utf8_to_utf32.cpp

#include <cstdint>
#include <cstdio>

inline bool isUTF8Tail(uint8_t c)
{
	return 0x80 <= c && c <= 0xBF;
}

int convertUTF8CharToUTF32Char(const char * p_utf8_str, char32_t & c)
{
	uint8_t c0 = static_cast<uint8_t>(p_utf8_str[0]);
	if(c0 <= 0x7F)
	{
		c = c0;
		return 1;
	}

	if(c0 < 0xC2)
	{
		return -1;
	}

	uint8_t c1 = static_cast<uint8_t>(p_utf8_str[1]);
	if(c0 <= 0xDF)
	{
		if(!isUTF8Tail(c1))
		{
			return -1;
		}

		c = (static_cast<char32_t>(c0) & 0x1F) << 6;
		c |= c1 & 0x3F;

		return 2;
	}

	uint8_t c2 = static_cast<uint8_t>(p_utf8_str[2]);
	if(c0 == 0xE0)
	{
		if(c1 < 0xA0 || 0xBF < c1 || !isUTF8Tail(c2))
		{
			return -1;
		}
	}
	else if((0xE1 <= c0 && c0 <= 0xEC) || c0 == 0xEE || c0 == 0xEF)
	{
		if(!isUTF8Tail(c1) || !isUTF8Tail(c2))
		{
			return -1;
		}
	}
	else if(c0 == 0xED)
	{
		if(c1 < 0x80 || 0x9F < c1 || !isUTF8Tail(c2))
		{
			return -1;
		}
	}

	if(c0 < 0xF0)
	{
		c = (static_cast<char32_t>(c0) & 0x0F) << 12;
		c |= (static_cast<char32_t>(c1) & 0x3F) << 6;
		c |= c2 & 0x3F;
		return 3;
	}

	uint8_t c3 = static_cast<uint8_t>(p_utf8_str[3]);
	if(c0 == 0xF0)
	{
		if(c1 < 0x90 || 0xBF > c1 || !isUTF8Tail(c2) || !isUTF8Tail(c3))
		{
			return -1;
		}
	}
	else if(c0 == 0xF1 || c0 == 0xF2 || c0 == 0xF3)
	{
		if(!isUTF8Tail(c1) || !isUTF8Tail(c2) || !isUTF8Tail(c3))
		{
			return false;
		}
	}
	else if(c0 == 0xF4)
	{
		if(c1 < 0x80 || 0x8F > c1 || !isUTF8Tail(c2) || !isUTF8Tail(c3))
		{
			return -1;
		}
	}

	if(c0 <= 0xF4)
	{
		c = (static_cast<char32_t>(c0) & 0x07) << 18;
		c |= (static_cast<char32_t>(c1) & 0x3F) << 12;
		c |= (static_cast<char32_t>(c2) & 0x3F) << 6;
		c |= c3 & 0x3F;
		return 4;
	}

	return -1;
}

int main(int argc, char * argv[])
{
	char s[] = u8"こんにちは";
	char32_t s32[] = U"こんにちは";

	int index = 0;
	char * p = s;
	while(*p)
	{
		char32_t c;
		int length = convertUTF8CharToUTF32Char(p, c);
		if(length <= 0)
		{
			return -1;
		}

		if(s32[index] == c)
		{
			printf("s32[%d] match\n", index);
		}
		else
		{
			printf("s32[%d] unmatch\n", index);
		}

		p += length;
		++index;
	}

	return 0;
}