1/*
2 * This is an implementation of wcwidth() and wcswidth() (defined in
3 * IEEE Std 1002.1-2001) for Unicode.
4 *
5 * http://www.opengroup.org/onlinepubs/007904975/functions/wcwidth.html
6 * http://www.opengroup.org/onlinepubs/007904975/functions/wcswidth.html
7 *
8 * In fixed-width output devices, Latin characters all occupy a single
9 * "cell" position of equal width, whereas ideographic CJK characters
10 * occupy two such cells. Interoperability between terminal-line
11 * applications and (teletype-style) character terminals using the
12 * UTF-8 encoding requires agreement on which character should advance
13 * the cursor by how many cell positions. No established formal
14 * standards exist at present on which Unicode character shall occupy
15 * how many cell positions on character terminals. These routines are
16 * a first attempt of defining such behavior based on simple rules
17 * applied to data provided by the Unicode Consortium.
18 *
19 * For some graphical characters, the Unicode standard explicitly
20 * defines a character-cell width via the definition of the East Asian
21 * FullWidth (F), Wide (W), Half-width (H), and Narrow (Na) classes.
22 * In all these cases, there is no ambiguity about which width a
23 * terminal shall use. For characters in the East Asian Ambiguous (A)
24 * class, the width choice depends purely on a preference of backward
25 * compatibility with either historic CJK or Western practice.
26 * Choosing single-width for these characters is easy to justify as
27 * the appropriate long-term solution, as the CJK practice of
28 * displaying these characters as double-width comes from historic
29 * implementation simplicity (8-bit encoded characters were displayed
30 * single-width and 16-bit ones double-width, even for Greek,
31 * Cyrillic, etc.) and not any typographic considerations.
32 *
33 * Much less clear is the choice of width for the Not East Asian
34 * (Neutral) class. Existing practice does not dictate a width for any
35 * of these characters. It would nevertheless make sense
36 * typographically to allocate two character cells to characters such
37 * as for instance EM SPACE or VOLUME INTEGRAL, which cannot be
38 * represented adequately with a single-width glyph. The following
39 * routines at present merely assign a single-cell width to all
40 * neutral characters, in the interest of simplicity. This is not
41 * entirely satisfactory and should be reconsidered before
42 * establishing a formal standard in this area. At the moment, the
43 * decision which Not East Asian (Neutral) characters should be
44 * represented by double-width glyphs cannot yet be answered by
45 * applying a simple rule from the Unicode database content. Setting
46 * up a proper standard for the behavior of UTF-8 character terminals
47 * will require a careful analysis not only of each Unicode character,
48 * but also of each presentation form, something the author of these
49 * routines has avoided to do so far.
50 *
51 * http://www.unicode.org/unicode/reports/tr11/
52 *
53 * Markus Kuhn -- 2007-05-26 (Unicode 5.0)
54 *
55 * Permission to use, copy, modify, and distribute this software
56 * for any purpose and without fee is hereby granted. The author
57 * disclaims all warranties with regard to this software.
58 *
59 * Latest version: http://www.cl.cam.ac.uk/~mgk25/ucs/wcwidth.c
60 */
61
62#include <apr_lib.h>
63
64#include "svn_utf.h"
65#include "private/svn_utf_private.h"
66
67#include "svn_private_config.h"
68
69struct interval {
70  apr_uint32_t first;
71  apr_uint32_t last;
72};
73
74/* auxiliary function for binary search in interval table */
75static int
76bisearch(apr_uint32_t ucs, const struct interval *table, apr_uint32_t max)
77{
78  apr_uint32_t min = 0;
79  apr_uint32_t mid;
80
81  if (ucs < table[0].first || ucs > table[max].last)
82    return 0;
83  while (max >= min) {
84    mid = (min + max) / 2;
85    if (ucs > table[mid].last)
86      min = mid + 1;
87    else if (ucs < table[mid].first)
88      max = mid - 1; /* this is safe because ucs >= table[0].first */
89    else
90      return 1;
91  }
92
93  return 0;
94}
95
96
97/* The following two functions define the column width of an ISO 10646
98 * character as follows:
99 *
100 *    - The null character (U+0000) has a column width of 0.
101 *
102 *    - Other C0/C1 control characters and DEL will lead to a return
103 *      value of -1.
104 *
105 *    - Non-spacing and enclosing combining characters (general
106 *      category code Mn or Me in the Unicode database) have a
107 *      column width of 0.
108 *
109 *    - SOFT HYPHEN (U+00AD) has a column width of 1.
110 *
111 *    - Other format characters (general category code Cf in the Unicode
112 *      database) and ZERO WIDTH SPACE (U+200B) have a column width of 0.
113 *
114 *    - Hangul Jamo medial vowels and final consonants (U+1160-U+11FF)
115 *      have a column width of 0.
116 *
117 *    - Spacing characters in the East Asian Wide (W) or East Asian
118 *      Full-width (F) category as defined in Unicode Technical
119 *      Report #11 have a column width of 2.
120 *
121 *    - All remaining characters (including all printable
122 *      ISO 8859-1 and WGL4 characters, Unicode control characters,
123 *      etc.) have a column width of 1.
