Forward chaining (source code)

= Forward chaining
{wiki}

Rule evaluation that repeatedly adds consequences of premises already known to hold. Positive <Datalog> evaluation computes its <least Herbrand model> in this way.