[gecode-users] Detecting a complete problem

[Mikael Zayenz Lagerkvist]

On Nov 21, 2007 12:45 AM, Malcolm Ryan <malcolmr at cse.unsw.edu.au> wrote:
> On 20/11/2007, at 6:13 PM, Malcolm Ryan wrote:
> > How does Gecode/J decide that a problem is complete? I've just tried
> > running Gist on a problem and it stopped at the root node and said it
> > was solved, even though many variables were still unbound.

Gecode considers a Space solved when there are no more banchings left.


> Aha, I found the problem. There's a missing line in the copy
> constructor:
>
> >      public Test1(Boolean share, Test1 old) {
>
>         super(share, old);

The explanation for this is that leaving out the call to the correct
super-constructor had the effect that neither propagators nor
branchings will be copied. Thus the new Space will have no branchings,
and is therefore "solved".


Cheers,
Mikael


-- 
Mikael Zayenz Lagerkvist, http://www.ict.kth.se/~zayenz/