1 | /*- |
2 | * This code is derived from OpenBSD's libc/regex, original license follows: |
3 | * |
4 | * Copyright (c) 1992, 1993, 1994 Henry Spencer. |
5 | * Copyright (c) 1992, 1993, 1994 |
6 | * The Regents of the University of California. All rights reserved. |
7 | * |
8 | * This code is derived from software contributed to Berkeley by |
9 | * Henry Spencer. |
10 | * |
11 | * Redistribution and use in source and binary forms, with or without |
12 | * modification, are permitted provided that the following conditions |
13 | * are met: |
14 | * 1. Redistributions of source code must retain the above copyright |
15 | * notice, this list of conditions and the following disclaimer. |
16 | * 2. Redistributions in binary form must reproduce the above copyright |
17 | * notice, this list of conditions and the following disclaimer in the |
18 | * documentation and/or other materials provided with the distribution. |
19 | * 3. Neither the name of the University nor the names of its contributors |
20 | * may be used to endorse or promote products derived from this software |
21 | * without specific prior written permission. |
22 | * |
23 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
24 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
25 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
26 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
27 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
28 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
29 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
30 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
31 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
32 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
33 | * SUCH DAMAGE. |
34 | * |
35 | * @(#)regcomp.c 8.5 (Berkeley) 3/20/94 |
36 | */ |
37 | |
38 | #include <sys/types.h> |
39 | #include <stdint.h> |
40 | #include <stdio.h> |
41 | #include <string.h> |
42 | #include <ctype.h> |
43 | #include <limits.h> |
44 | #include <stdlib.h> |
45 | #include "regex_impl.h" |
46 | |
47 | #include "regutils.h" |
48 | #include "regex2.h" |
49 | |
50 | #include "llvm/Config/config.h" |
51 | #include "llvm/Support/Compiler.h" |
52 | |
53 | /* character-class table */ |
54 | static struct cclass { |
55 | const char *name; |
56 | const char *chars; |
57 | const char *multis; |
58 | } cclasses[] = { |
59 | { "alnum" , "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz\ |
60 | 0123456789" , "" } , |
61 | { "alpha" , "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz" , |
62 | "" } , |
63 | { "blank" , " \t" , "" } , |
64 | { "cntrl" , "\007\b\t\n\v\f\r\1\2\3\4\5\6\16\17\20\21\22\23\24\ |
65 | \25\26\27\30\31\32\33\34\35\36\37\177" , "" } , |
66 | { "digit" , "0123456789" , "" } , |
67 | { "graph" , "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz\ |
68 | 0123456789!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~" , |
69 | "" } , |
70 | { "lower" , "abcdefghijklmnopqrstuvwxyz" , |
71 | "" } , |
72 | { "print" , "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz\ |
73 | 0123456789!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~ " , |
74 | "" } , |
75 | { "punct" , "!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~" , |
76 | "" } , |
77 | { "space" , "\t\n\v\f\r " , "" } , |
78 | { "upper" , "ABCDEFGHIJKLMNOPQRSTUVWXYZ" , |
79 | "" } , |
80 | { "xdigit" , "0123456789ABCDEFabcdef" , |
81 | "" } , |
82 | { NULL, 0, "" } |
83 | }; |
84 | |
85 | /* character-name table */ |
86 | static struct cname { |
87 | const char *name; |
88 | char code; |
89 | } cnames[] = { |
90 | { "NUL" , '\0' }, |
91 | { "SOH" , '\001' }, |
92 | { "STX" , '\002' }, |
93 | { "ETX" , '\003' }, |
94 | { "EOT" , '\004' }, |
95 | { "ENQ" , '\005' }, |
96 | { "ACK" , '\006' }, |
97 | { "BEL" , '\007' }, |
98 | { "alert" , '\007' }, |
99 | { "BS" , '\010' }, |
100 | { "backspace" , '\b' }, |
101 | { "HT" , '\011' }, |
102 | { "tab" , '\t' }, |
103 | { "LF" , '\012' }, |
104 | { "newline" , '\n' }, |
105 | { "VT" , '\013' }, |
106 | { "vertical-tab" , '\v' }, |
107 | { "FF" , '\014' }, |
108 | { "form-feed" , '\f' }, |
109 | { "CR" , '\015' }, |
110 | { "carriage-return" , '\r' }, |
111 | { "SO" , '\016' }, |
112 | { "SI" , '\017' }, |
113 | { "DLE" , '\020' }, |
114 | { "DC1" , '\021' }, |
115 | { "DC2" , '\022' }, |
116 | { "DC3" , '\023' }, |
117 | { "DC4" , '\024' }, |
118 | { "NAK" , '\025' }, |
119 | { "SYN" , '\026' }, |
120 | { "ETB" , '\027' }, |
121 | { "CAN" , '\030' }, |
122 | { "EM" , '\031' }, |
123 | { "SUB" , '\032' }, |
124 | { "ESC" , '\033' }, |
125 | { "IS4" , '\034' }, |
126 | { "FS" , '\034' }, |
127 | { "IS3" , '\035' }, |
128 | { "GS" , '\035' }, |
129 | { "IS2" , '\036' }, |
130 | { "RS" , '\036' }, |
131 | { "IS1" , '\037' }, |
132 | { "US" , '\037' }, |
133 | { "space" , ' ' }, |
134 | { "exclamation-mark" , '!' }, |
135 | { "quotation-mark" , '"' }, |
136 | { "number-sign" , '#' }, |
137 | { "dollar-sign" , '$' }, |
138 | { "percent-sign" , '%' }, |
139 | { "ampersand" , '&' }, |
140 | { "apostrophe" , '\'' }, |
141 | { "left-parenthesis" , '(' }, |
142 | { "right-parenthesis" , ')' }, |
143 | { "asterisk" , '*' }, |
144 | { "plus-sign" , '+' }, |
145 | { "comma" , ',' }, |
146 | { "hyphen" , '-' }, |
147 | { "hyphen-minus" , '-' }, |
148 | { "period" , '.' }, |
149 | { "full-stop" , '.' }, |
150 | { "slash" , '/' }, |
151 | { "solidus" , '/' }, |
152 | { "zero" , '0' }, |
153 | { "one" , '1' }, |
154 | { "two" , '2' }, |
155 | { "three" , '3' }, |
156 | { "four" , '4' }, |
157 | { "five" , '5' }, |
158 | { "six" , '6' }, |
159 | { "seven" , '7' }, |
160 | { "eight" , '8' }, |
161 | { "nine" , '9' }, |
162 | { "colon" , ':' }, |
163 | { "semicolon" , ';' }, |
164 | { "less-than-sign" , '<' }, |
165 | { "equals-sign" , '=' }, |
166 | { "greater-than-sign" , '>' }, |
167 | { "question-mark" , '?' }, |
168 | { "commercial-at" , '@' }, |
169 | { "left-square-bracket" , '[' }, |
170 | { "backslash" , '\\' }, |
171 | { "reverse-solidus" , '\\' }, |
172 | { "right-square-bracket" , ']' }, |
173 | { "circumflex" , '^' }, |
174 | { "circumflex-accent" , '^' }, |
175 | { "underscore" , '_' }, |
176 | { "low-line" , '_' }, |
177 | { "grave-accent" , '`' }, |
178 | { "left-brace" , '{' }, |
179 | { "left-curly-bracket" , '{' }, |
180 | { "vertical-line" , '|' }, |
181 | { "right-brace" , '}' }, |
182 | { "right-curly-bracket" , '}' }, |
183 | { "tilde" , '~' }, |
184 | { "DEL" , '\177' }, |
185 | { NULL, 0 } |
186 | }; |
187 | |
188 | /* |
