Stroke count, stroke order, and stroke-based sorting for Chinese characters in Typst. Wraps a Rust → WASM plugin with compile-time lookup tables for Unihan kTotalStrokes (~103,000 CJK characters) and cnchar stroke order (6,939 simplified characters).

English

Installation

#import "@preview/auto-bihua:0.1.0": *

bihua-count

Stroke count of one or more Chinese characters.

#bihua-count("汉字")                  // → 11
#bihua-count("汉字", split: true)     // → (5, 6)
#bihua-count("龘")                    // → 48
#bihua-count("Hello 世界", split: true)
// → (none, none, none, none, none, none, 5, 9)
#bihua-count("漢字")                  // → 20  (traditional supported)
#bihua-count("")                      // → 0

• chars accepts str or content (Typst text content with a .text field).
• Non-Chinese characters contribute 0 to the total, and none in split mode.
• Counts come from Unihan, so coverage extends to traditional characters and CJK extensions.

bihua-order

Stroke order of a single character. Four output styles:

#bihua-order("汉")                       // "44154"          (digit, default)
#bihua-order("汉", style: "name")        // ("点","点","提","横撇","捺")
#bihua-order("汉", style: "shape")       // ("丶","丶","㇀","㇇","㇏")
#bihua-order("汉", style: "letter")      // "kkiel"          (raw cnchar code)

Errors and edge cases:

• Passing a multi-character string panics. Iterate yourself if needed.
• Characters not in the stroke-order dictionary (traditional Hanzi, rare CJK extension chars, single-stroke radicals like 丿, 丶) return "" for string styles and () for array styles.

bihua-sort

Sort an array of strings (or contents) by stroke count, then by stroke order as tiebreaker. Comparison is character-by-character, mirroring the standard Chinese-publishing convention of “first by total strokes, then by stroke type.”

#bihua-sort(("汉","字","人","一","乙","二"))
// → ("一", "乙", "二", "人", "汉", "字")

#bihua-sort(("人民","汉字","一二","北京","上海","天津"))
// → ("一二", "人民", "上海", "天津", "北京", "汉字")

• Items with characters outside both data sets sort to the end.
• Tiebreaker uses the GB13000 digit sequence (not the raw cnchar letter encoding), so 提 sorts before 竖 as the standard prescribes.

Real-world recipes

Sort Chinese author names by surname strokes (common journal style):

#bihua-sort(("王伟","李娜","张敏","刘洋","陈静","杨磊","黄勇","赵丽","周强","吴芳"))
// → ("王伟","刘洋","李娜","杨磊","吴芳","张敏","陈静","周强","赵丽","黄勇")

Stroke order card for character study:

#let card(c) = block(
  stroke: 0.5pt + gray, inset: 8pt, radius: 4pt,
  [
    *#text(24pt, c)* #h(1em) #bihua-count(c) strokes \
    #bihua-order(c, style: "name")
      .enumerate()
      .map(((i, n)) => [#(i+1).#n])
      .join("  ")
  ]
)
#card("永")

Index a glossary by first-character stroke count:

#let terms = ("汉字","人民","一致","中国","北京","笔画")
#let groups = (:)
#for t in bihua-sort(terms) {
  let n = str(bihua-count(t.first()))
  groups.insert(n, groups.at(n, default: ()) + (t,))
}
#table(columns: 2,
  ..groups.pairs().map(((n, ts)) => (n + " 画", ts.join("、"))).flatten()
)

Data sources

• Stroke count: Unicode 17.0 Unihan_IRGSources.txt, field kTotalStrokes (~103k CJK code points, includes traditional and CJK extensions).
• Stroke order: cnchar package’s stroke-order-jian.json (6,939 simplified characters), MIT license.
• Stroke type encoding: cnchar’s 26-letter alphabet, projected to the GB13000 5-type system at lookup time.

Building from source

cd wasm
rustup target add wasm32-unknown-unknown
cargo build --release
cp target/wasm32-unknown-unknown/release/auto_bihua.wasm ../auto-bihua.wasm

Or with just: just build. The data files are processed by build.rs into static phf maps at compile time, so no runtime parsing.

Limitations

• Stroke order data covers only ~6,939 simplified characters. Traditional characters have stroke counts but no order; bihua-order returns empty.
• Single-stroke radicals (丿, 丶) and some CJK extension characters are absent from the stroke-order dictionary even though their stroke count is known.

