threading.rst 38.3 KB
Newer Older
1 2
:mod:`threading` --- Thread-based parallelism
=============================================
3 4

.. module:: threading
5
   :synopsis: Thread-based parallelism.
6

7
**Source code:** :source:`Lib/threading.py`
8

9 10
--------------

11 12
This module constructs higher-level threading interfaces on top of the lower
level :mod:`_thread` module.  See also the :mod:`queue` module.
13

14 15
.. versionchanged:: 3.7
   This module used to be optional, it is now always available.
16

17 18
.. note::

19 20 21
   While they are not listed below, the ``camelCase`` names used for some
   methods and functions in this module in the Python 2.x series are still
   supported by this module.
22

23

24
This module defines the following functions:
25 26


27
.. function:: active_count()
28 29

   Return the number of :class:`Thread` objects currently alive.  The returned
Benjamin Peterson's avatar
Benjamin Peterson committed
30
   count is equal to the length of the list returned by :func:`.enumerate`.
31 32


33
.. function:: current_thread()
34 35 36 37 38 39 40

   Return the current :class:`Thread` object, corresponding to the caller's thread
   of control.  If the caller's thread of control was not created through the
   :mod:`threading` module, a dummy thread object with limited functionality is
   returned.


41 42 43 44 45 46 47 48 49 50 51
.. function:: get_ident()

   Return the 'thread identifier' of the current thread.  This is a nonzero
   integer.  Its value has no direct meaning; it is intended as a magic cookie
   to be used e.g. to index a dictionary of thread-specific data.  Thread
   identifiers may be recycled when a thread exits and another thread is
   created.

   .. versionadded:: 3.3


52 53
.. function:: enumerate()

54 55 56 57
   Return a list of all :class:`Thread` objects currently alive.  The list
   includes daemonic threads, dummy thread objects created by
   :func:`current_thread`, and the main thread.  It excludes terminated threads
   and threads that have not yet been started.
58 59


60 61 62 63 64 65 66 67 68
.. function:: main_thread()

   Return the main :class:`Thread` object.  In normal conditions, the
   main thread is the thread from which the Python interpreter was
   started.

   .. versionadded:: 3.4


69 70 71 72 73 74
.. function:: settrace(func)

   .. index:: single: trace function

   Set a trace function for all threads started from the :mod:`threading` module.
   The *func* will be passed to  :func:`sys.settrace` for each thread, before its
75
   :meth:`~Thread.run` method is called.
76 77 78 79 80 81 82 83


.. function:: setprofile(func)

   .. index:: single: profile function

   Set a profile function for all threads started from the :mod:`threading` module.
   The *func* will be passed to  :func:`sys.setprofile` for each thread, before its
84
   :meth:`~Thread.run` method is called.
85 86 87 88 89 90 91


.. function:: stack_size([size])

   Return the thread stack size used when creating new threads.  The optional
   *size* argument specifies the stack size to be used for subsequently created
   threads, and must be 0 (use platform or configured default) or a positive
92 93
   integer value of at least 32,768 (32 KiB). If *size* is not specified,
   0 is used.  If changing the thread stack size is
94
   unsupported, a :exc:`RuntimeError` is raised.  If the specified stack size is
95
   invalid, a :exc:`ValueError` is raised and the stack size is unmodified.  32 KiB
96 97 98
   is currently the minimum supported stack size value to guarantee sufficient
   stack space for the interpreter itself.  Note that some platforms may have
   particular restrictions on values for the stack size, such as requiring a
99
   minimum stack size > 32 KiB or requiring allocation in multiples of the system
100
   memory page size - platform documentation should be referred to for more
101
   information (4 KiB pages are common; using multiples of 4096 for the stack size is
102
   the suggested approach in the absence of more specific information).
103 104

   .. availability:: Windows, systems with POSIX threads.
105 106


107 108 109 110 111 112
This module also defines the following constant:

.. data:: TIMEOUT_MAX

   The maximum value allowed for the *timeout* parameter of blocking functions
   (:meth:`Lock.acquire`, :meth:`RLock.acquire`, :meth:`Condition.wait`, etc.).
Georg Brandl's avatar
Georg Brandl committed
113
   Specifying a timeout greater than this value will raise an
114 115
   :exc:`OverflowError`.

116
   .. versionadded:: 3.2
117

118

119 120
This module defines a number of classes, which are detailed in the sections
below.
121 122 123 124 125 126 127 128 129 130 131 132

The design of this module is loosely based on Java's threading model. However,
where Java makes locks and condition variables basic behavior of every object,
they are separate objects in Python.  Python's :class:`Thread` class supports a
subset of the behavior of Java's Thread class; currently, there are no
priorities, no thread groups, and threads cannot be destroyed, stopped,
suspended, resumed, or interrupted.  The static methods of Java's Thread class,
when implemented, are mapped to module-level functions.

All of the methods described below are executed atomically.


