1"""Thread module emulating a subset of Java's threading model.""" 2 3import os as _os 4import sys as _sys 5import _thread 6import functools 7 8from time import monotonic as _time 9from _weakrefset import WeakSet 10from itertools import islice as _islice, count as _count 11try: 12 from _collections import deque as _deque 13except ImportError: 14 from collections import deque as _deque 15 16# Note regarding PEP 8 compliant names 17# This threading model was originally inspired by Java, and inherited 18# the convention of camelCase function and method names from that 19# language. Those original names are not in any imminent danger of 20# being deprecated (even for Py3k),so this module provides them as an 21# alias for the PEP 8 compliant names 22# Note that using the new PEP 8 compliant names facilitates substitution 23# with the multiprocessing module, which doesn't provide the old 24# Java inspired names. 25 26__all__ = ['get_ident', 'active_count', 'Condition', 'current_thread', 27 'enumerate', 'main_thread', 'TIMEOUT_MAX', 28 'Event', 'Lock', 'RLock', 'Semaphore', 'BoundedSemaphore', 'Thread', 29 'Barrier', 'BrokenBarrierError', 'Timer', 'ThreadError', 30 'setprofile', 'settrace', 'local', 'stack_size', 31 'excepthook', 'ExceptHookArgs'] 32 33# Rename some stuff so "from threading import *" is safe 34_start_new_thread = _thread.start_new_thread 35_allocate_lock = _thread.allocate_lock 36_set_sentinel = _thread._set_sentinel 37get_ident = _thread.get_ident 38try: 39 get_native_id = _thread.get_native_id 40 _HAVE_THREAD_NATIVE_ID = True 41 __all__.append('get_native_id') 42except AttributeError: 43 _HAVE_THREAD_NATIVE_ID = False 44ThreadError = _thread.error 45try: 46 _CRLock = _thread.RLock 47except AttributeError: 48 _CRLock = None 49TIMEOUT_MAX = _thread.TIMEOUT_MAX 50del _thread 51 52 53# Support for profile and trace hooks 54 55_profile_hook = None 56_trace_hook = None 57 58def setprofile(func): 59 """Set a profile function for all threads started from the threading module. 60 61 The func will be passed to sys.setprofile() for each thread, before its 62 run() method is called. 63 64 """ 65 global _profile_hook 66 _profile_hook = func 67 68def settrace(func): 69 """Set a trace function for all threads started from the threading module. 70 71 The func will be passed to sys.settrace() for each thread, before its run() 72 method is called. 73 74 """ 75 global _trace_hook 76 _trace_hook = func 77 78# Synchronization classes 79 80Lock = _allocate_lock 81 82def RLock(*args, **kwargs): 83 """Factory function that returns a new reentrant lock. 84 85 A reentrant lock must be released by the thread that acquired it. Once a 86 thread has acquired a reentrant lock, the same thread may acquire it again 87 without blocking; the thread must release it once for each time it has 88 acquired it. 89 90 """ 91 if _CRLock is None: 92 return _PyRLock(*args, **kwargs) 93 return _CRLock(*args, **kwargs) 94 95class _RLock: 96 """This class implements reentrant lock objects. 97 98 A reentrant lock must be released by the thread that acquired it. Once a 99 thread has acquired a reentrant lock, the same thread may acquire it 100 again without blocking; the thread must release it once for each time it 101 has acquired it. 102 103 """ 104 105 def __init__(self): 106 self._block = _allocate_lock() 107 self._owner = None 108 self._count = 0 109 110 def __repr__(self): 111 owner = self._owner 112 try: 113 owner = _active[owner].name 114 except KeyError: 115 pass 116 return "<%s %s.%s object owner=%r count=%d at %s>" % ( 117 "locked" if self._block.locked() else "unlocked", 118 self.__class__.__module__, 119 self.__class__.__qualname__, 120 owner, 121 self._count, 122 hex(id(self)) 123 ) 124 125 def _at_fork_reinit(self): 126 self._block._at_fork_reinit() 127 self._owner = None 128 self._count = 0 129 130 def acquire(self, blocking=True, timeout=-1): 131 """Acquire a lock, blocking or non-blocking. 132 133 When invoked without arguments: if this thread already owns the lock, 134 increment the recursion level by one, and return immediately. Otherwise, 135 if another thread owns the lock, block until the lock is unlocked. Once 136 the lock is unlocked (not owned by any thread), then grab ownership, set 137 the recursion level to one, and return. If more than one thread is 138 blocked waiting until the lock is unlocked, only one at a time will be 139 able to grab ownership of the lock. There is no return value in this 140 case. 141 142 When invoked with the blocking argument set to true, do the same thing 143 as when called without arguments, and return true. 144 145 When invoked with the blocking argument set to false, do not block. If a 146 call without an argument would block, return false immediately; 147 otherwise, do the same thing as when called without arguments, and 148 return true. 149 150 When invoked with the floating-point timeout argument set to a positive 151 value, block for at most the number of seconds specified by timeout 152 and as long as the lock cannot be acquired. Return true if the lock has 153 been acquired, false if the timeout has elapsed. 154 155 """ 156 me = get_ident() 157 if self._owner == me: 158 self._count += 1 159 return 1 160 rc = self._block.acquire(blocking, timeout) 161 if rc: 162 self._owner = me 163 self._count = 1 164 return rc 165 166 __enter__ = acquire 167 168 def release(self): 169 """Release a lock, decrementing the recursion level. 170 171 If after the decrement it is zero, reset the lock to unlocked (not owned 172 by any thread), and if any other threads are blocked waiting for the 173 lock to become unlocked, allow exactly one of them to proceed. If after 174 the decrement the recursion level is still nonzero, the lock remains 175 locked and owned by the calling thread. 176 177 Only call this method when the calling thread owns the lock. A 178 RuntimeError is raised if this method is called when the lock is 179 unlocked. 180 181 There is no return value. 