关闭

线程

442人阅读 评论(0) 收藏 举报
  Doug是一个无私的人,他深知分享知识和分享苹果是不一样的,苹果会越分越少,而自己的知识并不会因为给了别人就减少了,知识的分享更能激荡出不一样的火花。这位并发编程的大师级人物的下一步,将会带给Java怎样的冲击,不禁令人屏息以待。 
     JDK5中的一个亮点就是将Doug Lea的并发库引入到Java标准库中。Doug Lea确实是一个牛人,能教书,能出书,能编码,不过这在国外还是比较普遍的,而国内的教授们就相差太远了。 

一般的服务器都需要线程池,比如Web、FTP等服务器,不过它们一般都自己实现了线程池,比如以前介绍过的Tomcat、Resin和Jetty等,现在有了JDK5,我们就没有必要重复造车轮了,直接使用就可以,何况使用也很方便,性能也非常高。 

例子一: 

package concurrent;   
import java.util.concurrent.ExecutorService;   
import java.util.concurrent.Executors;   
public class TestThreadPool {   
public static void main(String args[]) throws InterruptedException {   
// only two threads   
ExecutorService exec = Executors.newFixedThreadPool(2);   
for(int index = 0; index < 100; index++) {   
Runnable run = new Runnable() {   
public void run() {   
long time = (long) (Math.random() * 1000);   
System.out.println(“Sleeping ” + time + “ms”);   
try {   
Thread.sleep(time);   
} catch (InterruptedException e) {   
}   
}   
};   
exec.execute(run);   
}   
// must shutdown   
exec.shutdown();   
}   
}  

package concurrent; 
import java.util.concurrent.ExecutorService; 
import java.util.concurrent.Executors; 
public class TestThreadPool { 
public static void main(String args[]) throws InterruptedException { 
// only two threads 
ExecutorService exec = Executors.newFixedThreadPool(2); 
for(int index = 0; index < 100; index++) { 
Runnable run = new Runnable() { 
public void run() { 
long time = (long) (Math.random() * 1000); 
System.out.println(“Sleeping ” + time + “ms”); 
try { 
Thread.sleep(time); 
} catch (InterruptedException e) { 


}; 
exec.execute(run); 

// must shutdown 
exec.shutdown(); 



上面是一个简单的例子,使用了2个大小的线程池来处理100个线程。但有一个问题:在for循环的过程中,会等待线程池有空闲的线程,所以主线程会阻塞的。为了解决这个问题,一般启动一个线程来做for循环,就是为了避免由于线程池满了造成主线程阻塞。不过在这里我没有这样处理。[重要修正:经过测试,即使线程池大小小于实际线程数大小,线程池也不会阻塞的,这与Tomcat的线程池不同,它将Runnable实例放到一个“无限”的BlockingQueue中,所以就不用一个线程启动for循环,Doug Lea果然厉害] 

另外它使用了Executors的静态函数生成一个固定的线程池,顾名思义,线程池的线程是不会释放的,即使它是Idle。这就会产生性能问题,比如如果线程池的大小为200,当全部使用完毕后,所有的线程会继续留在池中,相应的内存和线程切换(while(true)+sleep循环)都会增加。如果要避免这个问题,就必须直接使用ThreadPoolExecutor()来构造。可以像Tomcat的线程池一样设置“最大线程数”、“最小线程数”和“空闲线程keepAlive的时间”。通过这些可以基本上替换Tomcat的线程池实现方案。 

需要注意的是线程池必须使用shutdown来显式关闭,否则主线程就无法退出。shutdown也不会阻塞主线程。 


例子二: 
许多长时间运行的应用有时候需要定时运行任务完成一些诸如统计、优化等工作,比如在电信行业中处理用户话单时,需要每隔1分钟处理话单;网站每天凌晨统计用户访问量、用户数;大型超时凌晨3点统计当天销售额、以及最热卖的商品;每周日进行数据库备份;公司每个月的10号计算工资并进行转帐等,这些都是定时任务。通过 java的并发库concurrent可以轻松的完成这些任务,而且非常的简单。 


