[rules-users] [planner] Hard constraints in drl, MoveFactory or isMoveDoable

[Geoffrey De Smet]

Hi Evert,

Op 24-03-11 15:10, Evert Penninckx schreef:
> Hi
>
> General question on the drools planner. If I'm not mistaken, I see three
> ways to impose hard constraints:
>
> 1. Generate violations in the scoring DRL.
This is the normal way for hard constraints.
> 2. Check them in isMoveDoable
This is a really bad idea.
Take a look in the latest snapshot/hudson reference manual in the 
section about "Delta Score calculation" to understand why.
   
https://hudson.jboss.org/hudson/view/Drools%20jBPM/job/drools-planner/lastSuccessfulBuild/artifact/
> 3. Do not create moves in the MoveFactory which cause hard constraints to be
> broken.
Then they are called "build-in hard constraints".
It's quite difficult to do this right, but as 
SimpleTravelingTournamentApp vs SmartTravelingTournamentApp proves,
this can be definitely worth doing, because the smart one is a lot better.

See "tip and tricks" section in the reference manual.

But still, in my experience, making some constraints "build-in hard 
constraints" is a short term gain, but a possible long term loss, 
because you're limiting the freedom of moves.
But this is use case specific.
Generally, in my experience, I prefer just mixing in course-grained 
moves together with the fine-grained moves first, before experimenting 
with build-in hard constraints.

> For safety I would always add a check in the DRL (1) in combination with 2
> or 3 if possible.
> I assume that checks in isMoveDoable are more expensive in terms of
> performance and should only be put there if they can vary with each
> solution.
>
> Does this make sense?
>
>
> Evert
>
> --
> View this message in context: http://drools-java-rules-engine.46999.n3.nabble.com/planner-Hard-constraints-in-drl-MoveFactory-or-isMoveDoable-tp2725305p2725305.html
> Sent from the Drools - User mailing list archive at Nabble.com.
> _______________________________________________
> rules-users mailing list
> rules-users at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-users
>

-- 
With kind regards,
Geoffrey De Smet