Python 3.6+：嵌套的多处理管理器导致 FileNotFoundError

Question

所以我试图在一个字典上使用多处理管理器，这是我最初的尝试：

from multiprocessing import Process, Manager

def task(stat):
    test['z'] += 1
    test['y']['Y0'] += 5


if __name__ == '__main__':
    test = Manager().dict({'x': {'X0': 10, 'X1': 20}, 'y': {'Y0': 0, 'Y1': 0}, 'z': 0})

    p = Process(target=task, args=(test,))
    p.start()
    p.join()

    print(test)

当然，当我运行它时，output 不是我所期望的， z正确更新而y不变：这是输出：

{'x': {'X0': 10, 'X1': 20}, 'y': {'Y0': 0, 'Y1': 0}, 'z': 1}

然后我用谷歌搜索， 在这里找到了一个解释，显然嵌套的字典也必须是Manager().dict()而不是正常的 python 字典（自 Python 3.6 以来可能）。 所以我做了以下事情：

from multiprocessing import Process, Manager

def task(stat):
    test['z'] += 1
    test['y']['Y0'] += 5


if __name__ == '__main__':
    test = Manager().dict({'x': Manager().dict({'X0': 10, 'X1': 20}), 'y': Manager().dict({'Y0': 0, 'Y1': 0}), 'z': 0})

    p = Process(target=task, args=(test,))
    p.start()
    p.join()

    print(test)
    print(test['y'])

但是它没有正常工作，而是出现了这个无法解释的错误，为了清楚起见，将其分为三个部分。 第一部分对应于test['y']['Y0'] += 5而第二部分只是print(test)最后一部分是print(test['y'])的 output

Process Process-4:
Traceback (most recent call last):
  File "/usr/lib/python3.7/multiprocessing/process.py", line 297, in _bootstrap
    self.run()
  File "/usr/lib/python3.7/multiprocessing/process.py", line 99, in run
    self._target(*self._args, **self._kwargs)
  File "shit.py", line 5, in task
    test['y']['Y0'] += 5
  File "<string>", line 2, in __getitem__
  File "/usr/lib/python3.7/multiprocessing/managers.py", line 796, in _callmethod
    kind, result = conn.recv()
  File "/usr/lib/python3.7/multiprocessing/connection.py", line 251, in recv
    return _ForkingPickler.loads(buf.getbuffer())
  File "/usr/lib/python3.7/multiprocessing/managers.py", line 920, in RebuildProxy
    return func(token, serializer, incref=incref, **kwds)
  File "/usr/lib/python3.7/multiprocessing/managers.py", line 770, in __init__
    self._incref()
  File "/usr/lib/python3.7/multiprocessing/managers.py", line 824, in _incref
    conn = self._Client(self._token.address, authkey=self._authkey)
  File "/usr/lib/python3.7/multiprocessing/connection.py", line 492, in Client
    c = SocketClient(address)
  File "/usr/lib/python3.7/multiprocessing/connection.py", line 619, in SocketClient
    s.connect(address)
FileNotFoundError: [Errno 2] No such file or directory




{'x': <DictProxy object, typeid 'dict' at 0x7f01de2c5860>, 'y': <DictProxy object, typeid 'dict' at 0x7f01de2c5898>, 'z': 1}




Traceback (most recent call last):
  File "test.py", line 16, in <module>
    print(test['y'])
  File "<string>", line 2, in __getitem__
  File "/usr/lib/python3.7/multiprocessing/managers.py", line 796, in _callmethod
    kind, result = conn.recv()
  File "/usr/lib/python3.7/multiprocessing/connection.py", line 251, in recv
    return _ForkingPickler.loads(buf.getbuffer())
  File "/usr/lib/python3.7/multiprocessing/managers.py", line 920, in RebuildProxy
    return func(token, serializer, incref=incref, **kwds)
  File "/usr/lib/python3.7/multiprocessing/managers.py", line 770, in __init__
    self._incref()
  File "/usr/lib/python3.7/multiprocessing/managers.py", line 824, in _incref
    conn = self._Client(self._token.address, authkey=self._authkey)
  File "/usr/lib/python3.7/multiprocessing/connection.py", line 492, in Client
    c = SocketClient(address)
  File "/usr/lib/python3.7/multiprocessing/connection.py", line 619, in SocketClient
    s.connect(address)
FileNotFoundError: [Errno 2] No such file or directory

我不确定为什么会这样。 内部字典显然已创建（如输出的第二部分所示）。 但由于某种原因，它们根本无法读取或写入。？ 为什么会这样？

额外：如果我通过 python 控制台（而不是脚本）运行相同的 python 代码，错误将从FileNotFoundError更改为ConnectionRefusedError 。 但是具有相同的精确追溯！

Answer 1

使用Manager() Manager().dict() Manager()中的Manager().dict() ，每次都会启动一个新的管理器进程，所以你真的是嵌套管理器（比如标题所说），这不是它应该的样子。 您需要做的是，实例化一个 Manager，然后在该管理器实例上创建字典：

from multiprocessing import Process, Manager
from multiprocessing.managers import DictProxy


def task(test):  # use parameter `test`, else you rely on forking
    test['z'] += 1
    test['y']['Y0'] += 5


if __name__ == '__main__':

    with Manager() as m:

        test = m.dict({'x': m.dict({'X0': 10, 'X1': 20}),
                       'y': m.dict({'Y0': 0, 'Y1': 0}),
                       'z': 0})

        p = Process(target=task, args=(test,))
        p.start()
        p.join()

        print(test)
        print(test['y'])

        # convert to normal dict before closing manager for persistence
        # in parent or for printing dict behind proxies
        d = {k: dict(v) if isinstance(v, DictProxy) else v
             for k, v in test.items()}

    print(d) # Manager already closed here

示例输出：

{'x': <DictProxy object, typeid 'dict' at 0x7f98cdaaa588>, 'y': <DictProxy object, typeid 'dict' at 0x7f98cda99c50>, 'z': 1}
{'Y0': 5, 'Y1': 0}
{'x': {'X0': 10, 'X1': 20}, 'y': {'Y0': 5, 'Y1': 0}, 'z': 1}

Process finished with exit code 0

如果您计划从多个进程修改管理器对象，则还需要使用Manager.Lock 。

Answer 2

我在 pycharm 中逐行运行 python 代码时遇到了这个问题。通过运行带有旁边绿色箭头的整个脚本解决了这个问题：

if __name__ == '__main__':

出于某种原因 python 不喜欢逐行运行池，但不确定为什么