Internationalization#

Warning

This theme is still in the process of setting up internationalization. Some of the text below may not yet be correct (for example, we do not yet have a locales/ directory). Follow these issues to track progress:

Internationalization (I18N) and localization (L10N) is performed using Gettext.

Types of files#

Gettext reads a program’s source and extracts text that has been marked as translatable, known as “source strings. Gettext uses three types of files:

PO file (.po)

A Portable Object (PO) file is made up of many entries. Each entry holds the relation between a source string and its translation. msgid contains the source string, and msgstr contains the translation. In a given PO file, all translations are expressed in a single target language. PO files are also known as “message catalogs”.

Entries begin with comments, on lines starting with the character #. Comments are created and maintained by Gettext. Comment lines starting with #: contain references to the program’s source. These references allow a human translator to find the source strings in their original context. Comment lines starting with #, contain flags like python-format, which indicates that the source string contains placeholders like %(copyright)s.

POT file (.pot)

A Portable Object Template (POT) file is the same as a PO file, except the translations are empty, so that it can be used as a template for new languages.

MO file (.mo)

A Machine Object (MO) file is a binary version of a PO file. PO files are compiled to MO files, which are required by Gettext.

Mark natural language text as translateable#

All natural language text must be marked as translatable, so that it can be extracted by Gettext and translated by humans.

Jinja2 provides a trans block and a _() function to mark text as translatable. Please refer to the Jinja2 documentation. Remember to manually escape variables if needed.

Any text that is marked in this way will be discoverable by gettext and used to generate .po files (see below for information). Once you’ve marked text as translateable, complete the steps for Add or change natural language text.

Add or change natural language text#

These steps cover how to add or change text that has been marked as translateable.

  1. Edit the natural language text as desired. Ensure that it is {ref}`marked as translateable <adding-natural-language-text>`.

  2. Generate/update the message catalog template (POT file) with the PyBabel extract command:

    pybabel extract . -F babel.cfg -o src/pydata_sphinx_theme/locale/sphinx.pot -k '_ __ l_ lazy_gettext'
    

    To run this in ``.nox``: nox -s translate -- extract.

  3. Update the message catalogs (PO files) with the PyBabel update command:

    pybabel update -i src/pydata_sphinx_theme/locale/sphinx.pot -d src/pydata_sphinx_theme/locale -D sphinx
    

    To run this in ``.nox``: nox -s translate -- update.

This will update these files with new information about the position and text of the language you have modified.

If you only change non-translatable text (like HTML markup), the extract and update commands will only update the positions (line numbers) of the translatable strings. Updating positions is optional - the line numbers are to inform the human translator, not to perform the translation.

If you change translatable strings, the extract command will extract the new or updated strings to the POT file, and the update command will try to fuzzy match the new or updated strings with existing translations in the PO files. If there is a fuzzy match, a comment like #, fuzzy is added before the matched entry. Otherwise, a new entry is added and needs to be translated.

Add translations to translateable text#

Once text has been marked as translateable, and PO files have been generated for it, we may add translations for new languages for the phrase. This section covers how to do so.

Note

These steps use the Spanish language as an example. To translate the theme to another language, replace es with the language’s two-letter lowercase ISO 639-1 code.

  1. If the language’s code matches no sub-directory of the pydata_sphinx_theme/locale directory, initialize the language’s message catalog (PO file) with PyBabel init:

    pybabel init -i src/pydata_sphinx_theme/locale/sphinx.pot -d src/pydata_sphinx_theme/locale -D sphinx -l es
    

    To run this in ``.nox``: nox -s translate -- init es

  2. Edit the language’s message catalog at pydata_sphinx_theme/locale/es/LC_MESSAGES/sphinx.po. For each source string introduced by the msgid keyword, add its translation after the msgstr keyword.

  3. Compile the message catalogs of every language. This creates or updates the MO files with PyBabel compile:

    pybabel compile -d src/pydata_sphinx_theme/locale -D sphinx
    

    To run this in ``.nox``: nox -s translate -- compile.

Translation tips#

Translate phrases, not words#

Full sentences and clauses must always be a single translatable string. Otherwise, you can get next page translated as suivant page instead of as page suivante, etc.

Deal with variables and markup in translations#

If a variable (like the edit_page_provider_name theme option) is used as part of a phrase, it must be included within the translatable string. Otherwise, the word order in other languages can be incorrect. In a Jinja template, simply surround the translatable string with {% trans variable=variable %} and {% endtrans %}}`. For example: ``{% trans provider=provider %}Edit on {{ provider }}{% endtrans %} The translatable string is extracted as the Python format string Edit on %(provider)s. This is so that the same translatable string can be used in both Python code and Jinja templates. It is the translator’s responsibility to use %(provider)s verbatim in the translation.

If a non-translatable word or token (like HTML markup) is used as part of a phrase, it must also be included within the translatable string. For example: {% trans theme_version=theme_version|e %}Built with the <a href="https://pydata-sphinx-theme.readthedocs.io/en/stable/index.html">PyData Sphinx Theme</a> {{ theme_version }}.{% endtrans %} It is the translator’s responsibility to use the HTML markup verbatim in the translation.

References#

I18N and L10N are deep topics. Here, we only cover the bare minimum needed to fulfill basics technical tasks. You might like: