Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!lll-crg!mordor!sri-spam!sri-unix!hplabs!sdcrdcf!burdvax!bigburd!dowding
From: dowding@bigburd.UUCP (John Dowding)
Newsgroups: comp.lang.prolog
Subject: Re: PROLOG Digest   V4 #80
Message-ID: <2215@bigburd.UUCP>
Date: Sun, 7-Dec-86 14:37:05 EST
Article-I.D.: bigburd.2215
Posted: Sun Dec  7 14:37:05 1986
Date-Received: Mon, 8-Dec-86 03:31:02 EST
References: <157@its63b.ed.ac.uk>
Reply-To: dowding@bigburd.UUCP (John Dowding)
Organization: System Development Corp., Paoli, PA
Lines: 48

In article <157@its63b.ed.ac.uk> csrdi@itspna.uucp (Rick Innis, CS3) writes:
>In PROLOG Digest V4 #80 John Cugini <cugini@nbs-vms.arpa> writes:
>>%%
>>%%  Many of these predicates expect certain of their arguments to be
>>%%  instantiated upon invocation.  When such restrictions apply it is
>>%%  usually the leading arguments which are thought of as input (and
>>%%  hence instantiated), and the trailing arguments as output (and hence
>>%%  allowed to be uninstantiated).
>>%%
>
>A standard way (at least, in Edinburgh) of  denoting the  status of arguments
>to a Prolog predicate is to  include a  comment line  before the  body of the
>clause, in which arguments expected to be  instantiated are  prefixed by '+',
>uninstantiated arguments by a '-', and arguments where it  doesn't matter (or
>where either can be used) by '?'.  For example, 
>
>%% Append(+L1, +L2, -L3)
>
>indicates  the  status  of  the arguments  to the  usual use  of the standard
>'append'  clause.    To  illustrate further  uses, further  comments could be
>added, viz:  
>
>%% Append(+L1, -L2, +L3) is the calling pattern for finding if L1 is a member
>%% of L3.  (?L2 would also be acceptable).  
>
>In use, I've found that this notation makes Prolog code much easier to read
>and understand, which given some of what is possible in Prolog is a very
>desirable attribute. What say anyone else?
>
>	--Rick Innis.

I have found this notation a little bit misleading because it is unclear
whether, for instance, when '+' is used, if the programmer intended the
argument to be fully instantiated (i.e. ground), or simply non-variable.
Also, it doesnt allow certain kinds of important distinctions to be made.
For instance, a predicate might have a modeing foo(?,?) because the
programmer intends it to be used with either the first or second
argument instantiated, however, may not intend that it be called with
both arguments uninstantiated.  To express this, the programmer would have to
write something like:  foo(+,-) | foo(-,+) | foo(+,+).
To me this seems kind of awkward.

I dont have a better notation in mind.  Any suggestions?

John Dowding

UUCP:  {psuvax1 | bpa | sdcrdcf } !burdvax!bigburd!dowding
ARPA:  dowding@cis.upenn.edu