#font #shape #glyph #text


Font introspection, complex text shaping and glyph rendering

10 releases

0.1.12 Feb 1, 2024
0.1.10 Jan 31, 2024
0.1.8 May 6, 2023
0.1.6 Dec 9, 2022
0.1.4 Jul 29, 2021

#52 in Text processing

Download history 5667/week @ 2023-11-01 5911/week @ 2023-11-08 4635/week @ 2023-11-15 5110/week @ 2023-11-22 5306/week @ 2023-11-29 5236/week @ 2023-12-06 5753/week @ 2023-12-13 4720/week @ 2023-12-20 4731/week @ 2023-12-27 5259/week @ 2024-01-03 7083/week @ 2024-01-10 10802/week @ 2024-01-17 7731/week @ 2024-01-24 9610/week @ 2024-01-31 10178/week @ 2024-02-07 9430/week @ 2024-02-14

38,292 downloads per month
Used in 79 crates (9 directly)


29K SLoC


Swash is a pure Rust, cross-platform crate that provides font introspection, complex text shaping and glyph rendering.

Crates.io Docs.rs MIT licensed Apache licensed


This crate aims to lay the foundation for a cross-platform, high performance set of components for beautiful typography. In particular, this library focuses on fonts and operations that are directly applicable to them. For the most part, the desire is to be unopinionated with respect to resource management, higher level layout and lower level rendering.

Non goals

Due to the intention of being generally useful and easy to integrate, the following areas of related interest are specifically avoided:

  • Text layout. This is highly application specific and the requirements for both features and performance differ greatly among web browsers, word processors, text editors, game engines, etc. There is a sibling crate in development that does provide general purpose text layout based on this library.

  • Composition. Like layout, this is also application specific in addition to being hardware dependent. Glyph caching, geometry batching and rendering all belong here and should integrate well with the application and the hardware environment.

General features

  • Simple borrowed font representation that imposes no requirements on resource management leading to...
  • Thread friendly architecture. Acceleration structures are completely separate from font data and can be retained per thread, thrown away and rematerialized at any time
  • Zero transient heap allocations. All scratch buffers and caches are maintained by contexts. Resources belonging to evicted cache entries are immediately reused


  • Enumerating font collections (ttc/otc)
  • Localized strings including names and other metadata
  • Font variation axes and named instances
  • Comprehensive font and per-glyph metrics (with synthesized vertical metrics if not provided)
  • Primary attributes (stretch, weight, style) with support for synthesis suggestions based on available variations and scaler transforms (faux bold and oblique)
  • Color palettes
  • Embedded color and alpha bitmap strikes
  • Character to nominal glyph identifier mapping with support for enumerating all pairs
  • Writing systems: provides a list of all supported script/language pairs and their associated typographic features
  • All introspection is zero allocation and zero copy

Complex text shaping

  • Full support for OpenType advanced typography (GSUB/GPOS)
  • Partial support for Apple advanced typography: glyph metamorphosis (morx) is fully supported while the current extended kerning support (kerx) covers most common cases (kerning and mark positioning)
  • Full support for variable fonts including positioning and feature substitutions
  • Implementation of the Universal Shaping Engine for complex scripts such as Devanagari, Malayalam, etc.
  • Arabic joining including Urdu style climbing runs
  • Basic shaping support: ligatures, marks, kerning, etc.
  • Enable/disable individual features with argument support for activating alternates such as swashes...
  • Pre-shaping cluster parsing with an iterative mapping technique (including normalization) allowing for sophisticated font fallback mechanisms without the expense of heuristically shaping runs
  • Shaper output is structured by cluster including original source ranges and provides simple identification of ligatures and complex multi-glyph clusters
  • Pass-through per character user data (a single u32) for accurate association of style properties with glyphs
  • Pass-through per cluster information for retaining text analysis results such as word and line boundaries, whitespace identification and emoji presentation modes


  • Scalable outlines with full variation support (TrueType and Postscript)
  • Asymmetric vertical hinting (TrueType and Postscript)
  • Horizontal subpixel rendering and fractional positioning
  • Full emoji support for Apple (sbix), Google (CBLC/CBDT) and Microsoft (COLR/CPAL) formats
  • Path effects (stroking and dashing)
  • Transforms including synthetic emboldening and affine transformations
  • Customizable glyph source prioritization (best fit color bitmap -> exact size alpha bitmap -> outline)

Text analysis

  • Unicode character properties related to layout and shaping
  • Character composition and decomposition (canonical and compatible)
  • Complex, script aware cluster segmentation
  • Single pass, iterator based analysis determines word and line boundaries and detects whether bidi resolution is necessary


Performance is a primary goal for this crate and preliminary microbenchmarks show a general improvement over FreeType and Harfbuzz by about 10-20% on average and some cases show substantial wins, particularly when scaling Postscript outlines or shaping text with complex features. Specifically, shaping runs with fonts like Calibri and Noto Sans Myanmar is almost twice as fast. Simple fonts and pure ASCII runs tend to show the smallest gains as those simply measure shaper initialization and glyph iteration. A comprehensive set of benchmarks (and test cases!) are needed here to gain more insight and track regressions.


Contributions are welcome and appreciated! All contributions must be granted under the licenses under which this project is released.


~43K SLoC