This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
regcomp.c: Rmv unnecessary branch
[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 */
134 char *copy_start_in_input; /* Position in input string
135 corresponding to copy_start */
ea3daa5d 136 SSize_t whilem_seen; /* number of WHILEM in this expr */
fac92740 137 regnode *emit_start; /* Start of emitted-code area */
f55b7b26 138 regnode_offset emit; /* Code-emit pointer */
830247a4
IZ
139 I32 naughty; /* How bad is this pattern? */
140 I32 sawback; /* Did we see \1, ...? */
141 U32 seen;
88f063b4
KW
142 SSize_t size; /* Number of regnode equivalents in
143 pattern */
7c932d07
KW
144
145 /* position beyond 'precomp' of the warning message furthest away from
146 * 'precomp'. During the parse, no warnings are raised for any problems
147 * earlier in the parse than this position. This works if warnings are
148 * raised the first time a given spot is parsed, and if only one
149 * independent warning is raised for any given spot */
150 Size_t latest_warn_offset;
151
152 I32 npar; /* Capture buffer count so far in the
153 parse, (OPEN) plus one. ("par" 0 is
154 the whole pattern)*/
155 I32 total_par; /* During initial parse, is either 0,
156 or -1; the latter indicating a
157 reparse is needed. After that pass,
158 it is what 'npar' became after the
159 pass. Hence, it being > 0 indicates
160 we are in a reparse situation */
538e84ed
KW
161 I32 nestroot; /* root parens we are in - used by
162 accept */
830247a4 163 I32 seen_zerolen;
f55b7b26
KW
164 regnode_offset *open_parens; /* offsets to open parens */
165 regnode_offset *close_parens; /* offsets to close parens */
d5a00e4a 166 regnode *end_op; /* END node in program */
02daf0ab
YO
167 I32 utf8; /* whether the pattern is utf8 or not */
168 I32 orig_utf8; /* whether the pattern was originally in utf8 */
169 /* XXX use this for future optimisation of case
170 * where pattern must be upgraded to utf8. */
e40e74fe
KW
171 I32 uni_semantics; /* If a d charset modifier should use unicode
172 rules, even if the pattern is not in
173 utf8 */
81714fb9 174 HV *paren_names; /* Paren names */
538e84ed 175
40d049e4 176 regnode **recurse; /* Recurse regops */
88f063b4 177 I32 recurse_count; /* Number of recurse regops we have generated */
4286711a 178 U8 *study_chunk_recursed; /* bitmap of which subs we have moved
538e84ed 179 through */
09a65838 180 U32 study_chunk_recursed_bytes; /* bytes in bitmap */
b57e4118 181 I32 in_lookbehind;
4624b182 182 I32 contains_locale;
bb3f3ed2 183 I32 override_recoding;
b6d67071
KW
184#ifdef EBCDIC
185 I32 recode_x_to_native;
186#endif
9d53c457 187 I32 in_multi_char_class;
1acab4c5 188 struct reg_code_blocks *code_blocks;/* positions of literal (?{})
68e2671b 189 within pattern */
b1603ef8 190 int code_index; /* next code_blocks[] slot */
ee273784 191 SSize_t maxlen; /* mininum possible number of chars in string to match */
3f910e62
YO
192 scan_frame *frame_head;
193 scan_frame *frame_last;
194 U32 frame_count;
7eec73eb 195 AV *warn_text;
dc6d7f5c 196#ifdef ADD_TO_REGEXEC
830247a4
IZ
197 char *starttry; /* -Dr: where regtry was called. */
198#define RExC_starttry (pRExC_state->starttry)
199#endif
d24ca0c5 200 SV *runtime_code_qr; /* qr with the runtime code blocks */
3dab1dad 201#ifdef DEBUGGING
be8e71aa 202 const char *lastparse;
3dab1dad 203 I32 lastnum;
1f1031fe 204 AV *paren_name_list; /* idx -> name */
d9a72fcc 205 U32 study_chunk_recursed_count;
c9f0d54c
YO
206 SV *mysv1;
207 SV *mysv2;
88f063b4 208
3dab1dad
YO
209#define RExC_lastparse (pRExC_state->lastparse)
210#define RExC_lastnum (pRExC_state->lastnum)
1f1031fe 211#define RExC_paren_name_list (pRExC_state->paren_name_list)
d9a72fcc 212#define RExC_study_chunk_recursed_count (pRExC_state->study_chunk_recursed_count)
c9f0d54c
YO
213#define RExC_mysv (pRExC_state->mysv1)
214#define RExC_mysv1 (pRExC_state->mysv1)
215#define RExC_mysv2 (pRExC_state->mysv2)
216
3dab1dad 217#endif
bd0b62a6 218 bool seen_d_op;
911bd04e 219 bool strict;
da7cf1cc 220 bool study_started;
034602eb 221 bool in_script_run;
7c932d07 222 bool use_BRANCHJ;
09b2b2e6 223};
830247a4 224
e2509266 225#define RExC_flags (pRExC_state->flags)
514a91f1 226#define RExC_pm_flags (pRExC_state->pm_flags)
830247a4 227#define RExC_precomp (pRExC_state->precomp)
51684e3c
KW
228#define RExC_copy_start_in_input (pRExC_state->copy_start_in_input)
229#define RExC_copy_start_in_constructed (pRExC_state->copy_start)
711b303b 230#define RExC_precomp_end (pRExC_state->precomp_end)
288b8c02 231#define RExC_rx_sv (pRExC_state->rx_sv)
830247a4 232#define RExC_rx (pRExC_state->rx)
f8fc2ecf 233#define RExC_rxi (pRExC_state->rxi)
fac92740 234#define RExC_start (pRExC_state->start)
830247a4
IZ
235#define RExC_end (pRExC_state->end)
236#define RExC_parse (pRExC_state->parse)
7c932d07 237#define RExC_latest_warn_offset (pRExC_state->latest_warn_offset )
830247a4 238#define RExC_whilem_seen (pRExC_state->whilem_seen)
bd0b62a6
KW
239#define RExC_seen_d_op (pRExC_state->seen_d_op) /* Seen something that differs
240 under /d from /u ? */
512c0f5a 241
512c0f5a 242
7122b237 243#ifdef RE_TRACK_PATTERN_OFFSETS
44f36cd3 244# define RExC_offsets (RExC_rxi->u.offsets) /* I am not like the
538e84ed 245 others */
7122b237 246#endif
830247a4 247#define RExC_emit (pRExC_state->emit)
fac92740 248#define RExC_emit_start (pRExC_state->emit_start)
830247a4
IZ
249#define RExC_sawback (pRExC_state->sawback)
250#define RExC_seen (pRExC_state->seen)
251#define RExC_size (pRExC_state->size)
ee273784 252#define RExC_maxlen (pRExC_state->maxlen)
830247a4 253#define RExC_npar (pRExC_state->npar)
1533d1d2 254#define RExC_total_parens (pRExC_state->total_par)
e2e6a0f1 255#define RExC_nestroot (pRExC_state->nestroot)
830247a4 256#define RExC_seen_zerolen (pRExC_state->seen_zerolen)
1aa99e6b 257#define RExC_utf8 (pRExC_state->utf8)
e40e74fe 258#define RExC_uni_semantics (pRExC_state->uni_semantics)
02daf0ab 259#define RExC_orig_utf8 (pRExC_state->orig_utf8)
40d049e4
YO
260#define RExC_open_parens (pRExC_state->open_parens)
261#define RExC_close_parens (pRExC_state->close_parens)
5bd2d46e 262#define RExC_end_op (pRExC_state->end_op)
81714fb9 263#define RExC_paren_names (pRExC_state->paren_names)
40d049e4
YO
264#define RExC_recurse (pRExC_state->recurse)
265#define RExC_recurse_count (pRExC_state->recurse_count)
09a65838 266#define RExC_study_chunk_recursed (pRExC_state->study_chunk_recursed)
538e84ed
KW
267#define RExC_study_chunk_recursed_bytes \
268 (pRExC_state->study_chunk_recursed_bytes)
b57e4118 269#define RExC_in_lookbehind (pRExC_state->in_lookbehind)
4624b182 270#define RExC_contains_locale (pRExC_state->contains_locale)
b6d67071
KW
271#ifdef EBCDIC
272# define RExC_recode_x_to_native (pRExC_state->recode_x_to_native)
273#endif
9d53c457 274#define RExC_in_multi_char_class (pRExC_state->in_multi_char_class)
3f910e62
YO
275#define RExC_frame_head (pRExC_state->frame_head)
276#define RExC_frame_last (pRExC_state->frame_last)
277#define RExC_frame_count (pRExC_state->frame_count)
67cdf558 278#define RExC_strict (pRExC_state->strict)
da7cf1cc 279#define RExC_study_started (pRExC_state->study_started)
7eec73eb 280#define RExC_warn_text (pRExC_state->warn_text)
034602eb 281#define RExC_in_script_run (pRExC_state->in_script_run)
7c932d07 282#define RExC_use_BRANCHJ (pRExC_state->use_BRANCHJ)
830247a4 283
99807a43
HS
284/* Heuristic check on the complexity of the pattern: if TOO_NAUGHTY, we set
285 * a flag to disable back-off on the fixed/floating substrings - if it's
286 * a high complexity pattern we assume the benefit of avoiding a full match
287 * is worth the cost of checking for the substrings even if they rarely help.
288 */
289#define RExC_naughty (pRExC_state->naughty)
290#define TOO_NAUGHTY (10)
291#define MARK_NAUGHTY(add) \
292 if (RExC_naughty < TOO_NAUGHTY) \
293 RExC_naughty += (add)
294#define MARK_NAUGHTY_EXP(exp, add) \
295 if (RExC_naughty < TOO_NAUGHTY) \
296 RExC_naughty += RExC_naughty / (exp) + (add)
cde0cee5 297
a687059c
LW
298#define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
299#define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
412f55bb 300 ((*s) == '{' && regcurly(s)))
a687059c
LW
301
302/*
303 * Flags to be passed up and down.
304 */
a687059c 305#define WORST 0 /* Worst case. */
c4ead770
KW
306#define HASWIDTH 0x01 /* Known to not match null strings, could match
307 non-null ones. */
fda99bee 308
e64f369d 309/* Simple enough to be STAR/PLUS operand; in an EXACTish node must be a single
2fd92675
KW
310 * character. (There needs to be a case: in the switch statement in regexec.c
311 * for any node marked SIMPLE.) Note that this is not the same thing as
312 * REGNODE_SIMPLE */
fda99bee 313#define SIMPLE 0x02
e64f369d 314#define SPSTART 0x04 /* Starts with * or + */
8d9c2815
NC
315#define POSTPONED 0x08 /* (?1),(?&name), (??{...}) or similar */
316#define TRYAGAIN 0x10 /* Weeded out a declaration. */
c36361b2
KW
317#define RESTART_PARSE 0x20 /* Need to redo the parse */
318#define NEED_UTF8 0x40 /* In conjunction with RESTART_PARSE, need to
319 calcuate sizes as UTF-8 */
a687059c 320
3dab1dad
YO
321#define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1)
322
07be1b83
YO
323/* whether trie related optimizations are enabled */
324#if PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION
325#define TRIE_STUDY_OPT
786e8c11 326#define FULL_TRIE_STUDY
07be1b83
YO
327#define TRIE_STCLASS
328#endif
1de06328
YO
329
330
40d049e4
YO
331
332#define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3]
333#define PBITVAL(paren) (1 << ((paren) & 7))
334#define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren))
335#define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren)
336#define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren))
337
82a6ada4 338#define REQUIRE_UTF8(flagp) STMT_START { \
8d9c2815 339 if (!UTF) { \
c36361b2 340 *flagp = RESTART_PARSE|NEED_UTF8; \
f55b7b26 341 return 0; \
8d9c2815 342 } \
82a6ada4 343 } STMT_END
40d049e4 344
d63d75f5 345/* Change from /d into /u rules, and restart the parse. RExC_uni_semantics is
f6e8f31e
KW
346 * a flag that indicates we need to override /d with /u as a result of
347 * something in the pattern. It should only be used in regards to calling
348 * set_regex_charset() or get_regex_charse() */
512c0f5a
KW
349#define REQUIRE_UNI_RULES(flagp, restart_retval) \
350 STMT_START { \
351 if (DEPENDS_SEMANTICS) { \
512c0f5a
KW
352 set_regex_charset(&RExC_flags, REGEX_UNICODE_CHARSET); \
353 RExC_uni_semantics = 1; \
bd0b62a6
KW
354 if (RExC_seen_d_op && LIKELY(RExC_total_parens >= 0)) { \
355 /* No need to restart the parse if we haven't seen \
356 * anything that differs between /u and /d, and no need \
357 * to restart immediately if we're going to reparse \
358 * anyway to count parens */ \
c36361b2 359 *flagp |= RESTART_PARSE; \
512c0f5a 360 return restart_retval; \
d1be211c 361 } \
512c0f5a
KW
362 } \
363 } STMT_END
364
7c932d07
KW
365#define BRANCH_MAX_OFFSET U16_MAX
366#define REQUIRE_BRANCHJ(flagp, restart_retval) \
367 STMT_START { \
368 RExC_use_BRANCHJ = 1; \
523a2924
KW
369 if (LIKELY(RExC_total_parens >= 0)) { \
370 /* No need to restart the parse immediately if we're \
371 * going to reparse anyway to count parens */ \
372 *flagp |= RESTART_PARSE; \
373 return restart_retval; \
374 } \
7c932d07
KW
375 } STMT_END
376
377#define REQUIRE_PARENS_PASS \
378 STMT_START { \
379 if (RExC_total_parens == 0) RExC_total_parens = -1; \
380 } STMT_END
381
f358228e
KW
382/* This is used to return failure (zero) early from the calling function if
383 * various flags in 'flags' are set. Two flags always cause a return:
384 * 'RESTART_PARSE' and 'NEED_UTF8'. 'extra' can be used to specify any
385 * additional flags that should cause a return; 0 if none. If the return will
386 * be done, '*flagp' is first set to be all of the flags that caused the
387 * return. */
388#define RETURN_FAIL_ON_RESTART_OR_FLAGS(flags,flagp,extra) \
fd7fc6ae 389 STMT_START { \
c36361b2
KW
390 if ((flags) & (RESTART_PARSE|NEED_UTF8|(extra))) { \
391 *(flagp) = (flags) & (RESTART_PARSE|NEED_UTF8|(extra)); \
f358228e 392 return 0; \
fd7fc6ae
YO
393 } \
394 } STMT_END
395
c36361b2 396#define MUST_RESTART(flags) ((flags) & (RESTART_PARSE))
fd7fc6ae 397
5d555bb8 398#define RETURN_FAIL_ON_RESTART(flags,flagp) \
f358228e 399 RETURN_FAIL_ON_RESTART_OR_FLAGS( flags, flagp, 0)
5d555bb8 400#define RETURN_FAIL_ON_RESTART_FLAGP(flagp) \
f358228e 401 if (MUST_RESTART(*(flagp))) return 0
fd7fc6ae 402
f19b1a63
KW
403/* This converts the named class defined in regcomp.h to its equivalent class
404 * number defined in handy.h. */
405#define namedclass_to_classnum(class) ((int) ((class) / 2))
406#define classnum_to_namedclass(classnum) ((classnum) * 2)
407
de92f5e6
KW
408#define _invlist_union_complement_2nd(a, b, output) \
409 _invlist_union_maybe_complement_2nd(a, b, TRUE, output)
410#define _invlist_intersection_complement_2nd(a, b, output) \
411 _invlist_intersection_maybe_complement_2nd(a, b, TRUE, output)
412
1de06328
YO
413/* About scan_data_t.
