?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Force a checkpoint in CREATE DATABASE before starting to copy the files,
[PostgreSQL.git] / src / backend / regex / regcomp.c
blob0da4c9e0700106a65f44f16af6a53c84b9e2340a
1 /*
2 * re_*comp and friends - compile REs
3 * This file #includes several others (see the bottom).
4 *
5 * Copyright (c) 1998, 1999 Henry Spencer. All rights reserved.
6 *
7 * Development of this software was funded, in part, by Cray Research Inc.,
8 * UUNET Communications Services Inc., Sun Microsystems Inc., and Scriptics
9 * Corporation, none of whom are responsible for the results. The author
10 * thanks all of them.
11 *
12 * Redistribution and use in source and binary forms -- with or without
13 * modification -- are permitted for any purpose, provided that
14 * redistributions in source form retain this entire copyright notice and
15 * indicate the origin and nature of any modifications.
16 *
17 * I'd appreciate being given credit for this package in the documentation
18 * of software which uses it, but that is not a requirement.
19 *
20 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES,
21 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
22 * AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL
23 * HENRY SPENCER BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
24 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
25 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
26 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
27 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
28 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
29 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30 *
31 * $PostgreSQL$
32 *
33 */
34
35 #include "regex/regguts.h"
36
37 /*
38 * forward declarations, up here so forward datatypes etc. are defined early
39 */
40 /* === regcomp.c === */
41 static void moresubs(struct vars *, int);
42 static int freev(struct vars *, int);
43 static void makesearch(struct vars *, struct nfa *);
44 static struct subre *parse(struct vars *, int, int, struct state *, struct state *);
45 static struct subre *parsebranch(struct vars *, int, int, struct state *, struct state *, int);
46 static void parseqatom(struct vars *, int, int, struct state *, struct state *, struct subre *);
47 static void nonword(struct vars *, int, struct state *, struct state *);
48 static void word(struct vars *, int, struct state *, struct state *);
49 static int scannum(struct vars *);
50 static void repeat(struct vars *, struct state *, struct state *, int, int);
51 static void bracket(struct vars *, struct state *, struct state *);
52 static void cbracket(struct vars *, struct state *, struct state *);
53 static void brackpart(struct vars *, struct state *, struct state *);
54 static const chr *scanplain(struct vars *);
55 static void onechr(struct vars *, chr, struct state *, struct state *);
56 static void dovec(struct vars *, struct cvec *, struct state *, struct state *);
57 static void wordchrs(struct vars *);
58 static struct subre *subre(struct vars *, int, int, struct state *, struct state *);
59 static void freesubre(struct vars *, struct subre *);
60 static void freesrnode(struct vars *, struct subre *);
61 static void optst(struct vars *, struct subre *);
62 static int numst(struct subre *, int);
63 static void markst(struct subre *);
64 static void cleanst(struct vars *);
65 static long nfatree(struct vars *, struct subre *, FILE *);
66 static long nfanode(struct vars *, struct subre *, FILE *);
67 static int newlacon(struct vars *, struct state *, struct state *, int);
68 static void freelacons(struct subre *, int);
69 static void rfree(regex_t *);
70
71 #ifdef REG_DEBUG
72 static void dump(regex_t *, FILE *);
73 static void dumpst(struct subre *, FILE *, int);
74 static void stdump(struct subre *, FILE *, int);
75 static const char *stid(struct subre *, char *, size_t);
76 #endif
77 /* === regc_lex.c === */
78 static void lexstart(struct vars *);
79 static void prefixes(struct vars *);
80 static void lexnest(struct vars *, const chr *, const chr *);
81 static void lexword(struct vars *);
82 static int next(struct vars *);
83 static int lexescape(struct vars *);
84 static chr lexdigits(struct vars *, int, int, int);
85 static int brenext(struct vars *, chr);
86 static void skip(struct vars *);
87 static chr newline(void);
88 static chr chrnamed(struct vars *, const chr *, const chr *, chr);
89
90 /* === regc_color.c === */
91 static void initcm(struct vars *, struct colormap *);
92 static void freecm(struct colormap *);
93 static void cmtreefree(struct colormap *, union tree *, int);
94 static color setcolor(struct colormap *, chr, pcolor);
95 static color maxcolor(struct colormap *);
96 static color newcolor(struct colormap *);
97 static void freecolor(struct colormap *, pcolor);
98 static color pseudocolor(struct colormap *);
99 static color subcolor(struct colormap *, chr c);
100 static color newsub(struct colormap *, pcolor);
101 static void subrange(struct vars *, chr, chr, struct state *, struct state *);
102 static void subblock(struct vars *, chr, struct state *, struct state *);
103 static void okcolors(struct nfa *, struct colormap *);
104 static void colorchain(struct colormap *, struct arc *);
105 static void uncolorchain(struct colormap *, struct arc *);
106 static void rainbow(struct nfa *, struct colormap *, int, pcolor, struct state *, struct state *);
107 static void colorcomplement(struct nfa *, struct colormap *, int, struct state *, struct state *, struct state *);
108
109 #ifdef REG_DEBUG
110 static void dumpcolors(struct colormap *, FILE *);
111 static void fillcheck(struct colormap *, union tree *, int, FILE *);
112 static void dumpchr(chr, FILE *);
113 #endif
114 /* === regc_nfa.c === */
115 static struct nfa *newnfa(struct vars *, struct colormap *, struct nfa *);
116 static void freenfa(struct nfa *);
117 static struct state *newstate(struct nfa *);
118 static struct state *newfstate(struct nfa *, int flag);
119 static void dropstate(struct nfa *, struct state *);
120 static void freestate(struct nfa *, struct state *);
121 static void destroystate(struct nfa *, struct state *);
122 static void newarc(struct nfa *, int, pcolor, struct state *, struct state *);
123 static struct arc *allocarc(struct nfa *, struct state *);
124 static void freearc(struct nfa *, struct arc *);
125 static struct arc *findarc(struct state *, int, pcolor);
126 static void cparc(struct nfa *, struct arc *, struct state *, struct state *);
127 static void moveins(struct nfa *, struct state *, struct state *);
128 static void copyins(struct nfa *, struct state *, struct state *);
129 static void moveouts(struct nfa *, struct state *, struct state *);
130 static void copyouts(struct nfa *, struct state *, struct state *);
131 static void cloneouts(struct nfa *, struct state *, struct state *, struct state *, int);
132 static void delsub(struct nfa *, struct state *, struct state *);
133 static void deltraverse(struct nfa *, struct state *, struct state *);
134 static void dupnfa(struct nfa *, struct state *, struct state *, struct state *, struct state *);
135 static void duptraverse(struct nfa *, struct state *, struct state *);
136 static void cleartraverse(struct nfa *, struct state *);
137 static void specialcolors(struct nfa *);
138 static long optimize(struct nfa *, FILE *);
139 static void pullback(struct nfa *, FILE *);
140 static int pull(struct nfa *, struct arc *);
141 static void pushfwd(struct nfa *, FILE *);
142 static int push(struct nfa *, struct arc *);
143
144 #define INCOMPATIBLE 1 /* destroys arc */
145 #define SATISFIED 2 /* constraint satisfied */
146 #define COMPATIBLE 3 /* compatible but not satisfied yet */
147 static int combine(struct arc *, struct arc *);
148 static void fixempties(struct nfa *, FILE *);
149 static int unempty(struct nfa *, struct arc *);
150 static void cleanup(struct nfa *);
