Integration of Relational and Native Approaches to XML Query Processing

Abstract

Existing XML twig pattern query processing algorithms fall into two classes: the relational approach and the native approach. Both kinds of approaches have their advantages and limitations. Particularly, the relational approach can search for data values (content search) efficiently using tables, but it is not efficient to match query structure to documents (structural search). The native approach processes structural search efficiently, but it has problem dealing with values. In this chapter, a hybrid approach for XML query processing is introduced. In this approach, the content search and the structural search in a twig pattern query are performed separately using the data structures in the relational approach and the native approach, i.e. relational tables and inverted lists. The authors show that this hybrid style technique can process both structural search and content search efficiently, and then improve the query processing performance comparing to the existing approaches. Furthermore, when more semantic information on object class and relationship between objects in the XML document is known, the relational tables used can be optimized according to such semantic information to achieve a better performance. Finally after performing twig pattern matching, value results can be extracted easily using relational tables, rather than navigating the document again in many other approaches.