Commit | Line | Data |
a687059c |
1 | /* NOTE: this is derived from Henry Spencer's regexp code, and should not |
2 | * confused with the original package (see point 3 below). Thanks, Henry! |
3 | */ |
4 | |
5 | /* Additional note: this code is very heavily munged from Henry's version |
6 | * in places. In some spots I've traded clarity for efficiency, so don't |
7 | * blame Henry for some of the lack of readability. |
8 | */ |
9 | |
ac58e20f |
10 | /* $Header: regcomp.c,v 3.0.1.2 90/02/28 18:08:35 lwall Locked $ |
a687059c |
11 | * |
12 | * $Log: regcomp.c,v $ |
ac58e20f |
13 | * Revision 3.0.1.2 90/02/28 18:08:35 lwall |
14 | * patch9: /[\200-\377]/ didn't work on machines with signed chars |
15 | * |
ae986130 |
16 | * Revision 3.0.1.1 89/11/11 04:51:04 lwall |
17 | * patch2: /[\000]/ didn't work |
18 | * |
a687059c |
19 | * Revision 3.0 89/10/18 15:22:29 lwall |
20 | * 3.0 baseline |
21 | * |
22 | */ |
23 | |
24 | /* |
25 | * regcomp and regexec -- regsub and regerror are not used in perl |
26 | * |
27 | * Copyright (c) 1986 by University of Toronto. |
28 | * Written by Henry Spencer. Not derived from licensed software. |
29 | * |
30 | * Permission is granted to anyone to use this software for any |
31 | * purpose on any computer system, and to redistribute it freely, |
32 | * subject to the following restrictions: |
33 | * |
34 | * 1. The author is not responsible for the consequences of use of |
35 | * this software, no matter how awful, even if they arise |
36 | * from defects in it. |
37 | * |
38 | * 2. The origin of this software must not be misrepresented, either |
39 | * by explicit claim or by omission. |
40 | * |
41 | * 3. Altered versions must be plainly marked as such, and must not |
42 | * be misrepresented as being the original software. |
43 | * |
44 | * |
45 | **** Alterations to Henry's code are... |
46 | **** |
47 | **** Copyright (c) 1989, Larry Wall |
48 | **** |
49 | **** You may distribute under the terms of the GNU General Public License |
50 | **** as specified in the README file that comes with the perl 3.0 kit. |
51 | * |
52 | * Beware that some of this code is subtly aware of the way operator |
53 | * precedence is structured in regular expressions. Serious changes in |
54 | * regular-expression syntax might require a total rethink. |
55 | */ |
56 | #include "EXTERN.h" |
57 | #include "perl.h" |
58 | #include "INTERN.h" |
59 | #include "regcomp.h" |
60 | |
61 | #ifndef STATIC |
62 | #define STATIC static |
63 | #endif |
64 | |
65 | #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?') |
66 | #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \ |
67 | ((*s) == '{' && regcurly(s))) |
68 | #define META "^$.[()|?+*\\" |
69 | |
70 | /* |
71 | * Flags to be passed up and down. |
72 | */ |
73 | #define HASWIDTH 01 /* Known never to match null string. */ |
74 | #define SIMPLE 02 /* Simple enough to be STAR/PLUS operand. */ |
75 | #define SPSTART 04 /* Starts with * or +. */ |
76 | #define WORST 0 /* Worst case. */ |
77 | |
78 | /* |
79 | * Global work variables for regcomp(). |
80 | */ |
81 | static char *regprecomp; /* uncompiled string. */ |
82 | static char *regparse; /* Input-scan pointer. */ |
83 | static char *regxend; /* End of input for compile */ |
84 | static int regnpar; /* () count. */ |
85 | static char *regcode; /* Code-emit pointer; ®dummy = don't. */ |
86 | static long regsize; /* Code size. */ |
87 | static int regfold; |
88 | static int regsawbracket; /* Did we do {d,d} trick? */ |
89 | |
90 | /* |
91 | * Forward declarations for regcomp()'s friends. |
92 | */ |
93 | STATIC int regcurly(); |
94 | STATIC char *reg(); |
95 | STATIC char *regbranch(); |
96 | STATIC char *regpiece(); |
97 | STATIC char *regatom(); |
98 | STATIC char *regclass(); |
99 | STATIC char *regnode(); |
100 | STATIC void regc(); |
101 | STATIC void reginsert(); |
102 | STATIC void regtail(); |
103 | STATIC void regoptail(); |
104 | |
105 | /* |
106 | - regcomp - compile a regular expression into internal code |
107 | * |
108 | * We can't allocate space until we know how big the compiled form will be, |
109 | * but we can't compile it (and thus know how big it is) until we've got a |
110 | * place to put the code. So we cheat: we compile it twice, once with code |
