Safe referential integrity and null constraint structures in relational databases

Abstract Referential integrity and null constraints are extensively used in relational databases for expressing existence dependencies between tuples and restrictions on the way nulls appear in relations, respectively. Referential integrity and null constraints can cause the following data manipulation anomalies: 1. (i) referential integrity and null constraints may block the execution of certain data manipulations, although these manipulations would result in consistent database states, and 2. (ii) depending on the order of enforcing the referential integrity constraints, certain data manipulations may result in different database states, thus being unpredictable. In this paper we examine these anomalies and formally develop safeness conditions for avoiding them.