[rules-users] Declared Types and Globals in different files with incremental KnowledgeAgent

De Rooms Brecht bderooms at vub.ac.be
Mon Sep 23 07:31:15 EDT 2013


Op 20/09/2013 21:23, Davide Sottara schreef:
> Brecht, thanks for reporting this, but I don't think these are bugs.
>
Thanks for the explanations. I tested it to be sure and it works as 
expected. The "drools.agent.useKBaseClassLoaderForCompiling" option 
indeed fixes the issue unless the agent detects several files at the 
same time in which case they are processed in alphabetical order (which 
is to be expected). In that case my preprocessing hack is still useful. 
The globals are just something different than what I expected globals to 
be.

Thank you very much for the support.
> 1) In order to set the value of a global from a rule's RHS, you have 
> to do somehting like this
>
> drools.getKnowledgeRuntime().setGlobal( "RULES_MATCHED", newValue );
>
> The reason is that the rule's RHS sees a reference to the original 
> global... a local variable with the
> same name initialized appropriately. Something like RULES_MATCHED = 0 
> will only change the local
> reference, not modify the global.
>
> 2) The KnowledgeAgent supports declared types incrementally.. (Esteban 
> and I spent a lot of time to make
> that work), but you have to set a flag in the KA's configuration:
>
> KnowledgeAgentConfiguration aconf = 
> KnowledgeAgentFactory.newKnowledgeAgentConfiguration();
> aconf.setProperty( "drools.agent.useKBaseClassLoaderForCompiling", true );
>
> The "incremental" mode (newInstance=false) only ensures that the 
> existing KnowlegeBase is updated,
> tue "useKBaseClassLoaderForCompiling" will ALSO make sure that, as the 
> new resources are compiled
> by a KnowledgeBuilder, this is initialized with the KB's classloader, 
> which has the classes for the declared types.
>
> 3) I have proposed a fix for the issue you reported the other day... 
> actually two in one.
> - 3a ) Runtime exceptions will be catched,
> - 3b ) you **WILL NOT** be able to use primitive types with globals 
> anymore.. (which would result in a RTE anyway)
>
> Hope this help
> Best
> Davide
>
>
>
> On 09/20/2013 02:18 AM, De Rooms Brecht wrote:
>> Dear Drools Users,
>>
>> I am building a network server for drools  since the existing 
>> drools-server did not meet my requirements.  Since I recently found a 
>> bug and find the people here very helpful I'll try to explain another 
>> issue I encountered in the hope that it improves drools 5.6 and 6.0.
>> I noticed that declared types and globals don't seem to be found when 
>> you access them from a rule that was written in a different file. For 
>> the declared types I hacked around this issue by preprocessing the 
>> files and placing every declared type at the top of each file that 
>> needs it. For globals this is of course not possible.
>>
>> An example is shown below. File1 is loaded from the moment the agent 
>> starts up, then file2 is loaded.
>> There is a difference when the knowledgeAgent detects the two files 
>> at once or one by one. In this case the KnowledgeAgent detects one 
>> file and then a few minutes later the other file and compiles them 
>> completely separately.
>> The idea is to keep how many rules are matched of a certain type in a 
>> global.
>>
>> *   FILE1: global_rules_matches.drl
>>     ----------------------------------------------
>>     package ellipsoidfacts
>>
>>     // declare
>>     global Integer RULES_MATCHED;
>>
>>     // initialize global
>>     rule "initRULESMATCHED"
>>           salience 999
>>           when
>>
>>           then
>>               RULES_MATCHED = 0;
>>     end*
>>
>> *   FILE2: testrule.drl
>>     ----------------------------------------------
>>     package ellipsoidfacts
>>
>>    rule "Gesture_lefthook"
>>     when
>>
>>         // ... any  precedent rules ...
>>
>>     then
>>         System.out.println("matched gesture:  lefthook"+ RULES_MATCHED);
>>     end*
>>
>> In this particular case, my rule is not matched. I load these rules 
>> using a changeset xml, my knowledgeagent is set to incremental (but 
>> either doesnt work).
>> The same happens when I declare types in FILE1 and use them in FILE2. 
>> When I write the type declaration in both files it works perfectly.
>> Being the same package I assumed that these two scenarios should 
>> work. Am I doing something wrong or is the agent not supposed to work 
>> like this and should a package be in one file?
>>
>> Kind Regards,
>> De Rooms Brecht
>>
>>
>> _______________________________________________
>> rules-users mailing list
>> rules-users at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/rules-users
>
>
>
> _______________________________________________
> rules-users mailing list
> rules-users at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-users


-- 
Brecht De Rooms
=================================
Phd Student
Vrije Universiteit Brussel
Research Group CISA(WISE) - DINF F.10.707
Department of Computer Science
Pleinlaan 2   - 1050 Brussels - Belgium
Tel. +32-2-629 1103
E-mail: bderooms at vub.ac.be
Website  http://wise.vub.ac.be/brecht-de-rooms
=================================

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/rules-users/attachments/20130923/e39b2453/attachment-0001.html 


More information about the rules-users mailing list