Internationalization and Localization

Internationalization (i18n) is the act of creating software with a user interface that can potentially be displayed in more than one language or cultural context. Localization (l10n) is the process of displaying the user interface of an internationalized application in a particular language or cultural context.

Pyramid offers internationalization and localization subsystems that can be used to translate the text of buttons, error messages, and other software- and template-defined values into the native language of a user of your application.

Creating a Translation String

While you write your software, you can insert specialized markup into your Python code that makes it possible for the system to translate text values into the languages used by your application's users. This markup creates a translation string. A translation string is an object that behaves mostly like a normal Unicode string, except that it also carries around extra information related to its job as part of the Pyramid translation machinery.

Using the TranslationString Class

The most primitive way to create a translation string is to use the pyramid.i18n.TranslationString callable:

from pyramid.i18n import TranslationString
ts = TranslationString('Add')

This creates a str-like object that is a TranslationString.

Note

For people more familiar with Zope i18n, a TranslationString is a lot like a zope.i18nmessageid.Message object. It is not a subclass, however. For people more familiar with Pylons or Django i18n, using a TranslationString is a lot like using "lazy" versions of related gettext APIs.

The first argument to TranslationString is the msgid; it is required. It represents the key into the translation mappings provided by a particular localization. The msgid argument must be a string. The msgid may optionally contain replacement markers. For instance:

from pyramid.i18n import TranslationString
ts = TranslationString('Add ${number}')

Within the string above, ${number} is a replacement marker. It will be replaced by whatever is in the mapping for a translation string. The mapping may be supplied at the same time as the replacement marker itself:

from pyramid.i18n import TranslationString
ts = TranslationString('Add ${number}', mapping={'number':1})

Any number of replacement markers can be present in the msgid value, any number of times. Only markers which can be replaced by the values in the mapping will be replaced at translation time. The others will not be interpolated and will be output literally.

A translation string should also usually carry a domain. The domain represents a translation category to disambiguate it from other translations of the same msgid, in case they conflict.

from pyramid.i18n import TranslationString
ts = TranslationString('Add ${number}', mapping={'number':1},
                       domain='form')

The above translation string named a domain of form. A translator function will often use the domain to locate the right translator file on the filesystem which contains translations for a given domain. In this case, if it were trying to translate our msgid to German, it might try to find a translation from a gettext file within a translation directory like this one:

locale/de/LC_MESSAGES/form.mo

In other words, it would want to take translations from the form.mo translation file in the German language.

Finally, the TranslationString constructor accepts a default argument. If a default argument is supplied, it replaces usages of the msgid as the default value for the translation string. When default is None, the msgid value passed to a TranslationString is used as an implicit message identifier. Message identifiers are matched with translations in translation files, so it is often useful to create translation strings with "opaque" message identifiers unrelated to their default text:

from pyramid.i18n import TranslationString
ts = TranslationString('add-number', default='Add ${number}',
                       domain='form', mapping={'number':1})

When default text is used, Default text objects may contain replacement values.

Using the TranslationStringFactory Class

Another way to generate a translation string is to use the TranslationStringFactory object. This object is a translation string factory. Basically a translation string factory presets the domain value of any translation string generated by using it. For example:

from pyramid.i18n import TranslationStringFactory
_ = TranslationStringFactory('pyramid')
ts = _('add-number', default='Add ${number}', mapping={'number':1})

Note

We assigned the translation string factory to the name _. This is a convention which will be supported by translation file generation tools.

After assigning _ to the result of a TranslationStringFactory(), the subsequent result of calling _ will be a TranslationString instance. Even though a domain value was not passed to _ (as would have been necessary if the TranslationString constructor were used instead of a translation string factory), the domain attribute of the resulting translation string will be pyramid. As a result, the previous code example is completely equivalent (except for spelling) to:

from pyramid.i18n import TranslationString as _
ts = _('add-number', default='Add ${number}', mapping={'number':1},
       domain='pyramid')

You can set up your own translation string factory much like the one provided above by using the TranslationStringFactory class. For example, if you'd like to create a translation string factory which presets the domain value of generated translation strings to form, you'd do something like this:

from pyramid.i18n import TranslationStringFactory
_ = TranslationStringFactory('form')
ts = _('add-number', default='Add ${number}', mapping={'number':1})

Creating a unique domain for your application via a translation string factory is best practice. Using your own unique translation domain allows another person to reuse your application without needing to merge your translation files with their own. Instead they can just include your package's translation directory via the pyramid.config.Configurator.add_translation_dirs() method.

