Failure to recognize negative scientific notation (with nargs:2) about argparse HOT 5 CLOSED

I submitted a patch to Python issue 9334 to

• accept negative scientific and complex numbers
• add the args_default_to_positional parser option

http://bugs.python.org/issue9334#msg184987

from argparse.

@hpaulj negative sentific floats now works. This fix does not cover complex numbers and values with leading dashes. I think, this can be safely postponed, until someone really needs it in real life.

from argparse.

I explored expanding the regex expression like this. Adding the exponential is easy. Adding complex makes the whole expression unwieldy. In the Python patch I chose instead to separate the two uses of _negativeNumberMatcher.

For identifying negative like option strings I left the original matcher (examples are all simple ones like '-1', not even '-1.23', much less '-1.23e4'.)

For identifying real negative numbers, I chose instead to let Python's own number parser do the job:

 if not self._has_negative_number_optionals:
     try:
         complex(arg_string)
         return None
 except ValueError:
        pass

The javascript equivalent could use !isNaN(argString)

if (!_.any(this._hasNegativeNumberOptionals) && !isNaN(argString)) {
  return null;
}

But since javascript does not have complex number type, expanding the regex to include exponentials should be just fine.

I've submitted a number of python argparse patches, but haven't gotten much feedback from their developers.

from argparse.

Yes, isNaN would be more simple. I didn't digged into algorythm, and just replaced existing regexp, after trebor89 pointed to bad line of code :) .

By the way, may be you have time to look into mincer project https://github.com/nodeca/mincer ? It has much more space for improvements, than argparse. I think, even in current state, argparse cover node developer's needs for 200% :) .

from argparse.

Argparse 2.0 was ported from scratch based on python 3.9.0. We aimed to keep source code as close as possible to the original, so this change didn't make it in.

Is it used by anybody? If so, maybe bug python developers again, issue still not resolved over there.

from argparse.