Ein japanisches Fachbuch: Kana, Kanji und Latein

// ═══ Postext Cookbook · Nº 123 · A Japanese technical manual: kana, kanji and Latin ═══
// https://postext.dev/en/cookbook/japanese-technical-manual
// Code: MIT · Text: original (CC BY 4.0) · Pictures: drawn in code
// Fonts: Noto Serif JP, Noto Sans JP, BIZ UDGothic (SIL OFL 1.1) · Needs postext ≥ 1.28.0
import {
  buildDocument, withLoadedFonts, renderPageToCanvas, defaultResourceTypes, parseTSV,
  registerResourceImage, inlineSvgFonts,
} from 'https://esm.sh/postext';
import { renderToPdf, decompressWoff2 } from 'https://esm.sh/postext-pdf';

const LANG = 'en'; // @lang: the language of the frame; the chapter is Japanese in both
const RECIPE = 'japanese-technical-manual';

// ─── 1 · Design ─────────────────────────────────────────────────────────────
// #region palette: ink, one deep teal for numbers, rules and labels, a pale tint for code
const palette = {
  ink: '#1d2327', // text: a cool near-black
  teal: '#0e5a6e', // the one accent: chapter number, heads' numbers, labels, the point box
  tint: '#e7f0f2', // the opener band, the listing's ground
  rule: '#b9c6cc', // hairlines: table rules, the note rule
  muted: '#5b666d', // running heads, folios, colophon
  paper: '#ffffff',
};
const col = (id) => ({ hex: palette[id], model: 'hex', paletteId: id });
const colorPalette = Object.entries({ ...palette, 'main-color': palette.teal })
  .map(([id, hex]) => ({ id, name: id, value: { hex, model: 'hex' } }));
// #endregion
const [MINCHO, GOTHIC, CODE] = ['Noto Serif JP', 'Noto Sans JP', 'BIZ UDGothic'];
const [BODY, PITCH] = [9, 16]; // pt: 9 pt text on a 16 pt line, 1.78 × the size
const [CHARS, LINES] = [36, 29]; // the type area in characters: 36 to a line, 29 lines
const MEASURE = CHARS * BODY * 25.4 / 72; // mm: 114.3

// #region answer: Japanese rules from the tag, written out; the quarter-em Latin space
// locale 'ja' turns on JLReq composition: kinsoku at its strictest, full-width marks squeezed
// where two meet (」、 takes one em, not two), a closing mark keeps its half em at a line
// end, 、。 may hang past it, and a paragraph opening with 「 sets the bracket in the indent.
// The values below are the ones 'auto' picks for 'ja'; they are spelled out to be seen.
const cjk = {
  lineBreak: 'ja-very-strict', // no line starts with ー, small kana, 々, 」、。？・ or ：
  punctuationWidth: 'fullwidth', // 、。「」 keep their em inside the line (JLReq §3.1.2)
  hangingPunctuation: 'allow', // 、。 hang only when the line would otherwise break before them
  paragraphStartBracket: 'half', // 「 at a paragraph start sits in the indent's second half
  latinSpacing: em(0.25), // 四分アキ between kana or kanji and Latin letters or digits
  grid: { enabled: true, charsPerLine: CHARS, linesPerPage: LINES }, // whole ems, whole lines
};
// The Latin is the Japanese face's own proportional Latin, at the text's size: never
// full-width Ａ or a second face for the words in parentheses.
const bodyText = {
  fontFamily: MINCHO, fontSize: pt(BODY), lineHeight: pt(PITCH), color: col('ink'),
  boldColor: col('ink'), italicColor: col('ink'), referenceColor: col('ink'),
  textAlign: 'justify', firstLineIndent: em(1), indentAfterHeading: true, // 1 字下げ
  referenceBold: false, // 図3-1 and 表3-1 in the text weight: the face loads no bold
  hyphenation: { enabled: false }, avoidRunts: true, // no one-character last line
  // The grid cuts a list where the page ends: the four forms stay in one piece.
  shortListSplit: 'avoid', shortListMaxItems: 4,
};
// #endregion

// #region listing: a fenced block is a code listing in a tinted box
// A ``` fence is set line by line as written, in codeStyle's face: the monospaced face,
// which has kana and kanji too, so a full-width character takes two half-width cells.
// The rest of the fence line (normalize.py) is the listing's title. Inline `code` is set
// in the same face, in the accent.
const codeStyle = {
  fontFamily: CODE, fontSize: pt(8), lineHeight: pt(12), color: col('ink'),
  background: col('tint'), snapToGrid: false, highlight: 'none', // one ink, as in print
  padding: { top: mm(2.5), right: mm(4), bottom: mm(3), left: mm(4) },
  marginTop: mm(2), marginBottom: mm(2),
  titleStyle: { fontFamily: CODE, fontSize: pt(7), fontWeight: 400, color: col('teal') },
  inline: { fontSize: pt(8.5), color: col('teal') },
};
// #endregion

// #region opener: a tinted band, the chapter number large in the accent, the title under it
const BAND = 74; // mm from the trim's top
const opener = { enabled: true, minHeight: pt(PITCH * 10), slot: { elements: [ // text: line 11
  { kind: 'box', id: 'band', reserve: false, style: { backgroundColor: col('tint') },
    placement: { anchor: { to: 'page', edge: 'top-left' },
      size: { width: 'fill', height: mm(BAND) } } },
  { kind: 'text', id: 'kicker', content: 'CHAPTER', fontFamily: GOTHIC, fontSize: pt(8),
    fontWeight: 700, letterSpacing: pt(1.6), color: col('teal'), align: 'left',
    placement: { anchor: { to: 'container', edge: 'top-left' }, offset: { y: mm(0) } } },
  { kind: 'text', id: 'number', content: '{numberDecimal}', fontFamily: GOTHIC, fontSize: pt(64),
    lineHeight: 1, fontWeight: 700, color: col('teal'), align: 'left',
    placement: { anchor: { to: '#kicker', edge: 'below' }, offset: { y: mm(1) } } },
  { kind: 'text', id: 'title', content: '{titleText}', fontFamily: GOTHIC,
    fontSize: pt(20), fontWeight: 700, color: col('ink'), align: 'left', overflow: 'wrap',
    placement: { anchor: { to: '#number', edge: 'below' }, offset: { y: mm(5) },
      size: { width: mm(MEASURE) } } },
] } };
// #endregion