Note

For people familiar with Zope internationalization, a TranslationStringFactory is a lot like a zope.i18nmessageid.MessageFactory object. It is not a subclass, however.

Working with gettext Translation Files

The basis of Pyramid translation services is GNU gettext. Once your application source code files and templates are marked up with translation markers, you can work on translations by creating various kinds of gettext files.

Note

The steps a developer must take to work with gettext message catalog files within a Pyramid application are very similar to the steps a Pylons developer must take to do the same. See the Pylons Internationalization and Localization documentation for more information.

GNU gettext uses three types of files in the translation framework, .pot files, .po files, and .mo files.

.pot (Portable Object Template) files

A .pot file is created by a program which searches through your project's source code and which picks out every message identifier passed to one of the _() functions (e.g., translation string constructions). The list of all message identifiers is placed into a .pot file, which serves as a template for creating .po files.

.po (Portable Object) files

The list of messages in a .pot file are translated by a human to a particular language; the result is saved as a .po file.

.mo (Machine Object) files

A .po file is turned into a machine-readable binary file, which is the .mo file. Compiling the translations to machine code makes the localized program start faster.

The tools for working with gettext translation files related to a Pyramid application are Lingua and Gettext. Lingua can scrape i18n references out of Python and Chameleon files and create the .pot file. Gettext includes msgmerge tool to update a .po file from an updated .pot file and msgfmt to compile .po files to .mo files.

Installing Lingua and Gettext

In order for the commands related to working with gettext translation files to work properly, you will need to have Lingua and Gettext installed into the same environment in which Pyramid is installed.

Installation on Unix

Gettext is often already installed on Unix systems. You can check if it is installed by testing if the msgfmt command is available. If it is not available you can install it through the packaging system from your OS; the package name is almost always gettext. For example on a Debian or Ubuntu system run this command:

sudo apt-get install gettext

Installing Lingua is done with the Python packaging tools. If the virtual environment into which you've installed your Pyramid application lives at the environment variable $VENV, you can install Lingua like so:

$VENV/bin/pip install lingua

Installation on Windows

There are several ways to install Gettext on Windows: it is included in the Cygwin collection, or you can use the installer from the GnuWin32, or compile it yourself. Make sure the installation path is added to your $PATH.

Installing Lingua is done with the Python packaging tools. If the virtual environment into which you've installed your Pyramid application lives at the environment variable %VENV%, you can install Lingua like so:

%VENV%\Scripts\pip install lingua

Extracting Messages from Code and Templates

Once Lingua is installed, you may extract a message catalog template from the code and Chameleon templates which reside in your Pyramid application. You run a pot-create command to extract the messages:

cd /file/path/to/myapplication_setup.py
mkdir -p myapplication/locale
$VENV/bin/pot-create -o myapplication/locale/myapplication.pot src

The message catalog .pot template will end up in myapplication/locale/myapplication.pot.

Initializing a Message Catalog File

Once you've extracted messages into a .pot file (see Extracting Messages from Code and Templates), to begin localizing the messages present in the .pot file, you need to generate at least one .po file. A .po file represents translations of a particular set of messages to a particular locale. Initialize a .po file for a specific locale from a pre-generated .pot template by using the msginit command from Gettext:

cd /file/path/to/myapplication_setup.py
cd myapplication/locale
mkdir -p es/LC_MESSAGES
msginit -l es -o es/LC_MESSAGES/myapplication.po

This will create a new message catalog .po file in myapplication/locale/es/LC_MESSAGES/myapplication.po.

Once the file is there, it can be worked on by a human translator. One tool which may help with this is Poedit.

Note that Pyramid itself ignores the existence of all .po files. For a running application to have translations available, a .mo file must exist. See Compiling a Message Catalog File.

Updating a Catalog File

If more translation strings are added to your application, or translation strings change, you will need to update existing .po files based on changes to the .pot file, so that the new and changed messages can also be translated or re-translated.

