考試指導(dǎo):java多線程設(shè)計(jì)模式詳解之三

字號(hào):

前面談了多線程應(yīng)用程序能極大地改善用戶相應(yīng)。例如對(duì)于一個(gè)Web應(yīng)用程序,每當(dāng)一個(gè)用戶請(qǐng)求服務(wù)器連接時(shí),服務(wù)器就可以啟動(dòng)一個(gè)新線程為用戶服務(wù)。
    然而,創(chuàng)建和銷毀線程本身就有一定的開銷,如果頻繁創(chuàng)建和銷毀線程,CPU和內(nèi)存開銷就不可忽略,垃圾收集器還必須負(fù)擔(dān)更多的工作。因此,線程池就是為了避免頻繁創(chuàng)建和銷毀線程。
    每當(dāng)服務(wù)器接受了一個(gè)新的請(qǐng)求后,服務(wù)器就從線程池中挑選一個(gè)等待的線程并執(zhí)行請(qǐng)求處理。處理完畢后,線程并不結(jié)束,而是轉(zhuǎn)為阻塞狀態(tài)再次被放入線程池中。這樣就避免了頻繁創(chuàng)建和銷毀線程。
    Worker Pattern實(shí)現(xiàn)了類似線程池的功能。首先定義Task接口:
    package com.crackj2ee.thread;
    public interface Task {
    void execute();
    }
    線程將負(fù)責(zé)執(zhí)行execute()方法。注意到任務(wù)是由子類通過實(shí)現(xiàn)execute()方法實(shí)現(xiàn)的,線程本身并不知道自己執(zhí)行的任務(wù)。它只負(fù)責(zé)運(yùn)行一個(gè)耗時(shí)的execute()方法。
    具體任務(wù)由子類實(shí)現(xiàn),我們定義了一個(gè)CalculateTask和一個(gè)TimerTask:
    // CalculateTask.java
    package com.crackj2ee.thread;
    public class CalculateTask implements Task {
    private static int count = 0;
    private int num = count;
    public CalculateTask() {
    count++;
    }
    public void execute() {
    System.out.println("[CalculateTask " + num + "] start...");
    try {
    Thread.sleep(3000);
    }
    catch(InterruptedException ie) {}
    System.out.println("[CalculateTask " + num + "] done.");
    }
    }
    // TimerTask.java
    package com.crackj2ee.thread;
    public class TimerTask implements Task {
    private static int count = 0;
    private int num = count;
    public TimerTask() {
    count++;
    }
    public void execute() {
    System.out.println("[TimerTask " + num + "] start...");
    try {
    Thread.sleep(2000);
    }
    catch(InterruptedException ie) {}
    System.out.println("[TimerTask " + num + "] done.");
    }
    }
    以上任務(wù)均簡單的sleep若干秒。
    TaskQueue實(shí)現(xiàn)了一個(gè)隊(duì)列,客戶端可以將請(qǐng)求放入隊(duì)列,服務(wù)器線程可以從隊(duì)列中取出任務(wù):
    package com.crackj2ee.thread;
    import java.util.*;
    public class TaskQueue {
    private List queue = new LinkedList();
    public synchronized Task getTask() {
    while(queue.size()==0) {
    try {
    this.wait();
    }
    catch(InterruptedException ie) {
    return null;
    }
    }
    return (Task)queue.remove(0);
    }
    public synchronized void putTask(Task task) {
    queue.add(task);
    this.notifyAll();
    }
    }
    終于到了真正的WorkerThread,這是真正執(zhí)行任務(wù)的服務(wù)器線程:
    package com.crackj2ee.thread;
    public class WorkerThread extends Thread {
    private static int count = 0;
    private boolean busy = false;
    private boolean stop = false;
    private TaskQueue queue;
    public WorkerThread(ThreadGroup group, TaskQueue queue) {
    super(group, "worker-" + count);
    count++;
    this.queue = queue;
    }
    public void shutdown() {
    stop = true;
    this.interrupt();
    try {
    this.join();
    }
    catch(InterruptedException ie) {}
    }
    public boolean isIdle() {
    return !busy;
    }
    public void run() {
    System.out.println(getName() + " start.");
    while(!stop) {
    Task task = queue.getTask();
    if(task!=null) {
    busy = true;
    task.execute();
    busy = false;
    }
    }
    System.out.println(getName() + " end.");
    }
    }
    前面已經(jīng)講過,queue.getTask()是一個(gè)阻塞方法,服務(wù)器線程可能在此wait()一段時(shí)間。此外,WorkerThread還有一個(gè)shutdown方法,用于安全結(jié)束線程。
    最后是ThreadPool,負(fù)責(zé)管理所有的服務(wù)器線程,還可以動(dòng)態(tài)增加和減少線程數(shù):
    package com.crackj2ee.thread;
    import java.util.*;
    public class ThreadPool extends ThreadGroup {
    private List threads = new LinkedList();
    private TaskQueue queue;
    public ThreadPool(TaskQueue queue) {
    super("Thread-Pool");
    this.queue = queue;
    }
    public synchronized void addWorkerThread() {
    Thread t = new WorkerThread(this, queue);
    threads.add(t);
    t.start();
    }
    public synchronized void removeWorkerThread() {
    if(threads.size()>0) {
    WorkerThread t = (WorkerThread)threads.remove(0);
    t.shutdown();
    }
    }
    public synchronized void currentStatus() {
    System.out.println("-----------------------------------------------");
    System.out.println("Thread count = " + threads.size());
    Iterator it = threads.iterator();