2
* This code is derived from OpenBSD's libc/regex, original license follows:
4
* Copyright (c) 1992, 1993, 1994 Henry Spencer.
5
* Copyright (c) 1992, 1993, 1994
6
* The Regents of the University of California. All rights reserved.
8
* This code is derived from software contributed to Berkeley by
11
* Redistribution and use in source and binary forms, with or without
12
* modification, are permitted provided that the following conditions
14
* 1. Redistributions of source code must retain the above copyright
15
* notice, this list of conditions and the following disclaimer.
16
* 2. Redistributions in binary form must reproduce the above copyright
17
* notice, this list of conditions and the following disclaimer in the
18
* documentation and/or other materials provided with the distribution.
19
* 3. Neither the name of the University nor the names of its contributors
20
* may be used to endorse or promote products derived from this software
21
* without specific prior written permission.
23
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35
* @(#)regcomp.c 8.5 (Berkeley) 3/20/94
38
#include <sys/types.h>
44
#include "regex_impl.h"
49
#include "regcclass.h"
53
* parse structure, passed up and down to avoid global variables and
57
char *next; /* next character in RE */
58
char *end; /* end of string (-> NUL normally) */
59
int error; /* has an error been seen? */
60
sop *strip; /* malloced strip */
61
sopno ssize; /* malloced strip size (allocated) */
62
sopno slen; /* malloced strip length (used) */
63
int ncsalloc; /* number of csets allocated */
65
# define NPAREN 10 /* we need to remember () 1-9 for back refs */
66
sopno pbegin[NPAREN]; /* -> ( ([0] unused) */
67
sopno pend[NPAREN]; /* -> ) ([0] unused) */
70
static void p_ere(struct parse *, int);
71
static void p_ere_exp(struct parse *);
72
static void p_str(struct parse *);
73
static void p_bre(struct parse *, int, int);
74
static int p_simp_re(struct parse *, int);
75
static int p_count(struct parse *);
76
static void p_bracket(struct parse *);
77
static void p_b_term(struct parse *, cset *);
78
static void p_b_cclass(struct parse *, cset *);
79
static void p_b_eclass(struct parse *, cset *);
80
static char p_b_symbol(struct parse *);
81
static char p_b_coll_elem(struct parse *, int);
82
static char othercase(int);
83
static void bothcases(struct parse *, int);
84
static void ordinary(struct parse *, int);
85
static void nonnewline(struct parse *);
86
static void repeat(struct parse *, sopno, int, int);
87
static int seterr(struct parse *, int);
88
static cset *allocset(struct parse *);
89
static void freeset(struct parse *, cset *);
90
static int freezeset(struct parse *, cset *);
91
static int firstch(struct parse *, cset *);
92
static int nch(struct parse *, cset *);
93
static void mcadd(struct parse *, cset *, const char *);
94
static void mcinvert(struct parse *, cset *);
95
static void mccase(struct parse *, cset *);
96
static int isinsets(struct re_guts *, int);
97
static int samesets(struct re_guts *, int, int);
98
static void categorize(struct parse *, struct re_guts *);
99
static sopno dupl(struct parse *, sopno, sopno);
100
static void doemit(struct parse *, sop, size_t);
101
static void doinsert(struct parse *, sop, size_t, sopno);
102
static void dofwd(struct parse *, sopno, sop);
103
static void enlarge(struct parse *, sopno);
104
static void stripsnug(struct parse *, struct re_guts *);
105
static void findmust(struct parse *, struct re_guts *);
106
static sopno pluscount(struct parse *, struct re_guts *);
108
static char nuls[10]; /* place to point scanner in event of error */
111
* macros for use with parse structure
112
* BEWARE: these know that the parse structure is named `p' !!!
