This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
PATCH: [perl #134325] Heap buffer overflow
[perl5.git] / regcomp.c
CommitLineData
a0d0e21e
LW
1/* regcomp.c
2 */
3
4/*
4ac71550
TC
5 * 'A fair jaw-cracker dwarf-language must be.' --Samwise Gamgee
6 *
7 * [p.285 of _The Lord of the Rings_, II/iii: "The Ring Goes South"]
a0d0e21e
LW
8 */
9
61296642
DM
10/* This file contains functions for compiling a regular expression. See
11 * also regexec.c which funnily enough, contains functions for executing
166f8a29 12 * a regular expression.
e4a054ea
DM
13 *
14 * This file is also copied at build time to ext/re/re_comp.c, where
15 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
16 * This causes the main functions to be compiled under new names and with
17 * debugging support added, which makes "use re 'debug'" work.
166f8a29
DM
18 */
19
a687059c
LW
20/* NOTE: this is derived from Henry Spencer's regexp code, and should not
21 * confused with the original package (see point 3 below). Thanks, Henry!
22 */
23
24/* Additional note: this code is very heavily munged from Henry's version
25 * in places. In some spots I've traded clarity for efficiency, so don't
26 * blame Henry for some of the lack of readability.
27 */
28
e50aee73 29/* The names of the functions have been changed from regcomp and
3b753521 30 * regexec to pregcomp and pregexec in order to avoid conflicts
e50aee73
AD
31 * with the POSIX routines of the same names.
32*/
33
b9d5759e 34#ifdef PERL_EXT_RE_BUILD
54df2634 35#include "re_top.h"
b81d288d 36#endif
56953603 37
a687059c 38/*
e50aee73 39 * pregcomp and pregexec -- regsub and regerror are not used in perl
a687059c
LW
40 *
41 * Copyright (c) 1986 by University of Toronto.
42 * Written by Henry Spencer. Not derived from licensed software.
43 *
44 * Permission is granted to anyone to use this software for any
45 * purpose on any computer system, and to redistribute it freely,
46 * subject to the following restrictions:
47 *
48 * 1. The author is not responsible for the consequences of use of
49 * this software, no matter how awful, even if they arise
50 * from defects in it.
51 *
52 * 2. The origin of this software must not be misrepresented, either
53 * by explicit claim or by omission.
54 *
55 * 3. Altered versions must be plainly marked as such, and must not
56 * be misrepresented as being the original software.
57 *
58 *
59 **** Alterations to Henry's code are...
60 ****
4bb101f2 61 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
1129b882
NC
62 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
63 **** by Larry Wall and others
a687059c 64 ****
9ef589d8
LW
65 **** You may distribute under the terms of either the GNU General Public
66 **** License or the Artistic License, as specified in the README file.
67
a687059c
LW
68 *
69 * Beware that some of this code is subtly aware of the way operator
70 * precedence is structured in regular expressions. Serious changes in
71 * regular-expression syntax might require a total rethink.
72 */
73#include "EXTERN.h"
864dbfa3 74#define PERL_IN_REGCOMP_C
a687059c 75#include "perl.h"
d06ea78c 76
c277df42 77#define REG_COMP_C
54df2634
NC
78#ifdef PERL_IN_XSUB_RE
79# include "re_comp.h"
afda64e8 80EXTERN_C const struct regexp_engine my_reg_engine;
54df2634
NC
81#else
82# include "regcomp.h"
83#endif
a687059c 84
f7e03a10 85#include "dquote_inline.h"
b992490d 86#include "invlist_inline.h"
1b0f46bf 87#include "unicode_constants.h"
04e98a4d 88
538e84ed
KW
89#define HAS_NONLATIN1_FOLD_CLOSURE(i) \
90 _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)
f12c0118
KW
91#define HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(i) \
92 _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)
26faadbd 93#define IS_NON_FINAL_FOLD(c) _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
2c61f163 94#define IS_IN_SOME_FOLD_L1(c) _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
94dc5c2d 95
a687059c
LW
96#ifndef STATIC
97#define STATIC static
98#endif
99
3f910e62
YO
100/* this is a chain of data about sub patterns we are processing that
101 need to be handled separately/specially in study_chunk. Its so
102 we can simulate recursion without losing state. */
103struct scan_frame;
104typedef struct scan_frame {
105 regnode *last_regnode; /* last node to process in this frame */
106 regnode *next_regnode; /* next node to process when last is reached */
107 U32 prev_recursed_depth;
108 I32 stopparen; /* what stopparen do we use */
3f910e62
YO
109
110 struct scan_frame *this_prev_frame; /* this previous frame */
111 struct scan_frame *prev_frame; /* previous frame */
112 struct scan_frame *next_frame; /* next frame */
113} scan_frame;
114
f4ae5a27
KW
115/* Certain characters are output as a sequence with the first being a
116 * backslash. */
eb7486de 117#define isBACKSLASHED_PUNCT(c) strchr("-[]\\^", c)
f4ae5a27
KW
118
119
09b2b2e6 120struct RExC_state_t {
514a91f1
DM
121 U32 flags; /* RXf_* are we folding, multilining? */
122 U32 pm_flags; /* PMf_* stuff from the calling PMOP */
830247a4 123 char *precomp; /* uncompiled string. */
711b303b 124 char *precomp_end; /* pointer to end of uncompiled string. */
288b8c02 125 REGEXP *rx_sv; /* The SV that is the regexp. */
f8fc2ecf 126 regexp *rx; /* perl core regexp structure */
538e84ed
KW
127 regexp_internal *rxi; /* internal data for regexp object
128 pprivate field */
fac92740 129 char *start; /* Start of input for compile */
830247a4
IZ
130 char *end; /* End of input for compile */
131 char *parse; /* Input-scan pointer. */
51684e3c
KW
132 char *copy_start; /* start of copy of input within
133 constructed parse string */
cc16d262
KW
134 char *save_copy_start; /* Provides one level of saving
135 and restoring 'copy_start' */
51684e3c
KW
136 char *copy_start_in_input; /* Position in input string
137 corresponding to copy_start */
ea3daa5d 138 SSize_t whilem_seen; /* number of WHILEM in this expr */
fac92740 139 regnode *emit_start; /* Start of emitted-code area */
f55b7b26 140 regnode_offset emit; /* Code-emit pointer */
830247a4
IZ
141 I32 naughty; /* How bad is this pattern? */
142 I32 sawback; /* Did we see \1, ...? */
143 U32 seen;
88f063b4
KW
144 SSize_t size; /* Number of regnode equivalents in
145 pattern */
7c932d07
KW
146
147 /* position beyond 'precomp' of the warning message furthest away from
148 * 'precomp'. During the parse, no warnings are raised for any problems
149 * earlier in the parse than this position. This works if warnings are
150 * raised the first time a given spot is parsed, and if only one
151 * independent warning is raised for any given spot */
152 Size_t latest_warn_offset;
153
154 I32 npar; /* Capture buffer count so far in the
155 parse, (OPEN) plus one. ("par" 0 is
156 the whole pattern)*/
157 I32 total_par; /* During initial parse, is either 0,
158 or -1; the latter indicating a
159 reparse is needed. After that pass,
160 it is what 'npar' became after the
161 pass. Hence, it being > 0 indicates
162 we are in a reparse situation */
538e84ed
KW
163 I32 nestroot; /* root parens we are in - used by
164 accept */
830247a4 165 I32 seen_zerolen;
f55b7b26
KW
166 regnode_offset *open_parens; /* offsets to open parens */
167 regnode_offset *close_parens; /* offsets to close parens */
0275405c 168 I32 parens_buf_size; /* #slots malloced open/close_parens */
d5a00e4a 169 regnode *end_op; /* END node in program */
02daf0ab
YO
170 I32 utf8; /* whether the pattern is utf8 or not */
171 I32 orig_utf8; /* whether the pattern was originally in utf8 */
172 /* XXX use this for future optimisation of case
173 * where pattern must be upgraded to utf8. */
e40e74fe
KW
174 I32 uni_semantics; /* If a d charset modifier should use unicode
175 rules, even if the pattern is not in
176 utf8 */
81714fb9 177 HV *paren_names; /* Paren names */
538e84ed 178
40d049e4 179 regnode **recurse; /* Recurse regops */
88f063b4 180 I32 recurse_count; /* Number of recurse regops we have generated */
4286711a 181 U8 *study_chunk_recursed; /* bitmap of which subs we have moved
538e84ed 182 through */
09a65838 183 U32 study_chunk_recursed_bytes; /* bytes in bitmap */
b57e4118 184 I32 in_lookbehind;
105c827d 185 I32 in_lookahead;
4624b182 186 I32 contains_locale;
bb3f3ed2 187 I32 override_recoding;
4caef9c3 188 I32 recode_x_to_native;
9d53c457 189 I32 in_multi_char_class;
1acab4c5 190 struct reg_code_blocks *code_blocks;/* positions of literal (?{})
68e2671b 191 within pattern */
b1603ef8 192 int code_index; /* next code_blocks[] slot */
ee273784 193 SSize_t maxlen; /* mininum possible number of chars in string to match */
3f910e62
YO
194 scan_frame *frame_head;
195 scan_frame *frame_last;
196 U32 frame_count;
7eec73eb 197 AV *warn_text;
945961fd 198 HV *unlexed_names;
dc6d7f5c 199#ifdef ADD_TO_REGEXEC
830247a4
IZ
200 char *starttry; /* -Dr: where regtry was called. */
201#define RExC_starttry (pRExC_state->starttry)
202#endif
d24ca0c5 203 SV *runtime_code_qr; /* qr with the runtime code blocks */
3dab1dad 204#ifdef DEBUGGING
be8e71aa 205 const char *lastparse;
3dab1dad 206 I32 lastnum;
1f1031fe 207 AV *paren_name_list; /* idx -> name */
d9a72fcc 208 U32 study_chunk_recursed_count;
c9f0d54c
YO
209 SV *mysv1;
210 SV *mysv2;
88f063b4 211
3dab1dad
YO
212#define RExC_lastparse (pRExC_state->lastparse)
213#define RExC_lastnum (pRExC_state->lastnum)
1f1031fe 214#define RExC_paren_name_list (pRExC_state->paren_name_list)
d9a72fcc 215#define RExC_study_chunk_recursed_count (pRExC_state->study_chunk_recursed_count)
c9f0d54c
YO
216#define RExC_mysv (pRExC_state->mysv1)
217#define RExC_mysv1 (pRExC_state->mysv1)
218#define RExC_mysv2 (pRExC_state->mysv2)
219
3dab1dad 220#endif
bd0b62a6 221 bool seen_d_op;
911bd04e 222 bool strict;
da7cf1cc 223 bool study_started;
034602eb 224 bool in_script_run;
7c932d07 225 bool use_BRANCHJ;
09b2b2e6 226};
830247a4 227
e2509266 228#define RExC_flags (pRExC_state->flags)
514a91f1 229#define RExC_pm_flags (pRExC_state->pm_flags)
830247a4 230#define RExC_precomp (pRExC_state->precomp)
51684e3c
KW
231#define RExC_copy_start_in_input (pRExC_state->copy_start_in_input)
232#define RExC_copy_start_in_constructed (pRExC_state->copy_start)
cc16d262 233#define RExC_save_copy_start_in_constructed (pRExC_state->save_copy_start)
711b303b 234#define RExC_precomp_end (pRExC_state->precomp_end)
288b8c02 235#define RExC_rx_sv (pRExC_state->rx_sv)
830247a4 236#define RExC_rx (pRExC_state->rx)
f8fc2ecf 237#define RExC_rxi (pRExC_state->rxi)
fac92740 238#define RExC_start (pRExC_state->start)
830247a4
IZ
239#define RExC_end (pRExC_state->end)
240#define RExC_parse (pRExC_state->parse)
7c932d07 241#define RExC_latest_warn_offset (pRExC_state->latest_warn_offset )
830247a4 242#define RExC_whilem_seen (pRExC_state->whilem_seen)
bd0b62a6
KW
243#define RExC_seen_d_op (pRExC_state->seen_d_op) /* Seen something that differs
244 under /d from /u ? */
512c0f5a 245
7122b237 246#ifdef RE_TRACK_PATTERN_OFFSETS
44f36cd3 247# define RExC_offsets (RExC_rxi->u.offsets) /* I am not like the
538e84ed 248 others */
7122b237 249#endif
830247a4 250#define RExC_emit (pRExC_state->emit)
fac92740 251#define RExC_emit_start (pRExC_state->emit_start)
830247a4
IZ
252#define RExC_sawback (pRExC_state->sawback)
253#define RExC_seen (pRExC_state->seen)
254#define RExC_size (pRExC_state->size)
ee273784 255#define RExC_maxlen (pRExC_state->maxlen)
830247a4 256#define RExC_npar (pRExC_state->npar)
1533d1d2 257#define RExC_total_parens (pRExC_state->total_par)
0275405c 258#define RExC_parens_buf_size (pRExC_state->parens_buf_size)
e2e6a0f1 259#define RExC_nestroot (pRExC_state->nestroot)
830247a4 260#define RExC_seen_zerolen (pRExC_state->seen_zerolen)
1aa99e6b 261#define RExC_utf8 (pRExC_state->utf8)
e40e74fe 262#define RExC_uni_semantics (pRExC_state->uni_semantics)
02daf0ab 263#define RExC_orig_utf8 (pRExC_state->orig_utf8)
40d049e4
YO
264#define RExC_open_parens (pRExC_state->open_parens)
265#define RExC_close_parens (pRExC_state->close_parens)
5bd2d46e 266#define RExC_end_op (pRExC_state->end_op)
81714fb9 267#define RExC_paren_names (pRExC_state->paren_names)
40d049e4
YO
268#define RExC_recurse (pRExC_state->recurse)
269#define RExC_recurse_count (pRExC_state->recurse_count)
09a65838 270#define RExC_study_chunk_recursed (pRExC_state->study_chunk_recursed)
538e84ed
KW
271#define RExC_study_chunk_recursed_bytes \
272 (pRExC_state->study_chunk_recursed_bytes)
b57e4118 273#define RExC_in_lookbehind (pRExC_state->in_lookbehind)
105c827d 274#define RExC_in_lookahead (pRExC_state->in_lookahead)
4624b182 275#define RExC_contains_locale (pRExC_state->contains_locale)
4caef9c3
KW
276#define RExC_recode_x_to_native (pRExC_state->recode_x_to_native)
277
b6d67071 278#ifdef EBCDIC
4caef9c3
KW
279# define SET_recode_x_to_native(x) \
280 STMT_START { RExC_recode_x_to_native = (x); } STMT_END
281#else
282# define SET_recode_x_to_native(x) NOOP
b6d67071 283#endif
4caef9c3 284
9d53c457 285#define RExC_in_multi_char_class (pRExC_state->in_multi_char_class)
3f910e62
YO
286#define RExC_frame_head (pRExC_state->frame_head)
287#define RExC_frame_last (pRExC_state->frame_last)
288#define RExC_frame_count (pRExC_state->frame_count)
67cdf558 289#define RExC_strict (pRExC_state->strict)
da7cf1cc 290#define RExC_study_started (pRExC_state->study_started)
7eec73eb 291#define RExC_warn_text (pRExC_state->warn_text)
034602eb 292#define RExC_in_script_run (pRExC_state->in_script_run)
7c932d07 293#define RExC_use_BRANCHJ (pRExC_state->use_BRANCHJ)
945961fd 294#define RExC_unlexed_names (pRExC_state->unlexed_names)
830247a4 295
99807a43
HS
296/* Heuristic check on the complexity of the pattern: if TOO_NAUGHTY, we set
297 * a flag to disable back-off on the fixed/floating substrings - if it's
298 * a high complexity pattern we assume the benefit of avoiding a full match
299 * is worth the cost of checking for the substrings even if they rarely help.
300 */
301#define RExC_naughty (pRExC_state->naughty)
302#define TOO_NAUGHTY (10)
303#define MARK_NAUGHTY(add) \
304 if (RExC_naughty < TOO_NAUGHTY) \
305 RExC_naughty += (add)
306#define MARK_NAUGHTY_EXP(exp, add) \
307 if (RExC_naughty < TOO_NAUGHTY) \
308 RExC_naughty += RExC_naughty / (exp) + (add)
cde0cee5 309
a687059c
LW
310#define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
311#define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
412f55bb 312 ((*s) == '{' && regcurly(s)))
a687059c
LW
313
314/*
315 * Flags to be passed up and down.
316 */
a687059c 317#define WORST 0 /* Worst case. */
c4ead770
KW
318#define HASWIDTH 0x01 /* Known to not match null strings, could match
319 non-null ones. */
fda99bee 320
e64f369d 321/* Simple enough to be STAR/PLUS operand; in an EXACTish node must be a single
2fd92675
KW
322 * character. (There needs to be a case: in the switch statement in regexec.c
323 * for any node marked SIMPLE.) Note that this is not the same thing as
324 * REGNODE_SIMPLE */
fda99bee 325#define SIMPLE 0x02
e64f369d 326#define SPSTART 0x04 /* Starts with * or + */
8d9c2815
NC
327#define POSTPONED 0x08 /* (?1),(?&name), (??{...}) or similar */
328#define TRYAGAIN 0x10 /* Weeded out a declaration. */
c36361b2
KW
329#define RESTART_PARSE 0x20 /* Need to redo the parse */
330#define NEED_UTF8 0x40 /* In conjunction with RESTART_PARSE, need to
331 calcuate sizes as UTF-8 */
a687059c 332
3dab1dad
YO
333#define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1)
334
07be1b83
YO
335/* whether trie related optimizations are enabled */
336#if PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION
337#define TRIE_STUDY_OPT
786e8c11 338#define FULL_TRIE_STUDY
07be1b83
YO
339#define TRIE_STCLASS
340#endif
1de06328
YO
341
342
40d049e4
YO
343
344#define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3]
345#define PBITVAL(paren) (1 << ((paren) & 7))
346#define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren))
347#define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren)
348#define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren))
349
82a6ada4 350#define REQUIRE_UTF8(flagp) STMT_START { \
8d9c2815 351 if (!UTF) { \
c36361b2 352 *flagp = RESTART_PARSE|NEED_UTF8; \
f55b7b26 353 return 0; \
8d9c2815 354 } \
82a6ada4 355 } STMT_END
40d049e4 356
d63d75f5 357/* Change from /d into /u rules, and restart the parse. RExC_uni_semantics is
f6e8f31e
KW
358 * a flag that indicates we need to override /d with /u as a result of
359 * something in the pattern. It should only be used in regards to calling
360 * set_regex_charset() or get_regex_charse() */
512c0f5a
KW
361#define REQUIRE_UNI_RULES(flagp, restart_retval) \
362 STMT_START { \
363 if (DEPENDS_SEMANTICS) { \
512c0f5a
KW
364 set_regex_charset(&RExC_flags, REGEX_UNICODE_CHARSET); \
365 RExC_uni_semantics = 1; \
b76e1c04 366 if (RExC_seen_d_op && LIKELY(! IN_PARENS_PASS)) { \
bd0b62a6
KW
367 /* No need to restart the parse if we haven't seen \
368 * anything that differs between /u and /d, and no need \
369 * to restart immediately if we're going to reparse \
370 * anyway to count parens */ \
c36361b2 371 *flagp |= RESTART_PARSE; \
512c0f5a 372 return restart_retval; \
d1be211c 373 } \
512c0f5a
KW
374 } \
375 } STMT_END
376
7c932d07
KW
377#define REQUIRE_BRANCHJ(flagp, restart_retval) \
378 STMT_START { \
379 RExC_use_BRANCHJ = 1; \
439a3bfe
KW
380 *flagp |= RESTART_PARSE; \
381 return restart_retval; \
7c932d07
KW
382 } STMT_END
383
b76e1c04
KW
384/* Until we have completed the parse, we leave RExC_total_parens at 0 or
385 * less. After that, it must always be positive, because the whole re is
386 * considered to be surrounded by virtual parens. Setting it to negative
387 * indicates there is some construct that needs to know the actual number of
388 * parens to be properly handled. And that means an extra pass will be
389 * required after we've counted them all */
390#define ALL_PARENS_COUNTED (RExC_total_parens > 0)
7c932d07 391#define REQUIRE_PARENS_PASS \
b76e1c04
KW
392 STMT_START { /* No-op if have completed a pass */ \
393 if (! ALL_PARENS_COUNTED) RExC_total_parens = -1; \
7c932d07 394 } STMT_END
b76e1c04
KW
395#define IN_PARENS_PASS (RExC_total_parens < 0)
396
7c932d07 397
f358228e
KW
398/* This is used to return failure (zero) early from the calling function if
399 * various flags in 'flags' are set. Two flags always cause a return:
400 * 'RESTART_PARSE' and 'NEED_UTF8'. 'extra' can be used to specify any
401 * additional flags that should cause a return; 0 if none. If the return will
402 * be done, '*flagp' is first set to be all of the flags that caused the
403 * return. */
404#define RETURN_FAIL_ON_RESTART_OR_FLAGS(flags,flagp,extra) \
fd7fc6ae 405 STMT_START { \
c36361b2
KW
406 if ((flags) & (RESTART_PARSE|NEED_UTF8|(extra))) { \
407 *(flagp) = (flags) & (RESTART_PARSE|NEED_UTF8|(extra)); \
f358228e 408 return 0; \
fd7fc6ae
YO
409 } \
410 } STMT_END
411
c36361b2 412#define MUST_RESTART(flags) ((flags) & (RESTART_PARSE))
fd7fc6ae 413
5d555bb8 414#define RETURN_FAIL_ON_RESTART(flags,flagp) \
f358228e 415 RETURN_FAIL_ON_RESTART_OR_FLAGS( flags, flagp, 0)
5d555bb8 416#define RETURN_FAIL_ON_RESTART_FLAGP(flagp) \
f358228e 417 if (MUST_RESTART(*(flagp))) return 0
fd7fc6ae 418
f19b1a63
KW
419/* This converts the named class defined in regcomp.h to its equivalent class
420 * number defined in handy.h. */
421#define namedclass_to_classnum(class) ((int) ((class) / 2))
422#define classnum_to_namedclass(classnum) ((classnum) * 2)
423
de92f5e6
KW
424#define _invlist_union_complement_2nd(a, b, output) \
425 _invlist_union_maybe_complement_2nd(a, b, TRUE, output)
426#define _invlist_intersection_complement_2nd(a, b, output) \
427 _invlist_intersection_maybe_complement_2nd(a, b, TRUE, output)
428
1de06328
YO
429/* About scan_data_t.
