ExecutorService深入理解

ExecutorService是Executor直接的扩展接口，也是最常用的线程池接口，我们通常见到的线程池定时任务线程池都是它的实现类。

ExecutorService是Java中对线程池定义的一个接口，它java.util.concurrent包中，在这个接口中定义了和后台任务执行相关的方法：

 

Java API对ExecutorService接口的实现有两个，所以这两个即是Java线程池具体实现类（详细了解这两个实现类，点击这里）：

1. ThreadPoolExecutor
2. ScheduledThreadPoolExecutor

此之外，ExecutorService还继承了Executor接口（注意区分Executor接口和Executors工厂类），这个接口只有一个execute()方法，最后我们看一下整个继承树：

 

二、ExecutorService的创建

创建一个什么样的ExecutorService的实例（即线程池）需要g根据具体应用场景而定，不过Java给我们提供了一个Executors工厂类，它可以帮助我们很方便的创建各种类型ExecutorService线程池，Executors一共可以创建下面这四类线程池：

1. newCachedThreadPool 创建一个可缓存线程池，如果线程池长度超过处理需要，可灵活回收空闲线程，若无可回收，则新建线程。
2. newFixedThreadPool 创建一个定长线程池，可控制线程最大并发数，超出的线程会在队列中等待。
3. newScheduledThreadPool 创建一个定长线程池，支持定时及周期性任务执行。
4. newSingleThreadExecutor 创建一个单线程化的线程池，它只会用唯一的工作线程来执行任务，保证所有任务按照指定顺序(FIFO, LIFO, 优先级)执行

以上四个方法底层都是实现 new ThreadPoolExecutor()类

备注：Executors只是一个工厂类，它所有的方法返回的都是ThreadPoolExecutor、ScheduledThreadPoolExecutor这两个类的实例。

三、ExecutorService的使用

ExecutorService executorService = Executors.newFixedThreadPool(10);

executorService.execute(new Runnable() {
public void run() {
    System.out.println("Asynchronous task");
}
});

executorService.shutdown();

四、ExecutorService的执行

ExecutorService有如下几个执行方法：

- execute(Runnable)
- submit(Runnable)
- submit(Callable)
- invokeAny(...)
- invokeAll(...)

4.1 execute(Runnable)

这个方法接收一个Runnable实例，并且异步的执行，请看下面的实例：

ExecutorService executorService = Executors.newSingleThreadExecutor();

executorService.execute(new Runnable() {
public void run() {
    System.out.println("Asynchronous task");
}
});

executorService.shutdown();

这个方法有个问题，就是没有办法获知task的执行结果。如果我们想获得task的执行结果，我们可以传入一个Callable的实例（下面会介绍）。

4.2 submit(Runnable)

submit(Runnable)和execute(Runnable)区别是前者可以返回一个Future对象，通过返回的Future对象，我们可以检查提交的任务是否执行完毕，请看下面执行的例子：

Future future = executorService.submit(new Runnable() {
public void run() {
    System.out.println("Asynchronous task");
}
});

future.get();  //returns null if the task has finished correctly.

package cn.day14;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.Future;
public class Test_02 implements Runnable {
	public void run() {
		// TODO Auto-generated method stub
		System.out.println("Runnable。。。。。");
	}
	public static void main(String[] args) throws Exception {
		// TODO Auto-generated method stub
		ExecutorService executorService = Executors.newFixedThreadPool(2);
		Future future = executorService.submit(new Test_02());
		future.get();
		System.out.println(future.isCancelled());
	}
}

如果任务执行完成，future.get()方法会返回一个null。注意，future.get()方法会产生阻塞。

4.3 submit(Callable)

submit(Callable)和submit(Runnable)类似，也会返回一个Future对象，但是除此之外，submit(Callable)接收的是一个Callable的实现，Callable接口中的call()方法有一个返回值，可以返回任务的执行结果，而Runnable接口中的run()方法是void的，没有返回值。请看下面实例：