114
#define PEEK() (*p->next)
115
#define PEEK2() (*(p->next+1))
116
#define MORE() (p->next < p->end)
117
#define MORE2() (p->next+1 < p->end)
118
#define SEE(c) (MORE() && PEEK() == (c))
119
#define SEETWO(a, b) (MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
120
#define EAT(c) ((SEE(c)) ? (NEXT(), 1) : 0)
121
#define EATTWO(a, b) ((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
122
#define NEXT() (p->next++)
123
#define NEXT2() (p->next += 2)
124
#define NEXTn(n) (p->next += (n))
125
#define GETNEXT() (*p->next++)
126
#define SETERROR(e) seterr(p, (e))
127
#define REQUIRE(co, e) (void)((co) || SETERROR(e))
128
#define MUSTSEE(c, e) (REQUIRE(MORE() && PEEK() == (c), e))
129
#define MUSTEAT(c, e) (REQUIRE(MORE() && GETNEXT() == (c), e))
130
#define MUSTNOTSEE(c, e) (REQUIRE(!MORE() || PEEK() != (c), e))
131
#define EMIT(op, sopnd) doemit(p, (sop)(op), (size_t)(sopnd))
132
#define INSERT(op, pos) doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
133
#define AHEAD(pos) dofwd(p, pos, HERE()-(pos))
134
#define ASTERN(sop, pos) EMIT(sop, HERE()-pos)
135
#define HERE() (p->slen)
136
#define THERE() (p->slen - 1)
137
#define THERETHERE() (p->slen - 2)
138
#define DROP(n) (p->slen -= (n))
140
#ifdef _POSIX2_RE_DUP_MAX
141
#define DUPMAX _POSIX2_RE_DUP_MAX
145
#define INFINITY (DUPMAX + 1)
148
static int never = 0; /* for use in asserts; shuts lint up */
150
#define never 0 /* some <assert.h>s have bugs too */
154
- llvm_regcomp - interface for parser and compilation
156
int /* 0 success, otherwise REG_something */
157
llvm_regcomp(llvm_regex_t *preg, const char *pattern, int cflags)
161
struct parse *p = &pa;
165
# define GOODFLAGS(f) (f)
167
# define GOODFLAGS(f) ((f)&~REG_DUMP)
170
cflags = GOODFLAGS(cflags);
171
if ((cflags®_EXTENDED) && (cflags®_NOSPEC))
174
if (cflags®_PEND) {
175
if (preg->re_endp < pattern)
177
len = preg->re_endp - pattern;
179
len = strlen((const char *)pattern);
181
/* do the mallocs early so failure handling is easy */
182
g = (struct re_guts *)malloc(sizeof(struct re_guts) +
183
(NC-1)*sizeof(cat_t));
186
p->ssize = len/(size_t)2*(size_t)3 + (size_t)1; /* ugh */
187
p->strip = (sop *)calloc(p->ssize, sizeof(sop));
189
if (p->strip == NULL) {
196
p->next = (char *)pattern; /* convenience; we do not modify it */
197
p->end = p->next + len;
200
for (i = 0; i < NPAREN; i++) {
215
g->ncategories = 1; /* category 0 is "everything else" */
216
g->categories = &g->catspace[-(CHAR_MIN)];
217
(void) memset((char *)g->catspace, 0, NC*sizeof(cat_t));
222
g->firststate = THERE();
223
if (cflags®_EXTENDED)
225
else if (cflags®_NOSPEC)
230
g->laststate = THERE();
232
/* tidy up loose ends and fill things in */
236
g->nplus = pluscount(p, g);
238
preg->re_nsub = g->nsub;
240
preg->re_magic = MAGIC1;
242
/* not debugging, so can't rely on the assert() in llvm_regexec() */
243
if (g->iflags®EX_BAD)
244
SETERROR(REG_ASSERT);
247
/* win or lose, we're done */
248
if (p->error != 0) /* lose */
254
- p_ere - ERE parser top level, concatenation and alternation
257
p_ere(struct parse *p, int stop) /* character this ERE should end at */
263
int first = 1; /* is this the first alternative? */
266
/* do a bunch of concatenated expressions */
268