Java代码  

Java代码 
package concurrent;   
import static java.util.concurrent.TimeUnit.SECONDS;   
import java.util.Date;   
import java.util.concurrent.Executors;   
import java.util.concurrent.ScheduledExecutorService;   
import java.util.concurrent.ScheduledFuture;   
public class TestScheduledThread {   
public static void main(String[] args) {   
final ScheduledExecutorService scheduler = Executors   
.newScheduledThreadPool(2);   
final Runnable beeper = new Runnable() {   
int count = 0;   
public void run() {   
System.out.println(new Date() + ” beep ” + (++count));   
}   
};   
// 1秒钟后运行,并每隔2秒运行一次   
final ScheduledFuture beeperHandle = scheduler.scheduleAtFixedRate(   
beeper, 1, 2, SECONDS);   
// 2秒钟后运行,并每次在上次任务运行完后等待5秒后重新运行   
final ScheduledFuture beeperHandle2 = scheduler   
.scheduleWithFixedDelay(beeper, 2, 5, SECONDS);   
// 30秒后结束关闭任务,并且关闭Scheduler   
scheduler.schedule(new Runnable() {   
public void run() {   
beeperHandle.cancel(true);   
beeperHandle2.cancel(true);   
scheduler.shutdown();   
}   
}, 30, SECONDS);   
}   
}  

package concurrent; 
import static java.util.concurrent.TimeUnit.SECONDS; 
import java.util.Date; 
import java.util.concurrent.Executors; 
import java.util.concurrent.ScheduledExecutorService; 
import java.util.concurrent.ScheduledFuture; 
public class TestScheduledThread { 
public static void main(String[] args) { 
final ScheduledExecutorService scheduler = Executors 
.newScheduledThreadPool(2); 
final Runnable beeper = new Runnable() { 
int count = 0; 
public void run() { 
System.out.println(new Date() + ” beep ” + (++count)); 

}; 
// 1秒钟后运行,并每隔2秒运行一次 
final ScheduledFuture beeperHandle = scheduler.scheduleAtFixedRate( 
beeper, 1, 2, SECONDS); 
// 2秒钟后运行,并每次在上次任务运行完后等待5秒后重新运行 
final ScheduledFuture beeperHandle2 = scheduler 
.scheduleWithFixedDelay(beeper, 2, 5, SECONDS); 
// 30秒后结束关闭任务,并且关闭Scheduler 
scheduler.schedule(new Runnable() { 
public void run() { 
beeperHandle.cancel(true); 
beeperHandle2.cancel(true); 
scheduler.shutdown(); 

}, 30, SECONDS); 



为了退出进程,上面的代码中加入了关闭Scheduler的操作。而对于24小时运行的应用而言,是没有必要关闭Scheduler的。 


例子三: 
在实际应用中,有时候需要多个线程同时工作以完成同一件事情,而且在完成过程中,往往会等待其他线程都完成某一阶段后再执行,等所有线程都到达某一个阶段后再统一执行。 

比如有几个旅行团需要途经深圳、广州、韶关、长沙最后到达武汉。旅行团中有自驾游的,有徒步的,有乘坐旅游大巴的;这些旅行团同时出发,并且每到一个目的地,都要等待其他旅行团到达此地后再同时出发,直到都到达终点站武汉。 

这时候CyclicBarrier就可以派上用场。CyclicBarrier最重要的属性就是参与者个数,另外最要方法是await()。当所有线程都调用了await()后,就表示这些线程都可以继续执行,否则就会等待。 

