That's weird...

Using gettext in Django

· Apostolis Bessas

Internationalization (i18n) is the process of enabling your software to be translated to other languages. This is done by making the software able to map all user-visible strings to their translations in a specific language and use them, when appropriate.

The i18n toolset django uses is gettext and is enabled with the USE_I18N setting. Gettext works as follows: First, the developer marks the translation strings in the source code and then he runs the xgettext command to create the source PO file. The xgettext command analyzes the source code to locate the strings that were marked as translation strings and extracts them into a PO file, which can be translated to the target languages. The last step is to compile the PO files into binary message files (.mo files) with the msgfmt command.

Django makes sure to ease the process by providing its own management commands: makemessages that extracts the strings from the source code and compilemessages that converts the files into the binary ones. It also expects to find the PO files for a language under the directory locale/<locale>/LC_MESSAGES/django.po, where <locale> is the language code.

In the runtime, whenever a gettext function is used, it will use the PO file that corresponds to the currently active locale and return the correct translation that corresponds to this string. If no translation is found, the original string is used instead.

The gettext functions

The following is a list of the most important gettext functions that django exposes and can support unicode strings correctly. Keep in mind that they all mark the containing literal string as translatable, that is, you need to pass the string itself as an argument, not a variable.

  • ugettext: The function returns the translation for the currently selected language.
  • ugettext_lazy: The function marks the string as translation string, but only fetches the translated string, when it is used in a string context, such as when rendering a template.
  • ugettext_noop: This function only marks a string as translation string, it does not have any other effect; that is, it always returns the string itself.

The ugettext_lazy function is useful in cases, where a string needs to be marked as translation string, but the user’s locale is not active yet. For instance, all strings in a model or in the settings are loaded, when the django process is started. During startup, there is no user request that sets the active locale. By using the ugettext_lazy function, you ensure that the translations will be actually fetched only during a user request, when the locale is known.

Given the ugettext_lazy function, the ugettext_noop function seems to be redundant. However, it is really useful in cases that you want to mark a string as translation string, but need to use it in other contexts in its original form as well, such as an error message that has to be logged, too. For instance, take the following view that returns an error message to the user (it is a standard practice to import the gettext functions as “_”):

import logging
from django.http import HttpResponse
from django.utils.translation import ugettext as _

def view(request):
    msg = _("An error has occurred")
    logging.error(msg)
    return HttpResponse(msg)

The response will contain the translated error message, but so will the logs. In order to circumvent this issue, you can use the ugettext_noop function instead:

import logging
from django.http import HttpResponse
from django.utils.translation import ugettext as _, ugettext_noop as _noop

def view(request):
    msg = _noop("An error has occurred")
    logging.error(msg)
    return HttpResponse(_(msg))

The ugettext_noop will force gettext to mark the error message as translation string, but it will return the original (English) string. As a result, the log messages will use the English phrase. However, we still want to present the translated message to the user. We achieve this by using the ugettext function to force evaluating the string and fetching the translated message. This technique is especially useful for exception messages, that need to be logged in a higher layer of the codebase but also presented to the user.

Plural support

Most languages have two plural forms: singular and plural. Some languages, however, have only one, like Japanese. Others can have up to six, like Arabic.

Which plural form to use depends on the number of objects the phrase refers to. For instance, English uses the singular form, when the phrase refers to one object (“1 language”), and the plural, when it refers to none (“no languages”) or more than one (“2 languages”). Other languages have different rules.

Django supports plurals with the n family of gettext functions: ungettext and ungettext_lazy.

Context support

Gettext allows developers to define the context of a string. For instance, the word “Read” can mean “read an email” or “emails read”. The developer can define and explain the context of a string to differentiate between the two uses. This way, gettext will create separate entries for the word “read”, one for each context.

This is supported in django with the p family of gettext functions: pgettext, pgettext_lazy, npgettext and npgettext_lazy.

Support in templates

Django provides two templatetags for marking translation strings in templates: trans and blocktrans. The main difference is that the trans templatetag only supports simple strings, while the blocktrans templatetag supports variable substitution (placeholders) as well.

Both templatetags support most of the gettext features. One thing to keep in mind for the blocktrans templatetag, though, is that the containing string is extracted as is. This means that the string will be extracted by gettext with any newline and space characters or indentation it might have. For this reason, all indentation should be avoided.

More details for how to internationalize your web application can be found at django’s documentation.