1 #include "git-compat-util.h"
5 /* This code is originally from http://www.cl.cam.ac.uk/~mgk25/ucs/ */
12 size_t display_mode_esc_sequence_len(const char *s)
19 while (isdigit(*p) || *p == ';')
26 /* auxiliary function for binary search in interval table */
27 static int bisearch(ucs_char_t ucs, const struct interval *table, int max)
32 if (ucs < table[0].first || ucs > table[max].last)
35 mid = (min + max) / 2;
36 if (ucs > table[mid].last)
38 else if (ucs < table[mid].first)
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
55 * - Non-spacing and enclosing combining characters (general
56 * category code Mn or Me in the Unicode database) have a
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
79 static int git_wcwidth(ucs_char_t ch)
82 * Sorted list of non-overlapping intervals of non-spacing characters,
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 }
129 static const struct interval double_width[] = {
139 { 0x20000, 0x2FFFD },
143 /* test for 8-bit control characters */
146 if (ch < 32 || (ch >= 0x7f && ch < 0xa0))
149 /* binary search in table of non-spacing characters */
150 if (bisearch(ch, combining, sizeof(combining)
151 / sizeof(struct interval) - 1))
154 /* binary search in table of double width characters */
155 if (bisearch(ch, double_width, sizeof(double_width)
156 / sizeof(struct interval) - 1))
163 * Pick one ucs character starting from the location *start points at,
164 * and return it, while updating the *start pointer to point at the
165 * end of that character. When remainder_p is not NULL, the location
166 * holds the number of bytes remaining in the string that we are allowed
167 * to pick from. Otherwise we are allowed to pick up to the NUL that
168 * would eventually appear in the string. *remainder_p is also reduced
169 * by the number of bytes we have consumed.
171 * If the string was not a valid UTF-8, *start pointer is set to NULL
172 * and the return value is undefined.
174 static ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p)
176 unsigned char *s = (unsigned char *)*start;
178 size_t remainder, incr;
181 * A caller that assumes NUL terminated text can choose
182 * not to bother with the remainder length. We will
183 * stop at the first NUL.
185 remainder = (remainder_p ? *remainder_p : 999);
189 } else if (*s < 0x80) {
193 } else if ((s[0] & 0xe0) == 0xc0) {
194 /* 110XXXXx 10xxxxxx */
196 (s[1] & 0xc0) != 0x80 ||
197 (s[0] & 0xfe) == 0xc0)
199 ch = ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
201 } else if ((s[0] & 0xf0) == 0xe0) {
202 /* 1110XXXX 10Xxxxxx 10xxxxxx */
204 (s[1] & 0xc0) != 0x80 ||
205 (s[2] & 0xc0) != 0x80 ||
207 (s[0] == 0xe0 && (s[1] & 0xe0) == 0x80) ||
209 (s[0] == 0xed && (s[1] & 0xe0) == 0xa0) ||
210 /* U+FFFE or U+FFFF? */
211 (s[0] == 0xef && s[1] == 0xbf &&
212 (s[2] & 0xfe) == 0xbe))
214 ch = ((s[0] & 0x0f) << 12) |
215 ((s[1] & 0x3f) << 6) | (s[2] & 0x3f);
217 } else if ((s[0] & 0xf8) == 0xf0) {
218 /* 11110XXX 10XXxxxx 10xxxxxx 10xxxxxx */
220 (s[1] & 0xc0) != 0x80 ||
221 (s[2] & 0xc0) != 0x80 ||
222 (s[3] & 0xc0) != 0x80 ||
224 (s[0] == 0xf0 && (s[1] & 0xf0) == 0x80) ||
226 (s[0] == 0xf4 && s[1] > 0x8f) || s[0] > 0xf4)
228 ch = ((s[0] & 0x07) << 18) | ((s[1] & 0x3f) << 12) |
229 ((s[2] & 0x3f) << 6) | (s[3] & 0x3f);
239 *remainder_p = remainder - incr;
244 * This function returns the number of columns occupied by the character
245 * pointed to by the variable start. The pointer is updated to point at
246 * the next character. When remainder_p is not NULL, it points at the
247 * location that stores the number of remaining bytes we can use to pick
248 * a character (see pick_one_utf8_char() above).
250 int utf8_width(const char **start, size_t *remainder_p)
252 ucs_char_t ch = pick_one_utf8_char(start, remainder_p);
255 return git_wcwidth(ch);
259 * Returns the total number of columns required by a null-terminated
260 * string, assuming that the string is utf8. Returns strlen() instead
261 * if the string does not look like a valid utf8 string.
263 int utf8_strnwidth(const char *string, int len, int skip_ansi)
266 const char *orig = string;
269 len = strlen(string);
270 while (string && string < orig + len) {
273 (skip = display_mode_esc_sequence_len(string)) != 0)
275 width += utf8_width(&string, NULL);
277 return string ? width : len;
280 int utf8_strwidth(const char *string)
282 return utf8_strnwidth(string, -1, 0);
285 int is_utf8(const char *text)
288 if (*text == '\n' || *text == '\t' || *text == '\r') {
292 utf8_width(&text, NULL);
299 static void strbuf_addchars(struct strbuf *sb, int c, size_t n)
302 memset(sb->buf + sb->len, c, n);
303 strbuf_setlen(sb, sb->len + n);
306 static void strbuf_add_indented_text(struct strbuf *buf, const char *text,
307 int indent, int indent2)
312 const char *eol = strchrnul(text, '\n');
315 strbuf_addchars(buf, ' ', indent);
316 strbuf_add(buf, text, eol - text);
323 * Wrap the text, if necessary. The variable indent is the indent for the
324 * first line, indent2 is the indent for all other lines.
