This paper concerns with combining the compiled approach with the extended relational
model to evaluate queries in a non-Horn deductive database (DDB). In
modelling the real world, non-Horn clauses are useful to represent incomplete information.
The non-Horn clauses are classified into inclusive and exclusive disjunctions
so as to capture the natural meaning of the real world. The exclusive disjunctions
require careful consideration because they may derive new provably false facts and
may be lead to an inconsistent database.
We extend the relational model to m anipulate ground negative clauses,
ground inclusive disjunctions and ground exclusive disjunctions. Our extended relational
model retrieves the tuples which are in the fixpoint of a non-Horn DDB
excluding the provably false facts. Under the disjunctive database rule (DDR), we
develop a query evaluation procedure which accomodates inclusive disjunctions and
exclusive disjunctions.
The same approach is applied to the stratified database in which each
predicate has relative priority and lower priority predicates of a clause are used to
define the highest priority predicate.