bin/less/charset.c

   1 /*
   2  * Copyright (C) 1984-2012  Mark Nudelman
   3  * Modified for use with illumos by Garrett D'Amore.
   4  * Copyright 2014 Garrett D'Amore <garrett@damore.org>
   5  *
   6  * You may distribute under the terms of either the GNU General Public
   7  * License or the Less License, as specified in the README file.
   8  *
   9  * For more information, see the README file.
  10  */
  11
  12 /*
  13  * Functions to define the character set
  14  * and do things specific to the character set.
  15  */
  16
  17 #include <ctype.h>
  18 #include <langinfo.h>
  19 #include <locale.h>
  20
  21 #include "charset.h"
  22 #include "less.h"
  23
  24 int utf_mode = 0;
  25
  26 static const char *binfmt = NULL;
  27 static const char *utfbinfmt = NULL;
  28 int binattr = AT_STANDOUT;
  29
  30 static int
  31 checkfmt(const char *s)
  32 {
  33         char c;
  34         int seen = 0;
  35
  36         /* %[<flags>][<width>][.<prec>][<type>]<fmt> */
  37
  38         if (*s == '*') {                /* skip leading attribute if there */
  39                 s++;
  40                 if (*s == '\0' || strchr("dksu", *s) == NULL) {
  41                         return (-1);
  42                 }
  43                 s++;
  44         }
  45
  46         while ((c = *s++) != 0) {
  47                 if (!isascii(c) || !isprint(c)) {
  48                         return (-1);
  49                 }
  50                 if (c != '%') {
  51                         continue;
  52                 }
  53                 if (*s == '%') {        /* % escaped with second % */
  54                         s++;
  55                         continue;
  56                 }
  57                 if (seen) {
  58                         return (-1);    /* 2nd % format item! */
  59                 }
  60                 /* skip flags */
  61                 while (*s != '\0' && strchr(" '+-0#", *s) != NULL) {
  62                         s++;
  63                 }
  64                 while (isdigit(*s)) {                   /* skip width */
  65                         s++;
  66                 }
  67                 if (*s == '.') {                /* skip precision */
  68                         s++;
  69                         while (isdigit(*s)) {
  70                                 s++;
  71                         }
  72                 }
  73                 /* type width specifications, only "l", "h", and "hh" valid */
  74                 if (*s == 'l') {
  75                         s++;
  76                 } else if (*s == 'h') {
  77                         s++;
  78                         if (*s == 'h')
  79                                 s++;
  80                 }
  81
  82                 if (*s == '\0' || strchr("cCdiouxX", *s) == NULL) {
  83                         /* bad or evil format character (%s, %n, etc.) */
  84                         return (-1);
  85                 }
  86
  87                 seen = 1;
  88         }
  89
  90         return (0);
  91 }
  92
  93 /*
  94  * Define the printing format for control (or binary utf) chars.
  95  */
  96 static void
  97 setbinfmt(char *e, const char **fmtvarptr, const char *default_fmt)
  98 {
  99         const char *s;
 100
 101         if (((s = lgetenv(e)) == NULL) || (*s == 0)) {
 102                 s = default_fmt;
 103                 goto attr;
 104         }
 105
 106         if (s != NULL && *s != 0) {
 107                 if (checkfmt(s) < 0) {
 108                         s = default_fmt;
 109                         goto attr;
 110                 }
 111         }
 112
 113         /*
 114          * Select the attributes if it starts with "*".
 115          */
 116 attr:
 117         if (*s == '*') {
 118                 switch (s[1]) {
 119                 case 'd':  binattr = AT_BOLD; break;
 120                 case 'k':  binattr = AT_BLINK; break;
 121                 case 's':  binattr = AT_STANDOUT; break;
 122                 case 'u':  binattr = AT_UNDERLINE; break;
 123                 default:   binattr = AT_NORMAL; break;
 124                 }
 125                 s += 2;
 126         }
 127         *fmtvarptr = s;
 128 }
 129
 130 /*
 131  * Initialize charset data structures.
 132  */
 133 void
 134 init_charset(void)
 135 {
 136         char *s;
 137
 138         setlocale(LC_ALL, "");
 139
 140         s = nl_langinfo(CODESET);
 141         if (s && strcasecmp(s, "utf-8") == 0)
 142                 utf_mode = 1;
 143
 144         setbinfmt("LESSBINFMT", &binfmt, "*s<%02X>");
 145         setbinfmt("LESSUTFBINFMT", &utfbinfmt, "<U+%04lX>");
 146 }
 147
 148 /*
 149  * Is a given character a "binary" character?
 150  */
 151 int
 152 binary_char(LWCHAR c)
 153 {
 154         if (utf_mode)
 155                 return (is_ubin_char(c));
 156         c &= 0377;
 157         return (!isprint((unsigned char)c) && !iscntrl((unsigned char)c));
 158 }
 159
 160 /*
 161  * Is a given character a "control" character?
 162  */
 163 int
 164 control_char(LWCHAR c)
 165 {
 166         c &= 0377;
 167         if (utf_mode)
 168                 return (iscntrl((unsigned char)c));
 169         return (iscntrl((unsigned char)c) || !isprint((unsigned char)c));
 170 }
 171
 172 /*
 173  * Return the printable form of a character.
 174  * For example, in the "ascii" charset '\3' is printed as "^C".
 175  */
 176 char *
 177 prchar(LWCHAR c)
 178 {
 179         /* {{ This buffer can be overrun if LESSBINFMT is a long string. }} */
 180         static char buf[32];
 181
 182         c &= 0377;
 183         if ((c < 128 || !utf_mode) && !control_char(c))
 184                 (void) snprintf(buf, sizeof (buf), "%c", (int)c);
 185         else if (c == ESC)
 186                 (void) strlcpy(buf, "ESC", sizeof (buf));
 187         else if (c < 128 && !control_char(c ^ 0100))
 188                 (void) snprintf(buf, sizeof (buf), "^%c", (int)(c ^ 0100));
 189         else
 190                 (void) snprintf(buf, sizeof (buf), binfmt, c);
 191         return (buf);
 192 }
 193
 194 /*
 195  * Return the printable form of a UTF-8 character.
 196  */
 197 char *
 198 prutfchar(LWCHAR ch)
 199 {
 200         static char buf[32];
 201
 202         if (ch == ESC) {
 203                 (void) strlcpy(buf, "ESC", sizeof (buf));
 204         } else if (ch < 128 && control_char(ch)) {
 205                 if (!control_char(ch ^ 0100))
 206                         (void) snprintf(buf, sizeof (buf), "^%c",
 207                             ((char)ch) ^ 0100);
 208                 else
 209                         (void) snprintf(buf, sizeof (buf), binfmt, (char)ch);
 210         } else if (is_ubin_char(ch)) {
 211                 (void) snprintf(buf, sizeof (buf), utfbinfmt, ch);
 212         } else {
 213                 int len;
 214                 if (ch >= 0x80000000) {
 215                         len = 3;
 216                         ch = 0xFFFD;
 217                 } else {
 218                         len =   (ch < 0x80) ? 1
 219                             : (ch < 0x800) ? 2
 220                             : (ch < 0x10000) ? 3
 221                             : (ch < 0x200000) ? 4
 222                             : (ch < 0x4000000) ? 5
 223                             : 6;
 224                 }
 225                 buf[len] = '\0';
 226                 if (len == 1) {
 227                         *buf = (char)ch;
 228                 } else {
 229                         *buf = ((1 << len) - 1) << (8 - len);
 230                         while (--len > 0) {
 231                                 buf[len] = (char)(0x80 | (ch & 0x3F));
 232                                 ch >>= 6;
 233                         }
 234                         *buf |= ch;
 235                 }
 236         }
 237         return (buf);
 238 }
 239
 240 /*
 241  * Get the length of a UTF-8 character in bytes.
