[Gecode] Space consumption of set variables

[Gabor Szokoli]

Guido Tack wrote:

>My conclusion 
>is that additional propagation conditions are expensive, and should only be 
>used if a lot of propagators can really make use of the stronger information. 
>This does not seem to be the case with cardinality PCs for set constraints.
>  
>
Hi all,

I know nothing is too expensive for the guy who just watches, and that 
this implies a horrible nightmare of a build process, with exponential 
number of scenarios to test and maintain, but I just have to say it:
Maybe these kind of alternative implementations could be factorised into 
build options of the library, so double-plus-expert users could tweak 
them to fit their problem domains?
It just seems like such a waste to permanently drop a setup which might 
benefit some (agreed, probably not many) propagators.


Gabor Szokoli