Commit | Line | Data |
---|---|---|
dd3b648e RP |
1 | /* Extended regular expression matching and search library. |
2 | Copyright (C) 1985, 1989 Free Software Foundation, Inc. | |
3 | ||
4 | This program is free software; you can redistribute it and/or modify | |
5 | it under the terms of the GNU General Public License as published by | |
6 | the Free Software Foundation; either version 1, or (at your option) | |
7 | any later version. | |
8 | ||
9 | This program is distributed in the hope that it will be useful, | |
10 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | GNU General Public License for more details. | |
13 | ||
14 | You should have received a copy of the GNU General Public License | |
15 | along with this program; if not, write to the Free Software | |
16 | Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. | |
17 | ||
18 | ||
19 | In other words, you are welcome to use, share and improve this program. | |
20 | You are forbidden to forbid anyone else to use, share and improve | |
21 | what you give them. Help stamp out software-hoarding! */ | |
22 | ||
23 | ||
24 | /* To test, compile with -Dtest. | |
25 | This Dtestable feature turns this into a self-contained program | |
26 | which reads a pattern, describes how it compiles, | |
27 | then reads a string and searches for it. */ | |
28 | ||
29 | #ifdef emacs | |
30 | ||
31 | /* The `emacs' switch turns on certain special matching commands | |
32 | that make sense only in emacs. */ | |
33 | ||
34 | #include "config.h" | |
35 | #include "lisp.h" | |
36 | #include "buffer.h" | |
37 | #include "syntax.h" | |
38 | ||
39 | #else /* not emacs */ | |
40 | ||
41 | #ifdef USG | |
42 | #ifndef BSTRING | |
43 | #define bcopy(s,d,n) memcpy((d),(s),(n)) | |
44 | #define bcmp(s1,s2,n) memcmp((s1),(s2),(n)) | |
45 | #define bzero(s,n) memset((s),0,(n)) | |
46 | #endif | |
47 | #endif | |
48 | ||
49 | /* Make alloca work the best possible way. */ | |
50 | #ifdef __GNUC__ | |
51 | #define alloca __builtin_alloca | |
52 | #else | |
53 | #ifdef sparc | |
54 | #include <alloca.h> | |
55 | #endif | |
56 | #endif | |
57 | ||
58 | /* | |
59 | * Define the syntax stuff, so we can do the \<...\> things. | |
60 | */ | |
61 | ||
62 | #ifndef Sword /* must be non-zero in some of the tests below... */ | |
63 | #define Sword 1 | |
64 | #endif | |
65 | ||
66 | #define SYNTAX(c) re_syntax_table[c] | |
67 | ||
68 | #ifdef SYNTAX_TABLE | |
69 | ||
70 | char *re_syntax_table; | |
71 | ||
72 | #else | |
73 | ||
74 | static char re_syntax_table[256]; | |
75 | ||
76 | static void | |
77 | init_syntax_once () | |
78 | { | |
79 | register int c; | |
80 | static int done = 0; | |
81 | ||
82 | if (done) | |
83 | return; | |
84 | ||
85 | bzero (re_syntax_table, sizeof re_syntax_table); | |
86 | ||
87 | for (c = 'a'; c <= 'z'; c++) | |
88 | re_syntax_table[c] = Sword; | |
89 | ||
90 | for (c = 'A'; c <= 'Z'; c++) | |
91 | re_syntax_table[c] = Sword; | |
92 | ||
93 | for (c = '0'; c <= '9'; c++) | |
94 | re_syntax_table[c] = Sword; | |
95 | ||
96 | done = 1; | |
97 | } | |
98 | ||
99 | #endif /* SYNTAX_TABLE */ | |
100 | #endif /* not emacs */ | |
101 | ||
102 | #include "regex.h" | |
103 | ||
104 | /* Number of failure points to allocate space for initially, | |
105 | when matching. If this number is exceeded, more space is allocated, | |
106 | so it is not a hard limit. */ | |
107 | ||
108 | #ifndef NFAILURES | |
109 | #define NFAILURES 80 | |
110 | #endif /* NFAILURES */ | |
111 | ||
112 | /* width of a byte in bits */ | |
113 | ||
114 | #define BYTEWIDTH 8 | |
115 | ||
116 | #ifndef SIGN_EXTEND_CHAR | |
117 | #define SIGN_EXTEND_CHAR(x) (x) | |
118 | #endif | |
119 | \f | |
120 | static int obscure_syntax = 0; | |
121 | ||
122 | /* Specify the precise syntax of regexp for compilation. | |
123 | This provides for compatibility for various utilities | |
124 | which historically have different, incompatible syntaxes. | |
125 | ||
126 | The argument SYNTAX is a bit-mask containing the two bits | |
127 | RE_NO_BK_PARENS and RE_NO_BK_VBAR. */ | |
128 | ||
129 | int | |
130 | re_set_syntax (syntax) | |
131 | { | |
132 | int ret; | |
133 | ||
134 | ret = obscure_syntax; | |
135 | obscure_syntax = syntax; | |
136 | return ret; | |
137 | } | |
138 | \f | |
139 | /* re_compile_pattern takes a regular-expression string | |
140 | and converts it into a buffer full of byte commands for matching. | |
141 | ||
142 | PATTERN is the address of the pattern string | |
143 | SIZE is the length of it. | |
144 | BUFP is a struct re_pattern_buffer * which points to the info | |
145 | on where to store the byte commands. | |
146 | This structure contains a char * which points to the | |
147 | actual space, which should have been obtained with malloc. | |
148 | re_compile_pattern may use realloc to grow the buffer space. | |
149 | ||
150 | The number of bytes of commands can be found out by looking in | |
151 | the struct re_pattern_buffer that bufp pointed to, | |
152 | after re_compile_pattern returns. | |
153 | */ | |
154 | ||
155 | #define PATPUSH(ch) (*b++ = (char) (ch)) | |
156 | ||
157 | #define PATFETCH(c) \ | |
158 | {if (p == pend) goto end_of_pattern; \ | |
159 | c = * (unsigned char *) p++; \ | |
160 | if (translate) c = translate[c]; } | |
161 | ||
162 | #define PATFETCH_RAW(c) \ | |
163 | {if (p == pend) goto end_of_pattern; \ | |
164 | c = * (unsigned char *) p++; } | |
165 | ||
166 | #define PATUNFETCH p-- | |
167 | ||
168 | #define EXTEND_BUFFER \ | |
169 | { char *old_buffer = bufp->buffer; \ | |
170 | if (bufp->allocated == (1<<16)) goto too_big; \ | |
171 | bufp->allocated *= 2; \ | |
172 | if (bufp->allocated > (1<<16)) bufp->allocated = (1<<16); \ | |
173 | if (!(bufp->buffer = (char *) realloc (bufp->buffer, bufp->allocated))) \ | |
174 | goto memory_exhausted; \ | |
175 | c = bufp->buffer - old_buffer; \ | |
176 | b += c; \ | |
177 | if (fixup_jump) \ | |
178 | fixup_jump += c; \ | |
179 | if (laststart) \ | |
180 | laststart += c; \ | |
181 | begalt += c; \ | |
182 | if (pending_exact) \ | |
183 | pending_exact += c; \ | |
184 | } | |
185 | ||
186 | static int store_jump (), insert_jump (); | |
187 | ||
188 | char * | |
189 | re_compile_pattern (pattern, size, bufp) | |
190 | char *pattern; | |
191 | int size; | |
192 | struct re_pattern_buffer *bufp; | |
193 | { | |
194 | register char *b = bufp->buffer; | |
195 | register char *p = pattern; | |
196 | char *pend = pattern + size; | |
197 | register unsigned c, c1; | |
198 | char *p1; | |
199 | unsigned char *translate = (unsigned char *) bufp->translate; | |
200 | ||
201 | /* address of the count-byte of the most recently inserted "exactn" command. | |
202 | This makes it possible to tell whether a new exact-match character | |
203 | can be added to that command or requires a new "exactn" command. */ | |
204 | ||
205 | char *pending_exact = 0; | |
206 | ||
207 | /* address of the place where a forward-jump should go | |
208 | to the end of the containing expression. | |
209 | Each alternative of an "or", except the last, ends with a forward-jump | |
210 | of this sort. */ | |
211 | ||
212 | char *fixup_jump = 0; | |