// Running heads: the book on the verso, the chapter on the recto, folios outside.
// Anchored to the header's container, which spans the measure, so they align with the text
// wherever the grid puts its margins.
const head = (id, content, parity, edge, x, extra) => ({ kind: 'text', id, content, parity,
  pages: 'body', fontFamily: GOTHIC, fontSize: pt(7.5), color: col('muted'),
  align: edge.endsWith('left') ? 'left' : 'right',
  placement: { anchor: { to: 'container', edge }, offset: { x: mm(x), y: mm(13) } }, ...extra });
const folio = { fontWeight: 700, color: col('teal') };
const header = { elements: [
  head('v-folio', '{pageNumber}', 'even', 'top-left', 0, folio),
  head('v-title', '{title}', 'even', 'top-left', 8),
  head('r-title', '{chapterNumber}　{chapterTitle}', 'odd', 'top-right', -8),
  head('r-folio', '{pageNumber}', 'odd', 'top-right', 0, folio),
] };

const config = () => ({
  locale: 'ja', // written out, never LANG (gotcha: ja-locale-tag)
  resourceTypes: defaultResourceTypes('ja'), // 図 and 表, numbered by chapter: 図3-1
  colorPalette,
  page: { sizePreset: 'custom', width: mm(148), height: mm(210), dpi: 150, // A5
    // Minimums: the grid grows them to centre its 36 × 29 area, the head deeper than the foot.
    margins: { top: mm(23), bottom: mm(19), left: mm(17), right: mm(14), mirror: true } },
  layout: { layoutType: 'single' },
  cjk,
  bodyText,
  headings: { fontFamily: GOTHIC, fontWeight: 700, color: col('ink'),
    balancing: { enabled: false }, // no lines added above heads: the grid holds
    levels: [
    // parity 'odd': the chapter opens on a recto, after a blank verso only when one is needed.
    { level: 1, numberingTemplate: '第{1}章', breakBefore: { enabled: true, parity: 'odd' },
      advancedDesign: opener },
    // 3行取り: each section head takes three body lines, so the grid holds across it.
    { level: 2, numberingTemplate: '{1}.{2}', numberSeparator: '　', fontSize: pt(11),
      lineSpan: 3 },
  ] },
  // #region notes: only their size and colour; placement and numbering are the 'ja' defaults
  footnotes: { fontSize: pt(7.5), lineHeight: pt(12), color: col('ink'),
    separator: { color: col('rule') } }, // at the column foot, 1 on each page, superscript
  // #endregion
  unorderedLists: { bulletChar: '・', color: col('teal'), fontWeight: 400,
    marginTop: pt(0), marginBottom: pt(0) },
  codeStyle,
  calloutStyles: [
    { id: 'point', backgroundEnabled: false,
      stripe: { enabled: true, side: 'left', width: pt(3), color: col('teal') },
      padding: { top: mm(0), right: mm(0), bottom: mm(0), left: mm(5) },
      marginTop: pt(PITCH), marginBottom: pt(PITCH),
      titleStyle: { fontFamily: GOTHIC, fontSize: pt(8), fontWeight: 700, color: col('teal') },
      body: { fontFamily: GOTHIC, fontSize: pt(BODY), lineHeight: pt(PITCH), color: col('ink'),
        firstLineIndent: pt(0), textAlign: 'justify' } },
  ],
  tableStyle: { rules: 'horizontal', borderColor: col('rule'), borderWidth: pt(0.5),
    headerBackground: col('teal'), headerColor: col('paper'), headerFontFamily: GOTHIC,
    bodyFontFamily: MINCHO, bodyFontSize: pt(8), bodyColor: col('ink'), cellPadding: mm(1.2) },
  captionStyle: { fontFamily: GOTHIC, fontSize: pt(8), color: col('ink'), labelBold: true,
    labelColor: col('teal') }, // 図3-1　…, the ja default
  paragraphStyles: [
    { id: 'lead', fontFamily: GOTHIC, fontSize: pt(BODY), lineHeight: pt(PITCH),
      color: col('ink'), firstLineIndent: pt(0), marginBottom: pt(PITCH) },
    { id: 'colophon', fontFamily: GOTHIC, fontSize: pt(6.5), lineHeight: pt(9),
      color: col('muted'), firstLineIndent: pt(0), textAlign: 'left', marginTop: pt(PITCH) },
  ],
  header,
  footer: { elements: [{ kind: 'text', id: 'drop-folio', content: '{pageNumber}',
    pages: 'opener', fontFamily: GOTHIC, fontSize: pt(7.5), fontWeight: 700, color: col('teal'),
    align: 'center', placement: { anchor: { to: 'container', edge: 'bottom' },
      offset: { y: mm(-10) } } }] }, // a drop folio on the opener
});

// ─── 2 · Content ────────────────────────────────────────────────────────────
const markdown = `---
Markdown-Beispiel · 69 Zeilen · content.en.md
title: "実践 日本語テキスト処理"
---

# 文字列の正規化

:::paragraphs{style="lead"}
画面では同じに見える二つの文字列が、プログラムの中では別物として扱われることがある。検索に引っかからない、重複したはずのデータが二件残る、ファイル名で並べると順序が崩れる。こうした不具合の多くは、Unicodeの正規化を知っていれば防げる。
:::

本章では、まず同じ文字に二つの表し方がある理由を見て、Unicodeが定める四つの正規化形式を整理する。次にPythonの標準ライブラリで実際に変換し、最後に、検索キーを作るときに正規化で失われる情報について述べる。

## 同じに見えて違う文字列

「が」という文字は、Unicodeでは二通りに表せる。一つは「が」そのものに割り当てられた符号位置U+304Cを使う方法、もう一つは「か」（U+304B）の後ろに結合用の濁点（U+3099）を置く方法である。前者を合成済み文字、後者を結合文字列と呼ぶ。どちらも画面には同じ「が」として表示されるが、符号位置の並びが違うので、単純な比較では等しくならない。UTF-8で書き出すと、前者は3バイト、後者は6バイトになる（[@fig:bytes]）。

::resource{id="fig:bytes"}

この違いが表に出やすいのはファイル名である。macOSの以前のファイルシステムHFS+は、ファイル名を分解した形で保存していた[^hfs]。Macで作った「データ.csv」をWindowsやLinuxのサーバーにコピーすると、「テ」と濁点が分かれた名前のまま届く。見た目は同じ名前のファイルが二つ並んだり、プログラムからファイルを開けなかったりするのはこのためだ。