430
431 During optimisation we recurse through the regexp program performing
432 various inplace (keyhole style) optimisations. In addition study_chunk
433 and scan_commit populate this data structure with information about
538e84ed 434 what strings MUST appear in the pattern. We look for the longest
3b753521 435 string that must appear at a fixed location, and we look for the
1de06328
YO
436 longest string that may appear at a floating location. So for instance
437 in the pattern:
538e84ed 438
1de06328 439 /FOO[xX]A.*B[xX]BAR/
538e84ed 440
1de06328
YO
441 Both 'FOO' and 'A' are fixed strings. Both 'B' and 'BAR' are floating
442 strings (because they follow a .* construct). study_chunk will identify
443 both FOO and BAR as being the longest fixed and floating strings respectively.
538e84ed 444
1de06328 445 The strings can be composites, for instance
538e84ed 446
1de06328 447 /(f)(o)(o)/
538e84ed 448
1de06328 449 will result in a composite fixed substring 'foo'.
538e84ed 450
1de06328 451 For each string some basic information is maintained:
538e84ed 452
6ea77169 453 - min_offset
1de06328
YO
454 This is the position the string must appear at, or not before.
455 It also implicitly (when combined with minlenp) tells us how many
3b753521
FN
456 characters must match before the string we are searching for.
457 Likewise when combined with minlenp and the length of the string it
538e84ed 458 tells us how many characters must appear after the string we have
1de06328 459 found.
538e84ed 460
1de06328
YO
461 - max_offset
462 Only used for floating strings. This is the rightmost point that
ea3daa5d 463 the string can appear at. If set to SSize_t_MAX it indicates that the
1de06328 464 string can occur infinitely far to the right.
6e12f832 465 For fixed strings, it is equal to min_offset.
538e84ed 466
1de06328 467 - minlenp
2d608413
KW
468 A pointer to the minimum number of characters of the pattern that the
469 string was found inside. This is important as in the case of positive
538e84ed 470 lookahead or positive lookbehind we can have multiple patterns
1de06328 471 involved. Consider
538e84ed 472
1de06328 473 /(?=FOO).*F/
538e84ed 474
1de06328
YO
475 The minimum length of the pattern overall is 3, the minimum length
476 of the lookahead part is 3, but the minimum length of the part that
538e84ed 477 will actually match is 1. So 'FOO's minimum length is 3, but the
1de06328 478 minimum length for the F is 1. This is important as the minimum length
538e84ed 479 is used to determine offsets in front of and behind the string being
1de06328 480 looked for. Since strings can be composites this is the length of the
486ec47a 481 pattern at the time it was committed with a scan_commit. Note that
1de06328 482 the length is calculated by study_chunk, so that the minimum lengths
538e84ed 483 are not known until the full pattern has been compiled, thus the
1de06328 484 pointer to the value.
538e84ed 485
1de06328 486 - lookbehind
538e84ed 487
1de06328 488 In the case of lookbehind the string being searched for can be
538e84ed 489 offset past the start point of the final matching string.
1de06328
YO
490 If this value was just blithely removed from the min_offset it would
491 invalidate some of the calculations for how many chars must match
492 before or after (as they are derived from min_offset and minlen and
538e84ed 493 the length of the string being searched for).
1de06328
YO
494 When the final pattern is compiled and the data is moved from the
495 scan_data_t structure into the regexp structure the information
538e84ed
KW
496 about lookbehind is factored in, with the information that would
497 have been lost precalculated in the end_shift field for the
1de06328
YO
498 associated string.
499
500 The fields pos_min and pos_delta are used to store the minimum offset
538e84ed 501 and the delta to the maximum offset at the current point in the pattern.
1de06328
YO
502
503*/
2c2d71f5 504
cb6aef8b
DM
505struct scan_data_substrs {
506 SV *str; /* longest substring found in pattern */
507 SSize_t min_offset; /* earliest point in string it can appear */
508 SSize_t max_offset; /* latest point in string it can appear */
509 SSize_t *minlenp; /* pointer to the minlen relevant to the string */
510 SSize_t lookbehind; /* is the pos of the string modified by LB */
11683ecb 511 I32 flags; /* per substring SF_* and SCF_* flags */
cb6aef8b
DM
512};
513
2c2d71f5 514typedef struct scan_data_t {
1de06328
YO
515 /*I32 len_min; unused */
516 /*I32 len_delta; unused */
49f55535 517 SSize_t pos_min;
ea3daa5d 518 SSize_t pos_delta;
2c2d71f5 519 SV *last_found;
ea3daa5d 520 SSize_t last_end; /* min value, <0 unless valid. */
49f55535 521 SSize_t last_start_min;
ea3daa5d 522 SSize_t last_start_max;
2099df82
DM
523 U8 cur_is_floating; /* whether the last_* values should be set as
524 * the next fixed (0) or floating (1)
525 * substring */
6ea77169 526
2099df82
DM
527 /* [0] is longest fixed substring so far, [1] is longest float so far */
528 struct scan_data_substrs substrs[2];
6ea77169 529
11683ecb 530 I32 flags; /* common SF_* and SCF_* flags */
2c2d71f5 531 I32 whilem_c;
49f55535 532 SSize_t *last_closep;
b8f7bb16 533 regnode_ssc *start_class;
2c2d71f5
JH
534} scan_data_t;
535
a687059c 536/*
e50aee73 537 * Forward declarations for pregcomp()'s friends.
a687059c 538 */
a0d0e21e 539
6ea77169 540static const scan_data_t zero_scan_data = {
a2ca2017 541 0, 0, NULL, 0, 0, 0, 0,
6ea77169 542 {
11683ecb
DM
543 { NULL, 0, 0, 0, 0, 0 },
544 { NULL, 0, 0, 0, 0, 0 },
6ea77169
DM
545 },
546 0, 0, NULL, NULL
547};
c277df42 548
11683ecb
DM
549/* study flags */
550
07be1b83
YO
551#define SF_BEFORE_SEOL 0x0001
552#define SF_BEFORE_MEOL 0x0002
11683ecb 553#define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
c277df42 554
07be1b83
YO
555#define SF_IS_INF 0x0040
556#define SF_HAS_PAR 0x0080
557#define SF_IN_PAR 0x0100
558#define SF_HAS_EVAL 0x0200
bc604ad8
KW
559
560
561/* SCF_DO_SUBSTR is the flag that tells the regexp analyzer to track the
562 * longest substring in the pattern. When it is not set the optimiser keeps
563 * track of position, but does not keep track of the actual strings seen,
564 *
565 * So for instance /foo/ will be parsed with SCF_DO_SUBSTR being true, but
566 * /foo/i will not.
567 *
568 * Similarly, /foo.*(blah|erm|huh).*fnorble/ will have "foo" and "fnorble"
569 * parsed with SCF_DO_SUBSTR on, but while processing the (...) it will be
570 * turned off because of the alternation (BRANCH). */
07be1b83 571#define SCF_DO_SUBSTR 0x0400
bc604ad8 572
653099ff
GS
573#define SCF_DO_STCLASS_AND 0x0800
574#define SCF_DO_STCLASS_OR 0x1000
575#define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
e1901655 576#define SCF_WHILEM_VISITED_POS 0x2000
c277df42 577
786e8c11 578#define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */
538e84ed 579#define SCF_SEEN_ACCEPT 0x8000
688e0391 580#define SCF_TRIE_DOING_RESTUDY 0x10000
4286711a
YO
581#define SCF_IN_DEFINE 0x20000
582
583
584
07be1b83 585
43fead97 586#define UTF cBOOL(RExC_utf8)
00b27cfc
KW
587
588/* The enums for all these are ordered so things work out correctly */
a62b1201 589#define LOC (get_regex_charset(RExC_flags) == REGEX_LOCALE_CHARSET)
538e84ed
KW
590#define DEPENDS_SEMANTICS (get_regex_charset(RExC_flags) \
591 == REGEX_DEPENDS_CHARSET)
00b27cfc 592#define UNI_SEMANTICS (get_regex_charset(RExC_flags) == REGEX_UNICODE_CHARSET)
538e84ed
KW
593#define AT_LEAST_UNI_SEMANTICS (get_regex_charset(RExC_flags) \
594 >= REGEX_UNICODE_CHARSET)
595#define ASCII_RESTRICTED (get_regex_charset(RExC_flags) \
596 == REGEX_ASCII_RESTRICTED_CHARSET)
597#define AT_LEAST_ASCII_RESTRICTED (get_regex_charset(RExC_flags) \
598 >= REGEX_ASCII_RESTRICTED_CHARSET)
599#define ASCII_FOLD_RESTRICTED (get_regex_charset(RExC_flags) \
600 == REGEX_ASCII_MORE_RESTRICTED_CHARSET)
a62b1201 601
43fead97 602#define FOLD cBOOL(RExC_flags & RXf_PMf_FOLD)
a0ed51b3 603
f2c2a6ab
KW
604/* For programs that want to be strictly Unicode compatible by dying if any
605 * attempt is made to match a non-Unicode code point against a Unicode
606 * property. */
607#define ALWAYS_WARN_SUPER ckDEAD(packWARN(WARN_NON_UNICODE))
608
93733859 609#define OOB_NAMEDCLASS -1
b8c5462f 610
8e661ac5
KW
611/* There is no code point that is out-of-bounds, so this is problematic. But
612 * its only current use is to initialize a variable that is always set before
613 * looked at. */
614#define OOB_UNICODE 0xDEADBEEF
615
a0ed51b3 616#define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
a0ed51b3 617
8615cb43 618
b45f050a
JF
619/* length of regex to show in messages that don't mark a position within */
620#define RegexLengthToShowInErrorMessages 127
621
622/*
623 * If MARKER[12] are adjusted, be sure to adjust the constants at the top
624 * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
625 * op/pragma/warn/regcomp.
626 */
7253e4e3
RK
627#define MARKER1 "<-- HERE" /* marker as it appears in the description */
628#define MARKER2 " <-- HERE " /* marker as it appears within the regex */
b81d288d 629
538e84ed 630#define REPORT_LOCATION " in regex; marked by " MARKER1 \
147e3846 631 " in m/%" UTF8f MARKER2 "%" UTF8f "/"
b45f050a 632
285b5ca0
KW
633/* The code in this file in places uses one level of recursion with parsing
634 * rebased to an alternate string constructed by us in memory. This can take
635 * the form of something that is completely different from the input, or
636 * something that uses the input as part of the alternate. In the first case,
637 * there should be no possibility of an error, as we are in complete control of
51684e3c
KW
638 * the alternate string. But in the second case we don't completely control
639 * the input portion, so there may be errors in that. Here's an example:
285b5ca0
KW
640 * /[abc\x{DF}def]/ui
641 * is handled specially because \x{df} folds to a sequence of more than one
51684e3c 642 * character: 'ss'. What is done is to create and parse an alternate string,
285b5ca0
KW
643 * which looks like this:
644 * /(?:\x{DF}|[abc\x{DF}def])/ui
645 * where it uses the input unchanged in the middle of something it constructs,
646 * which is a branch for the DF outside the character class, and clustering
647 * parens around the whole thing. (It knows enough to skip the DF inside the
648 * class while in this substitute parse.) 'abc' and 'def' may have errors that
649 * need to be reported. The general situation looks like this:
650 *
51684e3c 651 * |<------- identical ------>|
285b5ca0 652 * sI tI xI eI
51684e3c 653 * Input: ---------------------------------------------------------------
285b5ca0
KW
654 * Constructed: ---------------------------------------------------
655 * sC tC xC eC EC
51684e3c 656 * |<------- identical ------>|
285b5ca0 657 *
51684e3c
KW
658 * sI..eI is the portion of the input pattern we are concerned with here.
659 * sC..EC is the constructed substitute parse string.
660 * sC..tC is constructed by us
661 * tC..eC is an exact duplicate of the portion of the input pattern tI..eI.
662 * In the diagram, these are vertically aligned.
663 * eC..EC is also constructed by us.
664 * xC is the position in the substitute parse string where we found a
665 * problem.
666 * xI is the position in the original pattern corresponding to xC.
285b5ca0 667 *
51684e3c
KW
668 * We want to display a message showing the real input string. Thus we need to
669 * translate from xC to xI. We know that xC >= tC, since the portion of the
670 * string sC..tC has been constructed by us, and so shouldn't have errors. We
671 * get:
672 * xI = tI + (xC - tC)
285b5ca0 673 *
51684e3c
KW
674 * When the substitute parse is constructed, the code needs to set:
675 * RExC_start (sC)
676 * RExC_end (eC)
677 * RExC_copy_start_in_input (tI)
678 * RExC_copy_start_in_constructed (tC)
679 * and restore them when done.
285b5ca0 680 *
51684e3c
KW
681 * During normal processing of the input pattern, both
682 * 'RExC_copy_start_in_input' and 'RExC_copy_start_in_constructed' are set to
683 * sI, so that xC equals xI.
285b5ca0
KW
684 */
685
51684e3c
KW
686#define sI RExC_precomp
687#define eI RExC_precomp_end
688#define sC RExC_start
689#define eC RExC_end
690#define tI RExC_copy_start_in_input
691#define tC RExC_copy_start_in_constructed
692#define xI(xC) (tI + (xC - tC))
693#define xI_offset(xC) (xI(xC) - sI)
285b5ca0
KW
694
695#define REPORT_LOCATION_ARGS(xC) \
696 UTF8fARG(UTF, \
51684e3c 697 (xI(xC) > eI) /* Don't run off end */ \
232b691f 698 ? eI - sI /* Length before the <--HERE */ \
bc929b61
KW
699 : ((xI_offset(xC) >= 0) \
700 ? xI_offset(xC) \
701 : (Perl_croak(aTHX_ "panic: %s: %d: negative offset: %" \
702 IVdf " trying to output message for " \
703 " pattern %.*s", \
51684e3c 704 __FILE__, __LINE__, (IV) xI_offset(xC), \
bc929b61 705 ((int) (eC - sC)), sC), 0)), \
51684e3c 706 sI), /* The input pattern printed up to the <--HERE */ \
285b5ca0 707 UTF8fARG(UTF, \
51684e3c
KW
708 (xI(xC) > eI) ? 0 : eI - xI(xC), /* Length after <--HERE */ \
709 (xI(xC) > eI) ? eI : xI(xC)) /* pattern after <--HERE */
c1d900c3 710
8a6d8ec6
HS
711/* Used to point after bad bytes for an error message, but avoid skipping
712 * past a nul byte. */
e3304382 713#define SKIP_IF_CHAR(s, e) (!*(s) ? 0 : UTF ? UTF8_SAFE_SKIP(s, e) : 1)
8a6d8ec6 714
ed68fe11
KW
715/* Set up to clean up after our imminent demise */
716#define PREPARE_TO_DIE \
717 STMT_START { \
718 if (RExC_rx_sv) \
719 SAVEFREESV(RExC_rx_sv); \
7c932d07
KW
720 if (RExC_open_parens) \
721 SAVEFREEPV(RExC_open_parens); \
722 if (RExC_close_parens) \
723 SAVEFREEPV(RExC_close_parens); \
ed68fe11
KW
724 } STMT_END
725
b45f050a
JF
726/*
727 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
728 * arg. Show regex, up to a maximum length. If it's too long, chop and add
729 * "...".
730 */
58e23c8d 731#define _FAIL(code) STMT_START { \
bfed75c6 732 const char *ellipses = ""; \
88f063b4 733 IV len = RExC_precomp_end - RExC_precomp; \
ccb2c380 734 \
ed68fe11 735 PREPARE_TO_DIE; \
ccb2c380
MP
736 if (len > RegexLengthToShowInErrorMessages) { \
737 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
738 len = RegexLengthToShowInErrorMessages - 10; \
739 ellipses = "..."; \
740 } \
58e23c8d 741 code; \
ccb2c380 742} STMT_END
8615cb43 743
58e23c8d 744#define FAIL(msg) _FAIL( \
147e3846 745 Perl_croak(aTHX_ "%s in regex m/%" UTF8f "%s/", \
c1d900c3 746 msg, UTF8fARG(UTF, len, RExC_precomp), ellipses))
58e23c8d
YO
747
748#define FAIL2(msg,arg) _FAIL( \
147e3846 749 Perl_croak(aTHX_ msg " in regex m/%" UTF8f "%s/", \
c1d900c3 750 arg, UTF8fARG(UTF, len, RExC_precomp), ellipses))
58e23c8d 751
4b7718d6
KW
752#define FAIL3(msg,arg1,arg2) _FAIL( \
753 Perl_croak(aTHX_ msg " in regex m/%" UTF8f "%s/", \
754 arg1, arg2, UTF8fARG(UTF, len, RExC_precomp), ellipses))
755
b45f050a 756/*
b45f050a
JF
757 * Simple_vFAIL -- like FAIL, but marks the current location in the scan
758 */
ccb2c380 759#define Simple_vFAIL(m) STMT_START { \
ccb2c380 760 Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
d528642a 761 m, REPORT_LOCATION_ARGS(RExC_parse)); \
ccb2c380 762} STMT_END
b45f050a
JF
763
764/*
765 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
766 */
ccb2c380 767#define vFAIL(m) STMT_START { \
ed68fe11 768 PREPARE_TO_DIE; \
ccb2c380
MP
769 Simple_vFAIL(m); \
770} STMT_END
b45f050a
JF
771
772/*
773 * Like Simple_vFAIL(), but accepts two arguments.
774 */
ccb2c380 775#define Simple_vFAIL2(m,a1) STMT_START { \
d528642a
KW
776 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, \
777 REPORT_LOCATION_ARGS(RExC_parse)); \
ccb2c380 778} STMT_END
b45f050a
JF
779
780/*
781 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
782 */
ccb2c380 783#define vFAIL2(m,a1) STMT_START { \
ed68fe11 784 PREPARE_TO_DIE; \
ccb2c380
MP
785 Simple_vFAIL2(m, a1); \
786} STMT_END
b45f050a
JF
787
788
789/*
790 * Like Simple_vFAIL(), but accepts three arguments.
791 */
ccb2c380 792#define Simple_vFAIL3(m, a1, a2) STMT_START { \
c1d900c3 793 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, \
d528642a 794 REPORT_LOCATION_ARGS(RExC_parse)); \
ccb2c380 795} STMT_END
b45f050a
JF
796
797/*
798 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
799 */
ccb2c380 800#define vFAIL3(m,a1,a2) STMT_START { \
ed68fe11 801 PREPARE_TO_DIE; \
ccb2c380
MP
802 Simple_vFAIL3(m, a1, a2); \
803} STMT_END
b45f050a
JF
804
805/*
806 * Like Simple_vFAIL(), but accepts four arguments.
