Mercurial > hg > xemacs-beta
annotate src/search.c @ 4539:061e030e3270
Fix some bugs in load-history construction, built-in symbol file names.
lib-src/ChangeLog addition:
2008-12-27 Aidan Kehoe <kehoea@parhasard.net>
* make-docfile.c (main): Allow more than one -d argument, followed
by a directory to change to.
(put_filename): Don't strip directory information; with previous
change, allows retrieval of Lisp function and variable origin
files from #'built-in-symbol-file relative to lisp-directory.
(scan_lisp_file): Don't add an extraneous newline after the file
name, put_filename has added the newline already.
lisp/ChangeLog addition:
2008-12-27 Aidan Kehoe <kehoea@parhasard.net>
* loadup.el (load-history):
Add the contents of current-load-list to load-history before
clearing it. Move the variable declarations earlier in the file to
a format understood by make-docfile.c.
* custom.el (custom-declare-variable): Add the variable's symbol
to the current file's load history entry correctly, don't use a
cons. Eliminate a comment that we don't need to worry about, we
don't need to check the `initialized' C variable in Lisp.
* bytecomp.el (byte-compile-output-file-form):
Merge Andreas Schwab's pre-GPLv3 GNU change of 19970831 here;
treat #'custom-declare-variable correctly, generating the
docstrings in a format understood by make-docfile.c.
* loadhist.el (symbol-file): Correct behaviour for checking
autoloaded macros and functions when supplied with a TYPE
argument. Accept fully-qualified paths from
#'built-in-symbol-file; if a path is not fully-qualified, return
it relative to lisp-directory if the filename corresponds to a
Lisp file, and relative to (concat source-directory "/src/")
otherwise.
* make-docfile.el (preloaded-file-list):
Rationalise some let bindings a little. Use the "-d" argument to
make-docfile.c to supply Lisp paths relative to lisp-directory,
not absolutely. Add in loadup.el explicitly to the list of files
to be processed by make-docfile.c--it doesn't make sense to add it
to preloaded-file-list, since that is used for purposes of
byte-compilation too.
src/ChangeLog addition:
2008-12-27 Aidan Kehoe <kehoea@parhasard.net>
* doc.c (Fbuilt_in_symbol_file):
Return a subr's filename immediately if we've found it. Check for
compiled function and compiled macro docstrings in DOC too, and
return them if they exist.
The branch of the if statement focused on functions may have
executed, but we may still want to check variable bindings; an
else clause isn't appropriate.
| author | Aidan Kehoe <kehoea@parhasard.net> |
|---|---|
| date | Sat, 27 Dec 2008 14:05:50 +0000 |
| parents | 69b803c646cd |
| children | 91a023144e72 19a72041c5ed |
| rev | line source |
|---|---|
| 428 | 1 /* String search routines for XEmacs. |
| 2 Copyright (C) 1985, 1986, 1987, 1992-1995 Free Software Foundation, Inc. | |
| 3 Copyright (C) 1995 Sun Microsystems, Inc. | |
| 793 | 4 Copyright (C) 2001, 2002 Ben Wing. |
| 428 | 5 |
| 6 This file is part of XEmacs. | |
| 7 | |
| 8 XEmacs is free software; you can redistribute it and/or modify it | |
| 9 under the terms of the GNU General Public License as published by the | |
| 10 Free Software Foundation; either version 2, or (at your option) any | |
| 11 later version. | |
| 12 | |
| 13 XEmacs is distributed in the hope that it will be useful, but WITHOUT | |
| 14 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
| 15 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License | |
| 16 for more details. | |
| 17 | |
| 18 You should have received a copy of the GNU General Public License | |
| 19 along with XEmacs; see the file COPYING. If not, write to | |
| 20 the Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
| 21 Boston, MA 02111-1307, USA. */ | |
| 22 | |
| 23 /* Synched up with: FSF 19.29, except for region-cache stuff. */ | |
| 24 | |
| 25 /* Hacked on for Mule by Ben Wing, December 1994 and August 1995. */ | |
| 26 | |
| 826 | 27 /* This file has been Mule-ized. */ |
| 428 | 28 |
| 29 #include <config.h> | |
| 30 #include "lisp.h" | |
| 31 | |
| 32 #include "buffer.h" | |
| 33 #include "insdel.h" | |
| 34 #include "opaque.h" | |
| 35 #ifdef REGION_CACHE_NEEDS_WORK | |
| 36 #include "region-cache.h" | |
| 37 #endif | |
| 38 #include "syntax.h" | |
| 39 | |
| 40 #include <sys/types.h> | |
| 41 #include "regex.h" | |
| 446 | 42 #include "casetab.h" |
| 43 #include "chartab.h" | |
| 44 | |
| 45 #define TRANSLATE(table, pos) \ | |
| 867 | 46 (!NILP (table) ? TRT_TABLE_OF (table, (Ichar) pos) : pos) |
| 428 | 47 |
| 48 #define REGEXP_CACHE_SIZE 20 | |
| 49 | |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
50 #ifdef DEBUG_XEMACS |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
51 |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
52 /* Used in tests/automated/case-tests.el if available. */ |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
53 Fixnum debug_xemacs_searches; |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
54 |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
55 Lisp_Object Qsearch_algorithm_used, Qboyer_moore, Qsimple_search; |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
56 |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
57 #endif |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
58 |
| 428 | 59 /* If the regexp is non-nil, then the buffer contains the compiled form |
| 60 of that regexp, suitable for searching. */ | |
| 446 | 61 struct regexp_cache |
| 62 { | |
| 428 | 63 struct regexp_cache *next; |
| 64 Lisp_Object regexp; | |
| 65 struct re_pattern_buffer buf; | |
| 66 char fastmap[0400]; | |
| 67 /* Nonzero means regexp was compiled to do full POSIX backtracking. */ | |
| 68 char posix; | |
| 69 }; | |
| 70 | |
| 71 /* The instances of that struct. */ | |
| 72 static struct regexp_cache searchbufs[REGEXP_CACHE_SIZE]; | |
| 73 | |
| 74 /* The head of the linked list; points to the most recently used buffer. */ | |
| 75 static struct regexp_cache *searchbuf_head; | |
| 76 | |
| 77 | |
| 78 /* Every call to re_match, etc., must pass &search_regs as the regs | |
| 79 argument unless you can show it is unnecessary (i.e., if re_match | |
| 80 is certainly going to be called again before region-around-match | |
| 81 can be called). | |
| 82 | |
| 83 Since the registers are now dynamically allocated, we need to make | |
| 84 sure not to refer to the Nth register before checking that it has | |
| 85 been allocated by checking search_regs.num_regs. | |
| 86 | |
| 87 The regex code keeps track of whether it has allocated the search | |
| 88 buffer using bits in the re_pattern_buffer. This means that whenever | |
| 89 you compile a new pattern, it completely forgets whether it has | |
| 90 allocated any registers, and will allocate new registers the next | |
| 91 time you call a searching or matching function. Therefore, we need | |
| 92 to call re_set_registers after compiling a new pattern or after | |
| 93 setting the match registers, so that the regex functions will be | |
| 94 able to free or re-allocate it properly. */ | |
| 95 | |
| 96 /* Note: things get trickier under Mule because the values returned from | |
| 826 | 97 the regexp routines are in Bytebpos's but we need them to be in Charbpos's. |
| 428 | 98 We take the easy way out for the moment and just convert them immediately. |
| 99 We could be more clever by not converting them until necessary, but | |
| 100 that gets real ugly real fast since the buffer might have changed and | |
| 101 the positions might be out of sync or out of range. | |
| 102 */ | |
| 103 static struct re_registers search_regs; | |
| 104 | |
| 1468 | 105 /* Every function that sets the match data _must_ clear unused search |
| 106 registers on success. An unsuccessful search or match _must_ preserve | |
| 107 the search registers. The traditional documentation implied that | |
| 108 any match operation might trash the registers, but in fact failures | |
| 109 have always preserved the match data (in GNU Emacs as well). Some | |
| 110 plausible code depends on this behavior (cf. `w3-configuration-data' | |
| 111 in library "w3-cfg"). | |
| 112 | |
| 113 Ordinary string searchs use set_search_regs to set the whole-string | |
| 114 match. That function takes care of clearing the unused subexpression | |
| 1425 | 115 registers. |
| 116 */ | |
| 117 static void set_search_regs (struct buffer *buf, Charbpos beg, Charcount len); | |
| 1468 | 118 static void clear_search_regs (void); |
| 1425 | 119 |
| 428 | 120 /* The buffer in which the last search was performed, or |
| 121 Qt if the last search was done in a string; | |
| 122 Qnil if no searching has been done yet. */ | |
| 123 static Lisp_Object last_thing_searched; | |
| 124 | |
| 125 /* error condition signalled when regexp compile_pattern fails */ | |
| 126 | |
| 127 Lisp_Object Qinvalid_regexp; | |
| 128 | |
| 129 /* Regular expressions used in forward/backward-word */ | |
| 130 Lisp_Object Vforward_word_regexp, Vbackward_word_regexp; | |
| 131 | |
| 507 | 132 Fixnum warn_about_possibly_incompatible_back_references; |
| 502 | 133 |
| 428 | 134 /* range table for use with skip_chars. Only needed for Mule. */ |
| 135 Lisp_Object Vskip_chars_range_table; | |
| 136 | |
| 867 | 137 static Charbpos simple_search (struct buffer *buf, Ibyte *base_pat, |
| 826 | 138 Bytecount len, Bytebpos pos, Bytebpos lim, |
| 139 EMACS_INT n, Lisp_Object trt); | |
| 867 | 140 static Charbpos boyer_moore (struct buffer *buf, Ibyte *base_pat, |
| 826 | 141 Bytecount len, Bytebpos pos, Bytebpos lim, |
| 142 EMACS_INT n, Lisp_Object trt, | |
| 143 Lisp_Object inverse_trt, int charset_base); | |
| 665 | 144 static Charbpos search_buffer (struct buffer *buf, Lisp_Object str, |
| 826 | 145 Charbpos charbpos, Charbpos buflim, EMACS_INT n, |
| 146 int RE, Lisp_Object trt, | |
| 147 Lisp_Object inverse_trt, int posix); | |
| 771 | 148 |
| 2268 | 149 static DECLARE_DOESNT_RETURN (matcher_overflow (void)); |
| 150 | |
| 151 static DOESNT_RETURN | |
| 152 matcher_overflow () | |
| 428 | 153 { |
| 563 | 154 stack_overflow ("Stack overflow in regexp matcher", Qunbound); |
| 428 | 155 } |
| 156 | |
| 157 /* Compile a regexp and signal a Lisp error if anything goes wrong. | |
| 158 PATTERN is the pattern to compile. | |
| 159 CP is the place to put the result. | |
| 826 | 160 TRANSLATE is a translation table for ignoring case, or Qnil for none. |
| 428 | 161 REGP is the structure that says where to store the "register" |
| 162 values that will result from matching this pattern. | |
| 163 If it is 0, we should compile the pattern not to record any | |
| 164 subexpression bounds. | |
| 165 POSIX is nonzero if we want full backtracking (POSIX style) | |
| 166 for this pattern. 0 means backtrack only enough to get a valid match. */ | |
| 167 | |
| 168 static int | |
| 169 compile_pattern_1 (struct regexp_cache *cp, Lisp_Object pattern, | |
| 2286 | 170 struct re_registers *UNUSED (regp), Lisp_Object translate, |
| 826 | 171 int posix, Error_Behavior errb) |
| 428 | 172 { |
| 442 | 173 const char *val; |
| 428 | 174 reg_syntax_t old; |
| 175 | |
| 176 cp->regexp = Qnil; | |
| 177 cp->buf.translate = translate; | |
| 178 cp->posix = posix; | |
| 179 old = re_set_syntax (RE_SYNTAX_EMACS | |
| 180 | (posix ? 0 : RE_NO_POSIX_BACKTRACKING)); | |
| 442 | 181 val = (const char *) |
| 428 | 182 re_compile_pattern ((char *) XSTRING_DATA (pattern), |
| 183 XSTRING_LENGTH (pattern), &cp->buf); | |
| 184 re_set_syntax (old); | |
| 185 if (val) | |
| 186 { | |
| 563 | 187 maybe_signal_error (Qinvalid_regexp, 0, build_string (val), |
| 428 | 188 Qsearch, errb); |
| 189 return 0; | |
| 190 } | |
| 191 | |
| 192 cp->regexp = Fcopy_sequence (pattern); | |
| 193 return 1; | |
| 194 } | |
| 195 | |
| 196 /* Compile a regexp if necessary, but first check to see if there's one in | |
| 197 the cache. | |
| 198 PATTERN is the pattern to compile. | |
| 826 | 199 TRANSLATE is a translation table for ignoring case, or Qnil for none. |
| 428 | 200 REGP is the structure that says where to store the "register" |
| 201 values that will result from matching this pattern. | |
| 202 If it is 0, we should compile the pattern not to record any | |
| 203 subexpression bounds. | |
| 204 POSIX is nonzero if we want full backtracking (POSIX style) | |
| 205 for this pattern. 0 means backtrack only enough to get a valid match. */ | |
| 206 | |
| 207 struct re_pattern_buffer * | |
| 208 compile_pattern (Lisp_Object pattern, struct re_registers *regp, | |
| 2286 | 209 Lisp_Object translate, Lisp_Object UNUSED (searchobj), |
| 210 struct buffer *UNUSED (searchbuf), int posix, | |
| 211 Error_Behavior errb) | |
| 428 | 212 { |
| 213 struct regexp_cache *cp, **cpp; | |
| 214 | |
| 215 for (cpp = &searchbuf_head; ; cpp = &cp->next) | |
| 216 { | |
| 217 cp = *cpp; | |
| 826 | 218 /* &&#### once we fix up the fastmap code in regex.c for 8-bit-fixed, |
| 219 we need to record and compare the buffer and format, since the | |
| 220 fastmap will reflect the state of the buffer -- and things get | |
| 221 more complicated if the buffer has changed formats or (esp.) has | |
| 222 kept the format but changed its interpretation! may need to have | |
| 223 the code that changes the interpretation go through and invalidate | |
| 224 cache entries for that buffer. */ | |
| 428 | 225 if (!NILP (Fstring_equal (cp->regexp, pattern)) |
| 446 | 226 && EQ (cp->buf.translate, translate) |
| 428 | 227 && cp->posix == posix) |
| 228 break; | |
| 229 | |
| 230 /* If we're at the end of the cache, compile into the last cell. */ | |
| 231 if (cp->next == 0) | |
| 232 { | |
| 826 | 233 if (!compile_pattern_1 (cp, pattern, regp, translate, |
| 234 posix, errb)) | |
| 428 | 235 return 0; |
| 236 break; | |
| 237 } | |
| 238 } | |
| 239 | |
| 240 /* When we get here, cp (aka *cpp) contains the compiled pattern, | |
| 241 either because we found it in the cache or because we just compiled it. | |
| 242 Move it to the front of the queue to mark it as most recently used. */ | |
| 243 *cpp = cp->next; | |
| 244 cp->next = searchbuf_head; | |
| 245 searchbuf_head = cp; | |
| 246 | |
| 247 /* Advise the searching functions about the space we have allocated | |
| 248 for register data. */ | |
| 249 if (regp) | |
| 250 re_set_registers (&cp->buf, regp, regp->num_regs, regp->start, regp->end); | |
| 251 | |
| 252 return &cp->buf; | |
| 253 } | |
| 254 | |
| 255 /* Error condition used for failing searches */ | |
| 256 Lisp_Object Qsearch_failed; | |
| 257 | |
| 2268 | 258 static DECLARE_DOESNT_RETURN (signal_failure (Lisp_Object)); |
| 259 | |
| 260 static DOESNT_RETURN | |
| 428 | 261 signal_failure (Lisp_Object arg) |
| 262 { | |
| 446 | 263 for (;;) |
| 264 Fsignal (Qsearch_failed, list1 (arg)); | |
| 428 | 265 } |
| 266 | |
| 826 | 267 /* Convert the search registers from Bytebpos's to Charbpos's. Needs to be |
| 428 | 268 done after each regexp match that uses the search regs. |
| 269 | |
| 270 We could get a potential speedup by not converting the search registers | |
| 271 until it's really necessary, e.g. when match-data or replace-match is | |
| 272 called. However, this complexifies the code a lot (e.g. the buffer | |
| 826 | 273 could have changed and the Bytebpos's stored might be invalid) and is |
| 428 | 274 probably not a great time-saver. */ |
| 275 | |
| 276 static void | |
| 277 fixup_search_regs_for_buffer (struct buffer *buf) | |
| 278 { | |
| 279 int i; | |
| 280 int num_regs = search_regs.num_regs; | |
| 281 | |
| 282 for (i = 0; i < num_regs; i++) | |
| 283 { | |
| 284 if (search_regs.start[i] >= 0) | |
| 826 | 285 search_regs.start[i] = bytebpos_to_charbpos (buf, |
| 286 search_regs.start[i]); | |
| 428 | 287 if (search_regs.end[i] >= 0) |
| 665 | 288 search_regs.end[i] = bytebpos_to_charbpos (buf, search_regs.end[i]); |
| 428 | 289 } |
| 290 } | |
| 291 | |
| 292 /* Similar but for strings. */ | |
| 293 static void | |
| 294 fixup_search_regs_for_string (Lisp_Object string) | |
| 295 { | |
| 296 int i; | |
| 297 int num_regs = search_regs.num_regs; | |
| 298 | |
| 299 /* #### bytecount_to_charcount() is not that efficient. This function | |
| 867 | 300 could be faster if it did its own conversion (using INC_IBYTEPTR() |
| 428 | 301 and such), because the register ends are likely to be somewhat ordered. |
| 302 (Even if not, you could sort them.) | |
| 303 | |
| 304 Think about this if this function is a time hog, which it's probably | |
| 305 not. */ | |
| 306 for (i = 0; i < num_regs; i++) | |
| 307 { | |
| 308 if (search_regs.start[i] > 0) | |
| 309 { | |
| 310 search_regs.start[i] = | |
| 793 | 311 string_index_byte_to_char (string, search_regs.start[i]); |
| 428 | 312 } |
| 313 if (search_regs.end[i] > 0) | |
| 314 { | |
| 315 search_regs.end[i] = | |
| 793 | 316 string_index_byte_to_char (string, search_regs.end[i]); |
| 428 | 317 } |
| 318 } | |
| 319 } | |
| 320 | |
| 321 | |
| 322 static Lisp_Object | |
| 323 looking_at_1 (Lisp_Object string, struct buffer *buf, int posix) | |
| 324 { | |
| 325 Lisp_Object val; | |
| 665 | 326 Bytebpos p1, p2; |
| 428 | 327 Bytecount s1, s2; |
| 328 REGISTER int i; | |
| 329 struct re_pattern_buffer *bufp; | |
| 826 | 330 struct syntax_cache scache_struct; |
| 331 struct syntax_cache *scache = &scache_struct; | |
| 332 | |
| 428 | 333 CHECK_STRING (string); |
| 334 bufp = compile_pattern (string, &search_regs, | |
| 335 (!NILP (buf->case_fold_search) | |
| 446 | 336 ? XCASE_TABLE_DOWNCASE (buf->case_table) : Qnil), |
| 826 | 337 wrap_buffer (buf), buf, posix, ERROR_ME); |
| 428 | 338 |
| 339 QUIT; | |
| 340 | |
| 341 /* Get pointers and sizes of the two strings | |
| 342 that make up the visible portion of the buffer. */ | |
| 343 | |
| 826 | 344 p1 = BYTE_BUF_BEGV (buf); |
| 345 p2 = BYTE_BUF_CEILING_OF (buf, p1); | |
| 428 | 346 s1 = p2 - p1; |
| 826 | 347 s2 = BYTE_BUF_ZV (buf) - p2; |
| 348 | |
