10:43 a.m.
Hello guys,
I have a very strange behaviour when i execute this fragment of code:
private void _clearKnowledge() {
try {
Collection<KnowledgePackage> knowledgePackages =
_knowledgeBase.getKnowledgePackages();
for (KnowledgePackage singlePackage : knowledgePackages) {
_log.debug("Removing knowledge in drools engine");
_log.debug("package name: " + singlePackage.getName());
_knowledgeBase.removeKnowledgePackage(singlePackage.getName());
}
} catch (NullPointerException e) {
// TODO An error about a missing package rarely occurs.
_log.warn("Drools Engine has tried to remove a missing knowledge
package.");
}
}
Sporadically the system raise this ugly exception:
Caused by: java.lang.NullPointerException
at
org.drools.reteoo.ReteooBuilder.removeRule(ReteooBuilder.java:252)
at
org.drools.reteoo.ReteooRuleBase.removeRule(ReteooRuleBase.java:459)
at
org.drools.common.AbstractRuleBase.removeRule(AbstractRuleBase.java:1099)
at
org.drools.common.AbstractRuleBase.removePackage(AbstractRuleBase.java:1017)
at
org.drools.impl.KnowledgeBaseImpl.removeKnowledgePackage(KnowledgeBaseImpl.java:202)
at
org.titan.frontend.memory.drools.DroolsEngine._clearKnowledge(DroolsEngine.java:304)
I thought about timing issues, but each group of interaction with the engine
is protected with a synchronized method on the facade of the module. The
try-catch is the unique solution that I have found so far.
Thanks
Andrea
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
10:54 a.m.
I forgot to say what version I'm using. The version of Drools is 5.5.
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
11:10 a.m.
Can you try 5.6 or a later version?
That part of logic was improved in 5.6, and rewritten in 6.x
Davide
On 07/14/2014 04:54 PM, braveheart85 wrote:
I forgot to say what version I'm using. The version of Drools is
5.5.
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
1:12 p.m.
Hi Davide,
Thanks for the reply.
Unfortunately I cannot migrate to version 5.6 or newer.
Today I have inspected that part of code, but it is quite huge and complex
in the Reeteo part.
I have seen the NullPointer is due to the "nodes" retrieved from the
"rules"
Map (line 250).
Moreover, this non-deterministic behaviour complicates everything.
65 private Map<Rule, BaseNode[]> rules;
244 public synchronized void More ...removeRule(final Rule rule) {
245 // reset working memories for potential propagation
246 InternalWorkingMemory[] workingMemories =
this.ruleBase.getWorkingMemories();
247
248 final Object object = this.rules.remove( rule );
249
250 final BaseNode[] nodes = (BaseNode[]) object;
251 final RuleRemovalContext context = new RuleRemovalContext( rule
);
252 for (final BaseNode node : nodes) {
253 node.remove( context,
254 this,
255 null,
256 workingMemories );
257 }
258
259 resetMasks(context);
260 }
I don't know if you can help me in some way...
Thanks in all cases.
Andrea
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
3:10 p.m.
I assume you can't recompile the source code of 5.5 either...
If not, the only workaround is 1) try to identify which rule or rules cause
the problem, 2) rewrite them to increase (or reduce) the node sharing
On Jul 14, 2014 2:13 PM, "braveheart85" <andrea.pacifico.85(a)gmail.com>
wrote:
Hi Davide,
Thanks for the reply.
Unfortunately I cannot migrate to version 5.6 or newer.
Today I have inspected that part of code, but it is quite huge and complex
in the Reeteo part.
I have seen the NullPointer is due to the "nodes" retrieved from the
"rules"
Map (line 250).
Moreover, this non-deterministic behaviour complicates everything.
65 private Map<Rule, BaseNode[]> rules;
244 public synchronized void More ...removeRule(final Rule rule) {
245 // reset working memories for potential propagation
246 InternalWorkingMemory[] workingMemories =
this.ruleBase.getWorkingMemories();
247
248 final Object object = this.rules.remove( rule );
249
250 final BaseNode[] nodes = (BaseNode[]) object;
251 final RuleRemovalContext context = new RuleRemovalContext( rule
);
252 for (final BaseNode node : nodes) {
253 node.remove( context,
254 this,
255 null,
256 workingMemories );
257 }
258
259 resetMasks(context);
260 }
I don't know if you can help me in some way...
