Does
generative programming work with
large-scale C++ software development?
As a software architect and software development manager of large C++ projects, I have found John Lakos's book
Large Scale C++ Software Design to be required reading for anyone building large C++ projects. One of the principles of the book is to build classes whose external interfaces are subject to infrequent change. Using techniques such as full and partial encapsulation, and the
pImpl idiom, this can be achieved. The result is reduced compile-time dependencies.
More recently, Kryzsztof Czarnecki and Ulrich W. Eisenecker wrote
Generative Programming. This book describes the use of C++ templates in a very interesting and novel approach to make sofware that is generated rather than built. Since this approach relies on C++ templates, which are typically defined in C++ header files, it breaks the encapsulation principles espoused in Lakos's book. What I have found is that when I use generative programming techniques, I have to build less encapsulated code.
Has anyone successfully applied generative programming concepts to large-scale C++ projects? If you have, I would be very interested in hearing from you. Please post your response
here.