414
415 During optimisation we recurse through the regexp program performing
416 various inplace (keyhole style) optimisations. In addition study_chunk
417 and scan_commit populate this data structure with information about
538e84ed 418 what strings MUST appear in the pattern. We look for the longest
3b753521 419 string that must appear at a fixed location, and we look for the
1de06328
YO
420 longest string that may appear at a floating location. So for instance
421 in the pattern:
538e84ed 422
1de06328 423 /FOO[xX]A.*B[xX]BAR/
538e84ed 424
1de06328
YO
425 Both 'FOO' and 'A' are fixed strings. Both 'B' and 'BAR' are floating
426 strings (because they follow a .* construct). study_chunk will identify
427 both FOO and BAR as being the longest fixed and floating strings respectively.
538e84ed 428
1de06328 429 The strings can be composites, for instance
538e84ed 430
1de06328 431 /(f)(o)(o)/
538e84ed 432
1de06328 433 will result in a composite fixed substring 'foo'.
538e84ed 434
1de06328 435 For each string some basic information is maintained:
538e84ed 436
6ea77169 437 - min_offset
1de06328
YO
438 This is the position the string must appear at, or not before.
439 It also implicitly (when combined with minlenp) tells us how many
3b753521
FN
440 characters must match before the string we are searching for.
441 Likewise when combined with minlenp and the length of the string it
538e84ed 442 tells us how many characters must appear after the string we have
1de06328 443 found.
538e84ed 444
1de06328
YO
445 - max_offset
446 Only used for floating strings. This is the rightmost point that
ea3daa5d 447 the string can appear at. If set to SSize_t_MAX it indicates that the
1de06328 448 string can occur infinitely far to the right.
6e12f832 449 For fixed strings, it is equal to min_offset.
538e84ed 450
1de06328 451 - minlenp
2d608413
KW
452 A pointer to the minimum number of characters of the pattern that the
453 string was found inside. This is important as in the case of positive
538e84ed 454 lookahead or positive lookbehind we can have multiple patterns
1de06328 455 involved. Consider
538e84ed 456
1de06328 457 /(?=FOO).*F/
538e84ed 458
1de06328
YO
459 The minimum length of the pattern overall is 3, the minimum length
460 of the lookahead part is 3, but the minimum length of the part that
538e84ed 461 will actually match is 1. So 'FOO's minimum length is 3, but the
1de06328 462 minimum length for the F is 1. This is important as the minimum length
538e84ed 463 is used to determine offsets in front of and behind the string being
1de06328 464 looked for. Since strings can be composites this is the length of the
486ec47a 465 pattern at the time it was committed with a scan_commit. Note that
1de06328 466 the length is calculated by study_chunk, so that the minimum lengths
538e84ed 467 are not known until the full pattern has been compiled, thus the
1de06328 468 pointer to the value.
538e84ed 469
1de06328 470 - lookbehind
538e84ed 471
1de06328 472 In the case of lookbehind the string being searched for can be
538e84ed 473 offset past the start point of the final matching string.
1de06328
YO
474 If this value was just blithely removed from the min_offset it would
475 invalidate some of the calculations for how many chars must match
476 before or after (as they are derived from min_offset and minlen and
538e84ed 477 the length of the string being searched for).
1de06328
YO
478 When the final pattern is compiled and the data is moved from the
479 scan_data_t structure into the regexp structure the information
538e84ed
KW
480 about lookbehind is factored in, with the information that would
481 have been lost precalculated in the end_shift field for the
1de06328
YO
482 associated string.
483
484 The fields pos_min and pos_delta are used to store the minimum offset
538e84ed 485 and the delta to the maximum offset at the current point in the pattern.
1de06328
YO
486
487*/
2c2d71f5 488
cb6aef8b
DM
489struct scan_data_substrs {
490 SV *str; /* longest substring found in pattern */
491 SSize_t min_offset; /* earliest point in string it can appear */
492 SSize_t max_offset; /* latest point in string it can appear */
493 SSize_t *minlenp; /* pointer to the minlen relevant to the string */
494 SSize_t lookbehind; /* is the pos of the string modified by LB */
11683ecb 495 I32 flags; /* per substring SF_* and SCF_* flags */
cb6aef8b
DM
496};
497
2c2d71f5 498typedef struct scan_data_t {
1de06328
YO
499 /*I32 len_min; unused */
500 /*I32 len_delta; unused */
49f55535 501 SSize_t pos_min;
ea3daa5d 502 SSize_t pos_delta;
2c2d71f5 503 SV *last_found;
ea3daa5d 504 SSize_t last_end; /* min value, <0 unless valid. */
49f55535 505 SSize_t last_start_min;
ea3daa5d 506 SSize_t last_start_max;
2099df82
DM
507 U8 cur_is_floating; /* whether the last_* values should be set as
508 * the next fixed (0) or floating (1)
509 * substring */
6ea77169 510
2099df82
DM
511 /* [0] is longest fixed substring so far, [1] is longest float so far */
512 struct scan_data_substrs substrs[2];
6ea77169 513
11683ecb 514 I32 flags; /* common SF_* and SCF_* flags */
2c2d71f5 515 I32 whilem_c;
49f55535 516 SSize_t *last_closep;
b8f7bb16 517 regnode_ssc *start_class;
2c2d71f5
JH
518} scan_data_t;
519
a687059c 520/*
e50aee73 521 * Forward declarations for pregcomp()'s friends.
a687059c 522 */
a0d0e21e 523
6ea77169 524static const scan_data_t zero_scan_data = {
a2ca2017 525 0, 0, NULL, 0, 0, 0, 0,
6ea77169 526 {
11683ecb
DM
527 { NULL, 0, 0, 0, 0, 0 },
528 { NULL, 0, 0, 0, 0, 0 },
6ea77169
DM
529 },
530 0, 0, NULL, NULL
531};
c277df42 532
11683ecb
DM
533/* study flags */
534
07be1b83
YO
535#define SF_BEFORE_SEOL 0x0001
536#define SF_BEFORE_MEOL 0x0002
11683ecb 537#define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
c277df42 538
07be1b83
YO
539#define SF_IS_INF 0x0040
540#define SF_HAS_PAR 0x0080
541#define SF_IN_PAR 0x0100
542#define SF_HAS_EVAL 0x0200
bc604ad8
KW
543
544
545/* SCF_DO_SUBSTR is the flag that tells the regexp analyzer to track the
546 * longest substring in the pattern. When it is not set the optimiser keeps
547 * track of position, but does not keep track of the actual strings seen,
548 *
549 * So for instance /foo/ will be parsed with SCF_DO_SUBSTR being true, but
550 * /foo/i will not.
551 *
552 * Similarly, /foo.*(blah|erm|huh).*fnorble/ will have "foo" and "fnorble"
553 * parsed with SCF_DO_SUBSTR on, but while processing the (...) it will be
554 * turned off because of the alternation (BRANCH). */
07be1b83 555#define SCF_DO_SUBSTR 0x0400
bc604ad8 556
653099ff
GS
557#define SCF_DO_STCLASS_AND 0x0800
558#define SCF_DO_STCLASS_OR 0x1000
559#define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
e1901655 560#define SCF_WHILEM_VISITED_POS 0x2000
c277df42 561
786e8c11 562#define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */
538e84ed 563#define SCF_SEEN_ACCEPT 0x8000
688e0391 564#define SCF_TRIE_DOING_RESTUDY 0x10000
4286711a
YO
565#define SCF_IN_DEFINE 0x20000
566
567
568
07be1b83 569
43fead97 570#define UTF cBOOL(RExC_utf8)
00b27cfc
KW
571
572/* The enums for all these are ordered so things work out correctly */
a62b1201 573#define LOC (get_regex_charset(RExC_flags) == REGEX_LOCALE_CHARSET)
538e84ed
KW
574#define DEPENDS_SEMANTICS (get_regex_charset(RExC_flags) \
575 == REGEX_DEPENDS_CHARSET)
00b27cfc 576#define UNI_SEMANTICS (get_regex_charset(RExC_flags) == REGEX_UNICODE_CHARSET)
538e84ed
KW
577#define AT_LEAST_UNI_SEMANTICS (get_regex_charset(RExC_flags) \
578 >= REGEX_UNICODE_CHARSET)
579#define ASCII_RESTRICTED (get_regex_charset(RExC_flags) \
580 == REGEX_ASCII_RESTRICTED_CHARSET)
581#define AT_LEAST_ASCII_RESTRICTED (get_regex_charset(RExC_flags) \
582 >= REGEX_ASCII_RESTRICTED_CHARSET)
583#define ASCII_FOLD_RESTRICTED (get_regex_charset(RExC_flags) \
584 == REGEX_ASCII_MORE_RESTRICTED_CHARSET)
a62b1201 585
43fead97 586#define FOLD cBOOL(RExC_flags & RXf_PMf_FOLD)
a0ed51b3 587
f2c2a6ab
KW
588/* For programs that want to be strictly Unicode compatible by dying if any
589 * attempt is made to match a non-Unicode code point against a Unicode
590 * property. */
591#define ALWAYS_WARN_SUPER ckDEAD(packWARN(WARN_NON_UNICODE))
592
93733859 593#define OOB_NAMEDCLASS -1
b8c5462f 594
8e661ac5
KW
595/* There is no code point that is out-of-bounds, so this is problematic. But
596 * its only current use is to initialize a variable that is always set before
597 * looked at. */
598#define OOB_UNICODE 0xDEADBEEF
599
a0ed51b3 600#define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
a0ed51b3 601
8615cb43 602
b45f050a
JF
603/* length of regex to show in messages that don't mark a position within */
604#define RegexLengthToShowInErrorMessages 127
605
606/*
607 * If MARKER[12] are adjusted, be sure to adjust the constants at the top
608 * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
609 * op/pragma/warn/regcomp.
610 */
7253e4e3
RK
611#define MARKER1 "<-- HERE" /* marker as it appears in the description */
612#define MARKER2 " <-- HERE " /* marker as it appears within the regex */
b81d288d 613
538e84ed 614#define REPORT_LOCATION " in regex; marked by " MARKER1 \
147e3846 615 " in m/%" UTF8f MARKER2 "%" UTF8f "/"
b45f050a 616
285b5ca0
KW
617/* The code in this file in places uses one level of recursion with parsing
618 * rebased to an alternate string constructed by us in memory. This can take
619 * the form of something that is completely different from the input, or
620 * something that uses the input as part of the alternate. In the first case,
621 * there should be no possibility of an error, as we are in complete control of
51684e3c
KW
622 * the alternate string. But in the second case we don't completely control
623 * the input portion, so there may be errors in that. Here's an example:
285b5ca0
KW
624 * /[abc\x{DF}def]/ui
625 * is handled specially because \x{df} folds to a sequence of more than one
51684e3c 626 * character: 'ss'. What is done is to create and parse an alternate string,
285b5ca0
KW
627 * which looks like this:
628 * /(?:\x{DF}|[abc\x{DF}def])/ui
629 * where it uses the input unchanged in the middle of something it constructs,
630 * which is a branch for the DF outside the character class, and clustering
631 * parens around the whole thing. (It knows enough to skip the DF inside the
632 * class while in this substitute parse.) 'abc' and 'def' may have errors that
633 * need to be reported. The general situation looks like this:
634 *
51684e3c 635 * |<------- identical ------>|
285b5ca0 636 * sI tI xI eI
51684e3c 637 * Input: ---------------------------------------------------------------
285b5ca0
KW
638 * Constructed: ---------------------------------------------------
639 * sC tC xC eC EC
51684e3c 640 * |<------- identical ------>|
285b5ca0 641 *
51684e3c
KW
642 * sI..eI is the portion of the input pattern we are concerned with here.
643 * sC..EC is the constructed substitute parse string.
644 * sC..tC is constructed by us
645 * tC..eC is an exact duplicate of the portion of the input pattern tI..eI.
646 * In the diagram, these are vertically aligned.
647 * eC..EC is also constructed by us.
648 * xC is the position in the substitute parse string where we found a
649 * problem.
650 * xI is the position in the original pattern corresponding to xC.
285b5ca0 651 *
51684e3c
KW
652 * We want to display a message showing the real input string. Thus we need to
653 * translate from xC to xI. We know that xC >= tC, since the portion of the
654 * string sC..tC has been constructed by us, and so shouldn't have errors. We
655 * get:
656 * xI = tI + (xC - tC)
285b5ca0 657 *
51684e3c
KW
658 * When the substitute parse is constructed, the code needs to set:
659 * RExC_start (sC)
660 * RExC_end (eC)
661 * RExC_copy_start_in_input (tI)
662 * RExC_copy_start_in_constructed (tC)
663 * and restore them when done.
285b5ca0 664 *
51684e3c
KW
665 * During normal processing of the input pattern, both
666 * 'RExC_copy_start_in_input' and 'RExC_copy_start_in_constructed' are set to
667 * sI, so that xC equals xI.
285b5ca0
KW
668 */
669
51684e3c
KW
670#define sI RExC_precomp
671#define eI RExC_precomp_end
672#define sC RExC_start
673#define eC RExC_end
674#define tI RExC_copy_start_in_input
675#define tC RExC_copy_start_in_constructed
676#define xI(xC) (tI + (xC - tC))
677#define xI_offset(xC) (xI(xC) - sI)
285b5ca0
KW
678
679#define REPORT_LOCATION_ARGS(xC) \
680 UTF8fARG(UTF, \
51684e3c 681 (xI(xC) > eI) /* Don't run off end */ \
232b691f 682 ? eI - sI /* Length before the <--HERE */ \
bc929b61
KW
683 : ((xI_offset(xC) >= 0) \
684 ? xI_offset(xC) \
685 : (Perl_croak(aTHX_ "panic: %s: %d: negative offset: %" \
686 IVdf " trying to output message for " \
687 " pattern %.*s", \
51684e3c 688 __FILE__, __LINE__, (IV) xI_offset(xC), \
bc929b61 689 ((int) (eC - sC)), sC), 0)), \
51684e3c 690 sI), /* The input pattern printed up to the <--HERE */ \
285b5ca0 691 UTF8fARG(UTF, \
51684e3c
KW
692 (xI(xC) > eI) ? 0 : eI - xI(xC), /* Length after <--HERE */ \
693 (xI(xC) > eI) ? eI : xI(xC)) /* pattern after <--HERE */
c1d900c3 694
8a6d8ec6
HS
695/* Used to point after bad bytes for an error message, but avoid skipping
696 * past a nul byte. */
697#define SKIP_IF_CHAR(s) (!*(s) ? 0 : UTF ? UTF8SKIP(s) : 1)
698
ed68fe11
KW
699/* Set up to clean up after our imminent demise */
700#define PREPARE_TO_DIE \
701 STMT_START { \
702 if (RExC_rx_sv) \
703 SAVEFREESV(RExC_rx_sv); \
7c932d07
KW
704 if (RExC_open_parens) \
705 SAVEFREEPV(RExC_open_parens); \
706 if (RExC_close_parens) \
707 SAVEFREEPV(RExC_close_parens); \
ed68fe11
KW
708 } STMT_END
709
b45f050a
JF
710/*
711 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
712 * arg. Show regex, up to a maximum length. If it's too long, chop and add
713 * "...".
