[本文是我对Java Concurrency In Practice 7.2的归纳和总结. 转载请注明作者和出处, 如有谬误, 欢迎在评论中指正. ]
以ExecutorService为例, 该类内部封装有多个线程, 类外部无法直接停止这些线程. 相反, 外部调用Service的shutDown和shutDownNow方法关闭Service, 而Service负责停止其拥有的线程.
大多数server应用会使用到log, 下例中的LogWriter是一个使用生产者消费者模式构建的log service, 需要打印log的线程将待打印的内容加入到阻塞队列中, 而logger线程则不断的从阻塞队列中取出数据输出:
public class LogWriter { private final BlockingQueue<String> queue; private final LoggerThread logger; public LogWriter(Writer writer) { this.queue = new LinkedBlockingQueue<String>(CAPACITY); this.logger = new LoggerThread(writer); } public void start() { logger.start(); } /** * 需要打印数据的线程调用该方法, 将待打印数据加入阻塞队列 */ public void log(String msg) throws InterruptedException { queue.put(msg); } /** * 负责从阻塞队列中取出数据输出的线程 */ private class LoggerThread extends Thread { private final PrintWriter writer; // ... public void run() { try { while (true) writer.println(queue.take()); } catch (InterruptedException ignored) { } finally { writer.close(); } } } }
LogWriter内部封装有LoggerThread线程, 所以LogWriter是一个基于线程构建的Service. 根据ExecutorService的经验, 我们需要在LogWriter中提供停止LoggerThread线程的方法. 看起来这并不是很难, 我们只需在LogWriter中添加shutDown方法:
/** * 该方法用于停止LoggerThread线程 */ public void shutDown() { logger.interrupt(); }
当LogWriter.shutDown方法被调用时, LoggerThread线程的中断标记被设置为true, 之后LoggerThread线程执行queue.take()方法时会抛出InterruptedException异常, 从而使得LoggerThread线程结束.
然而这样的shutDown方法并不是很恰当:
1. 丢弃了队列中尚未来得及输出的数据.
2. 更严重的是, 假如线程A对LogWriter.log方法的调用因为队列已满而阻塞, 此时停止LoggerThread线程将导致线程A永远阻塞在queue.put方法上.
对上例的改进:
public class LogWriter { private final BlockingQueue<String> queue; private final LoggerThread loggerThread; private final PrintWriter writer; /** * 表示是否关闭Service */ private boolean isShutdown; /** * 队列中待处理数据的数量 */ private int reservations; public void start() { loggerThread.start(); } public void shutDown() { synchronized (this) { isShutdown = true; } loggerThread.interrupt(); } public void log(String msg) throws InterruptedException { synchronized (this) { // service已关闭后调用log方法直接抛出异常 if (isShutdown) throw new IllegalStateException("Service has been shut down"); ++reservations; } // BlockingQueue本身就是线程安全的, put方法的调用不在同步代码块中 // 我们只需要保证isShutdown和reservations是线程安全的即可 queue.put(msg); } private class LoggerThread extends Thread { public void run() { try { while (true) { try { synchronized (this) { // 当service已关闭且处理完队列中的所有数据时才跳出while循环 if (isShutdown && reservations == 0) break; } String msg = queue.take(); synchronized (this) { --reservations; } writer.println(msg); } catch (InterruptedException e) { // 发生InterruptedException异常时不应该立刻跳出while循环 // 而应该继续输出log, 直到处理完队列中的所有数据 } } } finally { writer.close(); } } } }
上面的处理显得过于复杂, 利用ExecutorService可以编写出相对更简洁的程序:
public class LogService { /** * 创建只包含单个线程的线程池, 提交给该线程池的任务将以串行的方式逐个运行 * 本例中, 此线程用于执行打印log的任务 */ private final ExecutorService exec = Executors.newSingleThreadExecutor(); private final PrintWriter writer; public void start() { } public void shutdown() throws InterruptedException { try { // 关闭ExecutorService后再调用其awaitTermination将导致当前线程阻塞, 直到所有已提交的任务执行完毕, 或者发生超时 exec.shutdown(); exec.awaitTermination(TIMEOUT, UNIT); } finally { writer.close(); } } public void log(String msg) { try { // 线程池关闭后再调用其execute方法将抛出RejectedExecutionException异常 exec.execute(new WriteTask(msg)); } catch (RejectedExecutionException ignored) { } } private final class WriteTask implements Runnable { private String msg; public WriteTask(String msg) { this.msg = msg; } @Override public void run() { writer.println(msg); } } }