cn-font-split

cn-font-split is a high-performance font subsetting tool used by the Chinese Web Font Project. It is designed to split massive font files into smaller, web-optimized chunks for efficient distribution.

Key features include:

• High Performance: Implemented via WebAssembly or Rust FFI, achieving sub-second build times (e.g., 2MB font in ~50ms).
• Cross-Platform: Runs in Browsers, WASI, Linux, macOS, and Windows. Compatible with Node.js, Deno, Bun, and Python.
• Broad Language Support: While optimized for Chinese, it intelligently subsets any characters present in the font, including Japanese, Korean, Arabic, and minority languages.
• Advanced Capabilities: Supports OTF formats, complex glyph rendering, variable fonts (automatic weight recognition), and font preview image generation.
• Consistency: Guarantees identical output across all platforms.

For high-performance needs (like landing pages), use "Minimalist Optimization". This mode scans your source files to collect only the characters actually used in your code, resulting in much smaller font files (approx. 10% of full size).

1. Configure scanning: Add scanFiles to the plugin options in your bundler config.
2. Import with query: Append ?subsets to your font import statement.

Scoped Subsets (using key)

You can define multiple subsets by providing an object to scanFiles. Use the key query parameter in your import to specify which subset to load.

Example: import { css } from 'font.ttf?subsets&key=subset-1'

// vite.config.js configuration for subsets
import Font from 'vite-plugin-font';

// Single scan scope
Font.vite({
    scanFiles: ['src/**/*.{vue,ts,tsx,js,jsx}'],
});

// Multiple scoped subsets
Font.vite({
    scanFiles: {
        default: ['src/**/*.{json,js,jsx,ts,tsx,vue}'],
        'subset-1': ['example/**/*.{json,js,jsx,ts,tsx,vue}'],
    },
});

// Usage in component
import { css } from '../../demo/public/SmileySans-Oblique.ttf?subsets&key=subset-1';

The cn-font-split system is organized into three distinct layers that separate core logic from communication and external interfaces:

1. Core Layer (core): Handles the actual font subsetting operations. It utilizes harfbuzz and opentype libraries for high-performance font manipulation. The core logic consists of:
   • pre_subset: Pre-processes font data and optimizes multi-language subsetting to determine characters.
   • run_subset: Manages large-scale batch subsetting operations.
   • link_subset: Links, integrates, and reports the subsetted font data to generate final results.
2. Message Channel Layer (message_channel): Acts as the bridge for external interaction using Protocol Buffers (protobuf). Definitions are managed via the cn-font-proto library. Data is exchanged either via generated protobuf structs (for Rust-to-Rust communication) or via binary data (for cross-language communication).
3. Wrapper Layer (wrapper): Provides the public interfaces for different deployment scenarios and programming languages.