shithub: femtolisp

ref: acc691da05c16e097220430143ea7cab2cc3465e
dir: /3rd/utf/rune.c/

#include "platform.h"

enum
{
	Bit1	= 7,
	Bitx	= 6,
	Bit2	= 5,
	Bit3	= 4,
	Bit4	= 3,
	Bit5	= 2,

	T1	= ((1<<(Bit1+1))-1) ^ 0xFF,	/* 0000 0000 */
	Tx	= ((1<<(Bitx+1))-1) ^ 0xFF,	/* 1000 0000 */
	T2	= ((1<<(Bit2+1))-1) ^ 0xFF,	/* 1100 0000 */
	T3	= ((1<<(Bit3+1))-1) ^ 0xFF,	/* 1110 0000 */
	T4	= ((1<<(Bit4+1))-1) ^ 0xFF,	/* 1111 0000 */
	T5	= ((1<<(Bit5+1))-1) ^ 0xFF,	/* 1111 1000 */

	Rune1	= (1<<(Bit1+0*Bitx))-1,		/* 0000 0000 0000 0000 0111 1111 */
	Rune2	= (1<<(Bit2+1*Bitx))-1,		/* 0000 0000 0000 0111 1111 1111 */
	Rune3	= (1<<(Bit3+2*Bitx))-1,		/* 0000 0000 1111 1111 1111 1111 */
	Rune4	= (1<<(Bit4+3*Bitx))-1,		/* 0011 1111 1111 1111 1111 1111 */

	Maskx	= (1<<Bitx)-1,			/* 0011 1111 */
	Testx	= Maskx ^ 0xFF,			/* 1100 0000 */

	Bad	= Runeerror,
};

int
chartorune(Rune *rune, const char *str)
{
	int c, c1, c2, c3;
	long l;

	/*
	 * one character sequence
	 *	00000-0007F => T1
	 */
	c = *(uchar*)str;
	if(c < Tx) {
		*rune = c;
		return 1;
	}

	/*
	 * two character sequence
	 *	0080-07FF => T2 Tx
	 */
	c1 = *(uchar*)(str+1) ^ Tx;
	if(c1 & Testx)
		goto bad;
	if(c < T3) {
		if(c < T2)
			goto bad;
		l = ((c << Bitx) | c1) & Rune2;
		if(l <= Rune1)
			goto bad;
		*rune = l;
		return 2;
	}

	/*
	 * three character sequence
	 *	0800-FFFF => T3 Tx Tx
	 */
	c2 = *(uchar*)(str+2) ^ Tx;
	if(c2 & Testx)
		goto bad;
	if(c < T4) {
		l = ((((c << Bitx) | c1) << Bitx) | c2) & Rune3;
		if(l <= Rune2)
			goto bad;
		*rune = l;
		return 3;
	}

	/*
	 * four character sequence
	 *	10000-10FFFF => T4 Tx Tx Tx
	 */
	if(UTFmax >= 4) {
		c3 = *(uchar*)(str+3) ^ Tx;
		if(c3 & Testx)
			goto bad;
		if(c < T5) {
			l = ((((((c << Bitx) | c1) << Bitx) | c2) << Bitx) | c3) & Rune4;
			if(l <= Rune3)
				goto bad;
			if(l > Runemax)
				goto bad;
			*rune = l;
			return 4;
		}
	}

	/*
	 * bad decoding
	 */
bad:
	*rune = Bad;
	return 1;
}

int
runetochar(char *str, const Rune *rune)
{
	long c;

	c = *rune;
	if(c > Runemax)
		c = Runeerror;

	/*
	 * one character sequence
	 *	00000-0007F => 00-7F
	 */
	if(c <= Rune1) {
		str[0] = c;
		return 1;
	}

	/*
	 * two character sequence
	 *	0080-07FF => T2 Tx
	 */
	if(c <= Rune2) {
		str[0] = T2 | (c >> 1*Bitx);
		str[1] = Tx | (c & Maskx);
		return 2;
	}

	/*
	 * three character sequence
	 *	0800-FFFF => T3 Tx Tx
	 */
	if(c <= Rune3) {
		str[0] = T3 |  (c >> 2*Bitx);
		str[1] = Tx | ((c >> 1*Bitx) & Maskx);
		str[2] = Tx |  (c & Maskx);
		return 3;
	}

	/*
	 * four character sequence
	 *	10000-1FFFFF => T4 Tx Tx Tx
	 */
	str[0] = T4 |  (c >> 3*Bitx);
	str[1] = Tx | ((c >> 2*Bitx) & Maskx);
	str[2] = Tx | ((c >> 1*Bitx) & Maskx);
	str[3] = Tx |  (c & Maskx);
	return 4;
}

int
runelen(Rune c)
{
	char str[UTFmax];
	return runetochar(str, &c);
}

int
runenlen(const Rune *r, int nrune)
{
	int nb, c;

	nb = 0;
	while(nrune--) {
		c = *r++;
		if(c <= Rune1)
			nb++;
		else
		if(c <= Rune2)
			nb += 2;
		else
		if(c <= Rune3 || c > Runemax)
			nb += 3;
		else
			nb += 4;
	}
	return nb;
}

int
fullrune(const char *str, int n)
{
	int c;

	if(n <= 0)
		return 0;
	c = *(uchar*)str;
	if(c < Tx)
		return 1;
	if(c < T3)
		return n >= 2;
	if(UTFmax == 3 || c < T4)
		return n >= 3;
	return n >= 4;
}