Data-Intensive XQuery Debugging with Instant Replay.pdf

Data-Intensive XQuery Debugging with Instant Replay Torsten Grust Jan Rittinger Jens Teubner Technische Universita?t Mu?nchen Munich, Germany torsten.grust | jan.rittinger | jens.teubner@in.tum.de ABSTRACT We explore the design and implementation of Rover , a post- mortem debugger for XQuery. Rather than being based on the traditional breakpoint model, Rover acknowledges XQuery’s nature as a functional language: the debugger fol- lows a declarative debugging paradigm in which a user is enabled to observe the values of selected XQuery subexpres- sions. Rover has been designed to hook into Pathfinder , an XQuery compiler that emits relational algebra plans for eval- uation on commodity relational database back-ends. The debugger instruments the subject query with fn:trace() calls which, at query runtime, populate database tables with relational representations of XQuery item sequences. Thanks to Pathfinder ’s loop-lifting compilation strategy, a Rover trace (1) may span multiple XQuery for iteration scopes and (2) allows for interactive debugging sessions that can arbitrarily replay iterations in a unique forward/backward fashion. Since the query runtime as well as the debugger are database-supported, Rover is scalable and supports the observation of very data-intensive XQuery expressions. 1. DEBUGGING XQUERY Our own experience has taught us that authoring a mod- erately complex XQuery expression from scratch more often than not tends to yield queries which exhibit unexpected behavior. Such bugs may sometimes have obvious fixes, but at times they might only occur with specific input XML in- stances or are, generally, hard to track down. While we do not postulate that XQuery authoring is inherently prone to error, there are a number of language characteristics and intricacies that may turn into pitfalls. Consider that (1) much of the XQuery semantics is implicit and does not surface at the language level (e.g., existential quantifi- cation or casting), (2) expression e