/[dtapublic]/projs/trunk/shared_source/tcl_base/regcomp.c
ViewVC logotype

Contents of /projs/trunk/shared_source/tcl_base/regcomp.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 42 - (show annotations) (download)
Fri Oct 14 01:50:00 2016 UTC (8 years, 2 months ago) by dashley
File MIME type: text/plain
File size: 62035 byte(s)
Move shared source code to commonize.
1 /* $Header: /cvsroot/esrg/sfesrg/esrgpcpj/shared/tcl_base/regcomp.c,v 1.1.1.1 2001/06/13 04:31:32 dtashley Exp $ */
2
3 /*
4 * re_*comp and friends - compile REs
5 * This file #includes several others (see the bottom).
6 *
7 * Copyright (c) 1998, 1999 Henry Spencer. All rights reserved.
8 *
9 * Development of this software was funded, in part, by Cray Research Inc.,
10 * UUNET Communications Services Inc., Sun Microsystems Inc., and Scriptics
11 * Corporation, none of whom are responsible for the results. The author
12 * thanks all of them.
13 *
14 * Redistribution and use in source and binary forms -- with or without
15 * modification -- are permitted for any purpose, provided that
16 * redistributions in source form retain this entire copyright notice and
17 * indicate the origin and nature of any modifications.
18 *
19 * I'd appreciate being given credit for this package in the documentation
20 * of software which uses it, but that is not a requirement.
21 *
22 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES,
23 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
24 * AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL
25 * HENRY SPENCER BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
26 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
27 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
28 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
29 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
30 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
31 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32 *
33 */
34
35 #include "regguts.h"
36
37 /*
38 * forward declarations, up here so forward datatypes etc. are defined early
39 */
40 /* =====^!^===== begin forwards =====^!^===== */
41 /* automatically gathered by fwd; do not hand-edit */
42 /* === regcomp.c === */
43 int compile _ANSI_ARGS_((regex_t *, CONST chr *, size_t, int));
44 static VOID moresubs _ANSI_ARGS_((struct vars *, int));
45 static int freev _ANSI_ARGS_((struct vars *, int));
46 static VOID makesearch _ANSI_ARGS_((struct vars *, struct nfa *));
47 static struct subre *parse _ANSI_ARGS_((struct vars *, int, int, struct state *, struct state *));
48 static struct subre *parsebranch _ANSI_ARGS_((struct vars *, int, int, struct state *, struct state *, int));
49 static VOID parseqatom _ANSI_ARGS_((struct vars *, int, int, struct state *, struct state *, struct subre *));
50 static VOID nonword _ANSI_ARGS_((struct vars *, int, struct state *, struct state *));
51 static VOID word _ANSI_ARGS_((struct vars *, int, struct state *, struct state *));
52 static int scannum _ANSI_ARGS_((struct vars *));
53 static VOID repeat _ANSI_ARGS_((struct vars *, struct state *, struct state *, int, int));
54 static VOID bracket _ANSI_ARGS_((struct vars *, struct state *, struct state *));
55 static VOID cbracket _ANSI_ARGS_((struct vars *, struct state *, struct state *));
56 static VOID brackpart _ANSI_ARGS_((struct vars *, struct state *, struct state *));
57 static chr *scanplain _ANSI_ARGS_((struct vars *));
58 static VOID leaders _ANSI_ARGS_((struct vars *, struct cvec *));
59 static VOID onechr _ANSI_ARGS_((struct vars *, pchr, struct state *, struct state *));
60 static VOID dovec _ANSI_ARGS_((struct vars *, struct cvec *, struct state *, struct state *));
61 static celt nextleader _ANSI_ARGS_((struct vars *, pchr, pchr));
62 static VOID wordchrs _ANSI_ARGS_((struct vars *));
63 static struct subre *subre _ANSI_ARGS_((struct vars *, int, int, struct state *, struct state *));
64 static VOID freesubre _ANSI_ARGS_((struct vars *, struct subre *));
65 static VOID freesrnode _ANSI_ARGS_((struct vars *, struct subre *));
66 static VOID optst _ANSI_ARGS_((struct vars *, struct subre *));
67 static int numst _ANSI_ARGS_((struct subre *, int));
68 static VOID markst _ANSI_ARGS_((struct subre *));
69 static VOID cleanst _ANSI_ARGS_((struct vars *));
70 static long nfatree _ANSI_ARGS_((struct vars *, struct subre *, FILE *));
71 static long nfanode _ANSI_ARGS_((struct vars *, struct subre *, FILE *));
72 static int newlacon _ANSI_ARGS_((struct vars *, struct state *, struct state *, int));
73 static VOID freelacons _ANSI_ARGS_((struct subre *, int));
74 static VOID rfree _ANSI_ARGS_((regex_t *));
75 static VOID dump _ANSI_ARGS_((regex_t *, FILE *));
76 static VOID dumpst _ANSI_ARGS_((struct subre *, FILE *, int));
77 static VOID stdump _ANSI_ARGS_((struct subre *, FILE *, int));
78 static char *stid _ANSI_ARGS_((struct subre *, char *, size_t));
79 /* === regc_lex.c === */
80 static VOID lexstart _ANSI_ARGS_((struct vars *));
81 static VOID prefixes _ANSI_ARGS_((struct vars *));
82 static VOID lexnest _ANSI_ARGS_((struct vars *, chr *, chr *));
83 static VOID lexword _ANSI_ARGS_((struct vars *));
84 static int next _ANSI_ARGS_((struct vars *));
85 static int lexescape _ANSI_ARGS_((struct vars *));
86 static chr lexdigits _ANSI_ARGS_((struct vars *, int, int, int));
87 static int brenext _ANSI_ARGS_((struct vars *, pchr));
88 static VOID skip _ANSI_ARGS_((struct vars *));
89 static chr newline _ANSI_ARGS_((NOPARMS));
90 #ifdef REG_DEBUG
91 static chr *ch _ANSI_ARGS_((NOPARMS));
92 #endif
93 static chr chrnamed _ANSI_ARGS_((struct vars *, chr *, chr *, pchr));
94 /* === regc_color.c === */
95 static VOID initcm _ANSI_ARGS_((struct vars *, struct colormap *));
96 static VOID freecm _ANSI_ARGS_((struct colormap *));
97 static VOID cmtreefree _ANSI_ARGS_((struct colormap *, union tree *, int));
98 static color setcolor _ANSI_ARGS_((struct colormap *, pchr, pcolor));
99 static color maxcolor _ANSI_ARGS_((struct colormap *));
100 static color newcolor _ANSI_ARGS_((struct colormap *));
101 static VOID freecolor _ANSI_ARGS_((struct colormap *, pcolor));
102 static color pseudocolor _ANSI_ARGS_((struct colormap *));
103 static color subcolor _ANSI_ARGS_((struct colormap *, pchr c));
104 static color newsub _ANSI_ARGS_((struct colormap *, pcolor));
105 static VOID subrange _ANSI_ARGS_((struct vars *, pchr, pchr, struct state *, struct state *));
106 static VOID subblock _ANSI_ARGS_((struct vars *, pchr, struct state *, struct state *));
107 static VOID okcolors _ANSI_ARGS_((struct nfa *, struct colormap *));
108 static VOID colorchain _ANSI_ARGS_((struct colormap *, struct arc *));
109 static VOID uncolorchain _ANSI_ARGS_((struct colormap *, struct arc *));
110 static int singleton _ANSI_ARGS_((struct colormap *, pchr c));
111 static VOID rainbow _ANSI_ARGS_((struct nfa *, struct colormap *, int, pcolor, struct state *, struct state *));
112 static VOID colorcomplement _ANSI_ARGS_((struct nfa *, struct colormap *, int, struct state *, struct state *, struct state *));
113 #ifdef REG_DEBUG
114 static VOID dumpcolors _ANSI_ARGS_((struct colormap *, FILE *));
115 static VOID fillcheck _ANSI_ARGS_((struct colormap *, union tree *, int, FILE *));
116 static VOID dumpchr _ANSI_ARGS_((pchr, FILE *));
117 #endif
118 /* === regc_nfa.c === */
119 static struct nfa *newnfa _ANSI_ARGS_((struct vars *, struct colormap *, struct nfa *));
120 static VOID freenfa _ANSI_ARGS_((struct nfa *));
121 static struct state *newstate _ANSI_ARGS_((struct nfa *));
122 static struct state *newfstate _ANSI_ARGS_((struct nfa *, int flag));
123 static VOID dropstate _ANSI_ARGS_((struct nfa *, struct state *));
124 static VOID freestate _ANSI_ARGS_((struct nfa *, struct state *));
125 static VOID destroystate _ANSI_ARGS_((struct nfa *, struct state *));
126 static VOID newarc _ANSI_ARGS_((struct nfa *, int, pcolor, struct state *, struct state *));
127 static struct arc *allocarc _ANSI_ARGS_((struct nfa *, struct state *));
128 static VOID freearc _ANSI_ARGS_((struct nfa *, struct arc *));
129 static struct arc *findarc _ANSI_ARGS_((struct state *, int, pcolor));
130 static VOID cparc _ANSI_ARGS_((struct nfa *, struct arc *, struct state *, struct state *));
131 static VOID moveins _ANSI_ARGS_((struct nfa *, struct state *, struct state *));
132 static VOID copyins _ANSI_ARGS_((struct nfa *, struct state *, struct state *));
133 static VOID moveouts _ANSI_ARGS_((struct nfa *, struct state *, struct state *));
134 static VOID copyouts _ANSI_ARGS_((struct nfa *, struct state *, struct state *));
135 static VOID cloneouts _ANSI_ARGS_((struct nfa *, struct state *, struct state *, struct state *, int));
136 static VOID delsub _ANSI_ARGS_((struct nfa *, struct state *, struct state *));
137 static VOID deltraverse _ANSI_ARGS_((struct nfa *, struct state *, struct state *));
138 static VOID dupnfa _ANSI_ARGS_((struct nfa *, struct state *, struct state *, struct state *, struct state *));
139 static VOID duptraverse _ANSI_ARGS_((struct nfa *, struct state *, struct state *));
140 static VOID cleartraverse _ANSI_ARGS_((struct nfa *, struct state *));
