A Survey of Software Code Review Practices in Brazil

Context: Software code review aims to early find code anomalies and to perform code improvements when they are less expensive. However, issues and challenges faced by developers who do not apply code review practices regularly are unclear. Goal: Investigate difficulties developers face to apply code review practices without limiting the target audience to developers who already use this practice regularly. Method: We conducted a web-based survey with 350 Brazilian practitioners engaged on the software development industry. Results: Code review practices are widespread among Brazilian practitioners who recognize its importance. However, there is no routine for applying these practices. In addition, they report difficulties to fit static analysis tools in the software development process. One possible reason recognized by practitioners is that most of these tools use a single metric threshold, which might be not adequate to evaluate all system classes. Conclusion:
Improving guidelines to fit code review practices into the software development process could help to make them widely used. Additionally, future studies should investigate whether multiple metric thresholds that take source code context into account reduce static analysis tool false alarms. Finally, these tools should allow their use in distinct phases of the software development process.

研究背景：软件代码评审（Software Code Review）旨在尽早发现代码异常，并在修改成本较低时开展代码优化工作。然而，尚未明确未定期采用代码评审实践的开发者所面临的问题与挑战。研究目标：探究开发者在应用代码评审实践过程中遇到的困难，且不将研究目标群体限定为已定期使用该实践的开发者。研究方法：我们针对350名投身软件开发行业的巴西从业者开展了网络问卷调查。研究结果：代码评审实践在巴西从业者中已得到广泛应用，从业者普遍认可其重要性，但尚未形成统一的应用流程。此外，从业者报告称难以将静态分析工具（static analysis tools）集成至软件开发流程中。从业者指出其中一个潜在原因是，多数此类工具仅采用单一指标阈值，这可能不足以对所有系统代码类进行评估。研究结论：优化适配软件开发流程的代码评审实践指南，有助于推动该实践的广泛应用。此外，未来研究可探讨结合源代码上下文的多指标阈值是否能够降低静态分析工具的误报率。最后，此类工具应支持在软件开发流程的不同阶段使用。