Managing Expressions as Data in Relational Database Systems

A wide-range of applications, including Publish/Subscribe, Workflow, and Web-site Personalization, require maintaining user’s interest in expected data as conditional expressions. This paper proposes to manage such expressions as data in Relational Database Systems (RDBMS). This is accomplished 1) by allowing expressions to be stored in a column of a database table and 2) by introducing a SQL EVALUATE operator to evaluate expressions for given data. Expressions when combined with predicates on other forms of data in a database, are just a flexible and powerful way of expressing interest in a data item. The ability to evaluate expressions (via EVALUATE operator) in SQL, enables applications to take advantage of the expressive power of SQL to support complex subscription models. The paper describes the key concepts, presents our approach of managing expressions in Oracle RDBMS, discusses a novel indexing scheme that allows efficient filtering of a large set of expressions, and outlines future directions.