| Python 线程 |
Threading是用于提供线程相关的操作，线程是应用程序中工作的最小单元。线程与进程的关系下图所示：
子线程是由主线程产生的，但两者并没有关联。
利用threading创建线程：
1 '''利用threading包创建'''
2 import threading
3 import time
4
5 def run(n):
6 time.sleep(2)
7 print("task:",n)
8
9 '''串行：一个运行完后，再运行另外一个'''
10 run("t1") #并不是线程，只是调用方法传参数
11 run("t2")
12
13 '''并发性'''
14 t1 = threading.Thread(target=run,args=("T1",)) #t1是线程，args为元组
15 t2 = threading.Thread(target=run,args=("T2",))
16 t1.start() #并发性地工作
17 t2.start()
18
19
20 '''运行结果'''
21 task: t1 #t1运行后会间隔两秒，然后运行t2
22 task: t2
23
24 task: T2 #T1,T2同时运行
25 task: T1
上述创建了两个线程t1和t2,然后控制器就交给了CPU，CPU根据指定算法进行调度，分片执行指令。
更多方法：
- start 线程准备就绪，等待CPU调度；启动线程的活动，每个线程对象最多只能调用一次。
- join 逐个执行每个线程，执行完毕后继续往下执行，该方法使得多线程变得无意义。
- run 表示线程活动的方法。可以在子类中重写此方法。标准run()方法调用传递给对象构造函数的可调用对象作为目标参数(如果有的话)，分别使用args和kwargs参数中的顺序参数和关键字参数。线程被cpu调度后自动执行线程对象的run方法
- get_ident() 获得线程地址
- setName 为线程设置名称
- getName 获取线程名称
- daemon 一个布尔值，指示此线程是否为守护线程。这必须在调用start()之前设置，否则会引发运行时错误。它的初始值继承自创建线程;主线程不是守护进程线程，因此在主线程中创建的所有线程默认为守护进程= False。当没有存活的非守护进程线程时，整个Python程序退出。
- setDaemon 设置为后台线程或前台线程（默认）如果是后台线程，主线程执行过程中，后台线程也在进行，主线程执行完毕后，后台线程不论成功与否，均停止；如果是前台线程，主线程执行过程中，前台线程也在进行，主线程执行完毕后，等待前台线程也执行完成后，程序停止
1 #子线程是由主线程产生的，但两者并没有关联
2 import threading
3 import time
4
5 def run(n):
6 print("task:",n)
7 time.sleep(0.1)
8 print("taskdone:",n)
9
10 Start_time = time.time()
11 for i in range(50): #共有51个线程，代码本身是一个主线程
12 t = threading.Thread(target=run,args=("t--%s" % i,))
13 t.start()
14 t.join() #join使得主线程与子线程成串行运行
15
16 print(time.time()-Start_time) #print为创建子线程所产生的时间，而非运行时间
1 import threading
2 import time
3
4 class My_Thread(threading.Thread):
5 def __init__(self,n):
6 super(My_Thread,self).__init__()
7 self.n = n
8
9 def run(self):
10 print("task:",self.n)
11 time.sleep(0.1)
12 t_obj=[]
13 start_time = time.time()
14 for i in range(50): #共有51个线程，代码本身是一个主线程
15 t = My_Thread("t--%s" % i)
16 t.setDaemon(True) #监听端口，当主程序执行完毕，将不会执行其他线程（前提是去掉join方法）
17 t.start()
18 t_obj.append(t)
19 print(time.time()-start_time)
20
21
22 '''运行结果'''
23 task: t--0
24 task: t--1
25 task: t--2
26 task: t--3
27 task: t--4
28 task: t--5
29 task: t--6
30 task: t--7
31 task: t--8
32 task: t--9
33 task: t--10
34 task: t--11
35 task: t--12
36 task: t--13
37 task: t--14
38 task: t--15
39 task: t--16
40 task: t--17
41 task: t--18
42 task: t--19
43 task: t--20
44 task: t--21
45 task: t--22
46 task: t--23
47 task: t--24
48 task: t--25
49 task: t--26
50 task: t--27
51 task: t--28
52 task: t--29
53 task: t--30
54 task: t--31
55 task: t--32
56 task: t--33
57 task: t--34
58 task: t--35
59 task: t--36
60 task: t--37
61 task: t--38
62 task: t--39
63 task: t--40
64 task: t--41
65 task: t--42
66 task: t--43
67 task: t--44
68 task: t--45
69 task: t--46
70 task: t--47
71 task: t--48
72 task: t--49
73 0.01196908950805664
线程锁（Lock）：
1 def acquire(self, blocking=True, timeout=None): 2 """Acquire a semaphore, decrementing the internal counter by one. 3 When invoked without arguments: if the internal counter is larger than 4 zero on entry, decrement it by one and return immediately. If it is zero 5 on entry, block, waiting until some other thread has called release() to 6 make it larger than zero. This is done with proper interlocking so that 7 if multiple acquire() calls are blocked, release() will wake exactly one 8 of them up. The implementation may pick one at random, so the order in 9 which blocked threads are awakened should not be relied on. There is no 10 return value in this case. 11 When invoked with blocking set to true, do the same thing as when called 12 without arguments, and return true. 13 When invoked with blocking set to false, do not block. If a call without 14 an argument would block, return false immediately; otherwise, do the 15 same thing as when called without arguments, and return true. 16 When invoked with a timeout other than None, it will block for at 17 most timeout seconds. If acquire does not complete successfully in 18 that interval, return false. Return true otherwise. 19 """ 20 #获得一个信号量，将内部计数器减1。在没有参数的情况下调用时:如果内部计数器在入口时 21 # 大于0，则将其递减1并立即返回。如果进入时为零，阻塞，等待其他线程调用release() 22 # 使其大于零。这是通过适当的联锁完成的，这样，如果多个acquire()调用被阻塞， 23 # release()就会唤醒其中一个调用。实现可以随机选择一个线程，因此不应该依赖于被阻塞 24 # 线程被唤醒的顺序。在本例中没有返回值。当阻塞集调用为true时，执行与没有参数调用 25 # 时相同的操作，并返回true。当阻塞设置为false时，不要阻塞。如果一个没有参数的 26 # 调用将阻塞，立即返回false;否则，执行与没有参数调用时相同的操作，并返回true。 27 # 当使用除None以外的超时调用时，它最多将阻塞超时秒。如果在那段时间里收购没有成功 28 # 完成，还假。否则返回true。 29 if not blocking and timeout is not None: 30 raise ValueError("can't specify timeout for non-blocking acquire") 31 rc = False 32 endtime = None 33 with self._cond: 34 while self._value == 0: 35 if not blocking: 36 break 37 if timeout is not None: 38 if endtime is None: 39 endtime = _time() + timeout 40 else: 41 timeout = endtime - _time() 42 if timeout <= 0: 43 break 44 self._cond.wait(timeout) 45 else: 46 self._value -= 1 47 rc = True 48 return rc 49 50 __enter__ = acquire 51 52 def release(self): 53 """Release a semaphore, incrementing the internal counter by one. 54 When the counter is zero on entry and another thread is waiting for it 55 to become larger than zero again, wake up that thread. 56 """ 57 #释放信号量，增加一个内部计数器。当进入时计数器为零，而另一个线程正在等待计数器 58 # 再次大于零时，唤醒该线程。 59 with self._cond: 60 self._value += 1 61 self._cond.notify() 62 63 def __exit__(self, t, v, tb): 64 self.release()
1 import threading
2 import time
3
4 lock = threading.Lock() #线程锁
5
6 def run(n):
7 lock.acquire() #锁定
8 global num
9 num+=1
10 lock.release() #释放锁
11 time.sleep(1)
12
13 t_obj = []
14 num = 0
15 for i in range(50):
16 t = threading.Thread(target=run,args=("t--%s" % i,))
17 t.start()
18 t_obj.append(t)
19
20 for i in t_obj:
21 i.join()
22
23 print("num:",num)
24
25
26 '''运行结果'''
27 num: 50
'''可用来做测试'''
if __name__ == "__main__"
#表示函数的开始位置，判断自主运行与否
线程池（信号量（semaphore））：
信号量管理一个计数器，该计数器表示release()调用的数量减去acquire()调用的数量，再加上一个初始值。acquire()方法如果有必要会阻塞，直到它可以返回而不会使计数器变为负数为止。如果未指定，值默认为1。
'''信号量'''
import threading
import time
def run(n):
Semaphore.acquire()
print("task:",n)
time.sleep(1)
Semaphore.release()
if __name__ == "__main__":
Semaphore = threading.BoundedSemaphore(5)
#每五个子进程运行一次，间隔一秒后，再运行下五个
for i in range(20):
t = threading.Thread(target=run,args=(i,))
t.start()
while threading.active_count()!=1:
pass
else:
print("--all threading has done")
1 """Thread module emulating a subset of Java's threading model."""
2 #线程模块模拟Java线程模型的一个子集。
3 import os as _os
4 import sys as _sys
5 import _thread
6
7 from time import monotonic as _time
8 from traceback import format_exc as _format_exc
9 from _weakrefset import WeakSet
10 from itertools import islice as _islice, count as _count
11 try:
12 from _collections import deque as _deque
13 except 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
32 # Rename some stuff so "from threading import *" is safe
33 _start_new_thread = _thread.start_new_thread
34 _allocate_lock = _thread.allocate_lock
35 _set_sentinel = _thread._set_sentinel
36 get_ident = _thread.get_ident
37 ThreadError = _thread.error
38 try:
39 _CRLock = _thread.RLock
40 except AttributeError:
41 _CRLock = None
42 TIMEOUT_MAX = _thread.TIMEOUT_MAX
43 del _thread
44
45
46 # Support for profile and trace hooks
47 #支持配置文件和跟踪挂钩
48
49 _profile_hook = None
50 _trace_hook = None
51
52 def setprofile(func):
53 """Set a profile function for all threads started from the threading module.
54 The func will be passed to sys.setprofile() for each thread, before its
55 run() method is called.
56 """
57 #为从线程模块启动的所有线程设置一个配置文件函数。在调用其run()方法之前，
58 # func将被传递给每个线程的sys.setprofile()。
59
60 global _profile_hook
61 _profile_hook = func
62
63 def settrace(func):
64 """Set a trace function for all threads started from the threading module.
65 The func will be passed to sys.settrace() for each thread, before its run()
66 method is called.
67 """
68 #为从线程模块启动的所有线程设置跟踪函数。在调用其run()方法之前，
69 # func将被传递给每个线程的sys.settrace()。
70
71 global _trace_hook
72 _trace_hook = func
73
74 # Synchronization classes
75 # 同步类
76
77 Lock = _allocate_lock
78
79 def RLock(*args, **kwargs):
80 """Factory function that returns a new reentrant lock.
81 A reentrant lock must be released by the thread that acquired it. Once a
82 thread has acquired a reentrant lock, the same thread may acquire it again
83 without blocking; the thread must release it once for each time it has
84 acquired it.
85 """
86 #返回一个新的可重入锁的工厂函数。可重入锁必须由获得它的线程释放。
87 # 一旦一个线程获得了可重入锁，该线程可以在不阻塞的情况下再次获得该锁;
88 # 线程每次获得它时都必须释放它一次。
89
90 if _CRLock is None:
91 return _PyRLock(*args, **kwargs)
92 return _CRLock(*args, **kwargs)
93
94 class _RLock:
95 """This class implements reentrant lock objects.
96 A reentrant lock must be released by the thread that acquired it. Once a
97 thread has acquired a reentrant lock, the same thread may acquire it
98 again without blocking; the thread must release it once for each time it
99 has acquired it.
100 """
101 #该类实现可重入锁对象。可重入锁必须由获得它的线程释放。一旦一个线程获得了可重入锁，
102 # 该线程可以在不阻塞的情况下再次获得该锁;线程每次获得它时都必须释放它一次。
103
104 def __init__(self):
105 self._block = _allocate_lock()
106 self._owner = None
107 self._count = 0
108
109 def __repr__(self):
110 owner = self._owner
111 try:
112 owner = _active[owner].name
113 except KeyError:
114 pass
115 return "<%s %s.%s object owner=%r count=%d at %s>" % (
116 "locked" if self._block.locked() else "unlocked",
117 self.__class__.__module__,
118 self.__class__.__qualname__,
119 owner,
120 self._count,
121 hex(id(self))
122 )
123
124 def acquire(self, blocking=True, timeout=-1):
125 """Acquire a lock, blocking or non-blocking.
126 When invoked without arguments: if this thread already owns the lock,
127 increment the recursion level by one, and return immediately. Otherwise,
128 if another thread owns the lock, block until the lock is unlocked. Once
129 the lock is unlocked (not owned by any thread), then grab ownership, set
130 the recursion level to one, and return. If more than one thread is
131 blocked waiting until the lock is unlocked, only one at a time will be
132 able to grab ownership of the lock. There is no return value in this
133 case.
134 When invoked with the blocking argument set to true, do the same thing
135 as when called without arguments, and return true.
136 When invoked with the blocking argument set to false, do not block. If a
137 call without an argument would block, return false immediately;
138 otherwise, do the same thing as when called without arguments, and
139 return true.
140 When invoked with the floating-point timeout argument set to a positive
141 value, block for at most the number of seconds specified by timeout
142 and as long as the lock cannot be acquired. Return true if the lock has
143 been acquired, false if the timeout has elapsed.
144 """
145 #获得一个锁，阻塞或非阻塞。在没有参数的情况下调用时:如果这个线程已经拥有锁，
146 # 那么将递归级别增加1，并立即返回。否则，如果另一个线程拥有锁，
147 # 则阻塞直到锁被解锁。一旦锁被解锁(不属于任何线程)，然后获取所有权，
148 # 将递归级别设置为1，然后返回。如果有多个线程被阻塞，等待锁被解锁，
149 # 每次只有一个线程能够获取锁的所有权。在本例中没有返回值。当阻塞参数设置
150 # 为true时，执行与没有参数时相同的操作，并返回true。当阻塞参数设置为false时，
151 # 不要阻塞。如果一个没有参数的调用将阻塞，立即返回false;否则，执行与没有
152 # 参数调用时相同的操作，并返回true。当将浮点超时参数设置为正值时，如果获得
153 # 了锁，则最多阻塞超时指定的秒数，如果超时已过，则返回true;如果超时已过，则返回false。
154
155 me = get_ident()
156 if self._owner == me:
157 self._count += 1
158 return 1
159 rc = self._block.acquire(blocking, timeout)
160 if rc:
161 self._owner = me
162 self._count = 1
163 return rc
164
165 __enter__ = acquire
166
167 def release(self):
168 """Release a lock, decrementing the recursion level.
169 If after the decrement it is zero, reset the lock to unlocked (not owned
170 by any thread), and if any other threads are blocked waiting for the
171 lock to become unlocked, allow exactly one of them to proceed. If after
172 the decrement the recursion level is still nonzero, the lock remains
173 locked and owned by the calling thread.
174 Only call this method when the calling thread owns the lock. A
175 RuntimeError is raised if this method is called when the lock is
176 unlocked.
