About QCSunny Lab
I’m QCSunny, a software engineer focused on modern web technology, algorithm design, and developer-productivity tools. QCSunny Lab is where I hand-build browser-native utilities and then write up how they actually behave — one address for each tool and the engineering notes behind it.
What This Site Actually Is
A dependency-free static site holding 135 in-browser tools and76 engineering posts. The tools run on hand-written, vanilla-TypeScript code — no runtime framework — and the engines behind them (the calculator parser, the SQL tokenizer, Reed–Solomon error correction) are first-party rather than imported. Each post is the engineering record of that work.
Data you type into a tool stays on the page: there is no backend to receive it, by construction. The whole source is public and rebuilt from CI on every push.
Built from Scratch, Then Measured
- Hand-written engines: the QR encoder is ~400 lines of self-authored Reed–Solomon; Markdown parsing, the SQL tokenizer, the mathematical expression engine, ODE solvers, and matrix algorithms are first-party code, not library wrappers.
- Measured, not assumed: posts cite numbers taken from the running site — a 3-D surface plot spends 18 ms of a 181 ms frame on rasterization; this homepage ships 50 KB, 58% of it fonts; KaTeX ships as stylesheet only, with no CDN script left to break.
- Gated by tests: every deploy waits on a browser E2E suite of 150+ cases, which is what catches a dead calculator button or a two-language label leak before it ships.
Design Decisions, With Reasons
Tools run entirely inside the page, so input never reaches a server — nothing to configure, nothing to leak, and it works offline once loaded.
Hand-written engines over imported ones, and the trade-offs are measured and written up rather than assumed.
A browser-test suite must pass before any build reaches the live edge; a red build blocks the deploy.
Every page is written in English and Simplified Chinese together, never translated after the fact; one attribute flips the whole site, tool errors and chart legends included.
Inside the Notebook
- QR Reed–Solomon error correction, ~400 lines, dependency-free
- Writing a Markdown parser, and why you should not write a LaTeX renderer
- Why modern database primary keys are moving to UUID v7
- The byte ledger of this very site
- What a 3% “handling fee” loan really costs — the IRR math
All 76 posts live under /blog/.
Open Source & Contact
The source code for this site is open and hosted on GitHub. Bug reports, feature requests, and suggestions are always welcome via GitHub Issues.
For general inquiries, collaboration, or feedback, you can reach me directly by email at: admin@qcsunny.org.
关于本站 (QCSunny Lab)
我是 QCSunny，一名专注于现代 Web 技术、算法设计与开发者效率工具的 软件工程师。这个站是我亲手实现浏览器端工具、再把它们的真实行为写出来的地方——每个工具和它背后的工程 手记，在同一个地址下。
这个站到底是什么
一个零依赖的静态站，放着 135 款纯浏览器工具与 76 篇工程手记。 工具都是为本站手写的原生 TypeScript——不依赖运行时框架；背后的引擎（计算器解析器、SQL 分词器、 Reed–Solomon 纠错）全是第一方实现，不是套壳。每篇手记就是这些工作的工程档案。
你在任何工具里输入的数据都只留在当前页面——结构上就没有后端去接收它。全部源码公开，每次推送由 CI 重新构建发布。
从零手写，再拿数据验证
- 引擎都是自己写的：二维码编码器约 400 行自写 Reed–Solomon；Markdown 解析、SQL 分词器、 数学表达式引擎、微分方程与矩阵算法都是第一方代码，不是库的壳。
- 先测量再下结论：手记引用的是跑在真站上的数字——三维曲面一帧 181ms 里只有 18ms 在光栅化；这份首页 50KB 里 58% 是字体；KaTeX 只发样式表，没有会失效的 CDN 脚本。
- 用测试守门：每次部署前要跑 150+ 条浏览器 E2E，计算器按键静默失效、双语标签泄漏 这类回归，就是靠它在上线前拦下来的。
每条设计决策都有理由
所有工具都在页面内运行，输入根本不会到服务器——没什么可配置，也没什么可泄漏，载入后断网可用。
能自己写的就自己写，代价用测量和文章讲清楚，而不是靠假设。
浏览器测试全绿才允许部署，出现红就直接阻断发布。
每页同时用中文与英文创作，而非发布后翻译；单一属性切换全站，计算器报错与图表图例一起切。
手记里都有什么
- 二维码 Reed–Solomon 纠错：约 400 行、零依赖手写
- 手写 Markdown 解析器，与一个不该自己写的 LaTeX 排版器
- 为什么现代数据库主键推荐 UUID v7
- 这份站自己的字节账：首屏成本实测
- 3% 手续费的分期，真实年化（IRR）有多高
全部 76 篇都在 博客。
开源与联系
本站源码完全公开开源，托管于 GitHub。若在使用过程中发现任何问题或有新工具需求，非常欢迎在 GitHub Issues 中提交反馈。
商务合作、技术交流或建议咨询，可通过电子邮箱直接与我联系：admin@qcsunny.org。