Kazimierz Subieta, Jacek Leszczylowski, Irek Ulidowski

Processing Semi-Structured Data in Object Bases

852

Abstract

We address the problem of null values and other forms of semi-structured data in object-oriented databases. Various aspects and issues concerning semi-structured data that are currently presented in the literature are discussed in the paper. We propose a new universal approach to semi-structured data based on the idea of absent objects. The idea covers null values and union types and can be smoothly combined with the idea of default values. We introduce a simple model of object store that is similar to the Tsimmis model. In contrast to the main stream of the research, our basic assumption is that semi-structured data are not only to be queried but also processed by an integrated query/programming language. To this end we discuss query language constructs that are relevant to query semi-structured data and corresponding issues in programming languages. The idea follows the stack-based approach to integrated query/programming languages that we have implemented in the LOQIS system. Finally we briefly discuss a new approach to polymorphic typing of semi-structured data that is implemented in LOQIS.

Key words: null values, union types, semi-structured data, object-oriented databases, objectstore, querying semi-structured data, query/programming language integration, semi-structured data typying.