Modeling Framework API Evolution as a Multi-objective Optimization Problem

Abstract

Today's software development depends greatly on frameworks and libraries. When their APIs evolve, developers must update their programs accordingly. Existing approaches facilitate the upgrading process by generating change -- rules based on various input data, such call dependency, text similarity, software metrics, etc. However, existing approaches do not provide 100% precision and recall because of the limited set of input data that they use to generate change -- rules. For example, an approach only considering text similarity usually discovers less change -- rules then that considering both text similarity and call dependency with similar precision. But adding more input data may increase the complexity of the change -- rule generating algorithms and make them unpractical. We propse MOFAE (Multi-Objective Framework API Evolution) by modeling framework API evolution as multi-objective optimization problem to take more input data into account while generating change -- rules and to control the algorithmic complexity.