XQuery Implementation in a Relational Database System

Many enterprise applications prefer to store XML data as a rich data type, i.e. a sequence of bytes, in a relational database system to avoid the complexity of decomposing the data into a large number of tables and the cost of reassembling the XML data. The upcoming release of Microsoft’s SQL Server supports XQuery as the query language over such XML data using its relational infrastructure. XQuery is an emerging W3C recommendation for querying XML data. It provides a set of language constructs (FLWOR), the ability to dynamically shape the query result, and a large set of functions and operators. It includes the emerging W3C recommendation XPath 2.0 for path-based navigational access. XQuery’s type system is compatible with that of XML Schema and allows static type checking. This paper describes the experiences and the challenges in implementing XQuery in Microsoft’s SQL Server 2005. XQuery language constructs are compiled into an enhanced set of relational operators while preserving the semantics of XQuery. The query tree is optimized using relational optimization techniques, such as cost-based decisions, and rewrite rules based on XML schemas. Novel techniques are used for efficiently managing document order and XML hierarchy.