projects / wxWidgets.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Cleaned up regex.cpp
[wxWidgets.git] / src / regex / regcomp.c
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 * $Header: /projects/cvsroot/pgsql-server/src/backend/regex/regcomp.c,v 1.38 2003/08/08 21:41:56 momjian Exp $
32 *
33 */
34
35 #include "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 chr *scanplain(struct vars *);
55 static void leaders(struct vars *, struct cvec *);
56 static void onechr(struct vars *, chr, struct state *, struct state *);
57 static void dovec(struct vars *, struct cvec *, struct state *, struct state *);
58 static celt nextleader(struct vars *, chr, chr);
59 static void wordchrs(struct vars *);
60 static struct subre *subre(struct vars *, int, int, struct state *, struct state *);
61 static void freesubre(struct vars *, struct subre *);
62 static void freesrnode(struct vars *, struct subre *);
63 static void optst(struct vars *, struct subre *);
64 static int numst(struct subre *, int);
65 static void markst(struct subre *);
66 static void cleanst(struct vars *);
67 static long nfatree(struct vars *, struct subre *, FILE *);
68 static long nfanode(struct vars *, struct subre *, FILE *);
69 static int newlacon(struct vars *, struct state *, struct state *, int);
70 static void freelacons(struct subre *, int);
71 static void rfree(regex_t *);
72
73 #ifdef REG_DEBUG
74 static void dump(regex_t *, FILE *);
75 static void dumpst(struct subre *, FILE *, int);
76 static void stdump(struct subre *, FILE *, int);
77 static char *stid(struct subre *, char *, size_t);
78 #endif
79 /* === regc_lex.c === */
80 static void lexstart(struct vars *);
81 static void prefixes(struct vars *);
82 static void lexnest(struct vars *, chr *, chr *);
83 static void lexword(struct vars *);
84 static int next(struct vars *);
85 static int lexescape(struct vars *);
86 static chr lexdigits(struct vars *, int, int, int);
87 static int brenext(struct vars *, chr);
88 static void skip(struct vars *);
89 static chr newline(void);
90 static chr chrnamed(struct vars *, chr *, chr *, chr);
91
92 /* === regc_color.c === */
93 static void initcm(struct vars *, struct colormap *);
94 static void freecm(struct colormap *);
95 static void cmtreefree(struct colormap *, union tree *, int);
96 static color setcolor(struct colormap *, chr, pcolor);
97 static color maxcolor(struct colormap *);
98 static color newcolor(struct colormap *);
99 static void freecolor(struct colormap *, pcolor);
100 static color pseudocolor(struct colormap *);
101 static color subcolor(struct colormap *, chr c);
102 static color newsub(struct colormap *, pcolor);
103 static void subrange(struct vars *, chr, chr, struct state *, struct state *);
104 static void subblock(struct vars *, chr, struct state *, struct state *);
105 static void okcolors(struct nfa *, struct colormap *);
106 static void colorchain(struct colormap *, struct arc *);
107 static void uncolorchain(struct colormap *, struct arc *);
108 static int singleton(struct colormap *, chr c);
109 static void rainbow(struct nfa *, struct colormap *, int, pcolor, struct state *, struct state *);
110 static void colorcomplement(struct nfa *, struct colormap *, int, struct state *, struct state *, struct state *);
111
112 #ifdef REG_DEBUG
113 static void dumpcolors(struct colormap *, FILE *);
114 static void fillcheck(struct colormap *, union tree *, int, FILE *);
115 static void dumpchr(chr, FILE *);
116 #endif
117 /* === regc_nfa.c === */
118 static struct nfa *newnfa(struct vars *, struct colormap *, struct nfa *);
119 static void freenfa(struct nfa *);
120 static struct state *newstate(struct nfa *);
121 static struct state *newfstate(struct nfa *, int flag);
122 static void dropstate(struct nfa *, struct state *);
123 static void freestate(struct nfa *, struct state *);
124 static void destroystate(struct nfa *, struct state *);
125 static void newarc(struct nfa *, int, pcolor, struct state *, struct state *);
126 static struct arc *allocarc(struct nfa *, struct state *);
127 static void freearc(struct nfa *, struct arc *);
128 static struct arc *findarc(struct state *, int, pcolor);
129 static void cparc(struct nfa *, struct arc *, struct state *, struct state *);
130 static void moveins(struct nfa *, struct state *, struct state *);
131 static void copyins(struct nfa *, struct state *, struct state *);
132 static void moveouts(struct nfa *, struct state *, struct state *);
133 static void copyouts(struct nfa *, struct state *, struct state *);
134 static void cloneouts(struct nfa *, struct state *, struct state *, struct state *, int);
135 static void delsub(struct nfa *, struct state *, struct state *);
136 static void deltraverse(struct nfa *, struct state *, struct state *);
137 static void dupnfa(struct nfa *, struct state *, struct state *, struct state *, struct state *);
138 static void duptraverse(struct nfa *, struct state *, struct state *);
139 static void cleartraverse(struct nfa *, struct state *);
140 static void specialcolors(struct nfa *);
141 static long optimize(struct nfa *, FILE *);
142 static void pullback(struct nfa *, FILE *);
143 static int pull(struct nfa *, struct arc *);
144 static void pushfwd(struct nfa *, FILE *);
145 static int push(struct nfa *, struct arc *);
146
147 #define INCOMPATIBLE 1 /* destroys arc */
148 #define SATISFIED 2 /* constraint satisfied */
149 #define COMPATIBLE 3 /* compatible but not satisfied yet */
150 static int combine(struct arc *, struct arc *);
151 static void fixempties(struct nfa *, FILE *);
152 static int unempty(struct nfa *, struct arc *);
153 static void cleanup(struct nfa *);
154 static void markreachable(struct nfa *, struct state *, struct state *, struct state *);
155 static void markcanreach(struct nfa *, struct state *, struct state *, struct state *);
156 static long analyze(struct nfa *);
157 static void compact(struct nfa *, struct cnfa *);
158 static void carcsort(struct carc *, struct carc *);
159 static void freecnfa(struct cnfa *);
160 static void dumpnfa(struct nfa *, FILE *);
161
162 #ifdef REG_DEBUG
163 static void dumpstate(struct state *, FILE *);
164 static void dumparcs(struct state *, FILE *);
165 static int dumprarcs(struct arc *, struct state *, FILE *, int);
166 static void dumparc(struct arc *, struct state *, FILE *);
167 static void dumpcnfa(struct cnfa *, FILE *);
168 static void dumpcstate(int, struct carc *, struct cnfa *, FILE *);
169 #endif
170 /* === regc_cvec.c === */
171 static struct cvec *newcvec(int, int, int);
172 static struct cvec *clearcvec(struct cvec *);
173 static void addchr(struct cvec *, chr);
174 static void addrange(struct cvec *, chr, chr);
175 static void addmcce(struct cvec *, chr *, chr *);
176 static int haschr(struct cvec *, chr);
177 static struct cvec *getcvec(struct vars *, int, int, int);
178 static void freecvec(struct cvec *);
179
180 /* === regc_locale.c === */
181 extern int wx_isdigit(wx_wchar c);
182 extern int wx_isalpha(wx_wchar c);
183 extern int wx_isalnum(wx_wchar c);
184 extern int wx_isupper(wx_wchar c);
185 extern int wx_islower(wx_wchar c);
186 extern int wx_isgraph(wx_wchar c);
187 extern int wx_ispunct(wx_wchar c);
188 extern int wx_isspace(wx_wchar c);
189 extern wx_wchar wx_toupper(wx_wchar c);
190 extern wx_wchar wx_tolower(wx_wchar c);
191 extern int wx_strlen(const wx_wchar* szString);
192 static int nmcces(struct vars *);
193 static int nleaders(struct vars *);
194 static struct cvec *allmcces(struct vars *, struct cvec *);
195 static celt element(struct vars *, chr *, chr *);
196 static struct cvec *range(struct vars *, celt, celt, int);
197 static int before(celt, celt);
198 static struct cvec *eclass(struct vars *, celt, int);
199 static struct cvec *cclass(struct vars *, chr *, chr *, int);
200 static struct cvec *allcases(struct vars *, chr);
201 static int cmp(const chr *, const chr *, size_t);
202 static int casecmp(const chr *, const chr *, size_t);
203
204
205 /* internal variables, bundled for easy passing around */
206 struct vars
207 {
208 regex_t *re;
209 chr *now; /* scan pointer into string */
210 chr *stop; /* end of string */
211 chr *savenow; /* saved now and stop for "subroutine
212 * call" */
213 chr *savestop;
214 int err; /* error code (0 if none) */
215 int cflags; /* copy of compile flags */
216 int lasttype; /* type of previous token */
217 int nexttype; /* type of next token */
218 chr nextvalue; /* value (if any) of next token */
219 int lexcon; /* lexical context type (see lex.c) */
220 int nsubexp; /* subexpression count */
221 struct subre **subs; /* subRE pointer vector */
222 size_t nsubs; /* length of vector */
223 struct subre *sub10[10]; /* initial vector, enough for most */
224 struct nfa *nfa; /* the NFA */
