BUG: pandas coerces `np.longlong` to `np.int64`

[randolf-scholz]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import numpy as np
import pandas as pd

s = pd.Series([1,2,3])
assert type(s.values.astype(np.ulonglong)[0]) == np.ulonglong  # ✔
assert type(s.astype(np.ulonglong).iloc[0]) == np.ulonglong  # ✔

assert type(s.values.astype(np.longlong)[0]) == np.longlong  # ✔
assert type(s.astype(np.longlong).iloc[0]) == np.longlong  # ✘

Issue Description

numpy longlong and int64 are separate data types, but pandas seems to coerce longlong to int64. Surprisingly the same does not happen for the unsigned variant ulonglong, which seems to indicate that this in unintentional behavior.

Expected Behavior

It should respect numpy.longlong data type

Installed Versions

INSTALLED VERSIONS
------------------
commit           : 0f437949513225922d851e9581723d82120684a6
python           : 3.11.3.final.0
python-bits      : 64
OS               : Linux
OS-release       : 5.19.0-46-generic
Version          : #47~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Wed Jun 21 15:35:31 UTC 2
machine          : x86_64
processor        : x86_64
byteorder        : little
LC_ALL           : None
LANG             : en_US.UTF-8
LOCALE           : en_US.UTF-8

pandas           : 2.0.3
numpy            : 1.24.3
pytz             : 2023.3
dateutil         : 2.8.2
setuptools       : 68.0.0
pip              : 23.2
Cython           : 0.29.36
pytest           : 7.4.0
hypothesis       : None
sphinx           : 7.0.1
blosc            : None
feather          : None
xlsxwriter       : None
lxml.etree       : 4.9.3
html5lib         : None
pymysql          : 1.4.6
psycopg2         : None
jinja2           : 3.1.2
IPython          : 8.14.0
pandas_datareader: None
bs4              : 4.12.2
bottleneck       : None
brotli           : None
fastparquet      : 2023.7.0
fsspec           : 2023.6.0
gcsfs            : None
matplotlib       : 3.7.2
numba            : 0.57.1
numexpr          : 2.8.4
odfpy            : None
openpyxl         : 3.1.2
pandas_gbq       : None
pyarrow          : 12.0.1
pyreadstat       : None
pyxlsb           : None
s3fs             : None
scipy            : 1.11.1
snappy           : None
sqlalchemy       : 1.4.49
tables           : 3.8.0
tabulate         : 0.9.0
xarray           : 2023.7.0
xlrd             : None
zstandard        : None
tzdata           : 2023.3
qtpy             : None
pyqt5            : None

[jbrockmendel]

numpy longlong and int64 are separate data types

how many bits are in longlong?

Pretty much all of the pandas cython code is written in terms of int64_t, float32_t, etc. i.e. there are no specializations for longlong. If we do stop coercing longlong->int64, many other operations will subsequently fail with TypeError. Not necessarily a reason to coerce, but something to be aware of.

[randolf-scholz]

@jbrockmendel According to the documentation numpy.longlong is an integer type compatible with C long long, whereas numpy.int64 is an alias for one of numpy.byte, numpy.short, numpy.intc, numpy.int_ and numpy.longlong, which might differ depending on the platform.

On Linux we have:

import numpy as np

print(np.int64 is np.int_)  # True
print(np.int64 is np.longlong)  # False

AFAIK, C's long long is at least 64bits, but the actual size is implementation defined and not guaranteed to be consistent across platforms (hence why numpy wants you to use intXX instead.)

In any case, whether one wants to coerce or not, the behavior should be consistent between ulonglong and longlong.

[lithomas1]

Maybe we should raise in astype for both ulonglong and longlong if np.int64/np.uint64 does not alias to it?

[cmp0xff]

Hi, is it still relevant? On my machine (Windows 11 + python 3.14.0 + numpy 2.3.5), np.longlong is np.int64 gives True.

[randolf-scholz]

@cmp0xff np.longlong is platform dependent, on my machine (Ubuntu 24.04 + python 3.13.8 + numpy 2.3.5), np.longlong is np.int64 evaluates to False.