Java并发包--线程池原理
转载请注明出处：http://www.cnblogs.com/skywang12345/p/3509954.html
线程池示例
在分析线程池之前，先看一个简单的线程池示例。
1 import java.util.concurrent.Executors;
2 import java.util.concurrent.ExecutorService;
3
4 public class ThreadPoolDemo1 {
5
6 public static void main(String[] args) {
7 // 创建一个可重用固定线程数的线程池
8 ExecutorService pool = Executors.newFixedThreadPool(2);
9 // 创建实现了Runnable接口对象，Thread对象当然也实现了Runnable接口
10 Thread ta = new MyThread();
11 Thread tb = new MyThread();
12 Thread tc = new MyThread();
13 Thread td = new MyThread();
14 Thread te = new MyThread();
15 // 将线程放入池中进行执行
16 pool.execute(ta);
17 pool.execute(tb);
18 pool.execute(tc);
19 pool.execute(td);
20 pool.execute(te);
21 // 关闭线程池
22 pool.shutdown();
23 }
24 }
25
26 class MyThread extends Thread {
27
28 @Override
29 public void run() {
30 System.out.println(Thread.currentThread().getName()+ " is running.");
31 }
32 }
运行结果：
pool-1-thread-1 is running.
pool-1-thread-2 is running.
pool-1-thread-1 is running.
pool-1-thread-2 is running.
pool-1-thread-1 is running.
示例中，包括了线程池的创建，将任务添加到线程池中，关闭线程池这3个主要的步骤。稍后，我们会从这3个方面来分析ThreadPoolExecutor。
参考代码(基于JDK1.7.0_40)
Executors完整源码
1 /*
2 * ORACLE PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
3 *
4 *
5 *
6 *
7 *
8 *
9 *
10 *
11 *
12 *
13 *
14 *
15 *
16 *
17 *
18 *
19 *
20 *
21 *
22 *
23 */
24
25 /*
26 *
27 *
28 *
29 *
30 *
31 * Written by Doug Lea with assistance from members of JCP JSR-166
32 * Expert Group and released to the public domain, as explained at
33 * http://creativecommons.org/publicdomain/zero/1.0/
34 */
35
36 package java.util.concurrent;
37 import java.util.*;
38 import java.util.concurrent.atomic.AtomicInteger;
39 import java.security.AccessControlContext;
40 import java.security.AccessController;
41 import java.security.PrivilegedAction;
42 import java.security.PrivilegedExceptionAction;
43 import java.security.PrivilegedActionException;
44 import java.security.AccessControlException;
45 import sun.security.util.SecurityConstants;
46
47 /**
48 * Factory and utility methods for {@link Executor}, {@link
49 * ExecutorService}, {@link ScheduledExecutorService}, {@link
50 * ThreadFactory}, and {@link Callable} classes defined in this
51 * package. This class supports the following kinds of methods:
52 *
53 * <ul>
54 * <li> Methods that create and return an {@link ExecutorService}
55 * set up with commonly useful configuration settings.
56 * <li> Methods that create and return a {@link ScheduledExecutorService}
57 * set up with commonly useful configuration settings.
58 * <li> Methods that create and return a "wrapped" ExecutorService, that
59 * disables reconfiguration by making implementation-specific methods
60 * inaccessible.
61 * <li> Methods that create and return a {@link ThreadFactory}
62 * that sets newly created threads to a known state.
63 * <li> Methods that create and return a {@link Callable}
64 * out of other closure-like forms, so they can be used
65 * in execution methods requiring <tt>Callable</tt>.
66 * </ul>
67 *
68 * @since 1.5
69 * @author Doug Lea
70 */
71 public class Executors {
72
73 /**
74 * Creates a thread pool that reuses a fixed number of threads
75 * operating off a shared unbounded queue. At any point, at most
76 * <tt>nThreads</tt> threads will be active processing tasks.
77 * If additional tasks are submitted when all threads are active,
78 * they will wait in the queue until a thread is available.
79 * If any thread terminates due to a failure during execution
80 * prior to shutdown, a new one will take its place if needed to
81 * execute subsequent tasks. The threads in the pool will exist
82 * until it is explicitly {@link ExecutorService#shutdown shutdown}.
83 *
84 * @param nThreads the number of threads in the pool
85 * @return the newly created thread pool
86 * @throws IllegalArgumentException if {@code nThreads <= 0}
87 */
88 public static ExecutorService newFixedThreadPool(int nThreads) {
89 return new ThreadPoolExecutor(nThreads, nThreads,
90 0L, TimeUnit.MILLISECONDS,
91 new LinkedBlockingQueue<Runnable>());
92 }
93
94 /**
95 * Creates a thread pool that reuses a fixed number of threads
96 * operating off a shared unbounded queue, using the provided
97 * ThreadFactory to create new threads when needed. At any point,
98 * at most <tt>nThreads</tt> threads will be active processing
99 * tasks. If additional tasks are submitted when all threads are
100 * active, they will wait in the queue until a thread is
101 * available. If any thread terminates due to a failure during
102 * execution prior to shutdown, a new one will take its place if
103 * needed to execute subsequent tasks. The threads in the pool will
104 * exist until it is explicitly {@link ExecutorService#shutdown
105 * shutdown}.
106 *
107 * @param nThreads the number of threads in the pool
108 * @param threadFactory the factory to use when creating new threads
109 * @return the newly created thread pool
110 * @throws NullPointerException if threadFactory is null
111 * @throws IllegalArgumentException if {@code nThreads <= 0}
112 */
113 public static ExecutorService newFixedThreadPool(int nThreads, ThreadFactory threadFactory) {
114 return new ThreadPoolExecutor(nThreads, nThreads,
115 0L, TimeUnit.MILLISECONDS,
116 new LinkedBlockingQueue<Runnable>(),
117 threadFactory);
118 }
119
120 /**
121 * Creates an Executor that uses a single worker thread operating
122 * off an unbounded queue. (Note however that if this single
123 * thread terminates due to a failure during execution prior to
124 * shutdown, a new one will take its place if needed to execute
125 * subsequent tasks.) Tasks are guaranteed to execute
126 * sequentially, and no more than one task will be active at any
127 * given time. Unlike the otherwise equivalent
128 * <tt>newFixedThreadPool(1)</tt> the returned executor is
129 * guaranteed not to be reconfigurable to use additional threads.
130 *
131 * @return the newly created single-threaded Executor
132 */
133 public static ExecutorService newSingleThreadExecutor() {
134 return new FinalizableDelegatedExecutorService
135 (new ThreadPoolExecutor(1, 1,
136 0L, TimeUnit.MILLISECONDS,
137 new LinkedBlockingQueue<Runnable>()));
138 }
139
140 /**
141 * Creates an Executor that uses a single worker thread operating
142 * off an unbounded queue, and uses the provided ThreadFactory to
143 * create a new thread when needed. Unlike the otherwise
144 * equivalent <tt>newFixedThreadPool(1, threadFactory)</tt> the
145 * returned executor is guaranteed not to be reconfigurable to use
146 * additional threads.
147 *
148 * @param threadFactory the factory to use when creating new
149 * threads
150 *
151 * @return the newly created single-threaded Executor
152 * @throws NullPointerException if threadFactory is null
153 */
154 public static ExecutorService newSingleThreadExecutor(ThreadFactory threadFactory) {
155 return new FinalizableDelegatedExecutorService
156 (new ThreadPoolExecutor(1, 1,
157 0L, TimeUnit.MILLISECONDS,
158 new LinkedBlockingQueue<Runnable>(),
159 threadFactory));
160 }
161
162 /**
163 * Creates a thread pool that creates new threads as needed, but
164 * will reuse previously constructed threads when they are
165 * available. These pools will typically improve the performance
166 * of programs that execute many short-lived asynchronous tasks.
167 * Calls to <tt>execute</tt> will reuse previously constructed
168 * threads if available. If no existing thread is available, a new
169 * thread will be created and added to the pool. Threads that have
170 * not been used for sixty seconds are terminated and removed from
171 * the cache. Thus, a pool that remains idle for long enough will
172 * not consume any resources. Note that pools with similar
173 * properties but different details (for example, timeout parameters)
174 * may be created using {@link ThreadPoolExecutor} constructors.
175 *
176 * @return the newly created thread pool
177 */
178 public static ExecutorService newCachedThreadPool() {
179 return new ThreadPoolExecutor(0, Integer.MAX_VALUE,
180 60L, TimeUnit.SECONDS,
181 new SynchronousQueue<Runnable>());
182 }
183
184 /**
185 * Creates a thread pool that creates new threads as needed, but
186 * will reuse previously constructed threads when they are
187 * available, and uses the provided
188 * ThreadFactory to create new threads when needed.
189 * @param threadFactory the factory to use when creating new threads
190 * @return the newly created thread pool
191 * @throws NullPointerException if threadFactory is null
192 */
193 public static ExecutorService newCachedThreadPool(ThreadFactory threadFactory) {
194 return new ThreadPoolExecutor(0, Integer.MAX_VALUE,
195 60L, TimeUnit.SECONDS,
196 new SynchronousQueue<Runnable>(),
197 threadFactory);
198 }
199
200 /**
201 * Creates a single-threaded executor that can schedule commands
202 * to run after a given delay, or to execute periodically.
203 * (Note however that if this single
204 * thread terminates due to a failure during execution prior to
205 * shutdown, a new one will take its place if needed to execute
206 * subsequent tasks.) Tasks are guaranteed to execute
207 * sequentially, and no more than one task will be active at any
208 * given time. Unlike the otherwise equivalent
209 * <tt>newScheduledThreadPool(1)</tt> the returned executor is
210 * guaranteed not to be reconfigurable to use additional threads.
211 * @return the newly created scheduled executor
212 */
213 public static ScheduledExecutorService newSingleThreadScheduledExecutor() {
214 return new DelegatedScheduledExecutorService
215 (new ScheduledThreadPoolExecutor(1));
216 }
217
218 /**
219 * Creates a single-threaded executor that can schedule commands
220 * to run after a given delay, or to execute periodically. (Note
221 * however that if this single thread terminates due to a failure
222 * during execution prior to shutdown, a new one will take its
223 * place if needed to execute subsequent tasks.) Tasks are
224 * guaranteed to execute sequentially, and no more than one task
225 * will be active at any given time. Unlike the otherwise
226 * equivalent <tt>newScheduledThreadPool(1, threadFactory)</tt>
227 * the returned executor is guaranteed not to be reconfigurable to
228 * use additional threads.
229 * @param threadFactory the factory to use when creating new
230 * threads
231 * @return a newly created scheduled executor
232 * @throws NullPointerException if threadFactory is null
233 */
234 public static ScheduledExecutorService newSingleThreadScheduledExecutor(ThreadFactory threadFactory) {
235 return new DelegatedScheduledExecutorService
236 (new ScheduledThreadPoolExecutor(1, threadFactory));
237 }
238
239 /**
240 * Creates a thread pool that can schedule commands to run after a
241 * given delay, or to execute periodically.
242 * @param corePoolSize the number of threads to keep in the pool,
243 * even if they are idle.
244 * @return a newly created scheduled thread pool
245 * @throws IllegalArgumentException if {@code corePoolSize < 0}
246 */
247 public static ScheduledExecutorService newScheduledThreadPool(int corePoolSize) {
248 return new ScheduledThreadPoolExecutor(corePoolSize);
249 }
250
251 /**
252 * Creates a thread pool that can schedule commands to run after a
253 * given delay, or to execute periodically.
254 * @param corePoolSize the number of threads to keep in the pool,
255 * even if they are idle.
256 * @param threadFactory the factory to use when the executor
257 * creates a new thread.
258 * @return a newly created scheduled thread pool
259 * @throws IllegalArgumentException if {@code corePoolSize < 0}
260 * @throws NullPointerException if threadFactory is null
261 */
262 public static ScheduledExecutorService newScheduledThreadPool(
263 int corePoolSize, ThreadFactory threadFactory) {
264 return new ScheduledThreadPoolExecutor(corePoolSize, threadFactory);
265 }
266
267
268 /**
269 * Returns an object that delegates all defined {@link
270 * ExecutorService} methods to the given executor, but not any
271 * other methods that might otherwise be accessible using
272 * casts. This provides a way to safely "freeze" configuration and
273 * disallow tuning of a given concrete implementation.
274 * @param executor the underlying implementation
275 * @return an <tt>ExecutorService</tt> instance
276 * @throws NullPointerException if executor null
277 */
278 public static ExecutorService unconfigurableExecutorService(ExecutorService executor) {
279 if (executor == null)
280 throw new NullPointerException();
281 return new DelegatedExecutorService(executor);
282 }
283
284 /**
285 * Returns an object that delegates all defined {@link
286 * ScheduledExecutorService} methods to the given executor, but
287 * not any other methods that might otherwise be accessible using
288 * casts. This provides a way to safely "freeze" configuration and
289 * disallow tuning of a given concrete implementation.
290 * @param executor the underlying implementation
291 * @return a <tt>ScheduledExecutorService</tt> instance
292 * @throws NullPointerException if executor null
293 */
294 public static ScheduledExecutorService unconfigurableScheduledExecutorService(ScheduledExecutorService executor) {
295 if (executor == null)
296 throw new NullPointerException();
297 return new DelegatedScheduledExecutorService(executor);
298 }
299
300 /**
301 * Returns a default thread factory used to create new threads.
302 * This factory creates all new threads used by an Executor in the
303 * same {@link ThreadGroup}. If there is a {@link
304 * java.lang.SecurityManager}, it uses the group of {@link
305 * System#getSecurityManager}, else the group of the thread
306 * invoking this <tt>defaultThreadFactory</tt> method. Each new
307 * thread is created as a non-daemon thread with priority set to
308 * the smaller of <tt>Thread.NORM_PRIORITY</tt> and the maximum
309 * priority permitted in the thread group. New threads have names
310 * accessible via {@link Thread#getName} of
311 * <em>pool-N-thread-M</em>, where <em>N</em> is the sequence
312 * number of this factory, and <em>M</em> is the sequence number
313 * of the thread created by this factory.
314 * @return a thread factory
315 */
316 public static ThreadFactory defaultThreadFactory() {
317 return new DefaultThreadFactory();
318 }
319
320 /**
321 * Returns a thread factory used to create new threads that
322 * have the same permissions as the current thread.
323 * This factory creates threads with the same settings as {@link
324 * Executors#defaultThreadFactory}, additionally setting the
325 * AccessControlContext and contextClassLoader of new threads to
326 * be the same as the thread invoking this
327 * <tt>privilegedThreadFactory</tt> method. A new
328 * <tt>privilegedThreadFactory</tt> can be created within an
329 * {@link AccessController#doPrivileged} action setting the
330 * current thread's access control context to create threads with
331 * the selected permission settings holding within that action.
332 *
333 * <p> Note that while tasks running within such threads will have
334 * the same access control and class loader settings as the
335 * current thread, they need not have the same {@link
336 * java.lang.ThreadLocal} or {@link
337 * java.lang.InheritableThreadLocal} values. If necessary,
338 * particular values of thread locals can be set or reset before
339 * any task runs in {@link ThreadPoolExecutor} subclasses using
340 * {@link ThreadPoolExecutor#beforeExecute}. Also, if it is
341 * necessary to initialize worker threads to have the same
342 * InheritableThreadLocal settings as some other designated
343 * thread, you can create a custom ThreadFactory in which that
344 * thread waits for and services requests to create others that
345 * will inherit its values.
346 *
347 * @return a thread factory
348 * @throws AccessControlException if the current access control
349 * context does not have permission to both get and set context
350 * class loader.
351 */
352 public static ThreadFactory privilegedThreadFactory() {
353 return new PrivilegedThreadFactory();
354 }
355
356 /**
357 * Returns a {@link Callable} object that, when
358 * called, runs the given task and returns the given result. This
359 * can be useful when applying methods requiring a
360 * <tt>Callable</tt> to an otherwise resultless action.
361 * @param task the task to run
362 * @param result the result to return
363 * @return a callable object
364 * @throws NullPointerException if task null
365 */
366 public static <T> Callable<T> callable(Runnable task, T result) {
367 if (task == null)
368 throw new NullPointerException();
369 return new RunnableAdapter<T>(task, result);
370 }
371
372 /**
373 * Returns a {@link Callable} object that, when
374 * called, runs the given task and returns <tt>null</tt>.
375 * @param task the task to run
376 * @return a callable object
377 * @throws NullPointerException if task null
378 */
379 public static Callable<Object> callable(Runnable task) {
380 if (task == null)
381 throw new NullPointerException();
382 return new RunnableAdapter<Object>(task, null);
383 }
384
385 /**
386 * Returns a {@link Callable} object that, when
387 * called, runs the given privileged action and returns its result.