177 There is no return value.
178 """
179 #释放锁，降低递归级别。如果减量后为零，则将锁重置为解锁(不属于任何线程)，
180 # 如果任何其他线程被阻塞，等待锁解锁，则只允许其中一个线程继续执行。如果在递减
181 # 之后递归级别仍然是非零，则锁仍然被锁定，并且由调用线程拥有。只有当调用线程拥有
182 # 锁时才调用此方法。如果在解锁锁时调用此方法，将引发运行时错误。没有返回值。
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 #条件变量使用的内部方法
196
197 def _acquire_restore(self, state):
198 self._block.acquire()
199 self._count, self._owner = state
200
201 def _release_save(self):
202 if self._count == 0:
203 raise RuntimeError("cannot release un-acquired lock")
204 count = self._count
205 self._count = 0
206 owner = self._owner
207 self._owner = None
208 self._block.release()
209 return (count, owner)
210
211 def _is_owned(self):
212 return self._owner == get_ident()
213
214 _PyRLock = _RLock
215
216
217 class Condition:
218 """Class that implements a condition variable.
219 A condition variable allows one or more threads to wait until they are
220 notified by another thread.
221 If the lock argument is given and not None, it must be a Lock or RLock
222 object, and it is used as the underlying lock. Otherwise, a new RLock object
223 is created and used as the underlying lock.
224 """
225 #实现条件变量的类。条件变量允许一个或多个线程等待，直到另一个线程通知它们。
226 # 如果锁参数是给定的而不是空的，那么它必须是一个锁或RLock对象，并且它被用作底层锁。
227 # 否则，将创建一个新的RLock对象并将其用作底层锁。
228
229 def __init__(self, lock=None):
230 if lock is None:
231 lock = RLock()
232 self._lock = lock
233 # Export the lock's acquire() and release() methods
234 #导出锁的acquire()和release()方法
235 self.acquire = lock.acquire
236 self.release = lock.release
237 # If the lock defines _release_save() and/or _acquire_restore(),
238 # these override the default implementations (which just call
239 # release() and acquire() on the lock). Ditto for _is_owned().
240 #如果锁定义了_release_save()和/或_acquire_restore()，就会覆盖默认的实现
241 # (它只调用release()和acquire()对锁进行访问)。_is_owned同上()。
242 try:
243 self._release_save = lock._release_save
244 except AttributeError:
245 pass
246 try:
247 self._acquire_restore = lock._acquire_restore
248 except AttributeError:
249 pass
250 try:
251 self._is_owned = lock._is_owned
252 except AttributeError:
253 pass
254 self._waiters = _deque()
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 没有状态保存
267
268 def _acquire_restore(self, x):
269 self._lock.acquire() # Ignore saved state 忽略保存的状态
270
271 def _is_owned(self):
272 # Return True if lock is owned by current_thread.
273 #如果锁属于current_thread，则返回True。
274 # This method is called only if _lock doesn't have _is_owned().
275 #只有当_lock没有_is_owned()时才调用该方法。
276 if self._lock.acquire(0):
277 self._lock.release()
278 return False
279 else:
280 return True
281
282 def wait(self, timeout=None):
283 """Wait until notified or until a timeout occurs.
284 If the calling thread has not acquired the lock when this method is
285 called, a RuntimeError is raised.
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 When the timeout argument is present and not None, it should be a
291 floating point number specifying a timeout for the operation in seconds
292 (or fractions thereof).
293 When the underlying lock is an RLock, it is not released using its
294 release() method, since this may not actually unlock the lock when it
295 was acquired multiple times recursively. Instead, an internal interface
296 of the RLock class is used, which really unlocks it even when it has
297 been recursively acquired several times. Another internal interface is
298 then used to restore the recursion level when the lock is reacquired.
299 """
300 #等待直到通知或超时发生。如果调用该方法时调用的线程没有获得锁，则会引发运行时错误。
301 # 该方法释放底层锁，然后阻塞，直到它被另一个线程中的notify()或notify_all()调用
302 # 唤醒，或者直到出现可选超时为止。一旦被唤醒或超时，它会重新获得锁并返回。
303 # 当出现timeout参数而不是None时，它应该是一个浮点数，以秒(或几分之一)为单位指定
304 # 操作的超时。当底层锁是RLock时，不会使用其release()方法释放它，因为当递归地多次
305 # 获取锁时，这可能不会真正解锁它。相反，使用了RLock类的内部接口，即使递归地获得了
306 # 多次，它也会真正地解锁它。然后使用另一个内部接口在重新获得锁时恢复递归级别。
307
308 if not self._is_owned():
309 raise RuntimeError("cannot wait on un-acquired lock")
310 waiter = _allocate_lock()
311 waiter.acquire()
312 self._waiters.append(waiter)
313 saved_state = self._release_save()
314 gotit = False
315 try: # restore state no matter what (e.g., KeyboardInterrupt)
316 #无论如何都要恢复状态(例如，键盘中断)
317 if timeout is None:
318 waiter.acquire()
319 gotit = True
320 else:
321 if timeout > 0:
322 gotit = waiter.acquire(True, timeout)
323 else:
324 gotit = waiter.acquire(False)
325 return gotit
326 finally:
327 self._acquire_restore(saved_state)
328 if not gotit:
329 try:
330 self._waiters.remove(waiter)
331 except ValueError:
332 pass
333
334 def wait_for(self, predicate, timeout=None):
335 """Wait until a condition evaluates to True.
336 predicate should be a callable which result will be interpreted as a
337 boolean value. A timeout may be provided giving the maximum time to
338 wait.
339 """
340 #等待，直到条件的值为True。谓词应该是可调用的，其结果将被解释为布尔值。
341 # 可能会提供一个超时，以提供最长的等待时间。
342
343 endtime = None
344 waittime = timeout
345 result = predicate()
346 while not result:
347 if waittime is not None:
348 if endtime is None:
349 endtime = _time() + waittime
350 else:
351 waittime = endtime - _time()
352 if waittime <= 0:
353 break
354 self.wait(waittime)
355 result = predicate()
356 return result
357
358 def notify(self, n=1):
359 """Wake up one or more threads waiting on this condition, if any.
360 If the calling thread has not acquired the lock when this method is
361 called, a RuntimeError is raised.
362 This method wakes up at most n of the threads waiting for the condition
363 variable; it is a no-op if no threads are waiting.
364 """
365 #唤醒在此条件下等待的一个或多个线程(如果有的话)。如果调用该方法时调用的线程没有获得锁，
366 # 则会引发运行时错误。该方法最多唤醒n个等待条件变量的线程;如果没有线程在等待，那么
367 # 这是一个no-op。
368 if not self._is_owned():
369 raise RuntimeError("cannot notify on un-acquired lock")
370 all_waiters = self._waiters
371 waiters_to_notify = _deque(_islice(all_waiters, n))
372 if not waiters_to_notify:
373 return
374 for waiter in waiters_to_notify:
375 waiter.release()
376 try:
377 all_waiters.remove(waiter)
378 except ValueError:
379 pass
380
381 def notify_all(self):
382 """Wake up all threads waiting on this condition.
383 If the calling thread has not acquired the lock when this method
384 is called, a RuntimeError is raised.
385 """
386 #唤醒在此条件下等待的所有线程。如果调用该方法时调用的线程没有获得锁，
387 # 则会引发运行时错误。
388 self.notify(len(self._waiters))
389
390 notifyAll = notify_all
391
392
393 class Semaphore:
394 """This class implements semaphore objects.
395 Semaphores manage a counter representing the number of release() calls minus
396 the number of acquire() calls, plus an initial value. The acquire() method
397 blocks if necessary until it can return without making the counter
398 negative. If not given, value defaults to 1.
399 """
400 #这个类实现信号量对象。信号量管理一个计数器，该计数器表示release()调用的数量减去
401 # acquire()调用的数量，再加上一个初始值。acquire()方法如果有必要会阻塞，直到它可以
402 # 返回而不会使计数器变为负数为止。如果未指定，值默认为1。
403
404 # After Tim Peters' semaphore class, but not quite the same (no maximum)
405 #在Tim Peters的信号量类之后，但不完全相同(没有最大值)
406
407 def __init__(self, value=1):
408 if value < 0:
409 raise ValueError("semaphore initial value must be >= 0")
410 self._cond = Condition(Lock())
411 self._value = value
412
413 def acquire(self, blocking=True, timeout=None):
414 """Acquire a semaphore, decrementing the internal counter by one.
415 When invoked without arguments: if the internal counter is larger than
416 zero on entry, decrement it by one and return immediately. If it is zero
417 on entry, block, waiting until some other thread has called release() to
418 make it larger than zero. This is done with proper interlocking so that
419 if multiple acquire() calls are blocked, release() will wake exactly one
420 of them up. The implementation may pick one at random, so the order in
421 which blocked threads are awakened should not be relied on. There is no
422 return value in this case.
423 When invoked with blocking set to true, do the same thing as when called
424 without arguments, and return true.
425 When invoked with blocking set to false, do not block. If a call without
426 an argument would block, return false immediately; otherwise, do the
427 same thing as when called without arguments, and return true.
428 When invoked with a timeout other than None, it will block for at
429 most timeout seconds. If acquire does not complete successfully in
430 that interval, return false. Return true otherwise.
431 """
432 #获得一个信号量，将内部计数器减1。在没有参数的情况下调用时:如果内部计数器在入口时
433 # 大于0，则将其递减1并立即返回。如果进入时为零，阻塞，等待其他线程调用release()
434 # 使其大于零。这是通过适当的联锁完成的，这样，如果多个acquire()调用被阻塞，
435 # release()就会唤醒其中一个调用。实现可以随机选择一个线程，因此不应该依赖于被阻塞
436 # 线程被唤醒的顺序。在本例中没有返回值。当阻塞集调用为true时，执行与没有参数调用
437 # 时相同的操作，并返回true。当阻塞设置为false时，不要阻塞。如果一个没有参数的
438 # 调用将阻塞，立即返回false;否则，执行与没有参数调用时相同的操作，并返回true。
439 # 当使用除None以外的超时调用时，它最多将阻塞超时秒。如果在那段时间里收购没有成功
440 # 完成，还假。否则返回true。
441 if not blocking and timeout is not None:
442 raise ValueError("can't specify timeout for non-blocking acquire")
443 rc = False
444 endtime = None
445 with self._cond:
446 while self._value == 0:
447 if not blocking:
448 break
449 if timeout is not None:
450 if endtime is None:
451 endtime = _time() + timeout
452 else:
453 timeout = endtime - _time()
454 if timeout <= 0:
455 break
456 self._cond.wait(timeout)
457 else:
458 self._value -= 1
459 rc = True
460 return rc
461
462 __enter__ = acquire
463
464 def release(self):
465 """Release a semaphore, incrementing the internal counter by one.
466 When the counter is zero on entry and another thread is waiting for it
467 to become larger than zero again, wake up that thread.
468 """
469 #释放信号量，增加一个内部计数器。当进入时计数器为零，而另一个线程正在等待计数器
470 # 再次大于零时，唤醒该线程。
471 with self._cond:
472 self._value += 1
473 self._cond.notify()
474
475 def __exit__(self, t, v, tb):
476 self.release()
477
478
479 class BoundedSemaphore(Semaphore):
480 """Implements a bounded semaphore.
481 A bounded semaphore checks to make sure its current value doesn't exceed its
482 initial value. If it does, ValueError is raised. In most situations
483 semaphores are used to guard resources with limited capacity.
484 If the semaphore is released too many times it's a sign of a bug. If not
485 given, value defaults to 1.
486 Like regular semaphores, bounded semaphores manage a counter representing
487 the number of release() calls minus the number of acquire() calls, plus an
488 initial value. The acquire() method blocks if necessary until it can return
489 without making the counter negative. If not given, value defaults to 1.
490 """
491 #实现有界信号量。有界信号量检查其当前值是否不超过初始值。如果是，则会引发ValueError。
492 # 在大多数情况下，信号量被用来保护有限容量的资源。如果信号量被释放了太多次，这是错误
493 # 的信号。如果未指定，值默认为1。与常规信号量一样，有界信号量管理一个计数器，
494 # 表示release()调用的数量减去acquire()调用的数量，再加上一个初始值。acquire()方法
495 # 如果有必要会阻塞，直到它可以返回而不会使计数器变为负数为止。如果未指定，值默认为1。
496
497 def __init__(self, value=1):
498 Semaphore.__init__(self, value)
499 self._initial_value = value
500
501 def release(self):
502 """Release a semaphore, incrementing the internal counter by one.
503
504 When the counter is zero on entry and another thread is waiting for it
505 to become larger than zero again, wake up that thread.
506
507 If the number of releases exceeds the number of acquires,
508 raise a ValueError.
509 """
510 #释放信号量，增加一个内部计数器。当进入时计数器为0，而另一个线程正在等待i再次
511 # 大于0时，唤醒那个线程。如果发布的数量超过了获得的数量，则引发一个ValueError。
512 with self._cond:
513 if self._value >= self._initial_value:
514 raise ValueError("Semaphore released too many times")
515 self._value += 1
516 self._cond.notify()
517
518
519 class Event:
520 """Class implementing event objects.
521
522 Events manage a flag that can be set to true with the set() method and reset
523 to false with the clear() method. The wait() method blocks until the flag is
524 true. The flag is initially false.
525 """
526 #类实现事件对象。事件管理的标志可以用set()方法设置为true，用clear()方法重置为false。
527 # wait()方法将阻塞，直到标记为true。标志最初是假的。
528
529 # After Tim Peters' event class (without is_posted())
530 #在Tim Peters的事件类之后(没有is_post ())
531
532 def __init__(self):
533 self._cond = Condition(Lock())
534 self._flag = False
535
536 def _reset_internal_locks(self):
537 # private! called by Thread._reset_internal_locks by _after_fork()
538 #私人!调用线程._reset_internal_locks _after_fork()
539 self._cond.__init__(Lock())
540
541 def is_set(self):
542 """Return true if and only if the internal flag is true."""
543 #当且仅当内部标志为true时返回true。
544 return self._flag
545
546 isSet = is_set
547
548 def set(self):
549 """Set the internal flag to true.
550 All threads waiting for it to become true are awakened. Threads
551 that call wait() once the flag is true will not block at all.
552 """
553 #将内部标志设置为true。等待它成真的所有线程都被唤醒。一旦标志为true，
554 # 调用wait()的线程将不会阻塞。
555 with self._cond:
556 self._flag = True
557 self._cond.notify_all()
558
559 def clear(self):
560 """Reset the internal flag to false.
561 Subsequently, threads calling wait() will block until set() is called to
562 set the internal flag to true again.
563 """
564 #将内部标志重置为false。随后，调用wait()的线程将阻塞，直到调用set()将内部标志再次设置为true。
565 with self._cond:
566 self._flag = False
567
568 def wait(self, timeout=None):
569 """Block until the internal flag is true.
570 If the internal flag is true on entry, return immediately. Otherwise,
571 block until another thread calls set() to set the flag to true, or until
572 the optional timeout occurs.