Thanks in all cases.
Andrea
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
4:19 a.m.
I suppose the problematic query is:
query "Get Fixed BusinessEntities"
$result : BusinessEntityData( queryable == true )
end
What you means with increase and decrease node sharing?
Sorry, maybe I didn't study very well the reteeo part of the documentation.
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
6:51 a.m.
What are you trying to do exactly when you "clear" the knowledge base?
If you need to replace all the rules, I'd just create a new KB altogether..
facts would be reasserted into the new rules anyway.
If you need to update some of the rules, Drools 5.x had the KnowledgeAgent
for this purpose. To be fair, however, in 5.5 it still had some issues
that were
fixed in 5.6.
I may be missing some of the requirements, though.
Davide
On 07/15/2014 10:19 AM, braveheart85 wrote:
I suppose the problematic query is:
query "Get Fixed BusinessEntities"
$result : BusinessEntityData( queryable == true )
end
What you means with increase and decrease node sharing?
Sorry, maybe I didn't study very well the reteeo part of the documentation.
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
8:29 a.m.
My users can apply a filter (query in Drools) and the result facts are tagged
to be shown.
Filters can be from a filter library or also directly written by users.
I need to clear the production memory and set new knowledge (filter) each
time in order to get filter result.
First solution you propose is a bit problematic for me because i'd have to
reinsert all facts inside the working memory but I don't keep these objects
outside Drools. In my context facts are fixed while rules/queries are
variables. I'd have to query Drools for all facts, then instantiate a new
KnowledgeBase and finally reinsert all facts...
What do you think?
Thanks
Andrea
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
1:43 p.m.
Conceptually you are right, but in practice, when you add a rule/query
to the KBase all the available facts are propagated through the newly
built piece of RETE. Since you clear all the rules first (which has some
cost)
and then you build a new set of rules, the effect is very similar - only,
the engine will implicitly re-insert the objects for you.
Now, a clean insertion has some costs (like the creation of the handles),
so it would be interesting to benchmark the two alternatives.
From a developer's perspective, moving the objects from a session
into
another is as simple as this:
for ( Object o : oldSession.getObjects() ) {
newSession.insert( o );
}
and if you can't (re)compile the source code, there is little that can be
done to solve your problem. Unless there is a multi-threading issue,
that is, you have different threads adding facts, removing rules/queries
and adding the new ones. If that was the case, I would not be surprised
since 5.5 was not thread safe.
Davide
On 07/15/2014 02:29 PM, braveheart85 wrote:
My users can apply a filter (query in Drools) and the result facts
are tagged
to be shown.
Filters can be from a filter library or also directly written by users.
I need to clear the production memory and set new knowledge (filter) each
time in order to get filter result.
First solution you propose is a bit problematic for me because i'd have to
reinsert all facts inside the working memory but I don't keep these objects
outside Drools. In my context facts are fixed while rules/queries are
variables. I'd have to query Drools for all facts, then instantiate a new
KnowledgeBase and finally reinsert all facts...
What do you think?
Thanks
Andrea
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
3:33 a.m.
Yesterday I tried version 5.6. I have had a problem with the MVEL syntax.
rule "RULE1"
$result: PersonData( businessActivity.status == "CLOSED" )
end
This rule that works in the version 5.5, now have a problem in version 5.6:
unable to invoke method:
org.titan.registry.filter.BusinessActivityData.getStatus: target of method
is null
The problem is that not all Person facts inside the working memory have a
business activity. NullPointer would be quite correct but in version 5.5 the
behaviour was different...
Maybe I'm abusing of your kindness.
Andrea
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
3:54 a.m.
Not checking for nulls is a debatable behavior :)
In 5.6 you should be able to use the null safe accessor
Business activity.!status == ...
Which will be rewritten to do the check for you
On Jul 16, 2014 4:33 AM, "braveheart85" <andrea.pacifico.85(a)gmail.com>
wrote:
Yesterday I tried version 5.6. I have had a problem with the MVEL
syntax.
rule "RULE1"
$result: PersonData( businessActivity.status == "CLOSED" )
end
This rule that works in the version 5.5, now have a problem in version 5.6:
unable to invoke method:
org.titan.registry.filter.BusinessActivityData.getStatus: target of method
is null
The problem is that not all Person facts inside the working memory have a
business activity. NullPointer would be quite correct but in version 5.5
the
behaviour was different...
