Projecting XML Documents Amélie Marian and Jérôme Siméon   View Paper (PDF)   Return to XML Query Processing II (Session A6) Abstract XQuery is not only useful to query XML in databases, but also to applications that must pro- cess XML documents as files or streams. These applications suffer from the limitations of current main-memory XQuery processors which break for rather small documents. In this paper we pro- pose techniques, based on a notion of projection for XML, which can be used to drastically reduce memory requirements in XQuery processors. The main contribution of the paper is a static analysis technique that can identify at compile time which parts of the input document are needed to answer an arbitrary XQuery. We present a loading algo- rithm that takes the resulting information to build a projected document, which is smaller than the original document, and on which the query yields the same result. We implemented projection in the Galax XQuery processor. Our experiments show that projection reduces memory requirements by a factor of 20 on average, and is effective for a wide variety of queries. In addition, projection results in some speedup during query evaluation. ©2004 Association for Computing Machinery