213 | ||
214 | /* address of start of the most recently finished expression. | |
215 | This tells postfix * where to find the start of its operand. */ | |
216 | ||
217 | char *laststart = 0; | |
218 | ||
219 | /* In processing a repeat, 1 means zero matches is allowed */ | |
220 | ||
221 | char zero_times_ok; | |
222 | ||
223 | /* In processing a repeat, 1 means many matches is allowed */ | |
224 | ||
225 | char many_times_ok; | |
226 | ||
227 | /* address of beginning of regexp, or inside of last \( */ | |
228 | ||
229 | char *begalt = b; | |
230 | ||
231 | /* Stack of information saved by \( and restored by \). | |
232 | Four stack elements are pushed by each \(: | |
233 | First, the value of b. | |
234 | Second, the value of fixup_jump. | |
235 | Third, the value of regnum. | |
236 | Fourth, the value of begalt. */ | |
237 | ||
238 | int stackb[40]; | |
239 | int *stackp = stackb; | |
240 | int *stacke = stackb + 40; | |
241 | int *stackt; | |
242 | ||
243 | /* Counts \('s as they are encountered. Remembered for the matching \), | |
244 | where it becomes the "register number" to put in the stop_memory command */ | |
245 | ||
246 | int regnum = 1; | |
247 | ||
248 | bufp->fastmap_accurate = 0; | |
249 | ||
250 | #ifndef emacs | |
251 | #ifndef SYNTAX_TABLE | |
252 | /* | |
253 | * Initialize the syntax table. | |
254 | */ | |
255 | init_syntax_once(); | |
256 | #endif | |
257 | #endif | |
258 | ||
259 | if (bufp->allocated == 0) | |
260 | { | |
261 | bufp->allocated = 28; | |
262 | if (bufp->buffer) | |
263 | /* EXTEND_BUFFER loses when bufp->allocated is 0 */ | |
264 | bufp->buffer = (char *) realloc (bufp->buffer, 28); | |
265 | else | |
266 | /* Caller did not allocate a buffer. Do it for him */ | |
267 | bufp->buffer = (char *) malloc (28); | |
268 | if (!bufp->buffer) goto memory_exhausted; | |
269 | begalt = b = bufp->buffer; | |
270 | } | |
271 | ||
272 | while (p != pend) | |
273 | { | |
274 | if (b - bufp->buffer > bufp->allocated - 10) | |
275 | /* Note that EXTEND_BUFFER clobbers c */ | |
276 | EXTEND_BUFFER; | |
277 | ||
278 | PATFETCH (c); | |
279 | ||
280 | switch (c) | |
281 | { | |
282 | case '$': | |
283 | if (obscure_syntax & RE_TIGHT_VBAR) | |
284 | { | |
285 | if (! (obscure_syntax & RE_CONTEXT_INDEP_OPS) && p != pend) | |
286 | goto normal_char; | |
287 | /* Make operand of last vbar end before this `$'. */ | |
288 | if (fixup_jump) | |
289 | store_jump (fixup_jump, jump, b); | |
290 | fixup_jump = 0; | |
291 | PATPUSH (endline); | |
292 | break; | |
293 | } | |
294 | ||
295 | /* $ means succeed if at end of line, but only in special contexts. | |
296 | If randomly in the middle of a pattern, it is a normal character. */ | |
297 | if (p == pend || *p == '\n' | |
298 | || (obscure_syntax & RE_CONTEXT_INDEP_OPS) | |
299 | || (obscure_syntax & RE_NO_BK_PARENS | |
300 | ? *p == ')' | |
301 | : *p == '\\' && p[1] == ')') | |
302 | || (obscure_syntax & RE_NO_BK_VBAR | |
303 | ? *p == '|' | |
304 | : *p == '\\' && p[1] == '|')) | |
305 | { | |
306 | PATPUSH (endline); | |
307 | break; | |
308 | } | |
309 | goto normal_char; | |
310 | ||
311 | case '^': | |
312 | /* ^ means succeed if at beg of line, but only if no preceding pattern. */ | |
313 | ||
314 | if (laststart && p[-2] != '\n' | |
315 | && ! (obscure_syntax & RE_CONTEXT_INDEP_OPS)) | |
316 | goto normal_char; | |
317 | if (obscure_syntax & RE_TIGHT_VBAR) | |
318 | { | |
319 | if (p != pattern + 1 | |
320 | && ! (obscure_syntax & RE_CONTEXT_INDEP_OPS)) | |
321 | goto normal_char; | |
322 | PATPUSH (begline); | |
323 | begalt = b; | |
324 | } | |
325 | else | |
326 | PATPUSH (begline); | |
327 | break; | |
328 | ||
329 | case '+': | |
330 | case '?': | |
331 | if (obscure_syntax & RE_BK_PLUS_QM) | |
332 | goto normal_char; | |
333 | handle_plus: | |
334 | case '*': | |
335 | /* If there is no previous pattern, char not special. */ | |
336 | if (!laststart && ! (obscure_syntax & RE_CONTEXT_INDEP_OPS)) | |
337 | goto normal_char; | |
338 | /* If there is a sequence of repetition chars, | |
339 | collapse it down to equivalent to just one. */ | |
340 | zero_times_ok = 0; | |
341 | many_times_ok = 0; | |
342 | while (1) | |
343 | { | |
344 | zero_times_ok |= c != '+'; | |
345 | many_times_ok |= c != '?'; | |
346 | if (p == pend) | |
347 | break; | |
348 | PATFETCH (c); | |
349 | if (c == '*') | |
350 | ; | |
351 | else if (!(obscure_syntax & RE_BK_PLUS_QM) | |
352 | && (c == '+' || c == '?')) | |
353 | ; | |
354 | else if ((obscure_syntax & RE_BK_PLUS_QM) | |
355 | && c == '\\') | |
356 | { | |
357 | int c1; | |
358 | PATFETCH (c1); | |
359 | if (!(c1 == '+' || c1 == '?')) | |
360 | { | |
361 | PATUNFETCH; | |
362 | PATUNFETCH; | |
363 | break; | |
364 | } | |
365 | c = c1; | |
366 | } | |
367 | else | |
368 | { | |
369 | PATUNFETCH; | |
370 | break; | |
371 | } | |
372 | } | |
373 | ||
374 | /* Star, etc. applied to an empty pattern is equivalent | |
375 | to an empty pattern. */ | |
376 | if (!laststart) | |
377 | break; | |
378 | ||
379 | /* Now we know whether 0 matches is allowed, | |
380 | and whether 2 or more matches is allowed. */ | |
381 | if (many_times_ok) | |
382 | { | |
383 | /* If more than one repetition is allowed, | |
384 | put in a backward jump at the end. */ | |
385 | store_jump (b, maybe_finalize_jump, laststart - 3); | |
386 | b += 3; | |
387 | } | |
388 | insert_jump (on_failure_jump, laststart, b + 3, b); | |
389 | pending_exact = 0; | |
390 | b += 3; | |
391 | if (!zero_times_ok) | |
392 | { | |
393 | /* At least one repetition required: insert before the loop | |
394 | a skip over the initial on-failure-jump instruction */ | |
395 | insert_jump (dummy_failure_jump, laststart, laststart + 6, b); | |
396 | b += 3; | |
397 | } | |
398 | break; | |
399 | ||
400 | case '.': | |
401 | laststart = b; | |
402 | PATPUSH (anychar); | |
403 | break; | |
404 | ||
405 | case '[': | |
406 | while (b - bufp->buffer | |
407 | > bufp->allocated - 3 - (1 << BYTEWIDTH) / BYTEWIDTH) | |
408 | /* Note that EXTEND_BUFFER clobbers c */ | |
409 | EXTEND_BUFFER; | |
410 | ||
411 | laststart = b; | |
412 | if (*p == '^') | |
413 | PATPUSH (charset_not), p++; | |
414 | else | |
415 | PATPUSH (charset); | |
416 | p1 = p; | |
417 | ||
418 | PATPUSH ((1 << BYTEWIDTH) / BYTEWIDTH); | |
419 | /* Clear the whole map */ | |
420 | bzero (b, (1 << BYTEWIDTH) / BYTEWIDTH); | |
421 | /* Read in characters and ranges, setting map bits */ | |
422 | while (1) | |
423 | { | |
424 | PATFETCH (c); | |
425 | if (c == ']' && p != p1 + 1) break; | |
426 | if (*p == '-' && p[1] != ']') | |
427 | { | |
428 | PATFETCH (c1); | |
429 | PATFETCH (c1); | |
430 | while (c <= c1) | |
431 | b[c / BYTEWIDTH] |= 1 << (c % BYTEWIDTH), c++; | |
432 | } | |
433 | else | |
434 | { | |
435 | b[c / BYTEWIDTH] |= 1 << (c % BYTEWIDTH); | |
436 | } | |
437 | } | |
438 | /* Discard any bitmap bytes that are all 0 at the end of the map. | |
439 | Decrement the map-length byte too. */ | |
440 | while ((int) b[-1] > 0 && b[b[-1] - 1] == 0) | |
441 | b[-1]--; | |
442 | b += b[-1]; | |
443 | break; | |
444 | ||
445 | case '(': | |