807 */
ccb2c380 808#define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \
d528642a
KW
809 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, a3, \
810 REPORT_LOCATION_ARGS(RExC_parse)); \
ccb2c380 811} STMT_END
b45f050a 812
95db3ffa 813#define vFAIL4(m,a1,a2,a3) STMT_START { \
ed68fe11 814 PREPARE_TO_DIE; \
95db3ffa
KW
815 Simple_vFAIL4(m, a1, a2, a3); \
816} STMT_END
817
946095af 818/* A specialized version of vFAIL2 that works with UTF8f */
d528642a 819#define vFAIL2utf8f(m, a1) STMT_START { \
ed68fe11 820 PREPARE_TO_DIE; \
d528642a
KW
821 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, \
822 REPORT_LOCATION_ARGS(RExC_parse)); \
946095af
BF
823} STMT_END
824
3ba22297 825#define vFAIL3utf8f(m, a1, a2) STMT_START { \
ed68fe11 826 PREPARE_TO_DIE; \
3ba22297 827 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, \
d528642a 828 REPORT_LOCATION_ARGS(RExC_parse)); \
3ba22297
KW
829} STMT_END
830
d62e749e 831/* Setting this to NULL is a signal to not output warnings */
cc16d262
KW
832#define TURN_OFF_WARNINGS_IN_SUBSTITUTE_PARSE \
833 STMT_START { \
834 RExC_save_copy_start_in_constructed = RExC_copy_start_in_constructed;\
835 RExC_copy_start_in_constructed = NULL; \
836 } STMT_END
837#define RESTORE_WARNINGS \
838 RExC_copy_start_in_constructed = RExC_save_copy_start_in_constructed
d62e749e 839
7c932d07
KW
840/* Since a warning can be generated multiple times as the input is reparsed, we
841 * output it the first time we come to that point in the parse, but suppress it
842 * otherwise. 'RExC_copy_start_in_constructed' being NULL is a flag to not
843 * generate any warnings */
55bcf857 844#define TO_OUTPUT_WARNINGS(loc) \
7c932d07
KW
845 ( RExC_copy_start_in_constructed \
846 && ((xI(loc)) - RExC_precomp) > (Ptrdiff_t) RExC_latest_warn_offset)
55bcf857 847
7c932d07
KW
848/* After we've emitted a warning, we save the position in the input so we don't
849 * output it again */
850#define UPDATE_WARNINGS_LOC(loc) \
851 STMT_START { \
852 if (TO_OUTPUT_WARNINGS(loc)) { \
853 RExC_latest_warn_offset = (xI(loc)) - RExC_precomp; \
854 } \
855 } STMT_END
77ccba3c 856
086f21bf 857/* 'warns' is the output of the packWARNx macro used in 'code' */
f8220fe2
KW
858#define _WARN_HELPER(loc, warns, code) \
859 STMT_START { \
d62e749e
KW
860 if (! RExC_copy_start_in_constructed) { \
861 Perl_croak( aTHX_ "panic! %s: %d: Tried to warn when none" \
862 " expected at '%s'", \
863 __FILE__, __LINE__, loc); \
864 } \
55bcf857 865 if (TO_OUTPUT_WARNINGS(loc)) { \
086f21bf
KW
866 if (ckDEAD(warns)) \
867 PREPARE_TO_DIE; \
338190a1 868 code; \
77ccba3c 869 UPDATE_WARNINGS_LOC(loc); \
338190a1 870 } \
f8220fe2 871 } STMT_END
946095af 872
5e0a247b 873/* m is not necessarily a "literal string", in this macro */
f8220fe2
KW
874#define reg_warn_non_literal_string(loc, m) \
875 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
876 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a 877 "%s" REPORT_LOCATION, \
f8220fe2 878 m, REPORT_LOCATION_ARGS(loc)))
5e0a247b 879
f8220fe2
KW
880#define ckWARNreg(loc,m) \
881 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
882 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a 883 m REPORT_LOCATION, \
f8220fe2 884 REPORT_LOCATION_ARGS(loc)))
ccb2c380 885
f8220fe2
KW
886#define vWARN(loc, m) \
887 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
888 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a 889 m REPORT_LOCATION, \
f8220fe2 890 REPORT_LOCATION_ARGS(loc))) \
b927b7e9 891
f8220fe2
KW
892#define vWARN_dep(loc, m) \
893 _WARN_HELPER(loc, packWARN(WARN_DEPRECATED), \
894 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), \
d528642a 895 m REPORT_LOCATION, \
f8220fe2 896 REPORT_LOCATION_ARGS(loc)))
0d6106aa 897
f8220fe2
KW
898#define ckWARNdep(loc,m) \
899 _WARN_HELPER(loc, packWARN(WARN_DEPRECATED), \
900 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
d528642a 901 m REPORT_LOCATION, \
f8220fe2 902 REPORT_LOCATION_ARGS(loc)))
147508a2 903
f8220fe2
KW
904#define ckWARNregdep(loc,m) \
905 _WARN_HELPER(loc, packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
906 Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, \
d528642a
KW
907 WARN_REGEXP), \
908 m REPORT_LOCATION, \
f8220fe2 909 REPORT_LOCATION_ARGS(loc)))
ccb2c380 910
f8220fe2
KW
911#define ckWARN2reg_d(loc,m, a1) \
912 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
913 Perl_ck_warner_d(aTHX_ packWARN(WARN_REGEXP), \
d528642a 914 m REPORT_LOCATION, \
f8220fe2 915 a1, REPORT_LOCATION_ARGS(loc)))
2335b3d3 916
f8220fe2
KW
917#define ckWARN2reg(loc, m, a1) \
918 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
919 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a 920 m REPORT_LOCATION, \
f8220fe2 921 a1, REPORT_LOCATION_ARGS(loc)))
ccb2c380 922
f8220fe2
KW
923#define vWARN3(loc, m, a1, a2) \
924 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
925 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a 926 m REPORT_LOCATION, \
f8220fe2 927 a1, a2, REPORT_LOCATION_ARGS(loc)))
ccb2c380 928
f8220fe2
KW
929#define ckWARN3reg(loc, m, a1, a2) \
930 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
931 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a
KW
932 m REPORT_LOCATION, \
933 a1, a2, \
f8220fe2 934 REPORT_LOCATION_ARGS(loc)))
668c081a 935
f8220fe2
KW
936#define vWARN4(loc, m, a1, a2, a3) \
937 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
938 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a
KW
939 m REPORT_LOCATION, \
940 a1, a2, a3, \
f8220fe2 941 REPORT_LOCATION_ARGS(loc)))
ccb2c380 942
f8220fe2
KW
943#define ckWARN4reg(loc, m, a1, a2, a3) \
944 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
945 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a
KW
946 m REPORT_LOCATION, \
947 a1, a2, a3, \
f8220fe2 948 REPORT_LOCATION_ARGS(loc)))
668c081a 949
f8220fe2
KW
950#define vWARN5(loc, m, a1, a2, a3, a4) \
951 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
952 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a
KW
953 m REPORT_LOCATION, \
954 a1, a2, a3, a4, \
f8220fe2 955 REPORT_LOCATION_ARGS(loc)))
9d1d55b5 956
aea5f181
KW
957#define ckWARNexperimental(loc, class, m) \
958 _WARN_HELPER(loc, packWARN(class), \
959 Perl_ck_warner_d(aTHX_ packWARN(class), \
960 m REPORT_LOCATION, \
961 REPORT_LOCATION_ARGS(loc)))
962
1a322bb4
KW
963/* Convert between a pointer to a node and its offset from the beginning of the
964 * program */
965#define REGNODE_p(offset) (RExC_emit_start + (offset))
966#define REGNODE_OFFSET(node) ((node) - RExC_emit_start)
967
538e84ed 968/* Macros for recording node offsets. 20001227 mjd@plover.com
fac92740
MJD
969 * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
970 * element 2*n-1 of the array. Element #2n holds the byte length node #n.
971 * Element 0 holds the number n.
07be1b83 972 * Position is 1 indexed.
fac92740 973 */
7122b237 974#ifndef RE_TRACK_PATTERN_OFFSETS
d7c442ce 975#define Set_Node_Offset_To_R(offset,byte)
7122b237
YO
976#define Set_Node_Offset(node,byte)
977#define Set_Cur_Node_Offset
978#define Set_Node_Length_To_R(node,len)
979#define Set_Node_Length(node,len)
6a86c6ad 980#define Set_Node_Cur_Length(node,start)
538e84ed
KW
981#define Node_Offset(n)
982#define Node_Length(n)
7122b237
YO
983#define Set_Node_Offset_Length(node,offset,len)
984#define ProgLen(ri) ri->u.proglen
985#define SetProgLen(ri,x) ri->u.proglen = x
4cbeff97 986#define Track_Code(code)
7122b237
YO
987#else
988#define ProgLen(ri) ri->u.offsets[0]
989#define SetProgLen(ri,x) ri->u.offsets[0] = x
d7c442ce 990#define Set_Node_Offset_To_R(offset,byte) STMT_START { \
ccb2c380 991 MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
d7c442ce
KW
992 __LINE__, (int)(offset), (int)(byte))); \
993 if((offset) < 0) { \
538e84ed 994 Perl_croak(aTHX_ "value of node is %d in Offset macro", \
d7c442ce 995 (int)(offset)); \
ccb2c380 996 } else { \
d7c442ce 997 RExC_offsets[2*(offset)-1] = (byte); \
ccb2c380 998 } \
ccb2c380
MP
999} STMT_END
1000
88f063b4 1001#define Set_Node_Offset(node,byte) \
1a322bb4 1002 Set_Node_Offset_To_R(REGNODE_OFFSET(node), (byte)-RExC_start)
ccb2c380
MP
1003#define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
1004
1005#define Set_Node_Length_To_R(node,len) STMT_START { \
ccb2c380 1006 MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \
551405c4 1007 __LINE__, (int)(node), (int)(len))); \
ccb2c380 1008 if((node) < 0) { \
538e84ed
KW
1009 Perl_croak(aTHX_ "value of node is %d in Length macro", \
1010 (int)(node)); \
ccb2c380
MP
1011 } else { \
1012 RExC_offsets[2*(node)] = (len); \
1013 } \
ccb2c380
MP
1014} STMT_END
1015
1016#define Set_Node_Length(node,len) \
1a322bb4 1017 Set_Node_Length_To_R(REGNODE_OFFSET(node), len)
6a86c6ad
NC
1018#define Set_Node_Cur_Length(node, start) \
1019 Set_Node_Length(node, RExC_parse - start)
fac92740
MJD
1020
1021/* Get offsets and lengths */
1a322bb4
KW
1022#define Node_Offset(n) (RExC_offsets[2*(REGNODE_OFFSET(n))-1])
1023#define Node_Length(n) (RExC_offsets[2*(REGNODE_OFFSET(n))])
fac92740 1024
07be1b83 1025#define Set_Node_Offset_Length(node,offset,len) STMT_START { \
1a322bb4
KW
1026 Set_Node_Offset_To_R(REGNODE_OFFSET(node), (offset)); \
1027 Set_Node_Length_To_R(REGNODE_OFFSET(node), (len)); \
07be1b83 1028} STMT_END
4cbeff97
TC
1029
1030#define Track_Code(code) STMT_START { code } STMT_END
7122b237 1031#endif
07be1b83
YO
1032
1033#if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS
1034#define EXPERIMENTAL_INPLACESCAN
f427392e 1035#endif /*PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS*/
07be1b83 1036
cb41e5d6
YO
1037#ifdef DEBUGGING
1038int
6ad9a8ab 1039Perl_re_printf(pTHX_ const char *fmt, ...)
cb41e5d6
YO
1040{
1041 va_list ap;
1042 int result;
1043 PerlIO *f= Perl_debug_log;
1044 PERL_ARGS_ASSERT_RE_PRINTF;
1045 va_start(ap, fmt);
1046 result = PerlIO_vprintf(f, fmt, ap);
1047 va_end(ap);
1048 return result;
1049}
1050
1051int
7b031478 1052Perl_re_indentf(pTHX_ const char *fmt, U32 depth, ...)
cb41e5d6
YO
1053{
1054 va_list ap;
1055 int result;
1056 PerlIO *f= Perl_debug_log;
1057 PERL_ARGS_ASSERT_RE_INDENTF;
1058 va_start(ap, depth);
daeb874b 1059 PerlIO_printf(f, "%*s", ( (int)depth % 20 ) * 2, "");
cb41e5d6
YO
1060 result = PerlIO_vprintf(f, fmt, ap);
1061 va_end(ap);
1062 return result;
1063}
1064#endif /* DEBUGGING */
1065
7b031478 1066#define DEBUG_RExC_seen() \
538e84ed 1067 DEBUG_OPTIMISE_MORE_r({ \
88f063b4 1068 Perl_re_printf( aTHX_ "RExC_seen: "); \
538e84ed 1069 \
e384d5c1 1070 if (RExC_seen & REG_ZERO_LEN_SEEN) \
88f063b4 1071 Perl_re_printf( aTHX_ "REG_ZERO_LEN_SEEN "); \
538e84ed 1072 \
e384d5c1 1073 if (RExC_seen & REG_LOOKBEHIND_SEEN) \
88f063b4 1074 Perl_re_printf( aTHX_ "REG_LOOKBEHIND_SEEN "); \
538e84ed 1075 \
e384d5c1 1076 if (RExC_seen & REG_GPOS_SEEN) \
88f063b4 1077 Perl_re_printf( aTHX_ "REG_GPOS_SEEN "); \
538e84ed 1078 \
e384d5c1 1079 if (RExC_seen & REG_RECURSE_SEEN) \
88f063b4 1080 Perl_re_printf( aTHX_ "REG_RECURSE_SEEN "); \
538e84ed 1081 \
7b031478 1082 if (RExC_seen & REG_TOP_LEVEL_BRANCHES_SEEN) \
88f063b4 1083 Perl_re_printf( aTHX_ "REG_TOP_LEVEL_BRANCHES_SEEN "); \
538e84ed 1084 \
e384d5c1 1085 if (RExC_seen & REG_VERBARG_SEEN) \
88f063b4 1086 Perl_re_printf( aTHX_ "REG_VERBARG_SEEN "); \
538e84ed 1087 \
e384d5c1 1088 if (RExC_seen & REG_CUTGROUP_SEEN) \
88f063b4 1089 Perl_re_printf( aTHX_ "REG_CUTGROUP_SEEN "); \
538e84ed 1090 \
e384d5c1 1091 if (RExC_seen & REG_RUN_ON_COMMENT_SEEN) \
88f063b4 1092 Perl_re_printf( aTHX_ "REG_RUN_ON_COMMENT_SEEN "); \
538e84ed 1093 \
e384d5c1 1094 if (RExC_seen & REG_UNFOLDED_MULTI_SEEN) \
88f063b4 1095 Perl_re_printf( aTHX_ "REG_UNFOLDED_MULTI_SEEN "); \
538e84ed 1096 \
7b031478 1097 if (RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) \
88f063b4 1098 Perl_re_printf( aTHX_ "REG_UNBOUNDED_QUANTIFIER_SEEN "); \
ee273784 1099 \
88f063b4 1100 Perl_re_printf( aTHX_ "\n"); \
9e9ecfdf
YO
1101 });
1102
fdfb4f21 1103#define DEBUG_SHOW_STUDY_FLAG(flags,flag) \
6ad9a8ab 1104 if ((flags) & flag) Perl_re_printf( aTHX_ "%s ", #flag)
fdfb4f21 1105
fdfb4f21 1106
f5a36d78
DM
1107#ifdef DEBUGGING
1108static void
1109S_debug_show_study_flags(pTHX_ U32 flags, const char *open_str,
1110 const char *close_str)
1111{
1112 if (!flags)
1113 return;
1114
1115 Perl_re_printf( aTHX_ "%s", open_str);
11683ecb
DM
1116 DEBUG_SHOW_STUDY_FLAG(flags, SF_BEFORE_SEOL);
1117 DEBUG_SHOW_STUDY_FLAG(flags, SF_BEFORE_MEOL);
f5a36d78
DM
1118 DEBUG_SHOW_STUDY_FLAG(flags, SF_IS_INF);
1119 DEBUG_SHOW_STUDY_FLAG(flags, SF_HAS_PAR);
1120 DEBUG_SHOW_STUDY_FLAG(flags, SF_IN_PAR);
1121 DEBUG_SHOW_STUDY_FLAG(flags, SF_HAS_EVAL);
1122 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_SUBSTR);
1123 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_STCLASS_AND);
1124 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_STCLASS_OR);
1125 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_STCLASS);
1126 DEBUG_SHOW_STUDY_FLAG(flags, SCF_WHILEM_VISITED_POS);
1127 DEBUG_SHOW_STUDY_FLAG(flags, SCF_TRIE_RESTUDY);
1128 DEBUG_SHOW_STUDY_FLAG(flags, SCF_SEEN_ACCEPT);
1129 DEBUG_SHOW_STUDY_FLAG(flags, SCF_TRIE_DOING_RESTUDY);
1130 DEBUG_SHOW_STUDY_FLAG(flags, SCF_IN_DEFINE);
1131 Perl_re_printf( aTHX_ "%s", close_str);
1132}
1133
1134
1135static void
1136S_debug_studydata(pTHX_ const char *where, scan_data_t *data,
1137 U32 depth, int is_inf)
1138{
1139 GET_RE_DEBUG_FLAGS_DECL;
1140
1141 DEBUG_OPTIMISE_MORE_r({
1142 if (!data)
1143 return;
1144 Perl_re_indentf(aTHX_ "%s: Pos:%" IVdf "/%" IVdf " Flags: 0x%" UVXf,
1145 depth,
1146 where,
1147 (IV)data->pos_min,
1148 (IV)data->pos_delta,
1149 (UV)data->flags
1150 );
1151
1152 S_debug_show_study_flags(aTHX_ data->flags," [","]");
fdfb4f21 1153
f5a36d78
DM
1154 Perl_re_printf( aTHX_
1155 " Whilem_c: %" IVdf " Lcp: %" IVdf " %s",
1156 (IV)data->whilem_c,
1157 (IV)(data->last_closep ? *((data)->last_closep) : -1),
1158 is_inf ? "INF " : ""
1159 );
1160
11683ecb 1161 if (data->last_found) {
37b6262f 1162 int i;
11683ecb
DM
1163 Perl_re_printf(aTHX_
1164 "Last:'%s' %" IVdf ":%" IVdf "/%" IVdf,
1165 SvPVX_const(data->last_found),
1166 (IV)data->last_end,
1167 (IV)data->last_start_min,
1168 (IV)data->last_start_max
1169 );
1170
2099df82 1171 for (i = 0; i < 2; i++) {
37b6262f
DM
1172 Perl_re_printf(aTHX_
1173 " %s%s: '%s' @ %" IVdf "/%" IVdf,
2099df82
DM
1174 data->cur_is_floating == i ? "*" : "",
1175 i ? "Float" : "Fixed",
37b6262f
DM
1176 SvPVX_const(data->substrs[i].str),
1177 (IV)data->substrs[i].min_offset,
1178 (IV)data->substrs[i].max_offset
1179 );
1180 S_debug_show_study_flags(aTHX_ data->substrs[i].flags," [","]");
1181 }
11683ecb 1182 }
f5a36d78
DM
1183
1184 Perl_re_printf( aTHX_ "\n");
1185 });
1186}
1187
1188
1189static void
1190S_debug_peep(pTHX_ const char *str, const RExC_state_t *pRExC_state,
1191 regnode *scan, U32 depth, U32 flags)
1192{
1193 GET_RE_DEBUG_FLAGS_DECL;
1194
1195 DEBUG_OPTIMISE_r({
1196 regnode *Next;
1197
1198 if (!scan)
1199 return;
1200 Next = regnext(scan);
1201 regprop(RExC_rx, RExC_mysv, scan, NULL, pRExC_state);
1202 Perl_re_indentf( aTHX_ "%s>%3d: %s (%d)",
1203 depth,
1204 str,
1205 REG_NODE_NUM(scan), SvPV_nolen_const(RExC_mysv),
1206 Next ? (REG_NODE_NUM(Next)) : 0 );
1207 S_debug_show_study_flags(aTHX_ flags," [ ","]");
1208 Perl_re_printf( aTHX_ "\n");
1209 });
1210}
1211
1212
1213# define DEBUG_STUDYDATA(where, data, depth, is_inf) \
1214 S_debug_studydata(aTHX_ where, data, depth, is_inf)
1215
1216# define DEBUG_PEEP(str, scan, depth, flags) \
1217 S_debug_peep(aTHX_ str, pRExC_state, scan, depth, flags)
1218
1219#else
1220# define DEBUG_STUDYDATA(where, data, depth, is_inf) NOOP
1221# define DEBUG_PEEP(str, scan, depth, flags) NOOP
1222#endif
1de06328 1223
cb41e5d6 1224
c6871b76
KW
1225/* =========================================================
1226 * BEGIN edit_distance stuff.
