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;
197 # define GOODFLAGS(f) (f)
199 # define GOODFLAGS(f) ((f)&~REG_DUMP)
202 cflags = GOODFLAGS(cflags);
203 if ((cflags®_EXTENDED) && (cflags®_NOSPEC))
206 if (cflags®_PEND) {
207 if (preg->re_endp < pattern)
209 len = preg->re_endp - pattern;
211 len = strlen((char *)pattern);
213 /* do the mallocs early so failure handling is easy */
214 g = (struct re_guts *)malloc(sizeof(struct re_guts));
218 * Limit the pattern space to avoid a 32-bit overflow on buffer
219 * extension. Also avoid any signed overflow in case of conversion
220 * so make the real limit based on a 31-bit overflow.
222 * Likely not applicable on 64-bit systems but handle the case
223 * generically (who are we to stop people from using ~715MB+
226 maxlen = ((size_t)-1 >> 1) / sizeof(sop) * 2 / 3;
231 p->ssize = len/(size_t)2*(size_t)3 + (size_t)1; /* ugh */
232 assert(p->ssize >= len);
234 p->strip = (sop *)malloc(p->ssize * sizeof(sop));
236 if (p->strip == NULL) {
243 p->next = (char *)pattern; /* convenience; we do not modify it */
244 p->end = p->next + len;
247 for (i = 0; i < NPAREN; i++) {
267 g->firststate = THERE();
268 if (cflags®_EXTENDED)
270 else if (cflags®_NOSPEC)
275 g->laststate = THERE();
277 /* tidy up loose ends and fill things in */
280 /* only use Boyer-Moore algorithm if the pattern is bigger
281 * than three characters
285 computematchjumps(p, g);
286 if(g->matchjump == NULL && g->charjump != NULL) {
291 g->nplus = pluscount(p, g);
293 preg->re_nsub = g->nsub;
295 preg->re_magic = MAGIC1;
297 /* not debugging, so can't rely on the assert() in regexec() */
299 SETERROR(REG_ASSERT);
302 /* win or lose, we're done */
303 if (p->error != 0) /* lose */
309 - p_ere - ERE parser top level, concatenation and alternation
310 == static void p_ere(struct parse *p, int_t stop);
313 p_ere(struct parse *p,
314 int stop) /* character this ERE should end at */
320 int first = 1; /* is this the first alternative? */
323 /* do a bunch of concatenated expressions */
325 while (MORE() && (c = PEEK()) != '|' && c != stop)
327 (void)REQUIRE(HERE() != conc, REG_EMPTY); /* require nonempty */
330 break; /* NOTE BREAK OUT */
333 INSERT(OCH_, conc); /* offset is wrong */
338 ASTERN(OOR1, prevback);
340 AHEAD(prevfwd); /* fix previous offset */
342 EMIT(OOR2, 0); /* offset is very wrong */
345 if (!first) { /* tail-end fixups */
347 ASTERN(O_CH, prevback);
350 assert(!MORE() || SEE(stop));
354 - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
355 == static void p_ere_exp(struct parse *p);
358 p_ere_exp(struct parse *p)
368 assert(MORE()); /* caller should have ensured this */
374 (void)REQUIRE(MORE(), REG_EPAREN);
378 p->pbegin[subno] = HERE();
379 EMIT(OLPAREN, subno);
382 if (subno < NPAREN) {
383 p->pend[subno] = HERE();
384 assert(p->pend[subno] != 0);
386 EMIT(ORPAREN, subno);
387 (void)MUSTEAT(')', REG_EPAREN);
389 #ifndef POSIX_MISTAKE
390 case ')': /* happens only if no current unmatched ( */
392 * You may ask, why the ifndef? Because I didn't notice
393 * this until slightly too late for 1003.2, and none of the
394 * other 1003.2 regular-expression reviewers noticed it at
395 * all. So an unmatched ) is legal POSIX, at least until
396 * we can get it fixed.
