• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Written by Doug Lea with assistance from members of JCP JSR-166
3  * Expert Group and released to the public domain, as explained at
4  * http://creativecommons.org/publicdomain/zero/1.0/
5  */
6 
7 package java.util.concurrent;
8 
9 import java.util.concurrent.locks.AbstractQueuedSynchronizer;
10 import java.util.concurrent.locks.Condition;
11 import java.util.concurrent.locks.ReentrantLock;
12 import java.util.concurrent.atomic.AtomicInteger;
13 import java.util.*;
14 
15 // BEGIN android-note
16 // removed security manager docs
17 // END android-note
18 
19 /**
20  * An {@link ExecutorService} that executes each submitted task using
21  * one of possibly several pooled threads, normally configured
22  * using {@link Executors} factory methods.
23  *
24  * <p>Thread pools address two different problems: they usually
25  * provide improved performance when executing large numbers of
26  * asynchronous tasks, due to reduced per-task invocation overhead,
27  * and they provide a means of bounding and managing the resources,
28  * including threads, consumed when executing a collection of tasks.
29  * Each {@code ThreadPoolExecutor} also maintains some basic
30  * statistics, such as the number of completed tasks.
31  *
32  * <p>To be useful across a wide range of contexts, this class
33  * provides many adjustable parameters and extensibility
34  * hooks. However, programmers are urged to use the more convenient
35  * {@link Executors} factory methods {@link
36  * Executors#newCachedThreadPool} (unbounded thread pool, with
37  * automatic thread reclamation), {@link Executors#newFixedThreadPool}
38  * (fixed size thread pool) and {@link
39  * Executors#newSingleThreadExecutor} (single background thread), that
40  * preconfigure settings for the most common usage
41  * scenarios. Otherwise, use the following guide when manually
42  * configuring and tuning this class:
43  *
44  * <dl>
45  *
46  * <dt>Core and maximum pool sizes</dt>
47  *
48  * <dd>A {@code ThreadPoolExecutor} will automatically adjust the
49  * pool size (see {@link #getPoolSize})
50  * according to the bounds set by
51  * corePoolSize (see {@link #getCorePoolSize}) and
52  * maximumPoolSize (see {@link #getMaximumPoolSize}).
53  *
54  * When a new task is submitted in method {@link #execute(Runnable)},
55  * and fewer than corePoolSize threads are running, a new thread is
56  * created to handle the request, even if other worker threads are
57  * idle.  If there are more than corePoolSize but less than
58  * maximumPoolSize threads running, a new thread will be created only
59  * if the queue is full.  By setting corePoolSize and maximumPoolSize
60  * the same, you create a fixed-size thread pool. By setting
61  * maximumPoolSize to an essentially unbounded value such as {@code
62  * Integer.MAX_VALUE}, you allow the pool to accommodate an arbitrary
63  * number of concurrent tasks. Most typically, core and maximum pool
64  * sizes are set only upon construction, but they may also be changed
65  * dynamically using {@link #setCorePoolSize} and {@link
66  * #setMaximumPoolSize}. </dd>
67  *
68  * <dt>On-demand construction</dt>
69  *
70  * <dd>By default, even core threads are initially created and
71  * started only when new tasks arrive, but this can be overridden
72  * dynamically using method {@link #prestartCoreThread} or {@link
73  * #prestartAllCoreThreads}.  You probably want to prestart threads if
74  * you construct the pool with a non-empty queue. </dd>
75  *
76  * <dt>Creating new threads</dt>
77  *
78  * <dd>New threads are created using a {@link ThreadFactory}.  If not
79  * otherwise specified, a {@link Executors#defaultThreadFactory} is
80  * used, that creates threads to all be in the same {@link
81  * ThreadGroup} and with the same {@code NORM_PRIORITY} priority and
82  * non-daemon status. By supplying a different ThreadFactory, you can
83  * alter the thread's name, thread group, priority, daemon status,
84  * etc. If a {@code ThreadFactory} fails to create a thread when asked
85  * by returning null from {@code newThread}, the executor will
86  * continue, but might not be able to execute any tasks.</dd>
87  *
88  * <dt>Keep-alive times</dt>
89  *
90  * <dd>If the pool currently has more than corePoolSize threads,
91  * excess threads will be terminated if they have been idle for more
92  * than the keepAliveTime (see {@link #getKeepAliveTime(TimeUnit)}).
93  * This provides a means of reducing resource consumption when the
94  * pool is not being actively used. If the pool becomes more active
95  * later, new threads will be constructed. This parameter can also be
96  * changed dynamically using method {@link #setKeepAliveTime(long,
97  * TimeUnit)}.  Using a value of {@code Long.MAX_VALUE} {@link
98  * TimeUnit#NANOSECONDS} effectively disables idle threads from ever
99  * terminating prior to shut down. By default, the keep-alive policy
100  * applies only when there are more than corePoolSize threads. But
101  * method {@link #allowCoreThreadTimeOut(boolean)} can be used to
102  * apply this time-out policy to core threads as well, so long as the
103  * keepAliveTime value is non-zero. </dd>
104  *
105  * <dt>Queuing</dt>
106  *
107  * <dd>Any {@link BlockingQueue} may be used to transfer and hold
108  * submitted tasks.  The use of this queue interacts with pool sizing:
109  *
110  * <ul>
111  *
112  * <li> If fewer than corePoolSize threads are running, the Executor
113  * always prefers adding a new thread
114  * rather than queuing.</li>
115  *
116  * <li> If corePoolSize or more threads are running, the Executor
117  * always prefers queuing a request rather than adding a new
118  * thread.</li>
119  *
120  * <li> If a request cannot be queued, a new thread is created unless
121  * this would exceed maximumPoolSize, in which case, the task will be
122  * rejected.</li>
123  *
124  * </ul>
125  *
126  * There are three general strategies for queuing:
127  * <ol>
128  *
129  * <li> <em> Direct handoffs.</em> A good default choice for a work
130  * queue is a {@link SynchronousQueue} that hands off tasks to threads
131  * without otherwise holding them. Here, an attempt to queue a task
132  * will fail if no threads are immediately available to run it, so a
133  * new thread will be constructed. This policy avoids lockups when
134  * handling sets of requests that might have internal dependencies.
135  * Direct handoffs generally require unbounded maximumPoolSizes to
136  * avoid rejection of new submitted tasks. This in turn admits the
137  * possibility of unbounded thread growth when commands continue to
138  * arrive on average faster than they can be processed.  </li>
139  *
140  * <li><em> Unbounded queues.</em> Using an unbounded queue (for
141  * example a {@link LinkedBlockingQueue} without a predefined
142  * capacity) will cause new tasks to wait in the queue when all
143  * corePoolSize threads are busy. Thus, no more than corePoolSize
144  * threads will ever be created. (And the value of the maximumPoolSize
145  * therefore doesn't have any effect.)  This may be appropriate when
146  * each task is completely independent of others, so tasks cannot
147  * affect each others execution; for example, in a web page server.
148  * While this style of queuing can be useful in smoothing out
149  * transient bursts of requests, it admits the possibility of
150  * unbounded work queue growth when commands continue to arrive on
151  * average faster than they can be processed.  </li>
152  *
153  * <li><em>Bounded queues.</em> A bounded queue (for example, an
154  * {@link ArrayBlockingQueue}) helps prevent resource exhaustion when
155  * used with finite maximumPoolSizes, but can be more difficult to
156  * tune and control.  Queue sizes and maximum pool sizes may be traded
157  * off for each other: Using large queues and small pools minimizes
158  * CPU usage, OS resources, and context-switching overhead, but can
159  * lead to artificially low throughput.  If tasks frequently block (for
160  * example if they are I/O bound), a system may be able to schedule
161  * time for more threads than you otherwise allow. Use of small queues
162  * generally requires larger pool sizes, which keeps CPUs busier but
163  * may encounter unacceptable scheduling overhead, which also
164  * decreases throughput.  </li>
165  *
166  * </ol>
167  *
168  * </dd>
169  *
170  * <dt>Rejected tasks</dt>
171  *
172  * <dd>New tasks submitted in method {@link #execute(Runnable)} will be
173  * <em>rejected</em> when the Executor has been shut down, and also when
174  * the Executor uses finite bounds for both maximum threads and work queue
175  * capacity, and is saturated.  In either case, the {@code execute} method
176  * invokes the {@link
177  * RejectedExecutionHandler#rejectedExecution(Runnable, ThreadPoolExecutor)}
178  * method of its {@link RejectedExecutionHandler}.  Four predefined handler
179  * policies are provided:
180  *
181  * <ol>
182  *
183  * <li> In the default {@link ThreadPoolExecutor.AbortPolicy}, the
184  * handler throws a runtime {@link RejectedExecutionException} upon
185  * rejection. </li>
186  *
187  * <li> In {@link ThreadPoolExecutor.CallerRunsPolicy}, the thread
188  * that invokes {@code execute} itself runs the task. This provides a
189  * simple feedback control mechanism that will slow down the rate that
190  * new tasks are submitted. </li>
191  *
192  * <li> In {@link ThreadPoolExecutor.DiscardPolicy}, a task that
193  * cannot be executed is simply dropped.  </li>
194  *
195  * <li>In {@link ThreadPoolExecutor.DiscardOldestPolicy}, if the
196  * executor is not shut down, the task at the head of the work queue
197  * is dropped, and then execution is retried (which can fail again,
198  * causing this to be repeated.) </li>
199  *
200  * </ol>
201  *
202  * It is possible to define and use other kinds of {@link
203  * RejectedExecutionHandler} classes. Doing so requires some care
204  * especially when policies are designed to work only under particular
205  * capacity or queuing policies. </dd>
206  *
207  * <dt>Hook methods</dt>
208  *
209  * <dd>This class provides {@code protected} overridable
210  * {@link #beforeExecute(Thread, Runnable)} and
211  * {@link #afterExecute(Runnable, Throwable)} methods that are called
212  * before and after execution of each task.  These can be used to
213  * manipulate the execution environment; for example, reinitializing
214  * ThreadLocals, gathering statistics, or adding log entries.