573 When the timeout argument is present and not None, it should be a
574 floating point number specifying a timeout for the operation in seconds
575 (or fractions thereof).
576 This method returns the internal flag on exit, so it will always return
577 True except if a timeout is given and the operation times out.
578 """
579 #阻塞，直到内部标志为true。如果进入时内部标志为true，则立即返回。否则，阻塞直到
580 # 另一个线程调用set()将标志设置为true，或者直到出现可选超时。当出现timeout参数
581 # 而不是None时，它应该是一个浮点数，以秒(或几分之一)为单位指定操作的超时。这个
582 # 方法在退出时返回内部标志，因此它总是返回True，除非超时和操作超时。
583 with self._cond:
584 signaled = self._flag
585 if not signaled:
586 signaled = self._cond.wait(timeout)
587 return signaled
588
589
590 # A barrier class. Inspired in part by the pthread_barrier_* api and
591 # the CyclicBarrier class from Java. See
592 '''一个障碍类。部分灵感来自于pthread_barrier_* api和来自Java的循环屏障类。看到'''
593 # http://sourceware.org/pthreads-win32/manual/pthread_barrier_init.html and
594 # http://java.sun.com/j2se/1.5.0/docs/api/java/util/concurrent/
595 # CyclicBarrier.html
596 # for information. ##获取信息
597 # We maintain two main states, 'filling' and 'draining' enabling the barrier
598 # to be cyclic. Threads are not allowed into it until it has fully drained
599 # since the previous cycle. In addition, a 'resetting' state exists which is
600 # similar to 'draining' except that threads leave with a BrokenBarrierError,
601 # and a 'broken' state in which all threads get the exception.
602 '''我们维持两种主要状态，“填充”和“排水”，使屏障是循环的。线程不允许进入它，直到它从
603 上一个循环中完全耗尽为止。此外，存在一种“重置”状态，类似于“耗尽”状态，只是线程留下了
604 一个故障的barriererror错误，以及所有线程都得到异常的“中断”状态。'''
605 class Barrier:
606 """Implements a Barrier.
607 Useful for synchronizing a fixed number of threads at known synchronization
608 points. Threads block on 'wait()' and are simultaneously once they have all
609 made that call.
610 """
611 #实现了一个障碍。用于在已知同步点同步固定数量的线程。线程阻塞在'wait()'上，
612 # 并且一旦它们都进行了该调用，就会同时阻塞。
613
614 def __init__(self, parties, action=None, timeout=None):
615 """Create a barrier, initialised to 'parties' threads.
616 'action' is a callable which, when supplied, will be called by one of
617 the threads after they have all entered the barrier and just prior to
618 releasing them all. If a 'timeout' is provided, it is uses as the
619 default for all subsequent 'wait()' calls.
620 """
621 #创建一个障碍，初始化为“party”线程。“action”是一个可调用的线程，当它被提供时，
622 # 它将被其中一个线程在它们全部进入壁垒并释放它们之前调用。如果提供了'timeout'，
623 # 那么它将用作所有后续'wait()'调用的默认值。
624 self._cond = Condition(Lock())
625 self._action = action
626 self._timeout = timeout
627 self._parties = parties
628 self._state = 0 #0 filling, 1, draining, -1 resetting, -2 broken
629 self._count = 0
630
631 def wait(self, timeout=None):
632 """Wait for the barrier.
633 When the specified number of threads have started waiting, they are all
634 simultaneously awoken. If an 'action' was provided for the barrier, one
635 of the threads will have executed that callback prior to returning.
636 Returns an individual index number from 0 to 'parties-1'.
637 """
638 #等待障碍。当指定数量的线程开始等待时，它们都同时被唤醒。如果为barrier提供了一个
639 # “操作”，其中一个线程将在返回之前执行该回调。返回从0到“parties-1”的单个索引号。
640
641 if timeout is None:
642 timeout = self._timeout
643 with self._cond:
644 self._enter() # Block while the barrier drains. 隔离墙排水时要进行隔离。
645 index = self._count
646 self._count += 1
647 try:
648 if index + 1 == self._parties:
649 # We release the barrier
650 self._release()
651 else:
652 # We wait until someone releases us
653 self._wait(timeout)
654 return index
655 finally:
656 self._count -= 1
657 # Wake up any threads waiting for barrier to drain.
658 #唤醒任何等待屏障耗尽的线程。
659 self._exit()
660
661 # Block until the barrier is ready for us, or raise an exception
662 # if it is broken.
663 #阻止，直到障碍为我们准备好，或提出一个例外，如果它被打破。
664 def _enter(self):
665 while self._state in (-1, 1):
666 # It is draining or resetting, wait until done正在排水或重置，等待完成
667 self._cond.wait()
668 #see if the barrier is in a broken state看看势垒是否处于破碎状态
669 if self._state < 0:
670 raise BrokenBarrierError
671 assert self._state == 0
672
673 # Optionally run the 'action' and release the threads waiting
674 # in the barrier.
675 #可以选择运行“action”，并释放等待在barrier中的线程。
676
677 def _release(self):
678 try:
679 if self._action:
680 self._action()
681 # enter draining state 进入排水状态
682 self._state = 1
683 self._cond.notify_all()
684 except:
685 #an exception during the _action handler. Break and reraise
686 #_action处理程序期间的异常。打破和reraise
687 self._break()
688 raise
689
690 # Wait in the barrier until we are released. Raise an exception
691 # if the barrier is reset or broken.
692 #在障碍物里等着，直到我们被释放。如果障碍被重置或破坏，则引发异常。
693 def _wait(self, timeout):
694 if not self._cond.wait_for(lambda : self._state != 0, timeout):
695 #timed out. Break the barrier
696 self._break()
697 raise BrokenBarrierError
698 if self._state < 0:
699 raise BrokenBarrierError
700 assert self._state == 1
701
702 # If we are the last thread to exit the barrier, signal any threads
703 # # waiting for the barrier to drain.
704 #如果我们是最后一个退出屏障的线程，那么向等待屏障流出的线程发出信号。
705 def _exit(self):
706 if self._count == 0:
707 if self._state in (-1, 1):
708 #resetting or draining
709 self._state = 0
710 self._cond.notify_all()
711
712 def reset(self):
713 """Reset the barrier to the initial state.
714 Any threads currently waiting will get the BrokenBarrier exception
715 raised.
716 """
717 #将势垒重置为初始状态。当前等待的任何线程都将引发故障障碍异常。
718 with self._cond:
719 if self._count > 0:
720 if self._state == 0:
721 #reset the barrier, waking up threads 重置障碍，唤醒线程
722 self._state = -1
723 elif self._state == -2:
724 #was broken, set it to reset state 被破坏，设置为重置状态
725 #which clears when the last thread exits 最后一个线程退出时哪个线程清除
726 self._state = -1
727 else:
728 self._state = 0
729 self._cond.notify_all()
730
731 def abort(self):
732 """Place the barrier into a 'broken' state.
733 Useful in case of error. Any currently waiting threads and threads
734 attempting to 'wait()' will have BrokenBarrierError raised.
735 """
736 #将障碍设置为“破碎”状态。在发生错误时很有用。任何当前正在等待的线程和
737 # 试图“wait()”的线程都会出现故障障碍。
738 with self._cond:
739 self._break()
740
741 def _break(self):
742 # An internal error was detected. The barrier is set to
743 # a broken state all parties awakened.
744 #检测到内部错误。障碍被设置为一个破碎的国家，所有各方都觉醒了。
745 self._state = -2
746 self._cond.notify_all()
747
748 @property
749 def parties(self):
750 """Return the number of threads required to trip the barrier."""
751 #返回跳闸所需的线程数。
752 return self._parties
753
754 @property
755 def n_waiting(self):
756 """Return the number of threads currently waiting at the barrier."""
757 #返回阻塞处当前等待的线程数。
758 # We don't need synchronization here since this is an ephemeral result
759 # anyway. It returns the correct value in the steady state.
760 #我们不需要同步，因为这是一个短暂的结果。它在稳定状态下返回正确的值。
761 if self._state == 0:
762 return self._count
763 return 0
764
765 @property
766 def broken(self):
767 """Return True if the barrier is in a broken state."""
768 #如果屏障处于破坏状态，返回True。
769 return self._state == -2
770
771 # exception raised by the Barrier class
772 #由Barrier类引发的异常
773 class BrokenBarrierError(RuntimeError):
774 pass
775
776
777 # Helper to generate new thread names
778 #帮助程序生成新的线程名称
779 _counter = _count().__next__
780 _counter() # Consume 0 so first non-main thread has id 1.
781 #消耗0，所以第一个非主线程id为1。
782 def _newname(template="Thread-%d"):
783 return template % _counter()
784
785 # Active thread administration #活动线程管理
786 _active_limbo_lock = _allocate_lock()
787 _active = {} # maps thread id to Thread object 将线程id映射到线程对象
788 _limbo = {}
789 _dangling = WeakSet()
790
791 # Main class for threads
792 '''线程的主类'''
793
794 class Thread:
795 """A class that represents a thread of control.
796 This class can be safely subclassed in a limited fashion. There are two ways
797 to specify the activity: by passing a callable object to the constructor, or
798 by overriding the run() method in a subclass.
799 """
800 #表示控制线程的类。这个类可以以有限的方式安全地子类化。有两种方法可以指定活动:
801 # 通过将可调用对象传递给构造函数，或者在子类中重写run()方法。
802
803 _initialized = False
804 # Need to store a reference to sys.exc_info for printing
805 # out exceptions when a thread tries to use a global var. during interp.
806 # shutdown and thus raises an exception about trying to perform some
807 # operation on/with a NoneType
808 #需要存储对sys的引用。exc_info用于在interp期间线程试图使用全局变量时打印异常。
809 # 关闭，因此引发了一个异常，即试图对/使用非etype执行某些操作
810 _exc_info = _sys.exc_info
811 # Keep sys.exc_clear too to clear the exception just before
812 # allowing .join() to return.
813 #Keep sys.ex_clear也可以在allowing.join()返回之前清除异常。
814 #XXX __exc_clear = _sys.exc_clear
815
816 def __init__(self, group=None, target=None, name=None,
817 args=(), kwargs=None, *, daemon=None):
818 """This constructor should always be called with keyword arguments. Arguments are:
819 *group* should be None; reserved for future extension when a ThreadGroup
820 class is implemented.
821 *target* is the callable object to be invoked by the run()
822 method. Defaults to None, meaning nothing is called.
823 *name* is the thread name. By default, a unique name is constructed of
824 the form "Thread-N" where N is a small decimal number.
825 *args* is the argument tuple for the target invocation. Defaults to ().
826 *kwargs* is a dictionary of keyword arguments for the target
827 invocation. Defaults to {}.
828 If a subclass overrides the constructor, it must make sure to invoke
829 the base class constructor (Thread.__init__()) before doing anything
830 else to the thread.
831 """
832 #这个构造函数应该总是使用关键字参数调用。论点是:*group*不应该是;在实现
833 # ThreadGroup类时为将来的扩展保留。*target*是run()方法调用的可调用对象。
834 # 默认为None，表示不调用任何东西。*name*是线程名。默认情况下，唯一的名称
835 # 是由“Thread-N”的形式构造的，其中N是一个小数。*args*是目标调用的参数元组。
836 # 默认为()。*kwargs*是目标调用的关键字参数字典。默认为{}。如果子类重写构造
837 # 函数，它必须确保在对线程执行其他操作之前调用基类构造函数(thread. __init__())。
838
839 assert group is None, "group argument must be None for now"
840 if kwargs is None:
841 kwargs = {}
842 self._target = target
843 self._name = str(name or _newname())
844 self._args = args
845 self._kwargs = kwargs
846 if daemon is not None:
847 self._daemonic = daemon
848 else:
849 self._daemonic = current_thread().daemon
850 self._ident = None
851 self._tstate_lock = None
852 self._started = Event()
853 self._is_stopped = False
854 self._initialized = True
855 # sys.stderr is not stored in the class like
856 # sys.exc_info since it can be changed between instances
857 self._stderr = _sys.stderr
858 # For debugging and _after_fork()
859 _dangling.add(self)
860
861 def _reset_internal_locks(self, is_alive):
862 # private! Called by _after_fork() to reset our internal locks as
863 # they may be in an invalid state leading to a deadlock or crash.
864 #私人!由_after_fork()调用，以重置内部锁，因为它们可能处于无效状态，导致死锁或崩溃。
865 self._started._reset_internal_locks()
866 if is_alive:
867 self._set_tstate_lock()
868 else:
869 # The thread isn't alive after fork: it doesn't have a tstate anymore.
870 #在fork之后，线程不再是活的:它不再有tstate。
871 self._is_stopped = True
872 self._tstate_lock = None
873
874 def __repr__(self):
875 assert self._initialized, "Thread.__init__() was not called"
876 status = "initial"
877 if self._started.is_set():
878 status = "started"
879 self.is_alive() # easy way to get ._is_stopped set when appropriate
880 #在适当的情况下，获得._is_stopped设置的简单方法
881 if self._is_stopped:
882 status = "stopped"
883 if self._daemonic:
884 status += " daemon"
885 if self._ident is not None:
886 status += " %s" % self._ident
887 return "<%s(%s, %s)>" % (self.__class__.__name__, self._name, status)
888
889 def start(self):
890 """Start the thread's activity.
891 It must be called at most once per thread object. It arranges for the
892 object's run() method to be invoked in a separate thread of control.
893 This method will raise a RuntimeError if called more than once on the
894 same thread object.
895 """
896 #启动线程的活动。每个线程对象最多只能调用一次。它安排在一个单独的控制线程中
897 # 调用对象的run()方法。如果在同一个线程对象上调用多次，此方法将引发运行时错误。
898 if not self._initialized:
899 raise RuntimeError("thread.__init__() not called")
900
901 if self._started.is_set():
902 raise RuntimeError("threads can only be started once")
903 with _active_limbo_lock:
904 _limbo[self] = self
905 try:
906 _start_new_thread(self._bootstrap, ())
907 except Exception:
908 with _active_limbo_lock:
909 del _limbo[self]
910 raise
911 self._started.wait()
912
913 def run(self):
914 """Method representing the thread's activity.
915 You may override this method in a subclass. The standard run() method
916 invokes the callable object passed to the object's constructor as the
917 target argument, if any, with sequential and keyword arguments taken
918 from the args and kwargs arguments, respectively.
919 """
920 #表示线程活动的方法。您可以在子类中重写此方法。标准run()方法调用传递给对象
921 # 构造函数的可调用对象作为目标参数(如果有的话)，分别使用args和kwargs参数
922 # 中的顺序参数和关键字参数。
923 try:
924 if self._target:
925 self._target(*self._args, **self._kwargs)
926 finally:
927 # Avoid a refcycle if the thread is running a function with
928 # an argument that has a member that points to the thread.
