[rules-users] A Drools CEP Scenario - did I get it right?

[Wolfgang Laun]

1. A fact doesn't change until it *is* changed or the knowledge base
destined to operate it changes: so why should reasoning due to
the unchanged fact be repeated, with all consequential actions.

Continuing your line of thought would result in a hopeless muddle
if all consequences are re-executed each time fireAllRules is
called even where facts and rules haven't changed.

2. This works fine to adjust the pseudo-clock to the time of the next event,
about to be inserted in a "playback" run:

    PseudoClockScheduler clock;

    private void advance( Date eventDate ){
        long currentTime = clock.getCurrentTime();
        long eventTime = eventDate.getTime();
        clock.advanceTime( eventTime - currentTime, TimeUnit.MILLISECONDS );
    }

-W


On 09/09/2013, Alexander Wolf <mail at alexander-wolf.net> wrote:
> The question is more or less where to begin.
>
> 1. If there is a Fact in working memory (e.g. "Person") and I invoke
> fireAllRules() twice, I would expect to have the following rule fire both
> times:
>
> rule "alwaysFire"
> when
> 	$room : Room( )
> then
> 	System.out.println("there is a room");
> end
>
> I would also expect this rule to fire every time I invoke fireAllRules()
>
> rule "alwaysFire2"
> when
> then
> 	System.out.println("always!");
> end
>
> Both rules fire only once (the first time) even though the Working Memory is
> the same. (The room still exists...). I expected the rule to fire each time
> fireAllRules() is invoked and for the number of Rooms e.g. with 4 rooms ->
> 4x output "there is a room".
>
> 2. It seems to be impossible to set the pseudo clock to a specific time. How
> can I test CRON timer rules?!
>
>
>
> On 08.09.2013, at 18:51, Michael Anstis <michael.anstis at gmail.com> wrote:
>
>> What did not happen as you expected?
>>
>> The HTTP session to Knowledge session association seems reasonable; but I
>> assume you are trying with a more simple configuration initially.
>>
>> Sent on the move
>>
>> On 8 Sep 2013 17:42, "amarok" <mail at alexander-wolf.net> wrote:
>> I got the following scenario:
>> - I need a Web Service that reasons over received sensor events (1000+
>> per
>> day and client) in realtime and send messages if certain patterns occur.
>> - The web service has many clients (100-1000) but their events / facts do
>> not interfere (each client has his own state and events that change his
>> state).
>> - Each client also has his own rules (10-100) (but they are parameterized
>> generic rules - I thought maybe I could use rule templates or DSL to let
>> clients individualize their rules).
>> - Some rules are time dependant (e.g. if event (type=x) does not occour
>> at
>> least within half an hour after 8:00 a.m. (where 8 o'clock is a rule
>> parameter) -> send a message), others are accumulative (e.g. if number of
>> events within 5h is > 3).
>>
>>
>> I'd like to use drools CEP for this. I already read some documentation
>> and
>> tutorials but my first examples do not really work as I expected. Now I
>> fear
>> to have some general misconception about Drools/CEP. If someone could have
>> a
>> look over my ideas, I'd be very glad!
>>
>>
>> How I would do it:
>> -> Every client would have his own KnowledgeBase with his
>> (individualized)
>> rules. Rules would be stored in guvnor or on a mysql database as plain
>> text.
>> -> Every client would have his own KnowledgeBase /
>> StatefullKnowledgeSession
>> on the service, that knows about recent sensor events and some facts
>> about
>> the client. Events automatically are disposed of once they are not
>> relevant
>> anymore.
>> -> A static hashmap will hold references to the session for each client.
>> (clientID -> KnowledgeSession)
>> -> Sessions are created when a new client joins the system, but are
>> normally
>> not disposed unless the client completely leaves the system.
>> -> When a new Event is received (HTTP PUT), the session corresponding to
>> the
>> client would be retrieved from the HashMap and the new event would be
>> inserted via an event stream. Then (for every new event)
>> session.fireAllRules() would be called on the clients KnowledgeSession.
>>
>> -> The Knowledge Base would run in STREAM mode to be able to reason over
>> time windows and automatically expire (=delete?!) events. This would keep
>> memory usage in within bounds(?)
>> -> pseudo clock allows me to write tests or simulate the system with
>> records
>> of old events. (Is it possible to test timer, cron and sliding-window
>> rules
>> with pseudo clock? Can I explicitly set the time of the pseudo clock?)
>>
>> I appreciate your help!
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> --
>> View this message in context:
>> http://drools.46999.n3.nabble.com/A-Drools-CEP-Scenario-did-I-get-it-right-tp4025863.html
>> Sent from the Drools: User forum mailing list archive at Nabble.com.
>> _______________________________________________
>> rules-users mailing list
>> rules-users at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/rules-users
>> _______________________________________________
>> rules-users mailing list
>> rules-users at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/rules-users
>
>