Merge branch 'sb/repack-in-c'
[alt-git.git] / utf8.c
blob0d20e0acb2b6fb2dd1d63abed676a036b2b7ec2f
1 #include "git-compat-util.h"
2 #include "strbuf.h"
3 #include "utf8.h"
5 /* This code is originally from http://www.cl.cam.ac.uk/~mgk25/ucs/ */
7 struct interval {
8 int first;
9 int last;
12 size_t display_mode_esc_sequence_len(const char *s)
14 const char *p = s;
15 if (*p++ != '\033')
16 return 0;
17 if (*p++ != '[')
18 return 0;
19 while (isdigit(*p) || *p == ';')
20 p++;
21 if (*p++ != 'm')
22 return 0;
23 return p - s;
26 /* auxiliary function for binary search in interval table */
27 static int bisearch(ucs_char_t ucs, const struct interval *table, int max)
29 int min = 0;
30 int mid;
32 if (ucs < table[0].first || ucs > table[max].last)
33 return 0;
34 while (max >= min) {
35 mid = (min + max) / 2;
36 if (ucs > table[mid].last)
37 min = mid + 1;
38 else if (ucs < table[mid].first)
39 max = mid - 1;
40 else
41 return 1;
44 return 0;
47 /* The following two functions define the column width of an ISO 10646
48 * character as follows:
50 * - The null character (U+0000) has a column width of 0.
52 * - Other C0/C1 control characters and DEL will lead to a return
53 * value of -1.
55 * - Non-spacing and enclosing combining characters (general
56 * category code Mn or Me in the Unicode database) have a
57 * column width of 0.
59 * - SOFT HYPHEN (U+00AD) has a column width of 1.
61 * - Other format characters (general category code Cf in the Unicode
62 * database) and ZERO WIDTH SPACE (U+200B) have a column width of 0.
64 * - Hangul Jamo medial vowels and final consonants (U+1160-U+11FF)
65 * have a column width of 0.
67 * - Spacing characters in the East Asian Wide (W) or East Asian
68 * Full-width (F) category as defined in Unicode Technical
69 * Report #11 have a column width of 2.
71 * - All remaining characters (including all printable
72 * ISO 8859-1 and WGL4 characters, Unicode control characters,
73 * etc.) have a column width of 1.
75 * This implementation assumes that ucs_char_t characters are encoded
76 * in ISO 10646.
79 static int git_wcwidth(ucs_char_t ch)
82 * Sorted list of non-overlapping intervals of non-spacing characters,
83 * generated by
84 * "uniset +cat=Me +cat=Mn +cat=Cf -00AD +1160-11FF +200B c".
86 static const struct interval combining[] = {
87 { 0x0300, 0x0357 }, { 0x035D, 0x036F }, { 0x0483, 0x0486 },
88 { 0x0488, 0x0489 }, { 0x0591, 0x05A1 }, { 0x05A3, 0x05B9 },
89 { 0x05BB, 0x05BD }, { 0x05BF, 0x05BF }, { 0x05C1, 0x05C2 },
90 { 0x05C4, 0x05C4 }, { 0x0600, 0x0603 }, { 0x0610, 0x0615 },
91 { 0x064B, 0x0658 }, { 0x0670, 0x0670 }, { 0x06D6, 0x06E4 },
92 { 0x06E7, 0x06E8 }, { 0x06EA, 0x06ED }, { 0x070F, 0x070F },
93 { 0x0711, 0x0711 }, { 0x0730, 0x074A }, { 0x07A6, 0x07B0 },
94 { 0x0901, 0x0902 }, { 0x093C, 0x093C }, { 0x0941, 0x0948 },
95 { 0x094D, 0x094D }, { 0x0951, 0x0954 }, { 0x0962, 0x0963 },
96 { 0x0981, 0x0981 }, { 0x09BC, 0x09BC }, { 0x09C1, 0x09C4 },
97 { 0x09CD, 0x09CD }, { 0x09E2, 0x09E3 }, { 0x0A01, 0x0A02 },
98 { 0x0A3C, 0x0A3C }, { 0x0A41, 0x0A42 }, { 0x0A47, 0x0A48 },
99 { 0x0A4B, 0x0A4D }, { 0x0A70, 0x0A71 }, { 0x0A81, 0x0A82 },
100 { 0x0ABC, 0x0ABC }, { 0x0AC1, 0x0AC5 }, { 0x0AC7, 0x0AC8 },
