Scalaz（44）－ concurrency ：scalaz Future，尚不完整的多线程类型

本文链接：https://blog.csdn.net/TIGER_XC/article/details/51668456

scala已经配备了自身的Future类。我们先举个例子来了解scala Future的具体操作：

import scala.concurrent._
import ExecutionContext.Implicits.global
object scalafuture {
  def dbl(i: Int): Future[Int] = Future { Thread.sleep(1000) ; i + i }
                                      //> dbl: (i: Int)scala.concurrent.Future[Int]
  val fdbl = dbl(3)                   //> fdbl  : scala.concurrent.Future[Int] = List()
  fdbl.onSuccess {
    case a => println(s"${a/2} + ${a/2} = $a")
  }
  println("calculating ...")          //> calculating ...
  Thread.sleep(2000)                  //> 3 + 3 = 6
}

这是一个标准的异步运算；在成功完成运算事件上绑定callback来获取在其它线程中的运算结果。我们也可以进行异常处理：

  val fdz = Future { 3 / 0 }      //> fdz  : scala.concurrent.Future[Int] = List()
  fdz.onFailure {
    case e => println(s"error message {${e.getMessage}}")
  }
  Thread.sleep(100)               //> error message {/ by zero}

又或者同时绑定运算成功和失败事件的callback函数：

  import scala.util.{Success, Failure}
  fdz onComplete {
    case Success(a) => println(s"${a/2} + ${a/2} = $a")
    case Failure(e) => println(s"error message {${e.getMessage}}")
  }
  Thread.sleep(100)               //> error message {/ by zero}

scala Future 实现了flatMap，我们可以把几个Future组合起来用：

  def dbl(i: Int): Future[Int] = Future { Thread.sleep(1000); i + i }
                                                  //> dbl: (i: Int)scala.concurrent.Future[Int]
  def sqr(i: Int): Future[Int] = Future { i * i } //> sqr: (i: Int)scala.concurrent.Future[Int]
  def sum(a: Int, b: Int): Future[Int] = Future { a + b }
                                          //> sum: (a: Int, b: Int)scala.concurrent.Future[Int]
  val fsum = for {
    a <- dbl(3)
    b <- sqr(a)
    c <- sum(a,b)
  } yield c                               //> fsum  : scala.concurrent.Future[Int] = List()
  
  fsum onSuccess { case c => println(s"the combined result is: $c") }
  Thread.sleep(2000)                     //> the combined result is: 42

scala Future利用flatMap实现了流程运算：先运算dbl再sqr再sum，这个顺序是固定的即使它们可能在不同的线程里运算，因为sqr依赖dbl的结果，而sum又依赖dbl和sqr的结果。

好了，既然scala Future的功能已经比较完善了，那么scalaz的Future又有什么不同的特点呢？首先，细心一点可以发现scala Future是即时运算的，从下面的例子里可以看出：

  import scala.concurrent.duration._
  val fs = Future {println("run now..."); System.currentTimeMillis() }
                                         //> run now...
                                         //| fs  : scala.concurrent.Future[Long] = List()
  Await.result(fs, 1.second)             //> res0: Long = 1465907784714
  Thread.sleep(1000)
  Await.result(fs, 1.second)             //> res1: Long = 1465907784714

可以看到fs是在Future构建时即时运算的，而且只会运算一次。如果scala Future中包括了能产生副作用的代码，在构建时就会立即产生副作用。所以我们是无法使用scala Future来编写纯函数的，那么在scalaz里就必须为并发编程提供一个与scala Future具同等功能但又不会立即产生副作用的类型了，这就是scalaz版本的Future。我们看看scalaz是如何定义Future的：scalaz.concurrent/Future.scala

sealed abstract class Future[+A] {
...
object Future {
  case class Now[+A](a: A) extends Future[A]
  case class Async[+A](onFinish: (A => Trampoline[Unit]) => Unit) extends Future[A]
  case class Suspend[+A](thunk: () => Future[A]) extends Future[A]
  case class BindSuspend[A,B](thunk: () => Future[A], f: A => Future[B]) extends Future[B]
  case class BindAsync[A,B](onFinish: (A => Trampoline[Unit]) => Unit,
                            f: A => Future[B]) extends Future[B]
...

Future[A]就是个Free Monad。它的结构化表达方式分别有Now,Async,Suspend,BindSuspend,BindAsync。我们可以用这些结构实现flatMap函数，所以Future就是Free Monad：

  def flatMap[B](f: A => Future[B]): Future[B] = this match {
    case Now(a) => Suspend(() => f(a))
    case Suspend(thunk) => BindSuspend(thunk, f)
    case Async(listen) => BindAsync(listen, f)
    case BindSuspend(thunk, g) =>
      Suspend(() => BindSuspend(thunk, g andThen (_ flatMap f)))
    case BindAsync(listen, g) =>
      Suspend(() => BindAsync(listen, g andThen (_ flatMap f)))
  }

free structure类型可以支持算式／算法关注分离，也就是说我们可以用scalaz Future来描述程序功能而不涉及正真运算。scalaz Future的构建方式如下：

import scalaz._
import Scalaz._
import scalaz.concurrent._
import scala.concurrent.duration._
object scalazFuture {
val fnow = Future.now {println("run..."); System.currentTimeMillis()}
                                          //> run...
                                          //| fnow  : scalaz.concurrent.Future[Long] = Now(1465909860301)
val fdelay = Future.delay {println("run..."); System.currentTimeMillis()}
                                          //> fdelay  : scalaz.concurrent.Future[Long] = Suspend(<function0>)
val fapply = Future {println("run..."); System.currentTimeMillis()}
                                          //> fapply  : scalaz.concurrent.Future[Long] = Async(<function1>)

