On Thu, Nov 20, 2008 at 7:52 AM, Malcolm Ryan <[EMAIL PROTECTED]> wrote: > I'm not sure I understand that code. Would you mind explaining in more > detail?
First of all, the loops in the code looping through layers can be ignored for exposition, since they are there to handle the cases when cards are in the bottom, middle, or top of a pile of three cards. Assume that we have two piles of card (X1, X2, X3) and (Y1, Y2, Y3) where Xv and Xv are integer variables, and where X2 and Y2 have the same function for purposes of the game. The values of the cards represent when they are played (note that I mistranslated the condition in my earlier post, X1 and Y1 should be exchanged, as should X3 and Y3). Now the code can be simplified to // cond is the condition for the symmetry BoolVarArgs ba(4); // Both cards played after the ones on top of them ba[0] = post(this, ~(X2 < Y3)); ba[1] = post(this, ~(Y2 < X3)); // Both cards played before the ones under them ba[2] = post(this, ~(Y1 < X2)); ba[3] = post(this, ~(X1 < Y2)); // Cond holds when all the above holds BoolVar cond(this, 0, 1); rel(this, BOT_AND, ba, cond); // If cond is fulfilled, then we can order the cards // cond -> (X2 < Y2) post(this, tt(!cond || ~(X2 < Y2))); The ba-array holds a set Boolean variables for reified relations, and the rel-function takes the conjunction of these. The final post just uses the standard unfolding of an implication since there is no implication-operator. You can also use the implication-function together with post to get it: post(this, tt(imp(cond, ~(X2 < Y2)))); > How do you do reified constraints in Gecode? By using the reified post-functions, such as rel(this, x, IRT_LE, y, b); with x and y integer variables and b a Boolean control variable. In MiniModel expressions, you can use the unary tilde-operator to reify an expression: BoolVar b = post(this, ~(x < y)); Cheers, Mikael -- Mikael Zayenz Lagerkvist, http://www.ict.kth.se/~zayenz/ _______________________________________________ Gecode users mailing list [EMAIL PROTECTED] https://www.gecode.org/mailman/listinfo/gecode-users
