Strategies for automatic code migration in the context of model-driven development

Abstract

In Model-Driven Software Development, models are not paper artifacts that serve only to communicate ideas. Instead, they are effectively used as input to code generators that produce complete applications. As a result, the developer works in a higher abstraction level, obtaining benefits such as increased productivity, quality and maintainability. In this context, the importance of the code generator goes beyond mere automation of repetitive tasks, and therefore there is the need to facilitate the process of creating and maintaining this artifact. The use of template-based generators, together with a reference implementation, is one of the most attractive approaches, because it is language-independent and looks like the final product, among other advantages. In this process, the reference implementation serves as an example of the code to be generated, and is the place where changes, debugging and testing occur. Once the reference code is validated, it is then migrated to the generator, in a process known as code migration. However, this causes code duplication (in the reference implementation and the template) and additional manual effort to keep both artifacts synchronized. This project proposes automatic strategies for code migration, aiming to solve these problems. Some techniques will be studied and applied, such as code annotation, specialized code editors and round-trip engineering. Even being a non-trivial problem, important contributions are expected, besides productivity gains due to the partial or total automation of this process. (AU)