 242  */
 243 int
 244 utf_len(char ch)
 245 {
 246         if ((ch & 0x80) == 0)
 247                 return (1);
 248         if ((ch & 0xE0) == 0xC0)
 249                 return (2);
 250         if ((ch & 0xF0) == 0xE0)
 251                 return (3);
 252         if ((ch & 0xF8) == 0xF0)
 253                 return (4);
 254         if ((ch & 0xFC) == 0xF8)
 255                 return (5);
 256         if ((ch & 0xFE) == 0xFC)
 257                 return (6);
 258         /* Invalid UTF-8 encoding. */
 259         return (1);
 260 }
 261
 262 /*
 263  * Is a UTF-8 character well-formed?
 264  */
 265 int
 266 is_utf8_well_formed(const char *s)
 267 {
 268         int i;
 269         int len;
 270
 271         if (IS_UTF8_INVALID(s[0]))
 272                 return (0);
 273
 274         len = utf_len((char)s[0]);
 275         if (len == 1)
 276                 return (1);
 277         if (len == 2) {
 278                 if ((unsigned char)(s[0]) < 0xC2)
 279                         return (0);
 280         } else {
 281                 unsigned char mask;
 282                 mask = (~((1 << (8-len)) - 1)) & 0xFF;
 283                 if (s[0] == mask && (s[1] & mask) == 0x80)
 284                         return (0);
 285         }
 286
 287         for (i = 1; i < len; i++)
 288                 if (!IS_UTF8_TRAIL(s[i]))
 289                         return (0);
 290         return (1);
 291 }
 292
 293 /*
 294  * Get the value of a UTF-8 character.
 295  */
 296 LWCHAR
 297 get_wchar(const char *p)
 298 {
 299         switch (utf_len(p[0])) {
 300         case 1:
 301         default:
 302                 /* 0xxxxxxx */
 303                 return (LWCHAR)
 304                     (p[0] & 0xFF);
 305         case 2:
 306                 /* 110xxxxx 10xxxxxx */
 307                 return (LWCHAR) (
 308                     ((p[0] & 0x1F) << 6) |
 309                     (p[1] & 0x3F));
 310         case 3:
 311                 /* 1110xxxx 10xxxxxx 10xxxxxx */
 312                 return (LWCHAR) (
 313                     ((p[0] & 0x0F) << 12) |
 314                     ((p[1] & 0x3F) << 6) |
 315                     (p[2] & 0x3F));
 316         case 4:
 317                 /* 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx */
 318                 return (LWCHAR) (
 319                     ((p[0] & 0x07) << 18) |
 320                     ((p[1] & 0x3F) << 12) |
 321                     ((p[2] & 0x3F) << 6) |
 322                     (p[3] & 0x3F));
 323         case 5:
 324                 /* 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx */
 325                 return (LWCHAR) (
 326                     ((p[0] & 0x03) << 24) |
 327                     ((p[1] & 0x3F) << 18) |
 328                     ((p[2] & 0x3F) << 12) |
 329                     ((p[3] & 0x3F) << 6) |
 330                     (p[4] & 0x3F));
 331         case 6:
 332                 /* 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx */
 333                 return (LWCHAR) (
 334                     ((p[0] & 0x01) << 30) |
 335                     ((p[1] & 0x3F) << 24) |
 336                     ((p[2] & 0x3F) << 18) |
 337                     ((p[3] & 0x3F) << 12) |
 338                     ((p[4] & 0x3F) << 6) |
 339                     (p[5] & 0x3F));
 340         }
 341 }
 342
 343 /*
 344  * Store a character into a UTF-8 string.
 345  */
 346 void
 347 put_wchar(char **pp, LWCHAR ch)
 348 {
 349         if (!utf_mode || ch < 0x80) {
 350                 /* 0xxxxxxx */
 351                 *(*pp)++ = (char)ch;
 352         } else if (ch < 0x800) {
 353                 /* 110xxxxx 10xxxxxx */
 354                 *(*pp)++ = (char)(0xC0 | ((ch >> 6) & 0x1F));
 355                 *(*pp)++ = (char)(0x80 | (ch & 0x3F));
 356         } else if (ch < 0x10000) {
 357                 /* 1110xxxx 10xxxxxx 10xxxxxx */
 358                 *(*pp)++ = (char)(0xE0 | ((ch >> 12) & 0x0F));
 359                 *(*pp)++ = (char)(0x80 | ((ch >> 6) & 0x3F));
 360                 *(*pp)++ = (char)(0x80 | (ch & 0x3F));
 361         } else if (ch < 0x200000) {
 362                 /* 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx */
 363                 *(*pp)++ = (char)(0xF0 | ((ch >> 18) & 0x07));
 364                 *(*pp)++ = (char)(0x80 | ((ch >> 12) & 0x3F));
 365                 *(*pp)++ = (char)(0x80 | ((ch >> 6) & 0x3F));
 366                 *(*pp)++ = (char)(0x80 | (ch & 0x3F));
 367         } else if (ch < 0x4000000) {
 368                 /* 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx */
 369                 *(*pp)++ = (char)(0xF0 | ((ch >> 24) & 0x03));
 370                 *(*pp)++ = (char)(0x80 | ((ch >> 18) & 0x3F));
 371                 *(*pp)++ = (char)(0x80 | ((ch >> 12) & 0x3F));
 372                 *(*pp)++ = (char)(0x80 | ((ch >> 6) & 0x3F));
 373                 *(*pp)++ = (char)(0x80 | (ch & 0x3F));
 374         } else {
 375                 /* 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx */
 376                 *(*pp)++ = (char)(0xF0 | ((ch >> 30) & 0x01));
 377                 *(*pp)++ = (char)(0x80 | ((ch >> 24) & 0x3F));
 378                 *(*pp)++ = (char)(0x80 | ((ch >> 18) & 0x3F));
 379                 *(*pp)++ = (char)(0x80 | ((ch >> 12) & 0x3F));
 380                 *(*pp)++ = (char)(0x80 | ((ch >> 6) & 0x3F));
 381                 *(*pp)++ = (char)(0x80 | (ch & 0x3F));
 382         }
 383 }
 384
 385 /*
 386  * Step forward or backward one character in a string.