## 四つの正規化形式

Unicodeは、こうした表し方の違いをそろえる手順を「正規化形式」として定めている[^uax15]。正規化形式は、二つの観点の組み合わせで四つある。

一つ目の観点は、合成するか分解するかである。分解（decomposition）は「が」を「か」と濁点に分け、合成（composition）は分けたものを一文字に戻す。二つ目の観点は、どこまでを同じ文字とみなすかである。正準等価は、見た目も意味も同じものだけを同一視する。互換等価は、半角カナと全角カナ、丸数字と数字のように、形は違っても同じ文字として扱えるものまで同一視する。

- NFC：正準分解したあと、正準合成する
- NFD：正準分解する
- NFKC：互換分解したあと、正準合成する
- NFKD：互換分解する

日本語の文字がそれぞれの形式でどう変わるかを[@tbl:forms]に示す。NFCとNFDが変えるのは濁点と半濁点の付け方だけで、文字の種類は変わらない。一方、NFKCとNFKDは、半角カナを全角に、全角英数字を半角に、「①」を「1」に、「㍻」を「平成」に置き換える。

::resource{id="tbl:forms"}

## Pythonで正規化する

Pythonでは、標準ライブラリの\`unicodedata\`モジュールにある\`normalize()\`関数で正規化できる。第1引数に形式の名前を、第2引数に文字列を渡す。

\`\`\`python normalize.py
import unicodedata

s = "ｶﾞｲﾄﾞ ＡＢＣ ①"
for form in ("NFC", "NFKC"):
    print(form, unicodedata.normalize(form, s))
\`\`\`

実行すると、NFCの行には入力がそのまま出力され、NFKCの行には「ガイド ABC 1」が出力される。半角の「ｶ」と「ﾞ」が、一文字の「ガ」になっている点に注意してほしい。互換分解で「ｶﾞ」が「カ」と結合用の濁点に分かれ、続く正準合成で「ガ」にまとめられるからだ。

Python 3.8で加わった\`is_normalized()\`関数を使うと、文字列がすでに正規化されているかどうかを調べられる。読み込んだデータのうち変換の要るものだけを選べるので、大量のファイル名やレコードを処理するときに役に立つ。

## 検索キーを作るときの注意

利用者が入力した語で検索する場合、検索キーと検索対象の両方をNFKCで正規化しておけば、半角と全角の違いを気にせずに照合できる。ただし、NFKCは情報を捨てる変換でもある。「①」は「1」に、「㈱」は「(株)」になり、元の文字には戻せない。

:::callout{type="point" title="ポイント"}
正規化した文字列は、検索と照合のためだけに使う。画面に表示する文字列と保存する文字列には、利用者が入力したものをそのまま残しておく。
:::

もう一つ、NFKCでもそろわない揺れがある。波ダッシュ「〜」（U+301C）と全角チルダ「～」（U+FF5E）だ。NFKCは全角チルダを半角の「~」に変えるが、波ダッシュは変えない[^wave]。そのため「10〜20」と「10～20」は、NFKCを通しても一致しない。こうした文字は、正規化とは別に対応表を用意して置き換える必要がある。

[^hfs]: HFS+が使うのは、Unicodeの規格のNFDに近い独自の分解形である。2017年に導入されたAPFSは、ファイル名を書かれたとおりに保存する。

[^uax15]: Unicode Standard Annex #15「Unicode Normalization Forms」。Unicodeの版ごとに改訂され、unicode.orgで公開されている。

[^wave]: Shift_JISの0x8160は、JISの対応表では波ダッシュに、マイクロソフトのCP932の表では全角チルダに変換される。同じ文書でも、変換に使った表によって符号位置が分かれる。

:::paragraphs{style="colophon"}
A specimen chapter written for the Postext Cookbook; the book and its other chapters are fictitious. Set in Noto Serif JP, Noto Sans JP and BIZ UDGothic (SIL OFL). Text: CC BY 4.0.
:::
`; // content.<lang>.md: the same Japanese chapter in both

// #region resources: the table of forms, and the bytes of が drawn in code
const FORMS = `入力\tNFC\tNFD\tNFKC
が（U+304C）\tU+304C\tU+304B U+3099\tU+304C
ｶﾞ（U+FF76 U+FF9E）\tそのまま\tそのまま\tガ（U+30AC）
ＡＢＣ（全角）\tそのまま\tそのまま\tABC
①\tそのまま\tそのまま\t1
㍻\tそのまま\tそのまま\t平成`;
const resources = [
  { id: 'tbl:forms', typeId: 'table', kind: 'table', createdAt: 0, updatedAt: 0,
    placement: { position: 'here' }, // at its ::resource line, under the paragraph citing it
    caption: '日本語の文字と四つの正規化形式（NFKDは、NFKCで合成された文字を分解した形になる）',
    table: { model: { ...parseTSV(FORMS), headerRowCount: 1, columnWidths: [34, 18, 26, 22] } } },
  { id: 'fig:bytes', typeId: 'figure', kind: 'svg', createdAt: 0, updatedAt: 0,
    caption: '「が」の二つの表し方。上段が符号位置、下段がUTF-8のバイト列',
    altText: 'が as one code point U+304C, three UTF-8 bytes E3 81 8C; and as U+304B and the '
      + 'combining voiced mark U+3099, six bytes E3 81 8B E3 82 99.',
    svg: { fileId: 'bytes.svg', width: 1175, height: 400 } },
];
// #endregion

// #region art: the figure's labels name the code face, and loadSvg embeds its file
const LABELS = `text{font-family:"${CODE}";font-size:3.2px;text-anchor:middle;fill:${palette.ink}}`;
function bytesArt() {
  const box = (x, y, w, h, fill, stroke, text) => `<rect x="${x}" y="${y}" width="${w}" `
    + `height="${h}" fill="${palette[fill]}" stroke="${palette[stroke]}" stroke-width=".3"/>`
    + `<text x="${x + w / 2}" y="${y + h / 2 + 1.1}">${text}</text>`;
  const side = (x, y, text, color) => `<text x="${x}" y="${y}" style="fill:${palette[color]};`
    + `font-size:3.6px">${text}</text>`;
  const row = (y, label, points, bytes) => side(14, y + 9, label, 'teal')
    + points.map((p, i) => box(26 + i * 31, y, 30, 7, 'tint', 'teal', p)).join('')
    + bytes.map((b, i) => box(26 + i * 10 + Math.floor(i / 3), y + 9, 9.5, 7, 'paper', 'rule', b))
      .join('') + side(103, y + 9, `${bytes.length} bytes`, 'muted');
  return `<svg xmlns="http://www.w3.org/2000/svg" width="1175" height="400" viewBox="0 0 117.5 40">`
    + `<style>${LABELS}</style>${row(3, 'NFC', ['U+304C'], ['E3', '81', '8C'])}`
    + `${row(22, 'NFD', ['U+304B', 'U+3099'], ['E3', '81', '8B', 'E3', '82', '99'])}</svg>`;
}
// #endregion

