Quantitative techniques for the assessment of correspondence between UML designs and implementations

Abstract

In this paper we discuss approaches to assess the correspondence between a software design and its implementation. We consider object oriented software systems which are designed using the UML notation. Correspondence is important for understanding the system since designs are easier to comprehend than large pieces of source code. To assess the correspondence of a system, we match entities from the design to pieces of source code. We define a matching based on classifiers. Several approaches are discussed to establish such a matching. These approaches are matching based on classifier names, matching based on metric profiles and matching based on structural properties of classifiers. Once this matching is defined, it is possible to detect and visualize the actual differences between design entities and parts of source code. The approaches have been validated through an industrial case study.