Coding for translation
Coding guidelines to enable easy and correct translation of strings on the User Interface.
Contents
Introduction
Gramps has always been internationalized (see: https://gramps-project.org/introduction-WP/2006/04/looking-back-over-5-years/ ) Therefore, all strings meant for the user should always be flagged for translation.
In order to be considered for inclusion in the official Gramps release, any piece of code must support internationalization. What this means is that the Python module must support translations into different languages. Gramps provides support to make this as easy as possible for the developer. For enabling, a language code must be set on the ALL_LINGUAS section.
How to allow translations
Gramps is a fully-internationalized application with translations in many languages. All code which presents text to users must provide for that text to be translated. Fortunately, Gramps provides an extension of gettext which makes this fairly painless. First, alias the gettext function from the single localization instance:
from gramps.gen.const import GRAMPS_LOCALE as glocale _ = glocale.translation.gettext
This statement imports the gettext
function and aliases it as _
. The translation tools treat strings wrapped in _() as translatable and assemble them into catalogs for the translators to work with; by aliasing it to gettext(), we also enable python to retrieve the translation appropriate for the user's locale.
Example 1:
print("Hello world!")
In this example, the string will always be printed as specified.
Example 1 internationalized:
print _("Hello world!")
In this example, Gramps will attempt to translate the string. If a translation exists, the call to the function will return the translation. If a translation does not exist, the original string is returned.
More complicated translations
In addition to gettext, GrampsTranslation offers two more specialized retrieval functions, ngettext and sgettext.
In some strings, it's necessary to specify different translations depending upon the number of an argument. For example,
George Smith and Annie Jones have 1 child George Smith and Annie Jones have 3 children
We'd code that in python as follows:
_ = glocale.translation.ngettext _(George Smith and Annie Jones have %(num)d child, George Smith and Annie Jones have %(num)d children, n) % {num : n}
In other cases, it's necessary to provide a hint to translators, e.g.,
_(Remaining names | rest)
We're making sure that the translators know that this message id means "what's left" rather than "take a nap". When the file is translated, this is no problem, because the translation doesn't include the hint -- but if the user is working in English, we don't want him to see the hint, so we need to alias _ to sgettext:
_ = glocale.translation.sgettext
Often you need to combine them. While ngettext and sgettext can each handle plain strings, neither can handle the other's strings. Fortunately the intltool message extractor is pretty stupid, so any function name that ends in either _ or gettext will work. This will work pretty well:
_ = glocale.translation.gettext N_ = glocale.translation.ngettext S_ = glocale.translation.sgettext
Obviously you would pass the translatable string to the right function.
Encoding
String handling can be a bit tricky in a localized environment so it's important that developers understand Unicode string handling in both versions of the language.
This is mostly a problem for Microsoft Windows™: Mac OSX and Linux use UTF8 for just about everything if the locale is set up correctly (and we try to do that when Gramps starts up), so one can get away with a lot of encoding mistakes on those platforms. Windows™ on the other hand uses a slightly modified version of UTF16 for file names and retains the old DOS code page system for encoding output to cmd.exe. The take-away is that if you need to mess with input or output encoding, be sure to test on both Linux and Windows before deciding that you're done. If you're not set up for multiple-platform testing arrange with someone,, who can test for you on the platform you don't have.
Python 2
Python 2.7 has two text classes, str and unicode. Unicode objects are encoded in UTF16 internally on most platforms, and most python output functions will do the right thing with them. One caveat here: passing both unicodes and strs to os.path.join() will return a str, so either make sure when constructing a path that all arguments are unicodes or convert the result.
The bsddb module that ships with Python2 is stupid about paths and requires that they be encoded in the file system encoding. This is handled in gramps/gen/db/write.py with _encode() and independently in a few other places.
Strings from the operating system, including environment variables, are a problem on Windows™; The os module uses for input the ANSI API to the Windows SDK, which interprets the value of the environment variable according to the active code page and produces a str, converting any codepoints > 0xff to ? and often misinterpreting those between 0x0f and 0xff if the encoding of the input happens to be something other than the active system codepage. Once this is done it is quite difficult to get non-ASCII pathnames back into a useable form, so gramps/gen/constfunc.py provides a get_env_var() function that uses the Unicode API to instead. Always use that function to read environment variables which might include non-ASCII characters and avoid using os-module functions for reading paths.
By default string constants in Python 2 are str.
Python 3
Python 3 also provides two test classes, str and bytes. In Python 3, str is the unicode type and bytes is text encoded some other way. Everything pretty much "just works".
