git.1: switch homepage for stats
[git.git] / utf8.c
blob77c28d492cccfcbcb8a302d168a81b21d909ef12
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 ucs_char_t first;
9 ucs_char_t 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, 0x036F }, { 0x0483, 0x0489 }, { 0x0591, 0x05BD },
88 { 0x05BF, 0x05BF }, { 0x05C1, 0x05C2 }, { 0x05C4, 0x05C5 },
89 { 0x05C7, 0x05C7 }, { 0x0600, 0x0604 }, { 0x0610, 0x061A },
90 { 0x064B, 0x065F }, { 0x0670, 0x0670 }, { 0x06D6, 0x06E4 },
91 { 0x06E7, 0x06E8 }, { 0x06EA, 0x06ED }, { 0x070F, 0x070F },
92 { 0x0711, 0x0711 }, { 0x0730, 0x074A }, { 0x07A6, 0x07B0 },
93 { 0x0901, 0x0902 }, { 0x093C, 0x093C }, { 0x0941, 0x0948 },
94 { 0x094D, 0x094D }, { 0x0951, 0x0954 }, { 0x0962, 0x0963 },
95 { 0x0981, 0x0981 }, { 0x09BC, 0x09BC }, { 0x09C1, 0x09C4 },
96 { 0x09CD, 0x09CD }, { 0x09E2, 0x09E3 }, { 0x0A01, 0x0A02 },
97 { 0x0A3C, 0x0A3C }, { 0x0A41, 0x0A42 }, { 0x0A47, 0x0A48 },
98 { 0x0A4B, 0x0A4D }, { 0x0A70, 0x0A71 }, { 0x0A81, 0x0A82 },
99 { 0x0ABC, 0x0ABC }, { 0x0AC1, 0x0AC5 }, { 0x0AC7, 0x0AC8 },
100 { 0x0ACD, 0x0ACD }, { 0x0AE2, 0x0AE3 }, { 0x0B01, 0x0B01 },
101 { 0x0B3C, 0x0B3C }, { 0x0B3F, 0x0B3F }, { 0x0B41, 0x0B43 },
102 { 0x0B4D, 0x0B4D }, { 0x0B56, 0x0B56 }, { 0x0B82, 0x0B82 },
103 { 0x0BC0, 0x0BC0 }, { 0x0BCD, 0x0BCD }, { 0x0C3E, 0x0C40 },
104 { 0x0C46, 0x0C48 }, { 0x0C4A, 0x0C4D }, { 0x0C55, 0x0C56 },
105 { 0x0CBC, 0x0CBC }, { 0x0CBF, 0x0CBF }, { 0x0CC6, 0x0CC6 },
106 { 0x0CCC, 0x0CCD }, { 0x0D41, 0x0D43 }, { 0x0D4D, 0x0D4D },
107 { 0x0DCA, 0x0DCA }, { 0x0DD2, 0x0DD4 }, { 0x0DD6, 0x0DD6 },
108 { 0x0E31, 0x0E31 }, { 0x0E34, 0x0E3A }, { 0x0E47, 0x0E4E },
109 { 0x0EB1, 0x0EB1 }, { 0x0EB4, 0x0EB9 }, { 0x0EBB, 0x0EBC },
110 { 0x0EC8, 0x0ECD }, { 0x0F18, 0x0F19 }, { 0x0F35, 0x0F35 },
111 { 0x0F37, 0x0F37 }, { 0x0F39, 0x0F39 }, { 0x0F71, 0x0F7E },
112 { 0x0F80, 0x0F84 }, { 0x0F86, 0x0F87 }, { 0x0F90, 0x0F97 },
113 { 0x0F99, 0x0FBC }, { 0x0FC6, 0x0FC6 }, { 0x102D, 0x1030 },
114 { 0x1032, 0x1032 }, { 0x1036, 0x1037 }, { 0x1039, 0x1039 },
115 { 0x1058, 0x1059 }, { 0x1160, 0x11FF }, { 0x1712, 0x1714 },
116 { 0x1732, 0x1734 }, { 0x1752, 0x1753 }, { 0x1772, 0x1773 },
117 { 0x17B4, 0x17B5 }, { 0x17B7, 0x17BD }, { 0x17C6, 0x17C6 },
118 { 0x17C9, 0x17D3 }, { 0x17DD, 0x17DD }, { 0x180B, 0x180D },
119 { 0x18A9, 0x18A9 }, { 0x1920, 0x1922 }, { 0x1927, 0x1928 },