 387  */
 388 LWCHAR
 389 step_char(char **pp, int dir, char *limit)
 390 {
 391         LWCHAR ch;
 392         int len;
 393         char *p = *pp;
 394
 395         if (!utf_mode) {
 396                 /* It's easy if chars are one byte. */
 397                 if (dir > 0)
 398                         ch = (LWCHAR) ((p < limit) ? *p++ : 0);
 399                 else
 400                         ch = (LWCHAR) ((p > limit) ? *--p : 0);
 401         } else if (dir > 0) {
 402                 len = utf_len(*p);
 403                 if (p + len > limit) {
 404                         ch = 0;
 405                         p = limit;
 406                 } else {
 407                         ch = get_wchar(p);
 408                         p += len;
 409                 }
 410         } else {
 411                 while (p > limit && IS_UTF8_TRAIL(p[-1]))
 412                         p--;
 413                 if (p > limit)
 414                         ch = get_wchar(--p);
 415                 else
 416                         ch = 0;
 417         }
 418         *pp = p;
 419         return (ch);
 420 }
 421
 422 /*
 423  * Unicode characters data
 424  */
 425 struct wchar_range { LWCHAR first, last; };
 426
 427 /*
 428  * Characters with general category values
 429  *      Mn: Mark, Nonspacing
 430  *      Me: Mark, Enclosing
 431  * Last synched with
 432  *      <http://www.unicode.org/Public/5.0.0/ucd/UnicodeData-5.0.0d7.txt>
 433  *      dated 2005-11-30T00:58:48Z
 434  */
 435 static struct wchar_range comp_table[] = {
 436         {  0x0300,  0x036F} /* Mn */, {  0x0483,  0x0486} /* Mn */,
 437         {  0x0488,  0x0489} /* Me */,
 438         {  0x0591,  0x05BD} /* Mn */, {  0x05BF,  0x05BF} /* Mn */,
 439         {  0x05C1,  0x05C2} /* Mn */, {  0x05C4,  0x05C5} /* Mn */,
 440         {  0x05C7,  0x05C7} /* Mn */, {  0x0610,  0x0615} /* Mn */,
 441         {  0x064B,  0x065E} /* Mn */, {  0x0670,  0x0670} /* Mn */,
 442         {  0x06D6,  0x06DC} /* Mn */,
 443         {  0x06DE,  0x06DE} /* Me */,
 444         {  0x06DF,  0x06E4} /* Mn */, {  0x06E7,  0x06E8} /* Mn */,
 445         {  0x06EA,  0x06ED} /* Mn */, {  0x0711,  0x0711} /* Mn */,
 446         {  0x0730,  0x074A} /* Mn */, {  0x07A6,  0x07B0} /* Mn */,
 447         {  0x07EB,  0x07F3} /* Mn */, {  0x0901,  0x0902} /* Mn */,
 448         {  0x093C,  0x093C} /* Mn */, {  0x0941,  0x0948} /* Mn */,
 449         {  0x094D,  0x094D} /* Mn */, {  0x0951,  0x0954} /* Mn */,
 450         {  0x0962,  0x0963} /* Mn */, {  0x0981,  0x0981} /* Mn */,
 451         {  0x09BC,  0x09BC} /* Mn */, {  0x09C1,  0x09C4} /* Mn */,
 452         {  0x09CD,  0x09CD} /* Mn */, {  0x09E2,  0x09E3} /* Mn */,
 453         {  0x0A01,  0x0A02} /* Mn */, {  0x0A3C,  0x0A3C} /* Mn */,
 454         {  0x0A41,  0x0A42} /* Mn */, {  0x0A47,  0x0A48} /* Mn */,
 455         {  0x0A4B,  0x0A4D} /* Mn */, {  0x0A70,  0x0A71} /* Mn */,
 456         {  0x0A81,  0x0A82} /* Mn */, {  0x0ABC,  0x0ABC} /* Mn */,
 457         {  0x0AC1,  0x0AC5} /* Mn */, {  0x0AC7,  0x0AC8} /* Mn */,
 458         {  0x0ACD,  0x0ACD} /* Mn */, {  0x0AE2,  0x0AE3} /* Mn */,
 459         {  0x0B01,  0x0B01} /* Mn */, {  0x0B3C,  0x0B3C} /* Mn */,
 460         {  0x0B3F,  0x0B3F} /* Mn */, {  0x0B41,  0x0B43} /* Mn */,
 461         {  0x0B4D,  0x0B4D} /* Mn */, {  0x0B56,  0x0B56} /* Mn */,
 462         {  0x0B82,  0x0B82} /* Mn */, {  0x0BC0,  0x0BC0} /* Mn */,
 463         {  0x0BCD,  0x0BCD} /* Mn */, {  0x0C3E,  0x0C40} /* Mn */,
 464         {  0x0C46,  0x0C48} /* Mn */, {  0x0C4A,  0x0C4D} /* Mn */,
 465         {  0x0C55,  0x0C56} /* Mn */, {  0x0CBC,  0x0CBC} /* Mn */,
 466         {  0x0CBF,  0x0CBF} /* Mn */, {  0x0CC6,  0x0CC6} /* Mn */,
 467         {  0x0CCC,  0x0CCD} /* Mn */, {  0x0CE2,  0x0CE3} /* Mn */,
 468         {  0x0D41,  0x0D43} /* Mn */, {  0x0D4D,  0x0D4D} /* Mn */,
 469         {  0x0DCA,  0x0DCA} /* Mn */, {  0x0DD2,  0x0DD4} /* Mn */,
 470         {  0x0DD6,  0x0DD6} /* Mn */, {  0x0E31,  0x0E31} /* Mn */,
 471         {  0x0E34,  0x0E3A} /* Mn */, {  0x0E47,  0x0E4E} /* Mn */,
 472         {  0x0EB1,  0x0EB1} /* Mn */, {  0x0EB4,  0x0EB9} /* Mn */,
 473         {  0x0EBB,  0x0EBC} /* Mn */, {  0x0EC8,  0x0ECD} /* Mn */,
 474         {  0x0F18,  0x0F19} /* Mn */, {  0x0F35,  0x0F35} /* Mn */,
 475         {  0x0F37,  0x0F37} /* Mn */, {  0x0F39,  0x0F39} /* Mn */,
 476         {  0x0F71,  0x0F7E} /* Mn */, {  0x0F80,  0x0F84} /* Mn */,
 477         {  0x0F86,  0x0F87} /* Mn */, {  0x0F90,  0x0F97} /* Mn */,
 478         {  0x0F99,  0x0FBC} /* Mn */, {  0x0FC6,  0x0FC6} /* Mn */,