Java代码  Java代码 
package concurrent;   
import java.text.SimpleDateFormat;   
import java.util.Date;   
import java.util.concurrent.BrokenBarrierException;   
import java.util.concurrent.CyclicBarrier;   
import java.util.concurrent.ExecutorService;   
import java.util.concurrent.Executors;   
public class TestCyclicBarrier {   
// 徒步需要的时间: Shenzhen, Guangzhou, Shaoguan, Changsha, Wuhan   
private static int[] timeWalk = { 5, 8, 15, 15, 10 };   
// 自驾游   
private static int[] timeSelf = { 1, 3, 4, 4, 5 };   
// 旅游大巴   
private static int[] timeBus = { 2, 4, 6, 6, 7 };   
  
static String now() {   
SimpleDateFormat sdf = new SimpleDateFormat(“HH:mm:ss”);   
return sdf.format(new Date()) + “: “;   
}   
  
static class Tour implements Runnable {   
private int[] times;   
private CyclicBarrier barrier;   
private String tourName;   
public Tour(CyclicBarrier barrier, String tourName, int[] times) {   
this.times = times;   
this.tourName = tourName;   
this.barrier = barrier;   
}   
public void run() {   
try {   
Thread.sleep(times[0] * 1000);   
System.out.println(now() + tourName + ” Reached Shenzhen”);   
barrier.await();   
Thread.sleep(times[1] * 1000);   
System.out.println(now() + tourName + ” Reached Guangzhou”);   
barrier.await();   
Thread.sleep(times[2] * 1000);   
System.out.println(now() + tourName + ” Reached Shaoguan”);   
barrier.await();   
Thread.sleep(times[3] * 1000);   
System.out.println(now() + tourName + ” Reached Changsha”);   
barrier.await();   
Thread.sleep(times[4] * 1000);   
System.out.println(now() + tourName + ” Reached Wuhan”);   
barrier.await();   
} catch (InterruptedException e) {   
} catch (BrokenBarrierException e) {   
}   
}   
}   
  
public static void main(String[] args) {   
// 三个旅行团   
CyclicBarrier barrier = new CyclicBarrier(3);   
ExecutorService exec = Executors.newFixedThreadPool(3);   
exec.submit(new Tour(barrier, “WalkTour”, timeWalk));   
exec.submit(new Tour(barrier, “SelfTour”, timeSelf));   
exec.submit(new Tour(barrier, “BusTour”, timeBus));   
exec.shutdown();   
}   
}  

package concurrent; 
import java.text.SimpleDateFormat; 
import java.util.Date; 
import java.util.concurrent.BrokenBarrierException; 
import java.util.concurrent.CyclicBarrier; 
import java.util.concurrent.ExecutorService; 
import java.util.concurrent.Executors; 
public class TestCyclicBarrier { 
// 徒步需要的时间: Shenzhen, Guangzhou, Shaoguan, Changsha, Wuhan 
private static int[] timeWalk = { 5, 8, 15, 15, 10 }; 
// 自驾游 
private static int[] timeSelf = { 1, 3, 4, 4, 5 }; 
// 旅游大巴 
private static int[] timeBus = { 2, 4, 6, 6, 7 }; 

static String now() { 
SimpleDateFormat sdf = new SimpleDateFormat(“HH:mm:ss”); 
return sdf.format(new Date()) + “: “; 


static class Tour implements Runnable { 
private int[] times; 
private CyclicBarrier barrier; 
private String tourName; 
public Tour(CyclicBarrier barrier, String tourName, int[] times) { 
this.times = times; 
this.tourName = tourName; 
this.barrier = barrier; 

public void run() { 
try { 
Thread.sleep(times[0] * 1000); 
System.out.println(now() + tourName + ” Reached Shenzhen”); 
barrier.await(); 
Thread.sleep(times[1] * 1000); 
System.out.println(now() + tourName + ” Reached Guangzhou”); 
barrier.await(); 
Thread.sleep(times[2] * 1000); 
System.out.println(now() + tourName + ” Reached Shaoguan”); 
barrier.await(); 
Thread.sleep(times[3] * 1000); 
System.out.println(now() + tourName + ” Reached Changsha”); 
barrier.await(); 
Thread.sleep(times[4] * 1000); 
System.out.println(now() + tourName + ” Reached Wuhan”); 
barrier.await(); 
} catch (InterruptedException e) { 
} catch (BrokenBarrierException e) { 