111 | * generation turned off and size counting turned on, and once "for real". |
112 | * This also means that we don't allocate space until we are sure that the |
113 | * thing really will compile successfully, and we never have to move the |
114 | * code and thus invalidate pointers into it. (Note that it has to be in |
115 | * one piece because free() must be able to free it all.) [NB: not true in perl] |
116 | * |
117 | * Beware that the optimization-preparation code in here knows about some |
118 | * of the structure of the compiled regexp. [I'll say.] |
119 | */ |
120 | regexp * |
121 | regcomp(exp,xend,fold,rare) |
122 | char *exp; |
123 | char *xend; |
124 | int fold; |
125 | int rare; |
126 | { |
127 | register regexp *r; |
128 | register char *scan; |
129 | register STR *longest; |
130 | register int len; |
131 | register char *first; |
132 | int flags; |
133 | int back; |
134 | int curback; |
135 | extern char *safemalloc(); |
136 | extern char *savestr(); |
137 | |
138 | if (exp == NULL) |
139 | fatal("NULL regexp argument"); |
140 | |
141 | /* First pass: determine size, legality. */ |
142 | regfold = fold; |
143 | regparse = exp; |
144 | regxend = xend; |
145 | regprecomp = nsavestr(exp,xend-exp); |
146 | regsawbracket = 0; |
147 | regnpar = 1; |
148 | regsize = 0L; |
149 | regcode = ®dummy; |
150 | regc(MAGIC); |
151 | if (reg(0, &flags) == NULL) { |
152 | Safefree(regprecomp); |
153 | return(NULL); |
154 | } |
155 | |
156 | /* Small enough for pointer-storage convention? */ |
157 | if (regsize >= 32767L) /* Probably could be 65535L. */ |
158 | FAIL("regexp too big"); |
159 | |
160 | /* Allocate space. */ |
161 | Newc(1001, r, sizeof(regexp) + (unsigned)regsize, char, regexp); |
162 | if (r == NULL) |
163 | FAIL("regexp out of space"); |
164 | |
165 | /* Second pass: emit code. */ |
166 | if (regsawbracket) |
167 | bcopy(regprecomp,exp,xend-exp); |
168 | r->precomp = regprecomp; |
169 | r->subbase = NULL; |
170 | regparse = exp; |
171 | regnpar = 1; |
172 | regcode = r->program; |
173 | regc(MAGIC); |
174 | if (reg(0, &flags) == NULL) |
175 | return(NULL); |
176 | |
177 | /* Dig out information for optimizations. */ |
178 | r->regstart = Nullstr; /* Worst-case defaults. */ |
179 | r->reganch = 0; |
180 | r->regmust = Nullstr; |
181 | r->regback = -1; |
182 | r->regstclass = Nullch; |
183 | scan = r->program+1; /* First BRANCH. */ |
184 | if (OP(regnext(scan)) == END) {/* Only one top-level choice. */ |
185 | scan = NEXTOPER(scan); |
186 | |
187 | first = scan; |
188 | while ((OP(first) > OPEN && OP(first) < CLOSE) || |
189 | (OP(first) == BRANCH && OP(regnext(first)) != BRANCH) || |
190 | (OP(first) == PLUS) ) |
191 | first = NEXTOPER(first); |
192 | |
193 | /* Starting-point info. */ |
194 | if (OP(first) == EXACTLY) { |
195 | r->regstart = |
196 | str_make(OPERAND(first)+1,*OPERAND(first)); |
197 | if (r->regstart->str_cur > !(sawstudy|fold)) |
198 | fbmcompile(r->regstart,fold); |
199 | } |
200 | else if ((exp = index(simple,OP(first))) && exp > simple) |
201 | r->regstclass = first; |
202 | else if (OP(first) == BOUND || OP(first) == NBOUND) |
203 | r->regstclass = first; |
204 | else if (OP(first) == BOL) |
205 | r->reganch++; |
206 | |
207 | #ifdef DEBUGGING |
208 | if (debug & 512) |
209 | fprintf(stderr,"first %d next %d offset %d\n", |
210 | OP(first), OP(NEXTOPER(first)), first - scan); |
211 | #endif |
212 | /* |
213 | * If there's something expensive in the r.e., find the |
214 | * longest literal string that must appear and make it the |
215 | * regmust. Resolve ties in favor of later strings, since |
216 | * the regstart check works with the beginning of the r.e. |
217 | * and avoiding duplication strengthens checking. Not a |
218 | * strong reason, but sufficient in the absence of others. |
219 | * [Now we resolve ties in favor of the earlier string if |
220 | * it happens that curback has been invalidated, since the |
221 | * earlier string may buy us something the later one won't.] |
222 | */ |
223 | longest = str_make("",0); |
224 | len = 0; |
225 | curback = 0; |
226 | back = 0; |
227 | while (scan != NULL) { |
228 | if (OP(scan) == BRANCH) { |
229 | if (OP(regnext(scan)) == BRANCH) { |
230 | curback = -30000; |
231 | while (OP(scan) == BRANCH) |
232 | scan = regnext(scan); |
233 | } |
