如何在MasterActor类中的receive函数之外将mapActors变量设置为全局变量？
将f1和f2函数从MapReduceApp发送到MastorActor。
另一种方法是在MastorActor中的receive函数之前创建mapActor。然后当从MapReduceApp接收f1/f2函数时，可以将f1和f2分别发送到mapActor和reduceActor。但如何实现这一点呢？将f1和f2函数传递到mapActor和reduceActor似乎是一个令人头痛的问题。

class MasterActor extends Actor {
  var reduceActors = List[ActorRef]()
  var mapActors: ActorRef = ???
  def receive = {
    case SetMapperReducer(f1, f2, data) =>
      for (i <- 0 until 4) {
        reduceActors = context.actorOf(
          Props(classOf[ReduceActor], f2), name = "reduce" + i
        ) :: reduceActors
      }
      mapActors = context.actorOf(
        RoundRobinPool(4).props(
          Props(classOf[MapActor], reduceActors, f1)
        )
      )
    case Flush =>
      mapActors ! Broadcast(Flush)
}

Map和归约两个类。
Map执行元：

class MapActor(
  reduceActors: List[ActorRef],
  mf: Any => List[Tuple2[Any,Any]]
) extends Actor {
  def receive = {
    case MapData(data: Any) =>
      println(data)
    case Flush =>
      println("Got flush")
  }
}

class ReduceActor(rf: Any => List[Tuple2[Any,Any]]) extends Actor {
  def receive = {
    case Data(data: Any) =>
      println(data)
    case Flush =>
      println("Got flush")
}

还有一个主程序：

object MapReduceApp extends App {
  val system = ActorSystem("TestApp")
  val master = system.actorOf(Props[MasterActor](), name = "master")
  val data = List(
    ("Episode 1", "Once upon a time, in a land far, far away...")
  )
  master ! SetMapperReducer(mapProcess1 _, reduceProcess1 _, data)
}