Portability Functions and constants
We've provided a couple of functions in gramps/gen/constfunc.py to ease conversion of strs to unicodes; these include the necessary tests to portably do the right thing regardless of what's passed to them and according to which version of Python is in use:
- cuni is an alias for unicode in Python 2 and for str in Python 3. This has no protective checks so use it with care.
- conv_to_unicode(string, encoding='utf8'): This ensures that its return value is a Unicode string which has been converted from a non-Unicode in the encoding, which defaults to UTF8 for ease of use with the GUI.
- get_env_var(string, default=None): On Windows™ in Python2, uses the ctypes module to invoke the Microsoft Unicode API to read the value of an environment variable and return a Unicode; otherwise returns the value from the os.environ array.
There are also two constants:
- STRTYPE is an alias for basestring in Python 2 and for str in Python 3. It can be used to test whether an object is a text-type.
- UNITYPE is an alias for unicode in Python 2 and for str in Python 3. It can be used to test whether an object is already encoded in Unicode.
For portable string handling on all platforms and for all locales
- Localized strings returned from gettext, ngettext, etc. are always unicode
- Text files should always be encoded in UTF8. The easy and portable way to do this is to:
import io
fh = io.open(filepath, mode, encoding='utf8')
- where mode is one of r, rw, r+, or w+. Don't open these files in binary mode! Pass unicode-type strings to fh.write() and expect the same from fh.read().
- Always read environment variables with constfuncs.get_env_var() if there's any chance that it will contain a non-ASCII character.
- Use from __future__ import unicode_literals in any source file which might present strings to the user or to the operating system.
- When creating string literals, don't do this:
print _(u"Eg, valid values are 12.0154, 50° 52′ 21.92″N")
- Because the u prefix was removed for Python 3.0-3.2. (It was restored in 3.3 for compatibility with 2.7, but it's not necessary.)
- Instead, put in the first line of the module
# *-* coding: utf-8 *-*
- then in the imports section
from __future__ import unicode_literals
- which makes all of the literals unicode. Make sure that your editor is set up to save utf-8!
Glade files
Just enable the translatable attribute on an XML element.
<property name="label" translatable="yes">_Family:</property> <property name="tooltip" translatable="yes">Abandon changes and close window</property> <property name="label" translatable="no"><b> - </b> </property>
Non ASCII characters
If you plan to use non ASCII characters in a string, that shall be translated, do not use escape sequences:
Eg, valid values are 12.0154, 50&
#xB0; 52' 21.92"N
use instead:
Eg, valid values are 12.0154, 50° 52′ 21.92″N
In this case note the unicode characters for deg, min, sec. Ensure that your editor is set up to encode the characters in UTF-8!
Accessibility
In addition to accelerators, GtkWidget also support a custom <accessible> element, which supports actions and relations. Properties on the accessible implementation of an object can be set by accessing the internal child "accessible" of a GtkWidget. See GtkBuilder UI.
- Gtk label
A GtkLabel with mnemonic support will automaticaly generate accessibility keys on linked GtkEntry and UndoableEntry fields. Remember that Gramps also uses custom widgets like StyledTextEditor and ValidatableMaskedEntry, which do not always have relation with a GtkLabel.
- Toggle buttons and Icons on toolbar
Gramps often uses GtkToggleButtons and alone GtkImage (image without label), this excludes blind people and generates a poor interface for accessibility.
See Accessibility.
Addons
External addons often need to provide their own message catalogs. To pull one in, use this instead of the usual.
from gramps.gen.const import GRAMPS_LOCALE as glocale _ = glocale.get_addon_translator(__file__).gettext
or if you need more than one retrieval function:
_translation = glocale.get_addon_translator(__file__) _ = _translation.gettext S_ = _translation.sgettext
The addon translator is another instance of GrampsTranslation, so the rules for creating translatable strings and for retrieving the translated values are the same as for internal modules.
See Addons development for more details.
How it works
We need at least GNU gettext, then msginit will generate a standard gettext header.
Gramps has used different environments according to versions for retrieving strings to translate:
There are two stages to getting a translation to work.
Files and directory
Translations are stored in a .po
file that contains the mappings between the original strings and the translated strings, see Translating Gramps.
Translators use a generic file gramps.pot
to generate their .po
file.
Gramps uses a utility that extracts the strings from the source code to build the .po
file. This utility examines the source files for strings that have been marked as translatable. In the python source, these are the strings enclosed in the _()
function calls.