133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153
Thread-Local Data
-----------------

Thread-local data is data whose values are thread specific.  To manage
thread-local data, just create an instance of :class:`local` (or a
subclass) and store attributes on it::

  mydata = threading.local()
  mydata.x = 1

The instance's values will be different for separate threads.


.. class:: local()

   A class that represents thread-local data.

   For more details and extensive examples, see the documentation string of the
   :mod:`_threading_local` module.


154 155 156 157 158
.. _thread-objects:

Thread Objects
--------------

159 160 161 162 163 164
The :class:`Thread` class represents an activity that is run in a separate
thread of control.  There are two ways to specify the activity: by passing a
callable object to the constructor, or by overriding the :meth:`~Thread.run`
method in a subclass.  No other methods (except for the constructor) should be
overridden in a subclass.  In other words, *only*  override the
:meth:`~Thread.__init__` and :meth:`~Thread.run` methods of this class.
165 166

Once a thread object is created, its activity must be started by calling the
167 168
thread's :meth:`~Thread.start` method.  This invokes the :meth:`~Thread.run`
method in a separate thread of control.
169 170

Once the thread's activity is started, the thread is considered 'alive'. It
171 172 173
stops being alive when its :meth:`~Thread.run` method terminates -- either
normally, or by raising an unhandled exception.  The :meth:`~Thread.is_alive`
method tests whether the thread is alive.
174

175 176 177
Other threads can call a thread's :meth:`~Thread.join` method.  This blocks
the calling thread until the thread whose :meth:`~Thread.join` method is
called is terminated.
178 179

A thread has a name.  The name can be passed to the constructor, and read or
180
changed through the :attr:`~Thread.name` attribute.
181 182 183 184

A thread can be flagged as a "daemon thread".  The significance of this flag is
that the entire Python program exits when only daemon threads are left.  The
initial value is inherited from the creating thread.  The flag can be set
Antoine Pitrou's avatar
Antoine Pitrou committed
185 186
through the :attr:`~Thread.daemon` property or the *daemon* constructor
argument.
187

188 189 190 191 192 193
.. note::
   Daemon threads are abruptly stopped at shutdown.  Their resources (such
   as open files, database transactions, etc.) may not be released properly.
   If you want your threads to stop gracefully, make them non-daemonic and
   use a suitable signalling mechanism such as an :class:`Event`.

194 195 196 197 198 199 200
There is a "main thread" object; this corresponds to the initial thread of
control in the Python program.  It is not a daemon thread.

There is the possibility that "dummy thread objects" are created. These are
thread objects corresponding to "alien threads", which are threads of control
started outside the threading module, such as directly from C code.  Dummy
thread objects have limited functionality; they are always considered alive and
201 202
daemonic, and cannot be :meth:`~Thread.join`\ ed.  They are never deleted,
since it is impossible to detect the termination of alien threads.
203 204


Ezio Melotti's avatar
Ezio Melotti committed
205 206
.. class:: Thread(group=None, target=None, name=None, args=(), kwargs={}, *, \
                  daemon=None)
207

208 209
   This constructor should always be called with keyword arguments.  Arguments
   are:
210 211 212 213 214 215 216

   *group* should be ``None``; reserved for future extension when a
   :class:`ThreadGroup` class is implemented.

   *target* is the callable object to be invoked by the :meth:`run` method.
   Defaults to ``None``, meaning nothing is called.

217 218
   *name* is the thread name.  By default, a unique name is constructed of the
   form "Thread-*N*" where *N* is a small decimal number.
219 220 221 222 223 224

   *args* is the argument tuple for the target invocation.  Defaults to ``()``.

   *kwargs* is a dictionary of keyword arguments for the target invocation.
   Defaults to ``{}``.

225 226 227 228
   If not ``None``, *daemon* explicitly sets whether the thread is daemonic.
   If ``None`` (the default), the daemonic property is inherited from the
   current thread.

229 230 231
   If the subclass overrides the constructor, it must make sure to invoke the
   base class constructor (``Thread.__init__()``) before doing anything else to
   the thread.
232

233 234 235
   .. versionchanged:: 3.3
      Added the *daemon* argument.

236
   .. method:: start()
237

238
      Start the thread's activity.
239

240
      It must be called at most once per thread object.  It arranges for the
241 242
      object's :meth:`~Thread.run` method to be invoked in a separate thread
      of control.
243

244
      This method will raise a :exc:`RuntimeError` if called more than once
245
      on the same thread object.
246

247
   .. method:: run()
248

249
      Method representing the thread's activity.
250

251 252
      You may override this method in a subclass.  The standard :meth:`run`
      method invokes the callable object passed to the object's constructor as
253
      the *target* argument, if any, with positional and keyword arguments taken
254
      from the *args* and *kwargs* arguments, respectively.
255

256
   .. method:: join(timeout=None)
257

258 259
      Wait until the thread terminates. This blocks the calling thread until
      the thread whose :meth:`~Thread.join` method is called terminates -- either
260
      normally or through an unhandled exception -- or until the optional
261
      timeout occurs.
262