151 static void markreachable(struct nfa *, struct state *, struct state *, struct state *);
152 static void markcanreach(struct nfa *, struct state *, struct state *, struct state *);
153 static long analyze(struct nfa *);
154 static void compact(struct nfa *, struct cnfa *);
155 static void carcsort(struct carc *, struct carc *);
156 static void freecnfa(struct cnfa *);
157 static void dumpnfa(struct nfa *, FILE *);
158
159 #ifdef REG_DEBUG
160 static void dumpstate(struct state *, FILE *);
161 static void dumparcs(struct state *, FILE *);
162 static int dumprarcs(struct arc *, struct state *, FILE *, int);
163 static void dumparc(struct arc *, struct state *, FILE *);
164 static void dumpcnfa(struct cnfa *, FILE *);
165 static void dumpcstate(int, struct carc *, struct cnfa *, FILE *);
166 #endif
167 /* === regc_cvec.c === */
168 static struct cvec *newcvec(int, int);
169 static struct cvec *clearcvec(struct cvec *);
170 static void addchr(struct cvec *, chr);
171 static void addrange(struct cvec *, chr, chr);
172 static struct cvec *getcvec(struct vars *, int, int);
173 static void freecvec(struct cvec *);
174
175 /* === regc_locale.c === */
176 static int pg_wc_isdigit(pg_wchar c);
177 static int pg_wc_isalpha(pg_wchar c);
178 static int pg_wc_isalnum(pg_wchar c);
179 static int pg_wc_isupper(pg_wchar c);
180 static int pg_wc_islower(pg_wchar c);
181 static int pg_wc_isgraph(pg_wchar c);
182 static int pg_wc_isprint(pg_wchar c);
183 static int pg_wc_ispunct(pg_wchar c);
184 static int pg_wc_isspace(pg_wchar c);
185 static pg_wchar pg_wc_toupper(pg_wchar c);
186 static pg_wchar pg_wc_tolower(pg_wchar c);
187 static celt element(struct vars *, const chr *, const chr *);
188 static struct cvec *range(struct vars *, celt, celt, int);
189 static int before(celt, celt);
190 static struct cvec *eclass(struct vars *, celt, int);
191 static struct cvec *cclass(struct vars *, const chr *, const chr *, int);
192 static struct cvec *allcases(struct vars *, chr);
193 static int cmp(const chr *, const chr *, size_t);
194 static int casecmp(const chr *, const chr *, size_t);
195
196
197 /* internal variables, bundled for easy passing around */
198 struct vars
199 {
200 regex_t *re;
201 const chr *now; /* scan pointer into string */
202 const chr *stop; /* end of string */
203 const chr *savenow; /* saved now and stop for "subroutine call" */
204 const chr *savestop;
205 int err; /* error code (0 if none) */
206 int cflags; /* copy of compile flags */
207 int lasttype; /* type of previous token */
208 int nexttype; /* type of next token */
209 chr nextvalue; /* value (if any) of next token */
210 int lexcon; /* lexical context type (see lex.c) */
211 int nsubexp; /* subexpression count */
212 struct subre **subs; /* subRE pointer vector */
213 size_t nsubs; /* length of vector */
214 struct subre *sub10[10]; /* initial vector, enough for most */
215 struct nfa *nfa; /* the NFA */
216 struct colormap *cm; /* character color map */
217 color nlcolor; /* color of newline */
218 struct state *wordchrs; /* state in nfa holding word-char outarcs */
219 struct subre *tree; /* subexpression tree */
220 struct subre *treechain; /* all tree nodes allocated */
221 struct subre *treefree; /* any free tree nodes */
222 int ntree; /* number of tree nodes */
223 struct cvec *cv; /* interface cvec */
224 struct cvec *cv2; /* utility cvec */
225 struct subre *lacons; /* lookahead-constraint vector */
226 int nlacons; /* size of lacons */
227 };
228
229 /* parsing macros; most know that `v' is the struct vars pointer */
230 #define NEXT() (next(v)) /* advance by one token */
231 #define SEE(t) (v->nexttype == (t)) /* is next token this? */
232 #define EAT(t) (SEE(t) && next(v)) /* if next is this, swallow it */
233 #define VISERR(vv) ((vv)->err != 0) /* have we seen an error yet? */
234 #define ISERR() VISERR(v)
235 #define VERR(vv,e) ((vv)->nexttype = EOS, ((vv)->err) ? (vv)->err :\
236 ((vv)->err = (e)))
237 #define ERR(e) VERR(v, e) /* record an error */
238 #define NOERR() {if (ISERR()) return;} /* if error seen, return */
239 #define NOERRN() {if (ISERR()) return NULL;} /* NOERR with retval */
240 #define NOERRZ() {if (ISERR()) return 0;} /* NOERR with retval */
241 #define INSIST(c, e) ((c) ? 0 : ERR(e)) /* if condition false, error */
242 #define NOTE(b) (v->re->re_info |= (b)) /* note visible condition */
243 #define EMPTYARC(x, y) newarc(v->nfa, EMPTY, 0, x, y)
244
245 /* token type codes, some also used as NFA arc types */
246 #define EMPTY 'n' /* no token present */
247 #define EOS 'e' /* end of string */
248 #define PLAIN 'p' /* ordinary character */
249 #define DIGIT 'd' /* digit (in bound) */
250 #define BACKREF 'b' /* back reference */
251 #define COLLEL 'I' /* start of [. */
252 #define ECLASS 'E' /* start of [= */
253 #define CCLASS 'C' /* start of [: */
254 #define END 'X' /* end of [. [= [: */
255 #define RANGE 'R' /* - within [] which might be range delim. */
256 #define LACON 'L' /* lookahead constraint subRE */
257 #define AHEAD 'a' /* color-lookahead arc */
258 #define BEHIND 'r' /* color-lookbehind arc */
259 #define WBDRY 'w' /* word boundary constraint */
260 #define NWBDRY 'W' /* non-word-boundary constraint */
261 #define SBEGIN 'A' /* beginning of string (even if not BOL) */
262 #define SEND 'Z' /* end of string (even if not EOL) */
263 #define PREFER 'P' /* length preference */
264
265 /* is an arc colored, and hence on a color chain? */
266 #define COLORED(a) \
267 ((a)->type == PLAIN || (a)->type == AHEAD || (a)->type == BEHIND)
268
269
270 /* static function list */
271 static struct fns functions = {
272 rfree, /* regfree insides */
273 };
274
275
276
277 /*
278 * pg_regcomp - compile regular expression
279 */
280 int
281 pg_regcomp(regex_t *re,
282 const chr *string,
283 size_t len,
284 int flags)
285 {
286 struct vars var;
287 struct vars *v = &var;
288 struct guts *g;
289 int i;
290 size_t j;
291
292 #ifdef REG_DEBUG
293 FILE *debug = (flags & REG_PROGRESS) ? stdout : (FILE *) NULL;
294 #else
295 FILE *debug = (FILE *) NULL;
296 #endif
297
298 #define CNOERR() { if (ISERR()) return freev(v, v->err); }
299
300 /* sanity checks */
301
302 if (re == NULL || string == NULL)
303 return REG_INVARG;
304 if ((flags & REG_QUOTE) &&
305 (flags & (REG_ADVANCED | REG_EXPANDED | REG_NEWLINE)))
306 return REG_INVARG;
307 if (!(flags & REG_EXTENDED) && (flags & REG_ADVF))
308 return REG_INVARG;
309
310 /* initial setup (after which freev() is callable) */
311 v->re = re;
312 v->now = string;
313 v->stop = v->now + len;
314 v->savenow = v->savestop = NULL;
315 v->err = 0;
316 v->cflags = flags;
317 v->nsubexp = 0;
318 v->subs = v->sub10;
319 v->nsubs = 10;
320 for (j = 0; j < v->nsubs; j++)
321 v->subs[j] = NULL;
322 v->nfa = NULL;
323 v->cm = NULL;
324 v->nlcolor = COLORLESS;
325 v->wordchrs = NULL;
326 v->tree = NULL;
327 v->treechain = NULL;
328 v->treefree = NULL;
329 v->cv = NULL;
330 v->cv2 = NULL;
331 v->lacons = NULL;
332 v->nlacons = 0;
333 re->re_magic = REMAGIC;
334 re->re_info = 0; /* bits get set during parse */
335 re->re_csize = sizeof(chr);
336 re->re_guts = NULL;
337 re->re_fns = VS(&functions);
338
339 /* more complex setup, malloced things */
340 re->re_guts = VS(MALLOC(sizeof(struct guts)));
341 if (re->re_guts == NULL)
342 return freev(v, REG_ESPACE);
343 g = (struct guts *) re->re_guts;
344 g->tree = NULL;
345 initcm(v, &g->cmap);
346 v->cm = &g->cmap;
347 g->lacons = NULL;
348 g->nlacons = 0;
349 ZAPCNFA(g->search);
350 v->nfa = newnfa(v, v->cm, (struct nfa *) NULL);
351 CNOERR();
352 v->cv = newcvec(100, 20);
353 if (v->cv == NULL)
354 return freev(v, REG_ESPACE);
355
356 /* parsing */
357 lexstart(v); /* also handles prefixes */
358 if ((v->cflags & REG_NLSTOP) || (v->cflags & REG_NLANCH))
359 {
360 /* assign newline a unique color */
361 v->nlcolor = subcolor(v->cm, newline());
362 okcolors(v->nfa, v->cm);
363 }
364 CNOERR();
365 v->tree = parse(v, EOS, PLAIN, v->nfa->init, v->nfa->final);
