Can We Avoid High Coupling?

Abstract

Abstract. It is considered good software design practice to organize source code into modules and to favour within-module connections (cohesion) over between-module connections (coupling), leading to the oftrepeated maxim "low coupling/high cohesion". Prior research into network theory and its application to software systems has found evidence that many important properties in real software systems exhibit approximately scale-free structure, including coupling; researchers have claimed that such scale-free structures are ubiquitous. This implies that high coupling must be unavoidable, statistically speaking, apparently contradicting standard ideas about software structure. We present a model that leads to the simple predictions that approximately scale-free structures ought to arise both for between-module connectivity and overall connectivity, and not as the result of poor design or optimization shortcuts. These predictions are borne out by our large-scale empirical study. Hence we conclude that high coupling is not avoidable-and that this is in fact quite reasonable.

Predicting performance via automated feature-interaction detection

“…These observations are analogous to previous work on coupling in feature-oriented and object-oriented software [17] [18], and footprint feature interaction [3]. We anticipate the same distribution for performance feature interactions, following a power law [18].…”

“…We anticipate the same distribution for performance feature interactions, following a power law [18].…”

Remodularization analysis using semantic clustering

“…In this work, we do not make use of coupling metrics. As stated by Taube-Schock et al [22], high coupling is not avoidable, and even natural, during the evolution of a software. In addition to the cohesion metrics, we also used two metrics for measuring the concentration and scattering of 2 http://code.google.com/p/topic-viewer.…”

Efficient Retrieval and Ranking of Undesired Package Cycles in Large Software Systems.

“…In this case it is very likely that cycles exist between these sub-packages. These subpackages are highly coupled, which does not represent a poor design but is necessary for readability [TSWW11].…”

“…For instance, the study of Taube-Schock el al. [TSWW11] shows that a high coupling does not always represent a poor design, and some high coupling might be necessary for a good design. In this case, cycles between highly coupled packages are desired.…”