git
/
utf8.c
824 строки · 20.1 Кб
1#include "git-compat-util.h"2#include "strbuf.h"3#include "utf8.h"4
5/* This code is originally from https://www.cl.cam.ac.uk/~mgk25/ucs/ */
6
7static const char utf16_be_bom[] = {'\xFE', '\xFF'};8static const char utf16_le_bom[] = {'\xFF', '\xFE'};9static const char utf32_be_bom[] = {'\0', '\0', '\xFE', '\xFF'};10static const char utf32_le_bom[] = {'\xFF', '\xFE', '\0', '\0'};11
12struct interval {13ucs_char_t first;14ucs_char_t last;15};16
17size_t display_mode_esc_sequence_len(const char *s)18{
19const char *p = s;20if (*p++ != '\033')21return 0;22if (*p++ != '[')23return 0;24while (isdigit(*p) || *p == ';')25p++;26if (*p++ != 'm')27return 0;28return p - s;29}
30
31/* auxiliary function for binary search in interval table */
32static int bisearch(ucs_char_t ucs, const struct interval *table, int max)33{
34int min = 0;35int mid;36
37if (ucs < table[0].first || ucs > table[max].last)38return 0;39while (max >= min) {40mid = min + (max - min) / 2;41if (ucs > table[mid].last)42min = mid + 1;43else if (ucs < table[mid].first)44max = mid - 1;45else46return 1;47}48
49return 0;50}
51
52/* The following two functions define the column width of an ISO 10646
53* character as follows:
54*
55* - The null character (U+0000) has a column width of 0.
56*
57* - Other C0/C1 control characters and DEL will lead to a return
58* value of -1.
59*
60* - Non-spacing and enclosing combining characters (general
61* category code Mn or Me in the Unicode database) have a
62* column width of 0.
63*
64* - SOFT HYPHEN (U+00AD) has a column width of 1.
65*
66* - Other format characters (general category code Cf in the Unicode
67* database) and ZERO WIDTH SPACE (U+200B) have a column width of 0.
68*
69* - Hangul Jamo medial vowels and final consonants (U+1160-U+11FF)
70* have a column width of 0.
71*
72* - Spacing characters in the East Asian Wide (W) or East Asian
73* Full-width (F) category as defined in Unicode Technical
74* Report #11 have a column width of 2.
75*
76* - All remaining characters (including all printable
77* ISO 8859-1 and WGL4 characters, Unicode control characters,
78* etc.) have a column width of 1.
79*
80* This implementation assumes that ucs_char_t characters are encoded
81* in ISO 10646.
82*/
83
84static int git_wcwidth(ucs_char_t ch)85{
86/*87* Sorted list of non-overlapping intervals of non-spacing characters,
88*/
89#include "unicode-width.h"90
91/* test for 8-bit control characters */92if (ch == 0)93return 0;94if (ch < 32 || (ch >= 0x7f && ch < 0xa0))95return -1;96
97/* binary search in table of non-spacing characters */98if (bisearch(ch, zero_width, ARRAY_SIZE(zero_width) - 1))99return 0;100
101/* binary search in table of double width characters */102if (bisearch(ch, double_width, ARRAY_SIZE(double_width) - 1))103return 2;104
105return 1;106}
107
108/*
109* Pick one ucs character starting from the location *start points at,
110* and return it, while updating the *start pointer to point at the
111* end of that character. When remainder_p is not NULL, the location
112* holds the number of bytes remaining in the string that we are allowed
113* to pick from. Otherwise we are allowed to pick up to the NUL that
114* would eventually appear in the string. *remainder_p is also reduced
115* by the number of bytes we have consumed.
116*
117* If the string was not a valid UTF-8, *start pointer is set to NULL
118* and the return value is undefined.
119*/
120static ucs_char_t pick_one_utf8_char(const char **start, size_t *remainder_p)121{
122unsigned char *s = (unsigned char *)*start;123ucs_char_t ch;124size_t remainder, incr;125
126/*127* A caller that assumes NUL terminated text can choose
128* not to bother with the remainder length. We will
129* stop at the first NUL.