366 assert(SEE(EOS)); /* even if error; ISERR() => SEE(EOS) */
367 CNOERR();
368 assert(v->tree != NULL);
369
370 /* finish setup of nfa and its subre tree */
371 specialcolors(v->nfa);
372 CNOERR();
373 #ifdef REG_DEBUG
374 if (debug != NULL)
375 {
376 fprintf(debug, "\n\n\n========= RAW ==========\n");
377 dumpnfa(v->nfa, debug);
378 dumpst(v->tree, debug, 1);
379 }
380 #endif
381 optst(v, v->tree);
382 v->ntree = numst(v->tree, 1);
383 markst(v->tree);
384 cleanst(v);
385 #ifdef REG_DEBUG
386 if (debug != NULL)
387 {
388 fprintf(debug, "\n\n\n========= TREE FIXED ==========\n");
389 dumpst(v->tree, debug, 1);
390 }
391 #endif
392
393 /* build compacted NFAs for tree and lacons */
394 re->re_info |= nfatree(v, v->tree, debug);
395 CNOERR();
396 assert(v->nlacons == 0 || v->lacons != NULL);
397 for (i = 1; i < v->nlacons; i++)
398 {
399 #ifdef REG_DEBUG
400 if (debug != NULL)
401 fprintf(debug, "\n\n\n========= LA%d ==========\n", i);
402 #endif
403 nfanode(v, &v->lacons[i], debug);
404 }
405 CNOERR();
406 if (v->tree->flags & SHORTER)
407 NOTE(REG_USHORTEST);
408
409 /* build compacted NFAs for tree, lacons, fast search */
410 #ifdef REG_DEBUG
411 if (debug != NULL)
412 fprintf(debug, "\n\n\n========= SEARCH ==========\n");
413 #endif
414 /* can sacrifice main NFA now, so use it as work area */
415 (DISCARD) optimize(v->nfa, debug);
416 CNOERR();
417 makesearch(v, v->nfa);
418 CNOERR();
419 compact(v->nfa, &g->search);
420 CNOERR();
421
422 /* looks okay, package it up */
423 re->re_nsub = v->nsubexp;
424 v->re = NULL; /* freev no longer frees re */
425 g->magic = GUTSMAGIC;
426 g->cflags = v->cflags;
427 g->info = re->re_info;
428 g->nsub = re->re_nsub;
429 g->tree = v->tree;
430 v->tree = NULL;
431 g->ntree = v->ntree;
432 g->compare = (v->cflags & REG_ICASE) ? casecmp : cmp;
433 g->lacons = v->lacons;
434 v->lacons = NULL;
435 g->nlacons = v->nlacons;
436
437 #ifdef REG_DEBUG
438 if (flags & REG_DUMP)
439 dump(re, stdout);
440 #endif
441
442 assert(v->err == 0);
443 return freev(v, 0);
444 }
445
446 /*
447 * moresubs - enlarge subRE vector
448 */
449 static void
450 moresubs(struct vars * v,
451 int wanted) /* want enough room for this one */
452 {
453 struct subre **p;
454 size_t n;
455
456 assert(wanted > 0 && (size_t) wanted >= v->nsubs);
457 n = (size_t) wanted *3 / 2 + 1;
458
459 if (v->subs == v->sub10)
460 {
461 p = (struct subre **) MALLOC(n * sizeof(struct subre *));
462 if (p != NULL)
463 memcpy(VS(p), VS(v->subs),
464 v->nsubs * sizeof(struct subre *));
465 }
466 else
467 p = (struct subre **) REALLOC(v->subs, n * sizeof(struct subre *));
468 if (p == NULL)
469 {
470 ERR(REG_ESPACE);
471 return;
472 }
473 v->subs = p;
474 for (p = &v->subs[v->nsubs]; v->nsubs < n; p++, v->nsubs++)
475 *p = NULL;
476 assert(v->nsubs == n);
477 assert((size_t) wanted < v->nsubs);
478 }
479
480 /*
481 * freev - free vars struct's substructures where necessary
482 *
483 * Optionally does error-number setting, and always returns error code
484 * (if any), to make error-handling code terser.
485 */
486 static int
487 freev(struct vars * v,
488 int err)
489 {
490 if (v->re != NULL)
491 rfree(v->re);
492 if (v->subs != v->sub10)
493 FREE(v->subs);
494 if (v->nfa != NULL)
495 freenfa(v->nfa);
496 if (v->tree != NULL)
497 freesubre(v, v->tree);
498 if (v->treechain != NULL)
499 cleanst(v);
500 if (v->cv != NULL)
501 freecvec(v->cv);
502 if (v->cv2 != NULL)
503 freecvec(v->cv2);
504 if (v->lacons != NULL)
505 freelacons(v->lacons, v->nlacons);
506 ERR(err); /* nop if err==0 */
507
508 return v->err;
509 }
510
511 /*
512 * makesearch - turn an NFA into a search NFA (implicit prepend of .*?)
513 * NFA must have been optimize()d already.
514 */
515 static void
516 makesearch(struct vars * v,
517 struct nfa * nfa)
518 {
519 struct arc *a;
520 struct arc *b;
521 struct state *pre = nfa->pre;
522 struct state *s;
523 struct state *s2;
524 struct state *slist;
525
526 /* no loops are needed if it's anchored */
527 for (a = pre->outs; a != NULL; a = a->outchain)
528 {
529 assert(a->type == PLAIN);
530 if (a->co != nfa->bos[0] && a->co != nfa->bos[1])
531 break;
532 }
533 if (a != NULL)
534 {
535 /* add implicit .* in front */
536 rainbow(nfa, v->cm, PLAIN, COLORLESS, pre, pre);
537
538 /* and ^* and \A* too -- not always necessary, but harmless */
539 newarc(nfa, PLAIN, nfa->bos[0], pre, pre);
540 newarc(nfa, PLAIN, nfa->bos[1], pre, pre);
541 }
542
543 /*
544 * Now here's the subtle part. Because many REs have no lookback
545 * constraints, often knowing when you were in the pre state tells you
546 * little; it's the next state(s) that are informative. But some of them
547 * may have other inarcs, i.e. it may be possible to make actual progress
548 * and then return to one of them. We must de-optimize such cases,
549 * splitting each such state into progress and no-progress states.
550 */
551
552 /* first, make a list of the states */
553 slist = NULL;
554 for (a = pre->outs; a != NULL; a = a->outchain)
555 {
556 s = a->to;
557 for (b = s->ins; b != NULL; b = b->inchain)
558 if (b->from != pre)
559 break;
560 if (b != NULL && s->tmp == NULL)
561 {
562 /*
563 * Must be split if not already in the list (fixes bugs 505048,
564 * 230589, 840258, 504785).
565 */
566 s->tmp = slist;
567 slist = s;
568 }
569 }
570
571 /* do the splits */
572 for (s = slist; s != NULL; s = s2)
573 {
574 s2 = newstate(nfa);
575 copyouts(nfa, s, s2);
576 for (a = s->ins; a != NULL; a = b)
577 {
578 b = a->inchain;
579 if (a->from != pre)
580 {
581 cparc(nfa, a, a->from, s2);
582 freearc(nfa, a);
583 }
584 }
585 s2 = s->tmp;
586 s->tmp = NULL; /* clean up while we're at it */
587 }
588 }
589
590 /*
591 * parse - parse an RE
592 *
593 * This is actually just the top level, which parses a bunch of branches
594 * tied together with '|'. They appear in the tree as the left children
595 * of a chain of '|' subres.
596 */
597 static struct subre *
598 parse(struct vars * v,
599 int stopper, /* EOS or ')' */
600 int type, /* LACON (lookahead subRE) or PLAIN */
601 struct state * init, /* initial state */
602 struct state * final) /* final state */
603 {
604 struct state *left; /* scaffolding for branch */
605 struct state *right;
606 struct subre *branches; /* top level */
607 struct subre *branch; /* current branch */
608 struct subre *t; /* temporary */
609 int firstbranch; /* is this the first branch? */
610
611 assert(stopper == ')' || stopper == EOS);
612
613 branches = subre(v, '|', LONGER, init, final);
614 NOERRN();
615 branch = branches;
616 firstbranch = 1;
617 do
618 { /* a branch */
619 if (!firstbranch)
620 {
621 /* need a place to hang it */
622 branch->right = subre(v, '|', LONGER, init, final);
623 NOERRN();
624 branch = branch->right;
625 }
626 firstbranch = 0;
627 left = newstate(v->nfa);
628 right = newstate(v->nfa);
629 NOERRN();
630 EMPTYARC(init, left);
631 EMPTYARC(right, final);
632 NOERRN();
633 branch->left = parsebranch(v, stopper, type, left, right, 0);
634 NOERRN();
635 branch->flags |= UP(branch->flags | branch->left->flags);
636 if ((branch->flags & ~branches->flags) != 0) /* new flags */
637 for (t = branches; t != branch; t = t->right)
638 t->flags |= branch->flags;
639 } while (EAT('|'));
640 assert(SEE(stopper) || SEE(EOS));
641
642 if (!SEE(stopper))
643 {
644 assert(stopper == ')' && SEE(EOS));
645 ERR(REG_EPAREN);
646 }
647
648 /* optimize out simple cases */
649 if (branch == branches)
650 { /* only one branch */
651 assert(branch->right == NULL);
652 t = branch->left;
653 branch->left = NULL;
654 freesubre(v, branches);
655 branches = t;
656 }
657 else if (!MESSY(branches->flags))
658 { /* no interesting innards */
659 freesubre(v, branches->left);
660 branches->left = NULL;
661 freesubre(v, branches->right);
662 branches->right = NULL;
663 branches->op = '=';
664 }
665
666 return branches;
667 }
668
669 /*
670 * parsebranch - parse one branch of an RE
671 *
672 * This mostly manages concatenation, working closely with parseqatom().