 479         {  0x102D,  0x1030} /* Mn */, {  0x1032,  0x1032} /* Mn */,
 480         {  0x1036,  0x1037} /* Mn */, {  0x1039,  0x1039} /* Mn */,
 481         {  0x1058,  0x1059} /* Mn */, {  0x135F,  0x135F} /* Mn */,
 482         {  0x1712,  0x1714} /* Mn */, {  0x1732,  0x1734} /* Mn */,
 483         {  0x1752,  0x1753} /* Mn */, {  0x1772,  0x1773} /* Mn */,
 484         {  0x17B7,  0x17BD} /* Mn */, {  0x17C6,  0x17C6} /* Mn */,
 485         {  0x17C9,  0x17D3} /* Mn */, {  0x17DD,  0x17DD} /* Mn */,
 486         {  0x180B,  0x180D} /* Mn */, {  0x18A9,  0x18A9} /* Mn */,
 487         {  0x1920,  0x1922} /* Mn */, {  0x1927,  0x1928} /* Mn */,
 488         {  0x1932,  0x1932} /* Mn */, {  0x1939,  0x193B} /* Mn */,
 489         {  0x1A17,  0x1A18} /* Mn */, {  0x1B00,  0x1B03} /* Mn */,
 490         {  0x1B34,  0x1B34} /* Mn */, {  0x1B36,  0x1B3A} /* Mn */,
 491         {  0x1B3C,  0x1B3C} /* Mn */, {  0x1B42,  0x1B42} /* Mn */,
 492         {  0x1B6B,  0x1B73} /* Mn */, {  0x1DC0,  0x1DCA} /* Mn */,
 493         {  0x1DFE,  0x1DFF} /* Mn */, {  0x20D0,  0x20DC} /* Mn */,
 494         {  0x20DD,  0x20E0} /* Me */,
 495         {  0x20E1,  0x20E1} /* Mn */,
 496         {  0x20E2,  0x20E4} /* Me */,
 497         {  0x20E5,  0x20EF} /* Mn */, {  0x302A,  0x302F} /* Mn */,
 498         {  0x3099,  0x309A} /* Mn */, {  0xA806,  0xA806} /* Mn */,
 499         {  0xA80B,  0xA80B} /* Mn */, {  0xA825,  0xA826} /* Mn */,
 500         {  0xFB1E,  0xFB1E} /* Mn */, {  0xFE00,  0xFE0F} /* Mn */,
 501         {  0xFE20,  0xFE23} /* Mn */, { 0x10A01, 0x10A03} /* Mn */,
 502         { 0x10A05, 0x10A06} /* Mn */, { 0x10A0C, 0x10A0F} /* Mn */,
 503         { 0x10A38, 0x10A3A} /* Mn */, { 0x10A3F, 0x10A3F} /* Mn */,
 504         { 0x1D167, 0x1D169} /* Mn */, { 0x1D17B, 0x1D182} /* Mn */,
 505         { 0x1D185, 0x1D18B} /* Mn */, { 0x1D1AA, 0x1D1AD} /* Mn */,
 506         { 0x1D242, 0x1D244} /* Mn */, { 0xE0100, 0xE01EF} /* Mn */,
 507 };
 508
 509 /*
 510  * Special pairs, not ranges.
 511  */
 512 static struct wchar_range comb_table[] = {
 513         {0x0644, 0x0622}, {0x0644, 0x0623}, {0x0644, 0x0625}, {0x0644, 0x0627},
 514 };
 515
 516 /*
 517  * Characters with general category values
 518  *      Cc: Other, Control
 519  *      Cf: Other, Format
 520  *      Cs: Other, Surrogate
 521  *      Co: Other, Private Use
 522  *      Cn: Other, Not Assigned
 523  *      Zl: Separator, Line
 524  *      Zp: Separator, Paragraph
 525  * Last synched with
 526  *      <http://www.unicode.org/Public/5.0.0/ucd/UnicodeData-5.0.0d7.txt>
 527  *      dated 2005-11-30T00:58:48Z
 528  */
 529 static struct wchar_range ubin_table[] = {
 530         {  0x0000,  0x0007} /* Cc */,
 531         {  0x000B,  0x000C} /* Cc */,
 532         {  0x000E,  0x001A} /* Cc */,
 533         {  0x001C,  0x001F} /* Cc */,
 534         {  0x007F,  0x009F} /* Cc */,
 535 #if 0
 536         {  0x00AD,  0x00AD} /* Cf */,
 537 #endif
 538         {  0x0370,  0x0373} /* Cn */, {  0x0376,  0x0379} /* Cn */,
 539         {  0x037F,  0x0383} /* Cn */, {  0x038B,  0x038B} /* Cn */,
 540         {  0x038D,  0x038D} /* Cn */, {  0x03A2,  0x03A2} /* Cn */,
 541         {  0x03CF,  0x03CF} /* Cn */, {  0x0487,  0x0487} /* Cn */,
 542         {  0x0514,  0x0530} /* Cn */, {  0x0557,  0x0558} /* Cn */,
 543         {  0x0560,  0x0560} /* Cn */, {  0x0588,  0x0588} /* Cn */,
 544         {  0x058B,  0x0590} /* Cn */, {  0x05C8,  0x05CF} /* Cn */,
 545         {  0x05EB,  0x05EF} /* Cn */, {  0x05F5,  0x05FF} /* Cn */,
 546 #if 0
 547         {  0x0600,  0x0603} /* Cf */,
 548 #endif
 549         {  0x0604,  0x060A} /* Cn */, {  0x0616,  0x061A} /* Cn */,
 550         {  0x061C,  0x061D} /* Cn */, {  0x0620,  0x0620} /* Cn */,
 551         {  0x063B,  0x063F} /* Cn */, {  0x065F,  0x065F} /* Cn */,
 552 #if 0
 553         {  0x06DD,  0x06DD} /* Cf */,
 554 #endif
 555         {  0x070E,  0x070E} /* Cn */,
 556 #if 0
 557         {  0x070F,  0x070F} /* Cf */,
 558 #endif
 559         {  0x074B,  0x074C} /* Cn */, {  0x076E,  0x077F} /* Cn */,
 560         {  0x07B2,  0x07BF} /* Cn */, {  0x07FB,  0x0900} /* Cn */,
 561         {  0x093A,  0x093B} /* Cn */, {  0x094E,  0x094F} /* Cn */,
 562         {  0x0955,  0x0957} /* Cn */, {  0x0971,  0x097A} /* Cn */,
 563         {  0x0980,  0x0980} /* Cn */, {  0x0984,  0x0984} /* Cn */,
 564         {  0x098D,  0x098E} /* Cn */, {  0x0991,  0x0992} /* Cn */,
 565         {  0x09A9,  0x09A9} /* Cn */, {  0x09B1,  0x09B1} /* Cn */,