225 struct colormap *cm; /* character color map */
226 color nlcolor; /* color of newline */
227 struct state *wordchrs; /* state in nfa holding word-char outarcs */
228 struct subre *tree; /* subexpression tree */
229 struct subre *treechain; /* all tree nodes allocated */
230 struct subre *treefree; /* any free tree nodes */
231 int ntree; /* number of tree nodes */
232 struct cvec *cv; /* interface cvec */
233 struct cvec *cv2; /* utility cvec */
234 struct cvec *mcces; /* collating-element information */
235 #define ISCELEADER(v,c) (v->mcces != NULL && haschr(v->mcces, (c)))
236 struct state *mccepbegin; /* in nfa, start of MCCE prototypes */
237 struct state *mccepend; /* in nfa, end of MCCE prototypes */
238 struct subre *lacons; /* lookahead-constraint vector */
239 int nlacons; /* size of lacons */
240 };
241
242 /* parsing macros; most know that `v' is the struct vars pointer */
243 #define NEXT() (next(v)) /* advance by one token */
244 #define SEE(t) (v->nexttype == (t)) /* is next token this? */
245 #define EAT(t) (SEE(t) && next(v)) /* if next is this, swallow it */
246 #define VISERR(vv) ((vv)->err != 0) /* have we seen an error yet? */
247 #define ISERR() VISERR(v)
248 #define VERR(vv,e) ((vv)->nexttype = EOS, ((vv)->err) ? (vv)->err :\
249 ((vv)->err = (e)))
250 #define ERR(e) VERR(v, e) /* record an error */
251 #define NOERR() {if (ISERR()) return;} /* if error seen, return */
252 #define NOERRN() {if (ISERR()) return NULL;} /* NOERR with retval */
253 #define NOERRZ() {if (ISERR()) return 0;} /* NOERR with retval */
254 #define INSIST(c, e) ((c) ? 0 : ERR(e)) /* if condition false,
255 * error */
256 #define NOTE(b) (v->re->re_info |= (b)) /* note visible condition */
257 #define EMPTYARC(x, y) newarc(v->nfa, EMPTY, 0, x, y)
258
259 /* token type codes, some also used as NFA arc types */
260 #define EMPTY 'n' /* no token present */
261 #define EOS 'e' /* end of string */
262 #define PLAIN 'p' /* ordinary character */
263 #define DIGIT 'd' /* digit (in bound) */
264 #define BACKREF 'b' /* back reference */
265 #define COLLEL 'I' /* start of [. */
266 #define ECLASS 'E' /* start of [= */
267 #define CCLASS 'C' /* start of [: */
268 #define END 'X' /* end of [. [= [: */
269 #define RANGE 'R' /* - within [] which might be range delim. */
270 #define LACON 'L' /* lookahead constraint subRE */
271 #define AHEAD 'a' /* color-lookahead arc */
272 #define BEHIND 'r' /* color-lookbehind arc */
273 #define WBDRY 'w' /* word boundary constraint */
274 #define NWBDRY 'W' /* non-word-boundary constraint */
275 #define SBEGIN 'A' /* beginning of string (even if not BOL) */
276 #define SEND 'Z' /* end of string (even if not EOL) */
277 #define PREFER 'P' /* length preference */
278
279 /* is an arc colored, and hence on a color chain? */
280 #define COLORED(a) ((a)->type == PLAIN || (a)->type == AHEAD || \
281 (a)->type == BEHIND)
282
283
284
285 /* static function list */
286 static struct fns functions = {
287 rfree, /* regfree insides */
288 };
289
290
291
292 /*
293 * regcomp - compile regular expression
294 */
295 int
296 regcomp(regex_t *re,
297 const chr *string,
298 int flags)
299 {
300 return wx_regcomp(re, string, wx_strlen(string), flags);
301 }
302 int
303 wx_regcomp(regex_t *re,
304 const chr *string,
305 size_t len,
306 int flags)
307 {
308 struct vars var;
309 struct vars *v = &var;
310 struct guts *g;
311 int i;
312 size_t j;
313
314 #ifdef REG_DEBUG
315 FILE *debug = (flags & REG_PROGRESS) ? stdout : (FILE *) NULL;
316
317 #else
318 FILE *debug = (FILE *) NULL;
319 #endif
320
321 #define CNOERR() { if (ISERR()) return freev(v, v->err); }
322
323 /* sanity checks */
324
325 if (re == NULL || string == NULL)
326 return REG_INVARG;
327 if ((flags & REG_QUOTE) &&
328 (flags & (REG_ADVANCED | REG_EXPANDED | REG_NEWLINE)))
329 return REG_INVARG;
330 if (!(flags & REG_EXTENDED) && (flags & REG_ADVF))
331 return REG_INVARG;
332
333 /* initial setup (after which freev() is callable) */
334 v->re = re;
335 v->now = (chr *) string;
336 v->stop = v->now + len;
337 v->savenow = v->savestop = NULL;
338 v->err = 0;
339 v->cflags = flags;
340 v->nsubexp = 0;
341 v->subs = v->sub10;
342 v->nsubs = 10;
343 for (j = 0; j < v->nsubs; j++)
344 v->subs[j] = NULL;
345 v->nfa = NULL;
346 v->cm = NULL;
347 v->nlcolor = COLORLESS;
348 v->wordchrs = NULL;
349 v->tree = NULL;
350 v->treechain = NULL;
351 v->treefree = NULL;
352 v->cv = NULL;
353 v->cv2 = NULL;
354 v->mcces = NULL;
355 v->lacons = NULL;
356 v->nlacons = 0;
357 re->re_magic = REMAGIC;
358 re->re_info = 0; /* bits get set during parse */
359 re->re_csize = sizeof(chr);
360 re->re_guts = NULL;
361 re->re_fns = VS(&functions);
362
363 /* more complex setup, malloced things */
364 re->re_guts = VS(MALLOC(sizeof(struct guts)));
365 if (re->re_guts == NULL)
366 return freev(v, REG_ESPACE);
367 g = (struct guts *) re->re_guts;
368 g->tree = NULL;
369 initcm(v, &g->cmap);
370 v->cm = &g->cmap;
371 g->lacons = NULL;
372 g->nlacons = 0;
373 ZAPCNFA(g->search);
374 v->nfa = newnfa(v, v->cm, (struct nfa *) NULL);
375 CNOERR();
376 v->cv = newcvec(100, 20, 10);
377 if (v->cv == NULL)
378 return freev(v, REG_ESPACE);
379 i = nmcces(v);
380 if (i > 0)
381 {
382 v->mcces = newcvec(nleaders(v), 0, i);
383 CNOERR();
384 v->mcces = allmcces(v, v->mcces);
385 leaders(v, v->mcces);
386 addmcce(v->mcces, (chr *) NULL, (chr *) NULL); /* dummy */
387 }
388 CNOERR();
389
390 /* parsing */
391 lexstart(v); /* also handles prefixes */
392 if ((v->cflags & REG_NLSTOP) || (v->cflags & REG_NLANCH))
393 {
394 /* assign newline a unique color */
395 v->nlcolor = subcolor(v->cm, newline());
396 okcolors(v->nfa, v->cm);
397 }
398 CNOERR();
399 v->tree = parse(v, EOS, PLAIN, v->nfa->init, v->nfa->final);
400 assert(SEE(EOS)); /* even if error; ISERR() => SEE(EOS) */
401 CNOERR();
402 assert(v->tree != NULL);
403
404 /* finish setup of nfa and its subre tree */
405 specialcolors(v->nfa);
406 CNOERR();
407 #ifdef REG_DEBUG
408 if (debug != NULL)
409 {
410 fprintf(debug, "\n\n\n========= RAW ==========\n");
411 dumpnfa(v->nfa, debug);
412 dumpst(v->tree, debug, 1);
413 }
414 #endif
415 optst(v, v->tree);
416 v->ntree = numst(v->tree, 1);
417 markst(v->tree);
418 cleanst(v);
419 #ifdef REG_DEBUG
420 if (debug != NULL)
421 {
422 fprintf(debug, "\n\n\n========= TREE FIXED ==========\n");
423 dumpst(v->tree, debug, 1);
424 }
425 #endif
426
427 /* build compacted NFAs for tree and lacons */
428 re->re_info |= nfatree(v, v->tree, debug);
429 CNOERR();
430 assert(v->nlacons == 0 || v->lacons != NULL);
431 for (i = 1; i < v->nlacons; i++)
432 {
433 #ifdef REG_DEBUG
434 if (debug != NULL)
435 fprintf(debug, "\n\n\n========= LA%d ==========\n", i);
436 #endif
437 nfanode(v, &v->lacons[i], debug);
438 }
439 CNOERR();
440 if (v->tree->flags & SHORTER)
441 NOTE(REG_USHORTEST);
442
443 /* build compacted NFAs for tree, lacons, fast search */
444 #ifdef REG_DEBUG
445 if (debug != NULL)
446 fprintf(debug, "\n\n\n========= SEARCH ==========\n");
447 #endif
448 /* can sacrifice main NFA now, so use it as work area */
449 (DISCARD) optimize(v->nfa, debug);
450 CNOERR();
451 makesearch(v, v->nfa);
452 CNOERR();
453 compact(v->nfa, &g->search);
454 CNOERR();
455
456 /* looks okay, package it up */
457 re->re_nsub = v->nsubexp;
458 v->re = NULL; /* freev no longer frees re */
459 g->magic = GUTSMAGIC;
460 g->cflags = v->cflags;
461 g->info = re->re_info;
462 g->nsub = re->re_nsub;
463 g->tree = v->tree;
464 v->tree = NULL;
465 g->ntree = v->ntree;
466 g->compare = (v->cflags & REG_ICASE) ? casecmp : cmp;
467 g->lacons = v->lacons;
468 v->lacons = NULL;
469 g->nlacons = v->nlacons;
470
471 #ifdef REG_DEBUG
472 if (flags & REG_DUMP)
473 dump(re, stdout);
474 #endif
475
476 assert(v->err == 0);
477 return freev(v, 0);
478 }
479
480 /*
481 * moresubs - enlarge subRE vector
482 */
483 static void
484 moresubs(struct vars * v,
485 int wanted) /* want enough room for this one */
486 {
487 struct subre **p;
488 size_t n;
489
490 assert(wanted > 0 && (size_t) wanted >= v->nsubs);
491 n = (size_t) wanted *3 / 2 + 1;
492
493 if (v->subs == v->sub10)
494 {
495 p = (struct subre **) MALLOC(n * sizeof(struct subre *));
496 if (p != NULL)
497 memcpy(VS(p), VS(v->subs),
498 v->nsubs * sizeof(struct subre *));
499 }
500 else
501 p = (struct subre **) REALLOC(v->subs, n * sizeof(struct subre *));
502 if (p == NULL)
503 {
504 ERR(REG_ESPACE);
505 return;
506 }
507 v->subs = p;
508 for (p = &v->subs[v->nsubs]; v->nsubs < n; p++, v->nsubs++)
509 *p = NULL;
510 assert(v->nsubs == n);
511 assert((size_t) wanted < v->nsubs);
512 }
513
514 /*
515 * freev - free vars struct's substructures where necessary
516 *
517 * Optionally does error-number setting, and always returns error code
518 * (if any), to make error-handling code terser.
