Paulo Borba

Mapping features to aspects: a model-based generative approach

Handling the various derivations of an aspect-oriented software family architecture can be a daunting and costly task if explicit support is not systematically provided throughout early and late development stages. Aspect-oriented software development (AOSD) has been recently explored as a technique that enables software product line customization. However, the application of AOSD has been limited to modularize specific crosscutting features encountered in the implementation of software product-line architectures or frameworks. Only a few works have investigated the development of product derivation approaches for AOSD. This paper presents a model-based generative approach to mapping features to aspects across different artifacts of a product line. Our main aim is to enable the smooth and systematic derivation of aspect-oriented software family architecture. Our approach is complementary to a set of previously-proposed modularization guidelines to implement aspect-oriented frameworks. We present details about the suite of mappings supported by our generative model, illustrate them in heterogeneous case studies, and discuss several implementation issues for its accomplishment.

Safe Evolution of Software Product Lines: Feature Extraction Scenarios

Software Product Lines can improve productivity and product quality, but product line maintenance is not simple, since a single change can impact several products. In many situations, it is desirable to provide some assurance that we can safely change a SPL in the sense that the behaviour of existing products is preserved. Developers can rely on previously proposed safe evolution notions, by means of transformation templates to ensure safe evolution. However, the existing templates were only applied in scenarios where a product line expands, and have not been evaluated in the context of extracting features from existing code. Therefore, we conducted a study using an industrial system developed in Java with 400 KLOC. This study revealed the need for new templates to address feature extraction scenarios, as well as improving the existing templates notation to address more expressive mappings between features and code assets. As a result of this study, we successfully extracted a product line from this existing system using the proposed templates, and also found evidence that the new templates can help to prevent defects during product line evolution.