// ─── 3 · Fonts ──────────────────────────────────────────────────────────────
const FONTS = { // every face the pages use, loaded before the build
  'Noto Serif JP': ['400'], // 明朝: the text, the notes, the table
  'Noto Sans JP': ['400', '700'], // ゴシック: heads, the lead, labels, the point box, folios
  'BIZ UDGothic': ['400'], // the code: fixed pitch, half-width Latin, inline and in the listing
};

// ─── 4 · Build & show ───────────────────────────────────────────────────────
// Each Japanese face loads the files of what it sets (gotcha: cjk-fonts-slices).
const all = (re) => (markdown.match(re) ?? []).join('');
const gothic = `${all(/^#+ .*$/gm)}${all(/:::paragraphs\{style="lead"\}\n[^\n]*/g)}`
  + `${all(/:::callout\{type="point"[\s\S]*?\n:::/g)}${resources.map((r) => r.caption).join('')}`
  + '入力NFCDK第章実践日本語テキスト処理CHAPTER0123456789';
await loadFonts(FONTS, markdown);
await loadCjkFonts({ [MINCHO]: FONTS[MINCHO] }, `${markdown}${FORMS}`);
await loadCjkFonts({ [GOTHIC]: FONTS[GOTHIC] }, gothic);
const code = (markdown.match(/^```[\s\S]*?^```$|`[^`\n]+`/gm) ?? []).join('');
await loadCjkFonts({ [CODE]: FONTS[CODE] }, code);
await loadSvg('bytes.svg', bytesArt());
const continuation = { pageIndexOffset: 40, pageNumbering: { startAt: 41 }, headings: { h1: 2 } };
const doc = await withLoadedFonts(() => buildDocument({ markdown, resources, continuation },
  config()),
  { ...kitFonts(FONTS), text: markdown });
showPages(doc, { title: t({ en: 'A Japanese technical manual',
  es: 'Un manual técnico japonés' }) });
offerPdf(() => renderToPdf(doc, { fontProvider: cjkPdfProvider, resourceBytes: imageBytes }),
  `${RECIPE}.pdf`);

Kit · core, fonts, viewer, pdf, images, cjk: in jedem Rezept gleich · 504 Zeilen
// ─── Kit ── helpers shared by every Cookbook recipe · postext.dev/cookbook ─────

// ─── Kit · core v1 ── the same in every recipe · postext.dev/cookbook
function mm(value) { return { value, unit: 'mm' }; }
function pt(value) { return { value, unit: 'pt' }; }
function em(value) { return { value, unit: 'em' }; }
/** The sample language's string: t({ en: 'Figure', es: 'Figura' }). */
function t(strings) { return strings[LANG] ?? Object.values(strings)[0]; }
/** A file in this recipe's assets folder, served from the Postext repo by jsDelivr. */
function asset(file) { return `https://cdn.jsdelivr.net/gh/drnachio/postext@main/cookbook/${RECIPE}/assets/${file}`; }

// ─── Kit · fonts v3 ── the same in every recipe · postext.dev/cookbook
// Postext loads the faces a config asks for before layout (prepareFonts,
// buildDocumentWithFonts) and asks the kit for the files: Fontsource's, the
// files the PDF embeds too.

/** The options of buildDocumentWithFonts / prepareFonts / withLoadedFonts:
 *  the Fontsource resolver, and the faces of FONTS loaded with the ones the
 *  config names. FONTS = { 'Family Name': ['400', '400i', '700'] }. */
function kitFonts(FONTS = {}) {
  const faces = Object.entries(FONTS).flatMap(([family, specs]) => specs.map((spec) =>
    ({ family, weight: parseInt(spec, 10), style: spec.endsWith('i') ? 'italic' : 'normal' })));
  return { resolve: fontsourceResolver, faces };
}

/** A face's Fontsource files, in a fixed order: latin, then latin-ext and greek
 *  when `text` needs them and the family ships them (kitSubsetsFor), each with
 *  its unicode-range. Null for a weight or slant Fontsource does not ship. */
async function fontsourceResolver(family, weight, style, { text = '' } = {}) {
  const id = fontsourceId(family);
  const meta = await fontsourceMeta(family);
  if (meta && !(meta.weights.includes(weight) && meta.styles.includes(style))) return null;
  kitStatus('Loading fonts…');
  const at = await fontsourceAt(id);
  return ['latin', ...kitSubsetsFor(text, meta)].map((subset) => ({
    source: `${at}files/${id}-${subset}-${weight}-${style}.woff2`,
    unicodeRange: kitRanges()[subset],
  }));
}

/** The unicode-range of each Fontsource file the kit loads (a function: the
 *  kit is inlined after the pen's code, which runs before a const would be set). */
function kitRanges() {
  return {
    latin: 'U+0000-00FF,U+0131,U+0152-0153,U+02BB-02BC,U+02C6,U+02DA,U+02DC,U+0304,U+0308,U+0329,'
      + 'U+2000-206F,U+20AC,U+2122,U+2191,U+2193,U+2212,U+2215,U+FEFF,U+FFFD',
    'latin-ext': 'U+0100-02BA,U+02BD-02C5,U+02C7-02CC,U+02CE-02D7,U+02DD-02FF,U+0304,U+0308,U+0329,'
      + 'U+1D00-1DBF,U+1E00-1E9F,U+1EF2-1EFF,U+2020,U+20A0-20AB,U+20AD-20C0,U+2113,U+2C60-2C7F,U+A720-A7FF',
    greek: 'U+0370-03FF',
  };
}

