-
Bug
-
Resolution: Done
-
Blocker
-
7.7.0.GA
-
None
-
macOS 10.15.4, openjdk version "11.0.6" 2020-01-14, Drools 7.35.0.Final
-
2
-
Release Notes
-
-
-
-
-
-
CR1
-
+
-
-
2020 Week 16-18 (from Apr 13)
This happens both in RHPAM 7.7.0.GA, Drools 7.35.0.Final and Kogito 0.9.1-SNAPSHOT (and master).
Reproducer here: https://github.com/DuncanDoyle/drools-canonical-model-loop-issue
When running the unit-test with the Canonical Model enabled (which is the default in RHPAM 7.7.0.GA), this rule goes into an infinite loop. This does not happen with the Canonical Model disabled (-DgenerateModel=NO).
The problem seems to be with incorrect application of property reactivity. It seems that this line https://github.com/DuncanDoyle/drools-canonical-model-loop-issue/blob/master/src/main/resources/persons.drl#L9 is the problem (when you remove that System.out, looping is gone).
The generated rule with that line enabled:
public static org.drools.model.Rule rule_Set_32Duncan_39s_32age() { final org.drools.model.Variable<com.redhat.Person> var_$p = D.declarationOf(com.redhat.Person.class, DomainClassesMetadatae5c379871a3c48e99da0b6663f475772.com_redhat_Person_Metadata_INSTANCE, "$p"); final org.drools.model.BitMask mask_$p = org.drools.model.BitMask.getPatternMask(DomainClassesMetadatae5c379871a3c48e99da0b6663f475772.com_redhat_Person_Metadata_INSTANCE, "name", "age"); org.drools.model.Rule rule = D.rule("com.redhat", "Set Duncan's age").build(D.pattern(var_$p).expr("C1EB57BF17B7A3967DB15FAF80D69BBB", (com.redhat.Person _this) -> org.drools.modelcompiler.util.EvaluationUtil.areNullSafeEquals(_this.getName(), "Duncan"), D.alphaIndexedBy(java.lang.String.class, org.drools.model.Index.ConstraintType.EQUAL, DomainClassesMetadatae5c379871a3c48e99da0b6663f475772.com_redhat_Person_Metadata_INSTANCE.getPropertyIndex("name"), (com.redhat.Person _this) -> _this.getName(), "Duncan"), D.reactOn("name")), D.on(var_$p).execute((org.drools.model.Drools drools, com.redhat.Person $p) -> { { { System.out.println("Found: " + $p.getName()); ($p).setAge(41); drools.update($p, mask_$p); } } })); return rule; }
And with that line commented out:
public static org.drools.model.Rule rule_Set_32Duncan_39s_32age() { final org.drools.model.Variable<com.redhat.Person> var_$p = D.declarationOf(com.redhat.Person.class, DomainClassesMetadata9e133d566da042288db1930397a53ae8.com_redhat_Person_Metadata_INSTANCE, "$p"); final org.drools.model.BitMask mask_$p = org.drools.model.BitMask.getPatternMask(DomainClassesMetadata9e133d566da042288db1930397a53ae8.com_redhat_Person_Metadata_INSTANCE, "age"); org.drools.model.Rule rule = D.rule("com.redhat", "Set Duncan's age").build(D.pattern(var_$p).expr("C1EB57BF17B7A3967DB15FAF80D69BBB", (com.redhat.Person _this) -> org.drools.modelcompiler.util.EvaluationUtil.areNullSafeEquals(_this.getName(), "Duncan"), D.alphaIndexedBy(java.lang.String.class, org.drools.model.Index.ConstraintType.EQUAL, DomainClassesMetadata9e133d566da042288db1930397a53ae8.com_redhat_Person_Metadata_INSTANCE.getPropertyIndex("name"), (com.redhat.Person _this) -> _this.getName(), "Duncan"), D.reactOn("name")), D.on(var_$p).execute((org.drools.model.Drools drools, com.redhat.Person $p) -> { { { ($p).setAge(41); drools.update($p, mask_$p); } } })); return rule; }
The main difference (from a glance) is that in the first code-snippet, the bit-mask includes the age AND name, where in the second snippet, the bit-mask only includes the age. And this difference seems to be causing the loop.
Since the Canonical Model has been set as the default mode in version 7.7.0.GA, I find this a pretty serious issue, and therefore have marked this as a blocker.
- clones
-
-
- Closed
-
- incorporates
-
-
- Closed
-
-
-
- Closed
-
-
-
- Closed
-
