Unexpected behavior of accumulate() and insertLogical() when the result() goes from matching a constraint to not matching
10:32 p.m.
This is a follow up from my original post @
http://drools.46999.n3.nabble.com/rules-users-Question-about-custom-accum...
Using Drools version 5.5.0.Final
I am experiencing behavior I did not expect, regarding Drools accumulate
functionality.
The following demonstrates:
With an example (contrived) rule such as:
file <<example.drl>>
package drools
rule "testing"
dialect "mvel"
when
$res : Long() from accumulate( $l : Long(),
init( Long highest = null; ),
action( highest = ((null == highest) || ($l > highest)) ? $l :
highest; ),
result( (highest < 10) ? highest : null; ))
then
insertLogical( new String("inserted " + $res.toString()) );
end
endfile <<example.drl>>
I have the following, simple test case:
file <<ExampleDroolsTest.java>>
package drools;
import java.io.IOException;
import org.drools.KnowledgeBase;
import org.drools.builder.KnowledgeBuilder;
import org.drools.builder.KnowledgeBuilderFactory;
import org.drools.builder.ResourceType;
import org.drools.io.ResourceFactory;
import org.drools.runtime.StatefulKnowledgeSession;
import org.junit.Test;
public class ExampleDroolsTest {
@Test
public void test() throws IOException {
final KnowledgeBuilder kbuilder =
KnowledgeBuilderFactory.newKnowledgeBuilder();
kbuilder.add(ResourceFactory.newClassPathResource("drools/example.drl"),
ResourceType.DRL);
final KnowledgeBase kbase = kbuilder.newKnowledgeBase();
final StatefulKnowledgeSession ksession =
kbase.newStatefulKnowledgeSession();
/** HERE the insertion order of these two facts, will result in
different outcomes for
* the rule "testing" in example.drl .
* In this order, I will get the RHS logic to insert the new
String("inserted" + $res.toString())
* The accumulate should be getting the max valued Long, in the end
and then the `result` of the
* `accumulate` should yield a null value.
*
* If I flip the inserts, no fact is inserted into the session.
*/
ksession.insert(new Long(5));
ksession.insert(new Long(15));
final int rulesFired = ksession.fireAllRules();
System.out.println("Rules fired: " + rulesFired);
System.out.println("Session objs:");
for (final Object o : ksession.getObjects()) {
System.out.println(o);
}
ksession.dispose();
}
}
endfile <<ExampleDroolsTest.java>>
As I note in the Java source comment above, the order of the insertions of
the Long's into the session, change the results of the rule execution. My
initial understanding of accumulate, would be that the LHS would be
invalidated once the highest Long value
was found to be 15 > 10, as the `result` portion of the `accumulate` tests.
Since the String object is inserted logically, I'd
expect the fact to be retracted from the working memory once the LHS of the
rule "testing" was found to be false.
If this were the case, the order of insertions above in the test case, would
result in the same working memory after rule execution.
Is this intended behavior of the accumulate node along with the
`insertLogical` operation?
Thanks!
--
View this message in context:
http://drools.46999.n3.nabble.com/Unexpected-behavior-of-accumulate-and-i...
Sent from the Drools: User forum mailing list archive at Nabble.com.
2:46 a.m.
The reason why you experience this behavior is rooted in the difference
between rule evaluation and rule firing.
The custom accumulate is reevaluated, in your case, for each Long you
insert: since you only have the accumulate,
the result of the accumulate will be used to generate a tuple for the
RuleTerminalNode, which will result in
an activation, to be fired later when you call fireAllRules. By
(somewhat debatable) design, activations from
the same accumulate will override each other if not fired in between.
This said, we have two scenarios here:
1) 5 first, 15 second :
when you insert 5, 5 is now the highest. It is also < 10, so an
activation is generated on the agenda.
now you insert 15: it becomes the new highest value, but it is NOT < 10,
so the accumulate returns
null. Unfortunately, null does NOT match "Long() from accumulate ..." so
there will be no new activation
to override the preexisting one. That will fire and you will see
"inserted 5" in the working memory.
2) 15 first, 5 second
when you insert 15, it becomes the highest. The result is null, and for
the same reason above NO
activation will be generated. You now insert 5, but this value won't
change highest, so you will
not see any activations either.