388 * @param action the privileged action to run
389 * @return a callable object
390 * @throws NullPointerException if action null
391 */
392 public static Callable<Object> callable(final PrivilegedAction<?> action) {
393 if (action == null)
394 throw new NullPointerException();
395 return new Callable<Object>() {
396 public Object call() { return action.run(); }};
397 }
398
399 /**
400 * Returns a {@link Callable} object that, when
401 * called, runs the given privileged exception action and returns
402 * its result.
403 * @param action the privileged exception action to run
404 * @return a callable object
405 * @throws NullPointerException if action null
406 */
407 public static Callable<Object> callable(final PrivilegedExceptionAction<?> action) {
408 if (action == null)
409 throw new NullPointerException();
410 return new Callable<Object>() {
411 public Object call() throws Exception { return action.run(); }};
412 }
413
414 /**
415 * Returns a {@link Callable} object that will, when
416 * called, execute the given <tt>callable</tt> under the current
417 * access control context. This method should normally be
418 * invoked within an {@link AccessController#doPrivileged} action
419 * to create callables that will, if possible, execute under the
420 * selected permission settings holding within that action; or if
421 * not possible, throw an associated {@link
422 * AccessControlException}.
423 * @param callable the underlying task
424 * @return a callable object
425 * @throws NullPointerException if callable null
426 *
427 */
428 public static <T> Callable<T> privilegedCallable(Callable<T> callable) {
429 if (callable == null)
430 throw new NullPointerException();
431 return new PrivilegedCallable<T>(callable);
432 }
433
434 /**
435 * Returns a {@link Callable} object that will, when
436 * called, execute the given <tt>callable</tt> under the current
437 * access control context, with the current context class loader
438 * as the context class loader. This method should normally be
439 * invoked within an {@link AccessController#doPrivileged} action
440 * to create callables that will, if possible, execute under the
441 * selected permission settings holding within that action; or if
442 * not possible, throw an associated {@link
443 * AccessControlException}.
444 * @param callable the underlying task
445 *
446 * @return a callable object
447 * @throws NullPointerException if callable null
448 * @throws AccessControlException if the current access control
449 * context does not have permission to both set and get context
450 * class loader.
451 */
452 public static <T> Callable<T> privilegedCallableUsingCurrentClassLoader(Callable<T> callable) {
453 if (callable == null)
454 throw new NullPointerException();
455 return new PrivilegedCallableUsingCurrentClassLoader<T>(callable);
456 }
457
458 // Non-public classes supporting the public methods
459
460 /**
461 * A callable that runs given task and returns given result
462 */
463 static final class RunnableAdapter<T> implements Callable<T> {
464 final Runnable task;
465 final T result;
466 RunnableAdapter(Runnable task, T result) {
467 this.task = task;
468 this.result = result;
469 }
470 public T call() {
471 task.run();
472 return result;
473 }
474 }
475
476 /**
477 * A callable that runs under established access control settings
478 */
479 static final class PrivilegedCallable<T> implements Callable<T> {
480 private final Callable<T> task;
481 private final AccessControlContext acc;
482
483 PrivilegedCallable(Callable<T> task) {
484 this.task = task;
485 this.acc = AccessController.getContext();
486 }
487
488 public T call() throws Exception {
489 try {
490 return AccessController.doPrivileged(
491 new PrivilegedExceptionAction<T>() {
492 public T run() throws Exception {
493 return task.call();
494 }
495 }, acc);
496 } catch (PrivilegedActionException e) {
497 throw e.getException();
498 }
499 }
500 }
501
502 /**
503 * A callable that runs under established access control settings and
504 * current ClassLoader
505 */
506 static final class PrivilegedCallableUsingCurrentClassLoader<T> implements Callable<T> {
507 private final Callable<T> task;
508 private final AccessControlContext acc;
509 private final ClassLoader ccl;
510
511 PrivilegedCallableUsingCurrentClassLoader(Callable<T> task) {
512 SecurityManager sm = System.getSecurityManager();
513 if (sm != null) {
514 // Calls to getContextClassLoader from this class
515 // never trigger a security check, but we check
516 // whether our callers have this permission anyways.
517 sm.checkPermission(SecurityConstants.GET_CLASSLOADER_PERMISSION);
518
519 // Whether setContextClassLoader turns out to be necessary
520 // or not, we fail fast if permission is not available.
521 sm.checkPermission(new RuntimePermission("setContextClassLoader"));
522 }
523 this.task = task;
524 this.acc = AccessController.getContext();
525 this.ccl = Thread.currentThread().getContextClassLoader();
526 }
527
528 public T call() throws Exception {
529 try {
530 return AccessController.doPrivileged(
531 new PrivilegedExceptionAction<T>() {
532 public T run() throws Exception {
533 Thread t = Thread.currentThread();
534 ClassLoader cl = t.getContextClassLoader();
535 if (ccl == cl) {
536 return task.call();
537 } else {
538 t.setContextClassLoader(ccl);
539 try {
540 return task.call();
541 } finally {
542 t.setContextClassLoader(cl);
543 }
544 }
545 }
546 }, acc);
547 } catch (PrivilegedActionException e) {
548 throw e.getException();
549 }
550 }
551 }
552
553 /**
554 * The default thread factory
555 */
556 static class DefaultThreadFactory implements ThreadFactory {
557 private static final AtomicInteger poolNumber = new AtomicInteger(1);
558 private final ThreadGroup group;
559 private final AtomicInteger threadNumber = new AtomicInteger(1);
560 private final String namePrefix;
561
562 DefaultThreadFactory() {
563 SecurityManager s = System.getSecurityManager();
564 group = (s != null) ? s.getThreadGroup() :
565 Thread.currentThread().getThreadGroup();
566 namePrefix = "pool-" +
567 poolNumber.getAndIncrement() +
568 "-thread-";
569 }
570
571 public Thread newThread(Runnable r) {
572 Thread t = new Thread(group, r,
573 namePrefix + threadNumber.getAndIncrement(),
574 0);
575 if (t.isDaemon())
576 t.setDaemon(false);
577 if (t.getPriority() != Thread.NORM_PRIORITY)
578 t.setPriority(Thread.NORM_PRIORITY);
579 return t;
580 }
581 }
582
583 /**
584 * Thread factory capturing access control context and class loader
585 */
586 static class PrivilegedThreadFactory extends DefaultThreadFactory {
587 private final AccessControlContext acc;
588 private final ClassLoader ccl;
589
590 PrivilegedThreadFactory() {
591 super();
592 SecurityManager sm = System.getSecurityManager();
593 if (sm != null) {
594 // Calls to getContextClassLoader from this class
595 // never trigger a security check, but we check
596 // whether our callers have this permission anyways.
597 sm.checkPermission(SecurityConstants.GET_CLASSLOADER_PERMISSION);
598
599 // Fail fast
600 sm.checkPermission(new RuntimePermission("setContextClassLoader"));
601 }
602 this.acc = AccessController.getContext();
603 this.ccl = Thread.currentThread().getContextClassLoader();
604 }
605
606 public Thread newThread(final Runnable r) {
607 return super.newThread(new Runnable() {
608 public void run() {
609 AccessController.doPrivileged(new PrivilegedAction<Void>() {
610 public Void run() {
611 Thread.currentThread().setContextClassLoader(ccl);
612 r.run();
613 return null;
614 }
615 }, acc);
616 }
617 });
618 }
619 }
620
621 /**
622 * A wrapper class that exposes only the ExecutorService methods
623 * of an ExecutorService implementation.
624 */
625 static class DelegatedExecutorService extends AbstractExecutorService {
626 private final ExecutorService e;
627 DelegatedExecutorService(ExecutorService executor) { e = executor; }
628 public void execute(Runnable command) { e.execute(command); }
629 public void shutdown() { e.shutdown(); }
630 public List<Runnable> shutdownNow() { return e.shutdownNow(); }
631 public boolean isShutdown() { return e.isShutdown(); }
632 public boolean isTerminated() { return e.isTerminated(); }
633 public boolean awaitTermination(long timeout, TimeUnit unit)
634 throws InterruptedException {
635 return e.awaitTermination(timeout, unit);
636 }
637 public Future<?> submit(Runnable task) {
638 return e.submit(task);
639 }
640 public <T> Future<T> submit(Callable<T> task) {
641 return e.submit(task);
642 }
643 public <T> Future<T> submit(Runnable task, T result) {
644 return e.submit(task, result);
645 }
646 public <T> List<Future<T>> invokeAll(Collection<? extends Callable<T>> tasks)
647 throws InterruptedException {
648 return e.invokeAll(tasks);
649 }
650 public <T> List<Future<T>> invokeAll(Collection<? extends Callable<T>> tasks,
651 long timeout, TimeUnit unit)
652 throws InterruptedException {
653 return e.invokeAll(tasks, timeout, unit);
654 }
655 public <T> T invokeAny(Collection<? extends Callable<T>> tasks)
656 throws InterruptedException, ExecutionException {
657 return e.invokeAny(tasks);
658 }
659 public <T> T invokeAny(Collection<? extends Callable<T>> tasks,
660 long timeout, TimeUnit unit)
661 throws InterruptedException, ExecutionException, TimeoutException {
662 return e.invokeAny(tasks, timeout, unit);
663 }
664 }
665
666 static class FinalizableDelegatedExecutorService
667 extends DelegatedExecutorService {
668 FinalizableDelegatedExecutorService(ExecutorService executor) {
669 super(executor);
670 }
671 protected void finalize() {
672 super.shutdown();
673 }
674 }
675
676 /**
677 * A wrapper class that exposes only the ScheduledExecutorService
678 * methods of a ScheduledExecutorService implementation.
679 */
680 static class DelegatedScheduledExecutorService
681 extends DelegatedExecutorService
682 implements ScheduledExecutorService {
683 private final ScheduledExecutorService e;
684 DelegatedScheduledExecutorService(ScheduledExecutorService executor) {
685 super(executor);
686 e = executor;
687 }
688 public ScheduledFuture<?> schedule(Runnable command, long delay, TimeUnit unit) {
689 return e.schedule(command, delay, unit);
690 }
691 public <V> ScheduledFuture<V> schedule(Callable<V> callable, long delay, TimeUnit unit) {
692 return e.schedule(callable, delay, unit);
693 }
694 public ScheduledFuture<?> scheduleAtFixedRate(Runnable command, long initialDelay, long period, TimeUnit unit) {
695 return e.scheduleAtFixedRate(command, initialDelay, period, unit);
696 }
697 public ScheduledFuture<?> scheduleWithFixedDelay(Runnable command, long initialDelay, long delay, TimeUnit unit) {
698 return e.scheduleWithFixedDelay(command, initialDelay, delay, unit);
699 }
700 }
701
702
703 /** Cannot instantiate. */
704 private Executors() {}
705 }
ThreadPoolExecutor完整源码
1 /*
2 * ORACLE PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
3 *
4 *
5 *
6 *
7 *
8 *
9 *
10 *
11 *
12 *
13 *
14 *
15 *
16 *
17 *
18 *
19 *
20 *
21 *
22 *
23 */
24
25 /*
26 *
27 *
28 *
29 *
30 *
31 * Written by Doug Lea with assistance from members of JCP JSR-166
32 * Expert Group and released to the public domain, as explained at
33 * http://creativecommons.org/publicdomain/zero/1.0/
34 */
35
36 package java.util.concurrent;
37 import java.util.concurrent.locks.AbstractQueuedSynchronizer;
38 import java.util.concurrent.locks.Condition;
39 import java.util.concurrent.locks.ReentrantLock;
40 import java.util.concurrent.atomic.AtomicInteger;
41 import java.util.*;
42
43 /**
44 * An {@link ExecutorService} that executes each submitted task using
45 * one of possibly several pooled threads, normally configured
46 * using {@link Executors} factory methods.
47 *
48 * <p>Thread pools address two different problems: they usually
49 * provide improved performance when executing large numbers of
50 * asynchronous tasks, due to reduced per-task invocation overhead,
51 * and they provide a means of bounding and managing the resources,
52 * including threads, consumed when executing a collection of tasks.
53 * Each {@code ThreadPoolExecutor} also maintains some basic
54 * statistics, such as the number of completed tasks.
55 *
56 * <p>To be useful across a wide range of contexts, this class
57 * provides many adjustable parameters and extensibility
58 * hooks. However, programmers are urged to use the more convenient
59 * {@link Executors} factory methods {@link
60 * Executors#newCachedThreadPool} (unbounded thread pool, with
61 * automatic thread reclamation), {@link Executors#newFixedThreadPool}
62 * (fixed size thread pool) and {@link
63 * Executors#newSingleThreadExecutor} (single background thread), that
64 * preconfigure settings for the most common usage
65 * scenarios. Otherwise, use the following guide when manually
66 * configuring and tuning this class:
67 *
68 * <dl>
69 *
70 * <dt>Core and maximum pool sizes</dt>
71 *
72 * <dd>A {@code ThreadPoolExecutor} will automatically adjust the
73 * pool size (see {@link #getPoolSize})
74 * according to the bounds set by
75 * corePoolSize (see {@link #getCorePoolSize}) and
76 * maximumPoolSize (see {@link #getMaximumPoolSize}).
77 *
78 * When a new task is submitted in method {@link #execute}, and fewer
79 * than corePoolSize threads are running, a new thread is created to
80 * handle the request, even if other worker threads are idle. If
81 * there are more than corePoolSize but less than maximumPoolSize
82 * threads running, a new thread will be created only if the queue is
83 * full. By setting corePoolSize and maximumPoolSize the same, you
84 * create a fixed-size thread pool. By setting maximumPoolSize to an
85 * essentially unbounded value such as {@code Integer.MAX_VALUE}, you
86 * allow the pool to accommodate an arbitrary number of concurrent
87 * tasks. Most typically, core and maximum pool sizes are set only
88 * upon construction, but they may also be changed dynamically using
89 * {@link #setCorePoolSize} and {@link #setMaximumPoolSize}. </dd>
90 *
91 * <dt>On-demand construction</dt>
92 *
93 * <dd> By default, even core threads are initially created and
94 * started only when new tasks arrive, but this can be overridden
95 * dynamically using method {@link #prestartCoreThread} or {@link
96 * #prestartAllCoreThreads}. You probably want to prestart threads if
97 * you construct the pool with a non-empty queue. </dd>
98 *
99 * <dt>Creating new threads</dt>
100 *
101 * <dd>New threads are created using a {@link ThreadFactory}. If not
102 * otherwise specified, a {@link Executors#defaultThreadFactory} is
103 * used, that creates threads to all be in the same {@link
104 * ThreadGroup} and with the same {@code NORM_PRIORITY} priority and
105 * non-daemon status. By supplying a different ThreadFactory, you can
106 * alter the thread's name, thread group, priority, daemon status,
107 * etc. If a {@code ThreadFactory} fails to create a thread when asked
108 * by returning null from {@code newThread}, the executor will
109 * continue, but might not be able to execute any tasks. Threads
110 * should possess the "modifyThread" {@code RuntimePermission}. If
111 * worker threads or other threads using the pool do not possess this
112 * permission, service may be degraded: configuration changes may not
113 * take effect in a timely manner, and a shutdown pool may remain in a
114 * state in which termination is possible but not completed.</dd>
115 *
116 * <dt>Keep-alive times</dt>
117 *
118 * <dd>If the pool currently has more than corePoolSize threads,
119 * excess threads will be terminated if they have been idle for more
120 * than the keepAliveTime (see {@link #getKeepAliveTime}). This
121 * provides a means of reducing resource consumption when the pool is
122 * not being actively used. If the pool becomes more active later, new
123 * threads will be constructed. This parameter can also be changed
124 * dynamically using method {@link #setKeepAliveTime}. Using a value
125 * of {@code Long.MAX_VALUE} {@link TimeUnit#NANOSECONDS} effectively
126 * disables idle threads from ever terminating prior to shut down. By
127 * default, the keep-alive policy applies only when there are more
128 * than corePoolSizeThreads. But method {@link
129 * #allowCoreThreadTimeOut(boolean)} can be used to apply this
130 * time-out policy to core threads as well, so long as the
131 * keepAliveTime value is non-zero. </dd>
132 *
133 * <dt>Queuing</dt>
134 *
135 * <dd>Any {@link BlockingQueue} may be used to transfer and hold
136 * submitted tasks. The use of this queue interacts with pool sizing:
137 *
138 * <ul>
139 *
140 * <li> If fewer than corePoolSize threads are running, the Executor
141 * always prefers adding a new thread
142 * rather than queuing.</li>
143 *
144 * <li> If corePoolSize or more threads are running, the Executor
145 * always prefers queuing a request rather than adding a new
146 * thread.</li>
147 *
148 * <li> If a request cannot be queued, a new thread is created unless
149 * this would exceed maximumPoolSize, in which case, the task will be
150 * rejected.</li>
151 *
152 * </ul>
153 *
154 * There are three general strategies for queuing:
155 * <ol>
156 *
157 * <li> <em> Direct handoffs.</em> A good default choice for a work
158 * queue is a {@link SynchronousQueue} that hands off tasks to threads
159 * without otherwise holding them. Here, an attempt to queue a task
160 * will fail if no threads are immediately available to run it, so a
161 * new thread will be constructed. This policy avoids lockups when
162 * handling sets of requests that might have internal dependencies.