263 264
      When the *timeout* argument is present and not ``None``, it should be a
      floating point number specifying a timeout for the operation in seconds
265 266 267 268
      (or fractions thereof). As :meth:`~Thread.join` always returns ``None``,
      you must call :meth:`~Thread.is_alive` after :meth:`~Thread.join` to
      decide whether a timeout happened -- if the thread is still alive, the
      :meth:`~Thread.join` call timed out.
269

270 271
      When the *timeout* argument is not present or ``None``, the operation will
      block until the thread terminates.
272

273
      A thread can be :meth:`~Thread.join`\ ed many times.
274

275 276 277 278
      :meth:`~Thread.join` raises a :exc:`RuntimeError` if an attempt is made
      to join the current thread as that would cause a deadlock. It is also
      an error to :meth:`~Thread.join` a thread before it has been started
      and attempts to do so raise the same exception.
279

280
   .. attribute:: name
281

282 283 284
      A string used for identification purposes only. It has no semantics.
      Multiple threads may be given the same name.  The initial name is set by
      the constructor.
285

286 287
   .. method:: getName()
               setName()
288

289 290
      Old getter/setter API for :attr:`~Thread.name`; use it directly as a
      property instead.
291

292
   .. attribute:: ident
293

294
      The 'thread identifier' of this thread or ``None`` if the thread has not
295 296 297 298
      been started.  This is a nonzero integer.  See the :func:`get_ident`
      function.  Thread identifiers may be recycled when a thread exits and
      another thread is created.  The identifier is available even after the
      thread has exited.
299

300
   .. method:: is_alive()
301

302
      Return whether the thread is alive.
303

304 305 306
      This method returns ``True`` just before the :meth:`~Thread.run` method
      starts until just after the :meth:`~Thread.run` method terminates.  The
      module function :func:`.enumerate` returns a list of all alive threads.
307

308
   .. attribute:: daemon
309

310
      A boolean value indicating whether this thread is a daemon thread (True)
311
      or not (False).  This must be set before :meth:`~Thread.start` is called,
312 313
      otherwise :exc:`RuntimeError` is raised.  Its initial value is inherited
      from the creating thread; the main thread is not a daemon thread and
314 315
      therefore all threads created in the main thread default to
      :attr:`~Thread.daemon` = ``False``.
316

317
      The entire Python program exits when no alive non-daemon threads are left.
318

319 320
   .. method:: isDaemon()
               setDaemon()
321

322 323
      Old getter/setter API for :attr:`~Thread.daemon`; use it directly as a
      property instead.
324 325


326 327
.. impl-detail::

Ezio Melotti's avatar
Ezio Melotti committed
328
   In CPython, due to the :term:`Global Interpreter Lock`, only one thread
329 330
   can execute Python code at once (even though certain performance-oriented
   libraries might overcome this limitation).
Ezio Melotti's avatar
Ezio Melotti committed
331
   If you want your application to make better use of the computational
332 333 334 335 336 337
   resources of multi-core machines, you are advised to use
   :mod:`multiprocessing` or :class:`concurrent.futures.ProcessPoolExecutor`.
   However, threading is still an appropriate model if you want to run
   multiple I/O-bound tasks simultaneously.


338 339 340 341 342 343 344
.. _lock-objects:

Lock Objects
------------

A primitive lock is a synchronization primitive that is not owned by a
particular thread when locked.  In Python, it is currently the lowest level
345
synchronization primitive available, implemented directly by the :mod:`_thread`
346 347 348
extension module.

A primitive lock is in one of two states, "locked" or "unlocked". It is created
349 350 351 352 353 354 355 356 357 358
in the unlocked state.  It has two basic methods, :meth:`~Lock.acquire` and
:meth:`~Lock.release`.  When the state is unlocked, :meth:`~Lock.acquire`
changes the state to locked and returns immediately.  When the state is locked,
:meth:`~Lock.acquire` blocks until a call to :meth:`~Lock.release` in another
thread changes it to unlocked, then the :meth:`~Lock.acquire` call resets it
to locked and returns.  The :meth:`~Lock.release` method should only be
called in the locked state; it changes the state to unlocked and returns
immediately. If an attempt is made to release an unlocked lock, a
:exc:`RuntimeError` will be raised.

359
Locks also support the :ref:`context management protocol <with-locks>`.
360

361 362 363 364
When more than one thread is blocked in :meth:`~Lock.acquire` waiting for the
state to turn to unlocked, only one thread proceeds when a :meth:`~Lock.release`
call resets the state to unlocked; which one of the waiting threads proceeds
is not defined, and may vary across implementations.
365 366 367 368

All methods are executed atomically.


