Massey Documents by Type

Permanent URI for this communityhttps://mro.massey.ac.nz/handle/10179/294

Browse

Has files: YesSubject: search.filters.subject.Relational databases

Search Results

Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    Design of a relational data base management system : a thesis presented in partial fulfilment of the requirement for the degree of Master of Science in Computer Science at Massey University
    (Massey University, 1980) Vujcich, John M
    This thesis explains the design of a relational data base management system. In an effort to achieve a system which is shared, on-line, easy to use, responsive, capable of growth, capable of change, and having extensive security facilities, many innovations have been introduced. All data needed for enterprise operation and DBMS operation are stored in data base relations; administrators are considered as users; and one language is given with facilities for defining data, declaring mappings, defining comprehensive security/integrity constraints, and declaring new DBMS operations. Finally, a primitive language is given, which allows for a practical implementation of these innovations with a result of increased overall system performance, greater flexibility, and use of modern micro-processor technology.
  • Thumbnail Image
    Item
    Facilitating evolution in relational database design : a procedure to evaluate and refine novice database designers' schemata : a thesis presented in partial fulfilment of the requirements for the degree of Master of Business Studies in Information Systems at Massey University
    (Massey University, 1996) Ryder, Michael Robert
    Relational database management systems (RDBMS) have become widely used by many industries in recent years. Latterly these systems have begun to expand their market by becoming readily available at minimal cost to most users of modern computing technology. The quality of applications developed from RDBMSs however is largely dependent upon the quality of the underlying schema. This research looks at the area of schema design and in particular schemata designed by people who have a minimal understanding of relational concepts. It uses a survey and case studies to help define some of the issues involved in the area. A procedure to modify existing schemata is described, and the schema from one of the case studies used to apply the schema re-design procedure to a real database design. The results are compared to the original schema as well as a schema designed using a conventional application of the NIAM analysis and design methodology. The research supports the hypothesis that database applications based on schemata designed by lay-persons are currently being used to support business data management requirements. The utility, reliability and longevity of these applications depend to some extent on the quality of the underlying schema and its ability to store the required data and maintain that data's integrity. The application of the schema re-design procedure presented in this thesis reveals refinements on the original schema and provides a method for lay-persons to evaluate and improve existing database designs. A number of issues and questions related to the focus of this research are raised and, although outside the scope of the research, are noted as suggestions for further work.
  • Thumbnail Image
    Item
    An implementation of domains and keys in SQL : a thesis presented in fulfilment of the requirements for the degree Master of Philosophy at Massey University
    (Massey University, 1987) Alexander, John Lindsay
    The relational Data Base Management System model has steadily acquired increasing acceptance over the years since it was first introduced in 1970, yet few – if any – of the many relational products currently available support the fundamental concepts of domains and the integrity constraints of primary and foreign keys. Over more recent years the SQL Relational Database Sub-Language has found most favour among users and vendors alike, and a standard for SQL has recently been produced. This standard provides no support for domains or foreign keys, and only indirect support for primary keys. This thesis first reviews MURDER, the relational dabatase management system used for teaching purposes at Massey, and then describes an implementation of the domain and key concepts, highlighting some of the problem areas still to be resolved. Also described is an implementation of the query and update facilities of SQL, including some extensions which it is claimed increase its functionality. Finally, refinements to the language definition are suggested, to remove some redundancies and ambiguities.
  • Thumbnail Image
    Item
    The model theoretic notion of type in relational databases : a thesis presented in partial fulfillment of the requirements for the degree of Master of Information Sciences at Massey University, Wellington, New Zealand
    (Massey University, 2005) Paoletti, Alejandra Lorena
