先看个例子:

import time
from concurrent.futures import ThreadPoolExecutor


def foo():
    print('enter at {} ...'.format(time.strftime('%X')))
    time.sleep(5)
    print('exit  at {} ...'.format(time.strftime('%X')))


executor = ThreadPoolExecutor()
executor.submit(foo)
executor.shutdown()

执行结果:

enter at 16:20:31 ...
exit  at 16:20:36 ...

shutdown(wait=True) 方法默认阻塞当前线程,等待子线程执行完毕。即使 shutdown(wait=Fasle)也只是非阻塞的关闭线程池,线程池中正在执行任务的子线程并不会被马上停止,而是会继续执行直到执行完毕。尝试在源码中给新开启的子线程调用t.join(0)来立马强制停止子线程t,也不行,到底是什么原因保证了线程池中的线程在关闭线程池时,线程池中正在执行任务的子线程们不会被关闭呢?

看一下ThreadPoolExecutor源码:

class ThreadPoolExecutor(_base.Executor):
    def __init__(self, max_workers=None, thread_name_prefix=''):
        """Initializes a new ThreadPoolExecutor instance.

        Args:
            max_workers: The maximum number of threads that can be used to
                execute the given calls.
            thread_name_prefix: An optional name prefix to give our threads.
        """
        if max_workers is None:
            # Use this number because ThreadPoolExecutor is often
            # used to overlap I/O instead of CPU work.
            max_workers = (os.cpu_count() or 1) * 5
        if max_workers <= 0:
            raise ValueError("max_workers must be greater than 0")

        self._max_workers = max_workers
        self._work_queue = queue.Queue()
        self._threads = set()
        self._shutdown = False
        self._shutdown_lock = threading.Lock()
        self._thread_name_prefix = thread_name_prefix

    def submit(self, fn, *args, **kwargs):
        with self._shutdown_lock:
            if self._shutdown:
                raise RuntimeError('cannot schedule new futures after shutdown')

            f = _base.Future()
            # 把目标函数f包装成worker对象,执行worker.run()会调用f()
            w = _WorkItem(f, fn, args, kwargs)

            # 把worker对象放入到队列中
            self._work_queue.put(w)
            # 开启一个新的线程不断的从queue中获取worker对象,获取到则调用worker.run()
            self._adjust_thread_count()
            return f
    submit.__doc__ = _base.Executor.submit.__doc__

    def _adjust_thread_count(self):
        # 当执行del executor时,这个回调方法会被调用,也就是说当executor对象被垃圾回收时调用
        def weakref_cb(_, q=self._work_queue):
            q.put(None)

        num_threads = len(self._threads)
        if num_threads < self._max_workers:
            thread_name = '%s_%d' % (self._thread_name_prefix or self,
                                     num_threads)
            # 把_worker函数作为新线程的执行函数
            t = threading.Thread(name=thread_name, target=_worker,
                                 args=(weakref.ref(self, weakref_cb),
                                       self._work_queue))
            t.daemon = True
            t.start()
            self._threads.add(t)
            # 这一步很重要,是确保该线程t不被t.join(0)强制中断的关键。具体查看_python_exit函数
            _threads_queues[t] = self._work_queue

    def shutdown(self, wait=True):
        with self._shutdown_lock:
            self._shutdown = True
            self._work_queue.put(None)
        if wait:
            for t in self._threads:
                t.join()
    shutdown.__doc__ = _base.Executor.shutdown.__doc__

submit(func) 干了两件事:

  • 把worker放入queue中
  • 开启一个新线程不断从queue中取出woker,执行woker.run(),即执行func()

_adjust_thread_count()干了两件事:

  • 开启一个新线程执行_worker函数,这个函数的作用就是不断去queue中取出worker, 执行woker.run(),即执行func()

  • 把新线程跟队列queue绑定,防止线程被join(0)强制中断。

来看一下_worker函数源码:

def _worker(executor_reference, work_queue):
    try:
        while True:
            # 不断从queue中取出worker对象
            work_item = work_queue.get(block=True)
            if work_item is not None:
                # 执行func()
                work_item.run()
                # Delete references to object. See issue16284
                del work_item
                continue
            # 从弱引用对象中返回executor
            executor = executor_reference()
            # Exit if:
            #   - The interpreter is shutting down OR
            #   - The executor that owns the worker has been collected OR
            #   - The executor that owns the worker has been shutdown.