Future future = executorService.submit(new Callable(){
public Object call() throws Exception {
    System.out.println("Asynchronous Callable");
    return "Callable Result";
}
});

System.out.println("future.get() = " + future.get());

package cn.day14;

import java.util.concurrent.Callable;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.Future;
public class Test_03 implements Callable<Integer> {

	public Integer call() throws Exception {
		// TODO Auto-generated method stub
		System.out.println("Callable......");
		return 2;
	}
	public static void main(String[] args) throws Exception {
		// TODO Auto-generated method stub
		ExecutorService executorService = Executors.newFixedThreadPool(2);
		Future<Integer> future = executorService.submit(new Test_03());
		Integer number = future.get();
		System.out.println(number);
	}
}

如果任务执行完成，future.get()方法会返回Callable任务的执行结果。注意，future.get()方法会产生阻塞。

4.4 invokeAny(…)

invokeAny(...)方法接收的是一个Callable的集合，执行这个方法不会返回Future，但是会返回所有Callable任务中其中一个任务的执行结果。这个方法也无法保证返回的是哪个任务的执行结果，反正是其中的某一个。请看下面实例：

ExecutorService executorService = Executors.newSingleThreadExecutor();

Set<Callable<String>> callables = new HashSet<Callable<String>>();

callables.add(new Callable<String>() {
public String call() throws Exception {
    return "Task 1";
}
});
callables.add(new Callable<String>() {
public String call() throws Exception {
    return "Task 2";
}
});
callables.add(new Callable<String>() {
    public String call() throws Exception {
    return "Task 3";
}
});

String result = executorService.invokeAny(callables);
System.out.println("result = " + result);
executorService.shutdown();

大家可以尝试执行上面代码，每次执行都会返回一个结果，并且返回的结果是变化的，可能会返回“Task2”也可是“Task1”或者其它。

4.5 invokeAll(…)

invokeAll(...)与 invokeAny(...)类似也是接收一个Callable集合，但是前者执行之后会返回一个Future的List，其中对应着每个Callable任务执行后的Future对象。情况下面这个实例：

ExecutorService executorService = Executors.newSingleThreadExecutor();

Set<Callable<String>> callables = new HashSet<Callable<String>>();

callables.add(new Callable<String>() {
public String call() throws Exception {
    return "Task 1";
}
});
callables.add(new Callable<String>() {
    public String call() throws Exception {
    return "Task 2";
}
});
callables.add(new Callable<String>() {
public String call() throws Exception {
    return "Task 3";
}
});

List<Future<String>> futures = executorService.invokeAll(callables);

for(Future<String> future : futures){
System.out.println("future.get = " + future.get());
}

executorService.shutdown();

五、ExecutorService的关闭

当我们使用完成ExecutorService之后应该关闭它，否则它里面的线程会一直处于运行状态。

举个例子，如果的应用程序是通过main()方法启动的，在这个main()退出之后，如果应用程序中的ExecutorService没有关闭，这个应用将一直运行。之所以会出现这种情况，是因为ExecutorService中运行的线程会阻止JVM关闭。

如果要关闭ExecutorService中执行的线程，我们可以调用ExecutorService.shutdown()方法。在调用shutdown()方法之后，ExecutorService不会立即关闭，但是它不再接收新的任务，直到当前所有线程执行完成才会关闭，所有在shutdown()执行之前提交的任务都会被执行。

如果我们想立即关闭ExecutorService，我们可以调用ExecutorService.shutdownNow()方法。这个动作将跳过所有正在执行的任务和被提交还没有执行的任务。但是它并不对正在执行的任务做任何保证，有可能它们都会停止，也有可能执行完成。

---

关于Java线程池类ThreadPoolExecutor、ScheduledThreadPoolExecutor及Executors工厂类更详细介绍，点击这里

一、前提

/**
 * 线程运行demo，运行时打出线程id以及传入线程中参数
 */
public class ThreadRunner implements Runnable {

    private final SimpleDateFormat format = new SimpleDateFormat("HH:mm:ss.SSS");

