1 /* searchutils.c - helper subroutines for grep's matchers.
2 Copyright 1992, 1998, 2000, 2007, 2009-2015 Free Software Foundation, Inc.
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 3, or (at your option)
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA
21 #define SEARCH_INLINE _GL_EXTERN_INLINE
22 #define SYSTEM_INLINE _GL_EXTERN_INLINE
27 #define NCHAR (UCHAR_MAX + 1)
29 size_t mbclen_cache[NCHAR];
32 kwsinit (kwset_t *kwset)
34 static char trans[NCHAR];
37 if (match_icase && MB_CUR_MAX == 1)
39 for (i = 0; i < NCHAR; ++i)
40 trans[i] = toupper (i);
42 *kwset = kwsalloc (trans);
45 *kwset = kwsalloc (NULL);
51 /* Convert BEG, an *N-byte string, to uppercase, and write the
52 NUL-terminated result into malloc'd storage. Upon success, set *N
53 to the length (in bytes) of the resulting string (not including the
54 trailing NUL byte), and return a pointer to the uppercase string.
55 Upon memory allocation failure, exit. *N must be positive.
57 Although this function returns a pointer to malloc'd storage,
58 the caller must not free it, since this function retains a pointer
59 to the buffer and reuses it on any subsequent call. As a consequence,
60 this function is not thread-safe.
62 When each character in the uppercase result string has the same length
63 as the corresponding character in the input string, set *LEN_MAP_P
64 to NULL. Otherwise, set it to a malloc'd buffer (like the returned
65 buffer, this must not be freed by caller) of the same length as the
66 result string. (*LEN_MAP_P)[J] is the change in byte-length of the
67 character in BEG that formed byte J of the result as it was converted to
68 uppercase. It is usually zero. For lowercase Turkish dotless I it
69 is -1, since the lowercase input occupies two bytes, while the
70 uppercase output occupies only one byte. For lowercase I in the
71 tr_TR.utf8 locale, it is 1 because the uppercase Turkish dotted I
72 is one byte longer than the original. When that happens, we have two
73 or more slots in *LEN_MAP_P for each such character. We store the
74 difference in the first one and 0's in any remaining slots.
76 This map is used by the caller to convert offset,length pairs that
77 reference the uppercase result to numbers that refer to the matched
78 part of the original buffer. */
81 mbtoupper (const char *beg, size_t *n, mb_len_map_t **len_map_p)
84 static mb_len_map_t *len_map;
85 static size_t outalloc;
86 size_t outlen, mb_cur_max;
91 bool lengths_differ = false;
93 if (*n > outalloc || outalloc == 0)
95 outalloc = MAX (1, *n);
96 out = xrealloc (out, outalloc);
97 len_map = xrealloc (len_map, outalloc);
100 /* appease clang-2.6 */
106 memset (&is, 0, sizeof (is));
107 memset (&os, 0, sizeof (os));
110 mb_cur_max = MB_CUR_MAX;
117 size_t mbclen = mbrtowc (&wc, beg, end - beg, &is);
119 /* Handle a UTF-8 sequence for a character beyond the base plane.
120 Cygwin's wchar_t is UTF-16, as in the underlying OS. This
121 results in surrogate pairs which need some extra attention. */
123 if (mbclen == 3 && (wc & 0xdc00) == 0xd800)
125 /* We got the start of a 4 byte UTF-8 sequence. This is returned
126 as a UTF-16 surrogate pair. The first call to mbrtowc returned 3
127 and wc has been set to a high surrogate value, now we're going
128 to fetch the matching low surrogate. This second call to mbrtowc
129 is supposed to return 1 to complete the 4 byte UTF-8 sequence. */
131 size_t mbclen_2 = mbrtowc (&wc_2, beg + mbclen, end - beg - mbclen,
133 if (mbclen_2 == 1 && (wc_2 & 0xdc00) == 0xdc00)
135 /* Match. Convert this to a 4 byte wint_t which constitutes
136 a 32-bit UTF-32 value. */
137 wci = ( (((wint_t) (wc - 0xd800)) << 10)
138 | ((wint_t) (wc_2 - 0xdc00)))
144 /* Invalid UTF-8 sequence. */
145 mbclen = (size_t) -1;
149 if (outlen + mb_cur_max >= outalloc)
151 size_t dm = m - len_map;
152 out = x2nrealloc (out, &outalloc, 1);
153 len_map = xrealloc (len_map, outalloc);
158 if (mbclen == (size_t) -1 || mbclen == (size_t) -2 || mbclen == 0)
160 /* An invalid sequence, or a truncated multi-octet character.