| 349 /* By making the regex object, regex buffer, and syntax cache arguments | |
| 350 to re_{search,match}{,_2}, we've removed the need to do nasty things | |
| 351 to deal with regex reentrancy. (See stack trace in signal.c for proof | |
| 352 that this can happen.) | |
| 353 | |
| 354 #### there is still a potential problem with the regex cache -- | |
| 355 the compiled regex could be overwritten. we'd need 20-fold | |
| 356 reentrancy, though. Fix this. */ | |
| 357 | |
| 358 i = re_match_2 (bufp, (char *) BYTE_BUF_BYTE_ADDRESS (buf, p1), | |
| 359 s1, (char *) BYTE_BUF_BYTE_ADDRESS (buf, p2), s2, | |
| 360 BYTE_BUF_PT (buf) - BYTE_BUF_BEGV (buf), &search_regs, | |
| 361 BYTE_BUF_ZV (buf) - BYTE_BUF_BEGV (buf), wrap_buffer (buf), | |
| 362 buf, scache); | |
| 428 | 363 |
| 364 if (i == -2) | |
| 365 matcher_overflow (); | |
| 366 | |
| 367 val = (0 <= i ? Qt : Qnil); | |
| 368 if (NILP (val)) | |
| 826 | 369 return Qnil; |
| 428 | 370 { |
| 371 int num_regs = search_regs.num_regs; | |
| 372 for (i = 0; i < num_regs; i++) | |
| 373 if (search_regs.start[i] >= 0) | |
| 374 { | |
| 826 | 375 search_regs.start[i] += BYTE_BUF_BEGV (buf); |
| 376 search_regs.end[i] += BYTE_BUF_BEGV (buf); | |
| 428 | 377 } |
| 378 } | |
| 793 | 379 last_thing_searched = wrap_buffer (buf); |
| 428 | 380 fixup_search_regs_for_buffer (buf); |
| 826 | 381 return val; |
| 428 | 382 } |
| 383 | |
| 384 DEFUN ("looking-at", Flooking_at, 1, 2, 0, /* | |
| 385 Return t if text after point matches regular expression REGEXP. | |
| 1468 | 386 When the match is successful, this function modifies the match data |
| 387 that `match-beginning', `match-end' and `match-data' access; save the | |
| 388 match data with `match-data' and restore it with `store-match-data' if | |
| 389 you want to preserve them. If the match fails, the match data from the | |
| 390 previous success match is preserved. | |
| 428 | 391 |
| 392 Optional argument BUFFER defaults to the current buffer. | |
| 393 */ | |
| 394 (regexp, buffer)) | |
| 395 { | |
| 396 return looking_at_1 (regexp, decode_buffer (buffer, 0), 0); | |
| 397 } | |
| 398 | |
| 399 DEFUN ("posix-looking-at", Fposix_looking_at, 1, 2, 0, /* | |
| 400 Return t if text after point matches regular expression REGEXP. | |
| 401 Find the longest match, in accord with Posix regular expression rules. | |
| 1468 | 402 When the match is successful, this function modifies the match data |
| 403 that `match-beginning', `match-end' and `match-data' access; save the | |
| 404 match data with `match-data' and restore it with `store-match-data' if | |
| 405 you want to preserve them. If the match fails, the match data from the | |
| 406 previous success match is preserved. | |
| 428 | 407 |
| 408 Optional argument BUFFER defaults to the current buffer. | |
| 409 */ | |
| 410 (regexp, buffer)) | |
| 411 { | |
| 826 | 412 return looking_at_1 (regexp, decode_buffer (buffer, 0), 1); |
| 428 | 413 } |
| 414 | |
| 415 static Lisp_Object | |
| 416 string_match_1 (Lisp_Object regexp, Lisp_Object string, Lisp_Object start, | |
| 2286 | 417 struct buffer *buf, int UNUSED (posix)) |
| 428 | 418 { |
| 419 Bytecount val; | |
| 420 Charcount s; | |
| 421 struct re_pattern_buffer *bufp; | |
| 422 | |
| 853 | 423 /* Some FSF junk with running_asynch_code, to preserve the match |
| 424 data. Not necessary because we don't call process filters | |
| 425 asynchronously (i.e. from within QUIT). */ | |
| 428 | 426 |
| 427 CHECK_STRING (regexp); | |
| 428 CHECK_STRING (string); | |
| 429 | |
| 430 if (NILP (start)) | |
| 431 s = 0; | |
| 432 else | |
| 433 { | |
| 826 | 434 Charcount len = string_char_length (string); |
| 428 | 435 |
| 436 CHECK_INT (start); | |
| 437 s = XINT (start); | |
| 438 if (s < 0 && -s <= len) | |
| 439 s = len + s; | |
| 440 else if (0 > s || s > len) | |
| 441 args_out_of_range (string, start); | |
| 442 } | |
| 443 | |
| 444 | |
| 445 bufp = compile_pattern (regexp, &search_regs, | |
| 446 (!NILP (buf->case_fold_search) | |
| 446 | 447 ? XCASE_TABLE_DOWNCASE (buf->case_table) : Qnil), |
| 826 | 448 string, buf, 0, ERROR_ME); |
| 428 | 449 QUIT; |
| 450 { | |
| 793 | 451 Bytecount bis = string_index_char_to_byte (string, s); |
| 826 | 452 struct syntax_cache scache_struct; |
| 453 struct syntax_cache *scache = &scache_struct; | |
| 454 | |
| 455 /* By making the regex object, regex buffer, and syntax cache arguments | |
| 456 to re_{search,match}{,_2}, we've removed the need to do nasty things | |
| 457 to deal with regex reentrancy. (See stack trace in signal.c for proof | |
| 458 that this can happen.) | |
| 459 | |
| 460 #### there is still a potential problem with the regex cache -- | |
| 461 the compiled regex could be overwritten. we'd need 20-fold | |
| 462 reentrancy, though. Fix this. */ | |
| 463 | |
| 428 | 464 val = re_search (bufp, (char *) XSTRING_DATA (string), |
| 465 XSTRING_LENGTH (string), bis, | |
| 466 XSTRING_LENGTH (string) - bis, | |
| 826 | 467 &search_regs, string, buf, scache); |
| 428 | 468 } |
| 469 if (val == -2) | |
| 470 matcher_overflow (); | |
| 826 | 471 if (val < 0) return Qnil; |
| 428 | 472 last_thing_searched = Qt; |
| 473 fixup_search_regs_for_string (string); | |
| 826 | 474 return make_int (string_index_byte_to_char (string, val)); |
| 428 | 475 } |
| 476 | |
| 477 DEFUN ("string-match", Fstring_match, 2, 4, 0, /* | |
| 478 Return index of start of first match for REGEXP in STRING, or nil. | |
| 479 If third arg START is non-nil, start search at that index in STRING. | |
| 480 For index of first char beyond the match, do (match-end 0). | |
| 481 `match-end' and `match-beginning' also give indices of substrings | |
| 482 matched by parenthesis constructs in the pattern. | |
| 483 | |
| 826 | 484 Optional arg BUFFER controls how case folding and syntax and category |
| 485 lookup is done (according to the value of `case-fold-search' in that buffer | |
| 486 and that buffer's case tables, syntax tables, and category table). If nil | |
| 487 or unspecified, it defaults *NOT* to the current buffer but instead: | |
| 488 | |
| 489 -- the value of `case-fold-search' in the current buffer is still respected | |
| 490 because of idioms like | |
| 491 | |
| 492 (let ((case-fold-search nil)) | |
| 493 (string-match "^foo.*bar" string)) | |
| 494 | |
| 495 but the case, syntax, and category tables come from the standard tables, | |
| 1468 | 496 which are accessed through functions `default-{case,syntax,category}-table' |
| 497 and serve as the parents of the tables in particular buffer. | |
| 498 | |
| 499 When the match is successful, this function modifies the match data | |
| 500 that `match-beginning', `match-end' and `match-data' access; save the | |
| 501 match data with `match-data' and restore it with `store-match-data' if | |
| 502 you want to preserve them. If the match fails, the match data from the | |
| 503 previous success match is preserved. | |
| 428 | 504 */ |
| 505 (regexp, string, start, buffer)) | |
| 506 { | |
| 826 | 507 /* &&#### implement new interp for buffer arg; check code to see if it |
| 508 makes more sense than prev */ | |
| 428 | 509 return string_match_1 (regexp, string, start, decode_buffer (buffer, 0), 0); |
| 510 } | |
| 511 | |
| 512 DEFUN ("posix-string-match", Fposix_string_match, 2, 4, 0, /* | |
| 513 Return index of start of first match for REGEXP in STRING, or nil. | |
| 514 Find the longest match, in accord with Posix regular expression rules. | |
| 515 If third arg START is non-nil, start search at that index in STRING. | |
| 516 For index of first char beyond the match, do (match-end 0). | |
| 517 `match-end' and `match-beginning' also give indices of substrings | |
| 518 matched by parenthesis constructs in the pattern. | |
| 519 | |
| 520 Optional arg BUFFER controls how case folding is done (according to | |
| 521 the value of `case-fold-search' in that buffer and that buffer's case | |
| 522 tables) and defaults to the current buffer. | |
| 1468 | 523 |
| 524 When the match is successful, this function modifies the match data | |
| 525 that `match-beginning', `match-end' and `match-data' access; save the | |
| 526 match data with `match-data' and restore it with `store-match-data' if | |
| 527 you want to preserve them. If the match fails, the match data from the | |
| 528 previous success match is preserved. | |
| 428 | 529 */ |
| 530 (regexp, string, start, buffer)) | |
| 531 { | |
| 532 return string_match_1 (regexp, string, start, decode_buffer (buffer, 0), 1); | |
| 533 } | |
| 534 | |
| 535 /* Match REGEXP against STRING, searching all of STRING, | |
| 536 and return the index of the match, or negative on failure. | |
| 537 This does not clobber the match data. */ | |
| 538 | |
| 539 Bytecount | |
| 1347 | 540 fast_string_match (Lisp_Object regexp, const Ibyte *nonreloc, |
| 428 | 541 Lisp_Object reloc, Bytecount offset, |
| 542 Bytecount length, int case_fold_search, | |
| 578 | 543 Error_Behavior errb, int no_quit) |
| 428 | 544 { |
| 545 Bytecount val; | |
| 867 | 546 Ibyte *newnonreloc = (Ibyte *) nonreloc; |
| 428 | 547 struct re_pattern_buffer *bufp; |
| 826 | 548 struct syntax_cache scache_struct; |
| 549 struct syntax_cache *scache = &scache_struct; | |
| 428 | 550 |
| 551 bufp = compile_pattern (regexp, 0, | |
| 552 (case_fold_search | |
| 771 | 553 ? XCASE_TABLE_DOWNCASE (Vstandard_case_table) |
| 446 | 554 : Qnil), |
| 826 | 555 reloc, 0, 0, errb); |
| 428 | 556 if (!bufp) |
| 557 return -1; /* will only do this when errb != ERROR_ME */ | |
| 558 if (!no_quit) | |
| 559 QUIT; | |
| 560 else | |
| 561 no_quit_in_re_search = 1; | |
| 562 | |
| 563 fixup_internal_substring (nonreloc, reloc, offset, &length); | |
| 564 | |
| 771 | 565 /* Don't need to protect against GC inside of re_search() due to QUIT; |
| 566 QUIT is GC-inhibited. */ | |
| 428 | 567 if (!NILP (reloc)) |
| 771 | 568 newnonreloc = XSTRING_DATA (reloc); |
| 569 | |
| 826 | 570 /* By making the regex object, regex buffer, and syntax cache arguments |
| 571 to re_{search,match}{,_2}, we've removed the need to do nasty things | |
| 572 to deal with regex reentrancy. (See stack trace in signal.c for proof | |
| 573 that this can happen.) | |
| 574 | |
| 575 #### there is still a potential problem with the regex cache -- | |
| 576 the compiled regex could be overwritten. we'd need 20-fold | |
| 577 reentrancy, though. Fix this. */ | |
| 578 | |
| 428 | 579 val = re_search (bufp, (char *) newnonreloc + offset, length, 0, |
| 826 | 580 length, 0, reloc, 0, scache); |
| 428 | 581 |
| 582 no_quit_in_re_search = 0; | |
| 583 return val; | |
| 584 } | |
| 585 | |
| 586 Bytecount | |
| 587 fast_lisp_string_match (Lisp_Object regex, Lisp_Object string) | |
| 588 { | |
| 589 return fast_string_match (regex, 0, string, 0, -1, 0, ERROR_ME, 0); | |
| 590 } | |
| 591 | |
| 592 | |
| 593 #ifdef REGION_CACHE_NEEDS_WORK | |
| 594 /* The newline cache: remembering which sections of text have no newlines. */ | |
| 595 | |
| 596 /* If the user has requested newline caching, make sure it's on. | |
| 597 Otherwise, make sure it's off. | |
| 598 This is our cheezy way of associating an action with the change of | |
| 599 state of a buffer-local variable. */ | |
| 600 static void | |
| 601 newline_cache_on_off (struct buffer *buf) | |
| 602 { | |
| 603 if (NILP (buf->cache_long_line_scans)) | |
| 604 { | |
| 605 /* It should be off. */ | |
| 606 if (buf->newline_cache) | |
| 607 { | |
| 608 free_region_cache (buf->newline_cache); | |
| 609 buf->newline_cache = 0; | |
| 610 } | |
| 611 } | |
| 612 else | |
| 613 { | |
| 614 /* It should be on. */ | |
| 615 if (buf->newline_cache == 0) | |
| 616 buf->newline_cache = new_region_cache (); | |
| 617 } | |
| 618 } | |
| 619 #endif | |
| 620 | |
| 621 /* Search in BUF for COUNT instances of the character TARGET between | |
| 622 START and END. | |
| 623 | |
| 624 If COUNT is positive, search forwards; END must be >= START. | |
| 625 If COUNT is negative, search backwards for the -COUNTth instance; | |
| 626 END must be <= START. | |
| 627 If COUNT is zero, do anything you please; run rogue, for all I care. | |
| 628 | |
| 629 If END is zero, use BEGV or ZV instead, as appropriate for the | |
| 630 direction indicated by COUNT. | |
| 631 | |
| 632 If we find COUNT instances, set *SHORTAGE to zero, and return the | |
| 633 position after the COUNTth match. Note that for reverse motion | |
| 634 this is not the same as the usual convention for Emacs motion commands. | |
| 635 | |
| 636 If we don't find COUNT instances before reaching END, set *SHORTAGE | |
| 637 to the number of TARGETs left unfound, and return END. | |
| 638 | |
| 639 If ALLOW_QUIT is non-zero, call QUIT periodically. */ | |
| 640 | |
| 665 | 641 static Bytebpos |
| 867 | 642 byte_scan_buffer (struct buffer *buf, Ichar target, Bytebpos st, Bytebpos en, |
| 872 | 643 EMACS_INT count, EMACS_INT *shortage, int allow_quit) |
| 428 | 644 { |
| 665 | 645 Bytebpos lim = en > 0 ? en : |
| 826 | 646 ((count > 0) ? BYTE_BUF_ZV (buf) : BYTE_BUF_BEGV (buf)); |
| 428 | 647 |
| 648 /* #### newline cache stuff in this function not yet ported */ | |
| 649 assert (count != 0); | |
| 650 | |
| 651 if (shortage) | |
| 652 *shortage = 0; | |
| 653 | |
| 654 if (count > 0) | |
| 655 { | |
| 656 #ifdef MULE | |
| 826 | 657 Internal_Format fmt = buf->text->format; |
| 658 /* Check for char that's unrepresentable in the buffer -- it | |
| 659 certainly can't be there. */ | |
| 867 | 660 if (!ichar_fits_in_format (target, fmt, wrap_buffer (buf))) |
| 428 | 661 { |
| 826 | 662 *shortage = count; |
| 663 return lim; | |
| 664 } | |
| 665 /* Due to the Mule representation of characters in a buffer, we can | |
| 666 simply search for characters in the range 0 - 127 directly; for | |
| 667 8-bit-fixed, we can do this for all characters. In other cases, | |
| 668 we do it the "hard" way. Note that this way works for all | |
| 669 characters and all formats, but the other way is faster. */ | |
| 670 else if (! (fmt == FORMAT_8_BIT_FIXED || | |
| 867 | 671 (fmt == FORMAT_DEFAULT && ichar_ascii_p (target)))) |
| 826 | 672 { |
| 867 | 673 Raw_Ichar raw = ichar_to_raw (target, fmt, wrap_buffer (buf)); |
| 428 | 674 while (st < lim && count > 0) |
| 675 { | |
| 826 | 676 if (BYTE_BUF_FETCH_CHAR_RAW (buf, st) == raw) |
| 428 | 677 count--; |
| 665 | 678 INC_BYTEBPOS (buf, st); |
| 428 | 679 } |
| 680 } | |
| 681 else | |
| 682 #endif | |
| 683 { | |
| 867 | 684 Raw_Ichar raw = ichar_to_raw (target, fmt, wrap_buffer (buf)); |
| 428 | 685 while (st < lim && count > 0) |
| 686 { | |
| 665 | 687 Bytebpos ceil; |
| 867 | 688 Ibyte *bufptr; |
| 428 | 689 |
| 826 | 690 ceil = BYTE_BUF_CEILING_OF (buf, st); |
| 428 | 691 ceil = min (lim, ceil); |
| 867 | 692 bufptr = (Ibyte *) memchr (BYTE_BUF_BYTE_ADDRESS (buf, st), |
| 826 | 693 raw, ceil - st); |
| 428 | 694 if (bufptr) |
| 695 { | |
| 696 count--; | |
| 826 | 697 st = BYTE_BUF_PTR_BYTE_POS (buf, bufptr) + 1; |
| 428 | 698 } |
| 699 else | |
| 700 st = ceil; | |
| 701 } | |
| 702 } | |
| 703 | |
| 704 if (shortage) | |
| 705 *shortage = count; | |
| 706 if (allow_quit) | |
| 707 QUIT; | |
| 708 return st; | |
| 709 } | |
| 710 else | |
| 711 { | |
| 712 #ifdef MULE | |
| 826 | 713 Internal_Format fmt = buf->text->format; |
| 714 /* Check for char that's unrepresentable in the buffer -- it | |
| 715 certainly can't be there. */ | |
| 867 | 716 if (!ichar_fits_in_format (target, fmt, wrap_buffer (buf))) |
| 428 | 717 { |
| 826 | 718 *shortage = -count; |
| 719 return lim; | |
| 720 } | |
| 721 else if (! (fmt == FORMAT_8_BIT_FIXED || | |
| 867 | 722 (fmt == FORMAT_DEFAULT && ichar_ascii_p (target)))) |
| 826 | 723 { |
| 867 | 724 Raw_Ichar raw = ichar_to_raw (target, fmt, wrap_buffer (buf)); |
| 428 | 725 while (st > lim && count < 0) |
| 726 { | |
| 665 | 727 DEC_BYTEBPOS (buf, st); |
| 826 | 728 if (BYTE_BUF_FETCH_CHAR_RAW (buf, st) == raw) |
| 428 | 729 count++; |
| 730 } | |
| 731 } | |
| 732 else | |
| 733 #endif | |
| 734 { | |
| 867 | 735 Raw_Ichar raw = ichar_to_raw (target, fmt, wrap_buffer (buf)); |
| 428 | 736 while (st > lim && count < 0) |
| 737 { | |
| 665 | 738 Bytebpos floor; |
| 867 | 739 Ibyte *bufptr; |
| 740 Ibyte *floorptr; | |
| 428 | 741 |
| 826 | 742 floor = BYTE_BUF_FLOOR_OF (buf, st); |
| 428 | 743 floor = max (lim, floor); |
| 744 /* No memrchr() ... */ | |
| 826 | 745 bufptr = BYTE_BUF_BYTE_ADDRESS_BEFORE (buf, st); |
| 746 floorptr = BYTE_BUF_BYTE_ADDRESS (buf, floor); | |
| 428 | 747 while (bufptr >= floorptr) |
| 748 { | |
| 749 st--; | |
| 750 /* At this point, both ST and BUFPTR refer to the same | |
| 751 character. When the loop terminates, ST will | |
| 752 always point to the last character we tried. */ | |
| 867 | 753 if (*bufptr == (Ibyte) raw) |
| 428 | 754 { |
| 755 count++; | |
| 756 break; | |
| 757 } | |
| 758 bufptr--; | |
| 759 } | |
| 760 } | |
| 761 } | |
| 762 | |
| 763 if (shortage) | |
| 764 *shortage = -count; | |
| 765 if (allow_quit) | |
| 766 QUIT; | |
| 767 if (count) | |
| 768 return st; | |
| 769 else | |
| 770 { | |
| 771 /* We found the character we were looking for; we have to return | |
| 772 the position *after* it due to the strange way that the return | |
| 773 value is defined. */ | |
| 665 | 774 INC_BYTEBPOS (buf, st); |
| 428 | 775 return st; |
| 776 } | |
| 777 } | |
| 778 } | |
| 779 | |
| 665 | 780 Charbpos |
| 867 | 781 scan_buffer (struct buffer *buf, Ichar target, Charbpos start, Charbpos end, |
| 428 | 782 EMACS_INT count, EMACS_INT *shortage, int allow_quit) |
| 783 { | |
| 826 | 784 Bytebpos byte_retval; |
| 785 Bytebpos byte_start, byte_end; | |
| 786 | |
| 787 byte_start = charbpos_to_bytebpos (buf, start); | |
| 428 | 788 if (end) |
| 826 | 789 byte_end = charbpos_to_bytebpos (buf, end); |
| 428 | 790 else |
| 826 | 791 byte_end = 0; |
| 792 byte_retval = byte_scan_buffer (buf, target, byte_start, byte_end, count, | |
| 428 | 793 shortage, allow_quit); |