Under the premises, the behavior is correct. You may want to replace the
result:
result( (highest < 10) ? highest : -1L; )
you should see a more coherent behavior :)
Let me know if you need more clarifications
Best,
Davide
On 10/26/2013 01:32 PM, mikerod wrote:
This is a follow up from my original post @
http://drools.46999.n3.nabble.com/rules-users-Question-about-custom-accum...
Using Drools version 5.5.0.Final
I am experiencing behavior I did not expect, regarding Drools accumulate
functionality.
The following demonstrates:
With an example (contrived) rule such as:
file <<example.drl>>
package drools
rule "testing"
dialect "mvel"
when
$res : Long() from accumulate( $l : Long(),
init( Long highest = null; ),
action( highest = ((null == highest) || ($l > highest)) ? $l :
highest; ),
result( (highest < 10) ? highest : null; ))
then
insertLogical( new String("inserted " + $res.toString()) );
end
endfile <<example.drl>>
I have the following, simple test case:
file <<ExampleDroolsTest.java>>
package drools;
import java.io.IOException;
import org.drools.KnowledgeBase;
import org.drools.builder.KnowledgeBuilder;
import org.drools.builder.KnowledgeBuilderFactory;
import org.drools.builder.ResourceType;
import org.drools.io.ResourceFactory;
import org.drools.runtime.StatefulKnowledgeSession;
import org.junit.Test;
public class ExampleDroolsTest {
@Test
public void test() throws IOException {
final KnowledgeBuilder kbuilder =
KnowledgeBuilderFactory.newKnowledgeBuilder();
kbuilder.add(ResourceFactory.newClassPathResource("drools/example.drl"),
ResourceType.DRL);
final KnowledgeBase kbase = kbuilder.newKnowledgeBase();
final StatefulKnowledgeSession ksession =
kbase.newStatefulKnowledgeSession();
/** HERE the insertion order of these two facts, will result in
different outcomes for
* the rule "testing" in example.drl .
* In this order, I will get the RHS logic to insert the new
String("inserted" + $res.toString())
* The accumulate should be getting the max valued Long, in the end
and then the `result` of the
* `accumulate` should yield a null value.
*
* If I flip the inserts, no fact is inserted into the session.
*/
ksession.insert(new Long(5));
ksession.insert(new Long(15));
final int rulesFired = ksession.fireAllRules();
System.out.println("Rules fired: " + rulesFired);
System.out.println("Session objs:");
for (final Object o : ksession.getObjects()) {
System.out.println(o);
}
ksession.dispose();
}
}
endfile <<ExampleDroolsTest.java>>
As I note in the Java source comment above, the order of the insertions of
the Long's into the session, change the results of the rule execution. My
initial understanding of accumulate, would be that the LHS would be
invalidated once the highest Long value
was found to be 15 > 10, as the `result` portion of the `accumulate` tests.
Since the String object is inserted logically, I'd
expect the fact to be retracted from the working memory once the LHS of the
rule "testing" was found to be false.
If this were the case, the order of insertions above in the test case, would
result in the same working memory after rule execution.
Is this intended behavior of the accumulate node along with the
`insertLogical` operation?
Thanks!
--
View this message in context:
http://drools.46999.n3.nabble.com/Unexpected-behavior-of-accumulate-and-i...
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
5:33 p.m.
I appreciate your input. I understand what you are saying regarding the
design of the activations caused by the accumulate.
Unfortunately, this is making it hard to express the logic I'm going for in
a concise manner.
I'll give a bit more of a "real", but similar example here.