519 */
520 static int
521 freev(struct vars * v,
522 int err)
523 {
524 if (v->re != NULL)
525 rfree(v->re);
526 if (v->subs != v->sub10)
527 FREE(v->subs);
528 if (v->nfa != NULL)
529 freenfa(v->nfa);
530 if (v->tree != NULL)
531 freesubre(v, v->tree);
532 if (v->treechain != NULL)
533 cleanst(v);
534 if (v->cv != NULL)
535 freecvec(v->cv);
536 if (v->cv2 != NULL)
537 freecvec(v->cv2);
538 if (v->mcces != NULL)
539 freecvec(v->mcces);
540 if (v->lacons != NULL)
541 freelacons(v->lacons, v->nlacons);
542 ERR(err); /* nop if err==0 */
543
544 return v->err;
545 }
546
547 /*
548 * makesearch - turn an NFA into a search NFA (implicit prepend of .*?)
549 * NFA must have been optimize()d already.
550 */
551 static void
552 makesearch(struct vars * v,
553 struct nfa * nfa)
554 {
555 struct arc *a;
556 struct arc *b;
557 struct state *pre = nfa->pre;
558 struct state *s;
559 struct state *s2;
560 struct state *slist;
561
562 /* no loops are needed if it's anchored */
563 for (a = pre->outs; a != NULL; a = a->outchain)
564 {
565 assert(a->type == PLAIN);
566 if (a->co != nfa->bos[0] && a->co != nfa->bos[1])
567 break;
568 }
569 if (a != NULL)
570 {
571 /* add implicit .* in front */
572 rainbow(nfa, v->cm, PLAIN, COLORLESS, pre, pre);
573
574 /* and ^* and \A* too -- not always necessary, but harmless */
575 newarc(nfa, PLAIN, nfa->bos[0], pre, pre);
576 newarc(nfa, PLAIN, nfa->bos[1], pre, pre);
577 }
578
579 /*
580 * Now here's the subtle part. Because many REs have no lookback
581 * constraints, often knowing when you were in the pre state tells you
582 * little; it's the next state(s) that are informative. But some of
583 * them may have other inarcs, i.e. it may be possible to make actual
584 * progress and then return to one of them. We must de-optimize such
585 * cases, splitting each such state into progress and no-progress
586 * states.
587 */
588
589 /* first, make a list of the states */
590 slist = NULL;
591 for (a = pre->outs; a != NULL; a = a->outchain)
592 {
593 s = a->to;
594 for (b = s->ins; b != NULL; b = b->inchain)
595 if (b->from != pre)
596 break;
597 if (b != NULL)
598 { /* must be split */
599 s->tmp = slist;
600 slist = s;
601 }
602 }
603
604 /* do the splits */
605 for (s = slist; s != NULL; s = s2)
606 {
607 s2 = newstate(nfa);
608 copyouts(nfa, s, s2);
609 for (a = s->ins; a != NULL; a = b)
610 {
611 b = a->inchain;
612 if (a->from != pre)
613 {
614 cparc(nfa, a, a->from, s2);
615 freearc(nfa, a);
616 }
617 }
618 s2 = s->tmp;
619 s->tmp = NULL; /* clean up while we're at it */
620 }
621 }
622
623 /*
624 * parse - parse an RE
625 *
626 * This is actually just the top level, which parses a bunch of branches
627 * tied together with '|'. They appear in the tree as the left children
628 * of a chain of '|' subres.
629 */
630 static struct subre *
631 parse(struct vars * v,
632 int stopper, /* EOS or ')' */
633 int type, /* LACON (lookahead subRE) or PLAIN */
634 struct state * init, /* initial state */
635 struct state * final) /* final state */
636 {
637 struct state *left; /* scaffolding for branch */
638 struct state *right;
639 struct subre *branches; /* top level */
640 struct subre *branch; /* current branch */
641 struct subre *t; /* temporary */
642 int firstbranch; /* is this the first branch? */
643
644 assert(stopper == ')' || stopper == EOS);
645
646 branches = subre(v, '|', LONGER, init, final);
647 NOERRN();
648 branch = branches;
649 firstbranch = 1;
650 do
651 { /* a branch */
652 if (!firstbranch)
653 {
654 /* need a place to hang it */
655 branch->right = subre(v, '|', LONGER, init, final);
656 NOERRN();
657 branch = branch->right;
658 }
659 firstbranch = 0;
660 left = newstate(v->nfa);
661 right = newstate(v->nfa);
662 NOERRN();
663 EMPTYARC(init, left);
664 EMPTYARC(right, final);
665 NOERRN();
666 branch->left = parsebranch(v, stopper, type, left, right, 0);
667 NOERRN();
668 branch->flags |= UP(branch->flags | branch->left->flags);
669 if ((branch->flags & ~branches->flags) != 0) /* new flags */
670 for (t = branches; t != branch; t = t->right)
671 t->flags |= branch->flags;
672 } while (EAT('|'));
673 assert(SEE(stopper) || SEE(EOS));
674
675 if (!SEE(stopper))
676 {
677 assert(stopper == ')' && SEE(EOS));
678 ERR(REG_EPAREN);
679 }
680
681 /* optimize out simple cases */
682 if (branch == branches)
683 { /* only one branch */
684 assert(branch->right == NULL);
685 t = branch->left;
686 branch->left = NULL;
687 freesubre(v, branches);
688 branches = t;
689 }
690 else if (!MESSY(branches->flags))
691 { /* no interesting innards */
692 freesubre(v, branches->left);
693 branches->left = NULL;
694 freesubre(v, branches->right);
695 branches->right = NULL;
696 branches->op = '=';
697 }
698
699 return branches;
700 }
701
702 /*
703 * parsebranch - parse one branch of an RE
704 *
705 * This mostly manages concatenation, working closely with parseqatom().
706 * Concatenated things are bundled up as much as possible, with separate
707 * ',' nodes introduced only when necessary due to substructure.
708 */
709 static struct subre *
710 parsebranch(struct vars * v,
711 int stopper, /* EOS or ')' */
712 int type, /* LACON (lookahead subRE) or PLAIN */
713 struct state * left, /* leftmost state */
714 struct state * right, /* rightmost state */
715 int partial) /* is this only part of a branch? */
716 {
717 struct state *lp; /* left end of current construct */
718 int seencontent; /* is there anything in this branch yet? */
719 struct subre *t;
720
721 lp = left;
722 seencontent = 0;
723 t = subre(v, '=', 0, left, right); /* op '=' is tentative */
724 NOERRN();
725 while (!SEE('|') && !SEE(stopper) && !SEE(EOS))
726 {
727 if (seencontent)
728 { /* implicit concat operator */
729 lp = newstate(v->nfa);
730 NOERRN();
731 moveins(v->nfa, right, lp);
732 }
733 seencontent = 1;
734
735 /* NB, recursion in parseqatom() may swallow rest of branch */
736 parseqatom(v, stopper, type, lp, right, t);
737 }
738
739 if (!seencontent)
740 { /* empty branch */
741 if (!partial)
742 NOTE(REG_UUNSPEC);
743 assert(lp == left);
744 EMPTYARC(left, right);
745 }
746
747 return t;
748 }
749
750 /*
751 * parseqatom - parse one quantified atom or constraint of an RE
752 *
753 * The bookkeeping near the end cooperates very closely with parsebranch();
754 * in particular, it contains a recursion that can involve parsing the rest
755 * of the branch, making this function's name somewhat inaccurate.
