4. 广播变量

一、分区规则（DataStream Broadcast）和广播变量（Flink Broadcast）

1.1 DataStream Broadcast（分区规则）

​ 分区规则是把元素广播给所有的分区，数据会被重复处理。

DataStream.broadcast()

1.2 Flink Broadcast（广播变量）

​ 类似于Spark广播变量，广播的数据是Dataset，接收广播的也是Dataset

import org.apache.flink.api.common.functions.RichMapFunction
import org.apache.flink.api.scala.{DataSet, ExecutionEnvironment}
import org.apache.flink.configuration.Configuration
import org.apache.flink.streaming.api.scala.{StreamExecutionEnvironment, createTypeInformation}


object BroadCastTest1 {
  def main(args: Array[String]): Unit = {
    val env = ExecutionEnvironment.getExecutionEnvironment  //创建批处理执行环境

    val broadcastData = List(("zs", 18), ("ls", 28), ("ww", 38)) // 创建要广播的dataset
    val tupleData = env.fromCollection(broadcastData)
    val toBroadcastData = tupleData.map(tup => {
      Map(tup._1->tup._2)
    })


    val text: DataSet[String] = env.fromElements("zs", "ls", "ww")  //创建接收广播的dataset
    val result = text.map(new RichMapFunction[String, String] {
      var listData: java.util.List[Map[String, Int]] = null
      var allMap = Map[String, Int]()

      override def open(parameters: Configuration): Unit = {
        this.listData = getRuntimeContext.getBroadcastVariable[Map[String, Int]]("bd")	//获取broadcast

        val it = listData.iterator()

        while(it.hasNext) {
          val next = it.next()
          allMap = allMap.++(next)
        }
      }
      override def map(value: String): String = {
        val age = allMap.getOrElse(value, 1)
        value + ", " + age
      }
    }).withBroadcastSet(toBroadcastData, "bd")

    result.print

    env.execute()
  }
}