Re: JESS: Linked list / Sorted List Design Pattern

[Gary Napier]

Hi Wolfgang,
thanks for the proposed solution. My case is a little more complex than
this.

Because of the data skew and possibility of missing events, i keep the
initial facts in working memory for a long period of time (48hrs at
current settings, to allow delayed comms)
So the computation would become burdensome as each new open close fact
entered working memory and was tested against every existing fact. I was
hoping to have one list type construct that each new fact can be judged
against and
then included in that construct, then when the 48hrs are up, pairs can
be picked off the construct like removing the head of a list, at which
point the initial facts can be retracted.

So pseudo code

assert OPEN 0845
open is added to list, no computation as list is empty
assert OPEN 0815
open is added to the list, ahead of OPEN 0845 as time0815 < time0845

(same happens for close on its list construct)
(48 hrs pass)
rule fires to remove head of both lists, assert an access fact and
retract the open and close facts that generated the initial list entries

I have been perusing this idea and trying to relate it to multislots or
have an extra slot for an index number to represent the elements of the
list, but i'm new to multislots

Any help would be appreciated.

Thanks
Gary


Wolfgang Laun wrote:

As you state it, there is no need to sort all open and close events so
that you can combine an OPEN to ist closest CLOSE. The conditions that
define, for any OPEN, the matching CLOSE are simple enough:

deftemplate Open   (slot time))
(deftemplate Close  (slot time))
(deftemplate Access (slot begin)(slot end))

(defrule MatchOpenClose
   "For any matching OpenClose, create an Access"
   ?open  <- (Open  (time ?otime))
   ; a matching CLOSE must not be earlier
   ?close <- (Close (time ?ctime &:(>= ?ctime ?otime)))
   ; there must not be a closer CLOSE
   (not (Close (time ?xtime &:(< (- ?xtime ?otime)(- ?ctime ?otime)
=>
   (retract ?open)
   (retract ?close)
   (assert (Access (begin ?otime)(end ?ctime)))
)

(deffacts Opcl
   (Open  (time  845))
   (Open  (time 1045))
   (Close (time  900))
   (Close (time 1115))
   (Open  (time 1215))
   (Close (time 1215))
)

(reset)
(run)
(facts)


f-0   (MAIN::initial-fact)
f-7   (MAIN::Access (begin 845) (end 900))
f-8   (MAIN::Access (begin 1045) (end 1115))
f-9   (MAIN::Access (begin 1215) (end 1215))
For a total of 4 facts in module MAIN.

Regards
Wolfgang


On Fri, May 9, 2008 at 8:41 PM, Gary Napier <[EMAIL PROTECTED]
> wrote:

My basic problem is this.
I have two events, OPEN and CLOSE, both timestamped

I would like to relate an OPEN and CLOSE to generate a new fact X
However with data skew, there may be a mix of OPEN and CLOSE facts
mixed, such that

OPEN 7:15
CLOSE 9:00
CLOSE 8:00
OPEN 8:15

Would there be any way of sorting these facts (possibly in a multi
slot
or using an extra slot) so that they were ordered.






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]

Re: JESS: Linked list / Sorted List Design Pattern

[Wolfgang Laun]

As you state it, there is no need to sort all open and close events so that
you can combine an OPEN to ist closest CLOSE. The conditions that define,
for any OPEN, the matching CLOSE are simple enough:

deftemplate Open   (slot time))
(deftemplate Close  (slot time))
(deftemplate Access (slot begin)(slot end))

(defrule MatchOpenClose
   "For any matching OpenClose, create an Access"
   ?open  <- (Open  (time ?otime))
   ; a matching CLOSE must not be earlier
   ?close <- (Close (time ?ctime &:(>= ?ctime ?otime)))
   ; there must not be a closer CLOSE
   (not (Close (time ?xtime &:(< (- ?xtime ?otime)(- ?ctime ?otime)
=>
   (retract ?open)
   (retract ?close)
   (assert (Access (begin ?otime)(end ?ctime)))
)

(deffacts Opcl
   (Open  (time  845))
   (Open  (time 1045))
   (Close (time  900))
   (Close (time 1115))
   (Open  (time 1215))
   (Close (time 1215))
)

(reset)
(run)
(facts)


f-0   (MAIN::initial-fact)
f-7   (MAIN::Access (begin 845) (end 900))
f-8   (MAIN::Access (begin 1045) (end 1115))
f-9   (MAIN::Access (begin 1215) (end 1215))
For a total of 4 facts in module MAIN.

Regards
Wolfgang


On Fri, May 9, 2008 at 8:41 PM, Gary Napier <[EMAIL PROTECTED]>
wrote:

> My basic problem is this.
> I have two events, OPEN and CLOSE, both timestamped
>
> I would like to relate an OPEN and CLOSE to generate a new fact X
> However with data skew, there may be a mix of OPEN and CLOSE facts
> mixed, such that
>
> OPEN 7:15
> CLOSE 9:00
> CLOSE 8:00
> OPEN 8:15
>
> Would there be any way of sorting these facts (possibly in a multi slot
> or using an extra slot) so that they were ordered.
>
>