文件排序影响代码评审效果？基于33万PR的实证研究
原标题：Does Order Matter? An Empirical Investigation into the Impact of File Ordering on Code Review Effectiveness
AI 摘要
该研究基于182个GitHub项目、330,343个多文件PR（756,814个文件实例）的挖掘数据，实证考察文件在PR中的展示顺序与评审效果的关系。结果显示文件位置与潜在缺陷率存在统计显著但幅度较小的关联，位置越靠后的文件潜在缺陷率略高（从第1位的56.7%升至第30位的61.5%）；PR规模与缺陷风险呈非线性关系，约10个文件的中等规模PR风险最低；每增加一个修改文件，文件获得任何评审评论的几率下降约8.7%。作者据此指出评审注意力与效果之间存在缺口，认为按字母排序并非中立的界面默认设置，并提出上下文感知排序、依赖感知分组、风险感知优先级和逐文件覆盖率指示等工具设计方向。
正文节选
Does Order Matter? An Empirical Investigation into the Impact of File Ordering on Code Review Effectiveness Abstract. Modern code review is central to quality assurance in software development practice, yet its effectiveness depends not only on reviewer expertise and change characteristics, but also on how review tools present code changes. Most code review platforms display modified files alphabetically by default, although our prior work shows that developers often perceive this ordering as co