Pattern matching in a functional transformation language using treeparsing

Abstract

This article describes the techniques used for the implementation of pattern matching in TrafoLa-H, a functional transformation language. TrafoLa-H was designed to allow for very short and suggestive formulations of program transformations. Therefore, TrafoLa-H has a very powerful pattern language. There exist “distributed” patterns matching a set of subterms without fixed distances; patterns can match at arbitrary distance from the root of the whole term and anywhere in lists.

These powerful language constructs require special implementation techniques. By transforming TrafoLa-H patterns into regular tree grammars, and generating fast bottom-up parsers, it is possible to match even complex patterns in times linear to the size of the term to be matched.

After a short introduction of TrafoLa-H, the translation of patterns into possibly ambiguous regular tree grammars is presented, and the mechanism of variable binding is demonstrated.