Analyzing the Complexity of Domain Model Representations

Software Product Lines (SPL) are a viable method for systematic reuse. An essential part of an SPL is the domain model. In order to be efficient the domain model should have an as low as possible complexity. In this way the usability and maintainability of the domain model can be improved. This is important because of the long lifecycle and the hopefully high number of derived products. One important influence factor for the complexity is the choice of the domain modeling paradigm. Another point is the design of the model. Various aspects can be modeled in different ways resulting in different levels of complexity. To the best of our knowledge there is no simple metric to measure and compare the complexity of different domain representations. To make it clear we do not measure the complexity of the domain itself, but of its representation. This work suggests simple metrics to estimate interface, element and property complexity, the main building blocks of domain models. These three values are simply summed up for an overall complexity. In this way they can be used to compare the complexity of different representations. In order to be able to show that our metrics yield useful values despite their simplicity we investigated several use cases. We show the influence of the modeling paradigm and various characteristics of the domain on the complexity of the representation. Finally, we suggest a method to reduce the complexity of complex, heterogeneous domains.