929 #如果线程正在运行一个具有指向线程的成员的参数的函数，请避免使用refcycle。
930 del self._target, self._args, self._kwargs
931
932 def _bootstrap(self):
933 # Wrapper around the real bootstrap code that ignores
934 # exceptions during interpreter cleanup. Those typically
935 # happen when a daemon thread wakes up at an unfortunate
936 # moment, finds the world around it destroyed, and raises some
937 # random exception *** while trying to report the exception in
938 # _bootstrap_inner() below ***. Those random exceptions
939 # don't help anybody, and they confuse users, so we suppress
940 # them. We suppress them only when it appears that the world
941 # indeed has already been destroyed, so that exceptions in
942 # _bootstrap_inner() during normal business hours are properly
943 # reported. Also, we only suppress them for daemonic threads;
944 # if a non-daemonic encounters this, something else is wrong.
945 '''包装真正的引导代码，在解释器清理期间忽略异常。这通常发生在守护进程线程
946 在一个不幸的时刻醒来，发现它周围的世界被破坏，并在试图报告***下面的异常
947 in_bootstrap_inner()时引发一些随机异常时。这些随机的异常对任何人都没有
948 帮助，而且它们混淆了用户，所以我们抑制了它们。只有当世界似乎确实已经被破坏
949 时，我们才会抑制它们，以便在正常工作时间内正确报告_bootstrap_inner()中
950 的异常。而且，我们只对daemonic线程禁止它们;如果一个非daemonic遇到了这个
951 问题，就会出现其他问题'''
952 try:
953 self._bootstrap_inner()
954 except:
955 if self._daemonic and _sys is None:
956 return
957 raise
958
959 def _set_ident(self):
960 self._ident = get_ident()
961
962 def _set_tstate_lock(self):
963 """
964 Set a lock object which will be released by the interpreter when
965 the underlying thread state (see pystate.h) gets deleted.
966 """
967 #设置一个锁对象，当底层线程状态(请参阅pystate.h)被删除时，解释器将释放这个锁对象。
968 self._tstate_lock = _set_sentinel()
969 self._tstate_lock.acquire()
970
971 def _bootstrap_inner(self):
972 try:
973 self._set_ident()
974 self._set_tstate_lock()
975 self._started.set()
976 with _active_limbo_lock:
977 _active[self._ident] = self
978 del _limbo[self]
979
980 if _trace_hook:
981 _sys.settrace(_trace_hook)
982 if _profile_hook:
983 _sys.setprofile(_profile_hook)
984
985 try:
986 self.run()
987 except SystemExit:
988 pass
989 except:
990 # If sys.stderr is no more (most likely from interpreter
991 # shutdown) use self._stderr. Otherwise still use sys (as in
992 # _sys) in case sys.stderr was redefined since the creation of
993 # self.
994 #如果系统。stderr不再使用self._stderr(很可能是由于解释器关闭)。否则，
995 # 在case sys中仍然使用sys(如in_sys)。stderr自自我创造以来被重新定义。
996 if _sys and _sys.stderr is not None:
997 print("Exception in thread %s:\n%s" %
998 (self.name, _format_exc()), file=_sys.stderr)
999 elif self._stderr is not None:
1000 # Do the best job possible w/o a huge amt. of code to
1001 # approximate a traceback (code ideas from Lib/traceback.py)
1002 #尽最大的努力做最好的工作。近似回溯的代码(来自Lib/traceback.py的代码思想)
1003 exc_type, exc_value, exc_tb = self._exc_info()
1004 try:
1005 print((
1006 "Exception in thread " + self.name +
1007 " (most likely raised during interpreter shutdown):"), file=self._stderr)
1008 print((
1009 "Traceback (most recent call last):"), file=self._stderr)
1010 while exc_tb:
1011 print((
1012 ' File "%s", line %s, in %s' %
1013 (exc_tb.tb_frame.f_code.co_filename,
1014 exc_tb.tb_lineno,
1015 exc_tb.tb_frame.f_code.co_name)), file=self._stderr)
1016 exc_tb = exc_tb.tb_next
1017 print(("%s: %s" % (exc_type, exc_value)), file=self._stderr)
1018 self._stderr.flush()
1019 # Make sure that exc_tb gets deleted since it is a memory
1020 # hog; deleting everything else is just for thoroughness
1021 #确保exc_tb被删除，因为它占用内存;删除所有其他内容只是为了彻底
1022 finally:
1023 del exc_type, exc_value, exc_tb
1024 finally:
1025 # Prevent a race in
1026 # test_threading.test_no_refcycle_through_target when
1027 # the exception keeps the target alive past when we
1028 # assert that it's dead.
1029 #防止test_threading中的竞争。test_no_refcycle_through_target，
1030 # 当异常断言目标已死时，该异常将使目标保持存活。
1031 #XXX self._exc_clear()
1032 pass
1033 finally:
1034 with _active_limbo_lock:
1035 try:
1036 # We don't call self._delete() because it also
1037 # grabs _active_limbo_lock.
1038 #我们不调用self._delete()，因为它也抓取_active_limbo_lock。
1039 del _active[get_ident()]
1040 except:
1041 pass
1042
1043 def _stop(self):
1044 # After calling ._stop(), .is_alive() returns False and .join() returns
1045 # immediately. ._tstate_lock must be released before calling ._stop().
1046 #调用._stop()后，.is_alive()返回False， .join()立即返回。
1047
1048 # Normal case: C code at the end of the thread's life
1049 # (release_sentinel in _threadmodule.c) releases ._tstate_lock, and
1050 # that's detected by our ._wait_for_tstate_lock(), called by .join()
1051 # and .is_alive(). Any number of threads _may_ call ._stop()
1052 # simultaneously (for example, if multiple threads are blocked in
1053 # .join() calls), and they're not serialized. That's harmless -
1054 # they'll just make redundant rebindings of ._is_stopped and
1055 # ._tstate_lock. Obscure: we rebind ._tstate_lock last so that the
1056 # "assert self._is_stopped" in ._wait_for_tstate_lock() always works
1057 # (the assert is executed only if ._tstate_lock is None).
1058 #正常情况:线程生命周期结束时的C代码(_threadmodule.c中的release_sentinel)
1059 # 释放了._tstate_lock，我们的._wait_for_tstate_lock()检测到这一点，
1060 # 它被.join()和.is_alive()调用。同时调用任意数量的线程_may_ ._stop()
1061 # (例如，如果多个线程在.join()调用中被阻塞，并且它们没有被序列化)。这是无害的，
1062 # 他们只会对._is_stopped和._tstate_lock进行冗余的重绑定。晦涩的:
1063 # 我们将._tstate_lock绑定到最后，以便“断言self”。_is_stopped()中
1064 # 的._wait_for_tstate_lock()总是有效的(只有当._tstate_lock为空时才执行断言)。
1065
1066 # Special case: _main_thread releases ._tstate_lock via this
1067 # module's _shutdown() function.
1068 #特殊情况:_main_thread通过这个模块的_shutdown()函数释放._tstate_lock。
1069 lock = self._tstate_lock
1070 if lock is not None:
1071 assert not lock.locked()
1072 self._is_stopped = True
1073 self._tstate_lock = None
1074
1075 def _delete(self):
1076 "Remove current thread from the dict of currently running threads."
1077 with _active_limbo_lock:
1078 del _active[get_ident()]
1079 # There must not be any python code between the previous line
1080 # and after the lock is released. Otherwise a tracing function
1081 # could try to acquire the lock again in the same thread, (in
1082 # current_thread()), and would block.
1083 #前一行和锁释放后之间不应该有任何python代码。否则，跟踪函数可以尝试在相
1084 # 同的线程(在current_thread()中)中再次获取锁，并将阻塞。
1085
1086 def join(self, timeout=None):
1087 """Wait until the thread terminates.
1088 This blocks the calling thread until the thread whose join() method is
1089 called terminates -- either normally or through an unhandled exception
1090 or until the optional timeout occurs.
1091 When the timeout argument is present and not None, it should be a
1092 floating point number specifying a timeout for the operation in seconds
1093 (or fractions thereof). As join() always returns None, you must call
1094 isAlive() after join() to decide whether a timeout happened -- if the
1095 thread is still alive, the join() call timed out.
1096 When the timeout argument is not present or None, the operation will
1097 block until the thread terminates.
1098 A thread can be join()ed many times.
1099 join() raises a RuntimeError if an attempt is made to join the current
1100 thread as that would cause a deadlock. It is also an error to join() a
1101 thread before it has been started and attempts to do so raises the same
1102 exception.
1103 """
1104 #等待直到线程终止。这将阻塞调用线程，直到调用join()方法的线程终止——通常或通过
1105 # 未处理的异常终止，或直到出现可选超时为止。当出现timeout参数而不是None时，
1106 # 它应该是一个浮点数，以秒(或几分之一)为单位指定操作的超时。因为join()总是
1107 # 返回None，所以必须在join()之后调用isAlive()，以决定是否发生超时——如果线程
1108 # 仍然活着，则join()调用超时。当timeout参数不存在或不存在时，操作将阻塞，
1109 # 直到线程终止。一个线程可以多次连接()ed。如果尝试连接当前线程，join()将引发
1110 # 一个运行时错误，因为这会导致死锁。在线程启动之前连接()线程也是一个错误，
1111 # 试图这样做会引发相同的异常。
1112 if not self._initialized:
1113 raise RuntimeError("Thread.__init__() not called")
1114 if not self._started.is_set():
1115 raise RuntimeError("cannot join thread before it is started")
1116 if self is current_thread():
1117 raise RuntimeError("cannot join current thread")
1118
1119 if timeout is None:
1120 self._wait_for_tstate_lock()
1121 else:
1122 # the behavior of a negative timeout isn't documented, but
1123 # historically .join(timeout=x) for x<0 has acted as if timeout=0
1124 #没有记录消极超时的行为，但是在历史上，x<0时的.join(timeout=x)就像timeout=0一样
1125 self._wait_for_tstate_lock(timeout=max(timeout, 0))
1126
1127 def _wait_for_tstate_lock(self, block=True, timeout=-1):
1128 # Issue #18808: wait for the thread state to be gone.
1129 # At the end of the thread's life, after all knowledge of the thread
1130 # is removed from C data structures, C code releases our _tstate_lock.
1131 # This method passes its arguments to _tstate_lock.acquire().
1132 # If the lock is acquired, the C code is done, and self._stop() is
1133 # called. That sets ._is_stopped to True, and ._tstate_lock to None.
1134 #问题#18808:等待线程状态消失。在线程生命周期结束时，在从C数据结构中删除所有
1135 # 线程知识之后，C代码释放我们的_tstate_lock。该方法将其参数
1136 # 传递给_tstate_lock.acquire()。如果获得了锁，则完成C代码，
1137 # 并调用self._stop()。这将._is_stopped设置为True，._tstate_lock设置为None。
1138 lock = self._tstate_lock
1139 if lock is None: # already determined that the C code is done 已经确定C代码已经完成
1140 assert self._is_stopped
1141 elif lock.acquire(block, timeout):
1142 lock.release()
1143 self._stop()
1144
1145 @property
1146 def name(self):
1147 """A string used for identification purposes only.
1148 It has no semantics. Multiple threads may be given the same name. The
1149 initial name is set by the constructor.
1150 """
1151 #仅用于识别目的的字符串。它没有语义。多个线程可能被赋予相同的名称。初始名称由构造函数设置。
1152 assert self._initialized, "Thread.__init__() not called"
1153 return self._name
1154
1155 @name.setter
1156 def name(self, name):
1157 assert self._initialized, "Thread.__init__() not called"
1158 self._name = str(name)
1159
1160 @property
1161 def ident(self):
1162 """Thread identifier of this thread or None if it has not been started.
1163 This is a nonzero integer. See the get_ident() function. Thread
1164 identifiers may be recycled when a thread exits and another thread is
1165 created. The identifier is available even after the thread has exited.
1166 """
1167 #此线程的线程标识符，如果没有启动，则为空。这是非零整数。请参阅get_ident()函数。
1168 # 当线程退出并创建另一个线程时，可以回收线程标识符。即使线程已经退出，标识符也是可用的。
1169 assert self._initialized, "Thread.__init__() not called"
1170 return self._ident
1171
1172 def is_alive(self):
1173 """Return whether the thread is alive.
1174 This method returns True just before the run() method starts until just
1175 after the run() method terminates. The module function enumerate()
1176 returns a list of all alive threads.
1177 """
1178 #返回线程是否存在。这个方法在run()方法开始之前返回True，直到run()方法终止之后。
1179 # 模块函数enumerate()返回一个包含所有活线程的列表。
1180 assert self._initialized, "Thread.__init__() not called"
1181 if self._is_stopped or not self._started.is_set():
1182 return False
1183 self._wait_for_tstate_lock(False)
1184 return not self._is_stopped
1185
1186 isAlive = is_alive
1187
1188 @property
1189 def daemon(self):
1190 """A boolean value indicating whether this thread is a daemon thread.
1191 This must be set before start() is called, otherwise RuntimeError is
1192 raised. Its initial value is inherited from the creating thread; the
1193 main thread is not a daemon thread and therefore all threads created in
1194 the main thread default to daemon = False.
1195 The entire Python program exits when no alive non-daemon threads are
1196 left.
1197 """
1198 #一个布尔值，指示此线程是否为守护线程。这必须在调用start()之前设置，否则会引发
1199 # 运行时错误。它的初始值继承自创建线程;主线程不是守护进程线程，因此在主线程中
1200 # 创建的所有线程默认为守护进程= False。当没有存活的非守护进程线程时，
1201 # 整个Python程序退出。
1202 assert self._initialized, "Thread.__init__() not called"
1203 return self._daemonic
1204
1205 @daemon.setter
1206 def daemon(self, daemonic):
1207 if not self._initialized:
1208 raise RuntimeError("Thread.__init__() not called")
1209 if self._started.is_set():
1210 raise RuntimeError("cannot set daemon status of active thread")
1211 self._daemonic = daemonic
1212
1213 def isDaemon(self): #Daemon:守护进程
1214 return self.daemon
1215
1216 def setDaemon(self, daemonic):
1217 self.daemon = daemonic
1218
1219 def getName(self):
1220 return self.name
1221
1222 def setName(self, name):
1223 self.name = name
1224
1225 # The timer class was contributed by Itamar Shtull-Trauring
1226 #计时器类由Itamar Shtull-Trauring贡献
1227
1228 class Timer(Thread):
1229 """Call a function after a specified number of seconds:
1230 t = Timer(30.0, f, args=None, kwargs=None)
1231 t.start()
1232 t.cancel() # stop the timer's action if it's still waiting
1233 """
1234 #在指定的秒数后调用一个函数:t = Timer(30.0, f, args=None, kwargs=None)
1235 #t.start() t.cancel()如果计时器仍在等待，则停止计时器的操作
1236
1237 def __init__(self, interval, function, args=None, kwargs=None):
1238 Thread.__init__(self)
1239 self.interval = interval
1240 self.function = function
1241 self.args = args if args is not None else []
1242 self.kwargs = kwargs if kwargs is not None else {}
1243 self.finished = Event()
1244
1245 def cancel(self):
1246 """Stop the timer if it hasn't finished yet."""
