Re: Embarrassing Arithmetic behaviour

I knew beforehand Richard's answer which I think has already been delivered here with the same didactic ease.

Now, to soothe the pedagogical pressure : I already suspected that swi prolog was not a liar. I have also been suspecting these last 32 years that floating point arithmetic was not real number arithmetic.

I use the C interface of swi to work with Pascal. It all came from this Pascal code (example):

var a,b : double; 
begin
    b := 4.9 * 100000.0 ;
    if b = 490000 then showmessage('true') else showmessage('false') ;    
    a := PL_compare_mine(b)

Pascal will answer 'true'. (unlike swi-prolog, C, Gambit...  and Python which I have just checked)

I thought both languages used 64bits float IEEE754, hence my "embarrassment" !

I stumbled upon the problem when I used PL_put_float and PL_get_float in something like:

(The last two lines are the test, k is calculated by Pascal and r by Swi prolog, you may skip the other lines)

function PL_compare_mine (k : double) : double;
var requete: qid_t;
     solution : integer;
     arg1,arg2: term_t;
     r : double;
begin
   arg1:=PL_new_term_refs(2);
   arg2:=arg1+1  ;

   PL_put_float(arg1,k);
   requete:=PL_open_query(nil,
                                        PL_Q_NODEBUG,PL_predicate(PChar('plcompare'),2,'user'),
                                       arg1);
   solution:=PL_next_solution(requete);
   PL_get_float(arg2, r);
   PL_close_query(requete);
   result := r;
// ------------------ tests  -----------------------------
   showmessage(floattostr(k)+'    '+floattostr(r));
   if k=r then showmessage('true') else showmessage('false');

end;                        

The prolog code which concerns us is something like :

plcompare(_In,Out) :-   Out is 4.9 * 100000.0.

When you run the program you get :

% pascal test : ok

'true'

%  display of both numbers : ok

 490000    490000


% failure when comparing pascal number and swi prolog number

'false'  

There may be a Pascal and jgressier issue there. Something to do with Pascal floatostr function and with a non investigation on my part of the Pascal arithmetic background.

All the best

J.Gressier

> Message du 20/02/14 13:37
> De : "Derek Jones"
> A : "Boris Vassilev"
> Copie à : "Richard A. O'Keefe" , "j.gressier" , "Prolog mailing list"
> Objet : Re: [SWIPL] Embarrassing Arithmetic behaviour
>
> I have to agree with Richard on this. I find it quite strange.
>
> I learned about this in Computer Science A-level 32 years ago… Let alone a degree - where I *also* learned about it - an Electronic Engineering degree at that!
>
> But - then again - I find that people who I teach programming language classes to often do not have pre-requisites either.
>
> General dumbing down of education I think. Focus on Java - That kind of thing… ;-)
>
> Kind regards
>
> Derek.
>
>
> On Feb 20, 2014, at 2:42 AM, Boris Vassilev wrote:
>
> > On Thu, Feb 20, 2014 at 1:06 AM, Richard A. O'Keefe wrote:
> >
> >> How does anyone get through a programming course
> >> without meeting this stuff?
> >>
> >
> > It is quite easy. A degree in Software Engineering, or a degree in
> > Bioinformatics (with all the courses in them), just to name two examples.
> > Or maybe just doing Web-Development?
> > -------------- next part --------------
> > HTML attachment scrubbed and removed
> > _______________________________________________
> > SWI-Prolog mailing list
> > [email protected]
> > https://lists.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog
>
>
-------------- next part --------------
HTML attachment scrubbed and removed
_______________________________________________
SWI-Prolog mailing list
[email protected]
https://lists.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog