Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!apple!agate!linus!linus!guttman
From: guttman@linus.mitre.org (Joshua D. Guttman)
Newsgroups: comp.lang.lisp
Subject: Re: FALSE vs empty list
Summary: an example to distinguish 'em
Message-ID: <119011@linus.mitre.org>
Date: 4 Sep 90 18:15:19 GMT
References: <24305@uflorida.cis.ufl.EDU>
Distribution: usa
Organization: The MITRE Corporation, Bedford MA
Lines: 39


Well here's an example of why it would be nice to have '() and #f (in scheme
notation) be distinct.  (Note:  I believe that in the new Scheme standard
(R4RS) they will be required to be distinct, with the null list counting as
non-false in conditional tests.  Just imagine!)

Suppose you're doing syntactic matching or unification or some procedure like
that which wants to return a substitution when it succeeds.  

So ((x . 23) (y .  34)) means substitute 23 for x and 34 for y and you have a
match.  Then it's natural for () to mean that you don't have to substitute
anything, because the expressions already match.  

However, what do you return when the expressions can't be made to match, e.g.
(* x 3) and (+ x 3)?  Well, you would normally want to return #f to indicate
failure, but if () = false you can't, because that indicates you have
succeeded.  

So you have to do a ghastly hack like return the symbol FAIL.  This is highly
unpleasant when you want to write code like 

(any
 (lambda (exp)
  (match exp fixed-exp))
 some-list)

since you have to rewrite it as an iteration with a test for non-FAILing return
value.  Even 

(any
 (lambda (exp)
  (let ((subst (match exp fixed-exp)))
    (and (succeed? subst)
    	 subst)))
 some-list)

is wrong.

	Josh