714 */
58e23c8d 715#define _FAIL(code) STMT_START { \
bfed75c6 716 const char *ellipses = ""; \
88f063b4 717 IV len = RExC_precomp_end - RExC_precomp; \
ccb2c380 718 \
ed68fe11 719 PREPARE_TO_DIE; \
ccb2c380
MP
720 if (len > RegexLengthToShowInErrorMessages) { \
721 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
722 len = RegexLengthToShowInErrorMessages - 10; \
723 ellipses = "..."; \
724 } \
58e23c8d 725 code; \
ccb2c380 726} STMT_END
8615cb43 727
58e23c8d 728#define FAIL(msg) _FAIL( \
147e3846 729 Perl_croak(aTHX_ "%s in regex m/%" UTF8f "%s/", \
c1d900c3 730 msg, UTF8fARG(UTF, len, RExC_precomp), ellipses))
58e23c8d
YO
731
732#define FAIL2(msg,arg) _FAIL( \
147e3846 733 Perl_croak(aTHX_ msg " in regex m/%" UTF8f "%s/", \
c1d900c3 734 arg, UTF8fARG(UTF, len, RExC_precomp), ellipses))
58e23c8d 735
b45f050a 736/*
b45f050a
JF
737 * Simple_vFAIL -- like FAIL, but marks the current location in the scan
738 */
ccb2c380 739#define Simple_vFAIL(m) STMT_START { \
ccb2c380 740 Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
d528642a 741 m, REPORT_LOCATION_ARGS(RExC_parse)); \
ccb2c380 742} STMT_END
b45f050a
JF
743
744/*
745 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
746 */
ccb2c380 747#define vFAIL(m) STMT_START { \
ed68fe11 748 PREPARE_TO_DIE; \
ccb2c380
MP
749 Simple_vFAIL(m); \
750} STMT_END
b45f050a
JF
751
752/*
753 * Like Simple_vFAIL(), but accepts two arguments.
754 */
ccb2c380 755#define Simple_vFAIL2(m,a1) STMT_START { \
d528642a
KW
756 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, \
757 REPORT_LOCATION_ARGS(RExC_parse)); \
ccb2c380 758} STMT_END
b45f050a
JF
759
760/*
761 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
762 */
ccb2c380 763#define vFAIL2(m,a1) STMT_START { \
ed68fe11 764 PREPARE_TO_DIE; \
ccb2c380
MP
765 Simple_vFAIL2(m, a1); \
766} STMT_END
b45f050a
JF
767
768
769/*
770 * Like Simple_vFAIL(), but accepts three arguments.
771 */
ccb2c380 772#define Simple_vFAIL3(m, a1, a2) STMT_START { \
c1d900c3 773 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, \
d528642a 774 REPORT_LOCATION_ARGS(RExC_parse)); \
ccb2c380 775} STMT_END
b45f050a
JF
776
777/*
778 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
779 */
ccb2c380 780#define vFAIL3(m,a1,a2) STMT_START { \
ed68fe11 781 PREPARE_TO_DIE; \
ccb2c380
MP
782 Simple_vFAIL3(m, a1, a2); \
783} STMT_END
b45f050a
JF
784
785/*
786 * Like Simple_vFAIL(), but accepts four arguments.
787 */
ccb2c380 788#define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \
d528642a
KW
789 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, a3, \
790 REPORT_LOCATION_ARGS(RExC_parse)); \
ccb2c380 791} STMT_END
b45f050a 792
95db3ffa 793#define vFAIL4(m,a1,a2,a3) STMT_START { \
ed68fe11 794 PREPARE_TO_DIE; \
95db3ffa
KW
795 Simple_vFAIL4(m, a1, a2, a3); \
796} STMT_END
797
946095af 798/* A specialized version of vFAIL2 that works with UTF8f */
d528642a 799#define vFAIL2utf8f(m, a1) STMT_START { \
ed68fe11 800 PREPARE_TO_DIE; \
d528642a
KW
801 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, \
802 REPORT_LOCATION_ARGS(RExC_parse)); \
946095af
BF
803} STMT_END
804
3ba22297 805#define vFAIL3utf8f(m, a1, a2) STMT_START { \
ed68fe11 806 PREPARE_TO_DIE; \
3ba22297 807 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, \
d528642a 808 REPORT_LOCATION_ARGS(RExC_parse)); \
3ba22297
KW
809} STMT_END
810
d62e749e
KW
811/* Setting this to NULL is a signal to not output warnings */
812#define TURN_OFF_WARNINGS_IN_SUBSTITUTE_PARSE RExC_copy_start_in_constructed = NULL
813#define RESTORE_WARNINGS RExC_copy_start_in_constructed = RExC_precomp
814
7c932d07
KW
815/* Since a warning can be generated multiple times as the input is reparsed, we
816 * output it the first time we come to that point in the parse, but suppress it
817 * otherwise. 'RExC_copy_start_in_constructed' being NULL is a flag to not
818 * generate any warnings */
55bcf857 819#define TO_OUTPUT_WARNINGS(loc) \
7c932d07
KW
820 ( RExC_copy_start_in_constructed \
821 && ((xI(loc)) - RExC_precomp) > (Ptrdiff_t) RExC_latest_warn_offset)
55bcf857 822
7c932d07
KW
823/* After we've emitted a warning, we save the position in the input so we don't
824 * output it again */
825#define UPDATE_WARNINGS_LOC(loc) \
826 STMT_START { \
827 if (TO_OUTPUT_WARNINGS(loc)) { \
828 RExC_latest_warn_offset = (xI(loc)) - RExC_precomp; \
829 } \
830 } STMT_END
77ccba3c 831
086f21bf 832/* 'warns' is the output of the packWARNx macro used in 'code' */
f8220fe2
KW
833#define _WARN_HELPER(loc, warns, code) \
834 STMT_START { \
d62e749e
KW
835 if (! RExC_copy_start_in_constructed) { \
836 Perl_croak( aTHX_ "panic! %s: %d: Tried to warn when none" \
837 " expected at '%s'", \
838 __FILE__, __LINE__, loc); \
839 } \
55bcf857 840 if (TO_OUTPUT_WARNINGS(loc)) { \
086f21bf
KW
841 if (ckDEAD(warns)) \
842 PREPARE_TO_DIE; \
338190a1 843 code; \
77ccba3c 844 UPDATE_WARNINGS_LOC(loc); \
338190a1 845 } \
f8220fe2 846 } STMT_END
946095af 847
5e0a247b 848/* m is not necessarily a "literal string", in this macro */
f8220fe2
KW
849#define reg_warn_non_literal_string(loc, m) \
850 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
851 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a 852 "%s" REPORT_LOCATION, \
f8220fe2 853 m, REPORT_LOCATION_ARGS(loc)))
5e0a247b 854
f8220fe2
KW
855#define ckWARNreg(loc,m) \
856 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
857 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a 858 m REPORT_LOCATION, \
f8220fe2 859 REPORT_LOCATION_ARGS(loc)))
ccb2c380 860
f8220fe2
KW
861#define vWARN(loc, m) \
862 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
863 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a 864 m REPORT_LOCATION, \
f8220fe2 865 REPORT_LOCATION_ARGS(loc))) \
b927b7e9 866
f8220fe2
KW
867#define vWARN_dep(loc, m) \
868 _WARN_HELPER(loc, packWARN(WARN_DEPRECATED), \
869 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), \
d528642a 870 m REPORT_LOCATION, \
f8220fe2 871 REPORT_LOCATION_ARGS(loc)))
0d6106aa 872
f8220fe2
KW
873#define ckWARNdep(loc,m) \
874 _WARN_HELPER(loc, packWARN(WARN_DEPRECATED), \
875 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
d528642a 876 m REPORT_LOCATION, \
f8220fe2 877 REPORT_LOCATION_ARGS(loc)))
147508a2 878
f8220fe2
KW
879#define ckWARNregdep(loc,m) \
880 _WARN_HELPER(loc, packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
881 Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, \
d528642a
KW
882 WARN_REGEXP), \
883 m REPORT_LOCATION, \
f8220fe2 884 REPORT_LOCATION_ARGS(loc)))
ccb2c380 885
f8220fe2
KW
886#define ckWARN2reg_d(loc,m, a1) \
887 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
888 Perl_ck_warner_d(aTHX_ packWARN(WARN_REGEXP), \
d528642a 889 m REPORT_LOCATION, \
f8220fe2 890 a1, REPORT_LOCATION_ARGS(loc)))
2335b3d3 891
f8220fe2
KW
892#define ckWARN2reg(loc, m, a1) \
893 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
894 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a 895 m REPORT_LOCATION, \
f8220fe2 896 a1, REPORT_LOCATION_ARGS(loc)))
ccb2c380 897
f8220fe2
KW
898#define vWARN3(loc, m, a1, a2) \
899 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
900 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a 901 m REPORT_LOCATION, \
f8220fe2 902 a1, a2, REPORT_LOCATION_ARGS(loc)))
ccb2c380 903
f8220fe2
KW
904#define ckWARN3reg(loc, m, a1, a2) \
905 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
906 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a
KW
907 m REPORT_LOCATION, \
908 a1, a2, \
f8220fe2 909 REPORT_LOCATION_ARGS(loc)))
668c081a 910
f8220fe2
KW
911#define vWARN4(loc, m, a1, a2, a3) \
912 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
913 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a
KW
914 m REPORT_LOCATION, \
915 a1, a2, a3, \
f8220fe2 916 REPORT_LOCATION_ARGS(loc)))
ccb2c380 917
f8220fe2
KW
918#define ckWARN4reg(loc, m, a1, a2, a3) \
919 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
920 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a
KW
921 m REPORT_LOCATION, \
922 a1, a2, a3, \
f8220fe2 923 REPORT_LOCATION_ARGS(loc)))
668c081a 924
f8220fe2
KW
925#define vWARN5(loc, m, a1, a2, a3, a4) \
926 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
927 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
d528642a
KW
928 m REPORT_LOCATION, \
929 a1, a2, a3, a4, \
f8220fe2 930 REPORT_LOCATION_ARGS(loc)))
9d1d55b5 931
aea5f181
KW
932#define ckWARNexperimental(loc, class, m) \
933 _WARN_HELPER(loc, packWARN(class), \
934 Perl_ck_warner_d(aTHX_ packWARN(class), \
935 m REPORT_LOCATION, \
936 REPORT_LOCATION_ARGS(loc)))
937
1a322bb4
KW
938/* Convert between a pointer to a node and its offset from the beginning of the
939 * program */
940#define REGNODE_p(offset) (RExC_emit_start + (offset))
941#define REGNODE_OFFSET(node) ((node) - RExC_emit_start)
942
538e84ed 943/* Macros for recording node offsets. 20001227 mjd@plover.com
fac92740
MJD
944 * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
945 * element 2*n-1 of the array. Element #2n holds the byte length node #n.
946 * Element 0 holds the number n.
07be1b83 947 * Position is 1 indexed.
fac92740 948 */
7122b237 949#ifndef RE_TRACK_PATTERN_OFFSETS
d7c442ce 950#define Set_Node_Offset_To_R(offset,byte)
7122b237
YO
951#define Set_Node_Offset(node,byte)
952#define Set_Cur_Node_Offset
953#define Set_Node_Length_To_R(node,len)
954#define Set_Node_Length(node,len)
6a86c6ad 955#define Set_Node_Cur_Length(node,start)
538e84ed
KW
956#define Node_Offset(n)
957#define Node_Length(n)
7122b237
YO
958#define Set_Node_Offset_Length(node,offset,len)
959#define ProgLen(ri) ri->u.proglen
960#define SetProgLen(ri,x) ri->u.proglen = x
4cbeff97 961#define Track_Code(code)
7122b237
YO
962#else
963#define ProgLen(ri) ri->u.offsets[0]
964#define SetProgLen(ri,x) ri->u.offsets[0] = x
d7c442ce 965#define Set_Node_Offset_To_R(offset,byte) STMT_START { \
ccb2c380 966 MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
d7c442ce
KW
967 __LINE__, (int)(offset), (int)(byte))); \
968 if((offset) < 0) { \
538e84ed 969 Perl_croak(aTHX_ "value of node is %d in Offset macro", \
d7c442ce 970 (int)(offset)); \
ccb2c380 971 } else { \
d7c442ce 972 RExC_offsets[2*(offset)-1] = (byte); \
ccb2c380 973 } \
ccb2c380
MP
974} STMT_END
975
88f063b4 976#define Set_Node_Offset(node,byte) \
1a322bb4 977 Set_Node_Offset_To_R(REGNODE_OFFSET(node), (byte)-RExC_start)
ccb2c380
MP
978#define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
979
980#define Set_Node_Length_To_R(node,len) STMT_START { \
ccb2c380 981 MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \
551405c4 982 __LINE__, (int)(node), (int)(len))); \
ccb2c380 983 if((node) < 0) { \
538e84ed
KW
984 Perl_croak(aTHX_ "value of node is %d in Length macro", \
985 (int)(node)); \
ccb2c380
MP
986 } else { \
987 RExC_offsets[2*(node)] = (len); \
988 } \
ccb2c380
MP
989} STMT_END
990
991#define Set_Node_Length(node,len) \
1a322bb4 992 Set_Node_Length_To_R(REGNODE_OFFSET(node), len)
6a86c6ad
NC
993#define Set_Node_Cur_Length(node, start) \
994 Set_Node_Length(node, RExC_parse - start)
fac92740
MJD
995
996/* Get offsets and lengths */
1a322bb4
KW
997#define Node_Offset(n) (RExC_offsets[2*(REGNODE_OFFSET(n))-1])
998#define Node_Length(n) (RExC_offsets[2*(REGNODE_OFFSET(n))])
fac92740 999
07be1b83 1000#define Set_Node_Offset_Length(node,offset,len) STMT_START { \
1a322bb4
KW
1001 Set_Node_Offset_To_R(REGNODE_OFFSET(node), (offset)); \
1002 Set_Node_Length_To_R(REGNODE_OFFSET(node), (len)); \
07be1b83 1003} STMT_END
4cbeff97
TC
1004
1005#define Track_Code(code) STMT_START { code } STMT_END
7122b237 1006#endif
07be1b83
YO
1007
1008#if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS
1009#define EXPERIMENTAL_INPLACESCAN
f427392e 1010#endif /*PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS*/
07be1b83 1011
cb41e5d6
YO
1012#ifdef DEBUGGING
1013int
6ad9a8ab 1014Perl_re_printf(pTHX_ const char *fmt, ...)
