Hi Rob,

another approach we're considering to version rulesets used to perform mortgage risk assessment looks roughly as follows:

- each of the rulesets is versioned as a whole (i. e. compile all *.drl pertaining to a ruleset into a *.pkg file and give this *.pkg a version number).
- the scoring service is aware that rules evolve and selects the ruleset which fits the "processing date". the processing date is part of the scoring request we send to the service
- scoring service loads selected ruleset and performs risk assessment, sending back risk indicators

We don't like the idea of time-dependent rules very much, as this introduces *versioning* into the rule source code itself, which is additional complexity we'd like to keep out of the *.drl files. We want to be able to drop old rules from the source code once they are not valid any more, but still have them compiled with a version number somewhere in a repository, just in case a mortgage application needs re-scoring based on an older version of the rules. We can afford this rather "static" approach as our rules don't change too often (like a few rules every week or two).

Best regards

Ansgar


Am 24.08.2011 21:42, schrieb Rob Fisher:
Looking for best practice recommendations around the versioning of a rule.  Our application will have 50-60 rule sets comprised of up to 70 rules each.  The activation of many of the rules will be based on a processing date (inserted fact attribute).  The application processing date can be prior, equal, or after the system date.
 
 
For example, 2 rules, with slightly different logic
 
Eligibility rule version 1
//to be potentially activated  for application processing time period prior to 2011
When
     Car(color = “blue”)
Then
Terms(eligible = “yes”)
 
Eligibility rule version 2
//to be potentially activated for application processing time period 2011 and forward
When
   Car(color = “blue”, type=”coupe”)
Then
  Terms(eligible = “yes”)
 
We’ve come up with a couple of different ways to handle this:
  1. Add Date logic to LHS of the rule . i.e add “Control(processingDate < ‘01/01/2011’)” to 1st rule and  “Control(processingDate >= ‘01/01/2011’)” to 2nd rule. First rule would activate with processing date = 12/31/2010 and 2nd rule would activate with processing date = 07/31/2011.  When originally authored, first rule would be written without date logic.  When 2nd rule is authored, data logic will need to get added to both.
  2. We explored a way of setting the session clock to the processing date and using the ‘date-effective’ and ‘date-expires’ attributes to do the filtering.  i.e. 1st rule is effective until 12/31/2010 and 2nd rule is effective from 01/01/2011 to 12/31/9999.  However, we could only figure out how to do this with a stateful knowledge session.  Is it possible in a Stateless?
  3. We added metadata to the rules (@EffectiveDate and @ExpirationDate), and using agendafilter, we compared processing date to the metadata and return true or false dependent on comparison.
 
What are your opinions on these three methods?  What are other options?
 
Thanks
Rob
 
 
 
 

_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users