Abstract

Online Analytical Processing (OLAP) applications are widely used in the components of contemporary Decision Support systems. However, existing OLAP query languages are neither efficient nor intuitive for developers. In particular, Microsoft’s Multidimensional Expressions language (MDX), the de-facto standard for OLAP, is essentially a string-based extension to SQL that hinders code refactoring, limits compile-time checking, and provides no object-oriented functionality whatsoever.

In this thesis, we present Native language OLAP query eXecution, or NOX, a framework that provides responsive and intuitive query facilities. To this end, we exploit the underlying OLAP conceptual data model and provide a clean integration between the server and the client language. NOX queries are object-oriented and support inheritance, refactoring and compile-time checking. Underlying this functionality is a domain specific algebra and language grammar that are used to transparently convert client side queries written in the native development language into algebraic operations understood by the server. In our prototype of NOX, JAVA is used as the native language. We provide client side libraries that define an API for programmers to use for writing OLAP queries. We investigate the design of NOX through a series of real world query examples. Specifically, we explore the following: fundamental SELECTION and PROJECTION, set operations, hierarchies, parametrization and query inheritance. We compare NOX queries to MDX and show the intuitiveness and robustness of NOX. We also investigate NOX expressiveness with respect to MDX from an algebraic point of view by demonstrating the correspondence of the two approaches
in terms of SELECTION and PROJECTION operations. We believe the practical benefit of NOX-style query processing is significant. In short, it largely reduces OLAP database access to the manipulation of client side, in-memory data objects