234 | else /* single branch is ok */ |
235 | scan = NEXTOPER(scan); |
236 | } |
237 | if (OP(scan) == EXACTLY) { |
238 | first = scan; |
239 | while (OP(regnext(scan)) >= CLOSE) |
240 | scan = regnext(scan); |
241 | if (curback - back == len) { |
242 | str_ncat(longest, OPERAND(first)+1, |
243 | *OPERAND(first)); |
244 | len += *OPERAND(first); |
245 | curback += *OPERAND(first); |
246 | first = regnext(scan); |
247 | } |
248 | else if (*OPERAND(first) >= len + (curback >= 0)) { |
249 | len = *OPERAND(first); |
250 | str_nset(longest, OPERAND(first)+1,len); |
251 | back = curback; |
252 | curback += len; |
253 | first = regnext(scan); |
254 | } |
255 | else |
256 | curback += *OPERAND(first); |
257 | } |
258 | else if (index(varies,OP(scan))) |
259 | curback = -30000; |
260 | else if (index(simple,OP(scan))) |
261 | curback++; |
262 | scan = regnext(scan); |
263 | } |
264 | if (len) { |
265 | r->regmust = longest; |
266 | if (back < 0) |
267 | back = -1; |
268 | r->regback = back; |
269 | if (len > !(sawstudy||fold||OP(first)==EOL)) |
270 | fbmcompile(r->regmust,fold); |
271 | r->regmust->str_u.str_useful = 100; |
272 | if (OP(first) == EOL) /* is match anchored to EOL? */ |
273 | r->regmust->str_pok |= SP_TAIL; |
274 | } |
275 | else |
276 | str_free(longest); |
277 | } |
278 | |
279 | r->do_folding = fold; |
280 | r->nparens = regnpar - 1; |
281 | #ifdef DEBUGGING |
282 | if (debug & 512) |
283 | regdump(r); |
284 | #endif |
285 | return(r); |
286 | } |
287 | |
288 | /* |
289 | - reg - regular expression, i.e. main body or parenthesized thing |
290 | * |
291 | * Caller must absorb opening parenthesis. |
292 | * |
293 | * Combining parenthesis handling with the base level of regular expression |
294 | * is a trifle forced, but the need to tie the tails of the branches to what |
295 | * follows makes it hard to avoid. |
296 | */ |
297 | static char * |
298 | reg(paren, flagp) |
299 | int paren; /* Parenthesized? */ |
300 | int *flagp; |
301 | { |
302 | register char *ret; |
303 | register char *br; |
304 | register char *ender; |
305 | register int parno; |
306 | int flags; |
307 | |
308 | *flagp = HASWIDTH; /* Tentatively. */ |
309 | |
310 | /* Make an OPEN node, if parenthesized. */ |
311 | if (paren) { |
312 | if (regnpar >= NSUBEXP) |
313 | FAIL("too many () in regexp"); |
314 | parno = regnpar; |
315 | regnpar++; |
316 | ret = regnode(OPEN+parno); |
317 | } else |
318 | ret = NULL; |
319 | |
320 | /* Pick up the branches, linking them together. */ |
321 | br = regbranch(&flags); |
322 | if (br == NULL) |
323 | return(NULL); |
324 | if (ret != NULL) |
325 | regtail(ret, br); /* OPEN -> first. */ |
326 | else |
327 | ret = br; |
328 | if (!(flags&HASWIDTH)) |
329 | *flagp &= ~HASWIDTH; |
330 | *flagp |= flags&SPSTART; |
331 | while (*regparse == '|') { |
332 | regparse++; |
333 | br = regbranch(&flags); |
334 | if (br == NULL) |
335 | return(NULL); |
336 | regtail(ret, br); /* BRANCH -> BRANCH. */ |
337 | if (!(flags&HASWIDTH)) |
338 | *flagp &= ~HASWIDTH; |
339 | *flagp |= flags&SPSTART; |
340 | } |
341 | |
342 | /* Make a closing node, and hook it on the end. */ |
343 | ender = regnode((paren) ? CLOSE+parno : END); |
344 | regtail(ret, ender); |
345 | |
346 | /* Hook the tails of the branches to the closing node. */ |
347 | for (br = ret; br != NULL; br = regnext(br)) |
348 | regoptail(br, ender); |
349 | |
350 | /* Check for proper termination. */ |
351 | if (paren && *regparse++ != ')') { |
352 | FAIL("unmatched () in regexp"); |
353 | } else if (!paren && regparse < regxend) { |
354 | if (*regparse == ')') { |
355 | FAIL("unmatched () in regexp"); |
356 | } else |
357 | FAIL("junk on end of regexp"); /* "Can't happen". */ |
358 | /* NOTREACHED */ |
359 | } |
360 | |
361 | return(ret); |
362 | } |
363 | |
364 | /* |
365 | - regbranch - one alternative of an | operator |
366 | * |
367 | * Implements the concatenation operator. |
368 | */ |
369 | static char * |
370 | regbranch(flagp) |
371 | int *flagp; |
372 | { |
373 | register char *ret; |
374 | register char *chain; |
375 | register char *latest; |
376 | int flags; |
377 | |
378 | *flagp = WORST; /* Tentatively. */ |
379 | |
380 | ret = regnode(BRANCH); |
381 | chain = NULL; |
382 | while (regparse < regxend && *regparse != '|' && *regparse != ')') { |
