Combining Related Products into Product Lines

Abstract

Abstract. We address the problem of refactoring existing, closely related products into product line representations. Our approach is based on comparing and matching artifacts of these existing products and merging those deemed similar while explicating those that vary. Our work focuses on formal specification of a product line refactoring operator called merge-in that puts individual products together into product lines. We state sufficient conditions of model compare, match and merge operators that allow application of merge-in. Based on these, we formally prove correctness of the merge-in operator. We also demonstrate its operation on a small but realistic example.

A Variability-Based Approach to Reusable and Efficient Model Transformations

“…The variability-based rules introduced in this paper are inspired by annotative representations of product lines [12][13][14] and augment representations proposed in earlier works.…”

Migrating Automotive Product Lines: A Case Study

“…With the help of our industrial partners, we validated that our exemplar is realistic in terms of its structure and size. Since our goal is to do transformation lifting, the product line we produced is annotative [10,18,26]. We formally review the definition of the annotative product line approach below.…”

“…The combination of product lines and model transformations has been extensively studied from the perspective of using transformations for configuring and refining product lines [10,15,17,11], and merging products and feature models [3,9,26], A theory of product line refinement along with a classification of commonly used refinement approaches is presented in [7]. Transformation lifting di↵ers from these works because it is about adapting existing product-level transformations to the level of entire product lines, as opposed to creating transformations specifically for product lines.…”

Quality of Merge-Refactorings for Product Lines

“…In our earlier work [22,24], we focused on refactoring model-level cloned product variants and proposed a configurable merge-refactoring algorithm, merge-in, applicable to refactoring models of different types (e.g., UML, EMF and Matlab/Simulink). Our algorithm identifies similar and different elements of the input models using parameterizable compare and match operators, and then constructs a refactored model using a merge operator.…”

“…The resulting product line model contains reusable elements representing corresponding merged elements of the original models. In [24], we formally proved that merge-in produces semantically correct refactorings for any set of input models and parameters: a refactored model can derive exactly the set of original products, regardless of particular parameters chosen and implementations of compare / match / merge used, if they satisfy well-defined correctness properties (e.g., "each element produced by merge originates from an element of at least one input model").…”