A set-oriented inference mechanism for large data-intensive logic-bases

Several strategies for processing queries in data-intensive logic bases (deductive databases) have been recently proposed. These strategies can be grouped into the one-tuple-at-a-time and the set-of-tuples-at-a-time classes. A strategy of the first class searches the logic base for one solution to the processed query at any point in time and uses backtracking to find alternative solutions. A strategy of the second class searches the logic base to find all of the solutions to the query simultaneously, thus eliminating the need for backtracking. The modification of Prolog's tuple-at-a-time processing strategy into a set-at-a-time one is investigated. The set-modified version of Prolog has been implemented, and the resulting interpreter has been benchmarked against a classical Prolog one, showing a dramatic improvement in performance.<>