[rules-dev] [rules-users] What would you call a Fact that is only evaluated once?
Mark Proctor
mproctor at codehaus.org
Fri Aug 12 11:15:52 EDT 2011
On 12/08/2011 15:45, Swindells, Thomas wrote:
>
> I'm slightly confused as to the lifetime of a flake.
>
> Reading below what I think you are saying is the following:
>
> 1.Flake is inserted
>
> 2.Activations are created
>
> 3.Flake is kind-of retracted
>
> 4.Activations triggered by flake are then fired
>
> 5.Activations carry on usual.
>
> Consider the rules:
>
> Rule 1
>
> Flake()
>
> Not Foo()
>
> Then
>
> Insert(new Foo());
>
> End
>
> Rule 2
>
> Flake()
>
> Foo()
>
> Then
>
> Print("Flake for foo");
>
> end
>
Rule2 would never fire.
"flakes" as a single network propagation. It propagates through the
network as a single pass, the resulting conflict set is produced. The
fact is then retracted, although the Activations are not removed and
will continue to reference facts so they can fire. "flakes" wouldn't
support incremental matching in the engine.
When you think about this more there are a number of retraction
strategies that could be useful, and worth exploring, atleast on paper
to flesh out the ideas.
1) single pass (flake)
2) retract when there are no activations
3) retract when there are no matches (joins)
And probably others I haven't thought about.
Mark
>
> If my understanding is correct (doubtful) then only rule 1 would
> trigger -- by rule 2 the flake has already gone? Or have I got this
> totally wrong?
>
> I think my gut expectation was that they would stay until fireAllRules
> had completed.
>
> Thomas
>
> *From:*rules-dev-bounces at lists.jboss.org
> [mailto:rules-dev-bounces at lists.jboss.org] *On Behalf Of *Mark Proctor
> *Sent:* 12 August 2011 14:40
> *To:* rules-dev at lists.jboss.org
> *Subject:* Re: [rules-dev] [rules-users] What would you call a Fact
> that is only evaluated once?
>
> On 12/08/2011 12:10, Wolfgang Laun wrote:
>
> It's also a little dangerous.
>
> Consider that two or more rules should be firing with a flake. If it's
> already out of WM at the time the first one fires, you can't do a
> modify, e.g., for controlling the firing of the other rules, or simply
> for passing them some data via the flake.
>
> Flake's cannot be modified as they will not exist in the WM by the
> time by the time the rule fires. Flakes must be considered "final" and
> immutable such that the rule can still fire on them though.
>
>
> Also, a query run from the RHS of any rule with a flake would not show
> the flake, which is surprising.
>
> A flake would never show up in a query no.
>
>
> Event handling would need to be adapted - certainly the melting of a
> flake would have to be reported to a listener, probably with an
> appropriate indication.
>
> We can probably add a new enum for this special retraction.
>
>
> What happens if one inserts two or more flakes in a row?
>
> Can you collect or accumulate flakes?
>
> You can, but it would result in no change. The object is inserted and
> then retracted straight away, resulting in no change on the
> accumulate. The only difference is we leave the resulting conflict
> available to fire.
>
>
> A logical insert of a fact in a RHS of a rule firing with a flake is
> pointless.
>
> the flake would never be able to falsify the logical insertion,
> although other facts in the match can still do so. But users would
> definitely neeed to be aware of that.
>
>
> Well, could be that "djinn" would be even better: it comes, creates
> mischief and disappears again...
>
> :)
>
> What we are trying to achieve is the abilty for users to insert
> objects into WM and not have to worry about retracting them. While
> this is similar to events, it's not quite the same, as the retraction
> of events is not enforced if it's matched against.
>
> Definitely room some more thought on this one, I think it would be a
> nice use case, if we can make sure we get it right.
>
> Mark
>
>
> -W
>
>
> On 12 August 2011 12:45, Mark Proctor <mproctor at codehaus.org
> <mailto:mproctor at codehaus.org>> wrote:
>
> On 12/08/2011 11:22, Wolfgang Laun wrote:
>
> One term (probably too long) would be "interjection".