446 | if (! (obscure_syntax & RE_NO_BK_PARENS)) | |
447 | goto normal_char; | |
448 | else | |
449 | goto handle_open; | |
450 | ||
451 | case ')': | |
452 | if (! (obscure_syntax & RE_NO_BK_PARENS)) | |
453 | goto normal_char; | |
454 | else | |
455 | goto handle_close; | |
456 | ||
457 | case '\n': | |
458 | if (! (obscure_syntax & RE_NEWLINE_OR)) | |
459 | goto normal_char; | |
460 | else | |
461 | goto handle_bar; | |
462 | ||
463 | case '|': | |
464 | if (! (obscure_syntax & RE_NO_BK_VBAR)) | |
465 | goto normal_char; | |
466 | else | |
467 | goto handle_bar; | |
468 | ||
469 | case '\\': | |
470 | if (p == pend) goto invalid_pattern; | |
471 | PATFETCH_RAW (c); | |
472 | switch (c) | |
473 | { | |
474 | case '(': | |
475 | if (obscure_syntax & RE_NO_BK_PARENS) | |
476 | goto normal_backsl; | |
477 | handle_open: | |
478 | if (stackp == stacke) goto nesting_too_deep; | |
479 | if (regnum < RE_NREGS) | |
480 | { | |
481 | PATPUSH (start_memory); | |
482 | PATPUSH (regnum); | |
483 | } | |
484 | *stackp++ = b - bufp->buffer; | |
485 | *stackp++ = fixup_jump ? fixup_jump - bufp->buffer + 1 : 0; | |
486 | *stackp++ = regnum++; | |
487 | *stackp++ = begalt - bufp->buffer; | |
488 | fixup_jump = 0; | |
489 | laststart = 0; | |
490 | begalt = b; | |
491 | break; | |
492 | ||
493 | case ')': | |
494 | if (obscure_syntax & RE_NO_BK_PARENS) | |
495 | goto normal_backsl; | |
496 | handle_close: | |
497 | if (stackp == stackb) goto unmatched_close; | |
498 | begalt = *--stackp + bufp->buffer; | |
499 | if (fixup_jump) | |
500 | store_jump (fixup_jump, jump, b); | |
501 | if (stackp[-1] < RE_NREGS) | |
502 | { | |
503 | PATPUSH (stop_memory); | |
504 | PATPUSH (stackp[-1]); | |
505 | } | |
506 | stackp -= 2; | |
507 | fixup_jump = 0; | |
508 | if (*stackp) | |
509 | fixup_jump = *stackp + bufp->buffer - 1; | |
510 | laststart = *--stackp + bufp->buffer; | |
511 | break; | |
512 | ||
513 | case '|': | |
514 | if (obscure_syntax & RE_NO_BK_VBAR) | |
515 | goto normal_backsl; | |
516 | handle_bar: | |
517 | insert_jump (on_failure_jump, begalt, b + 6, b); | |
518 | pending_exact = 0; | |
519 | b += 3; | |
520 | if (fixup_jump) | |
521 | store_jump (fixup_jump, jump, b); | |
522 | fixup_jump = b; | |
523 | b += 3; | |
524 | laststart = 0; | |
525 | begalt = b; | |
526 | break; | |
527 | ||
528 | #ifdef emacs | |
529 | case '=': | |
530 | PATPUSH (at_dot); | |
531 | break; | |
532 | ||
533 | case 's': | |
534 | laststart = b; | |
535 | PATPUSH (syntaxspec); | |
536 | PATFETCH (c); | |
537 | PATPUSH (syntax_spec_code[c]); | |
538 | break; | |
539 | ||
540 | case 'S': | |
541 | laststart = b; | |
542 | PATPUSH (notsyntaxspec); | |
543 | PATFETCH (c); | |
544 | PATPUSH (syntax_spec_code[c]); | |
545 | break; | |
546 | #endif /* emacs */ | |
547 | ||
548 | case 'w': | |
549 | laststart = b; | |
550 | PATPUSH (wordchar); | |
551 | break; | |
552 | ||
553 | case 'W': | |
554 | laststart = b; | |
555 | PATPUSH (notwordchar); | |
556 | break; | |
557 | ||
558 | case '<': | |
559 | PATPUSH (wordbeg); | |
560 | break; | |
561 | ||
562 | case '>': | |
563 | PATPUSH (wordend); | |
564 | break; | |
565 | ||
566 | case 'b': | |
567 | PATPUSH (wordbound); | |
568 | break; | |
569 | ||
570 | case 'B': | |
571 | PATPUSH (notwordbound); | |
572 | break; | |
573 | ||
574 | case '`': | |
575 | PATPUSH (begbuf); | |
576 | break; | |
577 | ||
578 | case '\'': | |
579 | PATPUSH (endbuf); | |
580 | break; | |
581 | ||
582 | case '1': | |
583 | case '2': | |
584 | case '3': | |
585 | case '4': | |
586 | case '5': | |
587 | case '6': | |
588 | case '7': | |
589 | case '8': | |
590 | case '9': | |
591 | c1 = c - '0'; | |
592 | if (c1 >= regnum) | |
593 | goto normal_char; | |
594 | for (stackt = stackp - 2; stackt > stackb; stackt -= 4) | |
595 | if (*stackt == c1) | |
596 | goto normal_char; | |
597 | laststart = b; | |
598 | PATPUSH (duplicate); | |
599 | PATPUSH (c1); | |
600 | break; | |
601 | ||
602 | case '+': | |
603 | case '?': | |
604 | if (obscure_syntax & RE_BK_PLUS_QM) | |
605 | goto handle_plus; | |
606 | ||
607 | default: | |
608 | normal_backsl: | |
609 | /* You might think it would be useful for \ to mean | |
610 | not to translate; but if we don't translate it | |
611 | it will never match anything. */ | |
612 | if (translate) c = translate[c]; | |
613 | goto normal_char; | |
614 | } | |
615 | break; | |
616 | ||
617 | default: | |
618 | normal_char: | |
619 | if (!pending_exact || pending_exact + *pending_exact + 1 != b | |
620 | || *pending_exact == 0177 || *p == '*' || *p == '^' | |
621 | || ((obscure_syntax & RE_BK_PLUS_QM) | |
622 | ? *p == '\\' && (p[1] == '+' || p[1] == '?') | |
623 | : (*p == '+' || *p == '?'))) | |
624 | { | |
625 | laststart = b; | |
626 | PATPUSH (exactn); | |
627 | pending_exact = b; | |
628 | PATPUSH (0); | |
629 | } | |
630 | PATPUSH (c); | |
631 | (*pending_exact)++; | |
632 | } | |
633 | } | |
634 | ||
635 | if (fixup_jump) | |
636 | store_jump (fixup_jump, jump, b); | |
637 | ||
638 | if (stackp != stackb) goto unmatched_open; | |
639 | ||
640 | bufp->used = b - bufp->buffer; | |
641 | return 0; | |
642 | ||
643 | invalid_pattern: | |
644 | return "Invalid regular expression"; | |
645 | ||
646 | unmatched_open: | |
647 | return "Unmatched \\("; | |
648 | ||
649 | unmatched_close: | |
650 | return "Unmatched \\)"; | |
651 | ||
652 | end_of_pattern: | |
653 | return "Premature end of regular expression"; | |
654 | ||
655 | nesting_too_deep: | |
656 | return "Nesting too deep"; | |
657 | ||
658 | too_big: | |
659 | return "Regular expression too big"; | |
660 | ||
661 | memory_exhausted: | |
662 | return "Memory exhausted"; | |
663 | } | |
664 | ||
665 | /* Store where `from' points a jump operation to jump to where `to' points. | |
666 | `opcode' is the opcode to store. */ | |
667 | ||
668 | static int | |
669 | store_jump (from, opcode, to) | |
670 | char *from, *to; | |
671 | char opcode; | |
672 | { | |
673 | from[0] = opcode; | |
674 | from[1] = (to - (from + 3)) & 0377; | |
675 | from[2] = (to - (from + 3)) >> 8; | |
676 | } | |
677 | ||
678 | /* Open up space at char FROM, and insert there a jump to TO. | |
679 | CURRENT_END gives te end of the storage no in use, | |
680 | so we know how much data to copy up. | |
681 | OP is the opcode of the jump to insert. | |
682 | ||
683 | If you call this function, you must zero out pending_exact. */ | |
684 | ||
685 | static int | |
686 | insert_jump (op, from, to, current_end) | |
687 | char op; | |
688 | char *from, *to, *current_end; | |
689 | { | |
690 | register char *pto = current_end + 3; | |
691 | register char *pfrom = current_end; | |
692 | while (pfrom != from) | |
693 | *--pto = *--pfrom; | |
694 | store_jump (from, op, to); | |
695 | } | |
696 | \f | |
697 | /* Given a pattern, compute a fastmap from it. | |
698 | The fastmap records which of the (1 << BYTEWIDTH) possible characters | |
699 | can start a string that matches the pattern. | |
700 | This fastmap is used by re_search to skip quickly over totally implausible text. | |
701 | ||
702 | The caller must supply the address of a (1 << BYTEWIDTH)-byte data area | |
703 | as bufp->fastmap. | |