101 { 0x0ACD, 0x0ACD }, { 0x0AE2, 0x0AE3 }, { 0x0B01, 0x0B01 },
102 { 0x0B3C, 0x0B3C }, { 0x0B3F, 0x0B3F }, { 0x0B41, 0x0B43 },
103 { 0x0B4D, 0x0B4D }, { 0x0B56, 0x0B56 }, { 0x0B82, 0x0B82 },
104 { 0x0BC0, 0x0BC0 }, { 0x0BCD, 0x0BCD }, { 0x0C3E, 0x0C40 },
105 { 0x0C46, 0x0C48 }, { 0x0C4A, 0x0C4D }, { 0x0C55, 0x0C56 },
106 { 0x0CBC, 0x0CBC }, { 0x0CBF, 0x0CBF }, { 0x0CC6, 0x0CC6 },
107 { 0x0CCC, 0x0CCD }, { 0x0D41, 0x0D43 }, { 0x0D4D, 0x0D4D },
108 { 0x0DCA, 0x0DCA }, { 0x0DD2, 0x0DD4 }, { 0x0DD6, 0x0DD6 },
109 { 0x0E31, 0x0E31 }, { 0x0E34, 0x0E3A }, { 0x0E47, 0x0E4E },
110 { 0x0EB1, 0x0EB1 }, { 0x0EB4, 0x0EB9 }, { 0x0EBB, 0x0EBC },
111 { 0x0EC8, 0x0ECD }, { 0x0F18, 0x0F19 }, { 0x0F35, 0x0F35 },
112 { 0x0F37, 0x0F37 }, { 0x0F39, 0x0F39 }, { 0x0F71, 0x0F7E },
113 { 0x0F80, 0x0F84 }, { 0x0F86, 0x0F87 }, { 0x0F90, 0x0F97 },
114 { 0x0F99, 0x0FBC }, { 0x0FC6, 0x0FC6 }, { 0x102D, 0x1030 },
115 { 0x1032, 0x1032 }, { 0x1036, 0x1037 }, { 0x1039, 0x1039 },
116 { 0x1058, 0x1059 }, { 0x1160, 0x11FF }, { 0x1712, 0x1714 },
117 { 0x1732, 0x1734 }, { 0x1752, 0x1753 }, { 0x1772, 0x1773 },
118 { 0x17B4, 0x17B5 }, { 0x17B7, 0x17BD }, { 0x17C6, 0x17C6 },
119 { 0x17C9, 0x17D3 }, { 0x17DD, 0x17DD }, { 0x180B, 0x180D },
120 { 0x18A9, 0x18A9 }, { 0x1920, 0x1922 }, { 0x1927, 0x1928 },
121 { 0x1932, 0x1932 }, { 0x1939, 0x193B }, { 0x200B, 0x200F },
122 { 0x202A, 0x202E }, { 0x2060, 0x2063 }, { 0x206A, 0x206F },
123 { 0x20D0, 0x20EA }, { 0x302A, 0x302F }, { 0x3099, 0x309A },
124 { 0xFB1E, 0xFB1E }, { 0xFE00, 0xFE0F }, { 0xFE20, 0xFE23 },
125 { 0xFEFF, 0xFEFF }, { 0xFFF9, 0xFFFB }, { 0x1D167, 0x1D169 },
126 { 0x1D173, 0x1D182 }, { 0x1D185, 0x1D18B },
127 { 0x1D1AA, 0x1D1AD }, { 0xE0001, 0xE0001 },
128 { 0xE0020, 0xE007F }, { 0xE0100, 0xE01EF }
131 /* test for 8-bit control characters */
132 if (ch == 0)
133 return 0;
134 if (ch < 32 || (ch >= 0x7f && ch < 0xa0))
135 return -1;
137 /* binary search in table of non-spacing characters */
138 if (bisearch(ch, combining, sizeof(combining)
139 / sizeof(struct interval) - 1))
140 return 0;
143 * If we arrive here, ch is neither a combining nor a C0/C1
144 * control character.
147 return 1 +
148 (ch >= 0x1100 &&
149 /* Hangul Jamo init. consonants */
150 (ch <= 0x115f ||
151 ch == 0x2329 || ch == 0x232a ||
152 /* CJK ... Yi */
153 (ch >= 0x2e80 && ch <= 0xa4cf &&
154 ch != 0x303f) ||
155 /* Hangul Syllables */
156 (ch >= 0xac00 && ch <= 0xd7a3) ||
157 /* CJK Compatibility Ideographs */
158 (ch >= 0xf900 && ch <= 0xfaff) ||
159 /* CJK Compatibility Forms */
160 (ch >= 0xfe30 && ch <= 0xfe6f) ||
161 /* Fullwidth Forms */
162 (ch >= 0xff00 && ch <= 0xff60) ||
163 (ch >= 0xffe0 && ch <= 0xffe6) ||
164 (ch >= 0x20000 && ch <= 0x2fffd) ||
165 (ch >= 0x30000 && ch <= 0x3fffd)));
169 * Pick one ucs character starting from the location *start points at,
170 * and return it, while updating the *start pointer to point at the
171 * end of that character. When remainder_p is not NULL, the location
172 * holds the number of bytes remaining in the string that we are allowed
173 * to pick from. Otherwise we are allowed to pick up to the NUL that
174 * would eventually appear in the string. *remainder_p is also reduced
175 * by the number of bytes we have consumed.
