7:58 a.m.
So I have setup my drools execution as follows:
1.A single thread is pooling Orders from the db and insert them into X
number of sessions(10) running on separate threads(fireUntilHalt()).
2. I have a set up rules which need to execute first to allow other rules
to fire so I have rules such as:
rule "main 1"
salience 1000
no-loop true
when
<some_conditions>
then
<modify order info>(update is invoked)
end
3. There is another set of rules which then check the Order for these
preconditions along with others
rule "sub 1"
salience 100
no-loop true
when
<some_conditions>
<main 1 consequences present>
then
<modify order info>(no update invoked as other rules in the
session don't care)
end
4. I then have a single rule which has a low salience and should be last
rule to fire in the activation which retracts the Order and hands it to a
service for post processsing
rule "END"
salience -100
no-loop true
when
ord:Order
then
retract(ord);
invokeService(ord);
end
The problem I am having is that my END rule is being invoked on certain
Orders in working memory before the other activation have taken any affect(I
have verified this by send a replicated order 1000 times and each time a
large majority his the service before anything was modified on the order).
Am I missing something?
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
8:40 a.m.
All sessions are created from the same Knowledge Base?
What does "replicate" mean, exactly?
Do you see the same effect when all facts are inserted into one session?
What happens if you do call update in rule "sub 1"?
-W
On 9 March 2012 14:58, gboro54 <gboro54(a)gmail.com> wrote:
So I have setup my drools execution as follows:
1.A single thread is pooling Orders from the db and insert them into X
number of sessions(10) running on separate threads(fireUntilHalt()).
2. I have a set up rules which need to execute first to allow other rules
to fire so I have rules such as:
rule "main 1"
salience 1000
no-loop true
when
<some_conditions>
then
<modify order info>(update is invoked)
end
3. There is another set of rules which then check the Order for these
preconditions along with others
rule "sub 1"
salience 100
no-loop true
when
<some_conditions>
<main 1 consequences present>
then
<modify order info>(no update invoked as other rules in the
session don't care)
end
4. I then have a single rule which has a low salience and should be last
rule to fire in the activation which retracts the Order and hands it to a
service for post processsing
rule "END"
salience -100
no-loop true
when
ord:Order
then
retract(ord);
invokeService(ord);
end
The problem I am having is that my END rule is being invoked on certain
Orders in working memory before the other activation have taken any
affect(I
have verified this by send a replicated order 1000 times and each time a
large majority his the service before anything was modified on the order).
Am I missing something?
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
8:53 a.m.
All sessions are against the same kBase. When I say replicate I mean I create
10000 separate Orders all the same way(same values for all properties). I
have not tried to insert all into one session and fire but can try if this
really makes a difference(though I am not sure why it should). I will add an
update to 1 as well but again I am not sure why this would matter to a low
ranking rule who does not care about the actual values from that rule
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
9:02 a.m.
Is salience ignore in a dslr? I changed the tech file to a regular DRL and
everything seems to work. Additionally I configured the kBase for streaming
mode. Does this have an affect on this as well?
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
9:08 a.m.
I take that back...Same issue still occuring...
gboro54 wrote
Is salience ignore in a dslr? I changed the tech file to a regular DRL and
everything seems to work. Additionally I configured the kBase for
streaming mode. Does this have an affect on this as well?
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
9:40 a.m.
It's difficult to imagine that doing the same thing with equal facts in
identically configured sessions would create differing objects. And it
was the intent of my questions to unearth some irregularity, and to
narrow it down.
STREAM mode shouldn't matter as you aren't using windows, expiry
or other CEP features - or are you?
-W
On 9 March 2012 15:53, gboro54 <gboro54(a)gmail.com> wrote:
All sessions are against the same kBase. When I say replicate I mean
I
create
10000 separate Orders all the same way(same values for all properties). I
have not tried to insert all into one session and fire but can try if this
really makes a difference(though I am not sure why it should). I will add
an
update to 1 as well but again I am not sure why this would matter to a low
ranking rule who does not care about the actual values from that rule
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
9:44 a.m.
