Hi! On Sun, 2013-03-24 at 20:51:39 +0100, Jakub Wilk wrote: > I initially imported all plural forms that were known to gettext, > but then did some reality checking and commented out those that > weren't used in practice or were otherwise problematic. If an entry > is commented out, it means that I haven't wrapped my head around it > yet. :)
Sure, makes sense. :) > Bosnian (together with Belarusian, Croatian, Russian, Serbian and > Ukrainian) is actually an interesting case. According to gettext > documentation and Translate Toolkit wiki, the correct Plural-Forms > for them is: > > nplurals=3; plural= \ > n%10==1 && n%100!=11 ? 0 : \ > n%10>=2 && n%10<=4 && (n%100<10 || n%100>=20) ? 1 : 2; > > This expression has an unpleasant property that plural(1) == > plural(21). Why is that a problem? Well, usually it isn't. But > sometimes software authors use plural forms in messages like this: Aha, thanks for pointing this out, had missed that when skimming over those sites/docs. > msgid "an apple" > msgid_plural "apples" > > or like this: > > msgid "an apple" > msgid_plural "%d apples" > > Then in translation you would get singular form, even though the > numeral is greater than 1. Ooops! Actually, it does not really matter how the original string was written, as long as the translated string matches the plural-form formula. The argument to the printf-like function will be passed anyway (if it is passed at all), so it can still be translated like this: msgid "an apple" msgid_plural "%d apples" msgstr[0] "%d poma" msgstr[1] "%d pomes" So to me this seems to end up being a matter of style and how the translators want to translate the strings. > Of course, I'm not the first one who noticed this is a problem. In > fact, majority of Russian, Serbian and Ukrainian MO files in Debian > use this Plural-Forms instead: > > nplurals=4; plural= \ > n==1 ? 3 : \ > n%10==1 && n%100!=11 ? 0 : \ > n%10>=2 && n%10<=4 && (n%100<10 || n%100>=20) ? 1 : 2; > > This is the same as above, but with special case for 1. Given the above, then I think a solution could be to accept both plural-forms as valid (because they are a matter of translation-style). And ideally, i18nslator would check that if the condensed form is used (nplurals=3), the msgstr matching n==1 contains the same amount of format directives as the msgstr matching f.ex. n==2, to detect the problematic case you brought up before, and to not trip on strings with more than one format directive, for example: msgstr[0] "numeral-or-article text-0 %d" msgstr[1] "%d text-1 %d" msgstr[2] "%d text-2 %d" This would then trigger a warning. Thanks, Guillem -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
