API: Restore getting name from MultiIndex level

[TomAugspurger]

xref https://issues.apache.org/jira/browse/ARROW-6922 /
#27242 (comment)
/ #29032

No docs yet, since it isn't clear how this will eventually sort out. But
we at least want to preserve this behavior for 1.0

cc @jorisvandenbossche @topper-123

[jreback]

what is exactly the error from pyarrow?

[topper-123]

I'm working on the implementation discussed in #27242, including new to be deprecated levels with names. Won't that fix this issue?

[jorisvandenbossche]

I'm working on the implementation discussed in #27242, including new to be deprecated levels with names. Won't that fix this issue?

As I mentioned in the other issue, IMO we should first properly discuss if we actually want to deprecate .levels
And in general, I would prefer to not have master in a "broken" state (meaning that we removed something that will only be temporary). For projects trying to be good citizens and testing against pandas master that is a bit annoying.

what is exactly the error from pyarrow?

pyarrow is using the name of a level, and this is now no longer preserved.

[TomAugspurger]

Let's merge this, and discuss future plans in #29032? This at least gets CI passing for downstream projects.

[jreback]

sgtm.

[jreback]

do we have sufficient tests for downstream though (can be added later in test_downstream.py)

[jorisvandenbossche]

do we have sufficient tests for downstream though (can be added later in test_downstream.py)

I am not sure additional tests are needed. We have tests for it in out own tests. We just broke purposefully our API (and change our own tests to reflect that), so it is to be expected that behaviour that relies on accessing the level names like this will be broken.

[jorisvandenbossche]

@TomAugspurger thanks for the quick PR!

@topper-123 As mentioned above, IMO we should still first have a discussion about a deprecation of .levels. Do you want to open a dedicated issue for that? (or is there already an issue to discuss that?)

[TomAugspurger]

The discussion about the future of .levels can happen in #29032.