163 * Direct handoffs generally require unbounded maximumPoolSizes to
164 * avoid rejection of new submitted tasks. This in turn admits the
165 * possibility of unbounded thread growth when commands continue to
166 * arrive on average faster than they can be processed. </li>
167 *
168 * <li><em> Unbounded queues.</em> Using an unbounded queue (for
169 * example a {@link LinkedBlockingQueue} without a predefined
170 * capacity) will cause new tasks to wait in the queue when all
171 * corePoolSize threads are busy. Thus, no more than corePoolSize
172 * threads will ever be created. (And the value of the maximumPoolSize
173 * therefore doesn't have any effect.) This may be appropriate when
174 * each task is completely independent of others, so tasks cannot
175 * affect each others execution; for example, in a web page server.
176 * While this style of queuing can be useful in smoothing out
177 * transient bursts of requests, it admits the possibility of
178 * unbounded work queue growth when commands continue to arrive on
179 * average faster than they can be processed. </li>
180 *
181 * <li><em>Bounded queues.</em> A bounded queue (for example, an
182 * {@link ArrayBlockingQueue}) helps prevent resource exhaustion when
183 * used with finite maximumPoolSizes, but can be more difficult to
184 * tune and control. Queue sizes and maximum pool sizes may be traded
185 * off for each other: Using large queues and small pools minimizes
186 * CPU usage, OS resources, and context-switching overhead, but can
187 * lead to artificially low throughput. If tasks frequently block (for
188 * example if they are I/O bound), a system may be able to schedule
189 * time for more threads than you otherwise allow. Use of small queues
190 * generally requires larger pool sizes, which keeps CPUs busier but
191 * may encounter unacceptable scheduling overhead, which also
192 * decreases throughput. </li>
193 *
194 * </ol>
195 *
196 * </dd>
197 *
198 * <dt>Rejected tasks</dt>
199 *
200 * <dd> New tasks submitted in method {@link #execute} will be
201 * <em>rejected</em> when the Executor has been shut down, and also
202 * when the Executor uses finite bounds for both maximum threads and
203 * work queue capacity, and is saturated. In either case, the {@code
204 * execute} method invokes the {@link
205 * RejectedExecutionHandler#rejectedExecution} method of its {@link
206 * RejectedExecutionHandler}. Four predefined handler policies are
207 * provided:
208 *
209 * <ol>
210 *
211 * <li> In the default {@link ThreadPoolExecutor.AbortPolicy}, the
212 * handler throws a runtime {@link RejectedExecutionException} upon
213 * rejection. </li>
214 *
215 * <li> In {@link ThreadPoolExecutor.CallerRunsPolicy}, the thread
216 * that invokes {@code execute} itself runs the task. This provides a
217 * simple feedback control mechanism that will slow down the rate that
218 * new tasks are submitted. </li>
219 *
220 * <li> In {@link ThreadPoolExecutor.DiscardPolicy}, a task that
221 * cannot be executed is simply dropped. </li>
222 *
223 * <li>In {@link ThreadPoolExecutor.DiscardOldestPolicy}, if the
224 * executor is not shut down, the task at the head of the work queue
225 * is dropped, and then execution is retried (which can fail again,
226 * causing this to be repeated.) </li>
227 *
228 * </ol>
229 *
230 * It is possible to define and use other kinds of {@link
231 * RejectedExecutionHandler} classes. Doing so requires some care
232 * especially when policies are designed to work only under particular
233 * capacity or queuing policies. </dd>
234 *
235 * <dt>Hook methods</dt>
236 *
237 * <dd>This class provides {@code protected} overridable {@link
238 * #beforeExecute} and {@link #afterExecute} methods that are called
239 * before and after execution of each task. These can be used to
240 * manipulate the execution environment; for example, reinitializing
241 * ThreadLocals, gathering statistics, or adding log
242 * entries. Additionally, method {@link #terminated} can be overridden
243 * to perform any special processing that needs to be done once the
244 * Executor has fully terminated.
245 *
246 * <p>If hook or callback methods throw exceptions, internal worker
247 * threads may in turn fail and abruptly terminate.</dd>
248 *
249 * <dt>Queue maintenance</dt>
250 *
251 * <dd> Method {@link #getQueue} allows access to the work queue for
252 * purposes of monitoring and debugging. Use of this method for any
253 * other purpose is strongly discouraged. Two supplied methods,
254 * {@link #remove} and {@link #purge} are available to assist in
255 * storage reclamation when large numbers of queued tasks become
256 * cancelled.</dd>
257 *
258 * <dt>Finalization</dt>
259 *
260 * <dd> A pool that is no longer referenced in a program <em>AND</em>
261 * has no remaining threads will be {@code shutdown} automatically. If
262 * you would like to ensure that unreferenced pools are reclaimed even
263 * if users forget to call {@link #shutdown}, then you must arrange
264 * that unused threads eventually die, by setting appropriate
265 * keep-alive times, using a lower bound of zero core threads and/or
266 * setting {@link #allowCoreThreadTimeOut(boolean)}. </dd>
267 *
268 * </dl>
269 *
270 * <p> <b>Extension example</b>. Most extensions of this class
271 * override one or more of the protected hook methods. For example,
272 * here is a subclass that adds a simple pause/resume feature:
273 *
274 * <pre> {@code
275 * class PausableThreadPoolExecutor extends ThreadPoolExecutor {
276 * private boolean isPaused;
277 * private ReentrantLock pauseLock = new ReentrantLock();
278 * private Condition unpaused = pauseLock.newCondition();
279 *
280 * public PausableThreadPoolExecutor(...) { super(...); }
281 *
282 * protected void beforeExecute(Thread t, Runnable r) {
283 * super.beforeExecute(t, r);
284 * pauseLock.lock();
285 * try {
286 * while (isPaused) unpaused.await();
287 * } catch (InterruptedException ie) {
288 * t.interrupt();
289 * } finally {
290 * pauseLock.unlock();
291 * }
292 * }
293 *
294 * public void pause() {
295 * pauseLock.lock();
296 * try {
297 * isPaused = true;
298 * } finally {
299 * pauseLock.unlock();
300 * }
301 * }
302 *
303 * public void resume() {
304 * pauseLock.lock();
305 * try {
306 * isPaused = false;
307 * unpaused.signalAll();
308 * } finally {
309 * pauseLock.unlock();
310 * }
311 * }
312 * }}</pre>
313 *
314 * @since 1.5
315 * @author Doug Lea
316 */
317 public class ThreadPoolExecutor extends AbstractExecutorService {
318 /**
319 * The main pool control state, ctl, is an atomic integer packing
320 * two conceptual fields
321 * workerCount, indicating the effective number of threads
322 * runState, indicating whether running, shutting down etc
323 *
324 * In order to pack them into one int, we limit workerCount to
325 * (2^29)-1 (about 500 million) threads rather than (2^31)-1 (2
326 * billion) otherwise representable. If this is ever an issue in
327 * the future, the variable can be changed to be an AtomicLong,
328 * and the shift/mask constants below adjusted. But until the need
329 * arises, this code is a bit faster and simpler using an int.
330 *
331 * The workerCount is the number of workers that have been
332 * permitted to start and not permitted to stop. The value may be
333 * transiently different from the actual number of live threads,
334 * for example when a ThreadFactory fails to create a thread when
335 * asked, and when exiting threads are still performing
336 * bookkeeping before terminating. The user-visible pool size is
337 * reported as the current size of the workers set.
338 *
339 * The runState provides the main lifecyle control, taking on values:
340 *
341 * RUNNING: Accept new tasks and process queued tasks
342 * SHUTDOWN: Don't accept new tasks, but process queued tasks
343 * STOP: Don't accept new tasks, don't process queued tasks,
344 * and interrupt in-progress tasks
345 * TIDYING: All tasks have terminated, workerCount is zero,
346 * the thread transitioning to state TIDYING
347 * will run the terminated() hook method
348 * TERMINATED: terminated() has completed
349 *
350 * The numerical order among these values matters, to allow
351 * ordered comparisons. The runState monotonically increases over
352 * time, but need not hit each state. The transitions are:
353 *
354 * RUNNING -> SHUTDOWN
355 * On invocation of shutdown(), perhaps implicitly in finalize()
356 * (RUNNING or SHUTDOWN) -> STOP
357 * On invocation of shutdownNow()
358 * SHUTDOWN -> TIDYING
359 * When both queue and pool are empty
360 * STOP -> TIDYING
361 * When pool is empty
362 * TIDYING -> TERMINATED
363 * When the terminated() hook method has completed
364 *
365 * Threads waiting in awaitTermination() will return when the
366 * state reaches TERMINATED.
367 *
368 * Detecting the transition from SHUTDOWN to TIDYING is less
369 * straightforward than you'd like because the queue may become
370 * empty after non-empty and vice versa during SHUTDOWN state, but
371 * we can only terminate if, after seeing that it is empty, we see
372 * that workerCount is 0 (which sometimes entails a recheck -- see
373 * below).
374 */
375 private final AtomicInteger ctl = new AtomicInteger(ctlOf(RUNNING, 0));
376 private static final int COUNT_BITS = Integer.SIZE - 3;
377 private static final int CAPACITY = (1 << COUNT_BITS) - 1;
378
379 // runState is stored in the high-order bits
380 private static final int RUNNING = -1 << COUNT_BITS;
381 private static final int SHUTDOWN = 0 << COUNT_BITS;
382 private static final int STOP = 1 << COUNT_BITS;
383 private static final int TIDYING = 2 << COUNT_BITS;
384 private static final int TERMINATED = 3 << COUNT_BITS;
385
386 // Packing and unpacking ctl
387 private static int runStateOf(int c) { return c & ~CAPACITY; }
388 private static int workerCountOf(int c) { return c & CAPACITY; }
389 private static int ctlOf(int rs, int wc) { return rs | wc; }
390
391 /*
392 * Bit field accessors that don't require unpacking ctl.
393 * These depend on the bit layout and on workerCount being never negative.
394 */
395
396 private static boolean runStateLessThan(int c, int s) {
397 return c < s;
398 }
399
400 private static boolean runStateAtLeast(int c, int s) {
401 return c >= s;
402 }
403
404 private static boolean isRunning(int c) {
405 return c < SHUTDOWN;
406 }
407
408 /**
409 * Attempt to CAS-increment the workerCount field of ctl.
410 */
411 private boolean compareAndIncrementWorkerCount(int expect) {
412 return ctl.compareAndSet(expect, expect + 1);
413 }
414
415 /**
416 * Attempt to CAS-decrement the workerCount field of ctl.
417 */
418 private boolean compareAndDecrementWorkerCount(int expect) {
419 return ctl.compareAndSet(expect, expect - 1);
420 }
421
422 /**
423 * Decrements the workerCount field of ctl. This is called only on
424 * abrupt termination of a thread (see processWorkerExit). Other
425 * decrements are performed within getTask.
426 */
427 private void decrementWorkerCount() {
428 do {} while (! compareAndDecrementWorkerCount(ctl.get()));
429 }
430
431 /**
432 * The queue used for holding tasks and handing off to worker
433 * threads. We do not require that workQueue.poll() returning
434 * null necessarily means that workQueue.isEmpty(), so rely
435 * solely on isEmpty to see if the queue is empty (which we must
436 * do for example when deciding whether to transition from
437 * SHUTDOWN to TIDYING). This accommodates special-purpose
438 * queues such as DelayQueues for which poll() is allowed to
439 * return null even if it may later return non-null when delays
440 * expire.
441 */
442 private final BlockingQueue<Runnable> workQueue;
443
444 /**
445 * Lock held on access to workers set and related bookkeeping.
446 * While we could use a concurrent set of some sort, it turns out
447 * to be generally preferable to use a lock. Among the reasons is
448 * that this serializes interruptIdleWorkers, which avoids
449 * unnecessary interrupt storms, especially during shutdown.
450 * Otherwise exiting threads would concurrently interrupt those
451 * that have not yet interrupted. It also simplifies some of the
452 * associated statistics bookkeeping of largestPoolSize etc. We
453 * also hold mainLock on shutdown and shutdownNow, for the sake of
454 * ensuring workers set is stable while separately checking
455 * permission to interrupt and actually interrupting.
456 */
457 private final ReentrantLock mainLock = new ReentrantLock();
458
459 /**
460 * Set containing all worker threads in pool. Accessed only when
461 * holding mainLock.
462 */
463 private final HashSet<Worker> workers = new HashSet<Worker>();
464
465 /**
466 * Wait condition to support awaitTermination
467 */
468 private final Condition termination = mainLock.newCondition();
469
470 /**
471 * Tracks largest attained pool size. Accessed only under
472 * mainLock.
473 */
474 private int largestPoolSize;
475
476 /**
477 * Counter for completed tasks. Updated only on termination of
478 * worker threads. Accessed only under mainLock.
479 */
480 private long completedTaskCount;
481
482 /*
483 * All user control parameters are declared as volatiles so that
484 * ongoing actions are based on freshest values, but without need
485 * for locking, since no internal invariants depend on them
486 * changing synchronously with respect to other actions.
487 */
488
489 /**
490 * Factory for new threads. All threads are created using this
491 * factory (via method addWorker). All callers must be prepared
492 * for addWorker to fail, which may reflect a system or user's
493 * policy limiting the number of threads. Even though it is not
494 * treated as an error, failure to create threads may result in
495 * new tasks being rejected or existing ones remaining stuck in
496 * the queue.
497 *
498 * We go further and preserve pool invariants even in the face of
499 * errors such as OutOfMemoryError, that might be thrown while
500 * trying to create threads. Such errors are rather common due to
501 * the need to allocate a native stack in Thread#start, and users
502 * will want to perform clean pool shutdown to clean up. There
503 * will likely be enough memory available for the cleanup code to
504 * complete without encountering yet another OutOfMemoryError.
505 */
506 private volatile ThreadFactory threadFactory;
507
508 /**
509 * Handler called when saturated or shutdown in execute.
510 */
511 private volatile RejectedExecutionHandler handler;
512
513 /**
514 * Timeout in nanoseconds for idle threads waiting for work.
515 * Threads use this timeout when there are more than corePoolSize
516 * present or if allowCoreThreadTimeOut. Otherwise they wait
517 * forever for new work.
518 */
519 private volatile long keepAliveTime;
520
521 /**
522 * If false (default), core threads stay alive even when idle.
523 * If true, core threads use keepAliveTime to time out waiting
524 * for work.
525 */
526 private volatile boolean allowCoreThreadTimeOut;
527
528 /**
529 * Core pool size is the minimum number of workers to keep alive
530 * (and not allow to time out etc) unless allowCoreThreadTimeOut
531 * is set, in which case the minimum is zero.
532 */
533 private volatile int corePoolSize;
534
535 /**
536 * Maximum pool size. Note that the actual maximum is internally
537 * bounded by CAPACITY.
538 */
539 private volatile int maximumPoolSize;
540
541 /**
542 * The default rejected execution handler
543 */
544 private static final RejectedExecutionHandler defaultHandler =
545 new AbortPolicy();
546
547 /**
548 * Permission required for callers of shutdown and shutdownNow.
549 * We additionally require (see checkShutdownAccess) that callers
550 * have permission to actually interrupt threads in the worker set
551 * (as governed by Thread.interrupt, which relies on
552 * ThreadGroup.checkAccess, which in turn relies on
553 * SecurityManager.checkAccess). Shutdowns are attempted only if
554 * these checks pass.
555 *
556 * All actual invocations of Thread.interrupt (see
557 * interruptIdleWorkers and interruptWorkers) ignore
558 * SecurityExceptions, meaning that the attempted interrupts
559 * silently fail. In the case of shutdown, they should not fail
560 * unless the SecurityManager has inconsistent policies, sometimes
561 * allowing access to a thread and sometimes not. In such cases,
562 * failure to actually interrupt threads may disable or delay full
563 * termination. Other uses of interruptIdleWorkers are advisory,
564 * and failure to actually interrupt will merely delay response to
565 * configuration changes so is not handled exceptionally.
566 */
567 private static final RuntimePermission shutdownPerm =
568 new RuntimePermission("modifyThread");
569
570 /**
571 * Class Worker mainly maintains interrupt control state for
572 * threads running tasks, along with other minor bookkeeping.
573 * This class opportunistically extends AbstractQueuedSynchronizer
574 * to simplify acquiring and releasing a lock surrounding each
575 * task execution. This protects against interrupts that are
576 * intended to wake up a worker thread waiting for a task from
577 * instead interrupting a task being run. We implement a simple
578 * non-reentrant mutual exclusion lock rather than use
579 * ReentrantLock because we do not want worker tasks to be able to
580 * reacquire the lock when they invoke pool control methods like
581 * setCorePoolSize. Additionally, to suppress interrupts until
582 * the thread actually starts running tasks, we initialize lock
583 * state to a negative value, and clear it upon start (in
584 * runWorker).
585 */
586 private final class Worker
587 extends AbstractQueuedSynchronizer
588 implements Runnable
589 {
590 /**
591 * This class will never be serialized, but we provide a
592 * serialVersionUID to suppress a javac warning.
593 */
594 private static final long serialVersionUID = 6138294804551838833L;
595
596 /** Thread this worker is running in. Null if factory fails. */
597 final Thread thread;
598 /** Initial task to run. Possibly null. */
599 Runnable firstTask;
600 /** Per-thread task counter */
601 volatile long completedTasks;
602
603 /**
604 * Creates with given first task and thread from ThreadFactory.
605 * @param firstTask the first task (null if none)
606 */
607 Worker(Runnable firstTask) {
608 setState(-1); // inhibit interrupts until runWorker
609 this.firstTask = firstTask;
610 this.thread = getThreadFactory().newThread(this);
611 }
612
613 /** Delegates main run loop to outer runWorker */
614 public void run() {
615 runWorker(this);
616 }
617
618 // Lock methods
619 //
620 // The value 0 represents the unlocked state.
621 // The value 1 represents the locked state.
622
623 protected boolean isHeldExclusively() {
624 return getState() != 0;
625 }
626
627 protected boolean tryAcquire(int unused) {
628 if (compareAndSetState(0, 1)) {
629 setExclusiveOwnerThread(Thread.currentThread());
630 return true;
631 }
632 return false;
633 }
634
635 protected boolean tryRelease(int unused) {
636 setExclusiveOwnerThread(null);
637 setState(0);
638 return true;
639 }
640
641 public void lock() { acquire(1); }
642 public boolean tryLock() { return tryAcquire(1); }
643 public void unlock() { release(1); }
644 public boolean isLocked() { return isHeldExclusively(); }
645
646 void interruptIfStarted() {
647 Thread t;
648 if (getState() >= 0 && (t = thread) != null && !t.isInterrupted()) {
649 try {
650 t.interrupt();
651 } catch (SecurityException ignore) {
652 }
653 }
654 }
655 }
656
657 /*
658 * Methods for setting control state
659 */
660
661 /**
662 * Transitions runState to given target, or leaves it alone if
663 * already at least the given target.
664 *
665 * @param targetState the desired state, either SHUTDOWN or STOP
666 * (but not TIDYING or TERMINATED -- use tryTerminate for that)
667 */
668 private void advanceRunState(int targetState) {
669 for (;;) {
670 int c = ctl.get();
671 if (runStateAtLeast(c, targetState) ||
672 ctl.compareAndSet(c, ctlOf(targetState, workerCountOf(c))))
673 break;
674 }
675 }
676
677 /**
678 * Transitions to TERMINATED state if either (SHUTDOWN and pool
679 * and queue empty) or (STOP and pool empty). If otherwise
680 * eligible to terminate but workerCount is nonzero, interrupts an
681 * idle worker to ensure that shutdown signals propagate. This
682 * method must be called following any action that might make
683 * termination possible -- reducing worker count or removing tasks
684 * from the queue during shutdown. The method is non-private to
685 * allow access from ScheduledThreadPoolExecutor.
686 */
687 final void tryTerminate() {
688 for (;;) {
689 int c = ctl.get();
690 if (isRunning(c) ||
691 runStateAtLeast(c, TIDYING) ||
692 (runStateOf(c) == SHUTDOWN && ! workQueue.isEmpty()))
693 return;
694 if (workerCountOf(c) != 0) { // Eligible to terminate
695 interruptIdleWorkers(ONLY_ONE);
696 return;
697 }
698
699 final ReentrantLock mainLock = this.mainLock;
700 mainLock.lock();
701 try {
702 if (ctl.compareAndSet(c, ctlOf(TIDYING, 0))) {
703 try {
704 terminated();
705 } finally {
706 ctl.set(ctlOf(TERMINATED, 0));
707 termination.signalAll();
708 }
709 return;
710 }
711 } finally {
712 mainLock.unlock();
713 }
714 // else retry on failed CAS
715 }
716 }
717
718 /*
719 * Methods for controlling interrupts to worker threads.
720 */
721
722 /**
723 * If there is a security manager, makes sure caller has
724 * permission to shut down threads in general (see shutdownPerm).
725 * If this passes, additionally makes sure the caller is allowed
726 * to interrupt each worker thread. This might not be true even if
727 * first check passed, if the SecurityManager treats some threads
728 * specially.
729 */
730 private void checkShutdownAccess() {
731 SecurityManager security = System.getSecurityManager();
732 if (security != null) {
733 security.checkPermission(shutdownPerm);
734 final ReentrantLock mainLock = this.mainLock;
735 mainLock.lock();
736 try {
737 for (Worker w : workers)
738 security.checkAccess(w.thread);
739 } finally {
740 mainLock.unlock();
741 }
742 }
743 }
744
745 /**
746 * Interrupts all threads, even if active. Ignores SecurityExceptions
747 * (in which case some threads may remain uninterrupted).
748 */
749 private void interruptWorkers() {
750 final ReentrantLock mainLock = this.mainLock;
751 mainLock.lock();
752 try {
753 for (Worker w : workers)
754 w.interruptIfStarted();
755 } finally {
756 mainLock.unlock();
757 }
758 }
759
760 /**
761 * Interrupts threads that might be waiting for tasks (as
762 * indicated by not being locked) so they can check for
763 * termination or configuration changes. Ignores
764 * SecurityExceptions (in which case some threads may remain
765 * uninterrupted).
766 *
767 * @param onlyOne If true, interrupt at most one worker. This is
768 * called only from tryTerminate when termination is otherwise
769 * enabled but there are still other workers. In this case, at
770 * most one waiting worker is interrupted to propagate shutdown
771 * signals in case all threads are currently waiting.
772 * Interrupting any arbitrary thread ensures that newly arriving
773 * workers since shutdown began will also eventually exit.
774 * To guarantee eventual termination, it suffices to always
775 * interrupt only one idle worker, but shutdown() interrupts all
776 * idle workers so that redundant workers exit promptly, not
777 * waiting for a straggler task to finish.
778 */
779 private void interruptIdleWorkers(boolean onlyOne) {
780 final ReentrantLock mainLock = this.mainLock;
781 mainLock.lock();
782 try {
783 for (Worker w : workers) {
784 Thread t = w.thread;
785 if (!t.isInterrupted() && w.tryLock()) {
786 try {
787 t.interrupt();
788 } catch (SecurityException ignore) {
789 } finally {
790 w.unlock();
791 }
792 }
793 if (onlyOne)
794 break;
795 }
796 } finally {
797 mainLock.unlock();
798 }
799 }
800
801 /**
802 * Common form of interruptIdleWorkers, to avoid having to
803 * remember what the boolean argument means.
804 */
805 private void interruptIdleWorkers() {
806 interruptIdleWorkers(false);
807 }
808
809 private static final boolean ONLY_ONE = true;
810
811 /*
812 * Misc utilities, most of which are also exported to
813 * ScheduledThreadPoolExecutor
814 */
815
816 /**
817 * Invokes the rejected execution handler for the given command.
818 * Package-protected for use by ScheduledThreadPoolExecutor.
819 */
820 final void reject(Runnable command) {
821 handler.rejectedExecution(command, this);
822 }
823
824 /**
825 * Performs any further cleanup following run state transition on
826 * invocation of shutdown. A no-op here, but used by
827 * ScheduledThreadPoolExecutor to cancel delayed tasks.
828 */
829 void onShutdown() {
830 }
831
832 /**
833 * State check needed by ScheduledThreadPoolExecutor to
834 * enable running tasks during shutdown.
835 *
836 * @param shutdownOK true if should return true if SHUTDOWN
837 */
838 final boolean isRunningOrShutdown(boolean shutdownOK) {
839 int rs = runStateOf(ctl.get());
840 return rs == RUNNING || (rs == SHUTDOWN && shutdownOK);
841 }
842
843 /**
844 * Drains the task queue into a new list, normally using
845 * drainTo. But if the queue is a DelayQueue or any other kind of
846 * queue for which poll or drainTo may fail to remove some
847 * elements, it deletes them one by one.
848 */
849 private List<Runnable> drainQueue() {
850 BlockingQueue<Runnable> q = workQueue;
851 List<Runnable> taskList = new ArrayList<Runnable>();
852 q.drainTo(taskList);
853 if (!q.isEmpty()) {
854 for (Runnable r : q.toArray(new Runnable[0])) {
855 if (q.remove(r))
856 taskList.add(r);
857 }
858 }
859 return taskList;
860 }
861
862 /*
863 * Methods for creating, running and cleaning up after workers
864 */
865
866 /**
867 * Checks if a new worker can be added with respect to current
868 * pool state and the given bound (either core or maximum). If so,
869 * the worker count is adjusted accordingly, and, if possible, a
870 * new worker is created and started, running firstTask as its
871 * first task. This method returns false if the pool is stopped or
872 * eligible to shut down. It also returns false if the thread
873 * factory fails to create a thread when asked. If the thread
874 * creation fails, either due to the thread factory returning
875 * null, or due to an exception (typically OutOfMemoryError in
876 * Thread#start), we roll back cleanly.
877 *
878 * @param firstTask the task the new thread should run first (or
879 * null if none). Workers are created with an initial first task
880 * (in method execute()) to bypass queuing when there are fewer
881 * than corePoolSize threads (in which case we always start one),
882 * or when the queue is full (in which case we must bypass queue).
883 * Initially idle threads are usually created via
884 * prestartCoreThread or to replace other dying workers.
885 *
886 * @param core if true use corePoolSize as bound, else
887 * maximumPoolSize. (A boolean indicator is used here rather than a
888 * value to ensure reads of fresh values after checking other pool
889 * state).
890 * @return true if successful
891 */
892 private boolean addWorker(Runnable firstTask, boolean core) {
893 retry:
894 for (;;) {
895 int c = ctl.get();
896 int rs = runStateOf(c);
897
898 // Check if queue empty only if necessary.
899 if (rs >= SHUTDOWN &&
900 ! (rs == SHUTDOWN &&
901 firstTask == null &&
902 ! workQueue.isEmpty()))
903 return false;
904
905 for (;;) {
906 int wc = workerCountOf(c);
907 if (wc >= CAPACITY ||
908 wc >= (core ? corePoolSize : maximumPoolSize))
909 return false;
910 if (compareAndIncrementWorkerCount(c))
911 break retry;
912 c = ctl.get(); // Re-read ctl
913 if (runStateOf(c) != rs)
914 continue retry;
915 // else CAS failed due to workerCount change; retry inner loop
916 }
917 }
918
919 boolean workerStarted = false;
920 boolean workerAdded = false;
921 Worker w = null;
922 try {
923 final ReentrantLock mainLock = this.mainLock;
924 w = new Worker(firstTask);
925 final Thread t = w.thread;
926 if (t != null) {
927 mainLock.lock();
928 try {
929 // Recheck while holding lock.
930 // Back out on ThreadFactory failure or if
931 // shut down before lock acquired.
932 int c = ctl.get();
933 int rs = runStateOf(c);
934
935 if (rs < SHUTDOWN ||
936 (rs == SHUTDOWN && firstTask == null)) {
937 if (t.isAlive()) // precheck that t is startable
938 throw new IllegalThreadStateException();
939 workers.add(w);
940 int s = workers.size();
941 if (s > largestPoolSize)
942 largestPoolSize = s;
943 workerAdded = true;
944 }
945 } finally {
946 mainLock.unlock();
947 }
948 if (workerAdded) {
949 t.start();
950 workerStarted = true;
951 }
952 }
953 } finally {
954 if (! workerStarted)
955 addWorkerFailed(w);
956 }
957 return workerStarted;
958 }
959
960 /**
961 * Rolls back the worker thread creation.
962 * - removes worker from workers, if present
963 * - decrements worker count
964 * - rechecks for termination, in case the existence of this
965 * worker was holding up termination
966 */
967 private void addWorkerFailed(Worker w) {
968 final ReentrantLock mainLock = this.mainLock;
969 mainLock.lock();
970 try {
971 if (w != null)
972 workers.remove(w);
973 decrementWorkerCount();
974 tryTerminate();
975 } finally {
976 mainLock.unlock();
977 }
978 }
979
980 /**
981 * Performs cleanup and bookkeeping for a dying worker. Called
982 * only from worker threads. Unless completedAbruptly is set,
983 * assumes that workerCount has already been adjusted to account
984 * for exit. This method removes thread from worker set, and
985 * possibly terminates the pool or replaces the worker if either
986 * it exited due to user task exception or if fewer than
987 * corePoolSize workers are running or queue is non-empty but
988 * there are no workers.
989 *
990 * @param w the worker
991 * @param completedAbruptly if the worker died due to user exception
992 */
993 private void processWorkerExit(Worker w, boolean completedAbruptly) {
994 if (completedAbruptly) // If abrupt, then workerCount wasn't adjusted
995 decrementWorkerCount();
996
997 final ReentrantLock mainLock = this.mainLock;
998 mainLock.lock();
999 try {
1000 completedTaskCount += w.completedTasks;
1001 workers.remove(w);
1002 } finally {
1003 mainLock.unlock();
1004 }
1005
1006 tryTerminate();
1007
1008 int c = ctl.get();
1009 if (runStateLessThan(c, STOP)) {
1010 if (!completedAbruptly) {
1011 int min = allowCoreThreadTimeOut ? 0 : corePoolSize;
1012 if (min == 0 && ! workQueue.isEmpty())
1013 min = 1;
1014 if (workerCountOf(c) >= min)
1015 return; // replacement not needed
1016 }
1017 addWorker(null, false);
1018 }
1019 }
1020
1021 /**
1022 * Performs blocking or timed wait for a task, depending on
1023 * current configuration settings, or returns null if this worker
1024 * must exit because of any of:
1025 * 1. There are more than maximumPoolSize workers (due to
1026 * a call to setMaximumPoolSize).
1027 * 2. The pool is stopped.
1028 * 3. The pool is shutdown and the queue is empty.
1029 * 4. This worker timed out waiting for a task, and timed-out
1030 * workers are subject to termination (that is,
1031 * {@code allowCoreThreadTimeOut || workerCount > corePoolSize})
1032 * both before and after the timed wait.
1033 *
1034 * @return task, or null if the worker must exit, in which case
1035 * workerCount is decremented
1036 */
1037 private Runnable getTask() {
1038 boolean timedOut = false; // Did the last poll() time out?
1039
1040 retry:
1041 for (;;) {
1042 int c = ctl.get();
1043 int rs = runStateOf(c);
1044
1045 // Check if queue empty only if necessary.
1046 if (rs >= SHUTDOWN && (rs >= STOP || workQueue.isEmpty())) {
1047 decrementWorkerCount();
1048 return null;
1049 }
1050
1051 boolean timed; // Are workers subject to culling?
1052
1053 for (;;) {
1054 int wc = workerCountOf(c);
1055 timed = allowCoreThreadTimeOut || wc > corePoolSize;
1056
1057 if (wc <= maximumPoolSize && ! (timedOut && timed))
1058 break;
1059 if (compareAndDecrementWorkerCount(c))
1060 return null;
1061 c = ctl.get(); // Re-read ctl
1062 if (runStateOf(c) != rs)
1063 continue retry;
1064 // else CAS failed due to workerCount change; retry inner loop
1065 }
1066
1067 try {
1068 Runnable r = timed ?
1069 workQueue.poll(keepAliveTime, TimeUnit.NANOSECONDS) :
1070 workQueue.take();
1071 if (r != null)
1072 return r;
1073 timedOut = true;
1074 } catch (InterruptedException retry) {
1075 timedOut = false;
1076 }
1077 }
1078 }
1079
1080 /**
1081 * Main worker run loop. Repeatedly gets tasks from queue and
1082 * executes them, while coping with a number of issues:
1083 *
1084 * 1. We may start out with an initial task, in which case we
1085 * don't need to get the first one. Otherwise, as long as pool is
1086 * running, we get tasks from getTask. If it returns null then the
1087 * worker exits due to changed pool state or configuration
1088 * parameters. Other exits result from exception throws in
1089 * external code, in which case completedAbruptly holds, which
1090 * usually leads processWorkerExit to replace this thread.
