xref: /freebsd/sys/teken/teken_wcwidth.h (revision 9e589b09)
19b934d09SEd Schouten /*
29b934d09SEd Schouten  * Markus Kuhn -- 2007-05-26 (Unicode 5.0)
39b934d09SEd Schouten  *
49b934d09SEd Schouten  * Permission to use, copy, modify, and distribute this software
59b934d09SEd Schouten  * for any purpose and without fee is hereby granted. The author
69b934d09SEd Schouten  * disclaims all warranties with regard to this software.
79b934d09SEd Schouten  *
89b934d09SEd Schouten  * Latest version: http://www.cl.cam.ac.uk/~mgk25/ucs/wcwidth.c
99b934d09SEd Schouten  */
109b934d09SEd Schouten 
119e589b09SBojan Novković #define TEKEN_UTF8_INVALID_CODEPOINT -1
129e589b09SBojan Novković 
139b934d09SEd Schouten struct interval {
149b934d09SEd Schouten   teken_char_t first;
159b934d09SEd Schouten   teken_char_t last;
169b934d09SEd Schouten };
179b934d09SEd Schouten 
189b934d09SEd Schouten /* auxiliary function for binary search in interval table */
199b934d09SEd Schouten static int bisearch(teken_char_t ucs, const struct interval *table, int max) {
209b934d09SEd Schouten   int min = 0;
219b934d09SEd Schouten   int mid;
229b934d09SEd Schouten 
239b934d09SEd Schouten   if (ucs < table[0].first || ucs > table[max].last)
249b934d09SEd Schouten     return 0;
259b934d09SEd Schouten   while (max >= min) {
269b934d09SEd Schouten     mid = (min + max) / 2;
279b934d09SEd Schouten     if (ucs > table[mid].last)
289b934d09SEd Schouten       min = mid + 1;
299b934d09SEd Schouten     else if (ucs < table[mid].first)
309b934d09SEd Schouten       max = mid - 1;
319b934d09SEd Schouten     else
329b934d09SEd Schouten       return 1;
339b934d09SEd Schouten   }
349b934d09SEd Schouten 
359b934d09SEd Schouten   return 0;
369b934d09SEd Schouten }
379b934d09SEd Schouten 
389b934d09SEd Schouten static int teken_wcwidth(teken_char_t ucs)
399b934d09SEd Schouten {
409b934d09SEd Schouten   /* sorted list of non-overlapping intervals of non-spacing characters */
419b934d09SEd Schouten   /* generated by "uniset +cat=Me +cat=Mn +cat=Cf -00AD +1160-11FF +200B c" */
429b934d09SEd Schouten   static const struct interval combining[] = {
439b934d09SEd Schouten     { 0x0300, 0x036F }, { 0x0483, 0x0486 }, { 0x0488, 0x0489 },
449b934d09SEd Schouten     { 0x0591, 0x05BD }, { 0x05BF, 0x05BF }, { 0x05C1, 0x05C2 },
459b934d09SEd Schouten     { 0x05C4, 0x05C5 }, { 0x05C7, 0x05C7 }, { 0x0600, 0x0603 },
469b934d09SEd Schouten     { 0x0610, 0x0615 }, { 0x064B, 0x065E }, { 0x0670, 0x0670 },
479b934d09SEd Schouten     { 0x06D6, 0x06E4 }, { 0x06E7, 0x06E8 }, { 0x06EA, 0x06ED },
489b934d09SEd Schouten     { 0x070F, 0x070F }, { 0x0711, 0x0711 }, { 0x0730, 0x074A },
499b934d09SEd Schouten     { 0x07A6, 0x07B0 }, { 0x07EB, 0x07F3 }, { 0x0901, 0x0902 },
509b934d09SEd Schouten     { 0x093C, 0x093C }, { 0x0941, 0x0948 }, { 0x094D, 0x094D },
519b934d09SEd Schouten     { 0x0951, 0x0954 }, { 0x0962, 0x0963 }, { 0x0981, 0x0981 },
529b934d09SEd Schouten     { 0x09BC, 0x09BC }, { 0x09C1, 0x09C4 }, { 0x09CD, 0x09CD },
539b934d09SEd Schouten     { 0x09E2, 0x09E3 }, { 0x0A01, 0x0A02 }, { 0x0A3C, 0x0A3C },
549b934d09SEd Schouten     { 0x0A41, 0x0A42 }, { 0x0A47, 0x0A48 }, { 0x0A4B, 0x0A4D },
559b934d09SEd Schouten     { 0x0A70, 0x0A71 }, { 0x0A81, 0x0A82 }, { 0x0ABC, 0x0ABC },