public static void main(String[] args) { 
// 三个旅行团 
CyclicBarrier barrier = new CyclicBarrier(3); 
ExecutorService exec = Executors.newFixedThreadPool(3); 
exec.submit(new Tour(barrier, “WalkTour”, timeWalk)); 
exec.submit(new Tour(barrier, “SelfTour”, timeSelf)); 
exec.submit(new Tour(barrier, “BusTour”, timeBus)); 
exec.shutdown(); 



运行结果: 
00:02:25: SelfTour Reached Shenzhen 
00:02:25: BusTour Reached Shenzhen 
00:02:27: WalkTour Reached Shenzhen 
00:02:30: SelfTour Reached Guangzhou 
00:02:31: BusTour Reached Guangzhou 
00:02:35: WalkTour Reached Guangzhou 
00:02:39: SelfTour Reached Shaoguan 
00:02:41: BusTour Reached Shaoguan 



例子四: 
并发库中的BlockingQueue是一个比较好玩的类,顾名思义,就是阻塞队列。该类主要提供了两个方法put()和take(),前者将一个对象放到队列中,如果队列已经满了,就等待直到有空闲节点;后者从head取一个对象,如果没有对象,就等待直到有可取的对象。 

下面的例子比较简单,一个读线程,用于将要处理的文件对象添加到阻塞队列中,另外四个写线程用于取出文件对象,为了模拟写操作耗时长的特点,特让线程睡眠一段随机长度的时间。另外,该Demo也使用到了线程池和原子整型(AtomicInteger),AtomicInteger可以在并发情况下达到原子化更新,避免使用了synchronized,而且性能非常高。由于阻塞队列的put和take操作会阻塞,为了使线程退出,特在队列中添加了一个“标识”,算法中也叫“哨兵”,当发现这个哨兵后,写线程就退出。 

当然线程池也要显式退出了。 

Java代码  Java代码 
package concurrent;   
import java.io.File;   
import java.io.FileFilter;   
import java.util.concurrent.BlockingQueue;   
import java.util.concurrent.ExecutorService;   
import java.util.concurrent.Executors;   
import java.util.concurrent.LinkedBlockingQueue;   
import java.util.concurrent.atomic.AtomicInteger;   
  
public class TestBlockingQueue {   
static long randomTime() {   
return (long) (Math.random() * 1000);   
}   
  
public static void main(String[] args) {   
// 能容纳100个文件   
final BlockingQueue queue = new LinkedBlockingQueue(100);   
// 线程池   
final ExecutorService exec = Executors.newFixedThreadPool(5);   
final File root = new File(“F:\\JavaLib”);   
// 完成标志   
final File exitFile = new File(“”);   
// 读个数   
final AtomicInteger rc = new AtomicInteger();   
// 写个数   
final AtomicInteger wc = new AtomicInteger();   
// 读线程   
Runnable read = new Runnable() {   
public void run() {   
scanFile(root);   
scanFile(exitFile);   
}   
  
public void scanFile(File file) {   
if (file.isDirectory()) {   
File[] files = file.listFiles(new FileFilter() {   
public boolean accept(File pathname) {   
return pathname.isDirectory()   
|| pathname.getPath().endsWith(“.java”);   
}   
});   
for (File one : files)   
scanFile(one);   
} else {   
try {   
int index = rc.incrementAndGet();   
System.out.println(“Read0: ” + index + ” “   
+ file.getPath());   
queue.put(file);   
} catch (InterruptedException e) {   
}   
}   
}   
};   
exec.submit(read);   
// 四个写线程   
for (int index = 0; index < 4; index++) {   
// write thread   
final int NO = index;   
Runnable write = new Runnable() {   
String threadName = “Write” + NO;   
public void run() {   
while (true) {   
try {   
Thread.sleep(randomTime());   
int index = wc.incrementAndGet();   
File file = queue.take();   
// 队列已经无对象   
if (file == exitFile) {   
// 再次添加”标志”,以让其他线程正常退出   
queue.put(exitFile);   
break;   
}   
System.out.println(threadName + “: ” + index + ” “   
+ file.getPath());   
} catch (InterruptedException e) {   
}   
}   
}   
};   
exec.submit(write);   
}   
exec.shutdown();   
}   
}  

