Thesis Abstract

Anyone given the job of designing a product must choose a suitable design technique and must then use it correctly. If design tools are to help them then the tools should ensure that the technique is used correctly without restricting proper use. NIAM is a conceptual data modelling technique for designing databases. A simple test to decide when NIAM can and cannot be used would benefit product designers. Simple rules to say when the construction and alteration of a data model are proper would benefit both product designers and design tool designers. As far as can be ascertained this information has not appeared in the literature. Therefore the purpose of this work is to answer some specific questions with the general theme

"When can I use NIAM and how might design tools help me?".

We start by demonstrating the diversity of objects that can rightfully be called databases. We do not restrict ourselves to computer systems. We then develop a new set-theoretical model of NIAM conceptual data models. The model is defined in Scheurer's Feature Notation and all results are proved rigorously. We prove that each data model we have modelled is well-formed in that it provides a well defined specification of a database. We also prove that we have modelled all possible well-formed NIAM data models, except those requiring the database to have unconventional mathematical properties. (Ordinary commercial and industrial databases are conventional). We use these results to devise a simple test to decide when NIAM can be used and when it cannot.

We define several editing operations on data models. Some do incremental changes; the others do more general "cut and paste" changes. We state simple preconditions for each operation and prove that if the preconditions are obeyed then the result of operating on a well-formed data model is also well-formed. We show that any editing action likely to be needed can be composed from these operations. It would be a straightforward matter to implement these operations and their precondition tests in a design tool. We also define a new class of equivalent constructions in NIAM data models. We devise and prove a simple test for equivalence.