Beyond Lazy XML Parsing

Abstract

XML has become the standard format for data representation and exchange in domains ranging from Web to desktop applications. However, wide adoption of XML is hindered by inefficient document-parsing methods. Recent work on lazy parsing is a major step towards alleviating this problem. However, lazy parsers must still read the entire XML document in order to extract the overall document structure, due to the lack of internal navigation pointers inside XML documents. Further, these parsers must load and parse the entire virtual document tree into memory during XML query processing. These overheads significantly degrade the performance of navigation operations. We have developed a framework for efficient XML parsing based on the idea of placing internal physical pointers within the document, which allows skipping large portions of the document during parsing. The internal pointers are generated in a way that optimizes parsing for common navigation patterns. A double-Lazy Parser (2LP) is then used to parse the document that exploits the internal pointers. To create the internal pointers, we use constructs supported by the current W3C XML standard. We study our pointer generation and parsing algorithms both theoretically and experimentally, and show that they perform considerably better than existing approaches.

This project was supported in part by the National Science Foundation Grant IIS-0534530 and by the Department of Energy Grant ER25739.