/** For the cjk, arabic and comics blocks, which add their own files: the
 *  faces of `faces` = { 'Family Name': ['400', '400i', '700'] }. `text` is the sample:
 *  č ł † α χ also load latin-ext and greek files (kitSubsetsFor). With
 *  `optional`, a face Fontsource does not ship is skipped instead of failing.
 *  The files are added once all have loaded, in the order asked (latin, then
 *  latin-ext, then greek), whatever order they arrive in: the last one added
 *  wins a character two files hold. Resolves to the number of faces added. */
async function loadFonts(faces, text = '', { optional = false } = {}) {
  kitStatus('Loading fonts…');
  const ranges = kitRanges();
  const jobs = [];
  for (const [family, specs] of Object.entries(faces)) {
    const id = fontsourceId(family);
    const todo = [...new Set(specs)].map((spec) => [parseInt(spec, 10), spec.endsWith('i') ? 'italic' : 'normal'])
      .filter(([weight, style]) => !hasFace(family, weight, style)); // before any await
    const at = fontsourceAt(id);
    const meta = optional || /[^\0-ÿ]/u.test(text) ? await fontsourceMeta(family) : null;
    const subsets = ['latin', ...kitSubsetsFor(text, meta)];
    for (const [weight, style] of todo) {
      if (optional && !(meta?.weights.includes(weight) && meta.styles.includes(style))) continue;
      for (const subset of subsets) {
        const url = `${await at}files/${id}-${subset}-${weight}-${style}.woff2`;
        const face = new FontFace(family, `url(${url}) format('woff2')`,
          { weight: String(weight), style, unicodeRange: ranges[subset] });
        jobs.push(face.load().catch(() => {
          if (subset === 'latin' && !optional) throw new Error(`Fontsource has no ${family} ${weight} ${style}`);
        }));
      }
    }
  }
  const ready = await Promise.all(jobs).catch((error) => { kitFail(error); throw error; });
  return ready.filter((face) => face && document.fonts.add(face)).length;
}

/** A loaded FontFace covers this family, weight and style (fonts.check() would
 *  also say yes for families nobody declared). */
function hasFace(family, weight, style) {
  for (const face of document.fonts) {
    if (face.status !== 'loaded' || face.style !== style) continue;
    if (face.family.replace(/^["']|["']$/g, '') !== family) continue;
    const [low, high = low] = face.weight.split(' ').map(Number);
    if (weight >= low && weight <= high) return true;
  }
  return false;
}

/** The files beyond latin `text` needs that `meta`'s family ships. */
function kitSubsetsFor(text, meta) {
  return [[/[Ā-˿ᴀ-ᶿḀ-ỿ†ℓⱠ-Ɀ꜠-ꟿ]/u, 'latin-ext'], [/[Ͱ-Ͽ]/u, 'greek']]
    .filter(([re, x]) => re.test(text) && meta?.subsets?.includes(x)).map(([, x]) => x);
}

/** Fontsource's id for a family: 'Source Serif 4' → 'source-serif-4'. */
function fontsourceId(family) { return family.toLowerCase().replace(/\s+/g, '-'); }

/** A package's folder at the one version @5 names now (#766). */
function fontsourceAt(id, u = `https://cdn.jsdelivr.net/npm/@fontsource/${id}@`) {
  return (fontsourceAt.c ??= {})[id] ??= fetch(u + '5/package.json').then((r) => r.json())
    .then((p) => p.version || 5, () => 5).then((v) => `${u}${v}/`);
}

/** The family's Fontsource metadata (weights, styles, subsets), or null. */
function fontsourceMeta(family) {
  fontsourceMeta.cache ??= new Map();
  const id = fontsourceId(family);
  if (!fontsourceMeta.cache.has(id)) {
    fontsourceMeta.cache.set(id, fetch(`https://api.fontsource.org/v1/fonts/${id}`)
      .then((res) => (res.ok ? res.json() : null), () => null));
  }
  return fontsourceMeta.cache.get(id);
}

// ─── Kit · viewer v2 ── the same in every recipe · postext.dev/cookbook
/** The pages as spreads on a dark desk, page 1 alone, then verso | recto,
 *  each painted when it scrolls near. Other options go to renderPageToCanvas. */
function showPages(docs, { title, width = 460, ...render } = {}) {
  const root = viewer(title);
  const pages = [docs].flat().flatMap((doc) =>
    doc.pages.map((page) => ({ doc, page, n: (doc.pageIndexOffset ?? 0) + page.index })));
  const spreads = [];
  let verso = null;
  for (const p of pages) {
    if (p.n % 2 === 1) { if (verso) spreads.push([verso, null]); verso = p; }
    else { spreads.push([verso, p]); verso = null; }
  }
  if (verso) spreads.push([verso, null]);
  const density = Math.min(window.devicePixelRatio || 1, 2);
  showPages.painter?.disconnect();
  const painter = new IntersectionObserver((entries) => {
    for (const { isIntersecting, target } of entries) {
      if (!isIntersecting) continue;
      painter.unobserve(target);
      const { doc, page } = target.postext;
      renderPageToCanvas(page, doc, target, { ...render, scale: (width * density) / page.width });
    }
  }, { rootMargin: '800px' });
  showPages.painter = painter;
  root.replaceChildren(...spreads.map((pair) => {
    const spread = document.createElement('div');
    spread.className = 'pt-spread';
    for (const p of pair) {
      const figure = document.createElement('figure');
      if (p) {
        const label = p.page.pageLabel || String(p.n + 1);
        const canvas = document.createElement('canvas');
        canvas.postext = p;
        canvas.style.aspectRatio = `${p.page.width} / ${p.page.height}`;
        canvas.setAttribute('role', 'img');
        canvas.setAttribute('aria-label', `Page ${label}`);
        const folio = document.createElement('figcaption');
        folio.textContent = label;
        figure.append(canvas, folio);
        painter.observe(canvas);
      } else figure.className = 'pt-blank';
      spread.append(figure);
    }
    return spread;
  }));
  kitStatus(`${pages.length} ${pages.length === 1 ? 'page' : 'pages'}`);
  document.documentElement.dataset.postext = 'ready';
  return pages.length;
}

