ngettext

["Vladimir 'φ-coder/phcoder' Serbinenko" <phcoder@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1541 bytes --]

Hello, all. Full gettext has an important feature of ngettext. In GRUB
we don't have this feature. There was a proposal patch however it lacked
the most important part: formula parser. In GRUB codebase there are 4
strings which would normally require ngettext.
Already gettextized: " - Total size %llu sectors", "Mirror with %d
children\n"
Can be gettextized in future (after 2.00 is out): "%d files couldn't be
read and hash of %d files mismatches", "%d files couldn't be read."
Possible solutions are
1) Do as if there wasn't any problem. The strings already gettextised
don't have a big problem with languages that use the same construct for
n >= 2 (English, French, Italian, Spanish, ...). However it makes it
slightly awkward for the languages like Russian.
2) Reformulate the strings in question to avoid requiring ngettext. Does
anyone have any propositions?

Solution 1 or 2 is adopted for 2.00

3) Have a list of possible plural forms. This can easily work as long as
we have access to all translations but it would imply that some
languages need new GRUB version and can't be simply installed in
(without getting awkward plural form)
4) Have a reasonably complete formula parser. It has an advantage of
simple and natural use but its maintenance cost is high independently of
whether we use hand-written parser, our bison parser or import the
parser from full gettext (which uses bison). It seems an overkill for 4
strings.

Any other ideas?

-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]