141 static VOID specialcolors _ANSI_ARGS_((struct nfa *));
142 static long optimize _ANSI_ARGS_((struct nfa *, FILE *));
143 static VOID pullback _ANSI_ARGS_((struct nfa *, FILE *));
144 static int pull _ANSI_ARGS_((struct nfa *, struct arc *));
145 static VOID pushfwd _ANSI_ARGS_((struct nfa *, FILE *));
146 static int push _ANSI_ARGS_((struct nfa *, struct arc *));
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 _ANSI_ARGS_((struct arc *, struct arc *));
151 static VOID fixempties _ANSI_ARGS_((struct nfa *, FILE *));
152 static int unempty _ANSI_ARGS_((struct nfa *, struct arc *));
153 static VOID cleanup _ANSI_ARGS_((struct nfa *));
154 static VOID markreachable _ANSI_ARGS_((struct nfa *, struct state *, struct state *, struct state *));
155 static VOID markcanreach _ANSI_ARGS_((struct nfa *, struct state *, struct state *, struct state *));
156 static long analyze _ANSI_ARGS_((struct nfa *));
157 static VOID compact _ANSI_ARGS_((struct nfa *, struct cnfa *));
158 static VOID carcsort _ANSI_ARGS_((struct carc *, struct carc *));
159 static VOID freecnfa _ANSI_ARGS_((struct cnfa *));
160 static VOID dumpnfa _ANSI_ARGS_((struct nfa *, FILE *));
161 #ifdef REG_DEBUG
162 static VOID dumpstate _ANSI_ARGS_((struct state *, FILE *));
163 static VOID dumparcs _ANSI_ARGS_((struct state *, FILE *));
164 static int dumprarcs _ANSI_ARGS_((struct arc *, struct state *, FILE *, int));
165 static VOID dumparc _ANSI_ARGS_((struct arc *, struct state *, FILE *));
166 #endif
167 static VOID dumpcnfa _ANSI_ARGS_((struct cnfa *, FILE *));
168 #ifdef REG_DEBUG
169 static VOID dumpcstate _ANSI_ARGS_((int, struct carc *, struct cnfa *, FILE *));
170 #endif
171 /* === regc_cvec.c === */
172 static struct cvec *newcvec _ANSI_ARGS_((int, int, int));
173 static struct cvec *clearcvec _ANSI_ARGS_((struct cvec *));
174 static VOID addchr _ANSI_ARGS_((struct cvec *, pchr));
175 static VOID addrange _ANSI_ARGS_((struct cvec *, pchr, pchr));
176 static VOID addmcce _ANSI_ARGS_((struct cvec *, chr *, chr *));
177 static int haschr _ANSI_ARGS_((struct cvec *, pchr));
178 static struct cvec *getcvec _ANSI_ARGS_((struct vars *, int, int, int));
179 static VOID freecvec _ANSI_ARGS_((struct cvec *));
180 /* === regc_locale.c === */
181 static int nmcces _ANSI_ARGS_((struct vars *));
182 static int nleaders _ANSI_ARGS_((struct vars *));
183 static struct cvec *allmcces _ANSI_ARGS_((struct vars *, struct cvec *));
184 static celt element _ANSI_ARGS_((struct vars *, chr *, chr *));
185 static struct cvec *range _ANSI_ARGS_((struct vars *, celt, celt, int));
186 static int before _ANSI_ARGS_((celt, celt));
187 static struct cvec *eclass _ANSI_ARGS_((struct vars *, celt, int));
188 static struct cvec *cclass _ANSI_ARGS_((struct vars *, chr *, chr *, int));
189 static struct cvec *allcases _ANSI_ARGS_((struct vars *, pchr));
190 static int cmp _ANSI_ARGS_((CONST chr *, CONST chr *, size_t));
191 static int casecmp _ANSI_ARGS_((CONST chr *, CONST chr *, size_t));
192 /* automatically gathered by fwd; do not hand-edit */
193 /* =====^!^===== end forwards =====^!^===== */
194
195
196
197 /* internal variables, bundled for easy passing around */
198 struct vars {
199 regex_t *re;
200 chr *now; /* scan pointer into string */
201 chr *stop; /* end of string */
202 chr *savenow; /* saved now and stop for "subroutine call" */
203 chr *savestop;
204 int err; /* error code (0 if none) */
205 int cflags; /* copy of compile flags */
206 int lasttype; /* type of previous token */
207 int nexttype; /* type of next token */
208 chr nextvalue; /* value (if any) of next token */
209 int lexcon; /* lexical context type (see lex.c) */
210 int nsubexp; /* subexpression count */
211 struct subre **subs; /* subRE pointer vector */
212 size_t nsubs; /* length of vector */
213 struct subre *sub10[10]; /* initial vector, enough for most */
214 struct nfa *nfa; /* the NFA */
215 struct colormap *cm; /* character color map */
216 color nlcolor; /* color of newline */
217 struct state *wordchrs; /* state in nfa holding word-char outarcs */
218 struct subre *tree; /* subexpression tree */
219 struct subre *treechain; /* all tree nodes allocated */
220 struct subre *treefree; /* any free tree nodes */
221 int ntree; /* number of tree nodes */
222 struct cvec *cv; /* interface cvec */
223 struct cvec *cv2; /* utility cvec */
224 struct cvec *mcces; /* collating-element information */
225 # define ISCELEADER(v,c) (v->mcces != NULL && haschr(v->mcces, (c)))
226 struct state *mccepbegin; /* in nfa, start of MCCE prototypes */
227 struct state *mccepend; /* in nfa, end of MCCE prototypes */
228 struct subre *lacons; /* lookahead-constraint vector */
229 int nlacons; /* size of lacons */
230 };
231
232 /* parsing macros; most know that `v' is the struct vars pointer */
233 #define NEXT() (next(v)) /* advance by one token */
234 #define SEE(t) (v->nexttype == (t)) /* is next token this? */
235 #define EAT(t) (SEE(t) && next(v)) /* if next is this, swallow it */
236 #define VISERR(vv) ((vv)->err != 0) /* have we seen an error yet? */
237 #define ISERR() VISERR(v)
238 #define VERR(vv,e) ((vv)->nexttype = EOS, ((vv)->err) ? (vv)->err :\
239 ((vv)->err = (e)))
240 #define ERR(e) VERR(v, e) /* record an error */
241 #define NOERR() {if (ISERR()) return;} /* if error seen, return */
242 #define NOERRN() {if (ISERR()) return NULL;} /* NOERR with retval */
243 #define NOERRZ() {if (ISERR()) return 0;} /* NOERR with retval */
244 #define INSIST(c, e) ((c) ? 0 : ERR(e)) /* if condition false, error */
245 #define NOTE(b) (v->re->re_info |= (b)) /* note visible condition */
246 #define EMPTYARC(x, y) newarc(v->nfa, EMPTY, 0, x, y)
247
248 /* token type codes, some also used as NFA arc types */
249 #define EMPTY 'n' /* no token present */
250 #define EOS 'e' /* end of string */
251 #define PLAIN 'p' /* ordinary character */
252 #define DIGIT 'd' /* digit (in bound) */
253 #define BACKREF 'b' /* back reference */
254 #define COLLEL 'I' /* start of [. */
255 #define ECLASS 'E' /* start of [= */
256 #define CCLASS 'C' /* start of [: */
257 #define END 'X' /* end of [. [= [: */
258 #define RANGE 'R' /* - within [] which might be range delim. */
259 #define LACON 'L' /* lookahead constraint subRE */
260 #define AHEAD 'a' /* color-lookahead arc */
261 #define BEHIND 'r' /* color-lookbehind arc */
262 #define WBDRY 'w' /* word boundary constraint */
263 #define NWBDRY 'W' /* non-word-boundary constraint */
264 #define SBEGIN 'A' /* beginning of string (even if not BOL) */
265 #define SEND 'Z' /* end of string (even if not EOL) */
266 #define PREFER 'P' /* length preference */
267
268 /* is an arc colored, and hence on a color chain? */
269 #define COLORED(a) ((a)->type == PLAIN || (a)->type == AHEAD || \
270 (a)->type == BEHIND)
271
272
273
274 /* static function list */
275 static struct fns functions = {
276 rfree, /* regfree insides */
277 };
278
279
280
281 /*
282 - compile - compile regular expression
283 ^ int compile(regex_t *, CONST chr *, size_t, int);
284 */
285 int
286 compile(re, string, len, flags)
287 regex_t *re;
288 CONST chr *string;
289 size_t len;
290 int flags;
291 {
292 struct vars var;
293 struct vars *v = &var;
294 struct guts *g;
295 int i;
296 size_t j;
297 FILE *debug = (flags&REG_PROGRESS) ? stdout : (FILE *)NULL;
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 = (chr *)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->mcces = NULL;
332 v->lacons = NULL;
333 v->nlacons = 0;
334 re->re_magic = REMAGIC;
335 re->re_info = 0; /* bits get set during parse */
336 re->re_csize = sizeof(chr);
337 re->re_guts = NULL;
338 re->re_fns = VS(&functions);
339
340 /* more complex setup, malloced things */
341 re->re_guts = VS(MALLOC(sizeof(struct guts)));
342 if (re->re_guts == NULL)
343 return freev(v, REG_ESPACE);
344 g = (struct guts *)re->re_guts;
345 g->tree = NULL;
346 initcm(v, &g->cmap);
347 v->cm = &g->cmap;
348 g->lacons = NULL;
349 g->nlacons = 0;
350 ZAPCNFA(g->search);
351 v->nfa = newnfa(v, v->cm, (struct nfa *)NULL);
352 CNOERR();
353 v->cv = newcvec(100, 20, 10);
354 if (v->cv == NULL)
355 return freev(v, REG_ESPACE);
356 i = nmcces(v);
357 if (i > 0) {
358 v->mcces = newcvec(nleaders(v), 0, i);
359 CNOERR();
360 v->mcces = allmcces(v, v->mcces);
361 leaders(v, v->mcces);
362 addmcce(v->mcces, (chr *)NULL, (chr *)NULL); /* dummy */
363 }
364 CNOERR();
365
366 /* parsing */
367 lexstart(v); /* also handles prefixes */
368 if ((v->cflags&REG_NLSTOP) || (v->cflags&REG_NLANCH)) {
369 /* assign newline a unique color */
370 v->nlcolor = subcolor(v->cm, newline());
371 okcolors(v->nfa, v->cm);
372 }
373 CNOERR();
374 v->tree = parse(v, EOS, PLAIN, v->nfa->init, v->nfa->final);
375 assert(SEE(EOS)); /* even if error; ISERR() => SEE(EOS) */
376 CNOERR();
377 assert(v->tree != NULL);
378
379 /* finish setup of nfa and its subre tree */
380 specialcolors(v->nfa);
381 CNOERR();
382 if (debug != NULL) {
383 fprintf(debug, "\n\n\n========= RAW ==========\n");
384 dumpnfa(v->nfa, debug);
385 dumpst(v->tree, debug, 1);
386 }
387 optst(v, v->tree);
388 v->ntree = numst(v->tree, 1);
389 markst(v->tree);
390 cleanst(v);
391 if (debug != NULL) {
392 fprintf(debug, "\n\n\n========= TREE FIXED ==========\n");
393 dumpst(v->tree, debug, 1);
394 }
395
396 /* build compacted NFAs for tree and lacons */
397 re->re_info |= nfatree(v, v->tree, debug);
398 CNOERR();
399 assert(v->nlacons == 0 || v->lacons != NULL);
400 for (i = 1; i < v->nlacons; i++) {
401 if (debug != NULL)
402 fprintf(debug, "\n\n\n========= LA%d ==========\n", i);
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 if (debug != NULL)
411 fprintf(debug, "\n\n\n========= SEARCH ==========\n");
412 /* can sacrifice main NFA now, so use it as work area */
413 (DISCARD)optimize(v->nfa, debug);
414 CNOERR();
415 makesearch(v, v->nfa);
416 CNOERR();
417 compact(v->nfa, &g->search);
418 CNOERR();
419
420 /* looks okay, package it up */
421 re->re_nsub = v->nsubexp;
422 v->re = NULL; /* freev no longer frees re */
423 g->magic = GUTSMAGIC;
424 g->cflags = v->cflags;
425 g->info = re->re_info;
426 g->nsub = re->re_nsub;
427 g->tree = v->tree;
428 v->tree = NULL;
429 g->ntree = v->ntree;
430 g->compare = (v->cflags&REG_ICASE) ? casecmp : cmp;
431 g->lacons = v->lacons;
432 v->lacons = NULL;
433 g->nlacons = v->nlacons;
434
435 if (flags&REG_DUMP)
436 dump(re, stdout);
437
438 assert(v->err == 0);
439 return freev(v, 0);
440 }
441
442 /*
443 - moresubs - enlarge subRE vector
444 ^ static VOID moresubs(struct vars *, int);
445 */
446 static VOID
447 moresubs(v, wanted)
448 struct vars *v;
449 int wanted; /* want enough room for this one */
450 {
451 struct subre **p;
452 size_t n;
453
454 assert(wanted > 0 && (size_t)wanted >= v->nsubs);
455 n = (size_t)wanted * 3 / 2 + 1;
456 if (v->subs == v->sub10) {
457 p = (struct subre **)MALLOC(n * sizeof(struct subre *));
458 if (p != NULL)
459 memcpy(VS(p), VS(v->subs),
460 v->nsubs * sizeof(struct subre *));
461 } else
462 p = (struct subre **)REALLOC(v->subs, n*sizeof(struct subre *));
463 if (p == NULL) {
464 ERR(REG_ESPACE);
465 return;
466 }
467 v->subs = p;
468 for (p = &v->subs[v->nsubs]; v->nsubs < n; p++, v->nsubs++)
469 *p = NULL;
470 assert(v->nsubs == n);
471 assert((size_t)wanted < v->nsubs);
472 }
473
474 /*
475 - freev - free vars struct's substructures where necessary
476 * Optionally does error-number setting, and always returns error code
477 * (if any), to make error-handling code terser.
478 ^ static int freev(struct vars *, int);
479 */
480 static int
481 freev(v, err)
482 struct vars *v;
483 int err;
484 {
485 if (v->re != NULL)
486 rfree(v->re);
487 if (v->subs != v->sub10)
488 FREE(v->subs);
489 if (v->nfa != NULL)
490 freenfa(v->nfa);
491 if (v->tree != NULL)
492 freesubre(v, v->tree);
493 if (v->treechain != NULL)
494 cleanst(v);
495 if (v->cv != NULL)
496 freecvec(v->cv);
497 if (v->cv2 != NULL)
498 freecvec(v->cv2);
499 if (v->mcces != NULL)
500 freecvec(v->mcces);
501 if (v->lacons != NULL)
502 freelacons(v->lacons, v->nlacons);
503 ERR(err); /* nop if err==0 */
504
505 return v->err;
506 }
507
508 /*
509 - makesearch - turn an NFA into a search NFA (implicit prepend of .*?)
510 * NFA must have been optimize()d already.
511 ^ static VOID makesearch(struct vars *, struct nfa *);
512 */
513 static VOID
514 makesearch(v, nfa)
515 struct vars *v;
516 struct nfa *nfa;
517 {
518 struct arc *a;
519 struct arc *b;
520 struct state *pre = nfa->pre;
521 struct state *s;
522 struct state *s2;
523 struct state *slist;
524
525 /* no loops are needed if it's anchored */
526 for (a = pre->outs; a != NULL; a = a->outchain) {
527 assert(a->type == PLAIN);
528 if (a->co != nfa->bos[0] && a->co != nfa->bos[1])
529 break;
530 }
531 if (a != NULL) {
532 /* add implicit .* in front */
533 rainbow(nfa, v->cm, PLAIN, COLORLESS, pre, pre);
534
535 /* and ^* and \A* too -- not always necessary, but harmless */
536 newarc(nfa, PLAIN, nfa->bos[0], pre, pre);
537 newarc(nfa, PLAIN, nfa->bos[1], pre, pre);
538 }
539
540 /*
541 * Now here's the subtle part. Because many REs have no lookback
542 * constraints, often knowing when you were in the pre state tells
543 * you little; it's the next state(s) that are informative. But
544 * some of them may have other inarcs, i.e. it may be possible to
545 * make actual progress and then return to one of them. We must
546 * de-optimize such cases, splitting each such state into progress
547 * and no-progress states.
548 */
549
550 /* first, make a list of the states */
551 slist = NULL;
552 for (a = pre->outs; a != NULL; a = a->outchain) {
553 s = a->to;
554 for (b = s->ins; b != NULL; b = b->inchain)
555 if (b->from != pre)
556 break;
557 if (b != NULL) { /* must be split */
558 s->tmp = slist;
559 slist = s;
560 }
561 }
562
563 /* do the splits */
564 for (s = slist; s != NULL; s = s2) {
565 s2 = newstate(nfa);
566 copyouts(nfa, s, s2);
567 for (a = s->ins; a != NULL; a = b) {
568 b = a->inchain;
569 if (a->from != pre) {
570 cparc(nfa, a, a->from, s2);
571 freearc(nfa, a);
572 }
573 }
574 s2 = s->tmp;
575 s->tmp = NULL; /* clean up while we're at it */
576 }
577 }
578
579 /*
580 - parse - parse an RE
581 * This is actually just the top level, which parses a bunch of branches
582 * tied together with '|'. They appear in the tree as the left children
583 * of a chain of '|' subres.
584 ^ static struct subre *parse(struct vars *, int, int, struct state *,
585 ^ struct state *);
586 */
587 static struct subre *
588 parse(v, stopper, type, init, final)
589 struct vars *v;
590 int stopper; /* EOS or ')' */
591 int type; /* LACON (lookahead subRE) or PLAIN */
592 struct state *init; /* initial state */
593 struct state *final; /* final state */
594 {
595 struct state *left; /* scaffolding for branch */
596 struct state *right;
597 struct subre *branches; /* top level */
598 struct subre *branch; /* current branch */
599 struct subre *t; /* temporary */
600 int firstbranch; /* is this the first branch? */
601
602 assert(stopper == ')' || stopper == EOS);
603
604 branches = subre(v, '|', LONGER, init, final);
605 NOERRN();
606 branch = branches;
607 firstbranch = 1;
608 do { /* a branch */
609 if (!firstbranch) {
610 /* need a place to hang it */
611 branch->right = subre(v, '|', LONGER, init, final);
612 NOERRN();
613 branch = branch->right;
614 }
615 firstbranch = 0;
616 left = newstate(v->nfa);
617 right = newstate(v->nfa);
618 NOERRN();
619 EMPTYARC(init, left);
620 EMPTYARC(right, final);
621 NOERRN();
622 branch->left = parsebranch(v, stopper, type, left, right, 0);
623 NOERRN();
624 branch->flags |= UP(branch->flags | branch->left->flags);
625 if ((branch->flags &~ branches->flags) != 0) /* new flags */
626 for (t = branches; t != branch; t = t->right)
627 t->flags |= branch->flags;
628 } while (EAT('|'));
629 assert(SEE(stopper) || SEE(EOS));
630
631 if (!SEE(stopper)) {
632 assert(stopper == ')' && SEE(EOS));
633 ERR(REG_EPAREN);
634 }
635
636 /* optimize out simple cases */
637 if (branch == branches) { /* only one branch */
638 assert(branch->right == NULL);
639 t = branch->left;
640 branch->left = NULL;
641 freesubre(v, branches);
642 branches = t;
643 } else if (!MESSY(branches->flags)) { /* no interesting innards */
644 freesubre(v, branches->left);
645 branches->left = NULL;
646 freesubre(v, branches->right);
647 branches->right = NULL;
648 branches->op = '=';
649 }
650
651 return branches;
652 }
653
654 /*
655 - parsebranch - parse one branch of an RE
656 * This mostly manages concatenation, working closely with parseqatom().
657 * Concatenated things are bundled up as much as possible, with separate
658 * ',' nodes introduced only when necessary due to substructure.
659 ^ static struct subre *parsebranch(struct vars *, int, int, struct state *,
660 ^ struct state *, int);
661 */
662 static struct subre *
663 parsebranch(v, stopper, type, left, right, partial)
664 struct vars *v;
665 int stopper; /* EOS or ')' */
666 int type; /* LACON (lookahead subRE) or PLAIN */
667 struct state *left; /* leftmost state */
668 struct state *right; /* rightmost state */
669 int partial; /* is this only part of a branch? */
670 {
671 struct state *lp; /* left end of current construct */
672 int seencontent; /* is there anything in this branch yet? */
673 struct subre *t;
674
675 lp = left;
676 seencontent = 0;
677 t = subre(v, '=', 0, left, right); /* op '=' is tentative */
678 NOERRN();
679 while (!SEE('|') && !SEE(stopper) && !SEE(EOS)) {
680 if (seencontent) { /* implicit concat operator */
681 lp = newstate(v->nfa);
682 NOERRN();
683 moveins(v->nfa, right, lp);
684 }
685 seencontent = 1;
686
687 /* NB, recursion in parseqatom() may swallow rest of branch */
688 parseqatom(v, stopper, type, lp, right, t);
689 }
690
691 if (!seencontent) { /* empty branch */
692 if (!partial)
693 NOTE(REG_UUNSPEC);
694 assert(lp == left);
695 EMPTYARC(left, right);
696 }
697
698 return t;
699 }
700
701 /*
702 - parseqatom - parse one quantified atom or constraint of an RE
703 * The bookkeeping near the end cooperates very closely with parsebranch();
704 * in particular, it contains a recursion that can involve parsing the rest
705 * of the branch, making this function's name somewhat inaccurate.
