Python多进程编程中全局变量不共享的常见误解

import multiprocessing as mp

# 全局配置，所有子进程共用
config = {"mode": "fast", "retry": 3}
counter = 0

def process_data(data):
    global counter
    # 根据config处理数据
    result = do_something(data, config)
    counter += 1  # 统计处理了多少条
    return result

if __name__ == "__main__":
    with mp.Pool(4) as pool:
        results = pool.map(process_data, big_data_list)
    print(f"一共处理了{counter}条数据")

深入解析：为何多进程不共享全局变量？

• 多线程：多个线程位于同一个进程内，共享同一块内存。修改全局变量后，所有线程都能感知到变化。
• 多进程：每个进程拥有独立的内存空间。子进程修改的是自己那份拷贝，主进程完全不可见。

三种典型场景，三种常见陷阱

陷阱1：子进程修改了，主进程却看不见

陷阱2：Windows环境下更隐蔽——全局变量直接“失踪”

陷阱3：进程池内修改全局变量，改了也等于白改

from multiprocessing import Pool

total = 0

def worker(x):
    global total
    total += x
    return x * 2

if __name__ == "__main__":
    with Pool(4) as pool:
        pool.map(worker, [1,2,3,4])
    print(total)  # 还是0

如何让多进程“共享”数据？

方案1：通过参数传递（最简洁，最推荐）

def process_data(data, config):
    # 使用传进来的config，避免全局变量
    return do_something(data, config)

if __name__ == "__main__":
    CONFIG = {"mode": "fast"}
    with mp.Pool(4) as pool:
        # 把config作为参数传给每个worker
        results = pool.map(lambda x: process_data(x, CONFIG), big_data_list)

方案2：使用multiprocessing.Value和Array（共享内存）

from multiprocessing import Pool, Value

counter = Value('i', 0)  # 'i'表示整数

def worker(x):
    with counter.get_lock():  # 加锁防止竞争条件
        counter.value += x

if __name__ == "__main__":
    with Pool(4) as pool:
        pool.map(worker, [1,2,3,4])
    print(counter.value)  # 10

方案3：使用Manager（共享复杂对象）

from multiprocessing import Pool, Manager

def worker(shared_list, x):
    shared_list.append(x)

if __name__ == "__main__":
    manager = Manager()
    shared_list = manager.list()
    with Pool(4) as pool:
        pool.map(lambda x: worker(shared_list, x), [1,2,3,4])
    print(list(shared_list))  # [1,2,3,4]

方案4：使用Queue（任务分发与结果收集）

from multiprocessing import Process, Queue

def worker(q):
    while True:
        item = q.get()
        if item is None:
            break
        # 处理item
        q.put(item * 2)

if __name__ == "__main__":
    q = Queue()
    # 放任务、启动进程、取结果...

一张表快速对照

from multiprocessing import Pool, Value

counter = Value('i', 0)

def process_data(data, config):
    global counter
    with counter.get_lock():
        counter.value += 1
    return do_something(data, config)