Java并发包--线程池原理

时间:2023-03-08 16:02:25

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。

Java并发包--线程池原理
 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 }
Java并发包--线程池原理

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

Java并发包--线程池原理
  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 }
Java并发包--线程池原理

ThreadPoolExecutor完整源码

Java并发包--线程池原理
   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 }
Java并发包--线程池原理

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:

Java并发包--线程池原理
public ThreadPoolExecutor(int corePoolSize,
int maximumPoolSize,
long keepAliveTime,
TimeUnit unit,
BlockingQueue<Runnable> workQueue) {
this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue,
Executors.defaultThreadFactory(), defaultHandler);
}
Java并发包--线程池原理

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:

Java并发包--线程池原理
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;
}
Java并发包--线程池原理

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:

Java并发包--线程池原理
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;
}
}
Java并发包--线程池原理

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:

Java并发包--线程池原理
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());
}
}
Java并发包--线程池原理

(二) 添加任务到“线程池”

1. execute()

execute()定义在ThreadPoolExecutor.java中,源码如下:

Java并发包--线程池原理
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);
}
Java并发包--线程池原理

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:

Java并发包--线程池原理
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;
}
Java并发包--线程池原理

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:

Java并发包--线程池原理
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();
}
Java并发包--线程池原理

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:

Java并发包--线程池原理
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; }
Java并发包--线程池原理

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:

Java并发包--线程池原理

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:AbortPolicyCallerRunsPolicyDiscardOldestPolicyDiscardPolicy

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

Java并发包--线程池原理
 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 }
Java并发包--线程池原理

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*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

Java并发包--线程池原理
 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 }
Java并发包--线程池原理

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

Java并发包--线程池原理
 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 }
Java并发包--线程池原理

(某一次)运行结果

Java并发包--线程池原理
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.
Java并发包--线程池原理

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

Java并发包--线程池原理
 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 }
Java并发包--线程池原理

(某一次)运行结果

Java并发包--线程池原理
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.
Java并发包--线程池原理

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。