The Dependency Graph Change Impact Analysis Based On Multi-granularity Method

Software change impact analysis is a key technique for identifying unpredicted and potential effects caused by changes made to software, which plays an important role in software maintenance.This paper proposes a static change impact analysis based on multi-level dependency graph. The contribution is:Proposed an internal representation for object-oriented software:multi-level dependency graph. Comparing to traditional class-method level dependency graph, multi-level dependency graph can depict encapsulation, inheritance and polymorphism much more precisely.Proposed a set of abstract representation deriving from multi-level dependency graph. It can depict the inheritance relationship between classes, the override relationship between methods, the Execute-After relationship between call sites, and the virtual function dispatch, which provide a foundation for better change impact analysis.Proposed a change impact analysis technique based on the multi-level dependency graph. It can calculate impact set in an iterative way, and the result is more precise than that deriving from traditional method which uses transitive closure on a dependency graph to measure impact.Different change types often have different impact mechanisms, even some changes do not impact other entities in programs in spite of some dependences existed between these entities and the modified entity. This paper proposes a change classification. After studying the impact mechanism of each change type, we give the abstract representation of calculation of impact set of each change type.