fwiw, my Day Job is semantics-centric, so I do anticipate reasoning
stuff to make its way into Drools at some point.
-Bob
On Jan 23, 2006, at 12:26 PM, Andreas Andreakis wrote:
I agree, that it is not a beautiful solution.
But the problem is, that you can´t tell if two Objects are equal if
you dont define equals() , hashcode() on your own. How will your
dublication-preventing Rule look like ?
Ontology-optimized rule engines such as Jena or Sesame (openrdf)
work internaly with Triple Objects (Resource, Property, Object)
using Strings. Thus the framework can hande the hashcode
transparently to the user and you never get dublicated Facts after
applying Rules on the Model. But, you need to map your
BusinessObjects into the RuleEngine by calling Framework-specific
creationMethods manually , which is currently a pretty annoying
task...reminds me on the days without Hibernate when working with
JDBC and SQL.....
Anyway...from my point of view, this is simmilar to Object-Rational
mapping. Here we *could* have Object-Fact mapping. An easier
solution would be to use Annotations..to mark the attributes who
are considert to define hashCode and thus the Objectidentity.
Because we know that Reflection is evil, you could generats a Proxy-
Class using lets say Javassist, which wrapps your JavaClass and
generates a hashcode Method based on the Annotations you have set.
I think that a mapping-file in XML should not be necessary.
Implementation-steps:
1) assertObject into workmemory
2) get Class of Object and check if Drools-related Annotations are set
3) Generate HashCodeProxy
4) have fun - without Dublications
Peter Van Weert schrieb:
This will probably work, but is no longer declarative, rule-based
programming. If you do not care about this semi-philosophical
argument, stop reading here...
The correct way to solve the problem in my opinion is to add an
extra rule that removes duplicates (in Drools you should probably
give it a higher salience...). So you simply add away, and /
declare/ that duplicates have to be removed. The responsibility of
using a hash-set (or any other efficient data-structure) to
efficiently look up duplicate facts should lie with the rule-
engine, not with the user.
As always, I do not know the details of the internals of Drools,
so your solution might in fact be faster, but it remains a hack in
my opinion...
Greets,
Pé
