Integrating with Weblate¶

Weblate basics¶

Project and component structure¶

In Weblate translations are organized into projects and components. Each project can contain number of components and those contain translations into individual languages. The component corresponds to one translatable file (for example GNU gettext PO (Portable Object) or Android string resources). The projects are there to help you organize component into logical sets (for example to group all translations used within one application).

Additionally, components within projects can be structured using categories. Components can belong to a category, and categories can be nested.

Internally, each project has translations to common strings propagated across other components within it by default. This lightens the burden of repetitive and multi version translation. The translation propagation can be disabled per Component configuration using Allow translation propagation in case the translations should diverge.

Repository integration¶

Weblate is built to integrate with upstream version control repository, Continuous localization describes building blocks and how the changes flow between them.

See also

Architecture overview describes how Weblate works internally.

User attribution¶

Weblate keeps the translations properly authored by translators in the version control repository by using name and e-mail. Having a real e-mail attached to the commit follows the distributed version control spirits and allows services like GitHub to associate your contributions done in Weblate with your GitHub profile.

This feature also brings in risk of misusing e-mail published in the version control commits. Moreover, once such a commit is published on public hosting (such as GitHub), there is effectively no way to redact it. Weblate allows choosing a private commit e-mail in Account to avoid this.

Therefore, admins should consider this while configuring Weblate:

Such a usage of e-mail should be clearly described in service terms in case such document is needed. Legal can help with that.
PRIVATE_COMMIT_EMAIL_OPT_IN can make e-mails private by default.

Importing a localization project into Weblate¶

Weblate has been developed with VCS integration in mind as it’s core feature, so the easiest way is to grant Weblate the access to your repository. The import process will guide you through configuring your translations into Weblate components.

Alternatively, you can let Weblate set up a local-only repository containing all the translations without integration.

Fetching updated translations from Weblate¶

Weblate stores updated strings in a database and commits them to a local version-control repository. You can add a Weblate repository (when Git exporter is turned on) as an additional remote repository and fetch translation updates from it.

Prior to this, you might want to commit any pending local changes made in Weblate (see Lazy commits). This can be done from the user interface (in the Repository maintenance), or from the command-line using Weblate Client.

Pushing changes can be automated if you grant Weblate push access to your repository and configure Repository push URL in the Component configuration, see Pushing changes from Weblate.

Alternatively, use Weblate’s REST API to update translations so that they match the latest version from the upstream in your remote VCS repository.

Fetching remote changes into Weblate¶

To fetch any strings recently updated in your remote VCS repository into Weblate, allow Weblate to pull from the upstream repository. This can be achieved in the user interface (in the Repository maintenance), or from the command-line using Weblate Client.

This can be automated by setting a webhook in your repository to trigger Weblate whenever there is a new commit. See Updating repositories for more details.

If not using VCS integration, you can use the UI or Weblate’s REST API to update the translations so that they match your codebase.

Adding new strings¶

If your translation files are stored in a remote VCS together with the code, you most likely have an existing workflow for developers to introduce new strings. Any way of adding strings will be picked up, but consider using Quality gateway for the source strings to avoid introducing errors.

When translation files are separated from the code, the following ways can introduce new strings into Weblate.

Manually, using Add new translation string from Tools menu in the source language. You can choose between the radio buttons Singular and Plural inside the form. Select the appropriate form of the new translation string to be added.
Programmatically, using the API POST /api/translations/(string:project)/(string:component)/(string:language)/units/.
By uploading source file as Replace existing translation file (this overwrites existing strings, so please ensure the file includes both old and new strings) or Add new strings, see Import methods.

Note

The ability to add strings in Weblate requires Manage strings.

Updating target-language files¶

For monolingual files (see Supported file formats), Weblate might add new translation strings present in the Monolingual base language file, and not in actual translations. It does not however perform any automatic cleanup of stale strings, as that might have unexpected results. If you still want to do this, please install the Cleanup translation files add-on, which handles cleanup according to your requirements.

Weblate will also not try to update bilingual files when the source changes, so if you need po files to be updated from pot, do it yourself by using Update source strings Import methods, or by using the Update PO files to match POT (msgmerge) add-on.

Hint

Source string extraction tools, such as xgettext or lupdate, need to be executed outside of Weblate.

Introducing new strings¶

You can add new strings in Weblate with Manage strings turned on, but it is usually better to introduce new strings together with the code changes that introduced them.

Monolingual formats need to be configured so that new strings are added to Monolingual base language file. This is typically done by developers, as they write the code. You might want to use a review process of those strings using Quality gateway for the source strings.

Bilingual formats typically extract strings from the source code using some tooling (like xgettext or intltool-update). Follow your localization framework documentation for instructions how to do that. Once the strings are extracted, there might be an additional step needed to update existing translations, see Updating target-language files.

Hint

Automating string extraction is presently out of scope for Weblate. It typically involves executing untrusted code what makes it more suitable for a generic continuous integration than localization-specific platform.

You might want to integrate this into your continuous integration pipelines to make new strings automatically appear for translation. Such pipeline should also cover Avoiding merge conflicts.

Managing the local VCS repository¶

Weblate stores all translations in its underlying version control repository. It is suggested to be connected to a remote one, but internal-only setup is also possible. The Repository maintenance allows controlling this repository.

Hint

With Continuous localization, any changes are automatically pushed from the repository, so there is usually no need to manually manage it manually.