loc misbehaves when Period is at start of 3-level MultiIndex

[mgoodhand]

Code Sample

>>> pbc = pd.DataFrame({ 'p': [pd.Period('2017')], 'b':['b1'], 'c': ['c1'], 'v': [1]}).set_index(['p','b','c'])
>>> pbc.loc[(pd.Period('2017'), 'b1', 'c1'), 'v']
b   c
b1  c1    1
Name: v, dtype: int64
>>> pbc.loc[(pd.Period('2017'), 'wibble', 'c1'), 'v']
b   c
b1  c1    1
Name: v, dtype: int64
>>> pbc.loc[(pd.Period('2017'), 'b1', 'wobble'), 'v']
b   c
b1  c1    1
Name: v, dtype: int64

Problem description

I'd expect loc to return a scalar value when there's a unique match, rather than a Series with one item in it.

I'd also expect a KeyError when the second and third levels of the index don't match (the 'wibble' and 'wobble' selections). Instead it appears that the additional levels are ignored, and matches succeed or fail based on the period alone.

Expected Output

Everything works as I expect when I use strings instead of Periods:

>>> ybc = pd.DataFrame({ 'y': ['2017'], 'b':['b1'], 'c': ['c1'], 'v': [1]}).set_index(['y','b','c'])
>>> ybc.loc[('2017', 'b1', 'c1'), 'v']
1
>>> ybc.loc[('2017', 'b1', 'wobble'), 'v']
...
KeyError: ('2017', 'b1', 'wobble')

Things also work as I'd expect when there are only two levels in the MultiIndex (or four, or five):

>>> pb = pd.DataFrame({ 'p': [pd.Period('2017')], 'b':['b1'], 'v': [1]}).set_index(['p','b'])
>>> pb.loc[(pd.Period('2017'), 'b1'), 'v']
1
>>> pb.loc[(pd.Period('2017'), 'wibble'), 'v']
...
KeyError: (Period('2017', 'A-DEC'), 'wibble')

It also works if I put Period last in the three-level MultiIndex:

>>> bcp = pd.DataFrame({ 'p': [pd.Period('2017')], 'b':['b1'], 'c': ['c1'], 'v': [1]}).set_index(['b','c', 'p'])
>>> bcp.loc[('b1', 'c1', pd.Period('2017')), 'v']
1
>>> bcp.loc[('b1', 'wibble', pd.Period('2017')), 'v']
...
KeyError: ('b1', 'wibble', Period('2017', 'A-DEC'))

... or second:

>>> bpc = pd.DataFrame({ 'p': [pd.Period('2017')], 'b':['b1'], 'c': ['c1'], 'v': [1]}).set_index(['b','p','c'])
>>> bpc.loc[('b1', pd.Period('2017'), 'c1'), 'v']
1
>>> bpc.loc[('b1', pd.Period('2017'), 'wibble'), 'v']
...
KeyError: ('b1', Period('2017', 'A-DEC'), 'wibble')

Output of pd.show_versions()

[paste the output of pd.show_versions() here below this line]

INSTALLED VERSIONS

commit: None
python: 2.7.10.final.0
python-bits: 64
OS: Darwin
OS-release: 17.4.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_GB.UTF-8
LOCALE: None.None

pandas: 0.22.0
pytest: None
pip: 9.0.3
setuptools: 39.0.1
Cython: None
numpy: 1.14.2
scipy: None
pyarrow: None
xarray: None
IPython: None
sphinx: None
patsy: None
dateutil: 2.7.2
pytz: 2018.4
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: None
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: 2.10
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: 0.5.0

[jreback]

periods in a multiindex are not very well tested / supported for indexing. you are welcome to submit a patch.

[proost]

May i take a look?

[TomAugspurger]

Yep, doesn't look like anyone else is working on this.

…

On Fri, Aug 16, 2019 at 12:27 AM proost ***@***.***> wrote: May i take a look? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#20684?email_source=notifications&email_token=AAKAOISLBI5HOZUAHYPOAPDQEY3EFA5CNFSM4E2NHBM2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4NWIBQ#issuecomment-521888774>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAKAOIUTCCOEZQENBFROKSTQEY3EFANCNFSM4E2NHBMQ> .

[TomAugspurger]

This was fixed by #28628