Question

Is there a way to write a python doctest string to test a script intended to be launched from the command line (terminal) that doesn’t pollute the documentation examples with os.popen calls?

#!/usr/bin/env python
# filename: add
"""
Example:
>>> import os
>>> os.popen('add -n 1 2').read().strip()
'3'
"""

if __name__ == '__main__':
    from argparse import ArgumentParser
    p = ArgumentParser(description=__doc__.strip())
    p.add_argument('-n',type = int, nargs   = 2, default = 0,help  = 'Numbers to add.')
    p.add_argument('--test',action = 'store_true',help  = 'Test script.')
    a = p.parse_args()
    if a.test:
        import doctest
        doctest.testmod()
    if a.n and len(a.n)==2:
        print a.n[0]+a.n[1]

Running doctest.testmod() without using popen just causes a test failure because the script is run within a python shell instead of a bash (or DOS) shell.

The advanced python course at LLNL suggests putting scripts in files that are separate from .py modules. But then the doctest strings only test the module, without the arg parsing. And my os.popen() approach pollutes the Examples documentation. Is there a better way?

Answer 1

Just found something looking like the answer you want:
shell-doctest.