130*/
131remainder = (remainder_p ? *remainder_p : 999);132
133if (remainder < 1) {134goto invalid;135} else if (*s < 0x80) {136/* 0xxxxxxx */137ch = *s;138incr = 1;139} else if ((s[0] & 0xe0) == 0xc0) {140/* 110XXXXx 10xxxxxx */141if (remainder < 2 ||142(s[1] & 0xc0) != 0x80 ||143(s[0] & 0xfe) == 0xc0)144goto invalid;145ch = ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);146incr = 2;147} else if ((s[0] & 0xf0) == 0xe0) {148/* 1110XXXX 10Xxxxxx 10xxxxxx */149if (remainder < 3 ||150(s[1] & 0xc0) != 0x80 ||151(s[2] & 0xc0) != 0x80 ||152/* overlong? */153(s[0] == 0xe0 && (s[1] & 0xe0) == 0x80) ||154/* surrogate? */155(s[0] == 0xed && (s[1] & 0xe0) == 0xa0) ||156/* U+FFFE or U+FFFF? */157(s[0] == 0xef && s[1] == 0xbf &&158(s[2] & 0xfe) == 0xbe))159goto invalid;160ch = ((s[0] & 0x0f) << 12) |161((s[1] & 0x3f) << 6) | (s[2] & 0x3f);162incr = 3;163} else if ((s[0] & 0xf8) == 0xf0) {164/* 11110XXX 10XXxxxx 10xxxxxx 10xxxxxx */165if (remainder < 4 ||166(s[1] & 0xc0) != 0x80 ||167(s[2] & 0xc0) != 0x80 ||168(s[3] & 0xc0) != 0x80 ||169/* overlong? */170(s[0] == 0xf0 && (s[1] & 0xf0) == 0x80) ||171/* > U+10FFFF? */172(s[0] == 0xf4 && s[1] > 0x8f) || s[0] > 0xf4)173goto invalid;174ch = ((s[0] & 0x07) << 18) | ((s[1] & 0x3f) << 12) |175((s[2] & 0x3f) << 6) | (s[3] & 0x3f);176incr = 4;177} else {178invalid:179*start = NULL;180return 0;181}182
183*start += incr;184if (remainder_p)185*remainder_p = remainder - incr;186return ch;187}
188
189/*
190* This function returns the number of columns occupied by the character
191* pointed to by the variable start. The pointer is updated to point at
192* the next character. When remainder_p is not NULL, it points at the
193* location that stores the number of remaining bytes we can use to pick
194* a character (see pick_one_utf8_char() above).
195*/
196int utf8_width(const char **start, size_t *remainder_p)197{
198ucs_char_t ch = pick_one_utf8_char(start, remainder_p);199if (!*start)200return 0;201return git_wcwidth(ch);202}
203
204/*
205* Returns the total number of columns required by a null-terminated
206* string, assuming that the string is utf8. Returns strlen() instead
207* if the string does not look like a valid utf8 string.
208*/
209int utf8_strnwidth(const char *string, size_t len, int skip_ansi)210{
211const char *orig = string;212size_t width = 0;213
214while (string && string < orig + len) {215int glyph_width;216size_t skip;217
218while (skip_ansi &&219(skip = display_mode_esc_sequence_len(string)) != 0)220string += skip;221
222glyph_width = utf8_width(&string, NULL);223if (glyph_width > 0)224width += glyph_width;225}226
227/*228* TODO: fix the interface of this function and `utf8_strwidth()` to
229* return `size_t` instead of `int`.
230*/
231return cast_size_t_to_int(string ? width : len);232}
233
234int utf8_strwidth(const char *string)235{
236return utf8_strnwidth(string, strlen(string), 0);237}
238
239int is_utf8(const char *text)240{
241while (*text) {242if (*text == '\n' || *text == '\t' || *text == '\r') {243text++;244continue;245}246utf8_width(&text, NULL);247if (!text)248return 0;249}250return 1;251}
252
253static void strbuf_add_indented_text(struct strbuf *buf, const char *text,254int indent, int indent2)255{
256if (indent < 0)257indent = 0;258while (*text) {259const char *eol = strchrnul(text, '\n');260if (*eol == '\n')261eol++;262strbuf_addchars(buf, ' ', indent);263strbuf_add(buf, text, eol - text);264text = eol;265indent = indent2;266}267}
268
269/*
270* Wrap the text, if necessary. The variable indent is the indent for the
271* first line, indent2 is the indent for all other lines.
