Interested in helping open source friends on HP-UX?
[git/mjg.git] / utf8.c
blob520fbb4994ab1bdcfa0bc4589a0ca5131a89c490
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,
84 #include "unicode_width.h"
86 /* test for 8-bit control characters */
87 if (ch == 0)
88 return 0;
89 if (ch < 32 || (ch >= 0x7f && ch < 0xa0))
90 return -1;
92 /* binary search in table of non-spacing characters */
93 if (bisearch(ch, zero_width, sizeof(zero_width)
94 / sizeof(struct interval) - 1))
95 return 0;
97 /* binary search in table of double width characters */
98 if (bisearch(ch, double_width, sizeof(double_width)
99 / sizeof(struct interval) - 1))
100 return 2;
102 return 1;
106 * Pick one ucs character starting from the location *start points at,
107 * and return it, while updating the *start pointer to point at the
108 * end of that character. When remainder_p is not NULL, the location
109 * holds the number of bytes remaining in the string that we are allowed
110 * to pick from. Otherwise we are allowed to pick up to the NUL that
111 * would eventually appear in the string. *remainder_p is also reduced
112 * by the number of bytes we have consumed.
114 * If the string was not a valid UTF-8, *start pointer is set to NULL
115 * and the return value is undefined.
117 static ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p)
119 unsigned char *s = (unsigned char *)*start;
120 ucs_char_t ch;
121 size_t remainder, incr;
124 * A caller that assumes NUL terminated text can choose
125 * not to bother with the remainder length. We will
126 * stop at the first NUL.
128 remainder = (remainder_p ? *remainder_p : 999);
130 if (remainder < 1) {
131 goto invalid;
132 } else if (*s < 0x80) {
133 /* 0xxxxxxx */
134 ch = *s;
135 incr = 1;
136 } else if ((s[0] & 0xe0) == 0xc0) {
137 /* 110XXXXx 10xxxxxx */
138 if (remainder < 2 ||
139 (s[1] & 0xc0) != 0x80 ||
140 (s[0] & 0xfe) == 0xc0)
141 goto invalid;
142 ch = ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
143 incr = 2;
144 } else if ((s[0] & 0xf0) == 0xe0) {
145 /* 1110XXXX 10Xxxxxx 10xxxxxx */
146 if (remainder < 3 ||
147 (s[1] & 0xc0) != 0x80 ||
148 (s[2] & 0xc0) != 0x80 ||
149 /* overlong? */
150 (s[0] == 0xe0 && (s[1] & 0xe0) == 0x80) ||
151 /* surrogate? */
152 (s[0] == 0xed && (s[1] & 0xe0) == 0xa0) ||
153 /* U+FFFE or U+FFFF? */
154 (s[0] == 0xef && s[1] == 0xbf &&
155 (s[2] & 0xfe) == 0xbe))
156 goto invalid;
157 ch = ((s[0] & 0x0f) << 12) |
158 ((s[1] & 0x3f) << 6) | (s[2] & 0x3f);
159 incr = 3;
160 } else if ((s[0] & 0xf8) == 0xf0) {
161 /* 11110XXX 10XXxxxx 10xxxxxx 10xxxxxx */
162 if (remainder < 4 ||
163 (s[1] & 0xc0) != 0x80 ||
164 (s[2] & 0xc0) != 0x80 ||
165 (s[3] & 0xc0) != 0x80 ||
166 /* overlong? */
167 (s[0] == 0xf0 && (s[1] & 0xf0) == 0x80) ||
168 /* > U+10FFFF? */
169 (s[0] == 0xf4 && s[1] > 0x8f) || s[0] > 0xf4)
170 goto invalid;
171 ch = ((s[0] & 0x07) << 18) | ((s[1] & 0x3f) << 12) |
172 ((s[2] & 0x3f) << 6) | (s[3] & 0x3f);
173 incr = 4;
174 } else {
175 invalid:
176 *start = NULL;
177 return 0;
180 *start += incr;
181 if (remainder_p)
182 *remainder_p = remainder - incr;
183 return ch;
187 * This function returns the number of columns occupied by the character
188 * pointed to by the variable start. The pointer is updated to point at
189 * the next character. When remainder_p is not NULL, it points at the
190 * location that stores the number of remaining bytes we can use to pick
191 * a character (see pick_one_utf8_char() above).