while (MORE() && (c = PEEK()) != '|' && c != stop)
270
REQUIRE(HERE() != conc, REG_EMPTY); /* require nonempty */
273
break; /* NOTE BREAK OUT */
276
INSERT(OCH_, conc); /* offset is wrong */
281
ASTERN(OOR1, prevback);
283
AHEAD(prevfwd); /* fix previous offset */
285
EMIT(OOR2, 0); /* offset is very wrong */
288
if (!first) { /* tail-end fixups */
290
ASTERN(O_CH, prevback);
293
assert(!MORE() || SEE(stop));
297
- p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
300
p_ere_exp(struct parse *p)
309
assert(MORE()); /* caller should have ensured this */
315
REQUIRE(MORE(), REG_EPAREN);
319
p->pbegin[subno] = HERE();
320
EMIT(OLPAREN, subno);
323
if (subno < NPAREN) {
324
p->pend[subno] = HERE();
325
assert(p->pend[subno] != 0);
327
EMIT(ORPAREN, subno);
328
MUSTEAT(')', REG_EPAREN);
330
#ifndef POSIX_MISTAKE
331
case ')': /* happens only if no current unmatched ( */
333
* You may ask, why the ifndef? Because I didn't notice
334
* this until slightly too late for 1003.2, and none of the
335
* other 1003.2 regular-expression reviewers noticed it at
336
* all. So an unmatched ) is legal POSIX, at least until
337
* we can get it fixed.
339
SETERROR(REG_EPAREN);
344
p->g->iflags |= USEBOL;
350
p->g->iflags |= USEEOL;
359
SETERROR(REG_BADRPT);
362
if (p->g->cflags®_NEWLINE)
371
REQUIRE(MORE(), REG_EESCAPE);
375
case '{': /* okay as ordinary except if digit follows */
376
REQUIRE(!MORE() || !isdigit((uch)PEEK()), REG_BADRPT);
386
/* we call { a repetition if followed by a digit */
387
if (!( c == '*' || c == '+' || c == '?' ||
388
(c == '{' && MORE2() && isdigit((uch)PEEK2())) ))
389
return; /* no repetition, we're done */
392
REQUIRE(!wascaret, REG_BADRPT);
394
case '*': /* implemented as +? */
395
/* this case does not require the (y|) trick, noKLUDGE */
398
INSERT(OQUEST_, pos);
399
ASTERN(O_QUEST, pos);
406
/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
407
INSERT(OCH_, pos); /* offset slightly wrong */
408
ASTERN(OOR1, pos); /* this one's right */
409
AHEAD(pos); /* fix the OCH_ */
410
EMIT(OOR2, 0); /* offset very wrong... */
411
AHEAD(THERE()); /* ...so fix it */
412
ASTERN(O_CH, THERETHERE());
417
if (isdigit((uch)PEEK())) {
419
REQUIRE(count <= count2, REG_BADBR);
420
} else /* single number with comma */
422
} else /* just a single number */
424
repeat(p, pos, count, count2);
425
if (!EAT('}')) { /* error heuristics */
426
while (MORE() && PEEK() != '}')
428
REQUIRE(MORE(), REG_EBRACE);
437
if (!( c == '*' || c == '+' || c == '?' ||
438
(c == '{' && MORE2() && isdigit((uch)PEEK2())) ) )
440
SETERROR(REG_BADRPT);
444
- p_str - string (no metacharacters) "parser"
447
p_str(struct parse *p)
449
REQUIRE(MORE(), REG_EMPTY);
451
ordinary(p, GETNEXT());
455
- p_bre - BRE parser top level, anchoring and concatenation
456
* Giving end1 as OUT essentially eliminates the end1/end2 check.
458
* This implementation is a bit of a kludge, in that a trailing $ is first
459
* taken as an ordinary character and then revised to be an anchor. The
460
* only undesirable side effect is that '$' gets included as a character
461
* category in such cases. This is fairly harmless; not worth fixing.
462
* The amount of lookahead needed to avoid this kludge is excessive.