First, regenerate the .pot file as per Extracting Messages from Code and Templates. Then use the msgmerge command from Gettext.

cd /file/path/to/myapplication_setup.py
cd myapplication/locale
msgmerge --update es/LC_MESSAGES/myapplication.po myapplication.pot

Compiling a Message Catalog File

Finally, to prepare an application for performing actual runtime translations, compile .po files to .mo files using the msgfmt command from Gettext:

cd /file/path/to/myapplication_setup.py
msgfmt -o myapplication/locale/es/LC_MESSAGES/myapplication.mo \
    myapplication/locale/es/LC_MESSAGES/myapplication.po

This will create a .mo file for each .po file in your application. As long as the translation directory in which the .mo file ends up in is configured into your application (see Adding a Translation Directory), these translations will be available to Pyramid.

Using a Localizer

A localizer is an object that allows you to perform translation or pluralization "by hand" in an application. You may use the pyramid.request.Request.localizer attribute to obtain a localizer. The localizer object will be configured to produce translations implied by the active locale negotiator, or a default localizer object if no explicit locale negotiator is registered.

def aview(request):
    localizer = request.localizer

Note

If you need to create a localizer for a locale, use the pyramid.i18n.make_localizer() function.

Performing a Translation

A localizer has a translate method which accepts either a translation string or a Unicode string and which returns a Unicode string representing the translation. Generating a translation in a view component of an application might look like so:

from pyramid.i18n import TranslationString

ts = TranslationString('Add ${number}', mapping={'number':1},
                       domain='pyramid')

def aview(request):
    localizer = request.localizer
    translated = localizer.translate(ts) # translation string
    # ... use translated ...

The request.localizer attribute will be a pyramid.i18n.Localizer object bound to the locale name represented by the request. The translation returned from its pyramid.i18n.Localizer.translate() method will depend on the domain attribute of the provided translation string as well as the locale of the localizer.

Note

If you're using Chameleon templates, you don't need to pre-translate translation strings this way. See Chameleon Template Support for Translation Strings.

Performing a Pluralization

A localizer has a pluralize method with the following signature:

def pluralize(singular, plural, n, domain=None, mapping=None):
    # ...

The simplest case is the singular and plural arguments being passed as Unicode literals. This returns the appropriate literal according to the locale pluralization rules for the number n, and interpolates mapping.

def aview(request):
    localizer = request.localizer
    translated = localizer.pluralize('Item', 'Items', 1, 'mydomain')
    # ... use translated ...

However, for support of other languages, the singular argument should be a Unicode value representing a message identifier. In this case the plural value is ignored. domain should be a translation domain, and mapping should be a dictionary that is used for replacement value interpolation of the translated string.

The value of n will be used to find the appropriate plural form for the current language, and pluralize will return a Unicode translation for the message id singular. The message file must have defined singular as a translation with plural forms.

The argument provided as singular may be a translation string object, but the domain and mapping information attached is ignored.

def aview(request):
    localizer = request.localizer
    num = 1
    translated = localizer.pluralize('item_plural', '${number} items',
        num, 'mydomain', mapping={'number':num})

The corresponding message catalog must have language plural definitions and plural alternatives set.

"Plural-Forms: nplurals=3; plural=n==0 ? 0 : n==1 ? 1 : 2;"

msgid "item_plural"
msgid_plural ""
msgstr[0] "No items"
msgstr[1] "${number} item"
msgstr[2] "${number} items"

More information on complex plurals can be found in the gettext documentation.

Obtaining the Locale Name for a Request

You can obtain the locale name related to a request by using the pyramid.request.Request.locale_name() attribute of the request.

def aview(request):
    locale_name = request.locale_name

The locale name of a request is dynamically computed; it will be the locale name negotiated by the currently active locale negotiator, or the default locale name if the locale negotiator returns None. You can change the default locale name by changing the pyramid.default_locale_name setting. See Default Locale Name.

Once locale_name() is first run, the locale name is stored on the request object. Subsequent calls to locale_name() will return the stored locale name without invoking the locale negotiator. To avoid this caching, you can use the pyramid.i18n.negotiate_locale_name() function:

from pyramid.i18n import negotiate_locale_name

def aview(request):
    locale_name = negotiate_locale_name(request)

You can also obtain the locale name related to a request using the locale_name attribute of a localizer.

def aview(request):
    localizer = request.localizer
    locale_name = localizer.locale_name

Obtaining the locale name as an attribute of a localizer is equivalent to obtaining a locale name by asking for the locale_name() attribute.

