Abstract
Interacting with businesses, searching for information, or accessing news and entertainment sources all have a common feature, they are predominately accessed nowadays from mobile applications. The software architecture used in building those kinds of products represents a major factor in their lifecycle, costs, and roadmap, as it affects their maintainability and extensibility. In this study, our novel approach designed for detecting MVC architectural layers from mobile codebases (that use SDK for building their UI interrfaces) is validated and analysed from various perspectives (Artificial Intelligence, architectural rules, empiric evaluation). Our proposal is validated on eight different-sized iOS codebases corresponding to different mobile applications that have different scopes (both open and closed source). The performance of the detection quality is measured by the accuracy of the system, as we compared to a manually constructed ground truth, achieving an average accuracy of 85% on all the analyzed codebases. Our hybrid approach for detecting architectural layers achieves good results by combining the accuracy of the deterministic methods with the flexibility for being used on other architectural patterns and platforms via the non-deterministic step. We also validate the workflow of the proposal from an empirical point of view through an interview with two mobile application developers.KeywordsMobile applications software architecture analyserAutomatic analysis of software architecturesStructural and lexical informationSoftware clusteringHybrid approach.
Citare
@Inproceedings{Dobrean2021ValidatingHI,
author = {D. Dobrean and L. Dioşan},
booktitle = {International Conference on Evaluation of Novel Approaches to Software Engineering},
title = {Validating HyDe: Intelligent Method for Inferring Software Architectures from Mobile Codebase},
year = {2021}
}