182 183 """ 184 if self._owner != get_ident(): 185 raise RuntimeError("cannot release un-acquired lock") 186 self._count = count = self._count - 1 187 if not count: 188 self._owner = None 189 self._block.release() 190 191 def __exit__(self, t, v, tb): 192 self.release() 193 194 # Internal methods used by condition variables 195 196 def _acquire_restore(self, state): 197 self._block.acquire() 198 self._count, self._owner = state 199 200 def _release_save(self): 201 if self._count == 0: 202 raise RuntimeError("cannot release un-acquired lock") 203 count = self._count 204 self._count = 0 205 owner = self._owner 206 self._owner = None 207 self._block.release() 208 return (count, owner) 209 210 def _is_owned(self): 211 return self._owner == get_ident() 212 213_PyRLock = _RLock 214 215 216class Condition: 217 """Class that implements a condition variable. 218 219 A condition variable allows one or more threads to wait until they are 220 notified by another thread. 221 222 If the lock argument is given and not None, it must be a Lock or RLock 223 object, and it is used as the underlying lock. Otherwise, a new RLock object 224 is created and used as the underlying lock. 225 226 """ 227 228 def __init__(self, lock=None): 229 if lock is None: 230 lock = RLock() 231 self._lock = lock 232 # Export the lock's acquire() and release() methods 233 self.acquire = lock.acquire 234 self.release = lock.release 235 # If the lock defines _release_save() and/or _acquire_restore(), 236 # these override the default implementations (which just call 237 # release() and acquire() on the lock). Ditto for _is_owned(). 238 try: 239 self._release_save = lock._release_save 240 except AttributeError: 241 pass 242 try: 243 self._acquire_restore = lock._acquire_restore 244 except AttributeError: 245 pass 246 try: 247 self._is_owned = lock._is_owned 248 except AttributeError: 249 pass 250 self._waiters = _deque() 251 252 def _at_fork_reinit(self): 253 self._lock._at_fork_reinit() 254 self._waiters.clear() 255 256 def __enter__(self): 257 return self._lock.__enter__() 258 259 def __exit__(self, *args): 260 return self._lock.__exit__(*args) 261 262 def __repr__(self): 263 return "<Condition(%s, %d)>" % (self._lock, len(self._waiters)) 264 265 def _release_save(self): 266 self._lock.release() # No state to save 267 268 def _acquire_restore(self, x): 269 self._lock.acquire() # Ignore saved state 270 271 def _is_owned(self): 272 # Return True if lock is owned by current_thread. 273 # This method is called only if _lock doesn't have _is_owned(). 274 if self._lock.acquire(False): 275 self._lock.release() 276 return False 277 else: 278 return True 279 280 def wait(self, timeout=None): 281 """Wait until notified or until a timeout occurs. 282 283 If the calling thread has not acquired the lock when this method is 284 called, a RuntimeError is raised. 285 286 This method releases the underlying lock, and then blocks until it is 287 awakened by a notify() or notify_all() call for the same condition 288 variable in another thread, or until the optional timeout occurs. Once 289 awakened or timed out, it re-acquires the lock and returns. 290 291 When the timeout argument is present and not None, it should be a 292 floating point number specifying a timeout for the operation in seconds 293 (or fractions thereof). 294 295 When the underlying lock is an RLock, it is not released using its 296 release() method, since this may not actually unlock the lock when it 297 was acquired multiple times recursively. Instead, an internal interface 298 of the RLock class is used, which really unlocks it even when it has 299 been recursively acquired several times. Another internal interface is 300 then used to restore the recursion level when the lock is reacquired. 301 302 """ 303 if not self._is_owned(): 304 raise RuntimeError("cannot wait on un-acquired lock") 305 waiter = _allocate_lock() 306 waiter.acquire() 307 self._waiters.append(waiter) 308 saved_state = self._release_save() 309 gotit = False 310 try: # restore state no matter what (e.g., KeyboardInterrupt) 311 if timeout is None: 312 waiter.acquire() 313 gotit = True 314 else: 315 if timeout > 0: 316 gotit = waiter.acquire(True, timeout) 317 else: 318 gotit = waiter.acquire(False) 319 return gotit 320 finally: 321 self._acquire_restore(saved_state) 322 if not gotit: 323 try: 324 self._waiters.remove(waiter) 325 except ValueError: 326 pass 327 328 def wait_for(self, predicate, timeout=None): 329 """Wait until a condition evaluates to True. 330 331 predicate should be a callable which result will be interpreted as a 332 boolean value. A timeout may be provided giving the maximum time to 333 wait. 334 335 """ 336 endtime = None 337 waittime = timeout 338 result = predicate() 339 while not result: 340 if waittime is not None: 341 if endtime is None: 342 endtime = _time() + waittime 343 else: 344 waittime = endtime - _time() 345 if waittime <= 0: 346 break 347 self.wait(waittime) 348 result = predicate() 349 return result 350 351 def notify(self, n=1): 352 """Wake up one or more threads waiting on this condition, if any. 353 354 If the calling thread has not acquired the lock when this method is 355 called, a RuntimeError is raised. 356 357 This method wakes up at most n of the threads waiting for the condition 358 variable; it is a no-op if no threads are waiting. 359 360 """ 361 if not self._is_owned(): 362 raise RuntimeError("cannot notify on un-acquired lock") 363 all_waiters = self._waiters 364 waiters_to_notify = _deque(_islice(all_waiters, n)) 365 if not waiters_to_notify: 366 return 367 for waiter in waiters_to_notify: 368 waiter.release() 369 try: 370 all_waiters.remove(waiter) 371 except ValueError: 372 pass 373 374 def notify_all(self): 375 """Wake up all threads waiting on this condition. 376 377 If the calling thread has not acquired the lock when this method 378 is called, a RuntimeError is raised. 379 380 """ 381 self.notify(len(self._waiters)) 382 383 notifyAll = notify_all 384 385 386class Semaphore: 387 """This class implements semaphore objects. 388 389 Semaphores manage a counter representing the number of release() calls minus 390 the number of acquire() calls, plus an initial value. The acquire() method 391 blocks if necessary until it can return without making the counter 392 negative. If not given, value defaults to 1. 