Non-Directional XPath Processing in a Tree-Unaware RDBMS

xml query languages use directional path expressions to locate data in an xml data collection. They are tightly coupled to the structure of a data collection, and can fail when evaluated on the same data in a different structure. This paper extends path expressions with a new non-directional axis called the rank-distance axis. Given a context node and two parameters, α and β , the rank-distance axis returns those nodes that are ranked between α and β in terms of closeness from the context node in any direction. This paper shows how to evaluate the rank-distance axis in a tree-unaware xml database. A tree-unaware implementation does not invade the database kernel to support xml queries, instead it uses an existing rdbms such as Microsoft’s sql server as a back-end and provides a front-end layer to translate xml queries to sql. This paper presents an algorithm that translates queries with a rankdistance axis to sql. It also reports on several experiments that demonstrate the efficiency of the query evaluation plan produced by the translation.