/** The desk, the bar and the error reporting, created once. */
function viewer(title) {
  if (!document.getElementById('pt-kit')) {
    document.head.insertAdjacentHTML('beforeend', `<style id="pt-kit">
      :root { color-scheme: dark; }
      body { margin: 0; background: #0e1014; color: #b9bcc4; font: 13px/1.45 system-ui, sans-serif; }
      #pt-bar { position: sticky; top: 0; z-index: 1; display: flex; flex-wrap: wrap; align-items: center;
        gap: 6px 16px; padding: 10px 16px; background: rgb(14 16 20 / .92); backdrop-filter: blur(6px);
        border-bottom: 1px solid #23262d; }
      #pt-bar strong { color: #f4f1ea; font-weight: 600; }
      #pt-actions { display: flex; gap: 12px; margin-left: auto; }
      #pt-actions a, #pt-actions button { color: #d8a21a; font: inherit; background: none; border: 0; padding: 0; cursor: pointer; }
      #pages { display: grid; justify-items: center; gap: 48px; padding: 32px 16px 72px; }
      .pt-spread { display: flex; }
      .pt-spread figure { margin: 0; width: min(460px, 44vw); }
      .pt-spread canvas { display: block; width: 100%; background: #fff;
        box-shadow: 0 1px 2px rgb(0 0 0 / .5), 0 22px 44px -16px rgb(0 0 0 / .8); }
      .pt-spread figure:first-child canvas { box-shadow: inset -14px 0 14px -14px rgb(0 0 0 / .18), 0 1px 2px rgb(0 0 0 / .5), 0 22px 44px -16px rgb(0 0 0 / .8); }
      .pt-spread figcaption { margin-top: 10px; text-align: center; font: 600 10px/1 system-ui, sans-serif;
        letter-spacing: .18em; text-transform: uppercase; color: #6c7079; }
      .pt-blank { visibility: hidden; }
      @media (max-width: 760px) {
        .pt-spread { flex-direction: column; gap: 32px; }
        .pt-spread figure { width: min(460px, 92vw); }
        .pt-blank { display: none; }
      }
    </style>`);
    document.body.insertAdjacentHTML('afterbegin',
      '<header id="pt-bar"><strong id="pt-title"></strong><span id="pt-status" role="status"></span><span id="pt-actions"></span></header>');
    document.getElementById('pt-title').textContent = document.title || 'Postext';
    addEventListener('error', (event) => kitFail(event.error ?? event.message));
    addEventListener('unhandledrejection', (event) => kitFail(event.reason));
  }
  if (title) document.getElementById('pt-title').textContent = title;
  return document.getElementById('pages')
    ?? document.body.appendChild(Object.assign(document.createElement('main'), { id: 'pages' }));
}

function kitStatus(text) {
  viewer();
  document.getElementById('pt-status').textContent = text;
}

function kitFail(error) {
  document.documentElement.dataset.postext = 'error';
  kitStatus(`Error: ${error?.message ?? error}`);
}

// ─── Kit · pdf v3 ── the same in every recipe that exports a PDF
/** The Fontsource files the screen used, as TrueType, with their range and
 *  place: the nearest weight the family ships, upright with no italic;
 *  latin, then what its letters need (kitSubsetsFor). */
async function fontsourceProvider(family, weight, style, request) {
  const id = fontsourceId(family);
  const meta = await fontsourceMeta(family);
  const weights = meta?.weights?.length ? meta.weights : [400, 700];
  const w = weights.reduce((a, b) => (Math.abs(b - weight) < Math.abs(a - weight) ? b : a));
  const s = style === 'italic' && meta && !meta.styles.includes('italic') ? 'normal' : style;
  const text = String.fromCodePoint(...(request?.codePoints ?? []));
  const files = await Promise.all(['latin', ...kitSubsetsFor(text, meta)].map(async (subset, order) => {
    const res = await fetch(`${await fontsourceAt(id)}files/${id}-${subset}-${w}-${s}.woff2`);
    if (!res.ok) throw new Error(`Fontsource has no ${family} ${w} ${s} ${subset}`);
    return Object.assign(await decompressWoff2(new Uint8Array(await res.arrayBuffer())),
      { unicodeRange: kitRanges()[subset], order });
  }));
  return files.length === 1 ? files[0] : files;
}

/** A "Build the PDF" button; then "Open the PDF" (a new tab: CodePen's frame
 *  shows no PDFs) and a download link. */
function offerPdf(makePdf, filename) {
  viewer();
  const button = Object.assign(document.createElement('button'), { type: 'button', textContent: 'Build the PDF' });
  button.dataset.postextPdf = filename;
  button.addEventListener('click', async () => {
    button.disabled = true;
    button.textContent = 'Building the PDF…';
    try {
      const bytes = await makePdf();
      const url = URL.createObjectURL(new Blob([bytes], { type: 'application/pdf' }));
      const size = `${Math.max(1, Math.round(bytes.length / 1024))} KB`;
      button.replaceWith(
        Object.assign(document.createElement('a'), { href: url, target: '_blank', rel: 'noopener', textContent: 'Open the PDF ↗' }),
        Object.assign(document.createElement('a'), { href: url, download: filename, textContent: `Download ${filename} · ${size}` }));
    } catch (error) {
      button.disabled = false;
      button.textContent = 'Build the PDF';
      kitFail(error);
    }
  });
  document.getElementById('pt-actions').append(button);
}

// ─── Kit · images v3 ── recipes with pictures · postext.dev/cookbook
/** Registers a photo or PNG for the canvas and keeps its bytes for the PDF.
 *  fetch → ImageBitmap never taints the canvas (a plain cross-origin <img> would). */
async function loadImage(fileId, url) {
  const res = await fetch(url);
  if (!res.ok) throw new Error(`Image not found (${res.status}): ${url}`);
  const bytes = new Uint8Array(await res.arrayBuffer());
  registerResourceImage(fileId, await createImageBitmap(new Blob([bytes])));
  (loadImage.bytes ??= new Map()).set(fileId, bytes);
}

/** Registers SVG markup as a vector image, its fonts inside (inlineSvgFonts, ≥ 1.25).
 *  `options` go to registerResourceImage: { singleInk: true } tints it on the canvas. */
async function loadSvg(fileId, svg, options) {
  if (typeof inlineSvgFonts === 'function') svg = await inlineSvgFonts(svg, svgFaces);
  const img = new Image();
  img.src = `data:image/svg+xml;charset=utf-8,${encodeURIComponent(svg)}`;
  await img.decode();
  registerResourceImage(fileId, img, options);
  (loadImage.bytes ??= new Map()).set(fileId, new TextEncoder().encode(svg));
}