398 SETERROR(REG_EPAREN);
403 p->g->iflags |= USEBOL;
409 p->g->iflags |= USEEOL;
418 SETERROR(REG_BADRPT);
421 if (p->g->cflags®_NEWLINE)
430 (void)REQUIRE(MORE(), REG_EESCAPE);
444 case '{': /* okay as ordinary except if digit follows */
445 (void)REQUIRE(!MORE() || !isdigit((uch)PEEK()), REG_BADRPT);
457 /* we call { a repetition if followed by a digit */
458 if (!( c == '*' || c == '+' || c == '?' ||
459 (c == '{' && MORE2() && isdigit((uch)PEEK2())) ))
460 return; /* no repetition, we're done */
463 (void)REQUIRE(!wascaret, REG_BADRPT);
465 case '*': /* implemented as +? */
466 /* this case does not require the (y|) trick, noKLUDGE */
469 INSERT(OQUEST_, pos);
470 ASTERN(O_QUEST, pos);
477 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
478 INSERT(OCH_, pos); /* offset slightly wrong */
479 ASTERN(OOR1, pos); /* this one's right */
480 AHEAD(pos); /* fix the OCH_ */
481 EMIT(OOR2, 0); /* offset very wrong... */
482 AHEAD(THERE()); /* ...so fix it */
483 ASTERN(O_CH, THERETHERE());
488 if (isdigit((uch)PEEK())) {
490 (void)REQUIRE(count <= count2, REG_BADBR);
491 } else /* single number with comma */
493 } else /* just a single number */
495 repeat(p, pos, count, count2);
496 if (!EAT('}')) { /* error heuristics */
497 while (MORE() && PEEK() != '}')
499 (void)REQUIRE(MORE(), REG_EBRACE);
508 if (!( c == '*' || c == '+' || c == '?' ||
509 (c == '{' && MORE2() && isdigit((uch)PEEK2())) ) )
511 SETERROR(REG_BADRPT);
515 - p_str - string (no metacharacters) "parser"
516 == static void p_str(struct parse *p);
519 p_str(struct parse *p)
521 (void)REQUIRE(MORE(), REG_EMPTY);
523 ordinary(p, WGETNEXT());
527 - p_bre - BRE parser top level, anchoring and concatenation
528 == static void p_bre(struct parse *p, int end1, \
530 * Giving end1 as OUT essentially eliminates the end1/end2 check.
532 * This implementation is a bit of a kludge, in that a trailing $ is first
533 * taken as an ordinary character and then revised to be an anchor.
534 * The amount of lookahead needed to avoid this kludge is excessive.
537 p_bre(struct parse *p,
538 int end1, /* first terminating character */
539 int end2) /* second terminating character */
541 sopno start = HERE();
542 int first = 1; /* first subexpression? */
547 p->g->iflags |= USEBOL;
550 while (MORE() && !SEETWO(end1, end2)) {
551 wasdollar = p_simp_re(p, first);
554 if (wasdollar) { /* oops, that was a trailing anchor */
557 p->g->iflags |= USEEOL;
561 (void)REQUIRE(HERE() != start, REG_EMPTY); /* require nonempty */
565 - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
566 == static int p_simp_re(struct parse *p, int starordinary);
568 static int /* was the simple RE an unbackslashed $? */
569 p_simp_re(struct parse *p,
570 int starordinary) /* is a leading * an ordinary character? */
579 # define BACKSL (1<<CHAR_BIT)
581 pos = HERE(); /* repetion op, if any, covers from here */
583 assert(MORE()); /* caller should have ensured this */
586 (void)REQUIRE(MORE(), REG_EESCAPE);
587 c = BACKSL | GETNEXT();
591 if (p->g->cflags®_NEWLINE)
606 SETERROR(REG_BADRPT);
612 p->pbegin[subno] = HERE();
613 EMIT(OLPAREN, subno);
614 /* the MORE here is an error heuristic */
615 if (MORE() && !SEETWO('\\', ')'))
617 if (subno < NPAREN) {
618 p->pend[subno] = HERE();
619 assert(p->pend[subno] != 0);
621 EMIT(ORPAREN, subno);
622 (void)REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
624 case BACKSL|')': /* should not get here -- must be user */
626 SETERROR(REG_EPAREN);
637 i = (c&~BACKSL) - '0';
639 if (p->pend[i] != 0) {
640 assert(i <= p->g->nsub);
642 assert(p->pbegin[i] != 0);
643 assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
644 assert(OP(p->strip[p->pend[i]]) == ORPAREN);
645 (void) dupl(p, p->pbegin[i]+1, p->pend[i]);
648 SETERROR(REG_ESUBREG);
652 (void)REQUIRE(starordinary, REG_BADRPT);
661 if (EAT('*')) { /* implemented as +? */
