[rules-users] Grouping rules
Jai Vasanth
jaivasanth at gmail.com
Wed Feb 6 19:01:13 EST 2008
Thanks.
Jai
On Feb 6, 2008 5:29 AM, Edson Tirelli <tirelli at post.com> wrote:
>
> The answer is not simple. :)
> All the alpha constraints are evaluated at fact assertion time, does
> not matter if they are early or late in the pattern order. Although, alpha
> constraints are usually quite cheap, since they are constraints that do not
> depend on other facts. Example:
>
> Person( name == 'bob' )
>
> "name == 'bob'" is a literal constraint and do not depend on any other
> pattern (so it is an alpha constraint) and will be evaluated at assert time,
> does not matter the order in which patterns are written down.
> Beta constraints are constraints that depend on previous patterns and
> will only be evaluated when previous patterns are matched. Example:
>
> when
> Person( $likes : likes )
> Cheese( type == $likes )
> then
> end
>
> It is obvious in the previous example that "type == $likes" depends on
> the value of $likes that is a variable bound in a previous pattern (so it is
> a Beta constraint). It will only be evaluated if ALL previous patterns
> match.
> "from" CE is a special case of beta "node", that is only evaluated when
> all previous patterns match.
>
>
> []s
> Edson
> 2008/2/6, Anstis, Michael (M.) <manstis1 at ford.com>:
> >
> > Great news!
> >
> > Is this specific to the use of "from" whereas another rule like:-
> >
> > rule "avoid expense"
> > when
> > ControlFact( phase == "do expensive call" )
> > AnotherFact( someExpensiveTimeConstantFunctionCall == someValue
> > )
> > then
> > // Do something
> > end
> >
> > Could not have the LHS not evaluated or (to avoid a double negative),
> > would always have the LHS fully evaluated ;-)
> >
> > ------------------------------
> > *From:* rules-users-bounces at lists.jboss.org [mailto:
> > rules-users-bounces at lists.jboss.org] *On Behalf Of *Edson Tirelli
> > *Sent:* 06 February 2008 12:29
> > *To:* Rules Users List
> > *Subject:* Re: [rules-users] Grouping rules
> >
> >
> > Yes, it is possible to prevent the evaluation of part of the LHS
> > patterns by the use of previous constraints and/or the introduction of
> > control facts.
> >
> > For instance:
> >
> > rule "avoid expense"
> > when
> > ControlFact( phase == "do expensive call" )
> > AnotherFact( ) from someExpensiveServiceCall
> > then
> > // Do something
> > end
> >
> > In the above example, the "someExpensiveServiceCall" will only be
> > executed when the previous patterns are matched, i.e., when phase
> > attribute of the ControlFact has the value "do expensive call".
> >
> > []s
> > Edson
> >
> > 2008/2/6, Anstis, Michael (M.) <manstis1 at ford.com>:
> > >
> > > I'm not sure as LHS's are evaluated when facts are inserted not when
> > > fireAllRules (or execute) is called.
> > >
> > > You can prevent the RHS from activating by using (for lack of a better
> > > name) "flag facts" (but this doesn't solve your problem); for example:-
> > >
> > > rule "avoid expense"
> > > when
> > > not exists Flag()
> > > AnotherFact( ) from someExpensiveServiceCall
> > > then
> > > // Do something
> > > end
> > >
> > > Don't forget though that if you share LHS patterns across multiple
> > > rules they will share the same nodes in the RETE network, so although you
> > > might be aware that the expensive calls are not required for some rules they
> > > might be required for other rules and hence you'll need to perform the
> > > expensive call when the facts are inserted in order for the other rules to
> > > activate. Perhaps if you better describe your use-case there might be a
> > > better solution.
> > >
> > > With kind regards,
> > >
> > > Mike
> > >
> > > ------------------------------
> > > *From:* rules-users-bounces at lists.jboss.org [mailto:
> > > rules-users-bounces at lists.jboss.org] *On Behalf Of *Jai Vasanth
> > > *Sent:* 05 February 2008 19:35
> > > *To:* Rules Users List
> > > *Subject:* Re: [rules-users] Grouping rules
> > >
> > > I had another question on these lines. Is there a way to preven even
> > > the LHS from evaluating for certain cases ? A few of my LHS computation
> > > involve some expensive service calls and I would like to avoid unless and
> > > until it is asolutely necessary.
> > >
> > > Thanks
> > >
> > > Jai
> > >
> > > On Feb 1, 2008 8:35 AM, Jai Vasanth <jaivasanth at gmail.com> wrote:
> > >
> > > > Thanks. That helped.
> > > >
> > > >
> > > > On Feb 1, 2008 7:31 AM, Anstis, Michael (M.) <manstis1 at ford.com>
> > > > wrote:
> > > >
> > > > > Excuse me if you already understand but there is a difference
> > > > > between rule patterns being evaluated and their consequence running.
> > > > >
> > > > > Rules LHS are evaluated when objects are inserted into working
> > > > > memory whereas the consequence fires when you call fireAllRules();
> > > > >
> > > > > The rules whose activations are executed are those in the given
> > > > > Agenda Group; thereafter those not in any agenda group execute. So if all
> > > > > rules are in an Agenda Group you should be OK.
> > > > >
> > > > > Of course they don't prevent all rule patterns from being checked
> > > > > as objects are inserted into WM (which has a performance impact) but this is
> > > > > what the RETE network was designed to optimise.
