彻底理解JDK异步

从同步到异步

如何创建一个新线程

xxxxxxxxxx

new Thread();

如何启动一个新线程

xxxxxxxxxx

new Thread().start();

如何在新线程运行下面方法？

xxxxxxxxxx

public class Task {

​

    public static final String HELLO_WORLD = "Hello,World!";

​

    public void say() {

        System.out.println(Thread.currentThread().getName() + " say : " + HELLO_WORLD);

    }

}

两种方式

xxxxxxxxxx

public class MultiThread {

​

    public static void main(String[] args) {

​

        new Thread(new Runnable() {

            public void run() {

                new Task().say();

            }

        }).start();

​

        new Thread() {

            public void run() {

                new Task().say();

            };

        }.start();

​

    }

​

}

应该使用哪种？

• 传入Runnable对象

  推荐。将需要执行的逻辑封装到Runnable子类中

• 重写run()方法

  需要扩展Thread的功能时使用。

我们干了什么？

• 我们执行了异步任务！
• 我们在主线程执行时，把一个任务交给了另一个线程执行。另一个任务执行的时候不会阻塞当前流程！

我们的任务通常不会如此简单

xxxxxxxxxx

public class Task {

​

    public String say(String param) {

      try {

            TimeUnit.SECONDS.sleep(10);

        } catch (InterruptedException e) {

            e.printStackTrace();

        }

        String result = Thread.currentThread().getName() + " say : " + param;

        System.out.println(result);

        return result;

    }

}

• sleep模拟任务耗时
• 有参数，有返回值的任务如何执行呢？
• 我们需要在主线程中指定不同的参数，同时获取到不同的结果

封装Runnable实现RunnableTask

xxxxxxxxxx

public class RunnableTask implements Runnable {

    private Task task;

    private String param;

    private volatile String result;

​

    public RunnableTask(Task task, String param) {

        this.task = task;

        this.param = param;

    }

​

    @Override

    public void run() {

        this.result = task.say(param);

    }

​

    public String getResult() {

        return result;

    }

​

}

• volatile用于排除线程可见性影响

奇怪的结果？

xxxxxxxxxx

    public static void main(String[] args) throws Exception {

        RunnableTask runnableTask1 = new RunnableTask(new Task(), "Hello,World!");

        new Thread(runnableTask1).start();

        RunnableTask runnableTask2 = new RunnableTask(new Task(), "Hello,zby!");

        new Thread(runnableTask2).start();

        // TimeUnit.SECONDS.sleep(15);

        System.out.println(Thread.currentThread().getName() + " say : " + runnableTask1.getResult());

        System.out.println(Thread.currentThread().getName() + " say : " + runnableTask2.getResult());

xxxxxxxxxx

main say : null

Thread-0 say : Hello,World!

main say : null

Thread-1 say : Hello,zby!

​

• 两个10s的任务，通过异步的方式10s左右就能完成
• 我没有获取到异步执行的结果
• 去掉注释的一行试试？

为什么呢？

• 我们的任务是异步执行的，执行打印结果的时候，我们异步任务可能还没执行完成，因此result尚未设置
• 主线程等待15s，异步线程都执行完毕，可以获取到异步执行结果

优化思路

• 任务本身是异步执行的，但是需要在主线程获取执行结果，那么主线程就需要阻塞等待执行完成

优化后的RunnableTask

xxxxxxxxxx

public class RunnableTask implements Runnable {

    private Task task;

    private String param;

    private volatile String result;

    private volatile boolean complated = false;

​

    public RunnableTask(Task task, String param) {

        this.task = task;

        this.param = param;

    }

​

    @Override

    public void run() {

        synchronized (this) {

            this.result = task.say(param);

            complated = true;

            this.notifyAll();

        }

​

    }

​

    public String getResult() {

        synchronized (this) {

            while (!complated) {

                try {

                    this.wait();

                } catch (InterruptedException e) {

                    e.printStackTrace();

                }

            }

        }

        return result;

