]>
Commit | Line | Data |
---|---|---|
9e832665 | 1 | #include "git-compat-util.h" |
a94410c8 | 2 | #include "strbuf.h" |
9e832665 JS |
3 | #include "utf8.h" |
4 | ||
5 | /* This code is originally from http://www.cl.cam.ac.uk/~mgk25/ucs/ */ | |
6 | ||
7 | struct interval { | |
a68a67de JK |
8 | ucs_char_t first; |
9 | ucs_char_t last; | |
9e832665 JS |
10 | }; |
11 | ||
1640632b | 12 | size_t display_mode_esc_sequence_len(const char *s) |
4247fe79 NTND |
13 | { |
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; | |
24 | } | |
25 | ||
9e832665 | 26 | /* auxiliary function for binary search in interval table */ |
f3fa1838 JH |
27 | static int bisearch(ucs_char_t ucs, const struct interval *table, int max) |
28 | { | |
9e832665 JS |
29 | int min = 0; |
30 | int mid; | |
31 | ||
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; | |
42 | } | |
43 | ||
44 | return 0; | |
45 | } | |
46 | ||
47 | /* The following two functions define the column width of an ISO 10646 | |
48 | * character as follows: | |
49 | * | |
50 | * - The null character (U+0000) has a column width of 0. | |
51 | * | |
52 | * - Other C0/C1 control characters and DEL will lead to a return | |
53 | * value of -1. | |
54 | * | |
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. | |
58 | * | |
59 | * - SOFT HYPHEN (U+00AD) has a column width of 1. | |
60 | * | |
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. | |
63 | * | |
64 | * - Hangul Jamo medial vowels and final consonants (U+1160-U+11FF) | |
65 | * have a column width of 0. | |
66 | * | |
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. | |
70 | * | |
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. | |
74 | * | |
28321145 | 75 | * This implementation assumes that ucs_char_t characters are encoded |
9e832665 JS |
76 | * in ISO 10646. |
77 | */ | |
78 | ||
b51be13c | 79 | static int git_wcwidth(ucs_char_t ch) |
9e832665 JS |
80 | { |
81 | /* | |
82 | * Sorted list of non-overlapping intervals of non-spacing characters, | |
9e832665 | 83 | */ |
9c94389c | 84 | #include "unicode_width.h" |
9e832665 JS |
85 | |
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; | |
91 | ||
92 | /* binary search in table of non-spacing characters */ | |
9c94389c | 93 | if (bisearch(ch, zero_width, sizeof(zero_width) |
9e832665 JS |
94 | / sizeof(struct interval) - 1)) |
95 | return 0; | |
96 | ||
08460345 TB |
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; | |
9e832665 | 101 | |
08460345 | 102 | return 1; |
9e832665 JS |
103 | } |
104 | ||
105 | /* | |
396ccf1f JH |
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 | |
44b25b87 JH |
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. | |
396ccf1f JH |
113 | * |
114 | * If the string was not a valid UTF-8, *start pointer is set to NULL | |
115 | * and the return value is undefined. | |
9e832665 | 116 | */ |
5e133b8c | 117 | static ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p) |
9e832665 JS |
118 | { |
119 | unsigned char *s = (unsigned char *)*start; | |
28321145 | 120 | ucs_char_t ch; |
44b25b87 | 121 | size_t remainder, incr; |
9e832665 | 122 | |
44b25b87 JH |
123 | /* |
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. | |
127 | */ | |
128 | remainder = (remainder_p ? *remainder_p : 999); | |
129 | ||
130 | if (remainder < 1) { | |
131 | goto invalid; | |
132 | } else if (*s < 0x80) { | |
9e832665 JS |
133 | /* 0xxxxxxx */ |
