Unicode or not?

Unicode or not?

[Krzysztof]

Does it happen that command line which is passed to program arguments is "unicoded"? In other words, when should "main" be defined as "main(int argc, wchar_t **argv)"?

-- 
Regards
Krzysztof J.

Re: Unicode or not?

[Andrej Gelenberg]

HI,

it's always char **, which encoding the arguments has, depends on system
locale (mostly LANG environment variable), but it must be always some
sort of multibyte encoding at most (like utf8). UCS 2 (UTF-16) would
break many existing programs.

http://stackoverflow.com/questions/1664476/is-it-possible-to-use-a-unicode-argv

and here, how to covert between utf8 and wchar:
http://www.ibm.com/developerworks/linux/library/l-linuni/index.html

On 03/05/2012 01:23 PM, Krzysztof wrote:
> Does it happen that command line which is passed to program arguments is
> "unicoded"? In other words, when should "main" be defined as "main(int
> argc, wchar_t **argv)"?
> 

Regards,
Andrej Gelenberg

Re: Unicode or not?

[Glynn Clements]

Krzysztof wrote:

> Does it happen that command line which is passed to program arguments
> is "unicoded"? In other words, when should "main" be defined as
> "main(int argc, wchar_t **argv)"?

Never. Windows supports a wmain() function which uses wchar_t** for
argv, but there is no equivalent on Unix.

Unix doesn't normally use wide strings for communication between
programs or between programs and the OS. Unicode text invariably uses
UTF-8.

-- 
Glynn Clements <glynn@gclements.plus.com>

Re: Unicode or not?

[Krzysztof]

So how to read effectively UTF-8 characters from char* passed as an 
argument under Linux?
Should one simply cast argv[n] to wchar_t*?

-- 
Regards
Krzysztof J.

Re: Unicode or not?

[Andrej Gelenberg]

Hi,

no, you can't simply cast it to wchar. I recommend to read this article
about unicode under linux:
http://www.ibm.com/developerworks/linux/library/l-linuni/index.html

There are 2 possible ways to deal with utf8: keep it char* and use as
simple c-string. Pro: it's simple and you can keep using standard str*
functions and it often smaller as wchar string. Cons: some non latin
symbols may consume more then one bytes, so strlen will report bigger
number as characters there, which can lead to problems with displaying
or counting the characters. You can steel do it with mblen, but it's bit
pain.
Second option is to convert it to wchar with mbstowcs() function. Pro:
characters are always fixed bit-width. Cons: you need to convert between
utf8 and wchar and you need additional buffer to hold wchar string (you
can't do in in-place, because wchar string will be often bigger then
utf8 string).

For example, if you need or just wont wchar string, you can do something
like this:

int l = strlen(argv[i]);
wchar_t *nbuf = calloc(sizeof(*nbuf), l);
if ( !nbuf ) return 1;
l = mbstowcs(nbuf, argv[i], l); // mbstowcs may return smaller value as
                                // l
if ( l == -1 ) {
  /* invalid multibyte sequence was encountered */
  free(nbuf);
  return 2;
}

Regards,
Andrej Gelenberg

On 03/05/2012 09:19 PM, Krzysztof wrote:
> So how to read effectively UTF-8 characters from char* passed as an
> argument under Linux?
> Should one simply cast argv[n] to wchar_t*?
>