I have a Python module that uses the argparse library. How do I write tests for that section of the code base?

You should refactor your code and move the parsing to a function:

def parse_args(args):
    parser = argparse.ArgumentParser(...)
    parser.add_argument...
    # ...Create your parser as you like...
    return parser.parse_args(args)

Then in your main function you should just call it with:

parser = parse_args(sys.argv[1:])

(where the first element of sys.argv that represents the script name is removed to not send it as an additional switch during CLI operation.)

In your tests, you can then call the parser function with whatever list of arguments you want to test it with:

def test_parser(self):
    parser = parse_args(['-l', '-m'])
    self.assertTrue(parser.long)
    # ...and so on.

This way you’ll never have to execute the code of your application just to test the parser.

If you need to change and/or add options to your parser later in your application, then create a factory method:

def create_parser():
    parser = argparse.ArgumentParser(...)
    parser.add_argument...
    # ...Create your parser as you like...
    return parser

You can later manipulate it if you want, and a test could look like:

class ParserTest(unittest.TestCase):
    def setUp(self):
        self.parser = create_parser()

    def test_something(self):
        parsed = self.parser.parse_args(['--something', 'test'])
        self.assertEqual(parsed.something, 'test')

“argparse portion” is a bit vague so this answer focuses on one part: the parse_args method. This is the method that interacts with your command line and gets all the passed values. Basically, you can mock what parse_args returns so that it doesn’t need to actually get values from the command line. The mock package can be installed via pip for python versions 2.6-3.2. It’s part of the standard library as unittest.mock from version 3.3 onwards.

import argparse
try:
    from unittest import mock  # python 3.3+
except ImportError:
    import mock  # python 2.6-3.2


@mock.patch('argparse.ArgumentParser.parse_args',
            return_value=argparse.Namespace(kwarg1=value, kwarg2=value))
def test_command(mock_args):
    pass

You have to include all your command method’s args in Namespace even if they’re not passed. Give those args a value of None. (see the docs) This style is useful for quickly doing testing for cases where different values are passed for each method argument. If you opt to mock Namespace itself for total argparse non-reliance in your tests, make sure it behaves similarly to the actual Namespace class.

Below is an example using the first snippet from the argparse library.

# test_mock_argparse.py
import argparse
try:
    from unittest import mock  # python 3.3+
except ImportError:
    import mock  # python 2.6-3.2


def main():
    parser = argparse.ArgumentParser(description='Process some integers.')
    parser.add_argument('integers', metavar='N', type=int, nargs='+',
                        help='an integer for the accumulator')
    parser.add_argument('--sum', dest='accumulate', action='store_const',
                        const=sum, default=max,
                        help='sum the integers (default: find the max)')

    args = parser.parse_args()
    print(args)  # NOTE: this is how you would check what the kwargs are if you're unsure
    return args.accumulate(args.integers)


@mock.patch('argparse.ArgumentParser.parse_args',
            return_value=argparse.Namespace(accumulate=sum, integers=[1,2,3]))
def test_command(mock_args):
    res = main()
    assert res == 6, "1 + 2 + 3 = 6"


if __name__ == "__main__":
    print(main())

Make your main() function take argv as an argument rather than letting it read from sys.argv as it will by default:

# mymodule.py
import argparse
import sys


def main(args):
    parser = argparse.ArgumentParser()
    parser.add_argument('-a')
    process(**vars(parser.parse_args(args)))
    return 0


def process(a=None):
    pass

if __name__ == "__main__":
    sys.exit(main(sys.argv[1:]))

Then you can test normally.

import mock

from mymodule import main


@mock.patch('mymodule.process')
def test_main(process):
    main([])
    process.assert_call_once_with(a=None)


@mock.patch('foo.process')
def test_main_a(process):
    main(['-a', '1'])
    process.assert_call_once_with(a='1')

1. Populate your arg list by using sys.argv.append() and then call parse(), check the results and repeat.
2. Call from a batch/bash file with your flags and a dump args flag.
3. Put all your argument parsing in a separate file and in the if __name__ == "__main__": call parse and dump/evaluate the results then test this from a batch/bash file.

I did not want to modify the original serving script so I just mocked out the sys.argv part in argparse.

from unittest.mock import patch

with patch('argparse._sys.argv', ['python', 'serve.py']):
    ...  # your test code here

This breaks if argparse implementation changes but enough for a quick test script. Sensibility is much more important than specificity in test scripts anyways.

A simple way of testing a parser is:

parser = ...
parser.add_argument('-a',type=int)
...
argv = '-a 1 foo'.split()  # or ['-a','1','foo']
args = parser.parse_args(argv)
assert(args.a == 1)
...

Another way is to modify sys.argv, and call args = parser.parse_args()

There are lots of examples of testing argparse in lib/test/test_argparse.py

parse_args throws a SystemExit and prints to stderr, you can catch both of these:

import contextlib
import io
import sys

@contextlib.contextmanager
def captured_output():
    new_out, new_err = io.StringIO(), io.StringIO()
    old_out, old_err = sys.stdout, sys.stderr
    try:
        sys.stdout, sys.stderr = new_out, new_err
        yield sys.stdout, sys.stderr
    finally:
        sys.stdout, sys.stderr = old_out, old_err

def validate_args(args):
    with captured_output() as (out, err):
        try:
            parser.parse_args(args)
            return True
        except SystemExit as e:
            return False

You inspect stderr (using err.seek(0); err.read() but generally that granularity isn’t required.

Now you can use assertTrue or whichever testing you like:

assertTrue(validate_args(["-l", "-m"]))

Alternatively you might like to catch and rethrow a different error (instead of SystemExit):

def validate_args(args):
    with captured_output() as (out, err):
        try:
            return parser.parse_args(args)
        except SystemExit as e:
            err.seek(0)
            raise argparse.ArgumentError(err.read())

When passing results from argparse.ArgumentParser.parse_args to a function, I sometimes use a namedtuple to mock arguments for testing.

import unittest
from collections import namedtuple
from my_module import main

class TestMyModule(TestCase):

    args_tuple = namedtuple('args', 'arg1 arg2 arg3 arg4')

    def test_arg1(self):
        args = TestMyModule.args_tuple("age > 85", None, None, None)
        res = main(args)
        assert res == ["55289-0524", "00591-3496"], 'arg1 failed'

    def test_arg2(self):
        args = TestMyModule.args_tuple(None, [42, 69], None, None)
        res = main(args)
        assert res == [], 'arg2 failed'

if __name__ == '__main__':
    unittest.main()

For testing CLI (command line interface), and not command output I did something like this

import pytest
from argparse import ArgumentParser, _StoreAction

ap = ArgumentParser(prog="cli")
ap.add_argument("cmd", choices=("spam", "ham"))
ap.add_argument("-a", "--arg", type=str, nargs="?", default=None, const=None)
...

def test_parser():
    assert isinstance(ap, ArgumentParser)
    assert isinstance(ap, list)
    args = {_.dest: _ for _ in ap._actions if isinstance(_, _StoreAction)}
    
    assert args.keys() == {"cmd", "arg"}
    assert args["cmd"] == ("spam", "ham")
    assert args["arg"].type == str
    assert args["arg"].nargs == "?"
    ...