Justus Magin

that would be ideal, yes. However, I think there are a couple of issues that prevent us from implementing it the way you suggested (I think): 1. `xp.result_type` returns the...

All the considerations above can be satisfied by ```python def _future_array_api_result_type(*arrays_and_dtypes, weakly_dtyped, xp): dtype = xp.result_type(*arrays_and_dtypes, *weakly_dtyped) if weakly_dtyped is None or is_object(dtype): return dtype possible_dtypes = { complex: "complex64",...

as a summary of the discussion we just had, and so I don't forget in the time until I implement this (cc @shoyer): - splitting out weakly-dtyped values from dtypes...

I decided to do this now rather than later. Good news is that this is finally ready for a review and possibly even merging (cc @shoyer). Edit: ~actually, no. Trying...

if my most recent changes are fine, this should be ready for merging (the remaining upstream-dev test failures will be fixed by #9081). Once that is done, I will cut...

I agree, the example with `center=True` from the original post now returns ```python In [1]: import xarray as xr ...: import numpy as np ...: ...: data = xr.DataArray( ...:...

indeed, the best way to resolve this would involve having true missing value support in our in-memory arrays (and adapting `skipna` to do the right thing for those) There's two...

I wonder if, instead of describing code from a third-party package, we should add a "see also" or similar to the section? I.e. more prominent than an entry in the...

I don't have any knowledge about `jax` specifically, but packages like `cubed` have been wrapped successfully by `xarray` using the existing Array API support. So I'd say that as long...

> so it sounds like Xarray should have compliance with the Array API? at least as far as dispatching to wrapped arrays goes, yes. The full Array API is not...