With my example rules:
file <<example.drl>>
package drools
rule "testing"
dialect "mvel"
when
$sample : SampleFact() from accumulate( $sf : SampleFact(),
init( SampleFact temp = null; ),
action( temp = ((null == temp) || ($sf.version > temp.version))
? $sf : temp; ),
result( ((null != temp) && (temp.value < 10)) ? temp : null;
))
then
insertLogical( "inserted" + $sample.toString() );
end
declare SampleFact
value : Long
version : Long
end
endfile <<example.drl>>
And test case:
file <<ExampleDroolsTest.java>>
package drools;
import java.io.IOException;
import org.drools.KnowledgeBase;
import org.drools.builder.KnowledgeBuilder;
import org.drools.builder.KnowledgeBuilderFactory;
import org.drools.builder.ResourceType;
import org.drools.definition.type.FactType;
import org.drools.io.ResourceFactory;
import org.drools.runtime.StatefulKnowledgeSession;
import org.junit.Test;
public class ExampleDroolsTest {
@Test
public void test() throws IOException, InstantiationException,
IllegalAccessException {
final KnowledgeBuilder kbuilder =
KnowledgeBuilderFactory.newKnowledgeBuilder();
kbuilder.add(ResourceFactory.newClassPathResource("drools/example.drl"),
ResourceType.DRL);
final KnowledgeBase kbase = kbuilder.newKnowledgeBase();
final StatefulKnowledgeSession ksession =
kbase.newStatefulKnowledgeSession();
final FactType sampleFactFt = kbase.getFactType("drools",
"SampleFact");
final Object sampleFact1 = sampleFactFt.newInstance();
sampleFactFt.set(sampleFact1, "version", 1L);
sampleFactFt.set(sampleFact1, "value", 5L);
final Object sampleFact2 = sampleFactFt.newInstance();
sampleFactFt.set(sampleFact2, "version", 2L);
sampleFactFt.set(sampleFact2, "value", 15L);
// final Object sampleFact3 = sampleFactFt.newInstance();
// sampleFactFt.set(sampleFact3, "version", 3L);
// sampleFactFt.set(sampleFact3, "value", 9L);
ksession.insert(sampleFact1);
ksession.insert(sampleFact2);
// Changes behavior to what I would expect, the "most recent"
SampleFact does not have
// a value < 10, so there should be no working memory element
inserted.
// ksession.insert(sampleFact2);
// ksession.insert(sampleFact1);
final int rulesFired = ksession.fireAllRules();
System.out.println("Rules fired: " + rulesFired);
System.out.println("Session objs:");
for (final Object o : ksession.getObjects()) {
System.out.println(o);
}
ksession.dispose();
}
}
endfile <<ExampleDroolsTest.java>>
My goal is to write a rule that only inserts a new object into working
memory when the
"most recent" SampleFact has value < 10. Most recent is defined here as the
greatest value for the "version" attribute of SampleFact.
Ultimately, I do not want to accept any SampleFact other than the most
recent to activate the rule and cause the String generated by
`"inserted" + $sample.toString()` to be inserted into working memory.
Of course, when inserting sampleFact1, then sampleFact2 as above in
ExampleDroolsTest.java, I get the String "insertedSampleFact( value=5,
version=1 )" inserted, and not retracted into working memory. I consider
this fact to be an intermediate result during accumulation of the
SampleFacts in the working memory.
I have observed that successive accumulate activations overwrite each other
if the rule is not fired in between. This generally gives me the desired
behavior of getting the most recent when all facts meet the `result`
criteria as well, however, in this example where the most recent fact does
not meet the additional `result` criteria, it fails my use-case.
What would be a good way to accomplish this, if I cannot use the accumulate
as I have currently?
--
View this message in context:
http://drools.46999.n3.nabble.com/Unexpected-behavior-of-accumulate-and-i...
Sent from the Drools: User forum mailing list archive at Nabble.com.
5:57 p.m.
By default the rule engine has no concept of latest, there is simply the data that is
currently present. It evaluates that data to produce activations, and then fires those.
If you need the notion of latest, and are unable to emulate this yourself, you can switch
the engine to steam mode and use events. If the fact is the event you can do something
like the follow:
s1 : SampleFact( )
not( SampleFact( this after s1 )
eval( s1.getValue < 10 )
Mark
On 27 Oct 2013, at 16:33, mikerod <mjr4184(a)gmail.com> wrote:
I appreciate your input. I understand what you are saying regarding
the
design of the activations caused by the accumulate.
Unfortunately, this is making it hard to express the logic I'm going for in
a concise manner.
I'll give a bit more of a "real", but similar example here.