193 int utf8_width(const char **start, size_t *remainder_p)
195 ucs_char_t ch = pick_one_utf8_char(start, remainder_p);
196 if (!*start)
197 return 0;
198 return git_wcwidth(ch);
202 * Returns the total number of columns required by a null-terminated
203 * string, assuming that the string is utf8. Returns strlen() instead
204 * if the string does not look like a valid utf8 string.
206 int utf8_strnwidth(const char *string, int len, int skip_ansi)
208 int width = 0;
209 const char *orig = string;
211 if (len == -1)
212 len = strlen(string);
213 while (string && string < orig + len) {
214 int skip;
215 while (skip_ansi &&
216 (skip = display_mode_esc_sequence_len(string)) != 0)
217 string += skip;
218 width += utf8_width(&string, NULL);
220 return string ? width : len;
223 int utf8_strwidth(const char *string)
225 return utf8_strnwidth(string, -1, 0);
228 int is_utf8(const char *text)
230 while (*text) {
231 if (*text == '\n' || *text == '\t' || *text == '\r') {
232 text++;
233 continue;
235 utf8_width(&text, NULL);
236 if (!text)
237 return 0;
239 return 1;
242 static void strbuf_add_indented_text(struct strbuf *buf, const char *text,
243 int indent, int indent2)
245 if (indent < 0)
246 indent = 0;
247 while (*text) {
248 const char *eol = strchrnul(text, '\n');
249 if (*eol == '\n')
250 eol++;
251 strbuf_addchars(buf, ' ', indent);
252 strbuf_add(buf, text, eol - text);
253 text = eol;
254 indent = indent2;
259 * Wrap the text, if necessary. The variable indent is the indent for the
260 * first line, indent2 is the indent for all other lines.
261 * If indent is negative, assume that already -indent columns have been
262 * consumed (and no extra indent is necessary for the first line).
264 void strbuf_add_wrapped_text(struct strbuf *buf,
265 const char *text, int indent1, int indent2, int width)
267 int indent, w, assume_utf8 = 1;
268 const char *bol, *space, *start = text;
269 size_t orig_len = buf->len;
271 if (width <= 0) {
272 strbuf_add_indented_text(buf, text, indent1, indent2);
273 return;
276 retry:
277 bol = text;
278 w = indent = indent1;
279 space = NULL;
280 if (indent < 0) {
281 w = -indent;
282 space = text;
285 for (;;) {
286 char c;
287 size_t skip;
289 while ((skip = display_mode_esc_sequence_len(text)))
290 text += skip;
292 c = *text;
293 if (!c || isspace(c)) {
294 if (w <= width || !space) {
295 const char *start = bol;
296 if (!c && text == start)
297 return;
298 if (space)
299 start = space;
300 else
301 strbuf_addchars(buf, ' ', indent);
302 strbuf_add(buf, start, text - start);
303 if (!c)
304 return;
305 space = text;
306 if (c == '\t')
307 w |= 0x07;
308 else if (c == '\n') {
309 space++;
310 if (*space == '\n') {
311 strbuf_addch(buf, '\n');
312 goto new_line;
314 else if (!isalnum(*space))
315 goto new_line;
316 else
317 strbuf_addch(buf, ' ');
319 w++;
320 text++;
322 else {
323 new_line:
324 strbuf_addch(buf, '\n');
325 text = bol = space + isspace(*space);
326 space = NULL;
327 w = indent = indent2;
329 continue;
331 if (assume_utf8) {
332 w += utf8_width(&text, NULL);
333 if (!text) {
334 assume_utf8 = 0;
335 text = start;
336 strbuf_setlen(buf, orig_len);
337 goto retry;
339 } else {
340 w++;
341 text++;
346 void strbuf_add_wrapped_bytes(struct strbuf *buf, const char *data, int len,
347 int indent, int indent2, int width)
349 char *tmp = xstrndup(data, len);
350 strbuf_add_wrapped_text(buf, tmp, indent, indent2, width);
351 free(tmp);
354 void strbuf_utf8_replace(struct strbuf *sb_src, int pos, int width,
355 const char *subst)