369
.. class:: Lock()
370

371 372 373
   The class implementing primitive lock objects.  Once a thread has acquired a
   lock, subsequent attempts to acquire it block, until it is released; any
   thread may release it.
374

375 376 377
   Note that ``Lock`` is actually a factory function which returns an instance
   of the most efficient version of the concrete Lock class that is supported
   by the platform.
378 379


380
   .. method:: acquire(blocking=True, timeout=-1)
381

382
      Acquire a lock, blocking or non-blocking.
383

384 385
      When invoked with the *blocking* argument set to ``True`` (the default),
      block until the lock is unlocked, then set it to locked and return ``True``.
386

387 388 389
      When invoked with the *blocking* argument set to ``False``, do not block.
      If a call with *blocking* set to ``True`` would block, return ``False``
      immediately; otherwise, set the lock to locked and return ``True``.
390

391 392
      When invoked with the floating-point *timeout* argument set to a positive
      value, block for at most the number of seconds specified by *timeout*
393
      and as long as the lock cannot be acquired.  A *timeout* argument of ``-1``
394 395
      specifies an unbounded wait.  It is forbidden to specify a *timeout*
      when *blocking* is false.
396

397 398
      The return value is ``True`` if the lock is acquired successfully,
      ``False`` if not (for example if the *timeout* expired).
399

400 401
      .. versionchanged:: 3.2
         The *timeout* parameter is new.
402

403
      .. versionchanged:: 3.2
404 405
         Lock acquisition can now be interrupted by signals on POSIX if the
         underlying threading implementation supports it.
406 407 408 409 410 411 412 413 414 415


   .. method:: release()

      Release a lock.  This can be called from any thread, not only the thread
      which has acquired the lock.

      When the lock is locked, reset it to unlocked, and return.  If any other threads
      are blocked waiting for the lock to become unlocked, allow exactly one of them
      to proceed.
416

417
      When invoked on an unlocked lock, a :exc:`RuntimeError` is raised.
418

419
      There is no return value.
420 421 422 423 424 425 426 427 428 429 430 431 432


.. _rlock-objects:

RLock Objects
-------------

A reentrant lock is a synchronization primitive that may be acquired multiple
times by the same thread.  Internally, it uses the concepts of "owning thread"
and "recursion level" in addition to the locked/unlocked state used by primitive
locks.  In the locked state, some thread owns the lock; in the unlocked state,
no thread owns it.

433 434 435 436 437 438
To lock the lock, a thread calls its :meth:`~RLock.acquire` method; this
returns once the thread owns the lock.  To unlock the lock, a thread calls
its :meth:`~Lock.release` method. :meth:`~Lock.acquire`/:meth:`~Lock.release`
call pairs may be nested; only the final :meth:`~Lock.release` (the
:meth:`~Lock.release` of the outermost pair) resets the lock to unlocked and
allows another thread blocked in :meth:`~Lock.acquire` to proceed.
439

440
Reentrant locks also support the :ref:`context management protocol <with-locks>`.
441 442


443
.. class:: RLock()
444

445 446 447 448
   This class implements reentrant lock objects.  A reentrant lock must be
   released by the thread that acquired it.  Once a thread has acquired a
   reentrant lock, the same thread may acquire it again without blocking; the
   thread must release it once for each time it has acquired it.
449

450 451 452
   Note that ``RLock`` is actually a factory function which returns an instance
   of the most efficient version of the concrete RLock class that is supported
   by the platform.
453 454


455
   .. method:: acquire(blocking=True, timeout=-1)
456

457
      Acquire a lock, blocking or non-blocking.
458

459 460 461 462 463 464 465
      When invoked without arguments: if this thread already owns the lock, increment
      the recursion level by one, and return immediately.  Otherwise, if another
      thread owns the lock, block until the lock is unlocked.  Once the lock is
      unlocked (not owned by any thread), then grab ownership, set the recursion level
      to one, and return.  If more than one thread is blocked waiting until the lock
      is unlocked, only one at a time will be able to grab ownership of the lock.
      There is no return value in this case.
466

467 468
      When invoked with the *blocking* argument set to true, do the same thing as when
      called without arguments, and return true.
469

470 471 472
      When invoked with the *blocking* argument set to false, do not block.  If a call
      without an argument would block, return false immediately; otherwise, do the
      same thing as when called without arguments, and return true.
473

474 475 476 477
      When invoked with the floating-point *timeout* argument set to a positive
      value, block for at most the number of seconds specified by *timeout*
      and as long as the lock cannot be acquired.  Return true if the lock has
      been acquired, false if the timeout has elapsed.
478

479 480
      .. versionchanged:: 3.2
         The *timeout* parameter is new.
481

482 483 484 485 486 487 488 489 490 491 492 493 494 495

   .. method:: release()

      Release a lock, decrementing the recursion level.  If after the decrement it is
      zero, reset the lock to unlocked (not owned by any thread), and if any other
      threads are blocked waiting for the lock to become unlocked, allow exactly one
      of them to proceed.  If after the decrement the recursion level is still
      nonzero, the lock remains locked and owned by the calling thread.

      Only call this method when the calling thread owns the lock. A
      :exc:`RuntimeError` is raised if this method is called when the lock is
      unlocked.