704 | The other components of bufp describe the pattern to be used. */ | |
705 | ||
706 | void | |
707 | re_compile_fastmap (bufp) | |
708 | struct re_pattern_buffer *bufp; | |
709 | { | |
710 | unsigned char *pattern = (unsigned char *) bufp->buffer; | |
711 | int size = bufp->used; | |
712 | register char *fastmap = bufp->fastmap; | |
713 | register unsigned char *p = pattern; | |
714 | register unsigned char *pend = pattern + size; | |
715 | register int j, k; | |
716 | unsigned char *translate = (unsigned char *) bufp->translate; | |
717 | ||
718 | unsigned char *stackb[NFAILURES]; | |
719 | unsigned char **stackp = stackb; | |
720 | ||
721 | bzero (fastmap, (1 << BYTEWIDTH)); | |
722 | bufp->fastmap_accurate = 1; | |
723 | bufp->can_be_null = 0; | |
724 | ||
725 | while (p) | |
726 | { | |
727 | if (p == pend) | |
728 | { | |
729 | bufp->can_be_null = 1; | |
730 | break; | |
731 | } | |
732 | #ifdef SWITCH_ENUM_BUG | |
733 | switch ((int) ((enum regexpcode) *p++)) | |
734 | #else | |
735 | switch ((enum regexpcode) *p++) | |
736 | #endif | |
737 | { | |
738 | case exactn: | |
739 | if (translate) | |
740 | fastmap[translate[p[1]]] = 1; | |
741 | else | |
742 | fastmap[p[1]] = 1; | |
743 | break; | |
744 | ||
745 | case begline: | |
746 | case before_dot: | |
747 | case at_dot: | |
748 | case after_dot: | |
749 | case begbuf: | |
750 | case endbuf: | |
751 | case wordbound: | |
752 | case notwordbound: | |
753 | case wordbeg: | |
754 | case wordend: | |
755 | continue; | |
756 | ||
757 | case endline: | |
758 | if (translate) | |
759 | fastmap[translate['\n']] = 1; | |
760 | else | |
761 | fastmap['\n'] = 1; | |
762 | if (bufp->can_be_null != 1) | |
763 | bufp->can_be_null = 2; | |
764 | break; | |
765 | ||
766 | case finalize_jump: | |
767 | case maybe_finalize_jump: | |
768 | case jump: | |
769 | case dummy_failure_jump: | |
770 | bufp->can_be_null = 1; | |
771 | j = *p++ & 0377; | |
772 | j += SIGN_EXTEND_CHAR (*(char *)p) << 8; | |
773 | p += j + 1; /* The 1 compensates for missing ++ above */ | |
774 | if (j > 0) | |
775 | continue; | |
776 | /* Jump backward reached implies we just went through | |
777 | the body of a loop and matched nothing. | |
778 | Opcode jumped to should be an on_failure_jump. | |
779 | Just treat it like an ordinary jump. | |
780 | For a * loop, it has pushed its failure point already; | |
781 | if so, discard that as redundant. */ | |
782 | if ((enum regexpcode) *p != on_failure_jump) | |
783 | continue; | |
784 | p++; | |
785 | j = *p++ & 0377; | |
786 | j += SIGN_EXTEND_CHAR (*(char *)p) << 8; | |
787 | p += j + 1; /* The 1 compensates for missing ++ above */ | |
788 | if (stackp != stackb && *stackp == p) | |
789 | stackp--; | |
790 | continue; | |
791 | ||
792 | case on_failure_jump: | |
793 | j = *p++ & 0377; | |
794 | j += SIGN_EXTEND_CHAR (*(char *)p) << 8; | |
795 | p++; | |
796 | *++stackp = p + j; | |
797 | continue; | |
798 | ||
799 | case start_memory: | |
800 | case stop_memory: | |
801 | p++; | |
802 | continue; | |
803 | ||
804 | case duplicate: | |
805 | bufp->can_be_null = 1; | |
806 | fastmap['\n'] = 1; | |
807 | case anychar: | |
808 | for (j = 0; j < (1 << BYTEWIDTH); j++) | |
809 | if (j != '\n') | |
810 | fastmap[j] = 1; | |
811 | if (bufp->can_be_null) | |
812 | return; | |
813 | /* Don't return; check the alternative paths | |
814 | so we can set can_be_null if appropriate. */ | |
815 | break; | |
816 | ||
817 | case wordchar: | |
818 | for (j = 0; j < (1 << BYTEWIDTH); j++) | |
819 | if (SYNTAX (j) == Sword) | |
820 | fastmap[j] = 1; | |
821 | break; | |
822 | ||
823 | case notwordchar: | |
824 | for (j = 0; j < (1 << BYTEWIDTH); j++) | |
825 | if (SYNTAX (j) != Sword) | |
826 | fastmap[j] = 1; | |
827 | break; | |
828 | ||
829 | #ifdef emacs | |
830 | case syntaxspec: | |
831 | k = *p++; | |
832 | for (j = 0; j < (1 << BYTEWIDTH); j++) | |
833 | if (SYNTAX (j) == (enum syntaxcode) k) | |
834 | fastmap[j] = 1; | |
835 | break; | |
836 | ||
837 | case notsyntaxspec: | |
838 | k = *p++; | |
839 | for (j = 0; j < (1 << BYTEWIDTH); j++) | |
840 | if (SYNTAX (j) != (enum syntaxcode) k) | |
841 | fastmap[j] = 1; | |
842 | break; | |
843 | #endif /* emacs */ | |
844 | ||
845 | case charset: | |
846 | for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--) | |
847 | if (p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))) | |
848 | { | |
849 | if (translate) | |
850 | fastmap[translate[j]] = 1; | |
851 | else | |
852 | fastmap[j] = 1; | |
853 | } | |
854 | break; | |
855 | ||
856 | case charset_not: | |
857 | /* Chars beyond end of map must be allowed */ | |
858 | for (j = *p * BYTEWIDTH; j < (1 << BYTEWIDTH); j++) | |
859 | if (translate) | |
860 | fastmap[translate[j]] = 1; | |
861 | else | |
862 | fastmap[j] = 1; | |
863 | ||
864 | for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--) | |
865 | if (!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH)))) | |
866 | { | |
867 | if (translate) | |
868 | fastmap[translate[j]] = 1; | |
869 | else | |
870 | fastmap[j] = 1; | |
871 | } | |
872 | break; | |
873 | } | |
874 | ||
875 | /* Get here means we have successfully found the possible starting characters | |
876 | of one path of the pattern. We need not follow this path any farther. | |
877 | Instead, look at the next alternative remembered in the stack. */ | |
878 | if (stackp != stackb) | |
879 | p = *stackp--; | |
880 | else | |
881 | break; | |
882 | } | |
883 | } | |
884 | \f | |
885 | /* Like re_search_2, below, but only one string is specified. */ | |
886 | ||
887 | int | |
888 | re_search (pbufp, string, size, startpos, range, regs) | |
889 | struct re_pattern_buffer *pbufp; | |
890 | char *string; | |
891 | int size, startpos, range; | |
892 | struct re_registers *regs; | |
893 | { | |
894 | return re_search_2 (pbufp, 0, 0, string, size, startpos, range, regs, size); | |
895 | } | |
896 | ||
897 | /* Like re_match_2 but tries first a match starting at index STARTPOS, | |
898 | then at STARTPOS + 1, and so on. | |
899 | RANGE is the number of places to try before giving up. | |
900 | If RANGE is negative, the starting positions tried are | |
901 | STARTPOS, STARTPOS - 1, etc. | |
902 | It is up to the caller to make sure that range is not so large | |
903 | as to take the starting position outside of the input strings. | |
904 | ||
905 | The value returned is the position at which the match was found, | |
906 | or -1 if no match was found, | |
907 | or -2 if error (such as failure stack overflow). */ | |
908 | ||
909 | int | |
910 | re_search_2 (pbufp, string1, size1, string2, size2, startpos, range, regs, mstop) | |
911 | struct re_pattern_buffer *pbufp; | |
912 | char *string1, *string2; | |
913 | int size1, size2; | |
914 | int startpos; | |
915 | register int range; | |
916 | struct re_registers *regs; | |
917 | int mstop; | |
918 | { | |
919 | register char *fastmap = pbufp->fastmap; | |
920 | register unsigned char *translate = (unsigned char *) pbufp->translate; | |
921 | int total = size1 + size2; | |
922 | int val; | |
923 | ||
924 | /* Update the fastmap now if not correct already */ | |
