Abstract:
SQL database designs can result from methodologies such as UML or Entity-Relationship modeling, Description Logic specifications, or relational normalization. Independently from the methodology, the use of good sample data is promoted by academia and commercial database design tools to visualize, validate and consolidate the database designs produced. Unfortunately, advice on what constitutes good sample data, or support to create good sample data are hard to come by. Armstrong databases provide a right notion of sample data that perfectly represent the domain semantics encoded in the form of SQL constraints. We present a tool that computes Armstrong sample tables for different classes of SQL constraints, and different interpretations of null markers. Armstrong tables illustrate the perceptions of an SQL database design about the semantics of an application domain. The tool exemplifies the impact of various design choices on Armstrong tables. These include the expressiveness of the classes of SQL constraints considered, and the semantics of null markers. Armstrong tables complement existing database design methodologies. In particular, they provide data samples that guide the transfer from relational approximations of an application domain to an actual real-life SQL table design.
