Ø
Between those, you should be able to knock up something which does an appropriate audit of what's going on in your rules without refactoring any of your DRL.
Ah. I think I get it.
Based on my understanding of the code you provided, the Listener’s objectInserted() method gets a callback from the DROOLs runtime whenever a KS fact insert
event is processed. I could then use the ObjectInsertedEvent reference (passed by the callback) to do my own accounting/auditing of the fact, including the ability to establish “did this rule fail?” by providing the equivalent of the .DRL L-value logic in
the objectInserted() method’s implementation? If my understanding is correct, I could then get an *exactly* precise answer to “why did this rule fail?” by iterating on the equivalent of each of the .DRL L-value predicate participants and evaluating
“did this specific predicate participant fail to match?”. Very interesting. That would, indeed, effectively unburden the .DRL from any audit/accounting logic – moving the burden to the Listener.
Is my understanding close to how you envisioned doing this?
Very interesting. Thanks again!
From: rules-users-bounces@lists.jboss.org [mailto:rules-users-bounces@lists.jboss.org]
On Behalf Of Stephen Masters
Sent: Thursday, January 31, 2013 4:49 AM
To: Rules Users List
Subject: Re: [rules-users] Non short circuit ANDing
fyi - if you're just learning your way around working memory event listeners and the like, here's some example code. It's an extended version of something I found in the Drools docs.
A tracking working memory event listener:
Attach it to your session to keep track of all facts as they are inserted/updated/retracted. It also supports filtering for fact classes or handles so that you can pick out which ones are relevant.
A tracking agenda event listener:
Keeps track of and logs every rule which fires.
I use them in this test helper class, which enables me to create unit tests which assert that specific rules have fired or that specific facts are in working memory.
Between those, you should be able to knock up something which does an appropriate audit of what's going on in your rules without refactoring any of your DRL. I prefer using listeners like this to monitor what's going on, as they don't intrude
on your code. I tend to think that the rules should be simple and specific to the decision being made, rather than needing them to be written in a specific way in order to achieve some side effects.
Steve
On 31 Jan 2013, at 00:53, Davide Sottara <dsotty@gmail.com> wrote:
Gents,
this is a really interesting discussion and I wish I had more time to participate.. I agree with Stephen that some constraints are really worthy monitoring,
while others are simply "filters".. Wolfgang's mention of "near misses" is potentially a hot topic, I have been experimenting with "imperfect" rules which
are more robust in that sense and provide better traceability, but that is not yet implemented in the core engine.
I would also point out that one thing is monitor constraints for "logging" or "auditing" purposes, another is the ability to specify different behaviours
(i.e. consequences) according to the success vs failure of an evaluation.
To this end, and following on the "extending" rule thread, it has actually been planned and partially implemented.. it's only an experimental feature
(@Mario, any update/clarification?), so it's probably not well documented. But I would point you to these links:
http://blog.athico.com/2012/09/conditional-branches-with-sub-blocks.html
https://community.jboss.org/wiki/BranchCEs
with the warning that it's still experimental and, while adopting "familiar" keywords (if, then, else, switch), their semantics is different from the
usual imperative language one, and in fact could enable some of the "extend" capabilities that have been invoked in the last couple of posts.
Best
Davide
On 01/30/2013 02:25 PM, Cotton, Ben wrote:
automatically rename each rule "part" (being a single-condition rule derived from the original) to some derivation of the original rule name, and need the consequence
only in the fully-extended rule, thereby giving the same effect, but with rule "parts" that would reflect failure at the individual condition level
This is *exactly* the ambition we are hoping that some Drools .DRL enrichment tool (or other Drools capability) can provide.
Wish I had sufficient competency to comment on the merits of your different suggested approaches. Hopeful that the Drools gurus will comment explicitly on all these approaches
(and/or the approach suggested by another of using Working Memory listeners).
_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users