No I am not using CEP and I didn't think STREAM mode made a difference. I am
just as perplexed as you on why this is happening. Could it be salience is
only respected in the same package(these are in different packages )? I am
just taking a stab in the dark at this point. I generated audit logs and it
can be seen that the removal rule activation is created first and executed
first(it looks like 75% of the time based on my test this occurs)
laune wrote
It's difficult to imagine that doing the same thing with equal facts in
identically configured sessions would create differing objects. And it
was the intent of my questions to unearth some irregularity, and to
narrow it down.
STREAM mode shouldn't matter as you aren't using windows, expiry
or other CEP features - or are you?
-W
On 9 March 2012 15:53, gboro54 <gboro54@> wrote:
> All sessions are against the same kBase. When I say replicate I mean I
> create
> 10000 separate Orders all the same way(same values for all properties). I
> have not tried to insert all into one session and fire but can try if
> this
> really makes a difference(though I am not sure why it should). I will add
> an
> update to 1 as well but again I am not sure why this would matter to a
> low
> ranking rule who does not care about the actual values from that rule
>
> --
> View this message in context:
> http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
> Sent from the Drools: User forum mailing list archive at Nabble.com.
> _______________________________________________
> rules-users mailing list
> rules-users@.jboss
> https://lists.jboss.org/mailman/listinfo/rules-users
>
_______________________________________________
rules-users mailing list
rules-users@.jboss
https://lists.jboss.org/mailman/listinfo/rules-users
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
10:07 a.m.
So I changed the package name to be the same and the situation did not occur
at all...Is salience not honored across packages?
gboro54 wrote
No I am not using CEP and I didn't think STREAM mode made a difference. I
am just as perplexed as you on why this is happening. Could it be salience
is only respected in the same package(these are in different packages )? I
am just taking a stab in the dark at this point. I generated audit logs
and it can be seen that the removal rule activation is created first and
executed first(it looks like 75% of the time based on my test this occurs)
laune wrote
>
> It's difficult to imagine that doing the same thing with equal facts in
> identically configured sessions would create differing objects. And it
> was the intent of my questions to unearth some irregularity, and to
> narrow it down.
>
> STREAM mode shouldn't matter as you aren't using windows, expiry
> or other CEP features - or are you?
>
> -W
>
>
> On 9 March 2012 15:53, gboro54 <gboro54@> wrote:
>
>> All sessions are against the same kBase. When I say replicate I mean I
>> create
>> 10000 separate Orders all the same way(same values for all properties).
>> I
>> have not tried to insert all into one session and fire but can try if
>> this
>> really makes a difference(though I am not sure why it should). I will
>> add
>> an
>> update to 1 as well but again I am not sure why this would matter to a
>> low
>> ranking rule who does not care about the actual values from that rule
>>
>> --
>> View this message in context:
>>
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
>> Sent from the Drools: User forum mailing list archive at Nabble.com.
>> _______________________________________________
>> rules-users mailing list
>> rules-users@.jboss
>> https://lists.jboss.org/mailman/listinfo/rules-users
>>
>
> _______________________________________________
> rules-users mailing list
> rules-users@.jboss
> https://lists.jboss.org/mailman/listinfo/rules-users
>
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
10:19 a.m.
Salience is a field in the Activation object, used for ranking another
entry when it is inserted into the List. Looking at the package name for
this or at this time would be futile.
-W
On 9 March 2012 16:44, gboro54 <gboro54(a)gmail.com> wrote:
No I am not using CEP and I didn't think STREAM mode made a
difference. I
am
just as perplexed as you on why this is happening. Could it be salience is
only respected in the same package(these are in different packages )? I am
just taking a stab in the dark at this point. I generated audit logs and it
can be seen that the removal rule activation is created first and executed
first(it looks like 75% of the time based on my test this occurs)
laune wrote
>
> It's difficult to imagine that doing the same thing with equal facts in
> identically configured sessions would create differing objects. And it
> was the intent of my questions to unearth some irregularity, and to
> narrow it down.
