Describe the issue:

When using intersect1d where the input arrays are of differing integer types, for example, int64 and uint64. The dtype of the resulting numpy array is float64. This feels like a bug where the correct dtype for the output should be the most general integer type (probably the signed integer).

Reproduce the code example:

import numpy
a = numpy.arange(100, dtype=numpy.int64)
b = numpy.arange(100, dtype=numpy.uint64)
intersect_aa = numpy.intersect1d(a,a)
intersect_bb = numpy.intersect1d(b,b)
intersect_ab = numpy.intersect1d(a,b)

print(intersect_aa.dtype)
print(intersect_bb.dtype)
print(intersect_ab.dtype)

Error message:

No response

Python and NumPy Versions:

1.26.2
3.12.0 | packaged by Anaconda, Inc. | (main, Oct 2 2023, 12:22:05) [Clang 14.0.6 ]

Runtime Environment:

[{'numpy_version': '1.26.2',
'python': '3.12.0 | packaged by Anaconda, Inc. | (main, Oct 2 2023, '
'12:22:05) [Clang 14.0.6 ]',
'uname': uname_result(system='Darwin', node='Arthurs-MacBook-Air.local', release='23.1.0', version='Darwin Kernel Version 23.1.0: Mon Oct 9 21:28:31 PDT 2023; root:xnu-10002.41.9~6/RELEASE_ARM64_T8112', machine='arm64')},
{'simd_extensions': {'baseline': ['NEON', 'NEON_FP16', 'NEON_VFPV4', 'ASIMD'],
'found': ['ASIMDHP'],
'not_found': ['ASIMDFHM']}},
{'architecture': 'armv8',
'filepath': '/Users/arthurtolley/miniconda3/envs/test_env/lib/libopenblasp-r0.3.21.dylib',
'internal_api': 'openblas',
'num_threads': 8,
'prefix': 'libopenblas',
'threading_layer': 'pthreads',
'user_api': 'blas',
'version': '0.3.21'}]
None

Context for the issue:

This bug was preventing the usage of the output of intersect1d from being used to index an array further along our script.

We have enforced explicit types for the inputs to intersect1d in a recent PR: gwastro/pycbc#4586