Towards Crosscutting Metrics for Aspect-Based Features

Features are increments in program functionality and are the building blocks of product lines. Typical implementation of features using Object Oriented techniques commonly crosscuts several classes and interfaces. There exist many techniques that implement crosscuts, of which Aspect Oriented ones distinguish themselves by their support of dynamic crosscuts expressed with advice. Despite being a core tenet of AOP, very little attention has been paid to measuring crosscuts and how they are implemented by different language constructs in particular advice. In this paper we present a semi-formal definition of a set of basic metrics to measure crosscutting in features that use aspects. Our metrics categorize features within a crosscutting spectrum that ranges from heterogeneous to homogeneous according to the relative number and types of crosscuts features implement. This categorization helps assessing the actual use of aspects for feature implementation and provides a quantitative framework to gauge at and analyze the impact of aspects for product line development. We apply our metrics to a nontrivial product line case study implemented using AspectJ and relate our results to the ongoing assessment of aspects vs. collaboration-based designs for feature implementation.