| 826 | 794 return bytebpos_to_charbpos (buf, byte_retval); |
| 428 | 795 } |
| 796 | |
| 665 | 797 Bytebpos |
| 826 | 798 byte_find_next_newline_no_quit (struct buffer *buf, Bytebpos from, int count) |
| 428 | 799 { |
| 826 | 800 return byte_scan_buffer (buf, '\n', from, 0, count, 0, 0); |
| 428 | 801 } |
| 802 | |
| 665 | 803 Charbpos |
| 804 find_next_newline_no_quit (struct buffer *buf, Charbpos from, int count) | |
| 428 | 805 { |
| 806 return scan_buffer (buf, '\n', from, 0, count, 0, 0); | |
| 807 } | |
| 808 | |
| 665 | 809 Charbpos |
| 810 find_next_newline (struct buffer *buf, Charbpos from, int count) | |
| 428 | 811 { |
| 812 return scan_buffer (buf, '\n', from, 0, count, 0, 1); | |
| 813 } | |
| 814 | |
| 826 | 815 Bytecount |
| 867 | 816 byte_find_next_ichar_in_string (Lisp_Object str, Ichar target, Bytecount st, |
| 428 | 817 EMACS_INT count) |
| 818 { | |
| 793 | 819 Bytebpos lim = XSTRING_LENGTH (str) -1; |
| 867 | 820 Ibyte *s = XSTRING_DATA (str); |
| 428 | 821 |
| 822 assert (count >= 0); | |
| 823 | |
| 824 #ifdef MULE | |
| 825 /* Due to the Mule representation of characters in a buffer, | |
| 826 we can simply search for characters in the range 0 - 127 | |
| 827 directly. For other characters, we do it the "hard" way. | |
| 828 Note that this way works for all characters but the other | |
| 829 way is faster. */ | |
| 830 if (target >= 0200) | |
| 831 { | |
| 832 while (st < lim && count > 0) | |
| 833 { | |
| 867 | 834 if (string_ichar (str, st) == target) |
| 428 | 835 count--; |
| 826 | 836 INC_BYTECOUNT (s, st); |
| 428 | 837 } |
| 838 } | |
| 839 else | |
| 840 #endif | |
| 841 { | |
| 842 while (st < lim && count > 0) | |
| 843 { | |
| 867 | 844 Ibyte *bufptr = (Ibyte *) memchr (itext_n_addr (s, st), |
| 428 | 845 (int) target, lim - st); |
| 846 if (bufptr) | |
| 847 { | |
| 848 count--; | |
| 826 | 849 st = (Bytebpos) (bufptr - s) + 1; |
| 428 | 850 } |
| 851 else | |
| 852 st = lim; | |
| 853 } | |
| 854 } | |
| 855 return st; | |
| 856 } | |
| 857 | |
| 858 /* Like find_next_newline, but returns position before the newline, | |
| 859 not after, and only search up to TO. This isn't just | |
| 860 find_next_newline (...)-1, because you might hit TO. */ | |
| 665 | 861 Charbpos |
| 826 | 862 find_before_next_newline (struct buffer *buf, Charbpos from, Charbpos to, |
| 863 int count) | |
| 428 | 864 { |
| 865 EMACS_INT shortage; | |
| 665 | 866 Charbpos pos = scan_buffer (buf, '\n', from, to, count, &shortage, 1); |
| 428 | 867 |
| 868 if (shortage == 0) | |
| 869 pos--; | |
| 870 | |
| 871 return pos; | |
| 872 } | |
| 873 | |
| 872 | 874 /* This function synched with FSF 21.1 */ |
| 428 | 875 static Lisp_Object |
| 876 skip_chars (struct buffer *buf, int forwardp, int syntaxp, | |
| 877 Lisp_Object string, Lisp_Object lim) | |
| 878 { | |
| 867 | 879 REGISTER Ibyte *p, *pend; |
| 880 REGISTER Ichar c; | |
| 428 | 881 /* We store the first 256 chars in an array here and the rest in |
| 882 a range table. */ | |
| 883 unsigned char fastmap[0400]; | |
| 884 int negate = 0; | |
| 885 REGISTER int i; | |
| 665 | 886 Charbpos limit; |
| 826 | 887 struct syntax_cache *scache; |
| 888 | |
| 428 | 889 if (NILP (lim)) |
| 890 limit = forwardp ? BUF_ZV (buf) : BUF_BEGV (buf); | |
| 891 else | |
| 892 { | |
| 893 CHECK_INT_COERCE_MARKER (lim); | |
| 894 limit = XINT (lim); | |
| 895 | |
| 896 /* In any case, don't allow scan outside bounds of buffer. */ | |
| 897 if (limit > BUF_ZV (buf)) limit = BUF_ZV (buf); | |
| 898 if (limit < BUF_BEGV (buf)) limit = BUF_BEGV (buf); | |
| 899 } | |
| 900 | |
| 901 CHECK_STRING (string); | |
| 902 p = XSTRING_DATA (string); | |
| 903 pend = p + XSTRING_LENGTH (string); | |
| 904 memset (fastmap, 0, sizeof (fastmap)); | |
| 905 | |
| 906 Fclear_range_table (Vskip_chars_range_table); | |
| 907 | |
| 908 if (p != pend && *p == '^') | |
| 909 { | |
| 910 negate = 1; | |
| 911 p++; | |
| 912 } | |
| 913 | |
| 914 /* Find the characters specified and set their elements of fastmap. | |
| 915 If syntaxp, each character counts as itself. | |
| 916 Otherwise, handle backslashes and ranges specially */ | |
| 917 | |
| 918 while (p != pend) | |
| 919 { | |
| 867 | 920 c = itext_ichar (p); |
| 921 INC_IBYTEPTR (p); | |
| 428 | 922 if (syntaxp) |
| 923 { | |
| 924 if (c < 0400 && syntax_spec_code[c] < (unsigned char) Smax) | |
| 925 fastmap[c] = 1; | |
| 926 else | |
| 831 | 927 invalid_argument ("Invalid syntax designator", make_char (c)); |
| 428 | 928 } |
| 929 else | |
| 930 { | |
| 931 if (c == '\\') | |
| 932 { | |
| 933 if (p == pend) break; | |
| 867 | 934 c = itext_ichar (p); |
| 935 INC_IBYTEPTR (p); | |
| 428 | 936 } |
| 937 if (p != pend && *p == '-') | |
| 938 { | |
| 867 | 939 Ichar cend; |
| 428 | 940 |
| 872 | 941 /* Skip over the dash. */ |
| 428 | 942 p++; |
| 943 if (p == pend) break; | |
| 867 | 944 cend = itext_ichar (p); |
| 428 | 945 while (c <= cend && c < 0400) |
| 946 { | |
| 947 fastmap[c] = 1; | |
| 948 c++; | |
| 949 } | |
| 950 if (c <= cend) | |
| 951 Fput_range_table (make_int (c), make_int (cend), Qt, | |
| 952 Vskip_chars_range_table); | |
| 867 | 953 INC_IBYTEPTR (p); |
| 428 | 954 } |
| 955 else | |
| 956 { | |
| 957 if (c < 0400) | |
| 958 fastmap[c] = 1; | |
| 959 else | |
| 960 Fput_range_table (make_int (c), make_int (c), Qt, | |
| 961 Vskip_chars_range_table); | |
| 962 } | |
| 963 } | |
| 964 } | |
| 965 | |
| 872 | 966 /* #### Not in FSF 21.1 */ |
| 428 | 967 if (syntaxp && fastmap['-'] != 0) |
| 968 fastmap[' '] = 1; | |
| 969 | |
| 970 /* If ^ was the first character, complement the fastmap. | |
| 971 We don't complement the range table, however; we just use negate | |
| 972 in the comparisons below. */ | |
| 973 | |
| 974 if (negate) | |
| 647 | 975 for (i = 0; i < (int) (sizeof (fastmap)); i++) |
| 428 | 976 fastmap[i] ^= 1; |
| 977 | |
| 978 { | |
| 665 | 979 Charbpos start_point = BUF_PT (buf); |
| 872 | 980 Charbpos pos = start_point; |
| 981 Charbpos pos_byte = BYTE_BUF_PT (buf); | |
| 428 | 982 |
| 983 if (syntaxp) | |
| 984 { | |
| 872 | 985 scache = setup_buffer_syntax_cache (buf, pos, forwardp ? 1 : -1); |
| 428 | 986 /* All syntax designators are normal chars so nothing strange |
| 987 to worry about */ | |
| 988 if (forwardp) | |
| 989 { | |
| 872 | 990 if (pos < limit) |
| 991 while (fastmap[(unsigned char) | |
| 992 syntax_code_spec | |
| 993 [(int) SYNTAX_FROM_CACHE | |
| 994 (scache, BYTE_BUF_FETCH_CHAR (buf, pos_byte))]]) | |
| 995 { | |
| 996 pos++; | |
| 997 INC_BYTEBPOS (buf, pos_byte); | |
| 879 | 998 if (pos >= limit) |
| 872 | 999 break; |
| 1000 UPDATE_SYNTAX_CACHE_FORWARD (scache, pos); | |
| 1001 } | |
| 428 | 1002 } |
| 1003 else | |
| 1004 { | |
| 872 | 1005 while (pos > limit) |
| 460 | 1006 { |
| 872 | 1007 Charbpos savepos = pos_byte; |
| 1008 pos--; | |
| 1009 DEC_BYTEBPOS (buf, pos_byte); | |
| 1010 UPDATE_SYNTAX_CACHE_BACKWARD (scache, pos); | |
| 1011 if (!fastmap[(unsigned char) | |
| 1012 syntax_code_spec | |
| 1013 [(int) SYNTAX_FROM_CACHE | |
| 1014 (scache, BYTE_BUF_FETCH_CHAR (buf, pos_byte))]]) | |
| 1015 { | |
| 1016 pos++; | |
| 1017 pos_byte = savepos; | |
| 1018 break; | |
| 1019 } | |
| 460 | 1020 } |
| 428 | 1021 } |
| 1022 } | |
| 1023 else | |
| 1024 { | |
| 1025 if (forwardp) | |
| 1026 { | |
| 872 | 1027 while (pos < limit) |
| 428 | 1028 { |
| 872 | 1029 Ichar ch = BYTE_BUF_FETCH_CHAR (buf, pos_byte); |
| 428 | 1030 if ((ch < 0400) ? fastmap[ch] : |
| 1031 (NILP (Fget_range_table (make_int (ch), | |
| 1032 Vskip_chars_range_table, | |
| 1033 Qnil)) | |
| 1034 == negate)) | |
| 872 | 1035 { |
| 1036 pos++; | |
| 1037 INC_BYTEBPOS (buf, pos_byte); | |
| 1038 } | |
| 428 | 1039 else |
| 1040 break; | |
| 1041 } | |
| 1042 } | |
| 1043 else | |
| 1044 { | |
| 872 | 1045 while (pos > limit) |
| 428 | 1046 { |
| 872 | 1047 Charbpos prev_pos_byte = pos_byte; |
| 1048 Ichar ch; | |
| 1049 | |
| 1050 DEC_BYTEBPOS (buf, prev_pos_byte); | |
| 1051 ch = BYTE_BUF_FETCH_CHAR (buf, prev_pos_byte); | |
| 428 | 1052 if ((ch < 0400) ? fastmap[ch] : |
| 1053 (NILP (Fget_range_table (make_int (ch), | |
| 1054 Vskip_chars_range_table, | |
| 1055 Qnil)) | |
| 1056 == negate)) | |
| 872 | 1057 { |
| 1058 pos--; | |
| 1059 pos_byte = prev_pos_byte; | |
| 1060 } | |
| 428 | 1061 else |
| 1062 break; | |
| 1063 } | |
| 1064 } | |
| 1065 } | |
| 1066 QUIT; | |
| 872 | 1067 BOTH_BUF_SET_PT (buf, pos, pos_byte); |
| 428 | 1068 return make_int (BUF_PT (buf) - start_point); |
| 1069 } | |
| 1070 } | |
| 1071 | |
| 1072 DEFUN ("skip-chars-forward", Fskip_chars_forward, 1, 3, 0, /* | |
| 444 | 1073 Move point forward, stopping before a char not in STRING, or at pos LIMIT. |
| 428 | 1074 STRING is like the inside of a `[...]' in a regular expression |
| 1075 except that `]' is never special and `\\' quotes `^', `-' or `\\'. | |
| 1076 Thus, with arg "a-zA-Z", this skips letters stopping before first nonletter. | |
| 1077 With arg "^a-zA-Z", skips nonletters stopping before first letter. | |
| 1078 Returns the distance traveled, either zero or positive. | |
| 1079 | |
| 1080 Optional argument BUFFER defaults to the current buffer. | |
| 1081 */ | |
| 444 | 1082 (string, limit, buffer)) |
| 428 | 1083 { |
| 444 | 1084 return skip_chars (decode_buffer (buffer, 0), 1, 0, string, limit); |
| 428 | 1085 } |
| 1086 | |
| 1087 DEFUN ("skip-chars-backward", Fskip_chars_backward, 1, 3, 0, /* | |
| 444 | 1088 Move point backward, stopping after a char not in STRING, or at pos LIMIT. |
| 428 | 1089 See `skip-chars-forward' for details. |
| 1090 Returns the distance traveled, either zero or negative. | |
| 1091 | |
| 1092 Optional argument BUFFER defaults to the current buffer. | |
| 1093 */ | |
| 444 | 1094 (string, limit, buffer)) |
| 428 | 1095 { |
| 444 | 1096 return skip_chars (decode_buffer (buffer, 0), 0, 0, string, limit); |
| 428 | 1097 } |
| 1098 | |
| 1099 | |
| 1100 DEFUN ("skip-syntax-forward", Fskip_syntax_forward, 1, 3, 0, /* | |
| 1101 Move point forward across chars in specified syntax classes. | |
| 1102 SYNTAX is a string of syntax code characters. | |
| 444 | 1103 Stop before a char whose syntax is not in SYNTAX, or at position LIMIT. |
| 428 | 1104 If SYNTAX starts with ^, skip characters whose syntax is NOT in SYNTAX. |
| 1105 This function returns the distance traveled, either zero or positive. | |
| 1106 | |
| 1107 Optional argument BUFFER defaults to the current buffer. | |
| 1108 */ | |
| 444 | 1109 (syntax, limit, buffer)) |
| 428 | 1110 { |
| 444 | 1111 return skip_chars (decode_buffer (buffer, 0), 1, 1, syntax, limit); |
| 428 | 1112 } |
| 1113 | |
| 1114 DEFUN ("skip-syntax-backward", Fskip_syntax_backward, 1, 3, 0, /* | |
| 1115 Move point backward across chars in specified syntax classes. | |
| 1116 SYNTAX is a string of syntax code characters. | |
| 444 | 1117 Stop on reaching a char whose syntax is not in SYNTAX, or at position LIMIT. |
| 428 | 1118 If SYNTAX starts with ^, skip characters whose syntax is NOT in SYNTAX. |
| 1119 This function returns the distance traveled, either zero or negative. | |
| 1120 | |
| 1121 Optional argument BUFFER defaults to the current buffer. | |
| 1122 */ | |
| 444 | 1123 (syntax, limit, buffer)) |
| 428 | 1124 { |
| 444 | 1125 return skip_chars (decode_buffer (buffer, 0), 0, 1, syntax, limit); |
| 428 | 1126 } |
| 1127 | |
| 1128 | |
| 1129 /* Subroutines of Lisp buffer search functions. */ | |
| 1130 | |
| 1131 static Lisp_Object | |
| 444 | 1132 search_command (Lisp_Object string, Lisp_Object limit, Lisp_Object noerror, |
| 428 | 1133 Lisp_Object count, Lisp_Object buffer, int direction, |
| 1134 int RE, int posix) | |
| 1135 { | |
| 665 | 1136 REGISTER Charbpos np; |
| 1137 Charbpos lim; | |
| 428 | 1138 EMACS_INT n = direction; |
| 1139 struct buffer *buf; | |
| 1140 | |
| 1141 if (!NILP (count)) | |
| 1142 { | |
| 1143 CHECK_INT (count); | |
| 1144 n *= XINT (count); | |
| 1145 } | |
| 1146 | |
| 1147 buf = decode_buffer (buffer, 0); | |
| 1148 CHECK_STRING (string); | |
| 444 | 1149 if (NILP (limit)) |
| 428 | 1150 lim = n > 0 ? BUF_ZV (buf) : BUF_BEGV (buf); |
| 1151 else | |
| 1152 { | |
| 444 | 1153 CHECK_INT_COERCE_MARKER (limit); |
| 1154 lim = XINT (limit); | |
| 428 | 1155 if (n > 0 ? lim < BUF_PT (buf) : lim > BUF_PT (buf)) |
| 563 | 1156 invalid_argument ("Invalid search limit (wrong side of point)", |
| 1157 Qunbound); | |
| 428 | 1158 if (lim > BUF_ZV (buf)) |
| 1159 lim = BUF_ZV (buf); | |
| 1160 if (lim < BUF_BEGV (buf)) | |
| 1161 lim = BUF_BEGV (buf); | |
| 1162 } | |
| 1163 | |
| 1164 np = search_buffer (buf, string, BUF_PT (buf), lim, n, RE, | |
| 1165 (!NILP (buf->case_fold_search) | |
| 446 | 1166 ? XCASE_TABLE_CANON (buf->case_table) |
| 1167 : Qnil), | |
| 428 | 1168 (!NILP (buf->case_fold_search) |
| 446 | 1169 ? XCASE_TABLE_EQV (buf->case_table) |
| 1170 : Qnil), posix); | |
| 428 | 1171 |
| 1172 if (np <= 0) | |
| 1173 { | |
| 444 | 1174 if (NILP (noerror)) |
| 2268 | 1175 { |
| 1176 signal_failure (string); | |
| 1177 RETURN_NOT_REACHED (Qnil); | |
| 1178 } | |
| 444 | 1179 if (!EQ (noerror, Qt)) |
| 428 | 1180 { |
| 1181 if (lim < BUF_BEGV (buf) || lim > BUF_ZV (buf)) | |
| 2500 | 1182 ABORT (); |
| 428 | 1183 BUF_SET_PT (buf, lim); |
| 1184 return Qnil; | |
| 1185 #if 0 /* This would be clean, but maybe programs depend on | |
| 1186 a value of nil here. */ | |
| 1187 np = lim; | |
| 1188 #endif | |
| 1189 } | |
| 1190 else | |
| 1191 return Qnil; | |
| 1192 } | |
| 1193 | |
| 1194 if (np < BUF_BEGV (buf) || np > BUF_ZV (buf)) | |
| 2500 | 1195 ABORT (); |
| 428 | 1196 |
| 1197 BUF_SET_PT (buf, np); | |
| 1198 | |
| 1199 return make_int (np); | |
| 1200 } | |
| 1201 | |
| 1202 static int | |
| 1203 trivial_regexp_p (Lisp_Object regexp) | |
| 1204 { | |
| 1205 Bytecount len = XSTRING_LENGTH (regexp); | |
| 867 | 1206 Ibyte *s = XSTRING_DATA (regexp); |
| 428 | 1207 while (--len >= 0) |
| 1208 { | |
| 1209 switch (*s++) | |
| 1210 { | |
| 1724 | 1211 /* #### howcum ']' doesn't appear here, but ... */ |
| 428 | 1212 case '.': case '*': case '+': case '?': case '[': case '^': case '$': |
| 1213 return 0; | |
| 1214 case '\\': | |
| 1215 if (--len < 0) | |
| 1216 return 0; | |
| 1217 switch (*s++) | |
| 1218 { | |
| 1724 | 1219 /* ... ')' does appear here? ('<' and '>' can appear singly.) */ |
| 1220 /* #### are there other constructs to check? */ | |
| 428 | 1221 case '|': case '(': case ')': case '`': case '\'': case 'b': |
| 1222 case 'B': case '<': case '>': case 'w': case 'W': case 's': | |
| 1724 | 1223 case 'S': case '=': case '{': case '}': |
| 428 | 1224 #ifdef MULE |
| 1225 /* 97/2/25 jhod Added for category matches */ | |
| 1226 case 'c': case 'C': | |
| 1227 #endif /* MULE */ | |
| 1228 case '1': case '2': case '3': case '4': case '5': | |
| 1229 case '6': case '7': case '8': case '9': | |
| 1230 return 0; | |
| 1231 } | |
| 1232 } | |
| 1233 } | |
| 1234 return 1; | |
| 1235 } | |
| 1236 | |
| 1237 /* Search for the n'th occurrence of STRING in BUF, | |
| 665 | 1238 starting at position CHARBPOS and stopping at position BUFLIM, |
| 428 | 1239 treating PAT as a literal string if RE is false or as |
| 1240 a regular expression if RE is true. | |
| 1241 | |
| 1242 If N is positive, searching is forward and BUFLIM must be greater | |
| 665 | 1243 than CHARBPOS. |
| 428 | 1244 If N is negative, searching is backward and BUFLIM must be less |
| 665 | 1245 than CHARBPOS. |
| 428 | 1246 |
| 1247 Returns -x if only N-x occurrences found (x > 0), | |
| 1248 or else the position at the beginning of the Nth occurrence | |
| 1249 (if searching backward) or the end (if searching forward). | |
| 1250 | |
| 1251 POSIX is nonzero if we want full backtracking (POSIX style) | |
| 1252 for this pattern. 0 means backtrack only enough to get a valid match. */ | |
| 665 | 1253 static Charbpos |
| 1254 search_buffer (struct buffer *buf, Lisp_Object string, Charbpos charbpos, | |
| 1255 Charbpos buflim, EMACS_INT n, int RE, Lisp_Object trt, | |
| 446 | 1256 Lisp_Object inverse_trt, int posix) |
| 428 | 1257 { |
| 1258 Bytecount len = XSTRING_LENGTH (string); | |
| 867 | 1259 Ibyte *base_pat = XSTRING_DATA (string); |
| 428 | 1260 REGISTER EMACS_INT i, j; |
| 665 | 1261 Bytebpos p1, p2; |
| 428 | 1262 Bytecount s1, s2; |
| 665 | 1263 Bytebpos pos, lim; |
| 428 | 1264 |
| 853 | 1265 /* Some FSF junk with running_asynch_code, to preserve the match |
| 1266 data. Not necessary because we don't call process filters | |
| 1267 asynchronously (i.e. from within QUIT). */ | |
| 428 | 1268 |
| 1425 | 1269 /* Searching 0 times means noop---don't move, don't touch registers. */ |
| 1270 if (n == 0) | |
| 1271 return charbpos; | |
| 1272 | |
| 428 | 1273 /* Null string is found at starting position. */ |
| 1274 if (len == 0) | |
| 1275 { | |
| 665 | 1276 set_search_regs (buf, charbpos, 0); |
| 1277 return charbpos; | |
| 428 | 1278 } |
| 1279 | |
| 665 | 1280 pos = charbpos_to_bytebpos (buf, charbpos); |
| 1281 lim = charbpos_to_bytebpos (buf, buflim); | |
| 428 | 1282 if (RE && !trivial_regexp_p (string)) |
| 1283 { | |
| 1284 struct re_pattern_buffer *bufp; | |
| 826 | 1285 |
| 1286 bufp = compile_pattern (string, &search_regs, trt, | |
| 1287 wrap_buffer (buf), buf, posix, ERROR_ME); | |
| 428 | 1288 |
| 1289 /* Get pointers and sizes of the two strings | |
| 1290 that make up the visible portion of the buffer. */ | |
| 1291 | |
| 826 | 1292 p1 = BYTE_BUF_BEGV (buf); |
| 1293 p2 = BYTE_BUF_CEILING_OF (buf, p1); | |
| 428 | 1294 s1 = p2 - p1; |
| 826 | 1295 s2 = BYTE_BUF_ZV (buf) - p2; |
| 1296 | |
| 1297 while (n != 0) | |
| 428 | 1298 { |
| 1299 Bytecount val; | |
| 826 | 1300 struct syntax_cache scache_struct; |
| 1301 struct syntax_cache *scache = &scache_struct; | |
| 1302 | |
| 428 | 1303 QUIT; |
| 826 | 1304 /* By making the regex object, regex buffer, and syntax cache |
| 1305 arguments to re_{search,match}{,_2}, we've removed the need to | |
| 1306 do nasty things to deal with regex reentrancy. (See stack | |
| 1307 trace in signal.c for proof that this can happen.) | |
| 1308 | |
| 1309 #### there is still a potential problem with the regex cache -- | |
| 1310 the compiled regex could be overwritten. we'd need 20-fold | |
| 1311 reentrancy, though. Fix this. */ | |
| 1312 | |
| 428 | 1313 val = re_search_2 (bufp, |
| 826 | 1314 (char *) BYTE_BUF_BYTE_ADDRESS (buf, p1), s1, |
| 1315 (char *) BYTE_BUF_BYTE_ADDRESS (buf, p2), s2, | |
| 1316 pos - BYTE_BUF_BEGV (buf), lim - pos, &search_regs, | |
| 1317 n > 0 ? lim - BYTE_BUF_BEGV (buf) : | |
| 1318 pos - BYTE_BUF_BEGV (buf), wrap_buffer (buf), | |
| 1319 buf, scache); | |
| 428 | 1320 |
| 1321 if (val == -2) | |
| 1322 { | |
| 1323 matcher_overflow (); | |
| 1324 } | |
| 1325 if (val >= 0) | |
| 1326 { | |
| 1327 int num_regs = search_regs.num_regs; | |
| 826 | 1328 j = BYTE_BUF_BEGV (buf); |
| 428 | 1329 for (i = 0; i < num_regs; i++) |
| 1330 if (search_regs.start[i] >= 0) | |
| 1331 { | |
| 1332 search_regs.start[i] += j; | |
| 1333 search_regs.end[i] += j; | |
| 1334 } | |
| 793 | 1335 last_thing_searched = wrap_buffer (buf); |
| 428 | 1336 /* Set pos to the new position. */ |
| 826 | 1337 pos = n > 0 ? search_regs.end[0] : search_regs.start[0]; |
| 428 | 1338 fixup_search_regs_for_buffer (buf); |
| 665 | 1339 /* And charbpos too. */ |
| 826 | 1340 charbpos = n > 0 ? search_regs.end[0] : search_regs.start[0]; |
| 428 | 1341 } |
| 1342 else | |
| 826 | 1343 return (n > 0 ? 0 - n : n); |
| 1344 if (n > 0) n--; else n++; | |
| 428 | 1345 } |
| 665 | 1346 return charbpos; |
| 428 | 1347 } |
| 1348 else /* non-RE case */ | |
| 1349 { | |
| 446 | 1350 int charset_base = -1; |
| 1351 int boyer_moore_ok = 1; | |
| 2367 | 1352 Ibyte *patbuf = alloca_ibytes (len * MAX_ICHAR_LEN); |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1353 Ibyte *pat = patbuf; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1354 |
| 446 | 1355 #ifdef MULE |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1356 int entirely_one_byte_p = buf->text->entirely_one_byte_p; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1357 int nothing_greater_than_0xff = |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1358 buf->text->num_8_bit_fixed_chars == BUF_Z(buf) - BUF_BEG (buf); |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1359 |
| 446 | 1360 while (len > 0) |
| 1361 { | |
| 867 | 1362 Ibyte tmp_str[MAX_ICHAR_LEN]; |
| 1363 Ichar c, translated, inverse; | |
| 446 | 1364 Bytecount orig_bytelen, new_bytelen, inv_bytelen; |
| 1365 | |
| 1366 /* If we got here and the RE flag is set, it's because | |
| 1367 we're dealing with a regexp known to be trivial, so the | |
| 1368 backslash just quotes the next character. */ | |
| 1369 if (RE && *base_pat == '\\') | |
| 1370 { | |
| 1371 len--; | |
| 1372 base_pat++; | |
| 1373 } | |
| 867 | 1374 c = itext_ichar (base_pat); |
| 446 | 1375 translated = TRANSLATE (trt, c); |
| 1376 inverse = TRANSLATE (inverse_trt, c); | |
| 1377 | |
| 867 | 1378 orig_bytelen = itext_ichar_len (base_pat); |
| 1379 inv_bytelen = set_itext_ichar (tmp_str, inverse); | |
| 1380 new_bytelen = set_itext_ichar (tmp_str, translated); | |
| 446 | 1381 |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1382 if (boyer_moore_ok |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1383 /* Only do the Boyer-Moore check for characters needing |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1384 translation. */ |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1385 && (translated != c || inverse != c)) |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1386 { |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1387 Ichar starting_c = c; |
|
4421
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1388 int charset_base_code, checked = 0; |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1389 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1390 do |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1391 { |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1392 c = TRANSLATE (inverse_trt, c); |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1393 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1394 /* If a character cannot occur in the buffer, ignore |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1395 it. */ |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1396 if (c > 0x7F && entirely_one_byte_p) |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1397 continue; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1398 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1399 if (c > 0xFF && nothing_greater_than_0xff) |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1400 continue; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1401 |
|
4421
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1402 checked = 1; |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1403 |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1404 if (-1 == charset_base) /* No charset yet specified. */ |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1405 { |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1406 /* Keep track of which charset and character set row |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1407 contains the characters that need translation. |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1408 |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1409 Zero out the bits corresponding to the last |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1410 byte. */ |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1411 charset_base = c & ~ICHAR_FIELD3_MASK; |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1412 } |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1413 else |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1414 { |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1415 charset_base_code = c & ~ICHAR_FIELD3_MASK; |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1416 |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1417 if (charset_base_code != charset_base) |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1418 { |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1419 /* If two different rows, or two different |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1420 charsets, appear, needing non-ASCII |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1421 translation, then we cannot use boyer_moore |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1422 search. See the comment at the head of |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1423 boyer_moore(). */ |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1424 boyer_moore_ok = 0; |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1425 break; |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1426 } |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1427 } |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1428 } while (c != starting_c); |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1429 |
|
4421
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1430 if (!checked) |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1431 { |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1432 #ifdef DEBUG_XEMACS |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1433 if (debug_xemacs_searches) |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1434 { |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1435 Lisp_Symbol *sym = XSYMBOL (Qsearch_algorithm_used); |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1436 sym->value = Qnil; |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1437 } |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1438 #endif |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1439 /* The "continue" clauses were used above, for every |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1440 translation of the character. As such, this character |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1441 is not to be found in the buffer and neither is the |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1442 string as a whole. Return immediately; also avoid |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1443 triggering the assertion a few lines down. */ |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1444 return n > 0 ? -n : n; |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1445 } |
|
69b803c646cd
Fail searches immediately if searching for non-representable characters.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4414
diff
changeset
|
1446 |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1447 if (boyer_moore_ok && charset_base != -1 && |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1448 charset_base != (translated & ~ICHAR_FIELD3_MASK)) |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1449 { |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1450 /* In the rare event that the CANON entry for this |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1451 character is not in the desired set, choose one |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1452 that is, from the equivalence set. It doesn't much |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1453 matter which. */ |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1454 Ichar starting_ch = translated; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1455 do |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1456 { |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1457 translated = TRANSLATE (inverse_trt, translated); |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1458 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1459 if (charset_base == (translated & ~ICHAR_FIELD3_MASK)) |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1460 break; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1461 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1462 } while (starting_ch != translated); |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1463 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1464 assert (starting_ch != translated); |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1465 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1466 new_bytelen = set_itext_ichar (tmp_str, translated); |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1467 } |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1468 } |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1469 |
| 446 | 1470 memcpy (pat, tmp_str, new_bytelen); |
| 1471 pat += new_bytelen; | |
| 1472 base_pat += orig_bytelen; | |
| 1473 len -= orig_bytelen; | |
| 1474 } | |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1475 |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1476 if (-1 == charset_base) |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1477 { |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1478 charset_base = 'a' & ~ICHAR_FIELD3_MASK; /* Default to ASCII. */ |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1479 } |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1480 |
| 446 | 1481 #else /* not MULE */ |
| 1482 while (--len >= 0) | |
| 1483 { | |
| 1484 /* If we got here and the RE flag is set, it's because | |
| 1485 we're dealing with a regexp known to be trivial, so the | |
| 1486 backslash just quotes the next character. */ | |
| 1487 if (RE && *base_pat == '\\') | |
| 1488 { | |
| 1489 len--; | |
| 1490 base_pat++; | |
| 1491 } | |
| 1492 *pat++ = TRANSLATE (trt, *base_pat++); | |
| 1493 } | |
| 1494 #endif /* MULE */ | |
| 1495 len = pat - patbuf; | |
| 1496 pat = base_pat = patbuf; | |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1497 |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1498 #ifdef DEBUG_XEMACS |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1499 if (debug_xemacs_searches) |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1500 { |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1501 Lisp_Symbol *sym = XSYMBOL (Qsearch_algorithm_used); |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1502 sym->value = boyer_moore_ok ? Qboyer_moore : Qsimple_search; |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1503 } |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1504 #endif |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1505 |
| 446 | 1506 if (boyer_moore_ok) |
| 1507 return boyer_moore (buf, base_pat, len, pos, lim, n, | |
| 1508 trt, inverse_trt, charset_base); | |
| 1509 else | |
| 1510 return simple_search (buf, base_pat, len, pos, lim, n, trt); | |
| 1511 } | |
| 1512 } | |
| 1513 | |
| 826 | 1514 /* Do a simple string search N times for the string PAT, whose length is |
| 1515 LEN/LEN_BYTE, from buffer position POS until LIM. TRT is the | |
| 1516 translation table. | |
| 446 | 1517 |
| 1518 Return the character position where the match is found. | |
| 1519 Otherwise, if M matches remained to be found, return -M. | |
| 1520 | |
| 1521 This kind of search works regardless of what is in PAT and | |
| 1522 regardless of what is in TRT. It is used in cases where | |
| 1523 boyer_moore cannot work. */ | |
| 1524 | |
| 665 | 1525 static Charbpos |
| 867 | 1526 simple_search (struct buffer *buf, Ibyte *base_pat, Bytecount len, |
| 826 | 1527 Bytebpos pos, Bytebpos lim, EMACS_INT n, Lisp_Object trt) |
| 446 | 1528 { |
| 1529 int forward = n > 0; | |
| 1530 Bytecount buf_len = 0; /* Shut up compiler. */ | |
| 1531 | |
| 826 | 1532 if (lim > pos) |
| 446 | 1533 while (n > 0) |
| 428 | 1534 { |
| 446 | 1535 while (1) |
| 428 | 1536 { |
| 826 | 1537 Bytecount this_len = len; |
| 1538 Bytebpos this_pos = pos; | |
| 867 | 1539 Ibyte *p = base_pat; |
| 826 | 1540 if (pos >= lim) |
| 446 | 1541 goto stop; |
| 1542 | |
| 1543 while (this_len > 0) | |
| 1544 { | |
| 867 | 1545 Ichar pat_ch, buf_ch; |
| 446 | 1546 Bytecount pat_len; |
| 1547 | |
| 867 | 1548 pat_ch = itext_ichar (p); |
| 826 | 1549 buf_ch = BYTE_BUF_FETCH_CHAR (buf, this_pos); |
| 446 | 1550 |
| 1551 buf_ch = TRANSLATE (trt, buf_ch); | |
| 1552 | |
| 1553 if (buf_ch != pat_ch) | |
| 1554 break; | |
| 1555 | |
| 867 | 1556 pat_len = itext_ichar_len (p); |
| 446 | 1557 p += pat_len; |
| 1558 this_len -= pat_len; | |
| 826 | 1559 INC_BYTEBPOS (buf, this_pos); |
| 446 | 1560 } |
| 1561 if (this_len == 0) | |
| 428 | 1562 { |
| 826 | 1563 buf_len = this_pos - pos; |
| 1564 pos = this_pos; | |
| 446 | 1565 break; |
| 428 | 1566 } |
| 826 | 1567 INC_BYTEBPOS (buf, pos); |
| 428 | 1568 } |
| 446 | 1569 n--; |
| 1570 } | |
| 1571 else | |
|
4322
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1572 { |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1573 /* If lim < len, then there are too few buffer positions to hold the |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1574 pattern between the beginning of the buffer and lim. Adjust to |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1575 ensure pattern fits. If we don't do this, we can assert in the |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1576 DEC_BYTEBPOS below. */ |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1577 if (lim < len) |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1578 lim = len; |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1579 while (n < 0) |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1580 { |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1581 while (1) |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1582 { |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1583 Bytecount this_len = len; |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1584 Bytebpos this_pos = pos; |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1585 Ibyte *p; |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1586 if (pos <= lim) |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1587 goto stop; |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1588 p = base_pat + len; |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1589 |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1590 while (this_len > 0) |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1591 { |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1592 Ichar pat_ch, buf_ch; |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1593 |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1594 DEC_IBYTEPTR (p); |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1595 DEC_BYTEBPOS (buf, this_pos); |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1596 pat_ch = itext_ichar (p); |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1597 buf_ch = BYTE_BUF_FETCH_CHAR (buf, this_pos); |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1598 |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1599 buf_ch = TRANSLATE (trt, buf_ch); |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1600 |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1601 if (buf_ch != pat_ch) |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1602 break; |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1603 |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1604 this_len -= itext_ichar_len (p); |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1605 } |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1606 if (this_len == 0) |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1607 { |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1608 buf_len = pos - this_pos; |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1609 pos = this_pos; |
| 446 | 1610 break; |
|
4322
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1611 } |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1612 DEC_BYTEBPOS (buf, pos); |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1613 } |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1614 n++; |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1615 } |
|
f70e56bb52a7
src/search.c (simple_search): Fix underrun in reverse search.