    /**
     * 线程私有属性，创建线程时创建
     */
    private Integer num;

    public ThreadRunner(Integer num) {
        this.num = num;
    }

    @Override
    public void run() {
        System.out.println("thread:" + Thread.currentThread().getName() + ",time:" + format.format(new Date()) + ",num:" + num);
        try {//使线程睡眠，模拟线程阻塞情况
            TimeUnit.SECONDS.sleep(1);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
    }
}

 

二、分类

1、FixedThreadPool-有一个固定大小的线程池

public class FixedThreadPoolDemo {

    public static void main(String[] args) {
        ExecutorService pool = Executors.newFixedThreadPool(4);
        for(int i = 0 ; i < 50 ; i++){
            pool.submit(new ThreadRunner((i + 1)));
        }
        pool.shutdown();
    }
}

•  

输出：

thread:pool-1-thread-2,time:16:14:45.677,num:2
thread:pool-1-thread-4,time:16:14:45.678,num:4
thread:pool-1-thread-3,time:16:14:45.680,num:3
thread:pool-1-thread-1,time:16:14:45.684,num:1
thread:pool-1-thread-4,time:16:14:46.680,num:5
thread:pool-1-thread-2,time:16:14:46.680,num:6
thread:pool-1-thread-3,time:16:14:46.680,num:7
thread:pool-1-thread-1,time:16:14:46.684,num:8
thread:pool-1-thread-4,time:16:14:47.680,num:9
thread:pool-1-thread-2,time:16:14:47.680,num:10
thread:pool-1-thread-3,time:16:14:47.681,num:11
thread:pool-1-thread-1,time:16:14:47.684,num:12
thread:pool-1-thread-4,time:16:14:48.681,num:13
thread:pool-1-thread-2,time:16:14:48.681,num:14
thread:pool-1-thread-3,time:16:14:48.681,num:15
thread:pool-1-thread-1,time:16:14:48.684,num:16
thread:pool-1-thread-4,time:16:14:49.681,num:17
thread:pool-1-thread-2,time:16:14:49.682,num:18
thread:pool-1-thread-3,time:16:14:49.682,num:19
thread:pool-1-thread-1,time:16:14:49.684,num:20
thread:pool-1-thread-4,time:16:14:50.681,num:21
thread:pool-1-thread-2,time:16:14:50.682,num:22
thread:pool-1-thread-3,time:16:14:50.682,num:23
thread:pool-1-thread-1,time:16:14:50.684,num:24
thread:pool-1-thread-4,time:16:14:51.681,num:25
thread:pool-1-thread-2,time:16:14:51.682,num:26
thread:pool-1-thread-3,time:16:14:51.682,num:27
thread:pool-1-thread-1,time:16:14:51.684,num:28
thread:pool-1-thread-4,time:16:14:52.681,num:29
thread:pool-1-thread-2,time:16:14:52.682,num:30
thread:pool-1-thread-3,time:16:14:52.682,num:31
thread:pool-1-thread-1,time:16:14:52.684,num:32
thread:pool-1-thread-4,time:16:14:53.681,num:33
thread:pool-1-thread-2,time:16:14:53.682,num:34
thread:pool-1-thread-3,time:16:14:53.683,num:35
thread:pool-1-thread-1,time:16:14:53.685,num:36
thread:pool-1-thread-2,time:16:14:54.682,num:38
thread:pool-1-thread-4,time:16:14:54.682,num:37
thread:pool-1-thread-3,time:16:14:54.683,num:39
thread:pool-1-thread-1,time:16:14:54.686,num:40
thread:pool-1-thread-2,time:16:14:55.682,num:41
thread:pool-1-thread-4,time:16:14:55.682,num:42
thread:pool-1-thread-3,time:16:14:55.683,num:43
thread:pool-1-thread-1,time:16:14:55.686,num:44
thread:pool-1-thread-2,time:16:14:56.682,num:45
thread:pool-1-thread-4,time:16:14:56.683,num:46
thread:pool-1-thread-3,time:16:14:56.684,num:47
thread:pool-1-thread-1,time:16:14:56.686,num:48
thread:pool-1-thread-2,time:16:14:57.683,num:49
thread:pool-1-thread-4,time:16:14:57.683,num:50

