BUG: MultiIndex.get_loc errors if get_loc on level returns a slice

[shoyer]

Code Sample, a copy-pastable example if possible

With the dev version of pandas:

In [1]: import pandas as pd

In [2]: index = pd.date_range('2001-01-01', periods=100)

In [3]: mindex = pd.MultiIndex.from_arrays([index])

In [4]: mindex.get_loc('2001-01')
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-4-1914bb512715> in <module>
----> 1 mindex.get_loc('2001-01')

~/dev/pandas/pandas/core/indexes/multi.py in get_loc(self, key, method)
   2257
   2258         if not isinstance(key, tuple):
-> 2259             loc = self._get_level_indexer(key, level=0)
   2260             return _maybe_to_slice(loc)
   2261

~/dev/pandas/pandas/core/indexes/multi.py in _get_level_indexer(self, key, level, indexer)
   2525                 return locs
   2526
-> 2527             i = labels.searchsorted(code, side='left')
   2528             j = labels.searchsorted(code, side='right')
   2529             if i == j:

~/dev/pandas/pandas/util/_decorators.py in wrapper(*args, **kwargs)
    175                 else:
    176                     kwargs[new_arg_name] = new_arg_value
--> 177             return func(*args, **kwargs)
    178         return wrapper
    179     return _deprecate_kwarg

~/dev/pandas/pandas/core/indexes/frozen.py in searchsorted(self, value, side, sorter)
    181         # xref: https://github.com/numpy/numpy/issues/5370
    182         try:
--> 183             value = self.dtype.type(value)
    184         except ValueError:
    185             pass

TypeError: int() argument must be a string, a bytes-like object or a number, not 'slice'

Problem description

This appears to have been introduced by #22230 by @toobaz, which deleted a check for isinstance(loc, slice). I'm not quite sure why, though it looks like this line didn't have test coverage.

I could not figure out how to trigger this directly on a pandas.Series. Indexing like series.loc['2001-01'] appears to go through a different code path not involving get_loc.

Expected Output

With pandas 0.23.4, this returns slice(0, 31, None) (which is the correct result)

Output of pd.show_versions()

INSTALLED VERSIONS

commit: 2f6d682
python: 3.7.1.final.0
python-bits: 64
OS: Darwin
OS-release: 18.2.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8

pandas: 0.23.0.dev0+2160.g2f6d682a0
pytest: 4.0.1
pip: 18.1
setuptools: 40.6.2
Cython: 0.29.1
numpy: 1.15.4
scipy: None
pyarrow: None
xarray: 0.11.0
IPython: 7.1.1
sphinx: None
patsy: None
dateutil: 2.7.5
pytz: 2018.7
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: 3.0.2
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml.etree: None
bs4: None
html5lib: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: None
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None
gcsfs: None

[toobaz]

Indeed I made a mistake in not considering the possibility that single levels might return more than one code for a single key.

@shoyer were you able to reproduce with anything else than 1-level MultiIndexes where the first level has dates? (partial date indexing is the only case I can think of in which a single key corresponds to multiple locations having different labels)

[toobaz]

As suggested by @jorisvandenbossche , IntervalIndex can have the same behavior... but yes, I think this is specific to 1-level MultiIndexes. Will write a PR making such cases follow the "list of labels" code path (depending on codes identified).