blob: 584754051d9dfe662537fe0d36e2873ec9ef7230 [file] [log] [blame] [raw]
/* $Id$ */
/*
* Copyright (c) 2008 Nicholas Marriott <nicm@users.sourceforge.net>
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF MIND, USE, DATA OR PROFITS, WHETHER
* IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING
* OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
#include <sys/types.h>
#include <stdlib.h>
#include <string.h>
#include "tmux.h"
struct utf8_width_entry {
u_int first;
u_int last;
int width;
struct utf8_width_entry *left;
struct utf8_width_entry *right;
};
/* Random order. Not optimal but it'll do for now... */
struct utf8_width_entry utf8_width_table[] = {
{ 0x00951, 0x00954, 0, NULL, NULL },
{ 0x00ccc, 0x00ccd, 0, NULL, NULL },
{ 0x0fff9, 0x0fffb, 0, NULL, NULL },
{ 0x20000, 0x2fffd, 2, NULL, NULL },
{ 0x00ebb, 0x00ebc, 0, NULL, NULL },
{ 0x01932, 0x01932, 0, NULL, NULL },
{ 0x0070f, 0x0070f, 0, NULL, NULL },
{ 0x00a70, 0x00a71, 0, NULL, NULL },
{ 0x02329, 0x02329, 2, NULL, NULL },
{ 0x00acd, 0x00acd, 0, NULL, NULL },
{ 0x00ac7, 0x00ac8, 0, NULL, NULL },
{ 0x00a3c, 0x00a3c, 0, NULL, NULL },
{ 0x009cd, 0x009cd, 0, NULL, NULL },
{ 0x00591, 0x005bd, 0, NULL, NULL },
{ 0x01058, 0x01059, 0, NULL, NULL },
{ 0x0ffe0, 0x0ffe6, 2, NULL, NULL },
{ 0x01100, 0x0115f, 2, NULL, NULL },
{ 0x0fe20, 0x0fe23, 0, NULL, NULL },
{ 0x0302a, 0x0302f, 0, NULL, NULL },
{ 0x01772, 0x01773, 0, NULL, NULL },
{ 0x005bf, 0x005bf, 0, NULL, NULL },
{ 0x006ea, 0x006ed, 0, NULL, NULL },
{ 0x00bc0, 0x00bc0, 0, NULL, NULL },
{ 0x00962, 0x00963, 0, NULL, NULL },
{ 0x01732, 0x01734, 0, NULL, NULL },
{ 0x00d41, 0x00d43, 0, NULL, NULL },
{ 0x01b42, 0x01b42, 0, NULL, NULL },
{ 0x00a41, 0x00a42, 0, NULL, NULL },
{ 0x00eb4, 0x00eb9, 0, NULL, NULL },
{ 0x00b01, 0x00b01, 0, NULL, NULL },
{ 0x00e34, 0x00e3a, 0, NULL, NULL },
{ 0x03040, 0x03098, 2, NULL, NULL },
{ 0x0093c, 0x0093c, 0, NULL, NULL },
{ 0x00c4a, 0x00c4d, 0, NULL, NULL },
{ 0x01032, 0x01032, 0, NULL, NULL },
{ 0x00f37, 0x00f37, 0, NULL, NULL },
{ 0x00901, 0x00902, 0, NULL, NULL },
{ 0x00cbf, 0x00cbf, 0, NULL, NULL },
{ 0x0a806, 0x0a806, 0, NULL, NULL },
{ 0x00dd2, 0x00dd4, 0, NULL, NULL },
{ 0x00f71, 0x00f7e, 0, NULL, NULL },
{ 0x01752, 0x01753, 0, NULL, NULL },
{ 0x1d242, 0x1d244, 0, NULL, NULL },
{ 0x005c1, 0x005c2, 0, NULL, NULL },
