(original) (raw)

On Fri, Jan 19, 2018 at 8:47 PM, Nick Coghlan <ncoghlan@gmail.com> wrote:

On 20 January 2018 at 07:49, Mario Corchero <mariocj89@gmail.com> wrote:
\> I am happy to put some work into this (and Pablo Galindo in CC offered to
\> pair on it) but it is not clear for me whether the next step is drafting a
\> new PEP or this is just blocked on "re-evaluating" the current one.

I think that would be a question for Larry,

I think you meant for Guido. It's not Larry's language (yet :-).

since there are two main options here:

\- proposing just the "/" part of PEP 457 (which allows positional-only
arguments, but doesn't allow the expression of all builtin and
standard library signatures)
\- proposing the full PEP 547,

I assume you meant PEP 457 again. :-)

including the "argument groups" feature
(which is a bigger change, but allows the expression of signatures
like "range(\[start,\] stop, \[step,\] /)")

One key benefit I'd see to a new subset-of-457 PEP is that it would
allow a decision to be made on the basic "/" proposal without deciding
one way or the other on whether or not to provide a native way to
express signatures like the one for range().

I personally don't think such signatures are common enough to warrant special syntax, and I don't want to encourage them. The few we have (basically range(), slice() and a few functions in the curses module) don't inspire a lot of copy-cat APIs.

OTOH the more plain positional-only arguments are a pretty common need -- for example, for methods that are conventionally used that way, and overridden with disregard for argument names.

(IOW I agree with you here. ;-)

--Guido van Rossum (python.org/\~guido)