393 394 """ 395 396 # After Tim Peters' semaphore class, but not quite the same (no maximum) 397 398 def __init__(self, value=1): 399 if value < 0: 400 raise ValueError("semaphore initial value must be >= 0") 401 self._cond = Condition(Lock()) 402 self._value = value 403 404 def acquire(self, blocking=True, timeout=None): 405 """Acquire a semaphore, decrementing the internal counter by one. 406 407 When invoked without arguments: if the internal counter is larger than 408 zero on entry, decrement it by one and return immediately. If it is zero 409 on entry, block, waiting until some other thread has called release() to 410 make it larger than zero. This is done with proper interlocking so that 411 if multiple acquire() calls are blocked, release() will wake exactly one 412 of them up. The implementation may pick one at random, so the order in 413 which blocked threads are awakened should not be relied on. There is no 414 return value in this case. 415 416 When invoked with blocking set to true, do the same thing as when called 417 without arguments, and return true. 418 419 When invoked with blocking set to false, do not block. If a call without 420 an argument would block, return false immediately; otherwise, do the 421 same thing as when called without arguments, and return true. 422 423 When invoked with a timeout other than None, it will block for at 424 most timeout seconds. If acquire does not complete successfully in 425 that interval, return false. Return true otherwise. 426 427 """ 428 if not blocking and timeout is not None: 429 raise ValueError("can't specify timeout for non-blocking acquire") 430 rc = False 431 endtime = None 432 with self._cond: 433 while self._value == 0: 434 if not blocking: 435 break 436 if timeout is not None: 437 if endtime is None: 438 endtime = _time() + timeout 439 else: 440 timeout = endtime - _time() 441 if timeout <= 0: 442 break 443 self._cond.wait(timeout) 444 else: 445 self._value -= 1 446 rc = True 447 return rc 448 449 __enter__ = acquire 450 451 def release(self, n=1): 452 """Release a semaphore, incrementing the internal counter by one or more. 453 454 When the counter is zero on entry and another thread is waiting for it 455 to become larger than zero again, wake up that thread. 456 457 """ 458 if n < 1: 459 raise ValueError('n must be one or more') 460 with self._cond: 461 self._value += n 462 for i in range(n): 463 self._cond.notify() 464 465 def __exit__(self, t, v, tb): 466 self.release() 467 468 469class BoundedSemaphore(Semaphore): 470 """Implements a bounded semaphore. 471 472 A bounded semaphore checks to make sure its current value doesn't exceed its 473 initial value. If it does, ValueError is raised. In most situations 474 semaphores are used to guard resources with limited capacity. 475 476 If the semaphore is released too many times it's a sign of a bug. If not 477 given, value defaults to 1. 478 479 Like regular semaphores, bounded semaphores manage a counter representing 480 the number of release() calls minus the number of acquire() calls, plus an 481 initial value. The acquire() method blocks if necessary until it can return 482 without making the counter negative. If not given, value defaults to 1. 483 484 """ 485 486 def __init__(self, value=1): 487 Semaphore.__init__(self, value) 488 self._initial_value = value 489 490 def release(self, n=1): 491 """Release a semaphore, incrementing the internal counter by one or more. 492 493 When the counter is zero on entry and another thread is waiting for it 494 to become larger than zero again, wake up that thread. 495 496 If the number of releases exceeds the number of acquires, 497 raise a ValueError. 498 499 """ 500 if n < 1: 501 raise ValueError('n must be one or more') 502 with self._cond: 503 if self._value + n > self._initial_value: 504 raise ValueError("Semaphore released too many times") 505 self._value += n 506 for i in range(n): 507 self._cond.notify() 508 509 510class Event: 511 """Class implementing event objects. 512 513 Events manage a flag that can be set to true with the set() method and reset 514 to false with the clear() method. The wait() method blocks until the flag is 515 true. The flag is initially false. 516 517 """ 518 519 # After Tim Peters' event class (without is_posted()) 520 521 def __init__(self): 522 self._cond = Condition(Lock()) 523 self._flag = False 524 525 def _at_fork_reinit(self): 526 # Private method called by Thread._reset_internal_locks() 527 self._cond._at_fork_reinit() 528 529 def is_set(self): 530 """Return true if and only if the internal flag is true.""" 531 return self._flag 532 533 isSet = is_set 534 535 def set(self): 536 """Set the internal flag to true. 537 538 All threads waiting for it to become true are awakened. Threads 539 that call wait() once the flag is true will not block at all. 540 541 """ 542 with self._cond: 543 self._flag = True 544 self._cond.notify_all() 545 546 def clear(self): 547 """Reset the internal flag to false. 548 549 Subsequently, threads calling wait() will block until set() is called to 550 set the internal flag to true again. 551 552 """ 553 with self._cond: 554 self._flag = False 555 556 def wait(self, timeout=None): 557 """Block until the internal flag is true. 558 559 If the internal flag is true on entry, return immediately. Otherwise, 560 block until another thread calls set() to set the flag to true, or until 561 the optional timeout occurs. 562 563 When the timeout argument is present and not None, it should be a 564 floating point number specifying a timeout for the operation in seconds 565 (or fractions thereof). 566 567 This method returns the internal flag on exit, so it will always return 568 True except if a timeout is given and the operation times out. 569 570 """ 571 with self._cond: 572 signaled = self._flag 573 if not signaled: 574 signaled = self._cond.wait(timeout) 575 return signaled 576 577 578# A barrier class. Inspired in part by the pthread_barrier_* api and 579# the CyclicBarrier class from Java. See 580# http://sourceware.org/pthreads-win32/manual/pthread_barrier_init.html and 581# http://java.sun.com/j2se/1.5.0/docs/api/java/util/concurrent/ 582# CyclicBarrier.html 583# for information. 