BUG/ENH: MI.get_level_values and MI.set_levels get confused if level name happens to be int

[charlesdong1991]

While working on a PR, find out a confusion when level names are integer/None, e.g.

>>> mi = pd.MultiIndex.from_frame(pd.DataFrame([[0.5,1],[0.5,0], [0.6, 0], [0.6,1]]), names=[None, 0])
>>> mi
MultiIndex([(0.5, 1),
            (0.5, 0),
            (0.6, 0),
            (0.6, 1)],
           )
>>> mi.get_level_values(0)
Int64Index([1, 0, 0, 1], dtype='int64', name=0)
>>> mi.set_levels([0.2, 0.3], level=0)
MultiIndex([(0.5, 0.3),
            (0.5, 0.2),
            (0.6, 0.2),
            (0.6, 0.3)],
           )

Pandas has quite big tolerrance of names setting, however, if names is set by int, and happens to be valid position, then using set_levels to change level value could be confusing, and also using level name to only change level values for None becomes difficult, because by default, None should be used to set values for all levels, the same happens if names is [0, 1], e.g.

>>> mi = pd.MultiIndex.from_frame(pd.DataFrame([[0.5,1],[0.5,0], [0.6, 0], [0.6,1]]), names=[1, 0])
>>> mi
MultiIndex([(0.5, 1),
            (0.5, 0),
            (0.6, 0),
            (0.6, 1)],
           names=[1, 0])
>>> mi.set_levels([0.2,0.3], level=0)
MultiIndex([(0.5, 0.3),
            (0.5, 0.2),
            (0.6, 0.2),
            (0.6, 0.3)],
           )

And yes, these are really corner cases, but might be nice to resolve to let users explicitly set/get level values! I am not sure what is the best way to figure it out, I am thinking of something like: get_level_values_by_name(0) or get_level_values_by_loc(0), or have an argument in get_level_values, e.g. get_level_values(0, by_name=True), default is False which uses loc to get values.

All feedbacks are very welcome, if there is kind of consensus, i would like to work on it!
@jreback @WillAyd @TomAugspurger @jorisvandenbossche

[TomAugspurger]

I think this is a duplicate of #21677.

[charlesdong1991]

ahh, indeed, thanks! @TomAugspurger

i close this one since there are discussions around already