/** A face's Fontsource files: nearest weight; latin + what the SVG sets. */
async function svgFaces(family, weight, style, { codePoints = [] } = {}) {
  const [id, m] = [fontsourceId(family), await fontsourceMeta(family)];
  const w = m?.weights.reduce((a, b) => (Math.abs(b - weight) < Math.abs(a - weight) ? b : a)) ?? weight;
  const s = m?.styles.includes(style) === false ? 'normal' : style;
  return Promise.all(['latin', ...kitSubsetsFor(String.fromCodePoint(...codePoints), m)].map(async (x) => new Uint8Array(
    await (await fetch(`${await fontsourceAt(id)}files/${id}-${x}-${w}-${s}.woff2`)).arrayBuffer())));
}

/** renderToPdf({ resourceBytes: imageBytes }) */
function imageBytes(fileId) { return loadImage.bytes?.get(fileId); }

/** renderToHtml({ resourceImageUrl: imageUrl }) */
function imageUrl(fileId) {
  const bytes = imageBytes(fileId);
  if (!bytes) return undefined;
  imageUrl.urls ??= new Map();
  if (!imageUrl.urls.has(fileId)) {
    const type = /\.svg$/i.test(fileId) ? 'image/svg+xml' : /\.png$/i.test(fileId) ? 'image/png' : 'image/jpeg';
    imageUrl.urls.set(fileId, URL.createObjectURL(new Blob([bytes], { type })));
  }
  return imageUrl.urls.get(fileId);
}

// ─── Kit · cjk v2 ── Chinese, Japanese and Korean books · postext.dev/cookbook
// Fontsource ships a CJK family as about a hundred files per weight, each
// declared in its stylesheet with the unicode-range it covers. The screen
// loads the files the sample touches; the PDF gets the same files for the
// characters its pages set in each face, and embeds each as a subset.
// A book bound on the right (vertical text) is shown with its spreads
// mirrored: page 1 alone on the left of the spine, then [3 | 2].

/** The files of a Fontsource face, read from its stylesheet: { url, range,
 *  ranges }, the last declared first (the order the browser tries them in). */
function cjkSlices(family, weight, style) {
  cjkSlices.cache ??= new Map();
  const id = fontsourceId(family);
  const css = `${id}/${weight}${style === 'italic' ? '-italic' : ''}.css`;
  if (!cjkSlices.cache.has(css)) {
    let at;
    cjkSlices.cache.set(css, fontsourceAt(id).then((a) => fetch(at = a + css.slice(id.length + 1)))
      .then((res) => {
        if (!res.ok) throw new Error(`Fontsource has no ${family} ${weight} ${style} (${res.status})`);
        return res.text();
      })
      .then((text) => [...text.matchAll(/@font-face\s*{([^}]*)}/g)].map(([, rule]) => {
        const range = /unicode-range:\s*([^;]+);/.exec(rule)?.[1].trim() ?? 'U+0-10FFFF';
        const ranges = range.split(',').map((part) => {
          const [lo, hi = lo] = part.trim().slice(2).split('-');
          return [parseInt(lo, 16), parseInt(hi, 16)];
        });
        return { url: new URL(/url\(([^)]+?\.woff2)\)/.exec(rule)[1], at).href, range, ranges };
      }).reverse()));
  }
  return cjkSlices.cache.get(css);
}

/** The file of `slices` the screen sets code point `cp` in, if any: the
 *  last one loadCjkFonts adds whose range holds it. A numbered file that
 *  repeats a letter of latin is added after latin, and wins it. */
function cjkSliceFor(slices, cp) {
  const holds = (slice) => slice.ranges.some(([lo, hi]) => cp >= lo && cp <= hi);
  return slices.reduce((last, slice) => (holds(slice) ? slice : last), undefined);
}

/** Whether Fontsource serves `family` as a Chinese, Japanese or Korean
 *  family (its subsets name the script). Fails when the API does not
 *  answer: a CJK face taken for a Latin one would paint in a system face. */
async function isCjkFamily(family) {
  const meta = await fontsourceMeta(family);
  if (!meta) throw new Error(`api.fontsource.org did not describe ${family}: reload to try again`);
  return !!meta.subsets?.some((subset) => /^(chinese|japanese|korean)/.test(subset));
}

/** faces = { 'Noto Serif TC': ['400', '700'] }, as for loadFonts: the
 *  whole FONTS object may be passed, its other families are left to
 *  loadFonts. Adds one FontFace per file of each CJK face with its
 *  unicodeRange, then loads the files `text` touches. `text` is what the
 *  faces set: the sample for the text face; a book in several voices calls
 *  it once per voice (loadCjkFonts({ 'LXGW WenKai TC': ['400'] }, quotes)),
 *  so the heading and quotation faces fetch and check only their own
 *  characters. Fails when a character of `text` is in no file of a face.
 *  List every weight the pages use: a weight left to the engine's later round
 *  gets the latin file only. With { vertical: true } it also loads each
 *  family's vertical forms (brackets, quotes, pause marks) for the canvas,
 *  which needs loadVerticalAlternates imported from postext. Resolves to
 *  the number of files loaded. */
async function loadCjkFonts(faces, text, { vertical = false } = {}) {
  kitStatus('Loading fonts…');
  let loaded = 0;
  try {
    if (vertical && typeof loadVerticalAlternates !== 'function') {
      throw new Error('loadCjkFonts(…, { vertical: true }) needs loadVerticalAlternates imported from postext');
    }
    for (const [family, specs] of Object.entries(faces)) {
      if (!(await isCjkFamily(family))) continue;
      const twin = [];
      for (const spec of new Set(specs)) {
        const weight = parseInt(spec, 10);
        const style = spec.endsWith('i') ? 'italic' : 'normal';
        const slices = await cjkSlices(family, weight, style);
        const missing = [...new Set(text)].filter((ch) => /\S/.test(ch) && !cjkSliceFor(slices, ch.codePointAt(0)));
        if (missing.length) {
          throw new Error(`${family} ${spec} has no file for ${missing.slice(0, 12).join(' ')}: `
            + `give each face the text it sets (loadCjkFonts({ '${family}': ['${spec}'] }, text))`);
        }
        for (const slice of slices) {
          document.fonts.add(new FontFace(family, `url(${slice.url}) format('woff2')`,
            { weight: String(weight), style, unicodeRange: slice.range }));
          twin.push({ source: slice.url, weight: String(weight), style, unicodeRange: slice.range });
        }
        const font = `${style === 'italic' ? 'italic ' : ''}${weight} 16px "${family}"`;
        loaded += (await document.fonts.load(font, text)).length;
        if (!document.fonts.check(font, text)) throw new Error(`${family} ${spec} did not load for the sample`);
      }
      // The same files under a twin name with the `vert` feature on: the
      // canvas paints the punctuation of vertical lines with it.
      if (vertical && twin.length) await loadVerticalAlternates(family, twin);
    }
  } catch (error) {
    kitFail(error);
    throw error;
  }
  return loaded;
}