cb41e5d6
YO
1015{
1016 va_list ap;
1017 int result;
1018 PerlIO *f= Perl_debug_log;
1019 PERL_ARGS_ASSERT_RE_PRINTF;
1020 va_start(ap, fmt);
1021 result = PerlIO_vprintf(f, fmt, ap);
1022 va_end(ap);
1023 return result;
1024}
1025
1026int
7b031478 1027Perl_re_indentf(pTHX_ const char *fmt, U32 depth, ...)
cb41e5d6
YO
1028{
1029 va_list ap;
1030 int result;
1031 PerlIO *f= Perl_debug_log;
1032 PERL_ARGS_ASSERT_RE_INDENTF;
1033 va_start(ap, depth);
daeb874b 1034 PerlIO_printf(f, "%*s", ( (int)depth % 20 ) * 2, "");
cb41e5d6
YO
1035 result = PerlIO_vprintf(f, fmt, ap);
1036 va_end(ap);
1037 return result;
1038}
1039#endif /* DEBUGGING */
1040
7b031478 1041#define DEBUG_RExC_seen() \
538e84ed 1042 DEBUG_OPTIMISE_MORE_r({ \
88f063b4 1043 Perl_re_printf( aTHX_ "RExC_seen: "); \
538e84ed 1044 \
e384d5c1 1045 if (RExC_seen & REG_ZERO_LEN_SEEN) \
88f063b4 1046 Perl_re_printf( aTHX_ "REG_ZERO_LEN_SEEN "); \
538e84ed 1047 \
e384d5c1 1048 if (RExC_seen & REG_LOOKBEHIND_SEEN) \
88f063b4 1049 Perl_re_printf( aTHX_ "REG_LOOKBEHIND_SEEN "); \
538e84ed 1050 \
e384d5c1 1051 if (RExC_seen & REG_GPOS_SEEN) \
88f063b4 1052 Perl_re_printf( aTHX_ "REG_GPOS_SEEN "); \
538e84ed 1053 \
e384d5c1 1054 if (RExC_seen & REG_RECURSE_SEEN) \
88f063b4 1055 Perl_re_printf( aTHX_ "REG_RECURSE_SEEN "); \
538e84ed 1056 \
7b031478 1057 if (RExC_seen & REG_TOP_LEVEL_BRANCHES_SEEN) \
88f063b4 1058 Perl_re_printf( aTHX_ "REG_TOP_LEVEL_BRANCHES_SEEN "); \
538e84ed 1059 \
e384d5c1 1060 if (RExC_seen & REG_VERBARG_SEEN) \
88f063b4 1061 Perl_re_printf( aTHX_ "REG_VERBARG_SEEN "); \
538e84ed 1062 \
e384d5c1 1063 if (RExC_seen & REG_CUTGROUP_SEEN) \
88f063b4 1064 Perl_re_printf( aTHX_ "REG_CUTGROUP_SEEN "); \
538e84ed 1065 \
e384d5c1 1066 if (RExC_seen & REG_RUN_ON_COMMENT_SEEN) \
88f063b4 1067 Perl_re_printf( aTHX_ "REG_RUN_ON_COMMENT_SEEN "); \
538e84ed 1068 \
e384d5c1 1069 if (RExC_seen & REG_UNFOLDED_MULTI_SEEN) \
88f063b4 1070 Perl_re_printf( aTHX_ "REG_UNFOLDED_MULTI_SEEN "); \
538e84ed 1071 \
7b031478 1072 if (RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) \
88f063b4 1073 Perl_re_printf( aTHX_ "REG_UNBOUNDED_QUANTIFIER_SEEN "); \
ee273784 1074 \
88f063b4 1075 Perl_re_printf( aTHX_ "\n"); \
9e9ecfdf
YO
1076 });
1077
fdfb4f21 1078#define DEBUG_SHOW_STUDY_FLAG(flags,flag) \
6ad9a8ab 1079 if ((flags) & flag) Perl_re_printf( aTHX_ "%s ", #flag)
fdfb4f21 1080
fdfb4f21 1081
f5a36d78
DM
1082#ifdef DEBUGGING
1083static void
1084S_debug_show_study_flags(pTHX_ U32 flags, const char *open_str,
1085 const char *close_str)
1086{
1087 if (!flags)
1088 return;
1089
1090 Perl_re_printf( aTHX_ "%s", open_str);
11683ecb
DM
1091 DEBUG_SHOW_STUDY_FLAG(flags, SF_BEFORE_SEOL);
1092 DEBUG_SHOW_STUDY_FLAG(flags, SF_BEFORE_MEOL);
f5a36d78
DM
1093 DEBUG_SHOW_STUDY_FLAG(flags, SF_IS_INF);
1094 DEBUG_SHOW_STUDY_FLAG(flags, SF_HAS_PAR);
1095 DEBUG_SHOW_STUDY_FLAG(flags, SF_IN_PAR);
1096 DEBUG_SHOW_STUDY_FLAG(flags, SF_HAS_EVAL);
1097 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_SUBSTR);
1098 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_STCLASS_AND);
1099 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_STCLASS_OR);
1100 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_STCLASS);
1101 DEBUG_SHOW_STUDY_FLAG(flags, SCF_WHILEM_VISITED_POS);
1102 DEBUG_SHOW_STUDY_FLAG(flags, SCF_TRIE_RESTUDY);
1103 DEBUG_SHOW_STUDY_FLAG(flags, SCF_SEEN_ACCEPT);
1104 DEBUG_SHOW_STUDY_FLAG(flags, SCF_TRIE_DOING_RESTUDY);
1105 DEBUG_SHOW_STUDY_FLAG(flags, SCF_IN_DEFINE);
1106 Perl_re_printf( aTHX_ "%s", close_str);
1107}
1108
1109
1110static void
1111S_debug_studydata(pTHX_ const char *where, scan_data_t *data,
1112 U32 depth, int is_inf)
1113{
1114 GET_RE_DEBUG_FLAGS_DECL;
1115
1116 DEBUG_OPTIMISE_MORE_r({
1117 if (!data)
1118 return;
1119 Perl_re_indentf(aTHX_ "%s: Pos:%" IVdf "/%" IVdf " Flags: 0x%" UVXf,
1120 depth,
1121 where,
1122 (IV)data->pos_min,
1123 (IV)data->pos_delta,
1124 (UV)data->flags
1125 );
1126
1127 S_debug_show_study_flags(aTHX_ data->flags," [","]");
fdfb4f21 1128
f5a36d78
DM
1129 Perl_re_printf( aTHX_
1130 " Whilem_c: %" IVdf " Lcp: %" IVdf " %s",
1131 (IV)data->whilem_c,
1132 (IV)(data->last_closep ? *((data)->last_closep) : -1),
1133 is_inf ? "INF " : ""
1134 );
1135
11683ecb 1136 if (data->last_found) {
37b6262f 1137 int i;
11683ecb
DM
1138 Perl_re_printf(aTHX_
1139 "Last:'%s' %" IVdf ":%" IVdf "/%" IVdf,
1140 SvPVX_const(data->last_found),
1141 (IV)data->last_end,
1142 (IV)data->last_start_min,
1143 (IV)data->last_start_max
1144 );
1145
2099df82 1146 for (i = 0; i < 2; i++) {
37b6262f
DM
1147 Perl_re_printf(aTHX_
1148 " %s%s: '%s' @ %" IVdf "/%" IVdf,
2099df82
DM
1149 data->cur_is_floating == i ? "*" : "",
1150 i ? "Float" : "Fixed",
37b6262f
DM
1151 SvPVX_const(data->substrs[i].str),
1152 (IV)data->substrs[i].min_offset,
1153 (IV)data->substrs[i].max_offset
1154 );
1155 S_debug_show_study_flags(aTHX_ data->substrs[i].flags," [","]");
1156 }
11683ecb 1157 }
f5a36d78
DM
1158
1159 Perl_re_printf( aTHX_ "\n");
1160 });
1161}
1162
1163
1164static void
1165S_debug_peep(pTHX_ const char *str, const RExC_state_t *pRExC_state,
1166 regnode *scan, U32 depth, U32 flags)
1167{
1168 GET_RE_DEBUG_FLAGS_DECL;
1169
1170 DEBUG_OPTIMISE_r({
1171 regnode *Next;
1172
1173 if (!scan)
1174 return;
1175 Next = regnext(scan);
1176 regprop(RExC_rx, RExC_mysv, scan, NULL, pRExC_state);
1177 Perl_re_indentf( aTHX_ "%s>%3d: %s (%d)",
1178 depth,
1179 str,
1180 REG_NODE_NUM(scan), SvPV_nolen_const(RExC_mysv),
1181 Next ? (REG_NODE_NUM(Next)) : 0 );
1182 S_debug_show_study_flags(aTHX_ flags," [ ","]");
1183 Perl_re_printf( aTHX_ "\n");
1184 });
1185}
1186
1187
1188# define DEBUG_STUDYDATA(where, data, depth, is_inf) \
1189 S_debug_studydata(aTHX_ where, data, depth, is_inf)
1190
1191# define DEBUG_PEEP(str, scan, depth, flags) \
1192 S_debug_peep(aTHX_ str, pRExC_state, scan, depth, flags)
1193
1194#else
1195# define DEBUG_STUDYDATA(where, data, depth, is_inf) NOOP
1196# define DEBUG_PEEP(str, scan, depth, flags) NOOP
1197#endif
1de06328 1198
cb41e5d6 1199
c6871b76
KW
1200/* =========================================================
1201 * BEGIN edit_distance stuff.
1202 *
1203 * This calculates how many single character changes of any type are needed to
1204 * transform a string into another one. It is taken from version 3.1 of
1205 *
1206 * https://metacpan.org/pod/Text::Levenshtein::Damerau::XS
1207 */
1208
1209/* Our unsorted dictionary linked list. */
1210/* Note we use UVs, not chars. */
1211
1212struct dictionary{
1213 UV key;
1214 UV value;
1215 struct dictionary* next;
1216};
1217typedef struct dictionary item;
1218
1219
1220PERL_STATIC_INLINE item*
88f063b4 1221push(UV key, item* curr)
c6871b76
KW
1222{
1223 item* head;
d09f14bf 1224 Newx(head, 1, item);
c6871b76
KW
1225 head->key = key;
1226 head->value = 0;
1227 head->next = curr;
1228 return head;
1229}
1230
1231
1232PERL_STATIC_INLINE item*
1233find(item* head, UV key)
1234{
1235 item* iterator = head;
1236 while (iterator){
1237 if (iterator->key == key){
1238 return iterator;
1239 }
1240 iterator = iterator->next;
1241 }
1242
1243 return NULL;
1244}
1245
1246PERL_STATIC_INLINE item*
88f063b4 1247uniquePush(item* head, UV key)
c6871b76
KW
1248{
1249 item* iterator = head;
1250
1251 while (iterator){
1252 if (iterator->key == key) {
1253 return head;
1254 }
1255 iterator = iterator->next;
1256 }
1257
88f063b4 1258 return push(key, head);
c6871b76
KW
1259}
1260
1261PERL_STATIC_INLINE void
1262dict_free(item* head)
1263{
1264 item* iterator = head;
1265
1266 while (iterator) {
1267 item* temp = iterator;
1268 iterator = iterator->next;
1269 Safefree(temp);
1270 }
1271
1272 head = NULL;
1273}
1274
1275/* End of Dictionary Stuff */
1276
1277/* All calculations/work are done here */
1278STATIC int
1279S_edit_distance(const UV* src,
1280 const UV* tgt,
1281 const STRLEN x, /* length of src[] */
1282 const STRLEN y, /* length of tgt[] */
1283 const SSize_t maxDistance
1284)
1285{
1286 item *head = NULL;
88f063b4 1287 UV swapCount, swapScore, targetCharCount, i, j;
c6871b76
KW
1288 UV *scores;
1289 UV score_ceil = x + y;
1290
1291 PERL_ARGS_ASSERT_EDIT_DISTANCE;
1292
1293 /* intialize matrix start values */
d09f14bf 1294 Newx(scores, ( (x + 2) * (y + 2)), UV);
c6871b76
KW
1295 scores[0] = score_ceil;
1296 scores[1 * (y + 2) + 0] = score_ceil;
1297 scores[0 * (y + 2) + 1] = score_ceil;
1298 scores[1 * (y + 2) + 1] = 0;
88f063b4 1299 head = uniquePush(uniquePush(head, src[0]), tgt[0]);
c6871b76
KW
1300
1301 /* work loops */
1302 /* i = src index */
1303 /* j = tgt index */
1304 for (i=1;i<=x;i++) {
1305 if (i < x)
88f063b4 1306 head = uniquePush(head, src[i]);
c6871b76
KW
1307 scores[(i+1) * (y + 2) + 1] = i;
1308 scores[(i+1) * (y + 2) + 0] = score_ceil;
1309 swapCount = 0;
1310
1311 for (j=1;j<=y;j++) {
1312 if (i == 1) {
1313 if(j < y)
88f063b4 1314 head = uniquePush(head, tgt[j]);
c6871b76
KW
1315 scores[1 * (y + 2) + (j + 1)] = j;
1316 scores[0 * (y + 2) + (j + 1)] = score_ceil;
1317 }
1318
88f063b4 1319 targetCharCount = find(head, tgt[j-1])->value;
c6871b76
KW
1320 swapScore = scores[targetCharCount * (y + 2) + swapCount] + i - targetCharCount - 1 + j - swapCount;
1321
1322 if (src[i-1] != tgt[j-1]){
1323 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));
1324 }
1325 else {
1326 swapCount = j;
1327 scores[(i+1) * (y + 2) + (j + 1)] = MIN(scores[i * (y + 2) + j], swapScore);
1328 }
1329 }
1330
88f063b4 1331 find(head, src[i-1])->value = i;
c6871b76
KW
1332 }
1333
1334 {
1335 IV score = scores[(x+1) * (y + 2) + (y + 1)];
1336 dict_free(head);
1337 Safefree(scores);
1338 return (maxDistance != 0 && maxDistance < score)?(-1):score;
1339 }
1340}
1341
1342/* END of edit_distance() stuff
1343 * ========================================================= */
1344
8e35b056
KW
1345/* is c a control character for which we have a mnemonic? */
1346#define isMNEMONIC_CNTRL(c) _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
1347
549b4e78
KW
1348STATIC const char *
1349S_cntrl_to_mnemonic(const U8 c)
1350{
1351 /* Returns the mnemonic string that represents character 'c', if one
1352 * exists; NULL otherwise. The only ones that exist for the purposes of
1353 * this routine are a few control characters */
1354
1355 switch (c) {
1356 case '\a': return "\\a";
1357 case '\b': return "\\b";
1358 case ESC_NATIVE: return "\\e";
1359 case '\f': return "\\f";
1360 case '\n': return "\\n";
1361 case '\r': return "\\r";
1362 case '\t': return "\\t";
1363 }
1364
1365 return NULL;
1366}
1367
653099ff 1368/* Mark that we cannot extend a found fixed substring at this point.
