[ACT-R-users] Error with variable binding on LHS

[db30 at andrew.cmu.edu]

I just realized that this was also sent to the mailing list so I
figured I should send a follow up.

There was a bug.  I suspect it's a fairly rare set of circumstances
so I don't think it will have affected anyone else, but a fix has
been committed to the subversion archive if you want to get it.

The issue occurred if a chunk-type hierarchy was created and a
subtype explicitly specified slots which already existed in the
parent type.  Thus, something like this:

(chunk-type parent slot1 slot2 slot3)
(chunk-type (child (:include parent)) slot2 slot4)

If a model had such declarations then productions which specified
the parent type in a LHS buffer test and used that duplicate slot
may not have matched properly to a chunk of the subtype in the
buffer.

Dan

--On Friday, November 13, 2009 2:24 PM -0500 "Stu @ AGS TechNet" 
<stu at agstechnet.com> wrote:

> Hi Dan,
>     In our model, we have a situation in which a variable that is
> introduced on the LHS of a production as shown below.
>     We have a bug where the variable (=wf2) is getting bound during the
> LHS test to the value of the wrong slot.  Specifically, I have cases
> where the variable =wf2 is getting bound to the value of the slot
> letter-1 instead of the value of the slot word-form-2.
>     I am starting to step through the procedural module code to see where
> we may be getting trouble.
>     Can you tell me the name of the function(s) in the procedural module
> where variable bindings occur during LHS tests?
> Thanks
> Stu
>
> (p a-production
>    =goal>
>        isa goal-type
>     =retrieval>
>         isa word-pos
>         word-form-2 =wf2
> ==>
> ;;; etc and other RHS  actions ....
> )
>
>