大数据学习——actor编程

1 概念

Scala中的Actor能够实现并行编程的强大功能,它是基于事件模型的并发机制,Scala是运用消息(message)的发送、接收来实现多线程的。使用Scala能够更容易地实现多线程应用的开发。

 

2 传统java并发编程与scala actor编程的区别

 

 

对于Java,我们都知道它的多线程实现需要对共享资源(变量、对象等)使用synchronized 关键字进行代码块同步、对象锁互斥等等。而且,常常一大块的try…catch语句块中加上wait方法、notify方法、notifyAll方法是让人很头疼的。原因就在于Java中多数使用的是可变状态的对象资源,对这些资源进行共享来实现多线程编程的话,控制好资源竞争与防止对象状态被意外修改是非常重要的,而对象状态的不变性也是较难以保证的。 而在Scala中,我们可以通过复制不可变状态的资源(即对象,Scala中一切都是对象,连函数、方法也是)的一个副本,再基于Actor的消息发送、接收机制进行并行编程

 

3 actor方法执行顺序

1.首先调用start()方法启动Actor

2.调用start()方法后其act()方法会被执行

3.向Actor发送消息

 

发送消息的方式

!

发送异步消息,没有返回值。

!?

发送同步消息,等待返回值。

!!

发送异步消息,返回值是 Future[Any]。

 

例子

添加依赖

<!--scala actor-->
<dependency>
    <groupId>org.scala-lang</groupId>
    <artifactId>scala-actors</artifactId>
    <version>2.10.5</version>
</dependency>

package main.scala.com

import scala.actors.Actor

/**
  * Created by Administrator on 2019/6/4.
  */
object MyActor1 extends Actor {

  //重写act方法

  def act(): Unit = {
    for (i <- 1 to 10) {
      println("actor-1" + i)
      Thread.sleep(2000)
    }
  }
}

object MyActor2 extends Actor {
  //重写act方法
  def act() {
    for (i <- 1 to 10) {
      println("actor-2 " + i)
      Thread.sleep(2000)
    }
  }
}
object ActorTest extends App{
  //启动Actor
  MyActor1.start()
  MyActor2.start()
}

 

运行结果

说明:上面分别调用了两个单例对象的start()方法,他们的act()方法会被执行,相同与在java中开启了两个线程,线程的run()方法会被执行

注意:这两个Actor是并行执行的,act()方法中的for循环执行完成后actor程序就退出了

 

 

可能遇见的问题

1 Exception in thread "main" java.lang.NoSuchMethodError: scala.actors.AbstractActor.$init$(Lscala/actors/AbstractActor;)V

解决办法

使用scala2.12.x的版本运行Actor,会报这种错误。

报错原因:scala版本不匹配,

解决方法:创建新工程,选择scala2.10.x的版本

 

 

解决方案:项目->open module setting->Modules->Dependencies  加上scala sdk的library

 

 

 

package main.scala.com

import scala.actors.Actor

/**
  * Created by Administrator on 2019/6/4.
  */
class MyActor extends Actor {

  override def act(): Unit = {
    while (true) {
      receive {
        case "start" => {
          println("starting ...")
          Thread.sleep(5000)
          println("started")
        }
        case "stop" => {
          println("stopping ...")
          Thread.sleep(5000)
          println("stopped ...")
        }
      }
    }
  }
}

object MyActor {
  def main(args: Array[String]) {
    val actor = new MyActor
    actor.start()
    actor ! "start"
    actor ! "stop"
    println("消息发送完成!")
  }
}

说明:在act()方法中加入了while (true) 循环,就可以不停的接收消息

注意:发送start消息和stop的消息是异步的,但是Actor接收到消息执行的过程是同步的按顺序执行

 

 

(react方式会复用线程,比receive更高效)

package main.scala.com

import scala.actors.Actor

/**
  * Created by Administrator on 2019/6/4.
  */
class YourActor extends Actor {

  override def act(): Unit = {
    loop {
      react {
        case "start" => {
          println("starting ...")
          Thread.sleep(5000)
          println("started")
        }
        case "stop" => {
          println("stopping ...")
          Thread.sleep(8000)
          println("stopped ...")
        }
      }
    }
  }
}


object YourActor {
  def main(args: Array[String]) {
    val actor = new YourActor
    actor.start()
    actor ! "start"
    actor ! "stop"
    println("消息发送完成!")
  }
}

说明: react 如果要反复执行消息处理,react外层要用loop,不能用while

 

 

4

package main.scala.com

import scala.actors.Actor

/**
  * Created by Administrator on 2019/6/4.
  */
class AppleActor extends Actor {

  def act(): Unit = {
    while (true) {
      receive {
        case "start" => println("starting ...")
        case SyncMsg(id, msg) => {
          println(id + ",sync " + msg)
          Thread.sleep(5000)
          sender ! ReplyMsg(3, "finished")
        }
        case AsyncMsg(id, msg) => {
          println(id + ",async " + msg)
          Thread.sleep(5000)
        }
      }
    }
  }
}

object AppleActor {
  def main(args: Array[String]) {
    val a = new AppleActor
    a.start()
    //异步消息
    a ! AsyncMsg(1, "hello actor")
    println("异步消息发送完成")
    //同步消息
    //val content = a.!?(1000, SyncMsg(2, "hello actor"))
    //println(content)
    val reply = a !! SyncMsg(2, "hello actor")
    println(reply.isSet)
    //println("123")
    val c = reply.apply()
    println(reply.isSet)
    println(c)
  }
}

case class SyncMsg(id: Int, msg: String)

case class AsyncMsg(id: Int, msg: String)

case class ReplyMsg(id: Int, msg: String)

 

 

5  用actor并发编程写一个单机版的WorldCount,将多个文件作为输入,计算完成后将多个任务汇总,得到最终的结果

package main.scala.com

import java.io.File

import scala.actors.{Actor, Future}
import scala.collection.mutable
import scala.io.Source


/**
  * Created by Administrator on 2019/6/4.
  */
class Task extends Actor {

  override def act(): Unit = {
    loop {
      react {
        case SubmitTask(fileName) => {
          val contents = Source.fromFile(new File(fileName)).mkString
          val arr = contents.split("\r\n")
          val result = arr.flatMap(_.split(" ")).map((_, 1)).groupBy(_._1).mapValues(_.length)
          //val result = arr.flatMap(_.split(" ")).map((_, 1)).groupBy(_._1).mapValues(_.foldLeft(0)(_ + _._2))
          sender ! ResultTask(result)
        }
        case StopTask => {
          exit()
        }
      }
    }
  }
}

object WorkCount {
  def main(args: Array[String]) {
    val files = Array("c://words.txt", "c://words.log")

    val replaySet = new mutable.HashSet[Future[Any]]
    val resultList = new mutable.ListBuffer[ResultTask]

    for (f <- files) {
      val t = new Task
      val replay = t.start() !! SubmitTask(f)
      replaySet += replay
    }

    while (replaySet.size > 0) {
      val toCumpute = replaySet.filter(_.isSet)
      for (r <- toCumpute) {
        val result = r.apply()
        resultList += result.asInstanceOf[ResultTask]
        replaySet.remove(r)
      }
      Thread.sleep(100)
    }
    val finalResult = resultList.map(_.result).flatten.groupBy(_._1).mapValues(x => x.foldLeft(0)(_ + _._2))
    println(finalResult)
  }
}

case class SubmitTask(fileName: String)

case object StopTask

case class ResultTask(result: Map[String, Int])

 

posted on 2019-06-04 10:31  o_0的园子  阅读(319)  评论(0编辑  收藏  举报