756 */
757 static void
758 parseqatom(struct vars * v,
759 int stopper, /* EOS or ')' */
760 int type, /* LACON (lookahead subRE) or PLAIN */
761 struct state * lp, /* left state to hang it on */
762 struct state * rp, /* right state to hang it on */
763 struct subre * top) /* subtree top */
764 {
765 struct state *s; /* temporaries for new states */
766 struct state *s2;
767
768 #define ARCV(t, val) newarc(v->nfa, t, val, lp, rp)
769 int m,
770 n;
771 struct subre *atom; /* atom's subtree */
772 struct subre *t;
773 int cap; /* capturing parens? */
774 int pos; /* positive lookahead? */
775 int subno; /* capturing-parens or backref number */
776 int atomtype;
777 int qprefer; /* quantifier short/long preference */
778 int f;
779 struct subre **atomp; /* where the pointer to atom is */
780
781 /* initial bookkeeping */
782 atom = NULL;
783 assert(lp->nouts == 0); /* must string new code */
784 assert(rp->nins == 0); /* between lp and rp */
785 subno = 0; /* just to shut lint up */
786
787 /* an atom or constraint... */
788 atomtype = v->nexttype;
789 switch (atomtype)
790 {
791 /* first, constraints, which end by returning */
792 case '^':
793 ARCV('^', 1);
794 if (v->cflags & REG_NLANCH)
795 ARCV(BEHIND, v->nlcolor);
796 NEXT();
797 return;
798 break;
799 case '$':
800 ARCV('$', 1);
801 if (v->cflags & REG_NLANCH)
802 ARCV(AHEAD, v->nlcolor);
803 NEXT();
804 return;
805 break;
806 case SBEGIN:
807 ARCV('^', 1); /* BOL */
808 ARCV('^', 0); /* or BOS */
809 NEXT();
810 return;
811 break;
812 case SEND:
813 ARCV('$', 1); /* EOL */
814 ARCV('$', 0); /* or EOS */
815 NEXT();
816 return;
817 break;
818 case '<':
819 wordchrs(v); /* does NEXT() */
820 s = newstate(v->nfa);
821 NOERR();
822 nonword(v, BEHIND, lp, s);
823 word(v, AHEAD, s, rp);
824 return;
825 break;
826 case '>':
827 wordchrs(v); /* does NEXT() */
828 s = newstate(v->nfa);
829 NOERR();
830 word(v, BEHIND, lp, s);
831 nonword(v, AHEAD, s, rp);
832 return;
833 break;
834 case WBDRY:
835 wordchrs(v); /* does NEXT() */
836 s = newstate(v->nfa);
837 NOERR();
838 nonword(v, BEHIND, lp, s);
839 word(v, AHEAD, s, rp);
840 s = newstate(v->nfa);
841 NOERR();
842 word(v, BEHIND, lp, s);
843 nonword(v, AHEAD, s, rp);
844 return;
845 break;
846 case NWBDRY:
847 wordchrs(v); /* does NEXT() */
848 s = newstate(v->nfa);
849 NOERR();
850 word(v, BEHIND, lp, s);
851 word(v, AHEAD, s, rp);
852 s = newstate(v->nfa);
853 NOERR();
854 nonword(v, BEHIND, lp, s);
855 nonword(v, AHEAD, s, rp);
856 return;
857 break;
858 case LACON: /* lookahead constraint */
859 pos = v->nextvalue;
860 NEXT();
861 s = newstate(v->nfa);
862 s2 = newstate(v->nfa);
863 NOERR();
864 t = parse(v, ')', LACON, s, s2);
865 freesubre(v, t); /* internal structure irrelevant */
866 assert(SEE(')') || ISERR());
867 NEXT();
868 n = newlacon(v, s, s2, pos);
869 NOERR();
870 ARCV(LACON, n);
871 return;
872 break;
873 /* then errors, to get them out of the way */
874 case '*':
875 case '+':
876 case '?':
877 case '{':
878 ERR(REG_BADRPT);
879 return;
880 break;
881 default:
882 ERR(REG_ASSERT);
883 return;
884 break;
885 /* then plain characters, and minor variants on that theme */
886 case ')': /* unbalanced paren */
887 if ((v->cflags & REG_ADVANCED) != REG_EXTENDED)
888 {
889 ERR(REG_EPAREN);
890 return;
891 }
892 /* legal in EREs due to specification botch */
893 NOTE(REG_UPBOTCH);
894 /* fallthrough into case PLAIN */
895 case PLAIN:
896 onechr(v, v->nextvalue, lp, rp);
897 okcolors(v->nfa, v->cm);
898 NOERR();
899 NEXT();
900 break;
901 case '[':
902 if (v->nextvalue == 1)
903 bracket(v, lp, rp);
904 else
905 cbracket(v, lp, rp);
906 assert(SEE(']') || ISERR());
907 NEXT();
908 break;
909 case '.':
910 rainbow(v->nfa, v->cm, PLAIN,
911 (v->cflags & REG_NLSTOP) ? v->nlcolor : COLORLESS,
912 lp, rp);
913 NEXT();
914 break;
915 /* and finally the ugly stuff */
916 case '(': /* value flags as capturing or non */
917 cap = (type == LACON) ? 0 : v->nextvalue;
918 if (cap)
919 {
920 v->nsubexp++;
921 subno = v->nsubexp;
922 if ((size_t) subno >= v->nsubs)
923 moresubs(v, subno);
924 assert((size_t) subno < v->nsubs);
925 }
926 else
927 atomtype = PLAIN; /* something that's not '(' */
928 NEXT();
929 /* need new endpoints because tree will contain pointers */
930 s = newstate(v->nfa);
931 s2 = newstate(v->nfa);
932 NOERR();
933 EMPTYARC(lp, s);
934 EMPTYARC(s2, rp);
935 NOERR();
936 atom = parse(v, ')', PLAIN, s, s2);
937 assert(SEE(')') || ISERR());
938 NEXT();
939 NOERR();
940 if (cap)
941 {
942 v->subs[subno] = atom;
943 t = subre(v, '(', atom->flags | CAP, lp, rp);
944 NOERR();
945 t->subno = subno;
946 t->left = atom;
947 atom = t;
948 }
949 /* postpone everything else pending possible {0} */
950 break;
951 case BACKREF: /* the Feature From The Black Lagoon */
952 INSIST(type != LACON, REG_ESUBREG);
953 INSIST(v->nextvalue < v->nsubs, REG_ESUBREG);
954 INSIST(v->subs[v->nextvalue] != NULL, REG_ESUBREG);
955 NOERR();
956 assert(v->nextvalue > 0);
957 atom = subre(v, 'b', BACKR, lp, rp);
958 subno = v->nextvalue;
959 atom->subno = subno;
960 EMPTYARC(lp, rp); /* temporarily, so there's something */
961 NEXT();
962 break;
963 }
964
965 /* ...and an atom may be followed by a quantifier */
966 switch (v->nexttype)
967 {
968 case '*':
969 m = 0;
970 n = INFINITY;
971 qprefer = (v->nextvalue) ? LONGER : SHORTER;
972 NEXT();
973 break;
974 case '+':
975 m = 1;
976 n = INFINITY;
977 qprefer = (v->nextvalue) ? LONGER : SHORTER;
978 NEXT();
979 break;
980 case '?':
981 m = 0;
982 n = 1;
983 qprefer = (v->nextvalue) ? LONGER : SHORTER;
984 NEXT();
985 break;
986 case '{':
987 NEXT();
988 m = scannum(v);
989 if (EAT(','))
990 {
991 if (SEE(DIGIT))
992 n = scannum(v);
993 else
994 n = INFINITY;
995 if (m > n)
996 {
997 ERR(REG_BADBR);
998 return;
999 }
1000 /* {m,n} exercises preference, even if it's {m,m} */
1001 qprefer = (v->nextvalue) ? LONGER : SHORTER;
1002 }
1003 else
1004 {
1005 n = m;
1006 /* {m} passes operand's preference through */
1007 qprefer = 0;
1008 }
1009 if (!SEE('}'))
1010 { /* catches errors too */
1011 ERR(REG_BADBR);
1012 return;
1013 }
1014 NEXT();
1015 break;
1016 default: /* no quantifier */
1017 m = n = 1;
1018 qprefer = 0;
1019 break;
1020 }
1021
1022 /* annoying special case: {0} or {0,0} cancels everything */
1023 if (m == 0 && n == 0)
1024 {
1025 if (atom != NULL)
1026 freesubre(v, atom);
1027 if (atomtype == '(')
1028 v->subs[subno] = NULL;
1029 delsub(v->nfa, lp, rp);
1030 EMPTYARC(lp, rp);
1031 return;
1032 }
1033
1034 /* if not a messy case, avoid hard part */
1035 assert(!MESSY(top->flags));
1036 f = top->flags | qprefer | ((atom != NULL) ? atom->flags : 0);
1037 if (atomtype != '(' && atomtype != BACKREF && !MESSY(UP(f)))
1038 {
1039 if (!(m == 1 && n == 1))
1040 repeat(v, lp, rp, m, n);
1041 if (atom != NULL)
1042 freesubre(v, atom);
1043 top->flags = f;
1044 return;
1045 }
1046
1047 /*
1048 * hard part: something messy That is, capturing parens, back
1049 * reference, short/long clash, or an atom with substructure
1050 * containing one of those.
