通过直接继承Thread, 实现Runnable接口来创建线程。但这两种方式都有一种缺陷:在执行完任务之后无法获得执行结果。
如果需要获得执行结果,就必须通过共享变量或者使用线程通信的方式来达到效果,这样使用起来比较麻烦,而jdk中Callable和Future,通过他们可以在任务执行完毕之后得到任务执行结果。先看看他们之间的组织关系:
Callable:
public interface Callable<V> {
/**
* Computes a result, or throws an exception if unable to do so.
*
* @return computed result
* @throws Exception if unable to compute a result
*/
V call() throws Exception;
}
源码可知,它也是个一个接口,在他里面也只是申明一个方法,只不过这个方法为call(),call方法返回的就是该泛型传递进来的V类型,他怎么使用呢?就是结合之前的ExecuteService:
<T> Future<T> submit(Callable<T> task); <T> Future<T> submit(Runnable task, T result); Future<?> submit(Runnable task);
第一个submit方法里面的参数类型就是Callable。
Future:
Future就是对于具体的Runnable或者Callable任务的执行进度的查看,取消,查询是否完成,获取结果(正确完成时的结果,或异常)。必要时可以通过get方法获取执行的结果,该方法会阻塞直到任务返回结果,或通过指定阻塞时间的版本。
public interface Future<V> { boolean cancel(boolean mayInterruptIfRunning); boolean isDone(); V get() throws InterruptedException, ExecutionException; V get(long timeout, TimeUnit unit)
throws InterruptedException, ExecutionException, TimeoutException;
}
其中cancel()方法用来取消任务,如果取消任务成功则返回true, 如果取消任务失败则返回false。 参数mayInterruptIfRunning表示是否允许取消真在执行去没有执行完毕的任务,如果设置true, 则表示可以取消正在执行过程的任务。 当任务已经完成,或者已经被取消过了,或者因为别的原因不能取消, 则返回false。 当取消时,该任务还没有开始执行,则该任务不会执行,并且总是返回true。
FutureTask:
public class FutureTask<V> implements RunnableFuture<V>
FutureTask类实现了RunnableFuture接口,看一下RunnableFuture接口的定义:
public interface RunnableFuture<V> extends Runnable, Future<V>
RunnableFuture接口接触了Runnable接口和Future接口, 而FutureTask实现了RunnableFuture接口,所以它既可作为Runnable被线程执行,也可以作为Future得到Callable的返回值。
构造器定义:
public FutureTask(Callable<V> callable)
public FutureTask(Runnable runnable, V result) {
再来看看第二个构造器中的参数怎么变身Callable的:
this.callable = Executors.callable(runnable, result);
调用Executors.callable方法:
public static <T> Callable<T> callable(Runnable task, T result) {
if (task == null)
throw new NullPointerException();
return new RunnableAdapter<T>(task, result);
}
简单实现Callable:
static final 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;
}
}
流程:
下面结合完整具体流程走一下FutureTask过程,并解析源码,草图如下:
实例代码如下:
public class Test {
public static void main(String[] args) {
//第一种方式
ExecutorService executor = Executors.newCachedThreadPool();
Task task = new Task();
FutureTask<Integer> futureTask = new FutureTask<Integer>(task);
executor.submit(futureTask);
executor.shutdown(); try {
Thread.sleep(1000);
} catch (InterruptedException e1) {
e1.printStackTrace();
} System.out.println("主线程在执行任务"); try {
System.out.println("task运行结果"+futureTask.get());
} catch (InterruptedException e) {
e.printStackTrace();
} catch (ExecutionException e) {
e.printStackTrace();
} System.out.println("所有任务执行完毕");
}
}
class Task implements Callable<Integer>{
@Override
public Integer call() throws Exception {
System.out.println("子线程在进行计算");
Thread.sleep(3000);
int sum = 0;
for(int i=0;i<100;i++)
sum += i;
return sum;
}
}
分析过程之前,先准备前准备知识,首先看一下FutureTask内部状态,以及之间的转变:
private volatile int state; // volatile 内存可见性
private static final int NEW = 0; //该状态为new FutureTask()时设定,同时也表示内部成员callable已经成功赋值,一直到worker thread完成FutureTask中run().
private static final int COMPLETING = 1; //该状态位worker thread完成task时设定的中间状态,处于该状态下,说明worker thread 真正准备设置result.
private static final int NORMAL = 2; //当设置result结果完成后,FutureTask处于该状态,代表过程结果,该状态为最终状态final state,(正确完成的最终状态)
private static final int EXCEPTIONAL = 3; // 同上,只不过task执行过程出现异常,此时结果设值为exception,也是final state
private static final int CANCELLED = 4; //final state, 表明task被cancel(task还没有执行就被cancel的状态).
