BUG: Series.loc[list] with uint64 keys raises KeyError (converted to floats)

[jasonmahr]

bug = pd.Series([0, 1, 2, 3, 4], index=[7606741985629028552, 17876870360202815256, 13106359306506049338, 8991270399732411471, 8991270399732411471])
assert bug.loc[7606741985629028552]==0
assert bug.loc[17876870360202815256]==1
# throws KeyError: 'None of [[7606741985629028552, 17876870360202815256]] are in the [index]'
bug.loc[[7606741985629028552, 17876870360202815256]]

Problem description

This code throws KeyError even though clearly both keys are present in the index, as the asserts demonstrate. It is essential for the bug that there is a repeated element in the index (not the one being extracted). The bug disappears if the values are smaller; perhaps it's some kind of a type cast thing that makes the values not found?

Expected Output

7.606742e+18 0
1.787687e+19 1

Output of pd.show_versions()

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

commit: None
python: 3.6.6.final.0
python-bits: 64
OS: Linux
OS-release: 4.14.72-68.55.amzn1.x86_64
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: None
LOCALE: None.None

pandas: 0.23.4
pytest: 3.7.3
pip: 18.0
setuptools: 40.2.0
Cython: 0.28.5
numpy: 1.15.0
scipy: 1.1.0
pyarrow: None
xarray: 0.10.8
IPython: 6.5.0
sphinx: 1.7.5
patsy: 0.5.0
dateutil: 2.7.3
pytz: 2018.5
blosc: None
bottleneck: 1.2.1
tables: 3.4.4
numexpr: 2.6.6
feather: None
matplotlib: 2.2.3
openpyxl: 2.5.5
xlrd: 1.1.0
xlwt: 1.3.0
xlsxwriter: 1.0.7
lxml: 4.2.4
bs4: 4.6.3
html5lib: 1.0.1
sqlalchemy: 1.2.11
pymysql: None
psycopg2: None
jinja2: 2.10
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[jorisvandenbossche]

Thanks for the report!

As a workaround, you can pass the list as an index:

In [36]: bug.loc[pd.Index([7606741985629028552, 17876870360202815256])]                                                                                                                                            
Out[36]: 
7606741985629028552     0
17876870360202815256    1
dtype: int64

The reason that it is failing is because the list with values is converted to a float array of labels (this is the default behaviour of np.array(..), but in the pd.Index(..) constructor we have a special path to check for large ints and to automatically infer a uint64 dtype).
So to fix this bug, we need to use this correct inference in the indexing code as well. Closer investigation / PR always welcome!

[oguzhanogreden]

I've looked into this a bit, Thanks to the useful pointer of @jorisvandenbossche, I got the solution ready.

In addition, the fact that this "works" when there's no repeated index actually is pointing to another bug. In that case, the returned data frame has a Float64Index.

bug = pd.Series([0, 1, 2, 3, 4], index=[7606741985629028552, 17876870360202815256, 13106359306506049338, 8991270399732411471, 8991270399732411472])
assert bug.loc[7606741985629028552]==0
assert bug.loc[17876870360202815256]==1

bug.loc[[7606741985629028552, 17876870360202815256]].index

I have an idea for a solution but not sure of its possible implications. I'll look into it a bit and make a PR.