1051 */
1052
1053 /* now we'll need a subre for the contents even if they're boring */
1054 if (atom == NULL)
1055 {
1056 atom = subre(v, '=', 0, lp, rp);
1057 NOERR();
1058 }
1059
1060 /*
1061 * prepare a general-purpose state skeleton
1062 *
1063 * ---> [s] ---prefix---> [begin] ---atom---> [end] ----rest---> [rp] / /
1064 * [lp] ----> [s2] ----bypass---------------------
1065 *
1066 * where bypass is an empty, and prefix is some repetitions of atom
1067 */
1068 s = newstate(v->nfa); /* first, new endpoints for the atom */
1069 s2 = newstate(v->nfa);
1070 NOERR();
1071 moveouts(v->nfa, lp, s);
1072 moveins(v->nfa, rp, s2);
1073 NOERR();
1074 atom->begin = s;
1075 atom->end = s2;
1076 s = newstate(v->nfa); /* and spots for prefix and bypass */
1077 s2 = newstate(v->nfa);
1078 NOERR();
1079 EMPTYARC(lp, s);
1080 EMPTYARC(lp, s2);
1081 NOERR();
1082
1083 /* break remaining subRE into x{...} and what follows */
1084 t = subre(v, '.', COMBINE(qprefer, atom->flags), lp, rp);
1085 t->left = atom;
1086 atomp = &t->left;
1087 /* here we should recurse... but we must postpone that to the end */
1088
1089 /* split top into prefix and remaining */
1090 assert(top->op == '=' && top->left == NULL && top->right == NULL);
1091 top->left = subre(v, '=', top->flags, top->begin, lp);
1092 top->op = '.';
1093 top->right = t;
1094
1095 /* if it's a backref, now is the time to replicate the subNFA */
1096 if (atomtype == BACKREF)
1097 {
1098 assert(atom->begin->nouts == 1); /* just the EMPTY */
1099 delsub(v->nfa, atom->begin, atom->end);
1100 assert(v->subs[subno] != NULL);
1101 /* and here's why the recursion got postponed: it must */
1102 /* wait until the skeleton is filled in, because it may */
1103 /* hit a backref that wants to copy the filled-in skeleton */
1104 dupnfa(v->nfa, v->subs[subno]->begin, v->subs[subno]->end,
1105 atom->begin, atom->end);
1106 NOERR();
1107 }
1108
1109 /* it's quantifier time; first, turn x{0,...} into x{1,...}|empty */
1110 if (m == 0)
1111 {
1112 EMPTYARC(s2, atom->end); /* the bypass */
1113 assert(PREF(qprefer) != 0);
1114 f = COMBINE(qprefer, atom->flags);
1115 t = subre(v, '|', f, lp, atom->end);
1116 NOERR();
1117 t->left = atom;
1118 t->right = subre(v, '|', PREF(f), s2, atom->end);
1119 NOERR();
1120 t->right->left = subre(v, '=', 0, s2, atom->end);
1121 NOERR();
1122 *atomp = t;
1123 atomp = &t->left;
1124 m = 1;
1125 }
1126
1127 /* deal with the rest of the quantifier */
1128 if (atomtype == BACKREF)
1129 {
1130 /* special case: backrefs have internal quantifiers */
1131 EMPTYARC(s, atom->begin); /* empty prefix */
1132 /* just stuff everything into atom */
1133 repeat(v, atom->begin, atom->end, m, n);
1134 atom->min = (short) m;
1135 atom->max = (short) n;
1136 atom->flags |= COMBINE(qprefer, atom->flags);
1137 }
1138 else if (m == 1 && n == 1)
1139 {
1140 /* no/vacuous quantifier: done */
1141 EMPTYARC(s, atom->begin); /* empty prefix */
1142 }
1143 else
1144 {
1145 /* turn x{m,n} into x{m-1,n-1}x, with capturing */
1146 /* parens in only second x */
1147 dupnfa(v->nfa, atom->begin, atom->end, s, atom->begin);
1148 assert(m >= 1 && m != INFINITY && n >= 1);
1149 repeat(v, s, atom->begin, m - 1, (n == INFINITY) ? n : n - 1);
1150 f = COMBINE(qprefer, atom->flags);
1151 t = subre(v, '.', f, s, atom->end); /* prefix and atom */
1152 NOERR();
1153 t->left = subre(v, '=', PREF(f), s, atom->begin);
1154 NOERR();
1155 t->right = atom;
1156 *atomp = t;
1157 }
1158
1159 /* and finally, look after that postponed recursion */
1160 t = top->right;
1161 if (!(SEE('|') || SEE(stopper) || SEE(EOS)))
1162 t->right = parsebranch(v, stopper, type, atom->end, rp, 1);
1163 else
1164 {
1165 EMPTYARC(atom->end, rp);
1166 t->right = subre(v, '=', 0, atom->end, rp);
1167 }
1168 assert(SEE('|') || SEE(stopper) || SEE(EOS));
1169 t->flags |= COMBINE(t->flags, t->right->flags);
1170 top->flags |= COMBINE(top->flags, t->flags);
1171 }
1172
1173 /*
1174 * nonword - generate arcs for non-word-character ahead or behind
1175 */
1176 static void
1177 nonword(struct vars * v,
1178 int dir, /* AHEAD or BEHIND */
1179 struct state * lp,
1180 struct state * rp)
1181 {
1182 int anchor = (dir == AHEAD) ? '$' : '^';
1183
1184 assert(dir == AHEAD || dir == BEHIND);
1185 newarc(v->nfa, anchor, 1, lp, rp);
1186 newarc(v->nfa, anchor, 0, lp, rp);
1187 colorcomplement(v->nfa, v->cm, dir, v->wordchrs, lp, rp);
1188 /* (no need for special attention to \n) */
1189 }
1190
1191 /*
1192 * word - generate arcs for word character ahead or behind
1193 */
1194 static void
1195 word(struct vars * v,
1196 int dir, /* AHEAD or BEHIND */
1197 struct state * lp,
1198 struct state * rp)
1199 {
1200 assert(dir == AHEAD || dir == BEHIND);
1201 cloneouts(v->nfa, v->wordchrs, lp, rp, dir);
1202 /* (no need for special attention to \n) */
1203 }
1204
1205 /*
1206 * scannum - scan a number
1207 */
1208 static int /* value, <= DUPMAX */
1209 scannum(struct vars * v)
1210 {
1211 int n = 0;
1212
1213 while (SEE(DIGIT) && n < DUPMAX)
1214 {
1215 n = n * 10 + v->nextvalue;
1216 NEXT();
1217 }
1218 if (SEE(DIGIT) || n > DUPMAX)
1219 {
1220 ERR(REG_BADBR);
1221 return 0;
1222 }
1223 return n;
1224 }
1225
1226 /*
1227 * repeat - replicate subNFA for quantifiers
1228 *
1229 * The duplication sequences used here are chosen carefully so that any
1230 * pointers starting out pointing into the subexpression end up pointing into
1231 * the last occurrence. (Note that it may not be strung between the same
1232 * left and right end states, however!) This used to be important for the
1233 * subRE tree, although the important bits are now handled by the in-line
1234 * code in parse(), and when this is called, it doesn't matter any more.
1235 */
1236 static void
1237 repeat(struct vars * v,
1238 struct state * lp,
1239 struct state * rp,
1240 int m,
1241 int n)
1242 {
1243 #define SOME 2
1244 #define INF 3
1245 #define PAIR(x, y) ((x)*4 + (y))
1246 #define REDUCE(x) ( ((x) == INFINITY) ? INF : (((x) > 1) ? SOME : (x)) )
1247 const int rm = REDUCE(m);
1248 const int rn = REDUCE(n);
1249 struct state *s;
1250 struct state *s2;
1251
1252 switch (PAIR(rm, rn))
1253 {
1254 case PAIR(0, 0): /* empty string */
1255 delsub(v->nfa, lp, rp);
1256 EMPTYARC(lp, rp);
1257 break;
1258 case PAIR(0, 1): /* do as x| */
1259 EMPTYARC(lp, rp);
1260 break;
1261 case PAIR(0, SOME): /* do as x{1,n}| */
1262 repeat(v, lp, rp, 1, n);
1263 NOERR();
1264 EMPTYARC(lp, rp);
1265 break;
1266 case PAIR(0, INF): /* loop x around */
1267 s = newstate(v->nfa);
1268 NOERR();
1269 moveouts(v->nfa, lp, s);
1270 moveins(v->nfa, rp, s);
1271 EMPTYARC(lp, s);
1272 EMPTYARC(s, rp);
1273 break;
1274 case PAIR(1, 1): /* no action required */
1275 break;
1276 case PAIR(1, SOME): /* do as x{0,n-1}x = (x{1,n-1}|)x */
1277 s = newstate(v->nfa);
1278 NOERR();
1279 moveouts(v->nfa, lp, s);
1280 dupnfa(v->nfa, s, rp, lp, s);
1281 NOERR();
1282 repeat(v, lp, s, 1, n - 1);
1283 NOERR();
1284 EMPTYARC(lp, s);
1285 break;
1286 case PAIR(1, INF): /* add loopback arc */
1287 s = newstate(v->nfa);
1288 s2 = newstate(v->nfa);
1289 NOERR();
1290 moveouts(v->nfa, lp, s);
1291 moveins(v->nfa, rp, s2);
1292 EMPTYARC(lp, s);
1293 EMPTYARC(s2, rp);
1294 EMPTYARC(s2, s);
1295 break;
1296 case PAIR(SOME, SOME): /* do as x{m-1,n-1}x */
1297 s = newstate(v->nfa);
1298 NOERR();
1299 moveouts(v->nfa, lp, s);
1300 dupnfa(v->nfa, s, rp, lp, s);
1301 NOERR();
1302 repeat(v, lp, s, m - 1, n - 1);
1303 break;
1304 case PAIR(SOME, INF): /* do as x{m-1,}x */
1305 s = newstate(v->nfa);
1306 NOERR();
1307 moveouts(v->nfa, lp, s);
1308 dupnfa(v->nfa, s, rp, lp, s);
1309 NOERR();
1310 repeat(v, lp, s, m - 1, n);
1311 break;
1312 default:
1313 ERR(REG_ASSERT);
1314 break;
1315 }
1316 }
1317
1318 /*
1319 * bracket - handle non-complemented bracket expression
1320 * Also called from cbracket for complemented bracket expressions.
1321 */
1322 static void
1323 bracket(struct vars * v,
1324 struct state * lp,
1325 struct state * rp)
1326 {
1327 assert(SEE('['));
1328 NEXT();
1329 while (!SEE(']') && !SEE(EOS))
1330 brackpart(v, lp, rp);
1331 assert(SEE(']') || ISERR());
1332 okcolors(v->nfa, v->cm);
1333 }
1334
1335 /*
1336 * cbracket - handle complemented bracket expression
1337 * We do it by calling bracket() with dummy endpoints, and then complementing
1338 * the result. The alternative would be to invoke rainbow(), and then delete
1339 * arcs as the b.e. is seen... but that gets messy.
