[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Emacs i18n

From: Paul Eggert
Subject: Re: Emacs i18n
Date: Thu, 7 Mar 2019 09:19:35 -0800
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.5.1

On 3/7/19 6:46 AM, Eli Zaretskii wrote:

>> From: Richard Stallman <address@hidden> Cc: address@hidden,
>> address@hidden, address@hidden Date: Wed, 06 Mar 2019 22:42:06
>> -0500 > If we do that, how do we deal with strings that are computed
>> by > concatenation or formatting? Feed them in through %s or
>> something like that.
> But then the strings that are formatted via %s will not be translated,
> they will remain in English.
Yes, but the scenario you describe should not occur in a properly
internationalized GNU application. We obviously can't assume that
Emacs's translation subroutine acts like Google Translate and can
translate any English-language string to the user's language. All we can
assume is that the translation subroutine converts one of a fixed set of
English-language strings to a string appropriate for the user's language.

This limitation will cause problems with Elisp code that does extensive
parsing or processing of English syntax (doctor.el, say), and that sort
of Elisp code will remain English-only (unless someone takes the time to
i18nize them specially). However most Elisp code does not parse English
or generate idiomatic English on the fly: instead, it uses a fixed,
stilted style that can routinely be converted to calls like (message
FORMAT ARG1 ARG2 ...) where FORMAT is translated and the ARG values are not.

To get a quick feel for this issue, I did a simple grep for the string
'(message (concat' in the Emacs source code. I found 41 instances of
this string. Of these, 8 were erroneous because the result of the
concatenation could contain an unwanted "%" or '`' that could cause
'message' to go awry, and I fixed them by installing the attached patch
(by the way, it's routine for i18n efforts to find trivial bugs like
this). The other 33 instances could easily be reworded to do proper i18n
when 'message' translates just its first argument and only simple,
xgettext-style static analysis is used to find the message strings. For
example, this code in calc-do-embedded:

  (message (concat
            "Embedded Calc mode enabled; "
            (if calc-embedded-quiet
                "Type `C-x * x'"
              "Give this command again")
            " to return to normal"))

can easily be rewritten to this:

   (if calc-embedded-quiet
       "Embedded Calc mode enabled; Type `C-x * x' to return to normal"
     "Embedded Calc mode enabled; Give this command again to return to

which is easier for translators to grok, and is arguably clearer even if
we don't want to translate at all.Obviously my '(message (concat'
exercises only a small part of the problem, but the results of this
little sample are encouraging.

So this problem is solvable. Sure, it'll require substantial work, but
the work is routine and this sort of thing has been done for other packages.

The main argument against doing all this is that it's too much work
overall and nobody will have the time to do it all, so let's not even
bother. I have some sympathy for this argument, as i18n is clearly too
much work for any single contributor and the work will distract us from
other things. On the other hand, there's no pressing need to do all the
work quickly, it's a low-level task that can be farmed out to non-expert
volunteers that could conceivably grow the volunteer population, and if
we never even start the work then it will never get done and Emacs will
remain unfriendly to users who don't grok English.

Attachment: 0001-Be-safer-about-in-message-formats.patch
Description: Text Data

reply via email to

[Prev in Thread] Current Thread [Next in Thread]