383 | latest = regpiece(&flags); |
384 | if (latest == NULL) |
385 | return(NULL); |
386 | *flagp |= flags&HASWIDTH; |
387 | if (chain == NULL) /* First piece. */ |
388 | *flagp |= flags&SPSTART; |
389 | else |
390 | regtail(chain, latest); |
391 | chain = latest; |
392 | } |
393 | if (chain == NULL) /* Loop ran zero times. */ |
394 | (void) regnode(NOTHING); |
395 | |
396 | return(ret); |
397 | } |
398 | |
399 | /* |
400 | - regpiece - something followed by possible [*+?] |
401 | * |
402 | * Note that the branching code sequences used for ? and the general cases |
403 | * of * and + are somewhat optimized: they use the same NOTHING node as |
404 | * both the endmarker for their branch list and the body of the last branch. |
405 | * It might seem that this node could be dispensed with entirely, but the |
406 | * endmarker role is not redundant. |
407 | */ |
408 | static char * |
409 | regpiece(flagp) |
410 | int *flagp; |
411 | { |
412 | register char *ret; |
413 | register char op; |
414 | register char *next; |
415 | int flags; |
416 | char *origparse = regparse; |
417 | int orignpar = regnpar; |
418 | char *max; |
419 | int iter; |
420 | char ch; |
421 | |
422 | ret = regatom(&flags); |
423 | if (ret == NULL) |
424 | return(NULL); |
425 | |
426 | op = *regparse; |
427 | |
428 | /* Here's a total kludge: if after the atom there's a {\d+,?\d*} |
429 | * then we decrement the first number by one and reset our |
430 | * parsing back to the beginning of the same atom. If the first number |
431 | * is down to 0, decrement the second number instead and fake up |
432 | * a ? after it. Given the way this compiler doesn't keep track |
433 | * of offsets on the first pass, this is the only way to replicate |
434 | * a piece of code. Sigh. |
435 | */ |
436 | if (op == '{' && regcurly(regparse)) { |
437 | next = regparse + 1; |
438 | max = Nullch; |
439 | while (isdigit(*next) || *next == ',') { |
440 | if (*next == ',') { |
441 | if (max) |
442 | break; |
443 | else |
444 | max = next; |
445 | } |
446 | next++; |
447 | } |
448 | if (*next == '}') { /* got one */ |
449 | regsawbracket++; /* remember we clobbered exp */ |
450 | if (!max) |
451 | max = next; |
452 | regparse++; |
453 | iter = atoi(regparse); |
454 | if (iter > 0) { |
455 | ch = *max; |
456 | sprintf(regparse,"%.*d", max-regparse, iter - 1); |
457 | *max = ch; |
458 | if (*max == ',' && atoi(max+1) > 0) { |
459 | ch = *next; |
460 | sprintf(max+1,"%.*d", next-(max+1), atoi(max+1) - 1); |
461 | *next = ch; |
462 | } |
463 | if (iter != 1 || (*max == ',' || atoi(max+1))) { |
464 | regparse = origparse; /* back up input pointer */ |
465 | regnpar = orignpar; /* don't make more parens */ |
466 | } |
467 | else { |
468 | regparse = next; |
469 | goto nest_check; |
470 | } |
471 | *flagp = flags; |
472 | return ret; |
473 | } |
474 | if (*max == ',') { |
475 | max++; |
476 | iter = atoi(max); |
477 | if (max == next) { /* any number more? */ |
478 | regparse = next; |
479 | op = '*'; /* fake up one with a star */ |
480 | } |
481 | else if (iter > 0) { |
482 | op = '?'; /* fake up optional atom */ |
483 | ch = *next; |
484 | sprintf(max,"%.*d", next-max, iter - 1); |
485 | *next = ch; |
486 | if (iter == 1) |
487 | regparse = next; |
488 | else { |
489 | regparse = origparse - 1; /* offset ++ below */ |
490 | regnpar = orignpar; |
491 | } |
492 | } |
493 | else |
494 | fatal("Can't do {n,0}"); |
495 | } |
496 | else |
497 | fatal("Can't do {0}"); |
498 | } |
499 | } |
500 | |
501 | if (!ISMULT1(op)) { |
502 | *flagp = flags; |
503 | return(ret); |
504 | } |
505 | |
506 | if (!(flags&HASWIDTH) && op != '?') |
507 | FAIL("regexp *+ operand could be empty"); |
508 | *flagp = (op != '+') ? (WORST|SPSTART) : (WORST|HASWIDTH); |
509 | |
510 | if (op == '*' && (flags&SIMPLE)) |
511 | reginsert(STAR, ret); |
512 | else if (op == '*') { |
513 | /* Emit x* as (x&|), where & means "self". */ |
514 | reginsert(BRANCH, ret); /* Either x */ |
515 | regoptail(ret, regnode(BACK)); /* and loop */ |
516 | regoptail(ret, ret); /* back */ |
517 | regtail(ret, regnode(BRANCH)); /* or */ |
518 | regtail(ret, regnode(NOTHING)); /* null. */ |
519 | } else if (op == '+' && (flags&SIMPLE)) |
520 | reginsert(PLUS, ret); |
521 | else if (op == '+') { |
522 | /* Emit x+ as x(&|), where & means "self". */ |