1340 */
1341 static void
1342 cbracket(struct vars * v,
1343 struct state * lp,
1344 struct state * rp)
1345 {
1346 struct state *left = newstate(v->nfa);
1347 struct state *right = newstate(v->nfa);
1348 struct state *s;
1349 struct arc *a; /* arc from lp */
1350 struct arc *ba; /* arc from left, from bracket() */
1351 struct arc *pa; /* MCCE-prototype arc */
1352 color co;
1353 chr *p;
1354 int i;
1355
1356 NOERR();
1357 bracket(v, left, right);
1358 if (v->cflags & REG_NLSTOP)
1359 newarc(v->nfa, PLAIN, v->nlcolor, left, right);
1360 NOERR();
1361
1362 assert(lp->nouts == 0); /* all outarcs will be ours */
1363
1364 /* easy part of complementing */
1365 colorcomplement(v->nfa, v->cm, PLAIN, left, lp, rp);
1366 NOERR();
1367 if (v->mcces == NULL)
1368 { /* no MCCEs -- we're done */
1369 dropstate(v->nfa, left);
1370 assert(right->nins == 0);
1371 freestate(v->nfa, right);
1372 return;
1373 }
1374
1375 /* but complementing gets messy in the presence of MCCEs... */
1376 NOTE(REG_ULOCALE);
1377 for (p = v->mcces->chrs, i = v->mcces->nchrs; i > 0; p++, i--)
1378 {
1379 co = GETCOLOR(v->cm, *p);
1380 a = findarc(lp, PLAIN, co);
1381 ba = findarc(left, PLAIN, co);
1382 if (ba == NULL)
1383 {
1384 assert(a != NULL);
1385 freearc(v->nfa, a);
1386 }
1387 else
1388 assert(a == NULL);
1389 s = newstate(v->nfa);
1390 NOERR();
1391 newarc(v->nfa, PLAIN, co, lp, s);
1392 NOERR();
1393 pa = findarc(v->mccepbegin, PLAIN, co);
1394 assert(pa != NULL);
1395 if (ba == NULL)
1396 { /* easy case, need all of them */
1397 cloneouts(v->nfa, pa->to, s, rp, PLAIN);
1398 newarc(v->nfa, '$', 1, s, rp);
1399 newarc(v->nfa, '$', 0, s, rp);
1400 colorcomplement(v->nfa, v->cm, AHEAD, pa->to, s, rp);
1401 }
1402 else
1403 { /* must be selective */
1404 if (findarc(ba->to, '$', 1) == NULL)
1405 {
1406 newarc(v->nfa, '$', 1, s, rp);
1407 newarc(v->nfa, '$', 0, s, rp);
1408 colorcomplement(v->nfa, v->cm, AHEAD, pa->to,
1409 s, rp);
1410 }
1411 for (pa = pa->to->outs; pa != NULL; pa = pa->outchain)
1412 if (findarc(ba->to, PLAIN, pa->co) == NULL)
1413 newarc(v->nfa, PLAIN, pa->co, s, rp);
1414 if (s->nouts == 0) /* limit of selectivity: none */
1415 dropstate(v->nfa, s); /* frees arc too */
1416 }
1417 NOERR();
1418 }
1419
1420 delsub(v->nfa, left, right);
1421 assert(left->nouts == 0);
1422 freestate(v->nfa, left);
1423 assert(right->nins == 0);
1424 freestate(v->nfa, right);
1425 }
1426
1427 /*
1428 * brackpart - handle one item (or range) within a bracket expression
1429 */
1430 static void
1431 brackpart(struct vars * v,
1432 struct state * lp,
1433 struct state * rp)
1434 {
1435 celt startc;
1436 celt endc;
1437 struct cvec *cv;
1438 chr *startp;
1439 chr *endp;
1440 chr c[1];
1441
1442 /* parse something, get rid of special cases, take shortcuts */
1443 switch (v->nexttype)
1444 {
1445 case RANGE: /* a-b-c or other botch */
1446 ERR(REG_ERANGE);
1447 return;
1448 break;
1449 case PLAIN:
1450 c[0] = v->nextvalue;
1451 NEXT();
1452 /* shortcut for ordinary chr (not range, not MCCE leader) */
1453 if (!SEE(RANGE) && !ISCELEADER(v, c[0]))
1454 {
1455 onechr(v, c[0], lp, rp);
1456 return;
1457 }
1458 startc = element(v, c, c + 1);
1459 NOERR();
1460 break;
1461 case COLLEL:
1462 startp = v->now;
1463 endp = scanplain(v);
1464 INSIST(startp < endp, REG_ECOLLATE);
1465 NOERR();
1466 startc = element(v, startp, endp);
1467 NOERR();
1468 break;
1469 case ECLASS:
1470 startp = v->now;
1471 endp = scanplain(v);
1472 INSIST(startp < endp, REG_ECOLLATE);
1473 NOERR();
1474 startc = element(v, startp, endp);
1475 NOERR();
1476 cv = eclass(v, startc, (v->cflags & REG_ICASE));
1477 NOERR();
1478 dovec(v, cv, lp, rp);
1479 return;
1480 break;
1481 case CCLASS:
1482 startp = v->now;
1483 endp = scanplain(v);
1484 INSIST(startp < endp, REG_ECTYPE);
1485 NOERR();
1486 cv = cclass(v, startp, endp, (v->cflags & REG_ICASE));
1487 NOERR();
1488 dovec(v, cv, lp, rp);
1489 return;
1490 break;
1491 default:
1492 ERR(REG_ASSERT);
1493 return;
1494 break;
1495 }
1496
1497 if (SEE(RANGE))
1498 {
1499 NEXT();
1500 switch (v->nexttype)
1501 {
1502 case PLAIN:
1503 case RANGE:
1504 c[0] = v->nextvalue;
1505 NEXT();
1506 endc = element(v, c, c + 1);
1507 NOERR();
1508 break;
1509 case COLLEL:
1510 startp = v->now;
1511 endp = scanplain(v);
1512 INSIST(startp < endp, REG_ECOLLATE);
1513 NOERR();
1514 endc = element(v, startp, endp);
1515 NOERR();
1516 break;
1517 default:
1518 ERR(REG_ERANGE);
1519 return;
1520 break;
1521 }
1522 }
1523 else
1524 endc = startc;
1525
1526 /*
1527 * Ranges are unportable. Actually, standard C does guarantee that
1528 * digits are contiguous, but making that an exception is just too
1529 * complicated.
1530 */
1531 if (startc != endc)
1532 NOTE(REG_UUNPORT);
1533 cv = range(v, startc, endc, (v->cflags & REG_ICASE));
1534 NOERR();
1535 dovec(v, cv, lp, rp);
1536 }
1537
1538 /*
1539 * scanplain - scan PLAIN contents of [. etc.
1540 *
1541 * Certain bits of trickery in lex.c know that this code does not try
1542 * to look past the final bracket of the [. etc.
1543 */
1544 static chr * /* just after end of sequence */
1545 scanplain(struct vars * v)
1546 {
1547 chr *endp;
1548
1549 assert(SEE(COLLEL) || SEE(ECLASS) || SEE(CCLASS));
1550 NEXT();
1551
1552 endp = v->now;
1553 while (SEE(PLAIN))
1554 {
1555 endp = v->now;
1556 NEXT();
1557 }
1558
1559 assert(SEE(END) || ISERR());
1560 NEXT();
1561
1562 return endp;
1563 }
1564
1565 /*
1566 * leaders - process a cvec of collating elements to also include leaders
1567 * Also gives all characters involved their own colors, which is almost
1568 * certainly necessary, and sets up little disconnected subNFA.
1569 */
1570 static void
1571 leaders(struct vars * v,
1572 struct cvec * cv)
1573 {
1574 int mcce;
1575 chr *p;
1576 chr leader;
1577 struct state *s;
1578 struct arc *a;
1579
1580 v->mccepbegin = newstate(v->nfa);
1581 v->mccepend = newstate(v->nfa);
1582 NOERR();
1583
1584 for (mcce = 0; mcce < cv->nmcces; mcce++)
1585 {
1586 p = cv->mcces[mcce];
1587 leader = *p;
1588 if (!haschr(cv, leader))
1589 {
1590 addchr(cv, leader);
1591 s = newstate(v->nfa);
1592 newarc(v->nfa, PLAIN, subcolor(v->cm, leader),
1593 v->mccepbegin, s);
1594 okcolors(v->nfa, v->cm);
1595 }
1596 else
1597 {
1598 a = findarc(v->mccepbegin, PLAIN,
1599 GETCOLOR(v->cm, leader));
1600 assert(a != NULL);
1601 s = a->to;
1602 assert(s != v->mccepend);
1603 }
1604 p++;
1605 assert(*p != 0 && *(p + 1) == 0); /* only 2-char MCCEs for
1606 * now */
1607 newarc(v->nfa, PLAIN, subcolor(v->cm, *p), s, v->mccepend);
1608 okcolors(v->nfa, v->cm);
1609 }
1610 }
1611
1612 /*
1613 * onechr - fill in arcs for a plain character, and possible case complements
1614 * This is mostly a shortcut for efficient handling of the common case.
1615 */
1616 static void
1617 onechr(struct vars * v,
1618 chr c,
1619 struct state * lp,
1620 struct state * rp)
1621 {
1622 if (!(v->cflags & REG_ICASE))
1623 {
1624 newarc(v->nfa, PLAIN, subcolor(v->cm, c), lp, rp);
1625 return;
1626 }
1627
1628 /* rats, need general case anyway... */
1629 dovec(v, allcases(v, c), lp, rp);
1630 }
1631
1632 /*
1633 * dovec - fill in arcs for each element of a cvec
1634 * This one has to handle the messy cases, like MCCEs and MCCE leaders.