584# We maintain two main states, 'filling' and 'draining' enabling the barrier 585# to be cyclic. Threads are not allowed into it until it has fully drained 586# since the previous cycle. In addition, a 'resetting' state exists which is 587# similar to 'draining' except that threads leave with a BrokenBarrierError, 588# and a 'broken' state in which all threads get the exception. 589class Barrier: 590 """Implements a Barrier. 591 592 Useful for synchronizing a fixed number of threads at known synchronization 593 points. Threads block on 'wait()' and are simultaneously awoken once they 594 have all made that call. 595 596 """ 597 598 def __init__(self, parties, action=None, timeout=None): 599 """Create a barrier, initialised to 'parties' threads. 600 601 'action' is a callable which, when supplied, will be called by one of 602 the threads after they have all entered the barrier and just prior to 603 releasing them all. If a 'timeout' is provided, it is used as the 604 default for all subsequent 'wait()' calls. 605 606 """ 607 self._cond = Condition(Lock()) 608 self._action = action 609 self._timeout = timeout 610 self._parties = parties 611 self._state = 0 #0 filling, 1, draining, -1 resetting, -2 broken 612 self._count = 0 613 614 def wait(self, timeout=None): 615 """Wait for the barrier. 616 617 When the specified number of threads have started waiting, they are all 618 simultaneously awoken. If an 'action' was provided for the barrier, one 619 of the threads will have executed that callback prior to returning. 620 Returns an individual index number from 0 to 'parties-1'. 621 622 """ 623 if timeout is None: 624 timeout = self._timeout 625 with self._cond: 626 self._enter() # Block while the barrier drains. 627 index = self._count 628 self._count += 1 629 try: 630 if index + 1 == self._parties: 631 # We release the barrier 632 self._release() 633 else: 634 # We wait until someone releases us 635 self._wait(timeout) 636 return index 637 finally: 638 self._count -= 1 639 # Wake up any threads waiting for barrier to drain. 640 self._exit() 641 642 # Block until the barrier is ready for us, or raise an exception 643 # if it is broken. 644 def _enter(self): 645 while self._state in (-1, 1): 646 # It is draining or resetting, wait until done 647 self._cond.wait() 648 #see if the barrier is in a broken state 649 if self._state < 0: 650 raise BrokenBarrierError 651 assert self._state == 0 652 653 # Optionally run the 'action' and release the threads waiting 654 # in the barrier. 655 def _release(self): 656 try: 657 if self._action: 658 self._action() 659 # enter draining state 660 self._state = 1 661 self._cond.notify_all() 662 except: 663 #an exception during the _action handler. Break and reraise 664 self._break() 665 raise 666 667 # Wait in the barrier until we are released. Raise an exception 668 # if the barrier is reset or broken. 669 def _wait(self, timeout): 670 if not self._cond.wait_for(lambda : self._state != 0, timeout): 671 #timed out. Break the barrier 672 self._break() 673 raise BrokenBarrierError 674 if self._state < 0: 675 raise BrokenBarrierError 676 assert self._state == 1 677 678 # If we are the last thread to exit the barrier, signal any threads 679 # waiting for the barrier to drain. 680 def _exit(self): 681 if self._count == 0: 682 if self._state in (-1, 1): 683 #resetting or draining 684 self._state = 0 685 self._cond.notify_all() 686 687 def reset(self): 688 """Reset the barrier to the initial state. 689 690 Any threads currently waiting will get the BrokenBarrier exception 691 raised. 692 693 """ 694 with self._cond: 695 if self._count > 0: 696 if self._state == 0: 697 #reset the barrier, waking up threads 698 self._state = -1 699 elif self._state == -2: 700 #was broken, set it to reset state 701 #which clears when the last thread exits 702 self._state = -1 703 else: 704 self._state = 0 705 self._cond.notify_all() 706 707 def abort(self): 708 """Place the barrier into a 'broken' state. 709 710 Useful in case of error. Any currently waiting threads and threads 711 attempting to 'wait()' will have BrokenBarrierError raised. 712 713 """ 714 with self._cond: 715 self._break() 716 717 def _break(self): 718 # An internal error was detected. The barrier is set to 719 # a broken state all parties awakened. 720 self._state = -2 721 self._cond.notify_all() 722 723 @property 724 def parties(self): 725 """Return the number of threads required to trip the barrier.""" 726 return self._parties 727 728 @property 729 def n_waiting(self): 730 """Return the number of threads currently waiting at the barrier.""" 731 # We don't need synchronization here since this is an ephemeral result 732 # anyway. It returns the correct value in the steady state. 733 if self._state == 0: 734 return self._count 735 return 0 736 737 @property 738 def broken(self): 739 """Return True if the barrier is in a broken state.""" 740 return self._state == -2 741 742# exception raised by the Barrier class 743class BrokenBarrierError(RuntimeError): 744 pass 745 746 747# Helper to generate new thread names 748_counter = _count().__next__ 749_counter() # Consume 0 so first non-main thread has id 1. 750def _newname(template="Thread-%d"): 751 return template % _counter() 752 753# Active thread administration 754_active_limbo_lock = _allocate_lock() 755_active = {} # maps thread id to Thread object 756_limbo = {} 757_dangling = WeakSet() 758# Set of Thread._tstate_lock locks of non-daemon threads used by _shutdown() 759# to wait until all Python thread states get deleted: 760# see Thread._set_tstate_lock(). 761_shutdown_locks_lock = _allocate_lock() 762_shutdown_locks = set() 763 764# Main class for threads 765 766class Thread: 767 """A class that represents a thread of control. 