177 * If the string was not a valid UTF-8, *start pointer is set to NULL
178 * and the return value is undefined.
180 static ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p)
182 unsigned char *s = (unsigned char *)*start;
183 ucs_char_t ch;
184 size_t remainder, incr;
187 * A caller that assumes NUL terminated text can choose
188 * not to bother with the remainder length. We will
189 * stop at the first NUL.
191 remainder = (remainder_p ? *remainder_p : 999);
193 if (remainder < 1) {
194 goto invalid;
195 } else if (*s < 0x80) {
196 /* 0xxxxxxx */
197 ch = *s;
198 incr = 1;
199 } else if ((s[0] & 0xe0) == 0xc0) {
200 /* 110XXXXx 10xxxxxx */
201 if (remainder < 2 ||
202 (s[1] & 0xc0) != 0x80 ||
203 (s[0] & 0xfe) == 0xc0)
204 goto invalid;
205 ch = ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
206 incr = 2;
207 } else if ((s[0] & 0xf0) == 0xe0) {
208 /* 1110XXXX 10Xxxxxx 10xxxxxx */
209 if (remainder < 3 ||
210 (s[1] & 0xc0) != 0x80 ||
211 (s[2] & 0xc0) != 0x80 ||
212 /* overlong? */
213 (s[0] == 0xe0 && (s[1] & 0xe0) == 0x80) ||
214 /* surrogate? */
215 (s[0] == 0xed && (s[1] & 0xe0) == 0xa0) ||
216 /* U+FFFE or U+FFFF? */
217 (s[0] == 0xef && s[1] == 0xbf &&
218 (s[2] & 0xfe) == 0xbe))
219 goto invalid;
220 ch = ((s[0] & 0x0f) << 12) |
221 ((s[1] & 0x3f) << 6) | (s[2] & 0x3f);
222 incr = 3;
223 } else if ((s[0] & 0xf8) == 0xf0) {
224 /* 11110XXX 10XXxxxx 10xxxxxx 10xxxxxx */
225 if (remainder < 4 ||
226 (s[1] & 0xc0) != 0x80 ||
227 (s[2] & 0xc0) != 0x80 ||
228 (s[3] & 0xc0) != 0x80 ||
229 /* overlong? */
230 (s[0] == 0xf0 && (s[1] & 0xf0) == 0x80) ||
231 /* > U+10FFFF? */
232 (s[0] == 0xf4 && s[1] > 0x8f) || s[0] > 0xf4)
233 goto invalid;
234 ch = ((s[0] & 0x07) << 18) | ((s[1] & 0x3f) << 12) |
235 ((s[2] & 0x3f) << 6) | (s[3] & 0x3f);
236 incr = 4;
237 } else {
238 invalid:
239 *start = NULL;
240 return 0;
243 *start += incr;
244 if (remainder_p)
245 *remainder_p = remainder - incr;
246 return ch;
250 * This function returns the number of columns occupied by the character
251 * pointed to by the variable start. The pointer is updated to point at
252 * the next character. When remainder_p is not NULL, it points at the
253 * location that stores the number of remaining bytes we can use to pick
254 * a character (see pick_one_utf8_char() above).
256 int utf8_width(const char **start, size_t *remainder_p)
258 ucs_char_t ch = pick_one_utf8_char(start, remainder_p);
259 if (!*start)
260 return 0;
261 return git_wcwidth(ch);
265 * Returns the total number of columns required by a null-terminated
266 * string, assuming that the string is utf8. Returns strlen() instead
267 * if the string does not look like a valid utf8 string.
