An approach to impact analysis in software maintenance

Abstract

Impact analysis is a software maintenance activity, which consists of determining the scope of a requested change, as a basis for planning and implementing it. After a change request has been specified (change understanding) and the initial part of the system to be changed has been identified (change localization), impact analysis helps to understand consequences of the change on other parts of the system. Induced changes, also named ripple effects, among software components are detected. Most existing approaches perform impact analysis for changes occurring at the code level. In this thesis, concepts developed to perform impact analysis at the code level are applied to trace changes occurring at the design level. The method consists of proposing an activity model addressing the different steps of impact analysis and a data model on which propagations of changes can be traced. The method is validated with a case study applied to a system from the aerospace field. The tools we developed on PCTE help for consistency checks in HOOD based designs during editing. Our data-model based on an Entity Relationship notation describes a way to model HOOD diagrams in PCTE and further on to propagate changes on the repository. Examples chosen address the design phase of a simple engine system. We show that addressing modifications at a higher level of abstraction than the code eases understanding and localization of changes. It also limits the propagation of ripple effects (i.e., unexpected behaviour of the system) by detecting secondary changes at an earlier stage.