768 769 This class can be safely subclassed in a limited fashion. There are two ways 770 to specify the activity: by passing a callable object to the constructor, or 771 by overriding the run() method in a subclass. 772 773 """ 774 775 _initialized = False 776 777 def __init__(self, group=None, target=None, name=None, 778 args=(), kwargs=None, *, daemon=None): 779 """This constructor should always be called with keyword arguments. Arguments are: 780 781 *group* should be None; reserved for future extension when a ThreadGroup 782 class is implemented. 783 784 *target* is the callable object to be invoked by the run() 785 method. Defaults to None, meaning nothing is called. 786 787 *name* is the thread name. By default, a unique name is constructed of 788 the form "Thread-N" where N is a small decimal number. 789 790 *args* is the argument tuple for the target invocation. Defaults to (). 791 792 *kwargs* is a dictionary of keyword arguments for the target 793 invocation. Defaults to {}. 794 795 If a subclass overrides the constructor, it must make sure to invoke 796 the base class constructor (Thread.__init__()) before doing anything 797 else to the thread. 798 799 """ 800 assert group is None, "group argument must be None for now" 801 if kwargs is None: 802 kwargs = {} 803 self._target = target 804 self._name = str(name or _newname()) 805 self._args = args 806 self._kwargs = kwargs 807 if daemon is not None: 808 self._daemonic = daemon 809 else: 810 self._daemonic = current_thread().daemon 811 self._ident = None 812 if _HAVE_THREAD_NATIVE_ID: 813 self._native_id = None 814 self._tstate_lock = None 815 self._started = Event() 816 self._is_stopped = False 817 self._initialized = True 818 # Copy of sys.stderr used by self._invoke_excepthook() 819 self._stderr = _sys.stderr 820 self._invoke_excepthook = _make_invoke_excepthook() 821 # For debugging and _after_fork() 822 _dangling.add(self) 823 824 def _reset_internal_locks(self, is_alive): 825 # private! Called by _after_fork() to reset our internal locks as 826 # they may be in an invalid state leading to a deadlock or crash. 827 self._started._at_fork_reinit() 828 if is_alive: 829 # bpo-42350: If the fork happens when the thread is already stopped 830 # (ex: after threading._shutdown() has been called), _tstate_lock 831 # is None. Do nothing in this case. 832 if self._tstate_lock is not None: 833 self._tstate_lock._at_fork_reinit() 834 self._tstate_lock.acquire() 835 else: 836 # The thread isn't alive after fork: it doesn't have a tstate 837 # anymore. 838 self._is_stopped = True 839 self._tstate_lock = None 840 841 def __repr__(self): 842 assert self._initialized, "Thread.__init__() was not called" 843 status = "initial" 844 if self._started.is_set(): 845 status = "started" 846 self.is_alive() # easy way to get ._is_stopped set when appropriate 847 if self._is_stopped: 848 status = "stopped" 849 if self._daemonic: 850 status += " daemon" 851 if self._ident is not None: 852 status += " %s" % self._ident 853 return "<%s(%s, %s)>" % (self.__class__.__name__, self._name, status) 854 855 def start(self): 856 """Start the thread's activity. 857 858 It must be called at most once per thread object. It arranges for the 859 object's run() method to be invoked in a separate thread of control. 860 861 This method will raise a RuntimeError if called more than once on the 862 same thread object. 863 864 """ 865 if not self._initialized: 866 raise RuntimeError("thread.__init__() not called") 867 868 if self._started.is_set(): 869 raise RuntimeError("threads can only be started once") 870 871 with _active_limbo_lock: 872 _limbo[self] = self 873 try: 874 _start_new_thread(self._bootstrap, ()) 875 except Exception: 876 with _active_limbo_lock: 877 del _limbo[self] 878 raise 879 self._started.wait() 880 881 def run(self): 882 """Method representing the thread's activity. 883 884 You may override this method in a subclass. The standard run() method 885 invokes the callable object passed to the object's constructor as the 886 target argument, if any, with sequential and keyword arguments taken 887 from the args and kwargs arguments, respectively. 888 889 """ 890 try: 891 if self._target: 892 self._target(*self._args, **self._kwargs) 893 finally: 894 # Avoid a refcycle if the thread is running a function with 895 # an argument that has a member that points to the thread. 896 del self._target, self._args, self._kwargs 897 898 def _bootstrap(self): 899 # Wrapper around the real bootstrap code that ignores 900 # exceptions during interpreter cleanup. Those typically 901 # happen when a daemon thread wakes up at an unfortunate 902 # moment, finds the world around it destroyed, and raises some 903 # random exception *** while trying to report the exception in 904 # _bootstrap_inner() below ***. Those random exceptions 905 # don't help anybody, and they confuse users, so we suppress 906 # them. We suppress them only when it appears that the world 907 # indeed has already been destroyed, so that exceptions in 908 # _bootstrap_inner() during normal business hours are properly 909 # reported. Also, we only suppress them for daemonic threads; 910 # if a non-daemonic encounters this, something else is wrong. 911 try: 912 self._bootstrap_inner() 913 except: 914 if self._daemonic and _sys is None: 915 return 916 raise 917 918 def _set_ident(self): 919 self._ident = get_ident() 920 921 if _HAVE_THREAD_NATIVE_ID: 922 def _set_native_id(self): 923 self._native_id = get_native_id() 924 925 def _set_tstate_lock(self): 926 """ 927 Set a lock object which will be released by the interpreter when 928 the underlying thread state (see pystate.h) gets deleted. 929 """ 930 self._tstate_lock = _set_sentinel() 931 self._tstate_lock.acquire() 932 933 if not self.daemon: 934 with _shutdown_locks_lock: 935 _shutdown_locks.add(self._tstate_lock) 936 937 def _bootstrap_inner(self): 938 try: 939 self._set_ident() 940 self._set_tstate_lock() 941 if _HAVE_THREAD_NATIVE_ID: 942 self._set_native_id() 943 self._started.set() 944 with _active_limbo_lock: 945 _active[self._ident] = self 946 del _limbo[self] 947 948 if _trace_hook: 949 _sys.settrace(_trace_hook) 950 if _profile_hook: 951 _sys.setprofile(_profile_hook) 952 953 try: 954 self.run() 955 except: 956 self._invoke_excepthook(self) 957 finally: 958 with _active_limbo_lock: 959 try: 960 # We don't call self._delete() because it also 961 # grabs _active_limbo_lock. 962 del _active[get_ident()] 963 except: 964 pass 965 966 def _stop(self): 967 # After calling ._stop(), .is_alive() returns False and .join() returns 968 # immediately. ._tstate_lock must be released before calling ._stop(). 