134 | ch = *s; | |
44b25b87 | 135 | incr = 1; |
9e832665 JS |
136 | } else if ((s[0] & 0xe0) == 0xc0) { |
137 | /* 110XXXXx 10xxxxxx */ | |
44b25b87 JH |
138 | if (remainder < 2 || |
139 | (s[1] & 0xc0) != 0x80 || | |
140 | (s[0] & 0xfe) == 0xc0) | |
9e832665 JS |
141 | goto invalid; |
142 | ch = ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); | |
44b25b87 | 143 | incr = 2; |
9e832665 JS |
144 | } else if ((s[0] & 0xf0) == 0xe0) { |
145 | /* 1110XXXX 10Xxxxxx 10xxxxxx */ | |
44b25b87 JH |
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)) | |
9e832665 JS |
156 | goto invalid; |
157 | ch = ((s[0] & 0x0f) << 12) | | |
158 | ((s[1] & 0x3f) << 6) | (s[2] & 0x3f); | |
44b25b87 | 159 | incr = 3; |
9e832665 JS |
160 | } else if ((s[0] & 0xf8) == 0xf0) { |
161 | /* 11110XXX 10XXxxxx 10xxxxxx 10xxxxxx */ | |
44b25b87 JH |
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) | |
9e832665 JS |
170 | goto invalid; |
171 | ch = ((s[0] & 0x07) << 18) | ((s[1] & 0x3f) << 12) | | |
172 | ((s[2] & 0x3f) << 6) | (s[3] & 0x3f); | |
44b25b87 | 173 | incr = 4; |
9e832665 JS |
174 | } else { |
175 | invalid: | |
176 | *start = NULL; | |
177 | return 0; | |
178 | } | |
179 | ||
44b25b87 JH |
180 | *start += incr; |
181 | if (remainder_p) | |
182 | *remainder_p = remainder - incr; | |
396ccf1f JH |
183 | return ch; |
184 | } | |
185 | ||
186 | /* | |
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 | |
44b25b87 JH |
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). | |
396ccf1f | 192 | */ |
44b25b87 | 193 | int utf8_width(const char **start, size_t *remainder_p) |
396ccf1f | 194 | { |
44b25b87 | 195 | ucs_char_t ch = pick_one_utf8_char(start, remainder_p); |
396ccf1f JH |
196 | if (!*start) |
197 | return 0; | |
b51be13c | 198 | return git_wcwidth(ch); |
9e832665 JS |
199 | } |
200 | ||
8a9391e9 GT |
201 | /* |
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. | |
205 | */ | |
2bc1e7ec | 206 | int utf8_strnwidth(const char *string, int len, int skip_ansi) |
8a9391e9 GT |
207 | { |
208 | int width = 0; | |
209 | const char *orig = string; | |
210 | ||
2bc1e7ec NTND |
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; | |
8a9391e9 GT |
218 | width += utf8_width(&string, NULL); |
219 | } | |
2bc1e7ec NTND |
220 | return string ? width : len; |
221 | } | |
222 | ||
223 | int utf8_strwidth(const char *string) | |
224 | { | |
225 | return utf8_strnwidth(string, -1, 0); | |
8a9391e9 GT |
226 | } |
227 | ||
9e832665 JS |
228 | int is_utf8(const char *text) |
229 | { | |
230 | while (*text) { | |
231 | if (*text == '\n' || *text == '\t' || *text == '\r') { | |
232 | text++; | |
233 | continue; | |
234 | } | |
44b25b87 | 235 | utf8_width(&text, NULL); |
9e832665 JS |
236 | if (!text) |
237 | return 0; | |
238 | } | |
239 | return 1; | |
240 | } | |
241 | ||
3c0ff44a | 242 | static void strbuf_addchars(struct strbuf *sb, int c, size_t n) |
a94410c8 | 243 | { |
3c0ff44a RS |
244 | strbuf_grow(sb, n); |
245 | memset(sb->buf + sb->len, c, n); | |
246 | strbuf_setlen(sb, sb->len + n); | |
9e832665 JS |
247 | } |
248 | ||
37bb5d74 RS |
249 | static void strbuf_add_indented_text(struct strbuf *buf, const char *text, |
250 | int indent, int indent2) | |
251 | { | |
252 | if (indent < 0) | |
253 | indent = 0; | |
254 | while (*text) { | |
255 | const char *eol = strchrnul(text, '\n'); | |
256 | if (*eol == '\n') | |
257 | eol++; | |
