Support 2023.12
version of the array API
#3901
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The array API has just had its
2023.12
release (data-apis/array-api#755)... in Feb 2024 馃槄 This PR simply acceptsapi_version="2023.12"
forextra.array_api.make_strategies_namespace()
.I believe there aren't any changes that should necessitate what the strategies namespace is doing for
2023.12
! I assume this will be an on-going trend, as I can't think of any existing discussions that would either shake-up the existing strategies or want us to add news ones.I did have the idea to maybe just warn instead of erroring when
api_version
is greater than the last item ofRELEASED_VERSIONS
(and notdraft
), so we don't "annoy" libraries by asking them to use an olderapi_version
kwarg formake_strategies_namespace()
whilst they wait on us to update one variable upstream for a very trivial change 馃 I'm really not sure and might be something to just see how future array API iterations play out first.It's also worth noting that
numpy.array_api
will most likely be removed at some point (numpy/numpy#25542) and superseded by the newarray-api-strict
third-party package. As that library is "official" and pretty stable (esp. for the purposes of whatextra.array_api
does), is that something we'd want for CI @Zac-HD? If so, I can try work out how to do CI things with some direction.There's also
array-api-compat
which could be a good start in covering libraries other than NumPy (CuPy, PyTorch, Dask, possibly more in the future). I'll be having a think on it generally for the array API test suite (data-apis/array-api-tests#235) anywho.