37b6262f 1369 Update the longest found anchored substring or the longest found
653099ff
GS
1370 floating substrings if needed. */
1371
4327152a 1372STATIC void
ea3daa5d
FC
1373S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data,
1374 SSize_t *minlenp, int is_inf)
c277df42 1375{
e1ec3a88 1376 const STRLEN l = CHR_SVLEN(data->last_found);
2099df82 1377 SV * const longest_sv = data->substrs[data->cur_is_floating].str;
a2ca2017 1378 const STRLEN old_l = CHR_SVLEN(longest_sv);
1de06328 1379 GET_RE_DEBUG_FLAGS_DECL;
b81d288d 1380
7918f24d
NC
1381 PERL_ARGS_ASSERT_SCAN_COMMIT;
1382
c277df42 1383 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
2099df82 1384 const U8 i = data->cur_is_floating;
a2ca2017 1385 SvSetMagicSV(longest_sv, data->last_found);
37b6262f
DM
1386 data->substrs[i].min_offset = l ? data->last_start_min : data->pos_min;
1387
2099df82
DM
1388 if (!i) /* fixed */
1389 data->substrs[0].max_offset = data->substrs[0].min_offset;
1390 else { /* float */
1391 data->substrs[1].max_offset = (l
646e8787
DM
1392 ? data->last_start_max
1393 : (data->pos_delta > SSize_t_MAX - data->pos_min
ea3daa5d
FC
1394 ? SSize_t_MAX
1395 : data->pos_min + data->pos_delta));
1396 if (is_inf
2099df82
DM
1397 || (STRLEN)data->substrs[1].max_offset > (STRLEN)SSize_t_MAX)
1398 data->substrs[1].max_offset = SSize_t_MAX;
37b6262f 1399 }
11683ecb 1400
37b6262f
DM
1401 if (data->flags & SF_BEFORE_EOL)
1402 data->substrs[i].flags |= (data->flags & SF_BEFORE_EOL);
1403 else
1404 data->substrs[i].flags &= ~SF_BEFORE_EOL;
1405 data->substrs[i].minlenp = minlenp;
1406 data->substrs[i].lookbehind = 0;
c277df42 1407 }
37b6262f 1408
c277df42 1409 SvCUR_set(data->last_found, 0);
0eda9292 1410 {
a28509cc 1411 SV * const sv = data->last_found;
097eb12c
AL
1412 if (SvUTF8(sv) && SvMAGICAL(sv)) {
1413 MAGIC * const mg = mg_find(sv, PERL_MAGIC_utf8);
1414 if (mg)
1415 mg->mg_len = 0;
1416 }
0eda9292 1417 }
c277df42
IZ
1418 data->last_end = -1;
1419 data->flags &= ~SF_BEFORE_EOL;
f5a36d78 1420 DEBUG_STUDYDATA("commit", data, 0, is_inf);
c277df42
IZ
1421}
1422
cdd87c1d
KW
1423/* An SSC is just a regnode_charclass_posix with an extra field: the inversion
1424 * list that describes which code points it matches */
1425
653099ff 1426STATIC void
3420edd7 1427S_ssc_anything(pTHX_ regnode_ssc *ssc)
653099ff 1428{
cdd87c1d
KW
1429 /* Set the SSC 'ssc' to match an empty string or any code point */
1430
557bd3fb 1431 PERL_ARGS_ASSERT_SSC_ANYTHING;
7918f24d 1432
71068078 1433 assert(is_ANYOF_SYNTHETIC(ssc));
3fffb88a 1434
0854ea0b
KW
1435 /* mortalize so won't leak */
1436 ssc->invlist = sv_2mortal(_add_range_to_invlist(NULL, 0, UV_MAX));
93e92956 1437 ANYOF_FLAGS(ssc) |= SSC_MATCHES_EMPTY_STRING; /* Plus matches empty */
653099ff
GS
1438}
1439
653099ff 1440STATIC int
dc3bf405 1441S_ssc_is_anything(const regnode_ssc *ssc)
653099ff 1442{
c144baaa
KW
1443 /* Returns TRUE if the SSC 'ssc' can match the empty string and any code
1444 * point; FALSE otherwise. Thus, this is used to see if using 'ssc' buys
1445 * us anything: if the function returns TRUE, 'ssc' hasn't been restricted
1446 * in any way, so there's no point in using it */
cdd87c1d
KW
1447
1448 UV start, end;
1449 bool ret;
653099ff 1450
557bd3fb 1451 PERL_ARGS_ASSERT_SSC_IS_ANYTHING;
7918f24d 1452
71068078 1453 assert(is_ANYOF_SYNTHETIC(ssc));
cdd87c1d 1454
93e92956 1455 if (! (ANYOF_FLAGS(ssc) & SSC_MATCHES_EMPTY_STRING)) {
cdd87c1d
KW
1456 return FALSE;
1457 }
1458
1459 /* See if the list consists solely of the range 0 - Infinity */
1460 invlist_iterinit(ssc->invlist);
1461 ret = invlist_iternext(ssc->invlist, &start, &end)
1462 && start == 0
1463 && end == UV_MAX;
1464
1465 invlist_iterfinish(ssc->invlist);
1466
1467 if (ret) {
1468 return TRUE;
1469 }
1470
1471 /* If e.g., both \w and \W are set, matches everything */
e0e1be5f 1472 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
cdd87c1d
KW
1473 int i;
1474 for (i = 0; i < ANYOF_POSIXL_MAX; i += 2) {
1475 if (ANYOF_POSIXL_TEST(ssc, i) && ANYOF_POSIXL_TEST(ssc, i+1)) {
1476 return TRUE;
1477 }
1478 }
1479 }
1480
1481 return FALSE;
653099ff
GS
1482}
1483
653099ff 1484STATIC void
cdd87c1d 1485S_ssc_init(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc)
653099ff 1486{
cdd87c1d
KW
1487 /* Initializes the SSC 'ssc'. This includes setting it to match an empty
1488 * string, any code point, or any posix class under locale */
1489
557bd3fb 1490 PERL_ARGS_ASSERT_SSC_INIT;
7918f24d 1491
557bd3fb 1492 Zero(ssc, 1, regnode_ssc);
71068078 1493 set_ANYOF_SYNTHETIC(ssc);
93e92956 1494 ARG_SET(ssc, ANYOF_ONLY_HAS_BITMAP);
3420edd7 1495 ssc_anything(ssc);
cdd87c1d 1496
2f306ab9
KW
1497 /* If any portion of the regex is to operate under locale rules that aren't
1498 * fully known at compile time, initialization includes it. The reason
1499 * this isn't done for all regexes is that the optimizer was written under
1500 * the assumption that locale was all-or-nothing. Given the complexity and
1501 * lack of documentation in the optimizer, and that there are inadequate
1502 * test cases for locale, many parts of it may not work properly, it is
1503 * safest to avoid locale unless necessary. */
cdd87c1d
KW
1504 if (RExC_contains_locale) {
1505 ANYOF_POSIXL_SETALL(ssc);
cdd87c1d
KW
1506 }
1507 else {
1508 ANYOF_POSIXL_ZERO(ssc);
1509 }
653099ff
GS
1510}
1511
b423522f 1512STATIC int
dc3bf405
BF
1513S_ssc_is_cp_posixl_init(const RExC_state_t *pRExC_state,
1514 const regnode_ssc *ssc)
b423522f
KW
1515{
1516 /* Returns TRUE if the SSC 'ssc' is in its initial state with regard only
1517 * to the list of code points matched, and locale posix classes; hence does
1518 * not check its flags) */
1519
1520 UV start, end;
1521 bool ret;
1522
1523 PERL_ARGS_ASSERT_SSC_IS_CP_POSIXL_INIT;
1524
71068078 1525 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
1526
1527 invlist_iterinit(ssc->invlist);
1528 ret = invlist_iternext(ssc->invlist, &start, &end)
1529 && start == 0
1530 && end == UV_MAX;
1531
1532 invlist_iterfinish(ssc->invlist);
1533
1534 if (! ret) {
1535 return FALSE;
1536 }
1537
e0e1be5f 1538 if (RExC_contains_locale && ! ANYOF_POSIXL_SSC_TEST_ALL_SET(ssc)) {
31f05a37 1539 return FALSE;
b423522f
KW
1540 }
1541
1542 return TRUE;
1543}
1544
4ebed06a
KW
1545#define INVLIST_INDEX 0
1546#define ONLY_LOCALE_MATCHES_INDEX 1
1547#define DEFERRED_USER_DEFINED_INDEX 2
1548
b423522f
KW
1549STATIC SV*
1550S_get_ANYOF_cp_list_for_ssc(pTHX_ const RExC_state_t *pRExC_state,
5c0f85ef 1551 const regnode_charclass* const node)
b423522f
KW
1552{
1553 /* Returns a mortal inversion list defining which code points are matched
1554 * by 'node', which is of type ANYOF. Handles complementing the result if
1555 * appropriate. If some code points aren't knowable at this time, the
31f05a37
KW
1556 * returned list must, and will, contain every code point that is a
1557 * possibility. */
b423522f 1558
1565c085 1559 dVAR;
e2506fa7 1560 SV* invlist = NULL;
1ee208c4 1561 SV* only_utf8_locale_invlist = NULL;
b423522f
KW
1562 unsigned int i;
1563 const U32 n = ARG(node);
31f05a37 1564 bool new_node_has_latin1 = FALSE;
b423522f
KW
1565
1566 PERL_ARGS_ASSERT_GET_ANYOF_CP_LIST_FOR_SSC;
1567
1568 /* Look at the data structure created by S_set_ANYOF_arg() */
93e92956 1569 if (n != ANYOF_ONLY_HAS_BITMAP) {
b423522f
KW
1570 SV * const rv = MUTABLE_SV(RExC_rxi->data->data[n]);
1571 AV * const av = MUTABLE_AV(SvRV(rv));
1572 SV **const ary = AvARRAY(av);
1573 assert(RExC_rxi->data->what[n] == 's');
1574
4ebed06a 1575 if (av_tindex_skip_len_mg(av) >= DEFERRED_USER_DEFINED_INDEX) {
b423522f 1576
6ed02bb6
KW
1577 /* Here there are things that won't be known until runtime -- we
1578 * have to assume it could be anything */
e2506fa7 1579 invlist = sv_2mortal(_new_invlist(1));
b423522f
KW
1580 return _add_range_to_invlist(invlist, 0, UV_MAX);
1581 }
4ebed06a 1582 else if (ary[INVLIST_INDEX]) {
b423522f 1583
6ed02bb6 1584 /* Use the node's inversion list */
4ebed06a 1585 invlist = sv_2mortal(invlist_clone(ary[INVLIST_INDEX], NULL));
1ee208c4
KW
1586 }
1587
1588 /* Get the code points valid only under UTF-8 locales */
6ed02bb6 1589 if ( (ANYOF_FLAGS(node) & ANYOFL_FOLD)
4ebed06a 1590 && av_tindex_skip_len_mg(av) >= ONLY_LOCALE_MATCHES_INDEX)
1ee208c4 1591 {
4ebed06a 1592 only_utf8_locale_invlist = ary[ONLY_LOCALE_MATCHES_INDEX];
b423522f
KW
1593 }
1594 }
1595
e2506fa7
KW
1596 if (! invlist) {
1597 invlist = sv_2mortal(_new_invlist(0));
1598 }
1599
dcb20b36
KW
1600 /* An ANYOF node contains a bitmap for the first NUM_ANYOF_CODE_POINTS
1601 * code points, and an inversion list for the others, but if there are code
1602 * points that should match only conditionally on the target string being
1603 * UTF-8, those are placed in the inversion list, and not the bitmap.
1604 * Since there are circumstances under which they could match, they are
1605 * included in the SSC. But if the ANYOF node is to be inverted, we have
1606 * to exclude them here, so that when we invert below, the end result
1607 * actually does include them. (Think about "\xe0" =~ /[^\xc0]/di;). We
1608 * have to do this here before we add the unconditionally matched code
1609 * points */
b423522f
KW
1610 if (ANYOF_FLAGS(node) & ANYOF_INVERT) {
1611 _invlist_intersection_complement_2nd(invlist,
1612 PL_UpperLatin1,
1613 &invlist);
1614 }
1615
1616 /* Add in the points from the bit map */
5aa750d1 1617 if (OP(node) != ANYOFH) {
90973738
KW
1618 for (i = 0; i < NUM_ANYOF_CODE_POINTS; i++) {
1619 if (ANYOF_BITMAP_TEST(node, i)) {
1620 unsigned int start = i++;
1621
1622 for (; i < NUM_ANYOF_CODE_POINTS
1623 && ANYOF_BITMAP_TEST(node, i); ++i)
1624 {
1625 /* empty */
1626 }
1627 invlist = _add_range_to_invlist(invlist, start, i-1);
1628 new_node_has_latin1 = TRUE;
6f8848d5 1629 }
b423522f
KW
1630 }
1631 }
1632
1633 /* If this can match all upper Latin1 code points, have to add them
ac33c516
KW
1634 * as well. But don't add them if inverting, as when that gets done below,
1635 * it would exclude all these characters, including the ones it shouldn't
1636 * that were added just above */
1637 if (! (ANYOF_FLAGS(node) & ANYOF_INVERT) && OP(node) == ANYOFD
f240c685
KW
1638 && (ANYOF_FLAGS(node) & ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER))
1639 {
b423522f
KW
1640 _invlist_union(invlist, PL_UpperLatin1, &invlist);
1641 }
1642
1643 /* Similarly for these */
93e92956 1644 if (ANYOF_FLAGS(node) & ANYOF_MATCHES_ALL_ABOVE_BITMAP) {
e0a1ff7a 1645 _invlist_union_complement_2nd(invlist, PL_InBitmap, &invlist);
b423522f
KW
1646 }
1647
1648 if (ANYOF_FLAGS(node) & ANYOF_INVERT) {
1649 _invlist_invert(invlist);
1650 }
35b8412f
KW
1651 else if (ANYOF_FLAGS(node) & ANYOFL_FOLD) {
1652 if (new_node_has_latin1) {
31f05a37 1653
26be5fe6
KW
1654 /* Under /li, any 0-255 could fold to any other 0-255, depending on
1655 * the locale. We can skip this if there are no 0-255 at all. */
1656 _invlist_union(invlist, PL_Latin1, &invlist);
35b8412f
KW
1657
1658 invlist = add_cp_to_invlist(invlist, LATIN_SMALL_LETTER_DOTLESS_I);
1659 invlist = add_cp_to_invlist(invlist, LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE);
1660 }
1661 else {
1662 if (_invlist_contains_cp(invlist, LATIN_SMALL_LETTER_DOTLESS_I)) {
1663 invlist = add_cp_to_invlist(invlist, 'I');
1664 }
1665 if (_invlist_contains_cp(invlist,
1666 LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE))
1667 {
1668 invlist = add_cp_to_invlist(invlist, 'i');
1669 }
1670 }
31f05a37
KW
1671 }
1672
1ee208c4
KW
1673 /* Similarly add the UTF-8 locale possible matches. These have to be
1674 * deferred until after the non-UTF-8 locale ones are taken care of just
1675 * above, or it leads to wrong results under ANYOF_INVERT */
1676 if (only_utf8_locale_invlist) {
31f05a37 1677 _invlist_union_maybe_complement_2nd(invlist,
1ee208c4 1678 only_utf8_locale_invlist,
31f05a37
KW
1679 ANYOF_FLAGS(node) & ANYOF_INVERT,
1680 &invlist);
1681 }
b423522f
KW
1682
1683 return invlist;
1684}
1685
1051e1c4 1686/* These two functions currently do the exact same thing */
557bd3fb 1687#define ssc_init_zero ssc_init
653099ff 1688
cdd87c1d
KW
1689#define ssc_add_cp(ssc, cp) ssc_add_range((ssc), (cp), (cp))
1690#define ssc_match_all_cp(ssc) ssc_add_range(ssc, 0, UV_MAX)
1691
557bd3fb 1692/* 'AND' a given class with another one. Can create false positives. 'ssc'
93e92956
KW
1693 * should not be inverted. 'and_with->flags & ANYOF_MATCHES_POSIXL' should be
1694 * 0 if 'and_with' is a regnode_charclass instead of a regnode_ssc. */
cdd87c1d 1695
653099ff 1696STATIC void
b423522f 1697S_ssc_and(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc,
7dcac5f6 1698 const regnode_charclass *and_with)
653099ff 1699{
cdd87c1d
KW
1700 /* Accumulate into SSC 'ssc' its 'AND' with 'and_with', which is either
1701 * another SSC or a regular ANYOF class. Can create false positives. */
40d049e4 1702
a0dd4231
KW
1703 SV* anded_cp_list;
1704 U8 anded_flags;
1e6ade67 1705
cdd87c1d 1706 PERL_ARGS_ASSERT_SSC_AND;
653099ff 1707
71068078 1708 assert(is_ANYOF_SYNTHETIC(ssc));
a0dd4231
KW
1709
1710 /* 'and_with' is used as-is if it too is an SSC; otherwise have to extract
1711 * the code point inversion list and just the relevant flags */
71068078 1712 if (is_ANYOF_SYNTHETIC(and_with)) {
7dcac5f6 1713 anded_cp_list = ((regnode_ssc *)and_with)->invlist;
a0dd4231 1714 anded_flags = ANYOF_FLAGS(and_with);
e9b08962
KW
1715
1716 /* XXX This is a kludge around what appears to be deficiencies in the
1717 * optimizer. If we make S_ssc_anything() add in the WARN_SUPER flag,
1718 * there are paths through the optimizer where it doesn't get weeded
1719 * out when it should. And if we don't make some extra provision for
1720 * it like the code just below, it doesn't get added when it should.