private static final int INTERRUPTING = 5; // 中间状态,task运行过程中被interrupt时,设置的中间状态;
private static final int INTERRUPTED = 6; // final state, 中断完毕的最终状态,几种情况,下面具体分析。
下面是状态之间的转变,贯穿主线:
* Possible state transitions:
1* NEW -> COMPLETING -> NORMAL
2* NEW -> COMPLETING -> EXCEPTIONAL
3* NEW -> CANCELLED
4* NEW -> INTERRUPTING -> INTERRUPTED
*/
其他重要的变量:
/** The underlying callable; nulled out after running */
private Callable<V> callable; // 具体run运行时会调用其方法call(),并获得结果,结果时置为null.
/** The result to return or exception to throw from get() */
private Object outcome; // non-volatile, protected by state reads/writes 没必要为votaile,因为其是伴随state 进行读写,而state是FutureTask的主导因素。
/** The thread running the callable; CASed during run() */
private volatile Thread runner; //具体的worker thread.
/** Treiber stack of waiting threads */
private volatile WaitNode waiters; //Treiber stack 并发stack数据结构,用于存放阻塞在该futuretask#get方法的线程。
OK,构造new FutureTask开始:
public FutureTask(Callable<V> callable) {
if (callable == null)
throw new NullPointerException();
this.callable = callable; //底层callable赋值
this.state = NEW; // 初始状态NEW,同时也标志了callable的赋值,可见性
}
ThreadPoolExecutor.submit(Runnable),ThreadPoolExecutor里面具体细节请见这里,这里就假设它直接new a thread来处理该任务了,因为FutureTask为Runnable的子类,所以worker thread调用该类的run()方法:
public void run() {
if (state != NEW ||
!UNSAFE.compareAndSwapObject(this, runnerOffset,
null, Thread.currentThread())) //状态检测,和当前worker Thread的cas原子赋值,有一个不成立,就直接返回。什么情况下还没run()呢?就不是NEW状态了呢?
return; //caller调用cancel了,此时状态为Interrupting,也说明了上面的cancel方法说明,task没运行时,就interrupt,task得不到运行。总是返回
try { //true。
//再来看看这里worker thread赋值为什么要用cas操作,有竞争racing? 竞争哪里来?难道threadPoolExecutor线程池多个线程可能抢同一个
Callable<V> c = callable; //任务?不可能 1:线程数 < coreThreadPool 时, 直接new thread, 2 : 大于 coreThreadpool时,放在blockingqueue里,取的话只能一
if (c != null && state == NEW) { //线程。能想到就是caller那边了,即多callers(多线程)提交同一FutureTask.
V result; //多线程同时提交同一FutureTask,确保该FutureTask的run()只被调用一次,
boolean ran;
try {
result = c.call(); //此处的if,1:当state == NEW(task没完成,中断) 并且 worker Thread为null时,才会得到运行
ran = true; // 2: task已经完成了 或者 该任务已经有worker thread来执行时,直接返回不会运行。
} catch (Throwable ex) { //调用callable的call方法
result = null; //执行task时有异常
ran = false; //附异常
setException(ex);
}
if (ran) //正常完成,则赋值
set(result);
}
} finally {
//注意!!什么这里吧runner置为null,此时run()方法还没运行完呢啊!现在置为null,不怕并发调用run()吗?注意此时state已经变化了(Comple
runner = null; //teing或者interrupting了,run()一开始state != NEW 直接return,不会运行。可以说通过state和 worker thread来一起控制并发调用run
int s = state; //必须再读一次,防止worker thread == null后,遗漏的interrup信号,底下具体分析中断的情况。
if (s >= INTERRUPTING)
handlePossibleCancellationInterrupt(s); //如果caller中断信号有的话,则处理该interrupt.
} //另外该任务是一致性任务,即state只要不为NEW,该任务就不会在运行,运行结束或cancel后,就不能在运行了,因为state状态在那不变哦!