357 struct strbuf sb_dst = STRBUF_INIT;
358 char *src = sb_src->buf;
359 char *end = src + sb_src->len;
360 char *dst;
361 int w = 0, subst_len = 0;
363 if (subst)
364 subst_len = strlen(subst);
365 strbuf_grow(&sb_dst, sb_src->len + subst_len);
366 dst = sb_dst.buf;
368 while (src < end) {
369 char *old;
370 size_t n;
372 while ((n = display_mode_esc_sequence_len(src))) {
373 memcpy(dst, src, n);
374 src += n;
375 dst += n;
378 if (src >= end)
379 break;
381 old = src;
382 n = utf8_width((const char**)&src, NULL);
383 if (!src) /* broken utf-8, do nothing */
384 return;
385 if (n && w >= pos && w < pos + width) {
386 if (subst) {
387 memcpy(dst, subst, subst_len);
388 dst += subst_len;
389 subst = NULL;
391 w += n;
392 continue;
394 memcpy(dst, old, src - old);
395 dst += src - old;
396 w += n;
398 strbuf_setlen(&sb_dst, dst - sb_dst.buf);
399 strbuf_swap(sb_src, &sb_dst);
400 strbuf_release(&sb_dst);
403 int is_encoding_utf8(const char *name)
405 if (!name)
406 return 1;
407 if (!strcasecmp(name, "utf-8") || !strcasecmp(name, "utf8"))
408 return 1;
409 return 0;
412 int same_encoding(const char *src, const char *dst)
414 if (is_encoding_utf8(src) && is_encoding_utf8(dst))
415 return 1;
416 return !strcasecmp(src, dst);
420 * Wrapper for fprintf and returns the total number of columns required
421 * for the printed string, assuming that the string is utf8.
423 int utf8_fprintf(FILE *stream, const char *format, ...)
425 struct strbuf buf = STRBUF_INIT;
426 va_list arg;
427 int columns;
429 va_start(arg, format);
430 strbuf_vaddf(&buf, format, arg);
431 va_end(arg);
433 columns = fputs(buf.buf, stream);
434 if (0 <= columns) /* keep the error from the I/O */
435 columns = utf8_strwidth(buf.buf);
436 strbuf_release(&buf);
437 return columns;
441 * Given a buffer and its encoding, return it re-encoded
442 * with iconv. If the conversion fails, returns NULL.
444 #ifndef NO_ICONV
445 #if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6))
446 typedef const char * iconv_ibp;
447 #else
448 typedef char * iconv_ibp;
449 #endif
450 char *reencode_string_iconv(const char *in, size_t insz, iconv_t conv, int *outsz_p)
452 size_t outsz, outalloc;
453 char *out, *outpos;
454 iconv_ibp cp;
456 outsz = insz;
457 outalloc = outsz + 1; /* for terminating NUL */
458 out = xmalloc(outalloc);
459 outpos = out;
460 cp = (iconv_ibp)in;
462 while (1) {
463 size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz);
465 if (cnt == (size_t) -1) {
466 size_t sofar;
467 if (errno != E2BIG) {
468 free(out);
469 return NULL;
471 /* insz has remaining number of bytes.
472 * since we started outsz the same as insz,
473 * it is likely that insz is not enough for
474 * converting the rest.
476 sofar = outpos - out;
477 outalloc = sofar + insz * 2 + 32;
478 out = xrealloc(out, outalloc);
479 outpos = out + sofar;
480 outsz = outalloc - sofar - 1;
482 else {
483 *outpos = '\0';
484 if (outsz_p)
485 *outsz_p = outpos - out;
486 break;
489 return out;
492 char *reencode_string_len(const char *in, int insz,
493 const char *out_encoding, const char *in_encoding,
494 int *outsz)
496 iconv_t conv;
497 char *out;
499 if (!in_encoding)
500 return NULL;
502 conv = iconv_open(out_encoding, in_encoding);
503 if (conv == (iconv_t) -1) {
505 * Some platforms do not have the variously spelled variants of
506 * UTF-8, so let's fall back to trying the most official
507 * spelling. We do so only as a fallback in case the platform
508 * does understand the user's spelling, but not our official
509 * one.