1091 *
1092 * 2. Before running any task, the lock is acquired to prevent
1093 * other pool interrupts while the task is executing, and
1094 * clearInterruptsForTaskRun called to ensure that unless pool is
1095 * stopping, this thread does not have its interrupt set.
1096 *
1097 * 3. Each task run is preceded by a call to beforeExecute, which
1098 * might throw an exception, in which case we cause thread to die
1099 * (breaking loop with completedAbruptly true) without processing
1100 * the task.
1101 *
1102 * 4. Assuming beforeExecute completes normally, we run the task,
1103 * gathering any of its thrown exceptions to send to
1104 * afterExecute. We separately handle RuntimeException, Error
1105 * (both of which the specs guarantee that we trap) and arbitrary
1106 * Throwables. Because we cannot rethrow Throwables within
1107 * Runnable.run, we wrap them within Errors on the way out (to the
1108 * thread's UncaughtExceptionHandler). Any thrown exception also
1109 * conservatively causes thread to die.
1110 *
1111 * 5. After task.run completes, we call afterExecute, which may
1112 * also throw an exception, which will also cause thread to
1113 * die. According to JLS Sec 14.20, this exception is the one that
1114 * will be in effect even if task.run throws.
1115 *
1116 * The net effect of the exception mechanics is that afterExecute
1117 * and the thread's UncaughtExceptionHandler have as accurate
1118 * information as we can provide about any problems encountered by
1119 * user code.
1120 *
1121 * @param w the worker
1122 */
1123 final void runWorker(Worker w) {
1124 Thread wt = Thread.currentThread();
1125 Runnable task = w.firstTask;
1126 w.firstTask = null;
1127 w.unlock(); // allow interrupts
1128 boolean completedAbruptly = true;
1129 try {
1130 while (task != null || (task = getTask()) != null) {
1131 w.lock();
1132 // If pool is stopping, ensure thread is interrupted;
1133 // if not, ensure thread is not interrupted. This
1134 // requires a recheck in second case to deal with
1135 // shutdownNow race while clearing interrupt
1136 if ((runStateAtLeast(ctl.get(), STOP) ||
1137 (Thread.interrupted() &&
1138 runStateAtLeast(ctl.get(), STOP))) &&
1139 !wt.isInterrupted())
1140 wt.interrupt();
1141 try {
1142 beforeExecute(wt, task);
1143 Throwable thrown = null;
1144 try {
1145 task.run();
1146 } catch (RuntimeException x) {
1147 thrown = x; throw x;
1148 } catch (Error x) {
1149 thrown = x; throw x;
1150 } catch (Throwable x) {
1151 thrown = x; throw new Error(x);
1152 } finally {
1153 afterExecute(task, thrown);
1154 }
1155 } finally {
1156 task = null;
1157 w.completedTasks++;
1158 w.unlock();
1159 }
1160 }
1161 completedAbruptly = false;
1162 } finally {
1163 processWorkerExit(w, completedAbruptly);
1164 }
1165 }
1166
1167 // Public constructors and methods
1168
1169 /**
1170 * Creates a new {@code ThreadPoolExecutor} with the given initial
1171 * parameters and default thread factory and rejected execution handler.
1172 * It may be more convenient to use one of the {@link Executors} factory
1173 * methods instead of this general purpose constructor.
1174 *
1175 * @param corePoolSize the number of threads to keep in the pool, even
1176 * if they are idle, unless {@code allowCoreThreadTimeOut} is set
1177 * @param maximumPoolSize the maximum number of threads to allow in the
1178 * pool
1179 * @param keepAliveTime when the number of threads is greater than
1180 * the core, this is the maximum time that excess idle threads
1181 * will wait for new tasks before terminating.
1182 * @param unit the time unit for the {@code keepAliveTime} argument
1183 * @param workQueue the queue to use for holding tasks before they are
1184 * executed. This queue will hold only the {@code Runnable}
1185 * tasks submitted by the {@code execute} method.
1186 * @throws IllegalArgumentException if one of the following holds:<br>
1187 * {@code corePoolSize < 0}<br>
1188 * {@code keepAliveTime < 0}<br>
1189 * {@code maximumPoolSize <= 0}<br>
1190 * {@code maximumPoolSize < corePoolSize}
1191 * @throws NullPointerException if {@code workQueue} is null
1192 */
1193 public ThreadPoolExecutor(int corePoolSize,
1194 int maximumPoolSize,
1195 long keepAliveTime,
1196 TimeUnit unit,
1197 BlockingQueue<Runnable> workQueue) {
1198 this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue,
1199 Executors.defaultThreadFactory(), defaultHandler);
1200 }
1201
1202 /**
1203 * Creates a new {@code ThreadPoolExecutor} with the given initial
1204 * parameters and default rejected execution handler.
1205 *
1206 * @param corePoolSize the number of threads to keep in the pool, even
1207 * if they are idle, unless {@code allowCoreThreadTimeOut} is set
1208 * @param maximumPoolSize the maximum number of threads to allow in the
1209 * pool
1210 * @param keepAliveTime when the number of threads is greater than
1211 * the core, this is the maximum time that excess idle threads
1212 * will wait for new tasks before terminating.
1213 * @param unit the time unit for the {@code keepAliveTime} argument
1214 * @param workQueue the queue to use for holding tasks before they are
1215 * executed. This queue will hold only the {@code Runnable}
1216 * tasks submitted by the {@code execute} method.
1217 * @param threadFactory the factory to use when the executor
1218 * creates a new thread
1219 * @throws IllegalArgumentException if one of the following holds:<br>
1220 * {@code corePoolSize < 0}<br>
1221 * {@code keepAliveTime < 0}<br>
1222 * {@code maximumPoolSize <= 0}<br>
1223 * {@code maximumPoolSize < corePoolSize}
1224 * @throws NullPointerException if {@code workQueue}
1225 * or {@code threadFactory} is null
1226 */
1227 public ThreadPoolExecutor(int corePoolSize,
1228 int maximumPoolSize,
1229 long keepAliveTime,
1230 TimeUnit unit,
1231 BlockingQueue<Runnable> workQueue,
1232 ThreadFactory threadFactory) {
1233 this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue,
1234 threadFactory, defaultHandler);
1235 }
1236
1237 /**
1238 * Creates a new {@code ThreadPoolExecutor} with the given initial
1239 * parameters and default thread factory.
1240 *
1241 * @param corePoolSize the number of threads to keep in the pool, even
1242 * if they are idle, unless {@code allowCoreThreadTimeOut} is set
1243 * @param maximumPoolSize the maximum number of threads to allow in the
1244 * pool
1245 * @param keepAliveTime when the number of threads is greater than
1246 * the core, this is the maximum time that excess idle threads
1247 * will wait for new tasks before terminating.
1248 * @param unit the time unit for the {@code keepAliveTime} argument
1249 * @param workQueue the queue to use for holding tasks before they are
1250 * executed. This queue will hold only the {@code Runnable}
1251 * tasks submitted by the {@code execute} method.
1252 * @param handler the handler to use when execution is blocked
1253 * because the thread bounds and queue capacities are reached
1254 * @throws IllegalArgumentException if one of the following holds:<br>
1255 * {@code corePoolSize < 0}<br>
1256 * {@code keepAliveTime < 0}<br>
1257 * {@code maximumPoolSize <= 0}<br>
1258 * {@code maximumPoolSize < corePoolSize}
1259 * @throws NullPointerException if {@code workQueue}
1260 * or {@code handler} is null
1261 */
1262 public ThreadPoolExecutor(int corePoolSize,
1263 int maximumPoolSize,
1264 long keepAliveTime,
1265 TimeUnit unit,
1266 BlockingQueue<Runnable> workQueue,
1267 RejectedExecutionHandler handler) {
1268 this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue,
1269 Executors.defaultThreadFactory(), handler);
1270 }
1271
1272 /**
1273 * Creates a new {@code ThreadPoolExecutor} with the given initial
1274 * parameters.
1275 *
1276 * @param corePoolSize the number of threads to keep in the pool, even
1277 * if they are idle, unless {@code allowCoreThreadTimeOut} is set
1278 * @param maximumPoolSize the maximum number of threads to allow in the
1279 * pool
1280 * @param keepAliveTime when the number of threads is greater than
1281 * the core, this is the maximum time that excess idle threads
1282 * will wait for new tasks before terminating.
1283 * @param unit the time unit for the {@code keepAliveTime} argument
1284 * @param workQueue the queue to use for holding tasks before they are
1285 * executed. This queue will hold only the {@code Runnable}
1286 * tasks submitted by the {@code execute} method.
1287 * @param threadFactory the factory to use when the executor
1288 * creates a new thread
1289 * @param handler the handler to use when execution is blocked
1290 * because the thread bounds and queue capacities are reached
1291 * @throws IllegalArgumentException if one of the following holds:<br>
1292 * {@code corePoolSize < 0}<br>
1293 * {@code keepAliveTime < 0}<br>
1294 * {@code maximumPoolSize <= 0}<br>
1295 * {@code maximumPoolSize < corePoolSize}
1296 * @throws NullPointerException if {@code workQueue}
1297 * or {@code threadFactory} or {@code handler} is null
1298 */
1299 public ThreadPoolExecutor(int corePoolSize,
1300 int maximumPoolSize,
1301 long keepAliveTime,
1302 TimeUnit unit,
1303 BlockingQueue<Runnable> workQueue,
1304 ThreadFactory threadFactory,
1305 RejectedExecutionHandler handler) {
1306 if (corePoolSize < 0 ||
1307 maximumPoolSize <= 0 ||
1308 maximumPoolSize < corePoolSize ||
1309 keepAliveTime < 0)
1310 throw new IllegalArgumentException();
1311 if (workQueue == null || threadFactory == null || handler == null)
1312 throw new NullPointerException();
1313 this.corePoolSize = corePoolSize;
1314 this.maximumPoolSize = maximumPoolSize;
1315 this.workQueue = workQueue;
1316 this.keepAliveTime = unit.toNanos(keepAliveTime);
1317 this.threadFactory = threadFactory;
1318 this.handler = handler;
1319 }
1320
1321 /**
1322 * Executes the given task sometime in the future. The task
1323 * may execute in a new thread or in an existing pooled thread.
1324 *
1325 * If the task cannot be submitted for execution, either because this
1326 * executor has been shutdown or because its capacity has been reached,
1327 * the task is handled by the current {@code RejectedExecutionHandler}.
1328 *
1329 * @param command the task to execute
1330 * @throws RejectedExecutionException at discretion of
1331 * {@code RejectedExecutionHandler}, if the task
1332 * cannot be accepted for execution
1333 * @throws NullPointerException if {@code command} is null
1334 */
1335 public void execute(Runnable command) {
1336 if (command == null)
1337 throw new NullPointerException();
1338 /*
1339 * Proceed in 3 steps:
1340 *
1341 * 1. If fewer than corePoolSize threads are running, try to
1342 * start a new thread with the given command as its first
1343 * task. The call to addWorker atomically checks runState and
1344 * workerCount, and so prevents false alarms that would add
1345 * threads when it shouldn't, by returning false.
1346 *
1347 * 2. If a task can be successfully queued, then we still need
1348 * to double-check whether we should have added a thread
1349 * (because existing ones died since last checking) or that
1350 * the pool shut down since entry into this method. So we
1351 * recheck state and if necessary roll back the enqueuing if
1352 * stopped, or start a new thread if there are none.
1353 *
1354 * 3. If we cannot queue task, then we try to add a new
1355 * thread. If it fails, we know we are shut down or saturated
1356 * and so reject the task.
1357 */
1358 int c = ctl.get();
1359 if (workerCountOf(c) < corePoolSize) {
1360 if (addWorker(command, true))
1361 return;
1362 c = ctl.get();
1363 }
1364 if (isRunning(c) && workQueue.offer(command)) {
1365 int recheck = ctl.get();
1366 if (! isRunning(recheck) && remove(command))
1367 reject(command);
1368 else if (workerCountOf(recheck) == 0)
1369 addWorker(null, false);
1370 }
1371 else if (!addWorker(command, false))
1372 reject(command);
1373 }
1374
1375 /**
1376 * Initiates an orderly shutdown in which previously submitted
1377 * tasks are executed, but no new tasks will be accepted.
1378 * Invocation has no additional effect if already shut down.
1379 *
1380 * <p>This method does not wait for previously submitted tasks to
1381 * complete execution. Use {@link #awaitTermination awaitTermination}
1382 * to do that.
1383 *
1384 * @throws SecurityException {@inheritDoc}
1385 */
1386 public void shutdown() {
1387 final ReentrantLock mainLock = this.mainLock;
1388 mainLock.lock();
1389 try {
1390 checkShutdownAccess();
1391 advanceRunState(SHUTDOWN);
1392 interruptIdleWorkers();
1393 onShutdown(); // hook for ScheduledThreadPoolExecutor
1394 } finally {
1395 mainLock.unlock();
1396 }
1397 tryTerminate();
1398 }
1399
1400 /**
1401 * Attempts to stop all actively executing tasks, halts the
1402 * processing of waiting tasks, and returns a list of the tasks
1403 * that were awaiting execution. These tasks are drained (removed)
1404 * from the task queue upon return from this method.
1405 *
1406 * <p>This method does not wait for actively executing tasks to
1407 * terminate. Use {@link #awaitTermination awaitTermination} to
1408 * do that.
1409 *
1410 * <p>There are no guarantees beyond best-effort attempts to stop
1411 * processing actively executing tasks. This implementation
1412 * cancels tasks via {@link Thread#interrupt}, so any task that
1413 * fails to respond to interrupts may never terminate.
1414 *
1415 * @throws SecurityException {@inheritDoc}
1416 */
1417 public List<Runnable> shutdownNow() {
1418 List<Runnable> tasks;
1419 final ReentrantLock mainLock = this.mainLock;
1420 mainLock.lock();
1421 try {
1422 checkShutdownAccess();
1423 advanceRunState(STOP);
1424 interruptWorkers();
1425 tasks = drainQueue();
1426 } finally {
1427 mainLock.unlock();
1428 }
1429 tryTerminate();
1430 return tasks;
1431 }
1432
1433 public boolean isShutdown() {
1434 return ! isRunning(ctl.get());
1435 }
1436
1437 /**
1438 * Returns true if this executor is in the process of terminating
1439 * after {@link #shutdown} or {@link #shutdownNow} but has not
1440 * completely terminated. This method may be useful for
1441 * debugging. A return of {@code true} reported a sufficient
1442 * period after shutdown may indicate that submitted tasks have
1443 * ignored or suppressed interruption, causing this executor not
1444 * to properly terminate.
1445 *
1446 * @return true if terminating but not yet terminated
1447 */
1448 public boolean isTerminating() {
1449 int c = ctl.get();
1450 return ! isRunning(c) && runStateLessThan(c, TERMINATED);
1451 }
1452
1453 public boolean isTerminated() {
1454 return runStateAtLeast(ctl.get(), TERMINATED);
1455 }
1456
1457 public boolean awaitTermination(long timeout, TimeUnit unit)
1458 throws InterruptedException {
1459 long nanos = unit.toNanos(timeout);
1460 final ReentrantLock mainLock = this.mainLock;
1461 mainLock.lock();
1462 try {
1463 for (;;) {
1464 if (runStateAtLeast(ctl.get(), TERMINATED))
1465 return true;
1466 if (nanos <= 0)
1467 return false;
1468 nanos = termination.awaitNanos(nanos);
1469 }
1470 } finally {
1471 mainLock.unlock();
1472 }
1473 }
1474
1475 /**
1476 * Invokes {@code shutdown} when this executor is no longer
1477 * referenced and it has no threads.
1478 */
1479 protected void finalize() {
1480 shutdown();
1481 }
1482
1483 /**
1484 * Sets the thread factory used to create new threads.
1485 *
1486 * @param threadFactory the new thread factory
1487 * @throws NullPointerException if threadFactory is null
1488 * @see #getThreadFactory
1489 */
1490 public void setThreadFactory(ThreadFactory threadFactory) {
1491 if (threadFactory == null)
1492 throw new NullPointerException();
1493 this.threadFactory = threadFactory;
1494 }
1495
1496 /**
1497 * Returns the thread factory used to create new threads.
1498 *
1499 * @return the current thread factory
1500 * @see #setThreadFactory
1501 */
1502 public ThreadFactory getThreadFactory() {
1503 return threadFactory;
1504 }
1505
1506 /**
1507 * Sets a new handler for unexecutable tasks.
1508 *
1509 * @param handler the new handler
1510 * @throws NullPointerException if handler is null
1511 * @see #getRejectedExecutionHandler
1512 */
1513 public void setRejectedExecutionHandler(RejectedExecutionHandler handler) {
1514 if (handler == null)
1515 throw new NullPointerException();
1516 this.handler = handler;
1517 }
1518
1519 /**
1520 * Returns the current handler for unexecutable tasks.
