【并发编程】Java5 - CompletionService,将异步执行与获取结果分离

1. 简介

  相比Future(【并发编程】Java5 - Future,基本使用),CompletionService除了支持并行执行任务并获取结果外,还支持优先获取到最快执行的任务结果,但CompletionService要求并行执行的任务是无序的。
  使用Future的实现类FutureTask获取多个任务的结果时,当任务未执行完成,主线程会阻塞,只到任务执行完成才会获取到结果;CompletionService的实现类ExecutorCompletionService内部维护了一个存储Future的BlockingQueue,任务执行完成后会把Future保存到队列中。当获取多个任务的结果时,会从BlockingQueue中调用take()方法获取到Future再调用get()方法获取结果,如果暂时没有任务执行完成,则阻塞直到有任务执行完成并保存到BlockingQueue中,这样获取到结果的顺序会按照任务执行快慢的顺序依次返回。

2. CompletionService接口API

方法及参数 描述
Future submit(Callable task) 提交实现Callable的线程任务
Future submit(Runnable task, V result) 提交实现Runnable的线程任务,并指定预期结果
Future take() 获取已完成任务的 Future,没有则阻塞
Future poll() 获取已完成任务的 Future,没有则返回null
Future poll(long timeout, TimeUnit unit) 在指定时间内获取已完成任务的 Future,没有则返回null

3. CompletionService使用场景

  • 获取到任务结果后还需要继续处理,如果获取到结果后立即返回,则无法提现出CompletionService的优势
  • 多个任务并发执行,只需要获取到任意一个任务的结果即可

4. CompletionService应用

  • 前期准备:实现Callable创建线程,支持指定执行时间
/**
 * 创建线程
 */
@RequiredArgsConstructor
class MyCallable implements Callable<Long> {

    /**
     * 任务执行时间
     */
    private final long execTime;

    @Override
    public Long call() throws Exception {
        Thread.sleep(execTime);
        return execTime;
    }
}
  • Future:测试并行执行多个任务获取结果后继续业务处理
/**
 * Future:测试并行执行多个任务获取结果后继续业务处理
 *
 * @throws InterruptedException
 * @throws ExecutionException
 */
@Test
public void testFuture() throws InterruptedException, ExecutionException {
    ExecutorService executorService = Executors.newFixedThreadPool(5);

    List<Future<Long>> futureList = new ArrayList<>(3);

    // 记录开始时间
    StopWatch stopWatch = new StopWatch();
    stopWatch.start();

    // 创建5个任务
    futureList.add(executorService.submit(new MyCallable(5000)));
    futureList.add(executorService.submit(new MyCallable(4000)));
    futureList.add(executorService.submit(new MyCallable(3000)));
    futureList.add(executorService.submit(new MyCallable(2000)));
    futureList.add(executorService.submit(new MyCallable(1000)));

    // 阻塞获取结果
    for (Future<Long> future : futureList) {
        Long result = future.get();
        log.debug("执行结果:" + result);

        // 模拟获取结果后业务处理
        Thread.sleep(500);
    }

    stopWatch.stop();
    log.info("总耗时:" + stopWatch.getTotalTimeMillis() + " ms");

    executorService.shutdown();
}

  控制台打印:

18:53:06.185 [main] DEBUG com.c3stones.test.CompletionServiceTest - 执行结果:5000
18:53:06.690 [main] DEBUG com.c3stones.test.CompletionServiceTest - 执行结果:4000
18:53:07.190 [main] DEBUG com.c3stones.test.CompletionServiceTest - 执行结果:3000
18:53:07.690 [main] DEBUG com.c3stones.test.CompletionServiceTest - 执行结果:2000
18:53:08.191 [main] DEBUG com.c3stones.test.CompletionServiceTest - 执行结果:1000
18:53:08.691 [main] INFO com.c3stones.test.CompletionServiceTest - 总耗时:7516 ms

  通过结果可以看出,结果按照提交的顺序依次返回,总耗时 = 每个任务执行时间 + 获取结果后业务处理时间。

  • CompletionService:测试并行执行多个任务获取结果后继续业务处理
/**
 * CompletionService:测试并行执行多个任务获取结果后继续业务处理
 *
 * @throws InterruptedException
 * @throws ExecutionException
 */
@Test
public void testCompletionService() throws ExecutionException, InterruptedException {
    ExecutorService executorService = Executors.newFixedThreadPool(5);
    CompletionService<Long> completionService = new ExecutorCompletionService<>(executorService);

    // 记录开始时间
    StopWatch stopWatch = new StopWatch();
    stopWatch.start();

    // 创建5个任务
    completionService.submit(new MyCallable(5000));
    completionService.submit(new MyCallable(4000));
    completionService.submit(new MyCallable(3000));
    completionService.submit(new MyCallable(2000));
    completionService.submit(new MyCallable(1000));

    for (int i = 0; i < 5; i++) {
        Long result = completionService.take().get();
        log.debug("执行结果:" + result);

        // 模拟获取结果后业务处理
        Thread.sleep(500);
    }

    stopWatch.stop();
    log.info("总耗时:" + stopWatch.getTotalTimeMillis() + " ms");

    executorService.shutdown();
}

  控制台打印:

18:55:25.727 [main] DEBUG com.c3stones.test.CompletionServiceTest - 执行结果:1000
18:55:26.724 [main] DEBUG com.c3stones.test.CompletionServiceTest - 执行结果:2000
18:55:27.725 [main] DEBUG com.c3stones.test.CompletionServiceTest - 执行结果:3000
18:55:28.725 [main] DEBUG com.c3stones.test.CompletionServiceTest - 执行结果:4000
18:55:29.724 [main] DEBUG com.c3stones.test.CompletionServiceTest - 执行结果:5000
18:55:30.225 [main] INFO com.c3stones.test.CompletionServiceTest - 总耗时:5505 ms

  通过结果可以看出,结果按照提交的任务执行快慢顺序依次返回,总耗时 = 最长任务处理时间 + 获取结果后业务处理时间。

  • CompletionService:测试快速获取任意一个结果
/**
 * CompletionService:测试快速获取任意一个结果
 */
@Test
public void testFastResult() throws InterruptedException, ExecutionException {
    ExecutorService executorService = Executors.newCachedThreadPool();
    CompletionService<Long> completionService = new ExecutorCompletionService<>(executorService);

    // 记录开始时间
    StopWatch stopWatch = new StopWatch();
    stopWatch.start();

    // 创建3个任务
    completionService.submit(new MyCallable(3000));
    completionService.submit(new MyCallable(2000));
    completionService.submit(new MyCallable(1000));

    Long result = completionService.take().get();
    log.debug("执行结果:" + result);

    stopWatch.stop();
    log.info("总耗时:" + stopWatch.getTotalTimeMillis() + " ms");

    executorService.shutdown();
}

  控制台打印:

18:59:58.104 [main] DEBUG com.c3stones.test.CompletionServiceTest - 执行结果:1000
18:59:58.112 [main] INFO com.c3stones.test.CompletionServiceTest - 总耗时:1126 ms

  通过结果可以看出,最快的任务执行完成后结束。

5. 项目地址

  thread-demo

posted @ 2023-03-28 16:55  C3Stones  阅读(109)  评论(0编辑  收藏  举报