XML query evaluation

XML is now widely used and management of XML data has become important. To this end, there has been work on the native management of XML data in a database to utilize the different capabilities of such a system like transaction management and indexing structures. At the heart of such a native XML database is the query evaluator, which provides access methods specifically tailored for XML data manipulation. The design of efficient access methods is the topic of this thesis. The most frequently used operation in an XML database is called structural join. Almost all XML queries contain at least one structural join. The structural join returns matches to a pattern from an XML document. We introduce a new efficient family of algorithms to address this task. These algorithms use a stack data structure that exploits the hierarchy of XML in favor of performance. We then develop variants that permit the combination of other operators, including projection, set difference, and universal quantification, with the structural join operation for greater efficiency. An important value provided by XML is the seamless representation of text and structured data. Querying the text with regard to the structure yields fast and accurate results. However, standard database query paradigms are not suitable for querying text. We introduce the TIX algebra for this purpose, and develop new access methods capable of efficiently computing and combining scores associated with intermediate results. In such applications, one is typically interested in only a few results with the highest scores. We develop new access methods to find results that score within a margin of error from the actual top results. These new access methods out-perform getting actual top results by at least an order of magnitude.