[QST] Updating indices creation for split

I am interested in using GridSearchCV machinery from sklearn with non-numpy libraries like cuDF or dask.array. In exploring this work, I think functions like _iter_test_indices and _iter_test_masks could be amended slightly to return slices instead of numpy arrays.

For example in _iter_test_indices: https://github.com/scikit-learn/scikit-learn/blob/ec35ed226ca104e841238f2fac24269c2f9f2730/sklearn/model_selection/_split.py#L423-L436

If there is a shuffle, much of the code could be encapsulated in the shuffle conditional. And when there is no shuffle, we can yield the appropriate slice object: slice(0, 71)…

I wanted to check here before proceeding with the work and/or if folks have any advice before I begin?

[amueller]

we kinda only need to distinguish it from an integer or None, right? We could check for a single common method existing. If we then try to use some obscure method and it’s not implemented, well that’s a reasonable error. If it has a uniform method it’s much more likely to be something that implements sampling than an integer, I think 😉

[ogrisel]

It would be great if someone wanted to try to prototype Array API spec support in scikit-learn Cross-Validation tools in a draft PR on top of the #22554 PR.