Object-oriented query language design and processing

ii Abstract This thesis proposes an object-oriented query language that is more powerful than many existing query languages. The language is formally speciied and its expressive power is demonstrated by giving four translation schemes from other prominent object-oriented query languages. Further, this query language can be supported by a query algebra and both the query language and query algebra can be optimised using meaning preserving transformation rules. Object-Oriented Query Languages. The functional requirements of high-level object-oriented query languages are identiied and they combine as well as supplement features found in existing object-oriented query languages. EEectively they formulate a query model against which existing query languages can be evaluated and compared. An evaluation of four representative query languages chosen from research prototypes and commercial products shows that none satisses all the requirements. On the basis of the requirements a new query language, object comprehensions, is developed to provide a concise, clear, powerful, and optimisable query language for object-oriented databases. Some op-timisation opportunities for the novel features are identiied. A set of translation schemes from the query languages studied to object comprehensions is presented. Such translations demonstrate that object comprehensions are at least as powerful as these query languages and a system supporting object comprehensions can potentially support multiple query languages by providing translations to object comprehensions. Algebraic Support. The canonical algebra provides an abstract execution engine with which object comprehension queries can be expressed using algebraic operations. The translation scheme from object comprehensions to the canonical algebra is very simple and is novel for supporting queries involving mixed collection classes. The canonical algebra shares many operations with other query algebras and is formally speciied. A set of transformation rules that can be used for optimisation is presented whose validity can be veriied given the formal speciication. Formal Data Model. The data model which forms the basis of investigation is formally deened using the speciication language Z. This reference data model captures all the essential features of existing object-oriented data models including multiple inheritance. However, unlike existing data models, it also supports a generalised form of method overloading. Static type checking of such overloaded methods is studied in this thesis. iii Authorship The material presented in the thesis is the product of the author's own independent research carried out at the However the author is responsible for the majority of the technical substance of these papers and of this thesis. iv Acknowledgements …