706 ^ static VOID parseqatom(struct vars *, int, int, struct state *,
707 ^ struct state *, struct subre *);
708 */
709 static VOID
710 parseqatom(v, stopper, type, lp, rp, top)
711 struct vars *v;
712 int stopper; /* EOS or ')' */
713 int type; /* LACON (lookahead subRE) or PLAIN */
714 struct state *lp; /* left state to hang it on */
715 struct state *rp; /* right state to hang it on */
716 struct subre *top; /* subtree top */
717 {
718 struct state *s; /* temporaries for new states */
719 struct state *s2;
720 # define ARCV(t, val) newarc(v->nfa, t, val, lp, rp)
721 int m, n;
722 struct subre *atom; /* atom's subtree */
723 struct subre *t;
724 int cap; /* capturing parens? */
725 int pos; /* positive lookahead? */
726 int subno; /* capturing-parens or backref number */
727 int atomtype;
728 int qprefer; /* quantifier short/long preference */
729 int f;
730 struct subre **atomp; /* where the pointer to atom is */
731
732 /* initial bookkeeping */
733 atom = NULL;
734 assert(lp->nouts == 0); /* must string new code */
735 assert(rp->nins == 0); /* between lp and rp */
736 subno = 0; /* just to shut lint up */
737
738 /* an atom or constraint... */
739 atomtype = v->nexttype;
740 switch (atomtype) {
741 /* first, constraints, which end by returning */
742 case '^':
743 ARCV('^', 1);
744 if (v->cflags&REG_NLANCH)
745 ARCV(BEHIND, v->nlcolor);
746 NEXT();
747 return;
748 break;
749 case '$':
750 ARCV('$', 1);
751 if (v->cflags&REG_NLANCH)
752 ARCV(AHEAD, v->nlcolor);
753 NEXT();
754 return;
755 break;
756 case SBEGIN:
757 ARCV('^', 1); /* BOL */
758 ARCV('^', 0); /* or BOS */
759 NEXT();
760 return;
761 break;
762 case SEND:
763 ARCV('$', 1); /* EOL */
764 ARCV('$', 0); /* or EOS */
765 NEXT();
766 return;
767 break;
768 case '<':
769 wordchrs(v); /* does NEXT() */
770 s = newstate(v->nfa);
771 NOERR();
772 nonword(v, BEHIND, lp, s);
773 word(v, AHEAD, s, rp);
774 return;
775 break;
776 case '>':
777 wordchrs(v); /* does NEXT() */
778 s = newstate(v->nfa);
779 NOERR();
780 word(v, BEHIND, lp, s);
781 nonword(v, AHEAD, s, rp);
782 return;
783 break;
784 case WBDRY:
785 wordchrs(v); /* does NEXT() */
786 s = newstate(v->nfa);
787 NOERR();
788 nonword(v, BEHIND, lp, s);
789 word(v, AHEAD, s, rp);
790 s = newstate(v->nfa);
791 NOERR();
792 word(v, BEHIND, lp, s);
793 nonword(v, AHEAD, s, rp);
794 return;
795 break;
796 case NWBDRY:
797 wordchrs(v); /* does NEXT() */
798 s = newstate(v->nfa);
799 NOERR();
800 word(v, BEHIND, lp, s);
801 word(v, AHEAD, s, rp);
802 s = newstate(v->nfa);
803 NOERR();
804 nonword(v, BEHIND, lp, s);
805 nonword(v, AHEAD, s, rp);
806 return;
807 break;
808 case LACON: /* lookahead constraint */
809 pos = v->nextvalue;
810 NEXT();
811 s = newstate(v->nfa);
812 s2 = newstate(v->nfa);
813 NOERR();
814 t = parse(v, ')', LACON, s, s2);
815 freesubre(v, t); /* internal structure irrelevant */
816 assert(SEE(')') || ISERR());
817 NEXT();
818 n = newlacon(v, s, s2, pos);
819 NOERR();
820 ARCV(LACON, n);
821 return;
822 break;
823 /* then errors, to get them out of the way */
824 case '*':
825 case '+':
826 case '?':
827 case '{':
828 ERR(REG_BADRPT);
829 return;
830 break;
831 default:
832 ERR(REG_ASSERT);
833 return;
834 break;
835 /* then plain characters, and minor variants on that theme */
836 case ')': /* unbalanced paren */
837 if ((v->cflags&REG_ADVANCED) != REG_EXTENDED) {
838 ERR(REG_EPAREN);
839 return;
840 }
841 /* legal in EREs due to specification botch */
842 NOTE(REG_UPBOTCH);
843 /* fallthrough into case PLAIN */
844 case PLAIN:
845 onechr(v, v->nextvalue, lp, rp);
846 okcolors(v->nfa, v->cm);
847 NOERR();
848 NEXT();
849 break;
850 case '[':
851 if (v->nextvalue == 1)
852 bracket(v, lp, rp);
853 else
854 cbracket(v, lp, rp);
855 assert(SEE(']') || ISERR());
856 NEXT();
857 break;
858 case '.':
859 rainbow(v->nfa, v->cm, PLAIN,
860 (v->cflags&REG_NLSTOP) ? v->nlcolor : COLORLESS,
861 lp, rp);
862 NEXT();
863 break;
864 /* and finally the ugly stuff */
865 case '(': /* value flags as capturing or non */
866 cap = (type == LACON) ? 0 : v->nextvalue;
867 if (cap) {
868 v->nsubexp++;
869 subno = v->nsubexp;
870 if ((size_t)subno >= v->nsubs)
871 moresubs(v, subno);
872 assert((size_t)subno < v->nsubs);
873 } else
874 atomtype = PLAIN; /* something that's not '(' */
875 NEXT();
876 /* need new endpoints because tree will contain pointers */
877 s = newstate(v->nfa);
878 s2 = newstate(v->nfa);
879 NOERR();
880 EMPTYARC(lp, s);
881 EMPTYARC(s2, rp);
882 NOERR();
883 atom = parse(v, ')', PLAIN, s, s2);
884 assert(SEE(')') || ISERR());
885 NEXT();
886 NOERR();
887 if (cap) {
888 v->subs[subno] = atom;
889 t = subre(v, '(', atom->flags|CAP, lp, rp);
890 NOERR();
891 t->subno = subno;
892 t->left = atom;
893 atom = t;
894 }
895 /* postpone everything else pending possible {0} */
896 break;
897 case BACKREF: /* the Feature From The Black Lagoon */
898 INSIST(type != LACON, REG_ESUBREG);
899 INSIST(v->nextvalue < v->nsubs, REG_ESUBREG);
900 INSIST(v->subs[v->nextvalue] != NULL, REG_ESUBREG);
901 NOERR();
902 assert(v->nextvalue > 0);
903 atom = subre(v, 'b', BACKR, lp, rp);
904 subno = v->nextvalue;
905 atom->subno = subno;
906 EMPTYARC(lp, rp); /* temporarily, so there's something */
907 NEXT();
908 break;
909 }
910
911 /* ...and an atom may be followed by a quantifier */
912 switch (v->nexttype) {
913 case '*':
914 m = 0;
915 n = INFINITY;
916 qprefer = (v->nextvalue) ? LONGER : SHORTER;
917 NEXT();
918 break;
919 case '+':
920 m = 1;
921 n = INFINITY;
922 qprefer = (v->nextvalue) ? LONGER : SHORTER;
923 NEXT();
924 break;
925 case '?':
926 m = 0;
927 n = 1;
928 qprefer = (v->nextvalue) ? LONGER : SHORTER;
929 NEXT();
930 break;
931 case '{':
932 NEXT();
933 m = scannum(v);
934 if (EAT(',')) {
935 if (SEE(DIGIT))
936 n = scannum(v);
937 else
938 n = INFINITY;
939 if (m > n) {
940 ERR(REG_BADBR);
941 return;
942 }
943 /* {m,n} exercises preference, even if it's {m,m} */
944 qprefer = (v->nextvalue) ? LONGER : SHORTER;
945 } else {
946 n = m;
947 /* {m} passes operand's preference through */
948 qprefer = 0;
949 }
950 if (!SEE('}')) { /* catches errors too */
951 ERR(REG_BADBR);
952 return;
953 }
954 NEXT();
955 break;
956 default: /* no quantifier */
957 m = n = 1;
958 qprefer = 0;
959 break;
960 }
961
962 /* annoying special case: {0} or {0,0} cancels everything */
963 if (m == 0 && n == 0) {
964 if (atom != NULL)
965 freesubre(v, atom);
966 if (atomtype == '(')
967 v->subs[subno] = NULL;
968 delsub(v->nfa, lp, rp);
969 EMPTYARC(lp, rp);
970 return;
971 }
972
973 /* if not a messy case, avoid hard part */
974 assert(!MESSY(top->flags));
975 f = top->flags | qprefer | ((atom != NULL) ? atom->flags : 0);
976 if (atomtype != '(' && atomtype != BACKREF && !MESSY(UP(f))) {
977 if (!(m == 1 && n == 1))
978 repeat(v, lp, rp, m, n);
979 if (atom != NULL)
980 freesubre(v, atom);
981 top->flags = f;
982 return;
983 }
984
985 /*
986 * hard part: something messy
987 * That is, capturing parens, back reference, short/long clash, or
988 * an atom with substructure containing one of those.
