Feature model recovery from product variants based on a cloning technique

A great number of Software Product Lines are not constructed from scratch, they are rather re-engineered from several similar product variants that have been in use. Existing methods for SPL feature model development from product variants may not be applicable when these products are developed independently and hence differ in their element names, methods code and structures. When different developers were involved in the development of the product variants, the naming assumption becomes too difficult to meet. In addition, the variants’ code structures are often different when developed separately and/or when one variant is derived from another through several modifications. Furthermore, while an operation may keep its signature and name in different code variants, its internal code may be different to meet the specifics of each product variant. This paper tackles these three differences by proposing a feature model extraction method that harmonizes the names in the product variants using semantic criteria, tolerates structural differences, and identifies code variability through a clone-code detection technique. It illustrates the method applicability through an example and a CASE tool automating the method’s steps.