Op 21-08-2023 om 16:20 schreef Julien Lepiller:
> Le 21 août 2023 14:09:14 GMT+02:00, Maxime Devos <maximedevos@telenet.be> a écrit :
>> Consider, e.g.,
>>
>> (format #t (G_ "~0@*~a should be set to ~1@*~a instead of ~2@*~a~%") "CC" "(cc-for-target)" "gcc")
>> ->
>> CC should be set to (cc-for-target) instead of gcc
>>
>> By using positional arguments like this, translators can reorder the sentence to:
>>
>> (format #t (G_ "It's not ~2@*~a that ~0@*~a should be set to, but ~1@*~a~%") "CC" (cc-for-target) "gcc"))
>>
>> ~0@*~a should be set to ~1@*~a instead of ~2@*~a~%") "CC" "(cc-for-target)" "gcc")
>> ->
>> It's not gcc that CC should be set to, but (cc-for-target).
>>
>> CC should be set to (cc-for-target) instead of gcc
>>
>> Such reorderings are occasionally useful, yet AFAIK nowhere (except po/guix/ta.po, the mcron service and de.po) is this used.
>>
>> Sure, you could as translator add these ~N@* afterwards, but you need to know that's possible in the first place (and if you know it's possible, you still need to remember or rediscover what exactly to write), and it would be much simpler and more discoverable if they were included from the start.  Also, IIRC, Weblate complains if you add these.
>>
>> p.s.: I'm writing a new linter, this particular example doesn't occur yet in Guix.
> 
> That sounds reasonnabe. The very least we could do is document this syntax in the manual. Weblate would complain indeed, sirce it won't find the same formats in the source and target strings. It might complain about the order too, but that's something we could contribute upstseam if it happens.


Proposed new documentation:

* 22.5.5 Translatable messages

When constructing translatable messages with 'format' (<reference to 
guile manual>), it is important to use positional arguments.  For 
example, instead of writing

   (format #t (G_ "The package '~a' is newer than '~a'.~%")
              '("foo" "bar")),

you should write

   (format #t (G_ "The package '~1@*~a' is newer than '~2@*~a'.~%")
              '("foo" "bar"))

instead.  The reason for this is that when translating between 
languages, the word order sometimes changes.  While a translator could 
manually add the '~N@*' to the translation, it is more straightforward 
to include it in the untranslated message from the beginning.

There is also a technical reason for this: Weblate doesn't cope well 
with the translated message and original message having a different 
number of '~'.

Update the following documentation:

      If there are multiple formatting symbols, make sure to respect the
      order.  Guile does not know in which order you intended the string
      to be read, so it will substitute the symbols in the same order as
      the English sentence.

      As an example, you cannot translate ‘package '~a' has been
      superseded by '~a'’ by ‘'~a' superseeds package '~a'’, because the
      meaning would be reversed.  If FOO is superseded by BAR, the
      translation would read ‘'foo' superseeds package 'bar'’.  To work
      around this problem, it is possible to use more advanced formatting
      to select a given piece of data, instead of following the default
      English order.  *Note (guile)Formatted Output::, for more
      information on formatting in Guile.

--->

      If there are multiple formatting symbols, and positional arguments
      are used, you should be able to reorder the formatting symbols if
      it makes more sense in your language.  The positional information
      tells Guile which formatting symbol needs to be replaced by which
      value.

      If positional arguments aren't used and there are multiple
      formatting symbols, that's a bug, and possibly to translate
      the message properly you may need to report the bug.  The use of
      positional arguments in Guix is fairly new, so you might need
      to report plenty of them!

      As an example, you can translate ‘package '~1@*~a' has been
      superseded by '~2@*~a'’ by ‘'~2@*~a' superseeds package '~1@*~a'’.
      In this example, you recognise the positional arguments by the
      prefixes '~1@*' and '~2@*'.

(To be explicit, the changes are Copyright 2023 Maxime Devos

Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3 or
any later version published by the Free Software Foundation; with no
Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts.  A
copy of the license is included in the section entitled ``GNU Free
Documentation License''.)