925 | if (fastmap && !pbufp->fastmap_accurate) | |
926 | re_compile_fastmap (pbufp); | |
927 | ||
928 | /* Don't waste time in a long search for a pattern | |
929 | that says it is anchored. */ | |
930 | if (pbufp->used > 0 && (enum regexpcode) pbufp->buffer[0] == begbuf | |
931 | && range > 0) | |
932 | { | |
933 | if (startpos > 0) | |
934 | return -1; | |
935 | else | |
936 | range = 1; | |
937 | } | |
938 | ||
939 | while (1) | |
940 | { | |
941 | /* If a fastmap is supplied, skip quickly over characters | |
942 | that cannot possibly be the start of a match. | |
943 | Note, however, that if the pattern can possibly match | |
944 | the null string, we must test it at each starting point | |
945 | so that we take the first null string we get. */ | |
946 | ||
947 | if (fastmap && startpos < total && pbufp->can_be_null != 1) | |
948 | { | |
949 | if (range > 0) | |
950 | { | |
951 | register int lim = 0; | |
952 | register unsigned char *p; | |
953 | int irange = range; | |
954 | if (startpos < size1 && startpos + range >= size1) | |
955 | lim = range - (size1 - startpos); | |
956 | ||
957 | p = ((unsigned char *) | |
958 | &(startpos >= size1 ? string2 - size1 : string1)[startpos]); | |
959 | ||
960 | if (translate) | |
961 | { | |
962 | while (range > lim && !fastmap[translate[*p++]]) | |
963 | range--; | |
964 | } | |
965 | else | |
966 | { | |
967 | while (range > lim && !fastmap[*p++]) | |
968 | range--; | |
969 | } | |
970 | startpos += irange - range; | |
971 | } | |
972 | else | |
973 | { | |
974 | register unsigned char c; | |
975 | if (startpos >= size1) | |
976 | c = string2[startpos - size1]; | |
977 | else | |
978 | c = string1[startpos]; | |
979 | c &= 0xff; | |
980 | if (translate ? !fastmap[translate[c]] : !fastmap[c]) | |
981 | goto advance; | |
982 | } | |
983 | } | |
984 | ||
985 | if (range >= 0 && startpos == total | |
986 | && fastmap && pbufp->can_be_null == 0) | |
987 | return -1; | |
988 | ||
989 | val = re_match_2 (pbufp, string1, size1, string2, size2, startpos, regs, mstop); | |
990 | if (0 <= val) | |
991 | { | |
992 | if (val == -2) | |
993 | return -2; | |
994 | return startpos; | |
995 | } | |
996 | ||
997 | #ifdef C_ALLOCA | |
998 | alloca (0); | |
999 | #endif /* C_ALLOCA */ | |
1000 | ||
1001 | advance: | |
1002 | if (!range) break; | |
1003 | if (range > 0) range--, startpos++; else range++, startpos--; | |
1004 | } | |
1005 | return -1; | |
1006 | } | |
1007 | \f | |
1008 | #ifndef emacs /* emacs never uses this */ | |
1009 | int | |
1010 | re_match (pbufp, string, size, pos, regs) | |
1011 | struct re_pattern_buffer *pbufp; | |
1012 | char *string; | |
1013 | int size, pos; | |
1014 | struct re_registers *regs; | |
1015 | { | |
1016 | return re_match_2 (pbufp, 0, 0, string, size, pos, regs, size); | |
1017 | } | |
1018 | #endif /* emacs */ | |
1019 | ||
1020 | /* Maximum size of failure stack. Beyond this, overflow is an error. */ | |
1021 | ||
1022 | int re_max_failures = 2000; | |
1023 | ||
1024 | static int bcmp_translate(); | |
1025 | /* Match the pattern described by PBUFP | |
1026 | against data which is the virtual concatenation of STRING1 and STRING2. | |
1027 | SIZE1 and SIZE2 are the sizes of the two data strings. | |
1028 | Start the match at position POS. | |
1029 | Do not consider matching past the position MSTOP. | |
1030 | ||
1031 | If pbufp->fastmap is nonzero, then it had better be up to date. | |
1032 | ||
1033 | The reason that the data to match are specified as two components | |
1034 | which are to be regarded as concatenated | |
1035 | is so this function can be used directly on the contents of an Emacs buffer. | |
1036 | ||
1037 | -1 is returned if there is no match. -2 is returned if there is | |
1038 | an error (such as match stack overflow). Otherwise the value is the length | |
1039 | of the substring which was matched. */ | |
1040 | ||
1041 | int | |
1042 | re_match_2 (pbufp, string1, size1, string2, size2, pos, regs, mstop) | |
1043 | struct re_pattern_buffer *pbufp; | |
1044 | unsigned char *string1, *string2; | |
1045 | int size1, size2; | |
1046 | int pos; | |
1047 | struct re_registers *regs; | |
1048 | int mstop; | |
1049 | { | |
1050 | register unsigned char *p = (unsigned char *) pbufp->buffer; | |
1051 | register unsigned char *pend = p + pbufp->used; | |
1052 | /* End of first string */ | |
1053 | unsigned char *end1; | |
1054 | /* End of second string */ | |
1055 | unsigned char *end2; | |
1056 | /* Pointer just past last char to consider matching */ | |
1057 | unsigned char *end_match_1, *end_match_2; | |
1058 | register unsigned char *d, *dend; | |
1059 | register int mcnt; | |
1060 | unsigned char *translate = (unsigned char *) pbufp->translate; | |
1061 | ||
1062 | /* Failure point stack. Each place that can handle a failure further down the line | |
1063 | pushes a failure point on this stack. It consists of two char *'s. | |
1064 | The first one pushed is where to resume scanning the pattern; | |
1065 | the second pushed is where to resume scanning the strings. | |
1066 | If the latter is zero, the failure point is a "dummy". | |
1067 | If a failure happens and the innermost failure point is dormant, | |
1068 | it discards that failure point and tries the next one. */ | |
1069 | ||
1070 | unsigned char *initial_stack[2 * NFAILURES]; | |
1071 | unsigned char **stackb = initial_stack; | |
1072 | unsigned char **stackp = stackb, **stacke = &stackb[2 * NFAILURES]; | |
1073 | ||
1074 | /* Information on the "contents" of registers. | |
1075 | These are pointers into the input strings; they record | |
1076 | just what was matched (on this attempt) by some part of the pattern. | |
1077 | The start_memory command stores the start of a register's contents | |
1078 | and the stop_memory command stores the end. | |
1079 | ||
1080 | At that point, regstart[regnum] points to the first character in the register, | |
1081 | regend[regnum] points to the first character beyond the end of the register, | |
1082 | regstart_seg1[regnum] is true iff regstart[regnum] points into string1, | |
1083 | and regend_seg1[regnum] is true iff regend[regnum] points into string1. */ | |
1084 | ||
1085 | unsigned char *regstart[RE_NREGS]; | |
1086 | unsigned char *regend[RE_NREGS]; | |
1087 | unsigned char regstart_seg1[RE_NREGS], regend_seg1[RE_NREGS]; | |
1088 | ||
1089 | /* Set up pointers to ends of strings. | |
1090 | Don't allow the second string to be empty unless both are empty. */ | |
1091 | if (!size2) | |
1092 | { | |
1093 | string2 = string1; | |
1094 | size2 = size1; | |
1095 | string1 = 0; | |
1096 | size1 = 0; | |
1097 | } | |
1098 | end1 = string1 + size1; | |
1099 | end2 = string2 + size2; | |
1100 | ||
1101 | /* Compute where to stop matching, within the two strings */ | |
1102 | if (mstop <= size1) | |
1103 | { | |
1104 | end_match_1 = string1 + mstop; | |
1105 | end_match_2 = string2; | |
1106 | } | |
1107 | else | |
1108 | { | |
1109 | end_match_1 = end1; | |
1110 | end_match_2 = string2 + mstop - size1; | |
1111 | } | |
1112 | ||
1113 | /* Initialize \) text positions to -1 | |
1114 | to mark ones that no \( or \) has been seen for. */ | |