            # 当executor执行shutdown()方法时executor._shutdown为True,同时会放入None到队列,
            # 当work_item.run()执行完毕时,又会进入到下一轮循环从queue中获取worker对象,但是
            # 由于shutdown()放入了None到queue,因此取出的对象是None,从而判断这里的if条件分支,
            # 发现executor._shutdown是True,又放入一个None到queue中,是来通知其他线程跳出while循环的
            # shutdown()中的添加None到队列是用来结束线程池中的某一个线程的,这个if分支中的添加None
            # 队列是用来通知其他线程中的某一个线程结束的,这样连锁反应使得所有线程执行完func中的逻辑后都会结束
            if _shutdown or executor is None or executor._shutdown:
                # Notice other workers
                work_queue.put(None)
                return
            del executor
    except BaseException:
        _base.LOGGER.critical('Exception in worker', exc_info=True)

可以看出,这个 _worker方法的作用就是在新新线程中不断获得queue中的worker对象,执行worker.run()方法,执行完毕后通过放入None到queue队列的方式来通知其他线程结束。

再来看看_adjust_thread_count()方法中的_threads_queues[t] = self._work_queue这个操作是如何实现防止join(0)的操作强制停止正在执行的线程的。

import atexit


_threads_queues = weakref.WeakKeyDictionary()
_shutdown = False

def _python_exit():
    global _shutdown
    _shutdown = True
    items = list(_threads_queues.items())
    for t, q in items:
        q.put(None)
    # 取出_threads_queues中的线程t,执行t.join()强制等待子线程完成
    for t, q in items:
        t.join()

atexit.register(_python_exit)

这个atexit模块的作用是用来注册一个函数,当MainThread中的逻辑执行完毕时,会执行注册的这个_python_exit函数。然后执行_python_exit中的逻辑,也就是说t.join()会被执行,强制阻塞。这里好奇,既然是在MainThread结束后执行,那这个t.join()是在什么线程中被执行的呢。其实是一个叫_DummyThread线程的虚拟线程中执行的。

import atexit
import threading
import weakref
import time

threads_queues = weakref.WeakKeyDictionary()

def foo():
    print('enter at {} ...'.format(time.strftime('%X')))
    time.sleep(5)
    print('exit  at {} ...'.format(time.strftime('%X')))

def _python_exit():
    items = list(threads_queues.items())
    print('current thread in _python_exit --> ', threading.current_thread())
    for t, _ in items:
        t.join()

atexit.register(_python_exit)

if __name__ == '__main__':

    t = threading.Thread(target=foo)
    t.setDaemon(True)
    t.start()

    threads_queues[t] = foo

    print(time.strftime('%X'))
    t.join(timeout=2)
    print(time.strftime('%X'))
    t.join(timeout=2)
    print(time.strftime('%X'))
    print('current thread in main -->', threading.current_thread())
    print(threading.current_thread(), 'end')

执行结果:

enter at 17:13:44 ...
17:13:44
17:13:46
17:13:48
current thread in main --> <_MainThread(MainThread, started 12688)>
<_MainThread(MainThread, started 12688)> end
current thread in _python_exit -->  <_DummyThread(Dummy-2, started daemon 12688)>
exit  at 17:13:49 ...

从这个例子可以看到,当线程t开启时foo函数阻塞5秒,在MainThread中2次调用t.join(timeout=2),分别的等待了2秒,总等待时间是4秒,但是当执行第二个t.join(timeout=2)后,线程t依然没有被强制停止,然后主线执行完毕,然后_python_exit方法被调用,在_DummyThread线程中由调用t.join(),继续等待子线程t的执行完毕,直到线程t打印exit at 17:13:49 ...才执行完毕。

总结:

join()是可以被一个线程多次调用的,相当是多次等待的叠加。把_python_exit函数注册到atexit模块后,其他线程即使企图调用t.jion(n)来终止线程t也不起作用,因为_python_exit总是在最后执行时调用t.jion()来保证让线程t执行完毕,而不是被中途强制停止。

07-16 01:01