 566         {  0x09B3,  0x09B5} /* Cn */, {  0x09BA,  0x09BB} /* Cn */,
 567         {  0x09C5,  0x09C6} /* Cn */, {  0x09C9,  0x09CA} /* Cn */,
 568         {  0x09CF,  0x09D6} /* Cn */, {  0x09D8,  0x09DB} /* Cn */,
 569         {  0x09DE,  0x09DE} /* Cn */, {  0x09E4,  0x09E5} /* Cn */,
 570         {  0x09FB,  0x0A00} /* Cn */, {  0x0A04,  0x0A04} /* Cn */,
 571         {  0x0A0B,  0x0A0E} /* Cn */, {  0x0A11,  0x0A12} /* Cn */,
 572         {  0x0A29,  0x0A29} /* Cn */, {  0x0A31,  0x0A31} /* Cn */,
 573         {  0x0A34,  0x0A34} /* Cn */, {  0x0A37,  0x0A37} /* Cn */,
 574         {  0x0A3A,  0x0A3B} /* Cn */, {  0x0A3D,  0x0A3D} /* Cn */,
 575         {  0x0A43,  0x0A46} /* Cn */, {  0x0A49,  0x0A4A} /* Cn */,
 576         {  0x0A4E,  0x0A58} /* Cn */, {  0x0A5D,  0x0A5D} /* Cn */,
 577         {  0x0A5F,  0x0A65} /* Cn */, {  0x0A75,  0x0A80} /* Cn */,
 578         {  0x0A84,  0x0A84} /* Cn */, {  0x0A8E,  0x0A8E} /* Cn */,
 579         {  0x0A92,  0x0A92} /* Cn */, {  0x0AA9,  0x0AA9} /* Cn */,
 580         {  0x0AB1,  0x0AB1} /* Cn */, {  0x0AB4,  0x0AB4} /* Cn */,
 581         {  0x0ABA,  0x0ABB} /* Cn */, {  0x0AC6,  0x0AC6} /* Cn */,
 582         {  0x0ACA,  0x0ACA} /* Cn */, {  0x0ACE,  0x0ACF} /* Cn */,
 583         {  0x0AD1,  0x0ADF} /* Cn */, {  0x0AE4,  0x0AE5} /* Cn */,
 584         {  0x0AF0,  0x0AF0} /* Cn */, {  0x0AF2,  0x0B00} /* Cn */,
 585         {  0x0B04,  0x0B04} /* Cn */, {  0x0B0D,  0x0B0E} /* Cn */,
 586         {  0x0B11,  0x0B12} /* Cn */, {  0x0B29,  0x0B29} /* Cn */,
 587         {  0x0B31,  0x0B31} /* Cn */, {  0x0B34,  0x0B34} /* Cn */,
 588         {  0x0B3A,  0x0B3B} /* Cn */, {  0x0B44,  0x0B46} /* Cn */,
 589         {  0x0B49,  0x0B4A} /* Cn */, {  0x0B4E,  0x0B55} /* Cn */,
 590         {  0x0B58,  0x0B5B} /* Cn */, {  0x0B5E,  0x0B5E} /* Cn */,
 591         {  0x0B62,  0x0B65} /* Cn */, {  0x0B72,  0x0B81} /* Cn */,
 592         {  0x0B84,  0x0B84} /* Cn */, {  0x0B8B,  0x0B8D} /* Cn */,
 593         {  0x0B91,  0x0B91} /* Cn */, {  0x0B96,  0x0B98} /* Cn */,
 594         {  0x0B9B,  0x0B9B} /* Cn */, {  0x0B9D,  0x0B9D} /* Cn */,
 595         {  0x0BA0,  0x0BA2} /* Cn */, {  0x0BA5,  0x0BA7} /* Cn */,
 596         {  0x0BAB,  0x0BAD} /* Cn */, {  0x0BBA,  0x0BBD} /* Cn */,
 597         {  0x0BC3,  0x0BC5} /* Cn */, {  0x0BC9,  0x0BC9} /* Cn */,
 598         {  0x0BCE,  0x0BD6} /* Cn */, {  0x0BD8,  0x0BE5} /* Cn */,
 599         {  0x0BFB,  0x0C00} /* Cn */, {  0x0C04,  0x0C04} /* Cn */,
 600         {  0x0C0D,  0x0C0D} /* Cn */, {  0x0C11,  0x0C11} /* Cn */,
 601         {  0x0C29,  0x0C29} /* Cn */, {  0x0C34,  0x0C34} /* Cn */,
 602         {  0x0C3A,  0x0C3D} /* Cn */, {  0x0C45,  0x0C45} /* Cn */,
 603         {  0x0C49,  0x0C49} /* Cn */, {  0x0C4E,  0x0C54} /* Cn */,
 604         {  0x0C57,  0x0C5F} /* Cn */, {  0x0C62,  0x0C65} /* Cn */,
 605         {  0x0C70,  0x0C81} /* Cn */, {  0x0C84,  0x0C84} /* Cn */,
 606         {  0x0C8D,  0x0C8D} /* Cn */, {  0x0C91,  0x0C91} /* Cn */,
 607         {  0x0CA9,  0x0CA9} /* Cn */, {  0x0CB4,  0x0CB4} /* Cn */,
 608         {  0x0CBA,  0x0CBB} /* Cn */, {  0x0CC5,  0x0CC5} /* Cn */,
 609         {  0x0CC9,  0x0CC9} /* Cn */, {  0x0CCE,  0x0CD4} /* Cn */,
 610         {  0x0CD7,  0x0CDD} /* Cn */, {  0x0CDF,  0x0CDF} /* Cn */,
 611         {  0x0CE4,  0x0CE5} /* Cn */, {  0x0CF0,  0x0CF0} /* Cn */,
 612         {  0x0CF3,  0x0D01} /* Cn */, {  0x0D04,  0x0D04} /* Cn */,
 613         {  0x0D0D,  0x0D0D} /* Cn */, {  0x0D11,  0x0D11} /* Cn */,
 614         {  0x0D29,  0x0D29} /* Cn */, {  0x0D3A,  0x0D3D} /* Cn */,
 615         {  0x0D44,  0x0D45} /* Cn */, {  0x0D49,  0x0D49} /* Cn */,
 616         {  0x0D4E,  0x0D56} /* Cn */, {  0x0D58,  0x0D5F} /* Cn */,
 617         {  0x0D62,  0x0D65} /* Cn */, {  0x0D70,  0x0D81} /* Cn */,
 618         {  0x0D84,  0x0D84} /* Cn */, {  0x0D97,  0x0D99} /* Cn */,
 619         {  0x0DB2,  0x0DB2} /* Cn */, {  0x0DBC,  0x0DBC} /* Cn */,
 620         {  0x0DBE,  0x0DBF} /* Cn */, {  0x0DC7,  0x0DC9} /* Cn */,
 621         {  0x0DCB,  0x0DCE} /* Cn */, {  0x0DD5,  0x0DD5} /* Cn */,
 622         {  0x0DD7,  0x0DD7} /* Cn */, {  0x0DE0,  0x0DF1} /* Cn */,
 623         {  0x0DF5,  0x0E00} /* Cn */, {  0x0E3B,  0x0E3E} /* Cn */,
 624         {  0x0E5C,  0x0E80} /* Cn */, {  0x0E83,  0x0E83} /* Cn */,
 625         {  0x0E85,  0x0E86} /* Cn */, {  0x0E89,  0x0E89} /* Cn */,
 626         {  0x0E8B,  0x0E8C} /* Cn */, {  0x0E8E,  0x0E93} /* Cn */,
 627         {  0x0E98,  0x0E98} /* Cn */, {  0x0EA0,  0x0EA0} /* Cn */,
 628         {  0x0EA4,  0x0EA4} /* Cn */, {  0x0EA6,  0x0EA6} /* Cn */,