1115 | ||
1116 | for (mcnt = 0; mcnt < sizeof (regend) / sizeof (*regend); mcnt++) | |
1117 | regend[mcnt] = (unsigned char *) -1; | |
1118 | ||
1119 | /* `p' scans through the pattern as `d' scans through the data. | |
1120 | `dend' is the end of the input string that `d' points within. | |
1121 | `d' is advanced into the following input string whenever necessary, | |
1122 | but this happens before fetching; | |
1123 | therefore, at the beginning of the loop, | |
1124 | `d' can be pointing at the end of a string, | |
1125 | but it cannot equal string2. */ | |
1126 | ||
1127 | if (pos <= size1) | |
1128 | d = string1 + pos, dend = end_match_1; | |
1129 | else | |
1130 | d = string2 + pos - size1, dend = end_match_2; | |
1131 | ||
1132 | /* Write PREFETCH; just before fetching a character with *d. */ | |
1133 | #define PREFETCH \ | |
1134 | while (d == dend) \ | |
1135 | { if (dend == end_match_2) goto fail; /* end of string2 => failure */ \ | |
1136 | d = string2; /* end of string1 => advance to string2. */ \ | |
1137 | dend = end_match_2; } | |
1138 | ||
1139 | /* This loop loops over pattern commands. | |
1140 | It exits by returning from the function if match is complete, | |
1141 | or it drops through if match fails at this starting point in the input data. */ | |
1142 | ||
1143 | while (1) | |
1144 | { | |
1145 | if (p == pend) | |
1146 | /* End of pattern means we have succeeded! */ | |
1147 | { | |
1148 | /* If caller wants register contents data back, convert it to indices */ | |
1149 | if (regs) | |
1150 | { | |
1151 | regs->start[0] = pos; | |
1152 | if (dend == end_match_1) | |
1153 | regs->end[0] = d - string1; | |
1154 | else | |
1155 | regs->end[0] = d - string2 + size1; | |
1156 | for (mcnt = 1; mcnt < RE_NREGS; mcnt++) | |
1157 | { | |
1158 | if (regend[mcnt] == (unsigned char *) -1) | |
1159 | { | |
1160 | regs->start[mcnt] = -1; | |
1161 | regs->end[mcnt] = -1; | |
1162 | continue; | |
1163 | } | |
1164 | if (regstart_seg1[mcnt]) | |
1165 | regs->start[mcnt] = regstart[mcnt] - string1; | |
1166 | else | |
1167 | regs->start[mcnt] = regstart[mcnt] - string2 + size1; | |
1168 | if (regend_seg1[mcnt]) | |
1169 | regs->end[mcnt] = regend[mcnt] - string1; | |
1170 | else | |
1171 | regs->end[mcnt] = regend[mcnt] - string2 + size1; | |
1172 | } | |
1173 | } | |
1174 | if (dend == end_match_1) | |
1175 | return (d - string1 - pos); | |
1176 | else | |
1177 | return d - string2 + size1 - pos; | |
1178 | } | |
1179 | ||
1180 | /* Otherwise match next pattern command */ | |
1181 | #ifdef SWITCH_ENUM_BUG | |
1182 | switch ((int) ((enum regexpcode) *p++)) | |
1183 | #else | |
1184 | switch ((enum regexpcode) *p++) | |
1185 | #endif | |
1186 | { | |
1187 | ||
1188 | /* \( is represented by a start_memory, \) by a stop_memory. | |
1189 | Both of those commands contain a "register number" argument. | |
1190 | The text matched within the \( and \) is recorded under that number. | |
1191 | Then, \<digit> turns into a `duplicate' command which | |
1192 | is followed by the numeric value of <digit> as the register number. */ | |
1193 | ||
1194 | case start_memory: | |
1195 | regstart[*p] = d; | |
1196 | regstart_seg1[*p++] = (dend == end_match_1); | |
1197 | break; | |
1198 | ||
1199 | case stop_memory: | |
1200 | regend[*p] = d; | |
1201 | regend_seg1[*p++] = (dend == end_match_1); | |
1202 | break; | |
1203 | ||
1204 | case duplicate: | |
1205 | { | |
1206 | int regno = *p++; /* Get which register to match against */ | |
1207 | register unsigned char *d2, *dend2; | |
1208 | ||
1209 | d2 = regstart[regno]; | |
1210 | dend2 = ((regstart_seg1[regno] == regend_seg1[regno]) | |
1211 | ? regend[regno] : end_match_1); | |
1212 | while (1) | |
1213 | { | |
1214 | /* Advance to next segment in register contents, if necessary */ | |
1215 | while (d2 == dend2) | |
1216 | { | |
1217 | if (dend2 == end_match_2) break; | |
1218 | if (dend2 == regend[regno]) break; | |
1219 | d2 = string2, dend2 = regend[regno]; /* end of string1 => advance to string2. */ | |
1220 | } | |
1221 | /* At end of register contents => success */ | |
1222 | if (d2 == dend2) break; | |
1223 | ||
1224 | /* Advance to next segment in data being matched, if necessary */ | |
1225 | PREFETCH; | |
1226 | ||
1227 | /* mcnt gets # consecutive chars to compare */ | |
1228 | mcnt = dend - d; | |
1229 | if (mcnt > dend2 - d2) | |
1230 | mcnt = dend2 - d2; | |
1231 | /* Compare that many; failure if mismatch, else skip them. */ | |
1232 | if (translate ? bcmp_translate (d, d2, mcnt, translate) : bcmp (d, d2, mcnt)) | |
1233 | goto fail; | |
1234 | d += mcnt, d2 += mcnt; | |
1235 | } | |
1236 | } | |
1237 | break; | |
1238 | ||
1239 | case anychar: | |
1240 | /* fetch a data character */ | |
1241 | PREFETCH; | |
1242 | /* Match anything but a newline. */ | |
1243 | if ((translate ? translate[*d++] : *d++) == '\n') | |
1244 | goto fail; | |
1245 | break; | |
1246 | ||
1247 | case charset: | |
1248 | case charset_not: | |
1249 | { | |
1250 | /* Nonzero for charset_not */ | |
1251 | int not = 0; | |
1252 | register int c; | |
1253 | if (*(p - 1) == (unsigned char) charset_not) | |
1254 | not = 1; | |
1255 | ||
1256 | /* fetch a data character */ | |
1257 | PREFETCH; | |
1258 | ||
1259 | if (translate) | |
1260 | c = translate [*d]; | |
1261 | else | |
1262 | c = *d; | |
1263 | ||
1264 | if (c < *p * BYTEWIDTH | |
1265 | && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) | |
1266 | not = !not; | |
1267 | ||
1268 | p += 1 + *p; | |
1269 | ||
1270 | if (!not) goto fail; | |
1271 | d++; | |
1272 | break; | |
1273 | } | |
1274 | ||
1275 | case begline: | |
1276 | if (d == string1 || d[-1] == '\n') | |
1277 | break; | |
1278 | goto fail; | |
1279 | ||
1280 | case endline: | |
1281 | if (d == end2 | |
1282 | || (d == end1 ? (size2 == 0 || *string2 == '\n') : *d == '\n')) | |
1283 | break; | |
1284 | goto fail; | |
1285 | ||
1286 | /* "or" constructs ("|") are handled by starting each alternative | |
1287 | with an on_failure_jump that points to the start of the next alternative. | |
1288 | Each alternative except the last ends with a jump to the joining point. | |
1289 | (Actually, each jump except for the last one really jumps | |
1290 | to the following jump, because tensioning the jumps is a hassle.) */ | |
1291 | ||
1292 | /* The start of a stupid repeat has an on_failure_jump that points | |
1293 | past the end of the repeat text. | |
1294 | This makes a failure point so that, on failure to match a repetition, | |
1295 | matching restarts past as many repetitions have been found | |
1296 | with no way to fail and look for another one. */ | |
1297 | ||
1298 | /* A smart repeat is similar but loops back to the on_failure_jump | |
1299 | so that each repetition makes another failure point. */ | |
1300 | ||
1301 | case on_failure_jump: | |
1302 | if (stackp == stacke) | |
1303 | { | |
1304 | unsigned char **stackx; | |
1305 | if (stacke - stackb > re_max_failures * 2) | |
1306 | return -2; | |
1307 | stackx = (unsigned char **) alloca (2 * (stacke - stackb) | |
1308 | * sizeof (char *)); | |
