最近把《java并发编程实战》-Java Consurrency in Practice 重温了一遍，把书中提到的一些常用工具记录于此：
一、闭锁（门栓）- CountDownLatch
适用场景：多线程测试时，通常为了精确计时，要求所有线程都ready后，才开始执行，防止有线程先起跑，造成不公平，类似的，所有线程执行完，整个程序才算运行完成。
/**
* 闭锁测试（菩提树下的杨过 http://yjmyzz.cnblogs.com/）
*
* @throws InterruptedException
*/
@Test
public void countdownLatch() throws InterruptedException {
CountDownLatch startLatch = new CountDownLatch(1); //类似发令枪
CountDownLatch endLatch = new CountDownLatch(10);//这里的数量，要与线程数相同
for (int i = 0; i < 10; i++) {
Thread t = new Thread(() -> {
try {
startLatch.await(); //先等着，直到发令枪响，防止有线程先run
System.out.println(Thread.currentThread().getName() + " is running...");
Thread.sleep(10);
} catch (InterruptedException e) {
Thread.currentThread().interrupt();
} finally {
endLatch.countDown(); //每个线程执行完成后，计数
}
});
t.setName("线程-" + i);
t.start();
}
long start = System.currentTimeMillis();
startLatch.countDown();//发令枪响，所有线程『开跑』
endLatch.await();//等所有线程都完成
long end = System.currentTimeMillis();
System.out.println("done! exec time => " + (end - start) + " ms");
}
执行结果：
线程-1 is running...
线程-5 is running...
线程-8 is running...
线程-4 is running...
线程-3 is running...
线程-0 is running...
线程-2 is running...
线程-9 is running...
线程-7 is running...
线程-6 is running...
done! exec time => 13 ms
注：大家可以把第14行注释掉，再看看运行结果有什么不同。
二、信号量（Semaphore）
适用场景：用于资源数有限制的并发访问场景。
public class BoundedHashSet<T> {
private final Set<T> set;
private final Semaphore semaphore;
public BoundedHashSet(int bound) {
this.set = Collections.synchronizedSet(new HashSet<T>());
this.semaphore = new Semaphore(bound);
}
public boolean add(T t) throws InterruptedException {
if (!semaphore.tryAcquire(5, TimeUnit.SECONDS)) {
return false;
}
;
boolean added = false;
try {
added = set.add(t);
return added;
} finally {
if (!added) {
semaphore.release();
}
}
}
public boolean remove(Object o) {
boolean removed = set.remove(o);
if (removed) {
semaphore.release();
}
return removed;
}
}
@Test
public void semaphoreTest() throws InterruptedException {
BoundedHashSet<String> set = new BoundedHashSet<>(5);
for (int i = 0; i < 6; i++) {
if (set.add(i + "")) {
System.out.println(i + " added !");
} else {
System.out.println(i + " not add to Set!");
}
}
}
上面的示例将一个普通的Set变成了有界容器。执行结果如下：
0 added !
1 added !
2 added !
3 added !
4 added !
5 not add to Set!
三、栅栏CyclicBarrier
这个跟闭锁类似，可以通过代码设置一个『屏障』点，其它线程到达该点后才能继续，常用于约束其它线程都到达某一状态后，才允许做后面的事情。
public class Worker extends Thread {
private CyclicBarrier cyclicBarrier;
public Worker(CyclicBarrier cyclicBarrier) {
this.cyclicBarrier = cyclicBarrier;
}
private void step1() {
System.out.println(this.getName() + " step 1 ...");
}
private void step2() {
System.out.println(this.getName() + " step 2 ...");
}
public void run() {
step1();
try {
cyclicBarrier.await();
} catch (InterruptedException e) {
e.printStackTrace();
} catch (BrokenBarrierException e) {
e.printStackTrace();
}
step2();
}
}
@Test
public void cyclicBarrierTest() throws InterruptedException, BrokenBarrierException {
CyclicBarrier cyclicBarrier = new CyclicBarrier(11);
for (int i = 0; i < 10; i++) {
Worker w = new Worker(cyclicBarrier);
w.start();
}
cyclicBarrier.await();
}
这里我们假设有一个worder线程，里面有2步操作，要求所有线程完成step1后，才能继续step2. 执行结果如下：
Thread-0 step 1 ...
Thread-1 step 1 ...
Thread-2 step 1 ...
Thread-3 step 1 ...
Thread-4 step 1 ...
Thread-5 step 1 ...