673 * Concatenated things are bundled up as much as possible, with separate
674 * ',' nodes introduced only when necessary due to substructure.
675 */
676 static struct subre *
677 parsebranch(struct vars * v,
678 int stopper, /* EOS or ')' */
679 int type, /* LACON (lookahead subRE) or PLAIN */
680 struct state * left, /* leftmost state */
681 struct state * right, /* rightmost state */
682 int partial) /* is this only part of a branch? */
683 {
684 struct state *lp; /* left end of current construct */
685 int seencontent; /* is there anything in this branch yet? */
686 struct subre *t;
687
688 lp = left;
689 seencontent = 0;
690 t = subre(v, '=', 0, left, right); /* op '=' is tentative */
691 NOERRN();
692 while (!SEE('|') && !SEE(stopper) && !SEE(EOS))
693 {
694 if (seencontent)
695 { /* implicit concat operator */
696 lp = newstate(v->nfa);
697 NOERRN();
698 moveins(v->nfa, right, lp);
699 }
700 seencontent = 1;
701
702 /* NB, recursion in parseqatom() may swallow rest of branch */
703 parseqatom(v, stopper, type, lp, right, t);
704 }
705
706 if (!seencontent)
707 { /* empty branch */
708 if (!partial)
709 NOTE(REG_UUNSPEC);
710 assert(lp == left);
711 EMPTYARC(left, right);
712 }
713
714 return t;
715 }
716
717 /*
718 * parseqatom - parse one quantified atom or constraint of an RE
719 *
720 * The bookkeeping near the end cooperates very closely with parsebranch();
721 * in particular, it contains a recursion that can involve parsing the rest
722 * of the branch, making this function's name somewhat inaccurate.
723 */
724 static void
725 parseqatom(struct vars * v,
726 int stopper, /* EOS or ')' */
727 int type, /* LACON (lookahead subRE) or PLAIN */
728 struct state * lp, /* left state to hang it on */
729 struct state * rp, /* right state to hang it on */
730 struct subre * top) /* subtree top */
731 {
732 struct state *s; /* temporaries for new states */
733 struct state *s2;
734
735 #define ARCV(t, val) newarc(v->nfa, t, val, lp, rp)
736 int m,
737 n;
738 struct subre *atom; /* atom's subtree */
739 struct subre *t;
740 int cap; /* capturing parens? */
741 int pos; /* positive lookahead? */
742 int subno; /* capturing-parens or backref number */
743 int atomtype;
744 int qprefer; /* quantifier short/long preference */
745 int f;
746 struct subre **atomp; /* where the pointer to atom is */
747
748 /* initial bookkeeping */
749 atom = NULL;
750 assert(lp->nouts == 0); /* must string new code */
751 assert(rp->nins == 0); /* between lp and rp */
752 subno = 0; /* just to shut lint up */
753
754 /* an atom or constraint... */
755 atomtype = v->nexttype;
756 switch (atomtype)
757 {
758 /* first, constraints, which end by returning */
759 case '^':
760 ARCV('^', 1);
761 if (v->cflags & REG_NLANCH)
762 ARCV(BEHIND, v->nlcolor);
763 NEXT();
764 return;
765 break;
766 case '$':
767 ARCV('$', 1);
768 if (v->cflags & REG_NLANCH)
769 ARCV(AHEAD, v->nlcolor);
770 NEXT();
771 return;
772 break;
773 case SBEGIN:
774 ARCV('^', 1); /* BOL */
775 ARCV('^', 0); /* or BOS */
776 NEXT();
777 return;
778 break;
779 case SEND:
780 ARCV('$', 1); /* EOL */
781 ARCV('$', 0); /* or EOS */
782 NEXT();
783 return;
784 break;
785 case '<':
786 wordchrs(v); /* does NEXT() */
787 s = newstate(v->nfa);
788 NOERR();
789 nonword(v, BEHIND, lp, s);
790 word(v, AHEAD, s, rp);
791 return;
792 break;
793 case '>':
794 wordchrs(v); /* does NEXT() */
795 s = newstate(v->nfa);
796 NOERR();
797 word(v, BEHIND, lp, s);
798 nonword(v, AHEAD, s, rp);
799 return;
800 break;
801 case WBDRY:
802 wordchrs(v); /* does NEXT() */
803 s = newstate(v->nfa);
804 NOERR();
805 nonword(v, BEHIND, lp, s);
806 word(v, AHEAD, s, rp);
807 s = newstate(v->nfa);
808 NOERR();
809 word(v, BEHIND, lp, s);
810 nonword(v, AHEAD, s, rp);
811 return;
812 break;
813 case NWBDRY:
814 wordchrs(v); /* does NEXT() */
815 s = newstate(v->nfa);
816 NOERR();
817 word(v, BEHIND, lp, s);
818 word(v, AHEAD, s, rp);
819 s = newstate(v->nfa);
820 NOERR();
821 nonword(v, BEHIND, lp, s);
822 nonword(v, AHEAD, s, rp);
823 return;
824 break;
825 case LACON: /* lookahead constraint */
826 pos = v->nextvalue;
827 NEXT();
828 s = newstate(v->nfa);
829 s2 = newstate(v->nfa);
830 NOERR();
831 t = parse(v, ')', LACON, s, s2);
832 freesubre(v, t); /* internal structure irrelevant */
833 assert(SEE(')') || ISERR());
834 NEXT();
835 n = newlacon(v, s, s2, pos);
836 NOERR();
837 ARCV(LACON, n);
838 return;
839 break;
840 /* then errors, to get them out of the way */
841 case '*':
842 case '+':
843 case '?':
844 case '{':
845 ERR(REG_BADRPT);
846 return;
847 break;
848 default:
849 ERR(REG_ASSERT);
850 return;
851 break;
852 /* then plain characters, and minor variants on that theme */
853 case ')': /* unbalanced paren */
854 if ((v->cflags & REG_ADVANCED) != REG_EXTENDED)
855 {
856 ERR(REG_EPAREN);
857 return;
858 }
859 /* legal in EREs due to specification botch */
860 NOTE(REG_UPBOTCH);
861 /* fallthrough into case PLAIN */
862 case PLAIN:
863 onechr(v, v->nextvalue, lp, rp);
864 okcolors(v->nfa, v->cm);
865 NOERR();
866 NEXT();
867 break;
868 case '[':
869 if (v->nextvalue == 1)
870 bracket(v, lp, rp);
871 else
872 cbracket(v, lp, rp);
873 assert(SEE(']') || ISERR());
874 NEXT();
875 break;
876 case '.':
877 rainbow(v->nfa, v->cm, PLAIN,
878 (v->cflags & REG_NLSTOP) ? v->nlcolor : COLORLESS,
879 lp, rp);
880 NEXT();
881 break;
882 /* and finally the ugly stuff */
883 case '(': /* value flags as capturing or non */
884 cap = (type == LACON) ? 0 : v->nextvalue;
885 if (cap)
886 {
887 v->nsubexp++;
888 subno = v->nsubexp;
889 if ((size_t) subno >= v->nsubs)
890 moresubs(v, subno);
891 assert((size_t) subno < v->nsubs);
892 }
893 else
894 atomtype = PLAIN; /* something that's not '(' */
895 NEXT();
896 /* need new endpoints because tree will contain pointers */
897 s = newstate(v->nfa);
898 s2 = newstate(v->nfa);
899 NOERR();
900 EMPTYARC(lp, s);
901 EMPTYARC(s2, rp);
902 NOERR();
903 atom = parse(v, ')', PLAIN, s, s2);
904 assert(SEE(')') || ISERR());
905 NEXT();
906 NOERR();
907 if (cap)
908 {
909 v->subs[subno] = atom;
910 t = subre(v, '(', atom->flags | CAP, lp, rp);
911 NOERR();
912 t->subno = subno;
913 t->left = atom;
914 atom = t;
915 }
916 /* postpone everything else pending possible {0} */
917 break;
918 case BACKREF: /* the Feature From The Black Lagoon */
919 INSIST(type != LACON, REG_ESUBREG);
920 INSIST(v->nextvalue < v->nsubs, REG_ESUBREG);
921 INSIST(v->subs[v->nextvalue] != NULL, REG_ESUBREG);
922 NOERR();
923 assert(v->nextvalue > 0);
924 atom = subre(v, 'b', BACKR, lp, rp);
925 subno = v->nextvalue;
926 atom->subno = subno;
927 EMPTYARC(lp, rp); /* temporarily, so there's something */
928 NEXT();
929 break;
930 }
931
932 /* ...and an atom may be followed by a quantifier */
933 switch (v->nexttype)
934 {
935 case '*':
936 m = 0;
937 n = INFINITY;
938 qprefer = (v->nextvalue) ? LONGER : SHORTER;
939 NEXT();
940 break;
941 case '+':
942 m = 1;
943 n = INFINITY;
944 qprefer = (v->nextvalue) ? LONGER : SHORTER;
945 NEXT();
946 break;
947 case '?':
948 m = 0;
949 n = 1;
950 qprefer = (v->nextvalue) ? LONGER : SHORTER;
951 NEXT();
952 break;