662 /* this case does not require the (y|) trick, noKLUDGE */
665 INSERT(OQUEST_, pos);
666 ASTERN(O_QUEST, pos);
667 } else if (EATTWO('\\', '{')) {
670 if (MORE() && isdigit((uch)PEEK())) {
672 (void)REQUIRE(count <= count2, REG_BADBR);
673 } else /* single number with comma */
675 } else /* just a single number */
677 repeat(p, pos, count, count2);
678 if (!EATTWO('\\', '}')) { /* error heuristics */
679 while (MORE() && !SEETWO('\\', '}'))
681 (void)REQUIRE(MORE(), REG_EBRACE);
684 } else if (c == '$') /* $ (but not \$) ends it */
691 - p_count - parse a repetition count
692 == static int p_count(struct parse *p);
694 static int /* the value */
695 p_count(struct parse *p)
700 while (MORE() && isdigit((uch)PEEK()) && count <= DUPMAX) {
701 count = count*10 + (GETNEXT() - '0');
705 (void)REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
710 - p_bracket - parse a bracketed character list
711 == static void p_bracket(struct parse *p);
714 p_bracket(struct parse *p)
719 /* Dept of Truly Sickening Special-Case Kludges */
720 if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
725 if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
731 if ((cs = allocset(p)) == NULL)
734 if (p->g->cflags®_ICASE)
742 while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
746 (void)MUSTEAT(']', REG_EBRACK);
748 if (p->error != 0) /* don't mess things up further */
751 if (cs->invert && p->g->cflags®_NEWLINE)
752 cs->bmp['\n' >> 3] |= 1 << ('\n' & 7);
754 if ((ch = singleton(cs)) != OUT) { /* optimize singleton sets */
758 EMIT(OANYOF, (int)(cs - p->g->sets));
762 - p_b_term - parse one term of a bracketed character list
763 == static void p_b_term(struct parse *p, cset *cs);
766 p_b_term(struct parse *p, cset *cs)
769 wint_t start, finish;
771 struct xlocale_collate *table =
772 (struct xlocale_collate*)__get_locale()->components[XLC_COLLATE];
774 /* classify what we've got */
775 switch ((MORE()) ? PEEK() : '\0') {
777 c = (MORE2()) ? PEEK2() : '\0';
780 SETERROR(REG_ERANGE);
781 return; /* NOTE RETURN */
788 case ':': /* character class */
790 (void)REQUIRE(MORE(), REG_EBRACK);
792 (void)REQUIRE(c != '-' && c != ']', REG_ECTYPE);
794 (void)REQUIRE(MORE(), REG_EBRACK);
795 (void)REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
797 case '=': /* equivalence class */
799 (void)REQUIRE(MORE(), REG_EBRACK);
801 (void)REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
803 (void)REQUIRE(MORE(), REG_EBRACK);
804 (void)REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
806 default: /* symbol, ordinary character, or range */
807 start = p_b_symbol(p);
808 if (SEE('-') && MORE2() && PEEK2() != ']') {
814 finish = p_b_symbol(p);
820 if (table->__collate_load_error) {
821 (void)REQUIRE((uch)start <= (uch)finish, REG_ERANGE);
822 CHaddrange(p, cs, start, finish);
824 (void)REQUIRE(__collate_range_cmp(table, start, finish) <= 0, REG_ERANGE);
825 for (i = 0; i <= UCHAR_MAX; i++) {
826 if ( __collate_range_cmp(table, start, i) <= 0
827 && __collate_range_cmp(table, i, finish) <= 0
838 - p_b_cclass - parse a character-class name and deal with it
839 == static void p_b_cclass(struct parse *p, cset *cs);
842 p_b_cclass(struct parse *p, cset *cs)
849 while (MORE() && isalpha((uch)PEEK()))
852 if (len >= sizeof(clname) - 1) {
853 SETERROR(REG_ECTYPE);
856 memcpy(clname, sp, len);
858 if ((wct = wctype(clname)) == 0) {
859 SETERROR(REG_ECTYPE);
862 CHaddtype(p, cs, wct);
866 - p_b_eclass - parse an equivalence-class name and deal with it
867 == static void p_b_eclass(struct parse *p, cset *cs);
869 * This implementation is incomplete. xxx
872 p_b_eclass(struct parse *p, cset *cs)
876 c = p_b_coll_elem(p, '=');
881 - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
882 == static wint_t p_b_symbol(struct parse *p);
884 static wint_t /* value of symbol */
885 p_b_symbol(struct parse *p)
889 (void)REQUIRE(MORE(), REG_EBRACK);
890 if (!EATTWO('[', '.'))
893 /* collating symbol */
894 value = p_b_coll_elem(p, '.');
895 (void)REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
900 - p_b_coll_elem - parse a collating-element name and look it up
901 == static wint_t p_b_coll_elem(struct parse *p, wint_t endc);
903 static wint_t /* value of collating element */
904 p_b_coll_elem(struct parse *p,
905 wint_t endc) /* name ended by endc,']' */
914 while (MORE() && !SEETWO(endc, ']'))
917 SETERROR(REG_EBRACK);
921 for (cp = cnames; cp->name != NULL; cp++)
922 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
923 return(cp->code); /* known name */
924 memset(&mbs, 0, sizeof(mbs));
925 if ((clen = mbrtowc(&wc, sp, len, &mbs)) == len)
926 return (wc); /* single character */
927 else if (clen == (size_t)-1 || clen == (size_t)-2)
928 SETERROR(REG_ILLSEQ);
930 SETERROR(REG_ECOLLATE); /* neither */
935 - othercase - return the case counterpart of an alphabetic
936 == static wint_t othercase(wint_t ch);
938 static wint_t /* if no counterpart, return ch */
941 assert(iswalpha(ch));
943 return(towlower(ch));
944 else if (iswlower(ch))
945 return(towupper(ch));
946 else /* peculiar, but could happen */
951 - bothcases - emit a dualcase version of a two-case character
952 == static void bothcases(struct parse *p, wint_t ch);
954 * Boy, is this implementation ever a kludge...
957 bothcases(struct parse *p, wint_t ch)
959 char *oldnext = p->next;
960 char *oldend = p->end;
961 char bracket[3 + MB_LEN_MAX];
965 assert(othercase(ch) != ch); /* p_bracket() would recurse */
967 memset(&mbs, 0, sizeof(mbs));
968 n = wcrtomb(bracket, ch, &mbs);
969 assert(n != (size_t)-1);
971 bracket[n + 1] = '\0';
972 p->end = bracket+n+1;
974 assert(p->next == p->end);
980 - ordinary - emit an ordinary character
981 == static void ordinary(struct parse *p, wint_t ch);
984 ordinary(struct parse *p, wint_t ch)
988 if ((p->g->cflags®_ICASE) && iswalpha(ch) && othercase(ch) != ch)
990 else if ((ch & OPDMASK) == ch)
994 * Kludge: character is too big to fit into an OCHAR operand.
995 * Emit a singleton set.
997 if ((cs = allocset(p)) == NULL)
1000 EMIT(OANYOF, (int)(cs - p->g->sets));
1005 - nonnewline - emit REG_NEWLINE version of OANY
1006 == static void nonnewline(struct parse *p);
1008 * Boy, is this implementation ever a kludge...
