9:36 p.m.
Hello,
searched the forum already for this question and didn't get clear answer.
I would like to insert collection of facts into session(instead of looping
over each fact inserting) and assert the rules on each fact as usual.please
share your knowledge.
--
View this message in context:
http://www.nabble.com/inserting-collection-of-facts-tp20374245p20374245.html
Sent from the drools - user mailing list archive at Nabble.com.
10:17 a.m.
--- On Fri, 11/7/08, techy <techluver007(a)gmail.com> wrote:
1. I can't have multiple session from same rulebase and
assign each session to single thread. right?
No reason why you couldn't. The Session is what holds the state. Now there may be
differences between StatefulSession and StatelessSession. StatelessSession you can
definately do this. The originating RuleBase maintains no reference to the
StatelessSessions spawned from it. A RuleBase does maintain references to the
StatefulSessions it spawns, so you should be aware of that. I'd say if you create the
StatefulSessions in a single thread, then execute them in their own child threads, you
should be fine.
2. Can I use drools 5 in production environment? if not,
when is first prod release expected?
That's a question for the dev team. :)
3. where can I do look for rule base partitioning
examples/info?
I looked last night through the docs in trunk and couldn't find any, but all I did was
"grep -i partitioning". Probably the best place to look is in the unit tests.
There's only one method you need to call to set it up,
setMultiThreadedEvaluation(boolean). And see this unit test in trunk for an example:
drools-compiler\src\test\java\org\drools\integrationtests\RulebasePartitioningTest.java
Greg Barton wrote:
>
> I'm assuming you'll be running each rule base
in a separate thread
> simultaneously? There are no dependencies between the
rules in the five
> rule bases? If so, that will work just fine.
>
> If your rules are simple enough you can use sequential
mode. It can speed
> up execution quite a bit.
>
>
http://downloads.jboss.com/drools/docs/4.0.7.19894.GA/html_single/index.h...
>
> This could be used in conjunction with the
simultaneous execution above.
>
> Finally, you can get some multithreaded benefit in a
single rule session
> if you're using drools 5 and turn on rule base
partitioning. As I
> understand it this works best when the rules have less
join conditions in
> them. (i.e. mostly single object conditions, and less
multi-object
> conditions.)
>
> --- On Thu, 11/6/08, techy
<techluver007(a)gmail.com> wrote:
>
>> From: techy <techluver007(a)gmail.com>
>> Subject: [rules-users] Rule processing for High
volume input
>> To: rules-users(a)lists.jboss.org
>> Date: Thursday, November 6, 2008, 3:41 PM
>> I have to run rules(1000 rules) against 20-30
million of
>> record each day.
>> I guess I would have memory issue If I insert all
of
>> records into working
>> memory.
>>
>> so I'm thinking to have multiple rule base
instance and
>> do batch processing
>> as given below.so that I can avoid memory issue
and quick
>> rule processing
>>
>>
>> a.Take first 1 million records and distribute 1
millon
>> across working
>> memory of 5 rule base.
>> b.fire rules in all 5 rule base
>> c.clear working memory of all rule base
>> d. take next 1 million record and start from
step-a till I
>> process all of
>> the records.
>>
>>
>> Please advise whether this is possible/recommended
way or
>> not.
>>
>> thanks
5:46 p.m.
Thanks a lot Greg.
Can you please advise which of two is best approach?
1. Having multiple rulebase and assign to child thread. each child thread
would a single stateful session and fire rule for given facts.
2. having one rulebase and create multiple stateful session and assign to
each child thread.
3. I would give batch of facts to child threads. once they complete the
firing rules. now I want to release these facts from memory before I process
next batch of facts using same or new child threads. Do I need to do any
explicit clean up in either rulebase or stateful session ? or would JVM GC
take care of the clean up as long as I don't maintain any ref to those facts
in my application?
Greg Barton wrote:
--- On Fri, 11/7/08, techy <techluver007(a)gmail.com> wrote:
> 1. I can't have multiple session from same rulebase and
> assign each session to single thread. right?
No reason why you couldn't. The Session is what holds the state. Now
there may be differences between StatefulSession and StatelessSession.
StatelessSession you can definately do this. The originating RuleBase
maintains no reference to the StatelessSessions spawned from it. A
RuleBase does maintain references to the StatefulSessions it spawns, so
you should be aware of that. I'd say if you create the StatefulSessions
in a single thread, then execute them in their own child threads, you
should be fine.
> 2. Can I use drools 5 in production environment? if not,
> when is first prod release expected?
That's a question for the dev team. :)
> 3. where can I do look for rule base partitioning
> examples/info?
I looked last night through the docs in trunk and couldn't find any, but
all I did was "grep -i partitioning". Probably the best place to look is
in the unit tests. There's only one method you need to call to set it up,
setMultiThreadedEvaluation(boolean). And see this unit test in trunk for
an example:
drools-compiler\src\test\java\org\drools\integrationtests\RulebasePartitioningTest.java
>
>
> Greg Barton wrote:
> >
> > I'm assuming you'll be running each rule base
> in a separate thread
> > simultaneously? There are no dependencies between the
> rules in the five
> > rule bases? If so, that will work just fine.