> > > > >
> > > > > You could try another configuration (but I think Agenda Groups are
> > > > > probably the preference by design). This gives complete isolation - but I
> > > > > haven't tried it so it might not even be possible ;-)
> > > > >
> > > > > global WorkingMemory wm;
> > > > > global RuleBase rbX;
> > > > > global RuleBase rbY;
> > > > >
> > > > > rule "Group X"
> > > > > when
> > > > > Fact( attribute == "condition1" )
> > > > > then
> > > > > wm = rbX.newStatefulSession();
> > > > > wm.fireAllRules();
> > > > > end
> > > > >
> > > > > rule "Group Y"
> > > > > when
> > > > > Fact( attribute == "condition2" )
> > > > > then
> > > > > wm = rbY.newStatefulSession();
> > > > > wm.fireAllRules();
> > > > > end
> > > > >
> > > > > Good luck.
> > > > >
> > > > > Mike
> > > > >
> > > > > ------------------------------
> > > > > *From:* rules-users-bounces at lists.jboss.org [mailto:
> > > > > rules-users-bounces at lists.jboss.org] *On Behalf Of *Jai Vasanth
> > > > > *Sent:* 01 February 2008 15:07
> > > > > *To:* Rules Users List
> > > > > *Subject:* Re: [rules-users] Grouping rules
> > > > >
> > > > >
> > > > > Thanks for responding
> > > > > Yes, I did consider agenda groups, but I thought agenda groups
> > > > > only ordered execution in a particular fashion, so if Rule "pick group X"
> > > > > fires then it would make sure that rules in Agenda Group X fire first before
> > > > > the remaining but the remaining would fire eventually. Is my understanding
> > > > > correct ?
> > > > > I am looking for a solution for completely isolating a rule set
> > > > > conditioned on an object attribute.
> > > > >
> > > > > On Feb 1, 2008 12:46 AM, Anstis, Michael (M.) <manstis1 at ford.com>
> > > > > wrote:
> > > > >
> > > > > > What about Agenda Groups?
> > > > > >
> > > > > > Rule "pick group X"
> > > > > > when
> > > > > > Fact( attribute = "condition1" )
> > > > > > then
> > > > > > drools.setFocus("Group X")
> > > > > > end
> > > > > >
> > > > > > Rule "pick group Y"
> > > > > > when
> > > > > > Fact( attribute = "condition2" )
> > > > > > then
> > > > > > drools.setFocus("Group Y")
> > > > > > end
> > > > > >
> > > > > > Rule "Group X1"
> > > > > > Agenda Group "Group X"
> > > > > > when
> > > > > > Smurf( )
> > > > > > then
> > > > > > // Do something
> > > > > > end
> > > > > >
> > > > > > Rule "Group X2"
> > > > > > Agenda Group "Group X"
> > > > > > when
> > > > > > Smurfette( )
> > > > > > then
> > > > > > // Something else
> > > > > > end
> > > > > >
> > > > > > etc
> > > > > >
> > > > > > ------------------------------
> > > > > > *From:* rules-users-bounces at lists.jboss.org [mailto:
> > > > > > rules-users-bounces at lists.jboss.org] *On Behalf Of *Jai Vasanth
> > > > > > *Sent:* 31 January 2008 19:45
> > > > > > *To:* rules-users at lists.jboss.org
> > > > > > *Subject:* [rules-users] Grouping rules
> > > > > >
> > > > > > Hi,
> > > > > >
> > > > > >
> > > > > > I am building a system where I would need to fire different
> > > > > > sets of rules based on some attribute in the fact object.
> > > > > >
> > > > > > Here are someways which I thought of, I was wondering if there
> > > > > > was something better than that.
> > > > > >
> > > > > > 1) I could do this by creating different sessions based on the
> > > > > > data attrbute (different sessions would have different rule packages based
> > > > > > on the attribute)
> > > > > >
> > > > > > or
> > > > > >
> > > > > > 2) Have all the rules fire (all the different sets of rules)
> > > > > > irrespective of the attribute and have them insert fact new objects into the
> > > > > > working memory. In the second round of rules, collect the (newly made) fact
> > > > > > objects based on the attribute and take that action.
> > > > > >
> > > > > >
> > > > > >
> > > > > > Thanks
> > > > > >
> > > > > > Jai
> > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > > > _______________________________________________
> > > > > > 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
> > > > >
> > > > >
> > > >
> > >
> > > _______________________________________________
> > > rules-users mailing list
> > > rules-users at lists.jboss.org
> > > https://lists.jboss.org/mailman/listinfo/rules-users
> > >
> > >
> > >
> >
> >
> > --
> > Edson Tirelli
> > JBoss Drools Core Development
> > Office: +55 11 3529-6000
> > Mobile: +55 11 9287-5646
> > JBoss, a division of Red Hat @ www.jboss.com
> >
> >
> > _______________________________________________
> > rules-users mailing list
> > rules-users at lists.jboss.org
> > https://lists.jboss.org/mailman/listinfo/rules-users
> >
> >
> >
>
>
> --
> Edson Tirelli
> JBoss Drools Core Development
> Office: +55 11 3529-6000
> Mobile: +55 11 9287-5646
> JBoss, a division of Red Hat @ www.jboss.com
>
> _______________________________________________
> rules-users mailing list
> rules-users at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/rules-users/attachments/20080206/378a282c/attachment.html
More information about the rules-users
mailing list