可以看到fnow是个即时运算的构建器，而这个now就是一个lift函数, 它负责把一个普通无副作用运算升格成Future。fdelay,fapply分别把运算存入trampoline进行结构化了。我们必须另外运算trampoline来运行结构内的运算：

fdelay.run                                        //> run...
                                                  //| res0: Long = 1465910524847
Thread.sleep(1000)
fdelay.run                                        //> run...
                                                  //| res1: Long = 1465910525881
fapply.run                                        //> run...
                                                  //| res2: Long = 1465910525883
Thread.sleep(1000)
fapply.run                                        //> run...
                                                  //| res3: Long = 1465910526884

scalaz Future只有在运算时才会产生副作用，而且可以多次运算。

我们可以用即时（blocking）、异步、定时方式来运算Future：

fapply.unsafePerformSync                          //> run...
                                                  //| res4: Long = 1465958049118
fapply.unsafePerformAsync {
  case a => println(a)
}
Thread.sleep(1000)
fapply.unsafePerformSyncFor(1 second)             //> run...
                                                  //| 1465958051126
                                                  //| run...
                                                  //| res5: Long = 1465958052172

结构化状态Async代表了scalaz Future的多线程处理特性：

/**
   * Create a `Future` from an asynchronous computation, which takes the form
   * of a function with which we can register a callback. This can be used
   * to translate from a callback-based API to a straightforward monadic
   * version. See `Task.async` for a version that allows for asynchronous
   * exceptions.
   */
  def async[A](listen: (A => Unit) => Unit): Future[A] =
    Async((cb: A => Trampoline[Unit]) => listen { a => cb(a).run })

  /** Create a `Future` that will evaluate `a` using the given `ExecutorService`. */
  def apply[A](a: => A)(implicit pool: ExecutorService = Strategy.DefaultExecutorService): Future[A] = Async { cb =>
    pool.submit { new Callable[Unit] { def call = cb(a).run }}
  }

  /** Create a `Future` that will evaluate `a` after at least the given delay. */
  def schedule[A](a: => A, delay: Duration)(implicit pool: ScheduledExecutorService =
      Strategy.DefaultTimeoutScheduler): Future[A] =
    Async { cb =>
      pool.schedule(new Callable[Unit] {
        def call = cb(a).run
      }, delay.toMillis, TimeUnit.MILLISECONDS)
    }

我们看到apply和schedule在构建Future时对运算线程进行了配置。

如果我们需要模仿scala Future的功效可以用unsafeStart:

val fs = fapply.unsafeStart              //> run...
                                         //| fs  : scalaz.concurrent.Future[Long] = Suspend(<function0>)
fs.run                                   //> res6: Long = 1465958922401
Thread.sleep(1000)
fs.run                                   //> res7: Long = 1465958922401

我们也可以用scala Future的callback方式用async函数把自定义的callback挂在构建的Future上：

def fu(t: Long): Future[String] =
  Future.async[String]{k => k(s"the curreent time is: ${t.toString}!!!")}
                                                  //> fu: (t: Long)scalaz.concurrent.Future[String]
fu(System.currentTimeMillis()).run                //> res8: String = the curreent time is: 1465958923415!!!

scala Future和scalaz Future之间可以相互转换：

import scala.concurrent.{Future => sFuture}
import scala.concurrent.ExecutionContext
import scala.util.{Success,Failure}
def futureTozFuture[A](sf: sFuture[A])(implicit ec: ExecutionContext): Future[A] =
  Future.async {cb => sf.onComplete {
    case Success(a) => cb(a)
//    case Failure(e) => cb(e)
  }}                            //> futureTozFuture: [A](sf: scala.concurrent.Future[A])(implicit ec: scala.con
                                //| current.ExecutionContext)scalaz.concurrent.Future[A]
def zFutureTosFuture[A](zf: Future[A]): sFuture[A] = {
  val prom = scala.concurrent.Promise[A]
  zf.unsafePerformAsync {
     case a => prom.success(a)是
  }
  prom.future
}

突然发现scalaz Future是没有异常处理（exception）功能的。scalaz提供了concurrent.Task类型填补了Future的这部分缺陷。我们会在下篇讨论Task。
我们用上面scala Future的例子来示范scalaz Future的函数组合能力：

  def dbl(i: Int): Future[Int] = Future { i + i } //> dbl: (i: Int)scalaz.concurrent.Future[Int]
  def sqr(i: Int): Future[Int] = Future { i * i } //> sqr: (i: Int)scalaz.concurrent.Future[Int]
  def sum(a: Int, b: Int): Future[Int] = Future { a + b }
                                  //> sum: (a: Int, b: Int)scalaz.concurrent.Future[Int]
  val fsum = for {
    a <- dbl(3)
    b <- sqr(a)
    c <- sum(a,b)
  } yield c                       //> fsum  : scalaz.concurrent.Future[Int] = BindAsync(<function1>,<function1>)

  fsum.unsafePerformAsync {
    case a => println(s"result c is:$a")
  }
  Thread.sleep(1000)              //> result c is:42