}
请看下例子,三个提交线程(提交同一个FutureTask):
public class Test {
public static void main(String[] args) {
ExecutorService executor = Executors.newCachedThreadPool();
Task task = new Task();
CountDownLatch latch = new CountDownLatch(1);
FutureTask<Integer> futureTask = new FutureTask<Integer>(task);
for (int i = 0 ; i < 3; i++) {
new Thread(new Submit(executor, futureTask, latch)).start();
}
try {
Thread.sleep(3000);
latch.countDown();
Thread.sleep(20000);
} catch (InterruptedException e1) {
e1.printStackTrace();
}
System.out.println("所有任务执行完毕");
executor.shutdown();
} } class Submit implements Runnable {
private CountDownLatch latch ;
private ExecutorService es ;
private FutureTask<Integer> task;
public Submit(ExecutorService es, FutureTask<Integer> task, CountDownLatch latch) {
this.latch = latch;
this.es = es;
this.task = task;
}
public void run() { try {
latch.await();
Future<?> future = (Future<?>) es.submit(task);
System.out.println("Thread name : " + Thread.currentThread().getName() + "go!");
future.get(3000, TimeUnit.MILLISECONDS);
} catch (InterruptedException e) {
e.printStackTrace();
} catch (ExecutionException e1) {
e1.printStackTrace();
} catch (TimeoutException e2) {
System.err.println("Thread name : " + Thread.currentThread().getName() + " " + e2);
}
}
} class Task implements Callable<Integer>{ public Integer call() throws Exception {
System.out.println("thread name : " + Thread.currentThread().getName() + "do the work!");
Thread.sleep(6000);
int sum = 0;
for(int i=0;i<100;i++)
sum += i;
return sum;
}
}
显示如下:
Thread name : Thread-1go!
Thread name : Thread-0go!
Thread name : Thread-2go!
thread name : pool-1-thread-1do the work!
Thread name : Thread-1 java.util.concurrent.TimeoutException
所有任务执行完毕
结果很显然,同一个任务多次提交(并发提交),FutureTask保证只是启一个线程来运行。
想运行多次(只要不cancel,和throw exception,因为他set(result),正常运行结束,state还是new),用这个:
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 = null;
s = state;
if (s >= INTERRUPTING)
handlePossibleCancellationInterrupt(s);
}
return ran && s == NEW;
}
再来看看setException()和set(result):
protected void set(V v) {
if (UNSAFE.compareAndSwapInt(this, stateOffset, NEW, COMPLETING)) { // cas原子操作,失败直接返回,成功的前提之前的状态必须为NEW.
outcome = v; //可能和什么冲突呢? caller已经cancel该task,状态位Interrupting或者Interrpted(这次Interrupted代表interrupt完成,这set()
UNSAFE.putOrderedInt(this, stateOffset, NORMAL); // final state // 不是在worker thread中调用的嘛,怎么intterupt都完成了,怎么worker thread还在运行呢?worker thread运行的代码中没有响
finishCompletion(); //应interrupt的代码。所以客户端cancel操作,对运行中的worker thread,并不一定让它停下来,不过此时即使运行完毕,也不能赋值。
}
} //new -> Completing-> NORMAL 或者NEW ->Interrupting->Intterrpted.
protected void setException(Throwable t) {
if (UNSAFE.compareAndSwapInt(this, stateOffset, NEW, COMPLETING)) {
outcome = t; //同上,不过附异常。
UNSAFE.putOrderedInt(this, stateOffset, EXCEPTIONAL); // final state
finishCompletion();
} //new ->completing ->exception 或者 同上
}
finishCompletion()等会细聊,主要是没说到get()阻塞呢!看看caller端线程调用cancel()和workerThread的handlePossibleCancellationInterrupt(int s)协调:
public boolean cancel(boolean mayInterruptIfRunning) {
if (state != NEW)
return false; //1:已经cancel(cancelled,Interruping, Interrupted)过了 2:正常完成 Completing(Completed) 3:异常完成completing(exception) 直接返回false;
if (mayInterruptIfRunning) { // flag : worker thread 已经启动运行了,是否可以中断
if (!UNSAFE.compareAndSwapInt(this, stateOffset, NEW, INTERRUPTING)) //再次检查state状态,完成的话(上面的三种),直接返回false;
return false;
Thread t = runner;
if (t != null) // t == null对应Future task还没启动, 跳过thread.interrupt(),直接由interrpting -> interrupted,成功的话
t.interrupt(); //调用worker thread的 interrupt() //mayInterrptIfRunning 为true ,interrupt 状态转变 new -> interrupting -> interrupted.