•  

总结：
- 池中线程数量固定，不会发生变化
- 使用无界的LinkedBlockingQueue，要综合考虑生成与消费能力，生成过剩，可能导致堆内存溢出。
- 适用一些很稳定很固定的正规并发线程，多用于服务器

2、CachedThreadPool

public class CachedThreadPoolDemo {

    public static void main(String[] args) {
        ExecutorService pool = Executors.newCachedThreadPool();
        for(int i = 0 ; i < 50 ; i++){
            pool.submit(new ThreadRunner((i + 1)));
        }
        pool.shutdown();
    }
}

•  

输出：

thread:pool-1-thread-2,time:16:17:21.289,num:2
thread:pool-1-thread-3,time:16:17:21.290,num:3
thread:pool-1-thread-4,time:16:17:21.290,num:4
thread:pool-1-thread-6,time:16:17:21.291,num:6
thread:pool-1-thread-7,time:16:17:21.291,num:7
thread:pool-1-thread-8,time:16:17:21.291,num:8
thread:pool-1-thread-1,time:16:17:21.292,num:1
thread:pool-1-thread-10,time:16:17:21.293,num:10
thread:pool-1-thread-5,time:16:17:21.294,num:5
thread:pool-1-thread-11,time:16:17:21.294,num:11
thread:pool-1-thread-15,time:16:17:21.294,num:15
thread:pool-1-thread-9,time:16:17:21.294,num:9
thread:pool-1-thread-16,time:16:17:21.295,num:16
thread:pool-1-thread-20,time:16:17:21.295,num:20
thread:pool-1-thread-14,time:16:17:21.296,num:14
thread:pool-1-thread-12,time:16:17:21.296,num:12
thread:pool-1-thread-19,time:16:17:21.297,num:19
thread:pool-1-thread-13,time:16:17:21.299,num:13
thread:pool-1-thread-17,time:16:17:21.300,num:17
thread:pool-1-thread-18,time:16:17:21.302,num:18
thread:pool-1-thread-22,time:16:17:21.304,num:22
thread:pool-1-thread-23,time:16:17:21.304,num:23
thread:pool-1-thread-24,time:16:17:21.305,num:24
thread:pool-1-thread-21,time:16:17:21.305,num:21
thread:pool-1-thread-26,time:16:17:21.305,num:26
thread:pool-1-thread-25,time:16:17:21.306,num:25
thread:pool-1-thread-29,time:16:17:21.307,num:29
thread:pool-1-thread-28,time:16:17:21.308,num:28
thread:pool-1-thread-30,time:16:17:21.308,num:30
thread:pool-1-thread-34,time:16:17:21.308,num:34
thread:pool-1-thread-35,time:16:17:21.308,num:35
thread:pool-1-thread-33,time:16:17:21.308,num:33
thread:pool-1-thread-27,time:16:17:21.309,num:27
thread:pool-1-thread-32,time:16:17:21.308,num:32
thread:pool-1-thread-31,time:16:17:21.309,num:31
thread:pool-1-thread-36,time:16:17:21.310,num:36
thread:pool-1-thread-37,time:16:17:21.310,num:37
thread:pool-1-thread-38,time:16:17:21.310,num:38
thread:pool-1-thread-42,time:16:17:21.310,num:42
thread:pool-1-thread-40,time:16:17:21.310,num:40
thread:pool-1-thread-41,time:16:17:21.311,num:41
thread:pool-1-thread-47,time:16:17:21.762,num:47
thread:pool-1-thread-43,time:16:17:21.762,num:43
thread:pool-1-thread-39,time:16:17:21.762,num:39
thread:pool-1-thread-45,time:16:17:21.762,num:45
thread:pool-1-thread-44,time:16:17:21.763,num:44
thread:pool-1-thread-46,time:16:17:21.761,num:46
thread:pool-1-thread-48,time:16:17:21.761,num:48
thread:pool-1-thread-49,time:16:17:21.765,num:49
thread:pool-1-thread-50,time:16:17:21.765,num:50