1721 * This solution is to add it only when AND'ing, which is here, and
1722 * only when what is being AND'ed is the pristine, original node
1723 * matching anything. Thus it is like adding it to ssc_anything() but
1724 * only when the result is to be AND'ed. Probably the same solution
1725 * could be adopted for the same problem we have with /l matching,
1726 * which is solved differently in S_ssc_init(), and that would lead to
1727 * fewer false positives than that solution has. But if this solution
1728 * creates bugs, the consequences are only that a warning isn't raised
1729 * that should be; while the consequences for having /l bugs is
1730 * incorrect matches */
7dcac5f6 1731 if (ssc_is_anything((regnode_ssc *)and_with)) {
f240c685 1732 anded_flags |= ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER;
e9b08962 1733 }
a0dd4231
KW
1734 }
1735 else {
5c0f85ef 1736 anded_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, and_with);
f240c685
KW
1737 if (OP(and_with) == ANYOFD) {
1738 anded_flags = ANYOF_FLAGS(and_with) & ANYOF_COMMON_FLAGS;
1739 }
1740 else {
1741 anded_flags = ANYOF_FLAGS(and_with)
1742 &( ANYOF_COMMON_FLAGS
108316fb
KW
1743 |ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
1744 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP);
d1c40ef5
KW
1745 if (ANYOFL_UTF8_LOCALE_REQD(ANYOF_FLAGS(and_with))) {
1746 anded_flags &=
1747 ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
1748 }
f240c685 1749 }
a0dd4231
KW
1750 }
1751
1752 ANYOF_FLAGS(ssc) &= anded_flags;
cdd87c1d
KW
1753
1754 /* Below, C1 is the list of code points in 'ssc'; P1, its posix classes.
1755 * C2 is the list of code points in 'and-with'; P2, its posix classes.
1756 * 'and_with' may be inverted. When not inverted, we have the situation of
1757 * computing:
1758 * (C1 | P1) & (C2 | P2)
1759 * = (C1 & (C2 | P2)) | (P1 & (C2 | P2))
1760 * = ((C1 & C2) | (C1 & P2)) | ((P1 & C2) | (P1 & P2))
1761 * <= ((C1 & C2) | P2)) | ( P1 | (P1 & P2))
1762 * <= ((C1 & C2) | P1 | P2)
1763 * Alternatively, the last few steps could be:
1764 * = ((C1 & C2) | (C1 & P2)) | ((P1 & C2) | (P1 & P2))
1765 * <= ((C1 & C2) | C1 ) | ( C2 | (P1 & P2))
1766 * <= (C1 | C2 | (P1 & P2))
1767 * We favor the second approach if either P1 or P2 is non-empty. This is
1768 * because these components are a barrier to doing optimizations, as what
1769 * they match cannot be known until the moment of matching as they are
1770 * dependent on the current locale, 'AND"ing them likely will reduce or
1771 * eliminate them.
1772 * But we can do better if we know that C1,P1 are in their initial state (a
1773 * frequent occurrence), each matching everything:
1774 * (<everything>) & (C2 | P2) = C2 | P2
1775 * Similarly, if C2,P2 are in their initial state (again a frequent
1776 * occurrence), the result is a no-op
1777 * (C1 | P1) & (<everything>) = C1 | P1
1778 *
1779 * Inverted, we have
1780 * (C1 | P1) & ~(C2 | P2) = (C1 | P1) & (~C2 & ~P2)
1781 * = (C1 & (~C2 & ~P2)) | (P1 & (~C2 & ~P2))
1782 * <= (C1 & ~C2) | (P1 & ~P2)
1783 * */
1aa99e6b 1784
a0dd4231 1785 if ((ANYOF_FLAGS(and_with) & ANYOF_INVERT)
71068078 1786 && ! is_ANYOF_SYNTHETIC(and_with))
a0dd4231 1787 {
cdd87c1d 1788 unsigned int i;
8951c461 1789
cdd87c1d
KW
1790 ssc_intersection(ssc,
1791 anded_cp_list,
1792 FALSE /* Has already been inverted */
1793 );
c6b76537 1794
cdd87c1d
KW
1795 /* If either P1 or P2 is empty, the intersection will be also; can skip
1796 * the loop */
93e92956 1797 if (! (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL)) {
cdd87c1d
KW
1798 ANYOF_POSIXL_ZERO(ssc);
1799 }
e0e1be5f 1800 else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
cdd87c1d
KW
1801
1802 /* Note that the Posix class component P from 'and_with' actually
1803 * looks like:
1804 * P = Pa | Pb | ... | Pn
1805 * where each component is one posix class, such as in [\w\s].
1806 * Thus
1807 * ~P = ~(Pa | Pb | ... | Pn)
1808 * = ~Pa & ~Pb & ... & ~Pn
1809 * <= ~Pa | ~Pb | ... | ~Pn
1810 * The last is something we can easily calculate, but unfortunately
1811 * is likely to have many false positives. We could do better
1812 * in some (but certainly not all) instances if two classes in
1813 * P have known relationships. For example
1814 * :lower: <= :alpha: <= :alnum: <= \w <= :graph: <= :print:
1815 * So
1816 * :lower: & :print: = :lower:
1817 * And similarly for classes that must be disjoint. For example,
1818 * since \s and \w can have no elements in common based on rules in
1819 * the POSIX standard,
1820 * \w & ^\S = nothing
1821 * Unfortunately, some vendor locales do not meet the Posix
1822 * standard, in particular almost everything by Microsoft.
1823 * The loop below just changes e.g., \w into \W and vice versa */
1824
1ee208c4 1825 regnode_charclass_posixl temp;
cdd87c1d
KW
1826 int add = 1; /* To calculate the index of the complement */
1827
b1234259 1828 Zero(&temp, 1, regnode_charclass_posixl);
cdd87c1d
KW
1829 ANYOF_POSIXL_ZERO(&temp);
1830 for (i = 0; i < ANYOF_MAX; i++) {
1831 assert(i % 2 != 0
7dcac5f6
KW
1832 || ! ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i)
1833 || ! ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i + 1));
cdd87c1d 1834
7dcac5f6 1835 if (ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i)) {
cdd87c1d
KW
1836 ANYOF_POSIXL_SET(&temp, i + add);
1837 }
1838 add = 0 - add; /* 1 goes to -1; -1 goes to 1 */
1839 }
1840 ANYOF_POSIXL_AND(&temp, ssc);
c6b76537 1841
cdd87c1d
KW
1842 } /* else ssc already has no posixes */
1843 } /* else: Not inverted. This routine is a no-op if 'and_with' is an SSC
1844 in its initial state */
71068078 1845 else if (! is_ANYOF_SYNTHETIC(and_with)
7dcac5f6 1846 || ! ssc_is_cp_posixl_init(pRExC_state, (regnode_ssc *)and_with))
cdd87c1d
KW
1847 {
1848 /* But if 'ssc' is in its initial state, the result is just 'and_with';
1849 * copy it over 'ssc' */
1850 if (ssc_is_cp_posixl_init(pRExC_state, ssc)) {
71068078 1851 if (is_ANYOF_SYNTHETIC(and_with)) {
cdd87c1d
KW
1852 StructCopy(and_with, ssc, regnode_ssc);
1853 }
1854 else {
1855 ssc->invlist = anded_cp_list;
1856 ANYOF_POSIXL_ZERO(ssc);
93e92956 1857 if (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL) {
7dcac5f6 1858 ANYOF_POSIXL_OR((regnode_charclass_posixl*) and_with, ssc);
cdd87c1d
KW
1859 }
1860 }
1861 }
e0e1be5f 1862 else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)
93e92956 1863 || (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL))
cdd87c1d
KW
1864 {
1865 /* One or the other of P1, P2 is non-empty. */
93e92956 1866 if (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL) {
1ea8b7fe
KW
1867 ANYOF_POSIXL_AND((regnode_charclass_posixl*) and_with, ssc);
1868 }
cdd87c1d
KW
1869 ssc_union(ssc, anded_cp_list, FALSE);
1870 }
1871 else { /* P1 = P2 = empty */
1872 ssc_intersection(ssc, anded_cp_list, FALSE);
1873 }
137165a6 1874 }
653099ff
GS
1875}
1876
653099ff 1877STATIC void
cdd87c1d 1878S_ssc_or(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc,
7dcac5f6 1879 const regnode_charclass *or_with)
653099ff 1880{
cdd87c1d
KW
1881 /* Accumulate into SSC 'ssc' its 'OR' with 'or_with', which is either
1882 * another SSC or a regular ANYOF class. Can create false positives if
1883 * 'or_with' is to be inverted. */
7918f24d 1884
a0dd4231
KW
1885 SV* ored_cp_list;
1886 U8 ored_flags;
c6b76537 1887
cdd87c1d 1888 PERL_ARGS_ASSERT_SSC_OR;
c6b76537 1889
71068078 1890 assert(is_ANYOF_SYNTHETIC(ssc));
a0dd4231
KW
1891
1892 /* 'or_with' is used as-is if it too is an SSC; otherwise have to extract
1893 * the code point inversion list and just the relevant flags */
71068078 1894 if (is_ANYOF_SYNTHETIC(or_with)) {
7dcac5f6 1895 ored_cp_list = ((regnode_ssc*) or_with)->invlist;
a0dd4231
KW
1896 ored_flags = ANYOF_FLAGS(or_with);
1897 }
1898 else {
5c0f85ef 1899 ored_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, or_with);
eff8b7dc 1900 ored_flags = ANYOF_FLAGS(or_with) & ANYOF_COMMON_FLAGS;
f240c685
KW
1901 if (OP(or_with) != ANYOFD) {
1902 ored_flags
1903 |= ANYOF_FLAGS(or_with)
108316fb
KW
1904 & ( ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
1905 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP);
d1c40ef5
KW
1906 if (ANYOFL_UTF8_LOCALE_REQD(ANYOF_FLAGS(or_with))) {
1907 ored_flags |=
1908 ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
1909 }
f240c685 1910 }
a0dd4231
KW
1911 }
1912
1913 ANYOF_FLAGS(ssc) |= ored_flags;
cdd87c1d
KW
1914
1915 /* Below, C1 is the list of code points in 'ssc'; P1, its posix classes.
1916 * C2 is the list of code points in 'or-with'; P2, its posix classes.
1917 * 'or_with' may be inverted. When not inverted, we have the simple
1918 * situation of computing:
1919 * (C1 | P1) | (C2 | P2) = (C1 | C2) | (P1 | P2)
1920 * If P1|P2 yields a situation with both a class and its complement are
1921 * set, like having both \w and \W, this matches all code points, and we
1922 * can delete these from the P component of the ssc going forward. XXX We
1923 * might be able to delete all the P components, but I (khw) am not certain
1924 * about this, and it is better to be safe.