325 * If indent is negative, assume that already -indent columns have been
326 * consumed (and no extra indent is necessary for the first line).
328 void strbuf_add_wrapped_text(struct strbuf *buf,
329 const char *text, int indent1, int indent2, int width)
331 int indent, w, assume_utf8 = 1;
332 const char *bol, *space, *start = text;
333 size_t orig_len = buf->len;
336 strbuf_add_indented_text(buf, text, indent1, indent2);
342 w = indent = indent1;
353 while ((skip = display_mode_esc_sequence_len(text)))
357 if (!c || isspace(c)) {
358 if (w <= width || !space) {
359 const char *start = bol;
360 if (!c && text == start)
365 strbuf_addchars(buf, ' ', indent);
366 strbuf_add(buf, start, text - start);
372 else if (c == '\n') {
374 if (*space == '\n') {
375 strbuf_addch(buf, '\n');
378 else if (!isalnum(*space))
381 strbuf_addch(buf, ' ');
388 strbuf_addch(buf, '\n');
389 text = bol = space + isspace(*space);
391 w = indent = indent2;
396 w += utf8_width(&text, NULL);
400 strbuf_setlen(buf, orig_len);
410 void strbuf_add_wrapped_bytes(struct strbuf *buf, const char *data, int len,
411 int indent, int indent2, int width)
413 char *tmp = xstrndup(data, len);
414 strbuf_add_wrapped_text(buf, tmp, indent, indent2, width);
418 void strbuf_utf8_replace(struct strbuf *sb_src, int pos, int width,
421 struct strbuf sb_dst = STRBUF_INIT;
422 char *src = sb_src->buf;
423 char *end = src + sb_src->len;
425 int w = 0, subst_len = 0;
428 subst_len = strlen(subst);
429 strbuf_grow(&sb_dst, sb_src->len + subst_len);
436 while ((n = display_mode_esc_sequence_len(src))) {
443 n = utf8_width((const char**)&src, NULL);
444 if (!src) /* broken utf-8, do nothing */
446 if (n && w >= pos && w < pos + width) {
448 memcpy(dst, subst, subst_len);
455 memcpy(dst, old, src - old);
459 strbuf_setlen(&sb_dst, dst - sb_dst.buf);
460 strbuf_swap(sb_src, &sb_dst);
461 strbuf_release(&sb_dst);
464 int is_encoding_utf8(const char *name)
468 if (!strcasecmp(name, "utf-8") || !strcasecmp(name, "utf8"))
473 int same_encoding(const char *src, const char *dst)
475 if (is_encoding_utf8(src) && is_encoding_utf8(dst))
477 return !strcasecmp(src, dst);
481 * Wrapper for fprintf and returns the total number of columns required
482 * for the printed string, assuming that the string is utf8.
484 int utf8_fprintf(FILE *stream, const char *format, ...)
486 struct strbuf buf = STRBUF_INIT;
490 va_start(arg, format);
491 strbuf_vaddf(&buf, format, arg);
494 columns = fputs(buf.buf, stream);
495 if (0 <= columns) /* keep the error from the I/O */
496 columns = utf8_strwidth(buf.buf);
497 strbuf_release(&buf);
502 * Given a buffer and its encoding, return it re-encoded
503 * with iconv. If the conversion fails, returns NULL.
506 #if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6))
507 typedef const char * iconv_ibp;
509 typedef char * iconv_ibp;
511 char *reencode_string_iconv(const char *in, size_t insz, iconv_t conv, int *outsz_p)
513 size_t outsz, outalloc;
518 outalloc = outsz + 1; /* for terminating NUL */
519 out = xmalloc(outalloc);
524 size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz);
528 if (errno != E2BIG) {
532 /* insz has remaining number of bytes.
533 * since we started outsz the same as insz,
534 * it is likely that insz is not enough for
535 * converting the rest.
537 sofar = outpos - out;
538 outalloc = sofar + insz * 2 + 32;
539 out = xrealloc(out, outalloc);
540 outpos = out + sofar;
541 outsz = outalloc - sofar - 1;
546 *outsz_p = outpos - out;
553 char *reencode_string_len(const char *in, int insz,
554 const char *out_encoding, const char *in_encoding,
563 conv = iconv_open(out_encoding, in_encoding);
564 if (conv == (iconv_t) -1) {
566 * Some platforms do not have the variously spelled variants of
567 * UTF-8, so let's fall back to trying the most official
568 * spelling. We do so only as a fallback in case the platform
569 * does understand the user's spelling, but not our official
572 if (is_encoding_utf8(in_encoding))
573 in_encoding = "UTF-8";
574 if (is_encoding_utf8(out_encoding))
575 out_encoding = "UTF-8";
576 conv = iconv_open(out_encoding, in_encoding);
577 if (conv == (iconv_t) -1)
581 out = reencode_string_iconv(in, insz, conv, outsz);
588 * Returns first character length in bytes for multi-byte `text` according to
591 * - The `text` pointer is updated to point at the next character.
592 * - When `remainder_p` is not NULL, on entry `*remainder_p` is how much bytes
593 * we can consume from text, and on exit `*remainder_p` is reduced by returned
594 * character length. Otherwise `text` is treated as limited by NUL.
596 int mbs_chrlen(const char **text, size_t *remainder_p, const char *encoding)
599 const char *p = *text;
600 size_t r = (remainder_p ? *remainder_p : SIZE_MAX);
605 if (is_encoding_utf8(encoding)) {
606 pick_one_utf8_char(&p, &r);
608 chrlen = p ? (p - *text)
609 : 1 /* not valid UTF-8 -> raw byte sequence */;
613 * TODO use iconv to decode one char and obtain its chrlen
614 * for now, let's treat encodings != UTF-8 as one-byte
621 *remainder_p -= chrlen;