215  * Additionally, method {@link #terminated} can be overridden to perform
216  * any special processing that needs to be done once the Executor has
217  * fully terminated.
218  *
219  * <p>If hook or callback methods throw exceptions, internal worker
220  * threads may in turn fail and abruptly terminate.</dd>
221  *
222  * <dt>Queue maintenance</dt>
223  *
224  * <dd>Method {@link #getQueue()} allows access to the work queue
225  * for purposes of monitoring and debugging.  Use of this method for
226  * any other purpose is strongly discouraged.  Two supplied methods,
227  * {@link #remove(Runnable)} and {@link #purge} are available to
228  * assist in storage reclamation when large numbers of queued tasks
229  * become cancelled.</dd>
230  *
231  * <dt>Finalization</dt>
232  *
233  * <dd>A pool that is no longer referenced in a program <em>AND</em>
234  * has no remaining threads will be {@code shutdown} automatically. If
235  * you would like to ensure that unreferenced pools are reclaimed even
236  * if users forget to call {@link #shutdown}, then you must arrange
237  * that unused threads eventually die, by setting appropriate
238  * keep-alive times, using a lower bound of zero core threads and/or
239  * setting {@link #allowCoreThreadTimeOut(boolean)}.  </dd>
240  *
241  * </dl>
242  *
243  * <p><b>Extension example</b>. Most extensions of this class
244  * override one or more of the protected hook methods. For example,
245  * here is a subclass that adds a simple pause/resume feature:
246  *
247  *  <pre> {@code
248  * class PausableThreadPoolExecutor extends ThreadPoolExecutor {
249  *   private boolean isPaused;
250  *   private ReentrantLock pauseLock = new ReentrantLock();
251  *   private Condition unpaused = pauseLock.newCondition();
252  *
253  *   public PausableThreadPoolExecutor(...) { super(...); }
254  *
255  *   protected void beforeExecute(Thread t, Runnable r) {
256  *     super.beforeExecute(t, r);
257  *     pauseLock.lock();
258  *     try {
259  *       while (isPaused) unpaused.await();
260  *     } catch (InterruptedException ie) {
261  *       t.interrupt();
262  *     } finally {
263  *       pauseLock.unlock();
264  *     }
265  *   }
266  *
267  *   public void pause() {
268  *     pauseLock.lock();
269  *     try {
270  *       isPaused = true;
271  *     } finally {
272  *       pauseLock.unlock();
273  *     }
274  *   }
275  *
276  *   public void resume() {
277  *     pauseLock.lock();
278  *     try {
279  *       isPaused = false;
280  *       unpaused.signalAll();
281  *     } finally {
282  *       pauseLock.unlock();
283  *     }
284  *   }
285  * }}</pre>
286  *
287  * @since 1.5
288  * @author Doug Lea
289  */
290 public class ThreadPoolExecutor extends AbstractExecutorService {
291     /**
292      * The main pool control state, ctl, is an atomic integer packing
293      * two conceptual fields
294      *   workerCount, indicating the effective number of threads
295      *   runState,    indicating whether running, shutting down etc
296      *
297      * In order to pack them into one int, we limit workerCount to
298      * (2^29)-1 (about 500 million) threads rather than (2^31)-1 (2
299      * billion) otherwise representable. If this is ever an issue in
300      * the future, the variable can be changed to be an AtomicLong,
301      * and the shift/mask constants below adjusted. But until the need
302      * arises, this code is a bit faster and simpler using an int.
303      *
304      * The workerCount is the number of workers that have been
305      * permitted to start and not permitted to stop.  The value may be
306      * transiently different from the actual number of live threads,
307      * for example when a ThreadFactory fails to create a thread when
308      * asked, and when exiting threads are still performing
309      * bookkeeping before terminating. The user-visible pool size is
310      * reported as the current size of the workers set.
311      *
312      * The runState provides the main lifecycle control, taking on values:
313      *
314      *   RUNNING:  Accept new tasks and process queued tasks
315      *   SHUTDOWN: Don't accept new tasks, but process queued tasks
316      *   STOP:     Don't accept new tasks, don't process queued tasks,
317      *             and interrupt in-progress tasks
318      *   TIDYING:  All tasks have terminated, workerCount is zero,
319      *             the thread transitioning to state TIDYING
320      *             will run the terminated() hook method
321      *   TERMINATED: terminated() has completed
322      *
323      * The numerical order among these values matters, to allow
324      * ordered comparisons. The runState monotonically increases over
325      * time, but need not hit each state. The transitions are:
326      *
327      * RUNNING -> SHUTDOWN
328      *    On invocation of shutdown(), perhaps implicitly in finalize()
329      * (RUNNING or SHUTDOWN) -> STOP
330      *    On invocation of shutdownNow()
331      * SHUTDOWN -> TIDYING
332      *    When both queue and pool are empty
333      * STOP -> TIDYING
334      *    When pool is empty
335      * TIDYING -> TERMINATED
336      *    When the terminated() hook method has completed
337      *
338      * Threads waiting in awaitTermination() will return when the
339      * state reaches TERMINATED.
340      *
341      * Detecting the transition from SHUTDOWN to TIDYING is less
342      * straightforward than you'd like because the queue may become
343      * empty after non-empty and vice versa during SHUTDOWN state, but
344      * we can only terminate if, after seeing that it is empty, we see
345      * that workerCount is 0 (which sometimes entails a recheck -- see
346      * below).
347      */
348     private final AtomicInteger ctl = new AtomicInteger(ctlOf(RUNNING, 0));
349     private static final int COUNT_BITS = Integer.SIZE - 3;
350     private static final int CAPACITY   = (1 << COUNT_BITS) - 1;
351 
352     // runState is stored in the high-order bits
353     private static final int RUNNING    = -1 << COUNT_BITS;
354     private static final int SHUTDOWN   =  0 << COUNT_BITS;
355     private static final int STOP       =  1 << COUNT_BITS;
356     private static final int TIDYING    =  2 << COUNT_BITS;
357     private static final int TERMINATED =  3 << COUNT_BITS;
358 
359     // Packing and unpacking ctl
runStateOf(int c)360     private static int runStateOf(int c)     { return c & ~CAPACITY; }
workerCountOf(int c)361     private static int workerCountOf(int c)  { return c & CAPACITY; }
ctlOf(int rs, int wc)362     private static int ctlOf(int rs, int wc) { return rs | wc; }
363 
364     /*
365      * Bit field accessors that don't require unpacking ctl.
366      * These depend on the bit layout and on workerCount being never negative.
367      */
368 
runStateLessThan(int c, int s)369     private static boolean runStateLessThan(int c, int s) {
370         return c < s;
371     }
372 
runStateAtLeast(int c, int s)373     private static boolean runStateAtLeast(int c, int s) {
374         return c >= s;
375     }
376 
isRunning(int c)377     private static boolean isRunning(int c) {
378         return c < SHUTDOWN;
379     }
380 
381     /**
382      * Attempts to CAS-increment the workerCount field of ctl.
383      */
compareAndIncrementWorkerCount(int expect)384     private boolean compareAndIncrementWorkerCount(int expect) {
385         return ctl.compareAndSet(expect, expect + 1);
386     }
387 
388     /**
389      * Attempts to CAS-decrement the workerCount field of ctl.
390      */
compareAndDecrementWorkerCount(int expect)391     private boolean compareAndDecrementWorkerCount(int expect) {
392         return ctl.compareAndSet(expect, expect - 1);
393     }
394 
395     /**
396      * Decrements the workerCount field of ctl. This is called only on
397      * abrupt termination of a thread (see processWorkerExit). Other
398      * decrements are performed within getTask.
399      */
decrementWorkerCount()400     private void decrementWorkerCount() {
401         do {} while (! compareAndDecrementWorkerCount(ctl.get()));
402     }
403 
404     /**
405      * The queue used for holding tasks and handing off to worker
406      * threads.  We do not require that workQueue.poll() returning
407      * null necessarily means that workQueue.isEmpty(), so rely
408      * solely on isEmpty to see if the queue is empty (which we must
409      * do for example when deciding whether to transition from
410      * SHUTDOWN to TIDYING).  This accommodates special-purpose
411      * queues such as DelayQueues for which poll() is allowed to
412      * return null even if it may later return non-null when delays
413      * expire.
414      */
415     private final BlockingQueue<Runnable> workQueue;
416 
417     /**
418      * Lock held on access to workers set and related bookkeeping.
419      * While we could use a concurrent set of some sort, it turns out
420      * to be generally preferable to use a lock. Among the reasons is
421      * that this serializes interruptIdleWorkers, which avoids
422      * unnecessary interrupt storms, especially during shutdown.
423      * Otherwise exiting threads would concurrently interrupt those
424      * that have not yet interrupted. It also simplifies some of the
425      * associated statistics bookkeeping of largestPoolSize etc. We
426      * also hold mainLock on shutdown and shutdownNow, for the sake of
427      * ensuring workers set is stable while separately checking
428      * permission to interrupt and actually interrupting.
429      */
430     private final ReentrantLock mainLock = new ReentrantLock();
431 
432     /**
433      * Set containing all worker threads in pool. Accessed only when
434      * holding mainLock.
435      */
436     private final HashSet<Worker> workers = new HashSet<Worker>();
437 
438     /**
439      * Wait condition to support awaitTermination
440      */
441     private final Condition termination = mainLock.newCondition();
442 
443     /**
444      * Tracks largest attained pool size. Accessed only under
445      * mainLock.
446      */
447     private int largestPoolSize;
448 
449     /**
450      * Counter for completed tasks. Updated only on termination of
451      * worker threads. Accessed only under mainLock.
452      */
453     private long completedTaskCount;
454 
455     /*
456      * All user control parameters are declared as volatiles so that
457      * ongoing actions are based on freshest values, but without need
458      * for locking, since no internal invariants depend on them
459      * changing synchronously with respect to other actions.
460      */
461 
462     /**
463      * Factory for new threads. All threads are created using this
464      * factory (via method addWorker).  All callers must be prepared
465      * for addWorker to fail, which may reflect a system or user's
466      * policy limiting the number of threads.  Even though it is not
467      * treated as an error, failure to create threads may result in
468      * new tasks being rejected or existing ones remaining stuck in
469      * the queue.