    }

​

}

• 获取结果的时候需要判断是否完成，没完成需要阻塞等待完成
• 使用新的标记complated而不判断result是否为null是因为，result可能执行完返回null

还存在的问题

1. 同步机制性能很低，编码灵活度低，如取消等操作无法响应
2. 无法获取或者控制异步执行过程
3. 我不想对每个任务都进行封装
4. 我的参数和返回值类型不确定

解决参数问题

• 可以看到我们的优化代码并没有对参数做处理，其实我们创建Runnable子类的时候就可以指定有参构造方法，并不需要额外处理

解决返回值和异步过程控制

• 抽象出执行结果获取和执行控制过程的接口

xxxxxxxxxx

package java.util.concurrent;

​

public interface Future<V> {

​

    boolean cancel(boolean mayInterruptIfRunning);

​

    boolean isCancelled();

​

    boolean isDone();

​

    V get() throws InterruptedException, ExecutionException;

​

    V get(long timeout, TimeUnit unit)

        throws InterruptedException, ExecutionException, TimeoutException;

}

​

• 这个接口是jdk提供的，可以满足普通需求
• 但是，功能有限，只能阻塞方式获取结果。如Netty对它进行了更丰富的扩展，使用观察者模式监听执行结果

Runnable升级为RunnableFuture

• 需要获取异步结果的Runnable子类，应该同时实现Runnable接口和Future接口，并提供对于方法的实现

xxxxxxxxxx

package java.util.concurrent;

​

public interface RunnableFuture<V> extends Runnable, Future<V> {

​

    void run();

}

​

• 我们不需要每次都去实现两个接口
• 这个接口也是JDK提供的，可以执行任务并获取结果

我需要自己实现吗？

• Future接口提供的功能，是跟业务代码无关的
• run()方法执行的逻辑应该是独特的
• 我们可以使用装饰模式对Runnable进行实现。
• 可以对Future提供的方法进行统一的实现

Runnable另一个问题

• 我们在Runnable接口的run()方法中写自己的逻辑代码，没办法直接返回结果
• 我们装饰的是Runnable接口，该接口无法获取返回值，需要从它的内部获取返回值
• 为了获取到我们任务执行的结果，可能需要这么做

xxxxxxxxxx

public interface ResultRunnable<V> extends Runnable {

    V getResult();

}

​

​

public class MyRunnableFuture<V> implements Runnable, Future<V> {

    private ResultRunnable<V> resultRunnable;

​

    public MyRunnableFuture(ResultRunnable<V> resultRunnable) {

        this.resultRunnable = resultRunnable;

    }

​

    @Override

    public boolean cancel(boolean mayInterruptIfRunning) {

        // TODO Auto-generated method stub

        return false;

    }

​

    @Override

    public boolean isCancelled() {

      // TODO Auto-generated method stub

        return false;

    }

​

    @Override

    public boolean isDone() {

      // TODO Auto-generated method stub

        return false;

    }

​

    @Override

    public V get() throws InterruptedException, ExecutionException {

      // TODO Auto-generated method stub

        // 阻塞直到执行完成

        return null;

    }

​

    @Override

    public V get(long timeout, TimeUnit unit)

            throws InterruptedException, ExecutionException, TimeoutException {

      // TODO Auto-generated method stub

        return null;

    }

​

    @Override

    public void run() {

        resultRunnable.run();

        resultRunnable.getResult();//获取结果并设置到当前的属性，以便get()能获取

        // 通知阻塞结束

​

    }

​

}

• 这样编码的话，我们需要手动设置执行结果
• 或者抽象结果的设置，我们执行完调用set设置，还是很麻烦，忘了咋办

来一个更好写逻辑代码的接口Callable

xxxxxxxxxx

 package java.util.concurrent;

​

@FunctionalInterface

public interface Callable<V> {

​

    V call() throws Exception;

