@libraz/mejiro

Japanese vertical text layout engine for the web. It handles line breaking, kinsoku shori, hanging punctuation, ruby, tate-chu-yoko, pagination, image exclusion, and EPUB parsing and authoring. The core has zero DOM dependencies. Pairs with @libraz/mejiro-react / @libraz/mejiro-vue for ready-to-use reader/editor components.

Install

npm install @libraz/mejiro

Quick start

import { DEFAULT_HEADING_STYLES, MejiroBook } from '@libraz/mejiro/book';
import { parseEpub } from '@libraz/mejiro/epub';

const book = new MejiroBook({
  fontFamily: '"Noto Serif JP"',
  fontSize: 16,
  lineSpacing: 1.8,
  headingStyles: DEFAULT_HEADING_STYLES,
});
book.computePageSize(document.querySelector('.reader')!);

const epub = await parseEpub(epubBuffer);
const layout = await book.layoutChapter(epub.chapters[0]);
const spread = layout.getSpread(0);

Subpath exports

| Import path | Description | |---|---| | @libraz/mejiro | Core: line breaking, kinsoku, hanging, ruby (preprocessRuby), tate-chu-yoko (preprocessTcy / buildTcyAnnotations), NFC normalization (normalizeText), pagination (paginate / getLineRanges), image exclusion, annotation / reading-position persistence, i18n message catalogs | | @libraz/mejiro/browser | Font measurement (MejiroBrowser, CharMeasurer), ruby font derivation, image drag and resize sessions (createOverlayDragSession) | | @libraz/mejiro/epub | EPUB parsing (parseEpub) with untrusted-archive resource limits, EditableEpub write-back, cloneEditableEpubBook / clampEditableEpubSelection, manuscript parsing (parseManuscript), EpubProject authoring | | @libraz/mejiro/render | Layout → RenderPage data, paragraphClassName, static HTML (renderEpubStatic), stylesheets | | @libraz/mejiro/book | High-level MejiroBook / ChapterLayout: pagination, full-text search (findText), reading anchors, layout snapshots, estimateReadingTime | | @libraz/mejiro/image | Browser-side image decode / downscale helpers (prepareImage) | | @libraz/mejiro/analysis | Optional morphological analyzer for analysis-driven line breaking (createSuzumeAnalyzer, backed by the optional peer @libraz/suzume) |

Stylesheets

Five stylesheets ship as separate subpaths, so a host app imports only what it renders:

| Import path | Purpose | |---|---| | @libraz/mejiro/render/mejiro.css | Base vertical text layout — required whenever RenderPage data is rendered | | @libraz/mejiro/render/mejiro-reader.css | Reader chrome: header, spread frame, page navigation | | @libraz/mejiro/render/mejiro-editor.css | Editor chrome: toolbars, inline editing, image overlays | | @libraz/mejiro/render/mejiro-fonts.css | Optional Japanese webfont declarations | | @libraz/mejiro/render/mejiro-print.css | Print / PDF output |

Untrusted EPUB archives

parseEpub() and parseEditableEpub() enforce resource limits by default, so a hostile or malformed archive cannot exhaust memory: caps on compressed input size, entry count, per-entry and total expanded size, and per-entry compression ratio. Limits are checked against the bytes the decompressor really produces, not the sizes the archive declares for itself, and reading stops as soon as a cap would be passed.

The defaults are exported as DEFAULT_EPUB_PARSE_LIMITS. In a trusted environment — a build step over your own files, say — raise individual caps through EpubParseOptions.limits:

import { parseEpub } from '@libraz/mejiro/epub';

const book = await parseEpub(buffer, { limits: { maxTotalBytes: 1024 * 1024 * 1024 } });

Documentation

• Project README
• Getting started
• API reference
• 日本語ドキュメント: docs/ja

License

Apache License 2.0