Testing

The library's tests

The suite tests/Unit tests each fixer on its own cases and across parts of a run, the typographer on HTML fragments, documents and plain text, the hyphenator on small hand-written patterns and on the shipped ones, every shipped language read once, and emoji sequences through every preset. Each language whose patterns are written for this library has a list of at least forty common words that must break exactly where its rules allow, within its minimum lengths, and a test checks that the files under data/hyphenation/ are those bin/generate-patterns.php writes from the rules.

From the monorepo root:

vendor/bin/phpunit packages/typography/tests

The library is held to 100 % covered mutation score by Infection.

Benchmarks

TypographerBench fixes HTML fragments of 1, 10 and 100 KB in English and French, with hyphenation, and texts made of quotes, dots and dashes left open; it also builds the French typographer. The time grows with the length of the fragment. From the monorepo root:

composer bench -- packages/typography/tests/Benchmark

In worker mode, tests/Worker/worker.php is a page FrankenPHP serves from one process across requests, fixing 10 KB of French and 10 KB of English HTML; the services it builds before its loop serve every request, as they do in an application's worker. It needs the intl extension in the FrankenPHP image:

composer bench:worker phil/typography

Testing code that fixes text

The output is deterministic: the same typographer gives the same string. Compare it exactly, writing the invisible characters by their escapes so the test says what it expects:

self::assertSame("<p>«&nbsp;Oui&nbsp;»\u{202F}!</p>", Typographer::for('fr_FR', false)->fix('<p>"Oui" !</p>'));

The HTML5 serializer writes the no-break space as   and every other character as itself: the narrow no-break space (U+202F) and the soft hyphen (U+00AD) appear raw. Build the typographer without hyphenation in tests that do not check it, so a change of patterns does not change unrelated expectations.