Splitting APIs: An exploratory study of software unbundling

Abstract

Software unbundling consists of dividing an existing software artifact into smaller ones. Unbundling can be useful for removing clutter from an application or separating different features that may not share the same purpose, or simply for isolating an emergent functionality that merits to be an application on its own. This phenomenon is frequent with mobile apps and it is also propagating to APIs. This research proposes a first empirical study on unbundling to understand its effects on popular APIs. The study explores the possibilities of splitting libraries into 2 or more bundles based on the use that their client projects make of them. I mine over than 71,000 client projects of 10 open source APIs and automatically generate 2,090 sub-APIs to then study their properties. Results show that it is possible to have sets of different ways of using a given API and to unbundle it accordingly; the bundles can vary their representativeness and uniqueness, which is analyzed thoroughly in this study.