[Python] multiprocessing.Pool(processes=0) not allowed?

In a parallelized script I would like to keep the ability to do the processing in a single CPU. In order to do that it seems that I need to duplicate the code that iterates over the results:

#!/usr/bin/env python3
import multiprocessing

# argparsed parameters:
num_cpus = 1
input_filename = 'file.txt'

def worker(line):
'''Do something with a line'''
# dummy processing that requires some CPU time
for x in range(1000000):
line.split()
return line.split()

def data_iter(filename):
'''iterate over the data records'''
with open(filename) as fh:
for line in fh:
yield line

if __name__ == '__main__':
if (num_cpus >= 2):
with multiprocessing.Pool( processes = num_cpus - 1 ) as pool:
for result in pool.imap(worker, data_iter(input_filename)):
# dummy post-processing (copy#1)
print(' '.join(result))
else:
for line in data_iter(input_filename):
result = worker(line)
# dummy post-processing (copy#2)
print(' '.join(result))


note: you can use any text file with multiple lines for testing the code.

Is there a valid reason for multiprocessing.Pool not allowing processes = 0?

POSTSCRIPT


I though of one plausible "technical" reason for not allowing it:

The worker processes in the Pool can't modify the environment of the main program, which wouldn't be true if the worker function is run without spawning a new process.

Continue reading...