A Superimposed Codeword Indexing Scheme for Handling Sets in Prolog Databases

While there has been growing interest in the use of Prolog for database applications, the size of these applications is limited by the capabilities of current Prolog systems for handling disk resident clauses. A major impediment is the inordinate amount of time required for retrieval and urujication of clauses from a large set stored on disk. Indexing is commonly used in conventional database systems to achieve fast data access to large sets of records. However, the use of variables and non-atomic arguments such as complex structures which may exist in a Prolog clause render conventional indexing strategies unsuitable. To overcome this, several indexing schemes have been developed. The Superimposed Codewor< plus Mask Bits (SCW+MB) is one such scheme which is eficient and relatively simple to implement. Sets which are finite lists consisting only of constant terms are useful structures for advanced database systems. Existing indexing schemes for Prolog databases, including SCW+MB, cannot handle sets. In this paper, a modified SCW+MB matching algorithm which can handle set arguments is proposed. Two hardware approaches, based on existing machines namely CURE-FSl and CAFS, for implementing the modified algorithm are suggested.