Stephen J. Turnbull <stephen@xemacs.org>
parents:
4199
diff
changeset
|
1616 } |
| 446 | 1617 stop: |
| 1618 if (n == 0) | |
| 1619 { | |
| 665 | 1620 Charbpos beg, end, retval; |
| 446 | 1621 if (forward) |
| 1622 { | |
| 826 | 1623 beg = bytebpos_to_charbpos (buf, pos - buf_len); |
| 1624 retval = end = bytebpos_to_charbpos (buf, pos); | |
| 446 | 1625 } |
| 1626 else | |
| 428 | 1627 { |
| 826 | 1628 retval = beg = bytebpos_to_charbpos (buf, pos); |
| 1629 end = bytebpos_to_charbpos (buf, pos + buf_len); | |
| 428 | 1630 } |
| 446 | 1631 set_search_regs (buf, beg, end - beg); |
| 1632 | |
| 1633 return retval; | |
| 1634 } | |
| 1635 else if (n > 0) | |
| 1636 return -n; | |
| 1637 else | |
| 1638 return n; | |
| 1639 } | |
| 1640 | |
| 1641 /* Do Boyer-Moore search N times for the string PAT, | |
| 1642 whose length is LEN/LEN_BYTE, | |
| 1643 from buffer position POS/POS_BYTE until LIM/LIM_BYTE. | |
| 1644 DIRECTION says which direction we search in. | |
| 1645 TRT and INVERSE_TRT are translation tables. | |
| 1646 | |
| 1647 This kind of search works if all the characters in PAT that have | |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1648 (non-ASCII) translation are the same aside from the last byte. This |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1649 makes it possible to translate just the last byte of a character, and do |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1650 so after just a simple test of the context. |
| 446 | 1651 |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1652 If that criterion is not satisfied, do not call this function. You will |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1653 get an assertion failure. */ |
| 446 | 1654 |
| 665 | 1655 static Charbpos |
| 867 | 1656 boyer_moore (struct buffer *buf, Ibyte *base_pat, Bytecount len, |
| 665 | 1657 Bytebpos pos, Bytebpos lim, EMACS_INT n, Lisp_Object trt, |
| 2333 | 1658 Lisp_Object inverse_trt, int USED_IF_MULE (charset_base)) |
| 446 | 1659 { |
| 1660 /* #### Someone really really really needs to comment the workings | |
| 1661 of this junk somewhat better. | |
| 1662 | |
| 1663 BTW "BM" stands for Boyer-Moore, which is one of the standard | |
| 1664 string-searching algorithms. It's the best string-searching | |
| 1665 algorithm out there, provided that: | |
| 1666 | |
| 1667 a) You're not fazed by algorithm complexity. (Rabin-Karp, which | |
| 1668 uses hashing, is much much easier to code but not as fast.) | |
| 1669 b) You can freely move backwards in the string that you're | |
| 1670 searching through. | |
| 1671 | |
| 1672 As the comment below tries to explain (but garbles in typical | |
| 1673 programmer-ese), the idea is that you don't have to do a | |
| 1674 string match at every successive position in the text. For | |
| 1675 example, let's say the pattern is "a very long string". We | |
| 1676 compare the last character in the string (`g') with the | |
| 1677 corresponding character in the text. If it mismatches, and | |
| 1678 it is, say, `z', then we can skip forward by the entire | |
| 1679 length of the pattern because `z' does not occur anywhere | |
| 1680 in the pattern. If the mismatching character does occur | |
| 1681 in the pattern, we can usually still skip forward by more | |
| 1682 than one: e.g. if it is `l', then we can skip forward | |
| 1683 by the length of the substring "ong string" -- i.e. the | |
| 1684 largest end section of the pattern that does not contain | |
| 1685 the mismatched character. So what we do is compute, for | |
| 1686 each possible character, the distance we can skip forward | |
| 1687 (the "stride") and use it in the string matching. This | |
| 1688 is what the BM_tab holds. */ | |
| 1689 REGISTER EMACS_INT *BM_tab; | |
| 1690 EMACS_INT *BM_tab_base; | |
| 1691 REGISTER Bytecount dirlen; | |
| 1692 EMACS_INT infinity; | |
| 665 | 1693 Bytebpos limit; |
| 446 | 1694 Bytecount stride_for_teases = 0; |
| 1695 REGISTER EMACS_INT i, j; | |
| 867 | 1696 Ibyte *pat, *pat_end; |
| 1697 REGISTER Ibyte *cursor, *p_limit, *ptr2; | |
| 1698 Ibyte simple_translate[0400]; | |
| 446 | 1699 REGISTER int direction = ((n > 0) ? 1 : -1); |
| 1700 #ifdef MULE | |
| 867 | 1701 Ibyte translate_prev_byte = 0; |
| 1702 Ibyte translate_anteprev_byte = 0; | |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1703 /* These need to be rethought in the event that the internal format |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1704 changes, or in the event that num_8_bit_fixed_chars disappears |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1705 (entirely_one_byte_p can be trivially worked out by checking is the |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1706 byte count equal to the char count.) */ |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1707 int buffer_entirely_one_byte_p = buf->text->entirely_one_byte_p; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1708 int buffer_nothing_greater_than_0xff = |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1709 buf->text->num_8_bit_fixed_chars == BUF_Z(buf) - BUF_BEG (buf); |
| 446 | 1710 #endif |
| 1711 #ifdef C_ALLOCA | |
| 1712 EMACS_INT BM_tab_space[0400]; | |
| 1713 BM_tab = &BM_tab_space[0]; | |
| 1714 #else | |
| 1715 BM_tab = alloca_array (EMACS_INT, 256); | |
| 1716 #endif | |
| 1717 | |
| 1718 /* The general approach is that we are going to maintain that we | |
| 1719 know the first (closest to the present position, in whatever | |
| 1720 direction we're searching) character that could possibly be | |
| 1721 the last (furthest from present position) character of a | |
| 1722 valid match. We advance the state of our knowledge by | |
| 1723 looking at that character and seeing whether it indeed | |
| 1724 matches the last character of the pattern. If it does, we | |
| 1725 take a closer look. If it does not, we move our pointer (to | |
| 1726 putative last characters) as far as is logically possible. | |
| 1727 This amount of movement, which I call a stride, will be the | |
| 1728 length of the pattern if the actual character appears nowhere | |
| 1729 in the pattern, otherwise it will be the distance from the | |
| 1730 last occurrence of that character to the end of the pattern. | |
| 1731 As a coding trick, an enormous stride is coded into the table | |
| 1732 for characters that match the last character. This allows | |
| 1733 use of only a single test, a test for having gone past the | |
| 1734 end of the permissible match region, to test for both | |
| 1735 possible matches (when the stride goes past the end | |
| 1736 immediately) and failure to match (where you get nudged past | |
| 1737 the end one stride at a time). | |
| 1738 | |
| 1739 Here we make a "mickey mouse" BM table. The stride of the | |
| 1740 search is determined only by the last character of the | |
| 1741 putative match. If that character does not match, we will | |
| 1742 stride the proper distance to propose a match that | |
| 1743 superimposes it on the last instance of a character that | |
| 1744 matches it (per trt), or misses it entirely if there is | |
| 1745 none. */ | |
| 1746 | |
| 1747 dirlen = len * direction; | |
| 1748 infinity = dirlen - (lim + pos + len + len) * direction; | |
| 1749 /* Record position after the end of the pattern. */ | |
| 1750 pat_end = base_pat + len; | |
| 1751 if (direction < 0) | |
| 1752 base_pat = pat_end - 1; | |
| 1753 BM_tab_base = BM_tab; | |
| 1754 BM_tab += 0400; | |
| 1755 j = dirlen; /* to get it in a register */ | |
| 1756 /* A character that does not appear in the pattern induces a | |
| 1757 stride equal to the pattern length. */ | |
| 1758 while (BM_tab_base != BM_tab) | |
| 1759 { | |
| 1760 *--BM_tab = j; | |
| 1761 *--BM_tab = j; | |
| 1762 *--BM_tab = j; | |
| 1763 *--BM_tab = j; | |
| 1764 } | |
| 1765 /* We use this for translation, instead of TRT itself. We | |
| 1766 fill this in to handle the characters that actually occur | |
| 1767 in the pattern. Others don't matter anyway! */ | |
| 1768 xzero (simple_translate); | |
| 1769 for (i = 0; i < 0400; i++) | |
| 867 | 1770 simple_translate[i] = (Ibyte) i; |
| 446 | 1771 i = 0; |
| 1425 | 1772 |
| 446 | 1773 while (i != infinity) |
| 1774 { | |
| 867 | 1775 Ibyte *ptr = base_pat + i; |
| 446 | 1776 i += direction; |
| 1777 if (i == dirlen) | |
| 1778 i = infinity; | |
| 1779 if (!NILP (trt)) | |
| 428 | 1780 { |
| 446 | 1781 #ifdef MULE |
| 867 | 1782 Ichar ch, untranslated; |
| 446 | 1783 int this_translated = 1; |
| 1784 | |
| 1785 /* Is *PTR the last byte of a character? */ | |
| 867 | 1786 if (pat_end - ptr == 1 || ibyte_first_byte_p (ptr[1])) |
| 428 | 1787 { |
| 867 | 1788 Ibyte *charstart = ptr; |
| 1789 while (!ibyte_first_byte_p (*charstart)) | |
| 446 | 1790 charstart--; |
| 867 | 1791 untranslated = itext_ichar (charstart); |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1792 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1793 ch = TRANSLATE (trt, untranslated); |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1794 if (!ibyte_first_byte_p (*ptr)) |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1795 { |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1796 translate_prev_byte = ptr[-1]; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1797 if (!ibyte_first_byte_p (translate_prev_byte)) |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1798 translate_anteprev_byte = ptr[-2]; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1799 } |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1800 |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1801 if (ch != untranslated && /* Was translation done? */ |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1802 charset_base != (ch & ~ICHAR_FIELD3_MASK)) |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1803 { |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1804 /* In the very rare event that the CANON entry for this |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1805 character is not in the desired set, choose one that |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1806 is, from the equivalence set. It doesn't much matter |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1807 which, since we're building our own cheesy equivalence |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1808 table instead of using that belonging to the case |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1809 table directly. |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1810 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1811 We can get here if search_buffer has worked out that |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1812 the buffer is entirely single width. */ |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1813 Ichar starting_ch = ch; |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1814 int count = 0; |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1815 do |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1816 { |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1817 ch = TRANSLATE (inverse_trt, ch); |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1818 if (charset_base == (ch & ~ICHAR_FIELD3_MASK)) |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1819 break; |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1820 ++count; |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1821 } while (starting_ch != ch); |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1822 |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1823 /* If starting_ch is equal to ch (and count is not one, |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1824 which means no translation is necessary), the case |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1825 table is corrupt. (Any mapping in the canon table |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1826 should be reflected in the equivalence table, and we |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1827 know from the canon table that untranslated maps to |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1828 starting_ch and that untranslated has the correct value |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1829 for charset_base.) */ |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
1830 assert (1 == count || starting_ch != ch); |
| 446 | 1831 } |
| 428 | 1832 } |
| 1833 else | |
| 1834 { | |
| 446 | 1835 ch = *ptr; |
| 1836 this_translated = 0; | |
| 1837 } | |
| 1838 if (ch > 0400) | |
| 1839 j = ((unsigned char) ch | 0200); | |
| 1840 else | |
| 1841 j = (unsigned char) ch; | |
| 1842 | |
| 1843 if (i == infinity) | |
| 1844 stride_for_teases = BM_tab[j]; | |
| 1845 BM_tab[j] = dirlen - i; | |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1846 /* A translation table is accompanied by its inverse -- see |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1847 comment in casetab.c. */ |
| 446 | 1848 if (this_translated) |
| 1849 { | |
| 867 | 1850 Ichar starting_ch = ch; |
| 446 | 1851 EMACS_INT starting_j = j; |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1852 do |
| 446 | 1853 { |
| 1854 ch = TRANSLATE (inverse_trt, ch); | |
|
4407
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1855 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1856 if (ch > 0x7F && buffer_entirely_one_byte_p) |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1857 continue; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1858 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1859 if (ch > 0xFF && buffer_nothing_greater_than_0xff) |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1860 continue; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1861 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1862 if (ch > 0400) |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1863 j = ((unsigned char) ch | 0200); |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1864 else |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1865 j = (unsigned char) ch; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1866 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1867 /* For all the characters that map into CH, set up |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1868 simple_translate to map the last byte into |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1869 STARTING_J. */ |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1870 simple_translate[j] = (Ibyte) starting_j; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1871 BM_tab[j] = dirlen - i; |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1872 |
|
4ee73bbe4f8e
Always use boyer_moore in ASCII or Latin-1 buffers with ASCII search strings.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4322
diff
changeset
|
1873 } while (ch != starting_ch); |
| 446 | 1874 } |
| 1875 #else | |
| 1876 EMACS_INT k; | |
| 1877 j = *ptr; | |
| 1878 k = (j = TRANSLATE (trt, j)); | |
| 1879 if (i == infinity) | |
| 1880 stride_for_teases = BM_tab[j]; | |
| 1881 BM_tab[j] = dirlen - i; | |
| 1882 /* A translation table is accompanied by its inverse -- | |
| 826 | 1883 see comment in casetab.c. */ |
| 446 | 1884 while ((j = TRANSLATE (inverse_trt, j)) != k) |
| 1885 { | |
| 867 | 1886 simple_translate[j] = (Ibyte) k; |
| 428 | 1887 BM_tab[j] = dirlen - i; |
| 1888 } | |
| 446 | 1889 #endif |
| 1890 } | |
| 1891 else | |
| 1892 { | |
| 1893 j = *ptr; | |
| 1894 | |
| 1895 if (i == infinity) | |
| 1896 stride_for_teases = BM_tab[j]; | |
| 1897 BM_tab[j] = dirlen - i; | |
| 428 | 1898 } |
| 446 | 1899 /* stride_for_teases tells how much to stride if we get a |
| 1900 match on the far character but are subsequently | |
| 1901 disappointed, by recording what the stride would have been | |
| 1902 for that character if the last character had been | |
| 1903 different. */ | |
| 1904 } | |
| 1905 infinity = dirlen - infinity; | |
| 1906 pos += dirlen - ((direction > 0) ? direction : 0); | |
| 1907 /* loop invariant - pos points at where last char (first char if | |
| 1908 reverse) of pattern would align in a possible match. */ | |
| 1909 while (n != 0) | |
| 1910 { | |
| 665 | 1911 Bytebpos tail_end; |
| 867 | 1912 Ibyte *tail_end_ptr; |
| 446 | 1913 /* It's been reported that some (broken) compiler thinks |
| 1914 that Boolean expressions in an arithmetic context are | |
| 1915 unsigned. Using an explicit ?1:0 prevents this. */ | |
| 1916 if ((lim - pos - ((direction > 0) ? 1 : 0)) * direction < 0) | |
| 1917 return n * (0 - direction); | |
| 1918 /* First we do the part we can by pointers (maybe | |
| 1919 nothing) */ | |
| 1920 QUIT; | |
| 1921 pat = base_pat; | |
| 1922 limit = pos - dirlen + direction; | |
| 1923 /* XEmacs change: definitions of CEILING_OF and FLOOR_OF | |
| 1924 have changed. See buffer.h. */ | |
| 1925 limit = ((direction > 0) | |
| 826 | 1926 ? BYTE_BUF_CEILING_OF (buf, limit) - 1 |
| 1927 : BYTE_BUF_FLOOR_OF (buf, limit + 1)); | |
| 446 | 1928 /* LIMIT is now the last (not beyond-last!) value POS can |
| 1929 take on without hitting edge of buffer or the gap. */ | |
| 1930 limit = ((direction > 0) | |
| 1931 ? min (lim - 1, min (limit, pos + 20000)) | |
| 1932 : max (lim, max (limit, pos - 20000))); | |
| 826 | 1933 tail_end = BYTE_BUF_CEILING_OF (buf, pos); |
| 1934 tail_end_ptr = BYTE_BUF_BYTE_ADDRESS (buf, tail_end); | |
| 446 | 1935 |
| 1936 if ((limit - pos) * direction > 20) | |
| 428 | 1937 { |
| 826 | 1938 /* We have to be careful because the code can generate addresses |
| 1939 that don't point to the beginning of characters. */ | |
| 1940 p_limit = BYTE_BUF_BYTE_ADDRESS_NO_VERIFY (buf, limit); | |
| 1941 ptr2 = (cursor = BYTE_BUF_BYTE_ADDRESS_NO_VERIFY (buf, pos)); | |
| 446 | 1942 /* In this loop, pos + cursor - ptr2 is the surrogate |
| 1943 for pos */ | |
| 1944 while (1) /* use one cursor setting as long as i can */ | |
| 1945 { | |
| 1946 if (direction > 0) /* worth duplicating */ | |
| 1947 { | |
| 1948 /* Use signed comparison if appropriate to make | |
| 1949 cursor+infinity sure to be > p_limit. | |
| 1950 Assuming that the buffer lies in a range of | |
| 1951 addresses that are all "positive" (as ints) | |
| 1952 or all "negative", either kind of comparison | |
| 1953 will work as long as we don't step by | |
| 1954 infinity. So pick the kind that works when | |
| 1955 we do step by infinity. */ | |
| 1956 if ((EMACS_INT) (p_limit + infinity) > | |
| 1957 (EMACS_INT) p_limit) | |
| 1958 while ((EMACS_INT) cursor <= | |
| 1959 (EMACS_INT) p_limit) | |
| 1960 cursor += BM_tab[*cursor]; | |
| 1961 else | |
| 1962 while ((EMACS_UINT) cursor <= | |
| 1963 (EMACS_UINT) p_limit) | |
| 1964 cursor += BM_tab[*cursor]; | |
| 1965 } | |
| 1966 else | |
| 1967 { | |
| 1968 if ((EMACS_INT) (p_limit + infinity) < | |
| 1969 (EMACS_INT) p_limit) | |
| 1970 while ((EMACS_INT) cursor >= | |
| 1971 (EMACS_INT) p_limit) | |
| 1972 cursor += BM_tab[*cursor]; | |
| 1973 else | |
| 1974 while ((EMACS_UINT) cursor >= | |
| 1975 (EMACS_UINT) p_limit) | |
| 1976 cursor += BM_tab[*cursor]; | |
| 1977 } | |
| 1978 /* If you are here, cursor is beyond the end of the | |
| 1979 searched region. This can happen if you match on | |
| 1980 the far character of the pattern, because the | |