}

​

• 这个接口就符合我们编写业务代码的习惯

Callable接口的优点

• 对Callable接口进行装饰，可以调用执行业务代码，直接获取到返回值
• 无需对业务代码再次封装

对RunnableFuture的实现大概就是这样

class FutureTask<V> implements RunnableFuture<V> {

​

    private Callable<V> callable;

​

    public FutureTask(Callable<V> callable) {

        this.callable = callable;

    }

​

    @Override

    public boolean cancel(boolean mayInterruptIfRunning) {

        // TODO Auto-generated method stub

        return false;

    }

​

    @Override

    public boolean isCancelled() {

        // TODO Auto-generated method stub

        return false;

    }

​

    @Override

    public boolean isDone() {

        // TODO Auto-generated method stub

        return false;

    }

​

    @Override

    public V get() throws InterruptedException, ExecutionException {

        // TODO Auto-generated method stub

        return null;

    }

​

    @Override

    public V get(long timeout, TimeUnit unit)

            throws InterruptedException, ExecutionException, TimeoutException {

        // TODO Auto-generated method stub

        return null;

    }

​

    @Override

    public void run() {

        try {

            V call = callable.call();// 把结果设置到当前类的字段中

            // 唤醒等待获取线程

        } catch (Exception e) {

            e.printStackTrace();

        }

    }

​

}

可是我也有没有返回值的业务代码

• 显然Callable比Runnable更强大，多一个返回值
• 为了兼容Runnable，可以使用适配器模式进行封装

xxxxxxxxxx

class RunnableAdapter<T> implements Callable<T> {

        final Runnable task;

        final T result;

        RunnableAdapter(Runnable task, T result) {

            this.task = task;

            this.result = result;

        }

        public T call() {

            task.run();

            return result;

        }

    }

• Runnable接口可以完美转换为Callable

JDK中FutureTask

xxxxxxxxxx

package java.util.concurrent;

​

public class FutureTask<V> implements RunnableFuture<V> {

    /*

     * Revision notes: This differs from previous versions of this

     * class that relied on AbstractQueuedSynchronizer, mainly to

     * avoid surprising users about retaining interrupt status during

     * cancellation races. Sync control in the current design relies

     * on a "state" field updated via CAS to track completion, along

     * with a simple Treiber stack to hold waiting threads.

     *

     * Style note: As usual, we bypass overhead of using

     * AtomicXFieldUpdaters and instead directly use Unsafe intrinsics.

     */

​

    /**

     * The run state of this task, initially NEW.  The run state

     * transitions to a terminal state only in methods set,

     * setException, and cancel.  During completion, state may take on

     * transient values of COMPLETING (while outcome is being set) or

     * INTERRUPTING (only while interrupting the runner to satisfy a

     * cancel(true)). Transitions from these intermediate to final

     * states use cheaper ordered/lazy writes because values are unique

     * and cannot be further modified.

     *

     * Possible state transitions:

     * NEW -> COMPLETING -> NORMAL

     * NEW -> COMPLETING -> EXCEPTIONAL

     * NEW -> CANCELLED

     * NEW -> INTERRUPTING -> INTERRUPTED

     */

    private volatile int state;

    private static final int NEW          = 0;

    private static final int COMPLETING   = 1;

    private static final int NORMAL       = 2;

    private static final int EXCEPTIONAL  = 3;

    private static final int CANCELLED    = 4;

    private static final int INTERRUPTING = 5;

    private static final int INTERRUPTED  = 6;

​

    /** The underlying callable; nulled out after running */

    private Callable<V> callable;

    /** The result to return or exception to throw from get() */

    private Object outcome; // non-volatile, protected by state reads/writes

    /** The thread running the callable; CASed during run() */

    private volatile Thread runner;

    /** Treiber stack of waiting threads */

    private volatile WaitNode waiters;

​

    /**

     * Returns result or throws exception for completed task.

     *

     * @param s completed state value

     */

    @SuppressWarnings("unchecked")

    private V report(int s) throws ExecutionException {

        Object x = outcome;

        if (s == NORMAL)

            return (V)x;

        if (s >= CANCELLED)

            throw new CancellationException();

        throw new ExecutionException((Throwable)x);

    }

​

    /**

     * Creates a {@code FutureTask} that will, upon running, execute the

     * given {@code Callable}.