161 We treat it as a single-octet character. */
165 memset (&is, 0, sizeof (is));
166 memset (&os, 0, sizeof (os));
173 /* Handle Unicode characters beyond the base plane. */
176 /* towupper, taking wint_t (4 bytes), handles UCS-4 values. */
177 wci = towupper (wci);
181 wc = (wci >> 10) | 0xd800;
182 /* No need to check the return value. When reading the
183 high surrogate, the return value will be 0 and only the
184 mbstate indicates that we're in the middle of reading a
185 surrogate pair. The next wcrtomb call reading the low
186 surrogate will then return 4 and reset the mbstate. */
187 wcrtomb (p, wc, &os);
188 wc = (wci & 0x3ff) | 0xdc00;
194 ombclen = wcrtomb (p, wc, &os);
198 ombclen = wcrtomb (p, towupper (wc), &os);
199 *m = mbclen - ombclen;
200 memset (m + 1, 0, ombclen - 1);
204 lengths_differ |= (mbclen != ombclen);
208 *len_map_p = lengths_differ ? len_map : NULL;
214 /* Initialize a cache of mbrlen values for each of its 1-byte inputs. */
216 build_mbclen_cache (void)
220 for (i = CHAR_MIN; i <= CHAR_MAX; ++i)
223 unsigned char uc = i;
224 mbstate_t mbs = { 0 };
225 size_t len = mbrlen (&c, 1, &mbs);
226 mbclen_cache[uc] = len ? len : 1;
230 /* In the buffer *MB_START, return the number of bytes needed to go
231 back from CUR to the previous boundary, where a "boundary" is the
232 start of a multibyte character or is an error-encoding byte. The
233 buffer ends at END (i.e., one past the address of the buffer's last
234 byte). If CUR is already at a boundary, return 0. If *MB_START is
235 greater than or equal to CUR, return the negative value CUR - *MB_START.
237 When returning zero, set *MB_START to CUR. When returning a
238 positive value, set *MB_START to the next boundary after CUR, or to
239 END if there is no such boundary. When returning a negative value,
240 leave *MB_START alone. */
242 mb_goback (char const **mb_start, char const *cur, char const *end)
244 const char *p = *mb_start;
248 memset (&cur_state, 0, sizeof cur_state);
252 size_t clen = mb_clen (p, end - p, &cur_state);
254 if ((size_t) -2 <= clen)
256 /* An invalid sequence, or a truncated multibyte character.
257 Treat it as a single byte character. */
259 memset (&cur_state, 0, sizeof cur_state);
266 return p == cur ? 0 : cur - p0;
269 /* In the buffer BUF, return the wide character that is encoded just
270 before CUR. The buffer ends at END. Return WEOF if there is no
271 wide character just before CUR. */
273 mb_prev_wc (char const *buf, char const *cur, char const *end)
279 cur -= mb_goback (&p, cur, end);
280 return mb_next_wc (cur, end);
283 /* Return the wide character that is encoded at CUR. The buffer ends
284 at END. Return WEOF if there is no wide character encoded at CUR. */
286 mb_next_wc (char const *cur, char const *end)
289 mbstate_t mbs = { 0 };
290 return (end - cur != 0 && mbrtowc (&wc, cur, end - cur, &mbs) < (size_t) -2