523 | next = regnode(BRANCH); /* Either */ |
524 | regtail(ret, next); |
525 | regtail(regnode(BACK), ret); /* loop back */ |
526 | regtail(next, regnode(BRANCH)); /* or */ |
527 | regtail(ret, regnode(NOTHING)); /* null. */ |
528 | } else if (op == '?') { |
529 | /* Emit x? as (x|) */ |
530 | reginsert(BRANCH, ret); /* Either x */ |
531 | regtail(ret, regnode(BRANCH)); /* or */ |
532 | next = regnode(NOTHING); /* null. */ |
533 | regtail(ret, next); |
534 | regoptail(ret, next); |
535 | } |
536 | nest_check: |
537 | regparse++; |
538 | if (ISMULT2(regparse)) |
539 | FAIL("nested *?+ in regexp"); |
540 | |
541 | return(ret); |
542 | } |
543 | |
544 | /* |
545 | - regatom - the lowest level |
546 | * |
547 | * Optimization: gobbles an entire sequence of ordinary characters so that |
548 | * it can turn them into a single node, which is smaller to store and |
549 | * faster to run. Backslashed characters are exceptions, each becoming a |
550 | * separate node; the code is simpler that way and it's not worth fixing. |
551 | * |
552 | * [Yes, it is worth fixing, some scripts can run twice the speed.] |
553 | */ |
554 | static char * |
555 | regatom(flagp) |
556 | int *flagp; |
557 | { |
558 | register char *ret; |
559 | int flags; |
560 | |
561 | *flagp = WORST; /* Tentatively. */ |
562 | |
563 | switch (*regparse++) { |
564 | case '^': |
565 | ret = regnode(BOL); |
566 | break; |
567 | case '$': |
568 | ret = regnode(EOL); |
569 | break; |
570 | case '.': |
571 | ret = regnode(ANY); |
572 | *flagp |= HASWIDTH|SIMPLE; |
573 | break; |
574 | case '[': |
575 | ret = regclass(); |
576 | *flagp |= HASWIDTH|SIMPLE; |
577 | break; |
578 | case '(': |
579 | ret = reg(1, &flags); |
580 | if (ret == NULL) |
581 | return(NULL); |
582 | *flagp |= flags&(HASWIDTH|SPSTART); |
583 | break; |
584 | case '|': |
585 | case ')': |
586 | FAIL("internal urp in regexp"); /* Supposed to be caught earlier. */ |
587 | break; |
588 | case '?': |
589 | case '+': |
590 | case '*': |
591 | FAIL("?+* follows nothing in regexp"); |
592 | break; |
593 | case '\\': |
594 | switch (*regparse) { |
595 | case 'w': |
596 | ret = regnode(ALNUM); |
597 | *flagp |= HASWIDTH|SIMPLE; |
598 | regparse++; |
599 | break; |
600 | case 'W': |
601 | ret = regnode(NALNUM); |
602 | *flagp |= HASWIDTH|SIMPLE; |
603 | regparse++; |
604 | break; |
605 | case 'b': |
606 | ret = regnode(BOUND); |
607 | *flagp |= SIMPLE; |
608 | regparse++; |
609 | break; |
610 | case 'B': |
611 | ret = regnode(NBOUND); |
612 | *flagp |= SIMPLE; |
613 | regparse++; |
614 | break; |
615 | case 's': |
616 | ret = regnode(SPACE); |
617 | *flagp |= HASWIDTH|SIMPLE; |
618 | regparse++; |
619 | break; |
620 | case 'S': |
621 | ret = regnode(NSPACE); |
622 | *flagp |= HASWIDTH|SIMPLE; |
623 | regparse++; |
624 | break; |
625 | case 'd': |
626 | ret = regnode(DIGIT); |
627 | *flagp |= HASWIDTH|SIMPLE; |
628 | regparse++; |
629 | break; |
630 | case 'D': |
631 | ret = regnode(NDIGIT); |
632 | *flagp |= HASWIDTH|SIMPLE; |
633 | regparse++; |
634 | break; |
635 | case 'n': |
636 | case 'r': |
637 | case 't': |
638 | case 'f': |
639 | goto defchar; |
640 | case '0': case '1': case '2': case '3': case '4': |
641 | case '5': case '6': case '7': case '8': case '9': |
642 | if (isdigit(regparse[1])) |
643 | goto defchar; |
644 | else { |
645 | ret = regnode(REF + *regparse++ - '0'); |
646 | *flagp |= SIMPLE; |
647 | } |
648 | break; |
649 | case '\0': |
650 | if (regparse >= regxend) |
651 | FAIL("trailing \\ in regexp"); |
652 | /* FALL THROUGH */ |
653 | default: |
654 | goto defchar; |
655 | } |
656 | break; |
657 | default: { |
658 | register int len; |
659 | register char ender; |
660 | register char *p; |
661 | char *oldp; |
662 | int foo; |
663 | |
664 | defchar: |
665 | ret = regnode(EXACTLY); |
666 | regc(0); /* save spot for len */ |
667 | for (len=0, p=regparse-1; |
668 | len < 127 && p < regxend; |
669 | len++) |
670 | { |
671 | oldp = p; |
672 | switch (*p) { |
673 | case '^': |
674 | case '$': |
675 | case '.': |
676 | case '[': |
677 | case '(': |
678 | case ')': |
679 | case '|': |
680 | goto loopdone; |
681 | case '\\': |
682 | switch (*++p) { |
683 | case 'w': |
684 | case 'W': |
685 | case 'b': |
686 | case 'B': |
687 | case 's': |
688 | case 'S': |
689 | case 'd': |