 629         {  0x0EA8,  0x0EA9} /* Cn */, {  0x0EAC,  0x0EAC} /* Cn */,
 630         {  0x0EBA,  0x0EBA} /* Cn */, {  0x0EBE,  0x0EBF} /* Cn */,
 631         {  0x0EC5,  0x0EC5} /* Cn */, {  0x0EC7,  0x0EC7} /* Cn */,
 632         {  0x0ECE,  0x0ECF} /* Cn */, {  0x0EDA,  0x0EDB} /* Cn */,
 633         {  0x0EDE,  0x0EFF} /* Cn */, {  0x0F48,  0x0F48} /* Cn */,
 634         {  0x0F6B,  0x0F70} /* Cn */, {  0x0F8C,  0x0F8F} /* Cn */,
 635         {  0x0F98,  0x0F98} /* Cn */, {  0x0FBD,  0x0FBD} /* Cn */,
 636         {  0x0FCD,  0x0FCE} /* Cn */, {  0x0FD2,  0x0FFF} /* Cn */,
 637         {  0x1022,  0x1022} /* Cn */, {  0x1028,  0x1028} /* Cn */,
 638         {  0x102B,  0x102B} /* Cn */, {  0x1033,  0x1035} /* Cn */,
 639         {  0x103A,  0x103F} /* Cn */, {  0x105A,  0x109F} /* Cn */,
 640         {  0x10C6,  0x10CF} /* Cn */, {  0x10FD,  0x10FF} /* Cn */,
 641         {  0x115A,  0x115E} /* Cn */, {  0x11A3,  0x11A7} /* Cn */,
 642         {  0x11FA,  0x11FF} /* Cn */, {  0x1249,  0x1249} /* Cn */,
 643         {  0x124E,  0x124F} /* Cn */, {  0x1257,  0x1257} /* Cn */,
 644         {  0x1259,  0x1259} /* Cn */, {  0x125E,  0x125F} /* Cn */,
 645         {  0x1289,  0x1289} /* Cn */, {  0x128E,  0x128F} /* Cn */,
 646         {  0x12B1,  0x12B1} /* Cn */, {  0x12B6,  0x12B7} /* Cn */,
 647         {  0x12BF,  0x12BF} /* Cn */, {  0x12C1,  0x12C1} /* Cn */,
 648         {  0x12C6,  0x12C7} /* Cn */, {  0x12D7,  0x12D7} /* Cn */,
 649         {  0x1311,  0x1311} /* Cn */, {  0x1316,  0x1317} /* Cn */,
 650         {  0x135B,  0x135E} /* Cn */, {  0x137D,  0x137F} /* Cn */,
 651         {  0x139A,  0x139F} /* Cn */, {  0x13F5,  0x1400} /* Cn */,
 652         {  0x1677,  0x167F} /* Cn */, {  0x169D,  0x169F} /* Cn */,
 653         {  0x16F1,  0x16FF} /* Cn */, {  0x170D,  0x170D} /* Cn */,
 654         {  0x1715,  0x171F} /* Cn */, {  0x1737,  0x173F} /* Cn */,
 655         {  0x1754,  0x175F} /* Cn */, {  0x176D,  0x176D} /* Cn */,
 656         {  0x1771,  0x1771} /* Cn */, {  0x1774,  0x177F} /* Cn */,
 657 #if 0
 658         {  0x17B4,  0x17B5} /* Cf */,
 659 #endif
 660         {  0x17DE,  0x17DF} /* Cn */, {  0x17EA,  0x17EF} /* Cn */,
 661         {  0x17FA,  0x17FF} /* Cn */, {  0x180F,  0x180F} /* Cn */,
 662         {  0x181A,  0x181F} /* Cn */, {  0x1878,  0x187F} /* Cn */,
 663         {  0x18AA,  0x18FF} /* Cn */, {  0x191D,  0x191F} /* Cn */,
 664         {  0x192C,  0x192F} /* Cn */, {  0x193C,  0x193F} /* Cn */,
 665         {  0x1941,  0x1943} /* Cn */, {  0x196E,  0x196F} /* Cn */,
 666         {  0x1975,  0x197F} /* Cn */, {  0x19AA,  0x19AF} /* Cn */,
 667         {  0x19CA,  0x19CF} /* Cn */, {  0x19DA,  0x19DD} /* Cn */,
 668         {  0x1A1C,  0x1A1D} /* Cn */, {  0x1A20,  0x1AFF} /* Cn */,
 669         {  0x1B4C,  0x1B4F} /* Cn */, {  0x1B7D,  0x1CFF} /* Cn */,
 670         {  0x1DCB,  0x1DFD} /* Cn */, {  0x1E9C,  0x1E9F} /* Cn */,
 671         {  0x1EFA,  0x1EFF} /* Cn */, {  0x1F16,  0x1F17} /* Cn */,
 672         {  0x1F1E,  0x1F1F} /* Cn */, {  0x1F46,  0x1F47} /* Cn */,
 673         {  0x1F4E,  0x1F4F} /* Cn */, {  0x1F58,  0x1F58} /* Cn */,
 674         {  0x1F5A,  0x1F5A} /* Cn */, {  0x1F5C,  0x1F5C} /* Cn */,
 675         {  0x1F5E,  0x1F5E} /* Cn */, {  0x1F7E,  0x1F7F} /* Cn */,
 676         {  0x1FB5,  0x1FB5} /* Cn */, {  0x1FC5,  0x1FC5} /* Cn */,
 677         {  0x1FD4,  0x1FD5} /* Cn */, {  0x1FDC,  0x1FDC} /* Cn */,
 678         {  0x1FF0,  0x1FF1} /* Cn */, {  0x1FF5,  0x1FF5} /* Cn */,
 679         {  0x1FFF,  0x1FFF} /* Cn */,
 680         {  0x200B,  0x200F} /* Cf */,
 681         {  0x2028,  0x2028} /* Zl */,
 682         {  0x2029,  0x2029} /* Zp */,
 683         {  0x202A,  0x202E} /* Cf */,
 684         {  0x2060,  0x2063} /* Cf */,
 685         {  0x2064,  0x2069} /* Cn */,
 686         {  0x206A,  0x206F} /* Cf */,
 687         {  0x2072,  0x2073} /* Cn */, {  0x208F,  0x208F} /* Cn */,
 688         {  0x2095,  0x209F} /* Cn */, {  0x20B6,  0x20CF} /* Cn */,
 689         {  0x20F0,  0x20FF} /* Cn */, {  0x214F,  0x2152} /* Cn */,
 690         {  0x2185,  0x218F} /* Cn */, {  0x23E8,  0x23FF} /* Cn */,
 691         {  0x2427,  0x243F} /* Cn */, {  0x244B,  0x245F} /* Cn */,
 692         {  0x269D,  0x269F} /* Cn */, {  0x26B3,  0x2700} /* Cn */,
 693         {  0x2705,  0x2705} /* Cn */, {  0x270A,  0x270B} /* Cn */,
 694         {  0x2728,  0x2728} /* Cn */, {  0x274C,  0x274C} /* Cn */,
 695         {  0x274E,  0x274E} /* Cn */, {  0x2753,  0x2755} /* Cn */,
 696         {  0x2757,  0x2757} /* Cn */, {  0x275F,  0x2760} /* Cn */,
