Equivalence of Program Dependence Graphs in Refactoring

Program dependence graphs are widely used in computer science for example when performing optimizations, parallelization and as an internal program representation in integrated development environments. In this paper, I will look into the usefulness of program dependence graphs regarding refactoring. Refactoring is a very useful technique to improve the internal structure of a program. Many modern integrated development environments provide automated refactoring support because the process of refactoring, when done by hand is time-consuming an error-prone. This article summarizes three papers which are then used as a basis to develop ideas on the usefulness of program dependence graphs in automated refactoring. These papers all define equivalence theorems for a range of languages which, for some programs, with the use of program dependence graphs, can show that the programs are strongly equivalent when considering a given initial and a resulting final state. This means that two programs can represent the same program behavior even if they are not exactly the same. Since this is often the case for program code before and after refactoring, I examine in this article weather the ideas behind those theorems could be useful in the process of automated refactoring. It is discussed whether an extension of those equivalence theorems would be possible and necessary, to make them feasible for automated refactorings by guaranteeing that an automatically applied refactoring indeed preserve the functionality of the program.