690 | case 'D': |
691 | --p; |
692 | goto loopdone; |
693 | case 'n': |
694 | ender = '\n'; |
695 | p++; |
696 | break; |
697 | case 'r': |
698 | ender = '\r'; |
699 | p++; |
700 | break; |
701 | case 't': |
702 | ender = '\t'; |
703 | p++; |
704 | break; |
705 | case 'f': |
706 | ender = '\f'; |
707 | p++; |
708 | break; |
709 | case '0': case '1': case '2': case '3':case '4': |
710 | case '5': case '6': case '7': case '8':case '9': |
711 | if (isdigit(p[1])) { |
712 | foo = *p++ - '0'; |
713 | foo <<= 3; |
714 | foo += *p - '0'; |
715 | if (isdigit(p[1])) |
716 | foo = (foo<<3) + *++p - '0'; |
717 | ender = foo; |
718 | p++; |
719 | } |
720 | else { |
721 | --p; |
722 | goto loopdone; |
723 | } |
724 | break; |
725 | case '\0': |
726 | if (p >= regxend) |
727 | FAIL("trailing \\ in regexp"); |
728 | /* FALL THROUGH */ |
729 | default: |
730 | ender = *p++; |
731 | break; |
732 | } |
733 | break; |
734 | default: |
735 | ender = *p++; |
736 | break; |
737 | } |
738 | if (regfold && isupper(ender)) |
739 | ender = tolower(ender); |
740 | if (ISMULT2(p)) { /* Back off on ?+*. */ |
741 | if (len) |
742 | p = oldp; |
743 | else { |
744 | len++; |
745 | regc(ender); |
746 | } |
747 | break; |
748 | } |
749 | regc(ender); |
750 | } |
751 | loopdone: |
752 | regparse = p; |
753 | if (len <= 0) |
754 | FAIL("internal disaster in regexp"); |
755 | *flagp |= HASWIDTH; |
756 | if (len == 1) |
757 | *flagp |= SIMPLE; |
758 | if (regcode != ®dummy) |
759 | *OPERAND(ret) = len; |
760 | regc('\0'); |
761 | } |
762 | break; |
763 | } |
764 | |
765 | return(ret); |
766 | } |
767 | |
768 | static void |
769 | regset(bits,def,c) |
770 | char *bits; |
771 | int def; |
772 | register int c; |
773 | { |
774 | if (regcode == ®dummy) |
775 | return; |
ac58e20f |
776 | c &= 255; |
a687059c |
777 | if (def) |
778 | bits[c >> 3] &= ~(1 << (c & 7)); |
779 | else |
780 | bits[c >> 3] |= (1 << (c & 7)); |
781 | } |
782 | |
783 | static char * |
784 | regclass() |
785 | { |
786 | register char *bits; |
787 | register int class; |
788 | register int lastclass; |
789 | register int range = 0; |
790 | register char *ret; |
791 | register int def; |
792 | |
793 | if (*regparse == '^') { /* Complement of range. */ |
794 | ret = regnode(ANYBUT); |
795 | regparse++; |
796 | def = 0; |
797 | } else { |
798 | ret = regnode(ANYOF); |
799 | def = 255; |
800 | } |
801 | bits = regcode; |
802 | for (class = 0; class < 32; class++) |
803 | regc(def); |
804 | if (*regparse == ']' || *regparse == '-') |
805 | regset(bits,def,lastclass = *regparse++); |
806 | while (regparse < regxend && *regparse != ']') { |
807 | class = UCHARAT(regparse++); |
808 | if (class == '\\') { |
809 | class = UCHARAT(regparse++); |
810 | switch (class) { |
811 | case 'w': |
812 | for (class = 'a'; class <= 'z'; class++) |
813 | regset(bits,def,class); |
814 | for (class = 'A'; class <= 'Z'; class++) |
815 | regset(bits,def,class); |
816 | for (class = '0'; class <= '9'; class++) |
817 | regset(bits,def,class); |
818 | regset(bits,def,'_'); |
819 | lastclass = 1234; |
820 | continue; |
821 | case 's': |
822 | regset(bits,def,' '); |
823 | regset(bits,def,'\t'); |
824 | regset(bits,def,'\r'); |
825 | regset(bits,def,'\f'); |
826 | regset(bits,def,'\n'); |
827 | lastclass = 1234; |
828 | continue; |
829 | case 'd': |
830 | for (class = '0'; class <= '9'; class++) |
831 | regset(bits,def,class); |
832 | lastclass = 1234; |
833 | continue; |
834 | case 'n': |
835 | class = '\n'; |
836 | break; |
837 | case 'r': |
838 | class = '\r'; |
839 | break; |
840 | case 't': |
841 | class = '\t'; |
842 | break; |
843 | case 'f': |
844 | class = '\f'; |
845 | break; |
846 | case 'b': |
847 | class = '\b'; |
848 | break; |
849 | case '0': case '1': case '2': case '3': case '4': |
850 | case '5': case '6': case '7': case '8': case '9': |
851 | class -= '0'; |
852 | if (isdigit(*regparse)) { |
853 | class <<= 3; |
854 | class += *regparse++ - '0'; |
855 | } |
856 | if (isdigit(*regparse)) { |
857 | class <<= 3; |
858 | class += *regparse++ - '0'; |
859 | } |
860 | break; |
861 | } |
862 | } |
863 | if (!range && class == '-' && regparse < regxend && |
864 | *regparse != ']') { |
865 | range = 1; |
866 | continue; |
867 | } |
868 | if (range) { |
869 | if (lastclass > class) |