With my example rules:
file <<example.drl>>
package drools
rule "testing"
dialect "mvel"
when
$sample : SampleFact() from accumulate( $sf : SampleFact(),
init( SampleFact temp = null; ),
action( temp = ((null == temp) || ($sf.version > temp.version))
? $sf : temp; ),
result( ((null != temp) && (temp.value < 10)) ? temp : null;
))
then
insertLogical( "inserted" + $sample.toString() );
end
declare SampleFact
value : Long
version : Long
end
endfile <<example.drl>>
And test case:
file <<ExampleDroolsTest.java>>
package drools;
import java.io.IOException;
import org.drools.KnowledgeBase;
import org.drools.builder.KnowledgeBuilder;
import org.drools.builder.KnowledgeBuilderFactory;
import org.drools.builder.ResourceType;
import org.drools.definition.type.FactType;
import org.drools.io.ResourceFactory;
import org.drools.runtime.StatefulKnowledgeSession;
import org.junit.Test;
public class ExampleDroolsTest {
@Test
public void test() throws IOException, InstantiationException,
IllegalAccessException {
final KnowledgeBuilder kbuilder =
KnowledgeBuilderFactory.newKnowledgeBuilder();
kbuilder.add(ResourceFactory.newClassPathResource("drools/example.drl"),
ResourceType.DRL);
final KnowledgeBase kbase = kbuilder.newKnowledgeBase();
final StatefulKnowledgeSession ksession =
kbase.newStatefulKnowledgeSession();
final FactType sampleFactFt = kbase.getFactType("drools",
"SampleFact");
final Object sampleFact1 = sampleFactFt.newInstance();
sampleFactFt.set(sampleFact1, "version", 1L);
sampleFactFt.set(sampleFact1, "value", 5L);
final Object sampleFact2 = sampleFactFt.newInstance();
sampleFactFt.set(sampleFact2, "version", 2L);
sampleFactFt.set(sampleFact2, "value", 15L);
// final Object sampleFact3 = sampleFactFt.newInstance();
// sampleFactFt.set(sampleFact3, "version", 3L);
// sampleFactFt.set(sampleFact3, "value", 9L);
ksession.insert(sampleFact1);
ksession.insert(sampleFact2);
// Changes behavior to what I would expect, the "most recent"
SampleFact does not have
// a value < 10, so there should be no working memory element
inserted.
// ksession.insert(sampleFact2);
// ksession.insert(sampleFact1);
final int rulesFired = ksession.fireAllRules();
System.out.println("Rules fired: " + rulesFired);
System.out.println("Session objs:");
for (final Object o : ksession.getObjects()) {
System.out.println(o);
}
ksession.dispose();
}
}
endfile <<ExampleDroolsTest.java>>
My goal is to write a rule that only inserts a new object into working
memory when the
"most recent" SampleFact has value < 10. Most recent is defined here as
the
greatest value for the "version" attribute of SampleFact.
Ultimately, I do not want to accept any SampleFact other than the most
recent to activate the rule and cause the String generated by
`"inserted" + $sample.toString()` to be inserted into working memory.
Of course, when inserting sampleFact1, then sampleFact2 as above in
ExampleDroolsTest.java, I get the String "insertedSampleFact( value=5,
version=1 )" inserted, and not retracted into working memory. I consider
this fact to be an intermediate result during accumulation of the
SampleFacts in the working memory.
I have observed that successive accumulate activations overwrite each other
if the rule is not fired in between. This generally gives me the desired
behavior of getting the most recent when all facts meet the `result`
criteria as well, however, in this example where the most recent fact does
not meet the additional `result` criteria, it fails my use-case.
What would be a good way to accomplish this, if I cannot use the accumulate
as I have currently?
--
View this message in context:
http://drools.46999.n3.nabble.com/Unexpected-behavior-of-accumulate-and-i...
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
7:52 p.m.
I understand that there is no concept within the engine of latest. I the
surprising behavior to me here was that the rule seems to become invalidated
in the LHS when the fact is matched with the a version than the previous
one. Using insertLogical, I generally expect the rule to retract previously
inserted facts when the rule is found to no longer be true for its
constraints.
This concept of stream mode and events is not something I am very familiar
with. I appreciate you pointing it out to me.