      There is no return value.
496 497 498 499 500 501 502 503


.. _condition-objects:

Condition Objects
-----------------

A condition variable is always associated with some kind of lock; this can be
504 505 506 507
passed in or one will be created by default.  Passing one in is useful when
several condition variables must share the same lock.  The lock is part of
the condition object: you don't have to track it separately.

508
A condition variable obeys the :ref:`context management protocol <with-locks>`:
509 510 511 512
using the ``with`` statement acquires the associated lock for the duration of
the enclosed block.  The :meth:`~Condition.acquire` and
:meth:`~Condition.release` methods also call the corresponding methods of
the associated lock.
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530

Other methods must be called with the associated lock held.  The
:meth:`~Condition.wait` method releases the lock, and then blocks until
another thread awakens it by calling :meth:`~Condition.notify` or
:meth:`~Condition.notify_all`.  Once awakened, :meth:`~Condition.wait`
re-acquires the lock and returns.  It is also possible to specify a timeout.

The :meth:`~Condition.notify` method wakes up one of the threads waiting for
the condition variable, if any are waiting.  The :meth:`~Condition.notify_all`
method wakes up all threads waiting for the condition variable.

Note: the :meth:`~Condition.notify` and :meth:`~Condition.notify_all` methods
don't release the lock; this means that the thread or threads awakened will
not return from their :meth:`~Condition.wait` call immediately, but only when
the thread that called :meth:`~Condition.notify` or :meth:`~Condition.notify_all`
finally relinquishes ownership of the lock.

The typical programming style using condition variables uses the lock to
531
synchronize access to some shared state; threads that are interested in a
532 533 534 535 536 537
particular change of state call :meth:`~Condition.wait` repeatedly until they
see the desired state, while threads that modify the state call
:meth:`~Condition.notify` or :meth:`~Condition.notify_all` when they change
the state in such a way that it could possibly be a desired state for one
of the waiters.  For example, the following code is a generic
producer-consumer situation with unlimited buffer capacity::
538 539

   # Consume one item
540 541 542 543
   with cv:
       while not an_item_is_available():
           cv.wait()
       get_an_available_item()
544 545

   # Produce one item
546 547
   with cv:
       make_an_item_available()
548
       cv.notify()
549 550 551

The ``while`` loop checking for the application's condition is necessary
because :meth:`~Condition.wait` can return after an arbitrary long time,
552 553 554 555
and the condition which prompted the :meth:`~Condition.notify` call may
no longer hold true.  This is inherent to multi-threaded programming.  The
:meth:`~Condition.wait_for` method can be used to automate the condition
checking, and eases the computation of timeouts::
556

557 558 559 560
   # Consume an item
   with cv:
       cv.wait_for(an_item_is_available)
       get_an_available_item()
561

562 563 564 565
To choose between :meth:`~Condition.notify` and :meth:`~Condition.notify_all`,
consider whether one state change can be interesting for only one or several
waiting threads.  E.g. in a typical producer-consumer situation, adding one
item to the buffer only needs to wake up one consumer thread.
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
566

567

568
.. class:: Condition(lock=None)
569

570 571 572
   This class implements condition variable objects.  A condition variable
   allows one or more threads to wait until they are notified by another thread.

573 574 575
   If the *lock* argument is given and not ``None``, it must be a :class:`Lock`
   or :class:`RLock` object, and it is used as the underlying lock.  Otherwise,
   a new :class:`RLock` object is created and used as the underlying lock.
576

577 578 579
   .. versionchanged:: 3.3
      changed from a factory function to a class.

580
   .. method:: acquire(*args)
581

582 583
      Acquire the underlying lock. This method calls the corresponding method on
      the underlying lock; the return value is whatever that method returns.
584

585
   .. method:: release()
586

587 588
      Release the underlying lock. This method calls the corresponding method on
      the underlying lock; there is no return value.
589

590
   .. method:: wait(timeout=None)
591

592 593 594
      Wait until notified or until a timeout occurs. If the calling thread has
      not acquired the lock when this method is called, a :exc:`RuntimeError` is
      raised.
595

596 597 598 599
      This method releases the underlying lock, and then blocks until it is
      awakened by a :meth:`notify` or :meth:`notify_all` call for the same
      condition variable in another thread, or until the optional timeout
      occurs.  Once awakened or timed out, it re-acquires the lock and returns.
600

601 602 603
      When the *timeout* argument is present and not ``None``, it should be a
      floating point number specifying a timeout for the operation in seconds
      (or fractions thereof).
604

605 606 607 608 609 610 611
      When the underlying lock is an :class:`RLock`, it is not released using
      its :meth:`release` method, since this may not actually unlock the lock
      when it was acquired multiple times recursively.  Instead, an internal
      interface of the :class:`RLock` class is used, which really unlocks it
      even when it has been recursively acquired several times. Another internal
      interface is then used to restore the recursion level when the lock is
      reacquired.
612

613 614 615 616 617 618
      The return value is ``True`` unless a given *timeout* expired, in which
      case it is ``False``.

      .. versionchanged:: 3.2
         Previously, the method always returned ``None``.

Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
619 620
   .. method:: wait_for(predicate, timeout=None)

621
      Wait until a condition evaluates to true.  *predicate* should be a
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
622 623 624 625 626 627 628 629 630 631 632 633 634 635 636
      callable which result will be interpreted as a boolean value.
      A *timeout* may be provided giving the maximum time to wait.

      This utility method may call :meth:`wait` repeatedly until the predicate
      is satisfied, or until a timeout occurs. The return value is
      the last return value of the predicate and will evaluate to
      ``False`` if the method timed out.

      Ignoring the timeout feature, calling this method is roughly equivalent to
      writing::

        while not predicate():
            cv.wait()

      Therefore, the same rules apply as with :meth:`wait`: The lock must be
637
      held when called and is re-acquired on return.  The predicate is evaluated
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
638 639 640 641
      with the lock held.

      .. versionadded:: 3.2

642
   .. method:: notify(n=1)
643

644 645
      By default, wake up one thread waiting on this condition, if any.  If the
      calling thread has not acquired the lock when this method is called, a
646
      :exc:`RuntimeError` is raised.
647

648 649
      This method wakes up at most *n* of the threads waiting for the condition
      variable; it is a no-op if no threads are waiting.
650

651 652 653 654
      The current implementation wakes up exactly *n* threads, if at least *n*
      threads are waiting.  However, it's not safe to rely on this behavior.
      A future, optimized implementation may occasionally wake up more than
      *n* threads.
655

656
      Note: an awakened thread does not actually return from its :meth:`wait`
657 658
      call until it can reacquire the lock.  Since :meth:`notify` does not
      release the lock, its caller should.
659

660
   .. method:: notify_all()
661

662 663 664 665
      Wake up all threads waiting on this condition.  This method acts like
      :meth:`notify`, but wakes up all waiting threads instead of one. If the
      calling thread has not acquired the lock when this method is called, a
      :exc:`RuntimeError` is raised.
666 667 668 669 670 671 672 673 674


.. _semaphore-objects:

Semaphore Objects
-----------------

This is one of the oldest synchronization primitives in the history of computer
science, invented by the early Dutch computer scientist Edsger W. Dijkstra (he
675 676
used the names ``P()`` and ``V()`` instead of :meth:`~Semaphore.acquire` and
:meth:`~Semaphore.release`).
677 678

A semaphore manages an internal counter which is decremented by each
679 680 681 682
:meth:`~Semaphore.acquire` call and incremented by each :meth:`~Semaphore.release`
call.  The counter can never go below zero; when :meth:`~Semaphore.acquire`
finds that it is zero, it blocks, waiting until some other thread calls
:meth:`~Semaphore.release`.
683

684
Semaphores also support the :ref:`context management protocol <with-locks>`.
685 686


687
.. class:: Semaphore(value=1)
688

689 690
   This class implements semaphore objects.  A semaphore manages an atomic
   counter representing the number of :meth:`release` calls minus the number of
691 692 693 694
   :meth:`acquire` calls, plus an initial value.  The :meth:`acquire` method
   blocks if necessary until it can return without making the counter negative.
   If not given, *value* defaults to 1.

695 696 697 698
   The optional argument gives the initial *value* for the internal counter; it
   defaults to ``1``. If the *value* given is less than 0, :exc:`ValueError` is
   raised.

699 700 701
   .. versionchanged:: 3.3
      changed from a factory function to a class.

702
   .. method:: acquire(blocking=True, timeout=None)
703

704
      Acquire a semaphore.
705

706 707 708 709 710 711 712 713 714
      When invoked without arguments:

      * If the internal counter is larger than zero on entry, decrement it by
        one and return true immediately.
      * If the internal counter is zero on entry, block until awoken by a call to
        :meth:`~Semaphore.release`.  Once awoken (and the counter is greater
        than 0), decrement the counter by 1 and return true.  Exactly one
        thread will be awoken by each call to :meth:`~Semaphore.release`.  The
        order in which threads are awoken should not be relied on.
715

716
      When invoked with *blocking* set to false, do not block.  If a call
717 718
      without an argument would block, return false immediately; otherwise, do
      the same thing as when called without arguments, and return true.
719

720
      When invoked with a *timeout* other than ``None``, it will block for at
721 722 723 724 725
      most *timeout* seconds.  If acquire does not complete successfully in
      that interval, return false.  Return true otherwise.

      .. versionchanged:: 3.2
         The *timeout* parameter is new.
726

727
   .. method:: release()
728

729 730 731
      Release a semaphore, incrementing the internal counter by one.  When it
      was zero on entry and another thread is waiting for it to become larger
      than zero again, wake up that thread.
732 733


734 735 736 737 738 739 740 741 742 743 744 745
.. class:: BoundedSemaphore(value=1)

   Class implementing bounded semaphore objects.  A bounded semaphore checks to
   make sure its current value doesn't exceed its initial value.  If it does,
   :exc:`ValueError` is raised. In most situations semaphores are used to guard
   resources with limited capacity.  If the semaphore is released too many times
   it's a sign of a bug.  If not given, *value* defaults to 1.