953 case '{':
954 NEXT();
955 m = scannum(v);
956 if (EAT(','))
957 {
958 if (SEE(DIGIT))
959 n = scannum(v);
960 else
961 n = INFINITY;
962 if (m > n)
963 {
964 ERR(REG_BADBR);
965 return;
966 }
967 /* {m,n} exercises preference, even if it's {m,m} */
968 qprefer = (v->nextvalue) ? LONGER : SHORTER;
969 }
970 else
971 {
972 n = m;
973 /* {m} passes operand's preference through */
974 qprefer = 0;
975 }
976 if (!SEE('}'))
977 { /* catches errors too */
978 ERR(REG_BADBR);
979 return;
980 }
981 NEXT();
982 break;
983 default: /* no quantifier */
984 m = n = 1;
985 qprefer = 0;
986 break;
987 }
988
989 /* annoying special case: {0} or {0,0} cancels everything */
990 if (m == 0 && n == 0)
991 {
992 if (atom != NULL)
993 freesubre(v, atom);
994 if (atomtype == '(')
995 v->subs[subno] = NULL;
996 delsub(v->nfa, lp, rp);
997 EMPTYARC(lp, rp);
998 return;
999 }
1000
1001 /* if not a messy case, avoid hard part */
1002 assert(!MESSY(top->flags));
1003 f = top->flags | qprefer | ((atom != NULL) ? atom->flags : 0);
1004 if (atomtype != '(' && atomtype != BACKREF && !MESSY(UP(f)))
1005 {
1006 if (!(m == 1 && n == 1))
1007 repeat(v, lp, rp, m, n);
1008 if (atom != NULL)
1009 freesubre(v, atom);
1010 top->flags = f;
1011 return;
1012 }
1013
1014 /*
1015 * hard part: something messy That is, capturing parens, back reference,
1016 * short/long clash, or an atom with substructure containing one of those.
1017 */
1018
1019 /* now we'll need a subre for the contents even if they're boring */
1020 if (atom == NULL)
1021 {
1022 atom = subre(v, '=', 0, lp, rp);
1023 NOERR();
1024 }
1025
1026 /*
1027 * prepare a general-purpose state skeleton
1028 *
1029 * ---> [s] ---prefix---> [begin] ---atom---> [end] ----rest---> [rp] / /
1030 * [lp] ----> [s2] ----bypass---------------------
1031 *
1032 * where bypass is an empty, and prefix is some repetitions of atom
1033 */
1034 s = newstate(v->nfa); /* first, new endpoints for the atom */
1035 s2 = newstate(v->nfa);
1036 NOERR();
1037 moveouts(v->nfa, lp, s);
1038 moveins(v->nfa, rp, s2);
1039 NOERR();
1040 atom->begin = s;
1041 atom->end = s2;
1042 s = newstate(v->nfa); /* and spots for prefix and bypass */
1043 s2 = newstate(v->nfa);
1044 NOERR();
1045 EMPTYARC(lp, s);
1046 EMPTYARC(lp, s2);
1047 NOERR();
1048
1049 /* break remaining subRE into x{...} and what follows */
1050 t = subre(v, '.', COMBINE(qprefer, atom->flags), lp, rp);
1051 t->left = atom;
1052 atomp = &t->left;
1053 /* here we should recurse... but we must postpone that to the end */
1054
1055 /* split top into prefix and remaining */
1056 assert(top->op == '=' && top->left == NULL && top->right == NULL);
1057 top->left = subre(v, '=', top->flags, top->begin, lp);
1058 top->op = '.';
1059 top->right = t;
1060
1061 /* if it's a backref, now is the time to replicate the subNFA */
1062 if (atomtype == BACKREF)
1063 {
1064 assert(atom->begin->nouts == 1); /* just the EMPTY */
1065 delsub(v->nfa, atom->begin, atom->end);
1066 assert(v->subs[subno] != NULL);
1067 /* and here's why the recursion got postponed: it must */
1068 /* wait until the skeleton is filled in, because it may */
1069 /* hit a backref that wants to copy the filled-in skeleton */
1070 dupnfa(v->nfa, v->subs[subno]->begin, v->subs[subno]->end,
1071 atom->begin, atom->end);
1072 NOERR();
1073 }
1074
1075 /* it's quantifier time; first, turn x{0,...} into x{1,...}|empty */
1076 if (m == 0)
1077 {
1078 EMPTYARC(s2, atom->end); /* the bypass */
1079 assert(PREF(qprefer) != 0);
1080 f = COMBINE(qprefer, atom->flags);
1081 t = subre(v, '|', f, lp, atom->end);
1082 NOERR();
1083 t->left = atom;
1084 t->right = subre(v, '|', PREF(f), s2, atom->end);
1085 NOERR();
1086 t->right->left = subre(v, '=', 0, s2, atom->end);
1087 NOERR();
1088 *atomp = t;
1089 atomp = &t->left;
1090 m = 1;
1091 }
1092
1093 /* deal with the rest of the quantifier */
1094 if (atomtype == BACKREF)
1095 {
1096 /* special case: backrefs have internal quantifiers */
1097 EMPTYARC(s, atom->begin); /* empty prefix */
1098 /* just stuff everything into atom */
1099 repeat(v, atom->begin, atom->end, m, n);
1100 atom->min = (short) m;
1101 atom->max = (short) n;
1102 atom->flags |= COMBINE(qprefer, atom->flags);
1103 }
1104 else if (m == 1 && n == 1)
1105 {
1106 /* no/vacuous quantifier: done */
1107 EMPTYARC(s, atom->begin); /* empty prefix */
1108 }
1109 else
1110 {
1111 /* turn x{m,n} into x{m-1,n-1}x, with capturing */
1112 /* parens in only second x */
1113 dupnfa(v->nfa, atom->begin, atom->end, s, atom->begin);
1114 assert(m >= 1 && m != INFINITY && n >= 1);
1115 repeat(v, s, atom->begin, m - 1, (n == INFINITY) ? n : n - 1);
1116 f = COMBINE(qprefer, atom->flags);
1117 t = subre(v, '.', f, s, atom->end); /* prefix and atom */
1118 NOERR();
1119 t->left = subre(v, '=', PREF(f), s, atom->begin);
1120 NOERR();
1121 t->right = atom;
1122 *atomp = t;
1123 }
1124
1125 /* and finally, look after that postponed recursion */
1126 t = top->right;
1127 if (!(SEE('|') || SEE(stopper) || SEE(EOS)))
1128 t->right = parsebranch(v, stopper, type, atom->end, rp, 1);
1129 else
1130 {
1131 EMPTYARC(atom->end, rp);
1132 t->right = subre(v, '=', 0, atom->end, rp);
1133 }
1134 assert(SEE('|') || SEE(stopper) || SEE(EOS));
1135 t->flags |= COMBINE(t->flags, t->right->flags);
1136 top->flags |= COMBINE(top->flags, t->flags);
1137 }
1138
1139 /*
1140 * nonword - generate arcs for non-word-character ahead or behind
1141 */
1142 static void
1143 nonword(struct vars * v,
1144 int dir, /* AHEAD or BEHIND */
1145 struct state * lp,
1146 struct state * rp)
1147 {
1148 int anchor = (dir == AHEAD) ? '$' : '^';
1149
1150 assert(dir == AHEAD || dir == BEHIND);
1151 newarc(v->nfa, anchor, 1, lp, rp);
1152 newarc(v->nfa, anchor, 0, lp, rp);
1153 colorcomplement(v->nfa, v->cm, dir, v->wordchrs, lp, rp);
1154 /* (no need for special attention to \n) */
1155 }
1156
1157 /*
1158 * word - generate arcs for word character ahead or behind
1159 */
1160 static void
1161 word(struct vars * v,
1162 int dir, /* AHEAD or BEHIND */
1163 struct state * lp,
1164 struct state * rp)
1165 {
1166 assert(dir == AHEAD || dir == BEHIND);
1167 cloneouts(v->nfa, v->wordchrs, lp, rp, dir);
1168 /* (no need for special attention to \n) */
1169 }
1170
1171 /*
1172 * scannum - scan a number
1173 */
1174 static int /* value, <= DUPMAX */
1175 scannum(struct vars * v)
1176 {
1177 int n = 0;
1178
1179 while (SEE(DIGIT) && n < DUPMAX)
1180 {
1181 n = n * 10 + v->nextvalue;
1182 NEXT();
1183 }
1184 if (SEE(DIGIT) || n > DUPMAX)
1185 {
1186 ERR(REG_BADBR);
1187 return 0;
1188 }
1189 return n;
1190 }
1191
1192 /*
1193 * repeat - replicate subNFA for quantifiers
1194 *
1195 * The duplication sequences used here are chosen carefully so that any
1196 * pointers starting out pointing into the subexpression end up pointing into
1197 * the last occurrence. (Note that it may not be strung between the same
1198 * left and right end states, however!) This used to be important for the
1199 * subRE tree, although the important bits are now handled by the in-line
1200 * code in parse(), and when this is called, it doesn't matter any more.
