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.
6 * Copyright (c) 2011 The FreeBSD Foundation
8 * Portions of this software were developed by David Chisnall
9 * under sponsorship from the FreeBSD Foundation.
11 * This code is derived from software contributed to Berkeley by
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
17 * 1. Redistributions of source code must retain the above copyright
18 * notice, this list of conditions and the following disclaimer.
19 * 2. Redistributions in binary form must reproduce the above copyright
20 * notice, this list of conditions and the following disclaimer in the
21 * documentation and/or other materials provided with the distribution.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * @(#)regcomp.c 8.5 (Berkeley) 3/20/94
41 #if defined(LIBC_SCCS) && !defined(lint)
42 static char sccsid[] = "@(#)regcomp.c 8.5 (Berkeley) 3/20/94";
43 #endif /* LIBC_SCCS and not lint */
44 #include <sys/cdefs.h>
45 __FBSDID("$FreeBSD$");
47 #include <sys/types.h>
66 * parse structure, passed up and down to avoid global variables and
70 char *next; /* next character in RE */
71 char *end; /* end of string (-> NUL normally) */
72 int error; /* has an error been seen? */
73 sop *strip; /* malloced strip */
74 sopno ssize; /* malloced strip size (allocated) */
75 sopno slen; /* malloced strip length (used) */
76 int ncsalloc; /* number of csets allocated */
78 # define NPAREN 10 /* we need to remember () 1-9 for back refs */
79 sopno pbegin[NPAREN]; /* -> ( ([0] unused) */
80 sopno pend[NPAREN]; /* -> ) ([0] unused) */
83 /* ========= begin header generated by ./mkh ========= */
88 /* === regcomp.c === */
89 static void p_ere(struct parse *p, int stop);
90 static void p_ere_exp(struct parse *p);
91 static void p_str(struct parse *p);
92 static void p_bre(struct parse *p, int end1, int end2);
93 static int p_simp_re(struct parse *p, int starordinary);
94 static int p_count(struct parse *p);
95 static void p_bracket(struct parse *p);
96 static void p_b_term(struct parse *p, cset *cs);
97 static void p_b_cclass(struct parse *p, cset *cs);
98 static void p_b_eclass(struct parse *p, cset *cs);
99 static wint_t p_b_symbol(struct parse *p);
100 static wint_t p_b_coll_elem(struct parse *p, wint_t endc);
101 static wint_t othercase(wint_t ch);
102 static void bothcases(struct parse *p, wint_t ch);
103 static void ordinary(struct parse *p, wint_t ch);
104 static void nonnewline(struct parse *p);
105 static void repeat(struct parse *p, sopno start, int from, int to);
106 static int seterr(struct parse *p, int e);
107 static cset *allocset(struct parse *p);
108 static void freeset(struct parse *p, cset *cs);
109 static void CHadd(struct parse *p, cset *cs, wint_t ch);
110 static void CHaddrange(struct parse *p, cset *cs, wint_t min, wint_t max);
111 static void CHaddtype(struct parse *p, cset *cs, wctype_t wct);
112 static wint_t singleton(cset *cs);
113 static sopno dupl(struct parse *p, sopno start, sopno finish);
114 static void doemit(struct parse *p, sop op, size_t opnd);
115 static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
116 static void dofwd(struct parse *p, sopno pos, sop value);
117 static int enlarge(struct parse *p, sopno size);
118 static void stripsnug(struct parse *p, struct re_guts *g);
119 static void findmust(struct parse *p, struct re_guts *g);
120 static int altoffset(sop *scan, int offset);
121 static void computejumps(struct parse *p, struct re_guts *g);
122 static void computematchjumps(struct parse *p, struct re_guts *g);
123 static sopno pluscount(struct parse *p, struct re_guts *g);
124 static wint_t wgetnext(struct parse *p);
129 /* ========= end header generated by ./mkh ========= */
131 static char nuls[10]; /* place to point scanner in event of error */
134 * macros for use with parse structure
135 * BEWARE: these know that the parse structure is named `p' !!!
137 #define PEEK() (*p->next)
138 #define PEEK2() (*(p->next+1))
139 #define MORE() (p->next < p->end)
140 #define MORE2() (p->next+1 < p->end)
141 #define SEE(c) (MORE() && PEEK() == (c))
142 #define SEETWO(a, b) (MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
143 #define EAT(c) ((SEE(c)) ? (NEXT(), 1) : 0)
144 #define EATTWO(a, b) ((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
145 #define NEXT() (p->next++)
146 #define NEXT2() (p->next += 2)
147 #define NEXTn(n) (p->next += (n))
148 #define GETNEXT() (*p->next++)
149 #define WGETNEXT() wgetnext(p)
150 #define SETERROR(e) seterr(p, (e))
151 #define REQUIRE(co, e) ((co) || SETERROR(e))
152 #define MUSTSEE(c, e) (REQUIRE(MORE() && PEEK() == (c), e))
153 #define MUSTEAT(c, e) (REQUIRE(MORE() && GETNEXT() == (c), e))
154 #define MUSTNOTSEE(c, e) (REQUIRE(!MORE() || PEEK() != (c), e))
155 #define EMIT(op, sopnd) doemit(p, (sop)(op), (size_t)(sopnd))
156 #define INSERT(op, pos) doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
157 #define AHEAD(pos) dofwd(p, pos, HERE()-(pos))
158 #define ASTERN(sop, pos) EMIT(sop, HERE()-pos)
159 #define HERE() (p->slen)
160 #define THERE() (p->slen - 1)
161 #define THERETHERE() (p->slen - 2)
162 #define DROP(n) (p->slen -= (n))
165 static int never = 0; /* for use in asserts; shuts lint up */
167 #define never 0 /* some <assert.h>s have bugs too */
170 /* Macro used by computejump()/computematchjump() */
171 #define MIN(a,b) ((a)<(b)?(a):(b))
174 - regcomp - interface for parser and compilation
175 = extern int regcomp(regex_t *, const char *, int);
176 = #define REG_BASIC 0000
177 = #define REG_EXTENDED 0001
178 = #define REG_ICASE 0002
179 = #define REG_NOSUB 0004
180 = #define REG_NEWLINE 0010
181 = #define REG_NOSPEC 0020
182 = #define REG_PEND 0040
183 = #define REG_DUMP 0200
185 int /* 0 success, otherwise REG_something */
186 regcomp(regex_t * __restrict preg,
187 const char * __restrict pattern,
192 struct parse *p = &pa;
196 # define GOODFLAGS(f) (f)
198 # define GOODFLAGS(f) ((f)&~REG_DUMP)
201 cflags = GOODFLAGS(cflags);
202 if ((cflags®_EXTENDED) && (cflags®_NOSPEC))
205 if (cflags®_PEND) {
206 if (preg->re_endp < pattern)
208 len = preg->re_endp - pattern;
210 len = strlen((char *)pattern);
212 /* do the mallocs early so failure handling is easy */
213 g = (struct re_guts *)malloc(sizeof(struct re_guts));
216 p->ssize = len/(size_t)2*(size_t)3 + (size_t)1; /* ugh */
217 p->strip = (sop *)malloc(p->ssize * sizeof(sop));
219 if (p->strip == NULL) {
226 p->next = (char *)pattern; /* convenience; we do not modify it */
227 p->end = p->next + len;
230 for (i = 0; i < NPAREN; i++) {
250 g->firststate = THERE();
251 if (cflags®_EXTENDED)
253 else if (cflags®_NOSPEC)
258 g->laststate = THERE();
260 /* tidy up loose ends and fill things in */
263 /* only use Boyer-Moore algorithm if the pattern is bigger
264 * than three characters
268 computematchjumps(p, g);
269 if(g->matchjump == NULL && g->charjump != NULL) {
274 g->nplus = pluscount(p, g);
276 preg->re_nsub = g->nsub;
278 preg->re_magic = MAGIC1;
280 /* not debugging, so can't rely on the assert() in regexec() */
282 SETERROR(REG_ASSERT);
285 /* win or lose, we're done */
286 if (p->error != 0) /* lose */
292 - p_ere - ERE parser top level, concatenation and alternation
293 == static void p_ere(struct parse *p, int_t stop);
296 p_ere(struct parse *p,
297 int stop) /* character this ERE should end at */
303 int first = 1; /* is this the first alternative? */
306 /* do a bunch of concatenated expressions */
308 while (MORE() && (c = PEEK()) != '|' && c != stop)
310 (void)REQUIRE(HERE() != conc, REG_EMPTY); /* require nonempty */
313 break; /* NOTE BREAK OUT */
316 INSERT(OCH_, conc); /* offset is wrong */
321 ASTERN(OOR1, prevback);
323 AHEAD(prevfwd); /* fix previous offset */
325 EMIT(OOR2, 0); /* offset is very wrong */
328 if (!first) { /* tail-end fixups */
330 ASTERN(O_CH, prevback);
333 assert(!MORE() || SEE(stop));
337 - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
338 == static void p_ere_exp(struct parse *p);
341 p_ere_exp(struct parse *p)
351 assert(MORE()); /* caller should have ensured this */
357 (void)REQUIRE(MORE(), REG_EPAREN);
361 p->pbegin[subno] = HERE();
362 EMIT(OLPAREN, subno);
365 if (subno < NPAREN) {
366 p->pend[subno] = HERE();
367 assert(p->pend[subno] != 0);
369 EMIT(ORPAREN, subno);
370 (void)MUSTEAT(')', REG_EPAREN);
372 #ifndef POSIX_MISTAKE
373 case ')': /* happens only if no current unmatched ( */
375 * You may ask, why the ifndef? Because I didn't notice
376 * this until slightly too late for 1003.2, and none of the
377 * other 1003.2 regular-expression reviewers noticed it at
378 * all. So an unmatched ) is legal POSIX, at least until
379 * we can get it fixed.
381 SETERROR(REG_EPAREN);
386 p->g->iflags |= USEBOL;
392 p->g->iflags |= USEEOL;
401 SETERROR(REG_BADRPT);
404 if (p->g->cflags®_NEWLINE)
413 (void)REQUIRE(MORE(), REG_EESCAPE);
417 case '{': /* okay as ordinary except if digit follows */
418 (void)REQUIRE(!MORE() || !isdigit((uch)PEEK()), REG_BADRPT);
430 /* we call { a repetition if followed by a digit */
431 if (!( c == '*' || c == '+' || c == '?' ||
432 (c == '{' && MORE2() && isdigit((uch)PEEK2())) ))
433 return; /* no repetition, we're done */
436 (void)REQUIRE(!wascaret, REG_BADRPT);
438 case '*': /* implemented as +? */
439 /* this case does not require the (y|) trick, noKLUDGE */
442 INSERT(OQUEST_, pos);
443 ASTERN(O_QUEST, pos);
450 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
451 INSERT(OCH_, pos); /* offset slightly wrong */
452 ASTERN(OOR1, pos); /* this one's right */
453 AHEAD(pos); /* fix the OCH_ */
454 EMIT(OOR2, 0); /* offset very wrong... */
455 AHEAD(THERE()); /* ...so fix it */
456 ASTERN(O_CH, THERETHERE());
461 if (isdigit((uch)PEEK())) {
463 (void)REQUIRE(count <= count2, REG_BADBR);
464 } else /* single number with comma */
466 } else /* just a single number */
468 repeat(p, pos, count, count2);
469 if (!EAT('}')) { /* error heuristics */
470 while (MORE() && PEEK() != '}')
472 (void)REQUIRE(MORE(), REG_EBRACE);
481 if (!( c == '*' || c == '+' || c == '?' ||
482 (c == '{' && MORE2() && isdigit((uch)PEEK2())) ) )
484 SETERROR(REG_BADRPT);
488 - p_str - string (no metacharacters) "parser"
489 == static void p_str(struct parse *p);
492 p_str(struct parse *p)
494 (void)REQUIRE(MORE(), REG_EMPTY);
496 ordinary(p, WGETNEXT());
500 - p_bre - BRE parser top level, anchoring and concatenation
501 == static void p_bre(struct parse *p, int end1, \
503 * Giving end1 as OUT essentially eliminates the end1/end2 check.
505 * This implementation is a bit of a kludge, in that a trailing $ is first
506 * taken as an ordinary character and then revised to be an anchor.
507 * The amount of lookahead needed to avoid this kludge is excessive.
510 p_bre(struct parse *p,
511 int end1, /* first terminating character */
512 int end2) /* second terminating character */
514 sopno start = HERE();
515 int first = 1; /* first subexpression? */
520 p->g->iflags |= USEBOL;
523 while (MORE() && !SEETWO(end1, end2)) {
524 wasdollar = p_simp_re(p, first);
527 if (wasdollar) { /* oops, that was a trailing anchor */
530 p->g->iflags |= USEEOL;
534 (void)REQUIRE(HERE() != start, REG_EMPTY); /* require nonempty */
538 - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
539 == static int p_simp_re(struct parse *p, int starordinary);
541 static int /* was the simple RE an unbackslashed $? */
542 p_simp_re(struct parse *p,
543 int starordinary) /* is a leading * an ordinary character? */
552 # define BACKSL (1<<CHAR_BIT)
554 pos = HERE(); /* repetion op, if any, covers from here */
556 assert(MORE()); /* caller should have ensured this */
559 (void)REQUIRE(MORE(), REG_EESCAPE);
560 c = BACKSL | GETNEXT();
564 if (p->g->cflags®_NEWLINE)
573 SETERROR(REG_BADRPT);
579 p->pbegin[subno] = HERE();
580 EMIT(OLPAREN, subno);
581 /* the MORE here is an error heuristic */
582 if (MORE() && !SEETWO('\\', ')'))
584 if (subno < NPAREN) {
585 p->pend[subno] = HERE();
586 assert(p->pend[subno] != 0);
588 EMIT(ORPAREN, subno);
589 (void)REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
591 case BACKSL|')': /* should not get here -- must be user */
593 SETERROR(REG_EPAREN);
604 i = (c&~BACKSL) - '0';
606 if (p->pend[i] != 0) {
607 assert(i <= p->g->nsub);
609 assert(p->pbegin[i] != 0);
610 assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
611 assert(OP(p->strip[p->pend[i]]) == ORPAREN);
612 (void) dupl(p, p->pbegin[i]+1, p->pend[i]);
615 SETERROR(REG_ESUBREG);
619 (void)REQUIRE(starordinary, REG_BADRPT);
628 if (EAT('*')) { /* implemented as +? */
629 /* this case does not require the (y|) trick, noKLUDGE */
632 INSERT(OQUEST_, pos);
633 ASTERN(O_QUEST, pos);
634 } else if (EATTWO('\\', '{')) {
637 if (MORE() && isdigit((uch)PEEK())) {
639 (void)REQUIRE(count <= count2, REG_BADBR);
640 } else /* single number with comma */
642 } else /* just a single number */
644 repeat(p, pos, count, count2);
645 if (!EATTWO('\\', '}')) { /* error heuristics */
646 while (MORE() && !SEETWO('\\', '}'))
648 (void)REQUIRE(MORE(), REG_EBRACE);
651 } else if (c == '$') /* $ (but not \$) ends it */
658 - p_count - parse a repetition count
659 == static int p_count(struct parse *p);
661 static int /* the value */
662 p_count(struct parse *p)
667 while (MORE() && isdigit((uch)PEEK()) && count <= DUPMAX) {
668 count = count*10 + (GETNEXT() - '0');
672 (void)REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
677 - p_bracket - parse a bracketed character list
678 == static void p_bracket(struct parse *p);
681 p_bracket(struct parse *p)
686 /* Dept of Truly Sickening Special-Case Kludges */
687 if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
692 if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
698 if ((cs = allocset(p)) == NULL)
701 if (p->g->cflags®_ICASE)
709 while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
713 (void)MUSTEAT(']', REG_EBRACK);
715 if (p->error != 0) /* don't mess things up further */
718 if (cs->invert && p->g->cflags®_NEWLINE)
719 cs->bmp['\n' >> 3] |= 1 << ('\n' & 7);
721 if ((ch = singleton(cs)) != OUT) { /* optimize singleton sets */
725 EMIT(OANYOF, (int)(cs - p->g->sets));
729 - p_b_term - parse one term of a bracketed character list
730 == static void p_b_term(struct parse *p, cset *cs);
733 p_b_term(struct parse *p, cset *cs)
736 wint_t start, finish;
738 struct xlocale_collate *table =
739 (struct xlocale_collate*)__get_locale()->components[XLC_COLLATE];
741 /* classify what we've got */
742 switch ((MORE()) ? PEEK() : '\0') {
744 c = (MORE2()) ? PEEK2() : '\0';
747 SETERROR(REG_ERANGE);
748 return; /* NOTE RETURN */
756 case ':': /* character class */
758 (void)REQUIRE(MORE(), REG_EBRACK);
760 (void)REQUIRE(c != '-' && c != ']', REG_ECTYPE);
762 (void)REQUIRE(MORE(), REG_EBRACK);
763 (void)REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
765 case '=': /* equivalence class */
767 (void)REQUIRE(MORE(), REG_EBRACK);
769 (void)REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
771 (void)REQUIRE(MORE(), REG_EBRACK);
772 (void)REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
774 default: /* symbol, ordinary character, or range */
775 start = p_b_symbol(p);
776 if (SEE('-') && MORE2() && PEEK2() != ']') {
782 finish = p_b_symbol(p);
788 if (table->__collate_load_error) {
789 (void)REQUIRE((uch)start <= (uch)finish, REG_ERANGE);
790 CHaddrange(p, cs, start, finish);
792 (void)REQUIRE(__collate_range_cmp(table, start, finish) <= 0, REG_ERANGE);
793 for (i = 0; i <= UCHAR_MAX; i++) {
794 if ( __collate_range_cmp(table, start, i) <= 0
795 && __collate_range_cmp(table, i, finish) <= 0
806 - p_b_cclass - parse a character-class name and deal with it
807 == static void p_b_cclass(struct parse *p, cset *cs);
810 p_b_cclass(struct parse *p, cset *cs)
817 while (MORE() && isalpha((uch)PEEK()))
820 if (len >= sizeof(clname) - 1) {
821 SETERROR(REG_ECTYPE);
824 memcpy(clname, sp, len);
826 if ((wct = wctype(clname)) == 0) {
827 SETERROR(REG_ECTYPE);
830 CHaddtype(p, cs, wct);
834 - p_b_eclass - parse an equivalence-class name and deal with it
835 == static void p_b_eclass(struct parse *p, cset *cs);
837 * This implementation is incomplete. xxx
840 p_b_eclass(struct parse *p, cset *cs)
844 c = p_b_coll_elem(p, '=');
849 - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
850 == static wint_t p_b_symbol(struct parse *p);
852 static wint_t /* value of symbol */
853 p_b_symbol(struct parse *p)
857 (void)REQUIRE(MORE(), REG_EBRACK);
858 if (!EATTWO('[', '.'))
861 /* collating symbol */
862 value = p_b_coll_elem(p, '.');
863 (void)REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
868 - p_b_coll_elem - parse a collating-element name and look it up
869 == static wint_t p_b_coll_elem(struct parse *p, wint_t endc);
871 static wint_t /* value of collating element */
872 p_b_coll_elem(struct parse *p,
873 wint_t endc) /* name ended by endc,']' */
882 while (MORE() && !SEETWO(endc, ']'))
885 SETERROR(REG_EBRACK);
889 for (cp = cnames; cp->name != NULL; cp++)
890 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
891 return(cp->code); /* known name */
892 memset(&mbs, 0, sizeof(mbs));
893 if ((clen = mbrtowc(&wc, sp, len, &mbs)) == len)
894 return (wc); /* single character */
895 else if (clen == (size_t)-1 || clen == (size_t)-2)
896 SETERROR(REG_ILLSEQ);
898 SETERROR(REG_ECOLLATE); /* neither */
903 - othercase - return the case counterpart of an alphabetic
904 == static wint_t othercase(wint_t ch);
906 static wint_t /* if no counterpart, return ch */
909 assert(iswalpha(ch));
911 return(towlower(ch));
912 else if (iswlower(ch))
913 return(towupper(ch));
914 else /* peculiar, but could happen */
919 - bothcases - emit a dualcase version of a two-case character
920 == static void bothcases(struct parse *p, wint_t ch);
922 * Boy, is this implementation ever a kludge...
925 bothcases(struct parse *p, wint_t ch)
927 char *oldnext = p->next;
928 char *oldend = p->end;
929 char bracket[3 + MB_LEN_MAX];
933 assert(othercase(ch) != ch); /* p_bracket() would recurse */
935 memset(&mbs, 0, sizeof(mbs));
936 n = wcrtomb(bracket, ch, &mbs);
937 assert(n != (size_t)-1);
939 bracket[n + 1] = '\0';
940 p->end = bracket+n+1;
942 assert(p->next == p->end);
948 - ordinary - emit an ordinary character
949 == static void ordinary(struct parse *p, wint_t ch);
952 ordinary(struct parse *p, wint_t ch)
956 if ((p->g->cflags®_ICASE) && iswalpha(ch) && othercase(ch) != ch)
958 else if ((ch & OPDMASK) == ch)
962 * Kludge: character is too big to fit into an OCHAR operand.
963 * Emit a singleton set.
965 if ((cs = allocset(p)) == NULL)
968 EMIT(OANYOF, (int)(cs - p->g->sets));
973 - nonnewline - emit REG_NEWLINE version of OANY
974 == static void nonnewline(struct parse *p);
976 * Boy, is this implementation ever a kludge...
979 nonnewline(struct parse *p)
981 char *oldnext = p->next;
982 char *oldend = p->end;
992 assert(p->next == bracket+3);
998 - repeat - generate code for a bounded repetition, recursively if needed
999 == static void repeat(struct parse *p, sopno start, int from, int to);
1002 repeat(struct parse *p,
1003 sopno start, /* operand from here to end of strip */
1004 int from, /* repeated from this number */
1005 int to) /* to this number of times (maybe INFINITY) */
1007 sopno finish = HERE();
1010 # define REP(f, t) ((f)*8 + (t))
1011 # define MAP(n) (((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
1014 if (p->error != 0) /* head off possible runaway recursion */
1019 switch (REP(MAP(from), MAP(to))) {
1020 case REP(0, 0): /* must be user doing this */
1021 DROP(finish-start); /* drop the operand */
1023 case REP(0, 1): /* as x{1,1}? */
1024 case REP(0, N): /* as x{1,n}? */
1025 case REP(0, INF): /* as x{1,}? */
1026 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1027 INSERT(OCH_, start); /* offset is wrong... */
1028 repeat(p, start+1, 1, to);
1029 ASTERN(OOR1, start);
1030 AHEAD(start); /* ... fix it */
1033 ASTERN(O_CH, THERETHERE());
1035 case REP(1, 1): /* trivial case */
1038 case REP(1, N): /* as x?x{1,n-1} */
1039 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1040 INSERT(OCH_, start);
1041 ASTERN(OOR1, start);
1043 EMIT(OOR2, 0); /* offset very wrong... */
1044 AHEAD(THERE()); /* ...so fix it */
1045 ASTERN(O_CH, THERETHERE());
1046 copy = dupl(p, start+1, finish+1);
1047 assert(copy == finish+4);
1048 repeat(p, copy, 1, to-1);
1050 case REP(1, INF): /* as x+ */
1051 INSERT(OPLUS_, start);
1052 ASTERN(O_PLUS, start);
1054 case REP(N, N): /* as xx{m-1,n-1} */
1055 copy = dupl(p, start, finish);
1056 repeat(p, copy, from-1, to-1);
1058 case REP(N, INF): /* as xx{n-1,INF} */
1059 copy = dupl(p, start, finish);
1060 repeat(p, copy, from-1, to);
1062 default: /* "can't happen" */
1063 SETERROR(REG_ASSERT); /* just in case */
1069 - wgetnext - helper function for WGETNEXT() macro. Gets the next wide
1070 - character from the parse struct, signals a REG_ILLSEQ error if the
1071 - character can't be converted. Returns the number of bytes consumed.
1074 wgetnext(struct parse *p)
1080 memset(&mbs, 0, sizeof(mbs));
1081 n = mbrtowc(&wc, p->next, p->end - p->next, &mbs);
1082 if (n == (size_t)-1 || n == (size_t)-2) {
1083 SETERROR(REG_ILLSEQ);
1093 - seterr - set an error condition
1094 == static int seterr(struct parse *p, int e);
1096 static int /* useless but makes type checking happy */
1097 seterr(struct parse *p, int e)
1099 if (p->error == 0) /* keep earliest error condition */
1101 p->next = nuls; /* try to bring things to a halt */
1103 return(0); /* make the return value well-defined */
1107 - allocset - allocate a set of characters for []
1108 == static cset *allocset(struct parse *p);
1111 allocset(struct parse *p)
1115 ncs = realloc(p->g->sets, (p->g->ncsets + 1) * sizeof(*ncs));
1117 SETERROR(REG_ESPACE);
1121 cs = &p->g->sets[p->g->ncsets++];
1122 memset(cs, 0, sizeof(*cs));
1128 - freeset - free a now-unused set
1129 == static void freeset(struct parse *p, cset *cs);
1132 freeset(struct parse *p, cset *cs)
1134 cset *top = &p->g->sets[p->g->ncsets];
1139 memset(cs, 0, sizeof(*cs));
1140 if (cs == top-1) /* recover only the easy case */
1145 - singleton - Determine whether a set contains only one character,
1146 - returning it if so, otherwise returning OUT.
1153 for (i = n = 0; i < NC; i++)
1160 if (cs->nwides == 1 && cs->nranges == 0 && cs->ntypes == 0 &&
1162 return (cs->wides[0]);
1163 /* Don't bother handling the other cases. */
1168 - CHadd - add character to character set.
1171 CHadd(struct parse *p, cset *cs, wint_t ch)
1173 wint_t nch, *newwides;
1176 cs->bmp[ch >> 3] |= 1 << (ch & 7);
1178 newwides = realloc(cs->wides, (cs->nwides + 1) *
1179 sizeof(*cs->wides));
1180 if (newwides == NULL) {
1181 SETERROR(REG_ESPACE);
1184 cs->wides = newwides;
1185 cs->wides[cs->nwides++] = ch;
1188 if ((nch = towlower(ch)) < NC)
1189 cs->bmp[nch >> 3] |= 1 << (nch & 7);
1190 if ((nch = towupper(ch)) < NC)
1191 cs->bmp[nch >> 3] |= 1 << (nch & 7);
1196 - CHaddrange - add all characters in the range [min,max] to a character set.
1199 CHaddrange(struct parse *p, cset *cs, wint_t min, wint_t max)
1203 for (; min < NC && min <= max; min++)
1207 newranges = realloc(cs->ranges, (cs->nranges + 1) *
1208 sizeof(*cs->ranges));
1209 if (newranges == NULL) {
1210 SETERROR(REG_ESPACE);
1213 cs->ranges = newranges;
1214 cs->ranges[cs->nranges].min = min;
1215 cs->ranges[cs->nranges].max = max;
1220 - CHaddtype - add all characters of a certain type to a character set.
1223 CHaddtype(struct parse *p, cset *cs, wctype_t wct)
1228 for (i = 0; i < NC; i++)
1229 if (iswctype(i, wct))
1231 newtypes = realloc(cs->types, (cs->ntypes + 1) *
1232 sizeof(*cs->types));
1233 if (newtypes == NULL) {
1234 SETERROR(REG_ESPACE);
1237 cs->types = newtypes;
1238 cs->types[cs->ntypes++] = wct;
1242 - dupl - emit a duplicate of a bunch of sops
1243 == static sopno dupl(struct parse *p, sopno start, sopno finish);
1245 static sopno /* start of duplicate */
1246 dupl(struct parse *p,
1247 sopno start, /* from here */
1248 sopno finish) /* to this less one */
1251 sopno len = finish - start;
1253 assert(finish >= start);
1256 if (!enlarge(p, p->ssize + len)) /* this many unexpected additions */
1258 (void) memcpy((char *)(p->strip + p->slen),
1259 (char *)(p->strip + start), (size_t)len*sizeof(sop));
1265 - doemit - emit a strip operator
1266 == static void doemit(struct parse *p, sop op, size_t opnd);
1268 * It might seem better to implement this as a macro with a function as
1269 * hard-case backup, but it's just too big and messy unless there are
1270 * some changes to the data structures. Maybe later.
1273 doemit(struct parse *p, sop op, size_t opnd)
1275 /* avoid making error situations worse */
1279 /* deal with oversize operands ("can't happen", more or less) */
1280 assert(opnd < 1<<OPSHIFT);
1282 /* deal with undersized strip */
1283 if (p->slen >= p->ssize)
1284 if (!enlarge(p, (p->ssize+1) / 2 * 3)) /* +50% */
1287 /* finally, it's all reduced to the easy case */
1288 p->strip[p->slen++] = SOP(op, opnd);
1292 - doinsert - insert a sop into the strip
1293 == static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
1296 doinsert(struct parse *p, sop op, size_t opnd, sopno pos)
1302 /* avoid making error situations worse */
1307 EMIT(op, opnd); /* do checks, ensure space */
1308 assert(HERE() == sn+1);
1311 /* adjust paren pointers */
1313 for (i = 1; i < NPAREN; i++) {
1314 if (p->pbegin[i] >= pos) {
1317 if (p->pend[i] >= pos) {
1322 memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1323 (HERE()-pos-1)*sizeof(sop));
1328 - dofwd - complete a forward reference
1329 == static void dofwd(struct parse *p, sopno pos, sop value);
1332 dofwd(struct parse *p, sopno pos, sop value)
1334 /* avoid making error situations worse */
1338 assert(value < 1<<OPSHIFT);
1339 p->strip[pos] = OP(p->strip[pos]) | value;
1343 - enlarge - enlarge the strip
1344 == static int enlarge(struct parse *p, sopno size);
1347 enlarge(struct parse *p, sopno size)
1351 if (p->ssize >= size)
1354 sp = (sop *)realloc(p->strip, size*sizeof(sop));
1356 SETERROR(REG_ESPACE);
1365 - stripsnug - compact the strip
1366 == static void stripsnug(struct parse *p, struct re_guts *g);
1369 stripsnug(struct parse *p, struct re_guts *g)
1371 g->nstates = p->slen;
1372 g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
1373 if (g->strip == NULL) {
1374 SETERROR(REG_ESPACE);
1375 g->strip = p->strip;
1380 - findmust - fill in must and mlen with longest mandatory literal string
1381 == static void findmust(struct parse *p, struct re_guts *g);
1383 * This algorithm could do fancy things like analyzing the operands of |
1384 * for common subsequences. Someday. This code is simple and finds most
1385 * of the interesting cases.
1387 * Note that must and mlen got initialized during setup.
1390 findmust(struct parse *p, struct re_guts *g)
1399 char buf[MB_LEN_MAX];
1403 /* avoid making error situations worse */
1408 * It's not generally safe to do a ``char'' substring search on
1409 * multibyte character strings, but it's safe for at least
1410 * UTF-8 (see RFC 3629).
1412 if (MB_CUR_MAX > 1 &&
1413 strcmp(_CurrentRuneLocale->__encoding, "UTF-8") != 0)
1416 /* find the longest OCHAR sequence in strip */
1420 scan = g->strip + 1;
1424 case OCHAR: /* sequence member */
1425 if (newlen == 0) { /* new sequence */
1426 memset(&mbs, 0, sizeof(mbs));
1427 newstart = scan - 1;
1429 clen = wcrtomb(buf, OPND(s), &mbs);
1430 if (clen == (size_t)-1)
1434 case OPLUS_: /* things that don't break one */
1438 case OQUEST_: /* things that must be skipped */
1440 offset = altoffset(scan, offset);
1445 /* assert() interferes w debug printouts */
1446 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1451 } while (OP(s) != O_QUEST && OP(s) != O_CH);
1453 case OBOW: /* things that break a sequence */
1460 if (newlen > g->mlen) { /* ends one */
1464 g->moffset += offset;
1467 g->moffset = offset;
1475 if (newlen > g->mlen) { /* ends one */
1479 g->moffset += offset;
1482 g->moffset = offset;
1491 case OANYOF: /* may or may not invalidate offset */
1492 /* First, everything as OANY */
1493 if (newlen > g->mlen) { /* ends one */
1497 g->moffset += offset;
1500 g->moffset = offset;
1511 /* Anything here makes it impossible or too hard
1512 * to calculate the offset -- so we give up;
1513 * save the last known good offset, in case the
1514 * must sequence doesn't occur later.
1516 if (newlen > g->mlen) { /* ends one */
1520 g->moffset += offset;
1522 g->moffset = offset;
1528 } while (OP(s) != OEND);
1530 if (g->mlen == 0) { /* there isn't one */
1535 /* turn it into a character string */
1536 g->must = malloc((size_t)g->mlen + 1);
1537 if (g->must == NULL) { /* argh; just forget it */
1544 memset(&mbs, 0, sizeof(mbs));
1545 while (cp < g->must + g->mlen) {
1546 while (OP(s = *scan++) != OCHAR)
1548 clen = wcrtomb(cp, OPND(s), &mbs);
1549 assert(clen != (size_t)-1);
1552 assert(cp == g->must + g->mlen);
1553 *cp++ = '\0'; /* just on general principles */
1557 - altoffset - choose biggest offset among multiple choices
1558 == static int altoffset(sop *scan, int offset);
1560 * Compute, recursively if necessary, the largest offset among multiple
1564 altoffset(sop *scan, int offset)
1570 /* If we gave up already on offsets, return */
1577 while (OP(s) != O_QUEST && OP(s) != O_CH) {
1586 try = altoffset(scan, try);
1593 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1596 } while (OP(s) != O_QUEST && OP(s) != O_CH);
1597 /* We must skip to the next position, or we'll
1598 * leave altoffset() too early.
1624 return largest+offset;
1628 - computejumps - compute char jumps for BM scan
1629 == static void computejumps(struct parse *p, struct re_guts *g);
1631 * This algorithm assumes g->must exists and is has size greater than
1632 * zero. It's based on the algorithm found on Computer Algorithms by
1635 * A char jump is the number of characters one needs to jump based on
1636 * the value of the character from the text that was mismatched.
1639 computejumps(struct parse *p, struct re_guts *g)
1644 /* Avoid making errors worse */
1648 g->charjump = (int*) malloc((NC + 1) * sizeof(int));
1649 if (g->charjump == NULL) /* Not a fatal error */
1651 /* Adjust for signed chars, if necessary */
1652 g->charjump = &g->charjump[-(CHAR_MIN)];
1654 /* If the character does not exist in the pattern, the jump
1655 * is equal to the number of characters in the pattern.
1657 for (ch = CHAR_MIN; ch < (CHAR_MAX + 1); ch++)
1658 g->charjump[ch] = g->mlen;
1660 /* If the character does exist, compute the jump that would
1661 * take us to the last character in the pattern equal to it
1662 * (notice that we match right to left, so that last character
1663 * is the first one that would be matched).
1665 for (mindex = 0; mindex < g->mlen; mindex++)
1666 g->charjump[(int)g->must[mindex]] = g->mlen - mindex - 1;
1670 - computematchjumps - compute match jumps for BM scan
1671 == static void computematchjumps(struct parse *p, struct re_guts *g);
1673 * This algorithm assumes g->must exists and is has size greater than
1674 * zero. It's based on the algorithm found on Computer Algorithms by
1677 * A match jump is the number of characters one needs to advance based
1678 * on the already-matched suffix.
1679 * Notice that all values here are minus (g->mlen-1), because of the way
1680 * the search algorithm works.
1683 computematchjumps(struct parse *p, struct re_guts *g)
1685 int mindex; /* General "must" iterator */
1686 int suffix; /* Keeps track of matching suffix */
1687 int ssuffix; /* Keeps track of suffixes' suffix */
1688 int* pmatches; /* pmatches[k] points to the next i
1689 * such that i+1...mlen is a substring
1690 * of k+1...k+mlen-i-1
1693 /* Avoid making errors worse */
1697 pmatches = (int*) malloc(g->mlen * sizeof(unsigned int));
1698 if (pmatches == NULL) {
1699 g->matchjump = NULL;
1703 g->matchjump = (int*) malloc(g->mlen * sizeof(unsigned int));
1704 if (g->matchjump == NULL) /* Not a fatal error */
1707 /* Set maximum possible jump for each character in the pattern */
1708 for (mindex = 0; mindex < g->mlen; mindex++)
1709 g->matchjump[mindex] = 2*g->mlen - mindex - 1;
1711 /* Compute pmatches[] */
1712 for (mindex = g->mlen - 1, suffix = g->mlen; mindex >= 0;
1713 mindex--, suffix--) {
1714 pmatches[mindex] = suffix;
1716 /* If a mismatch is found, interrupting the substring,
1717 * compute the matchjump for that position. If no
1718 * mismatch is found, then a text substring mismatched
1719 * against the suffix will also mismatch against the
1722 while (suffix < g->mlen
1723 && g->must[mindex] != g->must[suffix]) {
1724 g->matchjump[suffix] = MIN(g->matchjump[suffix],
1725 g->mlen - mindex - 1);
1726 suffix = pmatches[suffix];
1730 /* Compute the matchjump up to the last substring found to jump
1731 * to the beginning of the largest must pattern prefix matching
1734 for (mindex = 0; mindex <= suffix; mindex++)
1735 g->matchjump[mindex] = MIN(g->matchjump[mindex],
1736 g->mlen + suffix - mindex);
1738 ssuffix = pmatches[suffix];
1739 while (suffix < g->mlen) {
1740 while (suffix <= ssuffix && suffix < g->mlen) {
1741 g->matchjump[suffix] = MIN(g->matchjump[suffix],
1742 g->mlen + ssuffix - suffix);
1745 if (suffix < g->mlen)
1746 ssuffix = pmatches[ssuffix];
1753 - pluscount - count + nesting
1754 == static sopno pluscount(struct parse *p, struct re_guts *g);
1756 static sopno /* nesting depth */
1757 pluscount(struct parse *p, struct re_guts *g)
1765 return(0); /* there may not be an OEND */
1767 scan = g->strip + 1;
1775 if (plusnest > maxnest)
1780 } while (OP(s) != OEND);