Using multiprocessing to read from a queue

Question

Here is code I use to populate and read from a queue using Python multiprocessing:

from multiprocessing import Lock, Process, Queue, Pool

import time
from random import randint

def add_to_queue(tasks_to_accomplish, name):
    while True:
        random_int = randint(0, 22)
        print('name', name , "adding" , random_int)
        tasks_to_accomplish.put(random_int)
        time.sleep(2)

def read_from_queue(tasks_to_accomplish, name):
    while True:
        item = tasks_to_accomplish.get()
        print('name' , name , item)

        time.sleep(.01)


if __name__ == '__main__':
    tasks_to_accomplish = Queue()

    p = Process(target=add_to_queue, args=(tasks_to_accomplish, "p"))
    p.start()

    p2 = Process(target=read_from_queue, args=(tasks_to_accomplish, "p2"))
    p2.start()
    p3 = Process(target=read_from_queue, args=(tasks_to_accomplish, "p3"))
    p3.start()

    p.join()
    p2.join()
    p3.join()

The code will execute inifinitely, here is partial output:

name p adding 3
name p2 3
name p adding 4
name p3 4
name p adding 0
name p2 0
name p adding 22
name p3 22
name p adding 2
name p2 2
name p adding 13
name p3 13
name p adding 0
name p2 0
name p adding 14
name p3 14
name p adding 20
name p2 20
name p adding 4
name p3 4

Reading from the queue the time taken is .01 seconds : time.sleep(.01). But the p2 & p3 processes do not appear to be reading the thread in .01 seconds as it's obvious they block for more than .01 seconds. Have I implemented the process threads correctly for reading from the queue ?

Answer 1

As Daniel pointed out, Queue.get() will block until data is available by default.

You can use q.get(block=True) to change that although this will raise an exception:

name p adding 12
name p2 12
Process Process-6:
Traceback (most recent call last):
  File "/home/user/.pyenv/versions/3.8.0/lib/python3.8/multiprocessing/process.py", line 313, in _bootstrap
    self.run()
  File "/home/user/.pyenv/versions/3.8.0/lib/python3.8/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "<ipython-input-2-4e6d57c64980>", line 15, in read_from_queue
    item = tasks_to_accomplish.get(block=False)
  File "/home/user/.pyenv/versions/3.8.0/lib/python3.8/multiprocessing/queues.py", line 110, in get
    raise Empty
_queue.Empty
Process Process-5:
Traceback (most recent call last):
  File "/home/user/.pyenv/versions/3.8.0/lib/python3.8/multiprocessing/process.py", line 313, in _bootstrap
    self.run()
  File "/home/user/.pyenv/versions/3.8.0/lib/python3.8/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "<ipython-input-2-4e6d57c64980>", line 15, in read_from_queue
    item = tasks_to_accomplish.get(block=False)
  File "/home/user/.pyenv/versions/3.8.0/lib/python3.8/multiprocessing/queues.py", line 110, in get
    raise Empty
_queue.Empty
name p adding 2
name p adding 12
name p adding 14
name p adding 21
name p adding 9
name p adding 13

You would need to:

def read_from_queue(tasks_to_accomplish, name):
    while True:
        try:
            item = tasks_to_accomplish.get(block=False)
        except:
            print('no data for', name)
        else:
            print('name' , name , item)
    
        time.sleep(.01)

to get:

name p adding 0
name p2 0
no data for p3
no data for p3
no data for p2
no data for p2
no data for p3
no data for p2
no data for p3
# about 350 more entries like this
name p adding 5
no data for p2
name p3 5
no data for p2
no data for p3
no data for p3
no data for p2
no data for p3
# ...

Unless you need to do some work in between reads I would say that yes, you've implemented correctly the reading processes (and you can safely remove the calls to sleep when reading).