MAINT: depend on array-api-compat

[crusaderky]

• ClosesENH/API: xp-bound namespaces, array-api-compat #6
• Blocked by pixi update #48
• Blocks implementation of at (ENH: Test for read-only arrays array-api-compat#205)

Follow-up: vendoring hook in scipy

[lucascolley]

WIP: vendoring hook in scipy (if you merge this now, scipy will break!)

Note that this will break all third-party libraries that vendor array-api-extra through a unpinned git subproject, like scipy does.

Why? I was under the impression that "unpinned" git submodules are not bumped unless someone deliberately does so. All of SciPy's git submodules are unpinned.

[crusaderky]

I was under the impression that "unpinned" git submodules are not bumped unless someone deliberately does so

I wasn't aware of that, good to know

[lucascolley]

I think we will want to document somewhere what we are relying on from array-api-compat, or otherwise indicate which versions of array-api-compat this can depend on or be co-vendored with.

I think I'm fine with something along the lines of "we aim for compatibility with the latest released version of array-api-compat, and your mileage may vary with older or dev versions".

[crusaderky]

@lucascolley this is ready for review. Could you please enable workflows for this PR?

[crusaderky]

Could you please unlock this?

[crusaderky]

I executed rm pixi.lock && pixi run tests to update the lock; not sure if it's the correct process?

[lucascolley]

rm pixi.lock && pixi run tests

pixi update should do

Could you please unlock this?

Unfortunately, I believe manual approval has to be given every time, for first-time contributors. Feel free to submit a separate PR with a trivial change if you want to unblock that.

[lucascolley]

looks like you may want to bump the version of pixi used by the CI workflows, if your local version is newer

[crusaderky]

Feel free to submit a separate PR with a trivial change if you want to unblock that.

#48

[crusaderky]

I've rebased this PR on top of #48 to get a more readable pixi.lock diff

[crusaderky]

@lucascolley all green; ready for final review and merge

[lucascolley]

thanks a lot @crusaderky, this looks great!

I suppose the scope of this package expanding is a blessing and a curse :). We will need to update the 'scope' section of the docs, but I'm happy to tackle this separately.

[crusaderky]

@lucascolley all comments addressed; this should be ready for merge

[lucascolley]

thanks!