A bi-labeling based XPath processing system

We present BLAS, a Bi-LAbeling based XPath processing System. BLAS uses two labeling schemes to speed up query processing: P-labeling for processing consecutive child (or parent) axis traversals, and D-labeling for processing descendant (or ancestor) axis traversals. XML data are stored in labeled form and indexed. Algorithms are presented for translating XPath queries to SQL expressions. BLAS reduces the number of joins in the SQL query translated from a given XPath query and reduces the number of disk accesses required to execute the SQL query compared with the traditional XPath processing using D-labeling alone. We also propose an approximate P-labeling scheme and the corresponding query translation algorithm to handle XML data trees that contain a large number of distinct tag names, and/or are very deep. This extension captures a spectrum of XPath-to-SQL query translation schemes, ranging from existing schemes that do not use P-labels to the one that uses exact P-labels. Experimental results demonstrate the efficiency of the BLAS system. Disciplines Computer Sciences Comments Chen, Y., Davidson, S., & Zheng, Y., A Bi-Labeling Based XPath Processing System, Information Systems 35, 2010, doi: 10.1016/j.is.2009.05.005 This conference paper is available at ScholarlyCommons: http://repository.upenn.edu/cis_papers/600 A bi-labeling based XPath processing system Yi Chen a, , Susan B. Davidson , Yifeng Zheng c a Arizona State University, United States b University of Pennsylvania, United States c Amazon.com, United States a r t i c l e i n f o