870 | FAIL("invalid [] range in regexp"); |
871 | } |
872 | else |
873 | lastclass = class - 1; |
874 | range = 0; |
875 | for (lastclass++; lastclass <= class; lastclass++) { |
876 | regset(bits,def,lastclass); |
877 | if (regfold && isupper(lastclass)) |
878 | regset(bits,def,tolower(lastclass)); |
879 | } |
880 | lastclass = class; |
881 | } |
882 | if (*regparse != ']') |
883 | FAIL("unmatched [] in regexp"); |
a687059c |
884 | regparse++; |
885 | return ret; |
886 | } |
887 | |
888 | /* |
889 | - regnode - emit a node |
890 | */ |
891 | static char * /* Location. */ |
892 | regnode(op) |
893 | char op; |
894 | { |
895 | register char *ret; |
896 | register char *ptr; |
897 | |
898 | ret = regcode; |
899 | if (ret == ®dummy) { |
900 | #ifdef REGALIGN |
901 | if (!(regsize & 1)) |
902 | regsize++; |
903 | #endif |
904 | regsize += 3; |
905 | return(ret); |
906 | } |
907 | |
908 | #ifdef REGALIGN |
909 | #ifndef lint |
910 | if (!((long)ret & 1)) |
911 | *ret++ = 127; |
912 | #endif |
913 | #endif |
914 | ptr = ret; |
915 | *ptr++ = op; |
916 | *ptr++ = '\0'; /* Null "next" pointer. */ |
917 | *ptr++ = '\0'; |
918 | regcode = ptr; |
919 | |
920 | return(ret); |
921 | } |
922 | |
923 | /* |
924 | - regc - emit (if appropriate) a byte of code |
925 | */ |
926 | static void |
927 | regc(b) |
928 | char b; |
929 | { |
930 | if (regcode != ®dummy) |
931 | *regcode++ = b; |
932 | else |
933 | regsize++; |
934 | } |
935 | |
936 | /* |
937 | - reginsert - insert an operator in front of already-emitted operand |
938 | * |
939 | * Means relocating the operand. |
940 | */ |
941 | static void |
942 | reginsert(op, opnd) |
943 | char op; |
944 | char *opnd; |
945 | { |
946 | register char *src; |
947 | register char *dst; |
948 | register char *place; |
949 | |
950 | if (regcode == ®dummy) { |
951 | #ifdef REGALIGN |
952 | regsize += 4; |
953 | #else |
954 | regsize += 3; |
955 | #endif |
956 | return; |
957 | } |
958 | |
959 | src = regcode; |
960 | #ifdef REGALIGN |
961 | regcode += 4; |
962 | #else |
963 | regcode += 3; |
964 | #endif |
965 | dst = regcode; |
966 | while (src > opnd) |
967 | *--dst = *--src; |
968 | |
969 | place = opnd; /* Op node, where operand used to be. */ |
970 | *place++ = op; |
971 | *place++ = '\0'; |
972 | *place++ = '\0'; |
973 | } |
974 | |
975 | /* |
976 | - regtail - set the next-pointer at the end of a node chain |
977 | */ |
978 | static void |
979 | regtail(p, val) |
980 | char *p; |
981 | char *val; |
982 | { |
983 | register char *scan; |
984 | register char *temp; |
985 | register int offset; |
986 | |
987 | if (p == ®dummy) |
988 | return; |
989 | |
990 | /* Find last node. */ |
991 | scan = p; |
992 | for (;;) { |
993 | temp = regnext(scan); |
994 | if (temp == NULL) |
995 | break; |
996 | scan = temp; |
997 | } |
998 | |
999 | #ifdef REGALIGN |
1000 | offset = val - scan; |
1001 | #ifndef lint |
1002 | *(short*)(scan+1) = offset; |
1003 | #else |
1004 | offset = offset; |
1005 | #endif |
1006 | #else |
1007 | if (OP(scan) == BACK) |
1008 | offset = scan - val; |
1009 | else |
1010 | offset = val - scan; |
1011 | *(scan+1) = (offset>>8)&0377; |
1012 | *(scan+2) = offset&0377; |
1013 | #endif |
1014 | } |
1015 | |
1016 | /* |
1017 | - regoptail - regtail on operand of first argument; nop if operandless |
1018 | */ |
1019 | static void |
1020 | regoptail(p, val) |
1021 | char *p; |
1022 | char *val; |
1023 | { |
1024 | /* "Operandless" and "op != BRANCH" are synonymous in practice. */ |
1025 | if (p == NULL || p == ®dummy || OP(p) != BRANCH) |
1026 | return; |
1027 | regtail(NEXTOPER(p), val); |
1028 | } |
1029 | |
1030 | /* |
1031 | - regcurly - a little FSA that accepts {\d+,?\d*} |
1032 | */ |
1033 | STATIC int |
1034 | regcurly(s) |
1035 | register char *s; |
1036 | { |
1037 | if (*s++ != '{') |
1038 | return FALSE; |
1039 | if (!isdigit(*s)) |
1040 | return FALSE; |
1041 | while (isdigit(*s)) |
1042 | s++; |
1043 | if (*s == ',') |
1044 | s++; |
1045 | while (isdigit(*s)) |
1046 | s++; |
1047 | if (*s != '}') |
1048 | return FALSE; |
1049 | return TRUE; |
1050 | } |
1051 | |
1052 | #ifdef DEBUGGING |
1053 | |
1054 | /* |
1055 | - regdump - dump a regexp onto stderr in vaguely comprehensible form |
1056 | */ |
1057 | void |
1058 | regdump(r) |
1059 | regexp *r; |
1060 | { |
1061 | register char *s; |
1062 | register char op = EXACTLY; /* Arbitrary non-END op. */ |
1063 | register char *next; |
1064 | extern char *index(); |
1065 | |
1066 | |
1067 | s = r->program + 1; |
1068 | while (op != END) { /* While that wasn't END last time... */ |
1069 | #ifdef REGALIGN |
1070 | if (!((long)s & 1)) |
1071 | s++; |
1072 | #endif |
1073 | op = OP(s); |
1074 | fprintf(stderr,"%2d%s", s-r->program, regprop(s)); /* Where, what. */ |
1075 | next = regnext(s); |
1076 | if (next == NULL) /* Next ptr. */ |
1077 | fprintf(stderr,"(0)"); |
1078 | else |
1079 | fprintf(stderr,"(%d)", (s-r->program)+(next-s)); |
1080 | s += 3; |
1081 | if (op == ANYOF || op == ANYBUT) { |
1082 | s += 32; |
1083 | } |
1084 | if (op == EXACTLY) { |
1085 | /* Literal string, where present. */ |
1086 | s++; |
1087 | while (*s != '\0') { |
1088 | (void)putchar(*s); |
1089 | s++; |
1090 | } |
1091 | s++; |
1092 | } |
1093 | (void)putchar('\n'); |
1094 | } |
1095 | |
1096 | /* Header fields of interest. */ |
1097 | if (r->regstart) |
1098 | fprintf(stderr,"start `%s' ", r->regstart->str_ptr); |
1099 | if (r->regstclass) |
1100 | fprintf(stderr,"stclass `%s' ", regprop(r->regstclass)); |
1101 | if (r->reganch) |
1102 | fprintf(stderr,"anchored "); |
1103 | if (r->regmust != NULL) |
1104 | fprintf(stderr,"must have \"%s\" back %d ", r->regmust->str_ptr, |
1105 | r->regback); |
1106 | fprintf(stderr,"\n"); |
1107 | } |
1108 | |
1109 | /* |
1110 | - regprop - printable representation of opcode |
1111 | */ |
1112 | char * |
1113 | regprop(op) |
1114 | char *op; |
1115 | { |
1116 | register char *p; |
1117 | |
1118 | (void) strcpy(buf, ":"); |
1119 | |
1120 | switch (OP(op)) { |
1121 | case BOL: |
1122 | p = "BOL"; |
1123 | break; |
1124 | case EOL: |
1125 | p = "EOL"; |
1126 | break; |
1127 | case ANY: |
1128 | p = "ANY"; |
1129 | break; |
1130 | case ANYOF: |
1131 | p = "ANYOF"; |
1132 | break; |
1133 | case ANYBUT: |
1134 | p = "ANYBUT"; |
1135 | break; |
1136 | case BRANCH: |
1137 | p = "BRANCH"; |
1138 | break; |
1139 | case EXACTLY: |
1140 | p = "EXACTLY"; |
1141 | break; |
1142 | case NOTHING: |
1143 | p = "NOTHING"; |
1144 | break; |
1145 | case BACK: |
1146 | p = "BACK"; |
1147 | break; |
1148 | case END: |
1149 | p = "END"; |
1150 | break; |
1151 | case ALNUM: |
1152 | p = "ALNUM"; |
1153 | break; |
1154 | case NALNUM: |
1155 | p = "NALNUM"; |
1156 | break; |
1157 | case BOUND: |
1158 | p = "BOUND"; |
1159 | break; |
1160 | case NBOUND: |
1161 | p = "NBOUND"; |
1162 | break; |
1163 | case SPACE: |
1164 | p = "SPACE"; |
1165 | break; |
1166 | case NSPACE: |
1167 | p = "NSPACE"; |
1168 | break; |
1169 | case DIGIT: |
1170 | p = "DIGIT"; |
1171 | break; |
1172 | case NDIGIT: |
1173 | p = "NDIGIT"; |
1174 | break; |
1175 | case REF: |
1176 | case REF+1: |
1177 | case REF+2: |
1178 | case REF+3: |
1179 | case REF+4: |
1180 | case REF+5: |
1181 | case REF+6: |
1182 | case REF+7: |
1183 | case REF+8: |
1184 | case REF+9: |
1185 | (void)sprintf(buf+strlen(buf), "REF%d", OP(op)-REF); |
1186 | p = NULL; |
1187 | break; |
1188 | case OPEN+1: |
1189 | case OPEN+2: |
1190 | case OPEN+3: |
1191 | case OPEN+4: |
1192 | case OPEN+5: |
1193 | case OPEN+6: |
1194 | case OPEN+7: |
1195 | case OPEN+8: |
1196 | case OPEN+9: |
1197 | (void)sprintf(buf+strlen(buf), "OPEN%d", OP(op)-OPEN); |
1198 | p = NULL; |
1199 | break; |
1200 | case CLOSE+1: |
1201 | case CLOSE+2: |
1202 | case CLOSE+3: |
1203 | case CLOSE+4: |
1204 | case CLOSE+5: |
1205 | case CLOSE+6: |
1206 | case CLOSE+7: |
1207 | case CLOSE+8: |
1208 | case CLOSE+9: |
1209 | (void)sprintf(buf+strlen(buf), "CLOSE%d", OP(op)-CLOSE); |
1210 | p = NULL; |
1211 | break; |
1212 | case STAR: |
1213 | p = "STAR"; |
1214 | break; |
1215 | case PLUS: |
1216 | p = "PLUS"; |
1217 | break; |
1218 | default: |
1219 | FAIL("corrupted regexp opcode"); |
1220 | } |
1221 | if (p != NULL) |
1222 | (void) strcat(buf, p); |
1223 | return(buf); |
1224 | } |
1225 | #endif /* DEBUGGING */ |
1226 | |
1227 | regfree(r) |
1228 | struct regexp *r; |
1229 | { |
1230 | if (r->precomp) |
1231 | Safefree(r->precomp); |
1232 | if (r->subbase) |
1233 | Safefree(r->subbase); |
1234 | if (r->regmust) |
1235 | str_free(r->regmust); |
1236 | if (r->regstart) |
1237 | str_free(r->regstart); |
1238 | Safefree(r); |
1239 | } |