1227 *
1228 * This calculates how many single character changes of any type are needed to
1229 * transform a string into another one. It is taken from version 3.1 of
1230 *
1231 * https://metacpan.org/pod/Text::Levenshtein::Damerau::XS
1232 */
1233
1234/* Our unsorted dictionary linked list. */
1235/* Note we use UVs, not chars. */
1236
1237struct dictionary{
1238 UV key;
1239 UV value;
1240 struct dictionary* next;
1241};
1242typedef struct dictionary item;
1243
1244
1245PERL_STATIC_INLINE item*
88f063b4 1246push(UV key, item* curr)
c6871b76
KW
1247{
1248 item* head;
d09f14bf 1249 Newx(head, 1, item);
c6871b76
KW
1250 head->key = key;
1251 head->value = 0;
1252 head->next = curr;
1253 return head;
1254}
1255
1256
1257PERL_STATIC_INLINE item*
1258find(item* head, UV key)
1259{
1260 item* iterator = head;
1261 while (iterator){
1262 if (iterator->key == key){
1263 return iterator;
1264 }
1265 iterator = iterator->next;
1266 }
1267
1268 return NULL;
1269}
1270
1271PERL_STATIC_INLINE item*
88f063b4 1272uniquePush(item* head, UV key)
c6871b76
KW
1273{
1274 item* iterator = head;
1275
1276 while (iterator){
1277 if (iterator->key == key) {
1278 return head;
1279 }
1280 iterator = iterator->next;
1281 }
1282
88f063b4 1283 return push(key, head);
c6871b76
KW
1284}
1285
1286PERL_STATIC_INLINE void
1287dict_free(item* head)
1288{
1289 item* iterator = head;
1290
1291 while (iterator) {
1292 item* temp = iterator;
1293 iterator = iterator->next;
1294 Safefree(temp);
1295 }
1296
1297 head = NULL;
1298}
1299
1300/* End of Dictionary Stuff */
1301
1302/* All calculations/work are done here */
1303STATIC int
1304S_edit_distance(const UV* src,
1305 const UV* tgt,
1306 const STRLEN x, /* length of src[] */
1307 const STRLEN y, /* length of tgt[] */
1308 const SSize_t maxDistance
1309)
1310{
1311 item *head = NULL;
88f063b4 1312 UV swapCount, swapScore, targetCharCount, i, j;
c6871b76
KW
1313 UV *scores;
1314 UV score_ceil = x + y;
1315
1316 PERL_ARGS_ASSERT_EDIT_DISTANCE;
1317
1318 /* intialize matrix start values */
d09f14bf 1319 Newx(scores, ( (x + 2) * (y + 2)), UV);
c6871b76
KW
1320 scores[0] = score_ceil;
1321 scores[1 * (y + 2) + 0] = score_ceil;
1322 scores[0 * (y + 2) + 1] = score_ceil;
1323 scores[1 * (y + 2) + 1] = 0;
88f063b4 1324 head = uniquePush(uniquePush(head, src[0]), tgt[0]);
c6871b76
KW
1325
1326 /* work loops */
1327 /* i = src index */
1328 /* j = tgt index */
1329 for (i=1;i<=x;i++) {
1330 if (i < x)
88f063b4 1331 head = uniquePush(head, src[i]);
c6871b76
KW
1332 scores[(i+1) * (y + 2) + 1] = i;
1333 scores[(i+1) * (y + 2) + 0] = score_ceil;
1334 swapCount = 0;
1335
1336 for (j=1;j<=y;j++) {
1337 if (i == 1) {
1338 if(j < y)
88f063b4 1339 head = uniquePush(head, tgt[j]);
c6871b76
KW
1340 scores[1 * (y + 2) + (j + 1)] = j;
1341 scores[0 * (y + 2) + (j + 1)] = score_ceil;
1342 }
1343
88f063b4 1344 targetCharCount = find(head, tgt[j-1])->value;
c6871b76
KW
1345 swapScore = scores[targetCharCount * (y + 2) + swapCount] + i - targetCharCount - 1 + j - swapCount;
1346
1347 if (src[i-1] != tgt[j-1]){
1348 scores[(i+1) * (y + 2) + (j + 1)] = MIN(swapScore,(MIN(scores[i * (y + 2) + j], MIN(scores[(i+1) * (y + 2) + j], scores[i * (y + 2) + (j + 1)])) + 1));
1349 }
1350 else {
1351 swapCount = j;
1352 scores[(i+1) * (y + 2) + (j + 1)] = MIN(scores[i * (y + 2) + j], swapScore);
1353 }
1354 }
1355
88f063b4 1356 find(head, src[i-1])->value = i;
c6871b76
KW
1357 }
1358
1359 {
1360 IV score = scores[(x+1) * (y + 2) + (y + 1)];
1361 dict_free(head);
1362 Safefree(scores);
1363 return (maxDistance != 0 && maxDistance < score)?(-1):score;
1364 }
1365}
1366
1367/* END of edit_distance() stuff
1368 * ========================================================= */
1369
8e35b056
KW
1370/* is c a control character for which we have a mnemonic? */
1371#define isMNEMONIC_CNTRL(c) _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
1372
549b4e78
KW
1373STATIC const char *
1374S_cntrl_to_mnemonic(const U8 c)
1375{
1376 /* Returns the mnemonic string that represents character 'c', if one
1377 * exists; NULL otherwise. The only ones that exist for the purposes of
1378 * this routine are a few control characters */
1379
1380 switch (c) {
1381 case '\a': return "\\a";
1382 case '\b': return "\\b";
1383 case ESC_NATIVE: return "\\e";
1384 case '\f': return "\\f";
1385 case '\n': return "\\n";
1386 case '\r': return "\\r";
1387 case '\t': return "\\t";
1388 }
1389
1390 return NULL;
1391}
1392
653099ff 1393/* Mark that we cannot extend a found fixed substring at this point.
37b6262f 1394 Update the longest found anchored substring or the longest found
653099ff
GS
1395 floating substrings if needed. */
1396
4327152a 1397STATIC void
ea3daa5d
FC
1398S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data,
1399 SSize_t *minlenp, int is_inf)
c277df42 1400{
e1ec3a88 1401 const STRLEN l = CHR_SVLEN(data->last_found);
2099df82 1402 SV * const longest_sv = data->substrs[data->cur_is_floating].str;
a2ca2017 1403 const STRLEN old_l = CHR_SVLEN(longest_sv);
1de06328 1404 GET_RE_DEBUG_FLAGS_DECL;
b81d288d 1405
7918f24d
NC
1406 PERL_ARGS_ASSERT_SCAN_COMMIT;
1407
c277df42 1408 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
2099df82 1409 const U8 i = data->cur_is_floating;
a2ca2017 1410 SvSetMagicSV(longest_sv, data->last_found);
37b6262f
DM
1411 data->substrs[i].min_offset = l ? data->last_start_min : data->pos_min;
1412
2099df82
DM
1413 if (!i) /* fixed */
1414 data->substrs[0].max_offset = data->substrs[0].min_offset;
1415 else { /* float */
1416 data->substrs[1].max_offset = (l
646e8787
DM
1417 ? data->last_start_max
1418 : (data->pos_delta > SSize_t_MAX - data->pos_min
ea3daa5d
FC
1419 ? SSize_t_MAX
1420 : data->pos_min + data->pos_delta));
1421 if (is_inf
2099df82
DM
1422 || (STRLEN)data->substrs[1].max_offset > (STRLEN)SSize_t_MAX)
1423 data->substrs[1].max_offset = SSize_t_MAX;
37b6262f 1424 }
11683ecb 1425
37b6262f
DM
1426 if (data->flags & SF_BEFORE_EOL)
1427 data->substrs[i].flags |= (data->flags & SF_BEFORE_EOL);
1428 else
1429 data->substrs[i].flags &= ~SF_BEFORE_EOL;
1430 data->substrs[i].minlenp = minlenp;
1431 data->substrs[i].lookbehind = 0;
c277df42 1432 }
37b6262f 1433
c277df42 1434 SvCUR_set(data->last_found, 0);
0eda9292 1435 {
a28509cc 1436 SV * const sv = data->last_found;
097eb12c
AL
1437 if (SvUTF8(sv) && SvMAGICAL(sv)) {
1438 MAGIC * const mg = mg_find(sv, PERL_MAGIC_utf8);
1439 if (mg)
1440 mg->mg_len = 0;
1441 }
0eda9292 1442 }
c277df42
IZ
1443 data->last_end = -1;
1444 data->flags &= ~SF_BEFORE_EOL;
f5a36d78 1445 DEBUG_STUDYDATA("commit", data, 0, is_inf);
c277df42
IZ
1446}
1447
cdd87c1d
KW
1448/* An SSC is just a regnode_charclass_posix with an extra field: the inversion
1449 * list that describes which code points it matches */
1450
653099ff 1451STATIC void
3420edd7 1452S_ssc_anything(pTHX_ regnode_ssc *ssc)
653099ff 1453{
cdd87c1d
KW
1454 /* Set the SSC 'ssc' to match an empty string or any code point */
1455
557bd3fb 1456 PERL_ARGS_ASSERT_SSC_ANYTHING;
7918f24d 1457
71068078 1458 assert(is_ANYOF_SYNTHETIC(ssc));
3fffb88a 1459
0854ea0b
KW
1460 /* mortalize so won't leak */
1461 ssc->invlist = sv_2mortal(_add_range_to_invlist(NULL, 0, UV_MAX));
93e92956 1462 ANYOF_FLAGS(ssc) |= SSC_MATCHES_EMPTY_STRING; /* Plus matches empty */
653099ff
GS
1463}
1464
653099ff 1465STATIC int
dc3bf405 1466S_ssc_is_anything(const regnode_ssc *ssc)
653099ff 1467{
c144baaa
KW
1468 /* Returns TRUE if the SSC 'ssc' can match the empty string and any code
1469 * point; FALSE otherwise. Thus, this is used to see if using 'ssc' buys
1470 * us anything: if the function returns TRUE, 'ssc' hasn't been restricted
1471 * in any way, so there's no point in using it */
cdd87c1d
KW
1472
1473 UV start, end;
1474 bool ret;
653099ff 1475
557bd3fb 1476 PERL_ARGS_ASSERT_SSC_IS_ANYTHING;
7918f24d 1477
71068078 1478 assert(is_ANYOF_SYNTHETIC(ssc));
cdd87c1d 1479
93e92956 1480 if (! (ANYOF_FLAGS(ssc) & SSC_MATCHES_EMPTY_STRING)) {
cdd87c1d
KW
1481 return FALSE;
1482 }
1483
1484 /* See if the list consists solely of the range 0 - Infinity */
1485 invlist_iterinit(ssc->invlist);
1486 ret = invlist_iternext(ssc->invlist, &start, &end)
1487 && start == 0
1488 && end == UV_MAX;
1489
1490 invlist_iterfinish(ssc->invlist);
1491
1492 if (ret) {
1493 return TRUE;
1494 }
1495
1496 /* If e.g., both \w and \W are set, matches everything */
e0e1be5f 1497 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
cdd87c1d
KW
1498 int i;
1499 for (i = 0; i < ANYOF_POSIXL_MAX; i += 2) {
1500 if (ANYOF_POSIXL_TEST(ssc, i) && ANYOF_POSIXL_TEST(ssc, i+1)) {
1501 return TRUE;
1502 }
1503 }
1504 }
1505
1506 return FALSE;
653099ff
GS
1507}
1508
653099ff 1509STATIC void
cdd87c1d 1510S_ssc_init(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc)
653099ff 1511{
cdd87c1d
KW
1512 /* Initializes the SSC 'ssc'. This includes setting it to match an empty
1513 * string, any code point, or any posix class under locale */
1514
557bd3fb 1515 PERL_ARGS_ASSERT_SSC_INIT;
7918f24d 1516
557bd3fb 1517 Zero(ssc, 1, regnode_ssc);
71068078 1518 set_ANYOF_SYNTHETIC(ssc);
93e92956 1519 ARG_SET(ssc, ANYOF_ONLY_HAS_BITMAP);
3420edd7 1520 ssc_anything(ssc);
cdd87c1d 1521
2f306ab9
KW
1522 /* If any portion of the regex is to operate under locale rules that aren't
1523 * fully known at compile time, initialization includes it. The reason
1524 * this isn't done for all regexes is that the optimizer was written under
1525 * the assumption that locale was all-or-nothing. Given the complexity and
1526 * lack of documentation in the optimizer, and that there are inadequate
1527 * test cases for locale, many parts of it may not work properly, it is
1528 * safest to avoid locale unless necessary. */
cdd87c1d
KW
1529 if (RExC_contains_locale) {
1530 ANYOF_POSIXL_SETALL(ssc);
cdd87c1d
KW
1531 }
1532 else {
1533 ANYOF_POSIXL_ZERO(ssc);
1534 }
653099ff
GS
1535}
1536
b423522f 1537STATIC int
dc3bf405
BF
1538S_ssc_is_cp_posixl_init(const RExC_state_t *pRExC_state,
1539 const regnode_ssc *ssc)
b423522f
KW
1540{
1541 /* Returns TRUE if the SSC 'ssc' is in its initial state with regard only
1542 * to the list of code points matched, and locale posix classes; hence does
1543 * not check its flags) */
1544
1545 UV start, end;
1546 bool ret;
1547
1548 PERL_ARGS_ASSERT_SSC_IS_CP_POSIXL_INIT;
1549
71068078 1550 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
1551
1552 invlist_iterinit(ssc->invlist);
1553 ret = invlist_iternext(ssc->invlist, &start, &end)
1554 && start == 0
1555 && end == UV_MAX;
1556
1557 invlist_iterfinish(ssc->invlist);
1558
1559 if (! ret) {
1560 return FALSE;
1561 }
1562
e0e1be5f 1563 if (RExC_contains_locale && ! ANYOF_POSIXL_SSC_TEST_ALL_SET(ssc)) {
31f05a37 1564 return FALSE;
b423522f
KW
1565 }
1566
1567 return TRUE;
1568}
1569
4ebed06a
KW
1570#define INVLIST_INDEX 0
1571#define ONLY_LOCALE_MATCHES_INDEX 1
1572#define DEFERRED_USER_DEFINED_INDEX 2
1573
b423522f
KW
1574STATIC SV*
1575S_get_ANYOF_cp_list_for_ssc(pTHX_ const RExC_state_t *pRExC_state,
5c0f85ef 1576 const regnode_charclass* const node)
b423522f
KW
1577{
1578 /* Returns a mortal inversion list defining which code points are matched
1579 * by 'node', which is of type ANYOF. Handles complementing the result if
1580 * appropriate. If some code points aren't knowable at this time, the
31f05a37
KW
1581 * returned list must, and will, contain every code point that is a
1582 * possibility. */
b423522f 1583
1565c085 1584 dVAR;
e2506fa7 1585 SV* invlist = NULL;
1ee208c4 1586 SV* only_utf8_locale_invlist = NULL;
b423522f
KW
1587 unsigned int i;
1588 const U32 n = ARG(node);
31f05a37 1589 bool new_node_has_latin1 = FALSE;
3146c00a 1590 const U8 flags = (inRANGE(OP(node), ANYOFH, ANYOFHr))
29a889ef
KW
1591 ? 0
1592 : ANYOF_FLAGS(node);
b423522f
KW
1593
1594 PERL_ARGS_ASSERT_GET_ANYOF_CP_LIST_FOR_SSC;
1595
1596 /* Look at the data structure created by S_set_ANYOF_arg() */
93e92956 1597 if (n != ANYOF_ONLY_HAS_BITMAP) {
b423522f
KW
1598 SV * const rv = MUTABLE_SV(RExC_rxi->data->data[n]);
1599 AV * const av = MUTABLE_AV(SvRV(rv));
1600 SV **const ary = AvARRAY(av);
1601 assert(RExC_rxi->data->what[n] == 's');
1602
4ebed06a 1603 if (av_tindex_skip_len_mg(av) >= DEFERRED_USER_DEFINED_INDEX) {
b423522f 1604
6ed02bb6
KW
1605 /* Here there are things that won't be known until runtime -- we
1606 * have to assume it could be anything */
e2506fa7 1607 invlist = sv_2mortal(_new_invlist(1));
b423522f
KW
1608 return _add_range_to_invlist(invlist, 0, UV_MAX);
1609 }
4ebed06a 1610 else if (ary[INVLIST_INDEX]) {
b423522f 1611
6ed02bb6 1612 /* Use the node's inversion list */
4ebed06a 1613 invlist = sv_2mortal(invlist_clone(ary[INVLIST_INDEX], NULL));
1ee208c4
KW
1614 }
1615
1616 /* Get the code points valid only under UTF-8 locales */
766d6d33 1617 if ( (flags & ANYOFL_FOLD)
4ebed06a 1618 && av_tindex_skip_len_mg(av) >= ONLY_LOCALE_MATCHES_INDEX)
1ee208c4 1619 {
4ebed06a 1620 only_utf8_locale_invlist = ary[ONLY_LOCALE_MATCHES_INDEX];
b423522f
KW
1621 }
1622 }
1623
e2506fa7
KW
1624 if (! invlist) {
1625 invlist = sv_2mortal(_new_invlist(0));
1626 }
1627
dcb20b36
KW
1628 /* An ANYOF node contains a bitmap for the first NUM_ANYOF_CODE_POINTS
1629 * code points, and an inversion list for the others, but if there are code
1630 * points that should match only conditionally on the target string being
1631 * UTF-8, those are placed in the inversion list, and not the bitmap.