package concurrent; 
import java.io.File; 
import java.io.FileFilter; 
import java.util.concurrent.BlockingQueue; 
import java.util.concurrent.ExecutorService; 
import java.util.concurrent.Executors; 
import java.util.concurrent.LinkedBlockingQueue; 
import java.util.concurrent.atomic.AtomicInteger; 

public class TestBlockingQueue { 
static long randomTime() { 
return (long) (Math.random() * 1000); 


public static void main(String[] args) { 
// 能容纳100个文件 
final BlockingQueue queue = new LinkedBlockingQueue(100); 
// 线程池 
final ExecutorService exec = Executors.newFixedThreadPool(5); 
final File root = new File(“F:\\JavaLib”); 
// 完成标志 
final File exitFile = new File(“”); 
// 读个数 
final AtomicInteger rc = new AtomicInteger(); 
// 写个数 
final AtomicInteger wc = new AtomicInteger(); 
// 读线程 
Runnable read = new Runnable() { 
public void run() { 
scanFile(root); 
scanFile(exitFile); 


public void scanFile(File file) { 
if (file.isDirectory()) { 
File[] files = file.listFiles(new FileFilter() { 
public boolean accept(File pathname) { 
return pathname.isDirectory() 
|| pathname.getPath().endsWith(“.java”); 

}); 
for (File one : files) 
scanFile(one); 
} else { 
try { 
int index = rc.incrementAndGet(); 
System.out.println(“Read0: ” + index + ” “ 
+ file.getPath()); 
queue.put(file); 
} catch (InterruptedException e) { 



}; 
exec.submit(read); 
// 四个写线程 
for (int index = 0; index < 4; index++) { 
// write thread 
final int NO = index; 
Runnable write = new Runnable() { 
String threadName = “Write” + NO; 
public void run() { 
while (true) { 
try { 
Thread.sleep(randomTime()); 
int index = wc.incrementAndGet(); 
File file = queue.take(); 
// 队列已经无对象 
if (file == exitFile) { 
// 再次添加”标志”,以让其他线程正常退出 
queue.put(exitFile); 
break; 

System.out.println(threadName + “: ” + index + ” “ 
+ file.getPath()); 
} catch (InterruptedException e) { 



}; 
exec.submit(write); 

exec.shutdown(); 






例子五: 
从名字可以看出,CountDownLatch是一个倒数计数的锁,当倒数到0时触发事件,也就是开锁,其他人就可以进入了。在一些应用场合中,需要等待某个条件达到要求后才能做后面的事情;同时当线程都完成后也会触发事件,以便进行后面的操作。 


CountDownLatch最重要的方法是countDown()和await(),前者主要是倒数一次,后者是等待倒数到0,如果没有到达0,就只有阻塞等待了。 

一个CountDouwnLatch实例是不能重复使用的,也就是说它是一次性的,锁一经被打开就不能再关闭使用了,如果想重复使用,请考虑使用CyclicBarrier。 