969 # 970 # Normal case: C code at the end of the thread's life 971 # (release_sentinel in _threadmodule.c) releases ._tstate_lock, and 972 # that's detected by our ._wait_for_tstate_lock(), called by .join() 973 # and .is_alive(). Any number of threads _may_ call ._stop() 974 # simultaneously (for example, if multiple threads are blocked in 975 # .join() calls), and they're not serialized. That's harmless - 976 # they'll just make redundant rebindings of ._is_stopped and 977 # ._tstate_lock. Obscure: we rebind ._tstate_lock last so that the 978 # "assert self._is_stopped" in ._wait_for_tstate_lock() always works 979 # (the assert is executed only if ._tstate_lock is None). 980 # 981 # Special case: _main_thread releases ._tstate_lock via this 982 # module's _shutdown() function. 983 lock = self._tstate_lock 984 if lock is not None: 985 assert not lock.locked() 986 self._is_stopped = True 987 self._tstate_lock = None 988 if not self.daemon: 989 with _shutdown_locks_lock: 990 _shutdown_locks.discard(lock) 991 992 def _delete(self): 993 "Remove current thread from the dict of currently running threads." 994 with _active_limbo_lock: 995 del _active[get_ident()] 996 # There must not be any python code between the previous line 997 # and after the lock is released. Otherwise a tracing function 998 # could try to acquire the lock again in the same thread, (in 999 # current_thread()), and would block. 1000 1001 def join(self, timeout=None): 1002 """Wait until the thread terminates. 1003 1004 This blocks the calling thread until the thread whose join() method is 1005 called terminates -- either normally or through an unhandled exception 1006 or until the optional timeout occurs. 1007 1008 When the timeout argument is present and not None, it should be a 1009 floating point number specifying a timeout for the operation in seconds 1010 (or fractions thereof). As join() always returns None, you must call 1011 is_alive() after join() to decide whether a timeout happened -- if the 1012 thread is still alive, the join() call timed out. 1013 1014 When the timeout argument is not present or None, the operation will 1015 block until the thread terminates. 1016 1017 A thread can be join()ed many times. 1018 1019 join() raises a RuntimeError if an attempt is made to join the current 1020 thread as that would cause a deadlock. It is also an error to join() a 1021 thread before it has been started and attempts to do so raises the same 1022 exception. 1023 1024 """ 1025 if not self._initialized: 1026 raise RuntimeError("Thread.__init__() not called") 1027 if not self._started.is_set(): 1028 raise RuntimeError("cannot join thread before it is started") 1029 if self is current_thread(): 1030 raise RuntimeError("cannot join current thread") 1031 1032 if timeout is None: 1033 self._wait_for_tstate_lock() 1034 else: 1035 # the behavior of a negative timeout isn't documented, but 1036 # historically .join(timeout=x) for x<0 has acted as if timeout=0 1037 self._wait_for_tstate_lock(timeout=max(timeout, 0)) 1038 1039 def _wait_for_tstate_lock(self, block=True, timeout=-1): 1040 # Issue #18808: wait for the thread state to be gone. 1041 # At the end of the thread's life, after all knowledge of the thread 1042 # is removed from C data structures, C code releases our _tstate_lock. 1043 # This method passes its arguments to _tstate_lock.acquire(). 1044 # If the lock is acquired, the C code is done, and self._stop() is 1045 # called. That sets ._is_stopped to True, and ._tstate_lock to None. 1046 lock = self._tstate_lock 1047 if lock is None: # already determined that the C code is done 1048 assert self._is_stopped 1049 elif lock.acquire(block, timeout): 1050 lock.release() 1051 self._stop() 1052 1053 @property 1054 def name(self): 1055 """A string used for identification purposes only. 1056 1057 It has no semantics. Multiple threads may be given the same name. The 1058 initial name is set by the constructor. 1059 1060 """ 1061 assert self._initialized, "Thread.__init__() not called" 1062 return self._name 1063 1064 @name.setter 1065 def name(self, name): 1066 assert self._initialized, "Thread.__init__() not called" 1067 self._name = str(name) 1068 1069 @property 1070 def ident(self): 1071 """Thread identifier of this thread or None if it has not been started. 1072 1073 This is a nonzero integer. See the get_ident() function. Thread 1074 identifiers may be recycled when a thread exits and another thread is 1075 created. The identifier is available even after the thread has exited. 1076 1077 """ 1078 assert self._initialized, "Thread.__init__() not called" 1079 return self._ident 1080 1081 if _HAVE_THREAD_NATIVE_ID: 1082 @property 1083 def native_id(self): 1084 """Native integral thread ID of this thread, or None if it has not been started. 1085 1086 This is a non-negative integer. See the get_native_id() function. 1087 This represents the Thread ID as reported by the kernel. 1088 1089 """ 1090 assert self._initialized, "Thread.__init__() not called" 1091 return self._native_id 1092 1093 def is_alive(self): 1094 """Return whether the thread is alive. 1095 1096 This method returns True just before the run() method starts until just 1097 after the run() method terminates. The module function enumerate() 1098 returns a list of all alive threads. 1099 1100 """ 1101 assert self._initialized, "Thread.__init__() not called" 1102 if self._is_stopped or not self._started.is_set(): 1103 return False 1104 self._wait_for_tstate_lock(False) 1105 return not self._is_stopped 1106 1107 @property 1108 def daemon(self): 1109 """A boolean value indicating whether this thread is a daemon thread. 