Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!usc!apple!portal!portal!cup.portal.com!pgl
From: pgl@cup.portal.com (Peter G Ludemann)
Newsgroups: comp.lang.prolog
Subject: Re: incrementing values
Message-ID: <26542@cup.portal.com>
Date: 3 Feb 90 02:41:19 GMT
References: <17467@megaron.cs.arizona.edu> <31462@shemp.CS.UCLA.EDU>
  <34699@iuvax.cs.indiana.edu>
Organization: The Portal System (TM)
Lines: 31

In IBM-Prolog, I wouldn't dream of using assert/retract; instead,
I would use the internal database (similar to record, but better):

	term_get(name, Value) &
	term_assert(name, ? Value + 1)	% "?" triggers functional evaluation

It's actually quite efficient (maybe it shouldn't be so efficient;
we don't really want to encourage this kind of thing).  
As a variant, if I do:

	term_get(name, Value) &
	term_put(name, ? Value + 1)

the change is undone on backtracking (or when the query has no more solutions).

Not only that, but I could tell the system that term_get_f(name) is a
functional abbreviation for term_get(name,Value) and do this:

	term_put(name, ? term_get_f(name) + 1)

By the way, IBM-Prolog allows multiple keys.  As an example, all the
pragma and scanner values are kept in this internal data base.  So,
to change the scanner to accept [a!b] instead of [a|b] (for example, on 
a Japanese keyboard which is missing the bar character), I just enter:

	<- term_assert(assignable,6,"!") .	% 6 is the cons char's index

In fact, that's how IBM-Prolog implements Edinburgh syntax: there's
a built-in predicate which just does a series of such term_assert()s.

- peter ludemann	pgl@cup.portal.com