3c0ff44a | 258 | strbuf_addchars(buf, ' ', indent); |
68ad5e1e | 259 | strbuf_add(buf, text, eol - text); |
37bb5d74 RS |
260 | text = eol; |
261 | indent = indent2; | |
262 | } | |
263 | } | |
264 | ||
9e832665 JS |
265 | /* |
266 | * Wrap the text, if necessary. The variable indent is the indent for the | |
267 | * first line, indent2 is the indent for all other lines. | |
094e03b0 JS |
268 | * If indent is negative, assume that already -indent columns have been |
269 | * consumed (and no extra indent is necessary for the first line). | |
9e832665 | 270 | */ |
e0db1765 | 271 | void strbuf_add_wrapped_text(struct strbuf *buf, |
462749b7 | 272 | const char *text, int indent1, int indent2, int width) |
9e832665 | 273 | { |
462749b7 RS |
274 | int indent, w, assume_utf8 = 1; |
275 | const char *bol, *space, *start = text; | |
276 | size_t orig_len = buf->len; | |
9e832665 | 277 | |
00d39473 | 278 | if (width <= 0) { |
462749b7 | 279 | strbuf_add_indented_text(buf, text, indent1, indent2); |
e0db1765 | 280 | return; |
00d39473 JH |
281 | } |
282 | ||
462749b7 RS |
283 | retry: |
284 | bol = text; | |
285 | w = indent = indent1; | |
286 | space = NULL; | |
094e03b0 JS |
287 | if (indent < 0) { |
288 | w = -indent; | |
289 | space = text; | |
290 | } | |
291 | ||
9e832665 | 292 | for (;;) { |
8a3c63e0 RS |
293 | char c; |
294 | size_t skip; | |
295 | ||
296 | while ((skip = display_mode_esc_sequence_len(text))) | |
297 | text += skip; | |
298 | ||
299 | c = *text; | |
9e832665 | 300 | if (!c || isspace(c)) { |
14e1a4e1 | 301 | if (w <= width || !space) { |
9e832665 | 302 | const char *start = bol; |
ae0b2702 | 303 | if (!c && text == start) |
e0db1765 | 304 | return; |
9e832665 JS |
305 | if (space) |
306 | start = space; | |
307 | else | |
3c0ff44a | 308 | strbuf_addchars(buf, ' ', indent); |
68ad5e1e | 309 | strbuf_add(buf, start, text - start); |
094e03b0 | 310 | if (!c) |
e0db1765 | 311 | return; |
9e832665 | 312 | space = text; |
ae0b2702 JS |
313 | if (c == '\t') |
314 | w |= 0x07; | |
315 | else if (c == '\n') { | |
316 | space++; | |
317 | if (*space == '\n') { | |
68ad5e1e | 318 | strbuf_addch(buf, '\n'); |
ae0b2702 JS |
319 | goto new_line; |
320 | } | |
321 | else if (!isalnum(*space)) | |
322 | goto new_line; | |
323 | else | |
68ad5e1e | 324 | strbuf_addch(buf, ' '); |
ae0b2702 | 325 | } |
9e832665 JS |
326 | w++; |
327 | text++; | |
328 | } | |
329 | else { | |
ae0b2702 | 330 | new_line: |
68ad5e1e | 331 | strbuf_addch(buf, '\n'); |
62273826 | 332 | text = bol = space + isspace(*space); |
9e832665 JS |
333 | space = NULL; |
334 | w = indent = indent2; | |
335 | } | |
336 | continue; | |
337 | } | |
462749b7 | 338 | if (assume_utf8) { |
44b25b87 | 339 | w += utf8_width(&text, NULL); |
462749b7 RS |
340 | if (!text) { |
341 | assume_utf8 = 0; | |
342 | text = start; | |
343 | strbuf_setlen(buf, orig_len); | |
344 | goto retry; | |
345 | } | |
346 | } else { | |
9e832665 JS |
347 | w++; |
348 | text++; | |
349 | } | |
350 | } | |
351 | } | |
b45974a6 | 352 | |
e0db1765 | 353 | void strbuf_add_wrapped_bytes(struct strbuf *buf, const char *data, int len, |
98acc837 JK |
354 | int indent, int indent2, int width) |
355 | { | |
356 | char *tmp = xstrndup(data, len); | |
e0db1765 | 357 | strbuf_add_wrapped_text(buf, tmp, indent, indent2, width); |
98acc837 | 358 | free(tmp); |
98acc837 JK |
359 | } |
360 | ||
a7f01c6b NTND |