| 1981 "stride" of that character is infinity, a number | |
| 1982 able to throw you well beyond the end of the | |
| 1983 search. It can also happen if you fail to match | |
| 1984 within the permitted region and would otherwise | |
| 1985 try a character beyond that region */ | |
| 1986 if ((cursor - p_limit) * direction <= len) | |
| 1987 break; /* a small overrun is genuine */ | |
| 1988 cursor -= infinity; /* large overrun = hit */ | |
| 1989 i = dirlen - direction; | |
| 1990 if (!NILP (trt)) | |
| 1991 { | |
| 1992 while ((i -= direction) + direction != 0) | |
| 1993 { | |
| 1994 #ifdef MULE | |
| 867 | 1995 Ichar ch; |
| 446 | 1996 cursor -= direction; |
| 1997 /* Translate only the last byte of a character. */ | |
| 1998 if ((cursor == tail_end_ptr | |
| 867 | 1999 || ibyte_first_byte_p (cursor[1])) |
| 2000 && (ibyte_first_byte_p (cursor[0]) | |
| 446 | 2001 || (translate_prev_byte == cursor[-1] |
| 867 | 2002 && (ibyte_first_byte_p (translate_prev_byte) |
| 446 | 2003 || translate_anteprev_byte == cursor[-2])))) |
| 2004 ch = simple_translate[*cursor]; | |
| 2005 else | |
| 2006 ch = *cursor; | |
| 2007 if (pat[i] != ch) | |
| 2008 break; | |
| 2009 #else | |
| 2010 if (pat[i] != TRANSLATE (trt, *(cursor -= direction))) | |
| 2011 break; | |
| 2012 #endif | |
| 2013 } | |
| 2014 } | |
| 2015 else | |
| 2016 { | |
| 2017 while ((i -= direction) + direction != 0) | |
| 2018 if (pat[i] != *(cursor -= direction)) | |
| 2019 break; | |
| 2020 } | |
| 2021 cursor += dirlen - i - direction; /* fix cursor */ | |
| 2022 if (i + direction == 0) | |
| 2023 { | |
| 2024 cursor -= direction; | |
| 2025 | |
| 2026 { | |
| 665 | 2027 Bytebpos bytstart = (pos + cursor - ptr2 + |
| 446 | 2028 ((direction > 0) |
| 2029 ? 1 - len : 0)); | |
| 665 | 2030 Charbpos bufstart = bytebpos_to_charbpos (buf, bytstart); |
| 2031 Charbpos bufend = bytebpos_to_charbpos (buf, bytstart + len); | |
| 446 | 2032 |
| 2033 set_search_regs (buf, bufstart, bufend - bufstart); | |
| 2034 } | |
| 2035 | |
| 2036 if ((n -= direction) != 0) | |
| 2037 cursor += dirlen; /* to resume search */ | |
| 2038 else | |
| 2039 return ((direction > 0) | |
| 2040 ? search_regs.end[0] : search_regs.start[0]); | |
| 2041 } | |
| 2042 else | |
| 2043 cursor += stride_for_teases; /* <sigh> we lose - */ | |
| 2044 } | |
| 2045 pos += cursor - ptr2; | |
| 2046 } | |
| 2047 else | |
| 2048 /* Now we'll pick up a clump that has to be done the hard | |
| 2049 way because it covers a discontinuity */ | |
| 2050 { | |
| 428 | 2051 /* XEmacs change: definitions of CEILING_OF and FLOOR_OF |
| 2052 have changed. See buffer.h. */ | |
| 2053 limit = ((direction > 0) | |
| 826 | 2054 ? BYTE_BUF_CEILING_OF (buf, pos - dirlen + 1) - 1 |
| 2055 : BYTE_BUF_FLOOR_OF (buf, pos - dirlen)); | |
| 428 | 2056 limit = ((direction > 0) |
| 446 | 2057 ? min (limit + len, lim - 1) |
| 2058 : max (limit - len, lim)); | |
| 2059 /* LIMIT is now the last value POS can have | |
| 2060 and still be valid for a possible match. */ | |
| 2061 while (1) | |
| 428 | 2062 { |
| 446 | 2063 /* This loop can be coded for space rather than |
| 2064 speed because it will usually run only once. | |
| 2065 (the reach is at most len + 21, and typically | |
| 2066 does not exceed len) */ | |
| 2067 while ((limit - pos) * direction >= 0) | |
| 826 | 2068 /* *not* BYTE_BUF_FETCH_CHAR. We are working here |
| 446 | 2069 with bytes, not characters. */ |
| 826 | 2070 pos += BM_tab[*BYTE_BUF_BYTE_ADDRESS_NO_VERIFY (buf, pos)]; |
| 446 | 2071 /* now run the same tests to distinguish going off |
| 2072 the end, a match or a phony match. */ | |
| 2073 if ((pos - limit) * direction <= len) | |
| 2074 break; /* ran off the end */ | |
| 2075 /* Found what might be a match. | |
| 2076 Set POS back to last (first if reverse) char pos. */ | |
| 2077 pos -= infinity; | |
| 2078 i = dirlen - direction; | |
| 2079 while ((i -= direction) + direction != 0) | |
| 428 | 2080 { |
| 446 | 2081 #ifdef MULE |
| 867 | 2082 Ichar ch; |
| 2083 Ibyte *ptr; | |
| 446 | 2084 #endif |
| 2085 pos -= direction; | |
| 2086 #ifdef MULE | |
| 826 | 2087 ptr = BYTE_BUF_BYTE_ADDRESS_NO_VERIFY (buf, pos); |
| 446 | 2088 if ((ptr == tail_end_ptr |
| 867 | 2089 || ibyte_first_byte_p (ptr[1])) |
| 2090 && (ibyte_first_byte_p (ptr[0]) | |
| 446 | 2091 || (translate_prev_byte == ptr[-1] |
| 867 | 2092 && (ibyte_first_byte_p (translate_prev_byte) |
| 446 | 2093 || translate_anteprev_byte == ptr[-2])))) |
| 2094 ch = simple_translate[*ptr]; | |
| 428 | 2095 else |
| 446 | 2096 ch = *ptr; |
| 2097 if (pat[i] != ch) | |
| 2098 break; | |
| 2099 | |
| 2100 #else | |
| 826 | 2101 if (pat[i] != |
| 2102 TRANSLATE (trt, | |
| 2103 *BYTE_BUF_BYTE_ADDRESS_NO_VERIFY (buf, pos))) | |
| 446 | 2104 break; |
| 2105 #endif | |
| 428 | 2106 } |
| 446 | 2107 /* Above loop has moved POS part or all the way back |
| 2108 to the first char pos (last char pos if reverse). | |
| 2109 Set it once again at the last (first if reverse) | |
| 2110 char. */ | |
| 2111 pos += dirlen - i- direction; | |
| 2112 if (i + direction == 0) | |
| 428 | 2113 { |
| 446 | 2114 pos -= direction; |
| 2115 | |
| 2116 { | |
| 665 | 2117 Bytebpos bytstart = (pos + |
| 446 | 2118 ((direction > 0) |
| 2119 ? 1 - len : 0)); | |
| 665 | 2120 Charbpos bufstart = bytebpos_to_charbpos (buf, bytstart); |
| 2121 Charbpos bufend = bytebpos_to_charbpos (buf, bytstart + len); | |
| 446 | 2122 |
| 2123 set_search_regs (buf, bufstart, bufend - bufstart); | |
| 2124 } | |
| 2125 | |
| 2126 if ((n -= direction) != 0) | |
| 2127 pos += dirlen; /* to resume search */ | |
| 428 | 2128 else |
| 446 | 2129 return ((direction > 0) |
| 2130 ? search_regs.end[0] : search_regs.start[0]); | |
| 428 | 2131 } |
| 446 | 2132 else |
| 2133 pos += stride_for_teases; | |
| 2134 } | |
| 428 | 2135 } |
| 446 | 2136 /* We have done one clump. Can we continue? */ |
| 2137 if ((lim - pos) * direction < 0) | |
| 2138 return (0 - n) * direction; | |
| 428 | 2139 } |
| 665 | 2140 return bytebpos_to_charbpos (buf, pos); |
| 428 | 2141 } |
| 2142 | |
| 1024 | 2143 /* Record the whole-match data (beginning BEG and end BEG + LEN) and the |
| 2144 buffer for a match just found. */ | |
| 428 | 2145 |
| 2146 static void | |
| 665 | 2147 set_search_regs (struct buffer *buf, Charbpos beg, Charcount len) |
| 428 | 2148 { |
| 2149 /* Make sure we have registers in which to store | |
| 2150 the match position. */ | |
| 2151 if (search_regs.num_regs == 0) | |
| 2152 { | |
| 2153 search_regs.start = xnew (regoff_t); | |
| 2154 search_regs.end = xnew (regoff_t); | |
| 2155 search_regs.num_regs = 1; | |
| 2156 } | |
| 2157 | |
| 1468 | 2158 clear_search_regs (); |
| 428 | 2159 search_regs.start[0] = beg; |
| 2160 search_regs.end[0] = beg + len; | |
| 793 | 2161 last_thing_searched = wrap_buffer (buf); |
| 428 | 2162 } |
| 2163 | |
| 1468 | 2164 /* Clear search registers so match data will be null. */ |
| 1024 | 2165 |
| 2166 static void | |
| 1468 | 2167 clear_search_regs (void) |
| 1024 | 2168 { |
| 2169 /* This function has been Mule-ized. */ | |
| 2170 int i; | |
| 2171 | |
| 1468 | 2172 for (i = 0; i < search_regs.num_regs; i++) |
| 2173 search_regs.start[i] = search_regs.end[i] = -1; | |
| 1024 | 2174 } |
| 2175 | |
| 428 | 2176 |
| 2177 /* Given a string of words separated by word delimiters, | |
| 442 | 2178 compute a regexp that matches those exact words |
| 2179 separated by arbitrary punctuation. */ | |
| 428 | 2180 |
| 2181 static Lisp_Object | |
| 2182 wordify (Lisp_Object buffer, Lisp_Object string) | |
| 2183 { | |
| 2184 Charcount i, len; | |
| 2185 EMACS_INT punct_count = 0, word_count = 0; | |
| 2186 struct buffer *buf = decode_buffer (buffer, 0); | |
| 826 | 2187 Lisp_Object syntax_table = buf->mirror_syntax_table; |
| 428 | 2188 |
| 2189 CHECK_STRING (string); | |
| 826 | 2190 len = string_char_length (string); |
| 428 | 2191 |
| 2192 for (i = 0; i < len; i++) | |
| 867 | 2193 if (!WORD_SYNTAX_P (syntax_table, string_ichar (string, i))) |
| 428 | 2194 { |
| 2195 punct_count++; | |
| 2196 if (i > 0 && WORD_SYNTAX_P (syntax_table, | |
| 867 | 2197 string_ichar (string, i - 1))) |
| 428 | 2198 word_count++; |
| 2199 } | |
| 867 | 2200 if (WORD_SYNTAX_P (syntax_table, string_ichar (string, len - 1))) |
| 428 | 2201 word_count++; |
| 2202 if (!word_count) return build_string (""); | |
| 2203 | |
| 2204 { | |
| 2205 /* The following value is an upper bound on the amount of storage we | |
| 2206 need. In non-Mule, it is exact. */ | |
| 867 | 2207 Ibyte *storage = |
| 2367 | 2208 alloca_ibytes (XSTRING_LENGTH (string) - punct_count + |
| 428 | 2209 5 * (word_count - 1) + 4); |
| 867 | 2210 Ibyte *o = storage; |
| 428 | 2211 |
| 2212 *o++ = '\\'; | |
| 2213 *o++ = 'b'; | |
| 2214 | |
| 2215 for (i = 0; i < len; i++) | |
| 2216 { | |
| 867 | 2217 Ichar ch = string_ichar (string, i); |
| 428 | 2218 |
| 2219 if (WORD_SYNTAX_P (syntax_table, ch)) | |
| 867 | 2220 o += set_itext_ichar (o, ch); |
| 428 | 2221 else if (i > 0 |
| 2222 && WORD_SYNTAX_P (syntax_table, | |
| 867 | 2223 string_ichar (string, i - 1)) |
| 428 | 2224 && --word_count) |
| 2225 { | |
| 2226 *o++ = '\\'; | |
| 2227 *o++ = 'W'; | |
| 2228 *o++ = '\\'; | |
| 2229 *o++ = 'W'; | |
| 2230 *o++ = '*'; | |
| 2231 } | |
| 2232 } | |
| 2233 | |
| 2234 *o++ = '\\'; | |
| 2235 *o++ = 'b'; | |
| 2236 | |
| 2237 return make_string (storage, o - storage); | |
| 2238 } | |
| 2239 } | |
| 2240 | |
| 2241 DEFUN ("search-backward", Fsearch_backward, 1, 5, "sSearch backward: ", /* | |
| 2242 Search backward from point for STRING. | |
| 2243 Set point to the beginning of the occurrence found, and return point. | |
| 444 | 2244 |
| 2245 Optional second argument LIMIT bounds the search; it is a buffer | |
| 2246 position. The match found must not extend before that position. | |
| 2247 The value nil is equivalent to (point-min). | |
| 2248 | |
| 2249 Optional third argument NOERROR, if t, means just return nil (no | |
| 2250 error) if the search fails. If neither nil nor t, set point to LIMIT | |
| 2251 and return nil. | |
| 2252 | |
| 2253 Optional fourth argument COUNT is a repeat count--search for | |
| 2254 successive occurrences. | |
| 2255 | |
| 428 | 2256 Optional fifth argument BUFFER specifies the buffer to search in and |
| 444 | 2257 defaults to the current buffer. |
| 2258 | |
| 1468 | 2259 When the match is successful, this function modifies the match data |
| 2260 that `match-beginning', `match-end' and `match-data' access; save the | |
| 2261 match data with `match-data' and restore it with `store-match-data' if | |
| 2262 you want to preserve them. If the match fails, the match data from the | |
| 2263 previous success match is preserved. | |
| 2264 | |
| 2265 See also the function `replace-match'. | |
| 428 | 2266 */ |
| 444 | 2267 (string, limit, noerror, count, buffer)) |
| 428 | 2268 { |
| 444 | 2269 return search_command (string, limit, noerror, count, buffer, -1, 0, 0); |
| 428 | 2270 } |
| 2271 | |
| 2272 DEFUN ("search-forward", Fsearch_forward, 1, 5, "sSearch: ", /* | |
| 2273 Search forward from point for STRING. | |
| 2274 Set point to the end of the occurrence found, and return point. | |
| 444 | 2275 |
| 2276 Optional second argument LIMIT bounds the search; it is a buffer | |
| 2277 position. The match found must not extend after that position. The | |
| 2278 value nil is equivalent to (point-max). | |
| 2279 | |
| 2280 Optional third argument NOERROR, if t, means just return nil (no | |
| 2281 error) if the search fails. If neither nil nor t, set point to LIMIT | |
| 2282 and return nil. | |
| 2283 | |
| 2284 Optional fourth argument COUNT is a repeat count--search for | |
| 2285 successive occurrences. | |
| 2286 | |
| 428 | 2287 Optional fifth argument BUFFER specifies the buffer to search in and |
| 444 | 2288 defaults to the current buffer. |
| 2289 | |
| 1468 | 2290 When the match is successful, this function modifies the match data |
| 2291 that `match-beginning', `match-end' and `match-data' access; save the | |
| 2292 match data with `match-data' and restore it with `store-match-data' if | |
| 2293 you want to preserve them. If the match fails, the match data from the | |
| 2294 previous success match is preserved. | |
| 2295 | |
| 2296 See also the function `replace-match'. | |
| 428 | 2297 */ |
| 444 | 2298 (string, limit, noerror, count, buffer)) |
| 428 | 2299 { |
| 444 | 2300 return search_command (string, limit, noerror, count, buffer, 1, 0, 0); |
| 428 | 2301 } |
| 2302 | |
| 2303 DEFUN ("word-search-backward", Fword_search_backward, 1, 5, | |
| 2304 "sWord search backward: ", /* | |
| 2305 Search backward from point for STRING, ignoring differences in punctuation. | |
| 2306 Set point to the beginning of the occurrence found, and return point. | |
| 444 | 2307 |
| 2308 Optional second argument LIMIT bounds the search; it is a buffer | |
| 2309 position. The match found must not extend before that position. | |
| 2310 The value nil is equivalent to (point-min). | |
| 2311 | |
| 2312 Optional third argument NOERROR, if t, means just return nil (no | |
| 2313 error) if the search fails. If neither nil nor t, set point to LIMIT | |
| 2314 and return nil. | |
| 2315 | |
| 2316 Optional fourth argument COUNT is a repeat count--search for | |
| 2317 successive occurrences. | |
| 2318 | |
| 428 | 2319 Optional fifth argument BUFFER specifies the buffer to search in and |
| 444 | 2320 defaults to the current buffer. |
| 2321 | |
| 1468 | 2322 When the match is successful, this function modifies the match data |
| 2323 that `match-beginning', `match-end' and `match-data' access; save the | |
| 2324 match data with `match-data' and restore it with `store-match-data' if | |
| 2325 you want to preserve them. If the match fails, the match data from the | |
| 2326 previous success match is preserved. | |
| 2327 | |
| 2328 See also the function `replace-match'. | |
| 428 | 2329 */ |
| 444 | 2330 (string, limit, noerror, count, buffer)) |
| 428 | 2331 { |
| 444 | 2332 return search_command (wordify (buffer, string), limit, noerror, count, |
| 428 | 2333 buffer, -1, 1, 0); |
| 2334 } | |
| 2335 | |
| 2336 DEFUN ("word-search-forward", Fword_search_forward, 1, 5, "sWord search: ", /* | |
| 2337 Search forward from point for STRING, ignoring differences in punctuation. | |
| 2338 Set point to the end of the occurrence found, and return point. | |
| 444 | 2339 |
| 2340 Optional second argument LIMIT bounds the search; it is a buffer | |
| 2341 position. The match found must not extend after that position. The | |
| 2342 value nil is equivalent to (point-max). | |
| 2343 | |
| 2344 Optional third argument NOERROR, if t, means just return nil (no | |
| 2345 error) if the search fails. If neither nil nor t, set point to LIMIT | |
| 2346 and return nil. | |
| 2347 | |
| 2348 Optional fourth argument COUNT is a repeat count--search for | |
| 2349 successive occurrences. | |
| 2350 | |
| 428 | 2351 Optional fifth argument BUFFER specifies the buffer to search in and |
| 444 | 2352 defaults to the current buffer. |
| 2353 | |
| 1468 | 2354 When the match is successful, this function modifies the match data |
| 2355 that `match-beginning', `match-end' and `match-data' access; save the | |
| 2356 match data with `match-data' and restore it with `store-match-data' if | |
| 2357 you want to preserve them. If the match fails, the match data from the | |
| 2358 previous success match is preserved. | |
| 2359 | |
| 2360 See also the function `replace-match'. | |
| 428 | 2361 */ |
| 444 | 2362 (string, limit, noerror, count, buffer)) |
| 428 | 2363 { |
| 444 | 2364 return search_command (wordify (buffer, string), limit, noerror, count, |
| 428 | 2365 buffer, 1, 1, 0); |
| 2366 } | |
| 2367 | |
| 2368 DEFUN ("re-search-backward", Fre_search_backward, 1, 5, | |
| 2369 "sRE search backward: ", /* | |
| 2370 Search backward from point for match for regular expression REGEXP. | |
| 2371 Set point to the beginning of the match, and return point. | |
| 2372 The match found is the one starting last in the buffer | |
| 2373 and yet ending before the origin of the search. | |
| 444 | 2374 |
| 2375 Optional second argument LIMIT bounds the search; it is a buffer | |
| 2376 position. The match found must not extend before that position. | |
| 2377 The value nil is equivalent to (point-min). | |
| 2378 | |
| 2379 Optional third argument NOERROR, if t, means just return nil (no | |
| 2380 error) if the search fails. If neither nil nor t, set point to LIMIT | |
| 2381 and return nil. | |
| 2382 | |
| 2383 Optional fourth argument COUNT is a repeat count--search for | |
| 2384 successive occurrences. | |
| 2385 | |
| 428 | 2386 Optional fifth argument BUFFER specifies the buffer to search in and |
| 444 | 2387 defaults to the current buffer. |
| 2388 | |
| 1468 | 2389 When the match is successful, this function modifies the match data |
| 2390 that `match-beginning', `match-end' and `match-data' access; save the | |
| 2391 match data with `match-data' and restore it with `store-match-data' if | |
| 2392 you want to preserve them. If the match fails, the match data from the | |
| 2393 previous success match is preserved. | |
| 2394 | |
| 2395 See also the function `replace-match'. | |
| 428 | 2396 */ |
| 444 | 2397 (regexp, limit, noerror, count, buffer)) |
| 428 | 2398 { |
| 444 | 2399 return search_command (regexp, limit, noerror, count, buffer, -1, 1, 0); |
| 428 | 2400 } |
| 2401 | |
| 2402 DEFUN ("re-search-forward", Fre_search_forward, 1, 5, "sRE search: ", /* | |
| 2403 Search forward from point for regular expression REGEXP. | |
| 2404 Set point to the end of the occurrence found, and return point. | |
| 444 | 2405 |
| 2406 Optional second argument LIMIT bounds the search; it is a buffer | |
| 2407 position. The match found must not extend after that position. The | |
| 2408 value nil is equivalent to (point-max). | |
| 2409 | |
| 2410 Optional third argument NOERROR, if t, means just return nil (no | |
| 2411 error) if the search fails. If neither nil nor t, set point to LIMIT | |
| 2412 and return nil. | |
| 2413 | |
| 2414 Optional fourth argument COUNT is a repeat count--search for | |
| 2415 successive occurrences. | |
| 2416 | |
| 428 | 2417 Optional fifth argument BUFFER specifies the buffer to search in and |
| 444 | 2418 defaults to the current buffer. |
| 2419 | |
| 1468 | 2420 When the match is successful, this function modifies the match data |
| 2421 that `match-beginning', `match-end' and `match-data' access; save the | |
| 2422 match data with `match-data' and restore it with `store-match-data' if | |
| 2423 you want to preserve them. If the match fails, the match data from the | |
| 2424 previous success match is preserved. | |
| 2425 | |
| 2426 See also the function `replace-match'. | |
| 428 | 2427 */ |
| 444 | 2428 (regexp, limit, noerror, count, buffer)) |
| 428 | 2429 { |
| 444 | 2430 return search_command (regexp, limit, noerror, count, buffer, 1, 1, 0); |
| 428 | 2431 } |
| 2432 | |
| 2433 DEFUN ("posix-search-backward", Fposix_search_backward, 1, 5, | |
| 2434 "sPosix search backward: ", /* | |
| 2435 Search backward from point for match for regular expression REGEXP. | |
| 2436 Find the longest match in accord with Posix regular expression rules. | |
