摘要
涉及多线程问题,往往绕不开锁。在 JDK 1.5 之前,Java 通过 synchronized 关键字来实现锁的功能,该方式是语法层面的,由 JVM 实现。JDK 1.5 增加了锁在 API 层面的实现,也就是 java.util.concurrent.locks.Lock 接口及其相关的实现类,它不仅具备 synchronized 的功能,而且还增加了更加丰富的功能。通常与其配合使用的还有 Condition 接口。
Lock接口的定义public interface Lock {
// 阻塞式获取锁,该方法与synchronized功能类似
void lock();
// 获取锁,可响应中断
void lockInterruptibly() throws InterruptedException;
// 尝试获取锁,若成功返回true;否则返回false
boolean tryLock();
// 尝试获取锁(在给定的时间内),若成功返回true;否则返回false
boolean tryLock(long time, TimeUnit unit) throws InterruptedException;
// 释放锁
void unlock();
// 创建一个与该锁绑定的Condition
Condition newCondition();
}
Condition 接口定义如下:
public interface Condition {
// 使当前线程等待,直到被signal唤醒或被中断
void await() throws InterruptedException;
// 使当前线程等待,直到被signal唤醒(不响应中断)
void awaitUninterruptibly();
// 使当前线程等待,直到被signal唤醒、或被中断、或到达等待时间
long awaitNanos(long nanosTimeout) throws InterruptedException;
// 使当前线程等待,直到被signal唤醒、或被中断、或到达等待时间(与上面方法类似)
boolean await(long time, TimeUnit unit) throws InterruptedException;
// // 使当前线程等待,直到被signal唤醒、或被中断、或到达给定的截止时间
boolean awaitUntil(Date deadline) throws InterruptedException;
// 唤醒一个等待的线程
void signal();
// 唤醒所有等待的线程
void signalAll();
}
Condition 的方法虽然不少,但其实就两类:
- await* 方法:让当前线程处于等待状态;
- signal* 方法:唤醒处于等待的线程。
定义了读锁和写锁,其中读锁是共享的,写锁是互斥的。
public interface ReadWriteLock {
// 读锁
Lock readLock();
// 写锁
Lock writeLock();
}
synchronized实战
public class ProdConsumerTest {
private static final Object monitor = new Object();
private Random random = new Random();
private static final int SIZE = 10;
private Queue queue = new LinkedList();
private void produce() throws InterruptedException {
for (; ; ) {
//使用 synchronized 锁
synchronized (monitor) {
if (queue.size() >= SIZE) {
monitor.wait();
}
int nextInt = random.nextInt(1000);
queue.offer(nextInt);
sleep(400);
System.out.println("size=" + queue.size() + ", 生产-->" + nextInt);
//通知
monitor.notify();
}
}
}
private void consume() throws InterruptedException {
for (; ; ) {
//使用 synchronized 锁
synchronized (monitor) {
if (queue.size() {
try {
test.produce();
} catch (InterruptedException e) {
e.printStackTrace();
}
});
Thread t2 = new Thread(() -> {
try {
test.consume();
} catch (InterruptedException e) {
e.printStackTrace();
}
});
t1.start();
t2.start();
}
}
Lock/Condition 实现
public class ProdConsumerTest {
private static final int SIZE = 10;
private Random random = new Random();
private Queue queue = new LinkedList();
// ReentrantLock 是 JDK 提供的 Lock 接口实现类,后文分析其原理
private Lock lock = new ReentrantLock();
private Condition notFull = lock.newCondition();
private Condition notEmpty = lock.newCondition();
private void produce() throws InterruptedException {
for (; ; ) {
//使用lock锁
lock.lock();
try {
if (queue.size() >= SIZE) {
notFull.await();
}
int nextInt = random.nextInt(1000);
queue.offer(nextInt);
sleep(400);
System.out.println("size=" + queue.size() + ", 生产-->" + nextInt);
notEmpty.signal();
} finally {
//使用unlock锁
lock.unlock();
}
}
}
private void consume() throws InterruptedException {
for (; ; ) {
//使用lock锁
lock.lock();
try {
if (queue.size() {
try {
test.produce();
} catch (InterruptedException e) {
e.printStackTrace();
}
});
Thread t2 = new Thread(() -> {
try {
test.consume();
} catch (InterruptedException e) {
e.printStackTrace();
}
});
t1.start();
t2.start();
}
}
通过对比发现,Lock/Condition 可以实现和 synchronized 一样的功能,而 Condition 的 await/signal 则相当于 Object 的 wait/notify。
博文参考