1925 *
1926 * Inverted, we have
1927 * (C1 | P1) | ~(C2 | P2) = (C1 | P1) | (~C2 & ~P2)
1928 * <= (C1 | P1) | ~C2
1929 * <= (C1 | ~C2) | P1
1930 * (which results in actually simpler code than the non-inverted case)
1931 * */
9826f543 1932
a0dd4231 1933 if ((ANYOF_FLAGS(or_with) & ANYOF_INVERT)
71068078 1934 && ! is_ANYOF_SYNTHETIC(or_with))
a0dd4231 1935 {
cdd87c1d 1936 /* We ignore P2, leaving P1 going forward */
1ea8b7fe 1937 } /* else Not inverted */
93e92956 1938 else if (ANYOF_FLAGS(or_with) & ANYOF_MATCHES_POSIXL) {
7dcac5f6 1939 ANYOF_POSIXL_OR((regnode_charclass_posixl*)or_with, ssc);
e0e1be5f 1940 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
cdd87c1d
KW
1941 unsigned int i;
1942 for (i = 0; i < ANYOF_MAX; i += 2) {
1943 if (ANYOF_POSIXL_TEST(ssc, i) && ANYOF_POSIXL_TEST(ssc, i + 1))
1944 {
1945 ssc_match_all_cp(ssc);
1946 ANYOF_POSIXL_CLEAR(ssc, i);
1947 ANYOF_POSIXL_CLEAR(ssc, i+1);
cdd87c1d
KW
1948 }
1949 }
1950 }
1aa99e6b 1951 }
cdd87c1d
KW
1952
1953 ssc_union(ssc,
1954 ored_cp_list,
1955 FALSE /* Already has been inverted */
1956 );
653099ff
GS
1957}
1958
b423522f
KW
1959PERL_STATIC_INLINE void
1960S_ssc_union(pTHX_ regnode_ssc *ssc, SV* const invlist, const bool invert2nd)
1961{
1962 PERL_ARGS_ASSERT_SSC_UNION;
1963
71068078 1964 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
1965
1966 _invlist_union_maybe_complement_2nd(ssc->invlist,
1967 invlist,
1968 invert2nd,
1969 &ssc->invlist);
1970}
1971
1972PERL_STATIC_INLINE void
1973S_ssc_intersection(pTHX_ regnode_ssc *ssc,
1974 SV* const invlist,
1975 const bool invert2nd)
1976{
1977 PERL_ARGS_ASSERT_SSC_INTERSECTION;
1978
71068078 1979 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
1980
1981 _invlist_intersection_maybe_complement_2nd(ssc->invlist,
1982 invlist,
1983 invert2nd,
1984 &ssc->invlist);
1985}
1986
1987PERL_STATIC_INLINE void
1988S_ssc_add_range(pTHX_ regnode_ssc *ssc, const UV start, const UV end)
1989{
1990 PERL_ARGS_ASSERT_SSC_ADD_RANGE;
1991
71068078 1992 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
1993
1994 ssc->invlist = _add_range_to_invlist(ssc->invlist, start, end);
1995}
1996
1997PERL_STATIC_INLINE void
1998S_ssc_cp_and(pTHX_ regnode_ssc *ssc, const UV cp)
1999{
2000 /* AND just the single code point 'cp' into the SSC 'ssc' */
2001
2002 SV* cp_list = _new_invlist(2);
2003
2004 PERL_ARGS_ASSERT_SSC_CP_AND;
2005
71068078 2006 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
2007
2008 cp_list = add_cp_to_invlist(cp_list, cp);
2009 ssc_intersection(ssc, cp_list,
2010 FALSE /* Not inverted */
2011 );
2012 SvREFCNT_dec_NN(cp_list);
2013}
2014
2015PERL_STATIC_INLINE void
dc3bf405 2016S_ssc_clear_locale(regnode_ssc *ssc)
b423522f
KW
2017{
2018 /* Set the SSC 'ssc' to not match any locale things */
b423522f
KW
2019 PERL_ARGS_ASSERT_SSC_CLEAR_LOCALE;
2020
71068078 2021 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f
KW
2022
2023 ANYOF_POSIXL_ZERO(ssc);
2024 ANYOF_FLAGS(ssc) &= ~ANYOF_LOCALE_FLAGS;
2025}
2026
b35552de
KW
2027#define NON_OTHER_COUNT NON_OTHER_COUNT_FOR_USE_ONLY_BY_REGCOMP_DOT_C
2028
2029STATIC bool
2030S_is_ssc_worth_it(const RExC_state_t * pRExC_state, const regnode_ssc * ssc)
2031{
2032 /* The synthetic start class is used to hopefully quickly winnow down
2033 * places where a pattern could start a match in the target string. If it
2034 * doesn't really narrow things down that much, there isn't much point to
2035 * having the overhead of using it. This function uses some very crude
2036 * heuristics to decide if to use the ssc or not.
2037 *
2038 * It returns TRUE if 'ssc' rules out more than half what it considers to
2039 * be the "likely" possible matches, but of course it doesn't know what the
2040 * actual things being matched are going to be; these are only guesses
2041 *
2042 * For /l matches, it assumes that the only likely matches are going to be
2043 * in the 0-255 range, uniformly distributed, so half of that is 127
2044 * For /a and /d matches, it assumes that the likely matches will be just
2045 * the ASCII range, so half of that is 63
2046 * For /u and there isn't anything matching above the Latin1 range, it
2047 * assumes that that is the only range likely to be matched, and uses
2048 * half that as the cut-off: 127. If anything matches above Latin1,
2049 * it assumes that all of Unicode could match (uniformly), except for
2050 * non-Unicode code points and things in the General Category "Other"
2051 * (unassigned, private use, surrogates, controls and formats). This
2052 * is a much large number. */
2053
b35552de
KW
2054 U32 count = 0; /* Running total of number of code points matched by
2055 'ssc' */
2056 UV start, end; /* Start and end points of current range in inversion
26be5fe6 2057 XXX outdated. UTF-8 locales are common, what about invert? list */
72400949
KW
2058 const U32 max_code_points = (LOC)
2059 ? 256
f6e8f31e 2060 : (( ! UNI_SEMANTICS
1a26cbcb 2061 || invlist_highest(ssc->invlist) < 256)
72400949
KW
2062 ? 128
2063 : NON_OTHER_COUNT);
2064 const U32 max_match = max_code_points / 2;
b35552de
KW
2065
2066 PERL_ARGS_ASSERT_IS_SSC_WORTH_IT;
2067
2068 invlist_iterinit(ssc->invlist);
2069 while (invlist_iternext(ssc->invlist, &start, &end)) {
72400949
KW
2070 if (start >= max_code_points) {
2071 break;
b35552de 2072 }
72400949 2073 end = MIN(end, max_code_points - 1);
b35552de 2074 count += end - start + 1;
72400949 2075 if (count >= max_match) {
b35552de
KW
2076 invlist_iterfinish(ssc->invlist);
2077 return FALSE;
2078 }
2079 }
2080
2081 return TRUE;
2082}
2083
2084
b423522f
KW
2085STATIC void
2086S_ssc_finalize(pTHX_ RExC_state_t *pRExC_state, regnode_ssc *ssc)
2087{
2088 /* The inversion list in the SSC is marked mortal; now we need a more
2089 * permanent copy, which is stored the same way that is done in a regular
dcb20b36
KW
2090 * ANYOF node, with the first NUM_ANYOF_CODE_POINTS code points in a bit
2091 * map */
b423522f 2092
28118b9c 2093 SV* invlist = invlist_clone(ssc->invlist, NULL);
b423522f
KW
2094
2095 PERL_ARGS_ASSERT_SSC_FINALIZE;
2096
71068078 2097 assert(is_ANYOF_SYNTHETIC(ssc));
b423522f 2098
a0dd4231 2099 /* The code in this file assumes that all but these flags aren't relevant
93e92956
KW
2100 * to the SSC, except SSC_MATCHES_EMPTY_STRING, which should be cleared
2101 * by the time we reach here */
f240c685
KW
2102 assert(! (ANYOF_FLAGS(ssc)
2103 & ~( ANYOF_COMMON_FLAGS
108316fb
KW
2104 |ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
2105 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP)));
a0dd4231 2106
b423522f
KW
2107 populate_ANYOF_from_invlist( (regnode *) ssc, &invlist);
2108
4c404f26 2109 set_ANYOF_arg(pRExC_state, (regnode *) ssc, invlist, NULL, NULL);
b423522f 2110
85c8e306
KW
2111 /* Make sure is clone-safe */
2112 ssc->invlist = NULL;
2113
e0e1be5f 2114 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
93e92956 2115 ANYOF_FLAGS(ssc) |= ANYOF_MATCHES_POSIXL;
d156f5cb 2116 OP(ssc) = ANYOFPOSIXL;
e0e1be5f 2117 }
d156f5cb 2118 else if (RExC_contains_locale) {
b2e90ddf
KW
2119 OP(ssc) = ANYOFL;
2120 }
2121
1462525b 2122 assert(! (ANYOF_FLAGS(ssc) & ANYOF_LOCALE_FLAGS) || RExC_contains_locale);
b423522f
KW
2123}
2124
a3621e74
YO
2125#define TRIE_LIST_ITEM(state,idx) (trie->states[state].trans.list)[ idx ]
2126#define TRIE_LIST_CUR(state) ( TRIE_LIST_ITEM( state, 0 ).forid )
2127#define TRIE_LIST_LEN(state) ( TRIE_LIST_ITEM( state, 0 ).newstate )
538e84ed
KW
2128#define TRIE_LIST_USED(idx) ( trie->states[state].trans.list \
2129 ? (TRIE_LIST_CUR( idx ) - 1) \
2130 : 0 )
a3621e74 2131
3dab1dad
YO
2132
2133#ifdef DEBUGGING
07be1b83 2134/*
2b8b4781
NC
2135 dump_trie(trie,widecharmap,revcharmap)
2136 dump_trie_interim_list(trie,widecharmap,revcharmap,next_alloc)
2137 dump_trie_interim_table(trie,widecharmap,revcharmap,next_alloc)
3dab1dad
YO
2138
2139 These routines dump out a trie in a somewhat readable format.
07be1b83
YO
2140 The _interim_ variants are used for debugging the interim
2141 tables that are used to generate the final compressed
2142 representation which is what dump_trie expects.
2143
486ec47a 2144 Part of the reason for their existence is to provide a form
3dab1dad 2145 of documentation as to how the different representations function.
07be1b83
YO
2146
2147*/
3dab1dad
YO
2148
2149/*
3dab1dad
YO
2150 Dumps the final compressed table form of the trie to Perl_debug_log.
2151 Used for debugging make_trie().
2152*/
b9a59e08 2153
3dab1dad 2154STATIC void
2b8b4781
NC
2155S_dump_trie(pTHX_ const struct _reg_trie_data *trie, HV *widecharmap,
2156 AV *revcharmap, U32 depth)
3dab1dad
YO
2157{
2158 U32 state;
ab3bbdeb 2159 SV *sv=sv_newmortal();
55eed653 2160 int colwidth= widecharmap ? 6 : 4;
2e64971a 2161 U16 word;
3dab1dad
YO
2162 GET_RE_DEBUG_FLAGS_DECL;
2163
7918f24d 2164 PERL_ARGS_ASSERT_DUMP_TRIE;
ab3bbdeb 2165
6ad9a8ab 2166 Perl_re_indentf( aTHX_ "Char : %-6s%-6s%-4s ",
1e37780e 2167 depth+1, "Match","Base","Ofs" );
3dab1dad
YO
2168
2169 for( state = 0 ; state < trie->uniquecharcount ; state++ ) {
2b8b4781 2170 SV ** const tmp = av_fetch( revcharmap, state, 0);
3dab1dad 2171 if ( tmp ) {
6ad9a8ab 2172 Perl_re_printf( aTHX_ "%*s",
ab3bbdeb 2173 colwidth,
538e84ed 2174 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
ab3bbdeb
YO
2175 PL_colors[0], PL_colors[1],
2176 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
538e84ed
KW
2177 PERL_PV_ESCAPE_FIRSTCHAR
2178 )
ab3bbdeb 2179 );
3dab1dad
YO
2180 }
2181 }
1e37780e
YO
2182 Perl_re_printf( aTHX_ "\n");
2183 Perl_re_indentf( aTHX_ "State|-----------------------", depth+1);
3dab1dad
YO
2184
2185 for( state = 0 ; state < trie->uniquecharcount ; state++ )
6ad9a8ab
YO
2186 Perl_re_printf( aTHX_ "%.*s", colwidth, "--------");
2187 Perl_re_printf( aTHX_ "\n");
3dab1dad 2188
1e2e3d02 2189 for( state = 1 ; state < trie->statecount ; state++ ) {
be8e71aa 2190 const U32 base = trie->states[ state ].trans.base;
3dab1dad 2191
147e3846 2192 Perl_re_indentf( aTHX_ "#%4" UVXf "|", depth+1, (UV)state);
3dab1dad
YO
2193
2194 if ( trie->states[ state ].wordnum ) {
1e37780e 2195 Perl_re_printf( aTHX_ " W%4X", trie->states[ state ].wordnum );
3dab1dad 2196 } else {
6ad9a8ab 2197 Perl_re_printf( aTHX_ "%6s", "" );
3dab1dad
YO
2198 }
2199
147e3846 2200 Perl_re_printf( aTHX_ " @%4" UVXf " ", (UV)base );
3dab1dad
YO
2201
2202 if ( base ) {
2203 U32 ofs = 0;
2204
2205 while( ( base + ofs < trie->uniquecharcount ) ||
2206 ( base + ofs - trie->uniquecharcount < trie->lasttrans
538e84ed
KW
2207 && trie->trans[ base + ofs - trie->uniquecharcount ].check
2208 != state))
3dab1dad
YO
2209 ofs++;
2210
147e3846 2211 Perl_re_printf( aTHX_ "+%2" UVXf "[ ", (UV)ofs);
3dab1dad
YO
2212
2213 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
538e84ed
KW
2214 if ( ( base + ofs >= trie->uniquecharcount )
2215 && ( base + ofs - trie->uniquecharcount
2216 < trie->lasttrans )
2217 && trie->trans[ base + ofs
2218 - trie->uniquecharcount ].check == state )
3dab1dad 2219 {
147e3846 2220 Perl_re_printf( aTHX_ "%*" UVXf, colwidth,
1e37780e
YO
2221 (UV)trie->trans[ base + ofs - trie->uniquecharcount ].next
2222 );
3dab1dad 2223 } else {
88f063b4 2224 Perl_re_printf( aTHX_ "%*s", colwidth," ." );
3dab1dad
YO
2225 }
2226 }
2227
6ad9a8ab 2228 Perl_re_printf( aTHX_ "]");
3dab1dad
YO
2229
2230 }
6ad9a8ab 2231 Perl_re_printf( aTHX_ "\n" );
3dab1dad 2232 }
6ad9a8ab 2233 Perl_re_indentf( aTHX_ "word_info N:(prev,len)=",
cb41e5d6 2234 depth);
2e64971a 2235 for (word=1; word <= trie->wordcount; word++) {
6ad9a8ab 2236 Perl_re_printf( aTHX_ " %d:(%d,%d)",
2e64971a
DM
2237 (int)word, (int)(trie->wordinfo[word].prev),
2238 (int)(trie->wordinfo[word].len));
2239 }
6ad9a8ab 2240 Perl_re_printf( aTHX_ "\n" );
538e84ed 2241}
3dab1dad 2242/*
3dab1dad 2243 Dumps a fully constructed but uncompressed trie in list form.
538e84ed 2244 List tries normally only are used for construction when the number of
3dab1dad
YO
2245 possible chars (trie->uniquecharcount) is very high.
