@mark1russell7/typeset

This package measures text and breaks it into lines. Each area has its own subpath.

| Subpath | What it has | | --- | --- | | @mark1russell7/typeset/breaking | The paragraph model of boxes, glue and penalties, and the greedy, total fit and balanced line breakers. | | @mark1russell7/typeset/fonts | The font model: it reads a font file and gives the width of a text as a browser engine gives it. | | @mark1russell7/typeset/hyphen | hyphenator: the matcher of Liang, with no language data. | | @mark1russell7/typeset/hyphen/en-us, /de, /fr | The hyphen patterns of one language, for the hook hyphenate. | | @mark1russell7/typeset/measurer | fontAdvance: the advance function of a layout engine, from the font files of the host. | | @mark1russell7/typeset/text | The Text rope: a text with its width, and edits that measure only the changed parts. |

Line breaking

import { greedy, knuthPlass, linesOf, paragraphOf } from "@mark1russell7/typeset/breaking";

const widthOf = (text: string): number => 8 * [...text].length;
const paragraph = paragraphOf("The quick brown fox jumps over the lazy dog", { widthOf });
const lines = linesOf(paragraph, knuthPlass.breaks(paragraph, 120));

• Each breaker takes any width function. The function gives the width of a run of text in any unit.
• greedy fills each line as a browser does. knuthPlass chooses the breaks with the least sum of demerits. balanced keeps the greedy line count with more equal lines.
• The option hyphenate gives the offsets of the hyphen breaks of a word. A language subpath gives a function for this option.
• The options stretch and shrink make a justified paragraph. Total fit then uses the adjustment ratio of Knuth and Plass.
• worstRatio measures the justified edge. docs/BREAKING.md in the repository defines it.

import { knuthPlass, linesOf, paragraphOf, worstRatio } from "@mark1russell7/typeset/breaking";
import { hyphenate } from "@mark1russell7/typeset/hyphen/en-us";

const paragraph = paragraphOf("The quick brown fox jumps over the lazy dog", { widthOf, hyphenate, stretch: 4, shrink: 2.5 });
const lines = linesOf(paragraph, knuthPlass.breaks(paragraph, 120));
const edge = worstRatio(paragraph, lines, 120);

The sizes and the licenses of the language sets

Each set is on its own subpath, so an application pays only for the sets that it imports. The bytes are for one bundled set with the matcher, minified, at gzip level 9.

| Subpath | Gzip bytes | Source file of tex-hyphen | License | | --- | --- | --- | --- | | ./hyphen/en-us | 18202 | hyph-en-us.tex | The all-permissive notice of Gerard D.C. Kuiken | | ./hyphen/de | 123851 | hyph-de-1996.tex | MIT | | ./hyphen/fr | 5655 | hyph-fr.tex | MIT | | ./hyphen/es | 15349 | hyph-es.tex | MIT |

The set of fr also has 18 exceptions. They correct the defects of the TeX patterns against the reference list Lexique 3.83.

The license of a pattern file applies to its data. The header of each file stays at the top of its module and in a bundle. docs/HYPHENATION.md in the repository has the sources and the results.

The Text rope

import { Text } from "@mark1russell7/typeset/text";

const widthOf = (text: string): number => 8 * [...text].length;
const text = Text.from("The quick brown fox", widthOf);
const edited = text.insert(4, "very ").remove(0, 4);
console.log(edited.toString(), edited.width);

• Each edit gives a new rope. The old rope does not change, and the two share the parts that the edit does not touch.
• An edit uses the width function of the order of log n times, on the changed leaves and on short windows.
• The rope keeps the widths of short windows, so an edit measures only the windows that it changes. Text.from clears them, so a new rope sees a font that loaded.
• The width is exact when the width function obeys the separator law: width(p + s + q) = width(p + s) + width(s + q) - width(s).
• The separator s is a space or a tab. The option separators changes the set.
• The font model of ./fonts obeys the law for its shaping dom and canvas, with the sum exact.

The font model and the measurer

import { readFileSync } from "node:fs";
import { createMeasurer, layoutText } from "@mark1russell7/layout-contract/text";
import { fontAdvance } from "@mark1russell7/typeset/measurer";

const advance = fontAdvance({ faces: [{ family: "Noto Sans", data: readFileSync("NotoSans.ttf") }], engine: "freetype" });
const layout = layoutText("The quick brown fox", { font: "16px Noto Sans", measurer: createMeasurer(advance), lineHeight: 20, wrapWidth: 80 });

• The host gives the bytes of each font file. The package has no font file.
• fontAdvance reads the CSS font, and finds the face with the CSS rules of the family, the style and the weight.
• Each run of the text gets the first family with a glyph for its first character.
• The width of a run is measureWidth of ./fonts, with the script of the run and the rules of one engine of engines.
• The engine rows are freetype (Chromium on Linux), coretext (macOS) and directwrite (Windows).
• readFont reads TrueType, OpenType and WOFF2 files. A WOFF2 file needs a Brotli decompress function, for example brotliDecompressSync of node:zlib.

The repository mark1russell7/typeset has the report of the breakers in docs/BREAKING.md.