Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!ucsd!ucselx!bionet!hayes.fai.alaska.edu!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: <1474@anaxagoras.ils.nwu.edu>
Date: 23 Aug 90 16:41:52 GMT
References: <81084@aerospace.AERO.ORG> <41670@think.Think.COM> <276@paradigm.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: 36
In-reply-to: gjc@paradigm.com

In article <276@paradigm.com>, gjc@paradigm writes:
>In article <41670@think.Think.COM>, barmar@think.com (Barry Margolin) writes:
>> Some Common Lisp implementations ...
>>  If the only reference to an object is as the key in such a
>> hash table it may be removed from the hash table (along with the value)
>> and may then be GCed.
>
>You *have* to mention that that would only be reasonable for EQ hash tables.
>
>Not for EQUAL hash tables, or any other predicate you can think of.


Not necessarily.

First of all, the behavior being discussed is only an option and is not
correct default behavior in any case.  Given this, it's clear that a user
may decide that he/she wants this behavior for EQUAL tables as well as for
EQ ones.  For example, if someone was using EQUAL hash tables to cache
computation results in a backward-chaining system (say, mapping propositions
onto the leaves of a proof tree that need to be true for the proposition to
be true) then the programmer may decide that the savings in computation is
only worthwhile in cases when there is no memory tradeoff (i.e., when the
structures have to be kept in memory for other reasons already), but when
they can be GC'ed then it's worth redoing the computation to keep memory
size down.

Regardless of whether you agree with my example here (which was a bit off
the cuff), it seems clear that this is a feature which would be an option,
and not the default behavior, and which a programmer may very well want as
an option for non-EQ-based hash tables.


Bruce Krulwich
Institute for the Learning Sciences