470      *
471      * We go further and preserve pool invariants even in the face of
472      * errors such as OutOfMemoryError, that might be thrown while
473      * trying to create threads.  Such errors are rather common due to
474      * the need to allocate a native stack in Thread.start, and users
475      * will want to perform clean pool shutdown to clean up.  There
476      * will likely be enough memory available for the cleanup code to
477      * complete without encountering yet another OutOfMemoryError.
478      */
479     private volatile ThreadFactory threadFactory;
480 
481     /**
482      * Handler called when saturated or shutdown in execute.
483      */
484     private volatile RejectedExecutionHandler handler;
485 
486     /**
487      * Timeout in nanoseconds for idle threads waiting for work.
488      * Threads use this timeout when there are more than corePoolSize
489      * present or if allowCoreThreadTimeOut. Otherwise they wait
490      * forever for new work.
491      */
492     private volatile long keepAliveTime;
493 
494     /**
495      * If false (default), core threads stay alive even when idle.
496      * If true, core threads use keepAliveTime to time out waiting
497      * for work.
498      */
499     private volatile boolean allowCoreThreadTimeOut;
500 
501     /**
502      * Core pool size is the minimum number of workers to keep alive
503      * (and not allow to time out etc) unless allowCoreThreadTimeOut
504      * is set, in which case the minimum is zero.
505      */
506     private volatile int corePoolSize;
507 
508     /**
509      * Maximum pool size. Note that the actual maximum is internally
510      * bounded by CAPACITY.
511      */
512     private volatile int maximumPoolSize;
513 
514     /**
515      * The default rejected execution handler
516      */
517     private static final RejectedExecutionHandler defaultHandler =
518         new AbortPolicy();
519 
520     /**
521      * Permission required for callers of shutdown and shutdownNow.
522      * We additionally require (see checkShutdownAccess) that callers
523      * have permission to actually interrupt threads in the worker set
524      * (as governed by Thread.interrupt, which relies on
525      * ThreadGroup.checkAccess, which in turn relies on
526      * SecurityManager.checkAccess). Shutdowns are attempted only if
527      * these checks pass.
528      *
529      * All actual invocations of Thread.interrupt (see
530      * interruptIdleWorkers and interruptWorkers) ignore
531      * SecurityExceptions, meaning that the attempted interrupts
532      * silently fail. In the case of shutdown, they should not fail
533      * unless the SecurityManager has inconsistent policies, sometimes
534      * allowing access to a thread and sometimes not. In such cases,
535      * failure to actually interrupt threads may disable or delay full
536      * termination. Other uses of interruptIdleWorkers are advisory,
537      * and failure to actually interrupt will merely delay response to
538      * configuration changes so is not handled exceptionally.
539      */
540     private static final RuntimePermission shutdownPerm =
541         new RuntimePermission("modifyThread");
542 
543     /**
544      * Class Worker mainly maintains interrupt control state for
545      * threads running tasks, along with other minor bookkeeping.
546      * This class opportunistically extends AbstractQueuedSynchronizer
547      * to simplify acquiring and releasing a lock surrounding each
548      * task execution.  This protects against interrupts that are
549      * intended to wake up a worker thread waiting for a task from
550      * instead interrupting a task being run.  We implement a simple
551      * non-reentrant mutual exclusion lock rather than use
552      * ReentrantLock because we do not want worker tasks to be able to
553      * reacquire the lock when they invoke pool control methods like
554      * setCorePoolSize.  Additionally, to suppress interrupts until
555      * the thread actually starts running tasks, we initialize lock
556      * state to a negative value, and clear it upon start (in
557      * runWorker).
558      */
559     private final class Worker
560         extends AbstractQueuedSynchronizer
561         implements Runnable
562     {
563         /**
564          * This class will never be serialized, but we provide a
565          * serialVersionUID to suppress a javac warning.
566          */
567         private static final long serialVersionUID = 6138294804551838833L;
568 
569         /** Thread this worker is running in.  Null if factory fails. */
570         final Thread thread;
571         /** Initial task to run.  Possibly null. */
572         Runnable firstTask;
573         /** Per-thread task counter */
574         volatile long completedTasks;
575 
576         /**
577          * Creates with given first task and thread from ThreadFactory.
578          * @param firstTask the first task (null if none)
579          */
Worker(Runnable firstTask)580         Worker(Runnable firstTask) {
581             setState(-1); // inhibit interrupts until runWorker
582             this.firstTask = firstTask;
583             this.thread = getThreadFactory().newThread(this);
584         }
585 
586         /** Delegates main run loop to outer runWorker. */
run()587         public void run() {
588             runWorker(this);
589         }
590 
591         // Lock methods
592         //
593         // The value 0 represents the unlocked state.
594         // The value 1 represents the locked state.
595 
isHeldExclusively()596         protected boolean isHeldExclusively() {
597             return getState() != 0;
598         }
599 
tryAcquire(int unused)600         protected boolean tryAcquire(int unused) {
601             if (compareAndSetState(0, 1)) {
602                 setExclusiveOwnerThread(Thread.currentThread());
603                 return true;
604             }
605             return false;
606         }
607 
tryRelease(int unused)608         protected boolean tryRelease(int unused) {
609             setExclusiveOwnerThread(null);
610             setState(0);
611             return true;
612         }
613 
lock()614         public void lock()        { acquire(1); }
tryLock()615         public boolean tryLock()  { return tryAcquire(1); }
unlock()616         public void unlock()      { release(1); }
isLocked()617         public boolean isLocked() { return isHeldExclusively(); }
618 
interruptIfStarted()619         void interruptIfStarted() {
620             Thread t;
621             if (getState() >= 0 && (t = thread) != null && !t.isInterrupted()) {
622                 try {
623                     t.interrupt();
624                 } catch (SecurityException ignore) {
625                 }
626             }
627         }
628     }
629 
630     /*
631      * Methods for setting control state
632      */
633 
634     /**
635      * Transitions runState to given target, or leaves it alone if
636      * already at least the given target.
637      *
638      * @param targetState the desired state, either SHUTDOWN or STOP
639      *        (but not TIDYING or TERMINATED -- use tryTerminate for that)
640      */
advanceRunState(int targetState)641     private void advanceRunState(int targetState) {
642         for (;;) {
643             int c = ctl.get();
644             if (runStateAtLeast(c, targetState) ||
645                 ctl.compareAndSet(c, ctlOf(targetState, workerCountOf(c))))
646                 break;
647         }
648     }
649 
650     /**
651      * Transitions to TERMINATED state if either (SHUTDOWN and pool
652      * and queue empty) or (STOP and pool empty).  If otherwise
653      * eligible to terminate but workerCount is nonzero, interrupts an
654      * idle worker to ensure that shutdown signals propagate. This
655      * method must be called following any action that might make
656      * termination possible -- reducing worker count or removing tasks
657      * from the queue during shutdown. The method is non-private to
658      * allow access from ScheduledThreadPoolExecutor.
659      */
tryTerminate()660     final void tryTerminate() {
661         for (;;) {
662             int c = ctl.get();
663             if (isRunning(c) ||
664                 runStateAtLeast(c, TIDYING) ||
665                 (runStateOf(c) == SHUTDOWN && ! workQueue.isEmpty()))
666                 return;
667             if (workerCountOf(c) != 0) { // Eligible to terminate
668                 interruptIdleWorkers(ONLY_ONE);
669                 return;
670             }
671 
672             final ReentrantLock mainLock = this.mainLock;
673             mainLock.lock();
674             try {
675                 if (ctl.compareAndSet(c, ctlOf(TIDYING, 0))) {
676                     try {
677                         terminated();
678                     } finally {
679                         ctl.set(ctlOf(TERMINATED, 0));
680                         termination.signalAll();
681                     }
682                     return;
683                 }
684             } finally {
685                 mainLock.unlock();
686             }
687             // else retry on failed CAS
688         }
689     }
690 
691     /*
692      * Methods for controlling interrupts to worker threads.
693      */
694 
695     /**
696      * If there is a security manager, makes sure caller has
697      * permission to shut down threads in general (see shutdownPerm).
698      * If this passes, additionally makes sure the caller is allowed
699      * to interrupt each worker thread. This might not be true even if
700      * first check passed, if the SecurityManager treats some threads
701      * specially.
702      */
checkShutdownAccess()703     private void checkShutdownAccess() {
704         SecurityManager security = System.getSecurityManager();
705         if (security != null) {
706             security.checkPermission(shutdownPerm);
707             final ReentrantLock mainLock = this.mainLock;
708             mainLock.lock();
709             try {
710                 for (Worker w : workers)
711                     security.checkAccess(w.thread);
712             } finally {
713                 mainLock.unlock();
714             }
715         }
716     }
717 
718     /**
719      * Interrupts all threads, even if active. Ignores SecurityExceptions
720      * (in which case some threads may remain uninterrupted).
721      */
interruptWorkers()722     private void interruptWorkers() {
723         final ReentrantLock mainLock = this.mainLock;
724         mainLock.lock();
725         try {
726             for (Worker w : workers)
727                 w.interruptIfStarted();
728         } finally {
729             mainLock.unlock();
730         }
731     }
732 
733     /**
734      * Interrupts threads that might be waiting for tasks (as
735      * indicated by not being locked) so they can check for
736      * termination or configuration changes. Ignores
737      * SecurityExceptions (in which case some threads may remain
738      * uninterrupted).
739      *
740      * @param onlyOne If true, interrupt at most one worker. This is
741      * called only from tryTerminate when termination is otherwise
742      * enabled but there are still other workers.  In this case, at
743      * most one waiting worker is interrupted to propagate shutdown
744      * signals in case all threads are currently waiting.
745      * Interrupting any arbitrary thread ensures that newly arriving
746      * workers since shutdown began will also eventually exit.
747      * To guarantee eventual termination, it suffices to always
748      * interrupt only one idle worker, but shutdown() interrupts all
749      * idle workers so that redundant workers exit promptly, not
750      * waiting for a straggler task to finish.