465
p_bre(struct parse *p,
466
int end1, /* first terminating character */
467
int end2) /* second terminating character */
469
sopno start = HERE();
470
int first = 1; /* first subexpression? */
475
p->g->iflags |= USEBOL;
478
while (MORE() && !SEETWO(end1, end2)) {
479
wasdollar = p_simp_re(p, first);
482
if (wasdollar) { /* oops, that was a trailing anchor */
485
p->g->iflags |= USEEOL;
489
REQUIRE(HERE() != start, REG_EMPTY); /* require nonempty */
493
- p_simp_re - parse a simple RE, an atom possibly followed by a repetition
495
static int /* was the simple RE an unbackslashed $? */
496
p_simp_re(struct parse *p,
497
int starordinary) /* is a leading * an ordinary character? */
505
# define BACKSL (1<<CHAR_BIT)
507
pos = HERE(); /* repetion op, if any, covers from here */
509
assert(MORE()); /* caller should have ensured this */
512
REQUIRE(MORE(), REG_EESCAPE);
513
c = BACKSL | GETNEXT();
517
if (p->g->cflags®_NEWLINE)
526
SETERROR(REG_BADRPT);
532
p->pbegin[subno] = HERE();
533
EMIT(OLPAREN, subno);
534
/* the MORE here is an error heuristic */
535
if (MORE() && !SEETWO('\\', ')'))
537
if (subno < NPAREN) {
538
p->pend[subno] = HERE();
539
assert(p->pend[subno] != 0);
541
EMIT(ORPAREN, subno);
542
REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
544
case BACKSL|')': /* should not get here -- must be user */
546
SETERROR(REG_EPAREN);
557
i = (c&~BACKSL) - '0';
559
if (p->pend[i] != 0) {
560
assert(i <= p->g->nsub);
562
assert(p->pbegin[i] != 0);
563
assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
564
assert(OP(p->strip[p->pend[i]]) == ORPAREN);
565
(void) dupl(p, p->pbegin[i]+1, p->pend[i]);
568
SETERROR(REG_ESUBREG);
572
REQUIRE(starordinary, REG_BADRPT);
575
ordinary(p, (char)c);
579
if (EAT('*')) { /* implemented as +? */
580
/* this case does not require the (y|) trick, noKLUDGE */
583
INSERT(OQUEST_, pos);
584
ASTERN(O_QUEST, pos);
585
} else if (EATTWO('\\', '{')) {
588
if (MORE() && isdigit((uch)PEEK())) {
590
REQUIRE(count <= count2, REG_BADBR);
591
} else /* single number with comma */
593
} else /* just a single number */
595
repeat(p, pos, count, count2);
596
if (!EATTWO('\\', '}')) { /* error heuristics */
597
while (MORE() && !SEETWO('\\', '}'))
599
REQUIRE(MORE(), REG_EBRACE);
602
} else if (c == '$') /* $ (but not \$) ends it */
609
- p_count - parse a repetition count
611
static int /* the value */
612
p_count(struct parse *p)
617
while (MORE() && isdigit((uch)PEEK()) && count <= DUPMAX) {
618
count = count*10 + (GETNEXT() - '0');
622
REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
627
- p_bracket - parse a bracketed character list
629
* Note a significant property of this code: if the allocset() did SETERROR,
630
* no set operations are done.