{ 0x0309b, 0x0a4cf, 2, NULL, NULL },
{ 0xe0100, 0xe01ef, 0, NULL, NULL },
{ 0x017dd, 0x017dd, 0, NULL, NULL },
{ 0x00600, 0x00603, 0, NULL, NULL },
{ 0x009e2, 0x009e3, 0, NULL, NULL },
{ 0x00cc6, 0x00cc6, 0, NULL, NULL },
{ 0x0a80b, 0x0a80b, 0, NULL, NULL },
{ 0x01712, 0x01714, 0, NULL, NULL },
{ 0x00b3c, 0x00b3c, 0, NULL, NULL },
{ 0x01b00, 0x01b03, 0, NULL, NULL },
{ 0x007eb, 0x007f3, 0, NULL, NULL },
{ 0xe0001, 0xe0001, 0, NULL, NULL },
{ 0x1d185, 0x1d18b, 0, NULL, NULL },
{ 0x0feff, 0x0feff, 0, NULL, NULL },
{ 0x01b36, 0x01b3a, 0, NULL, NULL },
{ 0x01920, 0x01922, 0, NULL, NULL },
{ 0x00670, 0x00670, 0, NULL, NULL },
{ 0x00f90, 0x00f97, 0, NULL, NULL },
{ 0x01927, 0x01928, 0, NULL, NULL },
{ 0x0200b, 0x0200f, 0, NULL, NULL },
{ 0x0ff00, 0x0ff60, 2, NULL, NULL },
{ 0x0f900, 0x0faff, 2, NULL, NULL },
{ 0x0fb1e, 0x0fb1e, 0, NULL, NULL },
{ 0x00cbc, 0x00cbc, 0, NULL, NULL },
{ 0x00eb1, 0x00eb1, 0, NULL, NULL },
{ 0x10a38, 0x10a3a, 0, NULL, NULL },
{ 0x007a6, 0x007b0, 0, NULL, NULL },
{ 0x00f80, 0x00f84, 0, NULL, NULL },
{ 0x005c4, 0x005c5, 0, NULL, NULL },
{ 0x0ac00, 0x0d7a3, 2, NULL, NULL },
{ 0x017c9, 0x017d3, 0, NULL, NULL },
{ 0x00d4d, 0x00d4d, 0, NULL, NULL },
{ 0x1d167, 0x1d169, 0, NULL, NULL },
{ 0x01036, 0x01037, 0, NULL, NULL },
{ 0xe0020, 0xe007f, 0, NULL, NULL },
{ 0x00f35, 0x00f35, 0, NULL, NULL },
{ 0x017b4, 0x017b5, 0, NULL, NULL },
{ 0x0206a, 0x0206f, 0, NULL, NULL },
{ 0x00c46, 0x00c48, 0, NULL, NULL },
{ 0x01939, 0x0193b, 0, NULL, NULL },
{ 0x01dc0, 0x01dca, 0, NULL, NULL },
{ 0x10a0c, 0x10a0f, 0, NULL, NULL },
{ 0x0102d, 0x01030, 0, NULL, NULL },
{ 0x017c6, 0x017c6, 0, NULL, NULL },
{ 0x00ec8, 0x00ecd, 0, NULL, NULL },
{ 0x00b41, 0x00b43, 0, NULL, NULL },
{ 0x017b7, 0x017bd, 0, NULL, NULL },
{ 0x1d173, 0x1d182, 0, NULL, NULL },
{ 0x00a47, 0x00a48, 0, NULL, NULL },
{ 0x0232a, 0x0232a, 2, NULL, NULL },
{ 0x01b3c, 0x01b3c, 0, NULL, NULL },
{ 0x10a01, 0x10a03, 0, NULL, NULL },
{ 0x00ae2, 0x00ae3, 0, NULL, NULL },
{ 0x00483, 0x00486, 0, NULL, NULL },
{ 0x0135f, 0x0135f, 0, NULL, NULL },
{ 0x01a17, 0x01a18, 0, NULL, NULL },
{ 0x006e7, 0x006e8, 0, NULL, NULL },
#ifndef __APPLE__
{ 0x03099, 0x0309a, 0, NULL, NULL },
#endif
{ 0x00b4d, 0x00b4d, 0, NULL, NULL },
{ 0x00ce2, 0x00ce3, 0, NULL, NULL },
{ 0x00bcd, 0x00bcd, 0, NULL, NULL },
