Assessing Modular Structure of Legacy Code Based on Mathematical Concept Analysis

We apply mathematical concept analysis in order to modularize legacy code. By analysing the relation between procedures and global variables, a so-called concept lattice is constructed. The paper explains how module structures show up in the lattice, and how the lattice can be used to assess cohesion and coupling between module candidates. Certain algebraic decompositions of the lattice can lead to automatic generation of modularization proposals. The method is applied to several examples written in Modula-2, Fortran, and Cobol; among them a >100 kloc aerodynamics program.