     *

     * @param  callable the callable task

     * @throws NullPointerException if the callable is null

     */

    public FutureTask(Callable<V> callable) {

        if (callable == null)

            throw new NullPointerException();

        this.callable = callable;

        this.state = NEW;       // ensure visibility of callable

    }

​

    /**

     * Creates a {@code FutureTask} that will, upon running, execute the

     * given {@code Runnable}, and arrange that {@code get} will return the

     * given result on successful completion.

     *

     * @param runnable the runnable task

     * @param result the result to return on successful completion. If

     * you don't need a particular result, consider using

     * constructions of the form:

     * {@code Future<?> f = new FutureTask<Void>(runnable, null)}

     * @throws NullPointerException if the runnable is null

     */

    public FutureTask(Runnable runnable, V result) {

        this.callable = Executors.callable(runnable, result);

        this.state = NEW;       // ensure visibility of callable

    }

​

    public boolean isCancelled() {

        return state >= CANCELLED;

    }

​

    public boolean isDone() {

        return state != NEW;

    }

​

    public boolean cancel(boolean mayInterruptIfRunning) {

        if (!(state == NEW &&

              UNSAFE.compareAndSwapInt(this, stateOffset, NEW,

                  mayInterruptIfRunning ? INTERRUPTING : CANCELLED)))

            return false;

        try {    // in case call to interrupt throws exception

            if (mayInterruptIfRunning) {

                try {

                    Thread t = runner;

                    if (t != null)

                        t.interrupt();

                } finally { // final state

                    UNSAFE.putOrderedInt(this, stateOffset, INTERRUPTED);

                }

            }

        } finally {

            finishCompletion();

        }

        return true;

    }

​

    /**

     * @throws CancellationException {@inheritDoc}

     */

    public V get() throws InterruptedException, ExecutionException {

        int s = state;

        if (s <= COMPLETING)

            s = awaitDone(false, 0L);

        return report(s);

    }

​

    /**

     * @throws CancellationException {@inheritDoc}

     */

    public V get(long timeout, TimeUnit unit)

        throws InterruptedException, ExecutionException, TimeoutException {

        if (unit == null)

            throw new NullPointerException();

        int s = state;

        if (s <= COMPLETING &&

            (s = awaitDone(true, unit.toNanos(timeout))) <= COMPLETING)

            throw new TimeoutException();

        return report(s);

    }

​

    /**

     * Protected method invoked when this task transitions to state

     * {@code isDone} (whether normally or via cancellation). The

     * default implementation does nothing.  Subclasses may override

     * this method to invoke completion callbacks or perform

     * bookkeeping. Note that you can query status inside the

     * implementation of this method to determine whether this task

     * has been cancelled.

     */

    protected void done() { }

​

    /**

     * Sets the result of this future to the given value unless

     * this future has already been set or has been cancelled.

     *

     * <p>This method is invoked internally by the {@link #run} method

     * upon successful completion of the computation.

     *

     * @param v the value

     */

    protected void set(V v) {

        if (UNSAFE.compareAndSwapInt(this, stateOffset, NEW, COMPLETING)) {

            outcome = v;

            UNSAFE.putOrderedInt(this, stateOffset, NORMAL); // final state

            finishCompletion();

        }

    }

​

    /**

     * Causes this future to report an {@link ExecutionException}

     * with the given throwable as its cause, unless this future has

     * already been set or has been cancelled.

     *

     * <p>This method is invoked internally by the {@link #run} method

     * upon failure of the computation.