1011 nonnewline(struct parse *p)
1013 char *oldnext = p->next;
1014 char *oldend = p->end;
1024 assert(p->next == bracket+3);
1030 - repeat - generate code for a bounded repetition, recursively if needed
1031 == static void repeat(struct parse *p, sopno start, int from, int to);
1034 repeat(struct parse *p,
1035 sopno start, /* operand from here to end of strip */
1036 int from, /* repeated from this number */
1037 int to) /* to this number of times (maybe INFINITY) */
1039 sopno finish = HERE();
1042 # define REP(f, t) ((f)*8 + (t))
1043 # define MAP(n) (((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
1046 if (p->error != 0) /* head off possible runaway recursion */
1051 switch (REP(MAP(from), MAP(to))) {
1052 case REP(0, 0): /* must be user doing this */
1053 DROP(finish-start); /* drop the operand */
1055 case REP(0, 1): /* as x{1,1}? */
1056 case REP(0, N): /* as x{1,n}? */
1057 case REP(0, INF): /* as x{1,}? */
1058 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1059 INSERT(OCH_, start); /* offset is wrong... */
1060 repeat(p, start+1, 1, to);
1061 ASTERN(OOR1, start);
1062 AHEAD(start); /* ... fix it */
1065 ASTERN(O_CH, THERETHERE());
1067 case REP(1, 1): /* trivial case */
1070 case REP(1, N): /* as x?x{1,n-1} */
1071 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1072 INSERT(OCH_, start);
1073 ASTERN(OOR1, start);
1075 EMIT(OOR2, 0); /* offset very wrong... */
1076 AHEAD(THERE()); /* ...so fix it */
1077 ASTERN(O_CH, THERETHERE());
1078 copy = dupl(p, start+1, finish+1);
1079 assert(copy == finish+4);
1080 repeat(p, copy, 1, to-1);
1082 case REP(1, INF): /* as x+ */
1083 INSERT(OPLUS_, start);
1084 ASTERN(O_PLUS, start);
1086 case REP(N, N): /* as xx{m-1,n-1} */
1087 copy = dupl(p, start, finish);
1088 repeat(p, copy, from-1, to-1);
1090 case REP(N, INF): /* as xx{n-1,INF} */
1091 copy = dupl(p, start, finish);
1092 repeat(p, copy, from-1, to);
1094 default: /* "can't happen" */
1095 SETERROR(REG_ASSERT); /* just in case */
1101 - wgetnext - helper function for WGETNEXT() macro. Gets the next wide
1102 - character from the parse struct, signals a REG_ILLSEQ error if the
1103 - character can't be converted. Returns the number of bytes consumed.
1106 wgetnext(struct parse *p)
1112 memset(&mbs, 0, sizeof(mbs));
1113 n = mbrtowc(&wc, p->next, p->end - p->next, &mbs);
1114 if (n == (size_t)-1 || n == (size_t)-2) {
1115 SETERROR(REG_ILLSEQ);
1125 - seterr - set an error condition
1126 == static int seterr(struct parse *p, int e);
1128 static int /* useless but makes type checking happy */
1129 seterr(struct parse *p, int e)
1131 if (p->error == 0) /* keep earliest error condition */
1133 p->next = nuls; /* try to bring things to a halt */
1135 return(0); /* make the return value well-defined */
1139 - allocset - allocate a set of characters for []
1140 == static cset *allocset(struct parse *p);
1143 allocset(struct parse *p)
1147 ncs = realloc(p->g->sets, (p->g->ncsets + 1) * sizeof(*ncs));
1149 SETERROR(REG_ESPACE);
1153 cs = &p->g->sets[p->g->ncsets++];
1154 memset(cs, 0, sizeof(*cs));
1160 - freeset - free a now-unused set
1161 == static void freeset(struct parse *p, cset *cs);
1164 freeset(struct parse *p, cset *cs)
1166 cset *top = &p->g->sets[p->g->ncsets];
1171 memset(cs, 0, sizeof(*cs));
1172 if (cs == top-1) /* recover only the easy case */
1177 - singleton - Determine whether a set contains only one character,
1178 - returning it if so, otherwise returning OUT.
1185 for (i = n = 0; i < NC; i++)
1192 if (cs->nwides == 1 && cs->nranges == 0 && cs->ntypes == 0 &&
1194 return (cs->wides[0]);
1195 /* Don't bother handling the other cases. */
1200 - CHadd - add character to character set.