1201 */
1202 static void
1203 repeat(struct vars * v,
1204 struct state * lp,
1205 struct state * rp,
1206 int m,
1207 int n)
1208 {
1209 #define SOME 2
1210 #define INF 3
1211 #define PAIR(x, y) ((x)*4 + (y))
1212 #define REDUCE(x) ( ((x) == INFINITY) ? INF : (((x) > 1) ? SOME : (x)) )
1213 const int rm = REDUCE(m);
1214 const int rn = REDUCE(n);
1215 struct state *s;
1216 struct state *s2;
1217
1218 switch (PAIR(rm, rn))
1219 {
1220 case PAIR(0, 0): /* empty string */
1221 delsub(v->nfa, lp, rp);
1222 EMPTYARC(lp, rp);
1223 break;
1224 case PAIR(0, 1): /* do as x| */
1225 EMPTYARC(lp, rp);
1226 break;
1227 case PAIR(0, SOME): /* do as x{1,n}| */
1228 repeat(v, lp, rp, 1, n);
1229 NOERR();
1230 EMPTYARC(lp, rp);
1231 break;
1232 case PAIR(0, INF): /* loop x around */
1233 s = newstate(v->nfa);
1234 NOERR();
1235 moveouts(v->nfa, lp, s);
1236 moveins(v->nfa, rp, s);
1237 EMPTYARC(lp, s);
1238 EMPTYARC(s, rp);
1239 break;
1240 case PAIR(1, 1): /* no action required */
1241 break;
1242 case PAIR(1, SOME): /* do as x{0,n-1}x = (x{1,n-1}|)x */
1243 s = newstate(v->nfa);
1244 NOERR();
1245 moveouts(v->nfa, lp, s);
1246 dupnfa(v->nfa, s, rp, lp, s);
1247 NOERR();
1248 repeat(v, lp, s, 1, n - 1);
1249 NOERR();
1250 EMPTYARC(lp, s);
1251 break;
1252 case PAIR(1, INF): /* add loopback arc */
1253 s = newstate(v->nfa);
1254 s2 = newstate(v->nfa);
1255 NOERR();
1256 moveouts(v->nfa, lp, s);
1257 moveins(v->nfa, rp, s2);
1258 EMPTYARC(lp, s);
1259 EMPTYARC(s2, rp);
1260 EMPTYARC(s2, s);
1261 break;
1262 case PAIR(SOME, SOME): /* do as x{m-1,n-1}x */
1263 s = newstate(v->nfa);
1264 NOERR();
1265 moveouts(v->nfa, lp, s);
1266 dupnfa(v->nfa, s, rp, lp, s);
1267 NOERR();
1268 repeat(v, lp, s, m - 1, n - 1);
1269 break;
1270 case PAIR(SOME, INF): /* do as x{m-1,}x */
1271 s = newstate(v->nfa);
1272 NOERR();
1273 moveouts(v->nfa, lp, s);
1274 dupnfa(v->nfa, s, rp, lp, s);
1275 NOERR();
1276 repeat(v, lp, s, m - 1, n);
1277 break;
1278 default:
1279 ERR(REG_ASSERT);
1280 break;
1281 }
1282 }
1283
1284 /*
1285 * bracket - handle non-complemented bracket expression
1286 * Also called from cbracket for complemented bracket expressions.
1287 */
1288 static void
1289 bracket(struct vars * v,
1290 struct state * lp,
1291 struct state * rp)
1292 {
1293 assert(SEE('['));
1294 NEXT();
1295 while (!SEE(']') && !SEE(EOS))
1296 brackpart(v, lp, rp);
1297 assert(SEE(']') || ISERR());
1298 okcolors(v->nfa, v->cm);
1299 }
1300
1301 /*
1302 * cbracket - handle complemented bracket expression
1303 * We do it by calling bracket() with dummy endpoints, and then complementing
1304 * the result. The alternative would be to invoke rainbow(), and then delete
1305 * arcs as the b.e. is seen... but that gets messy.
1306 */
1307 static void
1308 cbracket(struct vars * v,
1309 struct state * lp,
1310 struct state * rp)
1311 {
1312 struct state *left = newstate(v->nfa);
1313 struct state *right = newstate(v->nfa);
1314
1315 NOERR();
1316 bracket(v, left, right);
1317 if (v->cflags & REG_NLSTOP)
1318 newarc(v->nfa, PLAIN, v->nlcolor, left, right);
1319 NOERR();
1320
1321 assert(lp->nouts == 0); /* all outarcs will be ours */
1322
1323 /*
1324 * Easy part of complementing, and all there is to do since the MCCE code
1325 * was removed.
1326 */
1327 colorcomplement(v->nfa, v->cm, PLAIN, left, lp, rp);
1328 NOERR();
1329 dropstate(v->nfa, left);
1330 assert(right->nins == 0);
1331 freestate(v->nfa, right);
1332 }
1333
1334 /*
1335 * brackpart - handle one item (or range) within a bracket expression
1336 */
1337 static void
1338 brackpart(struct vars * v,
1339 struct state * lp,
1340 struct state * rp)
1341 {
1342 celt startc;
1343 celt endc;
1344 struct cvec *cv;
1345 const chr *startp;
1346 const chr *endp;
1347 chr c[1];
1348
1349 /* parse something, get rid of special cases, take shortcuts */
1350 switch (v->nexttype)
1351 {
1352 case RANGE: /* a-b-c or other botch */
1353 ERR(REG_ERANGE);
1354 return;
1355 break;
1356 case PLAIN:
1357 c[0] = v->nextvalue;
1358 NEXT();
1359 /* shortcut for ordinary chr (not range) */
1360 if (!SEE(RANGE))
1361 {
1362 onechr(v, c[0], lp, rp);
1363 return;
1364 }
1365 startc = element(v, c, c + 1);
1366 NOERR();
1367 break;
1368 case COLLEL:
1369 startp = v->now;
1370 endp = scanplain(v);
1371 INSIST(startp < endp, REG_ECOLLATE);
1372 NOERR();
1373 startc = element(v, startp, endp);
1374 NOERR();
1375 break;
1376 case ECLASS:
1377 startp = v->now;
1378 endp = scanplain(v);
1379 INSIST(startp < endp, REG_ECOLLATE);
1380 NOERR();
1381 startc = element(v, startp, endp);
1382 NOERR();
1383 cv = eclass(v, startc, (v->cflags & REG_ICASE));
1384 NOERR();
1385 dovec(v, cv, lp, rp);
1386 return;
1387 break;
1388 case CCLASS:
1389 startp = v->now;
1390 endp = scanplain(v);
1391 INSIST(startp < endp, REG_ECTYPE);
1392 NOERR();
1393 cv = cclass(v, startp, endp, (v->cflags & REG_ICASE));
1394 NOERR();
1395 dovec(v, cv, lp, rp);
1396 return;
1397 break;
1398 default:
1399 ERR(REG_ASSERT);
1400 return;
1401 break;
1402 }
1403
1404 if (SEE(RANGE))
1405 {
1406 NEXT();
1407 switch (v->nexttype)
1408 {
1409 case PLAIN:
1410 case RANGE:
1411 c[0] = v->nextvalue;
1412 NEXT();
1413 endc = element(v, c, c + 1);
1414 NOERR();
1415 break;
1416 case COLLEL:
1417 startp = v->now;
1418 endp = scanplain(v);
1419 INSIST(startp < endp, REG_ECOLLATE);
1420 NOERR();
1421 endc = element(v, startp, endp);
1422 NOERR();
1423 break;
1424 default:
1425 ERR(REG_ERANGE);
1426 return;
1427 break;
1428 }
1429 }
1430 else
1431 endc = startc;
1432
1433 /*
1434 * Ranges are unportable. Actually, standard C does guarantee that digits
1435 * are contiguous, but making that an exception is just too complicated.
