An extension to the cyclomatic measure of program complexity

Abstract

A recent paper has described a graph-theoretic measure of program complexity, where a program's complexity is assumed to be only a factor of the program's decision structure. However several anomalies have been found where a higher complexity measure would be calculated for a program of lesser complexity than for a more-complex program. This paper discusses these anomalies, describes a simple extension to the measure to eliminate them, and applies the measure to several programs in the literature.

High-MCC Functions in the Linux Kernel

“…Here "predicate" should be understood as the full condition that is checked in an if or while statement. If this is actually composed of multiple atomic predicates, we could also count them individually; this approach is sometimes called the "extended" MCC [19]. Note that the metric counts points of divergence, but not joins.…”

Measurement and experimentation in software engineering

“…From experience he believes that testing and reliability will become greater problems in a section of code whose v(G) exceeds 10. 64 have developed different counting methods for computing cyclomatic complexity. These differences involved counting rules for CASE statements and compound predicates.…”

Cyclomatic Complexity: theme and variations

“…Whilst relying on "intuition" to support his case (as do other authors e.g. Myers (1977) cf. Shepperd (1988)), Sagri (1989) fails to comment on three distinct anomalies of the mode of presentation of "three IPs" in Figure 2.…”

“…In other words, it has a constant complexity in the sense that all runs make the same number of decisions. This gives it the highest overall operating complexity, reflecting the fact that this is a measure of computational resources used on average and has little to do with comprehension since graphs g2 and g3 are unstructured (Section 3.1) and graphs g4 and g6 contain nested structures, commonly regarded as adding complexity (Magel, 1981;Myers, 1977;Piwowarski, 1982) (see also discussion in Section Modularization and Nesting). Whilst relying on "intuition" to support his case (as do other authors e.g.…”

“…In this paper, we focus on the "McCabe family" of measures for the decision/logic structure of a program, which are those encapsulating the cyclomatic complexity (e.g. Myers, 1977;Piwowarski, 1982;Sagri, 1989) including extensions to describe modularization (Henderson-Sellers and Tegarden, 1993), nesting (Piwowarski, 1982) and object-oriented program structures -although we focus here on the simplest case of single-entry, single-exit modules. (Extensions to multi-entry/multi-exit modules are described in Henderson-Sellers and Tegarden, 1993).…”