Appropriate inferences of data dependencies in relational databases (Q429431): Difference between revisions

The paper discusses axiomatization systems for inferring data dependencies in relational databases. These dependencies are used to constraint data stored in a relational database to fulfill certain desiderata of an application domain which could not be enforced by a relational schema per se. Specifically, the authors consider functional and full hierarchical dependencies. A functional dependency specifies that the value of a set of attributes uniquely determines the value of another set of attributes in a relation; keys are well-known examples of functional dependencies. Full hierarchical dependencies are more general than functional dependencies and allow to decompose a relation into a set of relations whose natural join gives back the original relation. The inference of data dependencies plays an important role in the implementation of database management systems, especially when views are defined over normalized schemata. Some data dependencies can be inferred from other data dependencies regardless of the underlying relational schemata. Such a dependency continues to hold if relations and attributes not involved in the dependency are modified. The axiomatization system introduced by the authors allows to imply functional and full hierarchical dependencies not depending on the underlying relational schemata by applying inference rules which only consider data dependencies. Functional and full hierarchical dependencies depending on the underlying relational schemata are instead inferred by referring to the schemata only at the last step of the inference. Correctness proofs are very technical, but the authors have included some examples in order to facilitate the understanding of the results. All in all, this is an interesting paper presenting nontrivial results on inference systems for data dependencies. It is interesting for researchers of the field, who can enjoy all technical details, but also for researchers in related fields, who instead can appreciate the main results.