751      */
interruptIdleWorkers(boolean onlyOne)752     private void interruptIdleWorkers(boolean onlyOne) {
753         final ReentrantLock mainLock = this.mainLock;
754         mainLock.lock();
755         try {
756             for (Worker w : workers) {
757                 Thread t = w.thread;
758                 if (!t.isInterrupted() && w.tryLock()) {
759                     try {
760                         t.interrupt();
761                     } catch (SecurityException ignore) {
762                     } finally {
763                         w.unlock();
764                     }
765                 }
766                 if (onlyOne)
767                     break;
768             }
769         } finally {
770             mainLock.unlock();
771         }
772     }
773 
774     /**
775      * Common form of interruptIdleWorkers, to avoid having to
776      * remember what the boolean argument means.
777      */
interruptIdleWorkers()778     private void interruptIdleWorkers() {
779         interruptIdleWorkers(false);
780     }
781 
782     private static final boolean ONLY_ONE = true;
783 
784     /*
785      * Misc utilities, most of which are also exported to
786      * ScheduledThreadPoolExecutor
787      */
788 
789     /**
790      * Invokes the rejected execution handler for the given command.
791      * Package-protected for use by ScheduledThreadPoolExecutor.
792      */
reject(Runnable command)793     final void reject(Runnable command) {
794         handler.rejectedExecution(command, this);
795     }
796 
797     /**
798      * Performs any further cleanup following run state transition on
799      * invocation of shutdown.  A no-op here, but used by
800      * ScheduledThreadPoolExecutor to cancel delayed tasks.
801      */
onShutdown()802     void onShutdown() {
803     }
804 
805     /**
806      * State check needed by ScheduledThreadPoolExecutor to
807      * enable running tasks during shutdown.
808      *
809      * @param shutdownOK true if should return true if SHUTDOWN
810      */
isRunningOrShutdown(boolean shutdownOK)811     final boolean isRunningOrShutdown(boolean shutdownOK) {
812         int rs = runStateOf(ctl.get());
813         return rs == RUNNING || (rs == SHUTDOWN && shutdownOK);
814     }
815 
816     /**
817      * Drains the task queue into a new list, normally using
818      * drainTo. But if the queue is a DelayQueue or any other kind of
819      * queue for which poll or drainTo may fail to remove some
820      * elements, it deletes them one by one.
821      */
drainQueue()822     private List<Runnable> drainQueue() {
823         BlockingQueue<Runnable> q = workQueue;
824         ArrayList<Runnable> taskList = new ArrayList<Runnable>();
825         q.drainTo(taskList);
826         if (!q.isEmpty()) {
827             for (Runnable r : q.toArray(new Runnable[0])) {
828                 if (q.remove(r))
829                     taskList.add(r);
830             }
831         }
832         return taskList;
833     }
834 
835     /*
836      * Methods for creating, running and cleaning up after workers
837      */
838 
839     /**
840      * Checks if a new worker can be added with respect to current
841      * pool state and the given bound (either core or maximum). If so,
842      * the worker count is adjusted accordingly, and, if possible, a
843      * new worker is created and started, running firstTask as its
844      * first task. This method returns false if the pool is stopped or
845      * eligible to shut down. It also returns false if the thread
846      * factory fails to create a thread when asked.  If the thread
847      * creation fails, either due to the thread factory returning
848      * null, or due to an exception (typically OutOfMemoryError in
849      * Thread.start()), we roll back cleanly.
850      *
851      * @param firstTask the task the new thread should run first (or
852      * null if none). Workers are created with an initial first task
853      * (in method execute()) to bypass queuing when there are fewer
854      * than corePoolSize threads (in which case we always start one),
855      * or when the queue is full (in which case we must bypass queue).
856      * Initially idle threads are usually created via
857      * prestartCoreThread or to replace other dying workers.
858      *
859      * @param core if true use corePoolSize as bound, else
860      * maximumPoolSize. (A boolean indicator is used here rather than a
861      * value to ensure reads of fresh values after checking other pool
862      * state).
863      * @return true if successful
864      */
addWorker(Runnable firstTask, boolean core)865     private boolean addWorker(Runnable firstTask, boolean core) {
866         retry:
867         for (;;) {
868             int c = ctl.get();
869             int rs = runStateOf(c);
870 
871             // Check if queue empty only if necessary.
872             if (rs >= SHUTDOWN &&
873                 ! (rs == SHUTDOWN &&
874                    firstTask == null &&
875                    ! workQueue.isEmpty()))
876                 return false;
877 
878             for (;;) {
879                 int wc = workerCountOf(c);
880                 if (wc >= CAPACITY ||
881                     wc >= (core ? corePoolSize : maximumPoolSize))
882                     return false;
883                 if (compareAndIncrementWorkerCount(c))
884                     break retry;
885                 c = ctl.get();  // Re-read ctl
886                 if (runStateOf(c) != rs)
887                     continue retry;
888                 // else CAS failed due to workerCount change; retry inner loop
889             }
890         }
891 
892         boolean workerStarted = false;
893         boolean workerAdded = false;
894         Worker w = null;
895         try {
896             w = new Worker(firstTask);
897             final Thread t = w.thread;
898             if (t != null) {
899                 final ReentrantLock mainLock = this.mainLock;
900                 mainLock.lock();
901                 try {
902                     // Recheck while holding lock.
903                     // Back out on ThreadFactory failure or if
904                     // shut down before lock acquired.
905                     int rs = runStateOf(ctl.get());
906 
907                     if (rs < SHUTDOWN ||
908                         (rs == SHUTDOWN && firstTask == null)) {
909                         if (t.isAlive()) // precheck that t is startable
910                             throw new IllegalThreadStateException();
911                         workers.add(w);
912                         int s = workers.size();
913                         if (s > largestPoolSize)
914                             largestPoolSize = s;
915                         workerAdded = true;
916                     }
917                 } finally {
918                     mainLock.unlock();
919                 }
920                 if (workerAdded) {
921                     t.start();
922                     workerStarted = true;
923                 }
924             }
925         } finally {
926             if (! workerStarted)
927                 addWorkerFailed(w);
928         }
929         return workerStarted;
930     }
931 
932     /**
933      * Rolls back the worker thread creation.
934      * - removes worker from workers, if present
935      * - decrements worker count
936      * - rechecks for termination, in case the existence of this
937      *   worker was holding up termination
938      */
addWorkerFailed(Worker w)939     private void addWorkerFailed(Worker w) {
940         final ReentrantLock mainLock = this.mainLock;
941         mainLock.lock();
942         try {
943             if (w != null)
944                 workers.remove(w);
945             decrementWorkerCount();
946             tryTerminate();
947         } finally {
948             mainLock.unlock();
949         }
950     }
951 
952     /**
953      * Performs cleanup and bookkeeping for a dying worker. Called
954      * only from worker threads. Unless completedAbruptly is set,
955      * assumes that workerCount has already been adjusted to account
956      * for exit.  This method removes thread from worker set, and
957      * possibly terminates the pool or replaces the worker if either
958      * it exited due to user task exception or if fewer than
959      * corePoolSize workers are running or queue is non-empty but
960      * there are no workers.
961      *
962      * @param w the worker
963      * @param completedAbruptly if the worker died due to user exception
964      */
processWorkerExit(Worker w, boolean completedAbruptly)965     private void processWorkerExit(Worker w, boolean completedAbruptly) {
966         if (completedAbruptly) // If abrupt, then workerCount wasn't adjusted
967             decrementWorkerCount();
968 
969         final ReentrantLock mainLock = this.mainLock;
970         mainLock.lock();
971         try {
972             completedTaskCount += w.completedTasks;
973             workers.remove(w);
974         } finally {
975             mainLock.unlock();
976         }
977 
978         tryTerminate();
979 
980         int c = ctl.get();
981         if (runStateLessThan(c, STOP)) {
982             if (!completedAbruptly) {
983                 int min = allowCoreThreadTimeOut ? 0 : corePoolSize;
984                 if (min == 0 && ! workQueue.isEmpty())
985                     min = 1;
986                 if (workerCountOf(c) >= min)
987                     return; // replacement not needed
988             }
989             addWorker(null, false);
990         }
991     }
992 
993     /**
994      * Performs blocking or timed wait for a task, depending on
995      * current configuration settings, or returns null if this worker
996      * must exit because of any of:
997      * 1. There are more than maximumPoolSize workers (due to
998      *    a call to setMaximumPoolSize).
999      * 2. The pool is stopped.
1000      * 3. The pool is shutdown and the queue is empty.
1001      * 4. This worker timed out waiting for a task, and timed-out
1002      *    workers are subject to termination (that is,
1003      *    {@code allowCoreThreadTimeOut || workerCount > corePoolSize})
1004      *    both before and after the timed wait, and if the queue is
1005      *    non-empty, this worker is not the last thread in the pool.
1006      *
1007      * @return task, or null if the worker must exit, in which case
1008      *         workerCount is decremented
1009      */
getTask()1010     private Runnable getTask() {
1011         boolean timedOut = false; // Did the last poll() time out?
1012 
1013         for (;;) {
1014             int c = ctl.get();
1015             int rs = runStateOf(c);
1016 
1017             // Check if queue empty only if necessary.
1018             if (rs >= SHUTDOWN && (rs >= STOP || workQueue.isEmpty())) {
1019                 decrementWorkerCount();
1020                 return null;
1021             }
1022 
1023             int wc = workerCountOf(c);
1024 
1025             // Are workers subject to culling?
1026             boolean timed = allowCoreThreadTimeOut || wc > corePoolSize;
1027 
1028             if ((wc > maximumPoolSize || (timed && timedOut))
1029                 && (wc > 1 || workQueue.isEmpty())) {
1030                 if (compareAndDecrementWorkerCount(c))
1031                     return null;
1032                 continue;
1033             }
1034 
1035             try {
1036                 Runnable r = timed ?
1037                     workQueue.poll(keepAliveTime, TimeUnit.NANOSECONDS) :
1038                     workQueue.take();
1039                 if (r != null)
1040                     return r;
1041                 timedOut = true;
1042             } catch (InterruptedException retry) {
1043                 timedOut = false;
1044             }
1045         }
1046     }
1047 
1048     /**
1049      * Main worker run loop.  Repeatedly gets tasks from queue and
1050      * executes them, while coping with a number of issues:
1051      *
1052      * 1. We may start out with an initial task, in which case we
1053      * don't need to get the first one. Otherwise, as long as pool is
1054      * running, we get tasks from getTask. If it returns null then the
1055      * worker exits due to changed pool state or configuration
1056      * parameters.  Other exits result from exception throws in
1057      * external code, in which case completedAbruptly holds, which
1058      * usually leads processWorkerExit to replace this thread.