>
> STREAM mode shouldn't matter as you aren't using windows, expiry
> or other CEP features - or are you?
>
> -W
>
>
> On 9 March 2012 15:53, gboro54 <gboro54@> wrote:
>
>> All sessions are against the same kBase. When I say replicate I mean I
>> create
>> 10000 separate Orders all the same way(same values for all properties).
I
>> have not tried to insert all into one session and fire but can try if
>> this
>> really makes a difference(though I am not sure why it should). I will
add
>> an
>> update to 1 as well but again I am not sure why this would matter to a
>> low
>> ranking rule who does not care about the actual values from that rule
>>
>> --
>> View this message in context:
>>
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
>> Sent from the Drools: User forum mailing list archive at Nabble.com.
>> _______________________________________________
>> rules-users mailing list
>> rules-users@.jboss
>> https://lists.jboss.org/mailman/listinfo/rules-users
>>
>
> _______________________________________________
> rules-users mailing list
> rules-users@.jboss
> https://lists.jboss.org/mailman/listinfo/rules-users
>
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
10:25 a.m.
Perhaps I am building my packages weird but if all my drl/dslr are under the
same package the salience is honored and all the rules are fired in order.
Why would it work in this way? Since changing all the files to the same
package I have done 10 runs of the test with 100% success. Am I building the
kBase in weird manner(I don't think so but look below):
if (logger.isDebugEnabled()) {
logger.debug("Knowledge base with the following changeset was requested
built "
+ changeset);
}
KnowledgeBuilder kbuilder = KnowledgeBuilderFactory
.newKnowledgeBuilder();
kbuilder.add(ResourceFactory.newClassPathResource(changeset),
ResourceType.CHANGE_SET);
KnowledgeBuilderErrors errors = kbuilder.getErrors();
if (errors.size() > 0) {
StringBuilder sb = new StringBuilder();
for (KnowledgeBuilderError error : errors) {
sb.append(error + "\n");
logger.error(error);
}
throw new IllegalArgumentException("Could not parse knowledge. \n"
+ sb);
}
KnowledgeBase kbase = KnowledgeBaseFactory.newKnowledgeBase();
kbase.addKnowledgePackages(kbuilder.getKnowledgePackages());
return kbase;
Would this cause any fragmenting when the engine is indexing the rules for
activation? I would think not. I am just going by the current results I have
gotten by declaring all rules in the same package.
laune wrote
Salience is a field in the Activation object, used for ranking another
entry when it is inserted into the List. Looking at the package name for
this or at this time would be futile.
-W
On 9 March 2012 16:44, gboro54 <gboro54@> wrote:
> No I am not using CEP and I didn't think STREAM mode made a difference. I
> am
> just as perplexed as you on why this is happening. Could it be salience
> is
> only respected in the same package(these are in different packages )? I
> am
> just taking a stab in the dark at this point. I generated audit logs and
> it
> can be seen that the removal rule activation is created first and
> executed
> first(it looks like 75% of the time based on my test this occurs)
>
>
> laune wrote
> >
> > It's difficult to imagine that doing the same thing with equal facts in
> > identically configured sessions would create differing objects. And it
> > was the intent of my questions to unearth some irregularity, and to
> > narrow it down.
> >
> > STREAM mode shouldn't matter as you aren't using windows, expiry
> > or other CEP features - or are you?
> >
> > -W
> >
> >
> > On 9 March 2012 15:53, gboro54 <gboro54@> wrote:
> >
> >> All sessions are against the same kBase. When I say replicate I mean I
> >> create
> >> 10000 separate Orders all the same way(same values for all
> properties).
> I
> >> have not tried to insert all into one session and fire but can try if
> >> this
> >> really makes a difference(though I am not sure why it should). I will
> add
> >> an
> >> update to 1 as well but again I am not sure why this would matter to a
> >> low
> >> ranking rule who does not care about the actual values from that rule
> >>
> >> --
> >> View this message in context:
> >>
> http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
> >> Sent from the Drools: User forum mailing list archive at Nabble.com.