--
View this message in context:
http://drools.46999.n3.nabble.com/Unexpected-behavior-of-accumulate-and-i...
Sent from the Drools: User forum mailing list archive at Nabble.com.
7:55 p.m.
insertLogical does retract the inserted fact, if it’s supported becomes false.
Mark
On 27 Oct 2013, at 18:52, mikerod <mjr4184(a)gmail.com> wrote:
I understand that there is no concept within the engine of latest. I
the
surprising behavior to me here was that the rule seems to become invalidated
in the LHS when the fact is matched with the a version than the previous
one. Using insertLogical, I generally expect the rule to retract previously
inserted facts when the rule is found to no longer be true for its
constraints.
This concept of stream mode and events is not something I am very familiar
with. I appreciate you pointing it out to me.
--
View this message in context:
http://drools.46999.n3.nabble.com/Unexpected-behavior-of-accumulate-and-i...
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
8:08 p.m.
I agree that it is bad form to introduce unnecessary nulls. I mostly did
that for the sake of simplicity in the example.
I tried out your initial proposal of how to implement this, where you put
the logic previously in my `result` segment of the `accumulate` into the
"result pattern" segment of the rule instead. This seems to have the effect
that I am going for, but I'm a little suprised that it produces different
results.
The reason I say this, is here:
$sample : SampleFact() from
accumulate( $sf : SampleFact(),
init( SampleFact temp = null; ),
action( temp = ((null == temp) || ($sf.version >
temp.version)) ? $sf : temp; ),
result( ((null != temp) && (temp.value < 10)) ?
temp : null; ))
When a SampleFact is matched in the accumulate that has the highest version,
but does not have a value < 10, the `result` should
produce the value null, which I'd not expect to match the "result pattern"
of SampleFact().
So basically, I would have expected this to have the same effect as putting
the restricitons in the result pattern segment itself, as your example
showed me (i.e. SampleFact( value < 10 ) ).
Going with Davide's previous reply;
the difference seems to be that if the `result` produces null, then it does
not match SampleFact(), and so *does not* override previous accumulate
activations that did match.
However, when `result` produces a SampleFact that does not match SampleFact(
value < 10), it *does* override previous activations from the accumulate.
So I feel that there is a gap in my understanding of what is happening here
at this point.
I appreciate all the helpful feedback on this topic!
--
View this message in context:
http://drools.46999.n3.nabble.com/Unexpected-behavior-of-accumulate-and-i...
Sent from the Drools: User forum mailing list archive at Nabble.com.
6:05 p.m.
It's generally a bad idea to introduce null when it isn't absolutely necessary.
$sample : SampleFact( value < 10 ) from
accumulate( $sf : SampleFact(),
init( SampleFact temp = null; long vers =
Long.MIN_VALUE ),
action( if ($sf.version > temp.version){
temp = $sf; vers = $sf.version } ),
result( temp) )
Or even better
$sample : SampleFact() from
accumulate( $sf : SampleFact(value < 10 ),
and continuing as above.
-W
On 27/10/2013, mikerod <mjr4184(a)gmail.com> wrote:
I appreciate your input. I understand what you are saying regarding
the
design of the activations caused by the accumulate.
Unfortunately, this is making it hard to express the logic I'm going for in
a concise manner.
I'll give a bit more of a "real", but similar example here.
