Python

Question

我需要在一段时间后终止一些进程，所以我使用休眠另一个进程来等待。但是我猜新进程无法从主进程访问全局变量。请问我该如何解决？

代码：

import os
from subprocess import Popen, PIPE
import time
import multiprocessing


log_file = open('stdout.log', 'a')
log_file.flush()

err_file = open('stderr.log', 'a')
err_file.flush()

processes = []


def processing():
    print "processing"
    global processes
    global log_file
    global err_file

    for i in range(0, 5):
        p = Popen(['java', '-jar', 'C:\Users\two\Documents\test.jar'], stdout=log_file, stderr=err_file) # something long running
        processes.append(p)
    print len(processes)    # returns 5

def waiting_service():
    name = multiprocessing.current_process().name
    print name, 'Starting'
    global processes
    print len(processes)    # returns 0

    time.sleep(2)

    for i in range(0, 5):
        processes[i].terminate()
    print name, 'Exiting'

if __name__ == '__main__':
    processing()

    service = multiprocessing.Process(name='waiting_service', target=waiting_service)

    service.start()

Answer 1

您应该使用 synchronization primitives。

可能您想要设置一个 Event，它会在一段时间后由主（父）进程触发。

您可能还想等待进程实际完成并 join 它们（就像线程一样）。

如果你有很多类似的任务，你可以使用像multiprocessing.Pool这样的处理池。

这是一个如何完成的小例子：

import multiprocessing
import time

kill_event = multiprocessing.Event()

def work(_id):
    while not kill_event.is_set():
        print "%d is doing stuff" % _id
        time.sleep(1)
    print "%d quit" % _id

def spawn_processes():
    processes = []

    # spawn 10 processes
    for i in xrange(10):
        # spawn process
        process = multiprocessing.Process(target=work, args=(i,))
        processes.append(process)
        process.start()
        time.sleep(1)

    # kill all processes by setting the kill event
    kill_event.set()

    # wait for all processes to complete
    for process in processes:
        process.join()

    print "done!"
spawn_processes()

Answer 2

整个问题都在 Windows' Python 中。 Python for Windows 正在阻止在函数中看到的全局变量。我已切换到 linux，我的脚本运行正常。

特别感谢@rchang的评论：

When I tested it, in both cases the print statement came up with 5. Perhaps we have a version mismatch in some way? I tested it with Python 2.7.6 on Linux kernel 3.13.0 (Mint distribution).

Python - 如何将全局变量传递给 multiprocessing.Process？

Python - How to pass global variable to multiprocessing.Process?

process

global-variables

multiprocessing

python-multiprocessing