@sgasse maybe you want to have a look at it. I had not realized that gh-15852 added these arguments to the method, but we forgot to also modify the documentation of np.ndarray.mean (i.e. the array attribute/method), which should be in _add_newdocs.py.
Also we use now in mean and var:
umr_any(ddof >= rcount, axis=None)
instead of just ddof >= rcount. This is fairly slow (about 30% of the call-time for tiny arrays). Just as all of the rest of the code, it should probably be ddof >= rcount if where is True else umr_any(ddof >= rcount, axis=None) to avoid most of the new overhead incurred by gh-15852.
@sgasse maybe you want to have a look at it. I had not realized that gh-15852 added these arguments to the method, but we forgot to also modify the documentation of
np.ndarray.mean(i.e. the array attribute/method), which should be in_add_newdocs.py.Also we use now in
meanandvar:instead of just
ddof >= rcount. This is fairly slow (about 30% of the call-time for tiny arrays). Just as all of the rest of the code, it should probably beddof >= rcount if where is True else umr_any(ddof >= rcount, axis=None)to avoid most of the new overhead incurred by gh-15852.