With my example rules:
file <<example.drl>>
package drools
rule "testing"
dialect "mvel"
when
$sample : SampleFact() from accumulate( $sf : SampleFact(),
init( SampleFact temp = null; ),
action( temp = ((null == temp) || ($sf.version >
temp.version))
? $sf : temp; ),
result( ((null != temp) && (temp.value < 10)) ? temp : null;
))
then
insertLogical( "inserted" + $sample.toString() );
end
declare SampleFact
value : Long
version : Long
end
endfile <<example.drl>>
And test case:
file <<ExampleDroolsTest.java>>
package drools;
import java.io.IOException;
import org.drools.KnowledgeBase;
import org.drools.builder.KnowledgeBuilder;
import org.drools.builder.KnowledgeBuilderFactory;
import org.drools.builder.ResourceType;
import org.drools.definition.type.FactType;
import org.drools.io.ResourceFactory;
import org.drools.runtime.StatefulKnowledgeSession;
import org.junit.Test;
public class ExampleDroolsTest {
@Test
public void test() throws IOException, InstantiationException,
IllegalAccessException {
final KnowledgeBuilder kbuilder =
KnowledgeBuilderFactory.newKnowledgeBuilder();
kbuilder.add(ResourceFactory.newClassPathResource("drools/example.drl"),
ResourceType.DRL);
final KnowledgeBase kbase = kbuilder.newKnowledgeBase();
final StatefulKnowledgeSession ksession =
kbase.newStatefulKnowledgeSession();
final FactType sampleFactFt = kbase.getFactType("drools",
"SampleFact");
final Object sampleFact1 = sampleFactFt.newInstance();
sampleFactFt.set(sampleFact1, "version", 1L);
sampleFactFt.set(sampleFact1, "value", 5L);
final Object sampleFact2 = sampleFactFt.newInstance();
sampleFactFt.set(sampleFact2, "version", 2L);
sampleFactFt.set(sampleFact2, "value", 15L);
// final Object sampleFact3 = sampleFactFt.newInstance();
// sampleFactFt.set(sampleFact3, "version", 3L);
// sampleFactFt.set(sampleFact3, "value", 9L);
ksession.insert(sampleFact1);
ksession.insert(sampleFact2);
// Changes behavior to what I would expect, the "most recent"
SampleFact does not have
// a value < 10, so there should be no working memory element
inserted.
// ksession.insert(sampleFact2);
// ksession.insert(sampleFact1);
final int rulesFired = ksession.fireAllRules();
System.out.println("Rules fired: " + rulesFired);
System.out.println("Session objs:");
for (final Object o : ksession.getObjects()) {
System.out.println(o);
}
ksession.dispose();
}
}
endfile <<ExampleDroolsTest.java>>
My goal is to write a rule that only inserts a new object into working
memory when the
"most recent" SampleFact has value < 10. Most recent is defined here as
the
greatest value for the "version" attribute of SampleFact.
Ultimately, I do not want to accept any SampleFact other than the most
recent to activate the rule and cause the String generated by
`"inserted" + $sample.toString()` to be inserted into working memory.
Of course, when inserting sampleFact1, then sampleFact2 as above in
ExampleDroolsTest.java, I get the String "insertedSampleFact( value=5,
version=1 )" inserted, and not retracted into working memory. I consider
this fact to be an intermediate result during accumulation of the
SampleFacts in the working memory.
I have observed that successive accumulate activations overwrite each other
if the rule is not fired in between. This generally gives me the desired
behavior of getting the most recent when all facts meet the `result`
criteria as well, however, in this example where the most recent fact does
not meet the additional `result` criteria, it fails my use-case.
What would be a good way to accomplish this, if I cannot use the accumulate
as I have currently?
--
View this message in context:
http://drools.46999.n3.nabble.com/Unexpected-behavior-of-accumulate-and-i...
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
6:07 p.m.
Oops - the "even better" doesn't quite achieve the same thing.
Sorry
-W
On 27/10/2013, Wolfgang Laun <wolfgang.laun(a)gmail.com> wrote:
It's generally a bad idea to introduce null when it isn't
absolutely
necessary.
$sample : SampleFact( value < 10 ) from
accumulate( $sf : SampleFact(),
init( SampleFact temp = null; long vers =
Long.MIN_VALUE ),
action( if ($sf.version > temp.version){
temp = $sf; vers = $sf.version } ),
result( temp) )
Or even better
$sample : SampleFact() from
accumulate( $sf : SampleFact(value < 10 ),
and continuing as above.
-W
On 27/10/2013, mikerod <mjr4184(a)gmail.com> wrote:
> I appreciate your input. I understand what you are saying regarding the
> design of the activations caused by the accumulate.
> Unfortunately, this is making it hard to express the logic I'm going for
> in
> a concise manner.
>
> I'll give a bit more of a "real", but similar example here.
