XML Transformation Language Based on Monadic Second Order Logic

Although monadic second-order logic (MSO) has been a foundation of XML queries, little work has attempted to take MSO formulae themselves as a programming construct. Indeed, MSO formulae are capable of expressing (1) all regular queries, (2) deep matching without explicit recursion, (3) queries in a “don’t-care semantics” for unmentioned nodes, and (4) n-ary queries for locating n-tuples of nodes. While previous frameworks for subtree extraction (path expressions, pattern matches, etc.) each had some of these properties, none has satisfied all of them. In this paper, we have designed and implemented a practical XML transformation language called MTran that fully exploits the expressiveness of MSO. MTran is a language based on “select-andtransform” templates similar in spirit to XSLT. However, we design our templates specially suitable for expressing structure-preserving transformation, eliminating the need for explicit recursive calls to be written. Moreover, we allow templates to be nested so as to make use of an n-ary query that depends on the n− 1 nodes selected by the preceding templates. For the implementation of the MTran language, we have developed, as the core part, an efficient evaluation strategy for n-ary MSO queries. This consists of (a) an exploitation of the existing MONA system for the translation from MSO formulae to tree automata and (b) a linear time query evaluation algorithm for tree automata. For the latter, our algorithm is similar to Flum-Frick-Grohe algorithm for MSO queries locating n-tuples of sets of nodes, except that ours is specialized to queries for tuples of nodes and employs a partially lazy set operations for attaining a simpler implementation with a fewer number of tree traversals. We have made experiments and confirmed that our strategy yields a practical performance.