DevOps Mobile Application Development

The main objective of this research is to classify the activities developers perform in real projects and map them to the phases of the DevOps cycle.

We analyzed commit messages and changes made during the development of Android applications within a dataset of 7,441 Android repositories, with 1,983,967 total commits. First, we performed a qualitative manual analysis of 2,000 sampled revisions to identify the activities of Android mobile application developers and correlate them with DevOps practices. The obtained data were then used to train an ML algorithm and extend the classification to the whole dataset. In addition, metadata extracted from the repositories was used to refine the results.

Finally, we categorized mobile development activities based on DevOps practices to understand how those practices are applied and potential open directions for practitioners and researchers.

本研究的核心目标是对开发者在实际项目中开展的各类活动进行分类，并将其映射至DevOps周期（DevOps cycle）的各个阶段。
我们基于包含7441个安卓（Android）代码仓库、总计1983967次提交的数据集，分析了安卓应用开发过程中的提交信息（commit messages）与代码变更。首先，我们对2000份抽样修订版本开展定性人工分析，以识别安卓移动应用开发者的活动类型，并将其与DevOps实践建立关联。随后，将获取的数据用于训练机器学习（ML）算法，进而将分类任务拓展至整个数据集。此外，我们还从代码仓库中提取元数据，以优化分析结果。
最终，我们基于DevOps实践对移动开发活动进行了分类，以期明晰此类实践的落地现状，并为从业者与研究者提供潜在的研究方向。