1247 #如果计时器还没有结束，请停止。
1248 self.finished.set()
1249
1250 def run(self):
1251 self.finished.wait(self.interval)
1252 if not self.finished.is_set():
1253 self.function(*self.args, **self.kwargs)
1254 self.finished.set()
1255
1256
1257 # Special thread class to represent the main thread
1258 '''表示主线程的特殊线程类'''
1259
1260 class _MainThread(Thread):
1261
1262 def __init__(self):
1263 Thread.__init__(self, name="MainThread", daemon=False)
1264 self._set_tstate_lock()
1265 self._started.set()
1266 self._set_ident()
1267 with _active_limbo_lock:
1268 _active[self._ident] = self
1269
1270
1271 # Dummy thread class to represent threads not started here.
1272 # These aren't garbage collected when they die, nor can they be waited for.
1273 # If they invoke anything in threading.py that calls current_thread(), they
1274 # leave an entry in the _active dict forever after.
1275 # Their purpose is to return *something* from current_thread().
1276 # They are marked as daemon threads so we won't wait for them
1277 # when we exit (conform previous semantics).
1278 #伪线程类来表示这里没有启动的线程。它们死后不会被垃圾收集，也不会被等待。如果它们在
1279 # 线程中调用任何东西。调用current_thread()的py在_active dict中永远留下一个条目。
1280 # 它们的目的是从current_thread()返回*something*。它们被标记为守护线程，因此在退出
1281 # 时我们不会等待它们(符合前面的语义)。
1282
1283 class _DummyThread(Thread):
1284
1285 def __init__(self):
1286 Thread.__init__(self, name=_newname("Dummy-%d"), daemon=True)
1287
1288 self._started.set()
1289 self._set_ident()
1290 with _active_limbo_lock:
1291 _active[self._ident] = self
1292
1293 def _stop(self):
1294 pass
1295
1296 def is_alive(self):
1297 assert not self._is_stopped and self._started.is_set()
1298 return True
1299
1300 def join(self, timeout=None):
1301 assert False, "cannot join a dummy thread"
1302
1303
1304 # Global API functions
1305 #全球API函数
1306
1307 def current_thread():
1308 """Return the current Thread object, corresponding to the caller's thread of control.
1309 If the caller's thread of control was not created through the threading
1310 module, a dummy thread object with limited functionality is returned.
1311 """
1312 #返回当前线程对象，对应于调用方的控制线程。如果没有通过线程模块创建调用者的控制线
1313 # 程，则返回具有有限功能的虚拟线程对象。
1314 try:
1315 return _active[get_ident()]
1316 except KeyError:
1317 return _DummyThread()
1318
1319 currentThread = current_thread
1320
1321 def active_count():
1322 """Return the number of Thread objects currently alive.
1323 The returned count is equal to the length of the list returned by
1324 enumerate().
1325 """
1326 #返回当前存活的线程对象的数量。返回的计数等于enumerate()返回的列表的长度。
1327 with _active_limbo_lock:
1328 return len(_active) + len(_limbo)
1329
1330 activeCount = active_count
1331
1332 def _enumerate():
1333 # Same as enumerate(), but without the lock. Internal use only.
1334 #与enumerate()相同，只是没有锁。内部使用。
1335 return list(_active.values()) + list(_limbo.values())
1336
1337 def enumerate():
1338 """Return a list of all Thread objects currently alive.
1339 The list includes daemonic threads, dummy thread objects created by
1340 current_thread(), and the main thread. It excludes terminated threads and
1341 threads that have not yet been started.
1342 """
1343 #返回当前所有线程对象的列表。该列表包括daemonic线程、current_thread()创建的虚拟
1344 # 线程对象和主线程。它排除终止的线程和尚未启动的线程。
1345 with _active_limbo_lock:
1346 return list(_active.values()) + list(_limbo.values())
1347
1348 from _thread import stack_size
1349
1350 # Create the main thread object,
1351 # and make it available for the interpreter
1352 # (Py_Main) as threading._shutdown.
1353 #创建主线程对象，并将其作为thread ._shutdown提供给解释器(Py_Main)。
1354
1355 _main_thread = _MainThread()
1356
1357 def _shutdown():
1358 # Obscure: other threads may be waiting to join _main_thread. That's
1359 # dubious, but some code does it. We can't wait for C code to release
1360 # the main thread's tstate_lock - that won't happen until the interpreter
1361 # is nearly dead. So we release it here. Note that just calling _stop()
1362 # isn't enough: other threads may already be waiting on _tstate_lock.
1363 #晦涩:其他线程可能正在等待加入_main_thread。这很可疑，但有些代码可以做到。
1364 # 我们不能等待C代码释放主线程的tstate_lock——这要等到解释器快死的时候才会发生。
1365 # 我们在这里释放它。注意，仅仅调用_stop()是不够的:其他线程可能已经在
1366 # 等待_tstate_lock了。
1367 if _main_thread._is_stopped:
1368 # _shutdown() was already called
1369 return
1370 tlock = _main_thread._tstate_lock
1371 # The main thread isn't finished yet, so its thread state lock can't have
1372 # been released.
1373 #主线程尚未完成，因此它的线程状态锁无法释放。
1374 assert tlock is not None
1375 assert tlock.locked()
1376 tlock.release()
1377 _main_thread._stop()
1378 t = _pickSomeNonDaemonThread()
1379 while t:
1380 t.join()
1381 t = _pickSomeNonDaemonThread()
1382
1383 def _pickSomeNonDaemonThread():
1384 for t in enumerate():
1385 if not t.daemon and t.is_alive():
1386 return t
1387 return None
1388
1389 def main_thread():
1390 """Return the main thread object.
1391 In normal conditions, the main thread is the thread from which the
1392 Python interpreter was started.
1393 """
1394 #返回主线程对象。在正常情况下，主线程是Python解释器启动的线程。
1395 return _main_thread
1396
1397 # get thread-local implementation, either from the thread
1398 # module, or from the python fallback
1399 #从线程模块或python回退中获取线程本地实现
1400
1401 try:
1402 from _thread import _local as local
1403 except ImportError:
1404 from _threading_local import local
1405
1406
1407 def _after_fork():
1408 """
1409 Cleanup threading module state that should not exist after a fork.
1410 """
1411 # Reset _active_limbo_lock, in case we forked while the lock was held
1412 # by another (non-forked) thread. http://bugs.python.org/issue874900
1413 #Reset _active_limbo_lock,以防我们分叉而锁被另一个(非分叉的)线程持有。
1414 global _active_limbo_lock, _main_thread
1415 _active_limbo_lock = _allocate_lock()
1416
1417 # fork() only copied the current thread; clear references to others.
1418 #fork()只复制当前线程;明确提及他人。
1419 new_active = {}
1420 current = current_thread()
1421 _main_thread = current
1422 with _active_limbo_lock:
1423 # Dangling thread instances must still have their locks reset,
1424 # because someone may join() them.
1425 #悬空线程实例必须重新设置它们的锁，因为有人可能会加入()它们。
1426 threads = set(_enumerate())
1427 threads.update(_dangling)
1428 for thread in threads:
1429 # Any lock/condition variable may be currently locked or in an
1430 # invalid state, so we reinitialize them.
1431 #任何锁/条件变量可能当前被锁定或处于无效状态，因此我们重新初始化它们。
1432 if thread is current:
1433 # There is only one active thread. We reset the ident to
1434 # its new value since it can have changed.
1435 #只有一个活动线程。我们将ident重置为它的新值，因为它可能已经更改。
1436 thread._reset_internal_locks(True)
1437 ident = get_ident()
1438 thread._ident = ident
1439 new_active[ident] = thread
1440 else:
1441 # All the others are already stopped.
1442 thread._reset_internal_locks(False)
1443 thread._stop()
1444
1445 _limbo.clear()
1446 _active.clear()
1447 _active.update(new_active)
1448 assert len(_active) == 1
1449
1450
1451 if hasattr(_os, "register_at_fork"):
1452 _os.register_at_fork(after_in_child=_after_fork)
队列：
Python的Queue模块中提供了同步的、线程安全的队列类，包括FIFO（先入先出)队列Queue，LIFO（后入先出）队列LifoQueue，和优先级队列PriorityQueue。这些队列都实现了锁原语，能够在多线程中直接使用。可以使用队列来实现线程间的同步。
1 '''队列'''
2 import queue
3 q =queue.Queue() #设置队列
4 q.put("q1") #队列中放入数据
5 q.put("q2")
6 q.put("q3")
7
8 # print(q.qsize()) #获取队列大小
9
10 '''队列中获取数据，取出的数据超出存入数据时会等待，不会报错'''
11 print(q.get())
12 print(q.get())
13 print(q.get())
14 # print(q.get())
15
16 '''获取队列，但不会等待，超出后直接报错'''
17 print(q.get_nowait())
18 print(q.get_nowait())
19 print(q.get_nowait())
20 # print(q.get_nowait())
21
22 '''设置优先级排序的依据'''
23 q = queue.PriorityQueue(maxsize=0)
24 q.put((3,"q1")) #当maxsizie<=0时，队列无限大，>0时，给定数据即为队列大小
25 q.put((1,"q2"))
26 q.put((-4,"q3"))
27 print(q.get()) #获取时会从小到大按顺序获取
28 print(q.get())
29 print(q.get())
上述代码只是队列的应用，下面将队列应用与线程之中：
1 import queue
2 import time
3 import threading
4
5 q = queue.Queue(maxsize=10)
6 def gave(name):
7 count = 1
8 while True:
9 q.put("--骨头--%s" % count)
10 print("%s 生产骨头 %s" % (name,count))
11 time.sleep(1)
12 count+=1
13
14 def consumer(name):
15 while q.qsize()>0:
16 # while True:
17 print("%s 吃掉 %s" % (name,q.get()))
18 # time.sleep(10)
19
20 g = threading.Thread(target=gave,args=("王二小",))
21 c = threading.Thread(target=consumer,args=("旺财",))
22 g.start()
23 c.start()
1 #print('\033[41;1m--red light on---\033[0m') #红灯
2 #print('\033[43;1m--yellow light on---\033[0m') #黄灯
3 #print('\033[42;1m--green light on---\033[0m') #绿灯
4 '''主要用在数据同步上'''
5 '''红绿灯事件'''
6
7 # import threading
8 # import time
9 # # import queue
10 # event = threading.Event()
11 # # q = queue.Queue()
12 #
13 # def light():
14 # count = 1
15 # while True:
16 # if count<=5:
17 # event.set()
18 # print('\033[42;1m--green light on---\033[0m')
19 # elif 5<count<=10:
20 # event.clear()
21 # print('\033[43;1m--yellow light on---\033[0m')
22 # else:
23 # print('\033[41;1m--red light on---\033[0m')
24 # if count>=15:
25 # count = 0
26 # time.sleep(1)
27 # count+=1
28 #
29 # def car(name):
30 # while True:
31 # if event.is_set():
32 # time.sleep(1)
33 # print("%s is running..." % name)
34 # else:
35 # print("car is waiting...")
36 # event.wait() #等待事件event对象发生变化
37 #
38 #
39 # Light = threading.Thread(target=light,)
40 # Light.start()
41 # Car = threading.Thread(target=car,args=("BENZ",))
42 # Car.start()
43
44
45 import threading
46 import time
47 import queue
48
49 event=threading.Event()
50 q=queue.PriorityQueue(maxsize=20)
51 #在循环之前先放入十辆车：
52 for i in range(10):
53 q.put("旧车辆，%s" % "QQ")
54
55 def light():
56 count=0
57 while True:
58 if count<10:
59 event.set()
60 print("\033[42;1m--green light on---\033[0m",10-count)
61 elif 10<=count<15:
62 event.clear()
63 print("\033[43;1m--yellow light on---\033[0m",15-count)
64 else:
65 event.clear()
66 if count>=25:
67 count=0
68 continue
69 print("\033[41;1m--red light on---\033[0m",25-count)
70 time.sleep(1)
71 count+=1
72
73 def car(name):
74 while True:
75 if event.is_set() and q.qsize()>=1:
76 print("%s is running..." % name)
77 time.sleep(1)
78 print("道路还有【%s】辆车" % q.qsize())
79 else:
80 print("car is waiting...")
81 print("现在道路中有车%s辆" % q.qsize())
82 event.wait() #等待事件event对象发生变化
83
84 #路口停车
85 def Put():
86 n=0
87 while q.qsize()<20:
88 time.sleep(2)
89 q.put("新车辆%s车辆" % n)
90 n+=1
91 print("车辆已驶入")
92 else:
93 event.wait()
94 print("停止驶入")
95 print("停止驶入后道路中有车%s" % q.qsize())
96
97 #车辆行驶
98 def Get():
99 while True:
100 if event.is_set():
101 time.sleep(2)
102 print("%s车辆--------通过" % q.get())
103 else:
104 print("禁止通行！！")
105 event.wait()
106
107
108
109 C=threading.Thread(target=car,args=("...T...",))
110 L=threading.Thread(target=light)
111 P=threading.Thread(target=Put)
112 G=threading.Thread(target=Get)
113 L.start()
114 C.start()
115 P.start()
116 G.start()
1 import threading
2
3 money = 0
4 lock = threading.Lock()
5
6 #存钱
7 def get_money(Sum):
8 global money
9 money+=Sum #x=money+sum;money=x
10
11 #取钱
12 def put_money(Sum):
13 global money
14 money-=Sum
15
16 def run(Sum):
17 lock.acquire()
18 for i in range(10000):
19 put_money(Sum)
20 get_money(Sum)
21 lock.release()
22
23 #单线程中不会存在问题
24 #然而在多线程中，操作系统交叉处理赋值语句，导致
25 # 全局变量被一个线程修改，而另一个线程却不知情。
26 m1 = threading.Thread(target=run,args=(100,))
27 m2 = threading.Thread(target=run,args=(1000,))
28 m1.start()
29 m2.start()
30 m1.join()
31 m2.join()
32 print(money)
1 '''A multi-producer, multi-consumer queue.'''
2 #多生产者、多消费者队列。
3 import threading
4 from collections import deque
5 from heapq import heappush, heappop
6 from time import monotonic as time
7 try:
8 from _queue import SimpleQueue
9 except ImportError:
10 SimpleQueue = None
11
12 __all__ = ['Empty', 'Full', 'Queue', 'PriorityQueue', 'LifoQueue', 'SimpleQueue']
13
14
15 try:
16 from _queue import Empty
17 except AttributeError:
18 class Empty(Exception):
19 'Exception raised by Queue.get(block=0)/get_nowait().'
20 pass
21
22 class Full(Exception):
23 'Exception raised by Queue.put(block=0)/put_nowait().'
24 pass
25
26
27 class Queue:
28 '''Create a queue object with a given maximum size.
29 If maxsize is <= 0, the queue size is infinite.
30 '''
31 #创建一个具有给定最大大小的队列对象。如果maxsize <= 0，则队列大小为无穷大。
32
33 def __init__(self, maxsize=0):
34 self.maxsize = maxsize
35 self._init(maxsize)
36
37 # mutex must be held whenever the queue is mutating. All methods
38 # that acquire mutex must release it before returning. mutex
39 # is shared between the three conditions, so acquiring and
40 # releasing the conditions also acquires and releases mutex.
