Support slices when using the tuple renaming shortcut in `sel`/`isel`?

[OriolAbril]

What is your issue?

I generally want indexed objects to have different dimension names than their original ones, so index quite often with .sel(original_dim_name=("new_dim_name", indexing_array)) instead of creating a dataarray for indexing_array with "new_dim_name" as dimension.

This doesn't seem to work with slice objects though: .sel(original_dim_name=("new_dim_name", slice(4, None, 2))) errors out with

ValueError: Variable None: Could not convert tuple of form (dims, data[, attrs, encoding]): ('new_dim_name', slice(4, None, 2)) to Variable.

I am not sure if the tuple syntax is discouraged and should be avoided altogether or if it would be possible to support that. From the error message and the indexing rules it seems whichever implementation actually happens for the conversion of slices into arrays could kick in for the 2nd element in the case of tuples and before the creation of the Variable object in order to support this case.

[dcherian]

tuples get converted to Variables which then triggers either outer / vectorized indexing or a combo of that. The error you see is because slices don't convert to an array, so they can't be put in a variable

Is there a reason .sel().rename() is not good enough? The indexing API is fairly complex and hard to teach already...