Don'ing a Trait Looses the Map-Based Fact it Proxys - drools

Question

We are running into an issue with Drools when we try to "don" a trait to a map-based model.
It appears that after don'ing the trait the map-based fact it is proxying disappears (or gets retracted).
There does not seem to be any issue using the trait properties to get to the model properties.
In the scenario below the expectation is:
A populated map-based fact is supplied to the engine.
We modify the map-based fact based on constraints in a rule.
We check the constraints against the map fact and apply the trait.
We use the trait and the model in subsequent rules.
In rule 5, regardless of whether I add a 'bmi' property to the trait and try to set that and/or put the property in the map-based fact, it still does not even get to this rule (or anything after rule 3).
We are trying to use the model, mostly for constraints and updates, then within the rules apply and use traits to give us a stronger-typed fact and easier rule constraining.
Test Case:
public class TraitDonningTest extends CommonTestMethodBase
{
#Test
public void should_see_map_after_don() throws IOException
{
final String rule = Files.asCharSource(new File("src/test/resources/rules/traits.drl"), Charset.defaultCharset())
.read();
KieSession ks = loadKnowledgeBaseFromString(rule).newKieSession();
TraitFactory.setMode(VirtualPropertyMode.MAP, ks.getKieBase());
MyModel model = new MyModel();
model.put("weight", 1.0);
ks.insert(model);
ks.fireAllRules();
//We should have an updated BMI here
assertEquals(3.0, model.get("bmi"));
}
/**
* The real model based off of a map
*/
public static class MyModel extends HashMap<String, Object>
{
}
}
Rules:
package drools
import TraitDonningTest.MyModel
declare MyModel
#Traitable
end
declare trait MyTrait
#PropertyReactive
height : Double
weight : Double
end
rule "1 - Update Model"
when
$model: MyModel(
this['weight'] != null,
this['height'] == null )
then
modify($model) {
put('height', 2.0)};
System.out.println("Rule 1");
end
rule "2 - Apply Trait"
dialect "mvel"
no-loop
when
$model: MyModel(
this['weight'] != null,
this['height'] != null)
then
System.out.println("Rule 2A: " + $model.get('weight') + " (before don)");
MyTrait trait = don($model, MyTrait.class);
System.out.println("Rule 2B: " + $model.get('weight') + " (after don)");
System.out.println("Rule 2C: " + trait.weight );
end
rule "3 - Check Trait"
dialect "mvel"
when
$trait : MyTrait()
then
System.out.println("Rule 3: " + $trait.weight);
end
rule "4 - Check Model"
dialect "mvel"
when
MyTrait()
$model: MyModel()
then
System.out.println("Rule 4A: " + $model.get('height'));
System.out.println("Rule 4B: " + $model.get('weight'));
end
rule "5 - Compute Map"
when
$model : MyModel()
MyTrait(
$height : height,
$weight : weight)
then
double compute = $height + $weight;
modify($model) {
put('bmi', compute)};
System.out.println("Rule 5: " + $model.get('bmi'));
end
rule "6 - Summarize Results"
when
$a: Number() from accumulate( MyModel(), count(1) )
$b: Number() from accumulate( MyTrait(), count(1) )
then
System.out.println("Model: " + $a);
System.out.println("Trait: " + $b);
end
Output:
Rule 1
Rule 2A: 1.0 (before don)
Rule 2B: 1.0 (after don)
Rule 2C: 1.0
Rule 3: 1.0
Model: 0
Trait: 1
More details and testing...
I tried a few different scenarios using a strong-typed class as the fact rather than the map.
If I replace the map-based fact with a strong-typed fact the output is what I would expect:
Rule 1
Rule 2A: 1.0 (before don)
Rule 2B: 1.0 (after don)
Rule 2C: 1.0
Rule 3: 1.0
Rule 4A: 2.0
Rule 4B: 1.0
Rule 5: 3.0
Model: 1
Trait: 1
If I make the strong-typed fact (class) extend HashMap, even though I have the properties in the trait, it reverts back to the non-working result set.
Rule 1
Rule 2A: 1.0 (before don)
Rule 2B: 1.0 (after don)
Rule 2C: null
Rule 3: null
Model: 0
Trait: 1