[Gecode] Views break ModEvent-driven propagators

[Christian Schulte]

Guido is absolutely right about that. The fix is that a view must be able to
translate a VarModEvent to a potential ModEvent. For an example see the int
stuff (as I am late Guido has also done it for the set stuff).

The comment with uniform propagators is misleading. Please keep in mind that
varmodevent information is always very weak as it makes a statement about
all variables of the same type -- regardless of its view.

By the way, a paper describing some ideas found in Gecode appears at CP 2004
and is available here:

http://www.imit.kth.se/~schulte/paper.html?id=SchulteStuckey:CP:2004

Cheers
Christian

--
Christian Schulte, http://www.imit.kth.se/~schulte/ 

-----Original Message-----
From: gecode-bounces at ps.uni-sb.de [mailto:gecode-bounces at ps.uni-sb.de] On
Behalf Of Guido Tack
Sent: Thursday, June 24, 2004 10:59 AM
To: Technical Discussions about Gecode
Subject: [Gecode] Views break ModEvent-driven propagators


Hi!

While implementing the selection constraints, I noticed a flaw in the
current 
design: If you have a parametric propagator, you can't use e.g. varmodevent 
to find out why it was invoked, because the events there only correspond to 
the core variable, not the view it's expecting. This could be solved by 
having a conversion function in the view, but that will only work for
uniform 
propagators (only using variables of the same view type).
Is there anything I missed? Are parametric propagators not supposed to
inspect 
the ModEvents?

Guido

P.S.: Christian, I don't want to disturb you during your holidays, I just 
thought I'd post this so that I don't forget about it ;-)

-- 
Guido Tack
Programming Systems Lab
http://www.ps.uni-sb.de/~tack
_______________________________________________
Gecode mailing list
Gecode at ps.uni-sb.de http://www.ps.uni-sb.de/mailman/listinfo/gecode