41 #当队列发生变化时，必须持有互斥锁。所有获得互斥锁的方法都必须在返回之前释放它。
42 # 互斥锁在这三个条件之间是共享的，因此获取和释放条件也获得和释放互斥锁。
43 self.mutex = threading.Lock()
44
45 # Notify not_empty whenever an item is added to the queue; a
46 # thread waiting to get is notified then.
47 #当一个项目被添加到队列中时，通知not_empty;然后会通知等待获取的线程。
48 self.not_empty = threading.Condition(self.mutex)
49
50 # Notify not_full whenever an item is removed from the queue;
51 # a thread waiting to put is notified then.
52 #当一个项目从队列中删除时，通知not_full;然后会通知等待放置的线程。
53 self.not_full = threading.Condition(self.mutex)
54
55 # Notify all_tasks_done whenever the number of unfinished tasks
56 # drops to zero; thread waiting to join() is notified to resume
57 #当未完成任务的数量降为零时，通知all_tasks_done;等待加入()的线程被通知恢复
58 self.all_tasks_done = threading.Condition(self.mutex)
59 self.unfinished_tasks = 0
60
61 def task_done(self):
62 '''Indicate that a formerly enqueued task is complete.
63 Used by Queue consumer threads. For each get() used to fetch a task,
64 a subsequent call to task_done() tells the queue that the processing
65 on the task is complete.
66 If a join() is currently blocking, it will resume when all items
67 have been processed (meaning that a task_done() call was received
68 for every item that had been put() into the queue).
69 Raises a ValueError if called more times than there were items
70 placed in the queue.
71 '''
72 #指示以前加入队列的任务已经完成。由队列使用者线程使用。对于用于获取任务的每个
73 # get()，对task_done()的后续调用将告诉队列任务的处理已经完成。如果一个join()
74 # 当前处于阻塞状态，那么当所有项都被处理完时(这意味着对于每个已将()放入队列的
75 # 项都接收了task_done()调用)，它将恢复。如果调用的次数超过了队列中放置的项的
76 # 次数，就会引发ValueError。
77 with self.all_tasks_done:
78 unfinished = self.unfinished_tasks - 1
79 if unfinished <= 0:
80 if unfinished < 0:
81 raise ValueError('task_done() called too many times')
82 self.all_tasks_done.notify_all()
83 self.unfinished_tasks = unfinished
84
85 def join(self):
86 '''Blocks until all items in the Queue have been gotten and processed.
87 The count of unfinished tasks goes up whenever an item is added to the
88 queue. The count goes down whenever a consumer thread calls task_done()
89 to indicate the item was retrieved and all work on it is complete.
90 When the count of unfinished tasks drops to zero, join() unblocks.
91 '''
92 #阻塞，直到获取和处理队列中的所有项。当一个项目被添加到队列中时，未完成任务的计数
93 # 就会上升。每当使用者线程调用task_done()时，计数就会下降，以指示检索了项目并完成
94 # 了对其的所有工作。当未完成任务的计数降为0时，join()将解块。
95 with self.all_tasks_done:
96 while self.unfinished_tasks:
97 self.all_tasks_done.wait()
98
99 def qsize(self):
100 '''Return the approximate size of the queue (not reliable!).'''
101 #返回队列的大致大小(不可靠!)
102 with self.mutex:
103 return self._qsize()
104
105 def empty(self):
106 '''Return True if the queue is empty, False otherwise (not reliable!).
107 This method is likely to be removed at some point. Use qsize() == 0
108 as a direct substitute, but be aware that either approach risks a race
109 condition where a queue can grow before the result of empty() or
110 qsize() can be used.
111 To create code that needs to wait for all queued tasks to be
112 completed, the preferred technique is to use the join() method.
113 '''
114 #如果队列为空，返回True，否则返回False(不可靠!)这种方法可能会在某个时候被删除。
115 # 使用qsize() == 0作为直接的替代，但是要注意，在使用empty()或qsize()的结果之前，
116 # 队列可能会增长，这可能会带来竞争条件的风险。要创建需要等待所有排队任务完成的代码，
117 # 首选技术是使用join()方法。
118 with self.mutex:
119 return not self._qsize()
120
121 def full(self):
122 '''Return True if the queue is full, False otherwise (not reliable!).
123 This method is likely to be removed at some point. Use qsize() >= n
124 as a direct substitute, but be aware that either approach risks a race
125 condition where a queue can shrink before the result of full() or
126 qsize() can be used.
127 '''
128 #如果队列满了，返回True，否则返回False(不可靠!)这种方法可能会在某个时候被删除。
129 # 使用qsize() >= n作为直接替代，但是要注意，在使用full()或qsize()的结果之前，
130 # 队列可能会收缩，这可能会导致竞争条件的风险。
131 with self.mutex:
132 return 0 < self.maxsize <= self._qsize()
133
134 def put(self, item, block=True, timeout=None):
135 '''Put an item into the queue.
136 If optional args 'block' is true and 'timeout' is None (the default),
137 block if necessary until a free slot is available. If 'timeout' is
138 a non-negative number, it blocks at most 'timeout' seconds and raises
139 the Full exception if no free slot was available within that time.
140 Otherwise ('block' is false), put an item on the queue if a free slot
141 is immediately available, else raise the Full exception ('timeout'
142 is ignored in that case).
143 '''
144 #将项目放入队列中。如果可选的args 'block'为true，而'timeout'为None(默认值)，
145 # 那么如果有必要，阻塞直到空闲的插槽可用为止。如果“timeout”是非负数，它最多会
146 # 阻塞“timeout”秒，如果在这段时间内没有可用的空闲时间，它就会引发完全异常。
147 # 否则(‘block’为false)，如果有空闲的插槽立即可用，就在队列中放置一个项目，
148 # 否则引发完整的异常(在这种情况下忽略‘timeout’)。
149 with self.not_full:
150 if self.maxsize > 0:
151 if not block:
152 if self._qsize() >= self.maxsize:
153 raise Full
154 elif timeout is None:
155 while self._qsize() >= self.maxsize:
156 self.not_full.wait()
157 elif timeout < 0:
158 raise ValueError("'timeout' must be a non-negative number")
159 else:
160 endtime = time() + timeout
161 while self._qsize() >= self.maxsize:
162 remaining = endtime - time()
163 if remaining <= 0.0:
164 raise Full
165 self.not_full.wait(remaining)
166 self._put(item)
167 self.unfinished_tasks += 1
168 self.not_empty.notify()
169
170 def get(self, block=True, timeout=None):
171 '''Remove and return an item from the queue.
172 If optional args 'block' is true and 'timeout' is None (the default),
173 block if necessary until an item is available. If 'timeout' is
174 a non-negative number, it blocks at most 'timeout' seconds and raises
175 the Empty exception if no item was available within that time.
176 Otherwise ('block' is false), return an item if one is immediately
177 available, else raise the Empty exception ('timeout' is ignored
178 in that case).
179 '''
180 #从队列中删除并返回项。如果可选的args 'block'为true，而'timeout'为None
181 # (默认值)，则在项可用之前，如果有必要，阻塞。如果“timeout”是非负数，它最多
182 # 会阻塞“timeout”秒，如果在这段时间内没有可用项，就会引发空异常。
183 # 否则(‘block’为false)，如果一个项立即可用，返回一个项，否则引发空异常
184 # (在这种情况下忽略'timeout')。
185 with self.not_empty:
186 if not block:
187 if not self._qsize():
188 raise Empty
189 elif timeout is None:
190 while not self._qsize():
191 self.not_empty.wait()
192 elif timeout < 0:
193 raise ValueError("'timeout' must be a non-negative number")
194 else:
195 endtime = time() + timeout
196 while not self._qsize():
197 remaining = endtime - time()
198 if remaining <= 0.0:
199 raise Empty
200 self.not_empty.wait(remaining)
201 item = self._get()
202 self.not_full.notify()
203 return item
204
205 def put_nowait(self, item):
206 '''Put an item into the queue without blocking.
207 Only enqueue the item if a free slot is immediately available.
208 Otherwise raise the Full exception.
209 '''
210 #将项目放入队列中而不阻塞。只有当一个空闲的插槽立即可用时，才将项目加入队列。否则引发完全异常。
211 return self.put(item, block=False)
212
213 def get_nowait(self):
214 '''Remove and return an item from the queue without blocking.
215 Only get an item if one is immediately available. Otherwise
216 raise the Empty exception.
217 '''
218 #在不阻塞的情况下从队列中删除并返回项。只有当一个项目是立即可用的。否则引发空异常。
219 return self.get(block=False)
220
221 # Override these methods to implement other queue organizations
222 # (e.g. stack or priority queue).
223 # These will only be called with appropriate locks held
224 #重写这些方法以实现其他队列组织(例如堆栈或优先队列)。只有在持有适当的锁时才会调用这些函数
225
226 # Initialize the queue representation
227 '''初始化队列表示'''
228 def _init(self, maxsize):
229 self.queue = deque()
230
231 def _qsize(self):
232 return len(self.queue)
233
234 # Put a new item in the queue
235 def _put(self, item):
236 self.queue.append(item)
237
238 # Get an item from the queue
239 def _get(self):
240 return self.queue.popleft()
241
242
243 class PriorityQueue(Queue):
244 '''Variant of Queue that retrieves open entries in priority order (lowest first).
245 Entries are typically tuples of the form: (priority number, data).
246 '''
247 #按优先级顺序(最低优先级)检索打开项的队列的变体。条目通常是表单的元组(优先级号、数据)。
248
249 def _init(self, maxsize):
250 self.queue = []
251
252 def _qsize(self):
253 return len(self.queue)
254
255 def _put(self, item):
256 heappush(self.queue, item)
257
258 def _get(self):
259 return heappop(self.queue)
260
261
262 class LifoQueue(Queue):
263 '''Variant of Queue that retrieves most recently added entries first.'''
264 #队列的变体，它首先检索最近添加的条目。
265
266 def _init(self, maxsize):
267 self.queue = []
268
269 def _qsize(self):
270 return len(self.queue)
271
272 def _put(self, item):
273 self.queue.append(item)
274
275 def _get(self):
276 return self.queue.pop()
277
278
279 class _PySimpleQueue:
280 '''Simple, unbounded FIFO queue.
281 This pure Python implementation is not reentrant.
282 '''
283 #简单、无界的FIFO队列。这个纯Python实现是不可重入的。
284
285 # Note: while this pure Python version provides fairness
286 # (by using a threading.Semaphore which is itself fair, being based
287 # on threading.Condition), fairness is not part of the API contract.
288 # This allows the C version to use a different implementation.
289 #注意:虽然这个纯Python版本提供了公平性(通过使用线程)。信号量本身是公平的，
290 # 基于thread . condition)，公平不是API契约的一部分。这允许C版本使用不同的实现。
291
292 def __init__(self):
293 self._queue = deque()
294 self._count = threading.Semaphore(0)
295
296 def put(self, item, block=True, timeout=None):
297 '''Put the item on the queue.
298 The optional 'block' and 'timeout' arguments are ignored, as this method
299 never blocks. They are provided for compatibility with the Queue class.
300 '''
301 #将项目放到队列中。可选的“block”和“timeout”参数被忽略，因为这个方法从不阻塞。
302 # 它们是为了与队列类兼容而提供的。
303 self._queue.append(item)
304 self._count.release()
305
306 def get(self, block=True, timeout=None):
307 '''Remove and return an item from the queue.
308 If optional args 'block' is true and 'timeout' is None (the default),
309 block if necessary until an item is available. If 'timeout' is
310 a non-negative number, it blocks at most 'timeout' seconds and raises
311 the Empty exception if no item was available within that time.
312 Otherwise ('block' is false), return an item if one is immediately
313 available, else raise the Empty exception ('timeout' is ignored
314 in that case).
315 '''
316 #从队列中删除并返回项。如果可选的args 'block'为true，而'timeout'为None
317 # (默认值)，则在项可用之前，如果有必要，阻塞。如果“timeout”是非负数，它最多
318 # 会阻塞“timeout”秒，如果在这段时间内没有可用项，就会引发空异常。否则
319 # (‘block’为false)，如果一个项立即可用，返回一个项，否则引发空异常
320 # (在这种情况下忽略'timeout')。
321 if timeout is not None and timeout < 0:
322 raise ValueError("'timeout' must be a non-negative number")
323 if not self._count.acquire(block, timeout):
324 raise Empty
325 return self._queue.popleft()
326
327 def put_nowait(self, item):
328 '''Put an item into the queue without blocking.
329 This is exactly equivalent to `put(item)` and is only provided
330 for compatibility with the Queue class.
331 '''
332 #将项目放入队列中而不阻塞。这完全等同于‘put(item)’，并且只提供与队列类的兼容性。
333 return self.put(item, block=False)
334
335 def get_nowait(self):
336 '''Remove and return an item from the queue without blocking.
337 Only get an item if one is immediately available. Otherwise
338 raise the Empty exception.
339 '''
340 #在不阻塞的情况下从队列中删除并返回项。只有当一个项目是立即可用的。否则引发空异常。
341 return self.get(block=False)
342
343 def empty(self):
344 '''Return True if the queue is empty, False otherwise (not reliable!).'''
345 #如果队列为空，返回True，否则返回False(不可靠!)
346 return len(self._queue) == 0
347
348 def qsize(self):
349 '''Return the approximate size of the queue (not reliable!).'''
350 #返回队列的大致大小(不可靠!)
351 return len(self._queue)
352
353
354 if SimpleQueue is None:
355 SimpleQueue = _PySimpleQueue
| Python进程 |
进程（multiprocessing）：
线程是进程最小的数据单元；每个进程都是相互独立的，它们之间不能共享数据。
启动单个进程：
1 '''启动一个进程'''
2 import multiprocessing
3 import time
4
5 def run(name):
6 time.sleep(2)
7 print("hello",name)
8
9 if __name__ == "__main__":
10 p = multiprocessing.Process(target=run,args=("pp",))
11 p.start()
12 p.join()
13
14
15 '''运行结果'''
16 hello pp
启动多个进程：
1 import multiprocessing
2 import time
3
4 def run(name):
5 time.sleep(2)
6 print("hello",name)
7
8 for i in range(10):
9 if __name__ == "__main__":
10 p = multiprocessing.Process(target=run, args=("pp",))
11 p.start()
12 p.join()
在进程中创建进程：
1 import multiprocessing
2 import threading
3 import time
4
5 def thread_run():
6 print(threading.get_ident()) #get_ident()获得线程地址
7
8 def run(name):
9 time.sleep(2)
10 print("hello",name)
11 t = threading.Thread(target=thread_run)
12 t.start()
13
14 if __name__ == "__main__":
15 p = multiprocessing.Process(target=run,args=("pp",))
16 p.start()
17 p.join()
18
19
20 '''运行结果'''
21 hello pp
22 2404
数据共享：
1 '''通过中间介质（pickle）使两个进程实现数据共享,实质上并不是完全的数据共享，
2 只是将子进程的对象（队列Queue）进行克隆'''
3 # import threading
4 # import queue
5 #
6 # def f():
7 # q.put("jfkdsljfkdls")
8 #
9 # if __name__=="__main__":
10 # q=queue.Queue()
11 # p=threading.Thread(target=f)
12 # p.start()
13 # print(q.get())
14 # p.join()
15
16 '''进程'''
17 from multiprocessing import Process,Queue
18 import queue
19
20 def f(q2):
21 q2.put("hkshhdjskajdksa")
22
23 if __name__=="__main__":
24 q=Queue()
25 p=Process(target=f,args=(q,))
26 p.start()
27 print(q.get())
28 p.join()
1 '''manager是用来传递对象'''
2 from multiprocessing import Process,Manager
3 import os
4
5 def f(d,l,l1):
6 d[os.getpid()] = os.getpid() #os.getpid():Return the current process id.
