Path: utzoo!attcan!uunet!husc6!cs.utexas.edu!sm.unisys.com!aero!abbott
From: abbott@aero.ARPA (Russell J. Abbott)
Newsgroups: comp.lang.prolog
Subject: Re: The meaning of "declarative"
Keywords: Declarative, Procedural
Message-ID: <39584@aero.ARPA>
Date: 17 Oct 88 18:06:57 GMT
Reply-To: abbott@aero.UUCP (Russell J. Abbott)
Organization: The Aerospace Corporation, El Segundo, CA
Lines: 20


In general, it would seem that any programming language for which there
is a semantics that maps programs onto mathematical relations could be
called declarative, although doing so does not seem particularly
insightful.  For example, it is not that unfair to say that to a great
extent denotational semantics translates programs to mathematical
functions by substituting functional composition for semi-colons.  While
that might make the resulting objects declarative rather than
procedural, it is not clear that making that transformation improves
one's intuition a great deal.

From a prolog perspective, I would guess that most people, when
programming in prolog, think relatively procedurally when writing
"rules" and relatively declaratively when writing "facts."  What is
particularly powerful about prolog as a programming language is the ease
with which one can move back and forth between the declarative and
procedural domains.  "Facts" can be made "smart" by embedding them
within rule predicates, and one can gain declarative leverage on "rules"
by storing them as facts and interpreting them.

-- Russ