License

MIT. Data licenses: Unihan (Unicode terms), cnchar (MIT).

中文

安装

#import "@preview/auto-bihua:0.1.0": *

bihua-count — 笔画数

输入单字或多字，返回总笔画数（默认）或按字拆分的数组。

#bihua-count("汉字")                  // → 11
#bihua-count("汉字", split: true)     // → (5, 6)
#bihua-count("龘")                    // → 48
#bihua-count("Hello 世界", split: true)
// → (none, none, none, none, none, none, 5, 9)
#bihua-count("漢字")                  // → 20  （繁体也支持）
#bihua-count("")                      // → 0

• 参数 chars 接受字符串或带 .text 字段的 content。
• 非汉字字符在总和模式贡献 0，在 split 模式返回 none。
• 数据来源是 Unihan，覆盖繁体字和 CJK 扩展区。

bihua-order — 笔顺

输入单个汉字，返回其笔顺。四种输出风格：

#bihua-order("汉")                       // "44154"          （数字, 默认）
#bihua-order("汉", style: "name")        // ("点","点","提","横撇","捺")
#bihua-order("汉", style: "shape")       // ("丶","丶","㇀","㇇","㇏")
#bihua-order("汉", style: "letter")      // "kkiel"          （cnchar 原始字母编码）

异常与边界：

• 传入多字字符串会 panic，请自行逐字遍历。
• 不在笔顺词典里的字符（繁体字、CJK 扩展生僻字、丿/丶 等单笔部首）： 字符串风格返回 ""，数组风格返回 ()。

bihua-sort — 笔画排序

对字符串（或 content）数组按笔画数排序，同笔画数时按笔顺数字序为次要键。 逐字比较，与中文出版界"先比总画数，同画数比笔形"的惯例一致。

#bihua-sort(("汉","字","人","一","乙","二"))
// → ("一", "乙", "二", "人", "汉", "字")

#bihua-sort(("人民","汉字","一二","北京","上海","天津"))
// → ("一二", "人民", "上海", "天津", "北京", "汉字")

• 包含两份数据都不收录字符的项会排到末尾。
• 次要键使用 GB13000 五笔位数字序（而非 cnchar 字母字典序），所以 提 排 在 竖 之前，符合国标顺序。

实战用法

中文作者按姓氏笔画排序（中文期刊投稿常用规范）：

#bihua-sort(("王伟","李娜","张敏","刘洋","陈静","杨磊","黄勇","赵丽","周强","吴芳"))
// → ("王伟","刘洋","李娜","杨磊","吴芳","张敏","陈静","周强","赵丽","黄勇")

笔顺展示卡片（适合识字教材）：

#let card(c) = block(
  stroke: 0.5pt + gray, inset: 8pt, radius: 4pt,
  [
    *#text(24pt, c)* #h(1em) 共 #bihua-count(c) 画 \
    #bihua-order(c, style: "name")
      .enumerate()
      .map(((i, n)) => [#(i+1).#n])
      .join("  ")
  ]
)
#card("永")

词条按首字笔画分组建索引：

#let terms = ("汉字","人民","一致","中国","北京","笔画")
#let groups = (:)
#for t in bihua-sort(terms) {
  let n = str(bihua-count(t.first()))
  groups.insert(n, groups.at(n, default: ()) + (t,))
}
#table(columns: 2,
  ..groups.pairs().map(((n, ts)) => (n + " 画", ts.join("、"))).flatten()
)

数据来源

• 笔画数：Unicode 17.0 Unihan_IRGSources.txt，字段 kTotalStrokes （约 10.3 万 CJK 码位，含繁体与扩展区）。
• 笔顺：cnchar 项目的 stroke-order-jian.json（6939 简体字），MIT 协议。
• 笔形编码：cnchar 26 字母方案，查询时映射到 GB13000 五类。

从源码构建

cd wasm
rustup target add wasm32-unknown-unknown
cargo build --release
cp target/wasm32-unknown-unknown/release/auto_bihua.wasm ../auto-bihua.wasm

或用 just build。build.rs 在编译期把数据文件转成静态 phf 表，运行时 零解析开销。

已知限制

• 笔顺数据只覆盖约 6939 个简体字。繁体字有笔画数但无笔顺；bihua-order 返回空。
• 单笔部首（如 丿、丶）和部分 CJK 扩展区字符虽有笔画数，但不在笔顺 词典中。

许可

MIT。数据：Unihan（Unicode 协议）、cnchar（MIT）。