val_highlight 是一个纯 Dart 包,支持 55 种语言的语法高亮,具备增量更新、可自定义主题以及导入 VS Code 主题的功能。它生成 HTML 输出,并可在 Android、iOS、Linux、macOS、Web 和 Windows 等所有主要平台上无缝运行。非常适合用于构建需要实时语言特定高亮且无需外部依赖的代码编辑器或文档工具的开发者。
适合场景
代码编辑器实现
文档生成
实时语法预览
教育类编程应用
带有高亮代码的静态站点生成
跨平台代码显示
主要特点
支持 55 种编程语言
增量高亮更新
VS Code 主题导入功能
HTML 输出生成
纯 Dart 实现
多平台兼容性
适合对象
开发代码编辑器的开发者
需要动态语法高亮的应用程序
带有代码片段的 Web 和移动应用
文档和教程平台
需要轻量级语法高亮的项目
使用建议
可能需要针对非常大的代码文件进行优化
主题自定义需要了解 CSS/JSON 格式
限于预定义的语言语法
无内置行号
性能取决于更新粒度
原始 README
以下为关联源码仓库的 README 原文快照,最新内容请访问源码仓库。
val_highlight
Syntax highlighting for Dart and Flutter.
A pure-Dart engine with 55 languages, incremental updates and themes,
plus Flutter widgets to show and edit code.
The engine, in pure Dart: 55 languages plus a generic fallback, flat tokens, incremental updates, language detection, 8 themes, VS Code theme import and HTML output. Runs on the VM, in Flutter and on the web.
Flutter widgets: CodeView to show code, CodeField to edit it, HighlightTextController, and CodeTheme for app-wide defaults.
A val_highlight_jaspr package is planned, built on the same engine.
// FlutterCodeView(source, language: dartLanguage)
// Pure Dartfinal html = constHighlighter().toHtml(source, language: dartLanguage);
CodeView Line numbers, highlighted lines, selection.
Themes Eight built in, or import a VS Code theme.
Languages inside languages Markdown fences, HTML with CSS and JavaScript.
CodeField An editor that highlights as you type.
⚡ Performance
just bench compiles to native code, highlights a Dart file and renders it
to HTML (best of 5 runs, on a MacBook):
Lines
Highlight + HTML
600
1.5 ms
2,400
4.3 ms
9,600
13.9 ms
just bench-languages measures about 256 KB of realistic source per
language on native code. It averages about 43 MB/s, from 30 MB/s (HTML) to
64 MB/s (Markdown). just bench-js runs the same test compiled to
JavaScript, where it averages about 62 MB/s.
With incremental highlighting, one keystroke in the 9,600-line file takes
about 0.4 ms. Frame timings for scrolling and typing in Flutter are in the
val_highlight_flutter README
(just bench-frames).
How it works
One pass, no copying. The scanner walks the source by offset, never
copies it, and time grows in step with input size.
First-character dispatch. Each rule's pattern is analysed once to find
the characters it can start with. The scanner skips positions where no
rule can start, and tries only the rules that can. Rules anchored with ^
are tried only at line starts. States with a rule that can start almost
anywhere fall back to one combined regular expression. Both paths give
identical results, and fuzz tests check this.
Plain-Dart matchers. On native platforms, most patterns run as small
Dart matchers instead of the regular expression engine, with identical
results (checked by fuzz tests). Browsers compile regular expressions to
machine code, so the web keeps using them.
Flat output. Tokens are two integer lists, not a tree of objects.
🛠️ Development
This repository is a pub workspace.
The Flutter version is pinned in .fvmrc (3.44.2), and the justfile runs
everything through FVM:
just setup # install the pinned SDK and fetch dependencies
just check # format check, analyze, all tests (VM and JS)
just devices # list available devices
just example-flutter # run the Flutter example; asks which device to use
just example-web # run the Flutter example in Chrome
just example-dart # run the pure-Dart example
just bench # benchmark
Run just with no arguments to list every command.
Note: with Xcode 27, Flutter 3.44.2 (the pinned version) fails to build
for iOS, and to build macOS apps in profile or release mode, such as the
frame benchmark. Its architecture check rejects Xcode 27's lipo output.
Flutter 3.47 builds both. Web, Android and macOS debug builds work with the
pin.
Built by Val
val_highlight is part of Val, the live visual layer
for AI agents.
💬 Community
Issues and pull requests are welcome. See CONTRIBUTING.md
for setup, tests and what we look for in a pull request.