361 | void strbuf_utf8_replace(struct strbuf *sb_src, int pos, int width, |
362 | const char *subst) | |
363 | { | |
364 | struct strbuf sb_dst = STRBUF_INIT; | |
365 | char *src = sb_src->buf; | |
366 | char *end = src + sb_src->len; | |
367 | char *dst; | |
368 | int w = 0, subst_len = 0; | |
369 | ||
370 | if (subst) | |
371 | subst_len = strlen(subst); | |
372 | strbuf_grow(&sb_dst, sb_src->len + subst_len); | |
373 | dst = sb_dst.buf; | |
374 | ||
375 | while (src < end) { | |
376 | char *old; | |
377 | size_t n; | |
378 | ||
379 | while ((n = display_mode_esc_sequence_len(src))) { | |
380 | memcpy(dst, src, n); | |
381 | src += n; | |
382 | dst += n; | |
383 | } | |
384 | ||
385 | old = src; | |
386 | n = utf8_width((const char**)&src, NULL); | |
387 | if (!src) /* broken utf-8, do nothing */ | |
388 | return; | |
389 | if (n && w >= pos && w < pos + width) { | |
390 | if (subst) { | |
391 | memcpy(dst, subst, subst_len); | |
392 | dst += subst_len; | |
393 | subst = NULL; | |
394 | } | |
395 | w += n; | |
396 | continue; | |
397 | } | |
398 | memcpy(dst, old, src - old); | |
399 | dst += src - old; | |
400 | w += n; | |
401 | } | |
402 | strbuf_setlen(&sb_dst, dst - sb_dst.buf); | |
980419b9 RJ |
403 | strbuf_swap(sb_src, &sb_dst); |
404 | strbuf_release(&sb_dst); | |
a7f01c6b NTND |
405 | } |
406 | ||
677cfed5 JH |
407 | int is_encoding_utf8(const char *name) |
408 | { | |
409 | if (!name) | |
410 | return 1; | |
411 | if (!strcasecmp(name, "utf-8") || !strcasecmp(name, "utf8")) | |
412 | return 1; | |
413 | return 0; | |
414 | } | |
415 | ||
0e18bcd5 JH |
416 | int same_encoding(const char *src, const char *dst) |
417 | { | |
418 | if (is_encoding_utf8(src) && is_encoding_utf8(dst)) | |
419 | return 1; | |
420 | return !strcasecmp(src, dst); | |
421 | } | |
422 | ||
c0821965 JX |
423 | /* |
424 | * Wrapper for fprintf and returns the total number of columns required | |
425 | * for the printed string, assuming that the string is utf8. | |
426 | */ | |
427 | int utf8_fprintf(FILE *stream, const char *format, ...) | |
428 | { | |
429 | struct strbuf buf = STRBUF_INIT; | |
430 | va_list arg; | |
431 | int columns; | |
432 | ||
433 | va_start(arg, format); | |
434 | strbuf_vaddf(&buf, format, arg); | |
435 | va_end(arg); | |
436 | ||
437 | columns = fputs(buf.buf, stream); | |
438 | if (0 <= columns) /* keep the error from the I/O */ | |
439 | columns = utf8_strwidth(buf.buf); | |
440 | strbuf_release(&buf); | |
441 | return columns; | |
442 | } | |
443 | ||
b45974a6 JH |
444 | /* |
445 | * Given a buffer and its encoding, return it re-encoded | |
446 | * with iconv. If the conversion fails, returns NULL. | |
447 | */ | |
448 | #ifndef NO_ICONV | |
309dbc82 | 449 | #if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6)) |
fd547a97 RJ |
450 | typedef const char * iconv_ibp; |
451 | #else | |
452 | typedef char * iconv_ibp; | |
453 | #endif | |
b782bbab | 454 | char *reencode_string_iconv(const char *in, size_t insz, iconv_t conv, int *outsz_p) |
b45974a6 | 455 | { |
76759c7d | 456 | size_t outsz, outalloc; |
fd547a97 RJ |
457 | char *out, *outpos; |
458 | iconv_ibp cp; | |
b45974a6 | 459 | |
b45974a6 JH |
460 | outsz = insz; |
461 | outalloc = outsz + 1; /* for terminating NUL */ | |
462 | out = xmalloc(outalloc); | |
463 | outpos = out; | |
fd547a97 | 464 | cp = (iconv_ibp)in; |
b45974a6 JH |
465 | |
466 | while (1) { | |