Performing Date Formatting and Currency Formatting

Pyramid does not itself perform date and currency formatting for different locales. However, Babel can help you do this via the babel.core.Locale class. The Babel documentation for this class provides minimal information about how to perform date and currency related locale operations. See Installing Lingua and Gettext for information about how to install Babel.

The babel.core.Locale class requires a locale name as an argument to its constructor. You can use Pyramid APIs to obtain the locale name for a request to pass to the babel.core.Locale constructor. See Obtaining the Locale Name for a Request. For example:

from babel.core import Locale

def aview(request):
    locale_name = request.locale_name
    locale = Locale(locale_name)

Chameleon Template Support for Translation Strings

When a translation string is used as the subject of textual rendering by a Chameleon template renderer, it will automatically be translated to the requesting user's language if a suitable translation exists. This is true of both the ZPT and text variants of the Chameleon template renderers.

For example, in a Chameleon ZPT template, the translation string represented by "some_translation_string" in each example below will go through translation before being rendered:

<span tal:content="some_translation_string"/>

<span tal:replace="some_translation_string"/>

<span>${some_translation_string}</span>

<a tal:attributes="href some_translation_string">Click here</a>

The features represented by attributes of the i18n namespace of Chameleon will also consult the Pyramid translations. See https://chameleon.readthedocs.io/en/latest/reference.html#translation-i18n.

Note

Unlike when Chameleon is used outside of Pyramid, when it is used within Pyramid, it does not support use of the zope.i18n translation framework. Applications which use Pyramid should use the features documented in this chapter rather than zope.i18n.

Third party Pyramid template renderers might not provide this support out of the box and may need special code to do an equivalent. For those, you can always use the more manual translation facility described in Performing a Translation.

Mako Pyramid i18n Support

There exists a recipe within the Pyramid Community Cookbook named Mako Internationalization which explains how to add idiomatic i18n support to Mako templates.

Jinja2 Pyramid i18n Support

The add-on pyramid_jinja2 provides a scaffold with an example of how to use internationalization with Jinja2 in Pyramid. See the documentation sections Internalization (i18n) and pcreate template i18n.

Localization-Related Deployment Settings

A Pyramid application will have a pyramid.default_locale_name setting. This value represents the default locale name used when the locale negotiator returns None. Pass it to the Configurator constructor at startup time:

from pyramid.config import Configurator
config = Configurator(settings={'pyramid.default_locale_name':'de'})

You may alternately supply a pyramid.default_locale_name via an application's .ini file:

[app:main]
use = egg:MyProject
pyramid.reload_templates = true
pyramid.debug_authorization = false
pyramid.debug_notfound = false
pyramid.default_locale_name = de

If this value is not supplied via the Configurator constructor or via a config file, it will default to en.

If this setting is supplied within the Pyramid application .ini file, it will be available as a settings key:

from pyramid.threadlocal import get_current_registry
settings = get_current_registry().settings
default_locale_name = settings['pyramid.default_locale_name']

"Detecting" Available Languages

Other systems provide an API that returns the set of "available languages" as indicated by the union of all languages in all translation directories on disk at the time of the call to the API.

It is by design that Pyramid doesn't supply such an API. Instead the application itself is responsible for knowing the "available languages". The rationale is this: any particular application deployment must always know which languages it should be translatable to anyway, regardless of which translation files are on disk.

Here's why: it's not a given that because translations exist in a particular language within the registered set of translation directories that this particular deployment wants to allow translation to that language. For example, some translations may exist but they may be incomplete or incorrect. Or there may be translations to a language but not for all translation domains.

Any nontrivial application deployment will always need to be able to selectively choose to allow only some languages even if that set of languages is smaller than all those detected within registered translation directories. The easiest way to allow for this is to make the application entirely responsible for knowing which languages are allowed to be translated to instead of relying on the framework to divine this information from translation directory file info.

You can set up a system to allow a deployer to select available languages based on convention by using the pyramid.settings mechanism.

Allow a deployer to modify your application's .ini file:

[app:main]
use = egg:MyProject
# ...
available_languages = fr de en ru

Then as a part of the code of a custom locale negotiator:

from pyramid.settings import aslist

def my_locale_negotiator(request):
    languages = aslist(request.registry.settings['available_languages'])
    # ...

