ListenableFuture的使用和原理

java中有几种实现异步的方式(FutureTask/ListenableFuture/CompletableFuture)
这篇介绍的是ListenableFuture，相比FutureTask，本质上只是增加了任务的回调函数，这个是google框架里面的一个东西，我用的版本是guava-27.0-jre.jar
首先还是说使用：

public class ListenableFutureTest {
    public static void main(String[] args) {
        testListenFuture();
    }

    public static void testListenFuture() {
        System.out.println("主线程start");
        ListeningExecutorService pool = MoreExecutors.listeningDecorator(Executors.newFixedThreadPool(5));

        Task task1 = new Task();
        task1.args = "task1";
        Task task2 = new Task();
        task2.args = "task2";
        ListenableFuture<String> future = pool.submit(task1);
        ListenableFuture<String> future2 = pool.submit(task2);

        future2.addListener(() -> System.out.println("addListener 不能带返回值"), pool);

        /**
         * FutureCallBack接口可以对每个任务的成功或失败单独做出响应
         */
        FutureCallback<String> futureCallback = new FutureCallback<String>() {
            @Override
            public void onSuccess(String result) {
                System.out.println("Futures.addCallback 能带返回值：" + result);
            }
            @Override
            public void onFailure(Throwable t) {
                System.out.println("出错,业务回滚或补偿");
            }
        };

        //为任务绑定回调接口
        Futures.addCallback(future, futureCallback, pool);
        System.out.println("主线程end");
    }
}

class Task implements Callable<String> {
    String args;
    @Override
    public String call() throws Exception {
        Thread.sleep(1000);
        System.out.println("任务：" + args);
        return "dong";
    }
}

future2.addListener(() -> System.out.println(“addListener 不能带返回值”), pool);
Futures.addCallback(future, futureCallback, pool);
两种方法都可以，一会说一下原理，其实Futures.addCallback里面也是调用addListener
然后从上到下说一下原理：
上一篇直接用的是Thread(futureTask).start()去执行一个任务
这一篇用了java中的线程池-ThreadPoolExecutor，里面主要是使用execute去执行任务的
MoreExecutors.listeningDecorator就是包装了一下ThreadPoolExecutor，目的是为了使用ListenableFuture

private static class ListeningDecorator extends AbstractListeningExecutorService {
        private final ExecutorService delegate;

        ListeningDecorator(ExecutorService delegate) {
            this.delegate = (ExecutorService)Preconditions.checkNotNull(delegate);
        }

        public final void execute(Runnable command) {
            this.delegate.execute(command);
        }
    }

这里的delegate就是ThreadPoolExecutor
虽然还重写了execute，不过还是直接调用ThreadPoolExecutor里面的execute
ListeningExecutorService pool=MoreExecutors.listeningDecorator(Executors.newFixedThreadPool(5));
这样一个执行器就被new出来了，现在需要往里面放任务了
ListenableFuture future = pool.submit(task1);看看submit代码

public <T> ListenableFuture<T> submit(Callable<T> task) {
        return (ListenableFuture)super.submit(task);
    }

然后调用父类的submit

public <T> Future<T> submit(Callable<T> task) {
        if (task == null) throw new NullPointerException();
        RunnableFuture<T> ftask = newTaskFor(task);
        execute(ftask);
        return ftask;
    }

这里会调用重写的newTaskFor

protected final <T> RunnableFuture<T> newTaskFor(Callable<T> callable) {
        return TrustedListenableFutureTask.create(callable);
    }

class TrustedListenableFutureTask<V> extends TrustedFuture<V> implements RunnableFuture<V> {
    private volatile InterruptibleTask<?> task;
    
    static <V> TrustedListenableFutureTask<V> create(Callable<V> callable) {
        return new TrustedListenableFutureTask(callable);
    }

    TrustedListenableFutureTask(Callable<V> callable) {
        this.task = new TrustedListenableFutureTask.TrustedFutureInterruptibleTask(callable);
    }

    public void run() {
        InterruptibleTask localTask = this.task;
        if (localTask != null) {
            localTask.run();
        }

        this.task = null;
    }
}    

创建了一个TrustedListenableFutureTask，里面有个task是TrustedFutureInterruptibleAsyncTask
这里重写了Runnable的run方法，调用的就是这个task得run方法（也就是我们真正的任务）

private final class TrustedFutureInterruptibleAsyncTask extends InterruptibleTask<ListenableFuture<V>> {
        private final AsyncCallable<V> callable;

        TrustedFutureInterruptibleAsyncTask(AsyncCallable<V> callable) {
            this.callable = (AsyncCallable)Preconditions.checkNotNull(callable);
        }
    }

RunnableFuture ftask = newTaskFor(task);
所以这里返回的是一个RunnableFuture，本质其实是TrustedListenableFutureTask（每个任务都有这么一个future）
execute(ftask);
然后会调用execute，让线程池中某个线程去执行

public void execute(Runnable command) {
        if (command == null)
            throw new NullPointerException();
        int c = ctl.get();
        if (workerCountOf(c) < corePoolSize) {
            if (addWorker(command, true))
                return;
            c = ctl.get();
        }
        if (isRunning(c) && workQueue.offer(command)) {
            int recheck = ctl.get();
            if (! isRunning(recheck) && remove(command))
                reject(command);
            else if (workerCountOf(recheck) == 0)
                addWorker(null, false);
        }
        else if (!addWorker(command, false))
            reject(command);
    }

