search

data-apis / array-api

RFC document, tooling and other content related to the array API standard
https://data-apis.github.io/array-api/latest/
MIT License
204 stars 42 forks source link

Referring to this project by name #759

Closed lucascolley closed 3 months ago

lucascolley commented 3 months ago

This project is referred to using a variety of names, including:

  1. (the) Python array API standard
  2. (the) array API standard
  3. (the) Array API standard
  4. (the) Array API

As far as I can tell, (1) is the closest we have to an official name for the project, as the title of the docs.

"Python" makes sense to drop in a lot of contexts. For example, (2) is used in the description of this repo and throughout NEP 56. I usually use (2).

"Array" is often capitalised. For example, the Quansight Labs Annual Report 2023 uses (3) throughout. Maybe this is more grammatically correct, I'm not sure.

"Standard" is often dropped. I have seen (4) a lot in discussions on consumer repos. I think that this has the potential to be confusing in some scenarios where other array APIs are also being discussed.

Maybe this is just bike-shedding to the extreme, but I think it would be good to improve the consistency of these references. I would suggest that we try to always use (2), with (1) where the context of Python isn't obvious, and the wider community should catch on. Such a well-thought-out API standard deserves a standard name :)

izaid commented 3 months ago

Weirdly, this has been on my mind as well whenever I write something about (the) array API standard. At some point, I mentally chose to go with 2).

rgommers commented 3 months ago

Yes, it's a little tricky indeed. I have just cleaned up all the NumPy 2.0 release notes to use (2). I'd say that (1) is the official name, and if it's clear we're within the Python world, then I think using (2) is fine.

I'd like to avoid (3) as an unnecessary deviation of capitalization, and (4) because it is too generic.