The utility of graph theoretic software metrics: a case study

The adoption of and adherence to object-oriented design and programming principles have allowed the software industry to create applications of ever-increasing complexity. A concomitant need arises for strategies to identify, manage, and, wherever possible, reduce this software complexity. One such strategy is the systematic collection, interpretation, and analysis of software metrics, mappings from software objects or constructs to sets of numerical features that quantify relevant software attributes. We describe a novel approach that employs various graph theoretic algorithms to analyze the higher level, application-wide class relationship graphs that emerge from object-oriented software. In addition to the software's overall inheritance tree characteristics, these algorithms will use metrics that reflect information on the import and export coupling of class-attribute and class-method relationships. Further, we incorporate information relating to the response sets for each object in the software, that is, the number of methods that can be executed in response to messages being received by objects.