   .. versionchanged:: 3.3
      changed from a factory function to a class.


746 747 748 749 750 751
.. _semaphore-examples:

:class:`Semaphore` Example
^^^^^^^^^^^^^^^^^^^^^^^^^^

Semaphores are often used to guard resources with limited capacity, for example,
Georg Brandl's avatar
Georg Brandl committed
752 753 754
a database server.  In any situation where the size of the resource is fixed,
you should use a bounded semaphore.  Before spawning any worker threads, your
main thread would initialize the semaphore::
755 756

   maxconnections = 5
757
   # ...
758 759 760 761 762
   pool_sema = BoundedSemaphore(value=maxconnections)

Once spawned, worker threads call the semaphore's acquire and release methods
when they need to connect to the server::

763 764 765
   with pool_sema:
       conn = connectdb()
       try:
766
           # ... use connection ...
767 768
       finally:
           conn.close()
769 770 771 772 773 774 775 776 777 778 779 780 781 782

The use of a bounded semaphore reduces the chance that a programming error which
causes the semaphore to be released more than it's acquired will go undetected.


.. _event-objects:

Event Objects
-------------

This is one of the simplest mechanisms for communication between threads: one
thread signals an event and other threads wait for it.

An event object manages an internal flag that can be set to true with the
783 784
:meth:`~Event.set` method and reset to false with the :meth:`~Event.clear`
method.  The :meth:`~Event.wait` method blocks until the flag is true.
785 786 787 788


.. class:: Event()

789 790 791 792 793 794 795
   Class implementing event objects.  An event manages a flag that can be set to
   true with the :meth:`~Event.set` method and reset to false with the
   :meth:`clear` method.  The :meth:`wait` method blocks until the flag is true.
   The flag is initially false.

   .. versionchanged:: 3.3
      changed from a factory function to a class.
796

797
   .. method:: is_set()
798

799
      Return true if and only if the internal flag is true.
800

801
   .. method:: set()
802

803 804 805
      Set the internal flag to true. All threads waiting for it to become true
      are awakened. Threads that call :meth:`wait` once the flag is true will
      not block at all.
806

807
   .. method:: clear()
808

809
      Reset the internal flag to false. Subsequently, threads calling
810
      :meth:`wait` will block until :meth:`.set` is called to set the internal
811
      flag to true again.
812

813
   .. method:: wait(timeout=None)
814

815 816
      Block until the internal flag is true.  If the internal flag is true on
      entry, return immediately.  Otherwise, block until another thread calls
817
      :meth:`.set` to set the flag to true, or until the optional timeout occurs.
818

819 820 821
      When the timeout argument is present and not ``None``, it should be a
      floating point number specifying a timeout for the operation in seconds
      (or fractions thereof).
822

823 824 825 826
      This method returns true if and only if the internal flag has been set to
      true, either before the wait call or after the wait starts, so it will
      always return ``True`` except if a timeout is given and the operation
      times out.
827

828 829
      .. versionchanged:: 3.1
         Previously, the method always returned ``None``.
Benjamin Peterson's avatar
Benjamin Peterson committed
830

831 832 833 834 835 836 837 838 839 840

.. _timer-objects:

Timer Objects
-------------

This class represents an action that should be run only after a certain amount
of time has passed --- a timer.  :class:`Timer` is a subclass of :class:`Thread`
and as such also functions as an example of creating custom threads.

841 842 843 844 845
Timers are started, as with threads, by calling their :meth:`~Timer.start`
method.  The timer can be stopped (before its action has begun) by calling the
:meth:`~Timer.cancel` method.  The interval the timer will wait before
executing its action may not be exactly the same as the interval specified by
the user.
846 847 848 849

For example::

   def hello():
850
       print("hello, world")
851 852

   t = Timer(30.0, hello)
853
   t.start()  # after 30 seconds, "hello, world" will be printed
854 855


856
.. class:: Timer(interval, function, args=None, kwargs=None)
857 858 859

   Create a timer that will run *function* with arguments *args* and  keyword
   arguments *kwargs*, after *interval* seconds have passed.
860 861
   If *args* is ``None`` (the default) then an empty list will be used.
   If *kwargs* is ``None`` (the default) then an empty dict will be used.
862

863 864 865
   .. versionchanged:: 3.3
      changed from a factory function to a class.

866
   .. method:: cancel()
867

868 869
      Stop the timer, and cancel the execution of the timer's action.  This will
      only work if the timer is still in its waiting stage.
870 871


Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
872 873 874
Barrier Objects
---------------

Georg Brandl's avatar
Georg Brandl committed
875 876 877 878
.. versionadded:: 3.2

This class provides a simple synchronization primitive for use by a fixed number
of threads that need to wait for each other.  Each of the threads tries to pass
879
the barrier by calling the :meth:`~Barrier.wait` method and will block until
880 881
all of the threads have made their :meth:`~Barrier.wait` calls. At this point,
the threads are released simultaneously.
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
882 883 884 885 886 887

The barrier can be reused any number of times for the same number of threads.

