The Python 2 implementation of hasattr
is fairly naive, it just tries to access that attribute and see whether it raises an exception or not.
Unfortunately, hasattr
will eat any exception type, not just an AttributeError
matching the name of the attribute which was attempted to access. It caught a NameError
in the example shown, which causes the incorrect result of False
to be returned there. To add insult to injury, any unhandled exceptions inside properties will get swallowed, and errors inside property code can get lost, masking bugs.
In Python 3.2+, the behavior has been corrected:
hasattr(object, name)
The arguments are an object and a string. The result is True
if the string is the name of one of the object’s attributes, False
if not. (This is implemented by calling getattr(object, name)
and seeing whether it raises an AttributeError
or not.)
The fix is here, but that change didn't get backported.
If the Python 2 behavior causes trouble for you, consider to avoid using hasattr
; instead you can use a try/except around getattr
, catching only the AttributeError
exception type and letting any other exceptions raise unhandled.
hasattr
behaviour may still be considered strange. It's true that arbitrary exceptions inside a property no longer cause the attribute to be considered absent (but rather get raised right in the caller's face), but if anAttributeError
originates from somewhere within the property, the result ofhasattr
will still beFalse
. This may not be intentional. (Or it may, so a property can determine itself whether it wants to be "there".) In any case, executing the property sounds wrong to me in the first place, considering side effects etc. – Anguine