Meta-level control for deductive database systems (Q1202083)

The volume is the author's doctoral dissertation accepted by the Technical University Munich. After description of the Horn clause langauge an expert deductive database system is proposed which offers possibilities for controlling the deduction process using the least fixpoint operator. The deduction control can be used for ensuring the termination of the deduction process, for increasing its efficiency and for refining the semantics of under-specified deductive databases. At the object level of the expert deductive database system the programmer defines the logical aspects of the problem. At the meta-level, control information about the object-level deduction process is specified. A deductive database is used both at the object-level and at the meta- level. The feedback from the meta-level to the object level is specified via so-called control predicates that -- if they are deduced at the meta- level -- have some pre-defined effect on the object-level deduction process, e.g., irrelevant rules are disregarded, irrelevant tuples are discarded, useful rules and tuples are preferred or recursion explicitly terminated. All these possibilities guarantee that the object-level deduction process remains sound. Completeness may be violated now, since it is no longer guaranteed that each tuple in the answer of an object- level query will eventually be deduced. Another important goal is efficiency of the object-level deduction process -- this can be achieved by the two sloppy delta-iteration schemes given in the appendix of the book. Some questions of the process remained open, e.g. semantics-changing use of control, the mutual interdependence of different control possibilities and/or the computing of the entire answer of the original query without having to computer the answer of the subquery as a whole (when preferring useful rules or tuples is used for a subquery).