1059      *
1060      * 2. Before running any task, the lock is acquired to prevent
1061      * other pool interrupts while the task is executing, and then we
1062      * ensure that unless pool is stopping, this thread does not have
1063      * its interrupt set.
1064      *
1065      * 3. Each task run is preceded by a call to beforeExecute, which
1066      * might throw an exception, in which case we cause thread to die
1067      * (breaking loop with completedAbruptly true) without processing
1068      * the task.
1069      *
1070      * 4. Assuming beforeExecute completes normally, we run the task,
1071      * gathering any of its thrown exceptions to send to afterExecute.
1072      * We separately handle RuntimeException, Error (both of which the
1073      * specs guarantee that we trap) and arbitrary Throwables.
1074      * Because we cannot rethrow Throwables within Runnable.run, we
1075      * wrap them within Errors on the way out (to the thread's
1076      * UncaughtExceptionHandler).  Any thrown exception also
1077      * conservatively causes thread to die.
1078      *
1079      * 5. After task.run completes, we call afterExecute, which may
1080      * also throw an exception, which will also cause thread to
1081      * die. According to JLS Sec 14.20, this exception is the one that
1082      * will be in effect even if task.run throws.
1083      *
1084      * The net effect of the exception mechanics is that afterExecute
1085      * and the thread's UncaughtExceptionHandler have as accurate
1086      * information as we can provide about any problems encountered by
1087      * user code.
1088      *
1089      * @param w the worker
1090      */
runWorker(Worker w)1091     final void runWorker(Worker w) {
1092         Thread wt = Thread.currentThread();
1093         Runnable task = w.firstTask;
1094         w.firstTask = null;
1095         w.unlock(); // allow interrupts
1096         boolean completedAbruptly = true;
1097         try {
1098             while (task != null || (task = getTask()) != null) {
1099                 w.lock();
1100                 // If pool is stopping, ensure thread is interrupted;
1101                 // if not, ensure thread is not interrupted.  This
1102                 // requires a recheck in second case to deal with
1103                 // shutdownNow race while clearing interrupt
1104                 if ((runStateAtLeast(ctl.get(), STOP) ||
1105                      (Thread.interrupted() &&
1106                       runStateAtLeast(ctl.get(), STOP))) &&
1107                     !wt.isInterrupted())
1108                     wt.interrupt();
1109                 try {
1110                     beforeExecute(wt, task);
1111                     Throwable thrown = null;
1112                     try {
1113                         task.run();
1114                     } catch (RuntimeException x) {
1115                         thrown = x; throw x;
1116                     } catch (Error x) {
1117                         thrown = x; throw x;
1118                     } catch (Throwable x) {
1119                         thrown = x; throw new Error(x);
1120                     } finally {
1121                         afterExecute(task, thrown);
1122                     }
1123                 } finally {
1124                     task = null;
1125                     w.completedTasks++;
1126                     w.unlock();
1127                 }
1128             }
1129             completedAbruptly = false;
1130         } finally {
1131             processWorkerExit(w, completedAbruptly);
1132         }
1133     }
1134 
1135     // Public constructors and methods
1136 
1137     /**
1138      * Creates a new {@code ThreadPoolExecutor} with the given initial
1139      * parameters and default thread factory and rejected execution handler.
1140      * It may be more convenient to use one of the {@link Executors} factory
1141      * methods instead of this general purpose constructor.
1142      *
1143      * @param corePoolSize the number of threads to keep in the pool, even
1144      *        if they are idle, unless {@code allowCoreThreadTimeOut} is set
1145      * @param maximumPoolSize the maximum number of threads to allow in the
1146      *        pool
1147      * @param keepAliveTime when the number of threads is greater than
1148      *        the core, this is the maximum time that excess idle threads
1149      *        will wait for new tasks before terminating.
1150      * @param unit the time unit for the {@code keepAliveTime} argument
1151      * @param workQueue the queue to use for holding tasks before they are
1152      *        executed.  This queue will hold only the {@code Runnable}
1153      *        tasks submitted by the {@code execute} method.
1154      * @throws IllegalArgumentException if one of the following holds:<br>
1155      *         {@code corePoolSize < 0}<br>
1156      *         {@code keepAliveTime < 0}<br>
1157      *         {@code maximumPoolSize <= 0}<br>
1158      *         {@code maximumPoolSize < corePoolSize}
1159      * @throws NullPointerException if {@code workQueue} is null
1160      */
ThreadPoolExecutor(int corePoolSize, int maximumPoolSize, long keepAliveTime, TimeUnit unit, BlockingQueue<Runnable> workQueue)1161     public ThreadPoolExecutor(int corePoolSize,
1162                               int maximumPoolSize,
1163                               long keepAliveTime,
1164                               TimeUnit unit,
1165                               BlockingQueue<Runnable> workQueue) {
1166         this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue,
1167              Executors.defaultThreadFactory(), defaultHandler);
1168     }
1169 
1170     /**
1171      * Creates a new {@code ThreadPoolExecutor} with the given initial
1172      * parameters and default rejected execution handler.
1173      *
1174      * @param corePoolSize the number of threads to keep in the pool, even
1175      *        if they are idle, unless {@code allowCoreThreadTimeOut} is set
1176      * @param maximumPoolSize the maximum number of threads to allow in the
1177      *        pool
1178      * @param keepAliveTime when the number of threads is greater than
1179      *        the core, this is the maximum time that excess idle threads
1180      *        will wait for new tasks before terminating.
1181      * @param unit the time unit for the {@code keepAliveTime} argument
1182      * @param workQueue the queue to use for holding tasks before they are
1183      *        executed.  This queue will hold only the {@code Runnable}
1184      *        tasks submitted by the {@code execute} method.
1185      * @param threadFactory the factory to use when the executor
1186      *        creates a new thread
1187      * @throws IllegalArgumentException if one of the following holds:<br>
1188      *         {@code corePoolSize < 0}<br>
1189      *         {@code keepAliveTime < 0}<br>
1190      *         {@code maximumPoolSize <= 0}<br>
1191      *         {@code maximumPoolSize < corePoolSize}
1192      * @throws NullPointerException if {@code workQueue}
1193      *         or {@code threadFactory} is null
1194      */
ThreadPoolExecutor(int corePoolSize, int maximumPoolSize, long keepAliveTime, TimeUnit unit, BlockingQueue<Runnable> workQueue, ThreadFactory threadFactory)1195     public ThreadPoolExecutor(int corePoolSize,
1196                               int maximumPoolSize,
1197                               long keepAliveTime,
1198                               TimeUnit unit,
1199                               BlockingQueue<Runnable> workQueue,
1200                               ThreadFactory threadFactory) {
1201         this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue,
1202              threadFactory, defaultHandler);
1203     }
1204 
1205     /**
1206      * Creates a new {@code ThreadPoolExecutor} with the given initial
1207      * parameters and default thread factory.
1208      *
1209      * @param corePoolSize the number of threads to keep in the pool, even
1210      *        if they are idle, unless {@code allowCoreThreadTimeOut} is set
1211      * @param maximumPoolSize the maximum number of threads to allow in the
1212      *        pool
1213      * @param keepAliveTime when the number of threads is greater than
1214      *        the core, this is the maximum time that excess idle threads
1215      *        will wait for new tasks before terminating.
1216      * @param unit the time unit for the {@code keepAliveTime} argument
1217      * @param workQueue the queue to use for holding tasks before they are
1218      *        executed.  This queue will hold only the {@code Runnable}
1219      *        tasks submitted by the {@code execute} method.
1220      * @param handler the handler to use when execution is blocked
1221      *        because the thread bounds and queue capacities are reached
1222      * @throws IllegalArgumentException if one of the following holds:<br>
1223      *         {@code corePoolSize < 0}<br>
1224      *         {@code keepAliveTime < 0}<br>
1225      *         {@code maximumPoolSize <= 0}<br>
1226      *         {@code maximumPoolSize < corePoolSize}
1227      * @throws NullPointerException if {@code workQueue}
1228      *         or {@code handler} is null
1229      */
ThreadPoolExecutor(int corePoolSize, int maximumPoolSize, long keepAliveTime, TimeUnit unit, BlockingQueue<Runnable> workQueue, RejectedExecutionHandler handler)1230     public ThreadPoolExecutor(int corePoolSize,
1231                               int maximumPoolSize,
1232                               long keepAliveTime,
1233                               TimeUnit unit,
1234                               BlockingQueue<Runnable> workQueue,
1235                               RejectedExecutionHandler handler) {
1236         this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue,
1237              Executors.defaultThreadFactory(), handler);
1238     }
1239 
1240     /**
1241      * Creates a new {@code ThreadPoolExecutor} with the given initial
1242      * parameters.
1243      *
1244      * @param corePoolSize the number of threads to keep in the pool, even
1245      *        if they are idle, unless {@code allowCoreThreadTimeOut} is set
1246      * @param maximumPoolSize the maximum number of threads to allow in the
1247      *        pool
1248      * @param keepAliveTime when the number of threads is greater than
1249      *        the core, this is the maximum time that excess idle threads
1250      *        will wait for new tasks before terminating.
1251      * @param unit the time unit for the {@code keepAliveTime} argument
1252      * @param workQueue the queue to use for holding tasks before they are
1253      *        executed.  This queue will hold only the {@code Runnable}
1254      *        tasks submitted by the {@code execute} method.
