java并发之读写锁ReentrantReadWriteLock的使用

Lock比传统线程模型中的synchronized方式更加面向对象，与生活中的锁类似，锁本身也应该是一个对象。两个线程执行的代码片段要实现同步互斥的效果，它们必须用同一个Lock对象。

　　读写锁：分为读锁和写锁，多个读锁不互斥，读锁与写锁互斥，这是由jvm自己控制的，你只要上好相应的锁即可。如果你的代码只读数据，可以很多人同时读，但不能同时写，那就上读锁；如果你的代码修改数据，只能有一个人在写，且不能同时读取，那就上写锁。总之，读的时候上读锁，写的时候上写锁！

　　ReentrantReadWriteLock会使用两把锁来解决问题，一个读锁，一个写锁
线程进入读锁的前提条件：
       没有其他线程的写锁，
       没有写请求或者有写请求，但调用线程和持有锁的线程是同一个

线程进入写锁的前提条件：
       没有其他线程的读锁
       没有其他线程的写锁

到ReentrantReadWriteLock，首先要做的是与ReentrantLock划清界限。它和后者都是单独的实现，彼此之间没有继承或实现的关系。然后就是总结这个锁机制的特性了： 
       (a).重入方面其内部的WriteLock可以获取ReadLock，但是反过来ReadLock想要获得WriteLock则永远都不要想。 
       (b).WriteLock可以降级为ReadLock，顺序是：先获得WriteLock再获得ReadLock，然后释放WriteLock，这时候线程将保持Readlock的持有。反过来ReadLock想要升级为WriteLock则不可能，为什么？参看(a)，呵呵. 
       (c).ReadLock可以被多个线程持有并且在作用时排斥任何的WriteLock，而WriteLock则是完全的互斥。这一特性最为重要，因为对于高读取频率而相对较低写入的数据结构，使用此类锁同步机制则可以提高并发量。 
       (d).不管是ReadLock还是WriteLock都支持Interrupt，语义与ReentrantLock一致。 
       (e).WriteLock支持Condition并且与ReentrantLock语义一致，而ReadLock则不能使用Condition，否则抛出UnsupportedOperationException异常。 

下面看一个读写锁的例子：

/**
 * 模拟数据库表 读数据 写数据
 * @author ko
 *
 */
public class DataQueue implements Runnable {

    private int randomNum;// 随机数
    private List<String> dataList;// 存放数据的集合
    private ReentrantReadWriteLock rwLock;// 读写锁
    
    public DataQueue(int randomNum, List<String> dataList, ReentrantReadWriteLock rwLock) {
        super();
        this.randomNum = randomNum;
        this.dataList = dataList;
        this.rwLock = rwLock;
    }

    public void getData(){
        rwLock.readLock().lock();// 开启读锁  只能允许读的线程访问
        System.out.println("read thread "+Thread.currentThread().getName()+" begin read data");
        StringBuffer sb = new StringBuffer();
        for (String data : dataList) {
            sb.append(data+" ");
        }
        System.out.println("read thread "+Thread.currentThread().getName()+" read data:"+sb.toString());
        System.out.println("read thread "+Thread.currentThread().getName()+" end read data");
        rwLock.readLock().unlock();// 释放读锁
    }
    
    public void setData(){
        rwLock.writeLock().lock();// 开启写锁  其它线程不管是读还是写都不能访问
        System.out.println("write thread "+Thread.currentThread().getName()+" begin write data");
        String data = UUID.randomUUID().toString();
        dataList.add(data);
        System.out.println("write thread "+Thread.currentThread().getName()+" write data:"+data);
        System.out.println("write thread "+Thread.currentThread().getName()+" end write data");
        rwLock.writeLock().unlock();// 释放读锁
    }

    @Override
    public void run() {
        if (randomNum%2 == 0) {
            getData();
        } else {
            setData();
        }
    }
}

/**
 * 利用ReentrantReadWriteLock模拟数据的读写分离
 * @author ko
 *
 */
public class DatabaseReadWriteSeparation {

    public static void main(String[] args) {
        List<String> dataList = new ArrayList<>();
        ReentrantReadWriteLock rwLock = new ReentrantReadWriteLock();
//        DataQueue dataQueue = ;
        ExecutorService exec = Executors.newCachedThreadPool();
        for (int i = 0; i < 10; i++) {
            exec.execute(new DataQueue(new Random().nextInt(10), dataList, rwLock));
        }
        exec.shutdown();
    }
}

read thread pool-1-thread-3 begin read data
read thread pool-1-thread-2 begin read data
read thread pool-1-thread-3 read data:
read thread pool-1-thread-2 read data:
read thread pool-1-thread-3 end read data
read thread pool-1-thread-2 end read data
write thread pool-1-thread-1 begin write data
write thread pool-1-thread-1 write data:73a8bfcc-7cb9-4a36-aa06-ecbf90c3e612
write thread pool-1-thread-1 end write data
write thread pool-1-thread-5 begin write data
write thread pool-1-thread-5 write data:c334bb7a-1dfe-4f64-a996-1ba6f714710e
write thread pool-1-thread-5 end write data
read thread pool-1-thread-4 begin read data
read thread pool-1-thread-6 begin read data
read thread pool-1-thread-4 read data:73a8bfcc-7cb9-4a36-aa06-ecbf90c3e612 c334bb7a-1dfe-4f64-a996-1ba6f714710e 
read thread pool-1-thread-4 end read data
read thread pool-1-thread-6 read data:73a8bfcc-7cb9-4a36-aa06-ecbf90c3e612 c334bb7a-1dfe-4f64-a996-1ba6f714710e 
read thread pool-1-thread-6 end read data
write thread pool-1-thread-7 begin write data
write thread pool-1-thread-7 write data:7266821f-dc72-4a17-8891-6b7ec80a047b
write thread pool-1-thread-7 end write data
write thread pool-1-thread-8 begin write data
write thread pool-1-thread-8 write data:e5fd7de9-3b5c-4a50-8dcb-539d3ca398fd
write thread pool-1-thread-8 end write data
read thread pool-1-thread-10 begin read data
read thread pool-1-thread-10 read data:73a8bfcc-7cb9-4a36-aa06-ecbf90c3e612 c334bb7a-1dfe-4f64-a996-1ba6f714710e 7266821f-dc72-4a17-8891-6b7ec80a047b e5fd7de9-3b5c-4a50-8dcb-539d3ca398fd 
read thread pool-1-thread-10 end read data
read thread pool-1-thread-9 begin read data
read thread pool-1-thread-9 read data:73a8bfcc-7cb9-4a36-aa06-ecbf90c3e612 c334bb7a-1dfe-4f64-a996-1ba6f714710e 7266821f-dc72-4a17-8891-6b7ec80a047b e5fd7de9-3b5c-4a50-8dcb-539d3ca398fd 
read thread pool-1-thread-9 end read data

　　从打印的结果可以看出当读的时候线程2 3、4 6、9 10分别是同时两两进行的，写的时候线程5、7、8分别是单独进行的。