谈谈java的BlockingQueue

最近在维护一个java工程，在群里面也就聊起来java的优劣！无奈一些Java的终极粉丝，总是号称性能已经不必C++差，并且很多标准类库都是大师级的人写的，如何如何稳定等等。索性就认真研究一番，他们给我的一项说明就是，在线程之间投递消息，用java已经封装好的BlockingQueue，就足够用了。

既然足够用那就写代码测试喽，简简单单写一个小程序做了一番测试：

//默认包
import java.util.concurrent.*;
 
import base.MyRunnable;
 
public class Test
{
    public static void main(String[] args)
    {
        BlockingQueue<Integer> queue = new LinkedBlockingQueue<Integer>();
        java.lang.Runnable r = new MyRunnable(queue);
        Thread t = new Thread(r);
        t.start();
         
        while(true)
        {
            try
            {
                while(true)
                {
                    for(int i =0;i < 10000;i++)
                    {
                        queue.offer(i);
                    }
                }
            }
            catch ( Exception e)
            {
                e.printStackTrace();
            }
        }
    }
}
 
 
//需要添加的包
package base;
 
import java.lang.Runnable;
import java.util.concurrent.*;
import java.util.*;
 
public class MyRunnable implements Runnable
{
    public MyRunnable(BlockingQueue<Integer> queue)
    {
        this.queue = queue;
    }
    public void run()
    {
        Date d = new Date();
        long starttime = d.getTime();
        System.err.println(starttime);
        int count = 0;
        while(true)
        {
            try
            {
                Integer i = this.queue.poll();
                if(i != null)
                {
                    count ++;
                }
                if(count == 100000)
                {
                    Date e = new Date();
                    long endtime = e.getTime();
                    System.err.println(count);
                    System.err.println(endtime);
                    System.err.print(endtime - starttime);
                    break;
                }
                 
            }
            catch (Exception e)
            {
                 
            }
        }
    }
    private BlockingQueue<Integer> queue;
}

传递十万条数据，在我的测试机上面，大概需要50ms左右，倒是还可以！索性就看了一下BlockingQueue的底层实现

我在上面的测试代码中使用的offer 和 poll，就看看这两个实现函数吧，首先是offer

public E poll() {
        final AtomicInteger count = this.count;
        if (count.get() == 0)
            return null;
        E x = null;
        int c = -1;
        final ReentrantLock takeLock = this.takeLock;
        takeLock.lock();
        try {
            if (count.get() > 0) {
                x = extract();
                c = count.getAndDecrement();
                if (c > 1)
                    notEmpty.signal();
            }
        } finally {
            takeLock.unlock();
        }
        if (c == capacity)
            signalNotFull();
        return x;
    }

和一般的同步线程类似，只是多加了一个signal,在学习unix环境高级编程时候，看到条件变量用于线程之间的同步，可以实现线程以竞争的方式实现同步！

poll函数的实现也是类似！

public boolean offer(E e) {
        if (e == null) throw new NullPointerException();
        final AtomicInteger count = this.count;
        if (count.get() == capacity)
            return false;
        int c = -1;
        final ReentrantLock putLock = this.putLock;
        putLock.lock();
        try {
            if (count.get() < capacity) {
                insert(e);
                c = count.getAndIncrement();
                if (c + 1 < capacity)
                    notFull.signal();
            }
        } finally {
            putLock.unlock();
        }
        if (c == 0)
            signalNotEmpty();
        return c >= 0;
    }