Towards a Streamed XPath Evaluation

XPath is a language for addressing fragments of XML documents, used in query and transformation languages such as XQuery and XSLT. For many applications it is desirable to process XPath on the fly and progressively against data streams. This diploma thesis is devoted to streamed and progressive evaluation of XPath. A streamed and progressive XPath evaluation considerably reduces the needed memory space. This is essential for XPath evaluation against very large documents. Furthermore, it enables the selection of informations from continuous services, such as stock market or meteorology data. The proposed approach enjoys attractive features. First, both translations, of general XPath into forward XPath, and of forward XPath into transducer networks generate output in linear time, with a size linear in the input size. Second, experiments with a straightforward prototype implementation suggests a remarkable efficiency of the approach. Third, the evaluation of an XPath expression can be performed in one pass over an XML stream and uses space at most quadratic in the document depth.