Dynamically updating XML data: Numbering scheme revisited

Almost all existing approaches use certain numbering scheme to encode XML elements to facilitate query processing when XML data is stored in databases. For example, under the most popular region-based numbering scheme, the starting and ending positions of an element in a document are used as the code to identify the element so that the ancestor/descendant relationship between two elements can be determined by merely examining their codes. While such numbering scheme can greatly improve query performance, renumbering large amount of elements caused by updates becomes a performance bottleneck if XML documents are frequently updated. Unfortunately, no satisfactory work has been reported for efficient update of XML data. In this paper, we first formalize the XML data update problem by defining the basic operators to support most XML update queries. We then present a new numbering scheme that not only requires minimal code-length in comparison with existing numbering schema but also improves update performance when XML data is frequently updated at arbitrary positions. The fundamental difference between our new scheme and existing ones is that, instead of maintaining the explicit codes for elements, we only store the necessary information and generate the codes when they are needed in query processing. In addition to present the basic scheme, we also discuss some optimization techniques to further reduce the update cost. Results of a comprehensive performance study are provided to show the advantages of the new scheme.