1632 * Since there are circumstances under which they could match, they are
1633 * included in the SSC. But if the ANYOF node is to be inverted, we have
1634 * to exclude them here, so that when we invert below, the end result
1635 * actually does include them. (Think about "\xe0" =~ /[^\xc0]/di;). We
1636 * have to do this here before we add the unconditionally matched code
1637 * points */
766d6d33 1638 if (flags & ANYOF_INVERT) {
b423522f
KW
1639 _invlist_intersection_complement_2nd(invlist,
1640 PL_UpperLatin1,
1641 &invlist);
1642 }
1643
1644 /* Add in the points from the bit map */
3146c00a 1645 if (! inRANGE(OP(node), ANYOFH, ANYOFHr)) {
90973738
KW
1646 for (i = 0; i < NUM_ANYOF_CODE_POINTS; i++) {
1647 if (ANYOF_BITMAP_TEST(node, i)) {
1648 unsigned int start = i++;
1649
1650 for (; i < NUM_ANYOF_CODE_POINTS
1651 && ANYOF_BITMAP_TEST(node, i); ++i)
1652 {
1653 /* empty */
1654 }
1655 invlist = _add_range_to_invlist(invlist, start, i-1);
1656 new_node_has_latin1 = TRUE;
6f8848d5 1657 }
b423522f
KW
1658 }
1659 }
1660
1661 /* If this can match all upper Latin1 code points, have to add them
ac33c516
KW
1662 * as well. But don't add them if inverting, as when that gets done below,
1663 * it would exclude all these characters, including the ones it shouldn't
1664 * that were added just above */
766d6d33
KW
1665 if (! (flags & ANYOF_INVERT) && OP(node) == ANYOFD
1666 && (flags & ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER))
f240c685 1667 {
b423522f
KW
1668 _invlist_union(invlist, PL_UpperLatin1, &invlist);
1669 }
1670
1671 /* Similarly for these */
766d6d33 1672 if (flags & ANYOF_MATCHES_ALL_ABOVE_BITMAP) {
e0a1ff7a 1673 _invlist_union_complement_2nd(invlist, PL_InBitmap, &invlist);
b423522f
KW
1674 }
1675
766d6d33 1676 if (flags & ANYOF_INVERT) {
b423522f
KW
1677 _invlist_invert(invlist);
1678 }
766d6d33 1679 else if (flags & ANYOFL_FOLD) {
35b8412f 1680 if (new_node_has_latin1) {
31f05a37 1681
26be5fe6
KW
1682 /* Under /li, any 0-255 could fold to any other 0-255, depending on
1683 * the locale. We can skip this if there are no 0-255 at all. */
1684 _invlist_union(invlist, PL_Latin1, &invlist);
35b8412f
KW
1685
1686 invlist = add_cp_to_invlist(invlist, LATIN_SMALL_LETTER_DOTLESS_I);
1687 invlist = add_cp_to_invlist(invlist, LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE);
1688 }
1689 else {
1690 if (_invlist_contains_cp(invlist, LATIN_SMALL_LETTER_DOTLESS_I)) {
1691 invlist = add_cp_to_invlist(invlist, 'I');
1692 }
1693 if (_invlist_contains_cp(invlist,
1694 LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE))
1695 {
1696 invlist = add_cp_to_invlist(invlist, 'i');
1697 }
1698 }
31f05a37
KW
1699 }
1700
1ee208c4
KW
1701 /* Similarly add the UTF-8 locale possible matches. These have to be
1702 * deferred until after the non-UTF-8 locale ones are taken care of just
1703 * above, or it leads to wrong results under ANYOF_INVERT */
1704 if (only_utf8_locale_invlist) {
31f05a37 1705 _invlist_union_maybe_complement_2nd(invlist,
1ee208c4 1706 only_utf8_locale_invlist,
766d6d33 1707 flags & ANYOF_INVERT,
31f05a37
KW
1708 &invlist);
1709 }
b423522f
KW
1710
1711 return invlist;
1712}
1713
1051e1c4 1714/* These two functions currently do the exact same thing */
557bd3fb 1715#define ssc_init_zero ssc_init
653099ff 1716
cdd87c1d
KW
1717#define ssc_add_cp(ssc, cp) ssc_add_range((ssc), (cp), (cp))
1718#define ssc_match_all_cp(ssc) ssc_add_range(ssc, 0, UV_MAX)
1719
557bd3fb 1720/* 'AND' a given class with another one. Can create false positives. 'ssc'
93e92956
KW
1721 * should not be inverted. 'and_with->flags & ANYOF_MATCHES_POSIXL' should be
1722 * 0 if 'and_with' is a regnode_charclass instead of a regnode_ssc. */
cdd87c1d 1723
653099ff 1724STATIC void
b423522f 1725S_ssc_and(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc,
7dcac5f6 1726 const regnode_charclass *and_with)
653099ff 1727{
cdd87c1d
KW
1728 /* Accumulate into SSC 'ssc' its 'AND' with 'and_with', which is either
1729 * another SSC or a regular ANYOF class. Can create false positives. */
40d049e4 1730
a0dd4231 1731 SV* anded_cp_list;
3146c00a 1732 U8 and_with_flags = inRANGE(OP(and_with), ANYOFH, ANYOFHr)
29a889ef
KW
1733 ? 0
1734 : ANYOF_FLAGS(and_with);
a0dd4231 1735 U8 anded_flags;
1e6ade67 1736
cdd87c1d 1737 PERL_ARGS_ASSERT_SSC_AND;
653099ff 1738
71068078 1739 assert(is_ANYOF_SYNTHETIC(ssc));
a0dd4231
KW
1740
1741 /* 'and_with' is used as-is if it too is an SSC; otherwise have to extract
1742 * the code point inversion list and just the relevant flags */
71068078 1743 if (is_ANYOF_SYNTHETIC(and_with)) {
7dcac5f6 1744 anded_cp_list = ((regnode_ssc *)and_with)->invlist;
765e6ecf 1745 anded_flags = and_with_flags;
e9b08962
KW
1746
1747 /* XXX This is a kludge around what appears to be deficiencies in the
1748 * optimizer. If we make S_ssc_anything() add in the WARN_SUPER flag,
1749 * there are paths through the optimizer where it doesn't get weeded
1750 * out when it should. And if we don't make some extra provision for
1751 * it like the code just below, it doesn't get added when it should.
1752 * This solution is to add it only when AND'ing, which is here, and
1753 * only when what is being AND'ed is the pristine, original node
1754 * matching anything. Thus it is like adding it to ssc_anything() but
1755 * only when the result is to be AND'ed. Probably the same solution
1756 * could be adopted for the same problem we have with /l matching,
1757 * which is solved differently in S_ssc_init(), and that would lead to
1758 * fewer false positives than that solution has. But if this solution
1759 * creates bugs, the consequences are only that a warning isn't raised
1760 * that should be; while the consequences for having /l bugs is
1761 * incorrect matches */
7dcac5f6 1762 if (ssc_is_anything((regnode_ssc *)and_with)) {
f240c685 1763 anded_flags |= ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER;
e9b08962 1764 }
a0dd4231
KW
1765 }
1766 else {
5c0f85ef 1767 anded_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, and_with);
f240c685 1768 if (OP(and_with) == ANYOFD) {
765e6ecf 1769 anded_flags = and_with_flags & ANYOF_COMMON_FLAGS;
f240c685
KW
1770 }
1771 else {
765e6ecf 1772 anded_flags = and_with_flags
f240c685 1773 &( ANYOF_COMMON_FLAGS
108316fb
KW
1774 |ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
1775 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP);
765e6ecf 1776 if (ANYOFL_UTF8_LOCALE_REQD(and_with_flags)) {
d1c40ef5
KW
1777 anded_flags &=
1778 ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
1779 }
f240c685 1780 }
a0dd4231
KW
1781 }
1782
1783 ANYOF_FLAGS(ssc) &= anded_flags;
cdd87c1d
KW
1784
1785 /* Below, C1 is the list of code points in 'ssc'; P1, its posix classes.
1786 * C2 is the list of code points in 'and-with'; P2, its posix classes.
1787 * 'and_with' may be inverted. When not inverted, we have the situation of
1788 * computing:
1789 * (C1 | P1) & (C2 | P2)
1790 * = (C1 & (C2 | P2)) | (P1 & (C2 | P2))
1791 * = ((C1 & C2) | (C1 & P2)) | ((P1 & C2) | (P1 & P2))
1792 * <= ((C1 & C2) | P2)) | ( P1 | (P1 & P2))
1793 * <= ((C1 & C2) | P1 | P2)
1794 * Alternatively, the last few steps could be:
1795 * = ((C1 & C2) | (C1 & P2)) | ((P1 & C2) | (P1 & P2))
1796 * <= ((C1 & C2) | C1 ) | ( C2 | (P1 & P2))
1797 * <= (C1 | C2 | (P1 & P2))
1798 * We favor the second approach if either P1 or P2 is non-empty. This is
1799 * because these components are a barrier to doing optimizations, as what
1800 * they match cannot be known until the moment of matching as they are
1801 * dependent on the current locale, 'AND"ing them likely will reduce or
1802 * eliminate them.
1803 * But we can do better if we know that C1,P1 are in their initial state (a
1804 * frequent occurrence), each matching everything:
1805 * (<everything>) & (C2 | P2) = C2 | P2
1806 * Similarly, if C2,P2 are in their initial state (again a frequent
1807 * occurrence), the result is a no-op
1808 * (C1 | P1) & (<everything>) = C1 | P1
1809 *
1810 * Inverted, we have
1811 * (C1 | P1) & ~(C2 | P2) = (C1 | P1) & (~C2 & ~P2)
1812 * = (C1 & (~C2 & ~P2)) | (P1 & (~C2 & ~P2))
1813 * <= (C1 & ~C2) | (P1 & ~P2)
1814 * */
1aa99e6b 1815
765e6ecf 1816 if ((and_with_flags & ANYOF_INVERT)
71068078 1817 && ! is_ANYOF_SYNTHETIC(and_with))
a0dd4231 1818 {
cdd87c1d 1819 unsigned int i;
8951c461 1820
cdd87c1d
KW
1821 ssc_intersection(ssc,
1822 anded_cp_list,
1823 FALSE /* Has already been inverted */
1824 );
c6b76537 1825
cdd87c1d
KW
1826 /* If either P1 or P2 is empty, the intersection will be also; can skip
1827 * the loop */
765e6ecf 1828 if (! (and_with_flags & ANYOF_MATCHES_POSIXL)) {
cdd87c1d
KW
1829 ANYOF_POSIXL_ZERO(ssc);
1830 }
e0e1be5f 1831 else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
cdd87c1d
KW
1832
1833 /* Note that the Posix class component P from 'and_with' actually
1834 * looks like:
1835 * P = Pa | Pb | ... | Pn
1836 * where each component is one posix class, such as in [\w\s].
1837 * Thus
1838 * ~P = ~(Pa | Pb | ... | Pn)
1839 * = ~Pa & ~Pb & ... & ~Pn
1840 * <= ~Pa | ~Pb | ... | ~Pn
1841 * The last is something we can easily calculate, but unfortunately
1842 * is likely to have many false positives. We could do better
1843 * in some (but certainly not all) instances if two classes in
1844 * P have known relationships. For example
1845 * :lower: <= :alpha: <= :alnum: <= \w <= :graph: <= :print:
1846 * So
1847 * :lower: & :print: = :lower:
1848 * And similarly for classes that must be disjoint. For example,
1849 * since \s and \w can have no elements in common based on rules in
1850 * the POSIX standard,
1851 * \w & ^\S = nothing
1852 * Unfortunately, some vendor locales do not meet the Posix
1853 * standard, in particular almost everything by Microsoft.
1854 * The loop below just changes e.g., \w into \W and vice versa */
1855
1ee208c4 1856 regnode_charclass_posixl temp;
cdd87c1d
KW
1857 int add = 1; /* To calculate the index of the complement */
1858
b1234259 1859 Zero(&temp, 1, regnode_charclass_posixl);
cdd87c1d
KW
1860 ANYOF_POSIXL_ZERO(&temp);
1861 for (i = 0; i < ANYOF_MAX; i++) {
1862 assert(i % 2 != 0
7dcac5f6
KW
1863 || ! ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i)
1864 || ! ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i + 1));
cdd87c1d 1865
7dcac5f6 1866 if (ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i)) {
cdd87c1d
KW
1867 ANYOF_POSIXL_SET(&temp, i + add);
1868 }
1869 add = 0 - add; /* 1 goes to -1; -1 goes to 1 */
1870 }
1871 ANYOF_POSIXL_AND(&temp, ssc);
c6b76537 1872
cdd87c1d
KW
1873 } /* else ssc already has no posixes */
1874 } /* else: Not inverted. This routine is a no-op if 'and_with' is an SSC
1875 in its initial state */
71068078 1876 else if (! is_ANYOF_SYNTHETIC(and_with)
7dcac5f6 1877 || ! ssc_is_cp_posixl_init(pRExC_state, (regnode_ssc *)and_with))
cdd87c1d
KW
1878 {
1879 /* But if 'ssc' is in its initial state, the result is just 'and_with';
1880 * copy it over 'ssc' */
1881 if (ssc_is_cp_posixl_init(pRExC_state, ssc)) {
71068078 1882 if (is_ANYOF_SYNTHETIC(and_with)) {
cdd87c1d
KW
1883 StructCopy(and_with, ssc, regnode_ssc);
1884 }
1885 else {
1886 ssc->invlist = anded_cp_list;
1887 ANYOF_POSIXL_ZERO(ssc);
765e6ecf 1888 if (and_with_flags & ANYOF_MATCHES_POSIXL) {
7dcac5f6 1889 ANYOF_POSIXL_OR((regnode_charclass_posixl*) and_with, ssc);
cdd87c1d
KW
1890 }
1891 }
1892 }
e0e1be5f 1893 else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)
765e6ecf 1894 || (and_with_flags & ANYOF_MATCHES_POSIXL))
cdd87c1d
KW
1895 {
1896 /* One or the other of P1, P2 is non-empty. */
765e6ecf 1897 if (and_with_flags & ANYOF_MATCHES_POSIXL) {
1ea8b7fe
KW
1898 ANYOF_POSIXL_AND((regnode_charclass_posixl*) and_with, ssc);
1899 }
cdd87c1d
KW
1900 ssc_union(ssc, anded_cp_list, FALSE);
1901 }
1902 else { /* P1 = P2 = empty */
1903 ssc_intersection(ssc, anded_cp_list, FALSE);
1904 }
137165a6 1905 }
653099ff
GS
1906}
1907
653099ff 1908STATIC void
cdd87c1d 1909S_ssc_or(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc,
7dcac5f6 1910 const regnode_charclass *or_with)
653099ff 1911{
cdd87c1d
KW
1912 /* Accumulate into SSC 'ssc' its 'OR' with 'or_with', which is either
1913 * another SSC or a regular ANYOF class. Can create false positives if
1914 * 'or_with' is to be inverted. */
7918f24d 1915
a0dd4231
KW
1916 SV* ored_cp_list;
1917 U8 ored_flags;
3146c00a 1918 U8 or_with_flags = inRANGE(OP(or_with), ANYOFH, ANYOFHr)
29a889ef
KW
1919 ? 0
1920 : ANYOF_FLAGS(or_with);
c6b76537 1921
cdd87c1d 1922 PERL_ARGS_ASSERT_SSC_OR;
c6b76537 1923
71068078 1924 assert(is_ANYOF_SYNTHETIC(ssc));
a0dd4231
KW
1925
1926 /* 'or_with' is used as-is if it too is an SSC; otherwise have to extract
1927 * the code point inversion list and just the relevant flags */
71068078 1928 if (is_ANYOF_SYNTHETIC(or_with)) {
7dcac5f6 1929 ored_cp_list = ((regnode_ssc*) or_with)->invlist;
765e6ecf 1930 ored_flags = or_with_flags;
a0dd4231
KW
1931 }
1932 else {
5c0f85ef 1933 ored_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, or_with);
765e6ecf 1934 ored_flags = or_with_flags & ANYOF_COMMON_FLAGS;
f240c685
KW
1935 if (OP(or_with) != ANYOFD) {
1936 ored_flags
765e6ecf 1937 |= or_with_flags
108316fb
KW
1938 & ( ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
1939 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP);
765e6ecf 1940 if (ANYOFL_UTF8_LOCALE_REQD(or_with_flags)) {
d1c40ef5
KW
1941 ored_flags |=
1942 ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
1943 }
f240c685 1944 }
a0dd4231
KW
1945 }
1946
1947 ANYOF_FLAGS(ssc) |= ored_flags;
cdd87c1d
KW
1948
1949 /* Below, C1 is the list of code points in 'ssc'; P1, its posix classes.
1950 * C2 is the list of code points in 'or-with'; P2, its posix classes.
1951 * 'or_with' may be inverted. When not inverted, we have the simple
1952 * situation of computing:
1953 * (C1 | P1) | (C2 | P2) = (C1 | C2) | (P1 | P2)
1954 * If P1|P2 yields a situation with both a class and its complement are
1955 * set, like having both \w and \W, this matches all code points, and we
1956 * can delete these from the P component of the ssc going forward. XXX We
1957 * might be able to delete all the P components, but I (khw) am not certain
1958 * about this, and it is better to be safe.
1959 *
1960 * Inverted, we have
1961 * (C1 | P1) | ~(C2 | P2) = (C1 | P1) | (~C2 & ~P2)
1962 * <= (C1 | P1) | ~C2
1963 * <= (C1 | ~C2) | P1
1964 * (which results in actually simpler code than the non-inverted case)
1965 * */
9826f543 1966
765e6ecf 1967 if ((or_with_flags & ANYOF_INVERT)
71068078 1968 && ! is_ANYOF_SYNTHETIC(or_with))
a0dd4231 1969 {
cdd87c1d 1970 /* We ignore P2, leaving P1 going forward */
1ea8b7fe 1971 } /* else Not inverted */
765e6ecf 1972 else if (or_with_flags & ANYOF_MATCHES_POSIXL) {
7dcac5f6 1973 ANYOF_POSIXL_OR((regnode_charclass_posixl*)or_with, ssc);
e0e1be5f 1974 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
cdd87c1d
KW
1975 unsigned int i;
1976 for (i = 0; i < ANYOF_MAX; i += 2) {
1977 if (ANYOF_POSIXL_TEST(ssc, i) && ANYOF_POSIXL_TEST(ssc, i + 1))
1978 {
1979 ssc_match_all_cp(ssc);
1980 ANYOF_POSIXL_CLEAR(ssc, i);
1981 ANYOF_POSIXL_CLEAR(ssc, i+1);
cdd87c1d
KW
1982 }
1983 }
1984 }
1aa99e6b 1985 }
cdd87c1d
KW
1986
1987 ssc_union(ssc,
1988 ored_cp_list,
1989 FALSE /* Already has been inverted */
1990 );
653099ff
GS
1991}
1992
b423522f
KW
1993PERL_STATIC_INLINE void
1994S_ssc_union(pTHX_ regnode_ssc *ssc, SV* const invlist, const bool invert2nd)
1995{
1996 PERL_ARGS_ASSERT_SSC_UNION;
1997
71068078 1998 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
1999
2000 _invlist_union_maybe_complement_2nd(ssc->invlist,
2001 invlist,
2002 invert2nd,
2003 &ssc->invlist);
2004}
2005
2006PERL_STATIC_INLINE void
2007S_ssc_intersection(pTHX_ regnode_ssc *ssc,
2008 SV* const invlist,
2009 const bool invert2nd)
2010{
2011 PERL_ARGS_ASSERT_SSC_INTERSECTION;
2012
71068078 2013 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
2014
2015 _invlist_intersection_maybe_complement_2nd(ssc->invlist,
2016 invlist,
2017 invert2nd,
2018 &ssc->invlist);
2019}
2020
2021PERL_STATIC_INLINE void
2022S_ssc_add_range(pTHX_ regnode_ssc *ssc, const UV start, const UV end)
2023{
2024 PERL_ARGS_ASSERT_SSC_ADD_RANGE;
2025
71068078 2026 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
2027
2028 ssc->invlist = _add_range_to_invlist(ssc->invlist, start, end);
2029}
2030
2031PERL_STATIC_INLINE void
2032S_ssc_cp_and(pTHX_ regnode_ssc *ssc, const UV cp)
2033{
2034 /* AND just the single code point 'cp' into the SSC 'ssc' */
2035
2036 SV* cp_list = _new_invlist(2);
2037
2038 PERL_ARGS_ASSERT_SSC_CP_AND;
2039
71068078 2040 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
2041
2042 cp_list = add_cp_to_invlist(cp_list, cp);
2043 ssc_intersection(ssc, cp_list,
2044 FALSE /* Not inverted */
2045 );
2046 SvREFCNT_dec_NN(cp_list);
2047}
2048
2049PERL_STATIC_INLINE void
dc3bf405 2050S_ssc_clear_locale(regnode_ssc *ssc)
b423522f
KW
2051{
2052 /* Set the SSC 'ssc' to not match any locale things */
b423522f
KW
2053 PERL_ARGS_ASSERT_SSC_CLEAR_LOCALE;
2054
71068078 2055 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
2056
2057 ANYOF_POSIXL_ZERO(ssc);
2058 ANYOF_FLAGS(ssc) &= ~ANYOF_LOCALE_FLAGS;
2059}
2060
b35552de
KW
2061#define NON_OTHER_COUNT NON_OTHER_COUNT_FOR_USE_ONLY_BY_REGCOMP_DOT_C
2062
2063STATIC bool
2064S_is_ssc_worth_it(const RExC_state_t * pRExC_state, const regnode_ssc * ssc)
2065{
2066 /* The synthetic start class is used to hopefully quickly winnow down
2067 * places where a pattern could start a match in the target string. If it
2068 * doesn't really narrow things down that much, there isn't much point to
2069 * having the overhead of using it. This function uses some very crude
2070 * heuristics to decide if to use the ssc or not.