189 | * parse structure, passed up and down to avoid global variables and |
190 | * other clumsinesses |
191 | */ |
192 | struct parse { |
193 | const char *next; /* next character in RE */ |
194 | const char *end; /* end of string (-> NUL normally) */ |
195 | int error; /* has an error been seen? */ |
196 | sop *strip; /* malloced strip */ |
197 | sopno ssize; /* malloced strip size (allocated) */ |
198 | sopno slen; /* malloced strip length (used) */ |
199 | int ncsalloc; /* number of csets allocated */ |
200 | struct re_guts *g; |
201 | # define NPAREN 10 /* we need to remember () 1-9 for back refs */ |
202 | sopno pbegin[NPAREN]; /* -> ( ([0] unused) */ |
203 | sopno pend[NPAREN]; /* -> ) ([0] unused) */ |
204 | }; |
205 | |
206 | static void p_ere(struct parse *, int); |
207 | static void p_ere_exp(struct parse *); |
208 | static void p_str(struct parse *); |
209 | static void p_bre(struct parse *, int, int); |
210 | static int p_simp_re(struct parse *, int); |
211 | static int p_count(struct parse *); |
212 | static void p_bracket(struct parse *); |
213 | static void p_b_term(struct parse *, cset *); |
214 | static void p_b_cclass(struct parse *, cset *); |
215 | static void p_b_eclass(struct parse *, cset *); |
216 | static char p_b_symbol(struct parse *); |
217 | static char p_b_coll_elem(struct parse *, int); |
218 | static char othercase(int); |
219 | static void bothcases(struct parse *, int); |
220 | static void ordinary(struct parse *, int); |
221 | static void nonnewline(struct parse *); |
222 | static void repeat(struct parse *, sopno, int, int); |
223 | static int seterr(struct parse *, int); |
224 | static cset *allocset(struct parse *); |
225 | static void freeset(struct parse *, cset *); |
226 | static int freezeset(struct parse *, cset *); |
227 | static int firstch(struct parse *, cset *); |
228 | static int nch(struct parse *, cset *); |
229 | static void mcadd(struct parse *, cset *, const char *); |
230 | static void mcinvert(struct parse *, cset *); |
231 | static void mccase(struct parse *, cset *); |
232 | static int isinsets(struct re_guts *, int); |
233 | static int samesets(struct re_guts *, int, int); |
234 | static void categorize(struct parse *, struct re_guts *); |
235 | static sopno dupl(struct parse *, sopno, sopno); |
236 | static void doemit(struct parse *, sop, size_t); |
237 | static void doinsert(struct parse *, sop, size_t, sopno); |
238 | static void dofwd(struct parse *, sopno, sop); |
239 | static void enlarge(struct parse *, sopno); |
240 | static void stripsnug(struct parse *, struct re_guts *); |
241 | static void findmust(struct parse *, struct re_guts *); |
242 | static sopno pluscount(struct parse *, struct re_guts *); |
243 | |
244 | static char nuls[10]; /* place to point scanner in event of error */ |
245 | |
246 | /* |
247 | * macros for use with parse structure |
248 | * BEWARE: these know that the parse structure is named `p' !!! |
249 | */ |
250 | #define PEEK() (*p->next) |
251 | #define PEEK2() (*(p->next+1)) |
252 | #define MORE() (p->end - p->next > 0) |
253 | #define MORE2() (p->end - p->next > 1) |
254 | #define SEE(c) (MORE() && PEEK() == (c)) |
255 | #define SEETWO(a, b) (MORE2() && PEEK() == (a) && PEEK2() == (b)) |
256 | #define EAT(c) ((SEE(c)) ? (NEXT(), 1) : 0) |
257 | #define EATTWO(a, b) ((SEETWO(a, b)) ? (NEXT2(), 1) : 0) |
258 | #define NEXT() (p->next++) |
259 | #define NEXT2() (p->next += 2) |
260 | #define NEXTn(n) (p->next += (n)) |
261 | #define GETNEXT() (*p->next++) |
262 | #define SETERROR(e) seterr(p, (e)) |
263 | #define REQUIRE(co, e) (void)((co) || SETERROR(e)) |
264 | #define MUSTSEE(c, e) (REQUIRE(MORE() && PEEK() == (c), e)) |
265 | #define MUSTEAT(c, e) (REQUIRE(MORE() && GETNEXT() == (c), e)) |
266 | #define MUSTNOTSEE(c, e) (REQUIRE(!MORE() || PEEK() != (c), e)) |
267 | #define EMIT(op, sopnd) doemit(p, (sop)(op), (size_t)(sopnd)) |
268 | #define INSERT(op, pos) doinsert(p, (sop)(op), HERE()-(pos)+1, pos) |
269 | #define AHEAD(pos) dofwd(p, pos, HERE()-(pos)) |
270 | #define ASTERN(sop, pos) EMIT(sop, HERE()-pos) |
271 | #define HERE() (p->slen) |
272 | #define THERE() (p->slen - 1) |
273 | #define THERETHERE() (p->slen - 2) |
274 | #define DROP(n) (p->slen -= (n)) |
275 | |
276 | #ifdef _POSIX2_RE_DUP_MAX |
277 | #define DUPMAX _POSIX2_RE_DUP_MAX |
278 | #else |
279 | #define DUPMAX 255 |
280 | #endif |
281 | #define REGINFINITY (DUPMAX + 1) |
282 | |
283 | #ifndef NDEBUG |
284 | static int never = 0; /* for use in asserts; shuts lint up */ |
285 | #else |
286 | #define never 0 /* some <assert.h>s have bugs too */ |
287 | #endif |
288 | |
289 | /* |
290 | - llvm_regcomp - interface for parser and compilation |
291 | */ |
292 | int /* 0 success, otherwise REG_something */ |
293 | llvm_regcomp(llvm_regex_t *preg, const char *pattern, int cflags) |
294 | { |
295 | struct parse pa; |
296 | struct re_guts *g; |
297 | struct parse *p = &pa; |
298 | int i; |
299 | size_t len; |
300 | #ifdef REDEBUG |
301 | # define GOODFLAGS(f) (f) |
302 | #else |
303 | # define GOODFLAGS(f) ((f)&~REG_DUMP) |
304 | #endif |
305 | |
306 | cflags = GOODFLAGS(cflags); |
307 | if ((cflags®_EXTENDED) && (cflags®_NOSPEC)) |
308 | return(REG_INVARG); |
309 | |
310 | if (cflags®_PEND) { |
311 | if (preg->re_endp < pattern) |
312 | return(REG_INVARG); |
313 | len = preg->re_endp - pattern; |
314 | } else |
315 | len = strlen(s: (const char *)pattern); |
316 | |
317 | /* do the mallocs early so failure handling is easy */ |
318 | g = (struct re_guts *)malloc(size: sizeof(struct re_guts) + |
319 | (NC-1)*sizeof(cat_t)); |
320 | if (g == NULL) |
321 | return(REG_ESPACE); |
322 | p->ssize = len/(size_t)2*(size_t)3 + (size_t)1; /* ugh */ |
323 | p->strip = (sop *)calloc(nmemb: p->ssize, size: sizeof(sop)); |
324 | p->slen = 0; |
325 | if (p->strip == NULL) { |
326 | free(ptr: (char *)g); |
327 | return(REG_ESPACE); |
328 | } |
329 | |
330 | /* set things up */ |
331 | p->g = g; |
332 | p->next = pattern; |
333 | p->end = p->next + len; |
334 | p->error = 0; |
335 | p->ncsalloc = 0; |
336 | for (i = 0; i < NPAREN; i++) { |
337 | p->pbegin[i] = 0; |
338 | p->pend[i] = 0; |
339 | } |
340 | g->csetsize = NC; |
341 | g->sets = NULL; |
342 | g->setbits = NULL; |
343 | g->ncsets = 0; |
344 | g->cflags = cflags; |
345 | g->iflags = 0; |
346 | g->nbol = 0; |
347 | g->neol = 0; |
348 | g->must = NULL; |
349 | g->mlen = 0; |
350 | g->nsub = 0; |
351 | g->ncategories = 1; /* category 0 is "everything else" */ |
352 | g->categories = &g->catspace[-(CHAR_MIN)]; |