•  

总结
- 池中线程时随着处理数据增加而增加
- 线程数并不是一直增加，如果有新任务需要执行时，首先查询池中是否有空闲线程并且还为到空闲截止时间，如果有，则使用空闲线程，如果没有，则创建新线程并放入池中。
- 用于执行一些生存期很短的异步型任务。不适用于IO等长延时操作，因为这可能会创建大量线程，导致系统崩溃。
- 使用SynchronousQueue作为阻塞队列，如果有新任务进入队列，必须队列中数据被其他线程处理，否则会等待。

3、SingleThreadExecutor

public class SingleThreadPoolDemo {

    public static void main(String[] args) {
        ExecutorService pool = Executors.newSingleThreadExecutor();
        for(int i = 0 ; i < 50 ; i++){
            pool.submit(new ThreadRunner((i + 1)));
        }
        pool.shutdown();
    }
}

•  

输出：

thread:pool-1-thread-1,time:16:20:10.194,num:1
thread:pool-1-thread-1,time:16:20:11.197,num:2
thread:pool-1-thread-1,time:16:20:12.197,num:3
thread:pool-1-thread-1,time:16:20:13.197,num:4
thread:pool-1-thread-1,time:16:20:14.197,num:5
thread:pool-1-thread-1,time:16:20:15.198,num:6
thread:pool-1-thread-1,time:16:20:16.198,num:7
thread:pool-1-thread-1,time:16:20:17.198,num:8
thread:pool-1-thread-1,time:16:20:18.198,num:9
thread:pool-1-thread-1,time:16:20:19.198,num:10
thread:pool-1-thread-1,time:16:20:20.198,num:11
thread:pool-1-thread-1,time:16:20:21.199,num:12
thread:pool-1-thread-1,time:16:20:22.200,num:13
thread:pool-1-thread-1,time:16:20:23.200,num:14
thread:pool-1-thread-1,time:16:20:24.200,num:15
thread:pool-1-thread-1,time:16:20:25.200,num:16
thread:pool-1-thread-1,time:16:20:26.201,num:17
thread:pool-1-thread-1,time:16:20:27.201,num:18
thread:pool-1-thread-1,time:16:20:28.201,num:19
thread:pool-1-thread-1,time:16:20:29.201,num:20
thread:pool-1-thread-1,time:16:20:30.202,num:21
thread:pool-1-thread-1,time:16:20:31.202,num:22
thread:pool-1-thread-1,time:16:20:32.203,num:23
thread:pool-1-thread-1,time:16:20:33.203,num:24
thread:pool-1-thread-1,time:16:20:34.203,num:25
thread:pool-1-thread-1,time:16:20:35.203,num:26
thread:pool-1-thread-1,time:16:20:36.203,num:27
thread:pool-1-thread-1,time:16:20:37.203,num:28
thread:pool-1-thread-1,time:16:20:38.203,num:29
thread:pool-1-thread-1,time:16:20:39.203,num:30
thread:pool-1-thread-1,time:16:20:40.203,num:31
thread:pool-1-thread-1,time:16:20:41.203,num:32
thread:pool-1-thread-1,time:16:20:42.203,num:33
thread:pool-1-thread-1,time:16:20:43.204,num:34
thread:pool-1-thread-1,time:16:20:44.204,num:35
thread:pool-1-thread-1,time:16:20:45.204,num:36
thread:pool-1-thread-1,time:16:20:46.204,num:37
thread:pool-1-thread-1,time:16:20:47.205,num:38
thread:pool-1-thread-1,time:16:20:48.205,num:39
thread:pool-1-thread-1,time:16:20:49.205,num:40
thread:pool-1-thread-1,time:16:20:50.206,num:41
thread:pool-1-thread-1,time:16:20:51.206,num:42
thread:pool-1-thread-1,time:16:20:52.207,num:43
thread:pool-1-thread-1,time:16:20:53.207,num:44
thread:pool-1-thread-1,time:16:20:54.207,num:45
thread:pool-1-thread-1,time:16:20:55.207,num:46
thread:pool-1-thread-1,time:16:20:56.207,num:47
thread:pool-1-thread-1,time:16:20:57.208,num:48
thread:pool-1-thread-1,time:16:20:58.208,num:49
thread:pool-1-thread-1,time:16:20:59.209,num:50

