Reverse Engineering Web Congurators

In many markets, being competitive echoes with the ability to propose customized products at the same cost and delivery rates as standard ones. As a result, companies provide their customers with online Web configurators to facilitate the product customization task. Web configurators offer a highly interactive configuration environment for customers to specify products that match their individual requirements and preferences. They provide capabilities to guide the customer through the multi-step and non linear configuration process, check consistency, and automatically complete partial configuration. To get a better grasp of what is the current practice in engineering Web configurators, we conducted a systematic empirical study of 111 configurators. We quantified their numerous properties, categorized patterns used in their engineering, and highlighted good and bad practices. We provided empirical evidence that Web configurators are complex information systems. Despite of this fact, this study revealed the absence of specific, adapted, and rigorous methods in their engineering. The lack of dedicated methods for efficiently engineering Web configurators leads to reliability, runtime efficiency, and maintainability issues. To migrate legacy Web configurators to more reliable, efficient, and maintainable solutions, we offer to systematically re-engineer these applications. This encompasses two main activities: (1) reverse engineering Web configurators to extract their configuration-specific data and encoding it into dedicated formalisms, and then (2) forward engineering new improved configurators. In this study, we are concerned with the reverse-engineering process. We developed a consistent set of methods, languages and tools to semi-automatically extract configuration-specific data from the Web pages of a configurator. Such data is stored in variability models (e.g., feature models). These models can later be used for verification purposes (e.g., checking the completeness and correctness of the configuration constraints) as well as input for forward-engineering techniques. To reverse engineer variability models from Web configurators, we developed techniques that target static structure and dynamic behaviour of Web configurators to locate and extract configuration-specific data. Experimental results on existing real Web configurators confirm the applicability of our contribution.