989 */
990
991 /* now we'll need a subre for the contents even if they're boring */
992 if (atom == NULL) {
993 atom = subre(v, '=', 0, lp, rp);
994 NOERR();
995 }
996
997 /*
998 * prepare a general-purpose state skeleton
999 *
1000 * ---> [s] ---prefix---> [begin] ---atom---> [end] ----rest---> [rp]
1001 * / /
1002 * [lp] ----> [s2] ----bypass---------------------
1003 *
1004 * where bypass is an empty, and prefix is some repetitions of atom
1005 */
1006 s = newstate(v->nfa); /* first, new endpoints for the atom */
1007 s2 = newstate(v->nfa);
1008 NOERR();
1009 moveouts(v->nfa, lp, s);
1010 moveins(v->nfa, rp, s2);
1011 NOERR();
1012 atom->begin = s;
1013 atom->end = s2;
1014 s = newstate(v->nfa); /* and spots for prefix and bypass */
1015 s2 = newstate(v->nfa);
1016 NOERR();
1017 EMPTYARC(lp, s);
1018 EMPTYARC(lp, s2);
1019 NOERR();
1020
1021 /* break remaining subRE into x{...} and what follows */
1022 t = subre(v, '.', COMBINE(qprefer, atom->flags), lp, rp);
1023 t->left = atom;
1024 atomp = &t->left;
1025 /* here we should recurse... but we must postpone that to the end */
1026
1027 /* split top into prefix and remaining */
1028 assert(top->op == '=' && top->left == NULL && top->right == NULL);
1029 top->left = subre(v, '=', top->flags, top->begin, lp);
1030 top->op = '.';
1031 top->right = t;
1032
1033 /* if it's a backref, now is the time to replicate the subNFA */
1034 if (atomtype == BACKREF) {
1035 assert(atom->begin->nouts == 1); /* just the EMPTY */
1036 delsub(v->nfa, atom->begin, atom->end);
1037 assert(v->subs[subno] != NULL);
1038 /* and here's why the recursion got postponed: it must */
1039 /* wait until the skeleton is filled in, because it may */
1040 /* hit a backref that wants to copy the filled-in skeleton */
1041 dupnfa(v->nfa, v->subs[subno]->begin, v->subs[subno]->end,
1042 atom->begin, atom->end);
1043 NOERR();
1044 }
1045
1046 /* it's quantifier time; first, turn x{0,...} into x{1,...}|empty */
1047 if (m == 0) {
1048 EMPTYARC(s2, atom->end); /* the bypass */
1049 assert(PREF(qprefer) != 0);
1050 f = COMBINE(qprefer, atom->flags);
1051 t = subre(v, '|', f, lp, atom->end);
1052 NOERR();
1053 t->left = atom;
1054 t->right = subre(v, '|', PREF(f), s2, atom->end);
1055 NOERR();
1056 t->right->left = subre(v, '=', 0, s2, atom->end);
1057 NOERR();
1058 *atomp = t;
1059 atomp = &t->left;
1060 m = 1;
1061 }
1062
1063 /* deal with the rest of the quantifier */
1064 if (atomtype == BACKREF) {
1065 /* special case: backrefs have internal quantifiers */
1066 EMPTYARC(s, atom->begin); /* empty prefix */
1067 /* just stuff everything into atom */
1068 repeat(v, atom->begin, atom->end, m, n);
1069 atom->min = (short)m;
1070 atom->max = (short)n;
1071 atom->flags |= COMBINE(qprefer, atom->flags);
1072 } else if (m == 1 && n == 1) {
1073 /* no/vacuous quantifier: done */
1074 EMPTYARC(s, atom->begin); /* empty prefix */
1075 } else {
1076 /* turn x{m,n} into x{m-1,n-1}x, with capturing */
1077 /* parens in only second x */
1078 dupnfa(v->nfa, atom->begin, atom->end, s, atom->begin);
1079 assert(m >= 1 && m != INFINITY && n >= 1);
1080 repeat(v, s, atom->begin, m-1, (n == INFINITY) ? n : n-1);
1081 f = COMBINE(qprefer, atom->flags);
1082 t = subre(v, '.', f, s, atom->end); /* prefix and atom */
1083 NOERR();
1084 t->left = subre(v, '=', PREF(f), s, atom->begin);
1085 NOERR();
1086 t->right = atom;
1087 *atomp = t;
1088 }
1089
1090 /* and finally, look after that postponed recursion */
1091 t = top->right;
1092 if (!(SEE('|') || SEE(stopper) || SEE(EOS)))
1093 t->right = parsebranch(v, stopper, type, atom->end, rp, 1);
1094 else {
1095 EMPTYARC(atom->end, rp);
1096 t->right = subre(v, '=', 0, atom->end, rp);
1097 }
1098 assert(SEE('|') || SEE(stopper) || SEE(EOS));
1099 t->flags |= COMBINE(t->flags, t->right->flags);
1100 top->flags |= COMBINE(top->flags, t->flags);
1101 }
1102
1103 /*
1104 - nonword - generate arcs for non-word-character ahead or behind
1105 ^ static VOID nonword(struct vars *, int, struct state *, struct state *);
1106 */
1107 static VOID
1108 nonword(v, dir, lp, rp)
1109 struct vars *v;
1110 int dir; /* AHEAD or BEHIND */
1111 struct state *lp;
1112 struct state *rp;
1113 {
1114 int anchor = (dir == AHEAD) ? '$' : '^';
1115
1116 assert(dir == AHEAD || dir == BEHIND);
1117 newarc(v->nfa, anchor, 1, lp, rp);
1118 newarc(v->nfa, anchor, 0, lp, rp);
1119 colorcomplement(v->nfa, v->cm, dir, v->wordchrs, lp, rp);
1120 /* (no need for special attention to \n) */
1121 }
1122
1123 /*
1124 - word - generate arcs for word character ahead or behind
1125 ^ static VOID word(struct vars *, int, struct state *, struct state *);
1126 */
1127 static VOID
1128 word(v, dir, lp, rp)
1129 struct vars *v;
1130 int dir; /* AHEAD or BEHIND */
1131 struct state *lp;
1132 struct state *rp;
1133 {
1134 assert(dir == AHEAD || dir == BEHIND);
1135 cloneouts(v->nfa, v->wordchrs, lp, rp, dir);
1136 /* (no need for special attention to \n) */
1137 }
1138
1139 /*
1140 - scannum - scan a number
1141 ^ static int scannum(struct vars *);
1142 */
1143 static int /* value, <= DUPMAX */
1144 scannum(v)
1145 struct vars *v;
1146 {
1147 int n = 0;
1148
1149 while (SEE(DIGIT) && n < DUPMAX) {
1150 n = n*10 + v->nextvalue;
1151 NEXT();
1152 }
1153 if (SEE(DIGIT) || n > DUPMAX) {
1154 ERR(REG_BADBR);
1155 return 0;
1156 }
1157 return n;
1158 }
1159
1160 /*
1161 - repeat - replicate subNFA for quantifiers
1162 * The duplication sequences used here are chosen carefully so that any
1163 * pointers starting out pointing into the subexpression end up pointing into
1164 * the last occurrence. (Note that it may not be strung between the same
1165 * left and right end states, however!) This used to be important for the
1166 * subRE tree, although the important bits are now handled by the in-line
1167 * code in parse(), and when this is called, it doesn't matter any more.
1168 ^ static VOID repeat(struct vars *, struct state *, struct state *, int, int);
1169 */
1170 static VOID
1171 repeat(v, lp, rp, m, n)
1172 struct vars *v;
1173 struct state *lp;
1174 struct state *rp;
1175 int m;
1176 int n;
1177 {
1178 # define SOME 2
1179 # define INF 3
1180 # define PAIR(x, y) ((x)*4 + (y))
1181 # define REDUCE(x) ( ((x) == INFINITY) ? INF : (((x) > 1) ? SOME : (x)) )
1182 CONST int rm = REDUCE(m);
1183 CONST int rn = REDUCE(n);
1184 struct state *s;
1185 struct state *s2;
1186
1187 switch (PAIR(rm, rn)) {
1188 case PAIR(0, 0): /* empty string */
1189 delsub(v->nfa, lp, rp);
1190 EMPTYARC(lp, rp);
1191 break;
1192 case PAIR(0, 1): /* do as x| */
1193 EMPTYARC(lp, rp);
1194 break;
1195 case PAIR(0, SOME): /* do as x{1,n}| */
1196 repeat(v, lp, rp, 1, n);
1197 NOERR();
1198 EMPTYARC(lp, rp);
1199 break;
1200 case PAIR(0, INF): /* loop x around */
1201 s = newstate(v->nfa);
1202 NOERR();
1203 moveouts(v->nfa, lp, s);
1204 moveins(v->nfa, rp, s);
1205 EMPTYARC(lp, s);
1206 EMPTYARC(s, rp);
1207 break;
1208 case PAIR(1, 1): /* no action required */
1209 break;
1210 case PAIR(1, SOME): /* do as x{0,n-1}x = (x{1,n-1}|)x */
1211 s = newstate(v->nfa);
1212 NOERR();
1213 moveouts(v->nfa, lp, s);
1214 dupnfa(v->nfa, s, rp, lp, s);
1215 NOERR();
1216 repeat(v, lp, s, 1, n-1);
1217 NOERR();
1218 EMPTYARC(lp, s);
1219 break;
1220 case PAIR(1, INF): /* add loopback arc */
1221 s = newstate(v->nfa);
1222 s2 = newstate(v->nfa);
1223 NOERR();
1224 moveouts(v->nfa, lp, s);
1225 moveins(v->nfa, rp, s2);
1226 EMPTYARC(lp, s);
1227 EMPTYARC(s2, rp);
1228 EMPTYARC(s2, s);
1229 break;
1230 case PAIR(SOME, SOME): /* do as x{m-1,n-1}x */
1231 s = newstate(v->nfa);
1232 NOERR();
1233 moveouts(v->nfa, lp, s);
1234 dupnfa(v->nfa, s, rp, lp, s);
1235 NOERR();
1236 repeat(v, lp, s, m-1, n-1);
1237 break;
1238 case PAIR(SOME, INF): /* do as x{m-1,}x */
1239 s = newstate(v->nfa);
1240 NOERR();
1241 moveouts(v->nfa, lp, s);
1242 dupnfa(v->nfa, s, rp, lp, s);
1243 NOERR();
1244 repeat(v, lp, s, m-1, n);
1245 break;
1246 default:
1247 ERR(REG_ASSERT);
1248 break;
1249 }
1250 }
1251
1252 /*
1253 - bracket - handle non-complemented bracket expression
1254 * Also called from cbracket for complemented bracket expressions.
1255 ^ static VOID bracket(struct vars *, struct state *, struct state *);
1256 */
1257 static VOID
1258 bracket(v, lp, rp)
1259 struct vars *v;
1260 struct state *lp;
1261 struct state *rp;
1262 {
1263 assert(SEE('['));
1264 NEXT();
1265 while (!SEE(']') && !SEE(EOS))
1266 brackpart(v, lp, rp);
1267 assert(SEE(']') || ISERR());
1268 okcolors(v->nfa, v->cm);
1269 }
1270
1271 /*
1272 - cbracket - handle complemented bracket expression
1273 * We do it by calling bracket() with dummy endpoints, and then complementing
1274 * the result. The alternative would be to invoke rainbow(), and then delete
1275 * arcs as the b.e. is seen... but that gets messy.
