python多进程队列的底层进程是什么？

Question

有人详细解释python多处理队列通信吗？当参数被放入队列时发生了什么？我有一段代码，这让我很困惑。

import time
import numpy as np
from multiprocessing import Queue, Process


def task(queue_in, queue_out):
    mutural_np = np.zeros((10, 2))
    while True:
        msg = queue_in.get()
        res = []
        i = 0
        for i in range(msg):
            newnp = np.ones((1, 2)) * (msg - i)
            mutural_np[i:i+1] = newnp
        res = mutural_np[:i]
        print("===> put: ", res)
        queue_out.put(res)


if __name__ == "__main__":
    queue_in = Queue(10)
    queue_out = Queue(1)
    p1 = Process(target=task, args=(queue_in, queue_out))
    p1.start()
    for i in range(5):
        queue_in.put(i + 1)

    while True:
        msg = queue_out.get()
        time.sleep(0.5)
        print("***> out: ", msg)

输出为：

===> put:  []
===> put:  [[2. 2.]]
===> put:  [[3. 3.]
 [2. 2.]]
***> out:  []
===> put:  [[4. 4.]
 [3. 3.]
 [2. 2.]]
***> out:  [[3. 3.]]
===> put:  [[5. 5.]
 [4. 4.]
 [3. 3.]
 [2. 2.]]
***> out:  [[4. 4.]
 [3. 3.]]
***> out:  [[5. 5.]
 [4. 4.]
 [3. 3.]]
***> out:  [[5. 5.]
 [4. 4.]
 [3. 3.]
 [2. 2.]]

为什么我有这种不一致？

我看到文档说

" When an object is put on a queue, the object is pickled and a background thread later flushes the pickled data to an underlying pipe. "

"The pickle module keeps track of the objects it has already serialized so that later references to the same object won’t be serialized again"

据我了解，当我将对象放入队列时，该对象已被 pickle 且不可变，但似乎 pickle 发生在它被清除后。

Answer 1

我想我通过阅读源代码找到了答案。

当我将对象放入进程队列时，Python 启动一个线程来序列化并发送数据。所以，为了安全起见，在放入队列之前复制一份。

python多进程队列的底层进程是什么？

what is the underlying process of python multiprocess queue?

python

multiprocessing

python-multiprocessing