> >> _______________________________________________
> >> rules-users mailing list
> >> rules-users@.jboss
> >> https://lists.jboss.org/mailman/listinfo/rules-users
> >>
> >
> > _______________________________________________
> > rules-users mailing list
> > rules-users@.jboss
> > https://lists.jboss.org/mailman/listinfo/rules-users
> >
>
>
> --
> View this message in context:
> http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
> Sent from the Drools: User forum mailing list archive at Nabble.com.
> _______________________________________________
> rules-users mailing list
> rules-users@.jboss
> https://lists.jboss.org/mailman/listinfo/rules-users
>
_______________________________________________
rules-users mailing list
rules-users@.jboss
https://lists.jboss.org/mailman/listinfo/rules-users
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
12:14 p.m.
Could be there is a race condition in the builder.
I would not use a changeset unless you really need dynamic changes of your
rules.
Simply feed the builder with DRL or whatever, create the Knowledge Base and
then
the sessions.
-W
On 9 March 2012 17:25, gboro54 <gboro54(a)gmail.com> wrote:
Perhaps I am building my packages weird but if all my drl/dslr are
under
the
same package the salience is honored and all the rules are fired in order.
Why would it work in this way? Since changing all the files to the same
package I have done 10 runs of the test with 100% success. Am I building
the
kBase in weird manner(I don't think so but look below):
if (logger.isDebugEnabled()) {
logger.debug("Knowledge base with the following
changeset was requested
built "
+ changeset);
}
KnowledgeBuilder kbuilder = KnowledgeBuilderFactory
.newKnowledgeBuilder();
kbuilder.add(ResourceFactory.newClassPathResource(changeset),
ResourceType.CHANGE_SET);
KnowledgeBuilderErrors errors = kbuilder.getErrors();
if (errors.size() > 0) {
StringBuilder sb = new StringBuilder();
for (KnowledgeBuilderError error : errors) {
sb.append(error + "\n");
logger.error(error);
}
throw new IllegalArgumentException("Could not parse
knowledge. \n"
+ sb);
}
KnowledgeBase kbase =
KnowledgeBaseFactory.newKnowledgeBase();
kbase.addKnowledgePackages(kbuilder.getKnowledgePackages());
return kbase;
Would this cause any fragmenting when the engine is indexing the rules for
activation? I would think not. I am just going by the current results I
have
gotten by declaring all rules in the same package.
laune wrote
>
> Salience is a field in the Activation object, used for ranking another
> entry when it is inserted into the List. Looking at the package name for
> this or at this time would be futile.
> -W
>
> On 9 March 2012 16:44, gboro54 <gboro54@> wrote:
>
>> No I am not using CEP and I didn't think STREAM mode made a difference.
I
>> am
>> just as perplexed as you on why this is happening. Could it be salience
>> is
>> only respected in the same package(these are in different packages )? I
>> am
>> just taking a stab in the dark at this point. I generated audit logs and
>> it
>> can be seen that the removal rule activation is created first and
>> executed
>> first(it looks like 75% of the time based on my test this occurs)
>>
>>
>> laune wrote
>> >
>> > It's difficult to imagine that doing the same thing with equal facts
in
>> > identically configured sessions would create differing objects. And it
>> > was the intent of my questions to unearth some irregularity, and to
>> > narrow it down.
>> >
>> > STREAM mode shouldn't matter as you aren't using windows, expiry
>> > or other CEP features - or are you?
>> >
>> > -W
>> >
>> >
>> > On 9 March 2012 15:53, gboro54 <gboro54@> wrote:
>> >
>> >> All sessions are against the same kBase. When I say replicate I mean
I
>> >> create
>> >> 10000 separate Orders all the same way(same values for all
>> properties).
