Hi there,
I personally think that this actually works around the basic idea of a
fact/rule system like Jess - a rule fires whenever all its conditions
are met and this will be tested whenever the facts in question are
modified. In other words - the behavior as it is right now is absolutely
correct.
What I would find favourable is a modified behavior in a way, that a
rule gets activated by a modified fact *only* if a slot applicable to
the rule had been changed, i.e. in this example
> (defrule my-rule
> ?fact <- (fact (slot ?x))
> =>
> (modify ?fact (slot (+ ?y 1))))"
should not fire again, but only if x has changed. I know, that this is
actually something different than initially proposed, but well ...
Regards,
tge
[EMAIL PROTECTED] wrote:
>
> Hi Folks,
>
> Very often, people post questions equivalent to the following to this
> list:
>
> "Why does this cause an infinite loop?
>
> (defrule my-rule
> ?fact <- (fact (slot ?x))
> =>
> (modify ?fact (slot (+ ?x 1))))"
>
> and then we have to explain why this happens, and how to avoid it
> (it's a side-effect of how the Rete algorithm works, and you can avoid
> it by making the rule more specific, so the modified fact won't match
> anymore.)
>
> Here's a proposed feature that would provide an alternative way to
> avoid the infinite loop -- and I think a more powerful way: "one-shot"
> rules. Basically, the feature is this: if you declare a rule like
> this:
>
> (defrule my-rule
> (declare (one-shot TRUE))
> ?fact <- (my-fact (slot ?x))
> =>
> (modify ?fact (slot (+ ?x 1))))
>
> then my-rule will keep track of all the combinations of facts it has
> ever fired on, and will never fire on the same combination twice, so
> that, for example, my-rule would fire exactly once for each my-fact.
>
> The performance overhead would bne relatively small, as the lists of
> past firings could be kept in a hash table and so you'd just see
> amortized constant-time overhead for each potential activation of a
> one-shot rule (and no impact at all for non-one-shot rules.)
>
> Does this sound like a useful feature? Do the semantics sound
> reasonable? Any suggestions or refinements? Should there be a global
> switch to make ALL rules one-shot?
>
> ---------------------------------------------------------
> Ernest Friedman-Hill
> Distributed Systems Research Phone: (925) 294-2154
> Sandia National Labs FAX: (925) 294-2234
> PO Box 969, MS 9012 [EMAIL PROTECTED]
> Livermore, CA 94550 http://herzberg.ca.sandia.gov
>
> --------------------------------------------------------------------
> To unsubscribe, send the words 'unsubscribe jess-users [EMAIL PROTECTED]'
> in the BODY of a message to [EMAIL PROTECTED], NOT to the list
> (use your own address!) List problems? Notify [EMAIL PROTECTED]
> --------------------------------------------------------------------
--
........................................................
Thomas Gentsch
--------------
Phone: +49 711 486948 E-mail: [EMAIL PROTECTED]
Mobil: +49 173 6620507 WWW: www.e-tge.de
Fax: +49 711 4687889 www.blue-elephant-systems.com
........................................................
--------------------------------------------------------------------
To unsubscribe, send the words 'unsubscribe jess-users [EMAIL PROTECTED]'
in the BODY of a message to [EMAIL PROTECTED], NOT to the list
(use your own address!) List problems? Notify [EMAIL PROTECTED]
--------------------------------------------------------------------