1635 */
1636 static void
1637 dovec(struct vars * v,
1638 struct cvec * cv,
1639 struct state * lp,
1640 struct state * rp)
1641 {
1642 chr ch,
1643 from,
1644 to;
1645 celt ce;
1646 chr *p;
1647 int i;
1648 color co;
1649 struct cvec *leads;
1650 struct arc *a;
1651 struct arc *pa; /* arc in prototype */
1652 struct state *s;
1653 struct state *ps; /* state in prototype */
1654
1655 /* need a place to store leaders, if any */
1656 if (nmcces(v) > 0)
1657 {
1658 assert(v->mcces != NULL);
1659 if (v->cv2 == NULL || v->cv2->nchrs < v->mcces->nchrs)
1660 {
1661 if (v->cv2 != NULL)
1662 free(v->cv2);
1663 v->cv2 = newcvec(v->mcces->nchrs, 0, v->mcces->nmcces);
1664 NOERR();
1665 leads = v->cv2;
1666 }
1667 else
1668 leads = clearcvec(v->cv2);
1669 }
1670 else
1671 leads = NULL;
1672
1673 /* first, get the ordinary characters out of the way */
1674 for (p = cv->chrs, i = cv->nchrs; i > 0; p++, i--)
1675 {
1676 ch = *p;
1677 if (!ISCELEADER(v, ch))
1678 newarc(v->nfa, PLAIN, subcolor(v->cm, ch), lp, rp);
1679 else
1680 {
1681 assert(singleton(v->cm, ch));
1682 assert(leads != NULL);
1683 if (!haschr(leads, ch))
1684 addchr(leads, ch);
1685 }
1686 }
1687
1688 /* and the ranges */
1689 for (p = cv->ranges, i = cv->nranges; i > 0; p += 2, i--)
1690 {
1691 from = *p;
1692 to = *(p + 1);
1693 while (from <= to && (ce = nextleader(v, from, to)) != NOCELT)
1694 {
1695 if (from < ce)
1696 subrange(v, from, ce - 1, lp, rp);
1697 assert(singleton(v->cm, ce));
1698 assert(leads != NULL);
1699 if (!haschr(leads, ce))
1700 addchr(leads, ce);
1701 from = ce + 1;
1702 }
1703 if (from <= to)
1704 subrange(v, from, to, lp, rp);
1705 }
1706
1707 if ((leads == NULL || leads->nchrs == 0) && cv->nmcces == 0)
1708 return;
1709
1710 /* deal with the MCCE leaders */
1711 NOTE(REG_ULOCALE);
1712 for (p = leads->chrs, i = leads->nchrs; i > 0; p++, i--)
1713 {
1714 co = GETCOLOR(v->cm, *p);
1715 a = findarc(lp, PLAIN, co);
1716 if (a != NULL)
1717 s = a->to;
1718 else
1719 {
1720 s = newstate(v->nfa);
1721 NOERR();
1722 newarc(v->nfa, PLAIN, co, lp, s);
1723 NOERR();
1724 }
1725 pa = findarc(v->mccepbegin, PLAIN, co);
1726 assert(pa != NULL);
1727 ps = pa->to;
1728 newarc(v->nfa, '$', 1, s, rp);
1729 newarc(v->nfa, '$', 0, s, rp);
1730 colorcomplement(v->nfa, v->cm, AHEAD, ps, s, rp);
1731 NOERR();
1732 }
1733
1734 /* and the MCCEs */
1735 for (i = 0; i < cv->nmcces; i++)
1736 {
1737 p = cv->mcces[i];
1738 assert(singleton(v->cm, *p));
1739 if (!singleton(v->cm, *p))
1740 {
1741 ERR(REG_ASSERT);
1742 return;
1743 }
1744 ch = *p++;
1745 co = GETCOLOR(v->cm, ch);
1746 a = findarc(lp, PLAIN, co);
1747 if (a != NULL)
1748 s = a->to;
1749 else
1750 {
1751 s = newstate(v->nfa);
1752 NOERR();
1753 newarc(v->nfa, PLAIN, co, lp, s);
1754 NOERR();
1755 }
1756 assert(*p != 0); /* at least two chars */
1757 assert(singleton(v->cm, *p));
1758 ch = *p++;
1759 co = GETCOLOR(v->cm, ch);
1760 assert(*p == 0); /* and only two, for now */
1761 newarc(v->nfa, PLAIN, co, s, rp);
1762 NOERR();
1763 }
1764 }
1765
1766 /*
1767 * nextleader - find next MCCE leader within range
1768 */
1769 static celt /* NOCELT means none */
1770 nextleader(struct vars * v,
1771 chr from,
1772 chr to)
1773 {
1774 int i;
1775 chr *p;
1776 chr ch;
1777 celt it = NOCELT;
1778
1779 if (v->mcces == NULL)
1780 return it;
1781
1782 for (i = v->mcces->nchrs, p = v->mcces->chrs; i > 0; i--, p++)
1783 {
1784 ch = *p;
1785 if (from <= ch && ch <= to)
1786 if (it == NOCELT || ch < it)
1787 it = ch;
1788 }
1789 return it;
1790 }
1791
1792 /*
1793 * wordchrs - set up word-chr list for word-boundary stuff, if needed
1794 *
1795 * The list is kept as a bunch of arcs between two dummy states; it's
1796 * disposed of by the unreachable-states sweep in NFA optimization.
1797 * Does NEXT(). Must not be called from any unusual lexical context.
1798 * This should be reconciled with the \w etc. handling in lex.c, and
1799 * should be cleaned up to reduce dependencies on input scanning.
1800 */
1801 static void
1802 wordchrs(struct vars * v)
1803 {
1804 struct state *left;
1805 struct state *right;
1806
1807 if (v->wordchrs != NULL)
1808 {
1809 NEXT(); /* for consistency */
1810 return;
1811 }
1812
1813 left = newstate(v->nfa);
1814 right = newstate(v->nfa);
1815 NOERR();
1816 /* fine point: implemented with [::], and lexer will set REG_ULOCALE */
1817 lexword(v);
1818 NEXT();
1819 assert(v->savenow != NULL && SEE('['));
1820 bracket(v, left, right);
1821 assert((v->savenow != NULL && SEE(']')) || ISERR());
1822 NEXT();
1823 NOERR();
1824 v->wordchrs = left;
1825 }
1826
1827 /*
1828 * subre - allocate a subre
1829 */
1830 static struct subre *
1831 subre(struct vars * v,
1832 int op,
1833 int flags,
1834 struct state * begin,
1835 struct state * end)
1836 {
1837 struct subre *ret;
1838
1839 ret = v->treefree;
1840 if (ret != NULL)
1841 v->treefree = ret->left;
1842 else
1843 {
1844 ret = (struct subre *) MALLOC(sizeof(struct subre));
1845 if (ret == NULL)
1846 {
1847 ERR(REG_ESPACE);
1848 return NULL;
1849 }
1850 ret->chain = v->treechain;
1851 v->treechain = ret;
1852 }
1853
1854 assert(strchr("|.b(=", op) != NULL);
1855
1856 ret->op = op;
1857 ret->flags = flags;
1858 ret->retry = 0;
1859 ret->subno = 0;
1860 ret->min = ret->max = 1;
1861 ret->left = NULL;
1862 ret->right = NULL;
1863 ret->begin = begin;
1864 ret->end = end;
1865 ZAPCNFA(ret->cnfa);
1866
1867 return ret;
1868 }
1869
1870 /*
1871 * freesubre - free a subRE subtree
1872 */
1873 static void
1874 freesubre(struct vars * v, /* might be NULL */
1875 struct subre * sr)
1876 {
1877 if (sr == NULL)
1878 return;
1879
1880 if (sr->left != NULL)
1881 freesubre(v, sr->left);
1882 if (sr->right != NULL)
1883 freesubre(v, sr->right);
1884
1885 freesrnode(v, sr);
1886 }
1887
1888 /*
1889 * freesrnode - free one node in a subRE subtree
1890 */
1891 static void
1892 freesrnode(struct vars * v, /* might be NULL */
1893 struct subre * sr)
1894 {
1895 if (sr == NULL)
1896 return;
1897
1898 if (!NULLCNFA(sr->cnfa))
1899 freecnfa(&sr->cnfa);
1900 sr->flags = 0;
1901
1902 if (v != NULL)
1903 {
1904 sr->left = v->treefree;
1905 v->treefree = sr;
1906 }
1907 else
1908 FREE(sr);
1909 }
1910
1911 /*
1912 * optst - optimize a subRE subtree
1913 */
1914 static void
1915 optst(struct vars * v,
1916 struct subre * t)
1917 {
1918 if (t == NULL)
1919 return;
1920
1921 /* recurse through children */
1922 if (t->left != NULL)
1923 optst(v, t->left);
1924 if (t->right != NULL)
1925 optst(v, t->right);
1926 }
1927
1928 /*
1929 * numst - number tree nodes (assigning retry indexes)
1930 */
1931 static int /* next number */
1932 numst(struct subre * t,
1933 int start) /* starting point for subtree numbers */
1934 {
1935 int i;
1936
1937 assert(t != NULL);
1938
1939 i = start;
1940 t->retry = (short) i++;
1941 if (t->left != NULL)
1942 i = numst(t->left, i);
1943 if (t->right != NULL)
1944 i = numst(t->right, i);
1945 return i;
1946 }
1947
1948 /*
1949 * markst - mark tree nodes as INUSE
1950 */
1951 static void
1952 markst(struct subre * t)
1953 {
1954 assert(t != NULL);
