Extracting refactoring trends from open-source software and a possible solution to the 'related refactoring' conundrum

摘要

Refactoring, as a software engineering discipline has emerged over recent years to become an important aspect of maintaining software. Refactoring refers to the restructuring of software according to specific mechanics and principles. In this paper, we describe a tool that allows refactoring data across multiple versions of seven open-source software systems to be collected. The tool automates the identification of refactorings as program transformations between consecutive software releases. The same tool thus allowed an empirical analysis of software development across versions from the perspective of those transformations. We describe results for the systems analysed and point to key conclusions from our analysis. In particular, we investigate a problematic empirical question as to whether certain refactorings are related, i.e., they cannot be undertaken in isolation without other refactorings being undertaken in parallel. In this context, we focus specifically on the four most common refactorings identified by the tool from three of the open-source systems and use a dependency graph to inform conclusions about the empirical data extracted by the tool. An interesting result relating to some common refactorings is described.