multiprocessing using global variable as read-only data (Bad idea)

mp_global_var.py

#!/usr/bin/env python3
#-*- coding:utf8 -*-
from __future__ import print_function
import numpy as np
import multiprocessing as mp
import sys
from multiprocessing.queues import Empty

SHARE_DATA = None

def worker(in_queue):
    pid = mp.current_process().pid
    while True:
        try:
            i = in_queue.get_nowait()
            print("[{}] {}".format(pid, SHARE_DATA[i]), flush=True)
            in_queue.task_done()
        except Empty:
            pass

def main():
    in_queue = mp.JoinableQueue()
    for _ in range(5):
        p = mp.Process(target=worker, args=(in_queue,))
        p.daemon = True
        p.start()
    global SHARE_DATA
    print(SHARE_DATA, flush=True)
    for i in range(5):
        in_queue.put(i)
    in_queue.join()
    return 0

if __name__ == "__main__":
    SHARE_DATA = np.arange(10).reshape(5, 2)
    sys.exit(main())

Sorry, it's is a bad example ~
Basically, processes do not 'share' global variables, they copy it.
I simply forgot this fact, my bad.

Maybe slice the data I need and only put slices in the queue?
As far as I know, multiprocess will pickle the args first before passing it to worker function.
So I think this may solve the memory issue a bit but not all.

The memory efficient ways I can think of now is using database or threading.
With threading, all threads can share a read-only data in the memory.
Or load the data in the worker function.

That's it.