353 | (void) memset(s: (char *)g->catspace, c: 0, NC*sizeof(cat_t)); |
354 | g->backrefs = 0; |
355 | |
356 | /* do it */ |
357 | EMIT(OEND, 0); |
358 | g->firststate = THERE(); |
359 | if (cflags®_EXTENDED) |
360 | p_ere(p, OUT); |
361 | else if (cflags®_NOSPEC) |
362 | p_str(p); |
363 | else |
364 | p_bre(p, OUT, OUT); |
365 | EMIT(OEND, 0); |
366 | g->laststate = THERE(); |
367 | |
368 | /* tidy up loose ends and fill things in */ |
369 | categorize(p, g); |
370 | stripsnug(p, g); |
371 | findmust(p, g); |
372 | g->nplus = pluscount(p, g); |
373 | g->magic = MAGIC2; |
374 | preg->re_nsub = g->nsub; |
375 | preg->re_g = g; |
376 | preg->re_magic = MAGIC1; |
377 | #ifndef REDEBUG |
378 | /* not debugging, so can't rely on the assert() in llvm_regexec() */ |
379 | if (g->iflags®EX_BAD) |
380 | SETERROR(REG_ASSERT); |
381 | #endif |
382 | |
383 | /* win or lose, we're done */ |
384 | if (p->error != 0) /* lose */ |
385 | llvm_regfree(preg); |
386 | return(p->error); |
387 | } |
388 | |
389 | /* |
390 | - p_ere - ERE parser top level, concatenation and alternation |
391 | */ |
392 | static void |
393 | p_ere(struct parse *p, int stop) /* character this ERE should end at */ |
394 | { |
395 | char c; |
396 | sopno prevback = 0; |
397 | sopno prevfwd = 0; |
398 | sopno conc; |
399 | int first = 1; /* is this the first alternative? */ |
400 | |
401 | for (;;) { |
402 | /* do a bunch of concatenated expressions */ |
403 | conc = HERE(); |
404 | while (MORE() && (c = PEEK()) != '|' && c != stop) |
405 | p_ere_exp(p); |
406 | REQUIRE(HERE() != conc, REG_EMPTY); /* require nonempty */ |
407 | |
408 | if (!EAT('|')) |
409 | break; /* NOTE BREAK OUT */ |
410 | |
411 | if (first) { |
412 | INSERT(OCH_, conc); /* offset is wrong */ |
413 | prevfwd = conc; |
414 | prevback = conc; |
415 | first = 0; |
416 | } |
417 | ASTERN(OOR1, prevback); |
418 | prevback = THERE(); |
419 | AHEAD(prevfwd); /* fix previous offset */ |
420 | prevfwd = HERE(); |
421 | EMIT(OOR2, 0); /* offset is very wrong */ |
422 | } |
423 | |
424 | if (!first) { /* tail-end fixups */ |
425 | AHEAD(prevfwd); |
426 | ASTERN(O_CH, prevback); |
427 | } |
428 | |
429 | assert(!MORE() || SEE(stop)); |
430 | } |
431 | |
432 | /* |
433 | - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op |
434 | */ |
435 | static void |
436 | p_ere_exp(struct parse *p) |
437 | { |
438 | char c; |
439 | sopno pos; |
440 | int count; |
441 | int count2; |
442 | int backrefnum; |
443 | sopno subno; |
444 | int wascaret = 0; |
445 | |
446 | assert(MORE()); /* caller should have ensured this */ |
447 | c = GETNEXT(); |
448 | |
449 | pos = HERE(); |
450 | switch (c) { |
451 | case '(': |
452 | REQUIRE(MORE(), REG_EPAREN); |
453 | p->g->nsub++; |
454 | subno = p->g->nsub; |
455 | if (subno < NPAREN) |
456 | p->pbegin[subno] = HERE(); |
457 | EMIT(OLPAREN, subno); |
458 | if (!SEE(')')) |
459 | p_ere(p, stop: ')'); |
460 | if (subno < NPAREN) { |
461 | p->pend[subno] = HERE(); |
462 | assert(p->pend[subno] != 0); |
463 | } |
464 | EMIT(ORPAREN, subno); |
465 | MUSTEAT(')', REG_EPAREN); |
466 | break; |
467 | #ifndef POSIX_MISTAKE |
468 | case ')': /* happens only if no current unmatched ( */ |
469 | /* |
470 | * You may ask, why the ifndef? Because I didn't notice |
471 | * this until slightly too late for 1003.2, and none of the |
472 | * other 1003.2 regular-expression reviewers noticed it at |
473 | * all. So an unmatched ) is legal POSIX, at least until |
474 | * we can get it fixed. |
475 | */ |
476 | SETERROR(REG_EPAREN); |
477 | break; |
478 | #endif |
479 | case '^': |
480 | EMIT(OBOL, 0); |
481 | p->g->iflags |= USEBOL; |
482 | p->g->nbol++; |
483 | wascaret = 1; |
484 | break; |
485 | case '$': |
486 | EMIT(OEOL, 0); |
487 | p->g->iflags |= USEEOL; |
488 | p->g->neol++; |
489 | break; |
490 | case '|': |
491 | SETERROR(REG_EMPTY); |
492 | break; |
493 | case '*': |
494 | case '+': |
495 | case '?': |
496 | SETERROR(REG_BADRPT); |
497 | break; |
498 | case '.': |
499 | if (p->g->cflags®_NEWLINE) |
500 | nonnewline(p); |
501 | else |
502 | EMIT(OANY, 0); |
503 | break; |
504 | case '[': |
505 | p_bracket(p); |
506 | break; |
507 | case '\\': |
508 | REQUIRE(MORE(), REG_EESCAPE); |
509 | c = GETNEXT(); |
510 | if (c >= '1' && c <= '9') { |
511 | /* \[0-9] is taken to be a back-reference to a previously specified |
512 | * matching group. backrefnum will hold the number. The matching |
513 | * group must exist (i.e. if \4 is found there must have been at |
514 | * least 4 matching groups specified in the pattern previously). |
515 | */ |
516 | backrefnum = c - '0'; |
517 | if (p->pend[backrefnum] == 0) { |
518 | SETERROR(REG_ESUBREG); |
519 | break; |
520 | } |
521 | |
522 | /* Make sure everything checks out and emit the sequence |
523 | * that marks a back-reference to the parse structure. |
524 | */ |
525 | assert(backrefnum <= p->g->nsub); |
526 | EMIT(OBACK_, backrefnum); |
527 | assert(p->pbegin[backrefnum] != 0); |
528 | assert(OP(p->strip[p->pbegin[backrefnum]]) == OLPAREN); |
529 | assert(OP(p->strip[p->pend[backrefnum]]) == ORPAREN); |
530 | (void) dupl(p, p->pbegin[backrefnum]+1, p->pend[backrefnum]); |
531 | EMIT(O_BACK, backrefnum); |
532 | p->g->backrefs = 1; |
533 | } else { |
534 | /* Other chars are simply themselves when escaped with a backslash. |
535 | */ |
536 | ordinary(p, c); |
537 | } |
538 | break; |
539 | case '{': /* okay as ordinary except if digit follows */ |
540 | REQUIRE(!MORE() || !isdigit((uch)PEEK()), REG_BADRPT); |
541 | LLVM_FALLTHROUGH; |
542 | default: |
543 | ordinary(p, c); |
544 | break; |
545 | } |
546 | |
547 | if (!MORE()) |
548 | return; |
549 | c = PEEK(); |
550 | /* we call { a repetition if followed by a digit */ |
551 | if (!( c == '*' || c == '+' || c == '?' || |
552 | (c == '{' && MORE2() && isdigit((uch)PEEK2())) )) |
553 | return; /* no repetition, we're done */ |
554 | NEXT(); |
555 | |
556 | REQUIRE(!wascaret, REG_BADRPT); |
557 | switch (c) { |
558 | case '*': /* implemented as +? */ |
559 | /* this case does not require the (y|) trick, noKLUDGE */ |
560 | INSERT(OPLUS_, pos); |
561 | ASTERN(O_PLUS, pos); |
562 | INSERT(OQUEST_, pos); |
563 | ASTERN(O_QUEST, pos); |
564 | break; |
565 | case '+': |
566 | INSERT(OPLUS_, pos); |
567 | ASTERN(O_PLUS, pos); |
568 | break; |
569 | case '?': |
570 | /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */ |
571 | INSERT(OCH_, pos); /* offset slightly wrong */ |
572 | ASTERN(OOR1, pos); /* this one's right */ |
573 | AHEAD(pos); /* fix the OCH_ */ |
574 | EMIT(OOR2, 0); /* offset very wrong... */ |