 697         {  0x2795,  0x2797} /* Cn */, {  0x27B0,  0x27B0} /* Cn */,
 698         {  0x27BF,  0x27BF} /* Cn */, {  0x27CB,  0x27CF} /* Cn */,
 699         {  0x27EC,  0x27EF} /* Cn */, {  0x2B1B,  0x2B1F} /* Cn */,
 700         {  0x2B24,  0x2BFF} /* Cn */, {  0x2C2F,  0x2C2F} /* Cn */,
 701         {  0x2C5F,  0x2C5F} /* Cn */, {  0x2C6D,  0x2C73} /* Cn */,
 702         {  0x2C78,  0x2C7F} /* Cn */, {  0x2CEB,  0x2CF8} /* Cn */,
 703         {  0x2D26,  0x2D2F} /* Cn */, {  0x2D66,  0x2D6E} /* Cn */,
 704         {  0x2D70,  0x2D7F} /* Cn */, {  0x2D97,  0x2D9F} /* Cn */,
 705         {  0x2DA7,  0x2DA7} /* Cn */, {  0x2DAF,  0x2DAF} /* Cn */,
 706         {  0x2DB7,  0x2DB7} /* Cn */, {  0x2DBF,  0x2DBF} /* Cn */,
 707         {  0x2DC7,  0x2DC7} /* Cn */, {  0x2DCF,  0x2DCF} /* Cn */,
 708         {  0x2DD7,  0x2DD7} /* Cn */, {  0x2DDF,  0x2DFF} /* Cn */,
 709         {  0x2E18,  0x2E1B} /* Cn */, {  0x2E1E,  0x2E7F} /* Cn */,
 710         {  0x2E9A,  0x2E9A} /* Cn */, {  0x2EF4,  0x2EFF} /* Cn */,
 711         {  0x2FD6,  0x2FEF} /* Cn */, {  0x2FFC,  0x2FFF} /* Cn */,
 712         {  0x3040,  0x3040} /* Cn */, {  0x3097,  0x3098} /* Cn */,
 713         {  0x3100,  0x3104} /* Cn */, {  0x312D,  0x3130} /* Cn */,
 714         {  0x318F,  0x318F} /* Cn */, {  0x31B8,  0x31BF} /* Cn */,
 715         {  0x31D0,  0x31EF} /* Cn */, {  0x321F,  0x321F} /* Cn */,
 716         {  0x3244,  0x324F} /* Cn */, {  0x32FF,  0x32FF} /* Cn */,
 717         {  0x4DB6,  0x4DBF} /* Cn */, {  0x9FBC,  0x9FFF} /* Cn */,
 718         {  0xA48D,  0xA48F} /* Cn */, {  0xA4C7,  0xA6FF} /* Cn */,
 719         {  0xA71B,  0xA71F} /* Cn */, {  0xA722,  0xA7FF} /* Cn */,
 720         {  0xA82C,  0xA83F} /* Cn */, {  0xA878,  0xABFF} /* Cn */,
 721         {  0xD7A4,  0xD7FF} /* Cn */,
 722         {  0xD800,  0xDFFF} /* Cs */,
 723         {  0xE000,  0xF8FF} /* Co */,
 724         {  0xFA2E,  0xFA2F} /* Cn */, {  0xFA6B,  0xFA6F} /* Cn */,
 725         {  0xFADA,  0xFAFF} /* Cn */, {  0xFB07,  0xFB12} /* Cn */,
 726         {  0xFB18,  0xFB1C} /* Cn */, {  0xFB37,  0xFB37} /* Cn */,
 727         {  0xFB3D,  0xFB3D} /* Cn */, {  0xFB3F,  0xFB3F} /* Cn */,
 728         {  0xFB42,  0xFB42} /* Cn */, {  0xFB45,  0xFB45} /* Cn */,
 729         {  0xFBB2,  0xFBD2} /* Cn */, {  0xFD40,  0xFD4F} /* Cn */,
 730         {  0xFD90,  0xFD91} /* Cn */, {  0xFDC8,  0xFDEF} /* Cn */,
 731         {  0xFDFE,  0xFDFF} /* Cn */, {  0xFE1A,  0xFE1F} /* Cn */,
 732         {  0xFE24,  0xFE2F} /* Cn */, {  0xFE53,  0xFE53} /* Cn */,
 733         {  0xFE67,  0xFE67} /* Cn */, {  0xFE6C,  0xFE6F} /* Cn */,
 734         {  0xFE75,  0xFE75} /* Cn */, {  0xFEFD,  0xFEFE} /* Cn */,
 735         {  0xFEFF,  0xFEFF} /* Cf */,
 736         {  0xFF00,  0xFF00} /* Cn */, {  0xFFBF,  0xFFC1} /* Cn */,
 737         {  0xFFC8,  0xFFC9} /* Cn */, {  0xFFD0,  0xFFD1} /* Cn */,
 738         {  0xFFD8,  0xFFD9} /* Cn */, {  0xFFDD,  0xFFDF} /* Cn */,
 739         {  0xFFE7,  0xFFE7} /* Cn */, {  0xFFEF,  0xFFF8} /* Cn */,
 740         {  0xFFF9,  0xFFFB} /* Cf */,
 741         {  0xFFFE,  0xFFFF} /* Cn */, { 0x1000C, 0x1000C} /* Cn */,
 742         { 0x10027, 0x10027} /* Cn */, { 0x1003B, 0x1003B} /* Cn */,
 743         { 0x1003E, 0x1003E} /* Cn */, { 0x1004E, 0x1004F} /* Cn */,
 744         { 0x1005E, 0x1007F} /* Cn */, { 0x100FB, 0x100FF} /* Cn */,
 745         { 0x10103, 0x10106} /* Cn */, { 0x10134, 0x10136} /* Cn */,
 746         { 0x1018B, 0x102FF} /* Cn */, { 0x1031F, 0x1031F} /* Cn */,
 747         { 0x10324, 0x1032F} /* Cn */, { 0x1034B, 0x1037F} /* Cn */,
 748         { 0x1039E, 0x1039E} /* Cn */, { 0x103C4, 0x103C7} /* Cn */,
 749         { 0x103D6, 0x103FF} /* Cn */,
 750         { 0x1049E, 0x1049F} /* Cn */, { 0x104AA, 0x107FF} /* Cn */,
 751         { 0x10806, 0x10807} /* Cn */, { 0x10809, 0x10809} /* Cn */,
 752         { 0x10836, 0x10836} /* Cn */, { 0x10839, 0x1083B} /* Cn */,
 753         { 0x1083D, 0x1083E} /* Cn */, { 0x10840, 0x108FF} /* Cn */,
 754         { 0x1091A, 0x1091E} /* Cn */, { 0x10920, 0x109FF} /* Cn */,
 755         { 0x10A04, 0x10A04} /* Cn */, { 0x10A07, 0x10A0B} /* Cn */,
 756         { 0x10A14, 0x10A14} /* Cn */, { 0x10A18, 0x10A18} /* Cn */,
 757         { 0x10A34, 0x10A37} /* Cn */, { 0x10A3B, 0x10A3E} /* Cn */,
 758         { 0x10A48, 0x10A4F} /* Cn */, { 0x10A59, 0x11FFF} /* Cn */,
 759         { 0x1236F, 0x123FF} /* Cn */, { 0x12463, 0x1246F} /* Cn */,
 760         { 0x12474, 0x1CFFF} /* Cn */, { 0x1D0F6, 0x1D0FF} /* Cn */,
 761         { 0x1D127, 0x1D129} /* Cn */,
 762         { 0x1D173, 0x1D17A} /* Cf */,
