Purely Relational FLWORs

We report on a compilation procedure that derives relational algebra plans from arbitrarily nested XQuery FLWOR blocks. While recent research was able to develop relational encodings of trees which may turn RDBMSs into highly efficient XPath and XML Schema processors, here we describe relational encodings of nested iteration, variables, and the item sequences to which variables are bound. The developed techniques are purely relational in more than one sense: (a) we rely on a standard (or rather: classical) algebra that is readily supported by relational engines, and (b) we use relational concepts like functional and multivalued dependencies to significantly simplify the emitted plans. This work blends well with the mentioned tree encodings and thus contributes a further important building block to investigations into XQuery processors based on relational database technology.

• Type: Conference
• Year: 2005
• Where: XIMEP
• Authors: Torsten Grust
• Keywords: Relational Algebra Plans, Relational Encodings, Relational Engines, XIMEP 2005