575 | AHEAD(THERE()); /* ...so fix it */ |
576 | ASTERN(O_CH, THERETHERE()); |
577 | break; |
578 | case '{': |
579 | count = p_count(p); |
580 | if (EAT(',')) { |
581 | if (isdigit((uch)PEEK())) { |
582 | count2 = p_count(p); |
583 | REQUIRE(count <= count2, REG_BADBR); |
584 | } else /* single number with comma */ |
585 | count2 = REGINFINITY; |
586 | } else /* just a single number */ |
587 | count2 = count; |
588 | repeat(p, pos, count, count2); |
589 | if (!EAT('}')) { /* error heuristics */ |
590 | while (MORE() && PEEK() != '}') |
591 | NEXT(); |
592 | REQUIRE(MORE(), REG_EBRACE); |
593 | SETERROR(REG_BADBR); |
594 | } |
595 | break; |
596 | } |
597 | |
598 | if (!MORE()) |
599 | return; |
600 | c = PEEK(); |
601 | if (!( c == '*' || c == '+' || c == '?' || |
602 | (c == '{' && MORE2() && isdigit((uch)PEEK2())) ) ) |
603 | return; |
604 | SETERROR(REG_BADRPT); |
605 | } |
606 | |
607 | /* |
608 | - p_str - string (no metacharacters) "parser" |
609 | */ |
610 | static void |
611 | p_str(struct parse *p) |
612 | { |
613 | REQUIRE(MORE(), REG_EMPTY); |
614 | while (MORE()) |
615 | ordinary(p, GETNEXT()); |
616 | } |
617 | |
618 | /* |
619 | - p_bre - BRE parser top level, anchoring and concatenation |
620 | * Giving end1 as OUT essentially eliminates the end1/end2 check. |
621 | * |
622 | * This implementation is a bit of a kludge, in that a trailing $ is first |
623 | * taken as an ordinary character and then revised to be an anchor. The |
624 | * only undesirable side effect is that '$' gets included as a character |
625 | * category in such cases. This is fairly harmless; not worth fixing. |
626 | * The amount of lookahead needed to avoid this kludge is excessive. |
627 | */ |
628 | static void |
629 | p_bre(struct parse *p, |
630 | int end1, /* first terminating character */ |
631 | int end2) /* second terminating character */ |
632 | { |
633 | sopno start = HERE(); |
634 | int first = 1; /* first subexpression? */ |
635 | int wasdollar = 0; |
636 | |
637 | if (EAT('^')) { |
638 | EMIT(OBOL, 0); |
639 | p->g->iflags |= USEBOL; |
640 | p->g->nbol++; |
641 | } |
642 | while (MORE() && !SEETWO(end1, end2)) { |
643 | wasdollar = p_simp_re(p, first); |
644 | first = 0; |
645 | } |
646 | if (wasdollar) { /* oops, that was a trailing anchor */ |
647 | DROP(1); |
648 | EMIT(OEOL, 0); |
649 | p->g->iflags |= USEEOL; |
650 | p->g->neol++; |
651 | } |
652 | |
653 | REQUIRE(HERE() != start, REG_EMPTY); /* require nonempty */ |
654 | } |
655 | |
656 | /* |
657 | - p_simp_re - parse a simple RE, an atom possibly followed by a repetition |
658 | */ |
659 | static int /* was the simple RE an unbackslashed $? */ |
660 | p_simp_re(struct parse *p, |
661 | int starordinary) /* is a leading * an ordinary character? */ |
662 | { |
663 | int c; |
664 | int count; |
665 | int count2; |
666 | sopno pos; |
667 | int i; |
668 | sopno subno; |
669 | # define BACKSL (1<<CHAR_BIT) |
670 | |
671 | pos = HERE(); /* repetition op, if any, covers from here */ |
672 | |
673 | assert(MORE()); /* caller should have ensured this */ |
674 | c = GETNEXT(); |
675 | if (c == '\\') { |
676 | REQUIRE(MORE(), REG_EESCAPE); |
677 | c = BACKSL | GETNEXT(); |
678 | } |
679 | switch (c) { |
680 | case '.': |
681 | if (p->g->cflags®_NEWLINE) |
682 | nonnewline(p); |
683 | else |
684 | EMIT(OANY, 0); |
685 | break; |
686 | case '[': |
687 | p_bracket(p); |
688 | break; |
689 | case BACKSL|'{': |
690 | SETERROR(REG_BADRPT); |
691 | break; |
692 | case BACKSL|'(': |
693 | p->g->nsub++; |
694 | subno = p->g->nsub; |
695 | if (subno < NPAREN) |
696 | p->pbegin[subno] = HERE(); |
697 | EMIT(OLPAREN, subno); |
698 | /* the MORE here is an error heuristic */ |
699 | if (MORE() && !SEETWO('\\', ')')) |
700 | p_bre(p, end1: '\\', end2: ')'); |
701 | if (subno < NPAREN) { |
702 | p->pend[subno] = HERE(); |
703 | assert(p->pend[subno] != 0); |
704 | } |
705 | EMIT(ORPAREN, subno); |
706 | REQUIRE(EATTWO('\\', ')'), REG_EPAREN); |
707 | break; |
708 | case BACKSL|')': /* should not get here -- must be user */ |
709 | case BACKSL|'}': |
710 | SETERROR(REG_EPAREN); |
711 | break; |
712 | case BACKSL|'1': |
713 | case BACKSL|'2': |
714 | case BACKSL|'3': |
715 | case BACKSL|'4': |
716 | case BACKSL|'5': |
717 | case BACKSL|'6': |
718 | case BACKSL|'7': |
719 | case BACKSL|'8': |
720 | case BACKSL|'9': |
721 | i = (c&~BACKSL) - '0'; |
722 | assert(i < NPAREN); |
723 | if (p->pend[i] != 0) { |
724 | assert(i <= p->g->nsub); |
725 | EMIT(OBACK_, i); |
726 | assert(p->pbegin[i] != 0); |
727 | assert(OP(p->strip[p->pbegin[i]]) == OLPAREN); |
728 | assert(OP(p->strip[p->pend[i]]) == ORPAREN); |
729 | (void) dupl(p, p->pbegin[i]+1, p->pend[i]); |
730 | EMIT(O_BACK, i); |
731 | } else |
732 | SETERROR(REG_ESUBREG); |
733 | p->g->backrefs = 1; |
734 | break; |
735 | case '*': |
736 | REQUIRE(starordinary, REG_BADRPT); |
737 | LLVM_FALLTHROUGH; |
738 | default: |
739 | ordinary(p, (char)c); |
740 | break; |
741 | } |
742 | |
743 | if (EAT('*')) { /* implemented as +? */ |
744 | /* this case does not require the (y|) trick, noKLUDGE */ |
745 | INSERT(OPLUS_, pos); |
746 | ASTERN(O_PLUS, pos); |
747 | INSERT(OQUEST_, pos); |
748 | ASTERN(O_QUEST, pos); |
749 | } else if (EATTWO('\\', '{')) { |
750 | count = p_count(p); |
751 | if (EAT(',')) { |
752 | if (MORE() && isdigit((uch)PEEK())) { |
753 | count2 = p_count(p); |
754 | REQUIRE(count <= count2, REG_BADBR); |
755 | } else /* single number with comma */ |
756 | count2 = REGINFINITY; |
757 | } else /* just a single number */ |
758 | count2 = count; |
759 | repeat(p, pos, count, count2); |
760 | if (!EATTWO('\\', '}')) { /* error heuristics */ |
761 | while (MORE() && !SEETWO('\\', '}')) |
762 | NEXT(); |
763 | REQUIRE(MORE(), REG_EBRACE); |
764 | SETERROR(REG_BADBR); |
765 | } |
766 | } else if (c == '$') /* $ (but not \$) ends it */ |
767 | return(1); |
768 | |
769 | return(0); |
770 | } |
771 | |
772 | /* |
773 | - p_count - parse a repetition count |
774 | */ |
775 | static int /* the value */ |
776 | p_count(struct parse *p) |
777 | { |
778 | int count = 0; |
779 | int ndigits = 0; |
780 | |
781 | while (MORE() && isdigit((uch)PEEK()) && count <= DUPMAX) { |
782 | count = count*10 + (GETNEXT() - '0'); |
783 | ndigits++; |
784 | } |
785 | |
786 | REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR); |
787 | return(count); |
788 | } |
789 | |
790 | /* |
791 | - p_bracket - parse a bracketed character list |
792 | * |
793 | * Note a significant property of this code: if the allocset() did SETERROR, |
794 | * no set operations are done. |
