Referential Constraints

Abstract

Inclusion dependencies support essential semantic aspects of the standard relational data model. An inclusion dependency is defined as the existence of attributes in a table whose values must be a subset of the values of the corresponding attributes in another table (Codd, 1990; Abiteboul, Hull, & Vianu, 1995; Connolly & Begg, 2004). Formally, it can be expressed as R[X]?S[Z]. R and S are relation names. With X and Z as compatible attributes, R[X] and S[Z] are the inclusion dependency’s left and right sides respectively. When Z is the primary key of S or it is restricted by a unique clause, the inclusion dependency is key-based (also named referential integrity restriction, rir). In this case, X is a foreign key (FK) for R. On the contrary, if Z does not constitute the key of the relation, the inclusion dependency is non-key-based (simply, an inclusion dependency, id). Both rirs and ids are referential constraints. Rirs are important because they contain basic local semantic characteristics, which have been elicited from the Universe of Discourse (UofD). They are sufficient to symbolize many natural semantic links such as the relationships and hierarchies that are captured by semantic models (Abiteboul et al., 1995). Conversely, ids do not appear as a product of the translation ‘conceptual schema?logical schema’, but because of ad-hoc changes made in the phase of detail design, some denormalization degree, or the presence of complex n-ary relationship constructs. In this scenario, ids frequently misrepresent objects and their corresponding inter-object relationships.