Matthew Barber

I'd be interested in this for the test suite too, where we [utilise](https://github.com/data-apis/array-api-tests/blob/master/array_api_tests/stubs.py) `signatures` already. Might help us with https://github.com/data-apis/array-api-tests/issues/107. I imagine this could be a completely automated process from...

I'd propose we first work on setting up a workflow where the [signatures folder](https://github.com/data-apis/array-api/tree/main/spec/API_specification/array_api) is uploaded as-is, available at say the `.draft` namespace. Then we could get onto uploading versioned...

Like `can_cast()` or `result_type()`, could these utils take both dtypes and arrays? I'd personally want these utils for dtype objects themselves, but definitely my own use cases are not quite...

A minor pro of dtype sets is that it could be a way for a library to communicate what dtypes they support—thinking of PyTorch and it only supporting `uint8` unsigned...

Thanks for the suggestions! Other than an updated TODO list, think I'm now happy with the API and general architecture things, but something to sleep on. I updated the `tests/array_api/`...

So in regards to `xp.__array_api_version__` https://github.com/HypothesisWorks/hypothesis/pull/3456#discussion_r976412249, the PR now uses that for inferrence, and ditches the funky `x.__array_namespace__()` method (save [here](https://gist.github.com/honno/6531d1e8d1acef9b3ef713200c76d91c) in case). Tests have been updated accordingly. For the...

Thanks for the great review comments, should be all addressed now. > I think after this round we're probably just waiting for a decision on complex-dtype introspection? Technically this PR...

Awesome, thanks again Zac!

FWIW `dataframe_protocol_tests.py` and the relevant dunder-df tests in [`dataframe-interchange-tests`](https://github.com/data-apis/dataframe-interchange-tests) pass for this PR locally. (If the bad gateway stuff gets resolved, note there's other issues with interchange stuff on win/mac...

> It should be specified for all types where casting to complexes is defined Yeah that makes sense, my only concern is it might end up looking slightly inconsistent. Currently:...