行组合
val config: Config = getIndexesConf(ctx, table)
out.collect(indexDocInfos)

  final var configBroadcast = new MapStateDescriptor[String, Int]("mb", classOf[String], classOf[Int])

  def main(args: Array[String]): Unit = {

    val env = StreamExecutionEnvironment.getExecutionEnvironment

    env.setParallelism(1)
    // 流
    val dataStream1: DataStream[String] = env.fromCollection(util.Arrays.asList(
      "aa", "bb", "cc", "aa", "bb", "cc", "aa", "bb", "cc", "aa", "bb", "cc", "aa", "bb", "cc", "aa", "bb", "cc", "aa", "bb", "cc", "aa", "bb", "cc", "aa", "bb", "cc"
    ))


    // 配置
    val dataStream2: BroadcastStream[(String, Int)] = env.fromCollection(util.Arrays.asList(
      ("aa", 1), ("bb", 2), ("cc", 3)
    )).setParallelism(1).broadcast(configBroadcast)

    //合关两个数据流
    val dataStream: DataStream[String] = dataStream1
      .connect(dataStream2)
      .process(new BroadcastProcessFunction[String, (String, Int), String] {

        override def processBroadcastElement(in2: (String, Int), ctx: BroadcastProcessFunction[String, (String, Int), String]#Context, collector: Collector[String]): Unit = {
          val state = ctx.getBroadcastState(configBroadcast)
          state.put(in2._1, in2._2)
        }

        override def processElement(in1: String, ctx: BroadcastProcessFunction[String, (String, Int), String]#ReadOnlyContext, cut: Collector[String]): Unit = {
          val broadCastState = ctx.getBroadcastState(configBroadcast)
          if (!broadCastState.contains(in1)) {
            cut.collect(in1+"123")
          } else {
            cut.collect(in1+broadCastState.get(in1))
          }
        }
      })

    dataStream.print()
    env.execute("job");
  }

结果，一开始会匹配不上。