Software Engineering in the Age of App Stores: Feature-Based Analyses to Guide Mobile Software Engineers

Mobile app stores are becoming the dominating distribution platform of mobile applications. Due to their rapid growth, their impact on software engineering practices is not yet well understood. There has been no comprehensive study that explores the mobile app store ecosystem's effect on software engineering practices. Therefore, this thesis, as its first contribution, empirically studies the app store as a phenomenon from the developers' perspective to investigate the extent to which app stores affect software engineering tasks. The study highlights the importance of a mobile application's features as a deliverable unit from developers to users. The study uncovers the involvement of app stores in eliciting requirements, perfective maintenance and domain analysis in the form of discoverable features written in text form in descriptions and user reviews. Developers discover possible features to include by searching the app store. Developers, through interviews, revealed the cost of such tasks given a highly prolific user base, which major app stores exhibit. Therefore, the thesis, in its second contribution, uses techniques to extract features from unstructured natural language artefacts. This is motivated by the indication that developers monitor similar applications, in terms of provided features, to understand user expectations in a certain application domain. This thesis then devises a semantic-aware technique of mobile application representation using textual functionality descriptions. This representation is then shown to successfully cluster mobile applications to uncover a finer-grained and functionality-based grouping of mobile apps. The thesis, furthermore, provides a comparison of baseline techniques of feature extraction from textual artefacts based on three main criteria: silhouette width measure, human judgement and execution time. Finally, this thesis, in its final contribution shows that features do indeed migrate in the app store beyond category boundaries and discovers a set of migratory characteristics and their relationship to price, rating and popularity in the app stores studied.