A program decomposition scheme with applications to software modification and testing

The authors use M. Weisers (1984) technique, to approximate the notion of direct-sum decomposition for the flowgraph representation of a program. Given a program slice, they define its complement and demonstrate how to obtain it; the slices and complements are used to define the components of the decomposition. These components are independent in the sense that they are executable projections of a subset of the program's specification and can have nonempty intersections. The authors characterize those projections that give a nontrivial decomposition and demonstrate how to extend any given projection to a maximal component with respect to the decomposition. Statements and variables within the components are characterized according to their appearance in other components. Variables and/or statements that appear in more than one component can be combined into one component that will completely delineate influence in the source. The decomposition yields a method and guidelines for maintainers to use so that changes can be assured to be completely contained in the modules under consideration and there are no undetected linkages between the modified and unmodified code.<
>