Simple SQL Validation of Generalized Entity Integrity

Abstract

Codd’s rule of entity integrity stipulates the existence of a primary key over every database table. That is, uniqueness and absence of null markers are enforced on the columns of the primary key. Key sets stipulate a generalized entity integrity rule that can be achieved on data sets where primary keys do not exist. Indeed, a key set means that different pairs of rows can be distinguished by unique non-null values on potentially different elements of the key set. While primary keys are a core feature of SQL databases, key sets have not been researched much at all. Our goal is to motivate the actual use of key sets in database systems. The use of key sets depends at least on the ability to identify those key sets that are meaningful in a given application domain, and to efficiently validate such key sets during the lifetime of the database. For this purpose, we analyze for the first time the performance of validating key sets in SQL experimentally, and also conduct experiments that provide insight on the time and size required to generate Armstrong relations for the implication of unary key sets by arbitrary key sets. Armstrong relations provide computational support for identifying key sets that are meaningful for a given application domain.