下面的例子简单的说明了CountDownLatch的使用方法,模拟了100米赛跑,10名选手已经准备就绪,只等裁判一声令下。当所有人都到达终点时,比赛结束。 

同样,线程池需要显式shutdown。 

Java代码  Java代码 
package concurrent;   
  
import java.util.concurrent.CountDownLatch;   
import java.util.concurrent.ExecutorService;   
import java.util.concurrent.Executors;   
  
public class TestCountDownLatch {   
public static void main(String[] args) throws InterruptedException {   
// 开始的倒数锁   
final CountDownLatch begin = new CountDownLatch(1);   
// 结束的倒数锁   
final CountDownLatch end = new CountDownLatch(10);   
// 十名选手   
final ExecutorService exec = Executors.newFixedThreadPool(10);   
for(int index = 0; index < 10; index++) {   
final int NO = index + 1;   
Runnable run = new Runnable(){   
public void run() {   
try {   
begin.await();   
Thread.sleep((long) (Math.random() * 10000));   
System.out.println(“No.” + NO + ” arrived”);   
} catch (InterruptedException e) {   
} finally {   
end.countDown();   
}   
}   
};   
exec.submit(run);   
}   
System.out.println(“Game Start”);   
begin.countDown();   
end.await();   
System.out.println(“Game Over”);   
exec.shutdown();   
}   
}  

package concurrent; 

import java.util.concurrent.CountDownLatch; 
import java.util.concurrent.ExecutorService; 
import java.util.concurrent.Executors; 

public class TestCountDownLatch { 
public static void main(String[] args) throws InterruptedException { 
// 开始的倒数锁 
final CountDownLatch begin = new CountDownLatch(1); 
// 结束的倒数锁 
final CountDownLatch end = new CountDownLatch(10); 
// 十名选手 
final ExecutorService exec = Executors.newFixedThreadPool(10); 
for(int index = 0; index < 10; index++) { 
final int NO = index + 1; 
Runnable run = new Runnable(){ 
public void run() { 
try { 
begin.await(); 
Thread.sleep((long) (Math.random() * 10000)); 
System.out.println(“No.” + NO + ” arrived”); 
} catch (InterruptedException e) { 
} finally { 
end.countDown(); 


}; 
exec.submit(run); 

System.out.println(“Game Start”); 
begin.countDown(); 
end.await(); 
System.out.println(“Game Over”); 
exec.shutdown(); 



运行结果: 
Game Start 
No.4 arrived 
No.1 arrived 
No.7 arrived 
No.9 arrived 
No.3 arrived 
No.2 arrived 
No.8 arrived 
No.10 arrived 
No.6 arrived 
No.5 arrived 
Game Over 



例子六: 
有时候在实际应用中,某些操作很耗时,但又不是不可或缺的步骤。比如用网页浏览器浏览新闻时,最重要的是要显示文字内容,至于与新闻相匹配的图片就没有那么重要的,所以此时首先保证文字信息先显示,而图片信息会后显示,但又不能不显示,由于下载图片是一个耗时的操作,所以必须一开始就得下载。 

Java的并发库的Future类就可以满足这个要求。Future的重要方法包括get()和cancel(),get()获取数据对象,如果数据没有加载,就会阻塞直到取到数据,而 cancel()是取消数据加载。另外一个get(timeout)操作,表示如果在timeout时间内没有取到就失败返回,而不再阻塞。 

下面的Demo简单的说明了Future的使用方法:一个非常耗时的操作必须一开始启动,但又不能一直等待;其他重要的事情又必须做,等完成后,就可以做不重要的事情。 
Java代码  Java代码 
package concurrent;   
  
import java.util.concurrent.Callable;   
import java.util.concurrent.ExecutionException;   
import java.util.concurrent.ExecutorService;   
import java.util.concurrent.Executors;   
import java.util.concurrent.Future;   
  
public class TestFutureTask {   
public static void main(String[] args)throws InterruptedException,   
ExecutionException {   
final ExecutorService exec = Executors.newFixedThreadPool(5);   
Callable call = new Callable() {   
public String call() throws Exception {   
Thread.sleep(1000 * 5);   
return “Other less important but longtime things.”;   
}   
};   
Future task = exec.submit(call);   
// 重要的事情   
Thread.sleep(1000 * 3);   
System.out.println(“Let’s do important things.”);   
// 其他不重要的事情   
String obj = task.get();   
System.out.println(obj);   
// 关闭线程池   
exec.shutdown();   
}   
}  

