Functional Dependencies for Object Databases

There is an increasing use of object oriented methods (e.g. [Ru91], [UML97]) for the design of database applications and databases themselves. Like the Entity–Relationship model and its variants, structural models of these methods provide rich modelling concepts for specifying object schemas. Some kinds of integrity constraints, e.g. referential integrity constraints or cardinality constraints for relationships, can be specified with the help of schema elements along with the schema definition. Constraints not coverable by schema elements are considered as “general integrity constraints”. They are often described by means of a declarative, logic based language or in a procedural style by using methods. In the relational data model several families of integrity constraints such as functional dependencies, inclusion dependencies, or null constraints are distinguished. Such a separation into families of constraints is less common for object models. One reason for this may be the fact that because of the concepts available for structuring schemas some families do not have the same relevance for the schema design process as they have in connection with relational schemas ([Kh93]). From a database perspective, however, features such as access paths for object types and value based identification criteria seem to require a more explicit treatment of constraints, especially of functional dependencies. Both are of great importance in connection with the implementation of object schemas using relational database systems. Not only into this regard a formal approach to functional dependencies for object schemas including inference rules could enjoy several advantages.