1955
1956 t->flags |= INUSE;
1957 if (t->left != NULL)
1958 markst(t->left);
1959 if (t->right != NULL)
1960 markst(t->right);
1961 }
1962
1963 /*
1964 * cleanst - free any tree nodes not marked INUSE
1965 */
1966 static void
1967 cleanst(struct vars * v)
1968 {
1969 struct subre *t;
1970 struct subre *next;
1971
1972 for (t = v->treechain; t != NULL; t = next)
1973 {
1974 next = t->chain;
1975 if (!(t->flags & INUSE))
1976 FREE(t);
1977 }
1978 v->treechain = NULL;
1979 v->treefree = NULL; /* just on general principles */
1980 }
1981
1982 /*
1983 * nfatree - turn a subRE subtree into a tree of compacted NFAs
1984 */
1985 static long /* optimize results from top node */
1986 nfatree(struct vars * v,
1987 struct subre * t,
1988 FILE *f) /* for debug output */
1989 {
1990 assert(t != NULL && t->begin != NULL);
1991
1992 if (t->left != NULL)
1993 (DISCARD) nfatree(v, t->left, f);
1994 if (t->right != NULL)
1995 (DISCARD) nfatree(v, t->right, f);
1996
1997 return nfanode(v, t, f);
1998 }
1999
2000 /*
2001 * nfanode - do one NFA for nfatree
2002 */
2003 static long /* optimize results */
2004 nfanode(struct vars * v,
2005 struct subre * t,
2006 FILE *f) /* for debug output */
2007 {
2008 struct nfa *nfa;
2009 long ret = 0;
2010
2011 assert(t->begin != NULL);
2012
2013 #ifdef REG_DEBUG
2014 if (f != NULL)
2015 {
2016 char idbuf[50];
2017
2018 fprintf(f, "\n\n\n========= TREE NODE %s ==========\n",
2019 stid(t, idbuf, sizeof(idbuf)));
2020 }
2021 #endif
2022 nfa = newnfa(v, v->cm, v->nfa);
2023 NOERRZ();
2024 dupnfa(nfa, t->begin, t->end, nfa->init, nfa->final);
2025 if (!ISERR())
2026 {
2027 specialcolors(nfa);
2028 ret = optimize(nfa, f);
2029 }
2030 if (!ISERR())
2031 compact(nfa, &t->cnfa);
2032
2033 freenfa(nfa);
2034 return ret;
2035 }
2036
2037 /*
2038 * newlacon - allocate a lookahead-constraint subRE
2039 */
2040 static int /* lacon number */
2041 newlacon(struct vars * v,
2042 struct state * begin,
2043 struct state * end,
2044 int pos)
2045 {
2046 int n;
2047 struct subre *sub;
2048
2049 if (v->nlacons == 0)
2050 {
2051 v->lacons = (struct subre *) MALLOC(2 * sizeof(struct subre));
2052 n = 1; /* skip 0th */
2053 v->nlacons = 2;
2054 }
2055 else
2056 {
2057 v->lacons = (struct subre *) REALLOC(v->lacons,
2058 (v->nlacons + 1) * sizeof(struct subre));
2059 n = v->nlacons++;
2060 }
2061 if (v->lacons == NULL)
2062 {
2063 ERR(REG_ESPACE);
2064 return 0;
2065 }
2066 sub = &v->lacons[n];
2067 sub->begin = begin;
2068 sub->end = end;
2069 sub->subno = pos;
2070 ZAPCNFA(sub->cnfa);
2071 return n;
2072 }
2073
2074 /*
2075 * freelacons - free lookahead-constraint subRE vector
2076 */
2077 static void
2078 freelacons(struct subre * subs,
2079 int n)
2080 {
2081 struct subre *sub;
2082 int i;
2083
2084 assert(n > 0);
2085 for (sub = subs + 1, i = n - 1; i > 0; sub++, i--) /* no 0th */
2086 if (!NULLCNFA(sub->cnfa))
2087 freecnfa(&sub->cnfa);
2088 FREE(subs);
2089 }
2090
2091 /*
2092 * rfree - free a whole RE (insides of regfree)
2093 */
2094 static void
2095 rfree(regex_t *re)
2096 {
2097 struct guts *g;
2098
2099 if (re == NULL || re->re_magic != REMAGIC)
2100 return;
2101
2102 re->re_magic = 0; /* invalidate RE */
2103 g = (struct guts *) re->re_guts;
2104 re->re_guts = NULL;
2105 re->re_fns = NULL;
2106 g->magic = 0;
2107 freecm(&g->cmap);
2108 if (g->tree != NULL)
2109 freesubre((struct vars *) NULL, g->tree);
2110 if (g->lacons != NULL)
2111 freelacons(g->lacons, g->nlacons);
2112 if (!NULLCNFA(g->search))
2113 freecnfa(&g->search);
2114 FREE(g);
2115 }
2116
2117 #ifdef REG_DEBUG
2118
2119 /*
2120 * dump - dump an RE in human-readable form
2121 */
2122 static void
2123 dump(regex_t *re,
2124 FILE *f)
2125 {
2126 struct guts *g;
2127 int i;
2128
2129 if (re->re_magic != REMAGIC)
2130 fprintf(f, "bad magic number (0x%x not 0x%x)\n", re->re_magic,
2131 REMAGIC);
2132 if (re->re_guts == NULL)
2133 {
2134 fprintf(f, "NULL guts!!!\n");
2135 return;
2136 }
2137 g = (struct guts *) re->re_guts;
2138 if (g->magic != GUTSMAGIC)
2139 fprintf(f, "bad guts magic number (0x%x not 0x%x)\n", g->magic,
2140 GUTSMAGIC);
2141
2142 fprintf(f, "\n\n\n========= DUMP ==========\n");
2143 fprintf(f, "nsub %d, info 0%lo, csize %d, ntree %d\n",
2144 re->re_nsub, re->re_info, re->re_csize, g->ntree);
2145
2146 dumpcolors(&g->cmap, f);
2147 if (!NULLCNFA(g->search))
2148 {
2149 printf("\nsearch:\n");
2150 dumpcnfa(&g->search, f);
2151 }
2152 for (i = 1; i < g->nlacons; i++)
2153 {
2154 fprintf(f, "\nla%d (%s):\n", i,
2155 (g->lacons[i].subno) ? "positive" : "negative");
2156 dumpcnfa(&g->lacons[i].cnfa, f);
2157 }
2158 fprintf(f, "\n");
2159 dumpst(g->tree, f, 0);
2160 }
2161
2162 /*
2163 * dumpst - dump a subRE tree
2164 */
2165 static void
2166 dumpst(struct subre * t,
2167 FILE *f,
2168 int nfapresent) /* is the original NFA still around? */
2169 {
2170 if (t == NULL)
2171 fprintf(f, "null tree\n");
2172 else
2173 stdump(t, f, nfapresent);
2174 fflush(f);
2175 }
2176
2177 /*
2178 * stdump - recursive guts of dumpst
2179 */
2180 static void
2181 stdump(struct subre * t,
2182 FILE *f,
2183 int nfapresent) /* is the original NFA still around? */
2184 {
2185 char idbuf[50];
2186
2187 fprintf(f, "%s. `%c'", stid(t, idbuf, sizeof(idbuf)), t->op);
2188 if (t->flags & LONGER)
2189 fprintf(f, " longest");
2190 if (t->flags & SHORTER)
2191 fprintf(f, " shortest");
2192 if (t->flags & MIXED)
2193 fprintf(f, " hasmixed");
2194 if (t->flags & CAP)
2195 fprintf(f, " hascapture");
2196 if (t->flags & BACKR)
2197 fprintf(f, " hasbackref");
2198 if (!(t->flags & INUSE))
2199 fprintf(f, " UNUSED");
2200 if (t->subno != 0)
2201 fprintf(f, " (#%d)", t->subno);
2202 if (t->min != 1 || t->max != 1)
2203 {
2204 fprintf(f, " {%d,", t->min);
2205 if (t->max != INFINITY)
2206 fprintf(f, "%d", t->max);
2207 fprintf(f, "}");
2208 }
2209 if (nfapresent)
2210 fprintf(f, " %ld-%ld", (long) t->begin->no, (long) t->end->no);
2211 if (t->left != NULL)
2212 fprintf(f, " L:%s", stid(t->left, idbuf, sizeof(idbuf)));
2213 if (t->right != NULL)
2214 fprintf(f, " R:%s", stid(t->right, idbuf, sizeof(idbuf)));
2215 if (!NULLCNFA(t->cnfa))
2216 {
2217 fprintf(f, "\n");
2218 dumpcnfa(&t->cnfa, f);
2219 fprintf(f, "\n");
2220 }
2221 if (t->left != NULL)
2222 stdump(t->left, f, nfapresent);
2223 if (t->right != NULL)
2224 stdump(t->right, f, nfapresent);
2225 }
2226
2227 /*
2228 * stid - identify a subtree node for dumping
2229 */
2230 static char * /* points to buf or constant string */
2231 stid(struct subre * t,
2232 char *buf,
2233 size_t bufsize)
2234 {
2235 /* big enough for hex int or decimal t->retry? */
2236 if (bufsize < sizeof(int) * 2 + 3 || bufsize < sizeof(t->retry) * 3 + 1)
2237 return "unable";
2238 if (t->retry != 0)
2239 sprintf(buf, "%d", t->retry);
2240 else
2241 sprintf(buf, "0x%x", (int) t); /* may lose bits, that's okay */
2242 return buf;
2243 }
2244 #endif /* REG_DEBUG */
2245
2246
2247 #include "regc_lex.c"
2248 #include "regc_color.c"
2249 #include "regc_nfa.c"
2250 #include "regc_cvec.c"
2251 #include "regc_locale.c"
wxWidgets (Impactor)