Partition Based Path Join Algorithms for XML Data

Abstract

Path expression is an important component in querying XML data. The extended preorder numbering scheme enables us to quickly determine the ancestor-descendant relationship between elements in the hierarchy of XML data. Using the numbering scheme, a path expression can be evaluated by join operations to avoid potentially high cost of tree traversals. In this paper, we first formulate XML path queries as range-point join queries. Then we discuss the partition based algorithms that can utilize the range containment property to efficiently process the range-point join queries. Under the partition based framework, we propose three algorithms, namely Descendant partition join, Segment-tree partition join and Ancestor Link partition join, which can be chosen by a query optimizer for different input data characteristics. The experimental results show that the partition based algorithms can make better use of the buffer memory than sort-merge algorithms, and the proposed Ancestor Link join algorithm yields the best performance by using small in-memory data structures and by taking advantage of unevenly sized inputs.

This work was sponsored in part by the National Science Foundation CAREER Award (IIS-9876037), NSF Grant No. IIS-0100436, and NSF Research Infrastructure program EIA-0080123. The authors assume all responsibility for the contents of the paper.