xarray.rolling_window Converts dims Argument from Tuple to List Causing Issues for Cupy-Xarray

[negin513]

What is your issue?

Hello,
I'm currently working on the development of cupy_xarray, and I've come across an issue with xarray.rolling_window. Specifically, it seems that the dims argument gets automatically converted to a list from a tuple.

While numpy.normalize_axis_tuple can handle lists, thereby not raising errors when using numpy-backed xarray objects, its cupy equivalent (cupy.core.internal._normalize_axis_indices) does not. When a list is passed, it raises the following error:

File cupy/core/internal.pyx:424, in cupy.core.internal._normalize_axis_indices()
TypeError: 'list' object cannot be interpreted as an integer.

It seems the line that might fix this is L2487:

xarray/xarray/core/variable.py

Lines 2487 to 2494 in c9d89e2

	axis = [self.get_axis_num(d) for d in dim]
	new_dims = self.dims + tuple(window_dim)
	return Variable(
	new_dims,
	duck_array_ops.sliding_window_view(
	padded.data, window_shape=window, axis=axis
	),
	)

For the time being, we've managed to devise a workaround in cupy, but I thought it would be worthwhile to bring this to your attention.

Relevant Cupy PR: cupy/cupy#7575
Contributor: @dcherian

[welcome]

Thanks for opening your first issue here at xarray! Be sure to follow the issue template!
If you have an idea for a solution, we would really welcome a Pull Request with proposed changes.
See the Contributing Guide for more.
It may take us a while to respond here, but we really value your contribution. Contributors like you help make xarray better.
Thank you!

[dcherian]

This seems like a real easy fix?

axis = tuple(self.get_axis_num(d) for d in dim)

EDIT: the Array API seems to type axis as Optional[Union[int, Tuple[int, ...]]] pretty consistently, so it seems like we should always pass tuples down to the array library

[negin513]

@dcherian : agreed! But I am afraid it might break other components.
Although numpy seems to be able to handle both tuple and list in normalize_axis_tuple and I cannot see any other issues rising from this: https://github.com/numpy/numpy/blob/f67467c21a1797becde3097661996f60df4080ff/numpy/core/numeric.py#L1328

[dcherian]

I think the only other one is dask, which should also work.

[headtr1ck]

@negin513 would you mind creating a PR for this?

[negin513]

@headtr1ck Thanks. I created the corresponding PR for this : #7938