Database models and query languages for relational data and metadata query processing.

摘要

Relational database systems do not effectively support queries which are independent of specific metadata context. The conventional relational query languages require that all queries, even the metadata independent queries, be formulated in strict accordance with the metadata of the database. Therefore, any change in metadata requires the reformulation of metadata independent queries--a very unnatural way of dealing with such queries. We call this phenomenon the metadata dependence problem of relational query languages. This dissertation describes our approaches to incorporate effective metadata query processing in relational database systems using reflective and meta-level techniques. We classify the approaches for achieving combined relational data and metadata querying capabilities in relational database systems into two categories: (1) extending the relational query languages with reflection principles, and (2) embedding the relational database model into a richer data model. We present our research in each of these categories, which serves to provide theoretical insights into the expressiveness and complexity of query languages that allow combined relational data and metadata querying capabilities.; Reflection in relational query languages is achieved by storing and manipulating queries as relations and by adding an evaluation operator to the query language. Van den Bussche, Van Gucht, and Vossen introduced a reflective relational algebra and showed that extending a query language with reflection mechanism not only increases its expressive power but also has several applications involving various forms of procedural data management. An useful application of a reflective query language is that it allows metadata independent expression of queries which suffer from metadata dependence problem. We extend the work done by Van den Bussche, Van Gucht, and Vossen and also present the reflective extension of the nested relational algebra.; As an approach belonging to the second category, we introduce a new model, the uniform database model, and its query languages. This approach permits the metadata to be treated as data. This uniform treatment of metadata and data allows for design of query languages that are immune to the metadata dependence problem. Besides providing metadata query processing capabilities, these uniform query languages can efficiently simulate conventional relational query languages. We present the properties of this model and its query languages and also situate this research in terms of expressiveness and complexity classes.