[论文解读] Exploring software developers work practices

Sherlock A. Licorish, Stephen G. MacDonell|arXiv (Cornell University)|Apr 1, 2017

Software Engineering Techniques and Practices参考文献 137被引用 18

一句话总结

本研究基于三年间来自474名IBM Jazz开发者的30,646项工作项，分析了多样化软件开发任务（尤其是缺陷修复）对开发者参与度及任务解决速度的影响。研究发现，行为、内在及外在因素共同预测了任务解决速度，且工作性质显著影响参与度与绩效表现。

ABSTRACT

We studied IBM Jazz practitioners interactions and task performance.This work mined logs from 30,646 work items (WIs) created since 3 years by 474 practitioners.Software practitioners studied were mostly involved in fixing defects.The nature of the work being performed influenced practitioners engagement.Combination of behavioral, intrinsic, and extrinsic factors affect task performance. In seeking to understand the processes enacted during software development, an increasing number of studies have mined software repositories. In particular, studies have endeavored to show how teams resolve software defects. Although much of this work has been useful, we contend that large-scale examinations across the range of activities that are commonly performed, beyond defect-related issues alone, would help us to more fully understand the reasons why defects occur as well as their consequences. More generally, these explorations would reveal how team processes occur during all software development efforts. We thus extend such studies by investigating how software practitioners work while undertaking the range of software tasks that are typically performed. Multiple forms of analyses of a longitudinal case study reveal that software practitioners were mostly involved in fixing defects, and that their engagement covaried depending on the nature of the work they were performing. Furthermore, multiple external factors affected speed of task resolution. Our outcomes suggest that behavioral and intrinsic issues may interact with extrinsic factors becoming significant predictors of the speed of software task resolution.

研究动机与目标

• 理解缺陷修复之外的软件开发活动全貌，包括其对开发者参与度与绩效的影响。
• 探究行为、内在与外在因素如何相互作用以影响任务解决速度。
• 通过考察多样化任务的纵向、多维度开发流程，拓展现有仓库挖掘研究。
• 揭示与任务解决效率及参与度水平相关的从业者行为模式。

提出的方法

• 对三年间来自IBM Jazz开发者的30,646项工作项（WIs）进行纵向分析。
• 挖掘软件仓库日志，提取任务类型、解决时间与交互模式。
• 应用多种分析方法——行为、内在与外在因素建模——以评估其对任务绩效的影响。
• 分析参与度水平与任务性质之间的协变关系，以识别行为与情境模式。
• 使用统计建模识别任务解决速度的关键预测因子。
• 将任务分类（如缺陷修复与功能开发）的定性洞察整合进绩效分析中。

实验结果

研究问题

• RQ1 软件开发任务的性质在多大程度上影响开发者的参与度与任务解决速度？
• RQ2 行为、内在与外在因素在多大程度上相互作用以预测任务解决速度？
• RQ3 与其它开发任务相比，与缺陷相关的任务在参与度与绩效方面有何差异？
• RQ4 从开发者交互的纵向分析中，工作项解决过程呈现出何种模式？

主要发现

• 开发者主要参与缺陷修复，该类任务在整个研究期间占据主要工作负荷。
• 参与度水平因任务性质而异，复杂或高影响性任务表现出更高的参与度。
• 行为、内在与外在因素的综合影响被证实是任务解决速度的重要预测因子。
• 外部因素（如团队背景与工具环境）显著影响解决速度，尤其在与行为特征结合时更为明显。
• 缺陷修复并非始终迅速完成；解决速度存在显著差异，表明存在未建模的影响因素，超越任务类型本身。
• 内在动机与行为模式与外部约束相互作用，共同塑造绩效结果。