569b934d09SEd Schouten     { 0x0AC1, 0x0AC5 }, { 0x0AC7, 0x0AC8 }, { 0x0ACD, 0x0ACD },
579b934d09SEd Schouten     { 0x0AE2, 0x0AE3 }, { 0x0B01, 0x0B01 }, { 0x0B3C, 0x0B3C },
589b934d09SEd Schouten     { 0x0B3F, 0x0B3F }, { 0x0B41, 0x0B43 }, { 0x0B4D, 0x0B4D },
599b934d09SEd Schouten     { 0x0B56, 0x0B56 }, { 0x0B82, 0x0B82 }, { 0x0BC0, 0x0BC0 },
609b934d09SEd Schouten     { 0x0BCD, 0x0BCD }, { 0x0C3E, 0x0C40 }, { 0x0C46, 0x0C48 },
619b934d09SEd Schouten     { 0x0C4A, 0x0C4D }, { 0x0C55, 0x0C56 }, { 0x0CBC, 0x0CBC },
629b934d09SEd Schouten     { 0x0CBF, 0x0CBF }, { 0x0CC6, 0x0CC6 }, { 0x0CCC, 0x0CCD },
639b934d09SEd Schouten     { 0x0CE2, 0x0CE3 }, { 0x0D41, 0x0D43 }, { 0x0D4D, 0x0D4D },
649b934d09SEd Schouten     { 0x0DCA, 0x0DCA }, { 0x0DD2, 0x0DD4 }, { 0x0DD6, 0x0DD6 },
659b934d09SEd Schouten     { 0x0E31, 0x0E31 }, { 0x0E34, 0x0E3A }, { 0x0E47, 0x0E4E },
669b934d09SEd Schouten     { 0x0EB1, 0x0EB1 }, { 0x0EB4, 0x0EB9 }, { 0x0EBB, 0x0EBC },
679b934d09SEd Schouten     { 0x0EC8, 0x0ECD }, { 0x0F18, 0x0F19 }, { 0x0F35, 0x0F35 },
689b934d09SEd Schouten     { 0x0F37, 0x0F37 }, { 0x0F39, 0x0F39 }, { 0x0F71, 0x0F7E },
699b934d09SEd Schouten     { 0x0F80, 0x0F84 }, { 0x0F86, 0x0F87 }, { 0x0F90, 0x0F97 },
709b934d09SEd Schouten     { 0x0F99, 0x0FBC }, { 0x0FC6, 0x0FC6 }, { 0x102D, 0x1030 },
719b934d09SEd Schouten     { 0x1032, 0x1032 }, { 0x1036, 0x1037 }, { 0x1039, 0x1039 },
729b934d09SEd Schouten     { 0x1058, 0x1059 }, { 0x1160, 0x11FF }, { 0x135F, 0x135F },
739b934d09SEd Schouten     { 0x1712, 0x1714 }, { 0x1732, 0x1734 }, { 0x1752, 0x1753 },
749b934d09SEd Schouten     { 0x1772, 0x1773 }, { 0x17B4, 0x17B5 }, { 0x17B7, 0x17BD },
759b934d09SEd Schouten     { 0x17C6, 0x17C6 }, { 0x17C9, 0x17D3 }, { 0x17DD, 0x17DD },
769b934d09SEd Schouten     { 0x180B, 0x180D }, { 0x18A9, 0x18A9 }, { 0x1920, 0x1922 },
779b934d09SEd Schouten     { 0x1927, 0x1928 }, { 0x1932, 0x1932 }, { 0x1939, 0x193B },
789b934d09SEd Schouten     { 0x1A17, 0x1A18 }, { 0x1B00, 0x1B03 }, { 0x1B34, 0x1B34 },
799b934d09SEd Schouten     { 0x1B36, 0x1B3A }, { 0x1B3C, 0x1B3C }, { 0x1B42, 0x1B42 },
809b934d09SEd Schouten     { 0x1B6B, 0x1B73 }, { 0x1DC0, 0x1DCA }, { 0x1DFE, 0x1DFF },
819b934d09SEd Schouten     { 0x200B, 0x200F }, { 0x202A, 0x202E }, { 0x2060, 0x2063 },
829b934d09SEd Schouten     { 0x206A, 0x206F }, { 0x20D0, 0x20EF }, { 0x302A, 0x302F },
839b934d09SEd Schouten     { 0x3099, 0x309A }, { 0xA806, 0xA806 }, { 0xA80B, 0xA80B },
849b934d09SEd Schouten     { 0xA825, 0xA826 }, { 0xFB1E, 0xFB1E }, { 0xFE00, 0xFE0F },
859b934d09SEd Schouten     { 0xFE20, 0xFE23 }, { 0xFEFF, 0xFEFF }, { 0xFFF9, 0xFFFB },
869b934d09SEd Schouten     { 0x10A01, 0x10A03 }, { 0x10A05, 0x10A06 }, { 0x10A0C, 0x10A0F },
879b934d09SEd Schouten     { 0x10A38, 0x10A3A }, { 0x10A3F, 0x10A3F }, { 0x1D167, 0x1D169 },
889b934d09SEd Schouten     { 0x1D173, 0x1D182 }, { 0x1D185, 0x1D18B }, { 0x1D1AA, 0x1D1AD },
899b934d09SEd Schouten     { 0x1D242, 0x1D244 }, { 0xE0001, 0xE0001 }, { 0xE0020, 0xE007F },
909b934d09SEd Schouten     { 0xE0100, 0xE01EF }
919b934d09SEd Schouten   };
929b934d09SEd Schouten 
939b934d09SEd Schouten   /* test for 8-bit control characters */
949b934d09SEd Schouten   if (ucs == 0)