{ 0x00610, 0x00615, 0, NULL, NULL },
{ 0x00f99, 0x00fbc, 0, NULL, NULL },
{ 0x009c1, 0x009c4, 0, NULL, NULL },
{ 0x00730, 0x0074a, 0, NULL, NULL },
{ 0x00300, 0x0036f, 0, NULL, NULL },
{ 0x03030, 0x0303e, 2, NULL, NULL },
{ 0x01b34, 0x01b34, 0, NULL, NULL },
{ 0x1d1aa, 0x1d1ad, 0, NULL, NULL },
{ 0x00dca, 0x00dca, 0, NULL, NULL },
{ 0x006d6, 0x006e4, 0, NULL, NULL },
{ 0x00f86, 0x00f87, 0, NULL, NULL },
{ 0x00b3f, 0x00b3f, 0, NULL, NULL },
{ 0x0fe30, 0x0fe6f, 2, NULL, NULL },
{ 0x01039, 0x01039, 0, NULL, NULL },
{ 0x0094d, 0x0094d, 0, NULL, NULL },
{ 0x00c55, 0x00c56, 0, NULL, NULL },
{ 0x00488, 0x00489, 0, NULL, NULL },
{ 0x00e47, 0x00e4e, 0, NULL, NULL },
{ 0x00a81, 0x00a82, 0, NULL, NULL },
{ 0x00ac1, 0x00ac5, 0, NULL, NULL },
{ 0x0202a, 0x0202e, 0, NULL, NULL },
{ 0x00dd6, 0x00dd6, 0, NULL, NULL },
{ 0x018a9, 0x018a9, 0, NULL, NULL },
{ 0x0064b, 0x0065e, 0, NULL, NULL },
{ 0x00abc, 0x00abc, 0, NULL, NULL },
{ 0x00b82, 0x00b82, 0, NULL, NULL },
{ 0x00f39, 0x00f39, 0, NULL, NULL },
{ 0x020d0, 0x020ef, 0, NULL, NULL },
{ 0x01dfe, 0x01dff, 0, NULL, NULL },
{ 0x30000, 0x3fffd, 2, NULL, NULL },
{ 0x00711, 0x00711, 0, NULL, NULL },
{ 0x0fe00, 0x0fe0f, 0, NULL, NULL },
{ 0x0180b, 0x0180d, 0, NULL, NULL },
{ 0x10a3f, 0x10a3f, 0, NULL, NULL },
{ 0x00981, 0x00981, 0, NULL, NULL },
{ 0x0a825, 0x0a826, 0, NULL, NULL },
{ 0x00941, 0x00948, 0, NULL, NULL },
{ 0x01b6b, 0x01b73, 0, NULL, NULL },
{ 0x00e31, 0x00e31, 0, NULL, NULL },
{ 0x0fe10, 0x0fe19, 2, NULL, NULL },
{ 0x00a01, 0x00a02, 0, NULL, NULL },
{ 0x00a4b, 0x00a4d, 0, NULL, NULL },
{ 0x00f18, 0x00f19, 0, NULL, NULL },
{ 0x00fc6, 0x00fc6, 0, NULL, NULL },
{ 0x02e80, 0x03029, 2, NULL, NULL },
{ 0x00b56, 0x00b56, 0, NULL, NULL },
{ 0x009bc, 0x009bc, 0, NULL, NULL },
{ 0x005c7, 0x005c7, 0, NULL, NULL },
{ 0x02060, 0x02063, 0, NULL, NULL },
{ 0x00c3e, 0x00c40, 0, NULL, NULL },
{ 0x10a05, 0x10a06, 0, NULL, NULL },
};
struct utf8_width_entry *utf8_width_root = NULL;
int utf8_overlap(struct utf8_width_entry *, struct utf8_width_entry *);
u_int utf8_combine(const struct utf8_data *);
u_int utf8_width(const struct utf8_data *);
/* Set a single character. */
void
utf8_set(struct utf8_data *utf8data, u_char ch)
{
*utf8data->data = ch;
utf8data->size = 1;
utf8data->width = 1;
}
/*
* Open UTF-8 sequence.
