Performing Algorithmic Refinement before Data Refinement in B

摘要

Algorithmic Refinement is part of the theory of the B method both at the refinement and implementation stages. It a sign of how little loop introduction is used in practice at the refinement stage that neither the B-Toolkit nor Atelier-B provide support for loop introduction until the implementation stage. This paper examines the use of algorithmic refinement in general before data refinement. This involves extending the usual scope of data refinement which usually happens before algorithmic refinement. Two case studies are used to compare and contrast the application of algorithmic refinement before data refinement and vice versa. Some extensions are needed in the B-Toolkit to implement this style (i.e., algorithmic before data refinement) and are proposed. Some workarounds are also presented when appropriate.