UNSAFE.putOrderedInt(this, stateOffset, INTERRUPTED); // final state
}
else if (!UNSAFE.compareAndSwapInt(this, stateOffset, NEW, CANCELLED)) //mayInterruptIfRunning 为false,interrupt成功的 状态转变 new -> Cancelled
return false;
finishCompletion();
return true;
}
由上面可知,客户端cancel()中不少cas操作,主要来自两方面的racing, 1:线程池worker Thread的完成(异常,正常)状态设置; 2:同一futuretask,不同客户端线程callers的cancel操作。
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();
}
当state处于Interrupting, 即caller即将调用worker thread.interrupt(), 所以worker thread自旋会,等会interrupt方法的调用,保留interrupt标志。
再来看看get()和带参数的get(timeout):
public V get() throws InterruptedException, ExecutionException {
int s = state;
if (s <= COMPLETING) //结果未设定的情况下
s = awaitDone(false, 0L); //无条件等待
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) //等到timeout时间内,没完成,throws TimeoutException
throw new TimeoutException();
return report(s);
}
awaitDone():
private int awaitDone(boolean timed, long nanos)
throws InterruptedException {
final long deadline = timed ? System.nanoTime() + nanos : 0L; //计算出等待时间。
WaitNode q = null;
boolean queued = false; //是否加入阻塞队列的标志
for (;;) {
if (Thread.interrupted()) { //阻塞该caller线程之前,caller线程被中断,直接throw 异常
removeWaiter(q); //在阻塞队列中移除该线程的封装node.此处无意义
throw new InterruptedException();
} int s = state; //读取state,阻塞前 recheck一下 是否完成?
if (s > COMPLETING) { //完成了,直接返回,不要阻塞了
if (q != null)
q.thread = null;
return s;
}
else if (s == COMPLETING) // cannot time out yet
Thread.yield(); //等会,快完成了。
else if (q == null)
q = new WaitNode(); //当前阻塞线程链表的简单封装
else if (!queued)
queued = UNSAFE.compareAndSwapObject(this, waitersOffset,
q.next = waiters, q); //设为当前FutureTask阻塞链表(stack结构)的栈顶。
else if (timed) {
nanos = deadline - System.nanoTime(); //计算当前要阻塞的等待时间
if (nanos <= 0L) {
removeWaiter(q); //小于0 直接返回,当前REMOVEWaiter无意义,并没有加入stack中。
return state;
}
LockSupport.parkNanos(this, nanos);本地native方法,阻塞当前线程。
}
else
LockSupport.park(this); //无时间条件阻塞
}
}
无时间限制阻塞,有时间阻塞(阻塞时间大于task完成时间)会等到任务完成而给通知,唤醒该线程,即finishCompletion();而有时间阻塞(阻塞时间在task完成之间就已经结束的)会通过for()退出(退出前,删除等待队列中的节点)。
WaiterNode定义:
static final class WaitNode {
volatile Thread thread;
volatile WaitNode next;
WaitNode() { thread = Thread.currentThread(); } //当前阻塞线程的引用
}
结合awaitDone()中的新阻塞节点加入顺序,其定位stack结构(Treiber stack);
removeWaiter():
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) // 检测竞争
continue retry; //发生重试
}
else if (!UNSAFE.compareAndSwapObject(this, waitersOffset,
q, s))
continue retry;
}
break;
}
}
}
finishCompletion():
private void finishCompletion() {
// assert state > COMPLETING;
for (WaitNode q; (q = waiters) != null;) { 得到栈顶阻塞节点的引用
if (UNSAFE.compareAndSwapObject(this, waitersOffset, q, null)) {
for (;;) {
Thread t = q.thread; //取得阻塞的线程引用,
if (t != null) {
q.thread = null;
LockSupport.unpark(t);//解阻塞
}
WaitNode next = q.next; //遍历解阻塞线程
if (next == null)
break;
q.next = null; // unlink to help gc
q = next;
}
break;
}
} done(); callable = null; // to reduce footprint
}
其实,前面的分析可知,多个caller线程并发提交同一个FutureTask, 并且所谓调用get()阻塞的话(阻塞在该FutureTask上),实际上也就一个caller线程阻塞,其他线程在调用该FutureTask的run()开始条件检查时,就直接return了,实际情况:三个并发线程提交同一个future task,对应生成三份FutureTask(不同于之前),三份FutureTask中对应三分Callable,而这三份Callable含有相同的FutureTask(所谓的相同任务) ,向ThreadPoolExecutor.submit(Runnable)实际上提交了三份Runnable(即生成的三分FutureTask), FutureTask实现了Runnable接口, 然后ThreadPoolExecutor生成三个线程来执行这所谓的三个任务,这三个任务run()中都是调用对应内部的callable的call(), 而callable的call方法调用的是他们共同引用的FutureTask(同一个对像)的run()方法,而run方法, 我们上面解析过了,通过cas和状态检测,只运行一个worker thread 调用run()(见上),另外两个线程直接从共同底层FutureTask的run方法开始直接返回。
晕了?从头再来看看提交的过程:
1:submit(FutureTask(Runnable)):AbstractExecutorService
public Future<?> submit(Runnable task) {
if (task == null) throw new NullPointerException();
RunnableFuture<Void> ftask = newTaskFor(task, null);
execute(ftask);
return ftask;
}