1203 CHadd(struct parse *p, cset *cs, wint_t ch)
1205 wint_t nch, *newwides;
1208 cs->bmp[ch >> 3] |= 1 << (ch & 7);
1210 newwides = realloc(cs->wides, (cs->nwides + 1) *
1211 sizeof(*cs->wides));
1212 if (newwides == NULL) {
1213 SETERROR(REG_ESPACE);
1216 cs->wides = newwides;
1217 cs->wides[cs->nwides++] = ch;
1220 if ((nch = towlower(ch)) < NC)
1221 cs->bmp[nch >> 3] |= 1 << (nch & 7);
1222 if ((nch = towupper(ch)) < NC)
1223 cs->bmp[nch >> 3] |= 1 << (nch & 7);
1228 - CHaddrange - add all characters in the range [min,max] to a character set.
1231 CHaddrange(struct parse *p, cset *cs, wint_t min, wint_t max)
1235 for (; min < NC && min <= max; min++)
1239 newranges = realloc(cs->ranges, (cs->nranges + 1) *
1240 sizeof(*cs->ranges));
1241 if (newranges == NULL) {
1242 SETERROR(REG_ESPACE);
1245 cs->ranges = newranges;
1246 cs->ranges[cs->nranges].min = min;
1247 cs->ranges[cs->nranges].max = max;
1252 - CHaddtype - add all characters of a certain type to a character set.
1255 CHaddtype(struct parse *p, cset *cs, wctype_t wct)
1260 for (i = 0; i < NC; i++)
1261 if (iswctype(i, wct))
1263 newtypes = realloc(cs->types, (cs->ntypes + 1) *
1264 sizeof(*cs->types));
1265 if (newtypes == NULL) {
1266 SETERROR(REG_ESPACE);
1269 cs->types = newtypes;
1270 cs->types[cs->ntypes++] = wct;
1274 - dupl - emit a duplicate of a bunch of sops
1275 == static sopno dupl(struct parse *p, sopno start, sopno finish);
1277 static sopno /* start of duplicate */
1278 dupl(struct parse *p,
1279 sopno start, /* from here */
1280 sopno finish) /* to this less one */
1283 sopno len = finish - start;
1285 assert(finish >= start);
1288 if (!enlarge(p, p->ssize + len)) /* this many unexpected additions */
1290 (void) memcpy((char *)(p->strip + p->slen),
1291 (char *)(p->strip + start), (size_t)len*sizeof(sop));
1297 - doemit - emit a strip operator
1298 == static void doemit(struct parse *p, sop op, size_t opnd);
1300 * It might seem better to implement this as a macro with a function as
1301 * hard-case backup, but it's just too big and messy unless there are
1302 * some changes to the data structures. Maybe later.
1305 doemit(struct parse *p, sop op, size_t opnd)
1307 /* avoid making error situations worse */
1311 /* deal with oversize operands ("can't happen", more or less) */
1312 assert(opnd < 1<<OPSHIFT);
1314 /* deal with undersized strip */
1315 if (p->slen >= p->ssize)
1316 if (!enlarge(p, (p->ssize+1) / 2 * 3)) /* +50% */
1319 /* finally, it's all reduced to the easy case */
1320 p->strip[p->slen++] = SOP(op, opnd);
1324 - doinsert - insert a sop into the strip
1325 == static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
1328 doinsert(struct parse *p, sop op, size_t opnd, sopno pos)
1334 /* avoid making error situations worse */
1339 EMIT(op, opnd); /* do checks, ensure space */
1340 assert(HERE() == sn+1);
1343 /* adjust paren pointers */
1345 for (i = 1; i < NPAREN; i++) {
1346 if (p->pbegin[i] >= pos) {
1349 if (p->pend[i] >= pos) {
1354 memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1355 (HERE()-pos-1)*sizeof(sop));
1360 - dofwd - complete a forward reference
1361 == static void dofwd(struct parse *p, sopno pos, sop value);
1364 dofwd(struct parse *p, sopno pos, sop value)
1366 /* avoid making error situations worse */
1370 assert(value < 1<<OPSHIFT);
1371 p->strip[pos] = OP(p->strip[pos]) | value;
1375 - enlarge - enlarge the strip
1376 == static int enlarge(struct parse *p, sopno size);
1379 enlarge(struct parse *p, sopno size)
1383 if (p->ssize >= size)
1386 sp = (sop *)realloc(p->strip, size*sizeof(sop));
1388 SETERROR(REG_ESPACE);
1397 - stripsnug - compact the strip
1398 == static void stripsnug(struct parse *p, struct re_guts *g);
1401 stripsnug(struct parse *p, struct re_guts *g)
1403 g->nstates = p->slen;
1404 g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
1405 if (g->strip == NULL) {
1406 SETERROR(REG_ESPACE);
1407 g->strip = p->strip;
1412 - findmust - fill in must and mlen with longest mandatory literal string
1413 == static void findmust(struct parse *p, struct re_guts *g);
1415 * This algorithm could do fancy things like analyzing the operands of |
1416 * for common subsequences. Someday. This code is simple and finds most
1417 * of the interesting cases.
1419 * Note that must and mlen got initialized during setup.
1422 findmust(struct parse *p, struct re_guts *g)
1431 char buf[MB_LEN_MAX];
1435 /* avoid making error situations worse */
1440 * It's not generally safe to do a ``char'' substring search on
1441 * multibyte character strings, but it's safe for at least
1442 * UTF-8 (see RFC 3629).
1444 if (MB_CUR_MAX > 1 &&
1445 strcmp(_CurrentRuneLocale->__encoding, "UTF-8") != 0)
1448 /* find the longest OCHAR sequence in strip */
1452 scan = g->strip + 1;
1456 case OCHAR: /* sequence member */
1457 if (newlen == 0) { /* new sequence */
1458 memset(&mbs, 0, sizeof(mbs));
1459 newstart = scan - 1;
1461 clen = wcrtomb(buf, OPND(s), &mbs);
1462 if (clen == (size_t)-1)
1466 case OPLUS_: /* things that don't break one */
1470 case OQUEST_: /* things that must be skipped */
1472 offset = altoffset(scan, offset);
1477 /* assert() interferes w debug printouts */
1478 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1483 } while (OP(s) != O_QUEST && OP(s) != O_CH);
1485 case OBOW: /* things that break a sequence */
1492 if (newlen > g->mlen) { /* ends one */
1496 g->moffset += offset;
1499 g->moffset = offset;
1507 if (newlen > g->mlen) { /* ends one */
1511 g->moffset += offset;
1514 g->moffset = offset;
1523 case OANYOF: /* may or may not invalidate offset */
1524 /* First, everything as OANY */
1525 if (newlen > g->mlen) { /* ends one */
1529 g->moffset += offset;
1532 g->moffset = offset;
1543 /* Anything here makes it impossible or too hard
1544 * to calculate the offset -- so we give up;
1545 * save the last known good offset, in case the
1546 * must sequence doesn't occur later.
1548 if (newlen > g->mlen) { /* ends one */
1552 g->moffset += offset;
1554 g->moffset = offset;
1560 } while (OP(s) != OEND);
1562 if (g->mlen == 0) { /* there isn't one */
1567 /* turn it into a character string */
1568 g->must = malloc((size_t)g->mlen + 1);
1569 if (g->must == NULL) { /* argh; just forget it */
1576 memset(&mbs, 0, sizeof(mbs));
1577 while (cp < g->must + g->mlen) {
1578 while (OP(s = *scan++) != OCHAR)
1580 clen = wcrtomb(cp, OPND(s), &mbs);
1581 assert(clen != (size_t)-1);
1584 assert(cp == g->must + g->mlen);
1585 *cp++ = '\0'; /* just on general principles */
1589 - altoffset - choose biggest offset among multiple choices
1590 == static int altoffset(sop *scan, int offset);
1592 * Compute, recursively if necessary, the largest offset among multiple
1596 altoffset(sop *scan, int offset)
1602 /* If we gave up already on offsets, return */
1609 while (OP(s) != O_QUEST && OP(s) != O_CH) {
1618 try = altoffset(scan, try);
1625 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1628 } while (OP(s) != O_QUEST && OP(s) != O_CH);
1629 /* We must skip to the next position, or we'll
1630 * leave altoffset() too early.
1656 return largest+offset;
1660 - computejumps - compute char jumps for BM scan
1661 == static void computejumps(struct parse *p, struct re_guts *g);
1663 * This algorithm assumes g->must exists and is has size greater than
1664 * zero. It's based on the algorithm found on Computer Algorithms by
1667 * A char jump is the number of characters one needs to jump based on
1668 * the value of the character from the text that was mismatched.
