Re: [PRD] need Assign action

I think a better solution is for PRD to support Skolem functions 
(logical functions that occur only in the conclusion), thus carving out 
a slightly larger Core.

Michael Kifer wrote:
> Apropos: we talked about _new (a.k.a. Skolem) for BLD use cases, but never
> decided anything. Should we try to put it in for the last call?
>
>
>  --michael  
>
>
> On Wed, 18 Jun 2008 09:52:28 -0700
> Gary Hallmark <gary.hallmark@oracle.com> wrote:
>
>   
>> Adrian Paschke wrote:
>>     
>>> It can not be simulated by a retract action followed by a new assert,
>>> because if PRD has a negation (not) the retract might trigger a production
>>> rule transition. But that is not the intended semantics of assign/update.
>>>   
>>>       
>> what I meant was to add Assign to the definition of the transition 
>> relation ?__RIF-PRD ? /P(W)/ × /L/ × /P(W)/ defined in Section 3.3 of 
>> the PRD spec.  Assign would be defined as a single transition, not as 
>> pair of (retract, assert) transitions.
>>
>> Section 3.3 (and associated syntax section) needs some work in any 
>> case.  There is no way to do what OBR calls "assert new", which means to 
>> create a new frame with a brand new OID.  The existing PRD assert can 
>> only take an existing frame and add a slot to it.
>> Also, in OBR and in Jess, Retract would not take a frame argument, but 
>> rather would take an OID and would retract all the frames that match 
>> that OID.
>>

Received on Wednesday, 18 June 2008 20:28:15 UTC