>> I
>> >> have not tried to insert all into one session and fire but can try if
>> >> this
>> >> really makes a difference(though I am not sure why it should). I will
>> add
>> >> an
>> >> update to 1 as well but again I am not sure why this would matter to
a
>> >> low
>> >> ranking rule who does not care about the actual values from that rule
>> >>
>> >> --
>> >> View this message in context:
>> >>
>>
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
>> >> Sent from the Drools: User forum mailing list archive at Nabble.com.
>> >> _______________________________________________
>> >> rules-users mailing list
>> >> rules-users@.jboss
>> >> https://lists.jboss.org/mailman/listinfo/rules-users
>> >>
>> >
>> > _______________________________________________
>> > rules-users mailing list
>> > rules-users@.jboss
>> > https://lists.jboss.org/mailman/listinfo/rules-users
>> >
>>
>>
>> --
>> View this message in context:
>>
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
>> Sent from the Drools: User forum mailing list archive at Nabble.com.
>> _______________________________________________
>> rules-users mailing list
>> rules-users@.jboss
>> https://lists.jboss.org/mailman/listinfo/rules-users
>>
>
> _______________________________________________
> rules-users mailing list
> rules-users@.jboss
> https://lists.jboss.org/mailman/listinfo/rules-users
>
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
12:18 p.m.
Each builder is building off multiple files. I guess I could provide a list
of these resources. Is there any worry about the order in which the packages
are built then?
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
12:28 p.m.
No.
(DSLs must be loaded before any DSLRs using them.)
-W
On 9 March 2012 19:18, gboro54 <gboro54(a)gmail.com> wrote:
Each builder is building off multiple files. I guess I could provide
a list
of these resources. Is there any worry about the order in which the
packages
are built then?
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
12:31 p.m.
That I figured .... Just out of curiosity why would a changeset cause a race
condition? I understand why a changeset may not be ideal for here but for
future implementations(I have used changsets mainly with Guvnor
built/manages packages).
laune wrote
No.
(DSLs must be loaded before any DSLRs using them.)
-W
On 9 March 2012 19:18, gboro54 <gboro54@> wrote:
> Each builder is building off multiple files. I guess I could provide a
> list
> of these resources. Is there any worry about the order in which the
> packages
> are built then?
>
> --
> View this message in context:
> http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
> Sent from the Drools: User forum mailing list archive at Nabble.com.
> _______________________________________________
> rules-users mailing list
> rules-users@.jboss
> https://lists.jboss.org/mailman/listinfo/rules-users
>
_______________________________________________
rules-users mailing list
rules-users@.jboss
https://lists.jboss.org/mailman/listinfo/rules-users
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
12:55 p.m.
The resource scanner runs in another thread. So when you start this
service, but don't monitor or sync with its events,
grab stuff - well, I would be very cagey there.
-W
On 9 March 2012 19:31, gboro54 <gboro54(a)gmail.com> wrote:
That I figured .... Just out of curiosity why would a changeset cause
a
race
condition? I understand why a changeset may not be ideal for here but for
future implementations(I have used changsets mainly with Guvnor
built/manages packages).
laune wrote
>
> No.
> (DSLs must be loaded before any DSLRs using them.)
> -W
>
> On 9 March 2012 19:18, gboro54 <gboro54@> wrote:
>
>> Each builder is building off multiple files. I guess I could provide a
>> list
>> of these resources. Is there any worry about the order in which the
>> packages
>> are built then?
>>
>> --
>> View this message in context:
>>
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
>> Sent from the Drools: User forum mailing list archive at Nabble.com.
>> _______________________________________________
>> rules-users mailing list
>> rules-users@.jboss
>> https://lists.jboss.org/mailman/listinfo/rules-users
>>
>
> _______________________________________________
> rules-users mailing list
> rules-users@.jboss
> https://lists.jboss.org/mailman/listinfo/rules-users
>
--
View this message in context:
http://drools.46999.n3.nabble.com/Drools-Salience-and-no-loop-tp3812723p3...
Sent from the Drools: User forum mailing list archive at Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
5049
days inactive
5049
days old
14 comments
2 participants
participants (2)
-
gboro54 -
Wolfgang Laun