再说一下addListener，其实就是给任务增加一个回调函数
future2.addListener(() -> System.out.println(“addListener 不能带返回值”), pool);
这里调用的是TrustedListenableFutureTask的addListener
实际上调用的是父类AbstractFuture的addListener

public void addListener(Runnable listener, Executor executor) {
    checkNotNull(listener, "Runnable was null.");
    checkNotNull(executor, "Executor was null.");
    if (!isDone()) {
      Listener oldHead = listeners;
      if (oldHead != Listener.TOMBSTONE) {
        Listener newNode = new Listener(listener, executor);
        do {
          newNode.next = oldHead;
          if (ATOMIC_HELPER.casListeners(this, oldHead, newNode)) {
            return;
          }
          oldHead = listeners; // re-read
        } while (oldHead != Listener.TOMBSTONE);
      }
    }
    // If we get here then the Listener TOMBSTONE was set, which means the future is done, call
    // the listener.
    executeListener(listener, executor);
  }

如果这个任务已经完成，执行executeListener

private static void executeListener(Runnable runnable, Executor executor) {
        try {
            executor.execute(runnable);
        } catch (RuntimeException var3) {
            log.log(java.util.logging.Level.SEVERE, "RuntimeException while executing runnable " + runnable + " with executor " + executor, var3);
        }

    }

这里就是调用执行器的execute方法，相当于给线程池扔一个任务，毕竟回调函数也是一个任务

如果这个任务没有完成，那么把回调函数保存起来，用一个链表保存，链表的头结点就是listener，这个listener里面的回调函数在哪执行呢？在任务执行完之后，我文章最后再说这个问题。
ATOMIC_HELPER.casListeners(this, oldHead, newNode)是给listener赋值的原子操作。

private static final class Listener {
    static final Listener TOMBSTONE = new Listener(null, null);
    final Runnable task;
    final Executor executor;

    // writes to next are made visible by subsequent CAS's on the listeners field
    @Nullable Listener next;

    Listener(Runnable task, Executor executor) {
      this.task = task;
      this.executor = executor;
    }
  }

为什么要用链表呢？因为一个任务可以有多个回调函数

再来看看Futures.addCallback(future, futureCallback, pool);

public static <V> void addCallback(
      final ListenableFuture<V> future,
      final FutureCallback<? super V> callback,
      Executor executor) {
    Preconditions.checkNotNull(callback);
    future.addListener(new CallbackListener<V>(future, callback), executor);
  }

底层用CallbackListener封装了一下，还是调用了addListener
为什么要封装一下呢？因为要拿到任务的返回值

private static final class CallbackListener<V> implements Runnable {
    final Future<V> future;
    final FutureCallback<? super V> callback;

    CallbackListener(Future<V> future, FutureCallback<? super V> callback) {
      this.future = future;
      this.callback = callback;
    }

    @Override
    public void run() {
      final V value;
      try {
        value = getDone(future);
      } catch (ExecutionException e) {
        callback.onFailure(e.getCause());
        return;
      } catch (RuntimeException | Error e) {
        callback.onFailure(e);
        return;
      }
      callback.onSuccess(value);
    }
  }

在这个run方法里面就调用getDone等待返回值，如果有异常就调用callback.onFailure，没有异常就调动callback.onSuccess(value);
这两个方法都是你传入的，这样就能接收到任务的返回值了

 public static <V> V getDone(Future<V> future) throws ExecutionException {
    checkState(future.isDone(), "Future was expected to be done: %s", future);
    return getUninterruptibly(future);
  }
  
public static <V> V getUninterruptibly(Future<V> future) throws ExecutionException {
    boolean interrupted = false;
    try {
      while (true) {
        try {
          return future.get();
        } catch (InterruptedException e) {
          interrupted = true;
        }
      }
    } finally {
      if (interrupted) {
        Thread.currentThread().interrupt();
      }
    }
  }

future.get()就是到等待任务执行结束

现在接着说listener中的回调函数在哪执行的

private final class TrustedFutureInterruptibleTask extends InterruptibleTask<V> {
    private final Callable<V> callable;

    @Override
    void afterRanInterruptibly(V result, Throwable error) {
      if (error == null) {
        TrustedListenableFutureTask.this.set(result);
      } else {
        setException(error);
      }
    }
  }

在任务执行完之后，这里调用了set方法

protected boolean set(@Nullable V value) {
    Object valueToSet = value == null ? NULL : value;
    if (ATOMIC_HELPER.casValue(this, null, valueToSet)) {
      complete(this);
      return true;
    }
    return false;
  }

然后又调用了comlete

private static void complete(AbstractFuture<?> future) {
    Listener next = null;
    outer:
    while (true) {
      future.releaseWaiters();
      next = future.clearListeners(next);
      future = null;
      while (next != null) {
        Listener curr = next;
        next = next.next;
        Runnable task = curr.task;
        if (task instanceof SetFuture) {
          SetFuture<?> setFuture = (SetFuture<?>) task;
          future = setFuture.owner;
          if (future.value == setFuture) {
            Object valueToSet = getFutureValue(setFuture.future);
            if (ATOMIC_HELPER.casValue(future, setFuture, valueToSet)) {
              continue outer;
            }
          }
          // other wise the future we were trying to set is already done.
        } else {
          executeListener(task, curr.executor);
        }
      }
      break;
    }
  }

就是在这里执行完所有的回调函数