Search Based Software Engineering for Variability Management

Because of economical, technological and marketing reasons today`s software systems are more frequently being built as Software Product Lines (SPLs) where each product implements a different combination of features - increments in program functionality. The effective management and realization of variability - the capacity of software artifacts to vary - is crucial to reap the benefits of SPL practices. Among these benefits are: increased software reuse, faster and easier product customization, and reduced time to market. An important challenge in Software Product Line Engineering (SPLE) is guaranteeing that all the desired software products (i.e. distinct feature combinations) can in fact be realized from the existing set of artifacts in a SPL. Verifying this guarantee is a non-trivial task because of the typically large number of software products and the inherent complexity of the artifacts involved. This guarantee is even more important when considering variability evolution, whether in effectively managing changes to an existing SPL or in extracting variability out of existing system variants - that may or may not have some common development history - to create a new SPL. Consequently, when this guarantee does not hold, it is vital to provide software designers with sets of fixes or other variability modeling and implementation alternatives to deal with this lack. Search-Based Software Engineering (SBSE) is an emerging discipline that focuses on the application of search- based optimization techniques to problems in Software Engineering. Among the techniques SBSE relies on are basic search algorithms such as hill climbing or simulated annealing, and evolutionary computation techniques such as genetic algorithms or genetic programming. SBSE techniques permit the elicitation of multiple software designs or module refactoring that can be evaluated and assessed using different software quality metrics as fitness metrics. Additionally, such techniques are generic, robust and scalable making them suitable for addressing the above challenges in variability management. The main goals of the project are to: perform a comprehensive review and analysis of SBSE techniques to identify the possible alternatives for tackling the identified SPL development problems and characterize their pros, cons and trade-offs involved; implement adequate tooling support that facilitates the application of the identified SBSE techniques; and conduct an assessment of the selected SBSE techniques with several cases studies of distinct sizes and application domains.

Software is present in almost every aspect of current modern life. To name a few examples, software controls our mobile phones, television sets, laundry machines, cars and the majority of public transportation and communication systems. However, product users have many similar needs but still may have special needs. This fact makes the manufacturing of products as part of product lines an economical necessity. A typical example is mobile phones that come with different options for cameras, memory sizes, battery duration, etc. Hence the software running behind these product lines must also be customized so that the product resources are effectively and efficiently used, without compromising any safety and security attributes of each individual type of product. This customization of software is a complex and demanding task which is currently done mostly by programmers in a manual and ad hoc way within the stringent technological and managerial constraints that demand high volumes of quality software delivered in short periods of time. Our work addressed this dire need by providing models and mechanisms to identify all possible combinations of features required for a software product line, assuring some basic functionality properties without exhaustive testing of all possible combinations, and supporting software engineers with information for the design and implementation decisions they have to make. The novelty of our work is the application of mechanisms inspired by nature to tackle the complexity of the software customization challenge in particular we explored techniques based on the basic notions of natural evolution proposed by Charles Darwin. The results obtained in our project proved the great potential and applicability of these kinds of techniques for software customization, mapped out the current state-of-the-art in this area, and laid down many open questions and research challenges thus defining a comprehensive research agenda. The adoption of these results by the software industry will materialize in more customizable software, tailored fit for the concrete needs of the users and products, while being more reliable, more environmentally-aware use of computational resources, and with reduced consumer costs.