Indiana University Bloomington

School of Informatics and Computing

Technical Report TR653:
View-Constraint Duality in Databases and Systems Engineering

John Springer
(Oct 2007), 164 pages
[Ph.D Dissertation]
Abstract:
In the database systems and systems engineering domains, the concepts of constraints and views are commonly and effectively used. Considered as highly distinct, they stand as well-established notions in both domains' bodies of knowledge. Constraints may be expressed as ``business rules'' independent of underlying models. In these situations, the correctness of the implementation of the constraint may be in jeopardy and potential development inefficiencies may result as well. Furthermore, the modeler or designer may employ a model or methodology that does not support natively the specification of a particular constraint. As a result, such constraints only manifest themselves in the information system, and consequently, one is only aware when the constraint has been violated.

The focus of this research is that in fact a duality does exist between views and constraints (hypothesis 1) and that this duality is a useful tool in the development of information systems (hypothesis 2). Employing both proof and empirical evidence, our investigation reveals that the accuracy with which the first hypothesis holds depends upon the degree to which the constraints can be formalized. In the case of the relational data model, the constraints can be formalized. This is even true of the semantic constraints that are expressible in relational algebra and relational calculus. In the case of the other models that we explore, the constraints have a less formal expression, and views prove to be a method for interjecting more formality into the expression of the constraints.

Our results concerning the accuracy of the first hypothesis also hold for the second hypothesis; that is, the accuracy with which the second hypothesis holds depends upon the degree to which the constraints can be formalized. In the case of the relational data model, the constraints can be formalized, and thus we can generate general solutions for them. Nevertheless, uses of the duality outside the relational domain may be possible, and this is a focus of future work.

Available as: