Re: [Core](PRD] safeness

I think that

Forall x if Not(Px) then print("hello world")

is equivalent to

if Not(Exists x (Px)) then print("hello world")

Both are safe.

But

if Not(Exists x (x=1)) then print("hello world")

is not safe.

Let's just focus for now on the formulas of the form Not(Exists ...)

I don't see how your safeness definition for PRD distinguishes the
safe ones from the unsafe ones.

On Wed, Jun 10, 2009 at 2:13 PM, Christian De Sainte
Marie<csma@fr.ibm.com> wrote:
>
> ********* NOTICE **********
> My new email address at IBM is: csma@fr.ibm.com
> My ILOG email address will not be forwarded after June 8
> *****************************
>
>
> gary.hallmark@gmail.com wrote on 10/06/2009 18:47:28:
>>
>> In production rule systems such as CLIPS and Jess (and JRules, too, I
>> think), if a rule such as
>>
>> if P then Q
>>
>> is safe, and if Q references no variables in P, then
>>
>> if not(P) then Q
>>
>> is also safe.
>
> I do not think so. At least, not always:
>
> Forall x If Px Then print("hello world") is safe
>
> and Forall x if Not(Px) then print("hello world") is not. I would be
> surprised if JRules accepted such a rule. Changhai?
>
> I do not know about Jess and CLIPS, but I would be surprised, too, if the
> accepted the rule.
>
>> I will not be happy with a PRD safeness definition that does not
>> properly account for this.
>
> I will not be happy with a definition that does not account for what the
> systems we want to cover do, either.
>
> If Q does not reference any variable that is in P, then the only restriction
> is that all the variable in the condition must be bound in the condition.
> Which means that, if your entire condition, P, is negated, then the only
> thing that is allowed for P is an existential formula, because existentially
> quantified variables need only be bound inside the scope of the quantifier;
> and you cannot move an existential quantifier out of a negation (not without
> changing it for an universal, which you do not want to do).
>
> So, if P is an existential formula, and there are no (universally
> quantified) rule variables outside of P, then, indeed, "if not(P) then Q" is
> safe.
>
> I think (but that must be confirmed by Changhai), that this (not exists) is
> the only case where an entirely negated condition is allowed in JRules. Can
> you give examples of what is allowed in CLIPS and Jess and that would go
> beyond that, so that we can repair the definition, if it is repairable?
>
> One thing that needs be added in PRD, anyway, if just for the sake of
> clarification, is that existential quantifiers cannot be moved out of a
> negation, in the preprocessing step.
>
> Cheers,
>
> Christian
>
>
> ILOG, an IBM Company
> 9 rue de Verdun
> 94253 - Gentilly cedex - FRANCE
> Tel. +33 1 49 08 35 00
> Fax +33 1 49 08 35 10
>
>
>
> Sauf indication contraire ci-dessus:/ Unless stated otherwise above:
> Compagnie IBM France
> Siège Social : Tour Descartes, 2, avenue Gambetta, La Défense 5, 92400
> Courbevoie
> RCS Nanterre 552 118 465
> Forme Sociale : S.A.S.
> Capital Social : 609.751.783,30 €
> SIREN/SIRET : 552 118 465 02430
>
>



-- 
Cheers,

Gary Hallmark

Received on Wednesday, 10 June 2009 23:47:32 UTC