I'd think that @ISA would be copied to .ISA on object instantiation, and
after that the two wouldn't have anything to do with each other. We could
set up one of those cool copy-on-write locks everyone's been talking about
to save memory too.
Or we could have it default to @ISA if .ISA doesn't exist, which is
basically what Bart's saying, so I guess I agree with him, maybe. (Although
I do think that when .ISA is created, it should get @ISA, whenever that
happens to be.)
-----Original Message-----
From: Bart Lateur [mailto:[EMAIL PROTECTED]]
Sent: Thursday, June 28, 2001 17:59
To: [EMAIL PROTECTED]
Subject: Re: Multiple classifications of an object
On Wed, 27 Jun 2001 13:48:38 -0400, Dan Sugalski wrote:
>And the current @ISA stuff is MI,
>albeit on a per-class basis rather than on a per-object one.
>
>Anyway, as Damian mentioned, setting the .ISA property is a perfectly
>reasonable sort of thing to do if the language supports this.
Just one question. If an object would have both per-object inheritance
(.ISA), and per class inheritance (@ISA), which one would have
precedence? If there's a conflict, a method exists both for a superclass
and for an object superclass (i.e. through .ISA), which list would be
checked first? Which method would be picked, and executed?
FWIW, I think I'd vote for .ISA, as it is more individually tied to this
particular object. The other one is generic.