120 { 0x1932, 0x1932 }, { 0x1939, 0x193B }, { 0x200B, 0x200F },
121 { 0x202A, 0x202E }, { 0x2060, 0x2063 }, { 0x206A, 0x206F },
122 { 0x20D0, 0x20EA }, { 0x302A, 0x302F }, { 0x3099, 0x309A },
123 { 0xFB1E, 0xFB1E }, { 0xFE00, 0xFE0F }, { 0xFE20, 0xFE23 },
124 { 0xFEFF, 0xFEFF }, { 0xFFF9, 0xFFFB }, { 0x1D167, 0x1D169 },
125 { 0x1D173, 0x1D182 }, { 0x1D185, 0x1D18B },
126 { 0x1D1AA, 0x1D1AD }, { 0xE0001, 0xE0001 },
127 { 0xE0020, 0xE007F }, { 0xE0100, 0xE01EF }
130 /* test for 8-bit control characters */
131 if (ch == 0)
132 return 0;
133 if (ch < 32 || (ch >= 0x7f && ch < 0xa0))
134 return -1;
136 /* binary search in table of non-spacing characters */
137 if (bisearch(ch, combining, sizeof(combining)
138 / sizeof(struct interval) - 1))
139 return 0;
142 * If we arrive here, ch is neither a combining nor a C0/C1
143 * control character.
146 return 1 +
147 (ch >= 0x1100 &&
148 /* Hangul Jamo init. consonants */
149 (ch <= 0x115f ||
150 ch == 0x2329 || ch == 0x232a ||
151 /* CJK ... Yi */
152 (ch >= 0x2e80 && ch <= 0xa4cf &&
153 ch != 0x303f) ||
154 /* Hangul Syllables */
155 (ch >= 0xac00 && ch <= 0xd7a3) ||
156 /* CJK Compatibility Ideographs */
157 (ch >= 0xf900 && ch <= 0xfaff) ||
158 /* CJK Compatibility Forms */
159 (ch >= 0xfe30 && ch <= 0xfe6f) ||
160 /* Fullwidth Forms */
161 (ch >= 0xff00 && ch <= 0xff60) ||
162 (ch >= 0xffe0 && ch <= 0xffe6) ||
163 (ch >= 0x20000 && ch <= 0x2fffd) ||
164 (ch >= 0x30000 && ch <= 0x3fffd)));
168 * Pick one ucs character starting from the location *start points at,
169 * and return it, while updating the *start pointer to point at the
170 * end of that character. When remainder_p is not NULL, the location
171 * holds the number of bytes remaining in the string that we are allowed
172 * to pick from. Otherwise we are allowed to pick up to the NUL that
173 * would eventually appear in the string. *remainder_p is also reduced
174 * by the number of bytes we have consumed.
176 * If the string was not a valid UTF-8, *start pointer is set to NULL
177 * and the return value is undefined.
179 static ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p)
181 unsigned char *s = (unsigned char *)*start;
182 ucs_char_t ch;
183 size_t remainder, incr;
186 * A caller that assumes NUL terminated text can choose
187 * not to bother with the remainder length. We will
188 * stop at the first NUL.
