189 lines
5.1 KiB
C
189 lines
5.1 KiB
C
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <wchar.h>
|
|
#include <locale.h>
|
|
#include "mbctype.h"
|
|
|
|
/* The following function concerns caracter coded in more than 0ne byte. For our small_printf we considered only
|
|
caracters coded in ASCII and therefore all our caracters are coded in 8 bits (One byte)
|
|
If you need the following treatment because caracter that you use are coded in more than one byte
|
|
comment the three following lines */
|
|
|
|
#if defined( _SMALL_PRINTF ) || defined(SMALL_SCANF)
|
|
#define _ASCII_CAR
|
|
#endif
|
|
|
|
|
|
/* for some conversions, we use the __count field as a place to store a state value */
|
|
#define __state __count
|
|
|
|
#ifndef _ASCII_CAR
|
|
extern char __lc_ctype[12];
|
|
#endif
|
|
int
|
|
_DEFUN (_wctomb_r, (r, s, wchar, state),
|
|
struct _reent *r _AND
|
|
char *s _AND
|
|
wchar_t wchar _AND
|
|
mbstate_t *state)
|
|
{
|
|
|
|
#ifndef _ASCII_CAR
|
|
|
|
if (strlen (__lc_ctype) <= 1)
|
|
{ /* fall-through */ }
|
|
else if (!strcmp (__lc_ctype, "C-UTF-8"))
|
|
{
|
|
if (s == NULL)
|
|
return 0; /* UTF-8 encoding is not state-dependent */
|
|
|
|
if (wchar <= 0x7f)
|
|
{
|
|
*s = wchar;
|
|
return 1;
|
|
}
|
|
else if (wchar >= 0x80 && wchar <= 0x7ff)
|
|
{
|
|
*s++ = 0xc0 | ((wchar & 0x7c0) >> 6);
|
|
*s = 0x80 | (wchar & 0x3f);
|
|
return 2;
|
|
}
|
|
else if (wchar >= 0x800 && wchar <= 0xffff)
|
|
{
|
|
/* UTF-16 surrogates -- must not occur in normal UCS-4 data */
|
|
if (wchar >= 0xd800 && wchar <= 0xdfff)
|
|
return -1;
|
|
|
|
*s++ = 0xe0 | ((wchar & 0xf000) >> 12);
|
|
*s++ = 0x80 | ((wchar & 0xfc0) >> 6);
|
|
*s = 0x80 | (wchar & 0x3f);
|
|
return 3;
|
|
}
|
|
else if (wchar >= 0x10000 && wchar <= 0x1fffff)
|
|
{
|
|
*s++ = 0xf0 | ((wchar & 0x1c0000) >> 18);
|
|
*s++ = 0x80 | ((wchar & 0x3f000) >> 12);
|
|
*s++ = 0x80 | ((wchar & 0xfc0) >> 6);
|
|
*s = 0x80 | (wchar & 0x3f);
|
|
return 4;
|
|
}
|
|
else if (wchar >= 0x200000 && wchar <= 0x3ffffff)
|
|
{
|
|
*s++ = 0xf8 | ((wchar & 0x3000000) >> 24);
|
|
*s++ = 0x80 | ((wchar & 0xfc0000) >> 18);
|
|
*s++ = 0x80 | ((wchar & 0x3f000) >> 12);
|
|
*s++ = 0x80 | ((wchar & 0xfc0) >> 6);
|
|
*s = 0x80 | (wchar & 0x3f);
|
|
return 5;
|
|
}
|
|
else if (wchar >= 0x4000000 && wchar <= 0x7fffffff)
|
|
{
|
|
*s++ = 0xfc | ((wchar & 0x40000000) >> 30);
|
|
*s++ = 0x80 | ((wchar & 0x3f000000) >> 24);
|
|
*s++ = 0x80 | ((wchar & 0xfc0000) >> 18);
|
|
*s++ = 0x80 | ((wchar & 0x3f000) >> 12);
|
|
*s++ = 0x80 | ((wchar & 0xfc0) >> 6);
|
|
*s = 0x80 | (wchar & 0x3f);
|
|
return 6;
|
|
}
|
|
else
|
|
return -1;
|
|
}
|
|
else if (!strcmp (__lc_ctype, "C-SJIS"))
|
|
{
|
|
unsigned char char2 = (unsigned char)wchar;
|
|
unsigned char char1 = (unsigned char)(wchar >> 8);
|
|
|
|
if (s == NULL)
|
|
return 0; /* not state-dependent */
|
|
|
|
if (char1 != 0x00)
|
|
{
|
|
/* first byte is non-zero..validate multi-byte char */
|
|
if (_issjis1(char1) && _issjis2(char2))
|
|
{
|
|
*s++ = (char)char1;
|
|
*s = (char)char2;
|
|
return 2;
|
|
}
|
|
else
|
|
return -1;
|
|
}
|
|
}
|
|
else if (!strcmp (__lc_ctype, "C-EUCJP"))
|
|
{
|
|
unsigned char char2 = (unsigned char)wchar;
|
|
unsigned char char1 = (unsigned char)(wchar >> 8);
|
|
|
|
if (s == NULL)
|
|
return 0; /* not state-dependent */
|
|
|
|
if (char1 != 0x00)
|
|
{
|
|
/* first byte is non-zero..validate multi-byte char */
|
|
if (_iseucjp (char1) && _iseucjp (char2))
|
|
{
|
|
*s++ = (char)char1;
|
|
*s = (char)char2;
|
|
return 2;
|
|
}
|
|
else
|
|
return -1;
|
|
}
|
|
}
|
|
else if (!strcmp (__lc_ctype, "C-JIS"))
|
|
{
|
|
int cnt = 0;
|
|
unsigned char char2 = (unsigned char)wchar;
|
|
unsigned char char1 = (unsigned char)(wchar >> 8);
|
|
|
|
if (s == NULL)
|
|
return 1; /* state-dependent */
|
|
|
|
if (char1 != 0x00)
|
|
{
|
|
/* first byte is non-zero..validate multi-byte char */
|
|
if (_isjis (char1) && _isjis (char2))
|
|
{
|
|
if (state->__state == 0)
|
|
{
|
|
/* must switch from ASCII to JIS state */
|
|
state->__state = 1;
|
|
*s++ = ESC_CHAR;
|
|
*s++ = '$';
|
|
*s++ = 'B';
|
|
cnt = 3;
|
|
}
|
|
*s++ = (char)char1;
|
|
*s = (char)char2;
|
|
return cnt + 2;
|
|
}
|
|
else
|
|
return -1;
|
|
}
|
|
else
|
|
{
|
|
if (state->__state != 0)
|
|
{
|
|
/* must switch from JIS to ASCII state */
|
|
state->__state = 0;
|
|
*s++ = ESC_CHAR;
|
|
*s++ = '(';
|
|
*s++ = 'B';
|
|
cnt = 3;
|
|
}
|
|
*s = (char)char2;
|
|
return cnt + 1;
|
|
}
|
|
}
|
|
|
|
if (s == NULL)
|
|
return 0;
|
|
#endif
|
|
/* otherwise we are dealing with a single byte character */
|
|
*s = (char) wchar;
|
|
return 1;
|
|
}
|
|
|
|
|