package concurrent; 

import java.util.concurrent.Callable; 
import java.util.concurrent.ExecutionException; 
import java.util.concurrent.ExecutorService; 
import java.util.concurrent.Executors; 
import java.util.concurrent.Future; 

public class TestFutureTask { 
public static void main(String[] args)throws InterruptedException, 
ExecutionException { 
final ExecutorService exec = Executors.newFixedThreadPool(5); 
Callable call = new Callable() { 
public String call() throws Exception { 
Thread.sleep(1000 * 5); 
return “Other less important but longtime things.”; 

}; 
Future task = exec.submit(call); 
// 重要的事情 
Thread.sleep(1000 * 3); 
System.out.println(“Let’s do important things.”); 
// 其他不重要的事情 
String obj = task.get(); 
System.out.println(obj); 
// 关闭线程池 
exec.shutdown(); 




运行结果: 
Let’s do important things. 
Other less important but longtime things. 



例子七: 
考虑以下场景:浏览网页时,浏览器了5个线程下载网页中的图片文件,由于图片大小、网站访问速度等诸多因素的影响,完成图片下载的时间就会有很大的不同。如果先下载完成的图片就会被先显示到界面上,反之,后下载的图片就后显示。 

Java的并发库的CompletionService可以满足这种场景要求。该接口有两个重要方法:submit()和take()。submit用于提交一个runnable或者callable,一般会提交给一个线程池处理;而take就是取出已经执行完毕runnable或者callable实例的Future对象,如果没有满足要求的,就等待了。 CompletionService还有一个对应的方法poll,该方法与take类似,只是不会等待,如果没有满足要求,就返回null对象。 

Java代码  

Java代码 
package concurrent;   
  
import java.util.concurrent.Callable;   
import java.util.concurrent.CompletionService;   
import java.util.concurrent.ExecutionException;   
import java.util.concurrent.ExecutorCompletionService;   
import java.util.concurrent.ExecutorService;   
import java.util.concurrent.Executors;   
import java.util.concurrent.Future;   
  
public class TestCompletionService {   
public static void main(String[] args) throws InterruptedException,   
ExecutionException {   
ExecutorService exec = Executors.newFixedThreadPool(10);   
CompletionService serv =   
new ExecutorCompletionService(exec);   
  
for (int index = 0; index < 5; index++) {   
final int NO = index;   
Callable downImg = new Callable() {   
public String call() throws Exception {   
Thread.sleep((long) (Math.random() * 10000));   
return “Downloaded Image ” + NO;   
}   
};   
serv.submit(downImg);   
}   
  
Thread.sleep(1000 * 2);   
System.out.println(“Show web content”);   
for (int index = 0; index < 5; index++) {   
Future task = serv.take();   
String img = task.get();   
System.out.println(img);   
}   
System.out.println(“End”);   
// 关闭线程池   
exec.shutdown();   
}   
}  

package concurrent; 

import java.util.concurrent.Callable; 
import java.util.concurrent.CompletionService; 
import java.util.concurrent.ExecutionException; 
import java.util.concurrent.ExecutorCompletionService; 
import java.util.concurrent.ExecutorService; 
import java.util.concurrent.Executors; 
import java.util.concurrent.Future; 

public class TestCompletionService { 
public static void main(String[] args) throws InterruptedException, 
ExecutionException { 
ExecutorService exec = Executors.newFixedThreadPool(10); 
CompletionService serv = 
new ExecutorCompletionService(exec); 

for (int index = 0; index < 5; index++) { 
final int NO = index; 
Callable downImg = new Callable() { 
public String call() throws Exception { 
Thread.sleep((long) (Math.random() * 10000)); 
return “Downloaded Image ” + NO; 

}; 
serv.submit(downImg); 


Thread.sleep(1000 * 2); 
System.out.println(“Show web content”); 
for (int index = 0; index < 5; index++) { 
Future task = serv.take(); 
String img = task.get(); 
System.out.println(img); 

