Is it possible to multiprocess a function that returns something in Python?

Question

In Python I have seen many examples where multiprocessing is called but the target just prints something. I have a scenario where the target returns 2 variables, which I need to use later. For example:

def foo(some args):
   a = someObject
   b = someObject
   return a,b

p1=multiprocess(target=foo,args(some args))
p2=multiprocess(target=foo,args(some args))
p3=multiprocess(target=foo,args(some args))

Now what? I can do .start and .join, but how do I retrieve the individual results? I need to catch the return a,b for all the jobs I execute and then work on it.

Answer 1

You are looking to do some embarrassingly parallel work using multiple processes, so why not use a Pool? A Pool will take care of starting up the processes, retrieving the results, and returning the results to you.

I use pathos, which has a fork of multiprocessing, because it has much better serialization than the version that standard library provides.

(.py) file

from pathos.multiprocessing import ProcessingPool as Pool

def foo(obj1, obj2):
    a = obj1.x**2
    b = obj2.x**2
    return a,b

class Bar(object):
    def __init__(self, x):
        self.x = x

Pool().map(foo, [Bar(1),Bar(2),Bar(3)], [Bar(4),Bar(5),Bar(6)])

Result

[(1, 16), (4, 25), (9, 36)]

And you see that foo takes two arguments, and returns a tuple of two objects. The map method of Pool submits foo to the underlying processes and returns the result as res.

You can get pathos here: https://github.com/uqfoundation

Answer 2

Yes, sure - you can use a number of methods. One of the easiest ones is a shared Queue. See an example here: http://eli.thegreenplace.net/2012/01/16/python-parallelizing-cpu-bound-tasks-with-multiprocessing/

Answer 3

I'm copying this example straight from the docs because I can't give you a direct link to it. Note that it prints out the results from the done_queue, but you can do whatever you like with it.

#
# Simple example which uses a pool of workers to carry out some tasks.
#
# Notice that the results will probably not come out of the output
# queue in the same in the same order as the corresponding tasks were
# put on the input queue.  If it is important to get the results back
# in the original order then consider using `Pool.map()` or
# `Pool.imap()` (which will save on the amount of code needed anyway).
#
# Copyright (c) 2006-2008, R Oudkerk
# All rights reserved.
#

import time
import random

from multiprocessing import Process, Queue, current_process, freeze_support

#
# Function run by worker processes
#

def worker(input, output):
    for func, args in iter(input.get, 'STOP'):
        result = calculate(func, args)
        output.put(result)

#
# Function used to calculate result
#

def calculate(func, args):
    result = func(*args)
    return '%s says that %s%s = %s' % \
        (current_process().name, func.__name__, args, result)

#
# Functions referenced by tasks
#

def mul(a, b):
    time.sleep(0.5*random.random())
    return a * b

def plus(a, b):
    time.sleep(0.5*random.random())
    return a + b

#
#
#

def test():
    NUMBER_OF_PROCESSES = 4
    TASKS1 = [(mul, (i, 7)) for i in range(20)]
    TASKS2 = [(plus, (i, 8)) for i in range(10)]

    # Create queues
    task_queue = Queue()
    done_queue = Queue()

    # Submit tasks
    for task in TASKS1:
        task_queue.put(task)

    # Start worker processes
    for i in range(NUMBER_OF_PROCESSES):
        Process(target=worker, args=(task_queue, done_queue)).start()

    # Get and print results
    print 'Unordered results:'
    for i in range(len(TASKS1)):
        print '\t', done_queue.get()

    # Add more tasks using `put()`
    for task in TASKS2:
        task_queue.put(task)

    # Get and print some more results
    for i in range(len(TASKS2)):
        print '\t', done_queue.get()

    # Tell child processes to stop
    for i in range(NUMBER_OF_PROCESSES):
        task_queue.put('STOP')


if __name__ == '__main__':
    freeze_support()
    test()

It is originally from the multiprocessing module docs.

Answer 4

Why nobody uses callback of multiprocessing.Pool?

Example:

from multiprocessing import Pool
from contextlib import contextmanager

from pprint import pprint
from requests import get as get_page

@contextmanager
def _terminating(thing):
    try:
        yield thing
    finally:
        thing.terminate()

def _callback(*args, **kwargs):
    print("CALBACK")
    pprint(args)
    pprint(kwargs)

print("Processing...")
with _terminating(Pool(processes=WORKERS)) as pool:
    results = pool.map_async(get_page, URLS, callback=_callback)

    start_time = time.time()
    results.wait()
    end_time = time.time()
    print("Time for Processing: %ssecs" % (end_time - start_time))

Here, I print both args and kwargs. But you can replace callback by:

def _callback2(responses):
    for r in responses:
        print(r.status_code) # or do whatever with response...

Answer 5

It won't work on windows but here is is my multiprocessing decorator for functions, it returns a queue that you can poll and collect returned data from

import os
from Queue import Queue
from multiprocessing import Process

def returning_wrapper(func, *args, **kwargs):
    queue = kwargs.get("multiprocess_returnable")
    del kwargs["multiprocess_returnable"]
    queue.put(func(*args, **kwargs))

class Multiprocess(object):
    """Cute decorator to run a function in multiple processes."""
    def __init__(self, func):
        self.func = func
        self.processes = []

    def __call__(self, *args, **kwargs):
        num_processes = kwargs.get("multiprocess_num_processes", 2) # default to two processes.
        return_obj = kwargs.get("multiprocess_returnable", Queue()) # default to stdlib Queue
        kwargs["multiprocess_returnable"] = return_obj
        for i in xrange(num_processes):
            pro = Process(target=returning_wrapper, args=tuple([self.func] + list(args)), kwargs=kwargs)
            self.processes.append(pro)
            pro.start()
        return return_obj


@Multiprocess
def info():
    print 'module name:', __name__
    print 'parent process:', os.getppid()
    print 'process id:', os.getpid()
    return 4 * 22

data = info()
print data.get(False)

Answer 6

Here is an example of multi-process search for huge files.