>
> With my example rules:
>
> file <<example.drl>>
>
> package drools
>
> rule "testing"
> dialect "mvel"
> when
> $sample : SampleFact() from accumulate( $sf : SampleFact(),
> init( SampleFact temp = null; ),
> action( temp = ((null == temp) || ($sf.version >
> temp.version))
> ? $sf : temp; ),
> result( ((null != temp) && (temp.value < 10)) ? temp : null;
> ))
> then
> insertLogical( "inserted" + $sample.toString() );
>
> end
>
> declare SampleFact
> value : Long
> version : Long
> end
>
> endfile <<example.drl>>
>
> And test case:
>
> file <<ExampleDroolsTest.java>>
> package drools;
>
> import java.io.IOException;
>
> import org.drools.KnowledgeBase;
> import org.drools.builder.KnowledgeBuilder;
> import org.drools.builder.KnowledgeBuilderFactory;
> import org.drools.builder.ResourceType;
> import org.drools.definition.type.FactType;
> import org.drools.io.ResourceFactory;
> import org.drools.runtime.StatefulKnowledgeSession;
> import org.junit.Test;
>
> public class ExampleDroolsTest {
>
> @Test
> public void test() throws IOException, InstantiationException,
> IllegalAccessException {
> final KnowledgeBuilder kbuilder =
> KnowledgeBuilderFactory.newKnowledgeBuilder();
>
> kbuilder.add(ResourceFactory.newClassPathResource("drools/example.drl"),
> ResourceType.DRL);
> final KnowledgeBase kbase = kbuilder.newKnowledgeBase();
>
> final StatefulKnowledgeSession ksession =
> kbase.newStatefulKnowledgeSession();
>
> final FactType sampleFactFt = kbase.getFactType("drools",
> "SampleFact");
>
> final Object sampleFact1 = sampleFactFt.newInstance();
> sampleFactFt.set(sampleFact1, "version", 1L);
> sampleFactFt.set(sampleFact1, "value", 5L);
> final Object sampleFact2 = sampleFactFt.newInstance();
> sampleFactFt.set(sampleFact2, "version", 2L);
> sampleFactFt.set(sampleFact2, "value", 15L);
> // final Object sampleFact3 = sampleFactFt.newInstance();
> // sampleFactFt.set(sampleFact3, "version", 3L);
> // sampleFactFt.set(sampleFact3, "value", 9L);
>
> ksession.insert(sampleFact1);
> ksession.insert(sampleFact2);
>
> // Changes behavior to what I would expect, the "most recent"
> SampleFact does not have
> // a value < 10, so there should be no working memory element
> inserted.
> // ksession.insert(sampleFact2);
> // ksession.insert(sampleFact1);
>
> final int rulesFired = ksession.fireAllRules();
>
> System.out.println("Rules fired: " + rulesFired);
>
> System.out.println("Session objs:");
> for (final Object o : ksession.getObjects()) {
> System.out.println(o);
> }
> ksession.dispose();
> }
>
> }
>
> endfile <<ExampleDroolsTest.java>>
>
> My goal is to write a rule that only inserts a new object into working
> memory when the
> "most recent" SampleFact has value < 10. Most recent is defined here
as
> the
> greatest value for the "version" attribute of SampleFact.
> Ultimately, I do not want to accept any SampleFact other than the most
> recent to activate the rule and cause the String generated by
> `"inserted" + $sample.toString()` to be inserted into working memory.
>
> Of course, when inserting sampleFact1, then sampleFact2 as above in
> ExampleDroolsTest.java, I get the String "insertedSampleFact( value=5,
> version=1 )" inserted, and not retracted into working memory. I consider
> this fact to be an intermediate result during accumulation of the
> SampleFacts in the working memory.
>
> I have observed that successive accumulate activations overwrite each
> other
> if the rule is not fired in between. This generally gives me the desired
> behavior of getting the most recent when all facts meet the `result`
> criteria as well, however, in this example where the most recent fact
> does
> not meet the additional `result` criteria, it fails my use-case.
>
> What would be a good way to accomplish this, if I cannot use the
> accumulate
> as I have currently?
>
>
>
> --
> View this message in context:
>
http://drools.46999.n3.nabble.com/Unexpected-behavior-of-accumulate-and-i...
> 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
>
3834
days inactive
3835
days old
8 comments
4 participants
participants (4)
-
Davide Sottara -
Mark Proctor -
mikerod -
Wolfgang Laun