 763         { 0x1D1DE, 0x1D1FF} /* Cn */, { 0x1D246, 0x1D2FF} /* Cn */,
 764         { 0x1D357, 0x1D35F} /* Cn */, { 0x1D372, 0x1D3FF} /* Cn */,
 765         { 0x1D455, 0x1D455} /* Cn */, { 0x1D49D, 0x1D49D} /* Cn */,
 766         { 0x1D4A0, 0x1D4A1} /* Cn */, { 0x1D4A3, 0x1D4A4} /* Cn */,
 767         { 0x1D4A7, 0x1D4A8} /* Cn */, { 0x1D4AD, 0x1D4AD} /* Cn */,
 768         { 0x1D4BA, 0x1D4BA} /* Cn */, { 0x1D4BC, 0x1D4BC} /* Cn */,
 769         { 0x1D4C4, 0x1D4C4} /* Cn */, { 0x1D506, 0x1D506} /* Cn */,
 770         { 0x1D50B, 0x1D50C} /* Cn */, { 0x1D515, 0x1D515} /* Cn */,
 771         { 0x1D51D, 0x1D51D} /* Cn */, { 0x1D53A, 0x1D53A} /* Cn */,
 772         { 0x1D53F, 0x1D53F} /* Cn */, { 0x1D545, 0x1D545} /* Cn */,
 773         { 0x1D547, 0x1D549} /* Cn */, { 0x1D551, 0x1D551} /* Cn */,
 774         { 0x1D6A6, 0x1D6A7} /* Cn */, { 0x1D7CC, 0x1D7CD} /* Cn */,
 775         { 0x1D800, 0x1FFFF} /* Cn */, { 0x2A6D7, 0x2F7FF} /* Cn */,
 776         { 0x2FA1E, 0xE0000} /* Cn */,
 777         { 0xE0001, 0xE0001} /* Cf */,
 778         { 0xE0002, 0xE001F} /* Cn */,
 779         { 0xE0020, 0xE007F} /* Cf */,
 780         { 0xE0080, 0xE00FF} /* Cn */, { 0xE01F0, 0xEFFFF} /* Cn */,
 781         { 0xF0000, 0xFFFFD} /* Co */,
 782         { 0xFFFFE, 0xFFFFF} /* Cn */,
 783         { 0x100000, 0x10FFFD} /* Co */,
 784         { 0x10FFFE, 0x10FFFF} /* Cn */,
 785         { 0x110000, 0x7FFFFFFF} /* ISO 10646?? */
 786 };
 787
 788 /*
 789  * Double width characters
 790  *      W: East Asian Wide
 791  *      F: East Asian Full-width
 792  * Unassigned code points may be included when they allow ranges to be merged.
 793  * Last synched with
 794  *      <http://www.unicode.org/Public/5.0.0/ucd/EastAsianWidth-5.0.0d2.txt>
 795  *      dated 2005-11-08T01:32:56Z
 796  */
 797 static struct wchar_range wide_table[] = {
 798         {  0x1100,  0x115F} /* W */, {  0x2329,  0x232A} /* W */,
 799         {  0x2E80,  0x2FFB} /* W */,
 800         {  0x3000,  0x3000} /* F */,
 801         {  0x3001,  0x303E} /* W */, {  0x3041,  0x4DB5} /* W */,
 802         {  0x4E00,  0x9FBB} /* W */, {  0xA000,  0xA4C6} /* W */,
 803         {  0xAC00,  0xD7A3} /* W */, {  0xF900,  0xFAD9} /* W */,
 804         {  0xFE10,  0xFE19} /* W */, {  0xFE30,  0xFE6B} /* W */,
 805         {  0xFF01,  0xFF60} /* F */, {  0xFFE0,  0xFFE6} /* F */,
 806         { 0x20000, 0x2FFFD} /* W */, { 0x30000, 0x3FFFD} /* W */,
 807 };
 808
 809 static int
 810 is_in_table(LWCHAR ch, struct wchar_range table[], int tsize)
 811 {
 812         int hi;
 813         int lo;
 814
 815         /* Binary search in the table. */
 816         if (ch < table[0].first)
 817                 return (0);
 818         lo = 0;
 819         hi = tsize - 1;
 820         while (lo <= hi) {
 821                 int mid = (lo + hi) / 2;
 822                 if (ch > table[mid].last)
 823                         lo = mid + 1;
 824                 else if (ch < table[mid].first)
 825                         hi = mid - 1;
 826                 else
 827                         return (1);
 828         }
 829         return (0);
 830 }
 831
 832 /*
 833  * Is a character a UTF-8 composing character?
 834  * If a composing character follows any char, the two combine into one glyph.
 835  */
 836 int
 837 is_composing_char(LWCHAR ch)
 838 {
 839         return (is_in_table(ch, comp_table,
 840             (sizeof (comp_table) / sizeof (*comp_table))));
 841 }
 842
 843 /*
 844  * Should this UTF-8 character be treated as binary?
 845  */
 846 int
 847 is_ubin_char(LWCHAR ch)
 848 {
 849         return (is_in_table(ch, ubin_table,
 850             (sizeof (ubin_table) / sizeof (*ubin_table))));
 851 }
 852
 853 /*
 854  * Is this a double width UTF-8 character?
 855  */
 856 int
 857 is_wide_char(LWCHAR ch)
 858 {
 859         return (is_in_table(ch, wide_table,
 860             (sizeof (wide_table) / sizeof (*wide_table))));
 861 }
 862
 863 /*
 864  * Is a character a UTF-8 combining character?
 865  * A combining char acts like an ordinary char, but if it follows
 866  * a specific char (not any char), the two combine into one glyph.
 867  */
 868 int
 869 is_combining_char(LWCHAR ch1, LWCHAR ch2)
 870 {
 871         /* The table is small; use linear search. */
 872         int i;
 873         for (i = 0; i < sizeof (comb_table) / sizeof (*comb_table); i++) {
 874                 if (ch1 == comb_table[i].first &&
 875                     ch2 == comb_table[i].last)
 876                         return (1);
 877         }
 878         return (0);
 879 }