System.out.println(“End”); 
// 关闭线程池 
exec.shutdown(); 



运行结果: 
Show web content 
Downloaded Image 1 
Downloaded Image 2 
Downloaded Image 4 
Downloaded Image 0 
Downloaded Image 3 
End 



例子八: 
操作系统的信号量是个很重要的概念,在进程控制方面都有应用。Java并发库的Semaphore可以很轻松完成信号量控制,Semaphore可以控制某个资源可被同时访问的个数,acquire()获取一个许可,如果没有就等待,而release()释放一个许可。比如在Windows下可以设置共享文件的最大客户端访问个数。 

Semaphore维护了当前访问的个数,提供同步机制,控制同时访问的个数。在数据结构中链表可以保存“无限”的节点,用Semaphore可以实现有限大小的链表。另外重入锁ReentrantLock也可以实现该功能,但实现上要负责些,代码也要复杂些。 

下面的Demo中申明了一个只有5个许可的Semaphore,而有20个线程要访问这个资源,通过acquire()和release()获取和释放访问许可。 
Java代码  Java代码 
package concurrent;   
  
import java.util.concurrent.ExecutorService;   
import java.util.concurrent.Executors;   
import java.util.concurrent.Semaphore;   
  
public class TestSemaphore {   
public static void main(String[] args) {   
// 线程池   
ExecutorService exec = Executors.newCachedThreadPool();   
// 只能5个线程同时访问   
final Semaphore semp = new Semaphore(5);   
// 模拟20个客户端访问   
for (int index = 0; index < 20; index++) {   
final int NO = index;   
Runnable run = new Runnable() {   
public void run() {   
try {   
// 获取许可   
semp.acquire();   
System.out.println(“Accessing: ” + NO);   
Thread.sleep((long) (Math.random() * 10000));   
// 访问完后,释放   
semp.release();   
} catch (InterruptedException e) {   
}   
}   
};   
exec.execute(run);   
}   
// 退出线程池   
exec.shutdown();   
}   
}  

package concurrent; 

import java.util.concurrent.ExecutorService; 
import java.util.concurrent.Executors; 
import java.util.concurrent.Semaphore; 

public class TestSemaphore { 
public static void main(String[] args) { 
// 线程池 
ExecutorService exec = Executors.newCachedThreadPool(); 
// 只能5个线程同时访问 
final Semaphore semp = new Semaphore(5); 
// 模拟20个客户端访问 
for (int index = 0; index < 20; index++) { 
final int NO = index; 
Runnable run = new Runnable() { 
public void run() { 
try { 
// 获取许可 
semp.acquire(); 
System.out.println(“Accessing: ” + NO); 
Thread.sleep((long) (Math.random() * 10000)); 
// 访问完后,释放 
semp.release(); 
} catch (InterruptedException e) { 


}; 
exec.execute(run); 

// 退出线程池 
exec.shutdown(); 



运行结果: 
Accessing: 0 
Accessing: 1 
Accessing: 2 
Accessing: 3 
Accessing: 4 
Accessing: 5 
Accessing: 6 
Accessing: 7 
Accessing: 8 
Accessing: 9 
Accessing: 10 
Accessing: 11 
Accessing: 12 
Accessing: 13 
Accessing: 14 
Accessing: 15 
Accessing: 16 
Accessing: 17 
Accessing: 18 
Accessing: 19 

http://www.ibm.com/developerworks/java/library/j-threads3.html 


原文链接:http://fsh430623.iteye.com/blog/855877

0
0

查看评论
* 以上用户言论只代表其个人观点,不代表CSDN网站的观点或立场
    个人资料
    • 访问:207942次
    • 积分:2875
    • 等级:
    • 排名:第13111名
    • 原创:35篇
    • 转载:315篇
    • 译文:0篇
    • 评论:18条
    文章分类
    最新评论