reactos/sdk/lib/ucrt/convert/c32rtomb.cpp

94 lines
2.3 KiB
C++
Raw Normal View History

//
// c32rtomb.cpp
//
// Copyright (c) Microsoft Corporation. All rights reserved.
//
#include <corecrt_internal_mbstring.h>
#include <corecrt_internal_ptd_propagation.h>
#include <stdint.h>
#include <uchar.h>
using namespace __crt_mbstring;
extern "C" size_t __cdecl c32rtomb(char* s, char32_t c32, mbstate_t* ps)
{
// TODO: Bug 13307590 says this is always assuming UTF-8.
__crt_cached_ptd_host ptd;
return __c32rtomb_utf8(s, c32, ps, ptd);
}
size_t __cdecl __crt_mbstring::__c32rtomb_utf8(char* s, char32_t c32, mbstate_t* ps, __crt_cached_ptd_host& ptd)
{
if (!s)
{
// Equivalent to c32rtomb(buf, U'\0', ps) for some internal buffer buf
*ps = {};
return 1;
}
if (c32 == U'\0')
{
*s = '\0';
*ps = {};
return 1;
}
// Fast path for ASCII
if ((c32 & ~0x7f) == 0)
{
*s = static_cast<char>(c32);
return 1;
}
// Figure out how many trail bytes we need
size_t trail_bytes;
uint8_t lead_byte;
if ((c32 & ~0x7ff) == 0)
{
trail_bytes = 1;
lead_byte = 0xc0;
}
else if ((c32 & ~0xffff) == 0)
{
// high/low surrogates are only valid in UTF-16 encoded data
if (0xd800 <= c32 && c32 <= 0xdfff)
{
return return_illegal_sequence(ps, ptd);
}
trail_bytes = 2;
lead_byte = 0xe0;
}
else if ((c32 & ~0x001fffff) == 0)
{
// Unicode's max code point is 0x10ffff
if (0x10ffff < c32)
{
return return_illegal_sequence(ps, ptd);
}
trail_bytes = 3;
lead_byte = 0xf0;
}
else
{
return return_illegal_sequence(ps, ptd);
}
_ASSERTE(1 <= trail_bytes && trail_bytes <= 3);
// Put six bits into each of the trail bytes
// Lowest bits are in the last UTF-8 byte.
// Filling back to front.
for (size_t i = trail_bytes; i > 0; --i)
{
s[i] = (c32 & 0x3f) | 0x80;
c32 >>= 6;
}
// The first byte needs the upper (trail_bytes + 1) bits to store the length
// And the lower (7 - trail_bytes) to store the upper bits of the code point
_ASSERTE(c32 < (1u << (7 - trail_bytes)));
s[0] = static_cast<uint8_t>(c32) | lead_byte;
return reset_and_return(trail_bytes + 1, ps);
}