1436 */
1437 if (startc != endc)
1438 NOTE(REG_UUNPORT);
1439 cv = range(v, startc, endc, (v->cflags & REG_ICASE));
1440 NOERR();
1441 dovec(v, cv, lp, rp);
1442 }
1443
1444 /*
1445 * scanplain - scan PLAIN contents of [. etc.
1446 *
1447 * Certain bits of trickery in lex.c know that this code does not try
1448 * to look past the final bracket of the [. etc.
1449 */
1450 static const chr * /* just after end of sequence */
1451 scanplain(struct vars * v)
1452 {
1453 const chr *endp;
1454
1455 assert(SEE(COLLEL) || SEE(ECLASS) || SEE(CCLASS));
1456 NEXT();
1457
1458 endp = v->now;
1459 while (SEE(PLAIN))
1460 {
1461 endp = v->now;
1462 NEXT();
1463 }
1464
1465 assert(SEE(END) || ISERR());
1466 NEXT();
1467
1468 return endp;
1469 }
1470
1471 /*
1472 * onechr - fill in arcs for a plain character, and possible case complements
1473 * This is mostly a shortcut for efficient handling of the common case.
1474 */
1475 static void
1476 onechr(struct vars * v,
1477 chr c,
1478 struct state * lp,
1479 struct state * rp)
1480 {
1481 if (!(v->cflags & REG_ICASE))
1482 {
1483 newarc(v->nfa, PLAIN, subcolor(v->cm, c), lp, rp);
1484 return;
1485 }
1486
1487 /* rats, need general case anyway... */
1488 dovec(v, allcases(v, c), lp, rp);
1489 }
1490
1491 /*
1492 * dovec - fill in arcs for each element of a cvec
1493 */
1494 static void
1495 dovec(struct vars * v,
1496 struct cvec * cv,
1497 struct state * lp,
1498 struct state * rp)
1499 {
1500 chr ch,
1501 from,
1502 to;
1503 const chr *p;
1504 int i;
1505
1506 /* ordinary characters */
1507 for (p = cv->chrs, i = cv->nchrs; i > 0; p++, i--)
1508 {
1509 ch = *p;
1510 newarc(v->nfa, PLAIN, subcolor(v->cm, ch), lp, rp);
1511 }
1512
1513 /* and the ranges */
1514 for (p = cv->ranges, i = cv->nranges; i > 0; p += 2, i--)
1515 {
1516 from = *p;
1517 to = *(p + 1);
1518 if (from <= to)
1519 subrange(v, from, to, lp, rp);
1520 }
1521 }
1522
1523 /*
1524 * wordchrs - set up word-chr list for word-boundary stuff, if needed
1525 *
1526 * The list is kept as a bunch of arcs between two dummy states; it's
1527 * disposed of by the unreachable-states sweep in NFA optimization.
1528 * Does NEXT(). Must not be called from any unusual lexical context.
1529 * This should be reconciled with the \w etc. handling in lex.c, and
1530 * should be cleaned up to reduce dependencies on input scanning.
1531 */
1532 static void
1533 wordchrs(struct vars * v)
1534 {
1535 struct state *left;
1536 struct state *right;
1537
1538 if (v->wordchrs != NULL)
1539 {
1540 NEXT(); /* for consistency */
1541 return;
1542 }
1543
1544 left = newstate(v->nfa);
1545 right = newstate(v->nfa);
1546 NOERR();
1547 /* fine point: implemented with [::], and lexer will set REG_ULOCALE */
1548 lexword(v);
1549 NEXT();
1550 assert(v->savenow != NULL && SEE('['));
1551 bracket(v, left, right);
1552 assert((v->savenow != NULL && SEE(']')) || ISERR());
1553 NEXT();
1554 NOERR();
1555 v->wordchrs = left;
1556 }
1557
1558 /*
1559 * subre - allocate a subre
1560 */
1561 static struct subre *
1562 subre(struct vars * v,
1563 int op,
1564 int flags,
1565 struct state * begin,
1566 struct state * end)
1567 {
1568 struct subre *ret = v->treefree;
1569
1570 if (ret != NULL)
1571 v->treefree = ret->left;
1572 else
1573 {
1574 ret = (struct subre *) MALLOC(sizeof(struct subre));
1575 if (ret == NULL)
1576 {
1577 ERR(REG_ESPACE);
1578 return NULL;
1579 }
1580 ret->chain = v->treechain;
1581 v->treechain = ret;
1582 }
1583
1584 assert(strchr("|.b(=", op) != NULL);
1585
1586 ret->op = op;
1587 ret->flags = flags;
1588 ret->retry = 0;
1589 ret->subno = 0;
1590 ret->min = ret->max = 1;
1591 ret->left = NULL;
1592 ret->right = NULL;
1593 ret->begin = begin;
1594 ret->end = end;
1595 ZAPCNFA(ret->cnfa);
1596
1597 return ret;
1598 }
1599
1600 /*
1601 * freesubre - free a subRE subtree
1602 */
1603 static void
1604 freesubre(struct vars * v, /* might be NULL */
1605 struct subre * sr)
1606 {
1607 if (sr == NULL)
1608 return;
1609
1610 if (sr->left != NULL)
1611 freesubre(v, sr->left);
1612 if (sr->right != NULL)
1613 freesubre(v, sr->right);
1614
1615 freesrnode(v, sr);
1616 }
1617
1618 /*
1619 * freesrnode - free one node in a subRE subtree
1620 */
1621 static void
1622 freesrnode(struct vars * v, /* might be NULL */
1623 struct subre * sr)
1624 {
1625 if (sr == NULL)
1626 return;
1627
1628 if (!NULLCNFA(sr->cnfa))
1629 freecnfa(&sr->cnfa);
1630 sr->flags = 0;
1631
1632 if (v != NULL)
1633 {
1634 sr->left = v->treefree;
1635 v->treefree = sr;
1636 }
1637 else
1638 FREE(sr);
1639 }
1640
1641 /*
1642 * optst - optimize a subRE subtree
1643 */
1644 static void
1645 optst(struct vars * v,
1646 struct subre * t)
1647 {
1648 /*
1649 * DGP (2007-11-13): I assume it was the programmer's intent to eventually
1650 * come back and add code to optimize subRE trees, but the routine coded
1651 * just spends effort traversing the tree and doing nothing. We can do
1652 * nothing with less effort.