1276 ^ static VOID cbracket(struct vars *, struct state *, struct state *);
1277 */
1278 static VOID
1279 cbracket(v, lp, rp)
1280 struct vars *v;
1281 struct state *lp;
1282 struct state *rp;
1283 {
1284 struct state *left = newstate(v->nfa);
1285 struct state *right = newstate(v->nfa);
1286 struct state *s;
1287 struct arc *a; /* arc from lp */
1288 struct arc *ba; /* arc from left, from bracket() */
1289 struct arc *pa; /* MCCE-prototype arc */
1290 color co;
1291 chr *p;
1292 int i;
1293
1294 NOERR();
1295 bracket(v, left, right);
1296 if (v->cflags&REG_NLSTOP)
1297 newarc(v->nfa, PLAIN, v->nlcolor, left, right);
1298 NOERR();
1299
1300 assert(lp->nouts == 0); /* all outarcs will be ours */
1301
1302 /* easy part of complementing */
1303 colorcomplement(v->nfa, v->cm, PLAIN, left, lp, rp);
1304 NOERR();
1305 if (v->mcces == NULL) { /* no MCCEs -- we're done */
1306 dropstate(v->nfa, left);
1307 assert(right->nins == 0);
1308 freestate(v->nfa, right);
1309 return;
1310 }
1311
1312 /* but complementing gets messy in the presence of MCCEs... */
1313 NOTE(REG_ULOCALE);
1314 for (p = v->mcces->chrs, i = v->mcces->nchrs; i > 0; p++, i--) {
1315 co = GETCOLOR(v->cm, *p);
1316 a = findarc(lp, PLAIN, co);
1317 ba = findarc(left, PLAIN, co);
1318 if (ba == NULL) {
1319 assert(a != NULL);
1320 freearc(v->nfa, a);
1321 } else {
1322 assert(a == NULL);
1323 }
1324 s = newstate(v->nfa);
1325 NOERR();
1326 newarc(v->nfa, PLAIN, co, lp, s);
1327 NOERR();
1328 pa = findarc(v->mccepbegin, PLAIN, co);
1329 assert(pa != NULL);
1330 if (ba == NULL) { /* easy case, need all of them */
1331 cloneouts(v->nfa, pa->to, s, rp, PLAIN);
1332 newarc(v->nfa, '$', 1, s, rp);
1333 newarc(v->nfa, '$', 0, s, rp);
1334 colorcomplement(v->nfa, v->cm, AHEAD, pa->to, s, rp);
1335 } else { /* must be selective */
1336 if (findarc(ba->to, '$', 1) == NULL) {
1337 newarc(v->nfa, '$', 1, s, rp);
1338 newarc(v->nfa, '$', 0, s, rp);
1339 colorcomplement(v->nfa, v->cm, AHEAD, pa->to,
1340 s, rp);
1341 }
1342 for (pa = pa->to->outs; pa != NULL; pa = pa->outchain)
1343 if (findarc(ba->to, PLAIN, pa->co) == NULL)
1344 newarc(v->nfa, PLAIN, pa->co, s, rp);
1345 if (s->nouts == 0) /* limit of selectivity: none */
1346 dropstate(v->nfa, s); /* frees arc too */
1347 }
1348 NOERR();
1349 }
1350
1351 delsub(v->nfa, left, right);
1352 assert(left->nouts == 0);
1353 freestate(v->nfa, left);
1354 assert(right->nins == 0);
1355 freestate(v->nfa, right);
1356 }
1357
1358 /*
1359 - brackpart - handle one item (or range) within a bracket expression
1360 ^ static VOID brackpart(struct vars *, struct state *, struct state *);
1361 */
1362 static VOID
1363 brackpart(v, lp, rp)
1364 struct vars *v;
1365 struct state *lp;
1366 struct state *rp;
1367 {
1368 celt startc;
1369 celt endc;
1370 struct cvec *cv;
1371 chr *startp;
1372 chr *endp;
1373 chr c[1];
1374
1375 /* parse something, get rid of special cases, take shortcuts */
1376 switch (v->nexttype) {
1377 case RANGE: /* a-b-c or other botch */
1378 ERR(REG_ERANGE);
1379 return;
1380 break;
1381 case PLAIN:
1382 c[0] = v->nextvalue;
1383 NEXT();
1384 /* shortcut for ordinary chr (not range, not MCCE leader) */
1385 if (!SEE(RANGE) && !ISCELEADER(v, c[0])) {
1386 onechr(v, c[0], lp, rp);
1387 return;
1388 }
1389 startc = element(v, c, c+1);
1390 NOERR();
1391 break;
1392 case COLLEL:
1393 startp = v->now;
1394 endp = scanplain(v);
1395 INSIST(startp < endp, REG_ECOLLATE);
1396 NOERR();
1397 startc = element(v, startp, endp);
1398 NOERR();
1399 break;
1400 case ECLASS:
1401 startp = v->now;
1402 endp = scanplain(v);
1403 INSIST(startp < endp, REG_ECOLLATE);
1404 NOERR();
1405 startc = element(v, startp, endp);
1406 NOERR();
1407 cv = eclass(v, startc, (v->cflags&REG_ICASE));
1408 NOERR();
1409 dovec(v, cv, lp, rp);
1410 return;
1411 break;
1412 case CCLASS:
1413 startp = v->now;
1414 endp = scanplain(v);
1415 INSIST(startp < endp, REG_ECTYPE);
1416 NOERR();
1417 cv = cclass(v, startp, endp, (v->cflags&REG_ICASE));
1418 NOERR();
1419 dovec(v, cv, lp, rp);
1420 return;
1421 break;
1422 default:
1423 ERR(REG_ASSERT);
1424 return;
1425 break;
1426 }
1427
1428 if (SEE(RANGE)) {
1429 NEXT();
1430 switch (v->nexttype) {
1431 case PLAIN:
1432 case RANGE:
1433 c[0] = v->nextvalue;
1434 NEXT();
1435 endc = element(v, c, c+1);
1436 NOERR();
1437 break;
1438 case COLLEL:
1439 startp = v->now;
1440 endp = scanplain(v);
1441 INSIST(startp < endp, REG_ECOLLATE);
1442 NOERR();
1443 endc = element(v, startp, endp);
1444 NOERR();
1445 break;
1446 default:
1447 ERR(REG_ERANGE);
1448 return;
1449 break;
1450 }
1451 } else
1452 endc = startc;
1453
1454 /*
1455 * Ranges are unportable. Actually, standard C does
1456 * guarantee that digits are contiguous, but making
1457 * that an exception is just too complicated.
1458 */
1459 if (startc != endc)
1460 NOTE(REG_UUNPORT);
1461 cv = range(v, startc, endc, (v->cflags&REG_ICASE));
1462 NOERR();
1463 dovec(v, cv, lp, rp);
1464 }
1465
1466 /*
1467 - scanplain - scan PLAIN contents of [. etc.
1468 * Certain bits of trickery in lex.c know that this code does not try
1469 * to look past the final bracket of the [. etc.
1470 ^ static chr *scanplain(struct vars *);
1471 */
1472 static chr * /* just after end of sequence */
1473 scanplain(v)
1474 struct vars *v;
1475 {
1476 chr *endp;
1477
1478 assert(SEE(COLLEL) || SEE(ECLASS) || SEE(CCLASS));
1479 NEXT();
1480
1481 endp = v->now;
1482 while (SEE(PLAIN)) {
1483 endp = v->now;
1484 NEXT();
1485 }
1486
1487 assert(SEE(END) || ISERR());
1488 NEXT();
1489
1490 return endp;
1491 }
1492
1493 /*
1494 - leaders - process a cvec of collating elements to also include leaders
1495 * Also gives all characters involved their own colors, which is almost
1496 * certainly necessary, and sets up little disconnected subNFA.
1497 ^ static VOID leaders(struct vars *, struct cvec *);
1498 */
1499 static VOID
1500 leaders(v, cv)
1501 struct vars *v;
1502 struct cvec *cv;
1503 {
1504 int mcce;
1505 chr *p;
1506 chr leader;
1507 struct state *s;
1508 struct arc *a;
1509
1510 v->mccepbegin = newstate(v->nfa);
1511 v->mccepend = newstate(v->nfa);
1512 NOERR();
1513
1514 for (mcce = 0; mcce < cv->nmcces; mcce++) {
1515 p = cv->mcces[mcce];
1516 leader = *p;
1517 if (!haschr(cv, leader)) {
1518 addchr(cv, leader);
1519 s = newstate(v->nfa);
1520 newarc(v->nfa, PLAIN, subcolor(v->cm, leader),
1521 v->mccepbegin, s);
1522 okcolors(v->nfa, v->cm);
1523 } else {
1524 a = findarc(v->mccepbegin, PLAIN,
1525 GETCOLOR(v->cm, leader));
1526 assert(a != NULL);
1527 s = a->to;
1528 assert(s != v->mccepend);
1529 }
1530 p++;
1531 assert(*p != 0 && *(p+1) == 0); /* only 2-char MCCEs for now */
1532 newarc(v->nfa, PLAIN, subcolor(v->cm, *p), s, v->mccepend);
1533 okcolors(v->nfa, v->cm);
1534 }
1535 }
1536
1537 /*
1538 - onechr - fill in arcs for a plain character, and possible case complements
1539 * This is mostly a shortcut for efficient handling of the common case.
1540 ^ static VOID onechr(struct vars *, pchr, struct state *, struct state *);
1541 */
1542 static VOID
1543 onechr(v, c, lp, rp)
1544 struct vars *v;
1545 pchr c;
1546 struct state *lp;
1547 struct state *rp;
1548 {
1549 if (!(v->cflags&REG_ICASE)) {
1550 newarc(v->nfa, PLAIN, subcolor(v->cm, c), lp, rp);
1551 return;
1552 }
1553
1554 /* rats, need general case anyway... */
1555 dovec(v, allcases(v, c), lp, rp);
1556 }
1557
1558 /*
1559 - dovec - fill in arcs for each element of a cvec
1560 * This one has to handle the messy cases, like MCCEs and MCCE leaders.