269 int utf8_strnwidth(const char *string, int len, int skip_ansi)
271 int width = 0;
272 const char *orig = string;
274 if (len == -1)
275 len = strlen(string);
276 while (string && string < orig + len) {
277 int skip;
278 while (skip_ansi &&
279 (skip = display_mode_esc_sequence_len(string)) != 0)
280 string += skip;
281 width += utf8_width(&string, NULL);
283 return string ? width : len;
286 int utf8_strwidth(const char *string)
288 return utf8_strnwidth(string, -1, 0);
291 int is_utf8(const char *text)
293 while (*text) {
294 if (*text == '\n' || *text == '\t' || *text == '\r') {
295 text++;
296 continue;
298 utf8_width(&text, NULL);
299 if (!text)
300 return 0;
302 return 1;
305 static void strbuf_addchars(struct strbuf *sb, int c, size_t n)
307 strbuf_grow(sb, n);
308 memset(sb->buf + sb->len, c, n);
309 strbuf_setlen(sb, sb->len + n);
312 static void strbuf_add_indented_text(struct strbuf *buf, const char *text,
313 int indent, int indent2)
315 if (indent < 0)
316 indent = 0;
317 while (*text) {
318 const char *eol = strchrnul(text, '\n');
319 if (*eol == '\n')
320 eol++;
321 strbuf_addchars(buf, ' ', indent);
322 strbuf_add(buf, text, eol - text);
323 text = eol;
324 indent = indent2;
329 * Wrap the text, if necessary. The variable indent is the indent for the
330 * first line, indent2 is the indent for all other lines.
331 * If indent is negative, assume that already -indent columns have been
332 * consumed (and no extra indent is necessary for the first line).
334 void strbuf_add_wrapped_text(struct strbuf *buf,
335 const char *text, int indent1, int indent2, int width)
337 int indent, w, assume_utf8 = 1;
338 const char *bol, *space, *start = text;
339 size_t orig_len = buf->len;
341 if (width <= 0) {
342 strbuf_add_indented_text(buf, text, indent1, indent2);
343 return;
346 retry:
347 bol = text;
348 w = indent = indent1;
349 space = NULL;
350 if (indent < 0) {
351 w = -indent;
352 space = text;
355 for (;;) {
356 char c;
357 size_t skip;
359 while ((skip = display_mode_esc_sequence_len(text)))
360 text += skip;
362 c = *text;
363 if (!c || isspace(c)) {
364 if (w <= width || !space) {
365 const char *start = bol;
366 if (!c && text == start)
367 return;
368 if (space)
369 start = space;
370 else
371 strbuf_addchars(buf, ' ', indent);
372 strbuf_add(buf, start, text - start);
373 if (!c)
374 return;
375 space = text;
376 if (c == '\t')
377 w |= 0x07;
378 else if (c == '\n') {
379 space++;
380 if (*space == '\n') {
381 strbuf_addch(buf, '\n');
382 goto new_line;
384 else if (!isalnum(*space))
385 goto new_line;
386 else
387 strbuf_addch(buf, ' ');
389 w++;
390 text++;
392 else {
393 new_line:
394 strbuf_addch(buf, '\n');
395 text = bol = space + isspace(*space);
396 space = NULL;
397 w = indent = indent2;
399 continue;
401 if (assume_utf8) {
402 w += utf8_width(&text, NULL);
403 if (!text) {
404 assume_utf8 = 0;
405 text = start;
406 strbuf_setlen(buf, orig_len);
407 goto retry;
409 } else {
410 w++;
411 text++;
416 void strbuf_add_wrapped_bytes(struct strbuf *buf, const char *data, int len,
417 int indent, int indent2, int width)
419 char *tmp = xstrndup(data, len);
420 strbuf_add_wrapped_text(buf, tmp, indent, indent2, width);
421 free(tmp);
424 void strbuf_utf8_replace(struct strbuf *sb_src, int pos, int width,
425 const char *subst)
427 struct strbuf sb_dst = STRBUF_INIT;
428 char *src = sb_src->buf;
429 char *end = src + sb_src->len;
430 char *dst;
431 int w = 0, subst_len = 0;
433 if (subst)
434 subst_len = strlen(subst);
435 strbuf_grow(&sb_dst, sb_src->len + subst_len);
436 dst = sb_dst.buf;
438 while (src < end) {
439 char *old;
440 size_t n;
442 while ((n = display_mode_esc_sequence_len(src))) {
443 memcpy(dst, src, n);
444 src += n;
445 dst += n;
448 old = src;
449 n = utf8_width((const char**)&src, NULL);
450 if (!src) /* broken utf-8, do nothing */
451 return;
452 if (n && w >= pos && w < pos + width) {
453 if (subst) {
454 memcpy(dst, subst, subst_len);
455 dst += subst_len;
456 subst = NULL;
458 w += n;
459 continue;
461 memcpy(dst, old, src - old);
462 dst += src - old;
463 w += n;
465 strbuf_setlen(&sb_dst, dst - sb_dst.buf);
466 strbuf_swap(sb_src, &sb_dst);
467 strbuf_release(&sb_dst);
470 int is_encoding_utf8(const char *name)
472 if (!name)
473 return 1;
474 if (!strcasecmp(name, "utf-8") || !strcasecmp(name, "utf8"))
475 return 1;
476 return 0;
479 int same_encoding(const char *src, const char *dst)
481 if (is_encoding_utf8(src) && is_encoding_utf8(dst))
482 return 1;
483 return !strcasecmp(src, dst);
487 * Wrapper for fprintf and returns the total number of columns required
488 * for the printed string, assuming that the string is utf8.