633
p_bracket(struct parse *p)
638
/* Dept of Truly Sickening Special-Case Kludges */
639
if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
644
if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
650
if ((cs = allocset(p)) == NULL) {
651
/* allocset did set error status in p */
656
invert++; /* make note to invert set at end */
661
while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
665
MUSTEAT(']', REG_EBRACK);
667
if (p->error != 0) { /* don't mess things up further */
672
if (p->g->cflags®_ICASE) {
676
for (i = p->g->csetsize - 1; i >= 0; i--)
677
if (CHIN(cs, i) && isalpha(i)) {
682
if (cs->multis != NULL)
688
for (i = p->g->csetsize - 1; i >= 0; i--)
693
if (p->g->cflags®_NEWLINE)
695
if (cs->multis != NULL)
699
assert(cs->multis == NULL); /* xxx */
701
if (nch(p, cs) == 1) { /* optimize singleton sets */
702
ordinary(p, firstch(p, cs));
705
EMIT(OANYOF, freezeset(p, cs));
709
- p_b_term - parse one term of a bracketed character list
712
p_b_term(struct parse *p, cset *cs)
718
/* classify what we've got */
719
switch ((MORE()) ? PEEK() : '\0') {
721
c = (MORE2()) ? PEEK2() : '\0';
724
SETERROR(REG_ERANGE);
725
return; /* NOTE RETURN */
733
case ':': /* character class */
735
REQUIRE(MORE(), REG_EBRACK);
737
REQUIRE(c != '-' && c != ']', REG_ECTYPE);
739
REQUIRE(MORE(), REG_EBRACK);
740
REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
742
case '=': /* equivalence class */
744
REQUIRE(MORE(), REG_EBRACK);
746
REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
748
REQUIRE(MORE(), REG_EBRACK);
749
REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
751
default: /* symbol, ordinary character, or range */
752
/* xxx revision needed for multichar stuff */
753
start = p_b_symbol(p);
754
if (SEE('-') && MORE2() && PEEK2() != ']') {
760
finish = p_b_symbol(p);
763
/* xxx what about signed chars here... */
764
REQUIRE(start <= finish, REG_ERANGE);
765
for (i = start; i <= finish; i++)
772
- p_b_cclass - parse a character-class name and deal with it
775
p_b_cclass(struct parse *p, cset *cs)
783
while (MORE() && isalpha(PEEK()))
786
for (cp = cclasses; cp->name != NULL; cp++)
787
if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
789
if (cp->name == NULL) {
790
/* oops, didn't find it */
791
SETERROR(REG_ECTYPE);
796
while ((c = *u++) != '\0')
798
for (u = cp->multis; *u != '\0'; u += strlen(u) + 1)
803
- p_b_eclass - parse an equivalence-class name and deal with it
805
* This implementation is incomplete. xxx
808
p_b_eclass(struct parse *p, cset *cs)
812
c = p_b_coll_elem(p, '=');
817
- p_b_symbol - parse a character or [..]ed multicharacter collating symbol
819
static char /* value of symbol */
820
p_b_symbol(struct parse *p)
824
REQUIRE(MORE(), REG_EBRACK);
825
if (!EATTWO('[', '.'))
828
/* collating symbol */
829
value = p_b_coll_elem(p, '.');
830
REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
835
- p_b_coll_elem - parse a collating-element name and look it up
837
static char /* value of collating element */
838
p_b_coll_elem(struct parse *p,
839
int endc) /* name ended by endc,']' */
845
while (MORE() && !SEETWO(endc, ']'))
848
SETERROR(REG_EBRACK);
852
for (cp = cnames; cp->name != NULL; cp++)
853
if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
854
return(cp->code); /* known name */
856
return(*sp); /* single character */
857
SETERROR(REG_ECOLLATE); /* neither */
862
- othercase - return the case counterpart of an alphabetic
864
static char /* if no counterpart, return ch */
870
return ((uch)tolower(ch));
871
else if (islower(ch))
872
return ((uch)toupper(ch));
873
else /* peculiar, but could happen */
878
- bothcases - emit a dualcase version of a two-case character
880
* Boy, is this implementation ever a kludge...
883
bothcases(struct parse *p, int ch)
885
char *oldnext = p->next;
886
char *oldend = p->end;
890
assert(othercase(ch) != ch); /* p_bracket() would recurse */
897
assert(p->next == bracket+2);
903
- ordinary - emit an ordinary character
906
ordinary(struct parse *p, int ch)
908
cat_t *cap = p->g->categories;
910
if ((p->g->cflags®_ICASE) && isalpha((uch)ch) && othercase(ch) != ch)
913
EMIT(OCHAR, (uch)ch);
915
cap[ch] = p->g->ncategories++;
920
- nonnewline - emit REG_NEWLINE version of OANY
922
* Boy, is this implementation ever a kludge...
