使用scala参与者框架作为fork-join计算？

Question

从理论上讲，有没有可能使用Scala参与者框架来进行一种类似于JDK 7的Fork-Join框架的异步分而治之的计算？如果是这样，我该如何用框架来表达FJ问题--例如，教程合并概念？欢迎使用代码片段。

(我的想法是基于我在另一台FJ related question上得到的resource video。)

Answer 1

Scala确实有FJ风格的并行性。它是call futures的一部分，也是演员库的一部分

import scala.actors.Future
import scala.actors.Futures._

def mergeSort[A <% Ordered[A]](xs : List[A]) : List[A] =  {
  // merge is not interesting, it's sequential.  The complexity lies in keeping it tail recursive
  def merge[A <% Ordered[A]](accum : List[A], left : List[A], right : List[A]) : List[A] = {
    (left, right) match {
      case (lhead::ltail, rhead::rtail) => 
        if (lhead <= rhead) merge(lhead :: accum, ltail, right)
        else merge(rhead :: accum, left, rtail)
      case (Nil, _) => accum reverse_::: right
      case _ => accum reverse_::: left
    }
  }

    // here's the parallel sort bit
  def sort[A <% Ordered[A]](xs : List[A], length : Int) : List[A] =  {
    if (length <= 1) xs
    else {
      val leftLength = length / 2
      val rightLength = length - leftLength
      val (left, right) = xs splitAt leftLength

      // fork
      val leftFork = future { sort(left, leftLength) }
      val rightFork = future { sort(right, rightLength) }

      // join
      val leftJoin = leftFork()
      val rightJoin = rightFork()

      // merge
      merge(Nil, leftJoin, rightJoin)
    }  
  }

  sort(xs, xs.length)
}

现在，进入问题的核心。如果Scala没有未来，你可以基于参与者自己编写一个吗？确实如此。它看起来或多或少是这样的。

import scala.actors.Actor 
import scala.actors.Actor._

object MyFuture {
  def apply[T](x : => T) : MyFuture[T] = {
    val future = new MyFuture[T]

    val act = actor {
      react {
        case sender : Actor => sender ! (future, x)
      }
    }

    act ! self

    future

  }
}

class MyFuture[A] extends Function0[A] {
  me => 

  lazy val result = receive {
    case (`me`, result) => result.asInstanceOf[A]
  }

  def apply() = result

}

你可以这样使用它

scala> val x = MyFuture(28 * 1000)
x: Foo.MyFuture[Int] = <function>

scala> x()
res4: Int = 28000