*
* 11000010-11011111 C2-DF start of 2-byte sequence
* 11100000-11101111 E0-EF start of 3-byte sequence
* 11110000-11110100 F0-F4 start of 4-byte sequence
*
* Returns 1 if more UTF-8 to come, 0 if not UTF-8.
*/
int
utf8_open(struct utf8_data *utf8data, u_char ch)
{
memset(utf8data, 0, sizeof *utf8data);
if (ch >= 0xc2 && ch <= 0xdf)
utf8data->size = 2;
else if (ch >= 0xe0 && ch <= 0xef)
utf8data->size = 3;
else if (ch >= 0xf0 && ch <= 0xf4)
utf8data->size = 4;
else
return (0);
utf8_append(utf8data, ch);
return (1);
}
/*
* Append character to UTF-8, closing if finished.
*
* Returns 1 if more UTF-8 data to come, 0 if finished.
*/
int
utf8_append(struct utf8_data *utf8data, u_char ch)
{
if (utf8data->have >= utf8data->size)
fatalx("UTF-8 character overflow");
if (utf8data->size > sizeof utf8data->data)
fatalx("UTF-8 character size too large");
utf8data->data[utf8data->have++] = ch;
if (utf8data->have != utf8data->size)
return (1);
utf8data->width = utf8_width(utf8data);
return (0);
}
/* Check if two width tree entries overlap. */
int
utf8_overlap(struct utf8_width_entry *item1, struct utf8_width_entry *item2)
{
if (item1->first >= item2->first && item1->first <= item2->last)
return (1);
if (item1->last >= item2->first && item1->last <= item2->last)
return (1);
if (item2->first >= item1->first && item2->first <= item1->last)
return (1);
if (item2->last >= item1->first && item2->last <= item1->last)
return (1);
return (0);
}
/* Build UTF-8 width tree. */
void
utf8_build(void)
{
struct utf8_width_entry **ptr, *item, *node;
u_int i, j;
for (i = 0; i < nitems(utf8_width_table); i++) {
item = &utf8_width_table[i];
for (j = 0; j < nitems(utf8_width_table); j++) {
if (i != j && utf8_overlap(item, &utf8_width_table[j]))
log_fatalx("utf8 overlap: %u %u", i, j);
}
ptr = &utf8_width_root;
while (*ptr != NULL) {
node = *ptr;
if (item->last < node->first)
ptr = &(node->left);
else if (item->first > node->last)
ptr = &(node->right);
}
*ptr = item;
}
}
/* Combine UTF-8 into 32-bit Unicode. */
u_int
utf8_combine(const struct utf8_data *utf8data)
{
u_int value;
value = 0xff;
switch (utf8data->size) {
case 1:
value = utf8data->data[0];
break;
case 2:
value = utf8data->data[1] & 0x3f;
value |= (utf8data->data[0] & 0x1f) << 6;
break;
case 3:
value = utf8data->data[2] & 0x3f;
value |= (utf8data->data[1] & 0x3f) << 6;
value |= (utf8data->data[0] & 0x0f) << 12;
break;
case 4:
value = utf8data->data[3] & 0x3f;
value |= (utf8data->data[2] & 0x3f) << 6;
value |= (utf8data->data[1] & 0x3f) << 12;
value |= (utf8data->data[0] & 0x07) << 18;
break;
}
return (value);
}
/* Split a two-byte UTF-8 character. */
u_int
utf8_split2(u_int uc, u_char *ptr)
{
if (uc > 0x7f) {
ptr[0] = (uc >> 6) | 0xc0;
ptr[1] = (uc & 0x3f) | 0x80;
return (2);
}
ptr[0] = uc;
return (1);
}
/* Lookup width of UTF-8 data in tree. */
u_int
utf8_width(const struct utf8_data *utf8data)
{
struct utf8_width_entry *item;
u_int value;
value = utf8_combine(utf8data);
item = utf8_width_root;
while (item != NULL) {
if (value < item->first)
item = item->left;
else if (value > item->last)
item = item->right;
else
return (item->width);
}
return (1);
}
/*
* Encode len characters from src into dst, which is guaranteed to have four
* bytes available for each character from src (for \abc or UTF-8) plus space
* for \0.
