class members vs instance members

[hdixon]

Bruno Desthuilliers wrote:
> hdixon a écrit :
> > Ive spent a few days going thru a couple of Python tutorials. No
> > problem until I got to classes. I guess my java mindset is blocking my
> > vision.
>
> Then http://dirtsimple.org/2004/12/python-is-not-java.html
>
> > I've borrowed another thread's code snippet and cannot explain
> > the results:
> > class MyClass:
>
>   class MyClass(object):
>
> >     list = []
>
> This may not be too harmful in the given context, nut shadowing builtin
> types may not be a good idea.
>
> >     myvar = 10
> >
> >     def add(self, x):
> >         self.list.append(x)
>
> "list" is a class attribute. It's shared by all instances of MyClass.
> This is a FAQ AFAICT.
>
> >         self.myvar = x
>
> This creates an instance attribute named myvar that shadows the class
> attribute of the same name
>
> (snip expected results)
>
> > WHY do the "class members" list and myvar seem to behave differently?
>
> cf above. And notice that in Python, binding (aka 'assignement') and
> name lookup on objects are different beasts. There have been quite a few
> threads about this recently.
>
> > Given the way that list[] is supposedly shared why doesnt myvar exhibit
> > the same behavior?
>
> Because binding to self.myvar creates the instance attribute "myvar".
> Notice that you have not rebound "list".
>
> > It seems that myvar is acting like a true "instance
> > member".
>
> It is one.
>
> > Is this simply because of the underlying types?
>
> Nope, it's because in the first case (MyClass.list, accessed as
> self.list) you just call methods on the list object, which has no effect
> on the binding.

ok - i "think" its beginning to dawn on me. BTW - the blog above was a
good read for me. I _am_ trying to bring java into the mix and its
screwing me up big time. I'll go back over the chapters that dicuss
binding/name lookups.


Thanks