1255      * @param threadFactory the factory to use when the executor
1256      *        creates a new thread
1257      * @param handler the handler to use when execution is blocked
1258      *        because the thread bounds and queue capacities are reached
1259      * @throws IllegalArgumentException if one of the following holds:<br>
1260      *         {@code corePoolSize < 0}<br>
1261      *         {@code keepAliveTime < 0}<br>
1262      *         {@code maximumPoolSize <= 0}<br>
1263      *         {@code maximumPoolSize < corePoolSize}
1264      * @throws NullPointerException if {@code workQueue}
1265      *         or {@code threadFactory} or {@code handler} is null
1266      */
ThreadPoolExecutor(int corePoolSize, int maximumPoolSize, long keepAliveTime, TimeUnit unit, BlockingQueue<Runnable> workQueue, ThreadFactory threadFactory, RejectedExecutionHandler handler)1267     public ThreadPoolExecutor(int corePoolSize,
1268                               int maximumPoolSize,
1269                               long keepAliveTime,
1270                               TimeUnit unit,
1271                               BlockingQueue<Runnable> workQueue,
1272                               ThreadFactory threadFactory,
1273                               RejectedExecutionHandler handler) {
1274         if (corePoolSize < 0 ||
1275             maximumPoolSize <= 0 ||
1276             maximumPoolSize < corePoolSize ||
1277             keepAliveTime < 0)
1278             throw new IllegalArgumentException();
1279         if (workQueue == null || threadFactory == null || handler == null)
1280             throw new NullPointerException();
1281         this.corePoolSize = corePoolSize;
1282         this.maximumPoolSize = maximumPoolSize;
1283         this.workQueue = workQueue;
1284         this.keepAliveTime = unit.toNanos(keepAliveTime);
1285         this.threadFactory = threadFactory;
1286         this.handler = handler;
1287     }
1288 
1289     /**
1290      * Executes the given task sometime in the future.  The task
1291      * may execute in a new thread or in an existing pooled thread.
1292      *
1293      * If the task cannot be submitted for execution, either because this
1294      * executor has been shutdown or because its capacity has been reached,
1295      * the task is handled by the current {@code RejectedExecutionHandler}.
1296      *
1297      * @param command the task to execute
1298      * @throws RejectedExecutionException at discretion of
1299      *         {@code RejectedExecutionHandler}, if the task
1300      *         cannot be accepted for execution
1301      * @throws NullPointerException if {@code command} is null
1302      */
execute(Runnable command)1303     public void execute(Runnable command) {
1304         if (command == null)
1305             throw new NullPointerException();
1306         /*
1307          * Proceed in 3 steps:
1308          *
1309          * 1. If fewer than corePoolSize threads are running, try to
1310          * start a new thread with the given command as its first
1311          * task.  The call to addWorker atomically checks runState and
1312          * workerCount, and so prevents false alarms that would add
1313          * threads when it shouldn't, by returning false.
1314          *
1315          * 2. If a task can be successfully queued, then we still need
1316          * to double-check whether we should have added a thread
1317          * (because existing ones died since last checking) or that
1318          * the pool shut down since entry into this method. So we
1319          * recheck state and if necessary roll back the enqueuing if
1320          * stopped, or start a new thread if there are none.
1321          *
1322          * 3. If we cannot queue task, then we try to add a new
1323          * thread.  If it fails, we know we are shut down or saturated
1324          * and so reject the task.
1325          */
1326         int c = ctl.get();
1327         if (workerCountOf(c) < corePoolSize) {
1328             if (addWorker(command, true))
1329                 return;
1330             c = ctl.get();
1331         }
1332         if (isRunning(c) && workQueue.offer(command)) {
1333             int recheck = ctl.get();
1334             if (! isRunning(recheck) && remove(command))
1335                 reject(command);
1336             else if (workerCountOf(recheck) == 0)
1337                 addWorker(null, false);
1338         }
1339         else if (!addWorker(command, false))
1340             reject(command);
1341     }
1342 
1343     /**
1344      * Initiates an orderly shutdown in which previously submitted
1345      * tasks are executed, but no new tasks will be accepted.
1346      * Invocation has no additional effect if already shut down.
1347      *
1348      * <p>This method does not wait for previously submitted tasks to
1349      * complete execution.  Use {@link #awaitTermination awaitTermination}
1350      * to do that.
1351      */
1352     // android-note: Removed @throws SecurityException
shutdown()1353     public void shutdown() {
1354         final ReentrantLock mainLock = this.mainLock;
1355         mainLock.lock();
1356         try {
1357             checkShutdownAccess();
1358             advanceRunState(SHUTDOWN);
1359             interruptIdleWorkers();
1360             onShutdown(); // hook for ScheduledThreadPoolExecutor
1361         } finally {
1362             mainLock.unlock();
1363         }
1364         tryTerminate();
1365     }
1366 
1367     /**
1368      * Attempts to stop all actively executing tasks, halts the
1369      * processing of waiting tasks, and returns a list of the tasks
1370      * that were awaiting execution. These tasks are drained (removed)
1371      * from the task queue upon return from this method.
1372      *
1373      * <p>This method does not wait for actively executing tasks to
1374      * terminate.  Use {@link #awaitTermination awaitTermination} to
1375      * do that.
1376      *
1377      * <p>There are no guarantees beyond best-effort attempts to stop
1378      * processing actively executing tasks.  This implementation
1379      * cancels tasks via {@link Thread#interrupt}, so any task that
1380      * fails to respond to interrupts may never terminate.
1381      */
1382     // android-note: Removed @throws SecurityException
shutdownNow()1383     public List<Runnable> shutdownNow() {
1384         List<Runnable> tasks;
1385         final ReentrantLock mainLock = this.mainLock;
1386         mainLock.lock();
1387         try {
1388             checkShutdownAccess();
1389             advanceRunState(STOP);
1390             interruptWorkers();
1391             tasks = drainQueue();
1392         } finally {
1393             mainLock.unlock();
1394         }
1395         tryTerminate();
1396         return tasks;
1397     }
1398 
isShutdown()1399     public boolean isShutdown() {
1400         return ! isRunning(ctl.get());
1401     }
1402 
1403     /**
1404      * Returns true if this executor is in the process of terminating
1405      * after {@link #shutdown} or {@link #shutdownNow} but has not
1406      * completely terminated.  This method may be useful for
1407      * debugging. A return of {@code true} reported a sufficient
1408      * period after shutdown may indicate that submitted tasks have
1409      * ignored or suppressed interruption, causing this executor not
1410      * to properly terminate.
1411      *
1412      * @return {@code true} if terminating but not yet terminated
1413      */
isTerminating()1414     public boolean isTerminating() {
1415         int c = ctl.get();
1416         return ! isRunning(c) && runStateLessThan(c, TERMINATED);
1417     }
1418 
isTerminated()1419     public boolean isTerminated() {
1420         return runStateAtLeast(ctl.get(), TERMINATED);
1421     }
1422 
awaitTermination(long timeout, TimeUnit unit)1423     public boolean awaitTermination(long timeout, TimeUnit unit)
1424         throws InterruptedException {
1425         long nanos = unit.toNanos(timeout);
1426         final ReentrantLock mainLock = this.mainLock;
1427         mainLock.lock();
1428         try {
1429             for (;;) {
1430                 if (runStateAtLeast(ctl.get(), TERMINATED))
1431                     return true;
1432                 if (nanos <= 0)
1433                     return false;
1434                 nanos = termination.awaitNanos(nanos);
1435             }
1436         } finally {
1437             mainLock.unlock();
1438         }
1439     }
1440 
1441     /**
1442      * Invokes {@code shutdown} when this executor is no longer
1443      * referenced and it has no threads.
1444      */
finalize()1445     protected void finalize() {
1446         shutdown();
1447     }
1448 
1449     /**
1450      * Sets the thread factory used to create new threads.
1451      *
1452      * @param threadFactory the new thread factory
1453      * @throws NullPointerException if threadFactory is null
1454      * @see #getThreadFactory
1455      */
setThreadFactory(ThreadFactory threadFactory)1456     public void setThreadFactory(ThreadFactory threadFactory) {
1457         if (threadFactory == null)
1458             throw new NullPointerException();
1459         this.threadFactory = threadFactory;
1460     }
1461 
1462     /**
1463      * Returns the thread factory used to create new threads.
1464      *
1465      * @return the current thread factory
1466      * @see #setThreadFactory(ThreadFactory)
1467      */
getThreadFactory()1468     public ThreadFactory getThreadFactory() {
1469         return threadFactory;
1470     }
1471 
1472     /**
1473      * Sets a new handler for unexecutable tasks.
1474      *
1475      * @param handler the new handler
1476      * @throws NullPointerException if handler is null
1477      * @see #getRejectedExecutionHandler
1478      */
setRejectedExecutionHandler(RejectedExecutionHandler handler)1479     public void setRejectedExecutionHandler(RejectedExecutionHandler handler) {
1480         if (handler == null)
1481             throw new NullPointerException();
1482         this.handler = handler;
1483     }
1484 
1485     /**
1486      * Returns the current handler for unexecutable tasks.
1487      *
1488      * @return the current handler
1489      * @see #setRejectedExecutionHandler(RejectedExecutionHandler)
1490      */
getRejectedExecutionHandler()1491     public RejectedExecutionHandler getRejectedExecutionHandler() {
1492         return handler;
1493     }
1494 
1495     /**
1496      * Sets the core number of threads.  This overrides any value set
1497      * in the constructor.  If the new value is smaller than the
1498      * current value, excess existing threads will be terminated when
1499      * they next become idle.  If larger, new threads will, if needed,
1500      * be started to execute any queued tasks.
1501      *
1502      * @param corePoolSize the new core size
1503      * @throws IllegalArgumentException if {@code corePoolSize < 0}
1504      * @see #getCorePoolSize
1505      */
setCorePoolSize(int corePoolSize)1506     public void setCorePoolSize(int corePoolSize) {
1507         if (corePoolSize < 0)
1508             throw new IllegalArgumentException();
1509         int delta = corePoolSize - this.corePoolSize;
1510         this.corePoolSize = corePoolSize;
1511         if (workerCountOf(ctl.get()) > corePoolSize)
1512             interruptIdleWorkers();
1513         else if (delta > 0) {
1514             // We don't really know how many new threads are "needed".
1515             // As a heuristic, prestart enough new workers (up to new
1516             // core size) to handle the current number of tasks in
1517             // queue, but stop if queue becomes empty while doing so.
