I am just starting to try and teach myself Prolog, using both Seven Languages in Seven Weeks and Learn Prolog Now! and I ran into something confusing. On day one of SLSW it presents the following knowledge base:
cat(lion).
cat(tiger).
dorothy(X,Y,Z):- X = lion, Y = tiger, Z = bear.
twin_cats(X,Y):- cat(X), cat(Y).
which when queried with twin_cats(X,Y). it produces all of the possibilities for two cats, including X = lion, Y = lion and X = tiger, Y = tiger. Messing around I was trying to add a goal to the twin_cats(X,Y) rule so that it would only report two, distinct cats. At first I tried
cat(lion).
cat(tiger).
dorothy(X,Y,Z):- X = lion, Y = tiger, Z = bear.
twin_cats(X,Y):- \+(X = Y), cat(X), cat(Y).
To which Prolog responded no to twin_cats(X,Y). However, when i just switched the order of the goals to
cat(lion).
cat(tiger).
dorothy(X,Y,Z):- X = lion, Y = tiger, Z = bear.
twin_cats(X,Y):- cat(X), cat(Y), \+(X = Y).
Prolog now gives me what I was looking for, two answers X = lion, Y = tiger and X = tiger, Y = lion.
So this is my question, is there some inherent ordering in the way that Prolog unifies variables? It would seem there is because the order of the goals (even though they are all seperated by and) changes the outcome of the query. In the second, working KB, i think I know how it is unifying. First it unifies for a cat(X) then cat(Y) then checks if they are the same. If they are it moves onto the next combo, etc, until there are no combos left and report the two cases that passed all three goals. But for the first KB, when it checks if X = Y first, shouldn't this goal be true because the variables haven't been unified yet?