925
nonnewline(struct parse *p)
927
char *oldnext = p->next;
928
char *oldend = p->end;
938
assert(p->next == bracket+3);
944
- repeat - generate code for a bounded repetition, recursively if needed
947
repeat(struct parse *p,
948
sopno start, /* operand from here to end of strip */
949
int from, /* repeated from this number */
950
int to) /* to this number of times (maybe INFINITY) */
952
sopno finish = HERE();
955
# define REP(f, t) ((f)*8 + (t))
956
# define MAP(n) (((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
959
if (p->error != 0) /* head off possible runaway recursion */
964
switch (REP(MAP(from), MAP(to))) {
965
case REP(0, 0): /* must be user doing this */
966
DROP(finish-start); /* drop the operand */
968
case REP(0, 1): /* as x{1,1}? */
969
case REP(0, N): /* as x{1,n}? */
970
case REP(0, INF): /* as x{1,}? */
971
/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
972
INSERT(OCH_, start); /* offset is wrong... */
973
repeat(p, start+1, 1, to);
975
AHEAD(start); /* ... fix it */
978
ASTERN(O_CH, THERETHERE());
980
case REP(1, 1): /* trivial case */
983
case REP(1, N): /* as x?x{1,n-1} */
984
/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
988
EMIT(OOR2, 0); /* offset very wrong... */
989
AHEAD(THERE()); /* ...so fix it */
990
ASTERN(O_CH, THERETHERE());
991
copy = dupl(p, start+1, finish+1);
992
assert(copy == finish+4);
993
repeat(p, copy, 1, to-1);
995
case REP(1, INF): /* as x+ */
996
INSERT(OPLUS_, start);
997
ASTERN(O_PLUS, start);
999
case REP(N, N): /* as xx{m-1,n-1} */
1000
copy = dupl(p, start, finish);
1001
repeat(p, copy, from-1, to-1);
1003
case REP(N, INF): /* as xx{n-1,INF} */
1004
copy = dupl(p, start, finish);
1005
repeat(p, copy, from-1, to);
1007
default: /* "can't happen" */
1008
SETERROR(REG_ASSERT); /* just in case */
1014
- seterr - set an error condition
1016
static int /* useless but makes type checking happy */
1017
seterr(struct parse *p, int e)
1019
if (p->error == 0) /* keep earliest error condition */
1021
p->next = nuls; /* try to bring things to a halt */
1023
return(0); /* make the return value well-defined */
1027
- allocset - allocate a set of characters for []
1030
allocset(struct parse *p)
1032
int no = p->g->ncsets++;
1036
size_t css = (size_t)p->g->csetsize;
1039
if (no >= p->ncsalloc) { /* need another column of space */
1042
p->ncsalloc += CHAR_BIT;
1044
assert(nc % CHAR_BIT == 0);
1045
nbytes = nc / CHAR_BIT * css;
1047
ptr = (cset *)realloc((char *)p->g->sets, nc * sizeof(cset));
1052
ptr = (uch *)realloc((char *)p->g->setbits, nbytes);
1055
p->g->setbits = ptr;
1057
for (i = 0; i < no; i++)
1058
p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
1060
(void) memset((char *)p->g->setbits + (nbytes - css), 0, css);
1062
/* XXX should not happen */
1063
if (p->g->sets == NULL || p->g->setbits == NULL)
1066
cs = &p->g->sets[no];
1067
cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
1068
cs->mask = 1 << ((no) % CHAR_BIT);
1077
free(p->g->setbits);
1078
p->g->setbits = NULL;
1080
SETERROR(REG_ESPACE);
1081
/* caller's responsibility not to do set ops */
1086
- freeset - free a now-unused set
1089
freeset(struct parse *p, cset *cs)
1092
cset *top = &p->g->sets[p->g->ncsets];
1093
size_t css = (size_t)p->g->csetsize;
1095
for (i = 0; i < css; i++)
1097
if (cs == top-1) /* recover only the easy case */
1102
- freezeset - final processing on a set of characters
1104
* The main task here is merging identical sets. This is usually a waste
1105
* of time (although the hash code minimizes the overhead), but can win
1106
