[jboss-jira] [JBoss JIRA] (DROOLS-1478) Call ActivationUnMatchListener also when a rule rematches

[ https://issues.jboss.org/browse/DROOLS-1478?page=com.atlassian.jira.plugi... ] Geoffrey De Smet updated DROOLS-1478: ------------------------------------- Description: ActivationUnMatchListener is part of kie-internal, so its *not public API*. *ActivationUnMatchListener is only used by OptaPlanner* (probably): jBPM and downstream drools module don't use it: https://github.com/search?q=org%3Akiegroup+ActivationUnMatchListener&... Community users probably don't use it either, because no one mentions it on StackOverflow: https://www.google.be/search?q=%2BActivationUnMatchListener+site:stackove... The current behaviour of ActivationUnMatchListener is a pain because it doesn't unmatch if a RHS fires again, causing an imbalance. OptaPlanner works around this through a hack (which fails for PLANNER-761). For example, given this rule {code} global int score = 0; rule R when Wine(age < 10, $age : age) then score += $age; addUnmatchListener(() -> score -= $age); end {code} This works for a normal match and unmatch event: {code} Wine w = new Wine(3); insert(w); fireAllRules(); => score = 7; w.age = 50; update(w); fireAllRules(); => unmatch triggers => score = 0; // GOOD {code} So events are like this: {code} RHS unmatch {code} However if we have a rematch: {code} Wine w = new Wine(3); insert(w); fireAllRules(); => score = 7; w.age = 4; update(w); fireAllRules(); => no unmatch triggers => score = 13; // BAD should be 6! w.age = 50; update(w); fireAllRules(); => unmatch triggers => score = 7; // should be 0! {code} Because the events are unbalanced: {code} RHS RHS unmatch {code} was: ActivationUnMatchListener is part of kie-internal, so its *not public API*. *ActivationUnMatchListener is only used by OptaPlanner* (probably): jBPM and downstream drools module don't use it: https://github.com/search?q=org%3Akiegroup+ActivationUnMatchListener&... Community users probably don't use it either, because no one mentions it on StackOverflow: https://www.google.be/search?q=%2BActivationUnMatchListener+site:stackove... The current behaviour of ActivationUnMatchListener is a pain because it doesn't unmatch if a RHS fires again, causing an imbalance. OptaPlanner works around this through a hack (which fails for PLANNER-761). For example, given this rule {code} global int score = 0; rule R when Wine(age < 10, $age : age) then score += $age; addUnmatchListener(() -> score -= $age); end {code} This works for a normal match and unmatch event: {code} Wine w = new Wine(3); insert(w); fireAllRules(); => score = 7; w.age = 50; update(w); fireAllRules(); => unmatch triggers => score = 0; // GOOD {code} So events are like this: {code} RHS unmatch {code} However if we have a rematch: {code} Wine w = new Wine(3); insert(w); fireAllRules(); => score = 7; w.age = 4; update(w); fireAllRules(); => no unmatch triggers => score = 13; // should be 6! w.age = 50; update(w); fireAllRules(); => unmatch triggers => score = 7; // should be 0! {code} Because the events are unbalanced: {code} RHS RHS unmatch {code} -- This message was sent by Atlassian JIRA (v7.2.3#72005)