1518             int k = Math.min(delta, workQueue.size());
1519             while (k-- > 0 && addWorker(null, true)) {
1520                 if (workQueue.isEmpty())
1521                     break;
1522             }
1523         }
1524     }
1525 
1526     /**
1527      * Returns the core number of threads.
1528      *
1529      * @return the core number of threads
1530      * @see #setCorePoolSize
1531      */
getCorePoolSize()1532     public int getCorePoolSize() {
1533         return corePoolSize;
1534     }
1535 
1536     /**
1537      * Starts a core thread, causing it to idly wait for work. This
1538      * overrides the default policy of starting core threads only when
1539      * new tasks are executed. This method will return {@code false}
1540      * if all core threads have already been started.
1541      *
1542      * @return {@code true} if a thread was started
1543      */
prestartCoreThread()1544     public boolean prestartCoreThread() {
1545         return workerCountOf(ctl.get()) < corePoolSize &&
1546             addWorker(null, true);
1547     }
1548 
1549     /**
1550      * Same as prestartCoreThread except arranges that at least one
1551      * thread is started even if corePoolSize is 0.
1552      */
ensurePrestart()1553     void ensurePrestart() {
1554         int wc = workerCountOf(ctl.get());
1555         if (wc < corePoolSize)
1556             addWorker(null, true);
1557         else if (wc == 0)
1558             addWorker(null, false);
1559     }
1560 
1561     /**
1562      * Starts all core threads, causing them to idly wait for work. This
1563      * overrides the default policy of starting core threads only when
1564      * new tasks are executed.
1565      *
1566      * @return the number of threads started
1567      */
prestartAllCoreThreads()1568     public int prestartAllCoreThreads() {
1569         int n = 0;
1570         while (addWorker(null, true))
1571             ++n;
1572         return n;
1573     }
1574 
1575     /**
1576      * Returns true if this pool allows core threads to time out and
1577      * terminate if no tasks arrive within the keepAlive time, being
1578      * replaced if needed when new tasks arrive. When true, the same
1579      * keep-alive policy applying to non-core threads applies also to
1580      * core threads. When false (the default), core threads are never
1581      * terminated due to lack of incoming tasks.
1582      *
1583      * @return {@code true} if core threads are allowed to time out,
1584      *         else {@code false}
1585      *
1586      * @since 1.6
1587      */
allowsCoreThreadTimeOut()1588     public boolean allowsCoreThreadTimeOut() {
1589         return allowCoreThreadTimeOut;
1590     }
1591 
1592     /**
1593      * Sets the policy governing whether core threads may time out and
1594      * terminate if no tasks arrive within the keep-alive time, being
1595      * replaced if needed when new tasks arrive. When false, core
1596      * threads are never terminated due to lack of incoming
1597      * tasks. When true, the same keep-alive policy applying to
1598      * non-core threads applies also to core threads. To avoid
1599      * continual thread replacement, the keep-alive time must be
1600      * greater than zero when setting {@code true}. This method
1601      * should in general be called before the pool is actively used.
1602      *
1603      * @param value {@code true} if should time out, else {@code false}
1604      * @throws IllegalArgumentException if value is {@code true}
1605      *         and the current keep-alive time is not greater than zero
1606      *
1607      * @since 1.6
1608      */
allowCoreThreadTimeOut(boolean value)1609     public void allowCoreThreadTimeOut(boolean value) {
1610         if (value && keepAliveTime <= 0)
1611             throw new IllegalArgumentException("Core threads must have nonzero keep alive times");
1612         if (value != allowCoreThreadTimeOut) {
1613             allowCoreThreadTimeOut = value;
1614             if (value)
1615                 interruptIdleWorkers();
1616         }
1617     }
1618 
1619     /**
1620      * Sets the maximum allowed number of threads. This overrides any
1621      * value set in the constructor. If the new value is smaller than
1622      * the current value, excess existing threads will be
1623      * terminated when they next become idle.
1624      *
1625      * @param maximumPoolSize the new maximum
1626      * @throws IllegalArgumentException if the new maximum is
1627      *         less than or equal to zero, or
1628      *         less than the {@linkplain #getCorePoolSize core pool size}
1629      * @see #getMaximumPoolSize
1630      */
setMaximumPoolSize(int maximumPoolSize)1631     public void setMaximumPoolSize(int maximumPoolSize) {
1632         if (maximumPoolSize <= 0 || maximumPoolSize < corePoolSize)
1633             throw new IllegalArgumentException();
1634         this.maximumPoolSize = maximumPoolSize;
1635         if (workerCountOf(ctl.get()) > maximumPoolSize)
1636             interruptIdleWorkers();
1637     }
1638 
1639     /**
1640      * Returns the maximum allowed number of threads.
1641      *
1642      * @return the maximum allowed number of threads
1643      * @see #setMaximumPoolSize
1644      */
getMaximumPoolSize()1645     public int getMaximumPoolSize() {
1646         return maximumPoolSize;
1647     }
1648 
1649     /**
1650      * Sets the time limit for which threads may remain idle before
1651      * being terminated.  If there are more than the core number of
1652      * threads currently in the pool, after waiting this amount of
1653      * time without processing a task, excess threads will be
1654      * terminated.  This overrides any value set in the constructor.
1655      *
1656      * @param time the time to wait.  A time value of zero will cause
1657      *        excess threads to terminate immediately after executing tasks.
1658      * @param unit the time unit of the {@code time} argument
1659      * @throws IllegalArgumentException if {@code time} less than zero or
1660      *         if {@code time} is zero and {@code allowsCoreThreadTimeOut}
1661      * @see #getKeepAliveTime(TimeUnit)
1662      */
setKeepAliveTime(long time, TimeUnit unit)1663     public void setKeepAliveTime(long time, TimeUnit unit) {
1664         if (time < 0)
1665             throw new IllegalArgumentException();
1666         if (time == 0 && allowsCoreThreadTimeOut())
1667             throw new IllegalArgumentException("Core threads must have nonzero keep alive times");
1668         long keepAliveTime = unit.toNanos(time);
1669         long delta = keepAliveTime - this.keepAliveTime;
1670         this.keepAliveTime = keepAliveTime;
1671         if (delta < 0)
1672             interruptIdleWorkers();
1673     }
1674 
1675     /**
1676      * Returns the thread keep-alive time, which is the amount of time
1677      * that threads in excess of the core pool size may remain
1678      * idle before being terminated.
1679      *
1680      * @param unit the desired time unit of the result
1681      * @return the time limit
1682      * @see #setKeepAliveTime(long, TimeUnit)
1683      */
getKeepAliveTime(TimeUnit unit)1684     public long getKeepAliveTime(TimeUnit unit) {
1685         return unit.convert(keepAliveTime, TimeUnit.NANOSECONDS);
1686     }
1687 
1688     /* User-level queue utilities */
1689 
1690     /**
1691      * Returns the task queue used by this executor. Access to the
1692      * task queue is intended primarily for debugging and monitoring.
1693      * This queue may be in active use.  Retrieving the task queue
1694      * does not prevent queued tasks from executing.
1695      *
1696      * @return the task queue
1697      */
getQueue()1698     public BlockingQueue<Runnable> getQueue() {
1699         return workQueue;
1700     }
1701 
1702     /**
1703      * Removes this task from the executor's internal queue if it is
1704      * present, thus causing it not to be run if it has not already
1705      * started.
1706      *
1707      * <p>This method may be useful as one part of a cancellation
1708      * scheme.  It may fail to remove tasks that have been converted
1709      * into other forms before being placed on the internal queue. For
1710      * example, a task entered using {@code submit} might be
1711      * converted into a form that maintains {@code Future} status.
1712      * However, in such cases, method {@link #purge} may be used to
1713      * remove those Futures that have been cancelled.
1714      *
1715      * @param task the task to remove
1716      * @return {@code true} if the task was removed
1717      */
remove(Runnable task)1718     public boolean remove(Runnable task) {
1719         boolean removed = workQueue.remove(task);
1720         tryTerminate(); // In case SHUTDOWN and now empty
1721         return removed;
1722     }
1723 
1724     /**
1725      * Tries to remove from the work queue all {@link Future}
1726      * tasks that have been cancelled. This method can be useful as a
1727      * storage reclamation operation, that has no other impact on
1728      * functionality. Cancelled tasks are never executed, but may
1729      * accumulate in work queues until worker threads can actively
1730      * remove them. Invoking this method instead tries to remove them now.
1731      * However, this method may fail to remove tasks in
1732      * the presence of interference by other threads.
1733      */
purge()1734     public void purge() {
1735         final BlockingQueue<Runnable> q = workQueue;
1736         try {
1737             Iterator<Runnable> it = q.iterator();
1738             while (it.hasNext()) {
1739                 Runnable r = it.next();
1740                 if (r instanceof Future<?> && ((Future<?>)r).isCancelled())
1741                     it.remove();
1742             }
1743         } catch (ConcurrentModificationException fallThrough) {
1744             // Take slow path if we encounter interference during traversal.
1745             // Make copy for traversal and call remove for cancelled entries.
1746             // The slow path is more likely to be O(N*N).
1747             for (Object r : q.toArray())
1748                 if (r instanceof Future<?> && ((Future<?>)r).isCancelled())
1749                     q.remove(r);
1750         }
1751 
1752         tryTerminate(); // In case SHUTDOWN and now empty
1753     }
1754 
1755     /* Statistics */
1756 
1757     /**
1758      * Returns the current number of threads in the pool.
1759      *
1760      * @return the number of threads
1761      */
getPoolSize()1762     public int getPoolSize() {
1763         final ReentrantLock mainLock = this.mainLock;
1764         mainLock.lock();
1765         try {
1766             // Remove rare and surprising possibility of
1767             // isTerminated() && getPoolSize() > 0
1768             return runStateAtLeast(ctl.get(), TIDYING) ? 0
1769                 : workers.size();
1770         } finally {
1771             mainLock.unlock();
1772         }
1773     }
1774 
1775     /**
1776      * Returns the approximate number of threads that are actively
1777      * executing tasks.
