Extracting input/output dependencies from XSLT 2.0 and XQuery 1.0

Abstract

The success of XML has come with the explosion of the number of vocabularies defined to represent similar information. In many domains, ranging from life sciences to e-commerce, many mappings have been defined between these vocabularies as a strategy for information integration. However, as the vocabularies evolve, maintaining these complex mappings and all artifacts that rely on them remains one of the major issues in information integration. A critical aspect of schema (or vocabulary) evolution is the ability to perform an impact analysis across all artifacts that directly or indirectly depend on the changing vocabularies. When an evolving vocabulary is not directly used, but rather mapped to another one through a complex mapping expressed in languages such as XSLT or XQuery, an impact analysis can be performed at a fine level of granularity only if the dependencies between the source and the target vocabulary are also known with the same level of granularity. In this paper, we present an architecture for the static analysis of both XSLT and XQuery and a conservative and sound constraint-based analysis of these two languages that effectively uncovers input/output dependency information necessary for a fine grain impact analysis. For each element or attribute declaration d in the output schema the analysis determines all the attribute or element declarations in the input schemas whose instances may influence the evaluation of an instance of d. We show results of early experimental evaluation illustrating the accuracy and scalability of the approach. Copyright © 2005 Achille Fokoue.