Is it possible for multiple python files to use one queue?

Question

one script(datamanger.py)

from multiprocessing import Manager
q = Manager().Queue()

The other two scripts are like this

from datamanager import q
import time
while True:
    time.sleep(1)
    q.put(1)

from datamanager import q
while True:
    if not q.empty():
        data = q.get()
        print(data)

Is it possible to realize the function only use queue instead of message queue such as kafka?

Asked By: 唐景麒

||

Source

Answer 1

in order to have the queue alive and not tied to either process, you need to spawn a server that manages it, this server should have a singleton queue, and everyone that contacts it will get a proxy to this queue, the server code looks as follows:

# queue_server.py

from multiprocessing.managers import SyncManager
from multiprocessing.managers import BaseProxy
import multiprocessing

address = ('127.0.0.1', 50000)  # you can change this
authkey = b"abc"  # you should change this

class SingletonQueue:
    instance = None
    def __new__(cls, *args, **kwargs):
        if SingletonQueue.instance is None:
            SingletonQueue.instance = object.__new__(SingletonQueue)
            return SingletonQueue.instance
        else:
            return SingletonQueue.instance

    def get_queue(self):
        if not hasattr(self, "queue"):
            manager = SyncManager(address=address, authkey=authkey)
            manager.connect()
            self.queue = manager.Queue()
        return self.queue

class CustomQueueProxy(BaseProxy):
    _exposed_ = ['get_queue']
    def get_queue(self):
        queue = self._callmethod('get_queue')
        return queue


def connect_manager():
    multiprocessing.current_process().authkey = authkey

    manager = SyncManager(address=address, authkey=authkey)
    manager.register("SingletonQueue", SingletonQueue, CustomQueueProxy)
    manager.connect()
    return manager

def start_server():
    manager = SyncManager(address=address, authkey=authkey)
    manager.register("SingletonQueue", SingletonQueue, CustomQueueProxy)
    server = manager.get_server()
    print(f"running on ip = {server.address[0]}, and port {server.address[1]}")
    multiprocessing.current_process().authkey = authkey
    server.serve_forever()

if __name__ == "__main__":
    start_server()

you need to run the server, after running the server you can connect to it with a client, the client code will look like this:

import multiprocessing
import queue_server  # the server python file

manager = queue_server.connect_manager()
queue: multiprocessing.Queue = manager.SingletonQueue().get_queue()
queue.put(1)
print(queue.get())

note that this sets the authentication key of your python process to a certain value, so you cannot use it for doing multiple connections with different authentication keys, you have to have a fixed authentication key.

Edit: i’d probably go with Charchit Agarwal answer if anyone is reading this in the future, or a mix of both answers. depending on whether you want to allow connection over network/docker boundaries, which my answer allows.

Answered By: Ahmed AEK

Answer 2

An alternative to Ahmed’s answer, which uses a simpler singleton, is below.

server.py:

from multiprocessing.managers import BaseManager
from multiprocessing import Queue

address = ('127.0.0.1', 50000)  # you can change this
authkey = b"abc"  # you should change this


class SharedQueue:

    def __init__(self):
        self._queue = Queue()
        self._queue.put("Something really important!")

    def __call__(self):
        return self._queue


if __name__ == "__main__":
    # Register our queue
    shared_queue = SharedQueue()
    BaseManager.register("get_queue", shared_queue)

    # Start server
    manager = BaseManager(address=address, authkey=authkey)
    srv = manager.get_server()
    srv.serve_forever()

client.py

from server import address, authkey
from multiprocessing.managers import BaseManager


if __name__ == "__main__":
    BaseManager.register("get_queue")
    manager = BaseManager(authkey=authkey, address=address)
    manager.connect()

    queue = manager.get_queue()
    print(queue.get())

Answered By: Charchit Agarwal

Is it possible for multiple python files to use one queue?

Question:

Answers: