Extras/regexp/regex/regcomp.c

   1 /*-
   2  * Copyright (c) 1992, 1993, 1994 Henry Spencer.
   3  * Copyright (c) 1992, 1993, 1994
   4  *      The Regents of the University of California.  All rights reserved.
   5  *
   6  * This code is derived from software contributed to Berkeley by
   7  * Henry Spencer.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  * 3. All advertising materials mentioning features or use of this software
  18  *    must display the following acknowledgement:
  19  *      This product includes software developed by the University of
  20  *      California, Berkeley and its contributors.
  21  * 4. Neither the name of the University nor the names of its contributors
  22  *    may be used to endorse or promote products derived from this software
  23  *    without specific prior written permission.
  24  *
  25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  35  * SUCH DAMAGE.
  36  *
  37  *      @(#)regcomp.c   8.5 (Berkeley) 3/20/94
  38  */
  39
  40 #if defined(LIBC_SCCS) && !defined(lint)
  41 static char sccsid[] = "@(#)regcomp.c   8.5 (Berkeley) 3/20/94";
  42 #endif /* LIBC_SCCS and not lint */
  43
  44 #include <sys/types.h>
  45 #include <stdio.h>
  46 #include <string.h>
  47 #include <ctype.h>
  48 #include <limits.h>
  49 #include <stdlib.h>
  50 #include <regex.h>
  51
  52 #include "utils.h"
  53 #include "regex2.h"
  54
  55 #include "cclass.h"
  56 #include "cname.h"
  57
  58 /*
  59  * parse structure, passed up and down to avoid global variables and
  60  * other clumsinesses
  61  */
  62 struct parse {
  63         char *next;             /* next character in RE */
  64         char *end;              /* end of string (-> NUL normally) */
  65         int error;              /* has an error been seen? */
  66         sop *strip;             /* malloced strip */
  67         sopno ssize;            /* malloced strip size (allocated) */
  68         sopno slen;             /* malloced strip length (used) */
  69         int ncsalloc;           /* number of csets allocated */
  70         struct re_guts *g;
  71 #       define  NPAREN  10      /* we need to remember () 1-9 for back refs */
  72         sopno pbegin[NPAREN];   /* -> ( ([0] unused) */
  73         sopno pend[NPAREN];     /* -> ) ([0] unused) */
  74 };
  75
  76 static void p_ere(struct parse *p, int stop);
  77 static void p_ere_exp(struct parse *p);
  78 static void p_str(struct parse *p);
  79 static void p_bre(struct parse *p, int end1, int end2);
  80 static int p_simp_re(struct parse *p, int starordinary);
  81 static int p_count(struct parse *p);
  82 static void p_bracket(struct parse *p);
  83 static void p_b_term(struct parse *p, cset *cs);
  84 static void p_b_cclass(struct parse *p, cset *cs);
  85 static void p_b_eclass(struct parse *p, cset *cs);
  86 static char p_b_symbol(struct parse *p);
  87 static char p_b_coll_elem(struct parse *p, int endc);
  88 static char othercase(int ch);
  89 static void bothcases(struct parse *p, int ch);
  90 static void ordinary(struct parse *p, int ch);
  91 static void nonnewline(struct parse *p);
  92 static void repeat(struct parse *p, sopno start, int from, int to);
  93 static int seterr(struct parse *p, int e);
  94 static cset *allocset(struct parse *p);
  95 static void freeset(struct parse *p, cset *cs);
  96 static int freezeset(struct parse *p, cset *cs);
  97 static int firstch(struct parse *p, cset *cs);
  98 static int nch(struct parse *p, cset *cs);
  99 static void mcadd(struct parse *p, cset *cs, char *cp);
 100 static void mcsub(cset *cs, char *cp);
 101 static int mcin(cset *cs, char *cp);
 102 static char *mcfind(cset *cs, char *cp);
 103 static void mcinvert(struct parse *p, cset *cs);
 104 static void mccase(struct parse *p, cset *cs);
 105 static int isinsets(struct re_guts *g, int c);
 106 static int samesets(struct re_guts *g, int c1, int c2);
 107 static void categorize(struct parse *p, struct re_guts *g);
 108 static sopno dupl(struct parse *p, sopno start, sopno finish);
 109 static void doemit(struct parse *p, sop op, size_t opnd);
 110 static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
 111 static void dofwd(struct parse *p, sopno pos, sop value);
 112 static void enlarge(struct parse *p, sopno size);
 113 static void stripsnug(struct parse *p, struct re_guts *g);
 114 static void findmust(struct parse *p, struct re_guts *g);
 115 static sopno pluscount(struct parse *p, struct re_guts *g);
 116
 117 static char nuls[10];           /* place to point scanner in event of error */
 118
 119 /*
 120  * macros for use with parse structure
 121  * BEWARE:  these know that the parse structure is named `p' !!!
 122  */
 123 #define PEEK()  (*p->next)
 124 #define PEEK2() (*(p->next+1))
 125 #define MORE()  (p->next < p->end)
 126 #define MORE2() (p->next+1 < p->end)
 127 #define SEE(c)  (MORE() && PEEK() == (c))
 128 #define SEETWO(a, b)    (MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
 129 #define EAT(c)  ((SEE(c)) ? (NEXT(), 1) : 0)
 130 #define EATTWO(a, b)    ((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
 131 #define NEXT()  (p->next++)
 132 #define NEXT2() (p->next += 2)
 133 #define NEXTn(n)        (p->next += (n))
 134 #define GETNEXT()       (*p->next++)
 135 #define SETERROR(e)     seterr(p, (e))
 136 #define REQUIRE(co, e)  ((co) || SETERROR(e))
 137 #define MUSTSEE(c, e)   (REQUIRE(MORE() && PEEK() == (c), e))
 138 #define MUSTEAT(c, e)   (REQUIRE(MORE() && GETNEXT() == (c), e))
 139 #define MUSTNOTSEE(c, e)        (REQUIRE(!MORE() || PEEK() != (c), e))
 140 #define EMIT(op, sopnd) doemit(p, (sop)(op), (size_t)(sopnd))
 141 #define INSERT(op, pos) doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
 142 #define AHEAD(pos)              dofwd(p, pos, HERE()-(pos))
 143 #define ASTERN(sop, pos)        EMIT(sop, HERE()-pos)
 144 #define HERE()          (p->slen)
 145 #define THERE()         (p->slen - 1)
 146 #define THERETHERE()    (p->slen - 2)
 147 #define DROP(n) (p->slen -= (n))
 148
 149 #ifndef NDEBUG
 150 static int never = 0;           /* for use in asserts; shuts lint up */
 151 #else
 152 #define never   0               /* some <assert.h>s have bugs too */
 153 #endif
 154
 155 /*
 156  - regcomp - interface for parser and compilation
 157  = extern int regcomp(regex_t *, const char *, int);
 158  = #define      REG_BASIC       0000
 159  = #define      REG_EXTENDED    0001
 160  = #define      REG_ICASE       0002
 161  = #define      REG_NOSUB       0004
 162  = #define      REG_NEWLINE     0010
 163  = #define      REG_NOSPEC      0020
 164  = #define      REG_PEND        0040
 165  = #define      REG_DUMP        0200
 166  */
 167 int                             /* 0 success, otherwise REG_something */
 168 regcomp(preg, pattern, cflags)
 169 regex_t *preg;
 170 const char *pattern;
 171 int cflags;
 172 {
 173         struct parse pa;
 174         register struct re_guts *g;
 175         register struct parse *p = &pa;
 176         register int i;
 177         register size_t len;
 178 #ifdef REDEBUG
 179 #       define  GOODFLAGS(f)    (f)
 180 #else
 181 #       define  GOODFLAGS(f)    ((f)&~REG_DUMP)
 182 #endif
 183
 184         cflags = GOODFLAGS(cflags);
 185         if ((cflags&REG_EXTENDED) && (cflags&REG_NOSPEC))
 186                 return(REG_INVARG);
 187
 188         if (cflags&REG_PEND) {
 189                 if (preg->re_endp < pattern)
 190                         return(REG_INVARG);
 191                 len = preg->re_endp - pattern;
 192         } else
 193                 len = strlen((char *)pattern);
 194
 195         /* do the mallocs early so failure handling is easy */
 196         g = (struct re_guts *)malloc(sizeof(struct re_guts) +
 197                                                         (NC-1)*sizeof(cat_t));
 198         if (g == NULL)
 199                 return(REG_ESPACE);
 200         p->ssize = len/(size_t)2*(size_t)3 + (size_t)1; /* ugh */
 201         p->strip = (sop *)malloc(p->ssize * sizeof(sop));
 202         p->slen = 0;
 203         if (p->strip == NULL) {
 204                 free((char *)g);
 205                 return(REG_ESPACE);
 206         }
 207
 208         /* set things up */
 209         p->g = g;
 210         p->next = (char *)pattern;      /* convenience; we do not modify it */
 211         p->end = p->next + len;
 212         p->error = 0;
 213         p->ncsalloc = 0;
 214         for (i = 0; i < NPAREN; i++) {
 215                 p->pbegin[i] = 0;
 216                 p->pend[i] = 0;
 217         }
 218         g->csetsize = NC;
 219         g->sets = NULL;
 220         g->setbits = NULL;
 221         g->ncsets = 0;
 222         g->cflags = cflags;
 223         g->iflags = 0;
 224         g->nbol = 0;
 225         g->neol = 0;
 226         g->must = NULL;
 227         g->mlen = 0;
 228         g->nsub = 0;
 229         g->ncategories = 1;     /* category 0 is "everything else" */
 230         g->categories = &g->catspace[-(CHAR_MIN)];
 231         (void) memset((char *)g->catspace, 0, NC*sizeof(cat_t));
 232         g->backrefs = 0;
 233
 234         /* do it */
 235         EMIT(OEND, 0);
 236         g->firststate = THERE();
 237         if (cflags&REG_EXTENDED)
 238                 p_ere(p, OUT);
 239         else if (cflags&REG_NOSPEC)
 240                 p_str(p);
 241         else
 242                 p_bre(p, OUT, OUT);
 243         EMIT(OEND, 0);
 244         g->laststate = THERE();
 245
 246         /* tidy up loose ends and fill things in */
 247         categorize(p, g);
 248         stripsnug(p, g);
 249         findmust(p, g);
 250         g->nplus = pluscount(p, g);
 251         g->magic = MAGIC2;
 252         preg->re_nsub = g->nsub;
 253         preg->re_g = g;
 254         preg->re_magic = MAGIC1;
 255 #ifndef REDEBUG
 256         /* not debugging, so can't rely on the assert() in regexec() */
 257         if (g->iflags&BAD)
 258                 SETERROR(REG_ASSERT);
 259 #endif
 260
 261         /* win or lose, we're done */
 262         if (p->error != 0)      /* lose */
 263                 regfree(preg);
 264         return(p->error);
 265 }
 266
 267 /*
 268  - p_ere - ERE parser top level, concatenation and alternation
 269  == static void p_ere(register struct parse *p, int stop);
 270  */
 271 static void
 272 p_ere(p, stop)
 273 register struct parse *p;
 274 int stop;                       /* character this ERE should end at */
 275 {
 276         register char c;
 277         register sopno prevback;
 278         register sopno prevfwd;
 279         register sopno conc;
 280         register int first = 1;         /* is this the first alternative? */
 281
 282         for (;;) {
 283                 /* do a bunch of concatenated expressions */
 284                 conc = HERE();
 285                 while (MORE() && (c = PEEK()) != '|' && c != stop)
 286                         p_ere_exp(p);
 287                 REQUIRE(HERE() != conc, REG_EMPTY);     /* require nonempty */
 288
 289                 if (!EAT('|'))
 290                         break;          /* NOTE BREAK OUT */
 291
 292                 if (first) {
 293                         INSERT(OCH_, conc);     /* offset is wrong */
 294                         prevfwd = conc;
 295                         prevback = conc;
 296                         first = 0;
 297                 }
 298                 ASTERN(OOR1, prevback);
 299                 prevback = THERE();
 300                 AHEAD(prevfwd);                 /* fix previous offset */
 301                 prevfwd = HERE();
 302                 EMIT(OOR2, 0);                  /* offset is very wrong */
 303         }
 304
 305         if (!first) {           /* tail-end fixups */
 306                 AHEAD(prevfwd);
 307                 ASTERN(O_CH, prevback);
 308         }
 309
 310         assert(!MORE() || SEE(stop));
 311 }
 312
 313 /*
 314  - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
 315  == static void p_ere_exp(register struct parse *p);
 316  */
 317 static void
 318 p_ere_exp(p)
 319 register struct parse *p;
 320 {
 321         register char c;
 322         register sopno pos;
 323         register int count;
 324         register int count2;
 325         register sopno subno;
 326         int wascaret = 0;
 327
 328         assert(MORE());         /* caller should have ensured this */
 329         c = GETNEXT();
 330
 331         pos = HERE();
 332         switch (c) {
 333         case '(':
 334                 REQUIRE(MORE(), REG_EPAREN);
 335                 p->g->nsub++;
 336                 subno = p->g->nsub;
 337                 if (subno < NPAREN)
 338                         p->pbegin[subno] = HERE();
 339                 EMIT(OLPAREN, subno);
 340                 if (!SEE(')'))
 341                         p_ere(p, ')');
 342                 if (subno < NPAREN) {
 343                         p->pend[subno] = HERE();
 344                         assert(p->pend[subno] != 0);
 345                 }
 346                 EMIT(ORPAREN, subno);
 347                 MUSTEAT(')', REG_EPAREN);
 348                 break;
 349 #ifndef POSIX_MISTAKE
 350         case ')':               /* happens only if no current unmatched ( */
 351                 /*
 352                  * You may ask, why the ifndef?  Because I didn't notice
 353                  * this until slightly too late for 1003.2, and none of the
 354                  * other 1003.2 regular-expression reviewers noticed it at
 355                  * all.  So an unmatched ) is legal POSIX, at least until
 356                  * we can get it fixed.
 357                  */
 358                 SETERROR(REG_EPAREN);
 359                 break;
 360 #endif
 361         case '^':
 362                 EMIT(OBOL, 0);
 363                 p->g->iflags |= USEBOL;
 364                 p->g->nbol++;
 365                 wascaret = 1;
 366                 break;
 367         case '$':
 368                 EMIT(OEOL, 0);
 369                 p->g->iflags |= USEEOL;
 370                 p->g->neol++;
 371                 break;
 372         case '|':
 373                 SETERROR(REG_EMPTY);
 374                 break;
 375         case '*':
 376         case '+':
 377         case '?':
 378                 SETERROR(REG_BADRPT);
 379                 break;
 380         case '.':
 381                 if (p->g->cflags&REG_NEWLINE)
 382                         nonnewline(p);
 383                 else
 384                         EMIT(OANY, 0);
 385                 break;
 386         case '[':
 387                 p_bracket(p);
 388                 break;
 389         case '\\':
 390                 REQUIRE(MORE(), REG_EESCAPE);
 391                 c = GETNEXT();
 392                 ordinary(p, c);
 393                 break;
 394         case '{':               /* okay as ordinary except if digit follows */
 395                 REQUIRE(!MORE() || !isdigit(PEEK()), REG_BADRPT);
 396                 /* FALLTHROUGH */
 397         default:
 398                 ordinary(p, c);
 399                 break;
 400         }
 401
 402         if (!MORE())
 403                 return;
 404         c = PEEK();
 405         /* we call { a repetition if followed by a digit */
 406         if (!( c == '*' || c == '+' || c == '?' ||
 407                                 (c == '{' && MORE2() && isdigit(PEEK2())) ))
 408                 return;         /* no repetition, we're done */
 409         NEXT();
 410
 411         REQUIRE(!wascaret, REG_BADRPT);
 412         switch (c) {
 413         case '*':       /* implemented as +? */
 414                 /* this case does not require the (y|) trick, noKLUDGE */
 415                 INSERT(OPLUS_, pos);
 416                 ASTERN(O_PLUS, pos);
 417                 INSERT(OQUEST_, pos);
 418                 ASTERN(O_QUEST, pos);
 419                 break;
 420         case '+':
 421                 INSERT(OPLUS_, pos);
 422                 ASTERN(O_PLUS, pos);
 423                 break;
 424         case '?':
 425                 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
 426                 INSERT(OCH_, pos);              /* offset slightly wrong */
 427                 ASTERN(OOR1, pos);              /* this one's right */
 428                 AHEAD(pos);                     /* fix the OCH_ */
 429                 EMIT(OOR2, 0);                  /* offset very wrong... */
 430                 AHEAD(THERE());                 /* ...so fix it */
 431                 ASTERN(O_CH, THERETHERE());
 432                 break;
 433         case '{':
 434                 count = p_count(p);
 435                 if (EAT(',')) {
 436                         if (isdigit(PEEK())) {
 437                                 count2 = p_count(p);
 438                                 REQUIRE(count <= count2, REG_BADBR);
 439                         } else          /* single number with comma */
 440                                 count2 = INFINITY;
 441                 } else          /* just a single number */
 442                         count2 = count;
 443                 repeat(p, pos, count, count2);
 444                 if (!EAT('}')) {        /* error heuristics */
 445                         while (MORE() && PEEK() != '}')
 446                                 NEXT();
 447                         REQUIRE(MORE(), REG_EBRACE);
 448                         SETERROR(REG_BADBR);
 449                 }
 450                 break;
 451         }
 452
 453         if (!MORE())
 454                 return;
 455         c = PEEK();
 456         if (!( c == '*' || c == '+' || c == '?' ||
 457                                 (c == '{' && MORE2() && isdigit(PEEK2())) ) )
 458                 return;
 459         SETERROR(REG_BADRPT);
 460 }
 461
 462 /*
 463  - p_str - string (no metacharacters) "parser"
 464  == static void p_str(register struct parse *p);
 465  */
 466 static void
 467 p_str(p)
 468 register struct parse *p;
 469 {
 470         REQUIRE(MORE(), REG_EMPTY);
 471         while (MORE())
 472                 ordinary(p, GETNEXT());
 473 }
 474
 475 /*
 476  - p_bre - BRE parser top level, anchoring and concatenation
 477  == static void p_bre(register struct parse *p, register int end1, \
 478  ==     register int end2);
 479  * Giving end1 as OUT essentially eliminates the end1/end2 check.
 480  *
 481  * This implementation is a bit of a kludge, in that a trailing $ is first
 482  * taken as an ordinary character and then revised to be an anchor.  The
 483  * only undesirable side effect is that '$' gets included as a character
 484  * category in such cases.  This is fairly harmless; not worth fixing.
 485  * The amount of lookahead needed to avoid this kludge is excessive.
 486  */
 487 static void
 488 p_bre(p, end1, end2)
 489 register struct parse *p;
 490 register int end1;              /* first terminating character */
 491 register int end2;              /* second terminating character */
 492 {
 493         register sopno start = HERE();
 494         register int first = 1;                 /* first subexpression? */
 495         register int wasdollar = 0;
 496
 497         if (EAT('^')) {
 498                 EMIT(OBOL, 0);
 499                 p->g->iflags |= USEBOL;
 500                 p->g->nbol++;
 501         }
 502         while (MORE() && !SEETWO(end1, end2)) {
 503                 wasdollar = p_simp_re(p, first);
 504                 first = 0;
 505         }
 506         if (wasdollar) {        /* oops, that was a trailing anchor */
 507                 DROP(1);
 508                 EMIT(OEOL, 0);
 509                 p->g->iflags |= USEEOL;
 510                 p->g->neol++;
 511         }
 512
 513         REQUIRE(HERE() != start, REG_EMPTY);    /* require nonempty */
 514 }
 515
 516 /*
 517  - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
 518  == static int p_simp_re(register struct parse *p, int starordinary);
 519  */
 520 static int                      /* was the simple RE an unbackslashed $? */
 521 p_simp_re(p, starordinary)
 522 register struct parse *p;
 523 int starordinary;               /* is a leading * an ordinary character? */
 524 {
 525         register int c;
 526         register int count;
 527         register int count2;
 528         register sopno pos;
 529         register int i;
 530         register sopno subno;
 531 #       define  BACKSL  (1<<CHAR_BIT)
 532
 533         pos = HERE();           /* repetion op, if any, covers from here */
 534
 535         assert(MORE());         /* caller should have ensured this */
 536         c = GETNEXT();
 537         if (c == '\\') {
 538                 REQUIRE(MORE(), REG_EESCAPE);
 539                 c = BACKSL | (unsigned char)GETNEXT();
 540         }
 541         switch (c) {
 542         case '.':
 543                 if (p->g->cflags&REG_NEWLINE)
 544                         nonnewline(p);
 545                 else
 546                         EMIT(OANY, 0);
 547                 break;
 548         case '[':
 549                 p_bracket(p);
 550                 break;
 551         case BACKSL|'{':
 552                 SETERROR(REG_BADRPT);
 553                 break;
 554         case BACKSL|'(':
 555                 p->g->nsub++;
 556                 subno = p->g->nsub;
 557                 if (subno < NPAREN)
 558                         p->pbegin[subno] = HERE();
 559                 EMIT(OLPAREN, subno);
 560                 /* the MORE here is an error heuristic */
 561                 if (MORE() && !SEETWO('\\', ')'))
 562                         p_bre(p, '\\', ')');
 563                 if (subno < NPAREN) {
 564                         p->pend[subno] = HERE();
 565                         assert(p->pend[subno] != 0);
 566                 }
 567                 EMIT(ORPAREN, subno);
 568                 REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
 569                 break;
 570         case BACKSL|')':        /* should not get here -- must be user */
 571         case BACKSL|'}':
 572                 SETERROR(REG_EPAREN);
 573                 break;
 574         case BACKSL|'1':
 575         case BACKSL|'2':
 576         case BACKSL|'3':
 577         case BACKSL|'4':
 578         case BACKSL|'5':
 579         case BACKSL|'6':
 580         case BACKSL|'7':
 581         case BACKSL|'8':
 582         case BACKSL|'9':
 583                 i = (c&~BACKSL) - '0';
 584                 assert(i < NPAREN);
 585                 if (p->pend[i] != 0) {
 586                         assert(i <= p->g->nsub);
 587                         EMIT(OBACK_, i);
 588                         assert(p->pbegin[i] != 0);
 589                         assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
 590                         assert(OP(p->strip[p->pend[i]]) == ORPAREN);
 591                         (void) dupl(p, p->pbegin[i]+1, p->pend[i]);
 592                         EMIT(O_BACK, i);
 593                 } else
 594                         SETERROR(REG_ESUBREG);
 595                 p->g->backrefs = 1;
 596                 break;
 597         case '*':
 598                 REQUIRE(starordinary, REG_BADRPT);
 599                 /* FALLTHROUGH */
 600         default:
 601                 ordinary(p, c &~ BACKSL);
 602                 break;
 603         }
 604
 605         if (EAT('*')) {         /* implemented as +? */
 606                 /* this case does not require the (y|) trick, noKLUDGE */
 607                 INSERT(OPLUS_, pos);
 608                 ASTERN(O_PLUS, pos);
 609                 INSERT(OQUEST_, pos);
 610                 ASTERN(O_QUEST, pos);
 611         } else if (EATTWO('\\', '{')) {
 612                 count = p_count(p);
 613                 if (EAT(',')) {
 614                         if (MORE() && isdigit(PEEK())) {
 615                                 count2 = p_count(p);
 616                                 REQUIRE(count <= count2, REG_BADBR);
 617                         } else          /* single number with comma */
 618                                 count2 = INFINITY;
 619                 } else          /* just a single number */
 620                         count2 = count;
 621                 repeat(p, pos, count, count2);
 622                 if (!EATTWO('\\', '}')) {       /* error heuristics */
 623                         while (MORE() && !SEETWO('\\', '}'))
 624                                 NEXT();
 625                         REQUIRE(MORE(), REG_EBRACE);
 626                         SETERROR(REG_BADBR);
 627                 }
 628         } else if (c == (unsigned char)'$')     /* $ (but not \$) ends it */
 629                 return(1);
 630
 631         return(0);
 632 }
 633
 634 /*
 635  - p_count - parse a repetition count
 636  == static int p_count(register struct parse *p);
 637  */
 638 static int                      /* the value */
 639 p_count(p)
 640 register struct parse *p;
 641 {
 642         register int count = 0;
 643         register int ndigits = 0;
 644
 645         while (MORE() && isdigit(PEEK()) && count <= DUPMAX) {
 646                 count = count*10 + (GETNEXT() - '0');
 647                 ndigits++;
 648         }
 649
 650         REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
 651         return(count);
 652 }
 653
 654 /*
 655  - p_bracket - parse a bracketed character list
 656  == static void p_bracket(register struct parse *p);
 657  *
 658  * Note a significant property of this code:  if the allocset() did SETERROR,
 659  * no set operations are done.
 660  */
 661 static void
 662 p_bracket(p)
 663 register struct parse *p;
 664 {
 665         register char c;
 666         register cset *cs = allocset(p);
 667         register int invert = 0;
 668
 669         /* Dept of Truly Sickening Special-Case Kludges */
 670         if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
 671                 EMIT(OBOW, 0);
 672                 NEXTn(6);
 673                 return;
 674         }
 675         if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
 676                 EMIT(OEOW, 0);
 677                 NEXTn(6);
 678                 return;
 679         }
 680
 681         if (EAT('^'))
 682                 invert++;       /* make note to invert set at end */
 683         if (EAT(']'))
 684                 CHadd(cs, ']');
 685         else if (EAT('-'))
 686                 CHadd(cs, '-');
 687         while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
 688                 p_b_term(p, cs);
 689         if (EAT('-'))
 690                 CHadd(cs, '-');
 691         MUSTEAT(']', REG_EBRACK);
 692
 693         if (p->error != 0)      /* don't mess things up further */
 694                 return;
 695
 696         if (p->g->cflags&REG_ICASE) {
 697                 register int i;
 698                 register int ci;
 699
 700                 for (i = p->g->csetsize - 1; i >= 0; i--)
 701                         if (CHIN(cs, i) && isalpha(i)) {
 702                                 ci = othercase(i);
 703                                 if (ci != i)
 704                                         CHadd(cs, ci);
 705                         }
 706                 if (cs->multis != NULL)
 707                         mccase(p, cs);
 708         }
 709         if (invert) {
 710                 register int i;
 711
 712                 for (i = p->g->csetsize - 1; i >= 0; i--)
 713                         if (CHIN(cs, i))
 714                                 CHsub(cs, i);
 715                         else
 716                                 CHadd(cs, i);
 717                 if (p->g->cflags&REG_NEWLINE)
 718                         CHsub(cs, '\n');
 719                 if (cs->multis != NULL)
 720                         mcinvert(p, cs);
 721         }
 722
 723         assert(cs->multis == NULL);             /* xxx */
 724
 725         if (nch(p, cs) == 1) {          /* optimize singleton sets */
 726                 ordinary(p, firstch(p, cs));
 727                 freeset(p, cs);
 728         } else
 729                 EMIT(OANYOF, freezeset(p, cs));
 730 }
 731
 732 /*
 733  - p_b_term - parse one term of a bracketed character list
 734  == static void p_b_term(register struct parse *p, register cset *cs);
 735  */
 736 static void
 737 p_b_term(p, cs)
 738 register struct parse *p;
 739 register cset *cs;
 740 {
 741         register char c;
 742         register char start, finish;
 743         register int i;
 744
 745         /* classify what we've got */
 746         switch ((MORE()) ? PEEK() : '\0') {
 747         case '[':
 748                 c = (MORE2()) ? PEEK2() : '\0';
 749                 break;
 750         case '-':
 751                 SETERROR(REG_ERANGE);
 752                 return;                 /* NOTE RETURN */
 753                 break;
 754         default:
 755                 c = '\0';
 756                 break;
 757         }
 758
 759         switch (c) {
 760         case ':':               /* character class */
 761                 NEXT2();
 762                 REQUIRE(MORE(), REG_EBRACK);
 763                 c = PEEK();
 764                 REQUIRE(c != '-' && c != ']', REG_ECTYPE);
 765                 p_b_cclass(p, cs);
 766                 REQUIRE(MORE(), REG_EBRACK);
 767                 REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
 768                 break;
 769         case '=':               /* equivalence class */
 770                 NEXT2();
 771                 REQUIRE(MORE(), REG_EBRACK);
 772                 c = PEEK();
 773                 REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
 774                 p_b_eclass(p, cs);
 775                 REQUIRE(MORE(), REG_EBRACK);
 776                 REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
 777                 break;
 778         default:                /* symbol, ordinary character, or range */
 779 /* xxx revision needed for multichar stuff */
 780                 start = p_b_symbol(p);
 781                 if (SEE('-') && MORE2() && PEEK2() != ']') {
 782                         /* range */
 783                         NEXT();
 784                         if (EAT('-'))
 785                                 finish = '-';
 786                         else
 787                                 finish = p_b_symbol(p);
 788                 } else
 789                         finish = start;
 790 /* xxx what about signed chars here... */
 791                 REQUIRE(start <= finish, REG_ERANGE);
 792                 for (i = start; i <= finish; i++)
 793                         CHadd(cs, i);
 794                 break;
 795         }
 796 }
 797
 798 /*
 799  - p_b_cclass - parse a character-class name and deal with it
 800  == static void p_b_cclass(register struct parse *p, register cset *cs);
 801  */
 802 static void
 803 p_b_cclass(p, cs)
 804 register struct parse *p;
 805 register cset *cs;
 806 {
 807         register char *sp = p->next;
 808         register struct cclass *cp;
 809         register size_t len;
 810         register char *u;
 811         register char c;
 812
 813         while (MORE() && isalpha(PEEK()))
 814                 NEXT();
 815         len = p->next - sp;
 816         for (cp = cclasses; cp->name != NULL; cp++)
 817                 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
 818                         break;
 819         if (cp->name == NULL) {
 820                 /* oops, didn't find it */
 821                 SETERROR(REG_ECTYPE);
 822                 return;
 823         }
 824
 825         u = cp->chars;
 826         while ((c = *u++) != '\0')
 827                 CHadd(cs, c);
 828         for (u = cp->multis; *u != '\0'; u += strlen(u) + 1)
 829                 MCadd(p, cs, u);
 830 }
 831
 832 /*
 833  - p_b_eclass - parse an equivalence-class name and deal with it
 834  == static void p_b_eclass(register struct parse *p, register cset *cs);
 835  *
 836  * This implementation is incomplete. xxx
 837  */
 838 static void
 839 p_b_eclass(p, cs)
 840 register struct parse *p;
 841 register cset *cs;
 842 {
 843         register char c;
 844
 845         c = p_b_coll_elem(p, '=');
 846         CHadd(cs, c);
 847 }
 848
 849 /*
 850  - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
 851  == static char p_b_symbol(register struct parse *p);
 852  */
 853 static char                     /* value of symbol */
 854 p_b_symbol(p)
 855 register struct parse *p;
 856 {
 857         register char value;
 858
 859         REQUIRE(MORE(), REG_EBRACK);
 860         if (!EATTWO('[', '.'))
 861                 return(GETNEXT());
 862
 863         /* collating symbol */
 864         value = p_b_coll_elem(p, '.');
 865         REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
 866         return(value);
 867 }
 868
 869 /*
 870  - p_b_coll_elem - parse a collating-element name and look it up
 871  == static char p_b_coll_elem(register struct parse *p, int endc);
 872  */
 873 static char                     /* value of collating element */
 874 p_b_coll_elem(p, endc)
 875 register struct parse *p;
 876 int endc;                       /* name ended by endc,']' */
 877 {
 878         register char *sp = p->next;
 879         register struct cname *cp;
 880         register int len;
 881         register char c;
 882
 883         while (MORE() && !SEETWO(endc, ']'))
 884                 NEXT();
 885         if (!MORE()) {
 886                 SETERROR(REG_EBRACK);
 887                 return(0);
 888         }
 889         len = p->next - sp;
 890         for (cp = cnames; cp->name != NULL; cp++)
 891                 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
 892                         return(cp->code);       /* known name */
 893         if (len == 1)
 894                 return(*sp);    /* single character */
 895         SETERROR(REG_ECOLLATE);                 /* neither */
 896         return(0);
 897 }
 898
 899 /*
 900  - othercase - return the case counterpart of an alphabetic
 901  == static char othercase(int ch);
 902  */
 903 static char                     /* if no counterpart, return ch */
 904 othercase(ch)
 905 int ch;
 906 {
 907         assert(isalpha(ch));
 908         if (isupper(ch))
 909                 return(tolower(ch));
 910         else if (islower(ch))
 911                 return(toupper(ch));
 912         else                    /* peculiar, but could happen */
 913                 return(ch);
 914 }
 915
 916 /*
 917  - bothcases - emit a dualcase version of a two-case character
 918  == static void bothcases(register struct parse *p, int ch);
 919  *
 920  * Boy, is this implementation ever a kludge...
 921  */
 922 static void
 923 bothcases(p, ch)
 924 register struct parse *p;
 925 int ch;
 926 {
 927         register char *oldnext = p->next;
 928         register char *oldend = p->end;
 929         char bracket[3];
 930
 931         assert(othercase(ch) != ch);    /* p_bracket() would recurse */
 932         p->next = bracket;
 933         p->end = bracket+2;
 934         bracket[0] = ch;
 935         bracket[1] = ']';
 936         bracket[2] = '\0';
 937         p_bracket(p);
 938         assert(p->next == bracket+2);
 939         p->next = oldnext;
 940         p->end = oldend;
 941 }
 942
 943 /*
 944  - ordinary - emit an ordinary character
 945  == static void ordinary(register struct parse *p, register int ch);
 946  */
 947 static void
 948 ordinary(p, ch)
 949 register struct parse *p;
 950 register int ch;
 951 {
 952         register cat_t *cap = p->g->categories;
 953
 954         if ((p->g->cflags&REG_ICASE) && isalpha(ch) && othercase(ch) != ch)
 955                 bothcases(p, ch);
 956         else {
 957                 EMIT(OCHAR, (unsigned char)ch);
 958                 if (cap[ch] == 0)
 959                         cap[ch] = p->g->ncategories++;
 960         }
 961 }
 962
 963 /*
 964  - nonnewline - emit REG_NEWLINE version of OANY
 965  == static void nonnewline(register struct parse *p);
 966  *
 967  * Boy, is this implementation ever a kludge...
 968  */
 969 static void
 970 nonnewline(p)
 971 register struct parse *p;
 972 {
 973         register char *oldnext = p->next;
 974         register char *oldend = p->end;
 975         char bracket[4];
 976
 977         p->next = bracket;
 978         p->end = bracket+3;
 979         bracket[0] = '^';
 980         bracket[1] = '\n';
 981         bracket[2] = ']';
 982         bracket[3] = '\0';
 983         p_bracket(p);
 984         assert(p->next == bracket+3);
 985         p->next = oldnext;
 986         p->end = oldend;
 987 }
 988
 989 /*
 990  - repeat - generate code for a bounded repetition, recursively if needed
 991  == static void repeat(register struct parse *p, sopno start, int from, int to);
 992  */
 993 static void
 994 repeat(p, start, from, to)
 995 register struct parse *p;
 996 sopno start;                    /* operand from here to end of strip */
 997 int from;                       /* repeated from this number */
 998 int to;                         /* to this number of times (maybe INFINITY) */
 999 {
1000         register sopno finish = HERE();
1001 #       define  N       2
1002 #       define  INF     3
1003 #       define  REP(f, t)       ((f)*8 + (t))
1004 #       define  MAP(n)  (((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
1005         register sopno copy;
1006
1007         if (p->error != 0)      /* head off possible runaway recursion */
1008                 return;
1009
1010         assert(from <= to);
1011
1012         switch (REP(MAP(from), MAP(to))) {
1013         case REP(0, 0):                 /* must be user doing this */
1014                 DROP(finish-start);     /* drop the operand */
1015                 break;
1016         case REP(0, 1):                 /* as x{1,1}? */
1017         case REP(0, N):                 /* as x{1,n}? */
1018         case REP(0, INF):               /* as x{1,}? */
1019                 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1020                 INSERT(OCH_, start);            /* offset is wrong... */
1021                 repeat(p, start+1, 1, to);
1022                 ASTERN(OOR1, start);
1023                 AHEAD(start);                   /* ... fix it */
1024                 EMIT(OOR2, 0);
1025                 AHEAD(THERE());
1026                 ASTERN(O_CH, THERETHERE());
1027                 break;
1028         case REP(1, 1):                 /* trivial case */
1029                 /* done */
1030                 break;
1031         case REP(1, N):                 /* as x?x{1,n-1} */
1032                 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1033                 INSERT(OCH_, start);
1034                 ASTERN(OOR1, start);
1035                 AHEAD(start);
1036                 EMIT(OOR2, 0);                  /* offset very wrong... */
1037                 AHEAD(THERE());                 /* ...so fix it */
1038                 ASTERN(O_CH, THERETHERE());
1039                 copy = dupl(p, start+1, finish+1);
1040                 assert(copy == finish+4);
1041                 repeat(p, copy, 1, to-1);
1042                 break;
1043         case REP(1, INF):               /* as x+ */
1044                 INSERT(OPLUS_, start);
1045                 ASTERN(O_PLUS, start);
1046                 break;
1047         case REP(N, N):                 /* as xx{m-1,n-1} */
1048                 copy = dupl(p, start, finish);
1049                 repeat(p, copy, from-1, to-1);
1050                 break;
1051         case REP(N, INF):               /* as xx{n-1,INF} */
1052                 copy = dupl(p, start, finish);
1053                 repeat(p, copy, from-1, to);
1054                 break;
1055         default:                        /* "can't happen" */
1056                 SETERROR(REG_ASSERT);   /* just in case */
1057                 break;
1058         }
1059 }
1060
1061 /*
1062  - seterr - set an error condition
1063  == static int seterr(register struct parse *p, int e);
1064  */
1065 static int                      /* useless but makes type checking happy */
1066 seterr(p, e)
1067 register struct parse *p;
1068 int e;
1069 {
1070         if (p->error == 0)      /* keep earliest error condition */
1071                 p->error = e;
1072         p->next = nuls;         /* try to bring things to a halt */
1073         p->end = nuls;
1074         return(0);              /* make the return value well-defined */
1075 }
1076
1077 /*
1078  - allocset - allocate a set of characters for []
1079  == static cset *allocset(register struct parse *p);
1080  */
1081 static cset *
1082 allocset(p)
1083 register struct parse *p;
1084 {
1085         register int no = p->g->ncsets++;
1086         register size_t nc;
1087         register size_t nbytes;
1088         register cset *cs;
1089         register size_t css = (size_t)p->g->csetsize;
1090         register int i;
1091
1092         if (no >= p->ncsalloc) {        /* need another column of space */
1093                 p->ncsalloc += CHAR_BIT;
1094                 nc = p->ncsalloc;
1095                 assert(nc % CHAR_BIT == 0);
1096                 nbytes = nc / CHAR_BIT * css;
1097                 if (p->g->sets == NULL)
1098                         p->g->sets = (cset *)malloc(nc * sizeof(cset));
1099                 else
1100                         p->g->sets = (cset *)realloc((char *)p->g->sets,
1101                                                         nc * sizeof(cset));
1102                 if (p->g->setbits == NULL)
1103                         p->g->setbits = (uch *)malloc(nbytes);
1104                 else {
1105                         p->g->setbits = (uch *)realloc((char *)p->g->setbits,
1106                                                                 nbytes);
1107                         /* xxx this isn't right if setbits is now NULL */
1108                         for (i = 0; i < no; i++)
1109                                 p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
1110                 }
1111                 if (p->g->sets != NULL && p->g->setbits != NULL)
1112                         (void) memset((char *)p->g->setbits + (nbytes - css),
1113                                                                 0, css);
1114                 else {
1115                         no = 0;
1116                         SETERROR(REG_ESPACE);
1117                         /* caller's responsibility not to do set ops */
1118                 }
1119         }
1120
1121         assert(p->g->sets != NULL);     /* xxx */
1122         cs = &p->g->sets[no];
1123         cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
1124         cs->mask = 1 << ((no) % CHAR_BIT);
1125         cs->hash = 0;
1126         cs->smultis = 0;
1127         cs->multis = NULL;
1128
1129         return(cs);
1130 }
1131
1132 /*
1133  - freeset - free a now-unused set
1134  == static void freeset(register struct parse *p, register cset *cs);
1135  */
1136 static void
1137 freeset(p, cs)
1138 register struct parse *p;
1139 register cset *cs;
1140 {
1141         register int i;
1142         register cset *top = &p->g->sets[p->g->ncsets];
1143         register size_t css = (size_t)p->g->csetsize;
1144
1145         for (i = 0; i < css; i++)
1146                 CHsub(cs, i);
1147         if (cs == top-1)        /* recover only the easy case */
1148                 p->g->ncsets--;
1149 }
1150
1151 /*
1152  - freezeset - final processing on a set of characters
1153  == static int freezeset(register struct parse *p, register cset *cs);
1154  *
1155  * The main task here is merging identical sets.  This is usually a waste
1156  * of time (although the hash code minimizes the overhead), but can win
1157  * big if REG_ICASE is being used.  REG_ICASE, by the way, is why the hash
1158  * is done using addition rather than xor -- all ASCII [aA] sets xor to
1159  * the same value!
1160  */
1161 static int                      /* set number */
1162 freezeset(p, cs)
1163 register struct parse *p;
1164 register cset *cs;
1165 {
1166         register uch h = cs->hash;
1167         register int i;
1168         register cset *top = &p->g->sets[p->g->ncsets];
1169         register cset *cs2;
1170         register size_t css = (size_t)p->g->csetsize;
1171
1172         /* look for an earlier one which is the same */
1173         for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
1174                 if (cs2->hash == h && cs2 != cs) {
1175                         /* maybe */
1176                         for (i = 0; i < css; i++)
1177                                 if (!!CHIN(cs2, i) != !!CHIN(cs, i))
1178                                         break;          /* no */
1179                         if (i == css)
1180                                 break;                  /* yes */
1181                 }
1182
1183         if (cs2 < top) {        /* found one */
1184                 freeset(p, cs);
1185                 cs = cs2;
1186         }
1187
1188         return((int)(cs - p->g->sets));
1189 }
1190
1191 /*
1192  - firstch - return first character in a set (which must have at least one)
1193  == static int firstch(register struct parse *p, register cset *cs);
1194  */
1195 static int                      /* character; there is no "none" value */
1196 firstch(p, cs)
1197 register struct parse *p;
1198 register cset *cs;
1199 {
1200         register int i;
1201         register size_t css = (size_t)p->g->csetsize;
1202
1203         for (i = 0; i < css; i++)
1204                 if (CHIN(cs, i))
1205                         return((char)i);
1206         assert(never);
1207         return(0);              /* arbitrary */
1208 }
1209
1210 /*
1211  - nch - number of characters in a set
1212  == static int nch(register struct parse *p, register cset *cs);
1213  */
1214 static int
1215 nch(p, cs)
1216 register struct parse *p;
1217 register cset *cs;
1218 {
1219         register int i;
1220         register size_t css = (size_t)p->g->csetsize;
1221         register int n = 0;
1222
1223         for (i = 0; i < css; i++)
1224                 if (CHIN(cs, i))
1225                         n++;
1226         return(n);
1227 }
1228
1229 /*
1230  - mcadd - add a collating element to a cset
1231  == static void mcadd(register struct parse *p, register cset *cs, \
1232  ==     register char *cp);
1233  */
1234 static void
1235 mcadd(p, cs, cp)
1236 register struct parse *p;
1237 register cset *cs;
1238 register char *cp;
1239 {
1240         register size_t oldend = cs->smultis;
1241
1242         cs->smultis += strlen(cp) + 1;
1243         if (cs->multis == NULL)
1244                 cs->multis = malloc(cs->smultis);
1245         else
1246                 cs->multis = realloc(cs->multis, cs->smultis);
1247         if (cs->multis == NULL) {
1248                 SETERROR(REG_ESPACE);
1249                 return;
1250         }
1251
1252         (void) strcpy(cs->multis + oldend - 1, cp);
1253         cs->multis[cs->smultis - 1] = '\0';
1254 }
1255
1256 /*
1257  - mcsub - subtract a collating element from a cset
1258  == static void mcsub(register cset *cs, register char *cp);
1259  */
1260 static void
1261 mcsub(cs, cp)
1262 register cset *cs;
1263 register char *cp;
1264 {
1265         register char *fp = mcfind(cs, cp);
1266         register size_t len = strlen(fp);
1267
1268         assert(fp != NULL);
1269         (void) memmove(fp, fp + len + 1,
1270                                 cs->smultis - (fp + len + 1 - cs->multis));
1271         cs->smultis -= len;
1272
1273         if (cs->smultis == 0) {
1274                 free(cs->multis);
1275                 cs->multis = NULL;
1276                 return;
1277         }
1278
1279         cs->multis = realloc(cs->multis, cs->smultis);
1280         assert(cs->multis != NULL);
1281 }
1282
1283 /*
1284  - mcin - is a collating element in a cset?
1285  == static int mcin(register cset *cs, register char *cp);
1286  */
1287 static int
1288 mcin(cs, cp)
1289 register cset *cs;
1290 register char *cp;
1291 {
1292         return(mcfind(cs, cp) != NULL);
1293 }
1294
1295 /*
1296  - mcfind - find a collating element in a cset
1297  == static char *mcfind(register cset *cs, register char *cp);
1298  */
1299 static char *
1300 mcfind(cs, cp)
1301 register cset *cs;
1302 register char *cp;
1303 {
1304         register char *p;
1305
1306         if (cs->multis == NULL)
1307                 return(NULL);
1308         for (p = cs->multis; *p != '\0'; p += strlen(p) + 1)
1309                 if (strcmp(cp, p) == 0)
1310                         return(p);
1311         return(NULL);
1312 }
1313
1314 /*
1315  - mcinvert - invert the list of collating elements in a cset
1316  == static void mcinvert(register struct parse *p, register cset *cs);
1317  *
1318  * This would have to know the set of possibilities.  Implementation
1319  * is deferred.
1320  */
1321 static void
1322 mcinvert(p, cs)
1323 register struct parse *p;
1324 register cset *cs;
1325 {
1326         assert(cs->multis == NULL);     /* xxx */
1327 }
1328
1329 /*
1330  - mccase - add case counterparts of the list of collating elements in a cset
1331  == static void mccase(register struct parse *p, register cset *cs);
1332  *
1333  * This would have to know the set of possibilities.  Implementation
1334  * is deferred.
1335  */
1336 static void
1337 mccase(p, cs)
1338 register struct parse *p;
1339 register cset *cs;
1340 {
1341         assert(cs->multis == NULL);     /* xxx */
1342 }
1343
1344 /*
1345  - isinsets - is this character in any sets?
1346  == static int isinsets(register struct re_guts *g, int c);
1347  */
1348 static int                      /* predicate */
1349 isinsets(g, c)
1350 register struct re_guts *g;
1351 int c;
1352 {
1353         register uch *col;
1354         register int i;
1355         register int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1356         register unsigned uc = (unsigned char)c;
1357
1358         for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1359                 if (col[uc] != 0)
1360                         return(1);
1361         return(0);
1362 }
1363
1364 /*
1365  - samesets - are these two characters in exactly the same sets?
1366  == static int samesets(register struct re_guts *g, int c1, int c2);
1367  */
1368 static int                      /* predicate */
1369 samesets(g, c1, c2)
1370 register struct re_guts *g;
1371 int c1;
1372 int c2;
1373 {
1374         register uch *col;
1375         register int i;
1376         register int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1377         register unsigned uc1 = (unsigned char)c1;
1378         register unsigned uc2 = (unsigned char)c2;
1379
1380         for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1381                 if (col[uc1] != col[uc2])
1382                         return(0);
1383         return(1);
1384 }
1385
1386 /*
1387  - categorize - sort out character categories
1388  == static void categorize(struct parse *p, register struct re_guts *g);
1389  */
1390 static void
1391 categorize(p, g)
1392 struct parse *p;
1393 register struct re_guts *g;
1394 {
1395         register cat_t *cats = g->categories;
1396         register int c;
1397         register int c2;
1398         register cat_t cat;
1399
1400         /* avoid making error situations worse */
1401         if (p->error != 0)
1402                 return;
1403
1404         for (c = CHAR_MIN; c <= CHAR_MAX; c++)
1405                 if (cats[c] == 0 && isinsets(g, c)) {
1406                         cat = g->ncategories++;
1407                         cats[c] = cat;
1408                         for (c2 = c+1; c2 <= CHAR_MAX; c2++)
1409                                 if (cats[c2] == 0 && samesets(g, c, c2))
1410                                         cats[c2] = cat;
1411                 }
1412 }
1413
1414 /*
1415  - dupl - emit a duplicate of a bunch of sops
1416  == static sopno dupl(register struct parse *p, sopno start, sopno finish);
1417  */
1418 static sopno                    /* start of duplicate */
1419 dupl(p, start, finish)
1420 register struct parse *p;
1421 sopno start;                    /* from here */
1422 sopno finish;                   /* to this less one */
1423 {
1424         register sopno ret = HERE();
1425         register sopno len = finish - start;
1426
1427         assert(finish >= start);
1428         if (len == 0)
1429                 return(ret);
1430         enlarge(p, p->ssize + len);     /* this many unexpected additions */
1431         assert(p->ssize >= p->slen + len);
1432         (void) memcpy((char *)(p->strip + p->slen),
1433                 (char *)(p->strip + start), (size_t)len*sizeof(sop));
1434         p->slen += len;
1435         return(ret);
1436 }
1437
1438 /*
1439  - doemit - emit a strip operator
1440  == static void doemit(register struct parse *p, sop op, size_t opnd);
1441  *
1442  * It might seem better to implement this as a macro with a function as
1443  * hard-case backup, but it's just too big and messy unless there are
1444  * some changes to the data structures.  Maybe later.
1445  */
1446 static void
1447 doemit(p, op, opnd)
1448 register struct parse *p;
1449 sop op;
1450 size_t opnd;
1451 {
1452         /* avoid making error situations worse */
1453         if (p->error != 0)
1454                 return;
1455
1456         /* deal with oversize operands ("can't happen", more or less) */
1457         assert(opnd < 1<<OPSHIFT);
1458
1459         /* deal with undersized strip */
1460         if (p->slen >= p->ssize)
1461                 enlarge(p, (p->ssize+1) / 2 * 3);       /* +50% */
1462         assert(p->slen < p->ssize);
1463
1464         /* finally, it's all reduced to the easy case */
1465         p->strip[p->slen++] = SOP(op, opnd);
1466 }
1467
1468 /*
1469  - doinsert - insert a sop into the strip
1470  == static void doinsert(register struct parse *p, sop op, size_t opnd, sopno pos);
1471  */
1472 static void
1473 doinsert(p, op, opnd, pos)
1474 register struct parse *p;
1475 sop op;
1476 size_t opnd;
1477 sopno pos;
1478 {
1479         register sopno sn;
1480         register sop s;
1481         register int i;
1482
1483         /* avoid making error situations worse */
1484         if (p->error != 0)
1485                 return;
1486
1487         sn = HERE();
1488         EMIT(op, opnd);         /* do checks, ensure space */
1489         assert(HERE() == sn+1);
1490         s = p->strip[sn];
1491
1492         /* adjust paren pointers */
1493         assert(pos > 0);
1494         for (i = 1; i < NPAREN; i++) {
1495                 if (p->pbegin[i] >= pos) {
1496                         p->pbegin[i]++;
1497                 }
1498                 if (p->pend[i] >= pos) {
1499                         p->pend[i]++;
1500                 }
1501         }
1502
1503         memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1504                                                 (HERE()-pos-1)*sizeof(sop));
1505         p->strip[pos] = s;
1506 }
1507
1508 /*
1509  - dofwd - complete a forward reference
1510  == static void dofwd(register struct parse *p, sopno pos, sop value);
1511  */
1512 static void
1513 dofwd(p, pos, value)
1514 register struct parse *p;
1515 register sopno pos;
1516 sop value;
1517 {
1518         /* avoid making error situations worse */
1519         if (p->error != 0)
1520                 return;
1521
1522         assert(value < 1<<OPSHIFT);
1523         p->strip[pos] = OP(p->strip[pos]) | value;
1524 }
1525
1526 /*
1527  - enlarge - enlarge the strip
1528  == static void enlarge(register struct parse *p, sopno size);
1529  */
1530 static void
1531 enlarge(p, size)
1532 register struct parse *p;
1533 register sopno size;
1534 {
1535         register sop *sp;
1536
1537         if (p->ssize >= size)
1538                 return;
1539
1540         sp = (sop *)realloc(p->strip, size*sizeof(sop));
1541         if (sp == NULL) {
1542                 SETERROR(REG_ESPACE);
1543                 return;
1544         }
1545         p->strip = sp;
1546         p->ssize = size;
1547 }
1548
1549 /*
1550  - stripsnug - compact the strip
1551  == static void stripsnug(register struct parse *p, register struct re_guts *g);
1552  */
1553 static void
1554 stripsnug(p, g)
1555 register struct parse *p;
1556 register struct re_guts *g;
1557 {
1558         g->nstates = p->slen;
1559         g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
1560         if (g->strip == NULL) {
1561                 SETERROR(REG_ESPACE);
1562                 g->strip = p->strip;
1563         }
1564 }
1565
1566 /*
1567  - findmust - fill in must and mlen with longest mandatory literal string
1568  == static void findmust(register struct parse *p, register struct re_guts *g);
1569  *
1570  * This algorithm could do fancy things like analyzing the operands of |
1571  * for common subsequences.  Someday.  This code is simple and finds most
1572  * of the interesting cases.
1573  *
1574  * Note that must and mlen got initialized during setup.
1575  */
1576 static void
1577 findmust(p, g)
1578 struct parse *p;
1579 register struct re_guts *g;
1580 {
1581         register sop *scan;
1582         sop *start;
1583         register sop *newstart;
1584         register sopno newlen;
1585         register sop s;
1586         register char *cp;
1587         register sopno i;
1588
1589         /* avoid making error situations worse */
1590         if (p->error != 0)
1591                 return;
1592
1593         /* find the longest OCHAR sequence in strip */
1594         newlen = 0;
1595         scan = g->strip + 1;
1596         do {
1597                 s = *scan++;
1598                 switch (OP(s)) {
1599                 case OCHAR:             /* sequence member */
1600                         if (newlen == 0)                /* new sequence */
1601                                 newstart = scan - 1;
1602                         newlen++;
1603                         break;
1604                 case OPLUS_:            /* things that don't break one */
1605                 case OLPAREN:
1606                 case ORPAREN:
1607                         break;
1608                 case OQUEST_:           /* things that must be skipped */
1609                 case OCH_:
1610                         scan--;
1611                         do {
1612                                 scan += OPND(s);
1613                                 s = *scan;
1614                                 /* assert() interferes w debug printouts */
1615                                 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1616                                                         OP(s) != OOR2) {
1617                                         g->iflags |= BAD;
1618                                         return;
1619                                 }
1620                         } while (OP(s) != O_QUEST && OP(s) != O_CH);
1621                         /* fallthrough */
1622                 default:                /* things that break a sequence */
1623                         if (newlen > g->mlen) {         /* ends one */
1624                                 start = newstart;
1625                                 g->mlen = newlen;
1626                         }
1627                         newlen = 0;
1628                         break;
1629                 }
1630         } while (OP(s) != OEND);
1631
1632         if (g->mlen == 0)               /* there isn't one */
1633                 return;
1634
1635         /* turn it into a character string */
1636         g->must = malloc((size_t)g->mlen + 1);
1637         if (g->must == NULL) {          /* argh; just forget it */
1638                 g->mlen = 0;
1639                 return;
1640         }
1641         cp = g->must;
1642         scan = start;
1643         for (i = g->mlen; i > 0; i--) {
1644                 while (OP(s = *scan++) != OCHAR)
1645                         continue;
1646                 assert(cp < g->must + g->mlen);
1647                 *cp++ = (char)OPND(s);
1648         }
1649         assert(cp == g->must + g->mlen);
1650         *cp++ = '\0';           /* just on general principles */
1651 }
1652
1653 /*
1654  - pluscount - count + nesting
1655  == static sopno pluscount(register struct parse *p, register struct re_guts *g);
1656  */
1657 static sopno                    /* nesting depth */
1658 pluscount(p, g)
1659 struct parse *p;
1660 register struct re_guts *g;
1661 {
1662         register sop *scan;
1663         register sop s;
1664         register sopno plusnest = 0;
1665         register sopno maxnest = 0;
1666
1667         if (p->error != 0)
1668                 return(0);      /* there may not be an OEND */
1669
1670         scan = g->strip + 1;
1671         do {
1672                 s = *scan++;
1673                 switch (OP(s)) {
1674                 case OPLUS_:
1675                         plusnest++;
1676                         break;
1677                 case O_PLUS:
1678                         if (plusnest > maxnest)
1679                                 maxnest = plusnest;
1680                         plusnest--;
1681                         break;
1682                 }
1683         } while (OP(s) != OEND);
1684         if (plusnest != 0)
1685                 g->iflags |= BAD;
1686         return(maxnest);
1687 }