[
http://jira.jboss.com/jira/browse/JBRULES-387?page=comments#action_12340904 ]
Edson Tirelli commented on JBRULES-387:
---------------------------------------
Ok,
I finally realized the full extension of the problem exposed by this report. There are in
fact 3 problems:
1. A problem in JoinNode modifyTuple() method where a reference was being lost, causing
logically asserts to stay in memory. (FIXED, as per my previous comment)
2. BetaNodes modify*() method algorithms are relying on full cross product iteration. It
means that when beta memory indexing is active, we eventually might end up with strange
behaviors. (Still fixing NotNode, but need to double check all BetaNodes)
3. TerminalNode.modifyTuple() is not removing previously logically asserted facts. It
means that when no equals()/hashcode() methods are implemented in the logically asserted
class, we might end with several copies of the logically asserted fact for the same
activation. (FIXED)
So, I hope I will have all this fixed by the weekend.
Thanks,
Edson
Using exists together with facts of the same type can break
auto-retraction of logically asserted objects
---------------------------------------------------------------------------------------------------------
Key: JBRULES-387
URL:
http://jira.jboss.com/jira/browse/JBRULES-387
Project: JBoss Rules
Issue Type: Bug
Security Level: Public(Everyone can see)
Components: Reteoo
Affects Versions: 3.0.3
Environment: Windows XP, java 1.5
Reporter: Geoffrey De Smet
Assigned To: Edson Tirelli
Attachments: JBRULES-387-testcase-trunk.patch, JBRULES-387.zip, screenshot-1.jpg
I assert 3 Lesson objects in my memory.
I got 1 rule, which creates logically asserted HardConstraint objects.
In some cases (not always), those logically asserted objects aren't retracted when
they should be, which leads to HardContaints just adding more and more. They aren't
logically retracted as they should be when:
- exists is combined with fact selections of the same class
- the modified object trigged the rule by the exists part, not by the the fact selection
part
I 'll attach a maven 2 configured proof of concept to repreduce it:
mvn install eclipse:eclipse idea:idea
Run net.sf.tabudrool.app.TabudroolApp
Click "Load unsolvedSchedule2.xml" (the second button on the left)
Notice the HardConstraint = 2 at the bottom
Click "Move lesson"
Move "[net.sf.tabudrool.domain.Lesson-1] 0: teacher0 + group0" to timeslot 1.
Notice that HardConstraint = 3 instead 1 at the bottom,
because - as you can see in the log - the previous HardConstraint objects didn't get
retracted.
Here's the drl:
rule teacherCanOnlyTeachOneGroupAtATime
when
$lesson : Lesson($id : id, $teacher : teacher, $timeslot : timeslot)
exists Lesson(id > $id, teacher == $teacher, timeslot == $timeslot)
then
System.out.println(" teacherCanOnlyTeachOneGroupAtATime: " + $lesson);
assertLogical(new HardConstraint(1));
end
Here's a log:
teacherCanOnlyTeachOneGroupAtATime: [net.sf.tabudrool.domain.Lesson-1] 0: teacher0 +
group0
teacherCanOnlyTeachOneGroupAtATime: [net.sf.tabudrool.domain.Lesson-0] 0: teacher0 +
group0
[ObjectAsserted: handle=[fid:9:9];
object=net.sf.tabudrool.domain.HardConstraint@1f42731]
[ObjectAsserted: handle=[fid:10:10];
object=net.sf.tabudrool.domain.HardConstraint@53abbb]
2006-08-03 19:24:35,796 [AWT-EventQueue-0] INFO
net.sf.tabudrool.swingui.WorkflowFrame$MoveAction.actionPerformed(WorkflowFrame.java:201)
Moving [[net.sf.tabudrool.domain.Lesson-1] 0: teacher0 + group0] to [1]
[ObjectModified: handle=[fid:7:11]; old_object=[net.sf.tabudrool.domain.Lesson-1] 1:
teacher0 + group0; new_object=[net.sf.tabudrool.domain.Lesson-1] 1: teacher0 + group0]
teacherCanOnlyTeachOneGroupAtATime: [net.sf.tabudrool.domain.Lesson-1] 1: teacher0 +
group0
[ObjectAsserted: handle=[fid:11:12];
object=net.sf.tabudrool.domain.HardConstraint@1c8b884]
As you can see: net.sf.tabudrool.domain.HardConstraint@1f42731 and
net.sf.tabudrool.domain.HardConstraint@53abbb weren't retracted (at least one should
be).
Earlier this week I posted a dud jira issue, as far as I can tell from discussing on irc
(with conan), this looks like a genuine issue...
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://jira.jboss.com/jira/secure/Administrators.jspa
-
For more information on JIRA, see:
http://www.atlassian.com/software/jira