gettext-spellchecker

gettext-spellchecker is a command line tool which checks the spell of msgid or msgstr strings in a gettext .po or .pot file.

It uses hunspell dictionaries.

It outputs the results to a CSV file or to colored output in the terminal.

Plural forms are checked; obsolete (#~) entries are skipped.

Installation

The hunspell Python module is compiled when installed, so it needs the hunspell development files first. In Debian and derivatives:

apt install libhunspell-dev
pip install gettext-spellchecker

From source code

git clone https://github.com/rgglez/gettext-spellchecker.git
cd gettext-spellchecker
pip install -e .

Execution

gettext-spellchecker --lang=de_DE -i examples/example_simple.po --check=msgstr

The exit code is 0 when no errors are found, 1 when possible spelling errors are found, and 2 on invalid arguments (for instance, a missing dictionary), so the tool can be used in CI pipelines or pre-commit hooks.

Command line options

• --version prints the version.
• --path the path to hunspell dictionaries. Default: "/usr/share/hunspell".
• --lang the name of the hunspell dictionary of the language of the strings, without extension. Usually in the form xx_YY where xx is in ISO 639-1 format and YY in ISO 3166-1 alpha-2 format, but any dictionary present in --path works (e.g. de_DE_frami).
• -i, --input the path of the .po or .pot file.
• -o, --output optional path of the output file in CSV format.
• --check which string do you want to check: 'msgid' or 'msgstr'. Default="msgid".

Dictionaries

This tool uses hunspell dictionaries.

In Debian and derivatives you install them with apt, for instance:

apt install hunspell-fr

Read your operating system's documentation to know how to install those packages from its repositories, or visit hunspell if you need to compile them by hand.

Development

make dev     # editable install of the package and the dev tools
make check   # lint, tests, build and twine check

The tests need the de_DE and en_US hunspell dictionaries (apt install hunspell-de-de hunspell-en-us).

Releases are published to PyPI by the Publish GitHub workflow when a vX.Y.Z tag is pushed (make release). Run make help for the other targets.

Notes

• Sample .po and .pot files are included in the examples directory.
• Why gettext?
  • First and most relevant reason: it uses the full strings in the original language as key, so I don't have to be searching for weird keys such as "page.title.hello" or "item.specification". If one translation doesn't exist, the original key string is used.
  • It's a GNU standard, tried and trusted.

License

Copyright 2026 Rodolfo González González.

Apache License 2.0. Please read the LICENSE file.