Importable `multiprocessing.Pool` function

问题

This is probably simple and I'm just not finding a suitable question.

If I do a stand-alone script with multiprocessing.Pool i know I'm supposed to do:

def foo(x):
    return x**2

if __name__ == '__main__':
    with Pool(n_jobs) as p:
        p.map(foo, list_of_inputs)

But if I want to then make it an importable function, I assume __name__ will no longer be '__main__'. Is it safe to just do:

def __foo(x):
    return x**2

def bar(list_of_inputs, n_jobs):
    with Pool(n_jobs) as p:
        out = p.map(__foo, list_of_inputs)
    return out

###################################################

from test.py import bar

baz = bar(mylist, 4)

I assume so, since bar will only be called in the calling process and then only foo will be called in the daughter processes, but I'm just wondering if this is a safe way to do this in general.

回答1:

I don't see any problem with your code that works fine for me in python3 and python2 (if I don't use the with statement for the Pool). In my opinion, this is a safe way to use multiprocessing and I don't understand why you experience this kind of infinite hanging. What version of Python are you using?

来源：https://stackoverflow.com/questions/57140777/importable-multiprocessing-pool-function