David a écrit :
(snip)
Out of 'Abc.message' and 'self.message', which is the favoured
convention? It would be very easy to accidentally override
'self.messages' with an instance attribute!
Only use 'Abc.message' if you want to make sure you get the Abc class
'message' attribute - that is, if you want to skip possible overrides in
subclasses. As far as I'm concerned, I'd tend to consider this bad style
unless it's a very very specific implementation point of an abstract
class (in which case it would probably be named '__message' to prevent
accidental override) - but YMMV of course.
Use 'self.__class__.message' (or 'type(self).message') if you want to
make sure you get the class 'message' attribute - that is, if you want
to honor possible overrides in subclasses, but not per-instance
override. But then - at least in your example code - I'd use a classmethod:
class Abc:
message = 'Hello World'
@classmethod
def print_message(cls):
print cls.message
Now the most common idiom - that is, outside classmethods - is to just
use 'self.message'. Someone (even you) might have a very valid reason to
override the 'message' attribute on a per-instance basis. FWIW, if you
start to worry about possible accidental overrides here, then Python
might not be the right language for you - nothing prevents "accidental"
overrides of method and even the class (the '__class__' attribute)
itself - yes, they are all attributes, and you can dynamically override
them !-)
--
http://mail.python.org/mailman/listinfo/python-list
