Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!mailrus!nrl-cmf!ames!amdcad!sun!pitstop!sundc!seismo!uunet!mcvax!prlb2!kulcs!bimbart
From: bimbart@kulcs.uucp (Bart Demoen)
Newsgroups: comp.lang.prolog
Subject: Seattle tutorial prolog
Message-ID: <1415@kulcs.kulcs.uucp>
Date: 29 Aug 88 12:42:27 GMT
Reply-To: bimbart@kulcs.UUCP (Bart Demoen)
Organization: Katholieke Universiteit Leuven, Dept. Computer Science
Lines: 114

Tutorial No:8 of LP'88 Seattle (R. O'Keefe) makes 'interesting' reading, but
be careful not to believe everything without checking it on the system
you use; here is an example

Section 14 p.159 talks about the foreign interface. It uses gen_symcount/2 as
an example of what you should use the foreign interface for instead of Prolog.

I took the Prolog version mentioned in the tutorial, the version using the
foreign interface mentioned in the tutorial (corrected and adapted to BIMprolog)
and I wrote a version in 'pure BIMprolog'; you find them here, suffixed with C,
B and OK


C version
---------

:- extern_load([make_counter,increment_counter] , ['t.o']) .

:- extern_predicate(make_counter(pointer : r)) .

:- extern_predicate(increment_counter(integer : r , pointer : i)) .

get_next_gensym_counterC(_prefix,_count) :-
	gen_symcount(_prefix,_pcount) , ! ,
	increment_counter(_count,_pcount) .
get_next_gensym_counterC(_prefix,_count) :-
	make_counter(_pcount) ,
	assert(gen_symcount(_prefix,_pcount)) ,
	_count = 1 .

unfortunately, I had to change the C code as it stands in the tutorial; perhaps
Qprolog is interfacing to some other C ?

long *make_counter()
{
        register long *p = (long *)malloc(sizeof *p) ;
        *p = 1 ;
        return(p) ;  /*WAS MISSING IN THE TUTORIAL
        		this goes right on SUN3 - d0 still contains the return
        		value of malloc - but on VAX this goes wrong
        		see also comp.lang.c these days: topic QuickC
        	     */
}

long increment_counter(p)
register long *p ;
{
        (*p)++ ;    /*IN THE TUTORIAL VERSION, THIS WAS:  *p++ ;
        		DOING THE SAME AS *(p++) WHICH IS WRONG
        	    */
        return(*p) ;
}


BIMprolog version
-----------------

get_next_gensym_counterB(_prefix,_count) :-
	assoc(_prefix,_key) , ! ,
	recorded(_key,_old_count) ,
	_count is _old_count + 1 ,
	rerecord(_key,_count) .
get_next_gensym_counterB(_prefix,_count) :-
	make_key(_prefix,_key) ,
	assert(assoc(_prefix,_key)) ,
	rerecord(_key,1) ,
	_count = 1 .

make_key(_prefix,_prefix) . % this could be a more involved predicate, but since
			    % it is only called once, that's not going to make
			    % a big difference


Prolog tutorial version
-----------------------

get_next_gensym_counterOK(_prefix,_count) :-
	retract(gen_symcount(_prefix,_old_count)) ,
	! ,
	_count0 is _old_count + 1 ,
	assert(gen_symcount(_prefix,_count0)) ,
	_count = _count0 .
get_next_gensym_counterOK(_prefix,_count) :-
	assert(gen_symcount(_prefix,1)) ,
	_count = 1 .


the timings for 10000 calls to get_next_gensym_counter*(a,_x) (on a SUN 3/50)
in seconds:

	C	OK	B
	-------------------
	5.8	30.4	4.9

(the loop overhead of 0.8 sec was not subtracted)


gensym is clearly a case in which the foreign interface should NOT be used ...

as the BIMprolog manual states: the use of the external interface entails some
overhead, resulting in slower programs when using an external routine for a
short job

in the same section, the tutorial makes the general statement:
'If you need global variables, you might as well implement them in a language
which is good at them.' (meaning not Prolog, I suppose). This is not true.
Prolog has already the record predicates to do that.

bimbart@kulcs.uucp
Bart Demoen
Dept. of Computer Science
Celestijnenlaan 200A
B-3030 Leuven
Belgium