protected <T> RunnableFuture<T> newTaskFor(Runnable runnable, T value) {
return new FutureTask<T>(runnable, value);
}
2:生成三个FutureTask(其中runnable就是同一个底层FutureTask任务):
public FutureTask(Runnable runnable, V result) {
this.callable = Executors.callable(runnable, result);
this.state = NEW; // ensure visibility of callable
}
3:调用Executors.callable():
public static <T> Callable<T> callable(Runnable task, T result) {
if (task == null)
throw new NullPointerException();
return new RunnableAdapter<T>(task, result);
}
static final 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() { //直接调用底层同一个FutureTask的run();
task.run();
return result;
}
}
即三次提交,生成三份FutureTask,每份FutureTask调用Executors.callable()为自己底层的callable赋值,而Executors.callable方法生成简单的Callable实现,其中call(),调用底层共同FutureTask的run(), 也就受共同futureTask内部状态(state, runThread)限制。所以,阻塞在底层共同FutureTask阻塞队列中的只有一个线程,看下例:
public class Test {
public static void main(String[] args) {
ExecutorService executor = Executors.newCachedThreadPool();
Task task = new Task();
int waitTime = 4000;
CountDownLatch latch = new CountDownLatch(1);
FutureTask<Integer> futureTask = new FutureTask<Integer>(task);
for (int i = 0 ; i < 3; i++) {
new Thread(new Submit(executor, futureTask, latch, waitTime)).start();
}
try {
Thread.sleep(3000);
latch.countDown();
Thread.sleep(8000);
} catch (InterruptedException e1) {
e1.printStackTrace();
}
System.out.println("所有任务执行完毕");
executor.shutdown();
} } class Submit implements Runnable {
private CountDownLatch latch ;
private ExecutorService es ;
private FutureTask<Integer> task;
private int waitTime ;
public Submit(ExecutorService es, FutureTask<Integer> task, CountDownLatch latch, int waitTime) {
this.latch = latch;
this.es = es;
this.task = task;
this.waitTime = waitTime;
}
public void run() {
try {
latch.await();
Future<?> future = es.submit(task);
System.out.println("Thread name : " + Thread.currentThread().getName() + " go!");
waitTime = new Random().nextInt(waitTime);
System.out.println("Thread name : " + Thread.currentThread().getName() + " , The wait time : = " + waitTime );
future.get(waitTime, TimeUnit.MILLISECONDS);
System.out.println("Thread name : " + Thread.currentThread().getName() + " run over!");
} catch (InterruptedException e) {
e.printStackTrace();
} catch (ExecutionException e1) {
e1.printStackTrace();
} catch (TimeoutException e2) {
System.err.println("Thread name : " + Thread.currentThread().getName() + " " + e2);
}
}
} class Task implements Callable<Integer>{ public Integer call() throws Exception {
System.out.println("thread name : " + Thread.currentThread().getName() + " do the work!");
Thread.sleep(4000);
int sum = 0;
for(int i=0;i<20;i++)
sum += i;
return sum;
}
} class Task1 implements Runnable{
int sum = 0;
@Override
public void run() {
System.out.println("Thread Name : " + Thread.currentThread().getName() + "do the work!");
try {
Thread.sleep(6000); for(int i=0;i<100;i++)
sum += i;
} catch (InterruptedException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
}
}
显示结果:
Thread name : Thread-2 go!
Thread name : Thread-0 go!
Thread name : Thread-0 , The wait time : = 2738
thread name : pool-1-thread-1 do the work!
Thread name : Thread-1 go!
Thread name : Thread-2 , The wait time : = 284
Thread name : Thread-1 , The wait time : = 678
Thread name : Thread-2 run over!
Thread name : Thread-0 run over!
Thread name : Thread-1 java.util.concurrent.TimeoutException
所有任务执行完毕
三个线程都是阻塞一段时间,但是只有一个超时,另外两个运行完毕,(他两实际工作那部分没运行,处理各自FutureTask那部分代码,所以只能看到线程池只有一个线程处理底层FutureTask);
但,如果直接并发提交Callable,或者Runnable,线程池会启动三个线程来分别处理这三个不同任务,朋友可以自行试验demo下。而FutureTask是自己的自身的限制。
后话,一般调用ThreadPoolExecutor.submit()提交的是Callable<T>和Runnable, 返回的Future<T>, Future<?>(返回Null,或者不要求返回值),提交FutureTask用不着,所以实际中不会遇见这种情况。
另外,本文源码基于jdk1.7,与网上1.7之前源码不同(1.6通过AQS实现)。