1521 *
1522 * @return the current handler
1523 * @see #setRejectedExecutionHandler
1524 */
1525 public RejectedExecutionHandler getRejectedExecutionHandler() {
1526 return handler;
1527 }
1528
1529 /**
1530 * Sets the core number of threads. This overrides any value set
1531 * in the constructor. If the new value is smaller than the
1532 * current value, excess existing threads will be terminated when
1533 * they next become idle. If larger, new threads will, if needed,
1534 * be started to execute any queued tasks.
1535 *
1536 * @param corePoolSize the new core size
1537 * @throws IllegalArgumentException if {@code corePoolSize < 0}
1538 * @see #getCorePoolSize
1539 */
1540 public void setCorePoolSize(int corePoolSize) {
1541 if (corePoolSize < 0)
1542 throw new IllegalArgumentException();
1543 int delta = corePoolSize - this.corePoolSize;
1544 this.corePoolSize = corePoolSize;
1545 if (workerCountOf(ctl.get()) > corePoolSize)
1546 interruptIdleWorkers();
1547 else if (delta > 0) {
1548 // We don't really know how many new threads are "needed".
1549 // As a heuristic, prestart enough new workers (up to new
1550 // core size) to handle the current number of tasks in
1551 // queue, but stop if queue becomes empty while doing so.
1552 int k = Math.min(delta, workQueue.size());
1553 while (k-- > 0 && addWorker(null, true)) {
1554 if (workQueue.isEmpty())
1555 break;
1556 }
1557 }
1558 }
1559
1560 /**
1561 * Returns the core number of threads.
1562 *
1563 * @return the core number of threads
1564 * @see #setCorePoolSize
1565 */
1566 public int getCorePoolSize() {
1567 return corePoolSize;
1568 }
1569
1570 /**
1571 * Starts a core thread, causing it to idly wait for work. This
1572 * overrides the default policy of starting core threads only when
1573 * new tasks are executed. This method will return {@code false}
1574 * if all core threads have already been started.
1575 *
1576 * @return {@code true} if a thread was started
1577 */
1578 public boolean prestartCoreThread() {
1579 return workerCountOf(ctl.get()) < corePoolSize &&
1580 addWorker(null, true);
1581 }
1582
1583 /**
1584 * Same as prestartCoreThread except arranges that at least one
1585 * thread is started even if corePoolSize is 0.
1586 */
1587 void ensurePrestart() {
1588 int wc = workerCountOf(ctl.get());
1589 if (wc < corePoolSize)
1590 addWorker(null, true);
1591 else if (wc == 0)
1592 addWorker(null, false);
1593 }
1594
1595 /**
1596 * Starts all core threads, causing them to idly wait for work. This
1597 * overrides the default policy of starting core threads only when
1598 * new tasks are executed.
1599 *
1600 * @return the number of threads started
1601 */
1602 public int prestartAllCoreThreads() {
1603 int n = 0;
1604 while (addWorker(null, true))
1605 ++n;
1606 return n;
1607 }
1608
1609 /**
1610 * Returns true if this pool allows core threads to time out and
1611 * terminate if no tasks arrive within the keepAlive time, being
1612 * replaced if needed when new tasks arrive. When true, the same
1613 * keep-alive policy applying to non-core threads applies also to
1614 * core threads. When false (the default), core threads are never
1615 * terminated due to lack of incoming tasks.
1616 *
1617 * @return {@code true} if core threads are allowed to time out,
1618 * else {@code false}
1619 *
1620 * @since 1.6
1621 */
1622 public boolean allowsCoreThreadTimeOut() {
1623 return allowCoreThreadTimeOut;
1624 }
1625
1626 /**
1627 * Sets the policy governing whether core threads may time out and
1628 * terminate if no tasks arrive within the keep-alive time, being
1629 * replaced if needed when new tasks arrive. When false, core
1630 * threads are never terminated due to lack of incoming
1631 * tasks. When true, the same keep-alive policy applying to
1632 * non-core threads applies also to core threads. To avoid
1633 * continual thread replacement, the keep-alive time must be
1634 * greater than zero when setting {@code true}. This method
1635 * should in general be called before the pool is actively used.
1636 *
1637 * @param value {@code true} if should time out, else {@code false}
1638 * @throws IllegalArgumentException if value is {@code true}
1639 * and the current keep-alive time is not greater than zero
1640 *
1641 * @since 1.6
1642 */
1643 public void allowCoreThreadTimeOut(boolean value) {
1644 if (value && keepAliveTime <= 0)
1645 throw new IllegalArgumentException("Core threads must have nonzero keep alive times");
1646 if (value != allowCoreThreadTimeOut) {
1647 allowCoreThreadTimeOut = value;
1648 if (value)
1649 interruptIdleWorkers();
1650 }
1651 }
1652
1653 /**
1654 * Sets the maximum allowed number of threads. This overrides any
1655 * value set in the constructor. If the new value is smaller than
1656 * the current value, excess existing threads will be
1657 * terminated when they next become idle.
1658 *
1659 * @param maximumPoolSize the new maximum
1660 * @throws IllegalArgumentException if the new maximum is
1661 * less than or equal to zero, or
1662 * less than the {@linkplain #getCorePoolSize core pool size}
1663 * @see #getMaximumPoolSize
1664 */
1665 public void setMaximumPoolSize(int maximumPoolSize) {
1666 if (maximumPoolSize <= 0 || maximumPoolSize < corePoolSize)
1667 throw new IllegalArgumentException();
1668 this.maximumPoolSize = maximumPoolSize;
1669 if (workerCountOf(ctl.get()) > maximumPoolSize)
1670 interruptIdleWorkers();
1671 }
1672
1673 /**
1674 * Returns the maximum allowed number of threads.
1675 *
1676 * @return the maximum allowed number of threads
1677 * @see #setMaximumPoolSize
1678 */
1679 public int getMaximumPoolSize() {
1680 return maximumPoolSize;
1681 }
1682
1683 /**
1684 * Sets the time limit for which threads may remain idle before
1685 * being terminated. If there are more than the core number of
1686 * threads currently in the pool, after waiting this amount of
1687 * time without processing a task, excess threads will be
1688 * terminated. This overrides any value set in the constructor.
1689 *
1690 * @param time the time to wait. A time value of zero will cause
1691 * excess threads to terminate immediately after executing tasks.
1692 * @param unit the time unit of the {@code time} argument
1693 * @throws IllegalArgumentException if {@code time} less than zero or
1694 * if {@code time} is zero and {@code allowsCoreThreadTimeOut}
1695 * @see #getKeepAliveTime
1696 */
1697 public void setKeepAliveTime(long time, TimeUnit unit) {
1698 if (time < 0)
1699 throw new IllegalArgumentException();
1700 if (time == 0 && allowsCoreThreadTimeOut())
1701 throw new IllegalArgumentException("Core threads must have nonzero keep alive times");
1702 long keepAliveTime = unit.toNanos(time);
1703 long delta = keepAliveTime - this.keepAliveTime;
1704 this.keepAliveTime = keepAliveTime;
1705 if (delta < 0)
1706 interruptIdleWorkers();
1707 }
1708
1709 /**
1710 * Returns the thread keep-alive time, which is the amount of time
1711 * that threads in excess of the core pool size may remain
1712 * idle before being terminated.
1713 *
1714 * @param unit the desired time unit of the result
1715 * @return the time limit
1716 * @see #setKeepAliveTime
1717 */
1718 public long getKeepAliveTime(TimeUnit unit) {
1719 return unit.convert(keepAliveTime, TimeUnit.NANOSECONDS);
1720 }
1721
1722 /* User-level queue utilities */
1723
1724 /**
1725 * Returns the task queue used by this executor. Access to the
1726 * task queue is intended primarily for debugging and monitoring.
1727 * This queue may be in active use. Retrieving the task queue
1728 * does not prevent queued tasks from executing.
1729 *
1730 * @return the task queue
1731 */
1732 public BlockingQueue<Runnable> getQueue() {
1733 return workQueue;
1734 }
1735
1736 /**
1737 * Removes this task from the executor's internal queue if it is
1738 * present, thus causing it not to be run if it has not already
1739 * started.
1740 *
1741 * <p> This method may be useful as one part of a cancellation
1742 * scheme. It may fail to remove tasks that have been converted
1743 * into other forms before being placed on the internal queue. For
1744 * example, a task entered using {@code submit} might be
1745 * converted into a form that maintains {@code Future} status.
1746 * However, in such cases, method {@link #purge} may be used to
1747 * remove those Futures that have been cancelled.
1748 *
1749 * @param task the task to remove
1750 * @return true if the task was removed
1751 */
1752 public boolean remove(Runnable task) {
1753 boolean removed = workQueue.remove(task);
1754 tryTerminate(); // In case SHUTDOWN and now empty
1755 return removed;
1756 }
1757
1758 /**
1759 * Tries to remove from the work queue all {@link Future}
1760 * tasks that have been cancelled. This method can be useful as a
1761 * storage reclamation operation, that has no other impact on
1762 * functionality. Cancelled tasks are never executed, but may
1763 * accumulate in work queues until worker threads can actively
1764 * remove them. Invoking this method instead tries to remove them now.
1765 * However, this method may fail to remove tasks in
1766 * the presence of interference by other threads.
1767 */
1768 public void purge() {
1769 final BlockingQueue<Runnable> q = workQueue;
1770 try {
1771 Iterator<Runnable> it = q.iterator();
1772 while (it.hasNext()) {
1773 Runnable r = it.next();
1774 if (r instanceof Future<?> && ((Future<?>)r).isCancelled())
1775 it.remove();
1776 }
1777 } catch (ConcurrentModificationException fallThrough) {
1778 // Take slow path if we encounter interference during traversal.
1779 // Make copy for traversal and call remove for cancelled entries.
1780 // The slow path is more likely to be O(N*N).
1781 for (Object r : q.toArray())
1782 if (r instanceof Future<?> && ((Future<?>)r).isCancelled())
1783 q.remove(r);
1784 }
1785
1786 tryTerminate(); // In case SHUTDOWN and now empty
1787 }
1788
1789 /* Statistics */
1790
1791 /**
1792 * Returns the current number of threads in the pool.
1793 *
1794 * @return the number of threads
1795 */
1796 public int getPoolSize() {
1797 final ReentrantLock mainLock = this.mainLock;
1798 mainLock.lock();
1799 try {
1800 // Remove rare and surprising possibility of
1801 // isTerminated() && getPoolSize() > 0
1802 return runStateAtLeast(ctl.get(), TIDYING) ? 0
1803 : workers.size();
1804 } finally {
1805 mainLock.unlock();
1806 }
1807 }
1808
1809 /**
1810 * Returns the approximate number of threads that are actively
1811 * executing tasks.
1812 *
1813 * @return the number of threads
1814 */
1815 public int getActiveCount() {
1816 final ReentrantLock mainLock = this.mainLock;
1817 mainLock.lock();
1818 try {
1819 int n = 0;
1820 for (Worker w : workers)
1821 if (w.isLocked())
1822 ++n;
1823 return n;
1824 } finally {
1825 mainLock.unlock();
1826 }
1827 }
1828
1829 /**
1830 * Returns the largest number of threads that have ever
1831 * simultaneously been in the pool.
1832 *
1833 * @return the number of threads
1834 */
1835 public int getLargestPoolSize() {
1836 final ReentrantLock mainLock = this.mainLock;
1837 mainLock.lock();
1838 try {
1839 return largestPoolSize;
1840 } finally {
1841 mainLock.unlock();
1842 }
1843 }
1844
1845 /**
1846 * Returns the approximate total number of tasks that have ever been
1847 * scheduled for execution. Because the states of tasks and
1848 * threads may change dynamically during computation, the returned
1849 * value is only an approximation.
1850 *
1851 * @return the number of tasks
1852 */
1853 public long getTaskCount() {
1854 final ReentrantLock mainLock = this.mainLock;
1855 mainLock.lock();
1856 try {
1857 long n = completedTaskCount;
1858 for (Worker w : workers) {
1859 n += w.completedTasks;
1860 if (w.isLocked())
1861 ++n;
1862 }
1863 return n + workQueue.size();
1864 } finally {
1865 mainLock.unlock();
1866 }
1867 }
1868
1869 /**
1870 * Returns the approximate total number of tasks that have
1871 * completed execution. Because the states of tasks and threads
1872 * may change dynamically during computation, the returned value
1873 * is only an approximation, but one that does not ever decrease
1874 * across successive calls.
1875 *
1876 * @return the number of tasks
1877 */
1878 public long getCompletedTaskCount() {
1879 final ReentrantLock mainLock = this.mainLock;
1880 mainLock.lock();
1881 try {
1882 long n = completedTaskCount;
1883 for (Worker w : workers)
1884 n += w.completedTasks;
1885 return n;
1886 } finally {
1887 mainLock.unlock();
1888 }
1889 }
1890
1891 /**
1892 * Returns a string identifying this pool, as well as its state,
1893 * including indications of run state and estimated worker and
1894 * task counts.
1895 *
1896 * @return a string identifying this pool, as well as its state
1897 */
1898 public String toString() {
1899 long ncompleted;
1900 int nworkers, nactive;
1901 final ReentrantLock mainLock = this.mainLock;
1902 mainLock.lock();
1903 try {
1904 ncompleted = completedTaskCount;
1905 nactive = 0;
1906 nworkers = workers.size();
1907 for (Worker w : workers) {
1908 ncompleted += w.completedTasks;
1909 if (w.isLocked())
1910 ++nactive;
1911 }
1912 } finally {
1913 mainLock.unlock();
1914 }
1915 int c = ctl.get();
1916 String rs = (runStateLessThan(c, SHUTDOWN) ? "Running" :
1917 (runStateAtLeast(c, TERMINATED) ? "Terminated" :
1918 "Shutting down"));
1919 return super.toString() +
1920 "[" + rs +
1921 ", pool size = " + nworkers +
1922 ", active threads = " + nactive +
1923 ", queued tasks = " + workQueue.size() +
1924 ", completed tasks = " + ncompleted +
1925 "]";
1926 }
1927
1928 /* Extension hooks */
1929
1930 /**
1931 * Method invoked prior to executing the given Runnable in the
1932 * given thread. This method is invoked by thread {@code t} that
1933 * will execute task {@code r}, and may be used to re-initialize
1934 * ThreadLocals, or to perform logging.
1935 *
1936 * <p>This implementation does nothing, but may be customized in
1937 * subclasses. Note: To properly nest multiple overridings, subclasses
1938 * should generally invoke {@code super.beforeExecute} at the end of
1939 * this method.
1940 *
1941 * @param t the thread that will run task {@code r}
1942 * @param r the task that will be executed
1943 */
1944 protected void beforeExecute(Thread t, Runnable r) { }
1945
1946 /**
1947 * Method invoked upon completion of execution of the given Runnable.
1948 * This method is invoked by the thread that executed the task. If
1949 * non-null, the Throwable is the uncaught {@code RuntimeException}
1950 * or {@code Error} that caused execution to terminate abruptly.
1951 *
1952 * <p>This implementation does nothing, but may be customized in
1953 * subclasses. Note: To properly nest multiple overridings, subclasses
1954 * should generally invoke {@code super.afterExecute} at the
1955 * beginning of this method.
1956 *
1957 * <p><b>Note:</b> When actions are enclosed in tasks (such as
1958 * {@link FutureTask}) either explicitly or via methods such as
1959 * {@code submit}, these task objects catch and maintain
1960 * computational exceptions, and so they do not cause abrupt
1961 * termination, and the internal exceptions are <em>not</em>
1962 * passed to this method. If you would like to trap both kinds of
1963 * failures in this method, you can further probe for such cases,
1964 * as in this sample subclass that prints either the direct cause
1965 * or the underlying exception if a task has been aborted:
1966 *
1967 * <pre> {@code
1968 * class ExtendedExecutor extends ThreadPoolExecutor {
1969 * // ...
1970 * protected void afterExecute(Runnable r, Throwable t) {
1971 * super.afterExecute(r, t);
1972 * if (t == null && r instanceof Future<?>) {
1973 * try {
1974 * Object result = ((Future<?>) r).get();
1975 * } catch (CancellationException ce) {
1976 * t = ce;
1977 * } catch (ExecutionException ee) {
1978 * t = ee.getCause();
1979 * } catch (InterruptedException ie) {
1980 * Thread.currentThread().interrupt(); // ignore/reset
1981 * }
1982 * }
1983 * if (t != null)
1984 * System.out.println(t);
1985 * }
1986 * }}</pre>
1987 *
1988 * @param r the runnable that has completed
1989 * @param t the exception that caused termination, or null if
1990 * execution completed normally
1991 */
1992 protected void afterExecute(Runnable r, Throwable t) { }
1993
1994 /**
1995 * Method invoked when the Executor has terminated. Default
1996 * implementation does nothing. Note: To properly nest multiple
1997 * overridings, subclasses should generally invoke
1998 * {@code super.terminated} within this method.