> >
> > If your rules are simple enough you can use sequential
> mode. It can speed
> > up execution quite a bit.
> >
> >
> http://downloads.jboss.com/drools/docs/4.0.7.19894.GA/html_single/index.h...
> >
> > This could be used in conjunction with the
> simultaneous execution above.
> >
> > Finally, you can get some multithreaded benefit in a
> single rule session
> > if you're using drools 5 and turn on rule base
> partitioning. As I
> > understand it this works best when the rules have less
> join conditions in
> > them. (i.e. mostly single object conditions, and less
> multi-object
> > conditions.)
> >
> > --- On Thu, 11/6/08, techy
> <techluver007(a)gmail.com> wrote:
> >
> >> From: techy <techluver007(a)gmail.com>
> >> Subject: [rules-users] Rule processing for High
> volume input
> >> To: rules-users(a)lists.jboss.org
> >> Date: Thursday, November 6, 2008, 3:41 PM
> >> I have to run rules(1000 rules) against 20-30
> million of
> >> record each day.
> >> I guess I would have memory issue If I insert all
> of
> >> records into working
> >> memory.
> >>
> >> so I'm thinking to have multiple rule base
> instance and
> >> do batch processing
> >> as given below.so that I can avoid memory issue
> and quick
> >> rule processing
> >>
> >>
> >> a.Take first 1 million records and distribute 1
> millon
> >> across working
> >> memory of 5 rule base.
> >> b.fire rules in all 5 rule base
> >> c.clear working memory of all rule base
> >> d. take next 1 million record and start from
> step-a till I
> >> process all of
> >> the records.
> >>
> >>
> >> Please advise whether this is possible/recommended
> way or
> >> not.
> >>
> >> thanks
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
--
View this message in context:
http://www.nabble.com/inserting-collection-of-facts-tp20374245p20389930.html
Sent from the drools - user mailing list archive at Nabble.com.
7:08 p.m.
--- On Fri, 11/7/08, techy <techluver007(a)gmail.com> wrote:
Can you please advise which of two is best approach?
1. Having multiple rulebase and assign to child thread.
each child thread would a single stateful session and fire rule for
given facts.
2. having one rulebase and create multiple stateful session
and assign to each child thread.
That would depend on the structure of the rules, and specifically on the number of object
join rules in your rule base. The only way to know, truthfully, is to try both ways with
typical data sets.
3. I would give batch of facts to child threads. once they
complete the firing rules. now I want to release these facts from
memory before I process next batch of facts using same or new child
threads. Do I need to do any explicit clean up in either rulebase or
stateful session ? or would JVM GC take care of the clean up as long
as I don't maintain any ref to those facts in my application?
You must retract them from the working memory for the objects to be GC'ed. Otherwise
the stateful session would maintain a reference to them. You could also just destroy the
stateful session itself. That might actually be quicker. (Assuming that you're
starting the stateful session from starting conditions during each batch run.) Although
at that point why have a stateful session? :) It sounds to me like a stateless session
might be more what you need, really. Have you tried it?
5:48 p.m.
I was under impression that in stateless session, If we modify the facts, it
would not re-evaluate rule(got confused with Sequential mode :-)). But I
tried stateless session as per your advice. It did work ! Thanks a lot Greg!
In Sequential Mode, Now It does not re-evaluate any rule if any facts
modified. I'm thinking of this new feature. If any facts modified, then
re-evaluate only subsequent rules( i.e all rules with less salience than
that of current rule's) and provide some flag to enable/disable this
feature.
this feature would be very useful If I we want to control execution of the
rules with less salience by rules with higher salience in Sequential Mode.
Actually I want to control rules execution this way. Is there any hack to
achieve this now?
Thanks again!
Greg Barton wrote:
--- On Fri, 11/7/08, techy <techluver007(a)gmail.com> wrote:
> Can you please advise which of two is best approach?
>
> 1. Having multiple rulebase and assign to child thread.
> each child thread would a single stateful session and fire rule for
> given facts.
> 2. having one rulebase and create multiple stateful session
> and assign to each child thread.
That would depend on the structure of the rules, and specifically on the
number of object join rules in your rule base. The only way to know,
truthfully, is to try both ways with typical data sets.
> 3. I would give batch of facts to child threads. once they
> complete the firing rules. now I want to release these facts from
> memory before I process next batch of facts using same or new child
> threads. Do I need to do any explicit clean up in either rulebase or
> stateful session ? or would JVM GC take care of the clean up as long
> as I don't maintain any ref to those facts in my application?
You must retract them from the working memory for the objects to be
GC'ed. Otherwise the stateful session would maintain a reference to them.
You could also just destroy the stateful session itself. That might
actually be quicker. (Assuming that you're starting the stateful session
from starting conditions during each batch run.) Although at that point
why have a stateful session? :) It sounds to me like a stateless session
might be more what you need, really. Have you tried it?
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
--
View this message in context:
http://www.nabble.com/inserting-collection-of-facts-tp20374245p20401807.html
Sent from the drools - user mailing list archive at Nabble.com.
