Download the PHP package php-spellcheck/spellcheck-core without Composer

PHP Spellcheck Core — spell checker for PHP code and translation files

Framework-agnostic PHP spell checking engine for source code and translation catalogues. It finds typos in PHP identifiers, docblocks, comments and in Symfony translation catalogues, with Hunspell, Aspell, pspell or a pure PHP backend, character-accurate line and column positions, ICU MessageFormat support, a baseline for legacy projects and PSR-6 caching.

Use it to fail CI on typos, to lint a translation catalogue in the language it is actually written in, or to build your own spell checking tool on top of the pipeline.

For the Symfony integration (console commands, bundle configuration, Translator support) use php-spellcheck/spellcheck-symfony-bundle.

Features

• Spell check PHP source code — class, method, property, variable, constant and function names are split on camelCase, snake_case, SCREAMING_SNAKE and acronyms before being checked.
• Spell check translation catalogues per locale — each message is checked in the language of its own catalogue, so messages.it.yaml is checked in Italian and messages.de.yaml in German. Keys are resolved back to file and line in YAML, XLIFF and PHP array files.
• ICU MessageFormat aware — a built-in recursive descent parser walks plural, select and selectordinal branches. No ext-intl required.
• Placeholders are never reported — %count%, {{ var }}, {name}, :attribute, sprintf directives, HTML tags, Markdown syntax, URLs and docblock tags are stripped before the words reach the speller.
• Character-accurate positions — every reported misspelling points at the line and column in the original file, not in the cleaned-up text.
• Five backends — Hunspell and Aspell over the Ispell pipe protocol, ext-pspell, a dependency-free pure PHP word list, or an automatic chain.
• Baseline for legacy projects — suppress the existing typos with a line-shift resistant fingerprint and fail only on the new ones.
• Seven report formats — table, JSON, GitHub Actions annotations, Checkstyle, JUnit, GitLab Code Quality and CSV.
• Custom and built-in dictionaries — technical, PHP and Symfony word lists ship with the package; add your own per project or per locale.
• Flat memory usage — the whole pipeline is lazy, so a repository with thousands of files costs the same as one file.

Requirements

• PHP ≥ 8.1 with ext-mbstring (tested up to PHP 8.5)
• nikic/php-parser 4 or 5
• No mandatory ext-intl (the ICU parser is built in)
• No mandatory system binary (the pure PHP wordlist backend always works)

Table of contents

• Why another spell checker
• Related packages
• Install
• Quick start
• Architecture
• Extending
• Tests
• Contributing
• Changelog
• License

Why

cspell does not know that messages.it.yaml must be checked in Italian, that %count% is a placeholder or that {n, plural, one {...}} has textual branches. tigitz/php-spellchecker is an excellent backend abstraction but knows nothing about the domain. This library is the missing middle layer: it turns catalogues and PHP identifiers into checkable text, accurately, and hands the words to whichever backend you have.

Related packages

Install

Quick start

Architecture

Everything is a generator: memory stays flat regardless of how many files or messages are processed.

Accurate positions

Every processor that removes text records the correspondence in an OffsetMap, and maps compose associatively. That is what makes the reported column point at the position in the original message:

Ispell pipe protocol

PipeSpeller keeps one process alive per language for the whole run. Words are sent one per line prefixed with ^, and the terminating blank line is the synchronisation point. Terse mode (!) suppresses the * lines and cuts the I/O by roughly 90%; if your binary does not emit the blank line in terse mode, construct the backend with terseMode: false.

Baseline

The fingerprint is sha256(schema | contextSeed | lowercase word | type) truncated to 16 hex characters. It deliberately excludes line and column, so moving code around does not invalidate the baseline. Renaming a file does, and that is documented behaviour.

Extending

Implement the relevant interface and pass it to the chain. In a Symfony application the bundle autoconfigures the six extension points; outside of it, compose the objects yourself as in the quick start above.

Tests

tests/smoke.php runs a subset of the suite with a hand written autoloader and no composer at all: useful to verify a checkout in a bare container.

Contributing

Bug reports, feature requests and pull requests are welcome. Read CONTRIBUTING.md first: the offset and determinism rules are not negotiable.

Security

Report vulnerabilities as described in SECURITY.md. Do not open a public issue for them.

License

Released under the MIT License.