1561 ^ static VOID dovec(struct vars *, struct cvec *, struct state *,
1562 ^ struct state *);
1563 */
1564 static VOID
1565 dovec(v, cv, lp, rp)
1566 struct vars *v;
1567 struct cvec *cv;
1568 struct state *lp;
1569 struct state *rp;
1570 {
1571 chr ch, from, to;
1572 celt ce;
1573 chr *p;
1574 int i;
1575 color co;
1576 struct cvec *leads;
1577 struct arc *a;
1578 struct arc *pa; /* arc in prototype */
1579 struct state *s;
1580 struct state *ps; /* state in prototype */
1581
1582 /* need a place to store leaders, if any */
1583 if (nmcces(v) > 0) {
1584 assert(v->mcces != NULL);
1585 if (v->cv2 == NULL || v->cv2->nchrs < v->mcces->nchrs) {
1586 if (v->cv2 != NULL)
1587 free(v->cv2);
1588 v->cv2 = newcvec(v->mcces->nchrs, 0, v->mcces->nmcces);
1589 NOERR();
1590 leads = v->cv2;
1591 } else
1592 leads = clearcvec(v->cv2);
1593 } else
1594 leads = NULL;
1595
1596 /* first, get the ordinary characters out of the way */
1597 for (p = cv->chrs, i = cv->nchrs; i > 0; p++, i--) {
1598 ch = *p;
1599 if (!ISCELEADER(v, ch))
1600 newarc(v->nfa, PLAIN, subcolor(v->cm, ch), lp, rp);
1601 else {
1602 assert(singleton(v->cm, ch));
1603 assert(leads != NULL);
1604 if (!haschr(leads, ch))
1605 addchr(leads, ch);
1606 }
1607 }
1608
1609 /* and the ranges */
1610 for (p = cv->ranges, i = cv->nranges; i > 0; p += 2, i--) {
1611 from = *p;
1612 to = *(p+1);
1613 while (from <= to && (ce = nextleader(v, from, to)) != NOCELT) {
1614 if (from < ce)
1615 subrange(v, from, ce - 1, lp, rp);
1616 assert(singleton(v->cm, ce));
1617 assert(leads != NULL);
1618 if (!haschr(leads, ce))
1619 addchr(leads, ce);
1620 from = ce + 1;
1621 }
1622 if (from <= to)
1623 subrange(v, from, to, lp, rp);
1624 }
1625
1626 if ((leads == NULL || leads->nchrs == 0) && cv->nmcces == 0)
1627 return;
1628
1629 /* deal with the MCCE leaders */
1630 NOTE(REG_ULOCALE);
1631 for (p = leads->chrs, i = leads->nchrs; i > 0; p++, i--) {
1632 co = GETCOLOR(v->cm, *p);
1633 a = findarc(lp, PLAIN, co);
1634 if (a != NULL)
1635 s = a->to;
1636 else {
1637 s = newstate(v->nfa);
1638 NOERR();
1639 newarc(v->nfa, PLAIN, co, lp, s);
1640 NOERR();
1641 }
1642 pa = findarc(v->mccepbegin, PLAIN, co);
1643 assert(pa != NULL);
1644 ps = pa->to;
1645 newarc(v->nfa, '$', 1, s, rp);
1646 newarc(v->nfa, '$', 0, s, rp);
1647 colorcomplement(v->nfa, v->cm, AHEAD, ps, s, rp);
1648 NOERR();
1649 }
1650
1651 /* and the MCCEs */
1652 for (i = 0; i < cv->nmcces; i++) {
1653 p = cv->mcces[i];
1654 assert(singleton(v->cm, *p));
1655 if (!singleton(v->cm, *p)) {
1656 ERR(REG_ASSERT);
1657 return;
1658 }
1659 ch = *p++;
1660 co = GETCOLOR(v->cm, ch);
1661 a = findarc(lp, PLAIN, co);
1662 if (a != NULL)
1663 s = a->to;
1664 else {
1665 s = newstate(v->nfa);
1666 NOERR();
1667 newarc(v->nfa, PLAIN, co, lp, s);
1668 NOERR();
1669 }
1670 assert(*p != 0); /* at least two chars */
1671 assert(singleton(v->cm, *p));
1672 ch = *p++;
1673 co = GETCOLOR(v->cm, ch);
1674 assert(*p == 0); /* and only two, for now */
1675 newarc(v->nfa, PLAIN, co, s, rp);
1676 NOERR();
1677 }
1678 }
1679
1680 /*
1681 - nextleader - find next MCCE leader within range
1682 ^ static celt nextleader(struct vars *, pchr, pchr);
1683 */
1684 static celt /* NOCELT means none */
1685 nextleader(v, from, to)
1686 struct vars *v;
1687 pchr from;
1688 pchr to;
1689 {
1690 int i;
1691 chr *p;
1692 chr ch;
1693 celt it = NOCELT;
1694
1695 if (v->mcces == NULL)
1696 return it;
1697
1698 for (i = v->mcces->nchrs, p = v->mcces->chrs; i > 0; i--, p++) {
1699 ch = *p;
1700 if (from <= ch && ch <= to)
1701 if (it == NOCELT || ch < it)
1702 it = ch;
1703 }
1704 return it;
1705 }
1706
1707 /*
1708 - wordchrs - set up word-chr list for word-boundary stuff, if needed
1709 * The list is kept as a bunch of arcs between two dummy states; it's
1710 * disposed of by the unreachable-states sweep in NFA optimization.
1711 * Does NEXT(). Must not be called from any unusual lexical context.
1712 * This should be reconciled with the \w etc. handling in lex.c, and
1713 * should be cleaned up to reduce dependencies on input scanning.
1714 ^ static VOID wordchrs(struct vars *);
1715 */
1716 static VOID
1717 wordchrs(v)
1718 struct vars *v;
1719 {
1720 struct state *left;
1721 struct state *right;
1722
1723 if (v->wordchrs != NULL) {
1724 NEXT(); /* for consistency */
1725 return;
1726 }
1727
1728 left = newstate(v->nfa);
1729 right = newstate(v->nfa);
1730 NOERR();
1731 /* fine point: implemented with [::], and lexer will set REG_ULOCALE */
1732 lexword(v);
1733 NEXT();
1734 assert(v->savenow != NULL && SEE('['));
1735 bracket(v, left, right);
1736 assert((v->savenow != NULL && SEE(']')) || ISERR());
1737 NEXT();
1738 NOERR();
1739 v->wordchrs = left;
1740 }
1741
1742 /*
1743 - subre - allocate a subre
1744 ^ static struct subre *subre(struct vars *, int, int, struct state *,
1745 ^ struct state *);
1746 */
1747 static struct subre *
1748 subre(v, op, flags, begin, end)
1749 struct vars *v;
1750 int op;
1751 int flags;
1752 struct state *begin;
1753 struct state *end;
1754 {
1755 struct subre *ret;
1756
1757 ret = v->treefree;
1758 if (ret != NULL)
1759 v->treefree = ret->left;
1760 else {
1761 ret = (struct subre *)MALLOC(sizeof(struct subre));
1762 if (ret == NULL) {
1763 ERR(REG_ESPACE);
1764 return NULL;
1765 }
1766 ret->chain = v->treechain;
1767 v->treechain = ret;
1768 }
1769
1770 assert(strchr("|.b(=", op) != NULL);
1771
1772 ret->op = op;
1773 ret->flags = flags;
1774 ret->retry = 0;
1775 ret->subno = 0;
1776 ret->min = ret->max = 1;
1777 ret->left = NULL;
1778 ret->right = NULL;
1779 ret->begin = begin;
1780 ret->end = end;
1781 ZAPCNFA(ret->cnfa);
1782
1783 return ret;
1784 }
1785
1786 /*
1787 - freesubre - free a subRE subtree
1788 ^ static VOID freesubre(struct vars *, struct subre *);
1789 */
1790 static VOID
1791 freesubre(v, sr)
1792 struct vars *v; /* might be NULL */
1793 struct subre *sr;
1794 {
1795 if (sr == NULL)
1796 return;
1797
1798 if (sr->left != NULL)
1799 freesubre(v, sr->left);
1800 if (sr->right != NULL)
1801 freesubre(v, sr->right);
1802
1803 freesrnode(v, sr);
1804 }
1805
1806 /*
1807 - freesrnode - free one node in a subRE subtree
1808 ^ static VOID freesrnode(struct vars *, struct subre *);
1809 */
1810 static VOID
1811 freesrnode(v, sr)
1812 struct vars *v; /* might be NULL */
1813 struct subre *sr;
1814 {
1815 if (sr == NULL)
1816 return;
1817
1818 if (!NULLCNFA(sr->cnfa))
1819 freecnfa(&sr->cnfa);
1820 sr->flags = 0;
1821
1822 if (v != NULL) {
1823 sr->left = v->treefree;
1824 v->treefree = sr;
1825 } else
1826 FREE(sr);
1827 }
1828
1829 /*
1830 - optst - optimize a subRE subtree
1831 ^ static VOID optst(struct vars *, struct subre *);
1832 */
1833 static VOID
1834 optst(v, t)
1835 struct vars *v;
1836 struct subre *t;
1837 {
1838 if (t == NULL)
1839 return;
1840
1841 /* recurse through children */
1842 if (t->left != NULL)
1843 optst(v, t->left);
1844 if (t->right != NULL)
1845 optst(v, t->right);
1846 }
1847
1848 /*
1849 - numst - number tree nodes (assigning retry indexes)
1850 ^ static int numst(struct subre *, int);
1851 */
1852 static int /* next number */
1853 numst(t, start)
1854 struct subre *t;
1855 int start; /* starting point for subtree numbers */
1856 {
1857 int i;
1858
1859 assert(t != NULL);
1860
1861 i = start;
1862 t->retry = (short)i++;
1863 if (t->left != NULL)
1864 i = numst(t->left, i);
1865 if (t->right != NULL)
1866 i = numst(t->right, i);
1867 return i;
1868 }
1869
1870 /*
1871 - markst - mark tree nodes as INUSE
1872 ^ static VOID markst(struct subre *);
1873 */
1874 static VOID
1875 markst(t)
1876 struct subre *t;
1877 {
1878 assert(t != NULL);
1879
1880 t->flags |= INUSE;
1881 if (t->left != NULL)
1882 markst(t->left);
1883 if (t->right != NULL)
1884 markst(t->right);
1885 }
