Guidelines for Use In OpenStack

The OpenStack I18N team has a limited capacity to translate messages, so we want to make their work as effective as possible by identifying the most useful text for them to translate. All text messages the user sees via exceptions or API calls should be marked for translation. However, some exceptions are used internally to signal error conditions between modules and are not intended to be presented to the user. Those do not need to be translated.

Gettext Contextual Form and Plural Form

Sometimes under different contexts, the same word should be translated into different phrases using TranslatorFactory.contextual_form.

And recommend the following code to use contextual form:

# The contextual translation function using the name "_C"
_C = _translators.contextual_form

...
msg = _C('context', 'string')

In some languages, sometimes the translated strings are different with different item counts using TranslatorFactory.plural_form

And recommend the following code to use plural form:

# The plural translation function using the name "_P"
_P = _translators.plural_form

...
msg = _P('single', 'plural', count)

The contextual form and plural form are used only when needed. By default, the translation should use the _().

Note

These two functions were only available in oslo.i18n >= 2.1.0.

Log Translation

OpenStack supports translating some log levels using separate message catalogs, and so has separate marker functions. These well-known names are used by the build system jobs that extract the messages from the source code and pass it to the translation tool.

Level Function
INFO _LI()
WARNING _LW()
ERROR _LE()
CRITICAL _LC()

Note

  • Debug level log messages are not translated.
  • LOG.exception creates an ERROR level log, so when a marker function is used (see below) _LE() should be used.

Using a Marker Function

The marker functions are used to mark the translatable strings in the code. The strings are extracted into catalogs using a tool that looks for these specific markers, so the function argument must just be a string.

For example: do not do this:

# WRONG
msg = _(variable_containing_msg)
w_msg = _LW(variable_warning_msg)

Instead, use this style:

# RIGHT
msg = _('My message.')
w_msg = _LW('My warning message')

Choosing a Marker Function

The purpose of the different marker functions is to separate the translatable messages into different catalogs, which the translation teams can prioritize translating. It is important to choose the right marker function, to ensure that strings the user sees will be translated and to help the translation team manage their work load.

Everything marked with _() will be translated. Prioritizing the catalogs created from strings marked with the log marker functions is up to the individual translation teams and their users, but it is expected that they will work on critical and error messages before warning or info.

_() is preferred for any user facing message, even if it is also going to a log file. This ensures that the translated version of the message will be available to the user.

The log marker functions (_LI(), _LW(), _LE(), and _LC()) must only be used when the message is only sent directly to the log. Anytime that the message will be passed outside of the current context (for example as part of an exception) the _() marker function must be used.

A common pattern is to define a single message object and use it more than once, for the log call and the exception. In that case, _() must be used because the message is going to appear in an exception that may be presented to the user.

For example, do not do this:

# WRONG
msg = _LE('There was an error.')
LOG.exception(msg)
raise LocalExceptionClass(msg)

Instead, use this style:

# RIGHT
msg = _('There was an error.')
LOG.exception(msg)
raise LocalExceptionClass(msg)

Except in the case above, _() should not be used for translating log messages. This avoids having the same string in two message catalogs, possibly translated differently by two different translators. The log message will translate properly because when the message is not found in the log specific catalog the _() catalog will be used.

If a common message is not being used, they should each be treated separately with respect to choosing a marker function.

For example, do not do this:

# WRONG
LOG.exception(_('There was an error.'))
raise LocalExceptionClass(_('An error occured.'))

Instead, use this style:

# RIGHT
LOG.exception(_LE('There was an error.'))
raise LocalExceptionClass(_('An error occured.'))

Adding Variables to Translated Messages

Translated messages should not be combined with other literal strings to create partially translated messages. For example, do not do this:

# WRONG
raise ValueError(_('some message') + ': variable=%s' % variable)

Instead, use this style:

# RIGHT
raise ValueError(_('some message: variable=%s') % variable)

Including the variable reference inside the translated message allows the translator to take into account grammar rules, differences in left-right vs. right-left rendering, and other factors to make the translated message more useful to the end user.

Any message with more than one variable should use named interpolation instead of positional, to allow translators to move the variables around in the string to account for differences in grammar and writing direction.

For example, do not do this:

# WRONG
raise ValueError(_('some message: v1=%s v2=%s') % (v1, v2))

Instead, use this style:

# RIGHT
raise ValueError(_('some message: v1=%(v1)s v2=%(v2)s') % {'v1': v1, 'v2': v2})

Adding Variables to Log Messages

String interpolation should be delayed to be handled by the logging code, rather than being done at the point of the logging call. For example, do not do this:

# WRONG
LOG.info(_LI('some message: variable=%s') % variable)

Instead, use this style:

# RIGHT
LOG.info(_LI('some message: variable=%s'), variable)

This allows the logging package to skip creating the formatted log message if the message is not going to be emitted because of the current log level.

Avoid Forcing the Translation of Translatable Variables

Translation can also be delayed for variables that potentially contain translatable objects such as exceptions.

Whenever possible translation should not be forced by use of str(), unicode(), or six.text_type() on a message being used with a format string.

For example, do not do this:

# WRONG
LOG.info(_LI('some message: exception=%s'), six.text_type(exc))

Instead, use this style:

# RIGHT
LOG.info(_LI('some message: exception=%s'), exc)

This allows the translation of the translatable replacement text to be delayed until the message is translated.