190 remainder = (remainder_p ? *remainder_p : 999);
192 if (remainder < 1) {
193 goto invalid;
194 } else if (*s < 0x80) {
195 /* 0xxxxxxx */
196 ch = *s;
197 incr = 1;
198 } else if ((s[0] & 0xe0) == 0xc0) {
199 /* 110XXXXx 10xxxxxx */
200 if (remainder < 2 ||
201 (s[1] & 0xc0) != 0x80 ||
202 (s[0] & 0xfe) == 0xc0)
203 goto invalid;
204 ch = ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
205 incr = 2;
206 } else if ((s[0] & 0xf0) == 0xe0) {
207 /* 1110XXXX 10Xxxxxx 10xxxxxx */
208 if (remainder < 3 ||
209 (s[1] & 0xc0) != 0x80 ||
210 (s[2] & 0xc0) != 0x80 ||
211 /* overlong? */
212 (s[0] == 0xe0 && (s[1] & 0xe0) == 0x80) ||
213 /* surrogate? */
214 (s[0] == 0xed && (s[1] & 0xe0) == 0xa0) ||
215 /* U+FFFE or U+FFFF? */
216 (s[0] == 0xef && s[1] == 0xbf &&
217 (s[2] & 0xfe) == 0xbe))
218 goto invalid;
219 ch = ((s[0] & 0x0f) << 12) |
220 ((s[1] & 0x3f) << 6) | (s[2] & 0x3f);
221 incr = 3;
222 } else if ((s[0] & 0xf8) == 0xf0) {
223 /* 11110XXX 10XXxxxx 10xxxxxx 10xxxxxx */
224 if (remainder < 4 ||
225 (s[1] & 0xc0) != 0x80 ||
226 (s[2] & 0xc0) != 0x80 ||
227 (s[3] & 0xc0) != 0x80 ||
228 /* overlong? */
229 (s[0] == 0xf0 && (s[1] & 0xf0) == 0x80) ||
230 /* > U+10FFFF? */
231 (s[0] == 0xf4 && s[1] > 0x8f) || s[0] > 0xf4)
232 goto invalid;
233 ch = ((s[0] & 0x07) << 18) | ((s[1] & 0x3f) << 12) |
234 ((s[2] & 0x3f) << 6) | (s[3] & 0x3f);
235 incr = 4;
236 } else {
237 invalid:
238 *start = NULL;
239 return 0;
242 *start += incr;
243 if (remainder_p)
244 *remainder_p = remainder - incr;
245 return ch;
249 * This function returns the number of columns occupied by the character
250 * pointed to by the variable start. The pointer is updated to point at
251 * the next character. When remainder_p is not NULL, it points at the
252 * location that stores the number of remaining bytes we can use to pick
253 * a character (see pick_one_utf8_char() above).
255 int utf8_width(const char **start, size_t *remainder_p)
257 ucs_char_t ch = pick_one_utf8_char(start, remainder_p);
258 if (!*start)
259 return 0;
260 return git_wcwidth(ch);
264 * Returns the total number of columns required by a null-terminated
265 * string, assuming that the string is utf8. Returns strlen() instead
266 * if the string does not look like a valid utf8 string.
268 int utf8_strnwidth(const char *string, int len, int skip_ansi)
270 int width = 0;
271 const char *orig = string;
273 if (len == -1)
274 len = strlen(string);
275 while (string && string < orig + len) {
276 int skip;
277 while (skip_ansi &&
278 (skip = display_mode_esc_sequence_len(string)) != 0)
279 string += skip;
280 width += utf8_width(&string, NULL);
282 return string ? width : len;
285 int utf8_strwidth(const char *string)
287 return utf8_strnwidth(string, -1, 0);
290 int is_utf8(const char *text)
292 while (*text) {
293 if (*text == '\n' || *text == '\t' || *text == '\r') {
294 text++;
295 continue;
297 utf8_width(&text, NULL);
298 if (!text)
299 return 0;
301 return 1;
304 static void strbuf_addchars(struct strbuf *sb, int c, size_t n)
306 strbuf_grow(sb, n);
307 memset(sb->buf + sb->len, c, n);
308 strbuf_setlen(sb, sb->len + n);
311 static void strbuf_add_indented_text(struct strbuf *buf, const char *text,
312 int indent, int indent2)
314 if (indent < 0)
315 indent = 0;
316 while (*text) {
317 const char *eol = strchrnul(text, '\n');
318 if (*eol == '\n')
319 eol++;
320 strbuf_addchars(buf, ' ', indent);
321 strbuf_add(buf, text, eol - text);
322 text = eol;
323 indent = indent2;
328 * Wrap the text, if necessary. The variable indent is the indent for the
329 * first line, indent2 is the indent for all other lines.
