Language support for feature-oriented product line engineering

Abstract

Product line engineering is an emerging paradigm of developing a family of products. While product line analysis and design mainly focus on reasoning about commonality and variability of family members, product line implementation gives its attention to mechanisms of managing variability. In many cases, however, product line methods do not impose any specific synthesis mechanisms on product line implementation, so implementation details are left to developers. In our previous work, we adopted feature-oriented product line engineering to build a family of compilers and managed variations using the Standard ML module system. We demonstrated the applicability of this module system to product line implementation. Although we have benefited from the product line engineering paradigm, it mostly served us as a design paradigm to change the way we think about a set of closely related compilers, not to change the way we build them. The problem was that Standard ML did not fully realize this paradigm at the code level, which caused some difficulties when we were developing a set of compilers.In this paper, we address such issues with a language-based solution. MLPolyR is our choice of an implementation language. It supports three different programming styles. First, its first-class cases facilitate composable extensions at the expression levels. Second, its module language provides extensible and parameterized modules, which make large-scale extensible programming possible. Third, its macro system simplifies specification and composition of feature related code. We will show how the combination of these language features work together to facilitate the product line engineering paradigm.