Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!zaphod.mps.ohio-state.edu!rpi!sci.ccny.cuny.edu!phri!cmcl2!yale!husc6!bunny!tomf
From: tomf@GTE.COM (Tom Fawcett)
Newsgroups: comp.lang.prolog
Subject: Re: Arrays in Prolog
Message-ID: <9653@bunny.GTE.COM>
Date: 27 Aug 90 23:33:04 GMT
References: <90239.175243SCHMIED@DB0TUI11.BITNET>
Organization: GTE Laboratories, Waltham MA
Lines: 34


I too have needed to use arrays in Prolog, and have been generally
dissatisfied with what I've found.  I'm using Quintus Prolog, which has two
separate library files that implement arrays in different ways.  Neither have
constant-time access and update, and the one that claims to be better requires
that you retract, modify, and re-assert the array for every modification.

When asking about this deficiency, the reaction I've gotten is that you
shouldn't need to use arrays - there is no logical interpretation of
destructive modifications to a data structure.  I suppose this is generally
true; for most applications there is a more natural and elegant way to
accomplish what you want without resorting to arrays.  Nevertheless, for some
applications Prolog's indexing mechanism simply isn't fast enough, and you
need the speed of arrays.

The solution I've settled on is to use a library file provided by Quintus
called vectors.pl, which allows you to create and access single dimension
arrays of arbitrary length.  I haven't looked closely at the code, but it uses
Unix's malloc and free routines to do storage management inside Prolog's heap.
The package was designed to be used to pass data between C and Prolog, and not
be a general array management package, but it seems to do the job.  It should
be fairly easy to write one yourself.

Unfortunately, besides having to write the code yourself, you'll end up with a
system split across two languages.  You have to make sure that the split is
fairly clean, since you don't want to spend a lot of time passing data between
C and Prolog.


-- 


-Tom Fawcett
GTE Labs, and UMass/Amherst