* big if REG_ICASE is being used. REG_ICASE, by the way, is why the hash
1107
* is done using addition rather than xor -- all ASCII [aA] sets xor to
1110
static int /* set number */
1111
freezeset(struct parse *p, cset *cs)
1115
cset *top = &p->g->sets[p->g->ncsets];
1117
size_t css = (size_t)p->g->csetsize;
1119
/* look for an earlier one which is the same */
1120
for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
1121
if (cs2->hash == h && cs2 != cs) {
1123
for (i = 0; i < css; i++)
1124
if (!!CHIN(cs2, i) != !!CHIN(cs, i))
1130
if (cs2 < top) { /* found one */
1135
return((int)(cs - p->g->sets));
1139
- firstch - return first character in a set (which must have at least one)
1141
static int /* character; there is no "none" value */
1142
firstch(struct parse *p, cset *cs)
1145
size_t css = (size_t)p->g->csetsize;
1147
for (i = 0; i < css; i++)
1151
return(0); /* arbitrary */
1155
- nch - number of characters in a set
1158
nch(struct parse *p, cset *cs)
1161
size_t css = (size_t)p->g->csetsize;
1164
for (i = 0; i < css; i++)
1171
- mcadd - add a collating element to a cset
1174
mcadd( struct parse *p, cset *cs, const char *cp)
1176
size_t oldend = cs->smultis;
1179
cs->smultis += strlen(cp) + 1;
1180
np = realloc(cs->multis, cs->smultis);
1185
SETERROR(REG_ESPACE);
1190
llvm_strlcpy(cs->multis + oldend - 1, cp, cs->smultis - oldend + 1);
1194
- mcinvert - invert the list of collating elements in a cset
1196
* This would have to know the set of possibilities. Implementation
1201
mcinvert(struct parse *p, cset *cs)
1203
assert(cs->multis == NULL); /* xxx */
1207
- mccase - add case counterparts of the list of collating elements in a cset
1209
* This would have to know the set of possibilities. Implementation
1214
mccase(struct parse *p, cset *cs)
1216
assert(cs->multis == NULL); /* xxx */
1220
- isinsets - is this character in any sets?
1222
static int /* predicate */
1223
isinsets(struct re_guts *g, int c)
1227
int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1228
unsigned uc = (uch)c;
1230
for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1237
- samesets - are these two characters in exactly the same sets?
1239
static int /* predicate */
1240
samesets(struct re_guts *g, int c1, int c2)
1244
int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1245
unsigned uc1 = (uch)c1;
1246
unsigned uc2 = (uch)c2;
1248
for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1249
if (col[uc1] != col[uc2])
1255
- categorize - sort out character categories
1258
categorize(struct parse *p, struct re_guts *g)
1260
cat_t *cats = g->categories;
1265
/* avoid making error situations worse */
1269
for (c = CHAR_MIN; c <= CHAR_MAX; c++)
1270
if (cats[c] == 0 && isinsets(g, c)) {
1271
cat = g->ncategories++;
1273
for (c2 = c+1; c2 <= CHAR_MAX; c2++)
1274
if (cats[c2] == 0 && samesets(g, c, c2))
1280
- dupl - emit a duplicate of a bunch of sops
1282
static sopno /* start of duplicate */
1283
dupl(struct parse *p,
1284
sopno start, /* from here */
1285
sopno finish) /* to this less one */
1288
sopno len = finish - start;
1290
assert(finish >= start);
1293
enlarge(p, p->ssize + len); /* this many unexpected additions */
1294
assert(p->ssize >= p->slen + len);
1295
(void) memmove((char *)(p->strip + p->slen),
1296
(char *)(p->strip + start), (size_t)len*sizeof(sop));
1302
- doemit - emit a strip operator
1304
* It might seem better to implement this as a macro with a function as
1305
* hard-case backup, but it's just too big and messy unless there are
1306
* some changes to the data structures. Maybe later.
