Hi all,
TL;DR: If nobody has concerns, I think we may give always returning
boolean values for `any()` and `all()` a shot soon (for object input).
Today in the triage meeting, and generally once in a while it comes up
that:
object_arr.any()
object_arr.all()
should always return booleans. Currently we have for example:
>>> np.array([None, "string"], dtype=object).any()
'string'
while we return booleans for anything that isn't an object dtype.
Note that our logical reductions, which are used as an implementation:
np.logical_or.reduce()
np.logical_and.reduce()
would still return the old result, so there would be a workaround
available (although maybe hard to find).
One PR for this is:
https://github.com/numpy/numpy/pull/11857
And while I think that may need some tweaks, I think we may push
something similar soon (with plenty of time to test before the release)
if nobody voices concern.
Cheers,
Sebastian
_______________________________________________
NumPy-Discussion mailing list -- [email protected]
To unsubscribe send an email to [email protected]
https://mail.python.org/mailman3/lists/numpy-discussion.python.org/
Member address: [email protected]