This is only a suggestion. You can create your own "available languages" configuration scheme as necessary.

Activating Translation

By default, a Pyramid application performs no translation. To turn translation on you must:

• add at least one translation directory to your application.

• ensure that your application sets the locale name correctly.

Adding a Translation Directory

gettext is the underlying machinery behind the Pyramid translation machinery. A translation directory is a directory organized to be useful to gettext. A translation directory usually includes a listing of language directories, each of which itself includes an LC_MESSAGES directory. Each LC_MESSAGES directory should contain one or more .mo files. Each .mo file represents a message catalog, which is used to provide translations to your application.

Adding a translation directory registers all of its constituent message catalog files within your Pyramid application to be available to use for translation services. This includes all of the .mo files found within all LC_MESSAGES directories within each locale directory in the translation directory.

You can add a translation directory imperatively by using the pyramid.config.Configurator.add_translation_dirs() during application startup. For example:

from pyramid.config import Configurator
config.add_translation_dirs('my.application:locale/',
                            'another.application:locale/')

A message catalog in a translation directory added via add_translation_dirs() will be merged into translations from a message catalog added earlier if both translation directories contain translations for the same locale and translation domain.

Setting the Locale

When the default locale negotiator (see The Default Locale Negotiator) is in use, you can inform Pyramid of the current locale name by doing any of these things before any translations need to be performed:

• Set the _LOCALE_ attribute of the request to a valid locale name (usually directly within view code), e.g., request._LOCALE_ = 'de'.

• Ensure that a valid locale name value is in the request.params dictionary under the key named _LOCALE_. This is usually the result of passing a _LOCALE_ value in the query string or in the body of a form post associated with a request. For example, visiting http://my.application?_LOCALE_=de.

• Ensure that a valid locale name value is in the request.cookies dictionary under the key named _LOCALE_. This is usually the result of setting a _LOCALE_ cookie in a prior response, e.g., response.set_cookie('_LOCALE_', 'de').

Note

If this locale negotiation scheme is inappropriate for a particular application, you can configure a custom locale negotiator function into that application as required. See Using a Custom Locale Negotiator.

Locale Negotiators

A locale negotiator informs the operation of a localizer by telling it what locale name is related to a particular request. A locale negotiator is a bit of code which accepts a request and which returns a locale name. It is consulted when pyramid.i18n.Localizer.translate() or pyramid.i18n.Localizer.pluralize() is invoked. It is also consulted when locale_name() is accessed or when negotiate_locale_name() is invoked.

The Default Locale Negotiator

Most applications can make use of the default locale negotiator, which requires no additional coding or configuration.

The default locale negotiator implementation named default_locale_negotiator uses the following set of steps to determine the locale name.

• First the negotiator looks for the _LOCALE_ attribute of the request object (possibly set directly by view code or by a listener for an event).

• Then it looks for the request.params['_LOCALE_'] value.

• Then it looks for the request.cookies['_LOCALE_'] value.

• If no locale can be found via the request, it falls back to using the default locale name (see Localization-Related Deployment Settings).

• Finally if the default locale name is not explicitly set, it uses the locale name en.

Using a Custom Locale Negotiator

Locale negotiation is sometimes policy-laden and complex. If the (simple) default locale negotiation scheme described in Activating Translation is inappropriate for your application, you may create a special locale negotiator. Subsequently you may override the default locale negotiator by adding your newly created locale negotiator to your application's configuration.

A locale negotiator is simply a callable which accepts a request and returns a single locale name or None if no locale can be determined.

Here's an implementation of a simple locale negotiator:

def my_locale_negotiator(request):
    locale_name = request.params.get('my_locale')
    return locale_name

If a locale negotiator returns None, it signifies to Pyramid that the default application locale name should be used.

You may add your newly created locale negotiator to your application's configuration by passing an object which can act as the negotiator (or a dotted Python name referring to the object) as the locale_negotiator argument of the Configurator instance during application startup. For example:

from pyramid.config import Configurator
config = Configurator(locale_negotiator=my_locale_negotiator)

Alternatively, use the pyramid.config.Configurator.set_locale_negotiator() method.

For example:

from pyramid.config import Configurator
config = Configurator()
config.set_locale_negotiator(my_locale_negotiator)