330 * If indent is negative, assume that already -indent columns have been
331 * consumed (and no extra indent is necessary for the first line).
333 void strbuf_add_wrapped_text(struct strbuf *buf,
334 const char *text, int indent1, int indent2, int width)
336 int indent, w, assume_utf8 = 1;
337 const char *bol, *space, *start = text;
338 size_t orig_len = buf->len;
340 if (width <= 0) {
341 strbuf_add_indented_text(buf, text, indent1, indent2);
342 return;
345 retry:
346 bol = text;
347 w = indent = indent1;
348 space = NULL;
349 if (indent < 0) {
350 w = -indent;
351 space = text;
354 for (;;) {
355 char c;
356 size_t skip;
358 while ((skip = display_mode_esc_sequence_len(text)))
359 text += skip;
361 c = *text;
362 if (!c || isspace(c)) {
363 if (w <= width || !space) {
364 const char *start = bol;
365 if (!c && text == start)
366 return;
367 if (space)
368 start = space;
369 else
370 strbuf_addchars(buf, ' ', indent);
371 strbuf_add(buf, start, text - start);
372 if (!c)
373 return;
374 space = text;
375 if (c == '\t')
376 w |= 0x07;
377 else if (c == '\n') {
378 space++;
379 if (*space == '\n') {
380 strbuf_addch(buf, '\n');
381 goto new_line;
383 else if (!isalnum(*space))
384 goto new_line;
385 else
386 strbuf_addch(buf, ' ');
388 w++;
389 text++;
391 else {
392 new_line:
393 strbuf_addch(buf, '\n');
394 text = bol = space + isspace(*space);
395 space = NULL;
396 w = indent = indent2;
398 continue;
400 if (assume_utf8) {
401 w += utf8_width(&text, NULL);
402 if (!text) {
403 assume_utf8 = 0;
404 text = start;
405 strbuf_setlen(buf, orig_len);
406 goto retry;
408 } else {
409 w++;
410 text++;
415 void strbuf_add_wrapped_bytes(struct strbuf *buf, const char *data, int len,
416 int indent, int indent2, int width)
418 char *tmp = xstrndup(data, len);
419 strbuf_add_wrapped_text(buf, tmp, indent, indent2, width);
420 free(tmp);
423 void strbuf_utf8_replace(struct strbuf *sb_src, int pos, int width,
424 const char *subst)
426 struct strbuf sb_dst = STRBUF_INIT;
427 char *src = sb_src->buf;
428 char *end = src + sb_src->len;
429 char *dst;
430 int w = 0, subst_len = 0;
432 if (subst)
433 subst_len = strlen(subst);
434 strbuf_grow(&sb_dst, sb_src->len + subst_len);
435 dst = sb_dst.buf;
437 while (src < end) {
438 char *old;
439 size_t n;
441 while ((n = display_mode_esc_sequence_len(src))) {
442 memcpy(dst, src, n);
443 src += n;
444 dst += n;
447 old = src;
448 n = utf8_width((const char**)&src, NULL);
449 if (!src) /* broken utf-8, do nothing */
450 return;
451 if (n && w >= pos && w < pos + width) {
452 if (subst) {
453 memcpy(dst, subst, subst_len);
454 dst += subst_len;
455 subst = NULL;
457 w += n;
458 continue;
460 memcpy(dst, old, src - old);
461 dst += src - old;
462 w += n;
464 strbuf_setlen(&sb_dst, dst - sb_dst.buf);
465 strbuf_swap(sb_src, &sb_dst);
466 strbuf_release(&sb_dst);
469 int is_encoding_utf8(const char *name)
471 if (!name)
472 return 1;
473 if (!strcasecmp(name, "utf-8") || !strcasecmp(name, "utf8"))
474 return 1;
475 return 0;
478 int same_encoding(const char *src, const char *dst)
480 if (is_encoding_utf8(src) && is_encoding_utf8(dst))
481 return 1;
482 return !strcasecmp(src, dst);
486 * Wrapper for fprintf and returns the total number of columns required
487 * for the printed string, assuming that the string is utf8.