| 2437 Set point to the beginning of the match, and return point. | |
| 2438 The match found is the one starting last in the buffer | |
| 2439 and yet ending before the origin of the search. | |
| 444 | 2440 |
| 2441 Optional second argument LIMIT bounds the search; it is a buffer | |
| 2442 position. The match found must not extend before that position. | |
| 2443 The value nil is equivalent to (point-min). | |
| 2444 | |
| 2445 Optional third argument NOERROR, if t, means just return nil (no | |
| 2446 error) if the search fails. If neither nil nor t, set point to LIMIT | |
| 2447 and return nil. | |
| 2448 | |
| 2449 Optional fourth argument COUNT is a repeat count--search for | |
| 2450 successive occurrences. | |
| 2451 | |
| 428 | 2452 Optional fifth argument BUFFER specifies the buffer to search in and |
| 444 | 2453 defaults to the current buffer. |
| 2454 | |
| 1468 | 2455 When the match is successful, this function modifies the match data |
| 2456 that `match-beginning', `match-end' and `match-data' access; save the | |
| 2457 match data with `match-data' and restore it with `store-match-data' if | |
| 2458 you want to preserve them. If the match fails, the match data from the | |
| 2459 previous success match is preserved. | |
| 2460 | |
| 2461 See also the function `replace-match'. | |
| 428 | 2462 */ |
| 444 | 2463 (regexp, limit, noerror, count, buffer)) |
| 428 | 2464 { |
| 444 | 2465 return search_command (regexp, limit, noerror, count, buffer, -1, 1, 1); |
| 428 | 2466 } |
| 2467 | |
| 2468 DEFUN ("posix-search-forward", Fposix_search_forward, 1, 5, "sPosix search: ", /* | |
| 2469 Search forward from point for regular expression REGEXP. | |
| 2470 Find the longest match in accord with Posix regular expression rules. | |
| 2471 Set point to the end of the occurrence found, and return point. | |
| 444 | 2472 |
| 2473 Optional second argument LIMIT bounds the search; it is a buffer | |
| 2474 position. The match found must not extend after that position. The | |
| 2475 value nil is equivalent to (point-max). | |
| 2476 | |
| 2477 Optional third argument NOERROR, if t, means just return nil (no | |
| 2478 error) if the search fails. If neither nil nor t, set point to LIMIT | |
| 2479 and return nil. | |
| 2480 | |
| 2481 Optional fourth argument COUNT is a repeat count--search for | |
| 2482 successive occurrences. | |
| 2483 | |
| 428 | 2484 Optional fifth argument BUFFER specifies the buffer to search in and |
| 444 | 2485 defaults to the current buffer. |
| 2486 | |
| 1468 | 2487 When the match is successful, this function modifies the match data |
| 2488 that `match-beginning', `match-end' and `match-data' access; save the | |
| 2489 match data with `match-data' and restore it with `store-match-data' if | |
| 2490 you want to preserve them. If the match fails, the match data from the | |
| 2491 previous success match is preserved. | |
| 2492 | |
| 2493 See also the function `replace-match'. | |
| 428 | 2494 */ |
| 444 | 2495 (regexp, limit, noerror, count, buffer)) |
| 428 | 2496 { |
| 444 | 2497 return search_command (regexp, limit, noerror, count, buffer, 1, 1, 1); |
| 428 | 2498 } |
| 2499 | |
| 2500 | |
| 2501 static Lisp_Object | |
| 2502 free_created_dynarrs (Lisp_Object cons) | |
| 2503 { | |
| 2504 Dynarr_free (get_opaque_ptr (XCAR (cons))); | |
| 2505 Dynarr_free (get_opaque_ptr (XCDR (cons))); | |
| 2506 free_opaque_ptr (XCAR (cons)); | |
| 2507 free_opaque_ptr (XCDR (cons)); | |
| 853 | 2508 free_cons (cons); |
| 428 | 2509 return Qnil; |
| 2510 } | |
| 2511 | |
| 2512 DEFUN ("replace-match", Freplace_match, 1, 5, 0, /* | |
| 444 | 2513 Replace text matched by last search with REPLACEMENT. |
| 4199 | 2514 Leaves point at end of replacement text. |
| 2515 Optional boolean FIXEDCASE inhibits matching case of REPLACEMENT to source. | |
| 2516 Optional boolean LITERAL inhibits interpretation of escape sequences. | |
| 2517 Optional STRING provides the source text to replace. | |
| 2518 Optional STRBUFFER may be a buffer, providing match context, or an integer | |
| 2519 specifying the subexpression to replace. | |
| 2520 | |
| 2521 If FIXEDCASE is non-nil, do not alter case of replacement text. | |
| 428 | 2522 Otherwise maybe capitalize the whole text, or maybe just word initials, |
| 2523 based on the replaced text. | |
| 4199 | 2524 If the replaced text has only capital letters and has at least one |
| 2525 multiletter word, convert REPLACEMENT to all caps. | |
| 428 | 2526 If the replaced text has at least one word starting with a capital letter, |
| 444 | 2527 then capitalize each word in REPLACEMENT. |
| 428 | 2528 |
| 4199 | 2529 If LITERAL is non-nil, insert REPLACEMENT literally. |
| 428 | 2530 Otherwise treat `\\' as special: |
| 444 | 2531 `\\&' in REPLACEMENT means substitute original matched text. |
| 428 | 2532 `\\N' means substitute what matched the Nth `\\(...\\)'. |
| 2533 If Nth parens didn't match, substitute nothing. | |
| 2534 `\\\\' means insert one `\\'. | |
| 2535 `\\u' means upcase the next character. | |
| 2536 `\\l' means downcase the next character. | |
| 2537 `\\U' means begin upcasing all following characters. | |
| 2538 `\\L' means begin downcasing all following characters. | |
| 2539 `\\E' means terminate the effect of any `\\U' or `\\L'. | |
| 2540 Case changes made with `\\u', `\\l', `\\U', and `\\L' override | |
| 2541 all other case changes that may be made in the replaced text. | |
| 4199 | 2542 |
| 2543 If non-nil, STRING is the source string, and a new string with the specified | |
| 2544 replacements is created and returned. Otherwise the current buffer is the | |
| 2545 source text. | |
| 2546 | |
| 2547 If non-nil, STRBUFFER may be an integer, interpreted as the index of the | |
| 2548 subexpression to replace in the source text, or a buffer to provide the | |
| 2549 syntax table and case table. If nil, then the \"subexpression\" is 0, i.e., | |
| 2550 the whole match, and the current buffer provides the syntax and case tables. | |
| 2551 If STRING is nil, STRBUFFER must be nil or an integer. | |
| 2552 | |
| 2553 Specifying a subexpression is only useful after a regular expression match, | |
| 2554 since a fixed string search has no non-trivial subexpressions. | |
| 2555 | |
| 2556 It is not possible to specify both a buffer and a subexpression. If that is | |
| 2557 desired, the idiom `(with-current-buffer BUFFER (replace-match ... INTEGER))' | |
| 2558 may be appropriate. | |
| 2559 | |
| 2560 If STRING is nil but the last thing matched (or searched) was a string, or | |
| 2561 STRING is a string but the last thing matched was a buffer, an | |
| 2562 `invalid-argument' error will be signaled. (XEmacs does not check that the | |
| 2563 last thing searched is the source string, but it is not useful to use a | |
| 2564 different string as source.) | |
| 2565 | |
| 2566 If no match (including searches) has been successful or the requested | |
| 1468 | 2567 subexpression was not matched, an `args-out-of-range' error will be |
| 2568 signaled. (If no match has ever been conducted in this instance of | |
| 2569 XEmacs, an `invalid-operation' error will be signaled. This is very | |
| 2570 rare.) | |
| 428 | 2571 */ |
| 444 | 2572 (replacement, fixedcase, literal, string, strbuffer)) |
| 428 | 2573 { |
| 2574 /* This function can GC */ | |
| 2575 enum { nochange, all_caps, cap_initial } case_action; | |
| 665 | 2576 Charbpos pos, last; |
| 428 | 2577 int some_multiletter_word; |
| 2578 int some_lowercase; | |
| 2579 int some_uppercase; | |
| 2580 int some_nonuppercase_initial; | |
| 867 | 2581 Ichar c, prevc; |
| 428 | 2582 Charcount inslen; |
| 2583 struct buffer *buf; | |
| 826 | 2584 Lisp_Object syntax_table; |
| 428 | 2585 int mc_count; |
| 2586 Lisp_Object buffer; | |
| 2587 int_dynarr *ul_action_dynarr = 0; | |
| 2588 int_dynarr *ul_pos_dynarr = 0; | |
| 502 | 2589 int sub = 0; |
| 428 | 2590 int speccount; |
| 2591 | |
| 444 | 2592 CHECK_STRING (replacement); |
| 428 | 2593 |
| 4199 | 2594 /* Because GNU decided to be incompatible here, we support the following |
| 2595 baroque and bogus API for the STRING and STRBUFFER arguments: | |
| 2596 types interpretations | |
| 2597 STRING STRBUFFER STRING STRBUFFER | |
| 2598 nil nil none 0 = index of subexpression to replace | |
| 2599 nil integer none index of subexpression to replace | |
| 2600 nil other ***** error ***** | |
| 2601 string nil source current buffer provides syntax table | |
| 2602 subexpression = 0 (whole match) | |
| 2603 string buffer source buffer providing syntax table | |
| 2604 subexpression = 0 (whole match) | |
| 2605 string integer source current buffer provides syntax table | |
| 2606 subexpression = STRBUFFER | |
| 2607 string other ***** error ***** | |
| 2608 */ | |
| 2609 | |
| 2610 /* Do STRBUFFER first; if STRING is nil, we'll overwrite BUF and BUFFER. */ | |
| 2611 | |
| 2612 /* If the match data were abstracted into a special "match data" type | |
| 2613 instead of the typical half-assed "let the implementation be visible" | |
| 2614 form it's in, we could extend it to include the last string matched | |
| 2615 and the buffer used for that matching. But of course we can't change | |
| 2616 it as it is. | |
| 2617 */ | |
| 2618 if (NILP (strbuffer) || BUFFERP (strbuffer)) | |
| 2619 { | |
| 2620 buf = decode_buffer (strbuffer, 0); | |
| 2621 } | |
| 2622 else if (!NILP (strbuffer)) | |
| 2623 { | |
| 2624 CHECK_INT (strbuffer); | |
| 2625 sub = XINT (strbuffer); | |
| 2626 if (sub < 0 || sub >= (int) search_regs.num_regs) | |
| 2627 invalid_argument ("match data register invalid", strbuffer); | |
| 2628 if (search_regs.start[sub] < 0) | |
| 2629 invalid_argument ("match data register not set", strbuffer); | |
| 2630 buf = current_buffer; | |
| 2631 } | |
| 2632 else | |
| 2633 invalid_argument ("STRBUFFER must be nil, a buffer, or an integer", | |
| 2634 strbuffer); | |
| 2635 buffer = wrap_buffer (buf); | |
| 2636 | |
| 428 | 2637 if (! NILP (string)) |
| 2638 { | |
| 2639 CHECK_STRING (string); | |
| 2640 if (!EQ (last_thing_searched, Qt)) | |
| 4199 | 2641 invalid_argument ("last thing matched was not a string", Qunbound); |
| 428 | 2642 } |
| 2643 else | |
| 2644 { | |
| 2645 if (!BUFFERP (last_thing_searched)) | |
| 4199 | 2646 invalid_argument ("last thing matched was not a buffer", Qunbound); |
| 428 | 2647 buffer = last_thing_searched; |
| 2648 buf = XBUFFER (buffer); | |
| 2649 } | |
| 2650 | |
| 826 | 2651 syntax_table = buf->mirror_syntax_table; |
| 428 | 2652 |
| 2653 case_action = nochange; /* We tried an initialization */ | |
| 2654 /* but some C compilers blew it */ | |
| 2655 | |
| 2656 if (search_regs.num_regs == 0) | |
| 826 | 2657 signal_error (Qinvalid_operation, |
| 2658 "replace-match called before any match found", Qunbound); | |
| 428 | 2659 |
| 2660 if (NILP (string)) | |
| 2661 { | |
| 469 | 2662 if (search_regs.start[sub] < BUF_BEGV (buf) |
| 2663 || search_regs.start[sub] > search_regs.end[sub] | |
| 2664 || search_regs.end[sub] > BUF_ZV (buf)) | |
| 2665 args_out_of_range (make_int (search_regs.start[sub]), | |
| 2666 make_int (search_regs.end[sub])); | |
| 428 | 2667 } |
| 2668 else | |
| 2669 { | |
| 2670 if (search_regs.start[0] < 0 | |
| 2671 || search_regs.start[0] > search_regs.end[0] | |
| 826 | 2672 || search_regs.end[0] > string_char_length (string)) |
| 428 | 2673 args_out_of_range (make_int (search_regs.start[0]), |
| 2674 make_int (search_regs.end[0])); | |
| 2675 } | |
| 2676 | |
| 2677 if (NILP (fixedcase)) | |
| 2678 { | |
| 2679 /* Decide how to casify by examining the matched text. */ | |
| 2680 | |
| 707 | 2681 last = search_regs.end[sub]; |
| 428 | 2682 prevc = '\n'; |
| 2683 case_action = all_caps; | |
| 2684 | |
| 2685 /* some_multiletter_word is set nonzero if any original word | |
| 2686 is more than one letter long. */ | |
| 2687 some_multiletter_word = 0; | |
| 2688 some_lowercase = 0; | |
| 2689 some_nonuppercase_initial = 0; | |
| 2690 some_uppercase = 0; | |
| 2691 | |
| 707 | 2692 for (pos = search_regs.start[sub]; pos < last; pos++) |
| 428 | 2693 { |
| 2694 if (NILP (string)) | |
| 2695 c = BUF_FETCH_CHAR (buf, pos); | |
| 2696 else | |
| 867 | 2697 c = string_ichar (string, pos); |
| 428 | 2698 |
| 2699 if (LOWERCASEP (buf, c)) | |
| 2700 { | |
| 2701 /* Cannot be all caps if any original char is lower case */ | |
| 2702 | |
| 2703 some_lowercase = 1; | |
| 2704 if (!WORD_SYNTAX_P (syntax_table, prevc)) | |
| 2705 some_nonuppercase_initial = 1; | |
| 2706 else | |
| 2707 some_multiletter_word = 1; | |
| 2708 } | |
| 2709 else if (!NOCASEP (buf, c)) | |
| 2710 { | |
| 2711 some_uppercase = 1; | |
| 2712 if (!WORD_SYNTAX_P (syntax_table, prevc)) | |
| 2713 ; | |
| 2714 else | |
| 2715 some_multiletter_word = 1; | |
| 2716 } | |
| 2717 else | |
| 2718 { | |
| 2719 /* If the initial is a caseless word constituent, | |
| 2720 treat that like a lowercase initial. */ | |
| 2721 if (!WORD_SYNTAX_P (syntax_table, prevc)) | |
| 2722 some_nonuppercase_initial = 1; | |
| 2723 } | |
| 2724 | |
| 2725 prevc = c; | |
| 2726 } | |
| 2727 | |
| 2728 /* Convert to all caps if the old text is all caps | |
| 2729 and has at least one multiletter word. */ | |
| 2730 if (! some_lowercase && some_multiletter_word) | |
| 2731 case_action = all_caps; | |
| 2732 /* Capitalize each word, if the old text has all capitalized words. */ | |
| 2733 else if (!some_nonuppercase_initial && some_multiletter_word) | |
| 2734 case_action = cap_initial; | |
| 2735 else if (!some_nonuppercase_initial && some_uppercase) | |
| 2736 /* Should x -> yz, operating on X, give Yz or YZ? | |
| 2737 We'll assume the latter. */ | |
| 2738 case_action = all_caps; | |
| 2739 else | |
| 2740 case_action = nochange; | |
| 2741 } | |
| 2742 | |
| 2743 /* Do replacement in a string. */ | |
| 2744 if (!NILP (string)) | |
| 2745 { | |
| 2746 Lisp_Object before, after; | |
| 2747 | |
| 2748 speccount = specpdl_depth (); | |
| 4199 | 2749 before = Fsubstring (string, Qzero, make_int (search_regs.start[sub])); |
| 2750 after = Fsubstring (string, make_int (search_regs.end[sub]), Qnil); | |
| 428 | 2751 |
| 444 | 2752 /* Do case substitution into REPLACEMENT if desired. */ |
| 428 | 2753 if (NILP (literal)) |
| 2754 { | |
| 826 | 2755 Charcount stlen = string_char_length (replacement); |
| 428 | 2756 Charcount strpos; |
| 2757 /* XEmacs change: rewrote this loop somewhat to make it | |
| 2758 cleaner. Also added \U, \E, etc. */ | |
| 2759 Charcount literal_start = 0; | |
| 2760 /* We build up the substituted string in ACCUM. */ | |
| 2761 Lisp_Object accum; | |
| 2762 | |
| 2763 accum = Qnil; | |
| 2764 | |
| 2765 /* OK, the basic idea here is that we scan through the | |
| 2766 replacement string until we find a backslash, which | |
| 2767 represents a substring of the original string to be | |
| 2768 substituted. We then append onto ACCUM the literal | |
| 2769 text before the backslash (LASTPOS marks the | |
| 2770 beginning of this) followed by the substring of the | |
| 2771 original string that needs to be inserted. */ | |
| 2772 for (strpos = 0; strpos < stlen; strpos++) | |
| 2773 { | |
| 2774 /* If LITERAL_END is set, we've encountered a backslash | |
| 2775 (the end of literal text to be inserted). */ | |
| 2776 Charcount literal_end = -1; | |
| 2777 /* If SUBSTART is set, we need to also insert the | |
| 2778 text from SUBSTART to SUBEND in the original string. */ | |
| 2779 Charcount substart = -1; | |
| 2780 Charcount subend = -1; | |
| 2781 | |
| 867 | 2782 c = string_ichar (replacement, strpos); |
| 428 | 2783 if (c == '\\' && strpos < stlen - 1) |
| 2784 { | |
| 867 | 2785 c = string_ichar (replacement, ++strpos); |
| 428 | 2786 if (c == '&') |
| 2787 { | |
| 2788 literal_end = strpos - 1; | |
| 2789 substart = search_regs.start[0]; | |
| 2790 subend = search_regs.end[0]; | |
| 2791 } | |
| 4199 | 2792 /* #### This logic is totally broken, |
| 2793 since we can have backrefs like "\99", right? */ | |
| 428 | 2794 else if (c >= '1' && c <= '9' && |
| 2795 c <= search_regs.num_regs + '0') | |
| 2796 { | |
| 2797 if (search_regs.start[c - '0'] >= 0) | |
| 2798 { | |
| 2799 literal_end = strpos - 1; | |
| 2800 substart = search_regs.start[c - '0']; | |
| 2801 subend = search_regs.end[c - '0']; | |
| 2802 } | |
| 2803 } | |
| 2804 else if (c == 'U' || c == 'u' || c == 'L' || c == 'l' || | |
| 2805 c == 'E') | |
| 2806 { | |
| 2807 /* Keep track of all case changes requested, but don't | |
| 2808 make them now. Do them later so we override | |
| 2809 everything else. */ | |
| 2810 if (!ul_pos_dynarr) | |
| 2811 { | |
| 2812 ul_pos_dynarr = Dynarr_new (int); | |
| 2813 ul_action_dynarr = Dynarr_new (int); | |
| 2814 record_unwind_protect | |
| 2815 (free_created_dynarrs, | |
| 2816 noseeum_cons | |
| 2817 (make_opaque_ptr (ul_pos_dynarr), | |
| 2818 make_opaque_ptr (ul_action_dynarr))); | |
| 2819 } | |
| 2820 literal_end = strpos - 1; | |
| 2821 Dynarr_add (ul_pos_dynarr, | |
| 2822 (!NILP (accum) | |
| 826 | 2823 ? string_char_length (accum) |
| 428 | 2824 : 0) + (literal_end - literal_start)); |
| 2825 Dynarr_add (ul_action_dynarr, c); | |
| 2826 } | |
| 2827 else if (c == '\\') | |
| 2828 /* So we get just one backslash. */ | |
| 2829 literal_end = strpos; | |
| 2830 } | |
| 2831 if (literal_end >= 0) | |
| 2832 { | |
| 2833 Lisp_Object literal_text = Qnil; | |
| 2834 Lisp_Object substring = Qnil; | |
| 2835 if (literal_end != literal_start) | |
| 444 | 2836 literal_text = Fsubstring (replacement, |
| 428 | 2837 make_int (literal_start), |
| 2838 make_int (literal_end)); | |
| 2839 if (substart >= 0 && subend != substart) | |
| 2840 substring = Fsubstring (string, | |
| 2841 make_int (substart), | |
| 2842 make_int (subend)); | |
| 2843 if (!NILP (literal_text) || !NILP (substring)) | |
| 2844 accum = concat3 (accum, literal_text, substring); | |
| 2845 literal_start = strpos + 1; | |
| 2846 } | |
| 2847 } | |
| 2848 | |
| 2849 if (strpos != literal_start) | |
| 2850 /* some literal text at end to be inserted */ | |
| 444 | 2851 replacement = concat2 (accum, Fsubstring (replacement, |
| 2852 make_int (literal_start), | |
| 2853 make_int (strpos))); | |
| 428 | 2854 else |
| 444 | 2855 replacement = accum; |
| 428 | 2856 } |
| 2857 | |
| 444 | 2858 /* replacement can be nil. */ |
| 2859 if (NILP (replacement)) | |
| 2860 replacement = build_string (""); | |
| 2861 | |
| 428 | 2862 if (case_action == all_caps) |
| 444 | 2863 replacement = Fupcase (replacement, buffer); |
| 428 | 2864 else if (case_action == cap_initial) |
| 444 | 2865 replacement = Fupcase_initials (replacement, buffer); |
| 428 | 2866 |
| 2867 /* Now finally, we need to process the \U's, \E's, etc. */ | |
| 2868 if (ul_pos_dynarr) | |
| 2869 { | |
| 2870 int i = 0; | |
| 2871 int cur_action = 'E'; | |
| 826 | 2872 Charcount stlen = string_char_length (replacement); |
| 428 | 2873 Charcount strpos; |
| 2874 | |
| 2875 for (strpos = 0; strpos < stlen; strpos++) | |
| 2876 { | |
| 867 | 2877 Ichar curchar = string_ichar (replacement, strpos); |
| 2878 Ichar newchar = -1; | |
| 428 | 2879 if (i < Dynarr_length (ul_pos_dynarr) && |
| 2880 strpos == Dynarr_at (ul_pos_dynarr, i)) | |
| 2881 { | |
| 2882 int new_action = Dynarr_at (ul_action_dynarr, i); | |