Note that because strings are extracted by a script from the source file, string constants and not variables must be enclosed in the _()
call. In the following example, the extraction script will not extract the string.
mystring = "Hello World!" print _(mystring)
The correct method would be to use one of the following:
mystring = _("Hello World!") print mystring
At run time, the _()
calls will translate the string by looking it up in the translation database (created from the .po
files) and returning the translated string.
Add the reference to the file
We need to also add a reference to this file for generating the translation template.
Tips for writing a translatable Python module
Use complete sentences
Don't build up a sentence from phrases. Because a sentence is ordered in a particular way in your language does not mean that it is ordered the same way in another. Providing the entire sentence as a single unit allows the translator to make a meaningful translation. Do not concatenate phrases or terms as they will then show up as separate phrases or terms to be translated and the complete sentence may then show up incorrectly, especially in right-to-left languages (Arabic, Hebrew, etc.).
Use named %s/%d values
Python provides a powerful mechanism that allows the reordering of %s values in a string. A translator may need to rearrange the structure of a sentence, and it may not match the order you chose. For example:
print "%s was born in %s" % ('Joe','Toronto')
In some languages it may make more sense to say:
print "%s is the city in which %s was born" % ('Toronto', 'Joe')
The problem is that this requires a change to the order of the arguments. Python provides a solution for this. By using named operators and dictionaries, we can say:
print "%(male_name)s was born in %(city)s" % { 'city' : 'Toronto', 'male_name' : 'Joe'}
In this case, the order of the %s formatters is not important, since the values will be looked up in the dictionary at run time to resolve the value. The translator can reorder the %s formatters, or even remove them without causing any problems.
Note that Python also allows a variation which some people find easier to read:
print "%(male_name)s was born in %(city)s" % dict( city = 'Toronto', male_name = 'Joe')
Some languages are using right-to-left text direction. It is important to use named arguments when there is more than one %s/%d value into a translation string.
Provide separate strings for masculine and feminine.
Many languages have the concept of gender, while others don't. A sentence may need to be phrased differently depending on whether the subject is male or female. By using the named %s values along with a bit of code, this problem can be solved.
if person.getGender() == Person.male: print _("%(male_name)s was born in %(city)s\n") % { 'male_name' : name, 'city' : city } else: print _("%(female_name)s was born in %(city)s\n") % { 'female_name' : name, 'city' : city }
This allows languages with gender differences to map nicely into your sentence.
Provide support for plural forms.
Plurals are handled differently in various languages. Whilst English or German have a singular and a plural form, other languages like Turkish don't distinguish between plural or singular and there are languages which use different plurals for different numbers, e.g., Polish.
Gramps provides a plural forms support, useful for locales with multiples plurals according to a number (often slavic based languages) or for Asian family languages (singular = plural).
Note, some locales need singular form with zero and plural form might be also used in this case.
We need to call module :
from gen.ggettext import ngettext
and code like this :
ngettext("singular %d", "plural %d", n) %n
Sample:
msg = ngettext('Import Complete: %d second', 'Import Complete: %d seconds', t ) % t
Provide a context support.
A translator needs context for a good translation. Keep in mind you can help him/her, by using context on translation string.
We need to call module :
from gen.ggettext import sgettext as _
or
from gen.ggettext import sngettext as _
(if you use ngettext) # not implemented
Translation string will use context, but this will be hidden on user interface.
_("context|string")
Translator will see the translation string and a help string without loading program. Program will only display the string in English or with another locale.
Object classes
Gramps often displays names of primary objects (Person, Family, Event, etc. ...), for being consistent on displayed strings (also in english!), there is a trans_objclass(objclass_str) function on TransUtils module.
So, when we need to display the primary object name in lower case into a sentence, we can use this function.
ex:
from gen.ggettext import sgettext as _ from TransUtils import trans_objclass
_("the object|See %s details") % trans_objclass(objclass) _("the object|Make %s active") % trans_objclass('Person')
will display:
See the person details # or See the family, the event, etc. ... details Make the person active
Genitive form
Genitive (and some other) forms need to modify the name itself into some locales, like Finnish or Swedish.
Instead of "free form" text that talks about e.g.,
son of %s
better would be for example some tabulated format like this:
son: %s daughter: %s
which doesn't require genitive.
Punctuation
Use of commas, semicolons and spacing can be different than into english. Remember, simple is better, maybe try to limit punctuation marks.
definition
$ python3 >>> import string >>> print(string.punctuation) !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~
locale case
In French, a space is required before or after some punctuation marks and symbols, like
: ; « » ! ? % $ # .
