Transformation of Documents and Schemas by Patterns and Contextual Conditions

On the basis of the tree-regular language theory, we study document transformation and schema transformation. A document is represented by a tree t, and a schema is represented by a tree-regular language L. Document transformation is de ned as a composition of a marking function mPC and a linear tree homomorphism h, where P is a pattern and C is a contextual condition. Pattern P is a tree-regular language, and contextual condition C is a pointed tree representation. Marking function mPC marks a node if the subtree rooted by this node matches P and the envelope (the rest of the tree) satis es C. Linear tree homomorphism h then rewrites the tree, for example, by deleting or renaming marked nodes. Schema transformation is de ned by naturally extending document transformation; that is, the result of transforming a schema L, denoted h(mPC (L)), is fh(m P C (t)) j t 2 Lg. Given a tree automaton that accepts L, we can e ectively construct a tree automaton that accepts h(mPC (L)). This observation provides a theoretical basis for document transformation engines and document database systems.