Re: [eclipse-users] Effect of line in "steiner" example

Hi Chuck,

Lutz, Charles D wrote:
>
> Hi folks,
>
> Some of the subtleties of how modeling in ECLiPSe works take a while to
>
> comprehend ;-)
>
> I am trying out the Steiner set example from the integer sets library 
> chapter
>
> (10) in the tutorial. The original example is shown below:
>
> :- lib(ic_sets).
>
> :- lib(ic).
>
> steiner(N, Sets) :-
>
> NB is N * (N-1) // 6, % compute number of triplets
>
> intsets(Sets, NB, 1, N), % initialise the set variables
>
> ( foreach(S,Sets) do
>
> #(S,3) % constrain their cardinality
>
> ),
>
> ( fromto(Sets,[S1|Ss],Ss,[]) do
>
> ( foreach(S2,Ss), param(S1) do
>
> #(S1 /\ S2, C), % constrain the cardinality
>
> C #=< 1 % of pairwise intersections
>
> )
>
> ),
>
> label_sets(Sets). % search
>
> label_sets([]).
>
> label_sets([S|Ss]) :-
>
> insetdomain(S,_,_,_),
>
> label_sets(Ss).
>
> I was curious why we needed the “C #=< 1” constraint, if we wanted to 
> constrain
>
> the cardinality of S1 /\ S2 to be 1, so I removed it and just replaced 
> the “C” in the
>
> cardinality constraint with a 1.
>
C #=< 1 does *not* mean constraining the cardinality to 1, but to *at 
most* 1, i.e. it can be either 0 or 1
(cardinality must be non-negative).
C #= 1 would be constraining the cardinality to 1 (and equivalent to 
your change).

So the difference is because you have changed the specification of the 
model, so you are no longer trying to solve the same problem.

Cheers,

Kish
>
> ECLiPSe now seems to run this in an “infinite” loop; I played with 
> both versions in
>
> the Tracer but I don’t know the execution details enough yet to tell 
> what is going
>
> on.
>
> Thanks,
>
> **Chuck Lutz**
>
> Lockheed Martin
>
> Systems of Systems Engineering
>
> BMC4I Modeling and Simulation
>
> Moorestown, NJ
>
> (856)638-7234 (office)
>
> charles.d.lutz_at_lmco.com <mailto:charles.d.lutz_at_lmco.com>
>
> “A graphic is never an end in itself; it is a moment
>
> in the process of decision making.”
> - Jacques Bertin
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> ECLiPSe-Users mailing list
> ECLiPSe-Users_at_crosscoreop.com
> http://www.crosscoreop.com/mailman/options/eclipse-users
>