1671 computejumps(struct parse *p, struct re_guts *g)
1676 /* Avoid making errors worse */
1680 g->charjump = (int*) malloc((NC + 1) * sizeof(int));
1681 if (g->charjump == NULL) /* Not a fatal error */
1683 /* Adjust for signed chars, if necessary */
1684 g->charjump = &g->charjump[-(CHAR_MIN)];
1686 /* If the character does not exist in the pattern, the jump
1687 * is equal to the number of characters in the pattern.
1689 for (ch = CHAR_MIN; ch < (CHAR_MAX + 1); ch++)
1690 g->charjump[ch] = g->mlen;
1692 /* If the character does exist, compute the jump that would
1693 * take us to the last character in the pattern equal to it
1694 * (notice that we match right to left, so that last character
1695 * is the first one that would be matched).
1697 for (mindex = 0; mindex < g->mlen; mindex++)
1698 g->charjump[(int)g->must[mindex]] = g->mlen - mindex - 1;
1702 - computematchjumps - compute match jumps for BM scan
1703 == static void computematchjumps(struct parse *p, struct re_guts *g);
1705 * This algorithm assumes g->must exists and is has size greater than
1706 * zero. It's based on the algorithm found on Computer Algorithms by
1709 * A match jump is the number of characters one needs to advance based
1710 * on the already-matched suffix.
1711 * Notice that all values here are minus (g->mlen-1), because of the way
1712 * the search algorithm works.
1715 computematchjumps(struct parse *p, struct re_guts *g)
1717 int mindex; /* General "must" iterator */
1718 int suffix; /* Keeps track of matching suffix */
1719 int ssuffix; /* Keeps track of suffixes' suffix */
1720 int* pmatches; /* pmatches[k] points to the next i
1721 * such that i+1...mlen is a substring
1722 * of k+1...k+mlen-i-1
1725 /* Avoid making errors worse */
1729 pmatches = (int*) malloc(g->mlen * sizeof(unsigned int));
1730 if (pmatches == NULL) {
1731 g->matchjump = NULL;
1735 g->matchjump = (int*) malloc(g->mlen * sizeof(unsigned int));
1736 if (g->matchjump == NULL) { /* Not a fatal error */
1741 /* Set maximum possible jump for each character in the pattern */
1742 for (mindex = 0; mindex < g->mlen; mindex++)
1743 g->matchjump[mindex] = 2*g->mlen - mindex - 1;
1745 /* Compute pmatches[] */
1746 for (mindex = g->mlen - 1, suffix = g->mlen; mindex >= 0;
1747 mindex--, suffix--) {
1748 pmatches[mindex] = suffix;
1750 /* If a mismatch is found, interrupting the substring,
1751 * compute the matchjump for that position. If no
1752 * mismatch is found, then a text substring mismatched
1753 * against the suffix will also mismatch against the
1756 while (suffix < g->mlen
1757 && g->must[mindex] != g->must[suffix]) {
1758 g->matchjump[suffix] = MIN(g->matchjump[suffix],
1759 g->mlen - mindex - 1);
1760 suffix = pmatches[suffix];
1764 /* Compute the matchjump up to the last substring found to jump
1765 * to the beginning of the largest must pattern prefix matching
1768 for (mindex = 0; mindex <= suffix; mindex++)
1769 g->matchjump[mindex] = MIN(g->matchjump[mindex],
1770 g->mlen + suffix - mindex);
1772 ssuffix = pmatches[suffix];
1773 while (suffix < g->mlen) {
1774 while (suffix <= ssuffix && suffix < g->mlen) {
1775 g->matchjump[suffix] = MIN(g->matchjump[suffix],
1776 g->mlen + ssuffix - suffix);
1779 if (suffix < g->mlen)
1780 ssuffix = pmatches[ssuffix];
1787 - pluscount - count + nesting
1788 == static sopno pluscount(struct parse *p, struct re_guts *g);
1790 static sopno /* nesting depth */
1791 pluscount(struct parse *p, struct re_guts *g)
1799 return(0); /* there may not be an OEND */
1801 scan = g->strip + 1;
1809 if (plusnest > maxnest)
1814 } while (OP(s) != OEND);