Exploring the Relationship between Cohesion and Complexity

Abstract

Many metrics have been proposed to measure the complexity or cohesion of object-oriented
 software. However, the complexity or cohesion of a piece of software is more difficult to capture than
 these metrics imply. In fact, studies have shown that existing metrics consistently fail to capture
 complexity or cohesion well. This study explores the reasons behind these results: cohesion is difficult
 to capture from syntactic elements of code, complexity is too multi-faceted to be captured by one
 metric and the qualities of complexity and cohesion are not independent. These factors have resulted in
 metrics that are purported to measure complexity or cohesion but are inadequate or misclassified. This
 study shows that there is overlap between some of the complexity and cohesion metrics and points to a
 more basic relationship between complexity and cohesion: that a lack of cohesion may be associated
 with high complexity

Object-oriented product metrics: A generic framework

“…The axiom is analogous to this argument, suggesting that the decision to use a metric to measure multiple properties does not necessarily make the metric equivalent for measuring these properties. Yet another analogy to the arguments from Stein et al [36] that is similar to our presentation is the distinction between internal and external properties. Theoretically, a metric cannot measure two different properties.…”

“…And third, while the framework can be used to generate a large number of variations of metrics, this does not mean that the metrics are indeed meaningful or useful as predictors. Finally, the framework is not intended to replace research related to analyses of individual metrics such as characterizing behaviors of metrics in response to ranges assumed by input parameters or complexity analyses of existing metrics (e.g., [36]). …”

“…It may, of course, be possible to interpret a metric to 'inform' multiple properties. For example, Stein et al [36] distinguish between syntactic and semantic properties, indicating that the two may be correlated but not equivalent. The axiom is analogous to this argument, suggesting that the decision to use a metric to measure multiple properties does not necessarily make the metric equivalent for measuring these properties.…”

Exploring a Bayesian and linear approach to requirements traceability

“…The conjecture is that the higher the effort the higher the complexity of the produced code artefact, and hence a more relevant artefact to analyse for the quality of the generated software. Furthermore, a number of studies have shown that complexity correlates with the fault proneness of code artefacts [23,34,35]. Our evaluation approach is twofold.…”

Open Source Programmers' Information Seeking During Software Maintenance

“…The amount of software lifecycle effort consumed during this phase has been estimated to range between 60% and 80% of the entire lifecycle effort (Lientz et al, 1978;Mayrhauser and Vans, 1993;Pressman, 2004;Zayour and Lethbridge, 2001). While the empirical basis for such statements are dated and suggestions that they should be revisited have been made (Kemerer and Slaughter, 1999), the increasing scale and complexity of newer software systems (Pressman, 2004;Sommerville, 2008;Stein et al, 2005) implies that the effort invested in maintenance of successful systems can only have increased. Thus research in this area is vital towards the discovery and evolution of supportive methods or tools, which could aid maintainers in their software maintenance efforts.…”