>
> A more pictorial word is "flake", which (thinking of snowflake)
> provides an inkling for the fast fade away.
>
> The context here is I'm working on an adventure game. You insert
> commands, the engine evaluates what to do with them, then you retract
> it again. Once the initial conflict set is evaluated there is no more
> use for the fact, and you end with redundant rules. I'd rather declare
> @liftetime(flake) @liftetime(durable) and have the engine handle that,
> or something along those lines/terms. Grindwork also adds something
> slightly different called "consume" for handling similar situations:
> http://www.grindwork.com/site/node/6
> "This rule fires when those conditions in the 'when' clause become
> true. When they become true, the "consume" causes the removal of the
> client message and the old alias (if one was set). The "rising"
> actions add (+ means add to the knowledge base) facts. One fact is the
> new alias, and the other is that an alias has changed. The alias
> changed fact allows others rules to notify people in the channel that
> the alias changed."
>
>
> Currently how Commands are handled:
> rule invalidMove no-loop when
> $c : MoveCommand($d : direction)
> $h : Here( $l : location)
> not ?connect( $d, $l; )
> then
> System.out.println( rule.name <http://rule.name> + ':' + $c );
> end
>
> rule validMove no-loop when
> $c : MoveCommand($d : direction)
> $h : Here( $l : location)
> exists ?connect( $d, $l; )
> then
> System.out.println( rule.name <http://rule.name> + ':' + $c );
> insert( new ExitEvent( $l ) );
> insert( new EnterEvent( $d ) );
> System.out.println( $d );
> modify( $h ) { location = $d };
> end
>
> rule retractCommand salience -100 when
> $c : Command()
> then
> retract( $c );
> end
>
>
>
>
> -W
>
> On 12 August 2011 12:00, Mark Proctor <mproctor at codehaus.org
> <mailto:mproctor at codehaus.org>> wrote:
>
> What would you call a fact that is inserted once and the conflict set
> computed (the rules that can fire). The fact is then retracted so no
> more matches can take place, but the conflict set itself is allowed to
> fire (assuming their other facts remain true).
>
> I think this is quite a common use case and most users will handle this
> via a lower salience and retracting the fact manually, but I think it's
> useful enough to build in as a keyword on type declaration. We just need
> a name for it :)
>
> Mark
>
>
> _______________________________________________
> rules-users mailing list
> rules-users at lists.jboss.org <mailto:rules-users at lists.jboss.org>
> https://lists.jboss.org/mailman/listinfo/rules-users
>
>
>
>
> _______________________________________________
> rules-dev mailing list
> rules-dev at lists.jboss.org <mailto:rules-dev at lists.jboss.org>
> https://lists.jboss.org/mailman/listinfo/rules-dev
>
>
> _______________________________________________
> rules-dev mailing list
> rules-dev at lists.jboss.org <mailto:rules-dev at lists.jboss.org>
> https://lists.jboss.org/mailman/listinfo/rules-dev
>
>
>
>
>
> _______________________________________________
> rules-dev mailing list
> rules-dev at lists.jboss.org <mailto:rules-dev at lists.jboss.org>
> https://lists.jboss.org/mailman/listinfo/rules-dev
>
>
> ------------------------------------------------------------------------
>
> **************************************************************************************
> This message is confidential and intended only for the addressee. If
> you have received this message in error, please immediately notify the
> postmaster at nds.com and delete it from your system as well as any
> copies. The content of e-mails as well as traffic data may be
> monitored by NDS for employment and security purposes. To protect the
> environment please do not print this e-mail unless necessary.
>
> NDS Limited. Registered Office: One London Road, Staines, Middlesex,
> TW18 4EX, United Kingdom. A company registered in England and Wales.
> Registered no. 3080780. VAT no. GB 603 8808 40-00
> **************************************************************************************
>
>
> _______________________________________________
> rules-dev mailing list
> rules-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/rules-dev/attachments/20110812/5f1947f5/attachment-0001.html
More information about the rules-dev
mailing list