Maybe I'm abusing of your kindness.
Andrea
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
4:38 a.m.
I have tried your hint and i got:
query "dis"
$result : PersonData( businessActivity.!status == "CLOSED")
end
[48,38]: [ERR 101] Line 48:38 no viable alternative at input '.' in query
[0,0]: Parser returned a null Package
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
4:42 a.m.
Oops, it's !., not .!
On Jul 16, 2014 5:39 AM, "braveheart85" <andrea.pacifico.85(a)gmail.com>
wrote:
I have tried your hint and i got:
query "dis"
$result : PersonData( businessActivity.!status == "CLOSED")
end
[48,38]: [ERR 101] Line 48:38 no viable alternative at input '.' in query
[0,0]: Parser returned a null Package
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
5:19 a.m.
Ok, now the query works.
However I have noticed that when that rule with (. instead of !.) fails,
Drools can't remove the rule in the cleaning phase like the version 5.5.0. I
have found these logs in my console:
09:47:31,989 INFO [http-bio-8080-exec-8][WorkingMemoryConsoleLogger:114]
BEFORE PACKAGE ADDED package: "org.titan.memory.drools.command.custom"
09:47:31,990 INFO [http-bio-8080-exec-8][WorkingMemoryConsoleLogger:114]
BEFORE PACKAGE ADDED package: "org.titan.registry.filter"
09:47:31,992 INFO [http-bio-8080-exec-8][WorkingMemoryConsoleLogger:114]
BEFORE RULE ADDED package: "org.titan.memory.drools.command.custom" rule:
"dis"
then:
java.lang.RuntimeException: cannot invoke getter: getBusinessActivity ....
then:
09:47:32,373 INFO [http-bio-8080-exec-8][WorkingMemoryConsoleLogger:114]
BEFORE PACKAGE REMOVED package: "org.titan.memory.drools.command.custom"
09:47:32,373 INFO [http-bio-8080-exec-8][WorkingMemoryConsoleLogger:114]
BEFORE RULE REMOVED package: "org.titan.memory.drools.command.custom" rule:
"dis"
and finally:
Caused by: java.lang.NullPointerException
at org.drools.reteoo.ReteooBuilder.removeRule(ReteooBuilder.java:252)
at org.drools.reteoo.ReteooRuleBase.removeRule(ReteooRuleBase.java:469)
at
org.drools.common.AbstractRuleBase.removeRule(AbstractRuleBase.java:1109)
at
org.drools.common.AbstractRuleBase.removePackage(AbstractRuleBase.java:1027)
at
org.drools.impl.KnowledgeBaseImpl.removeKnowledgePackage(KnowledgeBaseImpl.java:210)
at
org.titan.frontend.memory.drools.DroolsEngine._clearKnowledge(DroolsEngine.java:316)
I don't have a log like AFTER RULE ADDED ... "dis". However Drools try to
remove "dis" when i execute another query after the exception have raised.
This is the last question. I promise!
Andrea
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
8:39 a.m.
Finally I resolve my issues destroing the Knowledge Base and then re-create
it when the problem occurs.
I confirm that version 5.6 is more stable than 5.5. In 5.6 i don't have the
non-deterministic behaviour anymore whereby it's impossible to remove a
knowledge package. In 5.6, however, the same issue is still present when i
add a knowledge package in the knowledge base that generate a
nullpointerexception when matching with facts already present inside the
session.
In 5.6 the issue can be accurately examined because have a deterministic
occurrences! The problem seems to be:
* KnowledgeBase fail to add the new package due to NullPointerException.
* When the cleaning is called KnowledgeBase want remove the package that
previously it couldn't add.
Probably I have had a lot of problems because I'm not doing things in the
right way.
I hope this experience can help someone.
--
View this message in context:
http://drools.46999.n3.nabble.com/Problem-when-I-cleaning-the-production-...
Sent from the Drools: User forum mailing list archive at Nabble.com.
3606
days inactive
3610
days old
14 comments
2 participants
participants (2)
-
braveheart85 -
Davide Sottara