511 if (is_encoding_utf8(in_encoding))
512 in_encoding = "UTF-8";
513 if (is_encoding_utf8(out_encoding))
514 out_encoding = "UTF-8";
515 conv = iconv_open(out_encoding, in_encoding);
516 if (conv == (iconv_t) -1)
517 return NULL;
520 out = reencode_string_iconv(in, insz, conv, outsz);
521 iconv_close(conv);
522 return out;
524 #endif
527 * Returns first character length in bytes for multi-byte `text` according to
528 * `encoding`.
530 * - The `text` pointer is updated to point at the next character.
531 * - When `remainder_p` is not NULL, on entry `*remainder_p` is how much bytes
532 * we can consume from text, and on exit `*remainder_p` is reduced by returned
533 * character length. Otherwise `text` is treated as limited by NUL.
535 int mbs_chrlen(const char **text, size_t *remainder_p, const char *encoding)
537 int chrlen;
538 const char *p = *text;
539 size_t r = (remainder_p ? *remainder_p : SIZE_MAX);
541 if (r < 1)
542 return 0;
544 if (is_encoding_utf8(encoding)) {
545 pick_one_utf8_char(&p, &r);
547 chrlen = p ? (p - *text)
548 : 1 /* not valid UTF-8 -> raw byte sequence */;
550 else {
552 * TODO use iconv to decode one char and obtain its chrlen
553 * for now, let's treat encodings != UTF-8 as one-byte
555 chrlen = 1;
558 *text += chrlen;
559 if (remainder_p)
560 *remainder_p -= chrlen;
562 return chrlen;
566 * Pick the next char from the stream, ignoring codepoints an HFS+ would.
567 * Note that this is _not_ complete by any means. It's just enough
568 * to make is_hfs_dotgit() work, and should not be used otherwise.
570 static ucs_char_t next_hfs_char(const char **in)
572 while (1) {
573 ucs_char_t out = pick_one_utf8_char(in, NULL);
575 * check for malformed utf8. Technically this
576 * gets converted to a percent-sequence, but
577 * returning 0 is good enough for is_hfs_dotgit
578 * to realize it cannot be .git
580 if (!*in)
581 return 0;
583 /* these code points are ignored completely */
584 switch (out) {
585 case 0x200c: /* ZERO WIDTH NON-JOINER */
586 case 0x200d: /* ZERO WIDTH JOINER */
587 case 0x200e: /* LEFT-TO-RIGHT MARK */
588 case 0x200f: /* RIGHT-TO-LEFT MARK */
589 case 0x202a: /* LEFT-TO-RIGHT EMBEDDING */
590 case 0x202b: /* RIGHT-TO-LEFT EMBEDDING */
591 case 0x202c: /* POP DIRECTIONAL FORMATTING */
592 case 0x202d: /* LEFT-TO-RIGHT OVERRIDE */
593 case 0x202e: /* RIGHT-TO-LEFT OVERRIDE */
594 case 0x206a: /* INHIBIT SYMMETRIC SWAPPING */
595 case 0x206b: /* ACTIVATE SYMMETRIC SWAPPING */
596 case 0x206c: /* INHIBIT ARABIC FORM SHAPING */
597 case 0x206d: /* ACTIVATE ARABIC FORM SHAPING */
598 case 0x206e: /* NATIONAL DIGIT SHAPES */
599 case 0x206f: /* NOMINAL DIGIT SHAPES */
600 case 0xfeff: /* ZERO WIDTH NO-BREAK SPACE */
601 continue;
604 return out;
608 int is_hfs_dotgit(const char *path)
610 ucs_char_t c;
612 c = next_hfs_char(&path);
613 if (c != '.')
614 return 0;
615 c = next_hfs_char(&path);
618 * there's a great deal of other case-folding that occurs
619 * in HFS+, but this is enough to catch anything that will
620 * convert to ".git"
622 if (c != 'g' && c != 'G')
623 return 0;
624 c = next_hfs_char(&path);
625 if (c != 'i' && c != 'I')
626 return 0;
627 c = next_hfs_char(&path);
628 if (c != 't' && c != 'T')
629 return 0;
630 c = next_hfs_char(&path);
631 if (c && !is_dir_sep(c))
632 return 0;
634 return 1;