De ning and Designing a Quality OO Metrics Suite

An eeective metric program is dependent upon high quality metric collection and analysis tools. One of the principal lessons to emerge from software measurement over the last twenty years is the need for metric collection tools (Fenton 1992). Conclusions reached from researching existing object-oriented metrics and supporting metric suite tools, indicate that there are certain qualities required of a high quality metrics suite which are currently not supported. Existing object-oriented design theories and principles highlight an approach that supports these qualities , and examples are presented to illustrate this. A metrics suite design is presented that supports these previously unsupported qualities, and it is concluded that object-oriented design has a particular aanity to the development of a high quality metrics suite. Seldom is there empirical evidence supporting object-oriented design metrics presented throughout the literature. Many have made claims, but few have substantiated these with empirical evidence. The object-oriented approach has matured to the stage where there is a need for fewer claims and more evidence. The metric suite design presented by this paper provides metric developers with a tool that supports empirical validation of object-oriented design metrics. 2 OO Design Quality During the past thirty years of software development , various principles have been proposed that, if adhered to, should lead to the development of quality products. These principles address various notions including: coupling , cohesion and modularity. They represent software engineering theories. Encapsulated in these principles is what Fenton (1992) deenes as the axiom of software engineering, whereby good internal structure) good external quality, illustrating that there are a myriad of factors impacting on software quality and that in order to develop quality software one must understand these factors. In order to do this, we must distinguish between internal software factors (McCall 1977) (those only perceptible to the computer professional, including coupling, cohesion and modularity) and external software factors (those whose presence or absence in a software product can be detected by the users of the products, including: maintainability, testability, reusability and reliability). This distinction is both important and subtle; as satisfaction of the internal qualities should ensure the satisfaction of external qualities. To this end, one can develop an abstract model depicting the relationships among quality characteristics , which can be used to qualitate and quantitate software products, by decomposing