7 l.append(os.getpid())
8 l1.append(os.getpid()) #l1属于直接传递，不能回传
9 # print(l)
10 # print("l1:***",l1) #普通列表在子进程中每次会获得一个新值，但都会被下一个值覆盖
11 # print(d)
12
13 if __name__ == "__main__":
14 with Manager() as mager:
15 d = mager.dict() #由manager生成的字典
16 l = mager.list(range(5)) #由manager生成的列表
17 l1 = [] #普通列表无法共享数据，最后仍旧是空列表
18 p_list = []
19 for i in range(10):
20 p = Process(target=f,args=(d,l,l1))
21 p.start()
22 p.join()
23 print("l:",l)
24 print(l1)
25 print("d:",d)
数据传递：
1 '''主进程(父类进程) 子进程'''
2 '''管道通信实现数据之间的传递'''
3 # from multiprocessing import Process,Pipe
4 #
5 # def f(con):
6 # con.send("hello from child1")
7 # con.send("hello from child2")
8 # print("parent news:",con.recv())
9 # con.close()
10 #
11 # if __name__ == "__main__":
12 # Parent_con,Child_con = Pipe()
13 # p = Process(target=f,args=(Child_con,))
14 # p.start()
15 # print(Parent_con.recv())
16 # print(Parent_con.recv())
17 # Parent_con.send("from parent")
18 # p.join()
19
20
21
22 '''运行结果'''
23 hello from child1
24 hello from child2
25 parent news: from parent
进程锁（Lock）：
屏幕存在共享，多进程可以同时使用屏幕，进程加锁的目的在于，确保屏幕被独个进程使用。
from multiprocessing import Process,Lock
def f(l,i):
l.acquire()
print("+++",i)
l.release()
if __name__ == "__main__":
lock = Lock() #加锁的目的是为了确保屏幕被单独占用
for num in range(100):
Process(target=f,args=(lock,num)).start()
'''运行结果'''
+++ 0
+++ 1
+++ 2
+++ 3
+++ 4
+++ 5
.
.
.(不在这里演示完所有内容)
进程池（pool）：
python中，进程池内部会维护一个进程序列。当需要时，程序会去进程池中获取一个进程。
如果进程池序列中没有可供使用的进程，那么程序就会等待，直到进程池中有可用进程为止。
from multiprocessing import Process,Pool
import time
import os
def foo(i):
time.sleep(2)
print("in the process:",os.getpid())
return i+100
def bar(args):
print("system done",args)
if __name__ == "__main__":
pool = Pool(5)
for i in range(10):
# pool.apply(func=foo,args=(i,))
#生成进程，把pool放入容器
#apply本身是一个串行方法,不受join影响
pool.apply_async(func=foo,args=(i,),callback=bar)
#apply_async是一个并行方法,受join影响
#callback()回调函数为主进程操作，进程池一旦开始运行，回调函数会自动执行
print("end")
pool.close() #pool关闭是需要时间的，所以在close之后再join
pool.join()
pool的内置方法：
- apply 串行方法。从进程池里取一个进程并同步执行，不受join影响
- apply_async 并行方法。从进程池里取出一个进程并异步执行，受join影响
- terminate 立刻关闭进程池
- join 主进程等待所有子进程执行完毕，必须在close或terminete之后（如上述代码）
- close 等待所有进程结束才关闭线程池
1 #
2 # Module providing the `Pool` class for managing a process pool
3 #模块提供用于管理进程池的“池”类
4 # multiprocessing/pool.py
5 #
6 # Copyright (c) 2006-2008, R Oudkerk
7 # Licensed to PSF under a Contributor Agreement.
8 #
9
10 __all__ = ['Pool', 'ThreadPool']
11
12 #
13 # Imports
14 #
15
16 import threading
17 import queue
18 import itertools
19 import collections
20 import os
21 import time
22 import traceback
23
24 # If threading is available then ThreadPool should be provided. Therefore
25 # we avoid top-level imports which are liable to fail on some systems.
26 from . import util
27 from . import get_context, TimeoutError
28
29 #
30 # Constants representing the state of a pool
31 #表示池状态的常数
32
33 RUN = 0
34 CLOSE = 1
35 TERMINATE = 2
36
37 #
38 # Miscellaneous
39 #
40
41 job_counter = itertools.count()
42
43 def mapstar(args):
44 return list(map(*args))
45
46 def starmapstar(args):
47 return list(itertools.starmap(args[0], args[1]))
48
49 #
50 # Hack to embed stringification of remote traceback in local traceback
51 #
52
53 class RemoteTraceback(Exception):
54 def __init__(self, tb):
55 self.tb = tb
56 def __str__(self):
57 return self.tb
58
59 class ExceptionWithTraceback:
60 def __init__(self, exc, tb):
61 tb = traceback.format_exception(type(exc), exc, tb)
62 tb = ''.join(tb)
63 self.exc = exc
64 self.tb = '\n"""\n%s"""' % tb
65 def __reduce__(self):
66 return rebuild_exc, (self.exc, self.tb)
67
68 def rebuild_exc(exc, tb):
69 exc.__cause__ = RemoteTraceback(tb)
70 return exc
71
72 #
73 # Code run by worker processes
74 #
75
76 class MaybeEncodingError(Exception):
77 """Wraps possible unpickleable errors, so they can be
78 safely sent through the socket."""
79 #包装可能出现的无法拾取的错误，以便通过套接字安全地发送这些错误。
80
81 def __init__(self, exc, value):
82 self.exc = repr(exc)
83 self.value = repr(value)
84 super(MaybeEncodingError, self).__init__(self.exc, self.value)
85
86 def __str__(self):
87 return "Error sending result: '%s'. Reason: '%s'" % (self.value,
88 self.exc)
89
90 def __repr__(self):
91 return "<%s: %s>" % (self.__class__.__name__, self)
92
93
94 def worker(inqueue, outqueue, initializer=None, initargs=(), maxtasks=None,
95 wrap_exception=False):
96 if (maxtasks is not None) and not (isinstance(maxtasks, int)
97 and maxtasks >= 1):
98 raise AssertionError("Maxtasks {!r} is not valid".format(maxtasks))
99 put = outqueue.put
100 get = inqueue.get
101 if hasattr(inqueue, '_writer'):
102 inqueue._writer.close()
103 outqueue._reader.close()
104
105 if initializer is not None:
106 initializer(*initargs)
107
108 completed = 0
109 while maxtasks is None or (maxtasks and completed < maxtasks):
110 try:
111 task = get()
112 except (EOFError, OSError):
113 util.debug('worker got EOFError or OSError -- exiting')
114 break
115
116 if task is None:
117 util.debug('worker got sentinel -- exiting')
118 break
119
120 job, i, func, args, kwds = task
121 try:
122 result = (True, func(*args, **kwds))
123 except Exception as e:
124 if wrap_exception and func is not _helper_reraises_exception:
125 e = ExceptionWithTraceback(e, e.__traceback__)
126 result = (False, e)
127 try:
128 put((job, i, result))
129 except Exception as e:
130 wrapped = MaybeEncodingError(e, result[1])
131 util.debug("Possible encoding error while sending result: %s" % (
132 wrapped))
133 put((job, i, (False, wrapped)))
134
135 task = job = result = func = args = kwds = None
136 completed += 1
137 util.debug('worker exiting after %d tasks' % completed)
138
139 def _helper_reraises_exception(ex):
140 'Pickle-able helper function for use by _guarded_task_generation.'
141 #用于_guarded_task_generation的可选择助手函数。
142 raise ex
143
144 #
145 # Class representing a process pool 类表示进程池
146 #
147
148 class Pool(object):
149 '''
150 Class which supports an async version of applying functions to arguments.
151 '''
152 #类，该类支持将函数应用于参数的异步版本。
153 _wrap_exception = True
154
155 def Process(self, *args, **kwds):
156 return self._ctx.Process(*args, **kwds)
157
158 def __init__(self, processes=None, initializer=None, initargs=(),
159 maxtasksperchild=None, context=None):
160 self._ctx = context or get_context()
161 self._setup_queues()
162 self._taskqueue = queue.SimpleQueue()
163 self._cache = {}
164 self._state = RUN
165 self._maxtasksperchild = maxtasksperchild
166 self._initializer = initializer
167 self._initargs = initargs
168
169 if processes is None:
170 processes = os.cpu_count() or 1
171 if processes < 1:
172 raise ValueError("Number of processes must be at least 1")
173
174 if initializer is not None and not callable(initializer):
175 raise TypeError('initializer must be a callable')
176
177 self._processes = processes
178 self._pool = []
179 self._repopulate_pool()
180
181 self._worker_handler = threading.Thread(
182 target=Pool._handle_workers,
183 args=(self, )
184 )
185 self._worker_handler.daemon = True
186 self._worker_handler._state = RUN
187 self._worker_handler.start()
188
189
190 self._task_handler = threading.Thread(
191 target=Pool._handle_tasks,
192 args=(self._taskqueue, self._quick_put, self._outqueue,
193 self._pool, self._cache)
194 )
195 self._task_handler.daemon = True
196 self._task_handler._state = RUN
197 self._task_handler.start()
198
199 self._result_handler = threading.Thread(
200 target=Pool._handle_results,
201 args=(self._outqueue, self._quick_get, self._cache)
202 )
203 self._result_handler.daemon = True
204 self._result_handler._state = RUN
205 self._result_handler.start()
206
207 self._terminate = util.Finalize(
208 self, self._terminate_pool,
209 args=(self._taskqueue, self._inqueue, self._outqueue, self._pool,
210 self._worker_handler, self._task_handler,
211 self._result_handler, self._cache),
212 exitpriority=15
213 )
214
215 def _join_exited_workers(self):
216 """Cleanup after any worker processes which have exited due to reaching
217 their specified lifetime. Returns True if any workers were cleaned up.
218 """
219 #在由于达到指定的生存期而退出的任何工作进程之后进行清理。如果有工人被清理干净，
220 # 返回True。
221 cleaned = False
222 for i in reversed(range(len(self._pool))):
223 worker = self._pool[i]
224 if worker.exitcode is not None:
225 # worker exited
226 util.debug('cleaning up worker %d' % i)
227 worker.join()
228 cleaned = True
229 del self._pool[i]
230 return cleaned
231
232 def _repopulate_pool(self):
233 """Bring the number of pool processes up to the specified number,
234 for use after reaping workers which have exited.
235 """
236 #将池进程的数量增加到指定的数量，以便在收割已退出的工人后使用。
237 for i in range(self._processes - len(self._pool)):
238 w = self.Process(target=worker,
239 args=(self._inqueue, self._outqueue,
240 self._initializer,
241 self._initargs, self._maxtasksperchild,
242 self._wrap_exception)
243 )
244 self._pool.append(w)
245 w.name = w.name.replace('Process', 'PoolWorker')
246 w.daemon = True
247 w.start()
248 util.debug('added worker')
249
250 def _maintain_pool(self):
251 """Clean up any exited workers and start replacements for them.
252 """
253 #清理所有离职的员工，并开始替换他们。
254 if self._join_exited_workers():
255 self._repopulate_pool()
256
257 def _setup_queues(self):
258 self._inqueue = self._ctx.SimpleQueue()
259 self._outqueue = self._ctx.SimpleQueue()
260 self._quick_put = self._inqueue._writer.send
261 self._quick_get = self._outqueue._reader.recv
262
263 def apply(self, func, args=(), kwds={}):
264 '''
265 Equivalent of `func(*args, **kwds)`.
266 Pool must be running.
267 '''
268 #相当于“func(*args， ** kwds)”。池必须正在运行。
269 return self.apply_async(func, args, kwds).get()
270
271 def map(self, func, iterable, chunksize=None):
272 '''
273 Apply `func` to each element in `iterable`, collecting the results
274 in a list that is returned.
275 '''
276 #对“iterable”中的每个元素应用“func”，在返回的列表中收集结果。
277 return self._map_async(func, iterable, mapstar, chunksize).get()
278
279 def starmap(self, func, iterable, chunksize=None):
280 '''
281 Like `map()` method but the elements of the `iterable` are expected to
282 be iterables as well and will be unpacked as arguments. Hence
283 `func` and (a, b) becomes func(a, b).
284 '''
285 #方法类似于“map()”，但“iterable”的元素也应该是可迭代的，并将作为参数解压缩。
286 # 因此“func”和(a, b)变成了func(a, b)。
287 return self._map_async(func, iterable, starmapstar, chunksize).get()
288
289 def starmap_async(self, func, iterable, chunksize=None, callback=None,
290 error_callback=None):
291 '''
292 Asynchronous version of `starmap()` method.
293 '''
294 #异步版本的“starmap()”方法。
295 return self._map_async(func, iterable, starmapstar, chunksize,
296 callback, error_callback)
297
298 def _guarded_task_generation(self, result_job, func, iterable):
299 '''Provides a generator of tasks for imap and imap_unordered with
300 appropriate handling for iterables which throw exceptions during
301 iteration.'''
