On Mar 5, 8:31 pm, Steven D'Aprano <[EMAIL PROTECTED] cybersource.com.au> wrote: > On Wed, 05 Mar 2008 21:05:31 -0500, Terry Reedy wrote: > > If I understand your question, classes are not singletons: > >>>> ll=[] > >>>> for i in range(2): > > import string > > ll[i]=string > > Where's the IndexError? :-) > > >>>> ll[0] is ll[1] > > True > > But yes, modules are singletons in that way, at least if you go through > the import mechanism. > > >>>> for i in range(2): > > class C: pass > > ll[i] = C > > >>>> ll[0] is ll[1] > > False > > Ah, but each time around the loop you create a *new class* that just > happens to be called C. An alternative way to see similar behaviour is: > > def foo(x=None): > class C(object): > X = x > return C > > Naturally foo() is foo() gives False -- although both classes are called > C, they are different classes that just happen to have the same state. > > I accept my question about classes being singletons is not well-formed, > not even in my own mind. I guess one way of asking is, for any two class > objects (not instances) C1 and C2, does "C1 == C2" imply "C1 is C2"?
>>> class metaC( type ): ... def __eq__( self, other ): ... return random.choice([ True, False ]) ... >>> class C( metaclass= metaC ): ... pass ... >>> import random >>> C==C True >>> C==C False >>> C==C False >>> C==C True >>> C==C False >>> You made me think of this. What relevance it has is open to debate. ...In a separate thread. http://en.wikipedia.org/wiki/Relevance It's my Signal-channel signal signal. -- http://mail.python.org/mailman/listinfo/python-list