272* If indent is negative, assume that already -indent columns have been
273* consumed (and no extra indent is necessary for the first line).
274*/
275void strbuf_add_wrapped_text(struct strbuf *buf,276const char *text, int indent1, int indent2, int width)277{
278int indent, w, assume_utf8 = 1;279const char *bol, *space, *start = text;280size_t orig_len = buf->len;281
282if (width <= 0) {283strbuf_add_indented_text(buf, text, indent1, indent2);284return;285}286
287retry:288bol = text;289w = indent = indent1;290space = NULL;291if (indent < 0) {292w = -indent;293space = text;294}295
296for (;;) {297char c;298size_t skip;299
300while ((skip = display_mode_esc_sequence_len(text)))301text += skip;302
303c = *text;304if (!c || isspace(c)) {305if (w <= width || !space) {306const char *start = bol;307if (!c && text == start)308return;309if (space)310start = space;311else312strbuf_addchars(buf, ' ', indent);313strbuf_add(buf, start, text - start);314if (!c)315return;316space = text;317if (c == '\t')318w |= 0x07;319else if (c == '\n') {320space++;321if (*space == '\n') {322strbuf_addch(buf, '\n');323goto new_line;324}325else if (!isalnum(*space))326goto new_line;327else328strbuf_addch(buf, ' ');329}330w++;331text++;332}333else {334new_line:335strbuf_addch(buf, '\n');336text = bol = space + isspace(*space);337space = NULL;338w = indent = indent2;339}340continue;341}342if (assume_utf8) {343w += utf8_width(&text, NULL);344if (!text) {345assume_utf8 = 0;346text = start;347strbuf_setlen(buf, orig_len);348goto retry;349}350} else {351w++;352text++;353}354}355}
356
357void strbuf_add_wrapped_bytes(struct strbuf *buf, const char *data, int len,358int indent, int indent2, int width)359{
360char *tmp = xstrndup(data, len);361strbuf_add_wrapped_text(buf, tmp, indent, indent2, width);362free(tmp);363}
364
365void strbuf_utf8_replace(struct strbuf *sb_src, int pos, int width,366const char *subst)367{
368const char *src = sb_src->buf, *end = sb_src->buf + sb_src->len;369struct strbuf dst;370int w = 0;371
372strbuf_init(&dst, sb_src->len);373
374while (src < end) {375const char *old;376int glyph_width;377size_t n;378
379while ((n = display_mode_esc_sequence_len(src))) {380strbuf_add(&dst, src, n);381src += n;382}383
384if (src >= end)385break;386
387old = src;388glyph_width = utf8_width((const char**)&src, NULL);389if (!src) /* broken utf-8, do nothing */390goto out;391
392/*393* In case we see a control character we copy it into the
394* buffer, but don't add it to the width.
395*/
396if (glyph_width < 0)397glyph_width = 0;398
399if (glyph_width && w >= pos && w < pos + width) {400if (subst) {401strbuf_addstr(&dst, subst);402subst = NULL;403}404} else {405strbuf_add(&dst, old, src - old);406}407
408w += glyph_width;409}410
411strbuf_swap(sb_src, &dst);412out:413strbuf_release(&dst);414}
415
416/*
417* Returns true (1) if the src encoding name matches the dst encoding
418* name directly or one of its alternative names. E.g. UTF-16BE is the
419* same as UTF16BE.
420*/
421static int same_utf_encoding(const char *src, const char *dst)422{
423if (skip_iprefix(src, "utf", &src) && skip_iprefix(dst, "utf", &dst)) {424skip_prefix(src, "-", &src);425skip_prefix(dst, "-", &dst);426return !strcasecmp(src, dst);427}428return 0;429}
430
431int is_encoding_utf8(const char *name)432{
433if (!name)434return 1;435if (same_utf_encoding("utf-8", name))436return 1;437return 0;438}
439
440int same_encoding(const char *src, const char *dst)441{
442static const char utf8[] = "UTF-8";443
444if (!src)445src = utf8;446if (!dst)447dst = utf8;448if (same_utf_encoding(src, dst))449return 1;450return !strcasecmp(src, dst);451}
452
453/*
454* Wrapper for fprintf and returns the total number of columns required
455* for the printed string, assuming that the string is utf8.