2071 *
2072 * It returns TRUE if 'ssc' rules out more than half what it considers to
2073 * be the "likely" possible matches, but of course it doesn't know what the
2074 * actual things being matched are going to be; these are only guesses
2075 *
2076 * For /l matches, it assumes that the only likely matches are going to be
2077 * in the 0-255 range, uniformly distributed, so half of that is 127
2078 * For /a and /d matches, it assumes that the likely matches will be just
2079 * the ASCII range, so half of that is 63
2080 * For /u and there isn't anything matching above the Latin1 range, it
2081 * assumes that that is the only range likely to be matched, and uses
2082 * half that as the cut-off: 127. If anything matches above Latin1,
2083 * it assumes that all of Unicode could match (uniformly), except for
2084 * non-Unicode code points and things in the General Category "Other"
2085 * (unassigned, private use, surrogates, controls and formats). This
2086 * is a much large number. */
2087
b35552de
KW
2088 U32 count = 0; /* Running total of number of code points matched by
2089 'ssc' */
2090 UV start, end; /* Start and end points of current range in inversion
26be5fe6 2091 XXX outdated. UTF-8 locales are common, what about invert? list */
72400949
KW
2092 const U32 max_code_points = (LOC)
2093 ? 256
f6e8f31e 2094 : (( ! UNI_SEMANTICS
1a26cbcb 2095 || invlist_highest(ssc->invlist) < 256)
72400949
KW
2096 ? 128
2097 : NON_OTHER_COUNT);
2098 const U32 max_match = max_code_points / 2;
b35552de
KW
2099
2100 PERL_ARGS_ASSERT_IS_SSC_WORTH_IT;
2101
2102 invlist_iterinit(ssc->invlist);
2103 while (invlist_iternext(ssc->invlist, &start, &end)) {
72400949
KW
2104 if (start >= max_code_points) {
2105 break;
b35552de 2106 }
72400949 2107 end = MIN(end, max_code_points - 1);
b35552de 2108 count += end - start + 1;
72400949 2109 if (count >= max_match) {
b35552de
KW
2110 invlist_iterfinish(ssc->invlist);
2111 return FALSE;
2112 }
2113 }
2114
2115 return TRUE;
2116}
2117
2118
b423522f
KW
2119STATIC void
2120S_ssc_finalize(pTHX_ RExC_state_t *pRExC_state, regnode_ssc *ssc)
2121{
2122 /* The inversion list in the SSC is marked mortal; now we need a more
2123 * permanent copy, which is stored the same way that is done in a regular
dcb20b36
KW
2124 * ANYOF node, with the first NUM_ANYOF_CODE_POINTS code points in a bit
2125 * map */
b423522f 2126
28118b9c 2127 SV* invlist = invlist_clone(ssc->invlist, NULL);
b423522f
KW
2128
2129 PERL_ARGS_ASSERT_SSC_FINALIZE;
2130
71068078 2131 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f 2132
a0dd4231 2133 /* The code in this file assumes that all but these flags aren't relevant
93e92956
KW
2134 * to the SSC, except SSC_MATCHES_EMPTY_STRING, which should be cleared
2135 * by the time we reach here */
f240c685
KW
2136 assert(! (ANYOF_FLAGS(ssc)
2137 & ~( ANYOF_COMMON_FLAGS
108316fb
KW
2138 |ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
2139 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP)));
a0dd4231 2140
b423522f
KW
2141 populate_ANYOF_from_invlist( (regnode *) ssc, &invlist);
2142
4c404f26 2143 set_ANYOF_arg(pRExC_state, (regnode *) ssc, invlist, NULL, NULL);
b423522f 2144
85c8e306
KW
2145 /* Make sure is clone-safe */
2146 ssc->invlist = NULL;
2147
e0e1be5f 2148 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
93e92956 2149 ANYOF_FLAGS(ssc) |= ANYOF_MATCHES_POSIXL;
d156f5cb 2150 OP(ssc) = ANYOFPOSIXL;
e0e1be5f 2151 }
d156f5cb 2152 else if (RExC_contains_locale) {
b2e90ddf
KW
2153 OP(ssc) = ANYOFL;
2154 }
2155
1462525b 2156 assert(! (ANYOF_FLAGS(ssc) & ANYOF_LOCALE_FLAGS) || RExC_contains_locale);
b423522f
KW
2157}
2158
a3621e74
YO
2159#define TRIE_LIST_ITEM(state,idx) (trie->states[state].trans.list)[ idx ]
2160#define TRIE_LIST_CUR(state) ( TRIE_LIST_ITEM( state, 0 ).forid )
2161#define TRIE_LIST_LEN(state) ( TRIE_LIST_ITEM( state, 0 ).newstate )
538e84ed
KW
2162#define TRIE_LIST_USED(idx) ( trie->states[state].trans.list \
2163 ? (TRIE_LIST_CUR( idx ) - 1) \
2164 : 0 )
a3621e74 2165
3dab1dad
YO
2166
2167#ifdef DEBUGGING
07be1b83 2168/*
2b8b4781
NC
2169 dump_trie(trie,widecharmap,revcharmap)
2170 dump_trie_interim_list(trie,widecharmap,revcharmap,next_alloc)
2171 dump_trie_interim_table(trie,widecharmap,revcharmap,next_alloc)
3dab1dad
YO
2172
2173 These routines dump out a trie in a somewhat readable format.
07be1b83
YO
2174 The _interim_ variants are used for debugging the interim
2175 tables that are used to generate the final compressed
2176 representation which is what dump_trie expects.
2177
486ec47a 2178 Part of the reason for their existence is to provide a form
3dab1dad 2179 of documentation as to how the different representations function.
07be1b83
YO
2180
2181*/
3dab1dad
YO
2182
2183/*
3dab1dad
YO
2184 Dumps the final compressed table form of the trie to Perl_debug_log.
2185 Used for debugging make_trie().
2186*/
b9a59e08 2187
3dab1dad 2188STATIC void
2b8b4781
NC
2189S_dump_trie(pTHX_ const struct _reg_trie_data *trie, HV *widecharmap,
2190 AV *revcharmap, U32 depth)
3dab1dad
YO
2191{
2192 U32 state;
ab3bbdeb 2193 SV *sv=sv_newmortal();
55eed653 2194 int colwidth= widecharmap ? 6 : 4;
2e64971a 2195 U16 word;
3dab1dad
YO
2196 GET_RE_DEBUG_FLAGS_DECL;
2197
7918f24d 2198 PERL_ARGS_ASSERT_DUMP_TRIE;
ab3bbdeb 2199
6ad9a8ab 2200 Perl_re_indentf( aTHX_ "Char : %-6s%-6s%-4s ",
1e37780e 2201 depth+1, "Match","Base","Ofs" );
3dab1dad
YO
2202
2203 for( state = 0 ; state < trie->uniquecharcount ; state++ ) {
2b8b4781 2204 SV ** const tmp = av_fetch( revcharmap, state, 0);
3dab1dad 2205 if ( tmp ) {
6ad9a8ab 2206 Perl_re_printf( aTHX_ "%*s",
ab3bbdeb 2207 colwidth,
538e84ed 2208 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
ab3bbdeb
YO
2209 PL_colors[0], PL_colors[1],
2210 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
538e84ed
KW
2211 PERL_PV_ESCAPE_FIRSTCHAR
2212 )
ab3bbdeb 2213 );
3dab1dad
YO
2214 }
2215 }
1e37780e
YO
2216 Perl_re_printf( aTHX_ "\n");
2217 Perl_re_indentf( aTHX_ "State|-----------------------", depth+1);
3dab1dad
YO
2218
2219 for( state = 0 ; state < trie->uniquecharcount ; state++ )
6ad9a8ab
YO
2220 Perl_re_printf( aTHX_ "%.*s", colwidth, "--------");
2221 Perl_re_printf( aTHX_ "\n");
3dab1dad 2222
1e2e3d02 2223 for( state = 1 ; state < trie->statecount ; state++ ) {
be8e71aa 2224 const U32 base = trie->states[ state ].trans.base;
3dab1dad 2225
147e3846 2226 Perl_re_indentf( aTHX_ "#%4" UVXf "|", depth+1, (UV)state);
3dab1dad
YO
2227
2228 if ( trie->states[ state ].wordnum ) {
1e37780e 2229 Perl_re_printf( aTHX_ " W%4X", trie->states[ state ].wordnum );
3dab1dad 2230 } else {
6ad9a8ab 2231 Perl_re_printf( aTHX_ "%6s", "" );
3dab1dad
YO
2232 }
2233
147e3846 2234 Perl_re_printf( aTHX_ " @%4" UVXf " ", (UV)base );
3dab1dad
YO
2235
2236 if ( base ) {
2237 U32 ofs = 0;
2238
2239 while( ( base + ofs < trie->uniquecharcount ) ||
2240 ( base + ofs - trie->uniquecharcount < trie->lasttrans
538e84ed
KW
2241 && trie->trans[ base + ofs - trie->uniquecharcount ].check
2242 != state))
3dab1dad
YO
2243 ofs++;
2244
147e3846 2245 Perl_re_printf( aTHX_ "+%2" UVXf "[ ", (UV)ofs);
3dab1dad
YO
2246
2247 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
538e84ed
KW
2248 if ( ( base + ofs >= trie->uniquecharcount )
2249 && ( base + ofs - trie->uniquecharcount
2250 < trie->lasttrans )
2251 && trie->trans[ base + ofs
2252 - trie->uniquecharcount ].check == state )
3dab1dad 2253 {
147e3846 2254 Perl_re_printf( aTHX_ "%*" UVXf, colwidth,
1e37780e
YO
2255 (UV)trie->trans[ base + ofs - trie->uniquecharcount ].next
2256 );
3dab1dad 2257 } else {
88f063b4 2258 Perl_re_printf( aTHX_ "%*s", colwidth," ." );
3dab1dad
YO
2259 }
2260 }
2261
6ad9a8ab 2262 Perl_re_printf( aTHX_ "]");
3dab1dad
YO
2263
2264 }
6ad9a8ab 2265 Perl_re_printf( aTHX_ "\n" );
3dab1dad 2266 }
6ad9a8ab 2267 Perl_re_indentf( aTHX_ "word_info N:(prev,len)=",
cb41e5d6 2268 depth);
2e64971a 2269 for (word=1; word <= trie->wordcount; word++) {
6ad9a8ab 2270 Perl_re_printf( aTHX_ " %d:(%d,%d)",
2e64971a
DM
2271 (int)word, (int)(trie->wordinfo[word].prev),
2272 (int)(trie->wordinfo[word].len));
2273 }
6ad9a8ab 2274 Perl_re_printf( aTHX_ "\n" );
538e84ed 2275}
3dab1dad 2276/*
3dab1dad 2277 Dumps a fully constructed but uncompressed trie in list form.
538e84ed 2278 List tries normally only are used for construction when the number of
3dab1dad
YO
2279 possible chars (trie->uniquecharcount) is very high.
2280 Used for debugging make_trie().
2281*/
2282STATIC void
55eed653 2283S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie,
2b8b4781
NC
2284 HV *widecharmap, AV *revcharmap, U32 next_alloc,
2285 U32 depth)
3dab1dad
YO
2286{
2287 U32 state;
ab3bbdeb 2288 SV *sv=sv_newmortal();
55eed653 2289 int colwidth= widecharmap ? 6 : 4;
3dab1dad 2290 GET_RE_DEBUG_FLAGS_DECL;
7918f24d
NC
2291
2292 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_LIST;
2293
3dab1dad 2294 /* print out the table precompression. */
6ad9a8ab 2295 Perl_re_indentf( aTHX_ "State :Word | Transition Data\n",
cb41e5d6 2296 depth+1 );
6ad9a8ab 2297 Perl_re_indentf( aTHX_ "%s",
cb41e5d6 2298 depth+1, "------:-----+-----------------\n" );
538e84ed 2299
3dab1dad
YO
2300 for( state=1 ; state < next_alloc ; state ++ ) {
2301 U16 charid;
538e84ed 2302
147e3846 2303 Perl_re_indentf( aTHX_ " %4" UVXf " :",
cb41e5d6 2304 depth+1, (UV)state );
3dab1dad 2305 if ( ! trie->states[ state ].wordnum ) {
6ad9a8ab 2306 Perl_re_printf( aTHX_ "%5s| ","");
3dab1dad 2307 } else {
6ad9a8ab 2308 Perl_re_printf( aTHX_ "W%4x| ",
3dab1dad
YO
2309 trie->states[ state ].wordnum
2310 );
2311 }
2312 for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) {
538e84ed 2313 SV ** const tmp = av_fetch( revcharmap,
88f063b4 2314 TRIE_LIST_ITEM(state, charid).forid, 0);
ab3bbdeb 2315 if ( tmp ) {
147e3846 2316 Perl_re_printf( aTHX_ "%*s:%3X=%4" UVXf " | ",
ab3bbdeb 2317 colwidth,
538e84ed
KW
2318 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp),
2319 colwidth,
2320 PL_colors[0], PL_colors[1],
2321 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
2322 | PERL_PV_ESCAPE_FIRSTCHAR
ab3bbdeb 2323 ) ,
88f063b4
KW
2324 TRIE_LIST_ITEM(state, charid).forid,
2325 (UV)TRIE_LIST_ITEM(state, charid).newstate
1e2e3d02 2326 );
538e84ed 2327 if (!(charid % 10))
6ad9a8ab 2328 Perl_re_printf( aTHX_ "\n%*s| ",
664e119d 2329 (int)((depth * 2) + 14), "");
1e2e3d02 2330 }
ab3bbdeb 2331 }
6ad9a8ab 2332 Perl_re_printf( aTHX_ "\n");
3dab1dad 2333 }
538e84ed 2334}
3dab1dad
YO
2335
2336/*
3dab1dad 2337 Dumps a fully constructed but uncompressed trie in table form.
538e84ed
KW
2338 This is the normal DFA style state transition table, with a few
2339 twists to facilitate compression later.
3dab1dad
YO
2340 Used for debugging make_trie().
2341*/
2342STATIC void
55eed653 2343S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie,
2b8b4781
NC
2344 HV *widecharmap, AV *revcharmap, U32 next_alloc,
2345 U32 depth)
3dab1dad
YO
2346{
2347 U32 state;
2348 U16 charid;
ab3bbdeb 2349 SV *sv=sv_newmortal();
55eed653 2350 int colwidth= widecharmap ? 6 : 4;
3dab1dad 2351 GET_RE_DEBUG_FLAGS_DECL;
7918f24d
NC
2352
2353 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_TABLE;
538e84ed 2354
3dab1dad
YO
2355 /*
2356 print out the table precompression so that we can do a visual check
2357 that they are identical.
2358 */
538e84ed 2359
6ad9a8ab 2360 Perl_re_indentf( aTHX_ "Char : ", depth+1 );
3dab1dad
YO
2361
2362 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
2b8b4781 2363 SV ** const tmp = av_fetch( revcharmap, charid, 0);
3dab1dad 2364 if ( tmp ) {
6ad9a8ab 2365 Perl_re_printf( aTHX_ "%*s",
ab3bbdeb 2366 colwidth,
538e84ed 2367 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
ab3bbdeb
YO
2368 PL_colors[0], PL_colors[1],
2369 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
538e84ed
KW
2370 PERL_PV_ESCAPE_FIRSTCHAR
2371 )
ab3bbdeb 2372 );
3dab1dad
YO
2373 }
2374 }
2375
4aaafc03
YO
2376 Perl_re_printf( aTHX_ "\n");
2377 Perl_re_indentf( aTHX_ "State+-", depth+1 );
3dab1dad
YO
2378
2379 for( charid=0 ; charid < trie->uniquecharcount ; charid++ ) {
6ad9a8ab 2380 Perl_re_printf( aTHX_ "%.*s", colwidth,"--------");
3dab1dad
YO
2381 }
2382
6ad9a8ab 2383 Perl_re_printf( aTHX_ "\n" );
3dab1dad
YO
2384
2385 for( state=1 ; state < next_alloc ; state += trie->uniquecharcount ) {
2386
147e3846 2387 Perl_re_indentf( aTHX_ "%4" UVXf " : ",
cb41e5d6 2388 depth+1,
3dab1dad
YO
2389 (UV)TRIE_NODENUM( state ) );
2390
2391 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
ab3bbdeb
YO
2392 UV v=(UV)SAFE_TRIE_NODENUM( trie->trans[ state + charid ].next );
2393 if (v)
147e3846 2394 Perl_re_printf( aTHX_ "%*" UVXf, colwidth, v );
ab3bbdeb 2395 else
6ad9a8ab 2396 Perl_re_printf( aTHX_ "%*s", colwidth, "." );
3dab1dad
YO
2397 }
2398 if ( ! trie->states[ TRIE_NODENUM( state ) ].wordnum ) {
147e3846 2399 Perl_re_printf( aTHX_ " (%4" UVXf ")\n",
538e84ed 2400 (UV)trie->trans[ state ].check );
3dab1dad 2401 } else {
147e3846 2402 Perl_re_printf( aTHX_ " (%4" UVXf ") W%4X\n",
538e84ed 2403 (UV)trie->trans[ state ].check,
3dab1dad
YO
2404 trie->states[ TRIE_NODENUM( state ) ].wordnum );
2405 }
2406 }
07be1b83 2407}
3dab1dad
YO
2408
2409#endif
2410
2e64971a 2411
786e8c11
YO
2412/* make_trie(startbranch,first,last,tail,word_count,flags,depth)
2413 startbranch: the first branch in the whole branch sequence
2414 first : start branch of sequence of branch-exact nodes.
2415 May be the same as startbranch
2416 last : Thing following the last branch.
2417 May be the same as tail.
2418 tail : item following the branch sequence
2419 count : words in the sequence
a4525e78 2420 flags : currently the OP() type we will be building one of /EXACT(|F|FA|FU|FU_SS|L|FLU8)/
786e8c11 2421 depth : indent depth
3dab1dad 2422
786e8c11 2423Inplace optimizes a sequence of 2 or more Branch-Exact nodes into a TRIE node.
07be1b83 2424
786e8c11
YO
2425A trie is an N'ary tree where the branches are determined by digital
2426decomposition of the key. IE, at the root node you look up the 1st character and
2427follow that branch repeat until you find the end of the branches. Nodes can be
2428marked as "accepting" meaning they represent a complete word. Eg:
07be1b83 2429
786e8c11 2430 /he|she|his|hers/
72f13be8 2431
786e8c11
YO
2432would convert into the following structure. Numbers represent states, letters
2433following numbers represent valid transitions on the letter from that state, if
2434the number is in square brackets it represents an accepting state, otherwise it
2435will be in parenthesis.
07be1b83 2436
786e8c11
YO
2437 +-h->+-e->[3]-+-r->(8)-+-s->[9]
2438 | |
2439 | (2)
2440 | |
2441 (1) +-i->(6)-+-s->[7]
2442 |
2443 +-s->(3)-+-h->(4)-+-e->[5]
07be1b83 2444
786e8c11
YO
2445 Accept Word Mapping: 3=>1 (he),5=>2 (she), 7=>3 (his), 9=>4 (hers)
2446
2447This shows that when matching against the string 'hers' we will begin at state 1
2448read 'h' and move to state 2, read 'e' and move to state 3 which is accepting,
2449then read 'r' and go to state 8 followed by 's' which takes us to state 9 which
2450is also accepting. Thus we know that we can match both 'he' and 'hers' with a
2451single traverse. We store a mapping from accepting to state to which word was
2452matched, and then when we have multiple possibilities we try to complete the
b8fda935 2453rest of the regex in the order in which they occurred in the alternation.
786e8c11
YO
2454
2455The only prior NFA like behaviour that would be changed by the TRIE support is
2456the silent ignoring of duplicate alternations which are of the form:
2457
2458 / (DUPE|DUPE) X? (?{ ... }) Y /x
2459
4b714af6 2460Thus EVAL blocks following a trie may be called a different number of times with
786e8c11 2461and without the optimisation. With the optimisations dupes will be silently
486ec47a 2462ignored. This inconsistent behaviour of EVAL type nodes is well established as
786e8c11
YO
2463the following demonstrates:
2464
2465 'words'=~/(word|word|word)(?{ print $1 })[xyz]/
2466
2467which prints out 'word' three times, but
2468
2469 'words'=~/(word|word|word)(?{ print $1 })S/
2470
2471which doesnt print it out at all. This is due to other optimisations kicking in.
