BUG/API: index.array for RangeIndex, MultiIndex

[jbrockmendel]

ATM MultiIndex.array raises ValueError, since there is no single array actually backing MI. RangeIndex creates and returns a PandasArray. I think RangeIndex should also raise (at least until/unless we implement a RangeArray).

This also has implications for extract_array, which currently raises on MultiIndex, which has led to some ugly checks in e.g. Index._cmp_method.

[jbrockmendel]

@TomAugspurger @jorisvandenbossche thoughts here on adding something like a preserve_mi=False kwarg to extract_array to return unchanged instead of raising?

[jorisvandenbossche]

I don't have a strong opinion about it, but so that would basically the same as:

arr = ...

if not isinstance(arr, MultiIndex):
    arr = extract_array(arr)

So it would save one line of boiler plate in every place where it is needed

[TomAugspurger]

Agreed that extract_array shouldn't materialize a PandasArray from a RangeIndex. IMO, it should only... extract, existing arrays. Not create new ones :)

…

On Tue, Oct 27, 2020 at 2:34 AM Joris Van den Bossche < ***@***.***> wrote: I don't have a strong opinion about it, but so that would basically the same as: arr = ... if not isinstance(arr, MultiIndex): arr = extract_array(arr) So it would save one line of boiler plate in every place where it is needed — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#37277 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAKAOIXGQIAJ6CA7HKHVFKDSMZZYHANCNFSM4SYRTYOQ> .