1110 1111 This must be set before start() is called, otherwise RuntimeError is 1112 raised. Its initial value is inherited from the creating thread; the 1113 main thread is not a daemon thread and therefore all threads created in 1114 the main thread default to daemon = False. 1115 1116 The entire Python program exits when only daemon threads are left. 1117 1118 """ 1119 assert self._initialized, "Thread.__init__() not called" 1120 return self._daemonic 1121 1122 @daemon.setter 1123 def daemon(self, daemonic): 1124 if not self._initialized: 1125 raise RuntimeError("Thread.__init__() not called") 1126 if self._started.is_set(): 1127 raise RuntimeError("cannot set daemon status of active thread") 1128 self._daemonic = daemonic 1129 1130 def isDaemon(self): 1131 return self.daemon 1132 1133 def setDaemon(self, daemonic): 1134 self.daemon = daemonic 1135 1136 def getName(self): 1137 return self.name 1138 1139 def setName(self, name): 1140 self.name = name 1141 1142 1143try: 1144 from _thread import (_excepthook as excepthook, 1145 _ExceptHookArgs as ExceptHookArgs) 1146except ImportError: 1147 # Simple Python implementation if _thread._excepthook() is not available 1148 from traceback import print_exception as _print_exception 1149 from collections import namedtuple 1150 1151 _ExceptHookArgs = namedtuple( 1152 'ExceptHookArgs', 1153 'exc_type exc_value exc_traceback thread') 1154 1155 def ExceptHookArgs(args): 1156 return _ExceptHookArgs(*args) 1157 1158 def excepthook(args, /): 1159 """ 1160 Handle uncaught Thread.run() exception. 1161 """ 1162 if args.exc_type == SystemExit: 1163 # silently ignore SystemExit 1164 return 1165 1166 if _sys is not None and _sys.stderr is not None: 1167 stderr = _sys.stderr 1168 elif args.thread is not None: 1169 stderr = args.thread._stderr 1170 if stderr is None: 1171 # do nothing if sys.stderr is None and sys.stderr was None 1172 # when the thread was created 1173 return 1174 else: 1175 # do nothing if sys.stderr is None and args.thread is None 1176 return 1177 1178 if args.thread is not None: 1179 name = args.thread.name 1180 else: 1181 name = get_ident() 1182 print(f"Exception in thread {name}:", 1183 file=stderr, flush=True) 1184 _print_exception(args.exc_type, args.exc_value, args.exc_traceback, 1185 file=stderr) 1186 stderr.flush() 1187 1188 1189def _make_invoke_excepthook(): 1190 # Create a local namespace to ensure that variables remain alive 1191 # when _invoke_excepthook() is called, even if it is called late during 1192 # Python shutdown. It is mostly needed for daemon threads. 1193 1194 old_excepthook = excepthook 1195 old_sys_excepthook = _sys.excepthook 1196 if old_excepthook is None: 1197 raise RuntimeError("threading.excepthook is None") 1198 if old_sys_excepthook is None: 1199 raise RuntimeError("sys.excepthook is None") 1200 1201 sys_exc_info = _sys.exc_info 1202 local_print = print 1203 local_sys = _sys 1204 1205 def invoke_excepthook(thread): 1206 global excepthook 1207 try: 1208 hook = excepthook 1209 if hook is None: 1210 hook = old_excepthook 1211 1212 args = ExceptHookArgs([*sys_exc_info(), thread]) 1213 1214 hook(args) 1215 except Exception as exc: 1216 exc.__suppress_context__ = True 1217 del exc 1218 1219 if local_sys is not None and local_sys.stderr is not None: 1220 stderr = local_sys.stderr 1221 else: 1222 stderr = thread._stderr 1223 1224 local_print("Exception in threading.excepthook:", 1225 file=stderr, flush=True) 1226 1227 if local_sys is not None and local_sys.excepthook is not None: 1228 sys_excepthook = local_sys.excepthook 1229 else: 1230 sys_excepthook = old_sys_excepthook 1231 1232 sys_excepthook(*sys_exc_info()) 1233 finally: 1234 # Break reference cycle (exception stored in a variable) 1235 args = None 1236 1237 return invoke_excepthook 1238 1239 1240# The timer class was contributed by Itamar Shtull-Trauring 1241 1242class Timer(Thread): 1243 """Call a function after a specified number of seconds: 1244 1245 t = Timer(30.0, f, args=None, kwargs=None) 1246 t.start() 1247 t.cancel() # stop the timer's action if it's still waiting 1248 1249 """ 1250 1251 def __init__(self, interval, function, args=None, kwargs=None): 1252 Thread.__init__(self) 1253 self.interval = interval 1254 self.function = function 1255 self.args = args if args is not None else [] 1256 self.kwargs = kwargs if kwargs is not None else {} 1257 self.finished = Event() 1258 1259 def cancel(self): 1260 """Stop the timer if it hasn't finished yet.""" 1261 self.finished.set() 1262 1263 def run(self): 1264 self.finished.wait(self.interval) 1265 if not self.finished.is_set(): 1266 self.function(*self.args, **self.kwargs) 1267 self.finished.set() 1268 1269 1270# Special thread class to represent the main thread 1271 1272class _MainThread(Thread): 1273 1274 def __init__(self): 1275 Thread.__init__(self, name="MainThread", daemon=False) 1276 self._set_tstate_lock() 1277 self._started.set() 1278 self._set_ident() 1279 if _HAVE_THREAD_NATIVE_ID: 1280 self._set_native_id() 1281 with _active_limbo_lock: 1282 _active[self._ident] = self 1283 1284 1285# Dummy thread class to represent threads not started here. 1286# These aren't garbage collected when they die, nor can they be waited for. 1287# If they invoke anything in threading.py that calls current_thread(), they 1288# leave an entry in the _active dict forever after. 1289# Their purpose is to return *something* from current_thread(). 1290# They are marked as daemon threads so we won't wait for them 1291# when we exit (conform previous semantics). 1292 1293class _DummyThread(Thread): 1294 1295 def __init__(self): 1296 Thread.__init__(self, name=_newname("Dummy-%d"), daemon=True) 1297 1298 self._started.set() 1299 self._set_ident() 1300 if _HAVE_THREAD_NATIVE_ID: 1301 self._set_native_id() 1302 with _active_limbo_lock: 1303 _active[self._ident] = self 1304 1305 def _stop(self): 1306 pass 1307 1308 def is_alive(self): 1309 assert not self._is_stopped and self._started.is_set() 1310 return True 1311 1312 def join(self, timeout=None): 1313 assert False, "cannot join a dummy thread" 1314 1315 1316# Global API functions 1317 1318def current_thread(): 1319 """Return the current Thread object, corresponding to the caller's thread of control. 