Storage and Retrieval of First-order Terms Using a Relational Database

We present several practical techniques for storing first-order terms (e.g. clauses) in a relational database, and for mechanically constructing SQL select queries which retrieve all terms matching a given pattern. First we outline the deductive database application which prompted us to develop the techniques, then we present our approach to clause storage and indexing, and consider three variants of the unifiability retrieval condition. We discuss several issues in interfacing Prolog to a RDBMS, including caching and re-entrancy. We then present four techniques: one for storing and retrieving general terms; a specialisation of this for storing ground terms; another specialisation for storing finite maps; and a quite different technique for ground terms. We comment on the complexity of some of the generated SQL queries.