489 int utf8_fprintf(FILE *stream, const char *format, ...)
491 struct strbuf buf = STRBUF_INIT;
492 va_list arg;
493 int columns;
495 va_start(arg, format);
496 strbuf_vaddf(&buf, format, arg);
497 va_end(arg);
499 columns = fputs(buf.buf, stream);
500 if (0 <= columns) /* keep the error from the I/O */
501 columns = utf8_strwidth(buf.buf);
502 strbuf_release(&buf);
503 return columns;
507 * Given a buffer and its encoding, return it re-encoded
508 * with iconv. If the conversion fails, returns NULL.
510 #ifndef NO_ICONV
511 #if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6))
512 typedef const char * iconv_ibp;
513 #else
514 typedef char * iconv_ibp;
515 #endif
516 char *reencode_string_iconv(const char *in, size_t insz, iconv_t conv, int *outsz_p)
518 size_t outsz, outalloc;
519 char *out, *outpos;
520 iconv_ibp cp;
522 outsz = insz;
523 outalloc = outsz + 1; /* for terminating NUL */
524 out = xmalloc(outalloc);
525 outpos = out;
526 cp = (iconv_ibp)in;
528 while (1) {
529 size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz);
531 if (cnt == (size_t) -1) {
532 size_t sofar;
533 if (errno != E2BIG) {
534 free(out);
535 return NULL;
537 /* insz has remaining number of bytes.
538 * since we started outsz the same as insz,
539 * it is likely that insz is not enough for
540 * converting the rest.
542 sofar = outpos - out;
543 outalloc = sofar + insz * 2 + 32;
544 out = xrealloc(out, outalloc);
545 outpos = out + sofar;
546 outsz = outalloc - sofar - 1;
548 else {
549 *outpos = '\0';
550 if (outsz_p)
551 *outsz_p = outpos - out;
552 break;
555 return out;
558 char *reencode_string_len(const char *in, int insz,
559 const char *out_encoding, const char *in_encoding,
560 int *outsz)
562 iconv_t conv;
563 char *out;
565 if (!in_encoding)
566 return NULL;
568 conv = iconv_open(out_encoding, in_encoding);
569 if (conv == (iconv_t) -1) {
571 * Some platforms do not have the variously spelled variants of
572 * UTF-8, so let's fall back to trying the most official
573 * spelling. We do so only as a fallback in case the platform
574 * does understand the user's spelling, but not our official
575 * one.
577 if (is_encoding_utf8(in_encoding))
578 in_encoding = "UTF-8";
579 if (is_encoding_utf8(out_encoding))
580 out_encoding = "UTF-8";
581 conv = iconv_open(out_encoding, in_encoding);
582 if (conv == (iconv_t) -1)
583 return NULL;
586 out = reencode_string_iconv(in, insz, conv, outsz);
587 iconv_close(conv);
588 return out;
590 #endif
593 * Returns first character length in bytes for multi-byte `text` according to
594 * `encoding`.
596 * - The `text` pointer is updated to point at the next character.
597 * - When `remainder_p` is not NULL, on entry `*remainder_p` is how much bytes
598 * we can consume from text, and on exit `*remainder_p` is reduced by returned
599 * character length. Otherwise `text` is treated as limited by NUL.
601 int mbs_chrlen(const char **text, size_t *remainder_p, const char *encoding)
603 int chrlen;
604 const char *p = *text;
605 size_t r = (remainder_p ? *remainder_p : SIZE_MAX);
607 if (r < 1)
608 return 0;
610 if (is_encoding_utf8(encoding)) {
611 pick_one_utf8_char(&p, &r);
613 chrlen = p ? (p - *text)
614 : 1 /* not valid UTF-8 -> raw byte sequence */;
616 else {
618 * TODO use iconv to decode one char and obtain its chrlen
619 * for now, let's treat encodings != UTF-8 as one-byte
621 chrlen = 1;
624 *text += chrlen;
625 if (remainder_p)
626 *remainder_p -= chrlen;
628 return chrlen;