Markus Kuhn writes:
> I just want to write in maximally portable ISO C99 code
>
> #if __STDC_ISO_10646__
> printf("%lc", 0x201c);
> #else
> putchar('"');
> #endif
>
> and it should do the right thing depending on the locale (including
> transliteration to " for ISO 8859-1, which glibc 2.1.96 fails to do at
> the moment unfortunately).
This piece of code is far from being portable. The printf line has
undefined behaviour, because the standards don't say what happens when
you output (wchar_t)0x201c (in case the conversion to multibyte
fails).
> But gettext is overkill for monolingual applications that want to run in
> different character encodings (and there are *many* of these).
If you only want the translation, not the message catalog lookup, you
can do it like this:
#if __STDC_ISO_10646__
const wchar_t[] wstr = { 0x201c };
char* str;
size_t length;
if (iconv_string (nl_langinfo (CODESET), "wchar_t",
(const char *) wstr, (const char *) (wstr + 1),
&str, &length) == 0)
fwrite (str, 1, length, stdout);
#else
putchar('"');
#endif
[iconv_string() is from libiconv/extras.]
But then it is not useful to check __STDC_ISO_10646__. The following
is more portable:
const uint16_t[] wstr = { 0x201c };
char* str;
size_t length;
if (iconv_string (nl_langinfo (CODESET), "UCS-2",
(const char *) wstr, (const char *) (wstr + 1),
&str, &length) == 0)
fwrite (str, 1, length, stdout);
> I do not want to have to be forced to use (the non-standard!)
> gettext
gettext is standard now: it's part of the LI18NUX specification.
Bruno
-
Linux-UTF8: i18n of Linux on all levels
Archive: http://mail.nl.linux.org/lists/
