Experiences in automatic levelling of data flow diagrams

The data flow diagram (DFD) is a valuable tool in redocumentation of software systems, showing the interaction of programs and data. When drawn naively, DFDs are likely to be too large and dense to be of any practical use. Levelling of DFDs provides a logical approach to the production of manageable documentation. We consider the objectives addressed in the automatic creation of levelled diagrams from a given syntactical structure. We present techniques for automatic levelling by partitioning diagram processes based on graph connectivity, and grouping and scoping of data stores based on data flows. An extensive case study confirms the usefulness of the approach outlined