1320 1321 If the caller's thread of control was not created through the threading 1322 module, a dummy thread object with limited functionality is returned. 1323 1324 """ 1325 try: 1326 return _active[get_ident()] 1327 except KeyError: 1328 return _DummyThread() 1329 1330currentThread = current_thread 1331 1332def active_count(): 1333 """Return the number of Thread objects currently alive. 1334 1335 The returned count is equal to the length of the list returned by 1336 enumerate(). 1337 1338 """ 1339 with _active_limbo_lock: 1340 return len(_active) + len(_limbo) 1341 1342activeCount = active_count 1343 1344def _enumerate(): 1345 # Same as enumerate(), but without the lock. Internal use only. 1346 return list(_active.values()) + list(_limbo.values()) 1347 1348def enumerate(): 1349 """Return a list of all Thread objects currently alive. 1350 1351 The list includes daemonic threads, dummy thread objects created by 1352 current_thread(), and the main thread. It excludes terminated threads and 1353 threads that have not yet been started. 1354 1355 """ 1356 with _active_limbo_lock: 1357 return list(_active.values()) + list(_limbo.values()) 1358 1359 1360_threading_atexits = [] 1361_SHUTTING_DOWN = False 1362 1363def _register_atexit(func, *arg, **kwargs): 1364 """CPython internal: register *func* to be called before joining threads. 1365 1366 The registered *func* is called with its arguments just before all 1367 non-daemon threads are joined in `_shutdown()`. It provides a similar 1368 purpose to `atexit.register()`, but its functions are called prior to 1369 threading shutdown instead of interpreter shutdown. 1370 1371 For similarity to atexit, the registered functions are called in reverse. 1372 """ 1373 if _SHUTTING_DOWN: 1374 raise RuntimeError("can't register atexit after shutdown") 1375 1376 call = functools.partial(func, *arg, **kwargs) 1377 _threading_atexits.append(call) 1378 1379 1380from _thread import stack_size 1381 1382# Create the main thread object, 1383# and make it available for the interpreter 1384# (Py_Main) as threading._shutdown. 1385 1386_main_thread = _MainThread() 1387 1388def _shutdown(): 1389 """ 1390 Wait until the Python thread state of all non-daemon threads get deleted. 1391 """ 1392 # Obscure: other threads may be waiting to join _main_thread. That's 1393 # dubious, but some code does it. We can't wait for C code to release 1394 # the main thread's tstate_lock - that won't happen until the interpreter 1395 # is nearly dead. So we release it here. Note that just calling _stop() 1396 # isn't enough: other threads may already be waiting on _tstate_lock. 1397 if _main_thread._is_stopped: 1398 # _shutdown() was already called 1399 return 1400 1401 global _SHUTTING_DOWN 1402 _SHUTTING_DOWN = True 1403 # Main thread 1404 tlock = _main_thread._tstate_lock 1405 # The main thread isn't finished yet, so its thread state lock can't have 1406 # been released. 1407 assert tlock is not None 1408 assert tlock.locked() 1409 tlock.release() 1410 _main_thread._stop() 1411 1412 # Call registered threading atexit functions before threads are joined. 1413 # Order is reversed, similar to atexit. 1414 for atexit_call in reversed(_threading_atexits): 1415 atexit_call() 1416 1417 # Join all non-deamon threads 1418 while True: 1419 with _shutdown_locks_lock: 1420 locks = list(_shutdown_locks) 1421 _shutdown_locks.clear() 1422 1423 if not locks: 1424 break 1425 1426 for lock in locks: 1427 # mimick Thread.join() 1428 lock.acquire() 1429 lock.release() 1430 1431 # new threads can be spawned while we were waiting for the other 1432 # threads to complete 1433 1434 1435def main_thread(): 1436 """Return the main thread object. 1437 1438 In normal conditions, the main thread is the thread from which the 1439 Python interpreter was started. 1440 """ 1441 return _main_thread 1442 1443# get thread-local implementation, either from the thread 1444# module, or from the python fallback 1445 1446try: 1447 from _thread import _local as local 1448except ImportError: 1449 from _threading_local import local 1450 1451 1452def _after_fork(): 1453 """ 1454 Cleanup threading module state that should not exist after a fork. 1455 """ 1456 # Reset _active_limbo_lock, in case we forked while the lock was held 1457 # by another (non-forked) thread. http://bugs.python.org/issue874900 1458 global _active_limbo_lock, _main_thread 1459 global _shutdown_locks_lock, _shutdown_locks 1460 _active_limbo_lock = _allocate_lock() 1461 1462 # fork() only copied the current thread; clear references to others. 1463 new_active = {} 1464 1465 try: 1466 current = _active[get_ident()] 1467 except KeyError: 1468 # fork() was called in a thread which was not spawned 1469 # by threading.Thread. For example, a thread spawned 1470 # by thread.start_new_thread(). 1471 current = _MainThread() 1472 1473 _main_thread = current 1474 1475 # reset _shutdown() locks: threads re-register their _tstate_lock below 1476 _shutdown_locks_lock = _allocate_lock() 1477 _shutdown_locks = set() 1478 1479 with _active_limbo_lock: 1480 # Dangling thread instances must still have their locks reset, 1481 # because someone may join() them. 1482 threads = set(_enumerate()) 1483 threads.update(_dangling) 1484 for thread in threads: 1485 # Any lock/condition variable may be currently locked or in an 1486 # invalid state, so we reinitialize them. 1487 if thread is current: 1488 # There is only one active thread. We reset the ident to 1489 # its new value since it can have changed. 1490 thread._reset_internal_locks(True) 1491 ident = get_ident() 1492 thread._ident = ident 1493 new_active[ident] = thread 1494 else: 1495 # All the others are already stopped. 1496 thread._reset_internal_locks(False) 1497 thread._stop() 1498 1499 _limbo.clear() 1500 _active.clear() 1501 _active.update(new_active) 1502 assert len(_active) == 1 1503 1504 1505if hasattr(_os, "register_at_fork"): 1506 _os.register_at_fork(after_in_child=_after_fork) 1507