Understanding Software Cohesion Metrics

Cohesion has been recognized as an important quality attribute of software design across decades. However measuring cohesion is not trivial. Several structural cohesion metrics have been proposed in the last decades and incorporated in industry-scale software measurement tools. Cohesion however is not always well represented by module internal structure. Therefore researchers have recently proposed an alternative way of measuring cohesion: the so-called conceptual cohesion metrics which are based on the explicit identification of the concerns realized by each module. However there is a lack of understanding on how conceptual cohesion metrics fit among a number of conventional structural cohesion metrics. In this context we performed a series of empirical studies in order to evaluate how and in which degree conceptual cohesion metrics differ from conventional structural cohesion metrics. Overall our research was based on data obtained from: (i) a web-survey with eighty software developers from nine countries and (ii) the repository of six open source long-lived industry-scale medium to large-sized systems with a change set of around thirty thousand source code revisions.