•  

总结：
- 线程中只有一个线程在执行
- 适用于有明确执行顺序但是不影响主线程的任务，压入池中的任务会按照队列顺序执行。
- 使用无界的LinkedBlockingQueue，要综合考虑生成与消费能力，生成过剩，可能导致堆内存溢出。

三、源码

public static ExecutorService newFixedThreadPool(int nThreads) {
        return new ThreadPoolExecutor(nThreads, nThreads,
                                      0L, TimeUnit.MILLISECONDS,
                                      new LinkedBlockingQueue<Runnable>());

•  

public static ExecutorService newCachedThreadPool() {
        return new ThreadPoolExecutor(0, Integer.MAX_VALUE,
                                      60L, TimeUnit.SECONDS,
                                      new SynchronousQueue<Runnable>());

•  

 public static ExecutorService newSingleThreadExecutor() {
        return new FinalizableDelegatedExecutorService
            (new ThreadPoolExecutor(1, 1,
                                    0L, TimeUnit.MILLISECONDS,
                                    new LinkedBlockingQueue<Runnable>()));
    }

•  

* ThreadPoolExecutor 构造方法*

public ThreadPoolExecutor(int corePoolSize,
                          int maximumPoolSize,
                          long keepAliveTime,
                          TimeUnit unit,
                          BlockingQueue<Runnable> workQueue) {
    this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue,
         Executors.defaultThreadFactory(), defaultHandler);
}


 public ThreadPoolExecutor(int corePoolSize,
                          int maximumPoolSize,
                          long keepAliveTime,
                          TimeUnit unit,
                          BlockingQueue<Runnable> workQueue,
                          ThreadFactory threadFactory,
                          RejectedExecutionHandler handler) {
    if (corePoolSize < 0 ||
        maximumPoolSize <= 0 ||
        maximumPoolSize < corePoolSize ||
        keepAliveTime < 0)
        throw new IllegalArgumentException();
    if (workQueue == null || threadFactory == null || handler == null)
        throw new NullPointerException();
    this.corePoolSize = corePoolSize;
    this.maximumPoolSize = maximumPoolSize;
    this.workQueue = workQueue;
    this.keepAliveTime = unit.toNanos(keepAliveTime);
    this.threadFactory = threadFactory;
    this.handler = handler;
}

•  
• corePoolSize:线程池核心线程数量
  1. 如果池中线程数量少于核心线程池数量，则直接新建线程处理当前任务。
  2. 核心线程池空闲不会被回收。
  3. 当池中无空闲线程时，新任务将被添加到阻塞队列
• maximumPoolSize：线程池最大线程数量
  1. 当阻塞队列已满，并且有新任务还在入队时，创建新的线程处理，直到线程数大于maximumPoolSize。
  2. 超出corePoolSize部分的线程超过空闲时间后会被回收
  3. 当线程已经超出corePoolSize，并且队列容量已满，则拒绝入队。
• keepAliveTime unit：线程存活时间
  1. 当线程超出corePoolSize时生效
  2. 线程空余keepAliveTime后，将被回收
• workQueue：线程使用阻塞队列
• threadFactory：创建线程池工厂
  1. 用于控制创建线程或者销毁线程时加入其它逻辑
• handler：线程池拒绝策略
  1. 直接丢弃（DiscardPolicy）
  2. 丢弃队列中最老的任务(DiscardOldestPolicy)。
  3. 抛异常(AbortPolicy)
  4. 将任务分给调用线程来执行(CallerRunsPolicy)