To present different translations properly, info about language name, text direction, plural definitions and language code is needed.
Parsing language codes
While parsing translations, Weblate attempts to map language code (usually the ISO 639-1 one) from the File mask to any existing language object.
You can further adjust this mapping at project level by Language aliases.
If no exact match can be found, an attempt will be made to best fit it into an existing language. Following steps are tried:
Case insensitive lookups.
Normalizing underscores and dashes.
Looking up built-in language aliases.
Looking up by language name.
Ignoring the default country code for a given language—choosing
Should that also fail, a new language definition will be created using the
defaults (left to right text direction, one plural). The automatically created
language with code
xx_XX will be named as xx_XX (generated).
You might want to change this in the admin interface later, (see
Changing language definitions) and report it to the issue tracker (see
Contributing to Weblate), so that the proper definition can be added to the
upcoming Weblate release.
In case you see something unwanted as a language, you might want to adjust Language filter to ignore such file when parsing translations.
Changing language definitions
You can change language definitions in the languages interface
While editing, make sure all fields are correct (especially plurals and text direction), otherwise translators will be unable to properly edit those translations.
Built-in language definitions
Definitions for about 600 languages are included in Weblate and the list is extended in every release. Whenever Weblate is upgraded (more specifically whenever weblate migrate is executed, see Generic upgrade instructions) the database of languages is updated to include all language definitions shipped in Weblate.
Ambiguous language codes and macrolanguages
In many cases it is not a good idea to use macrolanguage code for a translation. The typical problematic case might be Kurdish language, which might be written in Arabic or Latin script, depending on actual variant. To get correct behavior in Weblate, it is recommended to use individual language codes only and avoid macrolanguages.
Each language consists of following fields:
Code identifying the language. Weblate prefers two letter codes as defined by ISO 639-1, but uses ISO 639-2 or ISO 639-3 codes for languages that do not have two letter code. It can also support extended codes as defined by BCP 47.
Visible name of the language. The language names included in Weblate are also being localized depending on user interface language.
Determines whether language is written right to left or left to right. This property is autodetected correctly for most of the languages.
Number of plurals used in the language.
Gettext compatible plural formula used to determine which plural form is used for given count.
Adding new translations
Changed in version 2.18: In versions prior to 2.18 the behaviour of adding new translations was file format specific.
Weblate can automatically start new translation for all of the file formats.
Some formats expect to start with an empty file and only translated strings to be included (for example Android string resources), while others expect to have all keys present (for example GNU gettext). The document-based formats (for example OpenDocument Format) start with a copy of the source document and all strings marked as needing editing. In some situations this really doesn’t depend on the format, but rather on the framework you use to handle the translation (for example with JSON files).
When Template for new translations is empty and the file format supports it, an empty file is created where new strings will be added once they are translated.
The Language code style allows you to customize language code used in generated filenames:
- Default based on the file format
Dependent on file format, for most of them POSIX is used.
- POSIX style using underscore as a separator
Typically used by gettext and related tools, produces language codes like
- POSIX style using underscore as a separator, including country code
POSIX style language code including the country code even when not necessary (for example
- BCP style using hyphen as a separator
Typically used on web platforms, produces language codes like
- BCP style using hyphen as a separator, including country code
BCP style language code including the country code even when not necessary (for example
- Android style
Only used in Android apps, produces language codes like
- Java style
Used by Java—mostly BCP with legacy codes for Chinese.
- Linux style
Locales as used by Linux, uses legacy codes for Chinese and POSIX style notation.
Additionally, any mappings defined in Language aliases are applied in reverse.
Weblate recognizes any of these when parsing translation files, the above settings only influences how new files are created.