456*/
457int utf8_fprintf(FILE *stream, const char *format, ...)458{
459struct strbuf buf = STRBUF_INIT;460va_list arg;461int columns;462
463va_start(arg, format);464strbuf_vaddf(&buf, format, arg);465va_end(arg);466
467columns = fputs(buf.buf, stream);468if (0 <= columns) /* keep the error from the I/O */469columns = utf8_strwidth(buf.buf);470strbuf_release(&buf);471return columns;472}
473
474/*
475* Given a buffer and its encoding, return it re-encoded
476* with iconv. If the conversion fails, returns NULL.
477*/
478#ifndef NO_ICONV479#if defined(OLD_ICONV) || (defined(__sun__) && !defined(_XPG6))480typedef const char * iconv_ibp;481#else482typedef char * iconv_ibp;483#endif484char *reencode_string_iconv(const char *in, size_t insz, iconv_t conv,485size_t bom_len, size_t *outsz_p)486{
487size_t outsz, outalloc;488char *out, *outpos;489iconv_ibp cp;490
491outsz = insz;492outalloc = st_add(outsz, 1 + bom_len); /* for terminating NUL */493out = xmalloc(outalloc);494outpos = out + bom_len;495cp = (iconv_ibp)in;496
497while (1) {498size_t cnt = iconv(conv, &cp, &insz, &outpos, &outsz);499
500if (cnt == (size_t) -1) {501size_t sofar;502if (errno != E2BIG) {503free(out);504return NULL;505}506/* insz has remaining number of bytes.507* since we started outsz the same as insz,
508* it is likely that insz is not enough for
509* converting the rest.
510*/
511sofar = outpos - out;512outalloc = st_add3(sofar, st_mult(insz, 2), 32);513out = xrealloc(out, outalloc);514outpos = out + sofar;515outsz = outalloc - sofar - 1;516}517else {518*outpos = '\0';519if (outsz_p)520*outsz_p = outpos - out;521break;522}523}524return out;525}
526
527static const char *fallback_encoding(const char *name)528{
529/*530* Some platforms do not have the variously spelled variants of
531* UTF-8, so let's fall back to trying the most official
532* spelling. We do so only as a fallback in case the platform
533* does understand the user's spelling, but not our official
534* one.
535*/
536if (is_encoding_utf8(name))537return "UTF-8";538
539/*540* Even though latin-1 is still seen in e-mail
541* headers, some platforms only install ISO-8859-1.
542*/
543if (!strcasecmp(name, "latin-1"))544return "ISO-8859-1";545
546return name;547}
548
549char *reencode_string_len(const char *in, size_t insz,550const char *out_encoding, const char *in_encoding,551size_t *outsz)552{
553iconv_t conv;554char *out;555const char *bom_str = NULL;556size_t bom_len = 0;557
558if (!in_encoding)559return NULL;560
561/* UTF-16LE-BOM is the same as UTF-16 for reading */562if (same_utf_encoding("UTF-16LE-BOM", in_encoding))563in_encoding = "UTF-16";564
565/*566* For writing, UTF-16 iconv typically creates "UTF-16BE-BOM"
567* Some users under Windows want the little endian version
568*
569* We handle UTF-16 and UTF-32 ourselves only if the platform does not
570* provide a BOM (which we require), since we want to match the behavior
571* of the system tools and libc as much as possible.
