Path: utzoo!attcan!uunet!samsung!zaphod.mps.ohio-state.edu!mips!apple!portal!portal!cup.portal.com!pgl
From: pgl@cup.portal.com (Peter G Ludemann)
Newsgroups: comp.lang.prolog
Subject: Re: incrementing values
Message-ID: <26713@cup.portal.com>
Date: 8 Feb 90 07:39:54 GMT
References: <17467@megaron.cs.arizona.edu> <31462@shemp.CS.UCLA.EDU>
  <4884@itivax.iti.org> <2484@cs-spool.calgary.UUCP>
Organization: The Portal System (TM)
Lines: 29

The question is posed:

>                                         ....  This is a big problem for
> Logic programming ingeneral, HOW DO YOU DIRECTLY MODEL MUTATION AND UPDATE
> IN LOGIC PROGRAMMING?
>
> I personally feel that the only way to do this is to explicitly represent
> time within Logic Programming.  ... 

It's not clear to me that a time variable is needed.
If you do have a "time variable", how is it different from
a "state variable" (which gets us into the "frame problem")?

As an alternative, consider:

For example, suppose I am writing a chess program.  At each
step in the game, I can assume as axioms the description of
the board.  I then run my program and compute the best next
move.  At this point, I assert the new board position (this
is "destructive assignment") and start over.

In other words, inside my main program (which need not even
be in Prolog), I am running a series of mini-queries, each
time with a different axiom set.  This seems to me to be a
kind of meta-programming.  I haven't worked out the detailed
semantics, but with backtrackable "assignment", it may even have
a reasonably clean second-order logical theory.

- peter ludemann	sun!portal!cup.portal.com!pgl