Can't .loc[label] on a CategoricalIndex with labels being integer

[aavanian]

I came around this and this seemed related to #9595 but could also be a bug in the above interacting with the CategoricalIndex. Using the same example as #15470 with pandas 0.20.3:

Code Sample

    s = pd.Series([2, 1, 0], index=pd.CategoricalIndex([2, 1, 0]))
    s[2]  # works (interpreting as label)
    s.loc[2]  # fails with TypeError: cannot do label indexing on <class 'pandas.core.indexes.category.CategoricalIndex'> with these indexers [2] of <class 'int'>

    # of course the below works!
    s = pd.Series([2, 1, 0], index=[2, 1, 0])
    s[2]  # works (interpreting as label)
    s.loc[2]  # works (interpreting as label)

Problem description

s.loc[2] shouldn't fail and should select the label '2'.
Note this happens also with a pd.DataFrame.

[jreback]

duplicate of #15470

[jreback]

why don't you just show this example there (IOW your top section). its the same issue.

[jorisvandenbossche]

Actually, this is about the TypeError that loc raises, not about __getitem__

[jreback]

and its the exact same issue

[jorisvandenbossche]

It's clearly not the same user facing issue, so I suppose you mean "it probably has the same underlying reason/bug for this failure" ? If so, can you give a bit more explanation?

The other issue is not necessarily a bug (although if we decide not to be a bug, very confusing behaviour), it is a API design question where we have to ask whether a categorical index with numerical indices is actually a numerical index or not