1309
doemit(struct parse *p, sop op, size_t opnd)
1311
/* avoid making error situations worse */
1315
/* deal with oversize operands ("can't happen", more or less) */
1316
assert(opnd < 1<<OPSHIFT);
1318
/* deal with undersized strip */
1319
if (p->slen >= p->ssize)
1320
enlarge(p, (p->ssize+1) / 2 * 3); /* +50% */
1321
assert(p->slen < p->ssize);
1323
/* finally, it's all reduced to the easy case */
1324
p->strip[p->slen++] = SOP(op, opnd);
1328
- doinsert - insert a sop into the strip
1331
doinsert(struct parse *p, sop op, size_t opnd, sopno pos)
1337
/* avoid making error situations worse */
1342
EMIT(op, opnd); /* do checks, ensure space */
1343
assert(HERE() == sn+1);
1346
/* adjust paren pointers */
1348
for (i = 1; i < NPAREN; i++) {
1349
if (p->pbegin[i] >= pos) {
1352
if (p->pend[i] >= pos) {
1357
memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1358
(HERE()-pos-1)*sizeof(sop));
1363
- dofwd - complete a forward reference
1366
dofwd(struct parse *p, sopno pos, sop value)
1368
/* avoid making error situations worse */
1372
assert(value < 1<<OPSHIFT);
1373
p->strip[pos] = OP(p->strip[pos]) | value;
1377
- enlarge - enlarge the strip
1380
enlarge(struct parse *p, sopno size)
1384
if (p->ssize >= size)
1387
sp = (sop *)realloc(p->strip, size*sizeof(sop));
1389
SETERROR(REG_ESPACE);
1397
- stripsnug - compact the strip
1400
stripsnug(struct parse *p, struct re_guts *g)
1402
g->nstates = p->slen;
1403
g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
1404
if (g->strip == NULL) {
1405
SETERROR(REG_ESPACE);
1406
g->strip = p->strip;
1411
- findmust - fill in must and mlen with longest mandatory literal string
1413
* This algorithm could do fancy things like analyzing the operands of |
1414
* for common subsequences. Someday. This code is simple and finds most
1415
* of the interesting cases.
1417
* Note that must and mlen got initialized during setup.
1420
findmust(struct parse *p, struct re_guts *g)
1423
sop *start = 0; /* start initialized in the default case, after that */
1424
sop *newstart = 0; /* newstart was initialized in the OCHAR case */
1430
/* avoid making error situations worse */
1434
/* find the longest OCHAR sequence in strip */
1436
scan = g->strip + 1;
1440
case OCHAR: /* sequence member */
1441
if (newlen == 0) /* new sequence */
1442
newstart = scan - 1;
1445
case OPLUS_: /* things that don't break one */
1449
case OQUEST_: /* things that must be skipped */
1455
/* assert() interferes w debug printouts */
1456
if (OP(s) != O_QUEST && OP(s) != O_CH &&
1458
g->iflags |= REGEX_BAD;
1461
} while (OP(s) != O_QUEST && OP(s) != O_CH);
1463
default: /* things that break a sequence */
1464
if (newlen > g->mlen) { /* ends one */
1471
} while (OP(s) != OEND);
1473
if (g->mlen == 0) /* there isn't one */
1476
/* turn it into a character string */
1477
g->must = malloc((size_t)g->mlen + 1);
1478
if (g->must == NULL) { /* argh; just forget it */
1484
for (i = g->mlen; i > 0; i--) {
1485
while (OP(s = *scan++) != OCHAR)
1487
assert(cp < g->must + g->mlen);
1488
*cp++ = (char)OPND(s);
1490
assert(cp == g->must + g->mlen);
1491
*cp++ = '\0'; /* just on general principles */
1495
- pluscount - count + nesting
1497
static sopno /* nesting depth */
1498
pluscount(struct parse *p, struct re_guts *g)
1506
return(0); /* there may not be an OEND */
1508
scan = g->strip + 1;
1516
if (plusnest > maxnest)
1521
} while (OP(s) != OEND);
1523
g->iflags |= REGEX_BAD;