467 | size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz); | |
468 | ||
df5213b7 | 469 | if (cnt == (size_t) -1) { |
b45974a6 JH |
470 | size_t sofar; |
471 | if (errno != E2BIG) { | |
472 | free(out); | |
b45974a6 JH |
473 | return NULL; |
474 | } | |
475 | /* insz has remaining number of bytes. | |
476 | * since we started outsz the same as insz, | |
477 | * it is likely that insz is not enough for | |
478 | * converting the rest. | |
479 | */ | |
480 | sofar = outpos - out; | |
481 | outalloc = sofar + insz * 2 + 32; | |
482 | out = xrealloc(out, outalloc); | |
483 | outpos = out + sofar; | |
484 | outsz = outalloc - sofar - 1; | |
485 | } | |
486 | else { | |
487 | *outpos = '\0'; | |
b782bbab NTND |
488 | if (outsz_p) |
489 | *outsz_p = outpos - out; | |
b45974a6 JH |
490 | break; |
491 | } | |
492 | } | |
76759c7d TB |
493 | return out; |
494 | } | |
495 | ||
b782bbab NTND |
496 | char *reencode_string_len(const char *in, int insz, |
497 | const char *out_encoding, const char *in_encoding, | |
498 | int *outsz) | |
76759c7d TB |
499 | { |
500 | iconv_t conv; | |
501 | char *out; | |
502 | ||
503 | if (!in_encoding) | |
504 | return NULL; | |
5c680be1 | 505 | |
76759c7d | 506 | conv = iconv_open(out_encoding, in_encoding); |
5c680be1 JK |
507 | if (conv == (iconv_t) -1) { |
508 | /* | |
509 | * Some platforms do not have the variously spelled variants of | |
510 | * UTF-8, so let's fall back to trying the most official | |
511 | * spelling. We do so only as a fallback in case the platform | |
512 | * does understand the user's spelling, but not our official | |
513 | * one. | |
514 | */ | |
515 | if (is_encoding_utf8(in_encoding)) | |
516 | in_encoding = "UTF-8"; | |
517 | if (is_encoding_utf8(out_encoding)) | |
518 | out_encoding = "UTF-8"; | |
519 | conv = iconv_open(out_encoding, in_encoding); | |
520 | if (conv == (iconv_t) -1) | |
521 | return NULL; | |
522 | } | |
523 | ||
b782bbab | 524 | out = reencode_string_iconv(in, insz, conv, outsz); |
b45974a6 JH |
525 | iconv_close(conv); |
526 | return out; | |
527 | } | |
528 | #endif | |
6cd3c053 KS |
529 | |
530 | /* | |
531 | * Returns first character length in bytes for multi-byte `text` according to | |
532 | * `encoding`. | |
533 | * | |
534 | * - The `text` pointer is updated to point at the next character. | |
535 | * - When `remainder_p` is not NULL, on entry `*remainder_p` is how much bytes | |
536 | * we can consume from text, and on exit `*remainder_p` is reduced by returned | |
537 | * character length. Otherwise `text` is treated as limited by NUL. | |
538 | */ | |
539 | int mbs_chrlen(const char **text, size_t *remainder_p, const char *encoding) | |
540 | { | |
541 | int chrlen; | |
542 | const char *p = *text; | |
543 | size_t r = (remainder_p ? *remainder_p : SIZE_MAX); | |
544 | ||
545 | if (r < 1) | |
546 | return 0; | |
547 | ||
548 | if (is_encoding_utf8(encoding)) { | |
549 | pick_one_utf8_char(&p, &r); | |
550 | ||
551 | chrlen = p ? (p - *text) | |
552 | : 1 /* not valid UTF-8 -> raw byte sequence */; | |
553 | } | |
554 | else { | |
555 | /* | |
556 | * TODO use iconv to decode one char and obtain its chrlen | |
557 | * for now, let's treat encodings != UTF-8 as one-byte | |
558 | */ | |
559 | chrlen = 1; | |
560 | } | |
561 | ||
562 | *text += chrlen; | |
563 | if (remainder_p) | |
564 | *remainder_p -= chrlen; | |
565 | ||
566 | return chrlen; | |
567 | } |