paperbot · PL 论文追踪

全部论文 待读 收藏

P2 COBook: An AI-Assisted Interactive COBOL Notebook with Visualization

COBook:一个AI辅助的交互式COBOL笔记本环境与可视化
FSE Proceedings of the 34th ACM International Conference on the Foundations of Software Engineering2026匹配度 2/10 · 论文聚焦 COBOL 笔记本环境与教学可视化/AI辅助,属于开发者工具与教育 HCI,未涉及静态分析、Agentic/LLM 程序分析、浮点或贝叶斯程序分析等核心方向;仅“机器学习与程序语言”标签有微弱关联。引用 0 inbox
Raghavendra Pappu, Satish Pati, Sridhar Chimalakonda
crossref: 2026-09-12dblp: 2026-09-12

AI 速览

一句话:COBook 是首个面向 COBOL 的笔记本环境,集成可视化与 AI 辅助,显著降低新手认知负荷。

COBOL 支撑着全球约 70% 的商业交易,但经验丰富的开发者逐渐退休、高校课程又普遍移除该语言,且传统工具仍停留在 1980 年代的工作流,难以吸引新一代开发者。为此,作者开发了 COBook——一个将现代交互式工具引入 COBOL 学习与开发的笔记本环境,也是最早在笔记本中为 COBOL 集成此类能力的工具之一。COBook 通过流程图、数据流图和内存布局等可视化呈现程序结构,帮助学习者建立心智模型;其 AI 辅助功能提供自然语言解释、错误建议和代码生成,以支持不熟悉 COBOL 语法的学习者。在 33 名新手程序员、5 个 COBOL 理解任务的初步评估中,相比传统基于文本的 COBOL IDE,COBook 将心理需求降低 53%、体力需求降低 42%、时间需求降低 38%、挫败感降低 59%,同时将感知绩效提升 91%、逻辑复杂度处理简化 53%。

问题
COBOL 生态面临技能断层,传统工具与教学方式陈旧,缺乏面向新手的现代交互式开发与学习环境。
方法
构建笔记本式 COBOL 开发环境 COBook,提供流程图、数据流图与内存布局等可视化,并集成自然语言解释、错误建议与代码生成等 AI 辅助功能。
评估
对 33 名新手程序员开展 5 项 COBOL 理解任务的初步用户研究,对比传统文本式 COBOL IDE,测量认知负荷与感知绩效。
原文摘要(Abstract)

COBOL forms the backbone of the global economy, with more than 220 billion lines of code supporting nearly 70% of all business transactions worldwide. But the ecosystem faces a serious skills crisis: a large proportion of experienced COBOL programmers are nearing retirement, while universities have largely removed the language from their curricula. At the same time, traditional COBOL tools rooted in 1980s-era workflows fail to appeal to a new generation of developers raised on modern, interactive environments. To address this gap, we developed COBook, a notebook-based environment that brings modern interactive tooling to COBOL development and learning. COBook is one of the first tools to integrate these capabilities for COBOL in a notebook environment. COBook makes the program structure visible through visualizations, including flowcharts, data flow diagrams, and memory layouts, which may help learners build mental models of COBOL programs. Its AI-Assisted features include plain-language explanations, error suggestions, and code generation to support learners working with unfamiliar COBOL syntax. In a preliminary evaluation with 33 novice programmers across five COBOL comprehension tasks, COBook reduced mental demand by 53%, physical demand by 42%, temporal demand by 38%, and frustration by 59%, while improving perceived performance by 91% and simplifying the handling of logic complexity by 53% compared to traditional text-based COBOL IDE programming. By bridging the gap between legacy systems and modern tooling, COBook aims to make COBOL more accessible and engaging for today's programmers. The tool demo video is available at https://youtu.be/kjdzJsCq_Lc, the tool website and the artifacts are at https://github.com/rishalab/COBook.

链接与引用

DOI 原文 · PDF(开放获取) · DBLP

BibTeX
@inproceedings{PappuPC26,
  title = {COBook: An AI-Assisted Interactive COBOL Notebook with Visualization},
  author = {Raghavendra Pappu and Satish Pati and Sridhar Chimalakonda},
  booktitle = {FSE},
  year = {2026},
  doi = {10.1145/3803437.3806430}
}