Edson - Thanks for the reply. I am setting the global prior to calling the rules. That said, I was able to fix my problem by removing the evals from my rules, using the method described by Mohammed Riyaz in a post yesterday (thanks Mohammed!). Removing the evals also remedied a problem we were seeing where our rules using evals would occasionally fail due to null pointer exceptions throw by PredicateConstraint when being run by a multi-thread job (very similar to what's noted in issue JBRULES-1751). I defined all the methods in our RulesUtil class as static, imported the functions, and referenced them inside an exists() clause like this: exists( Boolean( booleanValue==true) from containsKeyword($myString) ) -David
________________________________ From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Edson Tirelli Sent: Tuesday, September 16, 2008 7:26 PM To: Rules Users List Subject: Re: [rules-users] Request for assistance in removing eval from a rule Are use setting the global in your application code? 2008/9/16 Warren, David [USA] <[EMAIL PROTECTED]> Edson - Thank you very much for the reply. I have one more question. In the example below I changed RulesUtil to be a global name 'ruleUtil' (since its member variables do not change), and reran the rule (with the 'ruleUtil:RulesUtil()' line omitted). When I do this, and execute the rule below, I get a null pointer exception on PredicateConstraint. The first and last parts of the stack trace are: org.drools.RuntimeDroolsException: Exception executing predicate [EMAIL PROTECTED] at org.drools.rule.PredicateConstraint.isAllowed(PredicateConstraint.java:2 16) at org.drools.reteoo.AlphaNode.assertObject(AlphaNode.java:132) at org.drools.reteoo.CompositeObjectSinkAdapter.propagateAssertObject(Compo siteObjectSinkAdapter.java:318) at org.drools.reteoo.AlphaNode.assertObject(AlphaNode.java:140) at org.drools.reteoo.CompositeObjectSinkAdapter.propagateAssertObject(Compo siteObjectSinkAdapter.java:299) at org.drools.reteoo.ObjectTypeNode.assertObject(ObjectTypeNode.java:153) ... Caused by: java.lang.NullPointerException at com.bah.aims.rules.Rule_my_rule__0.returnValue0(Rule_my_rule_0.java:27) at com.bah.aims.rules.Rule_my_rule_0ReturnValue0Invoker.evaluate(Rule_myRul e_0ReturnValue0Invoker.java:21) at org.drools.rule.PredicateConstraint.isAllowed(PredicateConstraint.java:2 10) ... 29 more I'm not sure what is causing this. Any thoughts? If it helps, RulesUtil.containsKeywordString() checks to see if a string passed in is present in a list of keywords held in the class. Thanks - David ________________________________ From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Edson Tirelli Sent: Tuesday, September 16, 2008 9:56 AM To: Rules Users List Subject: Re: [rules-users] Request for assistance in removing eval from a rule David, Eval is really bad for perf, but still, sometimes it is the only way to do things like call methods. So, in your case, it all depends on what "containsKeywordString()" method is doing? Also, is RulesUtil just a utility class? I mean with no data you reason over? if so, I strongly advise you to make it a global, instead of matching it in the rule like a fact. Regarding evals, inline evals are just a bit lighter than top level evals, but there is an important difference: * inline evals MUST be time consistent, i.e., they must evaluate to the same value everytime they are called. * top level evals can deal with changes appropriately []s Edson 2008/9/15 Warren, David [USA] <[EMAIL PROTECTED]> Folks - After seeing the thread on evals hurting performance a couple of weeks ago, I tried to remove them from a rule set we are using (running Drools 4.0.4), and have had limited success. An example is shown below. We're using evals every time we make a call to "rulesUtil", a helper class we have for checking strings for keywords. Two questions: 1. Does using inline evals (like below) hurt performance as much as using non-inline evals? 2. If so, any thoughts for how to rewrite this rule to avoid using the eval? rule 'my rule' salience 790 activation-group "priorityRule" when ruleUtil : RulesUtil( ) s1 : Sensor( source == "X, $rfp : RFP , eval( ruleUtil.containsKeywordString($rfp, "Y" )) , $tcn : TCN) s2 : Sensor( TCN != $tcn, source == "Z" then System.out.println("my rule"); priority.setPriorityName("A"); end Thanks, David Warren ________________________________ From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Edson Tirelli Sent: Friday, September 05, 2008 2:59 PM To: Rules Users List Subject: Re: [rules-users] Drools, Performance issues on 4.0.7 Vs 2.5 Start by eliminating the evals and writing your constraints properly inside the patterns. Drools 3+ is orders of magnitude faster than Drools 2.x, but you need to leverage its power in your rules. Please read the manual as the version 3 was a completely rewrite of version 2. Version 4 is an improvement over 3. As an example, look at this: http://blog.athico.com/2006/11/rush-hour-and-content-based-routing.html []s Edson 2008/9/5 Rout, Sushanta (ThoughtMill) <[EMAIL PROTECTED]> We were using Drools 2.5 version earlier. Now we have switched to drools 4.0.7. But we see significant issues with performance like drools 4.0.7 is three times slower than 2.5 . Has anybody encountered the issue? Here is a sample of the rule, we have some more similar to this. rule "test" dialect "java" activation-group "group1" when $croNumberDetailsRequest : CRONumberDetailsRequest() $resdirectPhoneNumber : ResdirectPhoneNumber() eval($resdirectPhoneNumber.getType().getId() == 5 && $resdirectPhoneNumber.isValidForBrand($croNumberDetailsRequest.getBrand( )) && $croNumberDetailsRequest.getRegion() != null && $resdirectPhoneNumber.isValidRegion($croNumberDetailsRequest.getRegion() ) && $croNumberDetailsRequest.getLocale() != null && $resdirectPhoneNumber.isValidForLocaleId($croNumberDetailsRequest.getLoc ale()) && $resdirectPhoneNumber.isValidSlot($croNumberDetailsRequest.getSlot())); then ResultList.add($resdirectPhoneNumber); end _______________________________________________ rules-users mailing list [email protected] https://lists.jboss.org/mailman/listinfo/rules-users -- Edson Tirelli JBoss Drools Core Development JBoss, a division of Red Hat @ www.jboss.com _______________________________________________ rules-users mailing list [email protected] https://lists.jboss.org/mailman/listinfo/rules-users -- Edson Tirelli JBoss Drools Core Development JBoss, a division of Red Hat @ www.jboss.com _______________________________________________ rules-users mailing list [email protected] https://lists.jboss.org/mailman/listinfo/rules-users -- Edson Tirelli JBoss Drools Core Development JBoss, a division of Red Hat @ www.jboss.com
_______________________________________________ rules-users mailing list [email protected] https://lists.jboss.org/mailman/listinfo/rules-users
