Relational database reverse engineering: a model-centric, transformational, interactive approach formalized in model theory

Approaches to relational database reverse engineering often expect that the input has desirable characteristics and that it is complete; they also often fail to provide formal guarantees that their results are faithful to the initial input. Both of these problems can be addressed by using an incremental approach based on a formally defined target model. The incremental approach we propose in this paper quickly produces an initial model instance that is provably equivalent to the original relational database, which is assumed to be correct but may lack desirable characteristics and may be incomplete. The approach then proceeds incrementally using provably correct transformations. These incremental transformations allow for user interaction to provide needed information that may be missing or hard to obtain because the input lacks some desirable characteristics