| In this thesis, the problem we tackle is how to facilitate the reuse of a set of independently developed C++ graph libraries. We are concerned more precisely with library users being offered an alternative to having to study in detail the code of a library before deciding whether or not and how to reuse its components. Our fundamental postulate is that domain engineering does facilitate reuse, as extensively argued in the corresponding literature. Thus, the question we ask is the following: can the notions of domain engineering, and more specifically the technique of feature modeling, be applied to a set of independently developed C++ graph libraries (in order to facilitate reusing these libraries)? Our thesis is that they can and we demonstrate this by building feature models addressing the representational aspects of a family of independently developed graph libraries.; First, we develop a partial feature diagram for each graph library. Our intention here is to have these models provide a conceptual "road map" (like the blue-print of a house) for each graph library. No such information is currently available in the documentation associated with each library. From a practical viewpoint, we believe that each such feature model helps in understanding the design of its corresponding library because this model provides a conceptual overview of its library. Then, as a second contribution, in order to highlight the differences between these libraries, we present an integrated feature model of the graph domain and explain how each library 'fits' in it. Though comparing libraries (e.g., with respect to requirements, variability, etc.) lies beyond the scope of this thesis, we hint at the usefulness of our feature models by using them to sketch a comparison between the libraries. |
