An absolutely watertight and foolproof system would know which attributes (or getter methods) change due to which setter method calls (and, beyond that, whether the setter really changes the value). This, and only this, would be the summit of "fidelity".

Anything else is an approximation, and, depending on the measure you apply, one procedure may be better than another one. Missing a change is more difficult to circumvent with legacy techniques than it is to avoid spurious firings.

Given that the aforementioned analysis is extremely difficult, it is obvious that a truly high-fidelity reaction to changes can't be done without users providing dependency information by means of annotations or some similar technique

-W
 


On 16 January 2012 14:46, Swindells, Thomas <TSwindells@nds.com> wrote:

The change is reducing the coarseness of the system and so making it more sensitive to the scope of the change so I’d say high fidelity is the correct term.

I agree with you I’m concerned how this works with non-pure java beans where some of the setters may update multiple fields, or there may be none bean methods like a reset() method – I don’t understand how these would interact in the system, is there a mechanism to explicitly invalidate a property or mark the whole lot as dirty.

 

Thomas

 

From: rules-dev-bounces@lists.jboss.org [mailto:rules-dev-bounces@lists.jboss.org] On Behalf Of Wolfgang Laun
Sent: 14 January 2012 08:59
To: Rules Dev List
Subject: Re: [rules-dev] Fine Grained Property Change Listeners (Slot Specific)

 

You really should follow naming conventions Java programmers are familiar

with. javax.xml.bind.annotation.XmlAccessType, for instance, uses the

term "member" for referring to "public getter/setter pairs", and this comes

close to what Drools does. Alternatively, there's "field" and "property",

and I wouldn't worry about the extra lip mileage - you can always abbreviate

to "prop-specific".

 

Is this new feature in some way configurable for the KnowledgeBase?

There are fact types where modifying one member results in the change

of more than one member. Consider the very plausible case:

     $f: Fact(...)

   then

      modify( $f ){ getList().add( $x ) }

   end

 

   when

     Fact( size > 10 )  # size implemented as getList().size()

   then

 

Wolfgang

 

PS: The new feature is actually restricting the sensitivity of the system.

Therefore, I feel that "high-fidelity" is actually counter-intuitive!

 

On 13 January 2012 23:30, Mark Proctor <mproctor@codehaus.org> wrote:

Mario just got a first cut working for fine grained property change
listeners. Previously when you call update() it will trigger revaluation
of all Patterns of the matching object type in the knowledeg base.

As some have found this can be a problem, forcing you to split up your
objects into smaller 1 to 1 objects, to avoid unwanted evaluation of
objects - i.e. recursion or excessive evaluation problems.

The new approach now means the pattern's will only react to fields
constrained or bound inside of the pattern. This will help with
performance and recursion and avoid artificial object splitting.  We
previously discussed this here:
http://blog.athico.com/2010/07/slot-specific-and-refraction.html
You can see the unit test here:
https://github.com/droolsjbpm/drools/blob/ca55c78429cbc0f14167c604c413cdc3faaf6988/drools-compiler/src/test/java/org/drools/integrationtests/MiscTest.java

The implementation is bit mask based, so very efficient. When the engine
executes a modify statement it uses a bit mask of fields being changed,
the pattern will only respond if it has an overlapping bit mask. This
does not work for update(), and is one of the reason why we promote
modify() as it encapsulates the field changes within the statement. You
can follow Mario's chain of work on this at his github activity feed:
https://github.com/mariofusco.atom

The adventerous amoung you can pick this up from hudson, or from maven,
and start playing now. My hope is that this will make drools much easier
to use:
https://hudson.jboss.org/hudson/job/drools/lastSuccessfulBuild/artifact/drools-distribution/target/

Btw we are after a name. Drools is not a frame based system, so "slot
specific" doesn't seem appropropriate. Property Specific seems a bit of
a mouth full. I'm quite liking High Fidelity Change Listeners :) any
other suggestions?

slot-specific is the name used by Jess for this feature,
http://www.jessrules.com/docs/71/constructs.html. It's also the standard
way that Clips COOL works, which is the Clips OO module. Although that's
partly a side effect of the triple representation of properties used in
COOL, and the modifications are triple based. I don't know what
mechanism Jess is using to enable this.

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

 



**************************************************************************************
This message is confidential and intended only for the addressee. If you have received this message in error, please immediately notify the postmaster@nds.com and delete it from your system as well as any copies. The content of e-mails as well as traffic data may be monitored by NDS for employment and security purposes. To protect the environment please do not print this e-mail unless necessary.

NDS Limited. Registered Office: One London Road, Staines, Middlesex, TW18 4EX, United Kingdom. A company registered in England and Wales. Registered no. 3080780. VAT no. GB 603 8808 40-00
**************************************************************************************

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