1778      *
1779      * @return the number of threads
1780      */
getActiveCount()1781     public int getActiveCount() {
1782         final ReentrantLock mainLock = this.mainLock;
1783         mainLock.lock();
1784         try {
1785             int n = 0;
1786             for (Worker w : workers)
1787                 if (w.isLocked())
1788                     ++n;
1789             return n;
1790         } finally {
1791             mainLock.unlock();
1792         }
1793     }
1794 
1795     /**
1796      * Returns the largest number of threads that have ever
1797      * simultaneously been in the pool.
1798      *
1799      * @return the number of threads
1800      */
getLargestPoolSize()1801     public int getLargestPoolSize() {
1802         final ReentrantLock mainLock = this.mainLock;
1803         mainLock.lock();
1804         try {
1805             return largestPoolSize;
1806         } finally {
1807             mainLock.unlock();
1808         }
1809     }
1810 
1811     /**
1812      * Returns the approximate total number of tasks that have ever been
1813      * scheduled for execution. Because the states of tasks and
1814      * threads may change dynamically during computation, the returned
1815      * value is only an approximation.
1816      *
1817      * @return the number of tasks
1818      */
getTaskCount()1819     public long getTaskCount() {
1820         final ReentrantLock mainLock = this.mainLock;
1821         mainLock.lock();
1822         try {
1823             long n = completedTaskCount;
1824             for (Worker w : workers) {
1825                 n += w.completedTasks;
1826                 if (w.isLocked())
1827                     ++n;
1828             }
1829             return n + workQueue.size();
1830         } finally {
1831             mainLock.unlock();
1832         }
1833     }
1834 
1835     /**
1836      * Returns the approximate total number of tasks that have
1837      * completed execution. Because the states of tasks and threads
1838      * may change dynamically during computation, the returned value
1839      * is only an approximation, but one that does not ever decrease
1840      * across successive calls.
1841      *
1842      * @return the number of tasks
1843      */
getCompletedTaskCount()1844     public long getCompletedTaskCount() {
1845         final ReentrantLock mainLock = this.mainLock;
1846         mainLock.lock();
1847         try {
1848             long n = completedTaskCount;
1849             for (Worker w : workers)
1850                 n += w.completedTasks;
1851             return n;
1852         } finally {
1853             mainLock.unlock();
1854         }
1855     }
1856 
1857     /**
1858      * Returns a string identifying this pool, as well as its state,
1859      * including indications of run state and estimated worker and
1860      * task counts.
1861      *
1862      * @return a string identifying this pool, as well as its state
1863      */
toString()1864     public String toString() {
1865         long ncompleted;
1866         int nworkers, nactive;
1867         final ReentrantLock mainLock = this.mainLock;
1868         mainLock.lock();
1869         try {
1870             ncompleted = completedTaskCount;
1871             nactive = 0;
1872             nworkers = workers.size();
1873             for (Worker w : workers) {
1874                 ncompleted += w.completedTasks;
1875                 if (w.isLocked())
1876                     ++nactive;
1877             }
1878         } finally {
1879             mainLock.unlock();
1880         }
1881         int c = ctl.get();
1882         String rs = (runStateLessThan(c, SHUTDOWN) ? "Running" :
1883                      (runStateAtLeast(c, TERMINATED) ? "Terminated" :
1884                       "Shutting down"));
1885         return super.toString() +
1886             "[" + rs +
1887             ", pool size = " + nworkers +
1888             ", active threads = " + nactive +
1889             ", queued tasks = " + workQueue.size() +
1890             ", completed tasks = " + ncompleted +
1891             "]";
1892     }
1893 
1894     /* Extension hooks */
1895 
1896     /**
1897      * Method invoked prior to executing the given Runnable in the
1898      * given thread.  This method is invoked by thread {@code t} that
1899      * will execute task {@code r}, and may be used to re-initialize
1900      * ThreadLocals, or to perform logging.
1901      *
1902      * <p>This implementation does nothing, but may be customized in
1903      * subclasses. Note: To properly nest multiple overridings, subclasses
1904      * should generally invoke {@code super.beforeExecute} at the end of
1905      * this method.
1906      *
1907      * @param t the thread that will run task {@code r}
1908      * @param r the task that will be executed
1909      */
beforeExecute(Thread t, Runnable r)1910     protected void beforeExecute(Thread t, Runnable r) { }
1911 
1912     /**
1913      * Method invoked upon completion of execution of the given Runnable.
1914      * This method is invoked by the thread that executed the task. If
1915      * non-null, the Throwable is the uncaught {@code RuntimeException}
1916      * or {@code Error} that caused execution to terminate abruptly.
1917      *
1918      * <p>This implementation does nothing, but may be customized in
1919      * subclasses. Note: To properly nest multiple overridings, subclasses
1920      * should generally invoke {@code super.afterExecute} at the
1921      * beginning of this method.
1922      *
1923      * <p><b>Note:</b> When actions are enclosed in tasks (such as
1924      * {@link FutureTask}) either explicitly or via methods such as
1925      * {@code submit}, these task objects catch and maintain
1926      * computational exceptions, and so they do not cause abrupt
1927      * termination, and the internal exceptions are <em>not</em>
1928      * passed to this method. If you would like to trap both kinds of
1929      * failures in this method, you can further probe for such cases,
1930      * as in this sample subclass that prints either the direct cause
1931      * or the underlying exception if a task has been aborted:
1932      *
1933      *  <pre> {@code
1934      * class ExtendedExecutor extends ThreadPoolExecutor {
1935      *   // ...
1936      *   protected void afterExecute(Runnable r, Throwable t) {
1937      *     super.afterExecute(r, t);
1938      *     if (t == null && r instanceof Future<?>) {
1939      *       try {
1940      *         Object result = ((Future<?>) r).get();
1941      *       } catch (CancellationException ce) {
1942      *           t = ce;
1943      *       } catch (ExecutionException ee) {
1944      *           t = ee.getCause();
1945      *       } catch (InterruptedException ie) {
1946      *           Thread.currentThread().interrupt(); // ignore/reset
1947      *       }
1948      *     }
1949      *     if (t != null)
1950      *       System.out.println(t);
1951      *   }
1952      * }}</pre>
1953      *
1954      * @param r the runnable that has completed
1955      * @param t the exception that caused termination, or null if
1956      * execution completed normally
1957      */
afterExecute(Runnable r, Throwable t)1958     protected void afterExecute(Runnable r, Throwable t) { }
1959 
1960     /**
1961      * Method invoked when the Executor has terminated.  Default
1962      * implementation does nothing. Note: To properly nest multiple
1963      * overridings, subclasses should generally invoke
1964      * {@code super.terminated} within this method.
1965      */
terminated()1966     protected void terminated() { }
1967 
1968     /* Predefined RejectedExecutionHandlers */
1969 
1970     /**
1971      * A handler for rejected tasks that runs the rejected task
1972      * directly in the calling thread of the {@code execute} method,
1973      * unless the executor has been shut down, in which case the task
1974      * is discarded.
1975      */
1976     public static class CallerRunsPolicy implements RejectedExecutionHandler {
1977         /**
1978          * Creates a {@code CallerRunsPolicy}.
1979          */
CallerRunsPolicy()1980         public CallerRunsPolicy() { }
1981 
1982         /**
1983          * Executes task r in the caller's thread, unless the executor
1984          * has been shut down, in which case the task is discarded.
1985          *
1986          * @param r the runnable task requested to be executed
1987          * @param e the executor attempting to execute this task
1988          */
rejectedExecution(Runnable r, ThreadPoolExecutor e)1989         public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
1990             if (!e.isShutdown()) {
1991                 r.run();
1992             }
1993         }
1994     }
1995 
1996     /**
1997      * A handler for rejected tasks that throws a
1998      * {@code RejectedExecutionException}.
1999      */
2000     public static class AbortPolicy implements RejectedExecutionHandler {
2001         /**
2002          * Creates an {@code AbortPolicy}.
2003          */
AbortPolicy()2004         public AbortPolicy() { }
2005 
2006         /**
2007          * Always throws RejectedExecutionException.
2008          *
2009          * @param r the runnable task requested to be executed
2010          * @param e the executor attempting to execute this task
2011          * @throws RejectedExecutionException always
2012          */
rejectedExecution(Runnable r, ThreadPoolExecutor e)2013         public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
2014             throw new RejectedExecutionException("Task " + r.toString() +
2015                                                  " rejected from " +
2016                                                  e.toString());
2017         }
2018     }
2019 
2020     /**
2021      * A handler for rejected tasks that silently discards the
2022      * rejected task.
2023      */
2024     public static class DiscardPolicy implements RejectedExecutionHandler {
2025         /**
2026          * Creates a {@code DiscardPolicy}.
2027          */
DiscardPolicy()2028         public DiscardPolicy() { }
2029 
2030         /**
2031          * Does nothing, which has the effect of discarding task r.
2032          *
2033          * @param r the runnable task requested to be executed
2034          * @param e the executor attempting to execute this task
2035          */
rejectedExecution(Runnable r, ThreadPoolExecutor e)2036         public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
2037         }
2038     }
2039 
2040     /**
2041      * A handler for rejected tasks that discards the oldest unhandled
2042      * request and then retries {@code execute}, unless the executor
2043      * is shut down, in which case the task is discarded.
2044      */
2045     public static class DiscardOldestPolicy implements RejectedExecutionHandler {
2046         /**
2047          * Creates a {@code DiscardOldestPolicy} for the given executor.
2048          */
DiscardOldestPolicy()2049         public DiscardOldestPolicy() { }
2050 
2051         /**
2052          * Obtains and ignores the next task that the executor
2053          * would otherwise execute, if one is immediately available,
2054          * and then retries execution of task r, unless the executor
2055          * is shut down, in which case task r is instead discarded.
2056          *
2057          * @param r the runnable task requested to be executed
2058          * @param e the executor attempting to execute this task
2059          */
rejectedExecution(Runnable r, ThreadPoolExecutor e)2060         public void rejectedExecution(Runnable r, ThreadPoolExecutor e) {
2061             if (!e.isShutdown()) {
2062                 e.getQueue().poll();
2063                 e.execute(r);
2064             }
2065         }
2066     }
2067 }
2068