[rules-dev] Locating objects by query/BC
Mark Proctor
mproctor at codehaus.org
Mon Jun 27 09:34:54 EDT 2011
On 27/06/2011 10:38, Wolfgang Laun wrote:
> I saw a commit and tried again, but the problem is still around.
> I added a test to BackwardChainingTest - see the pull request.
I have created a jira to track this:
https://issues.jboss.org/browse/JBRULES-3101
I'm going to first write core unit tests around the rete dsl, as it's
easier to test multi permutations for left/right input triggering and
will better stop future regressions.
commit here shows me updating the rete dsl testing engine for queries:
https://github.com/droolsjbpm/drools/commit/279c9df98434d668d6adc837982a40ae6b9c97d3
i've added a basic test to start with that shows it working for right
then left triggering, but failing for left then right triggering.
https://github.com/droolsjbpm/drools/blob/279c9df98434d668d6adc837982a40ae6b9c97d3/drools-core/src/test/resources/org/drools/reteoo/test/JoinNodeUnificationTest.nodeTestCase
At the same time i'm applying the fixes, which needs to be done to each
node. In essence the code to get the memory iterator and the first entry
for each left tuple assertion needs to be applied to each right object
assertions and all the modifies.
Mark
>
> -W
>
> On 25 June 2011 14:10, Mark Proctor <mproctor at codehaus.org
> <mailto:mproctor at codehaus.org>> wrote:
>
> I have isolated the issue and know what it is, and why it works
> one way and not the other. In short on unification if the variable
> is unbound we need to turn off indexing for that specific join
> attempt. We do this when it's left triggered, we don't do this
> when it's right triggered. Updated the code to also do this on
> right triggering. Should have it fixed tonight.
>
> Thanks
>
> Mark
>
> On 25/06/2011 12:38, Mark Proctor wrote:
>> ok thanks, I reproduced the problem with the following below. If
>> you change the String insertion order it works. I'll get this
>> fixed over the weekend:
>> @Test
>> public void testQueryWithObject() throws Exception {
>> String str = "" +
>> "package org.drools.test \n" +
>>
>> "import java.util.List\n" +
>> "import java.util.ArrayList\n" +
>>
>> "global List list\n" +
>>
>> "dialect \"mvel\"\n" +
>> "\n" +
>>
>> "import
>> org.drools.integrationtests.BackwardChainingTest.Q\n" +
>> "import
>> org.drools.integrationtests.BackwardChainingTest.R\n" +
>> "import
>> org.drools.integrationtests.BackwardChainingTest.S\n" +
>>
>> "query object(Object o)\n" +
>> " o := Object() \n" +
>> "end\n" +
>>
>> "rule collectObjects when\n" +
>> " String( this == 'go1' )\n" +
>> " object( o; )\n" +
>> "then\n" +
>> " list.add( o );\n" +
>> "end\n" +
>>
>>
>> "rule init when\n" +
>> " String( this == 'init' )\n" +
>> "then\n" +
>> " insert( new Q(1) );\n " +
>> " insert( new Q(5) );\n " +
>> " insert( new Q(6) );\n " +
>> " insert( new R(1) );\n " +
>> " insert( new R(4) );\n " +
>> " insert( new R(6) );\n " +
>> " insert( new R(2) );\n " +
>> " insert( new S(2) );\n " +
>> " insert( new S(3) );\n " +
>> " insert( new S(6) );\n " +
>> "end\n" +
>> "";
>>
>> KnowledgeBuilder kbuilder =
>> KnowledgeBuilderFactory.newKnowledgeBuilder();
>> kbuilder.add( ResourceFactory.newByteArrayResource(
>> str.getBytes() ),
>> ResourceType.DRL );
>>
>> if ( kbuilder.hasErrors() ) {
>> fail( kbuilder.getErrors().toString() );
>> }
>>
>> KnowledgeBase kbase =
>> KnowledgeBaseFactory.newKnowledgeBase();
>> kbase.addKnowledgePackages(
>> kbuilder.getKnowledgePackages() );
>>
>> kbase = SerializationHelper.serializeObject( kbase );
>>
>> StatefulKnowledgeSession ksession =
>> kbase.newStatefulKnowledgeSession();
>> List<Integer> list = new ArrayList<Integer>();
>> ksession.setGlobal( "list", list );
>>
>> ksession.insert( "go1" );
>> ksession.fireAllRules();
>>
>> ksession.insert( "init" );
>> ksession.fireAllRules();
>>
>> System.out.println( list );
>> }
>>
>> Mark
>> On 25/06/2011 11:22, Wolfgang Laun wrote:
>>> // This rule prints all objects in WM.
>>> rule findObject
>>> when
>>> $o: Object()
>>> then
>>> System.out.println( "Object: " + $o );
>>> end
>>>
>>> Happily I get
>>> Object: 42
>>> Object: a String object
>>> Object: org.drools.reteoo.InitialFactImpl at 4dde85f0
>>>
>>> Now the query way:
>>> query object( Object o )
>>> o := Object()
>>> end
>>> rule findObjectByQuery
>>> when
>>> object( $a ; )
>>> then
>>> System.out.println( "Object by query: " + $a );
>>> end
>>>
>>> Surprisingly, I get just one:
>>> Object by query: org.drools.reteoo.InitialFactImpl at 4dde85f0
>>>
>>> A marvellous effect can be observed when the condition in rule
>>> findObjectByQuery is modified to
>>> when
>>> Object()
>>> object( $a ; )
>>> then
>>> I'd have guessed that the above result would appear repeatedly,
>>> once for each Object in WM,
>>> but that's not so. For N-1 facts the rule fires N*(N+1)/2 times,
>>> showing the "initial object"
>>> N times, the first inserted object N-1 times, the second one N-2
>>> times and, finally, the last one
>>> once.
>>> Object by query: 42
>>> Object by query: org.drools.reteoo.InitialFactImpl at 4dde85f0
>>> Object by query: a String object
>>> Object by query: 100
>>> Object by query: 42
>>> Object by query: org.drools.reteoo.InitialFactImpl at 4dde85f0
>>> Object by query: a String object
>>> Object by query: org.drools.reteoo.InitialFactImpl at 4dde85f0
>>> Object by query: a String object
>>> Object by query: org.drools.reteoo.InitialFactImpl at 4dde85f0
>>>
>>> -W
>>>
>>>
>>> _______________________________________________
>>> rules-dev mailing list
>>> rules-dev at lists.jboss.org <mailto:rules-dev at lists.jboss.org>
>>> https://lists.jboss.org/mailman/listinfo/rules-dev
>>
>>
>> _______________________________________________
>> rules-dev mailing list
>> rules-dev at lists.jboss.org <mailto:rules-dev at lists.jboss.org>
>> https://lists.jboss.org/mailman/listinfo/rules-dev
>
>
> _______________________________________________
> rules-dev mailing list
> rules-dev at lists.jboss.org <mailto:rules-dev at lists.jboss.org>
> https://lists.jboss.org/mailman/listinfo/rules-dev
>
>
>
> _______________________________________________
> rules-dev mailing list
> rules-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/rules-dev/attachments/20110627/0e96659b/attachment-0001.html
More information about the rules-dev
mailing list