Code for list of partial indexers should be in MultiIndex.get_indexer

[toobaz]

Code Sample, a copy-pastable example if possible

From #27180

In [5]: midx = pd.MultiIndex.from_product([[0, 1], [2, 3]])                                                                                                                                                                                   

In [6]: midx.get_loc(0)                                                                                                                                                                                                                       
Out[6]: slice(0, 2, None)

In [7]: midx.get_indexer([0])                                                                                                                                                                                                                 
Out[7]: array([-1])

Problem description

Indexing with lists of partial keys works (except for #18345 )

In [5]: midx = pd.MultiIndex.from_product([[0, 1], [2, 3]])                                                                                                                                                                                   

In [6]: midx.get_loc(0)                                                                                                                                                                                                                       
Out[6]: slice(0, 2, None)

In [7]: midx.get_indexer([0])                                                                                                                                                                                                                 
Out[7]: array([-1])

but it is not implemented in get_indexer() - where it should be.

Expected Output

slice(0, 2, None)

Output of pd.show_versions()

INSTALLED VERSIONS

commit : 7ceefb3
python : 3.7.3.candidate.1
python-bits : 64
OS : Linux
OS-release : 4.9.0-9-amd64
machine : x86_64
processor :
byteorder : little
LC_ALL : None
LANG : it_IT.UTF-8
LOCALE : it_IT.UTF-8

pandas : 0.25.0.dev0+861.g7ceefb3f2.dirty
numpy : 1.16.4
pytz : 2016.7
dateutil : 2.8.0
pip : 9.0.1
setuptools : 41.0.1
Cython : 0.29.2
pytest : 4.6.3
hypothesis : 3.71.11
sphinx : 1.4.9
blosc : None
feather : None
xlsxwriter : 0.9.6
lxml.etree : 4.3.2
html5lib : 0.999999999
pymysql : None
psycopg2 : None
jinja2 : 2.10.1
IPython : 7.5.0
pandas_datareader: 0.2.1
bs4 : 4.5.3
bottleneck : 1.2.1
fastparquet : None
gcsfs : None
lxml.etree : 4.3.2
matplotlib : 3.0.2
numexpr : 2.6.9
openpyxl : 2.3.0
pandas_gbq : None
pyarrow : None
pytables : None
s3fs : None
scipy : 1.1.0
sqlalchemy : 1.0.15
tables : 3.4.4
xarray : None
xlrd : 1.0.0
xlwt : 1.3.0
xlsxwriter : 0.9.6

[toobaz]

In fact, it seems to be in MultiIndex.get_locs. Which seems to do a subset of what MultiIndex.get_indexer does, with the only difference that it raises on missing keys. Any objection to deprecating it?

[toobaz]

According to the docstring for get_locs, it accepts

• label (duplicates get_loc)
• slice (not supported by other indexes, implemented directly in indexing code)
• list (duplicates get_indexer)
• mask - doesn't work, but list of masks does (hardly useful - equivalent to calling np.where on each item and combining)

[jbrockmendel]

Which seems to do a subset of what MultiIndex.get_indexer does

A lot of the action also goes on in MultiIndex._convert_listlike_indexer, which handles (some) cases where we are indexing only along the first level of the MultiIndex. In those cases, get_indexer would raise. i.e. for MultiIndex get_indexer does not behave like a vectorized get_loc

mi = pd.MultiIndex.from_product([range(3), ["A", "B"]])
key = mi.get_level_values(0)
>>> mi.get_indexer(key)
array([-1, -1, -1, -1, -1, -1])

If we edit MultiIndex.get_indexer to behave like vectorized get_loc, that breaks MultiIndex.reindex and set operations, for which we dont want to allow partial indexing