Friday, March 9, 2018

Lazy and unlazy gettext

I still didn’t commit my yesterday’s work because I noticed that there was at least one subtle but important problem. I often used the construct:


That is, I call ugettext_lazy() to produce a Promise object which will look up my translatable text lasily when needed, and then I do nothing else than passing it to str() in order to force it to resolve my text. Which is nonsense. In that case we can directly call ugettext() instead of ugettext_lazy():


To make this more easily usable, I added a third import in lino.api:

from django.utils.translation import ugettext_lazy as _
from django.utils.translation import pgettext_lazy as pgettext
from django.utils.translation import ugettext as gettext

I would have preferred to use __() (a double underscore) as alias for gettext():

from django.utils.translation import ugettext as __

But this causes the message extractor to not find these strings. The xgettext program looks for calls to gettext(), pgettext(), and _(), but not __() (the –keyword option to xgettext has the following default values for Python: gettext, ugettext, dgettext:2, ngettext:1,2, ungettext:1,2, dngettext:2,3, _.

Lino uses the message extractors that come with Babel. I didn’t find any configuration option to tell Babel that it should also extract calls to __().


  • replace str(_(“Foo”)) constructs by gettext(“Foo”)
  • repair the welfare test suite which is still broken