1999 */
2000 protected void terminated() { }
2001
2002 /* Predefined RejectedExecutionHandlers */
2003
2004 /**
2005 * A handler for rejected tasks that runs the rejected task
2006 * directly in the calling thread of the {@code execute} method,
2007 * unless the executor has been shut down, in which case the task
2008 * is discarded.
2009 */
2010 public static class CallerRunsPolicy implements RejectedExecutionHandler {
2011 /**
2012 * Creates a {@code CallerRunsPolicy}.
2013 */
2014 public CallerRunsPolicy() { }
2015
2016 /**
2017 * Executes task r in the caller's thread, unless the executor
2018 * has been shut down, in which case the task is discarded.
2019 *
2020 * @param r the runnable task requested to be executed
2021 * @param e the executor attempting to execute this task
2022 */
2023 public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
2024 if (!e.isShutdown()) {
2025 r.run();
2026 }
2027 }
2028 }
2029
2030 /**
2031 * A handler for rejected tasks that throws a
2032 * {@code RejectedExecutionException}.
2033 */
2034 public static class AbortPolicy implements RejectedExecutionHandler {
2035 /**
2036 * Creates an {@code AbortPolicy}.
2037 */
2038 public AbortPolicy() { }
2039
2040 /**
2041 * Always throws RejectedExecutionException.
2042 *
2043 * @param r the runnable task requested to be executed
2044 * @param e the executor attempting to execute this task
2045 * @throws RejectedExecutionException always.
2046 */
2047 public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
2048 throw new RejectedExecutionException("Task " + r.toString() +
2049 " rejected from " +
2050 e.toString());
2051 }
2052 }
2053
2054 /**
2055 * A handler for rejected tasks that silently discards the
2056 * rejected task.
2057 */
2058 public static class DiscardPolicy implements RejectedExecutionHandler {
2059 /**
2060 * Creates a {@code DiscardPolicy}.
2061 */
2062 public DiscardPolicy() { }
2063
2064 /**
2065 * Does nothing, which has the effect of discarding task r.
2066 *
2067 * @param r the runnable task requested to be executed
2068 * @param e the executor attempting to execute this task
2069 */
2070 public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
2071 }
2072 }
2073
2074 /**
2075 * A handler for rejected tasks that discards the oldest unhandled
2076 * request and then retries {@code execute}, unless the executor
2077 * is shut down, in which case the task is discarded.
2078 */
2079 public static class DiscardOldestPolicy implements RejectedExecutionHandler {
2080 /**
2081 * Creates a {@code DiscardOldestPolicy} for the given executor.
2082 */
2083 public DiscardOldestPolicy() { }
2084
2085 /**
2086 * Obtains and ignores the next task that the executor
2087 * would otherwise execute, if one is immediately available,
2088 * and then retries execution of task r, unless the executor
2089 * is shut down, in which case task r is instead discarded.
2090 *
2091 * @param r the runnable task requested to be executed
2092 * @param e the executor attempting to execute this task
2093 */
2094 public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
2095 if (!e.isShutdown()) {
2096 e.getQueue().poll();
2097 e.execute(r);
2098 }
2099 }
2100 }
2101 }
线程池源码分析
(一) 创建“线程池”
下面以newFixedThreadPool()介绍线程池的创建过程。
1. newFixedThreadPool()
newFixedThreadPool()在Executors.java中定义，源码如下：
public static ExecutorService newFixedThreadPool(int nThreads) {
return new ThreadPoolExecutor(nThreads, nThreads,
0L, TimeUnit.MILLISECONDS,
new LinkedBlockingQueue<Runnable>());
}
说明：newFixedThreadPool(int nThreads)的作用是创建一个线程池，线程池的容量是nThreads。
newFixedThreadPool()在调用ThreadPoolExecutor()时，会传递一个LinkedBlockingQueue()对象，而LinkedBlockingQueue是单向链表实现的阻塞队列。在线程池中，就是通过该阻塞队列来实现"当线程池中任务数量超过允许的任务数量时，部分任务会阻塞等待"。
关于LinkedBlockingQueue的实现细节，读者可以参考"Java多线程系列--“JUC集合”08之 LinkedBlockingQueue"。
2. ThreadPoolExecutor()
ThreadPoolExecutor()在ThreadPoolExecutor.java中定义，源码如下：
public ThreadPoolExecutor(int corePoolSize,
int maximumPoolSize,
long keepAliveTime,
TimeUnit unit,
BlockingQueue<Runnable> workQueue) {
this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue,
Executors.defaultThreadFactory(), defaultHandler);
}
说明：该函数实际上是调用ThreadPoolExecutor的另外一个构造函数。该函数的源码如下：
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();
// 核心池大小
this.corePoolSize = corePoolSize;
// 最大池大小
this.maximumPoolSize = maximumPoolSize;
// 线程池的等待队列
this.workQueue = workQueue;
this.keepAliveTime = unit.toNanos(keepAliveTime);
// 线程工厂对象
this.threadFactory = threadFactory;
// 拒绝策略的句柄
this.handler = handler;
}
说明：在ThreadPoolExecutor()的构造函数中，进行的是初始化工作。
corePoolSize, maximumPoolSize, unit, keepAliveTime和workQueue这些变量的值是已知的，它们都是通过newFixedThreadPool()传递而来。下面看看threadFactory和handler对象。
2.1 ThreadFactory
线程池中的ThreadFactory是一个线程工厂，线程池创建线程都是通过线程工厂对象(threadFactory)来完成的。
上面所说的threadFactory对象，是通过 Executors.defaultThreadFactory()返回的。Executors.java中的defaultThreadFactory()源码如下：
public static ThreadFactory defaultThreadFactory() {
return new DefaultThreadFactory();
}
defaultThreadFactory()返回DefaultThreadFactory对象。Executors.java中的DefaultThreadFactory()源码如下：
static class DefaultThreadFactory implements ThreadFactory {
private static final AtomicInteger poolNumber = new AtomicInteger(1);
private final ThreadGroup group;
private final AtomicInteger threadNumber = new AtomicInteger(1);
private final String namePrefix;
DefaultThreadFactory() {
SecurityManager s = System.getSecurityManager();
group = (s != null) ? s.getThreadGroup() :
Thread.currentThread().getThreadGroup();
namePrefix = "pool-" +
poolNumber.getAndIncrement() +
"-thread-";
}
// 提供创建线程的API。
public Thread newThread(Runnable r) {
// 线程对应的任务是Runnable对象r
Thread t = new Thread(group, r,
namePrefix + threadNumber.getAndIncrement(),
0);
// 设为“非守护线程”
if (t.isDaemon())
t.setDaemon(false);
// 将优先级设为“Thread.NORM_PRIORITY”
if (t.getPriority() != Thread.NORM_PRIORITY)
t.setPriority(Thread.NORM_PRIORITY);
return t;
}
}
说明：ThreadFactory的作用就是提供创建线程的功能的线程工厂。
它是通过newThread()提供创建线程功能的，下面简单说说newThread()。newThread()创建的线程对应的任务是Runnable对象，它创建的线程都是“非守护线程”而且“线程优先级都是Thread.NORM_PRIORITY”。
2.2 RejectedExecutionHandler
handler是ThreadPoolExecutor中拒绝策略的处理句柄。所谓拒绝策略，是指将任务添加到线程池中时，线程池拒绝该任务所采取的相应策略。
线程池默认会采用的是defaultHandler策略，即AbortPolicy策略。在AbortPolicy策略中，线程池拒绝任务时会抛出异常！
defaultHandler的定义如下：
private static final RejectedExecutionHandler defaultHandler = new AbortPolicy();
AbortPolicy的源码如下：
public static class AbortPolicy implements RejectedExecutionHandler {
public AbortPolicy() { }
// 抛出异常
public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
throw new RejectedExecutionException("Task " + r.toString() +
" rejected from " +
e.toString());
}
}
(二) 添加任务到“线程池”
1. execute()
execute()定义在ThreadPoolExecutor.java中，源码如下：
public void execute(Runnable command) {
// 如果任务为null，则抛出异常。
if (command == null)
throw new NullPointerException();
// 获取ctl对应的int值。该int值保存了"线程池中任务的数量"和"线程池状态"信息
int c = ctl.get();
// 当线程池中的任务数量 < "核心池大小"时，即线程池中少于corePoolSize个任务。
// 则通过addWorker(command, true)新建一个线程，并将任务(command)添加到该线程中；然后，启动该线程从而执行任务。
if (workerCountOf(c) < corePoolSize) {
if (addWorker(command, true))
return;
c = ctl.get();
}
// 当线程池中的任务数量 >= "核心池大小"时，
// 而且，"线程池处于允许状态"时，则尝试将任务添加到阻塞队列中。
if (isRunning(c) && workQueue.offer(command)) {
// 再次确认“线程池状态”，若线程池异常终止了，则删除任务；然后通过reject()执行相应的拒绝策略的内容。
int recheck = ctl.get();
if (! isRunning(recheck) && remove(command))
reject(command);
// 否则，如果"线程池中任务数量"为0，则通过addWorker(null, false)尝试新建一个线程，新建线程对应的任务为null。
else if (workerCountOf(recheck) == 0)
addWorker(null, false);
}
// 通过addWorker(command, false)新建一个线程，并将任务(command)添加到该线程中；然后，启动该线程从而执行任务。
// 如果addWorker(command, false)执行失败，则通过reject()执行相应的拒绝策略的内容。
else if (!addWorker(command, false))
reject(command);
}
说明：execute()的作用是将任务添加到线程池中执行。它会分为3种情况进行处理：
情况1 -- 如果"线程池中任务数量" < "核心池大小"时，即线程池中少于corePoolSize个任务；此时就新建一个线程，并将该任务添加到线程中进行执行。
情况2 -- 如果"线程池中任务数量" >= "核心池大小"，并且"线程池是允许状态"；此时，则将任务添加到阻塞队列中阻塞等待。在该情况下，会再次确认"线程池的状态"，如果"第2次读到的线程池状态"和"第1次读到的线程池状态"不同，则从阻塞队列中删除该任务。
情况3 -- 非以上两种情况。在这种情况下，尝试新建一个线程，并将该任务添加到线程中进行执行。如果执行失败，则通过reject()拒绝该任务。
2. addWorker()
addWorker()的源码如下：
private boolean addWorker(Runnable firstTask, boolean core) {
retry:
// 更新"线程池状态和计数"标记，即更新ctl。
for (;;) {
// 获取ctl对应的int值。该int值保存了"线程池中任务的数量"和"线程池状态"信息
int c = ctl.get();
// 获取线程池状态。
int rs = runStateOf(c);
// 有效性检查
if (rs >= SHUTDOWN &&
! (rs == SHUTDOWN &&
firstTask == null &&
! workQueue.isEmpty()))
return false;
for (;;) {
// 获取线程池中任务的数量。
int wc = workerCountOf(c);
// 如果"线程池中任务的数量"超过限制，则返回false。
if (wc >= CAPACITY ||
wc >= (core ? corePoolSize : maximumPoolSize))
return false;
// 通过CAS函数将c的值+1。操作失败的话，则退出循环。
if (compareAndIncrementWorkerCount(c))
break retry;
c = ctl.get(); // Re-read ctl
// 检查"线程池状态"，如果与之前的状态不同，则从retry重新开始。
if (runStateOf(c) != rs)
continue retry;
// else CAS failed due to workerCount change; retry inner loop
}
}
boolean workerStarted = false;
boolean workerAdded = false;
Worker w = null;
// 添加任务到线程池，并启动任务所在的线程。
try {
final ReentrantLock mainLock = this.mainLock;
// 新建Worker，并且指定firstTask为Worker的第一个任务。
w = new Worker(firstTask);
// 获取Worker对应的线程。
final Thread t = w.thread;
if (t != null) {
// 获取锁
mainLock.lock();
try {
int c = ctl.get();
int rs = runStateOf(c);
// 再次确认"线程池状态"
if (rs < SHUTDOWN ||
(rs == SHUTDOWN && firstTask == null)) {
if (t.isAlive()) // precheck that t is startable
throw new IllegalThreadStateException();
// 将Worker对象(w)添加到"线程池的Worker集合(workers)"中
workers.add(w);
// 更新largestPoolSize
int s = workers.size();
if (s > largestPoolSize)
largestPoolSize = s;
workerAdded = true;
}
} finally {
// 释放锁
mainLock.unlock();
}
// 如果"成功将任务添加到线程池"中，则启动任务所在的线程。
if (workerAdded) {
t.start();
workerStarted = true;
}
}
} finally {
if (! workerStarted)
addWorkerFailed(w);
}
// 返回任务是否启动。
return workerStarted;
}
说明：
addWorker(Runnable firstTask, boolean core) 的作用是将任务(firstTask)添加到线程池中，并启动该任务。
core为true的话，则以corePoolSize为界限，若"线程池中已有任务数量>=corePoolSize"，则返回false；core为false的话，则以maximumPoolSize为界限，若"线程池中已有任务数量>=maximumPoolSize"，则返回false。
addWorker()会先通过for循环不断尝试更新ctl状态，ctl记录了"线程池中任务数量和线程池状态"。
更新成功之后，再通过try模块来将任务添加到线程池中，并启动任务所在的线程。
从addWorker()中，我们能清晰的发现：线程池在添加任务时，会创建任务对应的Worker对象；而一个Workder对象包含一个Thread对象。(01) 通过将Worker对象添加到"线程的workers集合"中，从而实现将任务添加到线程池中。 (02) 通过启动Worker对应的Thread线程，则执行该任务。
3. submit()
补充说明一点，submit()实际上也是通过调用execute()实现的，源码如下：
public Future<?> submit(Runnable task) {
if (task == null) throw new NullPointerException();
RunnableFuture<Void> ftask = newTaskFor(task, null);
execute(ftask);
return ftask;
}
(三) 关闭“线程池”
shutdown()的源码如下：
public void shutdown() {
final ReentrantLock mainLock = this.mainLock;
// 获取锁
mainLock.lock();
try {
// 检查终止线程池的“线程”是否有权限。
checkShutdownAccess();
// 设置线程池的状态为关闭状态。
advanceRunState(SHUTDOWN);
// 中断线程池中空闲的线程。
interruptIdleWorkers();
// 钩子函数，在ThreadPoolExecutor中没有任何动作。
onShutdown(); // hook for ScheduledThreadPoolExecutor
} finally {
// 释放锁
mainLock.unlock();
}
// 尝试终止线程池
tryTerminate();
}
说明：shutdown()的作用是关闭线程池。
线程有5种状态：新建状态，就绪状态，运行状态，阻塞状态，死亡状态。线程池也有5种状态；然而，线程池不同于线程，线程池的5种状态是：Running, SHUTDOWN, STOP, TIDYING, TERMINATED。
线程池状态定义代码如下：
private final AtomicInteger ctl = new AtomicInteger(ctlOf(RUNNING, 0));
private static final int COUNT_BITS = Integer.SIZE - 3;
private static final int CAPACITY = (1 << COUNT_BITS) - 1; private static final int RUNNING = -1 << COUNT_BITS;
private static final int SHUTDOWN = 0 << COUNT_BITS;
private static final int STOP = 1 << COUNT_BITS;
private static final int TIDYING = 2 << COUNT_BITS;
private static final int TERMINATED = 3 << COUNT_BITS;
private static int ctlOf(int rs, int wc) { return rs | wc; }
说明：
ctl是一个AtomicInteger类型的原子对象。ctl记录了"线程池中的任务数量"和"线程池状态"2个信息。
ctl共包括32位。其中，高3位表示"线程池状态"，低29位表示"线程池中的任务数量"。
RUNNING -- 对应的高3位值是111。
SHUTDOWN -- 对应的高3位值是000。
STOP -- 对应的高3位值是001。
TIDYING -- 对应的高3位值是010。
TERMINATED -- 对应的高3位值是011。
线程池各个状态之间的切换如下图所示：
1. RUNNING
(01) 状态说明：线程池处在RUNNING状态时，能够接收新任务，以及对已添加的任务进行处理。
(02) 状态切换：线程池的初始化状态是RUNNING。换句话说，线程池被一旦被创建，就处于RUNNING状态！
道理很简单，在ctl的初始化代码中(如下)，就将它初始化为RUNNING状态，并且"任务数量"初始化为0。
private final AtomicInteger ctl = new AtomicInteger(ctlOf(RUNNING, 0));
2. SHUTDOWN
(01) 状态说明：线程池处在SHUTDOWN状态时，不接收新任务，但能处理已添加的任务。
(02) 状态切换：调用线程池的shutdown()接口时，线程池由RUNNING -> SHUTDOWN。
3. STOP
(01) 状态说明：线程池处在STOP状态时，不接收新任务，不处理已添加的任务，并且会中断正在处理的任务。
(02) 状态切换：调用线程池的shutdownNow()接口时，线程池由(RUNNING or SHUTDOWN ) -> STOP。
4. TIDYING
(01) 状态说明：当所有的任务已终止，ctl记录的"任务数量"为0，线程池会变为TIDYING状态。当线程池变为TIDYING状态时，会执行钩子函数terminated()。terminated()在ThreadPoolExecutor类中是空的，若用户想在线程池变为TIDYING时，进行相应的处理；可以通过重载terminated()函数来实现。
(02) 状态切换：当线程池在SHUTDOWN状态下，阻塞队列为空并且线程池中执行的任务也为空时，就会由 SHUTDOWN -> TIDYING。
当线程池在STOP状态下，线程池中执行的任务为空时，就会由STOP -> TIDYING。
5. TERMINATED
(01) 状态说明：线程池彻底终止，就变成TERMINATED状态。
(02) 状态切换：线程池处在TIDYING状态时，执行完terminated()之后，就会由 TIDYING -> TERMINATED。
拒绝策略介绍
线程池的拒绝策略，是指当任务添加到线程池中被拒绝，而采取的处理措施。
当任务添加到线程池中之所以被拒绝，可能是由于：第一，线程池异常关闭。第二，任务数量超过线程池的最大限制。
线程池共包括4种拒绝策略，它们分别是：AbortPolicy, CallerRunsPolicy, DiscardOldestPolicy和DiscardPolicy。
AbortPolicy -- 当任务添加到线程池中被拒绝时，它将抛出 RejectedExecutionException 异常。
CallerRunsPolicy -- 当任务添加到线程池中被拒绝时，会在线程池当前正在运行的Thread线程池中处理被拒绝的任务。
DiscardOldestPolicy -- 当任务添加到线程池中被拒绝时，线程池会放弃等待队列中最旧的未处理任务，然后将被拒绝的任务添加到等待队列中。
DiscardPolicy -- 当任务添加到线程池中被拒绝时，线程池将丢弃被拒绝的任务。
线程池默认的处理策略是AbortPolicy！
拒绝策略对比和示例
下面通过示例，分别演示线程池的4种拒绝策略。
1. DiscardPolicy 示例
2. DiscardOldestPolicy 示例
3. AbortPolicy 示例
4. CallerRunsPolicy 示例
1. DiscardPolicy 示例
1 import java.lang.reflect.Field;
2 import java.util.concurrent.ArrayBlockingQueue;
3 import java.util.concurrent.ThreadPoolExecutor;
4 import java.util.concurrent.TimeUnit;
5 import java.util.concurrent.ThreadPoolExecutor.DiscardPolicy;
6
7 public class DiscardPolicyDemo {
8
9 private static final int THREADS_SIZE = 1;
10 private static final int CAPACITY = 1;
11
12 public static void main(String[] args) throws Exception {
13
14 // 创建线程池。线程池的"最大池大小"和"核心池大小"都为1(THREADS_SIZE)，"线程池"的阻塞队列容量为1(CAPACITY)。
15 ThreadPoolExecutor pool = new ThreadPoolExecutor(THREADS_SIZE, THREADS_SIZE, 0, TimeUnit.SECONDS,
16 new ArrayBlockingQueue<Runnable>(CAPACITY));
17 // 设置线程池的拒绝策略为"丢弃"
18 pool.setRejectedExecutionHandler(new ThreadPoolExecutor.DiscardPolicy());
19
20 // 新建10个任务，并将它们添加到线程池中。
21 for (int i = 0; i < 10; i++) {
22 Runnable myrun = new MyRunnable("task-"+i);
23 pool.execute(myrun);
24 }
25 // 关闭线程池
26 pool.shutdown();
27 }
28 }
29
30 class MyRunnable implements Runnable {
31 private String name;
32 public MyRunnable(String name) {
33 this.name = name;
34 }
35 @Override
36 public void run() {
37 try {
38 System.out.println(this.name + " is running.");
39 Thread.sleep(100);
40 } catch (Exception e) {
41 e.printStackTrace();
42 }
43 }
44 }
运行结果：
task-0 is running.
