any() for object-dtype arrays returns non-booleans

The any function seems to not deal well for arrays of dtype object. According to the documentation it should return a boolean. However, consider the following code:

print(numpy.array([5, None]).any())

To my surprise, this prints 5, rather than True. It seems to always return the first element of which the boolean value is True, or, if no such elements exist, the last element. (This means that numpy.array([False, None]).any() evaluates to None.)

Appears to happen in Python 3.6.1, Numpy 1.12.1, on Linux, as well as Python 2.7.10, numpy 1.9.2. on Windows (10).

[rkern]

I think we should just document the current behavior. .all() is essentially defined as logical_or.reduce(), and logical_or’s object implementation chose by design to replicate Python’s or shortcut semantics.

[acdr000]

Editing the documentation is fine as well, of course. But I’d still find it a little bit funky that array([5,2]).any() would evaluate to something other than array([5,2], dtype=object).any().