Re: GNUProlog vs. ECLiPSe

*snip*
> 
> > In GNUProlog, I get possibilities for the value of the variables by
> > simply running the file I attached. Than I can modify the file with one
> > of the resulting values until all variables have a value.
> 
> How do you achieve this in GNUProlog, it's most likely very similar.
> 
> > I must write a java-program that generates the input to ECLiPSe, get the
> > output (the possible values for the variables) and use that values to
> > generate a new input and so on , until I've a value for every variable.
> 

If what you are hoping to achieve by finding a value for every
variable, is to find values that satisfy all the constraints then
there is no need to write a Java program and genereate new input.  You
can find the values much more simply within ECLiPSe (or GNUProlog for
that matter).

e.g.

Here is the standard SEND+MORE=MONEY example

:-lib(ic).

send(Digits):-
	Digits=[S,E,N,D,M,O,R,Y],

         % Give initial domains to variable
	Digits :: [0..9],

         % setup constraints on variables
	alldifferent(Digits),
	S #\=0, M #\= 0,
	          S*1000 + E*100 + N*10 + D +
                   M*1000 + O*100 + R*10 + E #=
         M*10000 + O*1000 + N*100 + E*10 + Y,

         % search for solution values
	labeling(Digits).


As you can see there are 3 stages to the problem.

1) Declare your problem variables and give them initial domains.
2) Setup the constraints of the problem
3) Search for a valid assignement of values to your variables.

Running the query "send(Digits)" gives the following valid assignment

?- send(Digits).
Digits = [9, 5, 6, 7, 1, 0, 8, 2]
Yes (0.00s cpu, solution 1, maybe more)

asking for more solutions says "no", since the above solution is unique.


The above example uses the labeling/1 predicate to find valid values
for the variabels, ECLiPSe has a number of more general search
mechanisms aswell (e.g. in the library "ic_search").  These are
explained in detail in the ECLiPSe documentation.

Hope that helps.

Andrew Sadler