795 | */ |
796 | static void |
797 | p_bracket(struct parse *p) |
798 | { |
799 | cset *cs; |
800 | int invert = 0; |
801 | |
802 | /* Dept of Truly Sickening Special-Case Kludges */ |
803 | if (p->end - p->next > 5) { |
804 | if (strncmp(s1: p->next, s2: "[:<:]]" , n: 6) == 0) { |
805 | EMIT(OBOW, 0); |
806 | NEXTn(6); |
807 | return; |
808 | } |
809 | if (strncmp(s1: p->next, s2: "[:>:]]" , n: 6) == 0) { |
810 | EMIT(OEOW, 0); |
811 | NEXTn(6); |
812 | return; |
813 | } |
814 | } |
815 | |
816 | if ((cs = allocset(p)) == NULL) { |
817 | /* allocset did set error status in p */ |
818 | return; |
819 | } |
820 | |
821 | if (EAT('^')) |
822 | invert++; /* make note to invert set at end */ |
823 | if (EAT(']')) |
824 | CHadd(cs, ']'); |
825 | else if (EAT('-')) |
826 | CHadd(cs, '-'); |
827 | while (MORE() && PEEK() != ']' && !SEETWO('-', ']')) |
828 | p_b_term(p, cs); |
829 | if (EAT('-')) |
830 | CHadd(cs, '-'); |
831 | MUSTEAT(']', REG_EBRACK); |
832 | |
833 | if (p->error != 0) { /* don't mess things up further */ |
834 | freeset(p, cs); |
835 | return; |
836 | } |
837 | |
838 | if (p->g->cflags®_ICASE) { |
839 | int i; |
840 | int ci; |
841 | |
842 | for (i = p->g->csetsize - 1; i >= 0; i--) |
843 | if (CHIN(cs, i) && isalpha(i)) { |
844 | ci = othercase(i); |
845 | if (ci != i) |
846 | CHadd(cs, ci); |
847 | } |
848 | if (cs->multis != NULL) |
849 | mccase(p, cs); |
850 | } |
851 | if (invert) { |
852 | int i; |
853 | |
854 | for (i = p->g->csetsize - 1; i >= 0; i--) |
855 | if (CHIN(cs, i)) |
856 | CHsub(cs, i); |
857 | else |
858 | CHadd(cs, i); |
859 | if (p->g->cflags®_NEWLINE) |
860 | CHsub(cs, '\n'); |
861 | if (cs->multis != NULL) |
862 | mcinvert(p, cs); |
863 | } |
864 | |
865 | assert(cs->multis == NULL); /* xxx */ |
866 | |
867 | if (nch(p, cs) == 1) { /* optimize singleton sets */ |
868 | ordinary(p, firstch(p, cs)); |
869 | freeset(p, cs); |
870 | } else |
871 | EMIT(OANYOF, freezeset(p, cs)); |
872 | } |
873 | |
874 | /* |
875 | - p_b_term - parse one term of a bracketed character list |
876 | */ |
877 | static void |
878 | p_b_term(struct parse *p, cset *cs) |
879 | { |
880 | char c; |
881 | char start, finish; |
882 | int i; |
883 | |
884 | /* classify what we've got */ |
885 | switch ((MORE()) ? PEEK() : '\0') { |
886 | case '[': |
887 | c = (MORE2()) ? PEEK2() : '\0'; |
888 | break; |
889 | case '-': |
890 | SETERROR(REG_ERANGE); |
891 | return; /* NOTE RETURN */ |
892 | break; |
893 | default: |
894 | c = '\0'; |
895 | break; |
896 | } |
897 | |
898 | switch (c) { |
899 | case ':': /* character class */ |
900 | NEXT2(); |
901 | REQUIRE(MORE(), REG_EBRACK); |
902 | c = PEEK(); |
903 | REQUIRE(c != '-' && c != ']', REG_ECTYPE); |
904 | p_b_cclass(p, cs); |
905 | REQUIRE(MORE(), REG_EBRACK); |
906 | REQUIRE(EATTWO(':', ']'), REG_ECTYPE); |
907 | break; |
908 | case '=': /* equivalence class */ |
909 | NEXT2(); |
910 | REQUIRE(MORE(), REG_EBRACK); |
911 | c = PEEK(); |
912 | REQUIRE(c != '-' && c != ']', REG_ECOLLATE); |
913 | p_b_eclass(p, cs); |
914 | REQUIRE(MORE(), REG_EBRACK); |
915 | REQUIRE(EATTWO('=', ']'), REG_ECOLLATE); |
916 | break; |
917 | default: /* symbol, ordinary character, or range */ |
918 | /* xxx revision needed for multichar stuff */ |
919 | start = p_b_symbol(p); |
920 | if (SEE('-') && MORE2() && PEEK2() != ']') { |
921 | /* range */ |
922 | NEXT(); |
923 | if (EAT('-')) |
924 | finish = '-'; |
925 | else |
926 | finish = p_b_symbol(p); |
927 | } else |
928 | finish = start; |
929 | /* xxx what about signed chars here... */ |
930 | REQUIRE(start <= finish, REG_ERANGE); |
931 | for (i = start; i <= finish; i++) |
932 | CHadd(cs, i); |
933 | break; |
934 | } |
935 | } |
936 | |
937 | /* |
938 | - p_b_cclass - parse a character-class name and deal with it |
939 | */ |
940 | static void |
941 | p_b_cclass(struct parse *p, cset *cs) |
942 | { |
943 | const char *sp = p->next; |
944 | struct cclass *cp; |
945 | size_t len; |
946 | const char *u; |
947 | char c; |
948 | |
949 | while (MORE() && isalpha((uch)PEEK())) |
950 | NEXT(); |
951 | len = p->next - sp; |
952 | for (cp = cclasses; cp->name != NULL; cp++) |
953 | if (strncmp(s1: cp->name, s2: sp, n: len) == 0 && cp->name[len] == '\0') |
954 | break; |
955 | if (cp->name == NULL) { |
956 | /* oops, didn't find it */ |
957 | SETERROR(REG_ECTYPE); |
958 | return; |
959 | } |
960 | |
961 | u = cp->chars; |
962 | while ((c = *u++) != '\0') |
963 | CHadd(cs, c); |
964 | for (u = cp->multis; *u != '\0'; u += strlen(s: u) + 1) |
965 | MCadd(p, cs, u); |
966 | } |
967 | |
968 | /* |
969 | - p_b_eclass - parse an equivalence-class name and deal with it |
970 | * |
971 | * This implementation is incomplete. xxx |
972 | */ |
973 | static void |
974 | p_b_eclass(struct parse *p, cset *cs) |
975 | { |
976 | char c; |
977 | |
978 | c = p_b_coll_elem(p, '='); |
979 | CHadd(cs, c); |
980 | } |
981 | |
982 | /* |
983 | - p_b_symbol - parse a character or [..]ed multicharacter collating symbol |
984 | */ |
985 | static char /* value of symbol */ |
986 | p_b_symbol(struct parse *p) |
987 | { |
988 | char value; |
989 | |
990 | REQUIRE(MORE(), REG_EBRACK); |
991 | if (!EATTWO('[', '.')) |
992 | return(GETNEXT()); |
993 | |
994 | /* collating symbol */ |
995 | value = p_b_coll_elem(p, '.'); |
996 | REQUIRE(EATTWO('.', ']'), REG_ECOLLATE); |
997 | return(value); |
998 | } |
999 | |
1000 | /* |
1001 | - p_b_coll_elem - parse a collating-element name and look it up |
1002 | */ |
1003 | static char /* value of collating element */ |
1004 | p_b_coll_elem(struct parse *p, |
1005 | int endc) /* name ended by endc,']' */ |
1006 | { |
1007 | const char *sp = p->next; |
1008 | struct cname *cp; |
1009 | size_t len; |
1010 | |
1011 | while (MORE() && !SEETWO(endc, ']')) |
1012 | NEXT(); |
1013 | if (!MORE()) { |
1014 | SETERROR(REG_EBRACK); |
1015 | return(0); |
1016 | } |
1017 | len = p->next - sp; |
1018 | for (cp = cnames; cp->name != NULL; cp++) |
1019 | if (strncmp(s1: cp->name, s2: sp, n: len) == 0 && strlen(s: cp->name) == len) |
1020 | return(cp->code); /* known name */ |
1021 | if (len == 1) |
1022 | return(*sp); /* single character */ |
1023 | SETERROR(REG_ECOLLATE); /* neither */ |
1024 | return(0); |
1025 | } |
1026 | |
1027 | /* |
1028 | - othercase - return the case counterpart of an alphabetic |
1029 | */ |
1030 | static char /* if no counterpart, return ch */ |
1031 | othercase(int ch) |
1032 | { |
1033 | ch = (uch)ch; |
1034 | assert(isalpha(ch)); |
1035 | if (isupper(ch)) |
1036 | return ((uch)tolower(ch)); |