7:04 p.m.
That would break the contract of sequential mode, as I understand it. Basically
sequential mode assumes you're working with the same set of data, the same set of
rules, and they apply in exactly the same way from start to finish. Anything that alters
execution based on prior execution is out. :) This allows sequential mode to be very
fast, but at the cost of functionality. Ain't there always a cost?
That being said, what you're talking about sounds close to activation groups. I'm
not sure if activation groups works in sequential mode, but it's worth a try.
--- On Sat, 11/8/08, techy <techluver007(a)gmail.com> wrote:
From: techy <techluver007(a)gmail.com>
Subject: Re: [rules-users] inserting collection of facts
To: rules-users(a)lists.jboss.org
Date: Saturday, November 8, 2008, 5:48 PM
I was under impression that in stateless session, If we
modify the facts, it
would not re-evaluate rule(got confused with Sequential
mode :-)). But I
tried stateless session as per your advice. It did work !
Thanks a lot Greg!
In Sequential Mode, Now It does not re-evaluate any rule if
any facts
modified. I'm thinking of this new feature. If any
facts modified, then
re-evaluate only subsequent rules( i.e all rules with less
salience than
that of current rule's) and provide some flag to
enable/disable this
feature.
this feature would be very useful If I we want to control
execution of the
rules with less salience by rules with higher salience in
Sequential Mode.
Actually I want to control rules execution this way. Is
there any hack to
achieve this now?
Thanks again!
Greg Barton wrote:
>
>
> --- On Fri, 11/7/08, techy
<techluver007(a)gmail.com> wrote:
>
>> Can you please advise which of two is best
approach?
>>
>> 1. Having multiple rulebase and assign to child
thread.
>> each child thread would a single stateful session
and fire rule for
>> given facts.
>> 2. having one rulebase and create multiple
stateful session
>> and assign to each child thread.
>
> That would depend on the structure of the rules, and
specifically on the
> number of object join rules in your rule base. The
only way to know,
> truthfully, is to try both ways with typical data
sets.
>
>> 3. I would give batch of facts to child threads.
once they
>> complete the firing rules. now I want to release
these facts from
>> memory before I process next batch of facts using
same or new child
>> threads. Do I need to do any explicit clean up in
either rulebase or
>> stateful session ? or would JVM GC take care of
the clean up as long
>> as I don't maintain any ref to those facts in
my application?
>
> You must retract them from the working memory for the
objects to be
> GC'ed. Otherwise the stateful session would
maintain a reference to them.
> You could also just destroy the stateful session
itself. That might
> actually be quicker. (Assuming that you're
starting the stateful session
> from starting conditions during each batch run.)
Although at that point
> why have a stateful session? :) It sounds to me like
a stateless session
> might be more what you need, really. Have you tried
it?
>
>
>
>
> _______________________________________________
> rules-users mailing list
> rules-users(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-users
>
>
--
View this message in context:
http://www.nabble.com/inserting-collection-of-facts-tp20374245p20401807.html
Sent from the drools - user mailing list archive at
Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
10:26 p.m.
New subject: Comparing objects with decision tables
Hi,
I am attempting to use decision tables to compare the attribute values in two objects. I
have been successful using drl files, but with an excel, I get a parse error, the excel
looks something like this -
RuleSet
krishna.drools.decisiontable
Import
krishna.drools.*
RuleTable KrishnasTest
CONDITION
ACTION
MyCompare($cn:name)
MyOriginal
name==$cn
System.out.println("$param");
CheckTheName
SetValueOfPercent
10
CheckTheName and SetValueOfPercent are column headers.
I am not sure how to define the MyCompare.name as the value being compared. Would
appreciate any pointers.
Thank you,
Krishna
_________________________________________________________________
Get 5 GB of storage with Windows Live Hotmail.
http://windowslive.com/Explore/Hotmail?ocid=TXT_TAGLM_WL_hotmail_acq_5gb_...
8:46 a.m.
New subject: Comparing objects with decision tables
(I am resending this message because the original one fetched no answers :-) )I am
attempting to use decision tables to compare the attribute values in two objects. I have
been successful using drl files, but with an excel, I get a parse error, the excel looks
something like this -
RuleSet
krishna.drools.decisiontable
Import
krishna.drools.*
RuleTable KrishnasTest
CONDITION
ACTION
MyCompare($cn:name)
MyOriginal
name==$cn
System.out.println("$param");
CheckTheName
SetValueOfPercent
10
CheckTheName and SetValueOfPercent are column headers.I am not sure how to define the
MyCompare.name as the value being compared. Am I making the correct use of decision tables
in this scenario?
Would appreciate any pointers.Thank you,Krishna
Get 5 GB of storage with Windows Live Hotmail. Sign up today.
_________________________________________________________________
Get 5 GB of storage with Windows Live Hotmail.
http://windowslive.com/Explore/Hotmail?ocid=TXT_TAGLM_WL_hotmail_acq_5gb_...
5889
days inactive
5894
days old
7 comments
3 participants
participants (3)
-
Bhamidi Krishna -
Greg Barton -
techy