     *

     * @param t the cause of failure

     */

    protected void setException(Throwable t) {

        if (UNSAFE.compareAndSwapInt(this, stateOffset, NEW, COMPLETING)) {

            outcome = t;

            UNSAFE.putOrderedInt(this, stateOffset, EXCEPTIONAL); // final state

            finishCompletion();

        }

    }

​

    public void run() {

        if (state != NEW ||

            !UNSAFE.compareAndSwapObject(this, runnerOffset,

                                         null, Thread.currentThread()))

            return;

        try {

            Callable<V> c = callable;

            if (c != null && state == NEW) {

                V result;

                boolean ran;

                try {

                    result = c.call();

                    ran = true;

                } catch (Throwable ex) {

                    result = null;

                    ran = false;

                    setException(ex);

                }

                if (ran)

                    set(result);

            }

        } finally {

            // runner must be non-null until state is settled to

            // prevent concurrent calls to run()

            runner = null;

            // state must be re-read after nulling runner to prevent

            // leaked interrupts

            int s = state;

            if (s >= INTERRUPTING)

                handlePossibleCancellationInterrupt(s);

        }

    }

​

    /**

     * Executes the computation without setting its result, and then

     * resets this future to initial state, failing to do so if the

     * computation encounters an exception or is cancelled.  This is

     * designed for use with tasks that intrinsically execute more

     * than once.

     *

     * @return {@code true} if successfully run and reset

     */

    protected boolean runAndReset() {

        if (state != NEW ||

            !UNSAFE.compareAndSwapObject(this, runnerOffset,

                                         null, Thread.currentThread()))

            return false;

        boolean ran = false;

        int s = state;

        try {

            Callable<V> c = callable;

            if (c != null && s == NEW) {

                try {

                    c.call(); // don't set result

                    ran = true;

                } catch (Throwable ex) {

                    setException(ex);

                }

            }

        } finally {

            // runner must be non-null until state is settled to

            // prevent concurrent calls to run()

            runner = null;

            // state must be re-read after nulling runner to prevent

            // leaked interrupts

            s = state;

            if (s >= INTERRUPTING)

                handlePossibleCancellationInterrupt(s);

        }

        return ran && s == NEW;

    }

​

    /**

     * Ensures that any interrupt from a possible cancel(true) is only

     * delivered to a task while in run or runAndReset.

     */

    private void handlePossibleCancellationInterrupt(int s) {

        // It is possible for our interrupter to stall before getting a

        // chance to interrupt us.  Let's spin-wait patiently.

        if (s == INTERRUPTING)

            while (state == INTERRUPTING)

                Thread.yield(); // wait out pending interrupt

​

        // assert state == INTERRUPTED;

​

        // We want to clear any interrupt we may have received from

        // cancel(true).  However, it is permissible to use interrupts

        // as an independent mechanism for a task to communicate with

        // its caller, and there is no way to clear only the

        // cancellation interrupt.

        //

        // Thread.interrupted();

    }

​

    /**

     * Simple linked list nodes to record waiting threads in a Treiber

     * stack.  See other classes such as Phaser and SynchronousQueue

     * for more detailed explanation.

     */

    static final class WaitNode {

        volatile Thread thread;

        volatile WaitNode next;

        WaitNode() { thread = Thread.currentThread(); }

    }

​

    /**

     * Removes and signals all waiting threads, invokes done(), and

     * nulls out callable.

     */

    private void finishCompletion() {

        // assert state > COMPLETING;

        for (WaitNode q; (q = waiters) != null;) {

            if (UNSAFE.compareAndSwapObject(this, waitersOffset, q, null)) {

                for (;;) {

                    Thread t = q.thread;

                    if (t != null) {

                        q.thread = null;

                        LockSupport.unpark(t);

                    }

                    WaitNode next = q.next;

                    if (next == null)

                        break;

                    q.next = null; // unlink to help gc

                    q = next;

                }

                break;

            }

        }

​

        done();

​

        callable = null;        // to reduce footprint

    }

​

    /**

     * Awaits completion or aborts on interrupt or timeout.

