Recommended way to test for argparse exceptions

[standage]

When using pytest to test my CLIs, I often want to capture exceptions thrown by argparse during argument parsing. Because argparse throws a SystemExit, this is what pytest catches. If I want to test the type and/or message of the original ArgumentError, I either have to use something like capsys or I have to do some under-the-hood context unpacking.

Option 1

def test_bad_args(capsys):
    with pytest.raises(SystemExit):
        args = get_parser().parse_args(["--intarg=NotAnInt"])
    terminal = capsys.readouterr()
    assert "SomeExceptionType" in terminal.err
    assert "Some error message" in terminal.err

Option 2 (stolen shamelessly from https://stackoverflow.com/a/67107620/459780)

def test_bad_args():
    with pytest.raises(SystemExit) as se:
        args = get_parser().parse_args(["--intarg=NotAnInt"])
    assert isinstance(se.value.__context__, SomeExceptionType)
    assert "Some error message" in se.value.__context__.message

Both of these feel like dirty hacks. Is there a better, cleaner way to write such tests?

[RonnyPfannschmidt]

Argparse is hostile to negative unittesting, as such those hacks are pretty much what I'd expect to crop up

[The-Compiler]

With Python 3.9+, set exit_on_errors=False on the parser, and you get a argparse.ArgumentError instead.

For older Python versions, override the error method of the parser, so that it e.g. raises a custom exception instead. Or override exit like in the example above that.

@RonnyPfannschmidt I really fail to see how this is "hostile to negative unittesting". Quite the opposite.

[RonnyPfannschmidt]

@The-Compiler you just demonstrated the hostility

Needing to override either default behaviour and/or methods to be able to test, which in turn means that the objects under test will be enforced to be monkeypatched and/or different from what you would use in a application

If there was a lower level entry to parsing then one could do this without needing to mangle the parsers