Thread-6 step 1 ...
Thread-7 step 1 ...
Thread-8 step 1 ...
Thread-9 step 1 ...
Thread-9 step 2 ...
Thread-0 step 2 ...
Thread-3 step 2 ...
Thread-4 step 2 ...
Thread-6 step 2 ...
Thread-2 step 2 ...
Thread-1 step 2 ...
Thread-8 step 2 ...
Thread-7 step 2 ...
Thread-5 step 2 ...
四、Exchanger
如果2个线程需要交换数据，Exchanger就能派上用场了，见下面的示例：
@Test
public void exchangerTest() {
Exchanger<String> exchanger = new Exchanger<>();
Thread t1 = new Thread(() -> {
String temp = "AAAAAA";
System.out.println("thread 1 交换前：" + temp);
try {
temp = exchanger.exchange(temp);
} catch (InterruptedException e) {
e.printStackTrace();
}
System.out.println("thread 1 交换后：" + temp);
});
Thread t2 = new Thread(() -> {
String temp = "BBBBBB";
System.out.println("thread 2 交换前：" + temp);
try {
temp = exchanger.exchange(temp);
} catch (InterruptedException e) {
e.printStackTrace();
}
System.out.println("thread 2 交换后：" + temp);
});
t1.start();
t2.start();
}
执行结果：
thread 1 交换前：AAAAAA
thread 2 交换前：BBBBBB
thread 2 交换后：AAAAAA
thread 1 交换后：BBBBBB
五、FutureTask/Future
一些很耗时的操作，可以用Future转化成异步，不阻塞后续的处理，直到真正需要返回结果时调用get拿到结果
@Test
public void futureTaskTest() throws ExecutionException, InterruptedException, TimeoutException {
Callable<String> callable = () -> {
System.out.println("很耗时的操作处理中。。。");
Thread.sleep(5000);
return "done";
};
FutureTask<String> futureTask = new FutureTask<>(callable);
System.out.println("就绪。。。");
new Thread(futureTask).start();
System.out.println("主线程其它处理。。。");
System.out.println(futureTask.get());
System.out.println("处理完成！");
System.out.println("-----------------");
System.out.println("executor 就绪。。。");
ExecutorService executorService = Executors.newSingleThreadExecutor();
Future<String> future = executorService.submit(callable);
System.out.println(future.get(10, TimeUnit.SECONDS));
}
执行结果：
就绪。。。
主线程其它处理。。。
很耗时的操作处理中。。。
done
处理完成！
-----------------
executor 就绪。。。
很耗时的操作处理中。。。
done
六、阻塞队列BlockingQueue
阻塞队列可以在线程间实现生产者-消费者模式。比如下面的示例：线程producer模拟快速生产数据，而线程consumer模拟慢速消费数据，当达到队列的上限时（即：生产者产生的数据，已经放不下了），队列就堵塞住了。
@Test
public void blockingQueueTest() throws InterruptedException {
final BlockingQueue<String> blockingDeque = new ArrayBlockingQueue<>(5);
Thread producer = new Thread() {
public void run() {
Random rnd = new Random();
while (true) {
try {
int i = rnd.nextInt(10000);
blockingDeque.put(i + "");
System.out.println(this.getName() + " 产生了一个数字：" + i);
Thread.sleep(rnd.nextInt(50));//模拟生产者快速生产
} catch (InterruptedException e) {
Thread.currentThread().interrupt();
}
}
}
};
producer.setName("producer 1");
Thread consumer = new Thread() {
public void run() {
while (true) {
Random rnd = new Random();
try {
String i = blockingDeque.take();
System.out.println(this.getName() + " 消费了一个数字：" + i);
Thread.sleep(rnd.nextInt(10000));//消费者模拟慢速消费
} catch (InterruptedException e) {
Thread.currentThread().interrupt();
}
}
}
};
consumer.setName("consumer 1");
producer.start();
consumer.start();
while (true) {
Thread.sleep(100);
}
}
执行结果：
producer 1 产生了一个数字：6773
consumer 1 消费了一个数字：6773
producer 1 产生了一个数字：4456
producer 1 产生了一个数字：8572
producer 1 产生了一个数字：5764
producer 1 产生了一个数字：2874
producer 1 产生了一个数字：780 # 注意这里就已经堵住了，直到有消费者消费一条数据，才能继续生产
consumer 1 消费了一个数字：4456
producer 1 产生了一个数字：4193