1037 | else if (islower(ch)) |
1038 | return ((uch)toupper(ch)); |
1039 | else /* peculiar, but could happen */ |
1040 | return(ch); |
1041 | } |
1042 | |
1043 | /* |
1044 | - bothcases - emit a dualcase version of a two-case character |
1045 | * |
1046 | * Boy, is this implementation ever a kludge... |
1047 | */ |
1048 | static void |
1049 | bothcases(struct parse *p, int ch) |
1050 | { |
1051 | const char *oldnext = p->next; |
1052 | const char *oldend = p->end; |
1053 | char bracket[3]; |
1054 | |
1055 | ch = (uch)ch; |
1056 | assert(othercase(ch) != ch); /* p_bracket() would recurse */ |
1057 | p->next = bracket; |
1058 | p->end = bracket+2; |
1059 | bracket[0] = ch; |
1060 | bracket[1] = ']'; |
1061 | bracket[2] = '\0'; |
1062 | p_bracket(p); |
1063 | assert(p->next == bracket+2); |
1064 | p->next = oldnext; |
1065 | p->end = oldend; |
1066 | } |
1067 | |
1068 | /* |
1069 | - ordinary - emit an ordinary character |
1070 | */ |
1071 | static void |
1072 | ordinary(struct parse *p, int ch) |
1073 | { |
1074 | cat_t *cap = p->g->categories; |
1075 | |
1076 | if ((p->g->cflags®_ICASE) && isalpha((uch)ch) && othercase(ch) != ch) |
1077 | bothcases(p, ch); |
1078 | else { |
1079 | EMIT(OCHAR, (uch)ch); |
1080 | if (cap[ch] == 0) |
1081 | cap[ch] = p->g->ncategories++; |
1082 | } |
1083 | } |
1084 | |
1085 | /* |
1086 | - nonnewline - emit REG_NEWLINE version of OANY |
1087 | * |
1088 | * Boy, is this implementation ever a kludge... |
1089 | */ |
1090 | static void |
1091 | nonnewline(struct parse *p) |
1092 | { |
1093 | const char *oldnext = p->next; |
1094 | const char *oldend = p->end; |
1095 | static const char bracket[4] = {'^', '\n', ']', '\0'}; |
1096 | |
1097 | p->next = bracket; |
1098 | p->end = bracket+3; |
1099 | p_bracket(p); |
1100 | assert(p->next == bracket+3); |
1101 | p->next = oldnext; |
1102 | p->end = oldend; |
1103 | } |
1104 | |
1105 | /* |
1106 | - repeat - generate code for a bounded repetition, recursively if needed |
1107 | */ |
1108 | static void |
1109 | repeat(struct parse *p, |
1110 | sopno start, /* operand from here to end of strip */ |
1111 | int from, /* repeated from this number */ |
1112 | int to) /* to this number of times (maybe INFINITY) */ |
1113 | { |
1114 | sopno finish = HERE(); |
1115 | # define N 2 |
1116 | # define INF 3 |
1117 | # define REP(f, t) ((f)*8 + (t)) |
1118 | # define MAP(n) (((n) <= 1) ? (n) : ((n) == REGINFINITY) ? INF : N) |
1119 | sopno copy; |
1120 | |
1121 | if (p->error != 0) /* head off possible runaway recursion */ |
1122 | return; |
1123 | |
1124 | assert(from <= to); |
1125 | |
1126 | switch (REP(MAP(from), MAP(to))) { |
1127 | case REP(0, 0): /* must be user doing this */ |
1128 | DROP(finish-start); /* drop the operand */ |
1129 | break; |
1130 | case REP(0, 1): /* as x{1,1}? */ |
1131 | case REP(0, N): /* as x{1,n}? */ |
1132 | case REP(0, INF): /* as x{1,}? */ |
1133 | /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */ |
1134 | INSERT(OCH_, start); /* offset is wrong... */ |
1135 | repeat(p, start: start+1, from: 1, to); |
1136 | ASTERN(OOR1, start); |
1137 | AHEAD(start); /* ... fix it */ |
1138 | EMIT(OOR2, 0); |
1139 | AHEAD(THERE()); |
1140 | ASTERN(O_CH, THERETHERE()); |
1141 | break; |
1142 | case REP(1, 1): /* trivial case */ |
1143 | /* done */ |
1144 | break; |
1145 | case REP(1, N): /* as x?x{1,n-1} */ |
1146 | /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */ |
1147 | INSERT(OCH_, start); |
1148 | ASTERN(OOR1, start); |
1149 | AHEAD(start); |
1150 | EMIT(OOR2, 0); /* offset very wrong... */ |
1151 | AHEAD(THERE()); /* ...so fix it */ |
1152 | ASTERN(O_CH, THERETHERE()); |
1153 | copy = dupl(p, start+1, finish+1); |
1154 | assert(copy == finish+4); |
1155 | repeat(p, start: copy, from: 1, to: to-1); |
1156 | break; |
1157 | case REP(1, INF): /* as x+ */ |
1158 | INSERT(OPLUS_, start); |
1159 | ASTERN(O_PLUS, start); |
1160 | break; |
1161 | case REP(N, N): /* as xx{m-1,n-1} */ |
1162 | copy = dupl(p, start, finish); |
1163 | repeat(p, start: copy, from: from-1, to: to-1); |
1164 | break; |
1165 | case REP(N, INF): /* as xx{n-1,INF} */ |
1166 | copy = dupl(p, start, finish); |
1167 | repeat(p, start: copy, from: from-1, to); |
1168 | break; |
1169 | default: /* "can't happen" */ |
1170 | SETERROR(REG_ASSERT); /* just in case */ |
1171 | break; |
1172 | } |
1173 | } |
1174 | |
1175 | /* |
1176 | - seterr - set an error condition |
1177 | */ |
1178 | static int /* useless but makes type checking happy */ |
1179 | seterr(struct parse *p, int e) |
1180 | { |
1181 | if (p->error == 0) /* keep earliest error condition */ |
1182 | p->error = e; |
1183 | p->next = nuls; /* try to bring things to a halt */ |
1184 | p->end = nuls; |
1185 | return(0); /* make the return value well-defined */ |
1186 | } |
1187 | |
1188 | /* |
1189 | - allocset - allocate a set of characters for [] |
1190 | */ |
1191 | static cset * |
1192 | allocset(struct parse *p) |
1193 | { |
1194 | int no = p->g->ncsets++; |
1195 | size_t nc; |
1196 | size_t nbytes; |
1197 | cset *cs; |
1198 | size_t css = (size_t)p->g->csetsize; |
1199 | int i; |
1200 | |
1201 | if (no >= p->ncsalloc) { /* need another column of space */ |
1202 | void *ptr; |
1203 | |
1204 | p->ncsalloc += CHAR_BIT; |
1205 | nc = p->ncsalloc; |
1206 | if (nc > SIZE_MAX / sizeof(cset)) |
1207 | goto nomem; |
1208 | assert(nc % CHAR_BIT == 0); |
1209 | nbytes = nc / CHAR_BIT * css; |
1210 | |
1211 | ptr = (cset *)realloc(ptr: (char *)p->g->sets, size: nc * sizeof(cset)); |
1212 | if (ptr == NULL) |
1213 | goto nomem; |
1214 | p->g->sets = ptr; |
1215 | |
1216 | ptr = (uch *)realloc(ptr: (char *)p->g->setbits, size: nbytes); |
1217 | if (ptr == NULL) |
1218 | goto nomem; |
1219 | p->g->setbits = ptr; |
1220 | |
1221 | for (i = 0; i < no; i++) |
1222 | p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT); |
1223 | |
1224 | (void) memset(s: (char *)p->g->setbits + (nbytes - css), c: 0, n: css); |
1225 | } |
1226 | /* XXX should not happen */ |
1227 | if (p->g->sets == NULL || p->g->setbits == NULL) |
1228 | goto nomem; |
1229 | |
1230 | cs = &p->g->sets[no]; |
1231 | cs->ptr = p->g->setbits + css*((no)/CHAR_BIT); |
1232 | cs->mask = 1 << ((no) % CHAR_BIT); |
1233 | cs->hash = 0; |
1234 | cs->smultis = 0; |
1235 | cs->multis = NULL; |
1236 | |
1237 | return(cs); |
1238 | nomem: |
1239 | free(ptr: p->g->sets); |
1240 | p->g->sets = NULL; |
1241 | free(ptr: p->g->setbits); |
1242 | p->g->setbits = NULL; |
1243 | |
1244 | SETERROR(REG_ESPACE); |
1245 | /* caller's responsibility not to do set ops */ |