1886
1887 /*
1888 - cleanst - free any tree nodes not marked INUSE
1889 ^ static VOID cleanst(struct vars *);
1890 */
1891 static VOID
1892 cleanst(v)
1893 struct vars *v;
1894 {
1895 struct subre *t;
1896 struct subre *next;
1897
1898 for (t = v->treechain; t != NULL; t = next) {
1899 next = t->chain;
1900 if (!(t->flags&INUSE))
1901 FREE(t);
1902 }
1903 v->treechain = NULL;
1904 v->treefree = NULL; /* just on general principles */
1905 }
1906
1907 /*
1908 - nfatree - turn a subRE subtree into a tree of compacted NFAs
1909 ^ static long nfatree(struct vars *, struct subre *, FILE *);
1910 */
1911 static long /* optimize results from top node */
1912 nfatree(v, t, f)
1913 struct vars *v;
1914 struct subre *t;
1915 FILE *f; /* for debug output */
1916 {
1917 assert(t != NULL && t->begin != NULL);
1918
1919 if (t->left != NULL)
1920 (DISCARD)nfatree(v, t->left, f);
1921 if (t->right != NULL)
1922 (DISCARD)nfatree(v, t->right, f);
1923
1924 return nfanode(v, t, f);
1925 }
1926
1927 /*
1928 - nfanode - do one NFA for nfatree
1929 ^ static long nfanode(struct vars *, struct subre *, FILE *);
1930 */
1931 static long /* optimize results */
1932 nfanode(v, t, f)
1933 struct vars *v;
1934 struct subre *t;
1935 FILE *f; /* for debug output */
1936 {
1937 struct nfa *nfa;
1938 long ret = 0;
1939 char idbuf[50];
1940
1941 assert(t->begin != NULL);
1942
1943 if (f != NULL)
1944 fprintf(f, "\n\n\n========= TREE NODE %s ==========\n",
1945 stid(t, idbuf, sizeof(idbuf)));
1946 nfa = newnfa(v, v->cm, v->nfa);
1947 NOERRZ();
1948 dupnfa(nfa, t->begin, t->end, nfa->init, nfa->final);
1949 if (!ISERR()) {
1950 specialcolors(nfa);
1951 ret = optimize(nfa, f);
1952 }
1953 if (!ISERR())
1954 compact(nfa, &t->cnfa);
1955
1956 freenfa(nfa);
1957 return ret;
1958 }
1959
1960 /*
1961 - newlacon - allocate a lookahead-constraint subRE
1962 ^ static int newlacon(struct vars *, struct state *, struct state *, int);
1963 */
1964 static int /* lacon number */
1965 newlacon(v, begin, end, pos)
1966 struct vars *v;
1967 struct state *begin;
1968 struct state *end;
1969 int pos;
1970 {
1971 int n;
1972 struct subre *sub;
1973
1974 if (v->nlacons == 0) {
1975 v->lacons = (struct subre *)MALLOC(2 * sizeof(struct subre));
1976 n = 1; /* skip 0th */
1977 v->nlacons = 2;
1978 } else {
1979 v->lacons = (struct subre *)REALLOC(v->lacons,
1980 (v->nlacons+1)*sizeof(struct subre));
1981 n = v->nlacons++;
1982 }
1983 if (v->lacons == NULL) {
1984 ERR(REG_ESPACE);
1985 return 0;
1986 }
1987 sub = &v->lacons[n];
1988 sub->begin = begin;
1989 sub->end = end;
1990 sub->subno = pos;
1991 ZAPCNFA(sub->cnfa);
1992 return n;
1993 }
1994
1995 /*
1996 - freelacons - free lookahead-constraint subRE vector
1997 ^ static VOID freelacons(struct subre *, int);
1998 */
1999 static VOID
2000 freelacons(subs, n)
2001 struct subre *subs;
2002 int n;
2003 {
2004 struct subre *sub;
2005 int i;
2006
2007 assert(n > 0);
2008 for (sub = subs + 1, i = n - 1; i > 0; sub++, i--) /* no 0th */
2009 if (!NULLCNFA(sub->cnfa))
2010 freecnfa(&sub->cnfa);
2011 FREE(subs);
2012 }
2013
2014 /*
2015 - rfree - free a whole RE (insides of regfree)
2016 ^ static VOID rfree(regex_t *);
2017 */
2018 static VOID
2019 rfree(re)
2020 regex_t *re;
2021 {
2022 struct guts *g;
2023
2024 if (re == NULL || re->re_magic != REMAGIC)
2025 return;
2026
2027 re->re_magic = 0; /* invalidate RE */
2028 g = (struct guts *)re->re_guts;
2029 re->re_guts = NULL;
2030 re->re_fns = NULL;
2031 g->magic = 0;
2032 freecm(&g->cmap);
2033 if (g->tree != NULL)
2034 freesubre((struct vars *)NULL, g->tree);
2035 if (g->lacons != NULL)
2036 freelacons(g->lacons, g->nlacons);
2037 if (!NULLCNFA(g->search))
2038 freecnfa(&g->search);
2039 FREE(g);
2040 }
2041
2042 /*
2043 - dump - dump an RE in human-readable form
2044 ^ static VOID dump(regex_t *, FILE *);
2045 */
2046 static VOID
2047 dump(re, f)
2048 regex_t *re;
2049 FILE *f;
2050 {
2051 #ifdef REG_DEBUG
2052 struct guts *g;
2053 int i;
2054
2055 if (re->re_magic != REMAGIC)
2056 fprintf(f, "bad magic number (0x%x not 0x%x)\n", re->re_magic,
2057 REMAGIC);
2058 if (re->re_guts == NULL) {
2059 fprintf(f, "NULL guts!!!\n");
2060 return;
2061 }
2062 g = (struct guts *)re->re_guts;
2063 if (g->magic != GUTSMAGIC)
2064 fprintf(f, "bad guts magic number (0x%x not 0x%x)\n", g->magic,
2065 GUTSMAGIC);
2066
2067 fprintf(f, "\n\n\n========= DUMP ==========\n");
2068 fprintf(f, "nsub %d, info 0%lo, csize %d, ntree %d\n",
2069 re->re_nsub, re->re_info, re->re_csize, g->ntree);
2070
2071 dumpcolors(&g->cmap, f);
2072 if (!NULLCNFA(g->search)) {
2073 printf("\nsearch:\n");
2074 dumpcnfa(&g->search, f);
2075 }
2076 for (i = 1; i < g->nlacons; i++) {
2077 fprintf(f, "\nla%d (%s):\n", i,
2078 (g->lacons[i].subno) ? "positive" : "negative");
2079 dumpcnfa(&g->lacons[i].cnfa, f);
2080 }
2081 fprintf(f, "\n");
2082 dumpst(g->tree, f, 0);
2083 #endif
2084 }
2085
2086 /*
2087 - dumpst - dump a subRE tree
2088 ^ static VOID dumpst(struct subre *, FILE *, int);
2089 */
2090 static VOID
2091 dumpst(t, f, nfapresent)
2092 struct subre *t;
2093 FILE *f;
2094 int nfapresent; /* is the original NFA still around? */
2095 {
2096 if (t == NULL)
2097 fprintf(f, "null tree\n");
2098 else
2099 stdump(t, f, nfapresent);
2100 fflush(f);
2101 }
2102
2103 /*
2104 - stdump - recursive guts of dumpst
2105 ^ static VOID stdump(struct subre *, FILE *, int);
2106 */
2107 static VOID
2108 stdump(t, f, nfapresent)
2109 struct subre *t;
2110 FILE *f;
2111 int nfapresent; /* is the original NFA still around? */
2112 {
2113 char idbuf[50];
2114
2115 fprintf(f, "%s. `%c'", stid(t, idbuf, sizeof(idbuf)), t->op);
2116 if (t->flags&LONGER)
2117 fprintf(f, " longest");
2118 if (t->flags&SHORTER)
2119 fprintf(f, " shortest");
2120 if (t->flags&MIXED)
2121 fprintf(f, " hasmixed");
2122 if (t->flags&CAP)
2123 fprintf(f, " hascapture");
2124 if (t->flags&BACKR)
2125 fprintf(f, " hasbackref");
2126 if (!(t->flags&INUSE))
2127 fprintf(f, " UNUSED");
2128 if (t->subno != 0)
2129 fprintf(f, " (#%d)", t->subno);
2130 if (t->min != 1 || t->max != 1) {
2131 fprintf(f, " {%d,", t->min);
2132 if (t->max != INFINITY)
2133 fprintf(f, "%d", t->max);
2134 fprintf(f, "}");
2135 }
2136 if (nfapresent)
2137 fprintf(f, " %ld-%ld", (long)t->begin->no, (long)t->end->no);
2138 if (t->left != NULL)
2139 fprintf(f, " L:%s", stid(t->left, idbuf, sizeof(idbuf)));
2140 if (t->right != NULL)
2141 fprintf(f, " R:%s", stid(t->right, idbuf, sizeof(idbuf)));
2142 if (!NULLCNFA(t->cnfa)) {
2143 fprintf(f, "\n");
2144 dumpcnfa(&t->cnfa, f);
2145 fprintf(f, "\n");
2146 }
2147 if (t->left != NULL)
2148 stdump(t->left, f, nfapresent);
2149 if (t->right != NULL)
2150 stdump(t->right, f, nfapresent);
2151 }
2152
2153 /*
2154 - stid - identify a subtree node for dumping
2155 ^ static char *stid(struct subre *, char *, size_t);
2156 */
2157 static char * /* points to buf or constant string */
2158 stid(t, buf, bufsize)
2159 struct subre *t;
2160 char *buf;
2161 size_t bufsize;
2162 {
2163 /* big enough for hex int or decimal t->retry? */
2164 if (bufsize < sizeof(int)*2 + 3 || bufsize < sizeof(t->retry)*3 + 1)
2165 return "unable";
2166 if (t->retry != 0)
2167 sprintf(buf, "%d", t->retry);
2168 else
2169 sprintf(buf, "0x%x", (int)t); /* may lose bits, that's okay */
2170 return buf;
2171 }
2172
2173 #include "regc_lex.c"
2174 #include "regc_color.c"
2175 #include "regc_nfa.c"
2176 #include "regc_cvec.c"
2177 #include "regc_locale.c"
2178
2179 /* $History: regcomp.c $
2180 *
2181 * ***************** Version 1 *****************
2182 * User: Dtashley Date: 1/02/01 Time: 12:06a
2183 * Created in $/IjuScripter, IjuConsole/Source/Tcl Base
2184 * Initial check-in.
2185 */
2186
2187 /* End of REGCOMP.C */

dashley@gmail.com
ViewVC Help
Powered by ViewVC 1.1.25