| 2883 i++; | |
| 2884 if (new_action == 'u') | |
| 2885 newchar = UPCASE (buf, curchar); | |
| 2886 else if (new_action == 'l') | |
| 2887 newchar = DOWNCASE (buf, curchar); | |
| 2888 else | |
| 2889 cur_action = new_action; | |
| 2890 } | |
| 2891 if (newchar == -1) | |
| 2892 { | |
| 2893 if (cur_action == 'U') | |
| 2894 newchar = UPCASE (buf, curchar); | |
| 2895 else if (cur_action == 'L') | |
| 2896 newchar = DOWNCASE (buf, curchar); | |
| 2897 else | |
| 2898 newchar = curchar; | |
| 2899 } | |
| 2900 if (newchar != curchar) | |
| 793 | 2901 set_string_char (replacement, strpos, newchar); |
| 428 | 2902 } |
| 2903 } | |
| 2904 | |
| 2905 /* frees the Dynarrs if necessary. */ | |
| 771 | 2906 unbind_to (speccount); |
| 444 | 2907 return concat3 (before, replacement, after); |
| 428 | 2908 } |
| 2909 | |
| 707 | 2910 mc_count = begin_multiple_change (buf, search_regs.start[sub], |
| 2911 search_regs.end[sub]); | |
| 428 | 2912 |
| 2913 /* begin_multiple_change() records an unwind-protect, so we need to | |
| 2914 record this value now. */ | |
| 2915 speccount = specpdl_depth (); | |
| 2916 | |
| 2917 /* We insert the replacement text before the old text, and then | |
| 2918 delete the original text. This means that markers at the | |
| 2919 beginning or end of the original will float to the corresponding | |
| 2920 position in the replacement. */ | |
| 707 | 2921 BUF_SET_PT (buf, search_regs.start[sub]); |
| 428 | 2922 if (!NILP (literal)) |
| 444 | 2923 Finsert (1, &replacement); |
| 428 | 2924 else |
| 2925 { | |
| 826 | 2926 Charcount stlen = string_char_length (replacement); |
| 428 | 2927 Charcount strpos; |
| 2928 struct gcpro gcpro1; | |
| 444 | 2929 GCPRO1 (replacement); |
| 428 | 2930 for (strpos = 0; strpos < stlen; strpos++) |
| 2931 { | |
| 707 | 2932 /* on the first iteration assert(offset==0), |
| 2933 exactly complementing BUF_SET_PT() above. | |
| 2934 During the loop, it keeps track of the amount inserted. | |
| 2935 */ | |
| 2936 Charcount offset = BUF_PT (buf) - search_regs.start[sub]; | |
| 428 | 2937 |
| 867 | 2938 c = string_ichar (replacement, strpos); |
| 428 | 2939 if (c == '\\' && strpos < stlen - 1) |
| 2940 { | |
| 707 | 2941 /* XXX FIXME: replacing just a substring non-literally |
| 2942 using backslash refs to the match looks dangerous. But | |
| 2943 <15366.18513.698042.156573@ns.caldera.de> from Torsten Duwe | |
| 2944 <duwe@caldera.de> claims Finsert_buffer_substring already | |
| 2945 handles this correctly. | |
| 2946 */ | |
| 867 | 2947 c = string_ichar (replacement, ++strpos); |
| 428 | 2948 if (c == '&') |
| 2949 Finsert_buffer_substring | |
| 2950 (buffer, | |
| 2951 make_int (search_regs.start[0] + offset), | |
| 2952 make_int (search_regs.end[0] + offset)); | |
| 4199 | 2953 /* #### This logic is totally broken, |
| 2954 since we can have backrefs like "\99", right? */ | |
| 428 | 2955 else if (c >= '1' && c <= '9' && |
| 2956 c <= search_regs.num_regs + '0') | |
| 2957 { | |
| 2958 if (search_regs.start[c - '0'] >= 1) | |
| 2959 Finsert_buffer_substring | |
| 2960 (buffer, | |
| 2961 make_int (search_regs.start[c - '0'] + offset), | |
| 2962 make_int (search_regs.end[c - '0'] + offset)); | |
| 2963 } | |
| 2964 else if (c == 'U' || c == 'u' || c == 'L' || c == 'l' || | |
| 2965 c == 'E') | |
| 2966 { | |
| 2967 /* Keep track of all case changes requested, but don't | |
| 2968 make them now. Do them later so we override | |
| 2969 everything else. */ | |
| 2970 if (!ul_pos_dynarr) | |
| 2971 { | |
| 2972 ul_pos_dynarr = Dynarr_new (int); | |
| 2973 ul_action_dynarr = Dynarr_new (int); | |
| 2974 record_unwind_protect | |
| 2975 (free_created_dynarrs, | |
| 2976 Fcons (make_opaque_ptr (ul_pos_dynarr), | |
| 2977 make_opaque_ptr (ul_action_dynarr))); | |
| 2978 } | |
| 2979 Dynarr_add (ul_pos_dynarr, BUF_PT (buf)); | |
| 2980 Dynarr_add (ul_action_dynarr, c); | |
| 2981 } | |
| 2982 else | |
| 2983 buffer_insert_emacs_char (buf, c); | |
| 2984 } | |
| 2985 else | |
| 2986 buffer_insert_emacs_char (buf, c); | |
| 2987 } | |
| 2988 UNGCPRO; | |
| 2989 } | |
| 2990 | |
| 707 | 2991 inslen = BUF_PT (buf) - (search_regs.start[sub]); |
| 2992 buffer_delete_range (buf, search_regs.start[sub] + inslen, | |
| 2993 search_regs.end[sub] + inslen, 0); | |
| 428 | 2994 |
| 2995 if (case_action == all_caps) | |
| 2996 Fupcase_region (make_int (BUF_PT (buf) - inslen), | |
| 2997 make_int (BUF_PT (buf)), buffer); | |
| 2998 else if (case_action == cap_initial) | |
| 2999 Fupcase_initials_region (make_int (BUF_PT (buf) - inslen), | |
| 3000 make_int (BUF_PT (buf)), buffer); | |
| 3001 | |
| 3002 /* Now go through and make all the case changes that were requested | |
| 3003 in the replacement string. */ | |
| 3004 if (ul_pos_dynarr) | |
| 3005 { | |
| 665 | 3006 Charbpos eend = BUF_PT (buf); |
| 428 | 3007 int i = 0; |
| 3008 int cur_action = 'E'; | |
| 3009 | |
| 3010 for (pos = BUF_PT (buf) - inslen; pos < eend; pos++) | |
| 3011 { | |
| 867 | 3012 Ichar curchar = BUF_FETCH_CHAR (buf, pos); |
| 3013 Ichar newchar = -1; | |
| 428 | 3014 if (i < Dynarr_length (ul_pos_dynarr) && |
| 3015 pos == Dynarr_at (ul_pos_dynarr, i)) | |
| 3016 { | |
| 3017 int new_action = Dynarr_at (ul_action_dynarr, i); | |
| 3018 i++; | |
| 3019 if (new_action == 'u') | |
| 3020 newchar = UPCASE (buf, curchar); | |
| 3021 else if (new_action == 'l') | |
| 3022 newchar = DOWNCASE (buf, curchar); | |
| 3023 else | |
| 3024 cur_action = new_action; | |
| 3025 } | |
| 3026 if (newchar == -1) | |
| 3027 { | |
| 3028 if (cur_action == 'U') | |
| 3029 newchar = UPCASE (buf, curchar); | |
| 3030 else if (cur_action == 'L') | |
| 3031 newchar = DOWNCASE (buf, curchar); | |
| 3032 else | |
| 3033 newchar = curchar; | |
| 3034 } | |
| 3035 if (newchar != curchar) | |
| 3036 buffer_replace_char (buf, pos, newchar, 0, 0); | |
| 3037 } | |
| 3038 } | |
| 3039 | |
| 3040 /* frees the Dynarrs if necessary. */ | |
| 771 | 3041 unbind_to (speccount); |
| 428 | 3042 end_multiple_change (buf, mc_count); |
| 3043 | |
| 3044 return Qnil; | |
| 3045 } | |
| 3046 | |
| 3047 static Lisp_Object | |
| 3048 match_limit (Lisp_Object num, int beginningp) | |
| 3049 { | |
| 3050 int n; | |
| 3051 | |
| 3052 CHECK_INT (num); | |
| 3053 n = XINT (num); | |
| 3054 if (n < 0 || n >= search_regs.num_regs) | |
| 3055 args_out_of_range (num, make_int (search_regs.num_regs)); | |
| 3056 if (search_regs.num_regs == 0 || | |
| 3057 search_regs.start[n] < 0) | |
| 3058 return Qnil; | |
| 3059 return make_int (beginningp ? search_regs.start[n] : search_regs.end[n]); | |
| 3060 } | |
| 3061 | |
| 3062 DEFUN ("match-beginning", Fmatch_beginning, 1, 1, 0, /* | |
| 3063 Return position of start of text matched by last regexp search. | |
| 3064 NUM, specifies which parenthesized expression in the last regexp. | |
| 3065 Value is nil if NUMth pair didn't match, or there were less than NUM pairs. | |
| 3066 Zero means the entire text matched by the whole regexp or whole string. | |
| 3067 */ | |
| 3068 (num)) | |
| 3069 { | |
| 3070 return match_limit (num, 1); | |
| 3071 } | |
| 3072 | |
| 3073 DEFUN ("match-end", Fmatch_end, 1, 1, 0, /* | |
| 3074 Return position of end of text matched by last regexp search. | |
| 3075 NUM specifies which parenthesized expression in the last regexp. | |
| 3076 Value is nil if NUMth pair didn't match, or there were less than NUM pairs. | |
| 3077 Zero means the entire text matched by the whole regexp or whole string. | |
| 3078 */ | |
| 3079 (num)) | |
| 3080 { | |
| 3081 return match_limit (num, 0); | |
| 3082 } | |
| 3083 | |
| 3084 DEFUN ("match-data", Fmatch_data, 0, 2, 0, /* | |
| 3085 Return a list containing all info on what the last regexp search matched. | |
| 3086 Element 2N is `(match-beginning N)'; element 2N + 1 is `(match-end N)'. | |
| 3087 All the elements are markers or nil (nil if the Nth pair didn't match) | |
| 3088 if the last match was on a buffer; integers or nil if a string was matched. | |
| 3089 Use `store-match-data' to reinstate the data in this list. | |
| 3090 | |
| 3091 If INTEGERS (the optional first argument) is non-nil, always use integers | |
| 3092 \(rather than markers) to represent buffer positions. | |
| 3093 If REUSE is a list, reuse it as part of the value. If REUSE is long enough | |
| 3094 to hold all the values, and if INTEGERS is non-nil, no consing is done. | |
| 3095 */ | |
| 3096 (integers, reuse)) | |
| 3097 { | |
| 3098 Lisp_Object tail, prev; | |
| 3099 Lisp_Object *data; | |
| 3100 int i; | |
| 3101 Charcount len; | |
| 3102 | |
| 3103 if (NILP (last_thing_searched)) | |
| 563 | 3104 /*error ("match-data called before any match found", Qunbound);*/ |
| 428 | 3105 return Qnil; |
| 3106 | |
| 3107 data = alloca_array (Lisp_Object, 2 * search_regs.num_regs); | |
| 3108 | |
| 3109 len = -1; | |
| 3110 for (i = 0; i < search_regs.num_regs; i++) | |
| 3111 { | |
| 665 | 3112 Charbpos start = search_regs.start[i]; |
| 428 | 3113 if (start >= 0) |
| 3114 { | |
| 3115 if (EQ (last_thing_searched, Qt) | |
| 3116 || !NILP (integers)) | |
| 3117 { | |
| 3118 data[2 * i] = make_int (start); | |
| 3119 data[2 * i + 1] = make_int (search_regs.end[i]); | |
| 3120 } | |
| 3121 else if (BUFFERP (last_thing_searched)) | |
| 3122 { | |
| 3123 data[2 * i] = Fmake_marker (); | |
| 3124 Fset_marker (data[2 * i], | |
| 3125 make_int (start), | |
| 3126 last_thing_searched); | |
| 3127 data[2 * i + 1] = Fmake_marker (); | |
| 3128 Fset_marker (data[2 * i + 1], | |
| 3129 make_int (search_regs.end[i]), | |
| 3130 last_thing_searched); | |
| 3131 } | |
| 3132 else | |
| 3133 /* last_thing_searched must always be Qt, a buffer, or Qnil. */ | |
| 2500 | 3134 ABORT (); |
| 428 | 3135 |
| 3136 len = i; | |
| 3137 } | |
| 3138 else | |
| 3139 data[2 * i] = data [2 * i + 1] = Qnil; | |
| 3140 } | |
| 3141 if (!CONSP (reuse)) | |
| 3142 return Flist (2 * len + 2, data); | |
| 3143 | |
| 3144 /* If REUSE is a list, store as many value elements as will fit | |
| 3145 into the elements of REUSE. */ | |
| 3146 for (prev = Qnil, i = 0, tail = reuse; CONSP (tail); i++, tail = XCDR (tail)) | |
| 3147 { | |
| 3148 if (i < 2 * len + 2) | |
| 3149 XCAR (tail) = data[i]; | |
| 3150 else | |
| 3151 XCAR (tail) = Qnil; | |
| 3152 prev = tail; | |
| 3153 } | |
| 3154 | |
| 3155 /* If we couldn't fit all value elements into REUSE, | |
| 3156 cons up the rest of them and add them to the end of REUSE. */ | |
| 3157 if (i < 2 * len + 2) | |
| 3158 XCDR (prev) = Flist (2 * len + 2 - i, data + i); | |
| 3159 | |
| 3160 return reuse; | |
| 3161 } | |
| 3162 | |
| 3163 | |
| 3164 DEFUN ("store-match-data", Fstore_match_data, 1, 1, 0, /* | |
| 3165 Set internal data on last search match from elements of LIST. | |
| 1468 | 3166 LIST should have been created by calling `match-data' previously, |
| 3167 or be nil, to clear the internal match data. | |
| 428 | 3168 */ |
| 3169 (list)) | |
| 3170 { | |
| 3171 REGISTER int i; | |
| 3172 REGISTER Lisp_Object marker; | |
| 3173 int num_regs; | |
| 3174 int length; | |
| 3175 | |
| 853 | 3176 /* Some FSF junk with running_asynch_code, to preserve the match |
| 3177 data. Not necessary because we don't call process filters | |
| 3178 asynchronously (i.e. from within QUIT). */ | |
| 428 | 3179 |
| 3180 CONCHECK_LIST (list); | |
| 3181 | |
| 3182 /* Unless we find a marker with a buffer in LIST, assume that this | |
| 3183 match data came from a string. */ | |
| 3184 last_thing_searched = Qt; | |
| 3185 | |
| 3186 /* Allocate registers if they don't already exist. */ | |
| 3187 length = XINT (Flength (list)) / 2; | |
| 3188 num_regs = search_regs.num_regs; | |
| 3189 | |
| 3190 if (length > num_regs) | |
| 3191 { | |
| 3192 if (search_regs.num_regs == 0) | |
| 3193 { | |
| 3194 search_regs.start = xnew_array (regoff_t, length); | |
| 3195 search_regs.end = xnew_array (regoff_t, length); | |
| 3196 } | |
| 3197 else | |
| 3198 { | |
| 3199 XREALLOC_ARRAY (search_regs.start, regoff_t, length); | |
| 3200 XREALLOC_ARRAY (search_regs.end, regoff_t, length); | |
| 3201 } | |
| 3202 | |
| 3203 search_regs.num_regs = length; | |
| 3204 } | |
| 3205 | |
| 3206 for (i = 0; i < num_regs; i++) | |
| 3207 { | |
| 3208 marker = Fcar (list); | |
| 3209 if (NILP (marker)) | |
| 3210 { | |
| 3211 search_regs.start[i] = -1; | |
| 3212 list = Fcdr (list); | |
| 3213 } | |
| 3214 else | |
| 3215 { | |
| 3216 if (MARKERP (marker)) | |
| 3217 { | |
| 3218 if (XMARKER (marker)->buffer == 0) | |
| 3219 marker = Qzero; | |
| 3220 else | |
| 793 | 3221 last_thing_searched = wrap_buffer (XMARKER (marker)->buffer); |
| 428 | 3222 } |
| 3223 | |
| 3224 CHECK_INT_COERCE_MARKER (marker); | |
| 3225 search_regs.start[i] = XINT (marker); | |
| 3226 list = Fcdr (list); | |
| 3227 | |
| 3228 marker = Fcar (list); | |
| 3229 if (MARKERP (marker) && XMARKER (marker)->buffer == 0) | |
| 3230 marker = Qzero; | |
| 3231 | |
| 3232 CHECK_INT_COERCE_MARKER (marker); | |
| 3233 search_regs.end[i] = XINT (marker); | |
| 3234 } | |
| 3235 list = Fcdr (list); | |
| 3236 } | |
| 3237 | |
| 3238 return Qnil; | |
| 3239 } | |
| 3240 | |
| 3241 /* Quote a string to inactivate reg-expr chars */ | |
| 3242 | |
| 3243 DEFUN ("regexp-quote", Fregexp_quote, 1, 1, 0, /* | |
| 3244 Return a regexp string which matches exactly STRING and nothing else. | |
| 3245 */ | |
| 444 | 3246 (string)) |
| 428 | 3247 { |
| 867 | 3248 REGISTER Ibyte *in, *out, *end; |
| 3249 REGISTER Ibyte *temp; | |
| 428 | 3250 |
| 444 | 3251 CHECK_STRING (string); |
| 428 | 3252 |
| 2367 | 3253 temp = alloca_ibytes (XSTRING_LENGTH (string) * 2); |
| 428 | 3254 |
| 3255 /* Now copy the data into the new string, inserting escapes. */ | |
| 3256 | |
| 444 | 3257 in = XSTRING_DATA (string); |
| 3258 end = in + XSTRING_LENGTH (string); | |
| 428 | 3259 out = temp; |
| 3260 | |
| 3261 while (in < end) | |
| 3262 { | |
| 867 | 3263 Ichar c = itext_ichar (in); |
| 428 | 3264 |
| 3265 if (c == '[' || c == ']' | |
| 3266 || c == '*' || c == '.' || c == '\\' | |
| 3267 || c == '?' || c == '+' | |
| 3268 || c == '^' || c == '$') | |
| 3269 *out++ = '\\'; | |
| 867 | 3270 out += set_itext_ichar (out, c); |
| 3271 INC_IBYTEPTR (in); | |
| 428 | 3272 } |
| 3273 | |
| 3274 return make_string (temp, out - temp); | |
| 3275 } | |
| 3276 | |
| 3277 DEFUN ("set-word-regexp", Fset_word_regexp, 1, 1, 0, /* | |
| 3278 Set the regexp to be used to match a word in regular-expression searching. | |
| 3279 #### Not yet implemented. Currently does nothing. | |
| 3280 #### Do not use this yet. Its calling interface is likely to change. | |
| 3281 */ | |
| 2286 | 3282 (UNUSED (regexp))) |
| 428 | 3283 { |
| 3284 return Qnil; | |
| 3285 } | |
| 3286 | |
| 3287 | |
| 3288 /************************************************************************/ | |
| 3289 /* initialization */ | |
| 3290 /************************************************************************/ | |
| 3291 | |
| 3292 void | |
| 3293 syms_of_search (void) | |
| 3294 { | |
| 3295 | |
| 442 | 3296 DEFERROR_STANDARD (Qsearch_failed, Qinvalid_operation); |
| 3297 DEFERROR_STANDARD (Qinvalid_regexp, Qsyntax_error); | |
| 563 | 3298 Fput (Qinvalid_regexp, Qerror_lacks_explanatory_string, Qt); |
| 428 | 3299 |
| 3300 DEFSUBR (Flooking_at); | |
| 3301 DEFSUBR (Fposix_looking_at); | |
| 3302 DEFSUBR (Fstring_match); | |
| 3303 DEFSUBR (Fposix_string_match); | |
| 3304 DEFSUBR (Fskip_chars_forward); | |
| 3305 DEFSUBR (Fskip_chars_backward); | |
| 3306 DEFSUBR (Fskip_syntax_forward); | |
| 3307 DEFSUBR (Fskip_syntax_backward); | |
| 3308 DEFSUBR (Fsearch_forward); | |
| 3309 DEFSUBR (Fsearch_backward); | |
| 3310 DEFSUBR (Fword_search_forward); | |
| 3311 DEFSUBR (Fword_search_backward); | |
| 3312 DEFSUBR (Fre_search_forward); | |
| 3313 DEFSUBR (Fre_search_backward); | |
| 3314 DEFSUBR (Fposix_search_forward); | |
| 3315 DEFSUBR (Fposix_search_backward); | |
| 3316 DEFSUBR (Freplace_match); | |
| 3317 DEFSUBR (Fmatch_beginning); | |
| 3318 DEFSUBR (Fmatch_end); | |
| 3319 DEFSUBR (Fmatch_data); | |
| 3320 DEFSUBR (Fstore_match_data); | |
| 3321 DEFSUBR (Fregexp_quote); | |
| 3322 DEFSUBR (Fset_word_regexp); | |
| 3323 } | |
| 3324 | |
| 3325 void | |
| 3326 reinit_vars_of_search (void) | |
| 3327 { | |
| 3328 int i; | |
| 3329 | |
| 3330 last_thing_searched = Qnil; | |
| 3331 staticpro_nodump (&last_thing_searched); | |
| 3332 | |
| 3333 for (i = 0; i < REGEXP_CACHE_SIZE; ++i) | |
| 3334 { | |
| 3335 searchbufs[i].buf.allocated = 100; | |
| 3336 searchbufs[i].buf.buffer = (unsigned char *) xmalloc (100); | |
| 3337 searchbufs[i].buf.fastmap = searchbufs[i].fastmap; | |
| 3338 searchbufs[i].regexp = Qnil; | |
| 3339 staticpro_nodump (&searchbufs[i].regexp); | |
| 3340 searchbufs[i].next = (i == REGEXP_CACHE_SIZE-1 ? 0 : &searchbufs[i+1]); | |
| 3341 } | |
| 3342 searchbuf_head = &searchbufs[0]; | |
| 3343 } | |
| 3344 | |
| 3345 void | |
| 3346 vars_of_search (void) | |
| 3347 { | |
| 3348 DEFVAR_LISP ("forward-word-regexp", &Vforward_word_regexp /* | |
| 3349 *Regular expression to be used in `forward-word'. | |
| 3350 #### Not yet implemented. | |
| 3351 */ ); | |
| 3352 Vforward_word_regexp = Qnil; | |
| 3353 | |
| 3354 DEFVAR_LISP ("backward-word-regexp", &Vbackward_word_regexp /* | |
| 3355 *Regular expression to be used in `backward-word'. | |
| 3356 #### Not yet implemented. | |
| 3357 */ ); | |
| 3358 Vbackward_word_regexp = Qnil; | |
| 502 | 3359 |
| 3360 DEFVAR_INT ("warn-about-possibly-incompatible-back-references", | |
| 3361 &warn_about_possibly_incompatible_back_references /* | |
| 3362 If true, issue warnings when new-semantics back references occur. | |
| 3363 This is to catch places where old code might inadvertently have changed | |
| 3364 semantics. This will occur in old code only where more than nine groups | |
| 3365 occur and a back reference to one of them is directly followed by a digit. | |
| 3366 */ ); | |
| 3367 warn_about_possibly_incompatible_back_references = 1; | |
| 814 | 3368 |
| 2421 | 3369 Vskip_chars_range_table = Fmake_range_table (Qstart_closed_end_closed); |
| 428 | 3370 staticpro (&Vskip_chars_range_table); |
|
4414
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
3371 #ifdef DEBUG_XEMACS |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
3372 DEFSYMBOL (Qsearch_algorithm_used); |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
3373 DEFSYMBOL (Qboyer_moore); |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
3374 DEFSYMBOL (Qsimple_search); |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
3375 |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
3376 DEFVAR_INT ("debug-xemacs-searches", &debug_xemacs_searches /* |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
3377 If non-zero, bind `search-algorithm-used' to `boyer-moore' or `simple-search', |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
3378 depending on the algorithm used for each search. Used for testing. |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
3379 */ ); |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
3380 debug_xemacs_searches = 0; |
|
df576f30c1d8
Correct case-insensitive search for non-case, non-ASCII chars. Add tests.
Aidan Kehoe <kehoea@parhasard.net>
parents:
4407
diff
changeset
|
3381 #endif |
| 428 | 3382 } |