Model clone detection based on tree comparison

Abstract

Model driven development has become a key industry practice. With higher levels of abstraction and advent of domain specific languages, models find their presence in every field. Latest software engineering practices lead to large models which are really hard to design and manage. Significant overlaps in large models are really a matter of concern. Anecdotal evidences suggest that clones in models poses similar threats as in code. The paper introduces an approach to detect clones in UML models. The technique is based on finding similarities between two object oriented diagrams. Firstly, UML models are encoded as XMI files. Subtree comparison is applied after the XMI file is filtered and represented as a tree. Similarity between two diagram elements in a model is determined and reported as a clone.