Cohesion Metrics

We consider ways of measuring the cohesion of program fragments based upon techniques for program slicing, following the work of Ott et al 12, 2, 13, 11, 10]. The approach is based on the idea that the intersection of a program's slices represents that part of the fragment which is cohesive. We produce cohesion metrics that are structurally identical to those of Ott and Thuss 13], the diierence is that we consider diierent ways of measuring the signiicance of the intersection of a program's slices. We introduce expression metrics, to calculate the signiicance of the code in the intersection of slices, arguing that this approach may provide better answers than thèLines of Code' approach implicit in much of the literature 12, 10, 13]. The substitution of thèLines of Code' metric with alternative metrics, within the metrics deened by Ott and Thuss, motivates the consideration of metrics as higher{order functions, expressed in a pure functional programming language. This provides us with many convenient notational conventions with which to construct new metrics by partial application and instantiation of function{valued parameters, within such higher{order metrics. We also raise some new issues concerning the computability of a cohesion metric, based upon program slices, indicating that the approach may, in some cases give an over{large value for cohesion.