Bringing Domain Knowledge to PatternMatching

摘要

This paper addresses the pattern matching problem for model transformation languages. Despite being an NP-complete problem, the pattern matching can be solved efficiently in typical areas of application. Prediction of actual cardinalities of model elements is the key to sufficient efficiency. The existing approaches aquire the actual cardinalities using complex run-time model analysis or using analysis of metamodel where the required information is poorly supplied. In the paper we show how the deeper understanding of domain which is targeted by model transformation language can dramatically reduce the complexity of pattern matching implementation. We propose a simple pattern matching algorithm for model transformation MOLA which is efficient for tasks related to the model driven software development. Additionaly a metamodel annotation mechanism is proposed. It refines the existing means of metamodelling by adding new classes of cardinalites. They make more efficient the pattern matching algorithms which do not use the complex run-time analysis.