轻量级Scala fork连接语法

时间:2010-01-18 08:58:27

标签: scala parallel-processing fork-join

尽管即将推出java 7标准的fork / join框架,但我正在构建一些语法轻量级的辅助方法,以便客户端并行运行代码。 这是一个可运行的主要方法来说明这个想法。

import actors.Futures

object ForkTest2 {



  def main(args: Array[String]) {
    test1
    test2
  }



  def test1 {
    val (a, b, c) =fork({
      Thread.sleep(500)
      println("inside fx1 ",+System.currentTimeMillis)
      true
    }, {
      Thread.sleep(1000)
      println("inside fx2 ",+System.currentTimeMillis)
      "stringResult"
    }, {
      Thread.sleep(1500)
      println("inside fx3 ",+System.currentTimeMillis)
      1
    })

    println(b, a, c)
    true
  }

  def test2 {
    val results = forkAll({
      () =>
              Thread.sleep(500)
              println("inside fx1 ",+System.currentTimeMillis)
              true
    }, {
      () =>
              Thread.sleep(1000)
              println("inside fx2 ",+System.currentTimeMillis)
              "stringResult"
    }, {
      () =>
              Thread.sleep(1500)
              println("inside fx3 ",+System.currentTimeMillis)
              1
    }, {
      () =>
              Thread.sleep(2000)
              println("inside fx4 ",+System.currentTimeMillis)
              1.023
    })

    println(results)
    true
  }

  val tenMinutes = 1000 * 60 * 10

  def fork[A, B, C](
          fx1: => A,
          fx2: => B,
          fx3: => C
          ) = {
    val re1 = Futures.future(fx1)
    val re2 = Futures.future(fx2)
    val re3 = Futures.future(fx3)
    //default wait 10 minutes
    val result = Futures.awaitAll(tenMinutes, re1, re2, re3)
    (
            result(0).asInstanceOf[Option[A]],
            result(1).asInstanceOf[Option[B]],
            result(2).asInstanceOf[Option[C]]

            )
  }

  type fxAny = () => Any

  def forkAll(
          fx1: fxAny*
          ): List[Any] = {
    val results = fx1.toList.map {fx: fxAny => Futures.future(fx())}
    Futures.awaitAll(tenMinutes, results: _*)
  }
}

示例输出

(inside fx1 ,1263804802301)
(inside fx2 ,1263804802801)
(inside fx3 ,1263804803301)
(Some(stringResult),Some(true),Some(1))
(inside fx1 ,1263804803818)
(inside fx2 ,1263804804318)
(inside fx3 ,1263804804818)
(inside fx4 ,1263804805318)
List(Some(true), Some(stringResult), Some(1), Some(1.023))

测试1说明了类型安全返回类型

测试2说明了一个任意输入参数

我希望结合两种测试方法,使客户端代码可以与类型安全返回类型并行运行任意函数。

关于任意函数参数的另一点是:

我认为该行

  type fxAny = () => Any

应该是代码

  type fxAny =  => Any

,但scala编译器不允许我这样做。

任何帮助都表示赞赏。

3 个答案:

答案 0 :(得分:4)

Eric Torreborre在@retronym:

提供的链接中写道
trait LazyParameters { 
  /** transform a value to a zero-arg function returning that value */ 
  implicit def toLazyParameter[T](value: =>T) = new LazyParameter(() => value) 
  /** class holding a value to be evaluated lazily */ 
  class LazyParameter[T](value: ()=>T) { 
    lazy val v = value() 
    def apply() = v 
  } 
} 

这是您的测试的LazyParameter版本:

object ForkTest2 extends LazyParameters {

...

def forkAll(fx1: LazyParameter[Any]*): List[Any] = {
  val results = fx1.toList.map {
    fx: LazyParameter[Any] => Futures.future(fx.apply())}
  Futures.awaitAll(tenMinutes, results: _*)
}

编辑:正如您所注意到的,隐式评估了by-name参数,并且它不会延迟评估延迟。为什么不使用future这个词呢?我个人认为它使代码更具可读性。

import actors.Futures
import actors.Futures.future
import actors.Future

...

def test2 {
  val results = forkAll(
    future {
      Thread.sleep(500)
      println("inside fx1 ",+System.currentTimeMillis)
      true
    },
    future {
      Thread.sleep(1000)
      println("inside fx2 ",+System.currentTimeMillis)
      "stringResult"
    },
    future {
      Thread.sleep(1500)
      println("inside fx3 ",+System.currentTimeMillis)
      1
    },
    future {
      Thread.sleep(2000)
      println("inside fx4 ",+System.currentTimeMillis)
      1.023
    })

  println(results)
  true
}

...

def forkAll(futures: Future[Any]*): List[Any] = {
  println("forkAll")
  Futures.awaitAll(tenMinutes, futures: _*)
}

答案 1 :(得分:2)

由于此错误,您不能将名称调用类型用作重复参数(也称为varargs):https://lampsvn.epfl.ch/trac/scala/ticket/237

请参阅最近的讨论:http://old.nabble.com/Lazy-varargs-td27169264.html

答案 2 :(得分:0)

任何感兴趣的人的示例实施,

对于带有一些函数参数的用例,可能只是定义具有不同参数个数的fork函数列表(遵循元组定义中的思路) 并享受类型安全返回类型

object ForkTest4 extends LazyParameters {
  def main(args: Array[String]) {
    test4
  }

  def test4 {
    println("Begin test 4")
    //Without the explicit call to type conversion would cause early evaluation
    //    val result4 = forkAll({
    val result4 = forkAll(l {
      Thread.sleep(500)
      println("inside fx4 ", +System.currentTimeMillis)
      true
    })
    println(result4)
    true
  }

  val tenMinutes = 1000 * 60 * 10

  def forkAll(fx1: (() => Any)*): List[Any] = {

    val results = fx1.toList.map {

      fx: (() => Any) => {

        val result = Futures.future(fx.apply())

        result
      }
    }
    Futures.awaitAll(tenMinutes, results: _*)
  }

  def l[T](t: => T): (() => T) = () => t

  implicit def implicitLazy[T](t: => T): (() => T) = () => t
}