Argparse does not identify negative numbers with underscores as a numerical value

[coppinri]

Bug report

Bug description:

Simple example:

$ python test_arguments.py --Value -1_000.0
error: argument -v/--Value: expected one argument

The regular expression to identify negative numbers in the argparse.py module uses the following regular expression:

#argparse.py:1353-1354
        # determines whether an "option" looks like a negative number
        self._negative_number_matcher = _re.compile(r'^-\d+$|^-\d*\.\d+$')

This does not identify negative numerical values with underscores as a numerical value

This could probably be resolved by changing the regex to:

        # determines whether an "option" looks like a negative number
        self._negative_number_matcher = _re.compile(r'^-\d[\d_]*$|^-[\d_]*\.\d+$')

CPython versions tested on:

3.9

Operating systems tested on:

Windows

Linked PRs

• GH-123945: Update regex for parsing negative numbers that contain underscores #123970
• [3.13] GH-123945: Update regex for parsing negative numbers that contain underscores (GH-123970) #124158
• [3.12] GH-123945: Update regex for parsing negative numbers that contain underscores (GH-123970) #124175

[hugovk]

Small reproducer:

import argparse

parser = argparse.ArgumentParser()
parser.add_argument("--int", type=int)
parser.add_argument("--float", type=float)
args = parser.parse_args()
print(args)

Reproducible on main. Positive values will accept underscore, but not negative:

❯ p 1.py --int 1000
Namespace(int=1000, float=None)

❯ p 1.py --int 1_000
Namespace(int=1000, float=None)

❯ p 1.py --int -1000
Namespace(int=-1000, float=None)

❯ p 1.py --int -1_000
usage: 1.py [-h] [--int INT] [--float FLOAT]
1.py: error: argument --int: expected one argument

❯ p 1.py --float 1000.1
Namespace(int=None, float=1000.1)

❯ p 1.py --float 1_000.1
Namespace(int=None, float=1000.1)

❯ p 1.py --float -1000.1
Namespace(int=None, float=-1000.1)

❯ p 1.py --float -1_000.1
usage: 1.py [-h] [--int INT] [--float FLOAT]
1.py: error: argument --float: expected one argument

[savannahostrowski]

I'll take a look at this!

[coppinri]

Thanks for picking this up.

Just a quick note on the proposed regex:

The proposed regex may accept numbers like: -_.001 that may not be inline with the python parser, but I haven't check with the

There is also no provision for underscores in the fractional part of the number.

[serhiy-storchaka]

The last change introduced a regression: -.5 no longer accepted as a negative number.