[PATCH 1 of 2] templating: fix age() function to return 1 minute, 1 hour, 1 day etc

[Martin Geisler]

Matt Mackall <mpm at selenic.com> writes:

> On Tue, 2011-12-20 at 04:56 +0700, Andrei Polushin wrote:
>> 20.12.2011 4:20, Matt Mackall wrote:
>> > On Mon, 2011-12-19 at 06:58 +0700, Andrei Polushin wrote:
>> >>
>> >> The existence of plural() function earlier in code show the intent
>> >> to make singular/plural forms, while the condition '>= 2' disables
>> >> all singular forms other than '1 second'.
>> > 
>> > Incorrect. It merely demonstrates the developer who committed
>> > plural() in hgweb (me, 23 May 2005) was not yet aware of our
>> > long-standing policy of avoiding the pointless complexity of
>> > pluralization. The vast bulk of the code intentionally makes no
>> > effort to deal with pluralization and we've rejected dozens of
>> > patches to fix "1 files" and other pluralization special-cases.
>> 
>> That's a pity :(
>> 
>> First, pluralization is not pointless for Russians, compare:
>> 
>>   %d ignored file(s)              -- English
>>   %d игнорируем(ый,ых) файл(а,ов) -- Russian has 2 or 3 possible forms
>> 
>>   %d year(s) -- English
>>   1 год      -- Russian for 1, 21, 31 ... 101, 121 ...
>>   2 года     -- Russian for 2, 3, 4, 22, 23, 24 ...
>>   5 лет      -- Russian for other numbers (different word used!)
>
> This is actually one of the reasons why we don't bother. The rules
> change from language to language and basically no one knows all the
> rules.

It sounds like a misunderstanding here: you don't need to know the rules
for all languages, you just need to come up with singlular and plural
strings for *English*.

The translators will then have to deal with the two or more cases there
exist for their favorite language.

> Code like plural() is hopeless. Yes, ngettext exists, but it means
> educating way too many people about these complexities for basically
> negligible gain.

I'm sorry to say this, but that's a quite English-centric point of view.
Yes, many languages have only two forms and so always using the plural
form works well -- it only fails when you have '1 years.

But for languages where there is a difference between '2 years and '5
years', it looks strange when the wrong form is used.

> Note that only a witless moron could ever actually be confused (rather
> than simply annoyed) by "1 files". Unfortunately, we actually deal
> with these witless morons on a daily basis: they're called computers.
> And as it happens, they're actually much more likely to be confused by
> the difference between "1 file" and "2 files", especially if we were
> to switch to using the latter 6 years in.

Maybe we could let the English translation use the plural form always
(like today) when HGPLAIN is set? That makes it easy for computers to
parse the output, while providing nice output for humans.

I'm not sure if that was what Andrei already suggested in his ngettext
patch, but it would be something like:

  if os.environ.get('HGPLAIN'):
      def ngettext(singular, plural, count):
          # always use English plural text for easy parsing by witless
          # morons (computers).
          return plural
  else:
      def ngettext(singular, plural, count):
          # call tungettext, encode as necessary

> So we have a choice between potentially crucial scripts falling after
> doing more work and annoying language pedants with less work. Pedants
> lose.

We could even turn of translations when stdout is not attached to a TTY.
But maybe it would be too confusing for a Russian sysadmin that the
output changes like that depending on how the script is executed.

On the other hand, I'm not sure that translators have kept the output as
stable as we keep the English output. That is, if they made sure to pick
a good word for 'changeset' at the beginning and never changed their
minds.

Finally, words that were introduced late (like 'bookmarks') must have
changed from English to a native language as translators got around to
translate them. So grepping for specific English words can suddenly fail
when a new stable version is released with better translations. This is
yet another reason why people must use HGPLAIN when interfacing with
Mercurial in scripts.

-- 
Martin Geisler

Mercurial links: http://mercurial.ch/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: not available
URL: <http://selenic.com/pipermail/mercurial-devel/attachments/20120105/90c48094/attachment.pgp>