572*/
573if (same_utf_encoding("UTF-16LE-BOM", out_encoding)) {574bom_str = utf16_le_bom;575bom_len = sizeof(utf16_le_bom);576out_encoding = "UTF-16LE";577} else if (same_utf_encoding("UTF-16BE-BOM", out_encoding)) {578bom_str = utf16_be_bom;579bom_len = sizeof(utf16_be_bom);580out_encoding = "UTF-16BE";581#ifdef ICONV_OMITS_BOM582} else if (same_utf_encoding("UTF-16", out_encoding)) {583bom_str = utf16_be_bom;584bom_len = sizeof(utf16_be_bom);585out_encoding = "UTF-16BE";586} else if (same_utf_encoding("UTF-32", out_encoding)) {587bom_str = utf32_be_bom;588bom_len = sizeof(utf32_be_bom);589out_encoding = "UTF-32BE";590#endif591}592
593conv = iconv_open(out_encoding, in_encoding);594if (conv == (iconv_t) -1) {595in_encoding = fallback_encoding(in_encoding);596out_encoding = fallback_encoding(out_encoding);597
598conv = iconv_open(out_encoding, in_encoding);599if (conv == (iconv_t) -1)600return NULL;601}602out = reencode_string_iconv(in, insz, conv, bom_len, outsz);603iconv_close(conv);604if (out && bom_str && bom_len)605memcpy(out, bom_str, bom_len);606return out;607}
608#endif609
610static int has_bom_prefix(const char *data, size_t len,611const char *bom, size_t bom_len)612{
613return data && bom && (len >= bom_len) && !memcmp(data, bom, bom_len);614}
615
616int has_prohibited_utf_bom(const char *enc, const char *data, size_t len)617{
618return (619(same_utf_encoding("UTF-16BE", enc) ||620same_utf_encoding("UTF-16LE", enc)) &&621(has_bom_prefix(data, len, utf16_be_bom, sizeof(utf16_be_bom)) ||622has_bom_prefix(data, len, utf16_le_bom, sizeof(utf16_le_bom)))623) || (624(same_utf_encoding("UTF-32BE", enc) ||625same_utf_encoding("UTF-32LE", enc)) &&626(has_bom_prefix(data, len, utf32_be_bom, sizeof(utf32_be_bom)) ||627has_bom_prefix(data, len, utf32_le_bom, sizeof(utf32_le_bom)))628);629}
630
631int is_missing_required_utf_bom(const char *enc, const char *data, size_t len)632{
633return (634(same_utf_encoding(enc, "UTF-16")) &&635!(has_bom_prefix(data, len, utf16_be_bom, sizeof(utf16_be_bom)) ||636has_bom_prefix(data, len, utf16_le_bom, sizeof(utf16_le_bom)))637) || (638(same_utf_encoding(enc, "UTF-32")) &&639!(has_bom_prefix(data, len, utf32_be_bom, sizeof(utf32_be_bom)) ||640has_bom_prefix(data, len, utf32_le_bom, sizeof(utf32_le_bom)))641);642}
643
644/*
645* Returns first character length in bytes for multi-byte `text` according to
646* `encoding`.
647*
648* - The `text` pointer is updated to point at the next character.
649* - When `remainder_p` is not NULL, on entry `*remainder_p` is how much bytes
650* we can consume from text, and on exit `*remainder_p` is reduced by returned
651* character length. Otherwise `text` is treated as limited by NUL.
652*/
653int mbs_chrlen(const char **text, size_t *remainder_p, const char *encoding)654{
655int chrlen;656const char *p = *text;657size_t r = (remainder_p ? *remainder_p : SIZE_MAX);658
659if (r < 1)660return 0;661
662if (is_encoding_utf8(encoding)) {663pick_one_utf8_char(&p, &r);664
665chrlen = p ? (p - *text)666: 1 /* not valid UTF-8 -> raw byte sequence */;667}668else {669/*670* TODO use iconv to decode one char and obtain its chrlen
671* for now, let's treat encodings != UTF-8 as one-byte
672*/
673chrlen = 1;674}675
676*text += chrlen;677if (remainder_p)678*remainder_p -= chrlen;679
680return chrlen;681}
682
683/*
684* Pick the next char from the stream, ignoring codepoints an HFS+ would.
685* Note that this is _not_ complete by any means. It's just enough
686* to make is_hfs_dotgit() work, and should not be used otherwise.