1246 | return(NULL); |
1247 | } |
1248 | |
1249 | /* |
1250 | - freeset - free a now-unused set |
1251 | */ |
1252 | static void |
1253 | freeset(struct parse *p, cset *cs) |
1254 | { |
1255 | size_t i; |
1256 | cset *top = &p->g->sets[p->g->ncsets]; |
1257 | size_t css = (size_t)p->g->csetsize; |
1258 | |
1259 | for (i = 0; i < css; i++) |
1260 | CHsub(cs, i); |
1261 | if (cs == top-1) /* recover only the easy case */ |
1262 | p->g->ncsets--; |
1263 | } |
1264 | |
1265 | /* |
1266 | - freezeset - final processing on a set of characters |
1267 | * |
1268 | * The main task here is merging identical sets. This is usually a waste |
1269 | * of time (although the hash code minimizes the overhead), but can win |
1270 | * big if REG_ICASE is being used. REG_ICASE, by the way, is why the hash |
1271 | * is done using addition rather than xor -- all ASCII [aA] sets xor to |
1272 | * the same value! |
1273 | */ |
1274 | static int /* set number */ |
1275 | freezeset(struct parse *p, cset *cs) |
1276 | { |
1277 | uch h = cs->hash; |
1278 | size_t i; |
1279 | cset *top = &p->g->sets[p->g->ncsets]; |
1280 | cset *cs2; |
1281 | size_t css = (size_t)p->g->csetsize; |
1282 | |
1283 | /* look for an earlier one which is the same */ |
1284 | for (cs2 = &p->g->sets[0]; cs2 < top; cs2++) |
1285 | if (cs2->hash == h && cs2 != cs) { |
1286 | /* maybe */ |
1287 | for (i = 0; i < css; i++) |
1288 | if (!!CHIN(cs2, i) != !!CHIN(cs, i)) |
1289 | break; /* no */ |
1290 | if (i == css) |
1291 | break; /* yes */ |
1292 | } |
1293 | |
1294 | if (cs2 < top) { /* found one */ |
1295 | freeset(p, cs); |
1296 | cs = cs2; |
1297 | } |
1298 | |
1299 | return((int)(cs - p->g->sets)); |
1300 | } |
1301 | |
1302 | /* |
1303 | - firstch - return first character in a set (which must have at least one) |
1304 | */ |
1305 | static int /* character; there is no "none" value */ |
1306 | firstch(struct parse *p, cset *cs) |
1307 | { |
1308 | size_t i; |
1309 | size_t css = (size_t)p->g->csetsize; |
1310 | |
1311 | for (i = 0; i < css; i++) |
1312 | if (CHIN(cs, i)) |
1313 | return((char)i); |
1314 | assert(never); |
1315 | return(0); /* arbitrary */ |
1316 | } |
1317 | |
1318 | /* |
1319 | - nch - number of characters in a set |
1320 | */ |
1321 | static int |
1322 | nch(struct parse *p, cset *cs) |
1323 | { |
1324 | size_t i; |
1325 | size_t css = (size_t)p->g->csetsize; |
1326 | int n = 0; |
1327 | |
1328 | for (i = 0; i < css; i++) |
1329 | if (CHIN(cs, i)) |
1330 | n++; |
1331 | return(n); |
1332 | } |
1333 | |
1334 | /* |
1335 | - mcadd - add a collating element to a cset |
1336 | */ |
1337 | static void |
1338 | mcadd( struct parse *p, cset *cs, const char *cp) |
1339 | { |
1340 | size_t oldend = cs->smultis; |
1341 | void *np; |
1342 | |
1343 | cs->smultis += strlen(s: cp) + 1; |
1344 | np = realloc(ptr: cs->multis, size: cs->smultis); |
1345 | if (np == NULL) { |
1346 | if (cs->multis) |
1347 | free(ptr: cs->multis); |
1348 | cs->multis = NULL; |
1349 | SETERROR(REG_ESPACE); |
1350 | return; |
1351 | } |
1352 | cs->multis = np; |
1353 | |
1354 | llvm_strlcpy(dst: cs->multis + oldend - 1, src: cp, siz: cs->smultis - oldend + 1); |
1355 | } |
1356 | |
1357 | /* |
1358 | - mcinvert - invert the list of collating elements in a cset |
1359 | * |
1360 | * This would have to know the set of possibilities. Implementation |
1361 | * is deferred. |
1362 | */ |
1363 | /* ARGSUSED */ |
1364 | static void |
1365 | mcinvert(struct parse *p, cset *cs) |
1366 | { |
1367 | assert(cs->multis == NULL); /* xxx */ |
1368 | } |
1369 | |
1370 | /* |
1371 | - mccase - add case counterparts of the list of collating elements in a cset |
1372 | * |
1373 | * This would have to know the set of possibilities. Implementation |
1374 | * is deferred. |
1375 | */ |
1376 | /* ARGSUSED */ |
1377 | static void |
1378 | mccase(struct parse *p, cset *cs) |
1379 | { |
1380 | assert(cs->multis == NULL); /* xxx */ |
1381 | } |
1382 | |
1383 | /* |
1384 | - isinsets - is this character in any sets? |
1385 | */ |
1386 | static int /* predicate */ |
1387 | isinsets(struct re_guts *g, int c) |
1388 | { |
1389 | uch *col; |
1390 | int i; |
1391 | int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT; |
1392 | unsigned uc = (uch)c; |
1393 | |
1394 | for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize) |
1395 | if (col[uc] != 0) |
1396 | return(1); |
1397 | return(0); |
1398 | } |
1399 | |
1400 | /* |
1401 | - samesets - are these two characters in exactly the same sets? |
1402 | */ |
1403 | static int /* predicate */ |
1404 | samesets(struct re_guts *g, int c1, int c2) |
1405 | { |
1406 | uch *col; |
1407 | int i; |
1408 | int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT; |
1409 | unsigned uc1 = (uch)c1; |
1410 | unsigned uc2 = (uch)c2; |
1411 | |
1412 | for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize) |
1413 | if (col[uc1] != col[uc2]) |
1414 | return(0); |
1415 | return(1); |
1416 | } |
1417 | |
1418 | /* |
1419 | - categorize - sort out character categories |
1420 | */ |
1421 | static void |
1422 | categorize(struct parse *p, struct re_guts *g) |
1423 | { |
1424 | cat_t *cats = g->categories; |
1425 | int c; |
1426 | int c2; |
1427 | cat_t cat; |
1428 | |
1429 | /* avoid making error situations worse */ |
1430 | if (p->error != 0) |
1431 | return; |
1432 | |
1433 | for (c = CHAR_MIN; c <= CHAR_MAX; c++) |
1434 | if (cats[c] == 0 && isinsets(g, c)) { |
1435 | cat = g->ncategories++; |
1436 | cats[c] = cat; |
1437 | for (c2 = c+1; c2 <= CHAR_MAX; c2++) |
1438 | if (cats[c2] == 0 && samesets(g, c1: c, c2)) |
1439 | cats[c2] = cat; |
1440 | } |
1441 | } |
1442 | |
1443 | /* |
1444 | - dupl - emit a duplicate of a bunch of sops |
1445 | */ |
1446 | static sopno /* start of duplicate */ |
1447 | dupl(struct parse *p, |
1448 | sopno start, /* from here */ |
1449 | sopno finish) /* to this less one */ |
1450 | { |
1451 | sopno ret = HERE(); |
1452 | sopno len = finish - start; |
1453 | |
1454 | assert(finish >= start); |
1455 | if (len == 0) |
1456 | return(ret); |
1457 | enlarge(p, p->ssize + len); /* this many unexpected additions */ |
1458 | assert(p->ssize >= p->slen + len); |
1459 | (void) memmove(dest: (char *)(p->strip + p->slen), |
1460 | src: (char *)(p->strip + start), n: (size_t)len*sizeof(sop)); |
1461 | p->slen += len; |
1462 | return(ret); |
1463 | } |
1464 | |
1465 | /* |
1466 | - doemit - emit a strip operator |
1467 | * |
1468 | * It might seem better to implement this as a macro with a function as |
