Towards a Unified Declarative and Imperative XQuery Processor

Since the birth of XML, the processing of XML query languages like XQuery/XQueryP has been widely researched in the academic and industrial communities. Most of the approaches consider XQuery as a declarative query language similar to SQL, for which the iterator-based (stream-based), lazy evaluation processing strategy can be applied. The processing is combined with XML indexing, materialized view, XML view query rewrite over source data. An alternative approach views XQuery as a procedural programming language associated with eager, step-based evaluation, where each expression is fully evaluated by the end of the corresponding expression execution step. Usually, this approach uses a virtual machine running byte-code for compiled programs. In this paper, we share our experience of building a unified XQuery engine for the Oracle XML DB integrating both approaches. The key contribution of our approach is that the unified XQuery processor integrates both declarative and imperative XQuery/XQueryP processing paradigms. Furthermore, the processor is designed with a clean separation between the logical XML data model and the physical representation so that it can be optimized with various physical XML storages and data index and view models. We also discuss the challenges in our approach and our overall vision of the evolution of XQuery/XQueryP processors.