Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!accuvax.nwu.edu!anaxagoras!krulwich
From: krulwich@ils.nwu.edu (Bruce Krulwich)
Newsgroups: comp.lang.lisp
Subject: Re: Weak Reference in Common Lisp
Message-ID: <1436@anaxagoras.ils.nwu.edu>
Date: 21 Aug 90 14:10:28 GMT
References: <81084@aerospace.AERO.ORG> <41670@think.Think.COM>
Sender: news@anaxagoras.ils.nwu.edu
Reply-To: krulwich@ils.nwu.edu (Bruce Krulwich)
Distribution: usa
Organization: Institute for the Learning Sciences, Northwestern University, Evanston, IL 60201
Lines: 29
In-reply-to: barmar@think.com (Barry Margolin)

Regarding weak reference:

In article <41670@think.Think.COM>, barmar@think (Barry Margolin) writes: 
>I think they're only feasible in cells that are allowed to reference
>nothing.  Hash table entries meet this requirement because entries may be
>removed; symbol value cells and CLOS slots can be "unbound", so weak
>references are possible there.  But it wouldn't make sense to have a weak
>reference from a cons cell -- what does CAR return if the object had been
>GCed (I suppose you could wimp out and say that cells that aren't allowed
>to be unbound must be set to NIL)?

This problem becomes easier to manage if dereferencing a weak pointer had to
be done explicitly, similar to T's OBJECT-HASH and OBJECT-UNHASH.
Abstracting away from T's version for a minute, you could have:

        (MAKE-WP <obj>) -- return a weak pointer to <obj>
        (WP-DEREF <wp>) -- return the value pointed to (weakly) by <wp>

Then the question would not be "what should be done to <wp> when its <obj>
is GC'ed," as barmar seems to say above, but rather "what should WP-DEREF
return when <wp>'s <obj> has been GC'ed."  One answer to this question is to
model WP-DEREF after GETHASH, and in the case where <wp>'s <obj> has been
GC'ed have it return NIL or a user-specified DEFAULT.


Bruce Krulwich
krulwich@ils.nwu.edu