1653 */
1654 return;
1655 }
1656
1657 /*
1658 * numst - number tree nodes (assigning retry indexes)
1659 */
1660 static int /* next number */
1661 numst(struct subre * t,
1662 int start) /* starting point for subtree numbers */
1663 {
1664 int i;
1665
1666 assert(t != NULL);
1667
1668 i = start;
1669 t->retry = (short) i++;
1670 if (t->left != NULL)
1671 i = numst(t->left, i);
1672 if (t->right != NULL)
1673 i = numst(t->right, i);
1674 return i;
1675 }
1676
1677 /*
1678 * markst - mark tree nodes as INUSE
1679 */
1680 static void
1681 markst(struct subre * t)
1682 {
1683 assert(t != NULL);
1684
1685 t->flags |= INUSE;
1686 if (t->left != NULL)
1687 markst(t->left);
1688 if (t->right != NULL)
1689 markst(t->right);
1690 }
1691
1692 /*
1693 * cleanst - free any tree nodes not marked INUSE
1694 */
1695 static void
1696 cleanst(struct vars * v)
1697 {
1698 struct subre *t;
1699 struct subre *next;
1700
1701 for (t = v->treechain; t != NULL; t = next)
1702 {
1703 next = t->chain;
1704 if (!(t->flags & INUSE))
1705 FREE(t);
1706 }
1707 v->treechain = NULL;
1708 v->treefree = NULL; /* just on general principles */
1709 }
1710
1711 /*
1712 * nfatree - turn a subRE subtree into a tree of compacted NFAs
1713 */
1714 static long /* optimize results from top node */
1715 nfatree(struct vars * v,
1716 struct subre * t,
1717 FILE *f) /* for debug output */
1718 {
1719 assert(t != NULL && t->begin != NULL);
1720
1721 if (t->left != NULL)
1722 (DISCARD) nfatree(v, t->left, f);
1723 if (t->right != NULL)
1724 (DISCARD) nfatree(v, t->right, f);
1725
1726 return nfanode(v, t, f);
1727 }
1728
1729 /*
1730 * nfanode - do one NFA for nfatree
1731 */
1732 static long /* optimize results */
1733 nfanode(struct vars * v,
1734 struct subre * t,
1735 FILE *f) /* for debug output */
1736 {
1737 struct nfa *nfa;
1738 long ret = 0;
1739
1740 assert(t->begin != NULL);
1741
1742 #ifdef REG_DEBUG
1743 if (f != NULL)
1744 {
1745 char idbuf[50];
1746
1747 fprintf(f, "\n\n\n========= TREE NODE %s ==========\n",
1748 stid(t, idbuf, sizeof(idbuf)));
1749 }
1750 #endif
1751 nfa = newnfa(v, v->cm, v->nfa);
1752 NOERRZ();
1753 dupnfa(nfa, t->begin, t->end, nfa->init, nfa->final);
1754 if (!ISERR())
1755 {
1756 specialcolors(nfa);
1757 ret = optimize(nfa, f);
1758 }
1759 if (!ISERR())
1760 compact(nfa, &t->cnfa);
1761
1762 freenfa(nfa);
1763 return ret;
1764 }
1765
1766 /*
1767 * newlacon - allocate a lookahead-constraint subRE
1768 */
1769 static int /* lacon number */
1770 newlacon(struct vars * v,
1771 struct state * begin,
1772 struct state * end,
1773 int pos)
1774 {
1775 int n;
1776 struct subre *sub;
1777
1778 if (v->nlacons == 0)
1779 {
1780 v->lacons = (struct subre *) MALLOC(2 * sizeof(struct subre));
1781 n = 1; /* skip 0th */
1782 v->nlacons = 2;
1783 }
1784 else
1785 {
1786 v->lacons = (struct subre *) REALLOC(v->lacons,
1787 (v->nlacons + 1) * sizeof(struct subre));
1788 n = v->nlacons++;
1789 }
1790 if (v->lacons == NULL)
1791 {
1792 ERR(REG_ESPACE);
1793 return 0;
1794 }
1795 sub = &v->lacons[n];
1796 sub->begin = begin;
1797 sub->end = end;
1798 sub->subno = pos;
1799 ZAPCNFA(sub->cnfa);
1800 return n;
1801 }
1802
1803 /*
1804 * freelacons - free lookahead-constraint subRE vector
1805 */
1806 static void
1807 freelacons(struct subre * subs,
1808 int n)
1809 {
1810 struct subre *sub;
1811 int i;
1812
1813 assert(n > 0);
1814 for (sub = subs + 1, i = n - 1; i > 0; sub++, i--) /* no 0th */
1815 if (!NULLCNFA(sub->cnfa))
1816 freecnfa(&sub->cnfa);
1817 FREE(subs);
1818 }
1819
1820 /*
1821 * rfree - free a whole RE (insides of regfree)
1822 */
1823 static void
1824 rfree(regex_t *re)
1825 {
1826 struct guts *g;
1827
1828 if (re == NULL || re->re_magic != REMAGIC)
1829 return;
1830
1831 re->re_magic = 0; /* invalidate RE */
1832 g = (struct guts *) re->re_guts;
1833 re->re_guts = NULL;
1834 re->re_fns = NULL;
1835 g->magic = 0;
1836 freecm(&g->cmap);
1837 if (g->tree != NULL)
1838 freesubre((struct vars *) NULL, g->tree);
1839 if (g->lacons != NULL)
1840 freelacons(g->lacons, g->nlacons);
1841 if (!NULLCNFA(g->search))
1842 freecnfa(&g->search);
1843 FREE(g);
1844 }
1845
1846 #ifdef REG_DEBUG
1847
1848 /*
1849 * dump - dump an RE in human-readable form
1850 */
1851 static void
1852 dump(regex_t *re,
1853 FILE *f)
1854 {
1855 struct guts *g;
1856 int i;
1857
1858 if (re->re_magic != REMAGIC)
1859 fprintf(f, "bad magic number (0x%x not 0x%x)\n", re->re_magic,
1860 REMAGIC);
1861 if (re->re_guts == NULL)
1862 {
1863 fprintf(f, "NULL guts!!!\n");
1864 return;
1865 }
1866 g = (struct guts *) re->re_guts;
1867 if (g->magic != GUTSMAGIC)
1868 fprintf(f, "bad guts magic number (0x%x not 0x%x)\n", g->magic,
1869 GUTSMAGIC);
1870
1871 fprintf(f, "\n\n\n========= DUMP ==========\n");
1872 fprintf(f, "nsub %d, info 0%lo, csize %d, ntree %d\n",
1873 (int) re->re_nsub, re->re_info, re->re_csize, g->ntree);
1874
1875 dumpcolors(&g->cmap, f);
1876 if (!NULLCNFA(g->search))
1877 {
1878 printf("\nsearch:\n");
1879 dumpcnfa(&g->search, f);
1880 }
1881 for (i = 1; i < g->nlacons; i++)
1882 {
1883 fprintf(f, "\nla%d (%s):\n", i,
1884 (g->lacons[i].subno) ? "positive" : "negative");
1885 dumpcnfa(&g->lacons[i].cnfa, f);
1886 }
1887 fprintf(f, "\n");
1888 dumpst(g->tree, f, 0);
1889 }
1890
1891 /*
1892 * dumpst - dump a subRE tree
1893 */
1894 static void
1895 dumpst(struct subre * t,
1896 FILE *f,
1897 int nfapresent) /* is the original NFA still around? */
1898 {
1899 if (t == NULL)
1900 fprintf(f, "null tree\n");
1901 else
1902 stdump(t, f, nfapresent);
1903 fflush(f);
1904 }
1905
1906 /*
1907 * stdump - recursive guts of dumpst
1908 */
1909 static void
1910 stdump(struct subre * t,
1911 FILE *f,
1912 int nfapresent) /* is the original NFA still around? */
1913 {
1914 char idbuf[50];
1915
1916 fprintf(f, "%s. `%c'", stid(t, idbuf, sizeof(idbuf)), t->op);
1917 if (t->flags & LONGER)
1918 fprintf(f, " longest");
1919 if (t->flags & SHORTER)
1920 fprintf(f, " shortest");
1921 if (t->flags & MIXED)
1922 fprintf(f, " hasmixed");
1923 if (t->flags & CAP)
1924 fprintf(f, " hascapture");
1925 if (t->flags & BACKR)
1926 fprintf(f, " hasbackref");
1927 if (!(t->flags & INUSE))
1928 fprintf(f, " UNUSED");
1929 if (t->subno != 0)
1930 fprintf(f, " (#%d)", t->subno);
1931 if (t->min != 1 || t->max != 1)
1932 {
1933 fprintf(f, " {%d,", t->min);
1934 if (t->max != INFINITY)
1935 fprintf(f, "%d", t->max);
1936 fprintf(f, "}");
1937 }
1938 if (nfapresent)
1939 fprintf(f, " %ld-%ld", (long) t->begin->no, (long) t->end->no);
1940 if (t->left != NULL)
1941 fprintf(f, " L:%s", stid(t->left, idbuf, sizeof(idbuf)));
1942 if (t->right != NULL)
1943 fprintf(f, " R:%s", stid(t->right, idbuf, sizeof(idbuf)));
1944 if (!NULLCNFA(t->cnfa))
1945 {
1946 fprintf(f, "\n");
1947 dumpcnfa(&t->cnfa, f);
1948 }
1949 fprintf(f, "\n");
1950 if (t->left != NULL)
1951 stdump(t->left, f, nfapresent);
1952 if (t->right != NULL)
1953 stdump(t->right, f, nfapresent);
1954 }
1955
1956 /*
1957 * stid - identify a subtree node for dumping
1958 */
1959 static const char * /* points to buf or constant string */
1960 stid(struct subre * t,
1961 char *buf,
1962 size_t bufsize)
1963 {
1964 /* big enough for hex int or decimal t->retry? */
1965 if (bufsize < sizeof(void *) * 2 + 3 || bufsize < sizeof(t->retry) * 3 + 1)
1966 return "unable";
1967 if (t->retry != 0)
1968 sprintf(buf, "%d", t->retry);
1969 else
1970 sprintf(buf, "%p", t);
1971 return buf;
1972 }
1973 #endif /* REG_DEBUG */
1974
1975
1976 #include "regc_lex.c"
1977 #include "regc_color.c"
1978 #include "regc_nfa.c"
1979 #include "regc_cvec.c"
1980 #include "regc_locale.c"
PostgreSQL from CVS