Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!samsung!think.com!snorkelwacker.mit.edu!bloom-beacon!eru!hagbard!sunic!sics.se!dan
From: dan@sics.se (Dan Sahlin)
Newsgroups: comp.lang.prolog
Subject: Re: '->' operator
Message-ID: <1991Jan10.103521.2278@sics.se>
Date: 10 Jan 91 10:35:21 GMT
References: <291@valverde.cs.utexas.edu> <1991Jan4.030918.24996@athena.cs.uga.edu> <1991Jan4.061142.26941@athena.cs.uga.edu> <6402@munnari.oz.au>
Sender: news@sics.se
Organization: SICS, Swedish Inst. of Computer Science
Lines: 35

Jeff Schultz wrote:

>I find the definition of (C -> T) as (once(C), T) much less useful in
>practice than the if-like one, but there's no question that it's the
>standard behaviour.  Sad, really.

I agree, and would add that not only is the semantics bad, but also the
syntax of if-then-else, especially if you do program transformations.

For instance, "A;(B;C)" is not equivalent to "(A;B);C" in Prolog, due to the
poor syntax of if-then-else. Let A be "(Test->Then)" and you see why.

Another example, "true,D" is not always the same as "D", if "D" is
"(Test->Then)".  The expression "(true,(Test->Then));E" is not the same as
"(Test->Then);E".

It really becomes complicated if you have a clause of the form
	p(G,X) :- G; X=2.
So what happens if you call "p((true->X=1),X)"?

Instead just getting X=1, you also get X=2 on backtracking, as the clause
when read in is first converted into
	p(G,X) :- call(G); X=2.

On the other hand, not all meta-goals G are converted into call(G), as in the
clause
	p2(G,X) :- call((G; X=2)).
The call "p2((true->X=1),X)" will only return the answer X=1.

The above examples are run in compiled and interpreted SICStus Prolog, but
many other Prolog systems show the same behavior.  If I am correctly
informed, the forthcoming Prolog standard will keep this syntax and
semantics of if-then-else.

	/Dan Sahlin