SERG talk: On the Relation of Method Popularity to Breaking Changes in the Maven Ecosystem

Abstract Software reuse is a common practice in modern software engineering, helping developers save time and energy while accelerating the process of delivering software products. Dependency managers like Maven offer a large ecosystem of reusable libraries that build the backbone of software reuse. Breaking changes, however, are troublesome barriers to this library reuse, i.e., when an update to a library introduces incompatible changes that break existing client programs. Semantic Versioning has been proposed as a practice to make it easier for users to find safe updates by encoding the change impact in the version number. While this practice is widely studied from the framework perspective, no detailed insights exist yet into the ecosystem perspective. In this work, we inspect the commonness of the method-level semantic versioning violations that occur within the Maven ecosystem. Then we study the popularity of methods to understand the extent of the effects these violations have on the ecosystem. We study violations of semantic versioning identified for the 2,766 versions of 384 target versioned packages in a release time frame of 6 months. We also investigate the method-level effects of breaking changes on 7,190 dependent versioned packages. We found that 28% of the artifacts introduce at least one type of semantic versioning violation, either a breaking change or an illegal API extension. To further understand the effects of breaking changes on the ecosystem we study their impacts. After building call graphs for all dependents that reference the target versioned package (directly or transitively), we found that 87% of the publicly defined methods are never used by the dependents. Moreover, among methods with at least one usage, a minority of 35% receive half of all unique dependent calls to the respective library. We also find that the method’s popularity does not seem to affect stability, as popular methods also break frequently. Our results suggest that breaking changes might be a sign of a lack of change-impact awareness on the ecosystem. Overall, we confirm the previous result that Semantic Versioning is violated repeatedly in practice. We believe that adequate information about popular framework methods could improve update strategies and make developers reconsider if breaking changes to widely used methods are indeed required.