Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!decwrl!uunet!mcsun!unido!tub!tubopal!db0tui11!schmied
From: SCHMIED@tubvm.cs.tu-berlin.de (Albrecht Schmiedel)
Newsgroups: comp.lang.prolog
Subject: Arrays in Prolog
Message-ID: <90239.175243SCHMIED@DB0TUI11.BITNET>
Date: 27 Aug 90 15:52:43 GMT
Organization: Technical University Berlin
Lines: 43

We are implementing a knowledge representation tool involving large
inheritance networks, and until now we have been using Prolog, mainly
because this is the language we are most familiar with, and because 'rapid
prototyping' seems to be easy (compared to C, or even Lisp).
Also, we would like to continue using Prolog (for exactly
those reasons), but we  are increasingly confronted with
performance problems.

Essentially, this has to do with the inability of Prolog to store,
retrieve, and replace data indexed by integers. This is actually
all we need to search and to modify large concept graphs; but
Prolog forces us (as far as I can see) to do asserts, retracts
and database calls where simple assignments and lookups in
arrays would suffice. Our program spends large portions of
the total run time doing assertions, retractions and calls in those
parts of the Prolog database that could easily be replaced by an
array storing Prolog terms.

My questions are the following:
1) Are there Prologs that adress this kind of need?
   (I.e. Prologs that make assertions, retractions, and calls
   involving  predicates that fulfill certain conditions
   - for example, a guaranteed instantiated integer as its first
   argument, one clause per integer, no backtracking -
   similarly fast as array operations.)
2) Are there Prologs which are easily extendable by predicates
   written in C that could do the job?
   (It should be possible to pass *terms* back and forth;
   having to decompose terms and lists before processing them
   by a foreign language predicate and then having to put them
   together again later on is definitely not what we want.)
3) Shouldn't we be using Prolog?
4) Is there a 'Prolog-way' of handling large graphs efficiently?
   (Graphs where data referencing other nodes is attached to nodes,
   and search as well as updates are required.)

-------
Albrecht Schmiedel                  +49 30/314-25494
Technische Universitaet Berlin      schmied@db0tui11.bitnet
Sekr. FR 5-12                       schmied@tubvm.cs.tu-berlin.de
Franklinstr. 28/29
1000 Berlin 10
West Germany