python multiprocessing, do the processes share a common variable?

Question

I have this:

#!/usr/bin/env python

import multiprocessing

class MultiprocessingTest(object):

    def __init__(self):
        self.cmd = ''

    def for_process_A(self):
        self.cmd = "AA"
        print "%s executing and cmd is %s" % (multiprocessing.current_process().name, self.cmd)


    def for_process_B(self):
        self.cmd = "BB"
        print "%s executing and cmd is %s" % (multiprocessing.current_process().name, self.cmd)


if __name__ == '__main__':

    obj =   MultiprocessingTest()

    process_A = multiprocessing.Process(target=obj.for_process_A, name='process_A')
    process_B = multiprocessing.Process(target=obj.for_process_B, name='process_B')

    process_A.start()
    process_B.start()


    process_A.join()
    process_B.join()

Question:

Do the two processes share the variable cmd?

Do both processes have a separate class MultiprocessingTest definition and work off of that?

Independent copies of which data exists in the two processes?

I am trying to understand from a theoretical standpoint what is actually happening here. Can you please comment on that?

Test Run o/p:

$ ./commonvar.py 

process_A executing and cmd is AA

process_B executing and cmd is BB

Now that is such fantastic test to run yourself! What did it show up? What is the conculsion? — Aida Paul
– Aida Paul, Commented Feb 26, 2014 at 4:14
So what does that test say to you? Not much right? You may want expand it to actually test what are you trying to, well, test ;) — Aida Paul
– Aida Paul, Commented Feb 26, 2014 at 4:21
When you make a Process, the entire process is forked - i.e. memory is copied. That's putting it a little simplistically, but for the purposes of a mental exercise, all your objects split into two - one copy stays in the parent process, the other heads off to the new process. — roippi
– roippi, Commented Feb 26, 2014 at 4:25
@Puciek Hey Puciek can you suggest what a good way to test this would be? I'd like to draw conclusion (test result) that inspires 100% confidence. — Ankur Agarwal
– Ankur Agarwal, Commented Feb 26, 2014 at 5:40

Umair Khan · Accepted Answer · 2014-02-26 05:31:54Z

1

Processes don't share data. Each process is a separate container with following resources, generally speaking:

Code to execute
Stack
Processor time

Processes interact with outside world through Pipes.

So to answer your questions:

Processes will not share cmd variable.
Processes will have separate copies of the class code.
All the program data will be independent.

Further Explanation:

Behind the scenes, fork system call is used to create a process (assuming you are using *nix). Processes are heavier compared to threads because of the overhead involved in switching the conext.

edited Feb 26, 2014 at 5:31

answered Feb 26, 2014 at 5:21

Umair Khan

1994 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

mgilson · Accepted Answer · 2014-02-26 04:18:36Z

0

Changes which happen inside of multiprocessing shouldn't propagate back to the calling "thread" (or any of the other multiprocessing processes). If you want that sort of "shared-memory-like" behavior, you'll need to look into using a multiprocessing.Manager.

answered Feb 26, 2014 at 4:18

mgilson

312k70 gold badges656 silver badges722 bronze badges

2 Comments

Ankur Agarwal Over a year ago

I am trying to understand from a theoretical standpoint what is actually happening here. Can you please comment on that?

mgilson Over a year ago

@abc -- Sure. Essentially, where you hit the process.start(), multiprocessing starts a new python process. It basically clones your current processes state and then executes mostly independently from the current process. When it's done doing what you asked it to, it returns the values and sends them back to the current process for whatever you want to do with them.

Collectives™ on Stack Overflow

python multiprocessing, do the processes share a common variable?

2 Answers 2

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related