TypeError in Index.get_indexer_non_unique if Index.is_all_dates

[toobaz]

Code Sample, a copy-pastable example if possible

In [2]: idx = pd.DatetimeIndex(['2012-02-01', '2012-02-02']).append(pd.Index(['not date']))[:2]

In [3]: idx.get_indexer_non_unique(['2012-02-01'])
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-3-df568b5df150> in <module>()
----> 1 idx.get_indexer_non_unique(['2012-02-01'])

/home/pietro/nobackup/repo/pandas/pandas/core/indexes/base.py in get_indexer_non_unique(self, target)
   2797 
   2798         if self.is_all_dates:
-> 2799             self = Index(self.asi8)
   2800             tgt_values = target.asi8
   2801         else:

/home/pietro/nobackup/repo/pandas/pandas/core/indexes/base.py in __new__(cls, data, dtype, copy, name, fastpath, tupleize_cols, **kwargs)
    351                          **kwargs)
    352         elif data is None or is_scalar(data):
--> 353             cls._scalar_data_error(data)
    354         else:
    355             if tupleize_cols and is_list_like(data) and data:

/home/pietro/nobackup/repo/pandas/pandas/core/indexes/base.py in _scalar_data_error(cls, data)
    702         raise TypeError('{0}(...) must be called with a collection of some '
    703                         'kind, {1} was passed'.format(cls.__name__,
--> 704                                                       repr(data)))
    705 
    706     @classmethod

TypeError: Index(...) must be called with a collection of some kind, None was passed

Problem description

Not a great example, but this can bite more reasonable code. It is due to the fact that asi8 is used on a non-DatetimeIndex. In fact, asi8 should really disappear in the non-datetime-specific parts of the code base.

Expected Output

(array([0]), array([]))

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.5.3.final.0
python-bits: 64
OS: Linux
OS-release: 4.9.0-5-amd64
machine: x86_64
processor:
byteorder: little
LC_ALL: None
LANG: en_GB.UTF-8
LOCALE: en_GB.UTF-8

pandas: 0.22.0.dev0+448.gdbec3c92e
pytest: 3.0.6
pip: 9.0.1
setuptools: 33.1.1
Cython: 0.25.2
numpy: 1.12.1
scipy: 0.18.1
pyarrow: None
xarray: None
IPython: 5.2.2
sphinx: None
patsy: 0.4.1+dev
dateutil: 2.6.0
pytz: 2016.10
blosc: None
bottleneck: 1.2.0
tables: 3.3.0
numexpr: 2.6.1
feather: 0.3.1
matplotlib: 2.0.0
openpyxl: 2.3.0
xlrd: 1.0.0
xlwt: 1.2.0
xlsxwriter: None
lxml: 3.7.1
bs4: 4.5.3
html5lib: 0.999999999
sqlalchemy: 1.0.15
pymysql: None
psycopg2: None
jinja2: 2.8
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[jreback]

hmm, I think we could remove the check from is_all_dates if its object dtype entirely (it should just return False). if we just use inferred type then this is not a problem (not sure how much this might break though).

[jreback]

this is pretty much a result of #19204

[jreback]

so will leave this open, but it is going to be tied to fixing that one.

[toobaz]

this is pretty much a result of #19204

Not really (unless I misunderstood your comment there): #19204 affects Timedelta and Period, while this affects only Timestamp. If #19204 is fixed, then this will affect all of them.

[jreback]

my point is we do not want to infer object type and simply use the inferred property (and include for period timedelta and timestamp)

then this issue is moot

[jbrockmendel]

This appears fixed on master, needs test.