Yup, it seems like quite a simple spec really, doesn't it?
If the first "choice" succeeds (and it could succeed more than once
since "Corresp" has its own choice points), then _don't_ try the second
choice.
Otherwise succeed using the second choice (which, btw, is unique - no
backtracking "choice" points here) _unless_ the previous parse level
also used the second choice. (That's the PrevIsNum variable).
A doddle imperatively, but declaratively...? in clp?
Hm...
Hm, I see. If you only need that some solutions imply that some other
should be discarded, think of using a "choice penalty" and SearchBest
instead of SearchAll.
%% Global counter
ChoicePenaltyCounter=0x4000000
ChoicePenalty={NewCell 0}
ChoicePenaltyVar={FD.decl}
fun {Choose2}
ChoicePenaltyInc = @ChoicePenaltyCounter
in
ChoicePenaltyCounter := ChoicePenaltyInc div 2
choice
true
[]
ChoicePenalty := @ChoicePenalty + ChoicePenaltyInc
ChoicePenaltyVar >=: @ChoicePenalty
false
end
end
then turn some of "choice A [] B end" into "if {Choose2} then A else B
end" and add ChoicePenalty, ChoicePenaltyVar and ChoicePenaltyCounter to
your search Root. Finally add
{FD.assign min [ChoicePenaltyVar]}
to your search Script and require that each new solution has a lower
ChoicePenalty:
proc {Better S1 S2}
S1.choicePenaltyVar >=: S2.choicePenaltyVar
end
{SearchBest Script Better}
See what I mean?
_________________________________________________________________________________
mozart-users mailing list
[email protected]
http://www.mozart-oz.org/mailman/listinfo/mozart-users
