/* utf-8 encoding and decoding library */

#ifndef UTF8_H
#define UTF8_H

#include "wrmr.h"

#define UTF8_INVALID 0xFFFD /* replacement character */

u32 utf8_decode_len(const char *src, u32 ch_count);
u32 utf8_encode_len(const u32 *src, u32 cp_count);
void utf8_decode(u32 *dst, const char *src, u32 cp_count);
void utf8_encode(char *dst, const u32 *src, u32 cp_count);
u32 utf8_decode_ckd(const char **src, u32 *srcn);

#define UTF8_IMPL
#ifdef UTF8_IMPL

#include <stdbit.h>

/* packed array of 2-bit lengths for codepoints 0..10FFFF */
#define UTF8_CP_LEN_BITS   ((uint64_t)0xFFEAA550000)
#define UTF8_CP_SHIFT(cp)  ((32 - stdc_leading_zeros((uint32_t)(cp))) << 1)
#define UTF8_CP_LEN(cp)    (1 + ((UTF8_CP_LEN_BITS >> UTF8_CP_SHIFT(cp)) & 3))

u32 utf8_encode_len(const u32 *src, u32 cp_count) {
	u32 len = 0;
	while (cp_count) len += UTF8_CP_LEN(src[--cp_count]);
	return len;
}

u32 utf8_decode_len(const char *src, u32 ch_count) {
	u32 i = 0, len = 0;
	while (i < ch_count) {
		i += stdc_leading_ones((u8)src[i]) + ((~src[i] & 0x80) >> 7);
		len++;
	}
	return len;
}

void utf8_encode(char *dst, const u32 *src, u32 cp_count) {
	while (cp_count--) {
		u32 c = *src++;
		ASSUME(c > 0 && c < 0x110000);
		u32 len = UTF8_CP_LEN(c);
		ASSUME(len > 0 && len < 5);
		if (len > 1) {
			for (u32 i = len; --i;) {
				dst[i] = 0x80 | (c & 0x3f);
				c >>= 6;
			}
			*dst = (0xf0 << (4 - len)) | c;
			dst += len;
		} else {
			*dst++ = c;
		}
	}
}

void utf8_decode(u32 *dst, const char *src, u32 cp_count) {
	while (cp_count--) {
		u8 c = *src++;
		u32 bits = stdc_leading_ones(c);
		ASSUME(bits < 5);
		u32 cp = c & (0xff >> bits);
		while (bits-- > 1) {
			c = *src++;
			cp = (cp << 6) | (c & 0x3F);
		}
		*dst++ = cp;
	}
}

u32 utf8_decode_ckd(const char **src, u32 *srcn) {
	u32 n = *srcn;
	if (!n) return 0;
	const u8 *s = (const u8 *)*src;
	u32 cp = *s++;
	n -= 1;
	u32 bits = stdc_leading_ones((u8)cp);
	if (bits) {
		if (bits >= 5 || n < bits) return UTF8_INVALID;
		cp &= 0xff >> bits;
		u32 cp_len = bits - 1;
		ASSUME(cp_len > 0 && cp_len <= 4); 
		for (u32 i = 0; i < cp_len; i++) {
			u8 c = s[i];
			if ((c & 0xC0) != 0x80) return UTF8_INVALID;
			cp = (cp << 6) | (c & 0x3F);
		}
		if (cp > 0x10FFFF) return UTF8_INVALID;
		if (UTF8_CP_LEN(cp) != cp_len) return UTF8_INVALID;
		s += cp_len;
		n -= cp_len;
	}
	*src = (const char *)s;
	*srcn = n;
	return cp;
}

#endif
#endif