1309 | bcopy (stackb, stackx, (stacke - stackb) * sizeof (char *)); | |
1310 | stackp = stackx + (stackp - stackb); | |
1311 | stacke = stackx + 2 * (stacke - stackb); | |
1312 | stackb = stackx; | |
1313 | } | |
1314 | mcnt = *p++ & 0377; | |
1315 | mcnt += SIGN_EXTEND_CHAR (*(char *)p) << 8; | |
1316 | p++; | |
1317 | *stackp++ = mcnt + p; | |
1318 | *stackp++ = d; | |
1319 | break; | |
1320 | ||
1321 | /* The end of a smart repeat has an maybe_finalize_jump back. | |
1322 | Change it either to a finalize_jump or an ordinary jump. */ | |
1323 | ||
1324 | case maybe_finalize_jump: | |
1325 | mcnt = *p++ & 0377; | |
1326 | mcnt += SIGN_EXTEND_CHAR (*(char *)p) << 8; | |
1327 | p++; | |
1328 | { | |
1329 | register unsigned char *p2 = p; | |
1330 | /* Compare what follows with the begining of the repeat. | |
1331 | If we can establish that there is nothing that they would | |
1332 | both match, we can change to finalize_jump */ | |
1333 | while (p2 != pend | |
1334 | && (*p2 == (unsigned char) stop_memory | |
1335 | || *p2 == (unsigned char) start_memory)) | |
1336 | p2++; | |
1337 | if (p2 == pend) | |
1338 | p[-3] = (unsigned char) finalize_jump; | |
1339 | else if (*p2 == (unsigned char) exactn | |
1340 | || *p2 == (unsigned char) endline) | |
1341 | { | |
1342 | register int c = *p2 == (unsigned char) endline ? '\n' : p2[2]; | |
1343 | register unsigned char *p1 = p + mcnt; | |
1344 | /* p1[0] ... p1[2] are an on_failure_jump. | |
1345 | Examine what follows that */ | |
1346 | if (p1[3] == (unsigned char) exactn && p1[5] != c) | |
1347 | p[-3] = (unsigned char) finalize_jump; | |
1348 | else if (p1[3] == (unsigned char) charset | |
1349 | || p1[3] == (unsigned char) charset_not) | |
1350 | { | |
1351 | int not = p1[3] == (unsigned char) charset_not; | |
1352 | if (c < p1[4] * BYTEWIDTH | |
1353 | && p1[5 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) | |
1354 | not = !not; | |
1355 | /* not is 1 if c would match */ | |
1356 | /* That means it is not safe to finalize */ | |
1357 | if (!not) | |
1358 | p[-3] = (unsigned char) finalize_jump; | |
1359 | } | |
1360 | } | |
1361 | } | |
1362 | p -= 2; | |
1363 | if (p[-1] != (unsigned char) finalize_jump) | |
1364 | { | |
1365 | p[-1] = (unsigned char) jump; | |
1366 | goto nofinalize; | |
1367 | } | |
1368 | ||
1369 | /* The end of a stupid repeat has a finalize-jump | |
1370 | back to the start, where another failure point will be made | |
1371 | which will point after all the repetitions found so far. */ | |
1372 | ||
1373 | case finalize_jump: | |
1374 | stackp -= 2; | |
1375 | ||
1376 | case jump: | |
1377 | nofinalize: | |
1378 | mcnt = *p++ & 0377; | |
1379 | mcnt += SIGN_EXTEND_CHAR (*(char *)p) << 8; | |
1380 | p += mcnt + 1; /* The 1 compensates for missing ++ above */ | |
1381 | break; | |
1382 | ||
1383 | case dummy_failure_jump: | |
1384 | if (stackp == stacke) | |
1385 | { | |
1386 | unsigned char **stackx | |
1387 | = (unsigned char **) alloca (2 * (stacke - stackb) | |
1388 | * sizeof (char *)); | |
1389 | bcopy (stackb, stackx, (stacke - stackb) * sizeof (char *)); | |
1390 | stackp = stackx + (stackp - stackb); | |
1391 | stacke = stackx + 2 * (stacke - stackb); | |
1392 | stackb = stackx; | |
1393 | } | |
1394 | *stackp++ = 0; | |
1395 | *stackp++ = 0; | |
1396 | goto nofinalize; | |
1397 | ||
1398 | case wordbound: | |
1399 | if (d == string1 /* Points to first char */ | |
1400 | || d == end2 /* Points to end */ | |
1401 | || (d == end1 && size2 == 0)) /* Points to end */ | |
1402 | break; | |
1403 | if ((SYNTAX (d[-1]) == Sword) | |
1404 | != (SYNTAX (d == end1 ? *string2 : *d) == Sword)) | |
1405 | break; | |
1406 | goto fail; | |
1407 | ||
1408 | case notwordbound: | |
1409 | if (d == string1 /* Points to first char */ | |
1410 | || d == end2 /* Points to end */ | |
1411 | || (d == end1 && size2 == 0)) /* Points to end */ | |
1412 | goto fail; | |
1413 | if ((SYNTAX (d[-1]) == Sword) | |
1414 | != (SYNTAX (d == end1 ? *string2 : *d) == Sword)) | |
1415 | goto fail; | |
1416 | break; | |
1417 | ||
1418 | case wordbeg: | |
1419 | if (d == end2 /* Points to end */ | |
1420 | || (d == end1 && size2 == 0) /* Points to end */ | |
1421 | || SYNTAX (* (d == end1 ? string2 : d)) != Sword) /* Next char not a letter */ | |
1422 | goto fail; | |
1423 | if (d == string1 /* Points to first char */ | |
1424 | || SYNTAX (d[-1]) != Sword) /* prev char not letter */ | |
1425 | break; | |
1426 | goto fail; | |
1427 | ||
1428 | case wordend: | |
1429 | if (d == string1 /* Points to first char */ | |
1430 | || SYNTAX (d[-1]) != Sword) /* prev char not letter */ | |
1431 | goto fail; | |
1432 | if (d == end2 /* Points to end */ | |
1433 | || (d == end1 && size2 == 0) /* Points to end */ | |
1434 | || SYNTAX (d == end1 ? *string2 : *d) != Sword) /* Next char not a letter */ | |
1435 | break; | |
1436 | goto fail; | |
1437 | ||
1438 | #ifdef emacs | |
1439 | case before_dot: | |
1440 | if (((d - string2 <= (unsigned) size2) | |
1441 | ? d - bf_p2 : d - bf_p1) | |
1442 | <= point) | |
1443 | goto fail; | |
1444 | break; | |
1445 | ||
1446 | case at_dot: | |
1447 | if (((d - string2 <= (unsigned) size2) | |
1448 | ? d - bf_p2 : d - bf_p1) | |
1449 | == point) | |
1450 | goto fail; | |
1451 | break; | |
1452 | ||
1453 | case after_dot: | |
1454 | if (((d - string2 <= (unsigned) size2) | |
1455 | ? d - bf_p2 : d - bf_p1) | |
1456 | >= point) | |
1457 | goto fail; | |
1458 | break; | |
1459 | ||
1460 | case wordchar: | |
1461 | mcnt = (int) Sword; | |
1462 | goto matchsyntax; | |
1463 | ||
1464 | case syntaxspec: | |
1465 | mcnt = *p++; | |
1466 | matchsyntax: | |
1467 | PREFETCH; | |
1468 | if (SYNTAX (*d++) != (enum syntaxcode) mcnt) goto fail; | |
1469 | break; | |
1470 | ||
1471 | case notwordchar: | |
1472 | mcnt = (int) Sword; | |
1473 | goto matchnotsyntax; | |
1474 | ||
1475 | case notsyntaxspec: | |
1476 | mcnt = *p++; | |
1477 | matchnotsyntax: | |
1478 | PREFETCH; | |
1479 | if (SYNTAX (*d++) == (enum syntaxcode) mcnt) goto fail; | |
1480 | break; | |
1481 | #else | |
1482 | case wordchar: | |
1483 | PREFETCH; | |
1484 | if (SYNTAX (*d++) == 0) goto fail; | |
1485 | break; | |
1486 | ||
1487 | case notwordchar: | |
1488 | PREFETCH; | |
1489 | if (SYNTAX (*d++) != 0) goto fail; | |
1490 | break; | |
1491 | #endif /* not emacs */ | |
1492 | ||
1493 | case begbuf: | |
1494 | if (d == string1) /* Note, d cannot equal string2 */ | |
1495 | break; /* unless string1 == string2. */ | |
1496 | goto fail; | |
1497 | ||
1498 | case endbuf: | |
1499 | if (d == end2 || (d == end1 && size2 == 0)) | |
1500 | break; | |
1501 | goto fail; | |
1502 | ||
1503 | case exactn: | |
1504 | /* Match the next few pattern characters exactly. | |
1505 | mcnt is how many characters to match. */ | |
1506 | mcnt = *p++; | |
1507 | if (translate) | |
1508 | { | |
1509 | do | |
1510 | { | |
1511 | PREFETCH; | |
1512 | if (translate[*d++] != *p++) goto fail; | |
1513 | } | |
1514 | while (--mcnt); | |
1515 | } | |
1516 | else | |
1517 | { | |
1518 | do | |
1519 | { | |
1520 | PREFETCH; | |
1521 | if (*d++ != *p++) goto fail; | |
1522 | } | |
1523 | while (--mcnt); | |
1524 | } | |
