This will not improve efficiency but here's a single rule doing the job, at the cost of one additional fact.
(deftemplate driver (slot driverId) (slot score) (slot rank)) (deftemplate rover (slot rank)) (defrule rank-drivers ?d <- (driver (score ?s1) (rank nil)) (not (driver (score ?s2&:(> ?s2 ?s1)) (rank nil))) ?r <- (rover (rank ?x)) => (modify ?d (rank ?x)) (modify ?r (rank (++ ?x))) ) (reset) .. (assert (rover (rank 1))) Cheers Wolfgang On 6 March 2012 16:55, Friedman-Hill, Ernest <ejfr...@sandia.gov> wrote: > That's about the best you can do purely with rules, I'm afraid. The only > way to do better would be to query for all the facts, sort them using an > explicit sort, and then run through the list, modifying them one at a time. > That wouldn't be too hard to code up in Java. > > Now that I'm saying this, it seems like it should be possible to write a > generic Userfunction which sorted facts based on some given criteria. That > would be an interesting addition to Jess's library. > > From: Felix Chan <chanfe...@gmail.com> > Reply-To: <jess-users@sandia.gov> > Date: Mon, 5 Mar 2012 17:57:25 -0800 > To: <jess-users@sandia.gov> > Subject: JESS: [EXTERNAL] question on sorting > > Hi, > > I need to rank a set of facts based on a score. I came up with the idea > that I first find the top ranked fact using one rule, followed by ranking > the remaining facts using another rule. > > It works. But I suspect that this method will not be very efficient if I > have to rank thousands of facts, rather than just 4 as in my example below. > I also suspect that someone must have solved this sorting problem in a much > better way. > > Any recommendation would be much appreciated. Many thanks. > > Felix > > > > /******************************************************************/ > > (deftemplate driver (slot driverId) (slot score) (slot rank)) > > (defrule rankFirstDriver > "find top ranked driver if none has been ranked" > (not (driver (rank ~nil))) ; if no drivers have been > ranked > ?d1 <- (driver (score ?s1)) ; if we can find a driver > with score s1 > (not (driver (score ?s2&:(> ?s2 ?s1)))) ; but no other drivers have a > score higher than s1 > => > (modify ?d1 (rank 1)) > ) > > (defrule rankOtherDrivers > "rank other drivers after top ranked drivers have been found" > (driver (rank 1)) ; if top ranked driver has been found > > ; find lowest ranked driver > (driver (rank ?r1&~nil)) > (not (driver (rank ?r2&~nil&:(> ?r2 ?r1)))) > > ; find unranked driver with highest score > ?d <- (driver (score ?s1) (rank nil)) > (not (driver (score ?s2&:(> ?s2 ?s1)) (rank nil))) > > => > (modify ?d (rank (+ ?r1 1))) > ) > > (reset) > (assert (driver (driverId 1) (score 100))) > (assert (driver (driverId 2) (score 50))) > (assert (driver (driverId 3) (score 200))) > (assert (driver (driverId 4) (score 75))) > > (run) > (facts) > > > > /*********************************/ > > /***** OUTPUT **************/ > > > f-0 (MAIN::initial-fact) > f-1 (MAIN::driver (driverId 1) (score 100) (rank 2)) > f-2 (MAIN::driver (driverId 2) (score 50) (rank 4)) > f-3 (MAIN::driver (driverId 3) (score 200) (rank 1)) > f-4 (MAIN::driver (driverId 4) (score 75) (rank 3)) > For a total of 5 facts in module MAIN. >