/** The PDF font provider for recipes with CJK faces: a family whose
 *  Fontsource subsets are Chinese, Japanese or Korean gets the files the
 *  screen set its pages' characters in (`request.codePoints`), each marked
 *  with its unicode-range and its place in the order loadCjkFonts added
 *  them: postext-pdf sets a character two files hold in the one added
 *  last, as the browser does. Any other family gets the latin file
 *  fontsourceProvider fetches (the "pdf" block) and, when the face sets
 *  letters only latin-ext has, that file too. */
async function cjkPdfProvider(family, weight, style, request) {
  if (!(await isCjkFamily(family))) return cjkLatinPdfFiles(family, weight, style, request);
  const meta = await fontsourceMeta(family);
  const weights = meta.weights?.length ? meta.weights : [400, 700];
  const w = weights.reduce((a, b) => (Math.abs(b - weight) < Math.abs(a - weight) ? b : a));
  const s = style === 'italic' && !meta.styles.includes('italic') ? 'normal' : style;
  const slices = await cjkSlices(family, w, s);
  const picked = new Set();
  for (const cp of request?.codePoints ?? []) {
    const slice = cjkSliceFor(slices, cp);
    if (slice) picked.add(slice);
  }
  if (!picked.size) picked.add(slices[0]);
  return Promise.all(slices.map(async (slice, order) => {
    if (!picked.has(slice)) return null;
    const res = await fetch(slice.url);
    if (!res.ok) throw new Error(`Fontsource file ${slice.url} (${res.status})`);
    const bytes = await decompressWoff2(new Uint8Array(await res.arrayBuffer()));
    return Object.assign(bytes, { unicodeRange: slice.range, order });
  })).then((files) => files.filter(Boolean));
}

/** A Latin family set next to the CJK faces: its latin file, then its
 *  latin-ext file when the face sets letters only latin-ext has (ō ū in
 *  Hepburn rōmaji, ǎ in pinyin), the file loadFonts adds on screen for
 *  them, each marked with its unicode-range and its place in that order:
 *  the PDF sets a character in the file the browser took it from. A face
 *  Fontsource ships without latin-ext, or whose file does not come, gets
 *  latin alone, and the PDF names the letters it lacks. */
async function cjkLatinPdfFiles(family, weight, style, request) {
  const meta = await fontsourceMeta(family);
  const beyond = [...(request?.codePoints ?? [])].some(cjkLatinExtOnly);
  if (!beyond || !meta?.subsets?.includes('latin-ext')) return fontsourceProvider(family, weight, style);
  const weights = meta.weights?.length ? meta.weights : [400, 700];
  const w = weights.reduce((a, b) => (Math.abs(b - weight) < Math.abs(a - weight) ? b : a));
  const s = style === 'italic' && !meta.styles.includes('italic') ? 'normal' : style;
  const id = fontsourceId(family);
  const url = `${await fontsourceAt(id)}files/${id}-latin-ext-${w}-${s}.woff2`;
  const [latin, ext] = await Promise.all([fontsourceProvider(family, weight, style), fetch(url)
    .then(async (res) => (res.ok ? decompressWoff2(new Uint8Array(await res.arrayBuffer())) : null), () => null)]);
  const ranges = kitRanges();
  return ext ? [Object.assign(latin, { unicodeRange: ranges.latin, order: 0 }),
    Object.assign(ext, { unicodeRange: ranges['latin-ext'], order: 1 })] : latin;
}

/** Whether code point `cp` is in Fontsource's latin-ext file and not in
 *  its latin file: Latin Extended-A and -B, IPA, the spacing modifiers and
 *  Latin Extended Additional (loadFonts's test for latin-ext), less the
 *  few latin holds too (ı Œ œ ʻ ʼ ˆ ˚ ˜). */
function cjkLatinExtOnly(cp) {
  if (!((cp >= 0x100 && cp <= 0x2ff) || (cp >= 0x1e00 && cp <= 0x1eff))) return false;
  return ![0x131, 0x152, 0x153, 0x2bb, 0x2bc, 0x2c6, 0x2da, 0x2dc].includes(cp);
}

/** showPages for a book bound on either edge. A right-bound book (the
 *  document says so: doc.binding is 'right' for page.binding 'right' and
 *  for vertical text) lies on the desk as it opens: page 1 alone on the
 *  left of the spine, then [3 | 2], the spine shade on each page's inner
 *  edge. `binding` ('left' | 'right') overrides the document's. */
function showBook(docs, { binding, ...options } = {}) {
  const count = showPages(docs, options);
  const right = (binding ?? [docs].flat()[0]?.binding) === 'right';
  if (!document.getElementById('pt-kit-cjk')) {
    // The pages keep direction ltr: a canvas draws text in the direction its
    // element inherits, and under rtl each run would end where the engine
    // starts it, its brackets mirrored.
    document.head.insertAdjacentHTML('beforeend', `<style id="pt-kit-cjk">
      .pt-spread[dir="rtl"] canvas { direction: ltr; }
      .pt-spread[dir="rtl"] figure:first-child canvas { box-shadow: inset 14px 0 14px -14px rgb(0 0 0 / .18),
        0 1px 2px rgb(0 0 0 / .5), 0 22px 44px -16px rgb(0 0 0 / .8); }
    </style>`);
  }
  // Each pair stays [verso, recto] in the page; right to left, the verso
  // sits on the right. Phones stack the pages in reading order either way.
  for (const spread of document.querySelectorAll('#pages > .pt-spread')) spread.dir = right ? 'rtl' : 'ltr';
  document.getElementById('pages').dataset.binding = right ? 'right' : 'left';
  return count;
}

// ─── /Kit ───────────────────────────────────────────────────────────────────────