124 *
125 * This implementation assumes that wchar_t characters are encoded
126 * in ISO 10646.
127 */
128
129static int
130mk_wcwidth(apr_uint32_t ucs)
131{
132  /* sorted list of non-overlapping intervals of non-spacing characters */
133  /* generated by "uniset +cat=Me +cat=Mn +cat=Cf -00AD +1160-11FF +200B c" */
134  static const struct interval combining[] = {
135    { 0x0300, 0x036F }, { 0x0483, 0x0486 }, { 0x0488, 0x0489 },
136    { 0x0591, 0x05BD }, { 0x05BF, 0x05BF }, { 0x05C1, 0x05C2 },
137    { 0x05C4, 0x05C5 }, { 0x05C7, 0x05C7 }, { 0x0600, 0x0603 },
138    { 0x0610, 0x0615 }, { 0x064B, 0x065E }, { 0x0670, 0x0670 },
139    { 0x06D6, 0x06E4 }, { 0x06E7, 0x06E8 }, { 0x06EA, 0x06ED },
140    { 0x070F, 0x070F }, { 0x0711, 0x0711 }, { 0x0730, 0x074A },
141    { 0x07A6, 0x07B0 }, { 0x07EB, 0x07F3 }, { 0x0901, 0x0902 },
142    { 0x093C, 0x093C }, { 0x0941, 0x0948 }, { 0x094D, 0x094D },
143    { 0x0951, 0x0954 }, { 0x0962, 0x0963 }, { 0x0981, 0x0981 },
144    { 0x09BC, 0x09BC }, { 0x09C1, 0x09C4 }, { 0x09CD, 0x09CD },
145    { 0x09E2, 0x09E3 }, { 0x0A01, 0x0A02 }, { 0x0A3C, 0x0A3C },
146    { 0x0A41, 0x0A42 }, { 0x0A47, 0x0A48 }, { 0x0A4B, 0x0A4D },
147    { 0x0A70, 0x0A71 }, { 0x0A81, 0x0A82 }, { 0x0ABC, 0x0ABC },
148    { 0x0AC1, 0x0AC5 }, { 0x0AC7, 0x0AC8 }, { 0x0ACD, 0x0ACD },
149    { 0x0AE2, 0x0AE3 }, { 0x0B01, 0x0B01 }, { 0x0B3C, 0x0B3C },
150    { 0x0B3F, 0x0B3F }, { 0x0B41, 0x0B43 }, { 0x0B4D, 0x0B4D },
151    { 0x0B56, 0x0B56 }, { 0x0B82, 0x0B82 }, { 0x0BC0, 0x0BC0 },
152    { 0x0BCD, 0x0BCD }, { 0x0C3E, 0x0C40 }, { 0x0C46, 0x0C48 },
153    { 0x0C4A, 0x0C4D }, { 0x0C55, 0x0C56 }, { 0x0CBC, 0x0CBC },
154    { 0x0CBF, 0x0CBF }, { 0x0CC6, 0x0CC6 }, { 0x0CCC, 0x0CCD },
155    { 0x0CE2, 0x0CE3 }, { 0x0D41, 0x0D43 }, { 0x0D4D, 0x0D4D },
156    { 0x0DCA, 0x0DCA }, { 0x0DD2, 0x0DD4 }, { 0x0DD6, 0x0DD6 },
157    { 0x0E31, 0x0E31 }, { 0x0E34, 0x0E3A }, { 0x0E47, 0x0E4E },
158    { 0x0EB1, 0x0EB1 }, { 0x0EB4, 0x0EB9 }, { 0x0EBB, 0x0EBC },
159    { 0x0EC8, 0x0ECD }, { 0x0F18, 0x0F19 }, { 0x0F35, 0x0F35 },
160    { 0x0F37, 0x0F37 }, { 0x0F39, 0x0F39 }, { 0x0F71, 0x0F7E },
161    { 0x0F80, 0x0F84 }, { 0x0F86, 0x0F87 }, { 0x0F90, 0x0F97 },
162    { 0x0F99, 0x0FBC }, { 0x0FC6, 0x0FC6 }, { 0x102D, 0x1030 },
163    { 0x1032, 0x1032 }, { 0x1036, 0x1037 }, { 0x1039, 0x1039 },
164    { 0x1058, 0x1059 }, { 0x1160, 0x11FF }, { 0x135F, 0x135F },
165    { 0x1712, 0x1714 }, { 0x1732, 0x1734 }, { 0x1752, 0x1753 },
166    { 0x1772, 0x1773 }, { 0x17B4, 0x17B5 }, { 0x17B7, 0x17BD },
167    { 0x17C6, 0x17C6 }, { 0x17C9, 0x17D3 }, { 0x17DD, 0x17DD },
168    { 0x180B, 0x180D }, { 0x18A9, 0x18A9 }, { 0x1920, 0x1922 },
169    { 0x1927, 0x1928 }, { 0x1932, 0x1932 }, { 0x1939, 0x193B },
170    { 0x1A17, 0x1A18 }, { 0x1B00, 0x1B03 }, { 0x1B34, 0x1B34 },
171    { 0x1B36, 0x1B3A }, { 0x1B3C, 0x1B3C }, { 0x1B42, 0x1B42 },
172    { 0x1B6B, 0x1B73 }, { 0x1DC0, 0x1DCA }, { 0x1DFE, 0x1DFF },
173    { 0x200B, 0x200F }, { 0x202A, 0x202E }, { 0x2060, 0x2063 },
174    { 0x206A, 0x206F }, { 0x20D0, 0x20EF }, { 0x302A, 0x302F },
175    { 0x3099, 0x309A }, { 0xA806, 0xA806 }, { 0xA80B, 0xA80B },
176    { 0xA825, 0xA826 }, { 0xFB1E, 0xFB1E }, { 0xFE00, 0xFE0F },
177    { 0xFE20, 0xFE23 }, { 0xFEFF, 0xFEFF }, { 0xFFF9, 0xFFFB },
