BUG: np.cov with rowvar=False returns the wrong shape for N=1 #27658
Description
Describe the issue:
For one observation of three variables, np.cov with rowvar=True returns the following:
>>> x = np.ones((3, 1))
>>> np.cov(x, ddof=0, rowvar=True)
array([[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]])For the same computation laid-out with rowvar=False, the return value is a scalar:
>>> np.cov(x.T, ddof=0, rowvar=False)
array(0.)This should return a 3x3 matrix of zeros, same as the computation above.
The problem is the special-casing of X.shape[0] == 1 in this line, which seems aimed at handling the x.ndim == 1 case:
numpy/numpy/lib/_function_base_impl.py
Lines 2739 to 2740 in 70fde29
It looks like this behavior was introduced in 959f36c, 19 years ago.
Python and NumPy Versions:
$ python -c "import sys, numpy; print(numpy.__version__); print(sys.version)"
2.1.1
3.12.3 (v3.12.3:f6650f9ad7, Apr 9 2024, 08:18:47) [Clang 13.0.0 (clang-1300.0.29.30)]
Runtime Environment:
$ python -c "import numpy; numpy.show_runtime()"
[{'numpy_version': '2.1.1',
'python': '3.12.3 (v3.12.3:f6650f9ad7, Apr 9 2024, 08:18:47) [Clang 13.0.0 '
'(clang-1300.0.29.30)]',
'uname': uname_result(system='Darwin', node='jmdg-macbookpro.roam.internal', release='23.6.0', version='Darwin Kernel Version 23.6.0: Wed Jul 31 20:49:39 PDT 2024; root:xnu-10063.141.1.700.5~1/RELEASE_ARM64_T6000', machine='arm64')},
{'simd_extensions': {'baseline': ['NEON', 'NEON_FP16', 'NEON_VFPV4', 'ASIMD'],
'found': ['ASIMDHP'],
'not_found': ['ASIMDFHM']}}]