- GtkBuilder (editors, configuration dialogs) can provide a default colon after the string without spacing,
so need some extra-testing and customization for some translators. e.g., in french
#: ../gramps/gen/plug/report/stdoptions.py:257 ../gramps/gui/configure.py:1222 msgid "Date format" msgstr "Format des dates "
# comté (Canada) #: ../gramps/gui/configure.py:617 #: ../gramps/gui/editors/displaytabs/addrembedlist.py:75 #: ../gramps/plugins/view/repoview.py:92 msgid "State/County" msgstr "Province/Comté "
# L'espace final est pour précéder le « : » codé en dur. #: ../gramps/gui/configure.py:1332 msgid "Status bar" msgstr "Barre d'état "
Deferred key on lists
In most coding situations, strings are translated where they are coded. Occasionally however, you need to mark strings for translation, but defer actual translation until later. A classic example is:
animals = ['mollusk', 'albatross', 'rat', 'penguin', 'python', ] for a in animals: print(a)
Here, you want to mark the strings in the animals list as being translatable, but you don’t actually want to translate them until they are printed.
Here is one way you can handle this situation:
def _(message): return message animals = [_('mollusk'), _('albatross'), _('rat'), _('penguin'), _('python'), ] del _ for a in animals: print(_(a))
This works because the dummy definition of _() simply returns the string unchanged. And this dummy definition will temporarily override any definition of _() in the built-in namespace (until the del command). Take care, though if you have a previous definition of _() in the local namespace.
Note that the second use of _() will not identify “a” as being translatable to the gettext program, because the parameter is not a string literal.
Another way to handle this is with the following example:
def _T_(message): return message animals = [_T_('mollusk'), _T_('albatross'), _T_('rat'), _T_('penguin'), _T_('python'), ] for a in animals: print(_(a))
In this case, you are marking translatable strings with the function _T_(), which won’t conflict with any definition of _().
Current custom key on gramps code is _T_. Set as xgettext flag on shell script and python interface, generating the translation strings template.
Changing translated text message in the source code
One of the severities in our bug tracker is "text", which ranks up as easier than "tweak" and "minor", but more difficult than "trivial". If a bug is concerned with readability or correctness of a text that Gramps outputs, whether in GUI, in a console error message, or in a produced report, then "text" is the severity to use. So why is it more than "trivial"?
As described above, any translated text in the source code gets reflected into tens of *.po files, maintained by the translators. So every time you just change it in the source, ALL the translators need to do the translation again. Normally, the translation environment will give a prudent suggestion, but there is still a manual approval step. If you check in the change, the string will not be translated until the translators pick it up.
This is why, if what you change is just a couple of spelling mistakes, a missing comma in the middle, or maybe an extra space somewhere in the message, it's a good idea to save the translators' work, by doing a global search and replace of your source message text in the *.po files, and committing these along with your change.
For short enough messages, that don't span multiple lines in the *po files, you can do it by executing
perl -pi -e 's/YOUR MESSAGE BEFORE CORRECTION/your message after correction/g;' *.po *.pot
in the po/ directory. Make sure you do a "git diff" and observe the results make sense. (You'll probably have to escape some characters in the regular expression, such as | or .).
To make it easier to port your changes across multiple branches, it's a good idea to separate the changes in the source tree from the po/ ones. This way, you'll be able to quickly re-apply the source changes using normal cross-branch porting workflow (such as `git cherry-pick'), and then adjust and re-run the search-and-replace in the po/ on the new branch, because, most likely, it won't reapply due to the differences in the .po layout.
Textual reports
Starting with Gramps-3.2 we are able to select the language for textual reports, see feature 2371.
For Gramps it was only available on Ancestor report (3.2.x) and detailed reports (3.3.x).
The capability for translated-output was added to some more (gramps core) reports, in the gramps40 branch, before gramps 4.0.0 was released. So more than the "three original reports" now have had this feature request implemented.
For providing this option:
- import EnumeratedListOption
- import libtranslate
from gen.plug.menu import EnumeratedListOption import TransUtils from libtranslate import Translator, get_language_string
Sample of code:
language = menu.get_option_by_name('trans').get_value() translator = Translator(language) self._ = translator.gettext self.__narrator = Narrator(self.database, self.verbose, use_call, use_fulldate, empty_date, empty_place, translator=translator, get_endnote_numbers=self.endnotes) self.__get_date = translator.get_date self.__get_type = translator.get_type
self._("") self.__get_date(event.get_date_object()) self.__get_type(event.get_type())