Ambiguity in behavior of ArgumentParser prefix_chars when same flag is specified with different prefixes

[joshnatis]

Environment

$ python --version
Python 3.8.10
$ uname -a
Linux 4.4.0-19041-Microsoft #1237-Microsoft Sat Sep 11 14:32:00 PST 2021 x86_64 x86_64 x86_64 GNU/Linux

Description

It seems that flags specified with different prefixes but the same 'name' (for lack of a better word) are treated as one flag.

Example

# example.py
import argparse

parser = argparse.ArgumentParser(prefix_chars='-+')
parser.add_argument('-e', metavar='<example>', action='append')
parser.add_argument('+e', metavar='<example>', action='append')
args = parser.parse_args()

print(args)

$ python example.py -e hello1 +e hello2
Namespace(e=['hello1', 'hello2'])

I was kind of hoping that somebody had thought of a smart way of treating them as two distinct flags : ). Regardless, I don't see anything about this on https://docs.python.org/3/library/argparse.html#prefix-chars, so I'm filing this as a bug in the documentation. It seems like a nitty detail but the page advertises itself as being thorough:

This page contains the API reference information. For a more gentle introduction to Python command-line parsing, have a look at the argparse tutorial.

Linked PRs

• gh-95836: Add custom type converter examples to argparse tutorial #125376
• [3.12] gh-95836: Add custom type converter examples to argparse tutorial (GH-125376) #125642
• [3.13] gh-95836: Add custom type converter examples to argparse tutorial (GH-125376) #125643

[serhiy-storchaka]

Prefix character has nothing to do with this. These options do the same because they are defined the same way.

I encountered such problem several times. You want to collect values in the same list, but with some tags depending on the option. Usually this is '--include'/'--exclude', but you write it as '+e'/'-e'.

One way is to use a custom action. I used this in Lib/test/libregrtest/cmdline.py:

class FilterAction(argparse.Action):
    def __call__(self, parser, namespace, value, option_string=None):
        items = getattr(namespace, self.dest)
        items.append((value, self.const))
...
    group.add_argument('-m', '--match', metavar='PAT',
                       dest='match_tests', action=FilterAction, const=True,
                       help='match test cases and methods with glob pattern PAT')
    group.add_argument('-i', '--ignore', metavar='PAT',
                       dest='match_tests', action=FilterAction, const=False,
                       help='ignore test cases and methods with glob pattern PAT')

But now I see that it would be easier to do with a custom type converter. Modifying your example:

import argparse
parser = argparse.ArgumentParser(prefix_chars='-+')
parser.add_argument('-e', metavar='<example>', action='append', type=lambda x: ('-', x))
parser.add_argument('+e', metavar='<example>', action='append', type=lambda x: ('+', x))
print(parser.parse_args('-e hello1 +e hello2'.split()))

The output is:

Namespace(e=[('-', 'hello1'), ('+', 'hello2')])

[serhiy-storchaka]

This could be added to the tutorial as an example of using custom type converter or two options with the same dest.

[savannahostrowski]

Agreed - I'm in the process of reworking the docs/tutorial to remove duplicative examples etc. I'll self-assign and make sure that this is incorporated.