Backward chaining stack overflow for recursive queries - drools

Question

I'm experimenting with Drools backward chaining mechanism and some simple Web Ontology Language (OWL)-RL logic. OWL supports inverse properties, which means I have to generate recursive queries from my TBox. The Drools documentation states that "The algorithm uses stacks to handle recursion, so the method stack will not blow up.", but when invoking my query, the CPU usage goes to 100% and the stack grows to infinity. I have three queries for the two inverse properties "tsEquivalen" and "phxEquivalent". The invoked query is the "bind_tsEquivalent_value" query.
query "bind_tsEquivalent_value"(Resource $subject, Resource $object)
#Abductive(target=ObjectPropertyQueryResult.class)
Statement(subject == $subject, predicate == tsEquivalent, $object := object)
or
$object := bind_phxEquivalent_inverse_value($subject;)
end
query "bind_tsEquivalent_inverse_value"(Resource $subject, Resource $object)
#Abductive(target=ObjectPropertyQueryResult.class)
Statement($object := subject, predicate == tsEquivalent, object == $subject)
or
$object := bind_phxEquivalent_inverse_value($subject;)
end
query "bind_phxEquivalent_inverse_value"(Resource $subject, Resource $object)
#Abductive(target=ObjectPropertyQueryResult.class)
Statement($object := subject, predicate == phxEquivalent, object == $subject)
or
$object := bind_tsEquivalent_inverse_value($subject;)
end
My ObjectPropertyQueryResult looks like this:
import com.hp.hpl.jena.rdf.model.Resource;
public class ObjectPropertyQueryResult {
private Resource subject;
private Resource object;
public ObjectPropertyQueryResult() {
super();
}
public ObjectPropertyQueryResult(Resource subject, Resource object) {
this.subject = subject;
this.object = object;
}
public Resource getSubject() {
return subject;
}
public void setSubject(Resource subject) {
this.subject = subject;
}
public Resource getObject() {
return object;
}
public void setObject(Resource object) {
this.object = object;
}
#Override
public int hashCode() {
final int prime = 31;
int result = 1;
result = prime * result + ((object == null) ? 0 : object.hashCode());
result = prime * result + ((subject == null) ? 0 : subject.hashCode());
return result;
}
#Override
public boolean equals(Object obj) {
if (this == obj)
return true;
if (obj == null)
return false;
if (getClass() != obj.getClass())
return false;
ObjectPropertyQueryResult other = (ObjectPropertyQueryResult) obj;
if (object == null) {
if (other.object != null)
return false;
} else if (!object.equals(other.object))
return false;
if (subject == null) {
if (other.subject != null)
return false;
} else if (!subject.equals(other.subject))
return false;
return true;
}
}

Answer 1

I think that this query combination is (if you pardon the expression) a wash-out. Running
bind_tsEquivalent_value"(sub, obj)
without a matching Statement leads to an evaluation of
bind_phxEquivalent_inverse_value(sub)
and this delegates to
bind_tsEquivalent_inverse_value(sub)
and then to
bind_phxEquivalent_inverse_value(sub)
and now you are caught in an infinite recursive loop.
A logical "or" without a constraint condition on the non-terminal branch is inadequate for breaking recursion. It is possible that the DRL compiler should terminate recursion when being caught in a loop like this, but basically (I think) this is a programmer's error.

Answer 2

"the CPU usage goes to 100% and the stack grows to infinity."
The heap or the stack? If your query never returns, it will recurse until you run out of heap space.
If it's a stack error, can you paste the trace.