Re: Iteratively calling Prolog from C

[Daniel Dudley]

Lindsey Spratt wrote:
> On Thursday, July 17, 2003, at 10:15  AM, Daniel Dudley wrote:
> > It would appear that gprolog does have something seriously
> > wrong with its implementation. :-(
> >
> >     Win-Prolog 4.320           ECLiPsE 5.6
> >     (no big number library)    (has big number library)
> >     -----------------------    ------------------------
> >     ?- factorial(17,F).        ?- factorial(17, F).
> >     F = 355687428096000        F = 355687428096000
> >                                Yes (0.00s cpu)
> >
> >     ?- factorial(18,F).        ?- factorial(18, F).
> >     F = 6.402373705728E015     F = 6402373705728000
> >                                Yes (0.00s cpu)
> >
> >     GNU Prolog
> >     ----------
> >     ?- factorial(17,F).
> >     F = 248348672
> >     yes
> >
> >     ?- factorial(18,F).
> >     F = 175308800
> >     yes
> 
> This is an integer/float issue. If you force all of the 
> factorial calculations to be done using floats, then you get
> the right answers in gprolog.
> 
> factorial(N, F) :-
> FN is float(N),
> factorial(FN, 1.0, F).
> 
> 
> factorial(1.0, N, N) :- !.
> 
> factorial(N, J, F) :-
> N > 1,
> K is N - 1.0,
> L is N * J,
> factorial(K, L, F).
> 
> ?- factorial(17, F)
> Success
> F = 355687428096000.0

Yes, that's was my first impression, Lindsey. However, it
doesn't explain:

    ?- factorial(12,F).
    F = -57869312
    yes

A signed integer on a 32-bit platform is -2147483648 to
2147483647. Compare this with Win-Prolog and ECLiPSe:

    ?- factorial(12,F).
    F = 479001600

    ?- factorial(12, F).
    F = 479001600
    Yes (0.00s cpu)

Daniel