    Abstract. It is well known that finite model theory and database theory are two disciplines intimately connected. A topic which has been deeply studied in the context of finite model theory as well as in classic model theory, but which has not received the same attention in the context of databases, is the concept of type of a tuple. We believe that it is very important to have a clear understanding of the implications of the concept of type in the field of databases. Therefore, we study here the notion of type in FO (first-order logic) and in FOk (the restriction of FO to formulas with up to k variables), as well as in their corresponding infinitary extensions Lcow and L~w' respectively. The FOk types (L~w types) are quite relevant in database theory since they characterize the discerning power of the class of reflective relational machines of S. Abiteboul, C. Papadimitriou and V. Vianu with variable complexity k. We examine the three main characterizations of FO equivalence. These characterizations are based in Ehrenfeucht-Frai"sse games, back and forth systems of partial isomorphisms and isolating formulas for types. We study in detail all of them because they give a clear idea of the relevance of the notion of type in the context of databases, and make the fact that the types of the tuples in a database reflect all the information contained in it very clear. We found that the concept of type turned out to be useful to examine from a new perspective a well known problem of the field of databases, consisting on the redundant storage of information. Specifically, we initiate a study towards a method of normalization of relational databases based on the detection of redundant relations, i.e., relations which can be eliminated from the database without loosing information, by using isolating formulas for the types of the tuples in the database.
  • Thumbnail Image
    Item
    Dependencies in complex-value databases : a dissertation presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Information Systems at Massey University
    (Massey University, 2004) Link, Sebastian
    The relational data model has been the dominant model in database design for more than three decades. It considers data to be stored in matrices where rows correspond to individuals, columns correspond to attributes, and every cell constains [sic] a single atomic value. However, today's database technology trends, e.g. spatial, genetic or web-based data, require extended data models. Within the last decade new, complex-value data models such as the higher-order entity-relationship model, object-oriented data models, semi-structured data models, and XML have evolved which allow cells to contain lists, sets, multisets, trees, matrices or even more complex type constructors, references to other cells (which lead to infinite structures), and null values (indicating missing, unknown or vague data). Matrices as such allow the storage of inconsistent data, invalid in the semantic sense. As this is not acceptable, additional requirements called dependencies have to be formulated when designing a database. The correct specification and use of dependencies needs a sound mathematical basis. For the relational data model more than 90 different classes of dependencies have been defined and studied intensively. The major problems in dependency theory are the axiomatisability of classes of dependencies, determination of the closure of a chosen set of dependencies (as certain dependencies can be implied by others) and the characterisation of semantically desirable properties for well-designed databases (such as absence of redundancies or abnormal update behavior) by syntactic properties on closed sets of dependencies. With few exceptions research has only dealt with dependencies for the relational data model. Only recently, the emergence of XML as the standard format for web-based data and the rapidly increasing usage of persistent XML databases revealed the lack of a sound mathematical basis for complex-value data models. If they are expected to serve as first class data models they require a theoretical investigation of issues like integrity, consistency, data independence, recovery, redundancy, access rights, views and integration. The goal of this thesis is to develop a dependency theory for complex-value databases that is independent from any individual data model. Therefore, an abstract algebraic approach is taken that can be adapted to the presence of different combinations of type constructors such as records, lists, sets and multisets. Data models are classified according to the data types they support. In this framework the major objective is to initiate research on the following problems - investigate the axiomatisation of important dependency classes, relevant to complex-value data models, by sound and complete sets of inference rules that permit the determination of all dependencies implied by some chosen set of dependencies. - characterise semantically desirable properties by normal forms for complex-value data models and investigate whether these normal forms can always be achieved without violating other desirable properties. - develop efficient algorithms for determining the closure of a chosen set of dependencies and for restructuring databases such that normal forms are satisfied and no information is lost. In a single thesis it is impossible to consider all classes of relational dependencies in all different combinations of type constructors. Therefore the focus is put on extending two popular classes of relational dependencies: functional and multi-valued dependencies. The axiomatisation and implication of functional dependencies is investigated for all combinations of record, list, set and multiset type. Furthermore, a normal form with respect to functional dependencies in the presence of records and lists is proposed and semantically justified. It is also shown how to obtain databases which are in this normal form. Finally, axiomatisation and implication for the class of multi-valued dependencies and the class of functional and multi-valued dependencies are studied in the context of records and lists. The work of this thesis may lead to a unified dependency theory for complex-value data models.