homepage

[Moved from http://code.google.com/p/argparse/issues/detail?id=73]
What steps will reproduce the problem?
parser = ArgumentParser(prefix_chars="-+")
parser.add_argument("-a",action="store_true")
parser.add_argument("+b",action="store_true")
parser.add_argument("+c",action="store_true")
print parser.parse_args("-abc".split())
What is the expected output? What do you see instead?
I would expect Namespace(a=True, b=True, c=True)
You get Namespace(a=True, b=False, c=True)
because in the loop that iterates through the prefix characters it builds an option_string to try by pulling the first character off explicit_arg. If it doesn't match any option then it will try the next prefix character, but it will have lost that character from the explicit_arg. This is also a problem even with only one prefix character because if it doesn't match an option the error message will be missing that one character.
The above match -a, tried -b and fails, but instead of trying +b, it tries +c because the b has been lost.
I've included multiple_short.patch which fixes just this bug. However, I also don't agree with the behavior of trying all prefix characters. I would expect that when merging short options together they would have to all share the same prefix character. If I have prefix_chars="-+" and I have options -a, +a, -b, and +b and I type +ab I would expect my action to be called with option_strings +a and +b, not +a and -b. The patch to fix the above bug *and* only try the same prefix character as the first option is multiple_short_same_prefix.patch.
[Only the latter patch (which sounds like the right behavior) is attached. It will need to be updated to work against Python trunk.]

Attaching a test to verify parse failure on mismatched prefix (-abc or +abc). Steven's patch makes it pass.

Updated Steven's patch; no changes, but now it knows the new context so that ``patch -p0`` won't fail.

Patches applied in r86090 (3.X) and r86091 (2.7). Thanks for your help Catherine, and sorry it took me so long to apply these.