*/
int
utf8_strvis(char *dst, const char *src, size_t len, int flag)
{
struct utf8_data utf8data;
const char *start, *end;
int more;
size_t i;
start = dst;
end = src + len;
while (src < end) {
if (utf8_open(&utf8data, *src)) {
more = 1;
while (++src < end && more)
more = utf8_append(&utf8data, *src);
if (!more) {
/* UTF-8 character finished. */
for (i = 0; i < utf8data.size; i++)
*dst++ = utf8data.data[i];
continue;
} else if (utf8data.have > 0) {
/* Not a complete UTF-8 character. */
src -= utf8data.have;
}
}
if (src < end - 1)
dst = vis(dst, src[0], flag, src[1]);
else if (src < end)
dst = vis(dst, src[0], flag, '\0');
src++;
}
*dst = '\0';
return (dst - start);
}
/*
* Convert a string into a buffer of UTF-8 characters. Terminated by size == 0.
* Caller frees.
*/
struct utf8_data *
utf8_fromcstr(const char *src)
{
struct utf8_data *dst;
size_t n;
int more;
dst = NULL;
n = 0;
while (*src != '\0') {
dst = xreallocarray(dst, n + 1, sizeof *dst);
if (utf8_open(&dst[n], *src)) {
more = 1;
while (*++src != '\0' && more)
more = utf8_append(&dst[n], *src);
if (!more) {
n++;
continue;
}
src -= dst[n].have;
}
utf8_set(&dst[n], *src);
src++;
n++;
}
dst = xreallocarray(dst, n + 1, sizeof *dst);
dst[n].size = 0;
return (dst);
}
/* Convert from a buffer of UTF-8 characters into a string. Caller frees. */
char *
utf8_tocstr(struct utf8_data *src)
{
char *dst;
size_t n;
dst = NULL;
n = 0;
for(; src->size != 0; src++) {
dst = xreallocarray(dst, n + src->size, 1);
memcpy(dst + n, src->data, src->size);
n += src->size;
}
dst = xreallocarray(dst, n + 1, 1);
dst[n] = '\0';
return (dst);
}
/* Get width of UTF-8 string. */
u_int
utf8_cstrwidth(const char *s)
{
struct utf8_data tmp;
u_int width;
int more;
width = 0;
while (*s != '\0') {
if (utf8_open(&tmp, *s)) {
more = 1;
while (*++s != '\0' && more)
more = utf8_append(&tmp, *s);
if (!more) {
width += tmp.width;
continue;
}
s -= tmp.have;
}
width++;
s++;
}
return (width);
}
/* Trim UTF-8 string to width. Caller frees. */
char *
utf8_trimcstr(const char *s, u_int width)
{
struct utf8_data *tmp, *next;
char *out;
u_int at;
tmp = utf8_fromcstr(s);
at = 0;
for (next = tmp; next->size != 0; next++) {
if (at + next->width > width) {
next->size = 0;
break;
}
at += next->width;
}
out = utf8_tocstr(tmp);
free(tmp);
return (out);
}