2472
2473Example of what happens on a structural level:
2474
486ec47a 2475The regexp /(ac|ad|ab)+/ will produce the following debug output:
786e8c11
YO
2476
2477 1: CURLYM[1] {1,32767}(18)
2478 5: BRANCH(8)
2479 6: EXACT <ac>(16)
2480 8: BRANCH(11)
2481 9: EXACT <ad>(16)
2482 11: BRANCH(14)
2483 12: EXACT <ab>(16)
2484 16: SUCCEED(0)
2485 17: NOTHING(18)
2486 18: END(0)
2487
2488This would be optimizable with startbranch=5, first=5, last=16, tail=16
2489and should turn into:
2490
2491 1: CURLYM[1] {1,32767}(18)
2492 5: TRIE(16)
2493 [Words:3 Chars Stored:6 Unique Chars:4 States:5 NCP:1]
2494 <ac>
2495 <ad>
2496 <ab>
2497 16: SUCCEED(0)
2498 17: NOTHING(18)
2499 18: END(0)
2500
2501Cases where tail != last would be like /(?foo|bar)baz/:
2502
2503 1: BRANCH(4)
2504 2: EXACT <foo>(8)
2505 4: BRANCH(7)
2506 5: EXACT <bar>(8)
2507 7: TAIL(8)
2508 8: EXACT <baz>(10)
2509 10: END(0)
2510
2511which would be optimizable with startbranch=1, first=1, last=7, tail=8
2512and would end up looking like:
2513
2514 1: TRIE(8)
2515 [Words:2 Chars Stored:6 Unique Chars:5 States:7 NCP:1]
2516 <foo>
2517 <bar>
2518 7: TAIL(8)
2519 8: EXACT <baz>(10)
2520 10: END(0)
2521
c80e42f3 2522 d = uvchr_to_utf8_flags(d, uv, 0);
786e8c11
YO
2523
2524is the recommended Unicode-aware way of saying
2525
2526 *(d++) = uv;
2527*/
2528
fab2782b 2529#define TRIE_STORE_REVCHAR(val) \
786e8c11 2530 STMT_START { \
73031816 2531 if (UTF) { \
668fcfea 2532 SV *zlopp = newSV(UTF8_MAXBYTES); \
88c9ea1e 2533 unsigned char *flrbbbbb = (unsigned char *) SvPVX(zlopp); \
1d84a256
DM
2534 unsigned char *const kapow = uvchr_to_utf8(flrbbbbb, val); \
2535 *kapow = '\0'; \
73031816
NC
2536 SvCUR_set(zlopp, kapow - flrbbbbb); \
2537 SvPOK_on(zlopp); \
2538 SvUTF8_on(zlopp); \
2539 av_push(revcharmap, zlopp); \
2540 } else { \
fab2782b 2541 char ooooff = (char)val; \
73031816
NC
2542 av_push(revcharmap, newSVpvn(&ooooff, 1)); \
2543 } \
2544 } STMT_END
786e8c11 2545
914a25d5
KW
2546/* This gets the next character from the input, folding it if not already
2547 * folded. */
2548#define TRIE_READ_CHAR STMT_START { \
2549 wordlen++; \
2550 if ( UTF ) { \
2551 /* if it is UTF then it is either already folded, or does not need \
2552 * folding */ \
1c1d615a 2553 uvc = valid_utf8_to_uvchr( (const U8*) uc, &len); \
914a25d5
KW
2554 } \
2555 else if (folder == PL_fold_latin1) { \
7d006b13
KW
2556 /* This folder implies Unicode rules, which in the range expressible \
2557 * by not UTF is the lower case, with the two exceptions, one of \
2558 * which should have been taken care of before calling this */ \
2559 assert(*uc != LATIN_SMALL_LETTER_SHARP_S); \
2560 uvc = toLOWER_L1(*uc); \
2561 if (UNLIKELY(uvc == MICRO_SIGN)) uvc = GREEK_SMALL_LETTER_MU; \
2562 len = 1; \
914a25d5
KW
2563 } else { \
2564 /* raw data, will be folded later if needed */ \
2565 uvc = (U32)*uc; \
2566 len = 1; \
2567 } \
786e8c11
YO
2568} STMT_END
2569
2570
2571
2572#define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \
2573 if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \
00195859 2574 U32 ging = TRIE_LIST_LEN( state ) * 2; \
f9003953 2575 Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \
00195859 2576 TRIE_LIST_LEN( state ) = ging; \
786e8c11
YO
2577 } \
2578 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \
2579 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \
2580 TRIE_LIST_CUR( state )++; \
2581} STMT_END
07be1b83 2582
786e8c11 2583#define TRIE_LIST_NEW(state) STMT_START { \
d09f14bf 2584 Newx( trie->states[ state ].trans.list, \
786e8c11
YO
2585 4, reg_trie_trans_le ); \
2586 TRIE_LIST_CUR( state ) = 1; \
2587 TRIE_LIST_LEN( state ) = 4; \
2588} STMT_END
07be1b83 2589
786e8c11
YO
2590#define TRIE_HANDLE_WORD(state) STMT_START { \
2591 U16 dupe= trie->states[ state ].wordnum; \
2592 regnode * const noper_next = regnext( noper ); \
2593 \
786e8c11
YO
2594 DEBUG_r({ \
2595 /* store the word for dumping */ \
2596 SV* tmp; \
2597 if (OP(noper) != NOTHING) \
740cce10 2598 tmp = newSVpvn_utf8(STRING(noper), STR_LEN(noper), UTF); \
786e8c11 2599 else \
740cce10 2600 tmp = newSVpvn_utf8( "", 0, UTF ); \
2b8b4781 2601 av_push( trie_words, tmp ); \
786e8c11
YO
2602 }); \
2603 \
2604 curword++; \
2e64971a
DM
2605 trie->wordinfo[curword].prev = 0; \
2606 trie->wordinfo[curword].len = wordlen; \
2607 trie->wordinfo[curword].accept = state; \
786e8c11
YO
2608 \
2609 if ( noper_next < tail ) { \
2610 if (!trie->jump) \
538e84ed
KW
2611 trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, \
2612 sizeof(U16) ); \
7f69552c 2613 trie->jump[curword] = (U16)(noper_next - convert); \
786e8c11
YO
2614 if (!jumper) \
2615 jumper = noper_next; \
2616 if (!nextbranch) \
2617 nextbranch= regnext(cur); \
2618 } \
2619 \
2620 if ( dupe ) { \
2e64971a
DM
2621 /* It's a dupe. Pre-insert into the wordinfo[].prev */\
2622 /* chain, so that when the bits of chain are later */\
2623 /* linked together, the dups appear in the chain */\
2624 trie->wordinfo[curword].prev = trie->wordinfo[dupe].prev; \
2625 trie->wordinfo[dupe].prev = curword; \
786e8c11
YO
2626 } else { \
2627 /* we haven't inserted this word yet. */ \
2628 trie->states[ state ].wordnum = curword; \
2629 } \
2630} STMT_END
07be1b83 2631
3dab1dad 2632
786e8c11
YO
2633#define TRIE_TRANS_STATE(state,base,ucharcount,charid,special) \
2634 ( ( base + charid >= ucharcount \
2635 && base + charid < ubound \
2636 && state == trie->trans[ base - ucharcount + charid ].check \
2637 && trie->trans[ base - ucharcount + charid ].next ) \
2638 ? trie->trans[ base - ucharcount + charid ].next \
2639 : ( state==1 ? special : 0 ) \
2640 )
3dab1dad 2641
8bcafbf4
YO
2642#define TRIE_BITMAP_SET_FOLDED(trie, uvc, folder) \
2643STMT_START { \
2644 TRIE_BITMAP_SET(trie, uvc); \
2645 /* store the folded codepoint */ \
2646 if ( folder ) \
2647 TRIE_BITMAP_SET(trie, folder[(U8) uvc ]); \
2648 \
2649 if ( !UTF ) { \
2650 /* store first byte of utf8 representation of */ \
2651 /* variant codepoints */ \
2652 if (! UVCHR_IS_INVARIANT(uvc)) { \
2653 TRIE_BITMAP_SET(trie, UTF8_TWO_BYTE_HI(uvc)); \
2654 } \
2655 } \
2656} STMT_END
786e8c11
YO
2657#define MADE_TRIE 1
2658#define MADE_JUMP_TRIE 2
2659#define MADE_EXACT_TRIE 4
3dab1dad 2660
a3621e74 2661STATIC I32
538e84ed
KW
2662S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch,
2663 regnode *first, regnode *last, regnode *tail,
2664 U32 word_count, U32 flags, U32 depth)
a3621e74
YO
2665{
2666 /* first pass, loop through and scan words */
2667 reg_trie_data *trie;
55eed653 2668 HV *widecharmap = NULL;
2b8b4781 2669 AV *revcharmap = newAV();
a3621e74 2670 regnode *cur;
a3621e74
YO
2671 STRLEN len = 0;
2672 UV uvc = 0;
2673 U16 curword = 0;
2674 U32 next_alloc = 0;
786e8c11
YO
2675 regnode *jumper = NULL;
2676 regnode *nextbranch = NULL;
7f69552c 2677 regnode *convert = NULL;
2e64971a 2678 U32 *prev_states; /* temp array mapping each state to previous one */
a3621e74 2679 /* we just use folder as a flag in utf8 */
1e696034 2680 const U8 * folder = NULL;
a3621e74 2681
3a611511
YO
2682 /* in the below add_data call we are storing either 'tu' or 'tuaa'
2683 * which stands for one trie structure, one hash, optionally followed
2684 * by two arrays */
2b8b4781 2685#ifdef DEBUGGING
3a611511 2686 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("tuaa"));
2b8b4781
NC
2687 AV *trie_words = NULL;
2688 /* along with revcharmap, this only used during construction but both are
2689 * useful during debugging so we store them in the struct when debugging.
8e11feef 2690 */
2b8b4781 2691#else
cf78de0b 2692 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("tu"));
3dab1dad 2693 STRLEN trie_charcount=0;
3dab1dad 2694#endif
2b8b4781 2695 SV *re_trie_maxbuff;
a3621e74 2696 GET_RE_DEBUG_FLAGS_DECL;
7918f24d
NC
2697
2698 PERL_ARGS_ASSERT_MAKE_TRIE;
72f13be8
YO
2699#ifndef DEBUGGING
2700 PERL_UNUSED_ARG(depth);
2701#endif
a3621e74 2702
1e696034 2703 switch (flags) {
f6b4b99d 2704 case EXACT: case EXACT_ONLY8: case EXACTL: break;
89829bb5 2705 case EXACTFAA:
0ea669f4 2706 case EXACTFUP:
a4525e78
KW
2707 case EXACTFU:
2708 case EXACTFLU8: folder = PL_fold_latin1; break;
1e696034 2709 case EXACTF: folder = PL_fold; break;
fab2782b 2710 default: Perl_croak( aTHX_ "panic! In trie construction, unknown node type %u %s", (unsigned) flags, PL_reg_name[flags] );
1e696034
KW
2711 }
2712
c944940b 2713 trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
a3621e74 2714 trie->refcount = 1;
3dab1dad 2715 trie->startstate = 1;
786e8c11 2716 trie->wordcount = word_count;
f8fc2ecf 2717 RExC_rxi->data->data[ data_slot ] = (void*)trie;
c944940b 2718 trie->charmap = (U16 *) PerlMemShared_calloc( 256, sizeof(U16) );
f6b4b99d 2719 if (flags == EXACT || flags == EXACT_ONLY8 || flags == EXACTL)
c944940b 2720 trie->bitmap = (char *) PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 );
2e64971a
DM
2721 trie->wordinfo = (reg_trie_wordinfo *) PerlMemShared_calloc(
2722 trie->wordcount+1, sizeof(reg_trie_wordinfo));
2723
a3621e74 2724 DEBUG_r({
2b8b4781 2725 trie_words = newAV();
a3621e74 2726 });
a3621e74 2727
3b89859a 2728 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, GV_ADD);
316ebaf2 2729 assert(re_trie_maxbuff);
a3621e74 2730 if (!SvIOK(re_trie_maxbuff)) {
0111c4fd 2731 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
a3621e74 2732 }
df826430 2733 DEBUG_TRIE_COMPILE_r({
6ad9a8ab 2734 Perl_re_indentf( aTHX_
cb41e5d6
YO
2735 "make_trie start==%d, first==%d, last==%d, tail==%d depth=%d\n",
2736 depth+1,
88f063b4 2737 REG_NODE_NUM(startbranch), REG_NODE_NUM(first),
538e84ed 2738 REG_NODE_NUM(last), REG_NODE_NUM(tail), (int)depth);
3dab1dad 2739 });
538e84ed 2740
7f69552c
YO
2741 /* Find the node we are going to overwrite */
2742 if ( first == startbranch && OP( last ) != BRANCH ) {
2743 /* whole branch chain */
2744 convert = first;
2745 } else {
2746 /* branch sub-chain */
2747 convert = NEXTOPER( first );
2748 }
538e84ed 2749
a3621e74
YO
2750 /* -- First loop and Setup --
2751
2752 We first traverse the branches and scan each word to determine if it
2753 contains widechars, and how many unique chars there are, this is
2754 important as we have to build a table with at least as many columns as we
2755 have unique chars.
2756
2757 We use an array of integers to represent the character codes 0..255
538e84ed
KW
2758 (trie->charmap) and we use a an HV* to store Unicode characters. We use
2759 the native representation of the character value as the key and IV's for
2760 the coded index.
a3621e74
YO
2761
2762 *TODO* If we keep track of how many times each character is used we can
2763 remap the columns so that the table compression later on is more
3b753521 2764 efficient in terms of memory by ensuring the most common value is in the
a3621e74
YO
2765 middle and the least common are on the outside. IMO this would be better
2766 than a most to least common mapping as theres a decent chance the most
2767 common letter will share a node with the least common, meaning the node
486ec47a 2768 will not be compressible. With a middle is most common approach the worst
a3621e74
YO
2769 case is when we have the least common nodes twice.
2770
2771 */
2772
a3621e74 2773 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
df826430 2774 regnode *noper = NEXTOPER( cur );
944e05e3
YO
2775 const U8 *uc;
2776 const U8 *e;
bc031a7d 2777 int foldlen = 0;
07be1b83 2778 U32 wordlen = 0; /* required init */
bc031a7d
KW
2779 STRLEN minchars = 0;
2780 STRLEN maxchars = 0;
538e84ed
KW
2781 bool set_bit = trie->bitmap ? 1 : 0; /*store the first char in the
2782 bitmap?*/
a3621e74 2783
3dab1dad 2784 if (OP(noper) == NOTHING) {
20ed8c88
YO
2785 /* skip past a NOTHING at the start of an alternation
2786 * eg, /(?:)a|(?:b)/ should be the same as /a|b/
2787 */
df826430 2788 regnode *noper_next= regnext(noper);
944e05e3
YO
2789 if (noper_next < tail)
2790 noper= noper_next;
2791 }
2792
f6b4b99d
KW
2793 if ( noper < tail
2794 && ( OP(noper) == flags
2795 || (flags == EXACT && OP(noper) == EXACT_ONLY8)
a9f8c7ac 2796 || (flags == EXACTFU && ( OP(noper) == EXACTFU_ONLY8
0ea669f4 2797 || OP(noper) == EXACTFUP))))
f6b4b99d 2798 {
944e05e3
YO
2799 uc= (U8*)STRING(noper);
2800 e= uc + STR_LEN(noper);
2801 } else {
2802 trie->minlen= 0;
2803 continue;
3dab1dad 2804 }
df826430 2805
944e05e3 2806
fab2782b 2807 if ( set_bit ) { /* bitmap only alloced when !(UTF&&Folding) */
02daf0ab
YO
2808 TRIE_BITMAP_SET(trie,*uc); /* store the raw first byte
2809 regardless of encoding */
0ea669f4 2810 if (OP( noper ) == EXACTFUP) {
fab2782b 2811 /* false positives are ok, so just set this */
0dc4a61d 2812 TRIE_BITMAP_SET(trie, LATIN_SMALL_LETTER_SHARP_S);
fab2782b
YO
2813 }
2814 }
dca5fc4c 2815
bc031a7d
KW
2816 for ( ; uc < e ; uc += len ) { /* Look at each char in the current
2817 branch */
3dab1dad 2818 TRIE_CHARCOUNT(trie)++;
a3621e74 2819 TRIE_READ_CHAR;
645de4ce 2820
bc031a7d
KW
2821 /* TRIE_READ_CHAR returns the current character, or its fold if /i
2822 * is in effect. Under /i, this character can match itself, or
2823 * anything that folds to it. If not under /i, it can match just
2824 * itself. Most folds are 1-1, for example k, K, and KELVIN SIGN
2825 * all fold to k, and all are single characters. But some folds
2826 * expand to more than one character, so for example LATIN SMALL
2827 * LIGATURE FFI folds to the three character sequence 'ffi'. If
2828 * the string beginning at 'uc' is 'ffi', it could be matched by
2829 * three characters, or just by the one ligature character. (It
2830 * could also be matched by two characters: LATIN SMALL LIGATURE FF
2831 * followed by 'i', or by 'f' followed by LATIN SMALL LIGATURE FI).
2832 * (Of course 'I' and/or 'F' instead of 'i' and 'f' can also
2833 * match.) The trie needs to know the minimum and maximum number
2834 * of characters that could match so that it can use size alone to
2835 * quickly reject many match attempts. The max is simple: it is
2836 * the number of folded characters in this branch (since a fold is
2837 * never shorter than what folds to it. */
2838
2839 maxchars++;
2840
2841 /* And the min is equal to the max if not under /i (indicated by
2842 * 'folder' being NULL), or there are no multi-character folds. If
2843 * there is a multi-character fold, the min is incremented just
2844 * once, for the character that folds to the sequence. Each
2845 * character in the sequence needs to be added to the list below of
2846 * characters in the trie, but we count only the first towards the
2847 * min number of characters needed. This is done through the
2848 * variable 'foldlen', which is returned by the macros that look
2849 * for these sequences as the number of bytes the sequence
2850 * occupies. Each time through the loop, we decrement 'foldlen' by
2851 * how many bytes the current char occupies. Only when it reaches
2852 * 0 do we increment 'minchars' or look for another multi-character
2853 * sequence. */
2854 if (folder == NULL) {
2855 minchars++;
2856 }
2857 else if (foldlen > 0) {
2858 foldlen -= (UTF) ? UTF8SKIP(uc) : 1;
645de4ce
KW
2859 }
2860 else {
bc031a7d
KW
2861 minchars++;
2862
2863 /* See if *uc is the beginning of a multi-character fold. If
2864 * so, we decrement the length remaining to look at, to account
2865 * for the current character this iteration. (We can use 'uc'
2866 * instead of the fold returned by TRIE_READ_CHAR because for
2867 * non-UTF, the latin1_safe macro is smart enough to account
2868 * for all the unfolded characters, and because for UTF, the
2869 * string will already have been folded earlier in the
2870 * compilation process */
2871 if (UTF) {
2872 if ((foldlen = is_MULTI_CHAR_FOLD_utf8_safe(uc, e))) {
2873 foldlen -= UTF8SKIP(uc);
645de4ce
KW
2874 }
2875 }
bc031a7d
KW
2876 else if ((foldlen = is_MULTI_CHAR_FOLD_latin1_safe(uc, e))) {
2877 foldlen--;
2878 }
645de4ce 2879 }
bc031a7d
KW
2880
2881 /* The current character (and any potential folds) should be added
2882 * to the possible matching characters for this position in this
2883 * branch */
a3621e74 2884 if ( uvc < 256 ) {
fab2782b
YO
2885 if ( folder ) {
2886 U8 folded= folder[ (U8) uvc ];
2887 if ( !trie->charmap[ folded ] ) {
2888 trie->charmap[ folded ]=( ++trie->uniquecharcount );
2889 TRIE_STORE_REVCHAR( folded );
2890 }
2891 }
a3621e74
YO
2892 if ( !trie->charmap[ uvc ] ) {
2893 trie->charmap[ uvc ]=( ++trie->uniquecharcount );
fab2782b 2894 TRIE_STORE_REVCHAR( uvc );
a3621e74 2895 }
02daf0ab 2896 if ( set_bit ) {
62012aee
KW
2897 /* store the codepoint in the bitmap, and its folded
2898 * equivalent. */
8bcafbf4 2899 TRIE_BITMAP_SET_FOLDED(trie, uvc, folder);
02daf0ab
YO
2900 set_bit = 0; /* We've done our bit :-) */
2901 }
a3621e74 2902 } else {
bc031a7d
KW
2903
2904 /* XXX We could come up with the list of code points that fold
2905 * to this using PL_utf8_foldclosures, except not for
2906 * multi-char folds, as there may be multiple combinations
2907 * there that could work, which needs to wait until runtime to
2908 * resolve (The comment about LIGATURE FFI above is such an
2909 * example */
2910
a3621e74 2911 SV** svpp;
55eed653
NC
2912 if ( !widecharmap )
2913 widecharmap = newHV();
a3621e74 2914
55eed653 2915 svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 1 );
a3621e74
YO
2916
2917 if ( !svpp )
147e3846 2918 Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%" UVXf, uvc );
a3621e74
YO
2919
2920 if ( !SvTRUE( *svpp ) ) {
2921 sv_setiv( *svpp, ++trie->uniquecharcount );
fab2782b 2922 TRIE_STORE_REVCHAR(uvc);
a3621e74
YO
2923 }
2924 }
bc031a7d
KW
2925 } /* end loop through characters in this branch of the trie */
2926
2927 /* We take the min and max for this branch and combine to find the min
2928 * and max for all branches processed so far */
3dab1dad 2929 if( cur == first ) {
bc031a7d
KW
2930 trie->minlen = minchars;
2931 trie->maxlen = maxchars;
2932 } else if (minchars < trie->minlen) {
2933 trie->minlen = minchars;
2934 } else if (maxchars > trie->maxlen) {
2935 trie->maxlen = maxchars;
fab2782b 2936 }
a3621e74
YO
2937 } /* end first pass */
2938 DEBUG_TRIE_COMPILE_r(
6ad9a8ab 2939 Perl_re_indentf( aTHX_
cb41e5d6
YO
2940 "TRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n",
2941 depth+1,
55eed653 2942 ( widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
be8e71aa
YO
2943 (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount,
2944 (int)trie->minlen, (int)trie->maxlen )
a3621e74 2945 );
a3621e74
YO
2946
2947 /*
2948 We now know what we are dealing with in terms of unique chars and
2949 string sizes so we can calculate how much memory a naive
0111c4fd
RGS
2950 representation using a flat table will take. If it's over a reasonable
2951 limit (as specified by ${^RE_TRIE_MAXBUF}) we use a more memory
a3621e74
YO
2952 conservative but potentially much slower representation using an array
2953 of lists.