178    { 0x10A01, 0x10A03 }, { 0x10A05, 0x10A06 }, { 0x10A0C, 0x10A0F },
179    { 0x10A38, 0x10A3A }, { 0x10A3F, 0x10A3F }, { 0x1D167, 0x1D169 },
180    { 0x1D173, 0x1D182 }, { 0x1D185, 0x1D18B }, { 0x1D1AA, 0x1D1AD },
181    { 0x1D242, 0x1D244 }, { 0xE0001, 0xE0001 }, { 0xE0020, 0xE007F },
182    { 0xE0100, 0xE01EF }
183  };
184
185  /* test for 8-bit control characters */
186  if (ucs == 0)
187    return 0;
188  if (ucs < 32 || (ucs >= 0x7f && ucs < 0xa0))
189    return -1;
190
191  /* binary search in table of non-spacing characters */
192  if (bisearch(ucs, combining,
193               sizeof(combining) / sizeof(struct interval) - 1))
194    return 0;
195
196  /* if we arrive here, ucs is not a combining or C0/C1 control character */
197
198  return 1 +
199    (ucs >= 0x1100 &&
200     (ucs <= 0x115f ||                    /* Hangul Jamo init. consonants */
201      ucs == 0x2329 || ucs == 0x232a ||
202      (ucs >= 0x2e80 && ucs <= 0xa4cf &&
203       ucs != 0x303f) ||                  /* CJK ... Yi */
204      (ucs >= 0xac00 && ucs <= 0xd7a3) || /* Hangul Syllables */
205      (ucs >= 0xf900 && ucs <= 0xfaff) || /* CJK Compatibility Ideographs */
206      (ucs >= 0xfe10 && ucs <= 0xfe19) || /* Vertical forms */
207      (ucs >= 0xfe30 && ucs <= 0xfe6f) || /* CJK Compatibility Forms */
208      (ucs >= 0xff00 && ucs <= 0xff60) || /* Fullwidth Forms */
209      (ucs >= 0xffe0 && ucs <= 0xffe6) ||
210      (ucs >= 0x20000 && ucs <= 0x2fffd) ||
211      (ucs >= 0x30000 && ucs <= 0x3fffd)));
212}
213
214int
215svn_utf_cstring_utf8_width(const char *cstr)
216{
217  int width = 0;
218
219  if (*cstr == '\0')
220    return 0;
221
222  /* Ensure the conversion below doesn't fail because of encoding errors. */
223  if (!svn_utf__cstring_is_valid(cstr))
224    return -1;
225
226  /* Convert the UTF-8 string to UTF-32 (UCS4) which is the format
227   * mk_wcwidth() expects, and get the width of each character.
228   * We don't need much error checking since the input is valid UTF-8. */
229  while (*cstr)
230    {
231      apr_uint32_t ucs;
232      int nbytes;
233      int lead_mask;
234      int w;
235      int i;
236
237      if ((*cstr & 0x80) == 0)
238        {
239          nbytes = 1;
240          lead_mask = 0x7f;
241        }
242      else if ((*cstr & 0xe0) == 0xc0)
243        {
244          nbytes = 2;
245          lead_mask = 0x1f;
246        }
247      else if ((*cstr & 0xf0) == 0xe0)
248        {
249          nbytes = 3;
250          lead_mask = 0x0f;
251        }
252      else if ((*cstr & 0xf8) == 0xf0)
253        {
254          nbytes = 4;
255          lead_mask = 0x07;
256        }
257      else
258        {
259          /* RFC 3629 restricts UTF-8 to max 4 bytes per character. */
260          return -1;
261        }
262
263      /* Parse character data from leading byte. */
264      ucs = (apr_uint32_t)(*cstr & lead_mask);
265
266      /* Parse character data from continuation bytes. */
267      for (i = 1; i < nbytes; i++)
268        {
269          ucs <<= 6;
270          ucs |= (cstr[i] & 0x3f);
271        }
272
273      cstr += nbytes;
274
275      /* Determine the width of this character and add it to the total. */
276      w = mk_wcwidth(ucs);
277      if (w == -1)
278        return -1;
279      width += w;
280    }
281
282  return width;
283}
284