1469 | * hard-case backup, but it's just too big and messy unless there are |
1470 | * some changes to the data structures. Maybe later. |
1471 | */ |
1472 | static void |
1473 | doemit(struct parse *p, sop op, size_t opnd) |
1474 | { |
1475 | /* avoid making error situations worse */ |
1476 | if (p->error != 0) |
1477 | return; |
1478 | |
1479 | /* deal with oversize operands ("can't happen", more or less) */ |
1480 | assert(opnd < 1<<OPSHIFT); |
1481 | |
1482 | /* deal with undersized strip */ |
1483 | if (p->slen >= p->ssize) |
1484 | enlarge(p, (p->ssize+1) / 2 * 3); /* +50% */ |
1485 | assert(p->slen < p->ssize); |
1486 | |
1487 | /* finally, it's all reduced to the easy case */ |
1488 | p->strip[p->slen++] = SOP(op, opnd); |
1489 | } |
1490 | |
1491 | /* |
1492 | - doinsert - insert a sop into the strip |
1493 | */ |
1494 | static void |
1495 | doinsert(struct parse *p, sop op, size_t opnd, sopno pos) |
1496 | { |
1497 | sopno sn; |
1498 | sop s; |
1499 | int i; |
1500 | |
1501 | /* avoid making error situations worse */ |
1502 | if (p->error != 0) |
1503 | return; |
1504 | |
1505 | sn = HERE(); |
1506 | EMIT(op, opnd); /* do checks, ensure space */ |
1507 | assert(HERE() == sn+1); |
1508 | s = p->strip[sn]; |
1509 | |
1510 | /* adjust paren pointers */ |
1511 | assert(pos > 0); |
1512 | for (i = 1; i < NPAREN; i++) { |
1513 | if (p->pbegin[i] >= pos) { |
1514 | p->pbegin[i]++; |
1515 | } |
1516 | if (p->pend[i] >= pos) { |
1517 | p->pend[i]++; |
1518 | } |
1519 | } |
1520 | |
1521 | memmove(dest: (char *)&p->strip[pos+1], src: (char *)&p->strip[pos], |
1522 | n: (HERE()-pos-1)*sizeof(sop)); |
1523 | p->strip[pos] = s; |
1524 | } |
1525 | |
1526 | /* |
1527 | - dofwd - complete a forward reference |
1528 | */ |
1529 | static void |
1530 | dofwd(struct parse *p, sopno pos, sop value) |
1531 | { |
1532 | /* avoid making error situations worse */ |
1533 | if (p->error != 0) |
1534 | return; |
1535 | |
1536 | assert(value < 1<<OPSHIFT); |
1537 | p->strip[pos] = OP(p->strip[pos]) | value; |
1538 | } |
1539 | |
1540 | /* |
1541 | - enlarge - enlarge the strip |
1542 | */ |
1543 | static void |
1544 | enlarge(struct parse *p, sopno size) |
1545 | { |
1546 | sop *sp; |
1547 | |
1548 | if (p->ssize >= size) |
1549 | return; |
1550 | |
1551 | if ((uintptr_t)size > SIZE_MAX / sizeof(sop)) { |
1552 | SETERROR(REG_ESPACE); |
1553 | return; |
1554 | } |
1555 | |
1556 | sp = (sop *)realloc(ptr: p->strip, size: size*sizeof(sop)); |
1557 | if (sp == NULL) { |
1558 | SETERROR(REG_ESPACE); |
1559 | return; |
1560 | } |
1561 | p->strip = sp; |
1562 | p->ssize = size; |
1563 | } |
1564 | |
1565 | /* |
1566 | - stripsnug - compact the strip |
1567 | */ |
1568 | static void |
1569 | stripsnug(struct parse *p, struct re_guts *g) |
1570 | { |
1571 | g->nstates = p->slen; |
1572 | if ((uintptr_t)p->slen > SIZE_MAX / sizeof(sop)) { |
1573 | g->strip = p->strip; |
1574 | SETERROR(REG_ESPACE); |
1575 | return; |
1576 | } |
1577 | |
1578 | g->strip = (sop *)realloc(ptr: (char *)p->strip, size: p->slen * sizeof(sop)); |
1579 | if (g->strip == NULL) { |
1580 | SETERROR(REG_ESPACE); |
1581 | g->strip = p->strip; |
1582 | } |
1583 | } |
1584 | |
1585 | /* |
1586 | - findmust - fill in must and mlen with longest mandatory literal string |
1587 | * |
1588 | * This algorithm could do fancy things like analyzing the operands of | |
1589 | * for common subsequences. Someday. This code is simple and finds most |
1590 | * of the interesting cases. |
1591 | * |
1592 | * Note that must and mlen got initialized during setup. |
1593 | */ |
1594 | static void |
1595 | findmust(struct parse *p, struct re_guts *g) |
1596 | { |
1597 | sop *scan; |
1598 | sop *start = 0; /* start initialized in the default case, after that */ |
1599 | sop *newstart = 0; /* newstart was initialized in the OCHAR case */ |
1600 | sopno newlen; |
1601 | sop s; |
1602 | char *cp; |
1603 | sopno i; |
1604 | |
1605 | /* avoid making error situations worse */ |
1606 | if (p->error != 0) |
1607 | return; |
1608 | |
1609 | /* find the longest OCHAR sequence in strip */ |
1610 | newlen = 0; |
1611 | scan = g->strip + 1; |
1612 | do { |
1613 | s = *scan++; |
1614 | switch (OP(s)) { |
1615 | case OCHAR: /* sequence member */ |
1616 | if (newlen == 0) /* new sequence */ |
1617 | newstart = scan - 1; |
1618 | newlen++; |
1619 | break; |
1620 | case OPLUS_: /* things that don't break one */ |
1621 | case OLPAREN: |
1622 | case ORPAREN: |
1623 | break; |
1624 | case OQUEST_: /* things that must be skipped */ |
1625 | case OCH_: |
1626 | scan--; |
1627 | do { |
1628 | scan += OPND(s); |
1629 | s = *scan; |
1630 | /* assert() interferes w debug printouts */ |
1631 | if (OP(s) != O_QUEST && OP(s) != O_CH && |
1632 | OP(s) != OOR2) { |
1633 | g->iflags |= REGEX_BAD; |
1634 | return; |
1635 | } |
1636 | } while (OP(s) != O_QUEST && OP(s) != O_CH); |
1637 | LLVM_FALLTHROUGH; |
1638 | default: /* things that break a sequence */ |
1639 | if (newlen > g->mlen) { /* ends one */ |
1640 | start = newstart; |
1641 | g->mlen = newlen; |
1642 | } |
1643 | newlen = 0; |
1644 | break; |
1645 | } |
1646 | } while (OP(s) != OEND); |
1647 | |
1648 | if (g->mlen == 0) /* there isn't one */ |
1649 | return; |
1650 | |
1651 | /* turn it into a character string */ |
1652 | g->must = malloc(size: (size_t)g->mlen + 1); |
1653 | if (g->must == NULL) { /* argh; just forget it */ |
1654 | g->mlen = 0; |
1655 | return; |
1656 | } |
1657 | cp = g->must; |
1658 | scan = start; |
1659 | for (i = g->mlen; i > 0; i--) { |
1660 | while (OP(s = *scan++) != OCHAR) |
1661 | continue; |
1662 | assert(cp < g->must + g->mlen); |
1663 | *cp++ = (char)OPND(s); |
1664 | } |
1665 | assert(cp == g->must + g->mlen); |
1666 | *cp++ = '\0'; /* just on general principles */ |
1667 | } |
1668 | |
1669 | /* |
1670 | - pluscount - count + nesting |
1671 | */ |
1672 | static sopno /* nesting depth */ |
1673 | pluscount(struct parse *p, struct re_guts *g) |
1674 | { |
1675 | sop *scan; |
1676 | sop s; |
1677 | sopno plusnest = 0; |
1678 | sopno maxnest = 0; |
1679 | |
1680 | if (p->error != 0) |
1681 | return(0); /* there may not be an OEND */ |
1682 | |
1683 | scan = g->strip + 1; |
1684 | do { |
1685 | s = *scan++; |
1686 | switch (OP(s)) { |
1687 | case OPLUS_: |
1688 | plusnest++; |
1689 | break; |
1690 | case O_PLUS: |
1691 | if (plusnest > maxnest) |
1692 | maxnest = plusnest; |
1693 | plusnest--; |
1694 | break; |
1695 | } |
1696 | } while (OP(s) != OEND); |
1697 | if (plusnest != 0) |
1698 | g->iflags |= REGEX_BAD; |
1699 | return(maxnest); |
1700 | } |
1701 | |