2954
2955 At the end we convert both representations into the same compressed
2956 form that will be used in regexec.c for matching with. The latter
2957 is a form that cannot be used to construct with but has memory
2958 properties similar to the list form and access properties similar
2959 to the table form making it both suitable for fast searches and
2960 small enough that its feasable to store for the duration of a program.
2961
2962 See the comment in the code where the compressed table is produced
2963 inplace from the flat tabe representation for an explanation of how
2964 the compression works.
2965
2966 */
2967
2968
2e64971a
DM
2969 Newx(prev_states, TRIE_CHARCOUNT(trie) + 2, U32);
2970 prev_states[1] = 0;
2971
538e84ed
KW
2972 if ( (IV)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1)
2973 > SvIV(re_trie_maxbuff) )
2974 {
a3621e74
YO
2975 /*
2976 Second Pass -- Array Of Lists Representation
2977
2978 Each state will be represented by a list of charid:state records
2979 (reg_trie_trans_le) the first such element holds the CUR and LEN
2980 points of the allocated array. (See defines above).
2981
2982 We build the initial structure using the lists, and then convert
2983 it into the compressed table form which allows faster lookups
2984 (but cant be modified once converted).
a3621e74
YO
2985 */
2986
a3621e74
YO
2987 STRLEN transcount = 1;
2988
6ad9a8ab 2989 DEBUG_TRIE_COMPILE_MORE_r( Perl_re_indentf( aTHX_ "Compiling trie using list compiler\n",
cb41e5d6 2990 depth+1));
686b73d4 2991
c944940b
JH
2992 trie->states = (reg_trie_state *)
2993 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
2994 sizeof(reg_trie_state) );
a3621e74
YO
2995 TRIE_LIST_NEW(1);
2996 next_alloc = 2;
2997
2998 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
2999
df826430 3000 regnode *noper = NEXTOPER( cur );
c445ea15
AL
3001 U32 state = 1; /* required init */
3002 U16 charid = 0; /* sanity init */
07be1b83 3003 U32 wordlen = 0; /* required init */
c445ea15 3004
df826430
YO
3005 if (OP(noper) == NOTHING) {
3006 regnode *noper_next= regnext(noper);
944e05e3
YO
3007 if (noper_next < tail)
3008 noper= noper_next;
df826430
YO
3009 }
3010
f6b4b99d
KW
3011 if ( noper < tail
3012 && ( OP(noper) == flags
3013 || (flags == EXACT && OP(noper) == EXACT_ONLY8)
a9f8c7ac 3014 || (flags == EXACTFU && ( OP(noper) == EXACTFU_ONLY8
0ea669f4 3015 || OP(noper) == EXACTFUP))))
f6b4b99d 3016 {
944e05e3
YO
3017 const U8 *uc= (U8*)STRING(noper);
3018 const U8 *e= uc + STR_LEN(noper);
3019
786e8c11 3020 for ( ; uc < e ; uc += len ) {
c445ea15 3021
786e8c11 3022 TRIE_READ_CHAR;
c445ea15 3023
786e8c11
YO
3024 if ( uvc < 256 ) {
3025 charid = trie->charmap[ uvc ];
c445ea15 3026 } else {
538e84ed
KW
3027 SV** const svpp = hv_fetch( widecharmap,
3028 (char*)&uvc,
3029 sizeof( UV ),
3030 0);
786e8c11
YO
3031 if ( !svpp ) {
3032 charid = 0;
3033 } else {
3034 charid=(U16)SvIV( *svpp );
3035 }
c445ea15 3036 }
538e84ed
KW
3037 /* charid is now 0 if we dont know the char read, or
3038 * nonzero if we do */
786e8c11 3039 if ( charid ) {
a3621e74 3040
786e8c11
YO
3041 U16 check;
3042 U32 newstate = 0;
a3621e74 3043
786e8c11
YO
3044 charid--;
3045 if ( !trie->states[ state ].trans.list ) {
3046 TRIE_LIST_NEW( state );
c445ea15 3047 }
538e84ed
KW
3048 for ( check = 1;
3049 check <= TRIE_LIST_USED( state );
3050 check++ )
3051 {
3052 if ( TRIE_LIST_ITEM( state, check ).forid
3053 == charid )
3054 {
786e8c11
YO
3055 newstate = TRIE_LIST_ITEM( state, check ).newstate;
3056 break;
3057 }
3058 }
3059 if ( ! newstate ) {
3060 newstate = next_alloc++;
2e64971a 3061 prev_states[newstate] = state;
786e8c11
YO
3062 TRIE_LIST_PUSH( state, charid, newstate );
3063 transcount++;
3064 }
3065 state = newstate;
3066 } else {
147e3846 3067 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %" IVdf, uvc );
c445ea15 3068 }
a28509cc 3069 }
c445ea15 3070 }
3dab1dad 3071 TRIE_HANDLE_WORD(state);
a3621e74
YO
3072
3073 } /* end second pass */
3074
1e2e3d02 3075 /* next alloc is the NEXT state to be allocated */
538e84ed 3076 trie->statecount = next_alloc;
c944940b
JH
3077 trie->states = (reg_trie_state *)
3078 PerlMemShared_realloc( trie->states,
3079 next_alloc
3080 * sizeof(reg_trie_state) );
a3621e74 3081
3dab1dad 3082 /* and now dump it out before we compress it */
2b8b4781
NC
3083 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_list(trie, widecharmap,
3084 revcharmap, next_alloc,
3085 depth+1)
1e2e3d02 3086 );
a3621e74 3087
c944940b
JH
3088 trie->trans = (reg_trie_trans *)
3089 PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) );
a3621e74
YO
3090 {
3091 U32 state;
a3621e74
YO
3092 U32 tp = 0;
3093 U32 zp = 0;
3094
3095
3096 for( state=1 ; state < next_alloc ; state ++ ) {
3097 U32 base=0;
3098
3099 /*
3100 DEBUG_TRIE_COMPILE_MORE_r(
6ad9a8ab 3101 Perl_re_printf( aTHX_ "tp: %d zp: %d ",tp,zp)
a3621e74
YO
3102 );
3103 */
3104
3105 if (trie->states[state].trans.list) {
3106 U16 minid=TRIE_LIST_ITEM( state, 1).forid;
3107 U16 maxid=minid;
a28509cc 3108 U16 idx;
a3621e74
YO
3109
3110 for( idx = 2 ; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
c445ea15
AL
3111 const U16 forid = TRIE_LIST_ITEM( state, idx).forid;
3112 if ( forid < minid ) {
3113 minid=forid;
3114 } else if ( forid > maxid ) {
3115 maxid=forid;
3116 }
a3621e74
YO
3117 }
3118 if ( transcount < tp + maxid - minid + 1) {
3119 transcount *= 2;
c944940b
JH
3120 trie->trans = (reg_trie_trans *)
3121 PerlMemShared_realloc( trie->trans,
446bd890
NC
3122 transcount
3123 * sizeof(reg_trie_trans) );
538e84ed
KW
3124 Zero( trie->trans + (transcount / 2),
3125 transcount / 2,
3126 reg_trie_trans );
a3621e74
YO
3127 }
3128 base = trie->uniquecharcount + tp - minid;
3129 if ( maxid == minid ) {
3130 U32 set = 0;
3131 for ( ; zp < tp ; zp++ ) {
3132 if ( ! trie->trans[ zp ].next ) {
3133 base = trie->uniquecharcount + zp - minid;
538e84ed
KW
3134 trie->trans[ zp ].next = TRIE_LIST_ITEM( state,
3135 1).newstate;
a3621e74
YO
3136 trie->trans[ zp ].check = state;
3137 set = 1;
3138 break;
3139 }
3140 }
3141 if ( !set ) {
538e84ed
KW
3142 trie->trans[ tp ].next = TRIE_LIST_ITEM( state,
3143 1).newstate;
a3621e74
YO
3144 trie->trans[ tp ].check = state;
3145 tp++;
3146 zp = tp;
3147 }
3148 } else {
3149 for ( idx=1; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
538e84ed
KW
3150 const U32 tid = base
3151 - trie->uniquecharcount
3152 + TRIE_LIST_ITEM( state, idx ).forid;
3153 trie->trans[ tid ].next = TRIE_LIST_ITEM( state,
3154 idx ).newstate;
a3621e74
YO
3155 trie->trans[ tid ].check = state;
3156 }
3157 tp += ( maxid - minid + 1 );
3158 }
3159 Safefree(trie->states[ state ].trans.list);
3160 }
3161 /*
3162 DEBUG_TRIE_COMPILE_MORE_r(
6ad9a8ab 3163 Perl_re_printf( aTHX_ " base: %d\n",base);
a3621e74
YO
3164 );
3165 */
3166 trie->states[ state ].trans.base=base;
3167 }
cc601c31 3168 trie->lasttrans = tp + 1;
a3621e74
YO
3169 }
3170 } else {
3171 /*
3172 Second Pass -- Flat Table Representation.
3173
b423522f
KW
3174 we dont use the 0 slot of either trans[] or states[] so we add 1 to
3175 each. We know that we will need Charcount+1 trans at most to store
3176 the data (one row per char at worst case) So we preallocate both
3177 structures assuming worst case.
a3621e74
YO
3178
3179 We then construct the trie using only the .next slots of the entry
3180 structs.
3181
b423522f
KW
3182 We use the .check field of the first entry of the node temporarily
3183 to make compression both faster and easier by keeping track of how
3184 many non zero fields are in the node.
a3621e74
YO
3185
3186 Since trans are numbered from 1 any 0 pointer in the table is a FAIL
3187 transition.
3188
b423522f
KW
3189 There are two terms at use here: state as a TRIE_NODEIDX() which is
3190 a number representing the first entry of the node, and state as a
3191 TRIE_NODENUM() which is the trans number. state 1 is TRIE_NODEIDX(1)
3192 and TRIE_NODENUM(1), state 2 is TRIE_NODEIDX(2) and TRIE_NODENUM(3)
3193 if there are 2 entrys per node. eg:
a3621e74
YO
3194
3195 A B A B
3196 1. 2 4 1. 3 7
3197 2. 0 3 3. 0 5
3198 3. 0 0 5. 0 0
3199 4. 0 0 7. 0 0
3200
b423522f
KW
3201 The table is internally in the right hand, idx form. However as we
3202 also have to deal with the states array which is indexed by nodenum
3203 we have to use TRIE_NODENUM() to convert.
a3621e74
YO
3204
3205 */
6ad9a8ab 3206 DEBUG_TRIE_COMPILE_MORE_r( Perl_re_indentf( aTHX_ "Compiling trie using table compiler\n",
cb41e5d6 3207 depth+1));
3dab1dad 3208
c944940b
JH
3209 trie->trans = (reg_trie_trans *)
3210 PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 )
3211 * trie->uniquecharcount + 1,
3212 sizeof(reg_trie_trans) );
3213 trie->states = (reg_trie_state *)
3214 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
3215 sizeof(reg_trie_state) );
a3621e74
YO
3216 next_alloc = trie->uniquecharcount + 1;
3217
3dab1dad 3218
a3621e74
YO
3219 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
3220
df826430 3221 regnode *noper = NEXTOPER( cur );
a3621e74
YO
3222
3223 U32 state = 1; /* required init */
3224
3225 U16 charid = 0; /* sanity init */
3226 U32 accept_state = 0; /* sanity init */
a3621e74 3227
07be1b83 3228 U32 wordlen = 0; /* required init */
a3621e74 3229
df826430
YO
3230 if (OP(noper) == NOTHING) {
3231 regnode *noper_next= regnext(noper);
944e05e3
YO
3232 if (noper_next < tail)
3233 noper= noper_next;
df826430 3234 }
fab2782b 3235
f6b4b99d
KW
3236 if ( noper < tail
3237 && ( OP(noper) == flags
3238 || (flags == EXACT && OP(noper) == EXACT_ONLY8)
a9f8c7ac 3239 || (flags == EXACTFU && ( OP(noper) == EXACTFU_ONLY8
0ea669f4 3240 || OP(noper) == EXACTFUP))))
f6b4b99d 3241 {
944e05e3
YO
3242 const U8 *uc= (U8*)STRING(noper);
3243 const U8 *e= uc + STR_LEN(noper);
3244
786e8c11 3245 for ( ; uc < e ; uc += len ) {
a3621e74 3246
786e8c11 3247 TRIE_READ_CHAR;
a3621e74 3248
786e8c11
YO
3249 if ( uvc < 256 ) {
3250 charid = trie->charmap[ uvc ];
3251 } else {
538e84ed
KW
3252 SV* const * const svpp = hv_fetch( widecharmap,
3253 (char*)&uvc,
3254 sizeof( UV ),
3255 0);
786e8c11 3256 charid = svpp ? (U16)SvIV(*svpp) : 0;
a3621e74 3257 }
786e8c11
YO
3258 if ( charid ) {
3259 charid--;
3260 if ( !trie->trans[ state + charid ].next ) {
3261 trie->trans[ state + charid ].next = next_alloc;
3262 trie->trans[ state ].check++;
2e64971a
DM
3263 prev_states[TRIE_NODENUM(next_alloc)]
3264 = TRIE_NODENUM(state);
786e8c11
YO
3265 next_alloc += trie->uniquecharcount;
3266 }
3267 state = trie->trans[ state + charid ].next;
3268 } else {
147e3846 3269 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %" IVdf, uvc );
786e8c11 3270 }
538e84ed
KW
3271 /* charid is now 0 if we dont know the char read, or
3272 * nonzero if we do */
a3621e74 3273 }
a3621e74 3274 }
3dab1dad
YO
3275 accept_state = TRIE_NODENUM( state );
3276 TRIE_HANDLE_WORD(accept_state);
a3621e74
YO
3277
3278 } /* end second pass */
3279
3dab1dad 3280 /* and now dump it out before we compress it */
2b8b4781
NC
3281 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_table(trie, widecharmap,
3282 revcharmap,
3283 next_alloc, depth+1));
a3621e74 3284
a3621e74
YO
3285 {
3286 /*
3287 * Inplace compress the table.*
3288
3289 For sparse data sets the table constructed by the trie algorithm will
3290 be mostly 0/FAIL transitions or to put it another way mostly empty.
3291 (Note that leaf nodes will not contain any transitions.)
3292
3293 This algorithm compresses the tables by eliminating most such
3294 transitions, at the cost of a modest bit of extra work during lookup:
3295
3296 - Each states[] entry contains a .base field which indicates the
3297 index in the state[] array wheres its transition data is stored.
3298
3b753521 3299 - If .base is 0 there are no valid transitions from that node.
a3621e74
YO
3300
3301 - If .base is nonzero then charid is added to it to find an entry in
3302 the trans array.
3303
3304 -If trans[states[state].base+charid].check!=state then the
3305 transition is taken to be a 0/Fail transition. Thus if there are fail
3306 transitions at the front of the node then the .base offset will point
3307 somewhere inside the previous nodes data (or maybe even into a node
3308 even earlier), but the .check field determines if the transition is
3309 valid.
3310
786e8c11 3311 XXX - wrong maybe?
a3621e74 3312 The following process inplace converts the table to the compressed
3b753521 3313 table: We first do not compress the root node 1,and mark all its
a3621e74 3314 .check pointers as 1 and set its .base pointer as 1 as well. This
3b753521
FN
3315 allows us to do a DFA construction from the compressed table later,
3316 and ensures that any .base pointers we calculate later are greater
3317 than 0.
a3621e74
YO
3318
3319 - We set 'pos' to indicate the first entry of the second node.
3320
3321 - We then iterate over the columns of the node, finding the first and
3322 last used entry at l and m. We then copy l..m into pos..(pos+m-l),
3323 and set the .check pointers accordingly, and advance pos
3324 appropriately and repreat for the next node. Note that when we copy
3325 the next pointers we have to convert them from the original
3326 NODEIDX form to NODENUM form as the former is not valid post
3327 compression.
3328
3329 - If a node has no transitions used we mark its base as 0 and do not
3330 advance the pos pointer.
3331
3332 - If a node only has one transition we use a second pointer into the
3333 structure to fill in allocated fail transitions from other states.
3334 This pointer is independent of the main pointer and scans forward
3335 looking for null transitions that are allocated to a state. When it
3336 finds one it writes the single transition into the "hole". If the
786e8c11 3337 pointer doesnt find one the single transition is appended as normal.
a3621e74
YO
3338
3339 - Once compressed we can Renew/realloc the structures to release the
3340 excess space.
3341
3342 See "Table-Compression Methods" in sec 3.9 of the Red Dragon,
3343 specifically Fig 3.47 and the associated pseudocode.
3344
3345 demq
3346 */
a3b680e6 3347 const U32 laststate = TRIE_NODENUM( next_alloc );
a28509cc 3348 U32 state, charid;
a3621e74 3349 U32 pos = 0, zp=0;
1e2e3d02 3350 trie->statecount = laststate;
a3621e74
YO
3351
3352 for ( state = 1 ; state < laststate ; state++ ) {
3353 U8 flag = 0;
a28509cc
AL
3354 const U32 stateidx = TRIE_NODEIDX( state );
3355 const U32 o_used = trie->trans[ stateidx ].check;
3356 U32 used = trie->trans[ stateidx ].check;
a3621e74
YO
3357 trie->trans[ stateidx ].check = 0;
3358
538e84ed
KW
3359 for ( charid = 0;
3360 used && charid < trie->uniquecharcount;
3361 charid++ )
3362 {
a3621e74
YO
3363 if ( flag || trie->trans[ stateidx + charid ].next ) {
3364 if ( trie->trans[ stateidx + charid ].next ) {
3365 if (o_used == 1) {
3366 for ( ; zp < pos ; zp++ ) {
3367 if ( ! trie->trans[ zp ].next ) {
3368 break;
3369 }
3370 }
538e84ed
KW
3371 trie->states[ state ].trans.base
3372 = zp
3373 + trie->uniquecharcount
3374 - charid ;
3375 trie->trans[ zp ].next
3376 = SAFE_TRIE_NODENUM( trie->trans[ stateidx
3377 + charid ].next );
a3621e74
YO
3378 trie->trans[ zp ].check = state;
3379 if ( ++zp > pos ) pos = zp;
3380 break;
3381 }
3382 used--;
3383 }
3384 if ( !flag ) {
3385 flag = 1;
538e84ed
KW
3386 trie->states[ state ].trans.base
3387 = pos + trie->uniquecharcount - charid ;
a3621e74 3388 }
538e84ed
KW
3389 trie->trans[ pos ].next
3390 = SAFE_TRIE_NODENUM(
3391 trie->trans[ stateidx + charid ].next );
a3621e74
YO
3392 trie->trans[ pos ].check = state;
3393 pos++;
3394 }
3395 }
3396 }
cc601c31 3397 trie->lasttrans = pos + 1;
c944940b
JH
3398 trie->states = (reg_trie_state *)
3399 PerlMemShared_realloc( trie->states, laststate
3400 * sizeof(reg_trie_state) );
a3621e74 3401 DEBUG_TRIE_COMPILE_MORE_r(
147e3846 3402 Perl_re_indentf( aTHX_ "Alloc: %d Orig: %" IVdf " elements, Final:%" IVdf ". Savings of %%%5.2f\n",
cb41e5d6 3403 depth+1,
538e84ed
KW
3404 (int)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount
3405 + 1 ),
3406 (IV)next_alloc,
3407 (IV)pos,
3408 ( ( next_alloc - pos ) * 100 ) / (double)next_alloc );
a3621e74
YO
3409 );
3410
3411 } /* end table compress */
3412 }
1e2e3d02 3413 DEBUG_TRIE_COMPILE_MORE_r(
147e3846 3414 Perl_re_indentf( aTHX_ "Statecount:%" UVxf " Lasttrans:%" UVxf "\n",
cb41e5d6 3415 depth+1,
1e2e3d02
YO
3416