[Gecode] Announcement: FS constraints prototype

[Guido Tack]

Hi everyone.

As you may have guessed from the CVS logs, Gabor and I have started working on 
an implementation of finite integer set constraints for Gecode. Development 
will take place on the CVS branch "gecode-fs", as we expect some changes to 
be intrusive and don't want to break the rest of the system.

Here's what we have so far:
* FS variables
   - implemented as bitsets of fixed size
   - only upper and lower bound of the set (no cardinality yet)
   - basic tells: include and exclude
* "Non-basic propagators" from Tobias Müller's thesis
   - S_1\supseteq S_2\cap S_3
   - S_1\subseteq S_2\cup S_3
   - Other set constraints expressed in terms of the two above

We are now working on cardinality constraints and a simple distributor.

The model that we think of for cardinality at the moment is the following: We 
want to introduce an FD variable together with every FS variable and post a 
propagator that is subscribed to both and implements the cardinality 
constraints. The reason for separating the cardinality and the set variable 
is the limited number of modification events: We just could not signal a 
change of cardinality in addition to the MEs for lower and upper bound.
Christian: Is it possible/feasible to post a propagator directly upon variable 
creation?

Guido

-- 
Guido Tack
Programming Systems Lab
http://www.ps.uni-sb.de/~tack