Index(data=..., dtype=CategoricalDtype(...)) doesn't maintain dtype

[topper-123]

Code Sample, a copy-pastable example if possible

>>> c = pd.api.types.CategoricalDtype(categories=['German', 'English', 'French'],
...                                   ordered=True)
>>> pd.Index(['German', 'English', 'French'], dtype=c)
CategoricalIndex(['German', 'English', 'French'],
                 categories=['English', 'French', 'German'],
                 ordered=False, dtype='category')

Problem description

Notice that the index isn't ordered and the categories have changed location.

If construction goes through CategoricalIndex, everything is ok.

Expected Output

Expected output is CategoricalIndex(['German', 'English', 'French'], categories=['German', 'English', 'French'], ordered=True).

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.6.3.final.0
python-bits: 32
OS: Windows
OS-release: 10
machine: AMD64
processor: Intel64 Family 6 Model 78 Stepping 3, GenuineIntel
byteorder: little
LC_ALL: None
LANG: None
LOCALE: None.None

pandas: 0.21.1
pytest: 3.2.1
pip: 9.0.1
setuptools: 36.5.0.post20170922
Cython: 0.26.1
numpy: 1.11.3
scipy: 1.0.0
pyarrow: None
xarray: None
IPython: 6.1.0
sphinx: 1.6.5
patsy: None
dateutil: 2.6.1
pytz: 2017.2
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: 2.1.0
openpyxl: 2.4.8
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: 1.0.1
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: 2.9.6
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[jreback]

thought we had an issue for this one, ok, IIRC cc @jschendel @jorisvandenbossche @TomAugspurger
discussed this on the last Cat PR.

[jschendel]

My previous PR's related to CDT were in the context of .astype(CDT), so this case wasn't something that was covered. Fix looks super simple though, dtype just isn't being passed through to the CI constructor here:

pandas/pandas/core/indexes/base.py

Lines 198 to 200 in 04beec7

	if is_categorical_dtype(data) or is_categorical_dtype(dtype):
	from .category import CategoricalIndex
	return CategoricalIndex(data, copy=copy, name=name, **kwargs)

Will open a PR to fix it later today.

[topper-123]

Could you maybe also do the same for IntervalIndex? That is, if the dtype is a pd.api.types.IntervalDtype, the constructor should be passed to IntervalIndex. (this doesn't do anything useful ATM, but as a principle (IMO) if dtype=IntervalDtype() should delegate to IntervalIndex.

This should only be 2 lines more....

[TomAugspurger]

Could you maybe also do the same for IntervalIndex?

FWIW, I'm thinking through this type of issue with #18767

I'd recommend avoiding more

if isinstance(foo, extension_type_1): ...
elif isinstance(foo, extension_type_2): ...

blocks for now.

[jschendel]

I'd recommend avoiding more blocks for now

The block is actually already there, so it's really just a matter of passing dtype through to the IntervalIndex constructor. Won't do anything right now, as the IntervalIndex constructor accepts a dtype kwarg, but never actually uses it.

Planning to open an issue in the near future to allow subtype conversion for IntervalIndex, e.g. converting interval[int64] to interval[float64], which currently isn't possible via the constructor or astype. I believe this same issue would come into play then, so it'd need to be addressed at some point in the near term. Can do it in this PR or later, doesn't really matter to me.

[TomAugspurger]

The block is actually already there, so it's really just a matter of passing dtype through to the IntervalIndex constructor

👍 then.