1525 | break; | |
1526 | } | |
1527 | continue; /* Successfully matched one pattern command; keep matching */ | |
1528 | ||
1529 | /* Jump here if any matching operation fails. */ | |
1530 | fail: | |
1531 | if (stackp != stackb) | |
1532 | /* A restart point is known. Restart there and pop it. */ | |
1533 | { | |
1534 | if (!stackp[-2]) | |
1535 | { /* If innermost failure point is dormant, flush it and keep looking */ | |
1536 | stackp -= 2; | |
1537 | goto fail; | |
1538 | } | |
1539 | d = *--stackp; | |
1540 | p = *--stackp; | |
1541 | if (d >= string1 && d <= end1) | |
1542 | dend = end_match_1; | |
1543 | } | |
1544 | else break; /* Matching at this starting point really fails! */ | |
1545 | } | |
1546 | return -1; /* Failure to match */ | |
1547 | } | |
1548 | ||
1549 | static int | |
1550 | bcmp_translate (s1, s2, len, translate) | |
1551 | unsigned char *s1, *s2; | |
1552 | register int len; | |
1553 | unsigned char *translate; | |
1554 | { | |
1555 | register unsigned char *p1 = s1, *p2 = s2; | |
1556 | while (len) | |
1557 | { | |
1558 | if (translate [*p1++] != translate [*p2++]) return 1; | |
1559 | len--; | |
1560 | } | |
1561 | return 0; | |
1562 | } | |
1563 | \f | |
1564 | /* Entry points compatible with bsd4.2 regex library */ | |
1565 | ||
1566 | #ifndef emacs | |
1567 | ||
1568 | static struct re_pattern_buffer re_comp_buf; | |
1569 | ||
1570 | char * | |
1571 | re_comp (s) | |
1572 | char *s; | |
1573 | { | |
1574 | if (!s) | |
1575 | { | |
1576 | if (!re_comp_buf.buffer) | |
1577 | return "No previous regular expression"; | |
1578 | return 0; | |
1579 | } | |
1580 | ||
1581 | if (!re_comp_buf.buffer) | |
1582 | { | |
1583 | if (!(re_comp_buf.buffer = (char *) malloc (200))) | |
1584 | return "Memory exhausted"; | |
1585 | re_comp_buf.allocated = 200; | |
1586 | if (!(re_comp_buf.fastmap = (char *) malloc (1 << BYTEWIDTH))) | |
1587 | return "Memory exhausted"; | |
1588 | } | |
1589 | return re_compile_pattern (s, strlen (s), &re_comp_buf); | |
1590 | } | |
1591 | ||
1592 | int | |
1593 | re_exec (s) | |
1594 | char *s; | |
1595 | { | |
1596 | int len = strlen (s); | |
1597 | return 0 <= re_search (&re_comp_buf, s, len, 0, len, 0); | |
1598 | } | |
1599 | ||
1600 | #endif /* emacs */ | |
1601 | \f | |
1602 | #ifdef test | |
1603 | ||
1604 | #include <stdio.h> | |
1605 | ||
1606 | /* Indexed by a character, gives the upper case equivalent of the character */ | |
1607 | ||
1608 | static char upcase[0400] = | |
1609 | { 000, 001, 002, 003, 004, 005, 006, 007, | |
1610 | 010, 011, 012, 013, 014, 015, 016, 017, | |
1611 | 020, 021, 022, 023, 024, 025, 026, 027, | |
1612 | 030, 031, 032, 033, 034, 035, 036, 037, | |
1613 | 040, 041, 042, 043, 044, 045, 046, 047, | |
1614 | 050, 051, 052, 053, 054, 055, 056, 057, | |
1615 | 060, 061, 062, 063, 064, 065, 066, 067, | |
1616 | 070, 071, 072, 073, 074, 075, 076, 077, | |
1617 | 0100, 0101, 0102, 0103, 0104, 0105, 0106, 0107, | |
1618 | 0110, 0111, 0112, 0113, 0114, 0115, 0116, 0117, | |
1619 | 0120, 0121, 0122, 0123, 0124, 0125, 0126, 0127, | |
1620 | 0130, 0131, 0132, 0133, 0134, 0135, 0136, 0137, | |
1621 | 0140, 0101, 0102, 0103, 0104, 0105, 0106, 0107, | |
1622 | 0110, 0111, 0112, 0113, 0114, 0115, 0116, 0117, | |
1623 | 0120, 0121, 0122, 0123, 0124, 0125, 0126, 0127, | |
1624 | 0130, 0131, 0132, 0173, 0174, 0175, 0176, 0177, | |
1625 | 0200, 0201, 0202, 0203, 0204, 0205, 0206, 0207, | |
1626 | 0210, 0211, 0212, 0213, 0214, 0215, 0216, 0217, | |
1627 | 0220, 0221, 0222, 0223, 0224, 0225, 0226, 0227, | |
1628 | 0230, 0231, 0232, 0233, 0234, 0235, 0236, 0237, | |
1629 | 0240, 0241, 0242, 0243, 0244, 0245, 0246, 0247, | |
1630 | 0250, 0251, 0252, 0253, 0254, 0255, 0256, 0257, | |
1631 | 0260, 0261, 0262, 0263, 0264, 0265, 0266, 0267, | |
1632 | 0270, 0271, 0272, 0273, 0274, 0275, 0276, 0277, | |
1633 | 0300, 0301, 0302, 0303, 0304, 0305, 0306, 0307, | |
1634 | 0310, 0311, 0312, 0313, 0314, 0315, 0316, 0317, | |
1635 | 0320, 0321, 0322, 0323, 0324, 0325, 0326, 0327, | |
1636 | 0330, 0331, 0332, 0333, 0334, 0335, 0336, 0337, | |
1637 | 0340, 0341, 0342, 0343, 0344, 0345, 0346, 0347, | |
1638 | 0350, 0351, 0352, 0353, 0354, 0355, 0356, 0357, | |
1639 | 0360, 0361, 0362, 0363, 0364, 0365, 0366, 0367, | |
1640 | 0370, 0371, 0372, 0373, 0374, 0375, 0376, 0377 | |
1641 | }; | |
1642 | ||
1643 | main (argc, argv) | |
1644 | int argc; | |
1645 | char **argv; | |
1646 | { | |
1647 | char pat[80]; | |
1648 | struct re_pattern_buffer buf; | |
1649 | int i; | |
1650 | char c; | |
1651 | char fastmap[(1 << BYTEWIDTH)]; | |
1652 | ||
1653 | /* Allow a command argument to specify the style of syntax. */ | |
1654 | if (argc > 1) | |
1655 | obscure_syntax = atoi (argv[1]); | |
1656 | ||
1657 | buf.allocated = 40; | |
1658 | buf.buffer = (char *) malloc (buf.allocated); | |
1659 | buf.fastmap = fastmap; | |
1660 | buf.translate = upcase; | |
1661 | ||
1662 | while (1) | |
1663 | { | |
1664 | gets (pat); | |
1665 | ||
1666 | if (*pat) | |
1667 | { | |
1668 | re_compile_pattern (pat, strlen(pat), &buf); | |
1669 | ||
1670 | for (i = 0; i < buf.used; i++) | |
1671 | printchar (buf.buffer[i]); | |
1672 | ||
1673 | putchar ('\n'); | |
1674 | ||
1675 | printf ("%d allocated, %d used.\n", buf.allocated, buf.used); | |
1676 | ||
1677 | re_compile_fastmap (&buf); | |
1678 | printf ("Allowed by fastmap: "); | |
1679 | for (i = 0; i < (1 << BYTEWIDTH); i++) | |
1680 | if (fastmap[i]) printchar (i); | |
1681 | putchar ('\n'); | |
1682 | } | |
1683 | ||
1684 | gets (pat); /* Now read the string to match against */ | |
1685 | ||
1686 | i = re_match (&buf, pat, strlen (pat), 0, 0); | |
1687 | printf ("Match value %d.\n", i); | |
1688 | } | |
1689 | } | |
1690 | ||
1691 | #ifdef NOTDEF | |
1692 | print_buf (bufp) | |
1693 | struct re_pattern_buffer *bufp; | |
1694 | { | |
1695 | int i; | |
1696 | ||
1697 | printf ("buf is :\n----------------\n"); | |
1698 | for (i = 0; i < bufp->used; i++) | |
1699 | printchar (bufp->buffer[i]); | |
1700 | ||
1701 | printf ("\n%d allocated, %d used.\n", bufp->allocated, bufp->used); | |
1702 | ||
1703 | printf ("Allowed by fastmap: "); | |
1704 | for (i = 0; i < (1 << BYTEWIDTH); i++) | |
1705 | if (bufp->fastmap[i]) | |
1706 | printchar (i); | |
1707 | printf ("\nAllowed by translate: "); | |
1708 | if (bufp->translate) | |
1709 | for (i = 0; i < (1 << BYTEWIDTH); i++) | |
1710 | if (bufp->translate[i]) | |
1711 | printchar (i); | |
1712 | printf ("\nfastmap is%s accurate\n", bufp->fastmap_accurate ? "" : "n't"); | |
1713 | printf ("can %s be null\n----------", bufp->can_be_null ? "" : "not"); | |
1714 | } | |
1715 | #endif | |
1716 | ||
1717 | printchar (c) | |
1718 | char c; | |
1719 | { | |
1720 | if (c < 041 || c >= 0177) | |
1721 | { | |
1722 | putchar ('\\'); | |
1723 | putchar (((c >> 6) & 3) + '0'); | |
1724 | putchar (((c >> 3) & 7) + '0'); | |
1725 | putchar ((c & 7) + '0'); | |
1726 | } | |
1727 | else | |
1728 | putchar (c); | |
1729 | } | |
1730 | ||
1731 | error (string) | |
1732 | char *string; | |
1733 | { | |
1734 | puts (string); | |
1735 | exit (1); | |
1736 | } | |
1737 | ||
1738 | #endif /* test */ |