959b934d09SEd Schouten     return 0;
969b934d09SEd Schouten   if (ucs < 32 || (ucs >= 0x7f && ucs < 0xa0))
979b934d09SEd Schouten     return -1;
989b934d09SEd Schouten 
999b934d09SEd Schouten   /* binary search in table of non-spacing characters */
1009b934d09SEd Schouten   if (bisearch(ucs, combining,
1019b934d09SEd Schouten 	       sizeof(combining) / sizeof(struct interval) - 1))
1029b934d09SEd Schouten     return 0;
1039b934d09SEd Schouten 
1049b934d09SEd Schouten   /* if we arrive here, ucs is not a combining or C0/C1 control character */
1059b934d09SEd Schouten 
1069b934d09SEd Schouten   return 1 +
1079b934d09SEd Schouten     (ucs >= 0x1100 &&
1089b934d09SEd Schouten      (ucs <= 0x115f ||                    /* Hangul Jamo init. consonants */
1099b934d09SEd Schouten       ucs == 0x2329 || ucs == 0x232a ||
1109b934d09SEd Schouten       (ucs >= 0x2e80 && ucs <= 0xa4cf &&
1119b934d09SEd Schouten        ucs != 0x303f) ||                  /* CJK ... Yi */
1129b934d09SEd Schouten       (ucs >= 0xac00 && ucs <= 0xd7a3) || /* Hangul Syllables */
1139b934d09SEd Schouten       (ucs >= 0xf900 && ucs <= 0xfaff) || /* CJK Compatibility Ideographs */
1149b934d09SEd Schouten       (ucs >= 0xfe10 && ucs <= 0xfe19) || /* Vertical forms */
1159b934d09SEd Schouten       (ucs >= 0xfe30 && ucs <= 0xfe6f) || /* CJK Compatibility Forms */
1169b934d09SEd Schouten       (ucs >= 0xff00 && ucs <= 0xff60) || /* Fullwidth Forms */
1179b934d09SEd Schouten       (ucs >= 0xffe0 && ucs <= 0xffe6) ||
1189b934d09SEd Schouten       (ucs >= 0x20000 && ucs <= 0x2fffd) ||
1199b934d09SEd Schouten       (ucs >= 0x30000 && ucs <= 0x3fffd)));
1209b934d09SEd Schouten }
1219e589b09SBojan Novković 
1229e589b09SBojan Novković /*
1239e589b09SBojan Novković  * Converts an UTF-8 byte sequence to a codepoint as specified in
1249e589b09SBojan Novković  * https://datatracker.ietf.org/doc/html/rfc3629#section-3 . The function
1259e589b09SBojan Novković  * expects the 'bytes' array to start with the leading character.
1269e589b09SBojan Novković  */
1279e589b09SBojan Novković static teken_char_t
1289e589b09SBojan Novković teken_utf8_bytes_to_codepoint(uint8_t bytes[4], int nbytes)
1299e589b09SBojan Novković {
1309e589b09SBojan Novković 
1319e589b09SBojan Novković   /* Check for malformed characters. */
1329e589b09SBojan Novković   if (bitcount(bytes[0] & 0xf0) != nbytes)
1339e589b09SBojan Novković     return (TEKEN_UTF8_INVALID_CODEPOINT);
1349e589b09SBojan Novković 
1359e589b09SBojan Novković   switch (nbytes) {
1369e589b09SBojan Novković   case 1:
1379e589b09SBojan Novković     return (bytes[0] & 0x7f);
1389e589b09SBojan Novković   case 2:
1399e589b09SBojan Novković     return (bytes[0] & 0xf) << 6 | (bytes[1] & 0x3f);
1409e589b09SBojan Novković   case 3:
1419e589b09SBojan Novković     return (bytes[0] & 0xf) << 12 | (bytes[1] & 0x3f) << 6 | (bytes[2] & 0x3f);
1429e589b09SBojan Novković   case 4:
1439e589b09SBojan Novković     return (bytes[0] & 0x7) << 18 | (bytes[1] & 0x3f) << 12 |
1449e589b09SBojan Novković 	(bytes[2] & 0x3f) << 6 | (bytes[3] & 0x3f);
1459e589b09SBojan Novković   default:
1469e589b09SBojan Novković     return (TEKEN_UTF8_INVALID_CODEPOINT);
1479e589b09SBojan Novković   }
1489e589b09SBojan Novković }
149