[Python] How do I stop pandas.aggregate(engine='numba') downcasting uint64 to int64?

Is there a way to prevent recasting the data from uint64 to int64?

NumbaTypeSafetyWarning correctly foretells a datatype overflow


C:\apps\mapFolding\.venv\Lib\site-packages\pandas\core\_numba\executor.py:82:
NumbaTypeSafetyWarning: unsafe cast from uint64 to int64. Precision may be lost.
na_positions = np.array(na_pos)

Relevant code


dictionaryCurveLocations: dict[int, int]

datatypeCurveLocations = numpy.uint64
datatypeDistinctCrossings = numpy.uint64

dataframeAnalyzed = pandas.DataFrame({
'analyzed': pandas.Series(name='analyzed'
, data=list(dictionaryCurveLocations.keys())
, dtype=datatypeCurveLocations)
, 'distinctCrossings': pandas.Series(name='distinctCrossings'
, data=list(dictionaryCurveLocations.values())
, dtype=datatypeDistinctCrossings)
}, dtype=datatypeCurveLocations
)

dataframeAnalyzed = dataframeAnalyzed.groupby('analyzed', sort=False)['distinctCrossings'
].aggregate('sum'
, engine='numba'
, engine_kwargs={'nogil': True, 'nopython': True, 'parallel': True}
).reset_index()

Environment


(.venv) C:\apps\mapFolding>py -VV
Python 3.13.7 (tags/v3.13.7:bcee1c3, Aug 14 2025, 14:15:11) [MSC v.1944 64 bit (AMD64)]

(.venv) C:\apps\mapFolding>pip show numba
Name: numba
Version: 0.61.2

(.venv) C:\apps\mapFolding>pip show pandas
Name: pandas
Version: 2.3.2

(.venv) C:\apps\mapFolding>ver

Microsoft Windows [Version 10.0.26100.5074]

Continue reading...