The role of inheritance in the maintainability of object-oriented systems

In this paper, we describe the empirical evaluation of the level of inheritance in five object-oriented systems. The systems studied vary in both size and application domain. Results from our analysis together with other recent investigations seem to support the thesis that inheritance is used either sparingly or incorrectly. Statistical correlation between four inheritance metrics and a set of dependent variables (non-comment source lines, software understandability, known errors and error density) provide evidence for this claim. It is also not clear that systems using inheritance will necessarily be more maintainable than those that do not. The data analysed from two of our systems suggests that deeper inheritance trees are attributes of systems which are harder to understand and (by implication) harder to maintain. We analyse why this might be the case, and propose ways of remedying this situation.