     *

     * @param timed true if use timed waits

     * @param nanos time to wait, if timed

     * @return state upon completion

     */

    private int awaitDone(boolean timed, long nanos)

        throws InterruptedException {

        final long deadline = timed ? System.nanoTime() + nanos : 0L;

        WaitNode q = null;

        boolean queued = false;

        for (;;) {

            if (Thread.interrupted()) {

                removeWaiter(q);

                throw new InterruptedException();

            }

​

            int s = state;

            if (s > COMPLETING) {

                if (q != null)

                    q.thread = null;

                return s;

            }

            else if (s == COMPLETING) // cannot time out yet

                Thread.yield();

            else if (q == null)

                q = new WaitNode();

            else if (!queued)

                queued = UNSAFE.compareAndSwapObject(this, waitersOffset,

                                                     q.next = waiters, q);

            else if (timed) {

                nanos = deadline - System.nanoTime();

                if (nanos <= 0L) {

                    removeWaiter(q);

                    return state;

                }

                LockSupport.parkNanos(this, nanos);

            }

            else

                LockSupport.park(this);

        }

    }

​

    /**

     * Tries to unlink a timed-out or interrupted wait node to avoid

     * accumulating garbage.  Internal nodes are simply unspliced

     * without CAS since it is harmless if they are traversed anyway

     * by releasers.  To avoid effects of unsplicing from already

     * removed nodes, the list is retraversed in case of an apparent

     * race.  This is slow when there are a lot of nodes, but we don't

     * expect lists to be long enough to outweigh higher-overhead

     * schemes.

     */

    private void removeWaiter(WaitNode node) {

        if (node != null) {

            node.thread = null;

            retry:

            for (;;) {          // restart on removeWaiter race

                for (WaitNode pred = null, q = waiters, s; q != null; q = s) {

                    s = q.next;

                    if (q.thread != null)

                        pred = q;

                    else if (pred != null) {

                        pred.next = s;

                        if (pred.thread == null) // check for race

                            continue retry;

                    }

                    else if (!UNSAFE.compareAndSwapObject(this, waitersOffset,

                                                          q, s))

                        continue retry;

                }

                break;

            }

        }

    }

​

    // Unsafe mechanics

    private static final sun.misc.Unsafe UNSAFE;

    private static final long stateOffset;

    private static final long runnerOffset;

    private static final long waitersOffset;

    static {

        try {

            UNSAFE = sun.misc.Unsafe.getUnsafe();

            Class<?> k = FutureTask.class;

            stateOffset = UNSAFE.objectFieldOffset

                (k.getDeclaredField("state"));

            runnerOffset = UNSAFE.objectFieldOffset

                (k.getDeclaredField("runner"));

            waitersOffset = UNSAFE.objectFieldOffset

                (k.getDeclaredField("waiters"));

        } catch (Exception e) {

            throw new Error(e);

        }

    }

​

}

​

• 使用int字段表示当前任务状态流转
• 使用链表对阻塞线程进行排队
• 提供通用的异步任务封装

执行有返回值的任务

x

    public static void main(String[] args) throws Exception {

        FutureTask<String> futureTask = new FutureTask<String>(new Callable<String>() {

​

            @Override

            public String call() throws Exception {

                return new Task().say("Hello,World!");

            }

        });

        new Thread(futureTask).start();

      System.out.println(futureTask.isDone());//直接返回

        System.out.println(Thread.currentThread().getName() + " say : " + futureTask.get());//等待结果

    }

• 自动阻塞等待获取结果
• 不需要处理同步、排队
• 可以对futureTask进行操作

总结

• JDK1.0开始对多线程只提供了Thread类和Runnable接口的简单支持
• JDK的创建线程本不支持获取线程执行结果的返回值，但是通过大师一系列封装，一切都变得简单了
• 另外可以看到Runnable接口是1.0，Callable接口是1.5，那么获取异步结果在这之间JDK是有空白的
• 如果带着Callable接口回到1.0，那么Thread类的启动线程直接运行Callable.call()方法，Thread.start()方法直接返回Future，会不会更简单容易理解了
• 要是Thread.start()支持参数传递，会不会更完美了