As an example, here is a simple way to synchronize a client and server thread::

   b = Barrier(2, timeout=5)
Georg Brandl's avatar
Georg Brandl committed
888 889

   def server():
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
890 891 892 893 894 895
       start_server()
       b.wait()
       while True:
           connection = accept_connection()
           process_server_connection(connection)

Georg Brandl's avatar
Georg Brandl committed
896
   def client():
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
897 898
       b.wait()
       while True:
Georg Brandl's avatar
Georg Brandl committed
899 900 901
           connection = make_connection()
           process_client_connection(connection)

Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
902 903 904

.. class:: Barrier(parties, action=None, timeout=None)

Georg Brandl's avatar
Georg Brandl committed
905 906 907 908
   Create a barrier object for *parties* number of threads.  An *action*, when
   provided, is a callable to be called by one of the threads when they are
   released.  *timeout* is the default timeout value if none is specified for
   the :meth:`wait` method.
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
909 910 911 912

   .. method:: wait(timeout=None)

      Pass the barrier.  When all the threads party to the barrier have called
Georg Brandl's avatar
Georg Brandl committed
913
      this function, they are all released simultaneously.  If a *timeout* is
Ezio Melotti's avatar
Ezio Melotti committed
914
      provided, it is used in preference to any that was supplied to the class
Georg Brandl's avatar
Georg Brandl committed
915
      constructor.
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
916

Georg Brandl's avatar
Georg Brandl committed
917
      The return value is an integer in the range 0 to *parties* -- 1, different
918
      for each thread.  This can be used to select a thread to do some special
Georg Brandl's avatar
Georg Brandl committed
919
      housekeeping, e.g.::
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
920 921 922

         i = barrier.wait()
         if i == 0:
Georg Brandl's avatar
Georg Brandl committed
923 924
             # Only one thread needs to print this
             print("passed the barrier")
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
925

Georg Brandl's avatar
Georg Brandl committed
926 927 928
      If an *action* was provided to the constructor, one of the threads will
      have called it prior to being released.  Should this call raise an error,
      the barrier is put into the broken state.
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
929 930 931 932

      If the call times out, the barrier is put into the broken state.

      This method may raise a :class:`BrokenBarrierError` exception if the
Georg Brandl's avatar
Georg Brandl committed
933
      barrier is broken or reset while a thread is waiting.
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
934 935 936

   .. method:: reset()

Georg Brandl's avatar
Georg Brandl committed
937 938
      Return the barrier to the default, empty state.  Any threads waiting on it
      will receive the :class:`BrokenBarrierError` exception.
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
939 940

      Note that using this function may can require some external
Georg Brandl's avatar
Georg Brandl committed
941 942
      synchronization if there are other threads whose state is unknown.  If a
      barrier is broken it may be better to just leave it and create a new one.
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
943 944 945 946

   .. method:: abort()

      Put the barrier into a broken state.  This causes any active or future
Georg Brandl's avatar
Georg Brandl committed
947 948 949
      calls to :meth:`wait` to fail with the :class:`BrokenBarrierError`.  Use
      this for example if one of the needs to abort, to avoid deadlocking the
      application.
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
950 951

      It may be preferable to simply create the barrier with a sensible
Georg Brandl's avatar
Georg Brandl committed
952 953
      *timeout* value to automatically guard against one of the threads going
      awry.
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
954 955 956 957 958 959 960 961 962 963 964 965 966 967

   .. attribute:: parties

      The number of threads required to pass the barrier.

   .. attribute:: n_waiting

      The number of threads currently waiting in the barrier.

   .. attribute:: broken

      A boolean that is ``True`` if the barrier is in the broken state.


Georg Brandl's avatar
Georg Brandl committed
968
.. exception:: BrokenBarrierError
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
969

Georg Brandl's avatar
Georg Brandl committed
970 971
   This exception, a subclass of :exc:`RuntimeError`, is raised when the
   :class:`Barrier` object is reset or broken.
Kristján Valur Jónsson's avatar
Kristján Valur Jónsson committed
972 973


974 975
.. _with-locks:

976 977
Using locks, conditions, and semaphores in the :keyword:`!with` statement
-------------------------------------------------------------------------
978 979 980

All of the objects provided by this module that have :meth:`acquire` and
:meth:`release` methods can be used as context managers for a :keyword:`with`
981 982 983
statement.  The :meth:`acquire` method will be called when the block is
entered, and :meth:`release` will be called when the block is exited.  Hence,
the following snippet::
984

985 986
   with some_lock:
       # do something...
987

988
is equivalent to::
989

990 991 992 993 994
   some_lock.acquire()
   try:
       # do something...
   finally:
       some_lock.release()
995

996 997 998
Currently, :class:`Lock`, :class:`RLock`, :class:`Condition`,
:class:`Semaphore`, and :class:`BoundedSemaphore` objects may be used as
:keyword:`with` statement context managers.