[Python-ideas] Argparse argument deprecation

[Michel Desmoulin]

Le 09/08/2017 à 12:59, Ned Batchelder a écrit :
> OK, then on a more pragmatic note: why is it easier to write a callback
> than to write a simple if statement after the parsing?  Generating help
> is complex, and a common task that is closely tied to the syntax of the
> options, so it makes sense for argparse to do it.  Deprecation is
> neither complex, common, nor closely tied to the syntax of the options.
> 
> Another note about the proposal: calling it "deprecated" seems odd,
> since the proposal is really just a general-purpose callback.  argparse
> isn't generating the warning, your callback function would be doing it. 
> Why name it "deprecated"?  How is this different than the "action"
> keyword argument that argparse already provides?

I imagine something like:

def _(warn, forbid):
    warn('This is deprecated') # for forbid to just put an error

parser.add_option(on_deprecated=deprecationCallback)

This does:

- provide an easy way to warn, or transition to forbid
- allow introspection to list the deprecated options
- deprecated options can be marked as such in the generated --help
- create a complex dynamic deprecation message, or just pass a short lambda

But indeed I'd like it to be able to do:

parser.add_option(on_deprecated=DeprecationWarning('meh'))
parser.add_option(on_deprecated=ValueError('meh'))

As a shortcut for simple use cases.

I still don't know how to make the distinction between deprecated and
removed from the introspection point of view.

All in all, I think it's an interesting proposal, but I'm not going to
fight over it. If it never happens, I can fit a bunch of "if" like you said.

> 
> --Ned.
> 
> 
> On 8/9/17 5:54 AM, Michel Desmoulin wrote:
>> Argparse is not just about parsing, it's about providing convenient
>> tooling associated with parsing.
>>
>> Otherwise you would not have automatically generated a "usage" message
>> or a "--help" command.
>>
>> Following your definition, those are not parsing. But there are here,
>> because we all end up coding them anyway.
>>
>> Le 09/08/2017 à 11:50, Ned Batchelder a écrit :
>>> On 8/9/17 3:56 AM, Tarek Ziadé wrote:
>>>> Hey,
>>>>
>>>> I don't think there's any helper to deprecate an argument in argparse
>>>>
>>>> Let's say you have a --foo option in your CLI and want to deprecate it
>>>> in the next release before you completely remove it later.
>>>>
>>>> My first though on how to do this by adding a new "deprecated" option to
>>>> https://docs.python.org/3/library/argparse.html#argparse.ArgumentParser.add_argument
>>>>
>>>> "deprecated" would be a callable that is called after the argument has
>>>> been parsed by argparse, 
>>>> so the developer can decide if they want to issue a deprecation warning,
>>>> use the parsed value or override it etc.
>>> I don't see why this is something that argparse has to do.  The
>>> semantics of options is handled by the rest of the program.  Why would
>>> the parser be issuing these warnings?  Let argparse parse the options,
>>> then let other code deal with what they *mean*.
>>>
>>> --Ned.
>>>
>>> _______________________________________________
>>> Python-ideas mailing list
>>> Python-ideas at python.org
>>> https://mail.python.org/mailman/listinfo/python-ideas
>>> Code of Conduct: http://python.org/psf/codeofconduct/
>>>
>> _______________________________________________
>> Python-ideas mailing list
>> Python-ideas at python.org
>> https://mail.python.org/mailman/listinfo/python-ideas
>> Code of Conduct: http://python.org/psf/codeofconduct/
> 
> _______________________________________________
> Python-ideas mailing list
> Python-ideas at python.org
> https://mail.python.org/mailman/listinfo/python-ideas
> Code of Conduct: http://python.org/psf/codeofconduct/
>