BUG: CategoricalIndex can't be a boolen mask

[TomAugspurger]

In [1]: import pandas as pd

In [2]: idx = pd.CategoricalIndex([True, False, True])

In [3]: pd.Series(range(3))[idx]
Out[3]:
True    NaN
False   NaN
True    NaN
dtype: float64

Expected:

In [4]: pd.Series(range(3))[idx.astype(object)]
Out[4]:
0    0
2    2
dtype: int64

Fix coming shortly.

[pganssle]

Is this another manifestation of this bug, or a separate bug?

>>> import pandas as pd
>>> pd.__version__
'0.23.3'
>>> pd.CategoricalIndex(categories=[True, False])
ValueError                                Traceback (most recent call last)
...
~/.virtualenvs/pandas/lib/python3.7/site-packages/pandas/core/arrays/categorical.py in _get_codes_for_values(values, categories)
   2431     (_, _), cats = _get_data_algo(categories, _hashtables)
   2432     t = hash_klass(len(cats))
-> 2433     t.map_locations(cats)
   2434     return coerce_indexer_dtype(t.lookup(vals), cats)
   2435 

pandas/_libs/hashtable_class_helper.pxi in pandas._libs.hashtable.PyObjectHashTable.map_locations()

ValueError: Buffer dtype mismatch, expected 'Python object' but got 'unsigned long'

Full traceback under details tag:

<ipython-input-4-c58e7bc5bb0f> in <module>()
----> 1 pd.CategoricalIndex(categories=[True])

~/.virtualenvs/pandas/lib/python3.7/site-packages/pandas/core/indexes/category.py in __new__(cls, data, categories, ordered, dtype, copy, name, fastpath)
    100                 data = []
    101             data = cls._create_categorical(cls, data, categories, ordered,
--> 102                                            dtype)
    103 
    104         if copy:

~/.virtualenvs/pandas/lib/python3.7/site-packages/pandas/core/indexes/category.py in _create_categorical(self, data, categories, ordered, dtype)
    165             from pandas.core.arrays import Categorical
    166             data = Categorical(data, categories=categories, ordered=ordered,
--> 167                                dtype=dtype)
    168         else:
    169             if categories is not None:

~/.virtualenvs/pandas/lib/python3.7/site-packages/pandas/core/arrays/categorical.py in __init__(self, values, categories, ordered, dtype, fastpath)
    368 
    369         else:
--> 370             codes = _get_codes_for_values(values, dtype.categories)
    371 
    372         if null_mask.any():

I'll note that my version of it is a regression that occurred between 0.18.0 and 0.23.0.

[TomAugspurger]

Hmm that looks different.

Two things

1. Index constructors require a data argument, so the proper error is like TypeError: Index(...) must be called with a collection of some kind, None was passed
2. something is up with booleans. ValueError in Categorical Constructor with empty data and boolean categories #22702 for that.