687*/
688static ucs_char_t next_hfs_char(const char **in)689{
690while (1) {691ucs_char_t out = pick_one_utf8_char(in, NULL);692/*693* check for malformed utf8. Technically this
694* gets converted to a percent-sequence, but
695* returning 0 is good enough for is_hfs_dotgit
696* to realize it cannot be .git
697*/
698if (!*in)699return 0;700
701/* these code points are ignored completely */702switch (out) {703case 0x200c: /* ZERO WIDTH NON-JOINER */704case 0x200d: /* ZERO WIDTH JOINER */705case 0x200e: /* LEFT-TO-RIGHT MARK */706case 0x200f: /* RIGHT-TO-LEFT MARK */707case 0x202a: /* LEFT-TO-RIGHT EMBEDDING */708case 0x202b: /* RIGHT-TO-LEFT EMBEDDING */709case 0x202c: /* POP DIRECTIONAL FORMATTING */710case 0x202d: /* LEFT-TO-RIGHT OVERRIDE */711case 0x202e: /* RIGHT-TO-LEFT OVERRIDE */712case 0x206a: /* INHIBIT SYMMETRIC SWAPPING */713case 0x206b: /* ACTIVATE SYMMETRIC SWAPPING */714case 0x206c: /* INHIBIT ARABIC FORM SHAPING */715case 0x206d: /* ACTIVATE ARABIC FORM SHAPING */716case 0x206e: /* NATIONAL DIGIT SHAPES */717case 0x206f: /* NOMINAL DIGIT SHAPES */718case 0xfeff: /* ZERO WIDTH NO-BREAK SPACE */719continue;720}721
722return out;723}724}
725
726static int is_hfs_dot_generic(const char *path,727const char *needle, size_t needle_len)728{
729ucs_char_t c;730
731c = next_hfs_char(&path);732if (c != '.')733return 0;734
735/*736* there's a great deal of other case-folding that occurs
737* in HFS+, but this is enough to catch our fairly vanilla
738* hard-coded needles.
739*/
740for (; needle_len > 0; needle++, needle_len--) {741c = next_hfs_char(&path);742
743/*744* We know our needles contain only ASCII, so we clamp here to
745* make the results of tolower() sane.
746*/
747if (c > 127)748return 0;749if (tolower(c) != *needle)750return 0;751}752
753c = next_hfs_char(&path);754if (c && !is_dir_sep(c))755return 0;756
757return 1;758}
759
760/*
761* Inline wrapper to make sure the compiler resolves strlen() on literals at
762* compile time.
763*/
764static inline int is_hfs_dot_str(const char *path, const char *needle)765{
766return is_hfs_dot_generic(path, needle, strlen(needle));767}
768
769int is_hfs_dotgit(const char *path)770{
771return is_hfs_dot_str(path, "git");772}
773
774int is_hfs_dotgitmodules(const char *path)775{
776return is_hfs_dot_str(path, "gitmodules");777}
778
779int is_hfs_dotgitignore(const char *path)780{
781return is_hfs_dot_str(path, "gitignore");782}
783
784int is_hfs_dotgitattributes(const char *path)785{
786return is_hfs_dot_str(path, "gitattributes");787}
788
789int is_hfs_dotmailmap(const char *path)790{
791return is_hfs_dot_str(path, "mailmap");792}
793
794const char utf8_bom[] = "\357\273\277";795
796int skip_utf8_bom(char **text, size_t len)797{
798if (len < strlen(utf8_bom) ||799memcmp(*text, utf8_bom, strlen(utf8_bom)))800return 0;801*text += strlen(utf8_bom);802return 1;803}
804
805void strbuf_utf8_align(struct strbuf *buf, align_type position, unsigned int width,806const char *s)807{
808size_t slen = strlen(s);809int display_len = utf8_strnwidth(s, slen, 0);810int utf8_compensation = slen - display_len;811
812if (display_len >= width) {813strbuf_addstr(buf, s);814return;815}816
817if (position == ALIGN_LEFT)818strbuf_addf(buf, "%-*s", width + utf8_compensation, s);819else if (position == ALIGN_MIDDLE) {820int left = (width - display_len) / 2;821strbuf_addf(buf, "%*s%-*s", left, "", width - left + utf8_compensation, s);822} else if (position == ALIGN_RIGHT)823strbuf_addf(buf, "%*s", width + utf8_compensation, s);824}
825