490 int utf8_fprintf(FILE *stream, const char *format, ...)
492 struct strbuf buf = STRBUF_INIT;
493 va_list arg;
494 int columns;
496 va_start(arg, format);
497 strbuf_vaddf(&buf, format, arg);
498 va_end(arg);
500 columns = fputs(buf.buf, stream);
501 if (0 <= columns) /* keep the error from the I/O */
502 columns = utf8_strwidth(buf.buf);
503 strbuf_release(&buf);
504 return columns;
508 * Given a buffer and its encoding, return it re-encoded
509 * with iconv. If the conversion fails, returns NULL.
511 #ifndef NO_ICONV
512 #if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6))
513 typedef const char * iconv_ibp;
514 #else
515 typedef char * iconv_ibp;
516 #endif
517 char *reencode_string_iconv(const char *in, size_t insz, iconv_t conv, int *outsz_p)
519 size_t outsz, outalloc;
520 char *out, *outpos;
521 iconv_ibp cp;
523 outsz = insz;
524 outalloc = outsz + 1; /* for terminating NUL */
525 out = xmalloc(outalloc);
526 outpos = out;
527 cp = (iconv_ibp)in;
529 while (1) {
530 size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz);
532 if (cnt == -1) {
533 size_t sofar;
534 if (errno != E2BIG) {
535 free(out);
536 return NULL;
538 /* insz has remaining number of bytes.
539 * since we started outsz the same as insz,
540 * it is likely that insz is not enough for
541 * converting the rest.
543 sofar = outpos - out;
544 outalloc = sofar + insz * 2 + 32;
545 out = xrealloc(out, outalloc);
546 outpos = out + sofar;
547 outsz = outalloc - sofar - 1;
549 else {
550 *outpos = '\0';
551 if (outsz_p)
552 *outsz_p = outpos - out;
553 break;
556 return out;
559 char *reencode_string_len(const char *in, int insz,
560 const char *out_encoding, const char *in_encoding,
561 int *outsz)
563 iconv_t conv;
564 char *out;
566 if (!in_encoding)
567 return NULL;
569 conv = iconv_open(out_encoding, in_encoding);
570 if (conv == (iconv_t) -1) {
572 * Some platforms do not have the variously spelled variants of
573 * UTF-8, so let's fall back to trying the most official
574 * spelling. We do so only as a fallback in case the platform
575 * does understand the user's spelling, but not our official
576 * one.
578 if (is_encoding_utf8(in_encoding))
579 in_encoding = "UTF-8";
580 if (is_encoding_utf8(out_encoding))
581 out_encoding = "UTF-8";
582 conv = iconv_open(out_encoding, in_encoding);
583 if (conv == (iconv_t) -1)
584 return NULL;
587 out = reencode_string_iconv(in, insz, conv, outsz);
588 iconv_close(conv);
589 return out;
591 #endif
594 * Returns first character length in bytes for multi-byte `text` according to
595 * `encoding`.
597 * - The `text` pointer is updated to point at the next character.
598 * - When `remainder_p` is not NULL, on entry `*remainder_p` is how much bytes
599 * we can consume from text, and on exit `*remainder_p` is reduced by returned
600 * character length. Otherwise `text` is treated as limited by NUL.
602 int mbs_chrlen(const char **text, size_t *remainder_p, const char *encoding)
604 int chrlen;
605 const char *p = *text;
606 size_t r = (remainder_p ? *remainder_p : SIZE_MAX);
608 if (r < 1)
609 return 0;
611 if (is_encoding_utf8(encoding)) {
612 pick_one_utf8_char(&p, &r);
614 chrlen = p ? (p - *text)
615 : 1 /* not valid UTF-8 -> raw byte sequence */;
617 else {
619 * TODO use iconv to decode one char and obtain its chrlen
620 * for now, let's treat encodings != UTF-8 as one-byte
622 chrlen = 1;
625 *text += chrlen;
626 if (remainder_p)
627 *remainder_p -= chrlen;
629 return chrlen;