302 #为imap和imap_unordered提供任务生成器，并为迭代期间抛出异常的迭代提供适当的处理。
303 try:
304 i = -1
305 for i, x in enumerate(iterable):
306 yield (result_job, i, func, (x,), {})
307 except Exception as e:
308 yield (result_job, i+1, _helper_reraises_exception, (e,), {})
309
310 def imap(self, func, iterable, chunksize=1):
311 '''
312 Equivalent of `map()` -- can be MUCH slower than `Pool.map()`.
313 '''
314 #等价于“map()”——可能比“Pool.map()”慢得多。
315 if self._state != RUN:
316 raise ValueError("Pool not running")
317 if chunksize == 1:
318 result = IMapIterator(self._cache)
319 self._taskqueue.put(
320 (
321 self._guarded_task_generation(result._job, func, iterable),
322 result._set_length
323 ))
324 return result
325 else:
326 if chunksize < 1:
327 raise ValueError(
328 "Chunksize must be 1+, not {0:n}".format(
329 chunksize))
330 task_batches = Pool._get_tasks(func, iterable, chunksize)
331 result = IMapIterator(self._cache)
332 self._taskqueue.put(
333 (
334 self._guarded_task_generation(result._job,
335 mapstar,
336 task_batches),
337 result._set_length
338 ))
339 return (item for chunk in result for item in chunk)
340
341 def imap_unordered(self, func, iterable, chunksize=1):
342 '''
343 Like `imap()` method but ordering of results is arbitrary.
344 '''
345 #Like `imap()`方法，但结果的顺序是任意的。
346 if self._state != RUN:
347 raise ValueError("Pool not running")
348 if chunksize == 1:
349 result = IMapUnorderedIterator(self._cache)
350 self._taskqueue.put(
351 (
352 self._guarded_task_generation(result._job, func, iterable),
353 result._set_length
354 ))
355 return result
356 else:
357 if chunksize < 1:
358 raise ValueError(
359 "Chunksize must be 1+, not {0!r}".format(chunksize))
360 task_batches = Pool._get_tasks(func, iterable, chunksize)
361 result = IMapUnorderedIterator(self._cache)
362 self._taskqueue.put(
363 (
364 self._guarded_task_generation(result._job,
365 mapstar,
366 task_batches),
367 result._set_length
368 ))
369 return (item for chunk in result for item in chunk)
370
371 def apply_async(self, func, args=(), kwds={}, callback=None,
372 error_callback=None):
373 '''
374 Asynchronous version of `apply()` method. “apply()”方法的异步版本。
375 '''
376 if self._state != RUN:
377 raise ValueError("Pool not running")
378 result = ApplyResult(self._cache, callback, error_callback)
379 self._taskqueue.put(([(result._job, 0, func, args, kwds)], None))
380 return result
381
382 def map_async(self, func, iterable, chunksize=None, callback=None,
383 error_callback=None):
384 '''
385 Asynchronous version of `map()` method. 方法的异步版本
386 '''
387 return self._map_async(func, iterable, mapstar, chunksize, callback,
388 error_callback)
389
390 def _map_async(self, func, iterable, mapper, chunksize=None, callback=None,
391 error_callback=None):
392 '''
393 Helper function to implement map, starmap and their async counterparts.
394 '''
395 #帮助函数实现映射，星图和他们的异步对等。
396 if self._state != RUN:
397 raise ValueError("Pool not running")
398 if not hasattr(iterable, '__len__'):
399 iterable = list(iterable)
400
401 if chunksize is None:
402 chunksize, extra = divmod(len(iterable), len(self._pool) * 4)
403 if extra:
404 chunksize += 1
405 if len(iterable) == 0:
406 chunksize = 0
407
408 task_batches = Pool._get_tasks(func, iterable, chunksize)
409 result = MapResult(self._cache, chunksize, len(iterable), callback,
410 error_callback=error_callback)
411 self._taskqueue.put(
412 (
413 self._guarded_task_generation(result._job,
414 mapper,
415 task_batches),
416 None
417 )
418 )
419 return result
420
421 @staticmethod
422 def _handle_workers(pool):
423 thread = threading.current_thread()
424
425 # Keep maintaining workers until the cache gets drained, unless the pool
426 # is terminated.
427 #继续维护worker，直到缓存耗尽，除非池终止。
428 while thread._state == RUN or (pool._cache and thread._state != TERMINATE):
429 pool._maintain_pool()
430 time.sleep(0.1)
431 # send sentinel to stop workers
432 pool._taskqueue.put(None)
433 util.debug('worker handler exiting')
434
435 @staticmethod
436 def _handle_tasks(taskqueue, put, outqueue, pool, cache):
437 thread = threading.current_thread()
438
439 for taskseq, set_length in iter(taskqueue.get, None):
440 task = None
441 try:
442 # iterating taskseq cannot fail
443 #迭代taskseq不会失败
444 for task in taskseq:
445 if thread._state:
446 util.debug('task handler found thread._state != RUN')
447 break
448 try:
449 put(task)
450 except Exception as e:
451 job, idx = task[:2]
452 try:
453 cache[job]._set(idx, (False, e))
454 except KeyError:
455 pass
456 else:
457 if set_length:
458 util.debug('doing set_length()')
459 idx = task[1] if task else -1
460 set_length(idx + 1)
461 continue
462 break
463 finally:
464 task = taskseq = job = None
465 else:
466 util.debug('task handler got sentinel')
467
468 try:
469 # tell result handler to finish when cache is empty
470 #告诉结果处理程序在缓存为空时结束
471 util.debug('task handler sending sentinel to result handler')
472 outqueue.put(None)
473
474 # tell workers there is no more work
475 util.debug('task handler sending sentinel to workers')
476 for p in pool:
477 put(None)
478 except OSError:
479 util.debug('task handler got OSError when sending sentinels')
480
481 util.debug('task handler exiting')
482
483 @staticmethod
484 def _handle_results(outqueue, get, cache):
485 thread = threading.current_thread()
486
487 while 1:
488 try:
489 task = get()
490 except (OSError, EOFError):
491 util.debug('result handler got EOFError/OSError -- exiting')
492 return
493
494 if thread._state:
495 assert thread._state == TERMINATE, "Thread not in TERMINATE"
496 util.debug('result handler found thread._state=TERMINATE')
497 break
498
499 if task is None:
500 util.debug('result handler got sentinel')
501 break
502
503 job, i, obj = task
504 try:
505 cache[job]._set(i, obj)
506 except KeyError:
507 pass
508 task = job = obj = None
509
510 while cache and thread._state != TERMINATE:
511 try:
512 task = get()
513 except (OSError, EOFError):
514 util.debug('result handler got EOFError/OSError -- exiting')
515 return
516
517 if task is None:
518 util.debug('result handler ignoring extra sentinel')
519 continue
520 job, i, obj = task
521 try:
522 cache[job]._set(i, obj)
523 except KeyError:
524 pass
525 task = job = obj = None
526
527 if hasattr(outqueue, '_reader'):
528 util.debug('ensuring that outqueue is not full')
529 # If we don't make room available in outqueue then
530 # attempts to add the sentinel (None) to outqueue may
531 # block. There is guaranteed to be no more than 2 sentinels.
532 #如果我们不在outqueue中留出可用的空间，那么尝试将sentinel (None)
533 # 添加到outqueue可能会阻塞。保证不超过2个哨兵。
534 try:
535 for i in range(10):
536 if not outqueue._reader.poll():
537 break
538 get()
539 except (OSError, EOFError):
540 pass
541
542 util.debug('result handler exiting: len(cache)=%s, thread._state=%s',
543 len(cache), thread._state)
544
545 @staticmethod
546 def _get_tasks(func, it, size):
547 it = iter(it)
548 while 1:
549 x = tuple(itertools.islice(it, size))
550 if not x:
551 return
552 yield (func, x)
553
554 def __reduce__(self):
555 raise NotImplementedError(
556 'pool objects cannot be passed between processes or pickled'
557 #不能在进程之间传递池对象或pickle池对象
558 )
559
560 def close(self):
561 util.debug('closing pool')
562 if self._state == RUN:
563 self._state = CLOSE
564 self._worker_handler._state = CLOSE
565
566 def terminate(self):
567 util.debug('terminating pool')
568 self._state = TERMINATE
569 self._worker_handler._state = TERMINATE
570 self._terminate()
571
572 def join(self):
573 util.debug('joining pool')
574 if self._state == RUN:
575 raise ValueError("Pool is still running")
576 elif self._state not in (CLOSE, TERMINATE):
577 raise ValueError("In unknown state")
578 self._worker_handler.join()
579 self._task_handler.join()
580 self._result_handler.join()
581 for p in self._pool:
582 p.join()
583
584 @staticmethod
585 def _help_stuff_finish(inqueue, task_handler, size):
586 # task_handler may be blocked trying to put items on inqueue
587 #试图将项放入inqueue时可能阻塞task_handler
588 util.debug('removing tasks from inqueue until task handler finished')
589 inqueue._rlock.acquire()
590 while task_handler.is_alive() and inqueue._reader.poll():
591 inqueue._reader.recv()
592 time.sleep(0)
593
594 @classmethod
595 def _terminate_pool(cls, taskqueue, inqueue, outqueue, pool,
596 worker_handler, task_handler, result_handler, cache):
597 # this is guaranteed to only be called once 这保证只调用一次
598 util.debug('finalizing pool')
599
600 worker_handler._state = TERMINATE
601 task_handler._state = TERMINATE
602
603 util.debug('helping task handler/workers to finish')
604 cls._help_stuff_finish(inqueue, task_handler, len(pool))
605
606 if (not result_handler.is_alive()) and (len(cache) != 0):
607 raise AssertionError(
608 "Cannot have cache with result_hander not alive")
609
610 result_handler._state = TERMINATE
611 outqueue.put(None) # sentinel
612
613 # We must wait for the worker handler to exit before terminating
614 # workers because we don't want workers to be restarted behind our back.
615 #我们必须在终止工人之前等待工人处理程序退出，因为我们不希望工人在我们背后重新启动。
616 util.debug('joining worker handler')
617 if threading.current_thread() is not worker_handler:
618 worker_handler.join()
619
620 # Terminate workers which haven't already finished.
621 if pool and hasattr(pool[0], 'terminate'):
622 util.debug('terminating workers')
623 for p in pool:
624 if p.exitcode is None:
625 p.terminate()
626
627 util.debug('joining task handler')
628 if threading.current_thread() is not task_handler:
629 task_handler.join()
630
631 util.debug('joining result handler')
632 if threading.current_thread() is not result_handler:
633 result_handler.join()
634
635 if pool and hasattr(pool[0], 'terminate'):
636 util.debug('joining pool workers')
637 for p in pool:
638 if p.is_alive():
639 # worker has not yet exited
640 util.debug('cleaning up worker %d' % p.pid)
641 p.join()
642
643 def __enter__(self):
644 return self
645
646 def __exit__(self, exc_type, exc_val, exc_tb):
647 self.terminate()
648
649 #
650 # Class whose instances are returned by `Pool.apply_async()`
651 #
652
653 class ApplyResult(object):
654
655 def __init__(self, cache, callback, error_callback):
656 self._event = threading.Event()
657 self._job = next(job_counter)
658 self._cache = cache
659 self._callback = callback
660 self._error_callback = error_callback
661 cache[self._job] = self
662
663 def ready(self):
664 return self._event.is_set()
665
666 def successful(self):
667 if not self.ready():
668 raise ValueError("{0!r} not ready".format(self))
669 return self._success
670
671 def wait(self, timeout=None):
672 self._event.wait(timeout)
673
674 def get(self, timeout=None):
675 self.wait(timeout)
676 if not self.ready():
677 raise TimeoutError
678 if self._success:
679 return self._value
680 else:
681 raise self._value
682
683 def _set(self, i, obj):
684 self._success, self._value = obj
685 if self._callback and self._success:
686 self._callback(self._value)
687 if self._error_callback and not self._success:
688 self._error_callback(self._value)
689 self._event.set()
690 del self._cache[self._job]
691
692 AsyncResult = ApplyResult # create alias -- see #17805
693
694 #
695 # Class whose instances are returned by `Pool.map_async()`
696 #
697
698 class MapResult(ApplyResult):
699
700 def __init__(self, cache, chunksize, length, callback, error_callback):
701 ApplyResult.__init__(self, cache, callback,
702 error_callback=error_callback)
703 self._success = True
704 self._value = [None] * length
705 self._chunksize = chunksize
706 if chunksize <= 0:
707 self._number_left = 0
708 self._event.set()
709 del cache[self._job]
710 else:
711 self._number_left = length//chunksize + bool(length % chunksize)
712
713 def _set(self, i, success_result):
714 self._number_left -= 1
715 success, result = success_result
716 if success and self._success:
717 self._value[i*self._chunksize:(i+1)*self._chunksize] = result
718 if self._number_left == 0:
719 if self._callback:
720 self._callback(self._value)
721 del self._cache[self._job]
722 self._event.set()
723 else:
724 if not success and self._success:
725 # only store first exception
726 self._success = False
727 self._value = result
728 if self._number_left == 0:
729 # only consider the result ready once all jobs are done
730 if self._error_callback:
731 self._error_callback(self._value)
732 del self._cache[self._job]
733 self._event.set()
734
735 #
736 # Class whose instances are returned by `Pool.imap()`
737 #
738
739 class IMapIterator(object):
740
741 def __init__(self, cache):
742 self._cond = threading.Condition(threading.Lock())
743 self._job = next(job_counter)
744 self._cache = cache
745 self._items = collections.deque()
746 self._index = 0
747 self._length = None
748 self._unsorted = {}
749 cache[self._job] = self
750
751 def __iter__(self):
752 return self
753
754 def next(self, timeout=None):
755 with self._cond:
756 try:
757 item = self._items.popleft()
758 except IndexError:
759 if self._index == self._length:
760 raise StopIteration from None
761 self._cond.wait(timeout)
762 try:
763 item = self._items.popleft()
764 except IndexError:
765 if self._index == self._length:
766 raise StopIteration from None
767 raise TimeoutError from None
768
769 success, value = item
770 if success:
771 return value
772 raise value
773
774 __next__ = next # XXX
775
776 def _set(self, i, obj):
777 with self._cond:
778 if self._index == i:
779 self._items.append(obj)
780 self._index += 1
781 while self._index in self._unsorted:
782 obj = self._unsorted.pop(self._index)
783 self._items.append(obj)
784 self._index += 1
785 self._cond.notify()
786 else:
787 self._unsorted[i] = obj
788
789 if self._index == self._length:
790 del self._cache[self._job]
791
792 def _set_length(self, length):
793 with self._cond:
794 self._length = length
795 if self._index == self._length:
796 self._cond.notify()
797 del self._cache[self._job]
798
799 #
800 # Class whose instances are returned by `Pool.imap_unordered()`
801 #类，其实例由' Pool.imap_unordered() '返回
802
803 class IMapUnorderedIterator(IMapIterator):
804
805 def _set(self, i, obj):
806 with self._cond:
807 self._items.append(obj)
808 self._index += 1
809 self._cond.notify()
810 if self._index == self._length:
811 del self._cache[self._job]
812
813 #
814 #
815 #
816
817 class ThreadPool(Pool):
818 _wrap_exception = False
819
820 @staticmethod
821 def Process(*args, **kwds):
822 from .dummy import Process
823 return Process(*args, **kwds)
824
825 def __init__(self, processes=None, initializer=None, initargs=()):
826 Pool.__init__(self, processes, initializer, initargs)
827
828 def _setup_queues(self):
829 self._inqueue = queue.SimpleQueue()
830 self._outqueue = queue.SimpleQueue()
831 self._quick_put = self._inqueue.put
832 self._quick_get = self._outqueue.get
833
834 @staticmethod
835 def _help_stuff_finish(inqueue, task_handler, size):
836 # drain inqueue, and put sentinels at its head to make workers finish
837 #排干队伍内的水，并在其头部放置哨兵，使工人完成工作
838 try:
839 while True:
840 inqueue.get(block=False)
841 except queue.Empty:
842 pass
843 for i in range(size):
844 inqueue.put(None)