Re: lisa bug: matching on retracted facts

Hi,

I have a much simpler test case which more clearly demonstrates what's 
going on, and a workaround:

-------- [lisp code] --------

(in-package :lisa-user)

(clear)

(deftemplate control-fact ())

(deftemplate multiple-facts ()
   (slot instancename))

(defrule broken-rule ()
   ;; NOTE: swapping the order of these two patterns fixes the problem
   (control-fact)
   (multiple-facts (instancename ?name))
   =>
   (format t "This rule shouldnt fire (instance name=~a)~%" ?name))

(defrule startup ()
   =>
   (assert (multiple-facts (instancename first))) ; broken-rule doesn't 
fire on the first instance
   (assert (multiple-facts (instancename second))) ; but it does fire on 
each subsequent one.
   (assert (multiple-facts (instancename third)))
   (assert (multiple-facts (instancename fourth)))
   (retract (assert (control-fact))))

(watch :all)
(reset)
(run)

-------- [output] --------

==> F-0 (INITIAL-FACT)
==> Activation: STARTUP : (F-0)
FIRE 1: STARTUP (F-0)
==> F-1 (MULTIPLE-FACTS (INSTANCENAME FIRST))
==> F-2 (MULTIPLE-FACTS (INSTANCENAME SECOND))
==> F-3 (MULTIPLE-FACTS (INSTANCENAME THIRD))
==> F-4 (MULTIPLE-FACTS (INSTANCENAME FOURTH))
==> F-5 (CONTROL-FACT)
==> Activation: BROKEN-RULE : (F-5 F-1)
==> Activation: BROKEN-RULE : (F-5 F-2)
==> Activation: BROKEN-RULE : (F-5 F-3)
==> Activation: BROKEN-RULE : (F-5 F-4)
<== F-5 (CONTROL-FACT)
   0: (LISA::REMOVE-FACT-FROM-NETWORK #<RETE-NETWORK {9EE0A99}>
                          #<FACT F-5, CONTROL-FACT, CF is 0.000>)
<== Activation: BROKEN-RULE : (F-5 F-1)
   0: LISA::REMOVE-FACT-FROM-NETWORK returned NIL
FIRE 2: BROKEN-RULE (F-5 F-2)
This rule shouldnt fire (instance name=SECOND)
FIRE 3: BROKEN-RULE (F-5 F-3)
This rule shouldnt fire (instance name=THIRD)
FIRE 4: BROKEN-RULE (F-5 F-4)
This rule shouldnt fire (instance name=FOURTH)

----------------

(I have (trace lisa::remove-fact-from-network) enabled there).

I suspect that the remove-fact token isn't being pushed down all the way 
through the network, but I have no idea why.  I'm currently trying to 
figure out how that code works.

However!  If you swap the order of the two patterns in broken-rule, it 
retracts all potential activations as expected.

-Andy


-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted by HP,
AMD, and NVIDIA.  To register visit http://www.hp.com/go/dualwebinar