task-1 is running.
结果说明：线程池pool的"最大池大小"和"核心池大小"都为1(THREADS_SIZE)，这意味着"线程池能同时运行的任务数量最大只能是1"。
线程池pool的阻塞队列是ArrayBlockingQueue，ArrayBlockingQueue是一个有界的阻塞队列，ArrayBlockingQueue的容量为1。这也意味着线程池的阻塞队列只能有一个线程池阻塞等待。
根据""中分析的execute()代码可知：线程池中共运行了2个任务。第1个任务直接放到Worker中，通过线程去执行；第2个任务放到阻塞队列中等待。其他的任务都被丢弃了！
2. DiscardOldestPolicy 示例
1 import java.lang.reflect.Field;
2 import java.util.concurrent.ArrayBlockingQueue;
3 import java.util.concurrent.ThreadPoolExecutor;
4 import java.util.concurrent.TimeUnit;
5 import java.util.concurrent.ThreadPoolExecutor.DiscardOldestPolicy;
6
7 public class DiscardOldestPolicyDemo {
8
9 private static final int THREADS_SIZE = 1;
10 private static final int CAPACITY = 1;
11
12 public static void main(String[] args) throws Exception {
13
14 // 创建线程池。线程池的"最大池大小"和"核心池大小"都为1(THREADS_SIZE)，"线程池"的阻塞队列容量为1(CAPACITY)。
15 ThreadPoolExecutor pool = new ThreadPoolExecutor(THREADS_SIZE, THREADS_SIZE, 0, TimeUnit.SECONDS,
16 new ArrayBlockingQueue<Runnable>(CAPACITY));
17 // 设置线程池的拒绝策略为"DiscardOldestPolicy"
18 pool.setRejectedExecutionHandler(new ThreadPoolExecutor.DiscardOldestPolicy());
19
20 // 新建10个任务，并将它们添加到线程池中。
21 for (int i = 0; i < 10; i++) {
22 Runnable myrun = new MyRunnable("task-"+i);
23 pool.execute(myrun);
24 }
25 // 关闭线程池
26 pool.shutdown();
27 }
28 }
29
30 class MyRunnable implements Runnable {
31 private String name;
32 public MyRunnable(String name) {
33 this.name = name;
34 }
35 @Override
36 public void run() {
37 try {
38 System.out.println(this.name + " is running.");
39 Thread.sleep(200);
40 } catch (Exception e) {
41 e.printStackTrace();
42 }
43 }
44 }
运行结果：
task-0 is running.
task-9 is running.
结果说明：将"线程池的拒绝策略"由DiscardPolicy修改为DiscardOldestPolicy之后，当有任务添加到线程池被拒绝时，线程池会丢弃阻塞队列中末尾的任务，然后将被拒绝的任务添加到末尾。
3. AbortPolicy 示例
1 import java.lang.reflect.Field;
2 import java.util.concurrent.ArrayBlockingQueue;
3 import java.util.concurrent.ThreadPoolExecutor;
4 import java.util.concurrent.TimeUnit;
5 import java.util.concurrent.ThreadPoolExecutor.AbortPolicy;
6 import java.util.concurrent.RejectedExecutionException;
7
8 public class AbortPolicyDemo {
9
10 private static final int THREADS_SIZE = 1;
11 private static final int CAPACITY = 1;
12
13 public static void main(String[] args) throws Exception {
14
15 // 创建线程池。线程池的"最大池大小"和"核心池大小"都为1(THREADS_SIZE)，"线程池"的阻塞队列容量为1(CAPACITY)。
16 ThreadPoolExecutor pool = new ThreadPoolExecutor(THREADS_SIZE, THREADS_SIZE, 0, TimeUnit.SECONDS,
17 new ArrayBlockingQueue<Runnable>(CAPACITY));
18 // 设置线程池的拒绝策略为"抛出异常"
19 pool.setRejectedExecutionHandler(new ThreadPoolExecutor.AbortPolicy());
20
21 try {
22
23 // 新建10个任务，并将它们添加到线程池中。
24 for (int i = 0; i < 10; i++) {
25 Runnable myrun = new MyRunnable("task-"+i);
26 pool.execute(myrun);
27 }
28 } catch (RejectedExecutionException e) {
29 e.printStackTrace();
30 // 关闭线程池
31 pool.shutdown();
32 }
33 }
34 }
35
36 class MyRunnable implements Runnable {
37 private String name;
38 public MyRunnable(String name) {
39 this.name = name;
40 }
41 @Override
42 public void run() {
43 try {
44 System.out.println(this.name + " is running.");
45 Thread.sleep(200);
46 } catch (Exception e) {
47 e.printStackTrace();
48 }
49 }
50 }
(某一次)运行结果：
java.util.concurrent.RejectedExecutionException
at java.util.concurrent.ThreadPoolExecutor$AbortPolicy.rejectedExecution(ThreadPoolExecutor.java:1774)
at java.util.concurrent.ThreadPoolExecutor.reject(ThreadPoolExecutor.java:768)
at java.util.concurrent.ThreadPoolExecutor.execute(ThreadPoolExecutor.java:656)
at AbortPolicyDemo.main(AbortPolicyDemo.java:27)
task-0 is running.
task-1 is running.
结果说明：将"线程池的拒绝策略"由DiscardPolicy修改为AbortPolicy之后，当有任务添加到线程池被拒绝时，会抛出RejectedExecutionException。
4. CallerRunsPolicy 示例
1 import java.lang.reflect.Field;
2 import java.util.concurrent.ArrayBlockingQueue;
3 import java.util.concurrent.ThreadPoolExecutor;
4 import java.util.concurrent.TimeUnit;
5 import java.util.concurrent.ThreadPoolExecutor.CallerRunsPolicy;
6
7 public class CallerRunsPolicyDemo {
8
9 private static final int THREADS_SIZE = 1;
10 private static final int CAPACITY = 1;
11
12 public static void main(String[] args) throws Exception {
13
14 // 创建线程池。线程池的"最大池大小"和"核心池大小"都为1(THREADS_SIZE)，"线程池"的阻塞队列容量为1(CAPACITY)。
15 ThreadPoolExecutor pool = new ThreadPoolExecutor(THREADS_SIZE, THREADS_SIZE, 0, TimeUnit.SECONDS,
16 new ArrayBlockingQueue<Runnable>(CAPACITY));
17 // 设置线程池的拒绝策略为"CallerRunsPolicy"
18 pool.setRejectedExecutionHandler(new ThreadPoolExecutor.CallerRunsPolicy());
19
20 // 新建10个任务，并将它们添加到线程池中。
21 for (int i = 0; i < 10; i++) {
22 Runnable myrun = new MyRunnable("task-"+i);
23 pool.execute(myrun);
24 }
25
26 // 关闭线程池
27 pool.shutdown();
28 }
29 }
30
31 class MyRunnable implements Runnable {
32 private String name;
33 public MyRunnable(String name) {
34 this.name = name;
35 }
36 @Override
37 public void run() {
38 try {
39 System.out.println(this.name + " is running.");
40 Thread.sleep(100);
41 } catch (Exception e) {
42 e.printStackTrace();
43 }
44 }
45 }
(某一次)运行结果：
task-2 is running.
task-3 is running.
task-4 is running.
task-5 is running.
task-6 is running.
task-7 is running.
task-8 is running.
task-9 is running.
task-0 is running.
task-1 is running.
结果说明：将"线程池的拒绝策略"由DiscardPolicy修改为CallerRunsPolicy之后，当有任务添加到线程池被拒绝时，线程池会将被拒绝的任务添加到"线程池正在运行的线程"中取运行。
Java并发包--线程池原理的更多相关文章
- java并发包&线程池原理分析&锁的深度化
java并发包&线程池原理分析&锁的深度化 并发包 同步容器类 Vector与ArrayList区别 1.ArrayList是最常用的List实现类,内部是通过数组实现的, ...
- Java并发包——线程池
Java并发包——线程池 摘要:本文主要学习了Java并发包中的线程池. 部分内容来自以下博客: https://www.cnblogs.com/dolphin0520/p/3932921.html ...
- Java并发包线程池之Executors、ExecutorCompletionService工具类
前言 前面介绍了Java并发包提供的三种线程池,它们用处各不相同,接下来介绍一些工具类,对这三种线程池的使用. Executors Executors是JDK1.5就开始存在是一个线程池工具类,它定义 ...
- Java并发包线程池之ForkJoinPool即ForkJoin框架(二)
前言 前面介绍了ForkJoinPool相关的两个类ForkJoinTask.ForkJoinWorkerThread,现在开始了解ForkJoinPool.ForkJoinPool也是实现了Exec ...
- Java并发包线程池之ForkJoinPool即ForkJoin框架(一)
前言 这是Java并发包提供的最后一个线程池实现,也是最复杂的一个线程池.针对这一部分的代码太复杂,由于目前理解有限,只做简单介绍.通常大家说的Fork/Join框架其实就是指由ForkJoinPoo ...
- Java并发包线程池之ScheduledThreadPoolExecutor
前言 它是一种可以安排在给定的延迟之后执行一次或周期性执行任务的ThreadPoolExecutor.因为它继承了ThreadPoolExecutor, 当然也具有处理普通Runnable.Calla ...
- java多线程：线程池原理、阻塞队列
一.线程池定义和使用 jdk 1.5 之后就引入了线程池. 1.1 定义 从上面的空间切换看得出来,线程是稀缺资源,它的创建与销毁是一个相对偏重且耗资源的操作,而Java线程依赖于内核线程,创建线程需 ...
- Java并发包--线程池框架
转载请注明出处:http://www.cnblogs.com/skywang12345/p/3509903.html 线程池架构图 线程池的架构图如下: 1. Executor 它是"执行者 ...
- Java多线程系列——线程池原理之 ThreadPoolExecutor
ThreadPoolExecutor 简介 ThreadPoolExecutor 是线程池类. 通俗的讲,它是一个存放一定数量线程的线程集合.线程池允许多个线程同时运行,同时运行的线程数量就是这个线程 ...
随机推荐
- 冲刺Noip2017模拟赛8 解题报告——五十岚芒果酱
1.鼎纹 [问题描述] 据说鼎纹的 种制造 式是 铜模印出来的,这是我国古代劳动 智慧 的结晶.铜模印过的地 ,会留下深深的印记,经过时间的炼化,洗 练成历史的遗存. 聪明的古代劳动人民拥有一个 a ...
- python lanbda匿名函数（20）
在python开发中常规的函数在调用之前都需要先声明,而python还有一种匿名函数,有速写函数的功能并且匿名函数不需要声明也没有函数名字,完全不需要担心函数名冲突,具体的妙用还需要从实战练习中多多积 ...
- C++编译器、链接器工作原理
1 几个基本概念 编译:编译器对源文件的编译过程,就是将源文件中的文本形式代码翻译为机器语言形式的目标文件的过程,此过程中会有一系列语法检查.指令优化等,生成目标(OBJ)文件. 编译单元:每一个CP ...
- BS
- linux系统下，在用户空间应用程序中模拟发送系统键盘事件
Linux 有自己的 input 子系统,可以统一管理鼠标和键盘事件. 如果想模拟键盘事件,但是系统没有键盘设备该如何是好? 基于输入子系统实现的 input 可以方便的在用户空间模拟鼠标和键盘事件. ...
- [转帖]PostgreSQL 昨天,今天和明天
PostgreSQL 昨天,今天和明天 http://www.postgres.cn/v2/news/viewone/1/52 原作者:何伟平(laser) 创作时间:2005-01-15 11:44 ...
- Word 查找替换高级玩法系列之 -- 用替换功能删除空白区域
当你遇到Word文档中时不时的出现一些空白区域的时候会怎么办呢?一个个删除吗?NO!NO!NO!!!那样也太慢了!仅使用替换功能就可以帮你一步搞定它! 下面这篇文档中含有半角空格.全角空格.不间断空格 ...
- Python基础运算符（算数、比较、赋值、逻辑、成员）
Python运算符有(算数运算符.比较运算符.赋值运算符.逻辑运算符.位运算符.成员运算符.身份运算符): 本程序包含算数.比较.赋值.逻辑.成员运算符. 1.运算符测试 #!/usr/bin/pyt ...
- Centos6.5镜像下载
CentOS6.5系统安装 1.首先打开网易开源镜像站: http://mirrors.163.com/ 当然,大家也可以使用阿里开源镜像站:http://mirrors.aliyun.com/ 2. ...
- vscode配置phpxdebug
打debug还是很有必要的,以前嫌麻烦,现在觉得,通过debug可以看自己写的代码的执行的逻辑,更容易理清别人代码的逻辑. 步骤: 下载phpdebug插件 查看自己的php版本信息,下载对应的deb ...