2246 Used for debugging make_trie().
2247*/
2248STATIC void
55eed653 2249S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie,
2b8b4781
NC
2250 HV *widecharmap, AV *revcharmap, U32 next_alloc,
2251 U32 depth)
3dab1dad
YO
2252{
2253 U32 state;
ab3bbdeb 2254 SV *sv=sv_newmortal();
55eed653 2255 int colwidth= widecharmap ? 6 : 4;
3dab1dad 2256 GET_RE_DEBUG_FLAGS_DECL;
7918f24d
NC
2257
2258 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_LIST;
2259
3dab1dad 2260 /* print out the table precompression. */
6ad9a8ab 2261 Perl_re_indentf( aTHX_ "State :Word | Transition Data\n",
cb41e5d6 2262 depth+1 );
6ad9a8ab 2263 Perl_re_indentf( aTHX_ "%s",
cb41e5d6 2264 depth+1, "------:-----+-----------------\n" );
538e84ed 2265
3dab1dad
YO
2266 for( state=1 ; state < next_alloc ; state ++ ) {
2267 U16 charid;
538e84ed 2268
147e3846 2269 Perl_re_indentf( aTHX_ " %4" UVXf " :",
cb41e5d6 2270 depth+1, (UV)state );
3dab1dad 2271 if ( ! trie->states[ state ].wordnum ) {
6ad9a8ab 2272 Perl_re_printf( aTHX_ "%5s| ","");
3dab1dad 2273 } else {
6ad9a8ab 2274 Perl_re_printf( aTHX_ "W%4x| ",
3dab1dad
YO
2275 trie->states[ state ].wordnum
2276 );
2277 }
2278 for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) {
538e84ed 2279 SV ** const tmp = av_fetch( revcharmap,
88f063b4 2280 TRIE_LIST_ITEM(state, charid).forid, 0);
ab3bbdeb 2281 if ( tmp ) {
147e3846 2282 Perl_re_printf( aTHX_ "%*s:%3X=%4" UVXf " | ",
ab3bbdeb 2283 colwidth,
538e84ed
KW
2284 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp),
2285 colwidth,
2286 PL_colors[0], PL_colors[1],
2287 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
2288 | PERL_PV_ESCAPE_FIRSTCHAR
ab3bbdeb 2289 ) ,
88f063b4
KW
2290 TRIE_LIST_ITEM(state, charid).forid,
2291 (UV)TRIE_LIST_ITEM(state, charid).newstate
1e2e3d02 2292 );
538e84ed 2293 if (!(charid % 10))
6ad9a8ab 2294 Perl_re_printf( aTHX_ "\n%*s| ",
664e119d 2295 (int)((depth * 2) + 14), "");
1e2e3d02 2296 }
ab3bbdeb 2297 }
6ad9a8ab 2298 Perl_re_printf( aTHX_ "\n");
3dab1dad 2299 }
538e84ed 2300}
3dab1dad
YO
2301
2302/*
3dab1dad 2303 Dumps a fully constructed but uncompressed trie in table form.
538e84ed
KW
2304 This is the normal DFA style state transition table, with a few
2305 twists to facilitate compression later.
3dab1dad
YO
2306 Used for debugging make_trie().
2307*/
2308STATIC void
55eed653 2309S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie,
2b8b4781
NC
2310 HV *widecharmap, AV *revcharmap, U32 next_alloc,
2311 U32 depth)
3dab1dad
YO
2312{
2313 U32 state;
2314 U16 charid;
ab3bbdeb 2315 SV *sv=sv_newmortal();
55eed653 2316 int colwidth= widecharmap ? 6 : 4;
3dab1dad 2317 GET_RE_DEBUG_FLAGS_DECL;
7918f24d
NC
2318
2319 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_TABLE;
538e84ed 2320
3dab1dad
YO
2321 /*
2322 print out the table precompression so that we can do a visual check
2323 that they are identical.
2324 */
538e84ed 2325
6ad9a8ab 2326 Perl_re_indentf( aTHX_ "Char : ", depth+1 );
3dab1dad
YO
2327
2328 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
2b8b4781 2329 SV ** const tmp = av_fetch( revcharmap, charid, 0);
3dab1dad 2330 if ( tmp ) {
6ad9a8ab 2331 Perl_re_printf( aTHX_ "%*s",
ab3bbdeb 2332 colwidth,
538e84ed 2333 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
ab3bbdeb
YO
2334 PL_colors[0], PL_colors[1],
2335 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
538e84ed
KW
2336 PERL_PV_ESCAPE_FIRSTCHAR
2337 )
ab3bbdeb 2338 );
3dab1dad
YO
2339 }
2340 }
2341
4aaafc03
YO
2342 Perl_re_printf( aTHX_ "\n");
2343 Perl_re_indentf( aTHX_ "State+-", depth+1 );
3dab1dad
YO
2344
2345 for( charid=0 ; charid < trie->uniquecharcount ; charid++ ) {
6ad9a8ab 2346 Perl_re_printf( aTHX_ "%.*s", colwidth,"--------");
3dab1dad
YO
2347 }
2348
6ad9a8ab 2349 Perl_re_printf( aTHX_ "\n" );
3dab1dad
YO
2350
2351 for( state=1 ; state < next_alloc ; state += trie->uniquecharcount ) {
2352
147e3846 2353 Perl_re_indentf( aTHX_ "%4" UVXf " : ",
cb41e5d6 2354 depth+1,
3dab1dad
YO
2355 (UV)TRIE_NODENUM( state ) );
2356
2357 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
ab3bbdeb
YO
2358 UV v=(UV)SAFE_TRIE_NODENUM( trie->trans[ state + charid ].next );
2359 if (v)
147e3846 2360 Perl_re_printf( aTHX_ "%*" UVXf, colwidth, v );
ab3bbdeb 2361 else
6ad9a8ab 2362 Perl_re_printf( aTHX_ "%*s", colwidth, "." );
3dab1dad
YO
2363 }
2364 if ( ! trie->states[ TRIE_NODENUM( state ) ].wordnum ) {
147e3846 2365 Perl_re_printf( aTHX_ " (%4" UVXf ")\n",
538e84ed 2366 (UV)trie->trans[ state ].check );
3dab1dad 2367 } else {
147e3846 2368 Perl_re_printf( aTHX_ " (%4" UVXf ") W%4X\n",
538e84ed 2369 (UV)trie->trans[ state ].check,
3dab1dad
YO
2370 trie->states[ TRIE_NODENUM( state ) ].wordnum );
2371 }
2372 }
07be1b83 2373}
3dab1dad
YO
2374
2375#endif
2376
2e64971a 2377
786e8c11
YO
2378/* make_trie(startbranch,first,last,tail,word_count,flags,depth)
2379 startbranch: the first branch in the whole branch sequence
2380 first : start branch of sequence of branch-exact nodes.
2381 May be the same as startbranch
2382 last : Thing following the last branch.
2383 May be the same as tail.
2384 tail : item following the branch sequence
2385 count : words in the sequence
a4525e78 2386 flags : currently the OP() type we will be building one of /EXACT(|F|FA|FU|FU_SS|L|FLU8)/
786e8c11 2387 depth : indent depth
3dab1dad 2388
786e8c11 2389Inplace optimizes a sequence of 2 or more Branch-Exact nodes into a TRIE node.
07be1b83 2390
786e8c11
YO
2391A trie is an N'ary tree where the branches are determined by digital
2392decomposition of the key. IE, at the root node you look up the 1st character and
2393follow that branch repeat until you find the end of the branches. Nodes can be
2394marked as "accepting" meaning they represent a complete word. Eg:
07be1b83 2395
786e8c11 2396 /he|she|his|hers/
72f13be8 2397
786e8c11
YO
2398would convert into the following structure. Numbers represent states, letters
2399following numbers represent valid transitions on the letter from that state, if
2400the number is in square brackets it represents an accepting state, otherwise it
2401will be in parenthesis.
07be1b83 2402
786e8c11
YO
2403 +-h->+-e->[3]-+-r->(8)-+-s->[9]
2404 | |
2405 | (2)
2406 | |
2407 (1) +-i->(6)-+-s->[7]
2408 |
2409 +-s->(3)-+-h->(4)-+-e->[5]
07be1b83 2410
786e8c11
YO
2411 Accept Word Mapping: 3=>1 (he),5=>2 (she), 7=>3 (his), 9=>4 (hers)
2412
2413This shows that when matching against the string 'hers' we will begin at state 1
2414read 'h' and move to state 2, read 'e' and move to state 3 which is accepting,
2415then read 'r' and go to state 8 followed by 's' which takes us to state 9 which
2416is also accepting. Thus we know that we can match both 'he' and 'hers' with a
2417single traverse. We store a mapping from accepting to state to which word was
2418matched, and then when we have multiple possibilities we try to complete the
b8fda935 2419rest of the regex in the order in which they occurred in the alternation.
786e8c11
YO
2420
2421The only prior NFA like behaviour that would be changed by the TRIE support is
2422the silent ignoring of duplicate alternations which are of the form:
2423
2424 / (DUPE|DUPE) X? (?{ ... }) Y /x
2425
4b714af6 2426Thus EVAL blocks following a trie may be called a different number of times with
786e8c11 2427and without the optimisation. With the optimisations dupes will be silently
486ec47a 2428ignored. This inconsistent behaviour of EVAL type nodes is well established as
786e8c11
YO
2429the following demonstrates:
2430
2431 'words'=~/(word|word|word)(?{ print $1 })[xyz]/
2432
2433which prints out 'word' three times, but
2434
2435 'words'=~/(word|word|word)(?{ print $1 })S/
2436
2437which doesnt print it out at all. This is due to other optimisations kicking in.
2438
2439Example of what happens on a structural level:
2440
486ec47a 2441The regexp /(ac|ad|ab)+/ will produce the following debug output:
786e8c11
YO
2442
2443 1: CURLYM[1] {1,32767}(18)
2444 5: BRANCH(8)
2445 6: EXACT <ac>(16)
2446 8: BRANCH(11)
2447 9: EXACT <ad>(16)
2448 11: BRANCH(14)
2449 12: EXACT <ab>(16)
2450 16: SUCCEED(0)
2451 17: NOTHING(18)
2452 18: END(0)
2453
2454This would be optimizable with startbranch=5, first=5, last=16, tail=16
2455and should turn into:
2456
2457 1: CURLYM[1] {1,32767}(18)
2458 5: TRIE(16)
2459 [Words:3 Chars Stored:6 Unique Chars:4 States:5 NCP:1]
2460 <ac>
2461 <ad>
2462 <ab>
2463 16: SUCCEED(0)
2464 17: NOTHING(18)
2465 18: END(0)
2466
2467Cases where tail != last would be like /(?foo|bar)baz/:
2468
2469 1: BRANCH(4)
2470 2: EXACT <foo>(8)
2471 4: BRANCH(7)
2472 5: EXACT <bar>(8)
2473 7: TAIL(8)
2474 8: EXACT <baz>(10)
2475 10: END(0)
2476
2477which would be optimizable with startbranch=1, first=1, last=7, tail=8
2478and would end up looking like:
2479
2480 1: TRIE(8)
2481 [Words:2 Chars Stored:6 Unique Chars:5 States:7 NCP:1]
2482 <foo>
2483 <bar>
2484 7: TAIL(8)
2485 8: EXACT <baz>(10)
2486 10: END(0)
2487
c80e42f3 2488 d = uvchr_to_utf8_flags(d, uv, 0);
786e8c11
YO
2489
2490is the recommended Unicode-aware way of saying
2491
2492 *(d++) = uv;
2493*/
2494
fab2782b 2495#define TRIE_STORE_REVCHAR(val) \
786e8c11 2496 STMT_START { \
73031816 2497 if (UTF) { \
668fcfea 2498 SV *zlopp = newSV(UTF8_MAXBYTES); \
88c9ea1e 2499 unsigned char *flrbbbbb = (unsigned char *) SvPVX(zlopp); \
c80e42f3 2500 unsigned const char *const kapow = uvchr_to_utf8(flrbbbbb, val); \
73031816
NC
2501 SvCUR_set(zlopp, kapow - flrbbbbb); \
2502 SvPOK_on(zlopp); \
2503 SvUTF8_on(zlopp); \
2504 av_push(revcharmap, zlopp); \
2505 } else { \
fab2782b 2506 char ooooff = (char)val; \
73031816
NC
2507 av_push(revcharmap, newSVpvn(&ooooff, 1)); \
2508 } \
2509 } STMT_END
786e8c11 2510
914a25d5
KW
2511/* This gets the next character from the input, folding it if not already
2512 * folded. */
2513#define TRIE_READ_CHAR STMT_START { \
2514 wordlen++; \
2515 if ( UTF ) { \
2516 /* if it is UTF then it is either already folded, or does not need \
2517 * folding */ \
1c1d615a 2518 uvc = valid_utf8_to_uvchr( (const U8*) uc, &len); \
914a25d5
KW
2519 } \
2520 else if (folder == PL_fold_latin1) { \
7d006b13
KW
2521 /* This folder implies Unicode rules, which in the range expressible \
2522 * by not UTF is the lower case, with the two exceptions, one of \
2523 * which should have been taken care of before calling this */ \
2524 assert(*uc != LATIN_SMALL_LETTER_SHARP_S); \
2525 uvc = toLOWER_L1(*uc); \
2526 if (UNLIKELY(uvc == MICRO_SIGN)) uvc = GREEK_SMALL_LETTER_MU; \
2527 len = 1; \
914a25d5
KW
2528 } else { \
2529 /* raw data, will be folded later if needed */ \
2530 uvc = (U32)*uc; \
2531 len = 1; \
2532 } \
786e8c11
YO
2533} STMT_END
2534
2535
2536
2537#define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \
2538 if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \
00195859 2539 U32 ging = TRIE_LIST_LEN( state ) * 2; \
f9003953 2540 Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \
00195859 2541 TRIE_LIST_LEN( state ) = ging; \
786e8c11
YO
2542 } \
2543 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \
2544 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \
2545 TRIE_LIST_CUR( state )++; \
2546} STMT_END
07be1b83 2547
786e8c11 2548#define TRIE_LIST_NEW(state) STMT_START { \
d09f14bf 2549 Newx( trie->states[ state ].trans.list, \
786e8c11
YO
2550 4, reg_trie_trans_le ); \
2551 TRIE_LIST_CUR( state ) = 1; \
2552 TRIE_LIST_LEN( state ) = 4; \
2553} STMT_END
07be1b83 2554
786e8c11
YO
2555#define TRIE_HANDLE_WORD(state) STMT_START { \
2556 U16 dupe= trie->states[ state ].wordnum; \
2557 regnode * const noper_next = regnext( noper ); \
2558 \
786e8c11
YO
2559 DEBUG_r({ \
2560 /* store the word for dumping */ \
2561 SV* tmp; \
2562 if (OP(noper) != NOTHING) \
740cce10 2563 tmp = newSVpvn_utf8(STRING(noper), STR_LEN(noper), UTF); \
786e8c11 2564 else \
740cce10 2565 tmp = newSVpvn_utf8( "", 0, UTF ); \
2b8b4781 2566 av_push( trie_words, tmp ); \
786e8c11
YO
2567 }); \
2568 \
2569 curword++; \
2e64971a
DM
2570 trie->wordinfo[curword].prev = 0; \
2571 trie->wordinfo[curword].len = wordlen; \
2572 trie->wordinfo[curword].accept = state; \
786e8c11
YO
2573 \
2574 if ( noper_next < tail ) { \
2575 if (!trie->jump) \
538e84ed
KW
2576 trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, \
2577 sizeof(U16) ); \
7f69552c 2578 trie->jump[curword] = (U16)(noper_next - convert); \
786e8c11
YO
2579 if (!jumper) \
2580 jumper = noper_next; \
2581 if (!nextbranch) \
2582 nextbranch= regnext(cur); \
2583 } \
2584 \
2585 if ( dupe ) { \
2e64971a
DM
2586 /* It's a dupe. Pre-insert into the wordinfo[].prev */\
2587 /* chain, so that when the bits of chain are later */\
2588 /* linked together, the dups appear in the chain */\
2589 trie->wordinfo[curword].prev = trie->wordinfo[dupe].prev; \
2590 trie->wordinfo[dupe].prev = curword; \
786e8c11
YO
2591 } else { \
2592 /* we haven't inserted this word yet. */ \
2593 trie->states[ state ].wordnum = curword; \
2594 } \
2595} STMT_END
07be1b83 2596
3dab1dad 2597
786e8c11
YO
2598#define TRIE_TRANS_STATE(state,base,ucharcount,charid,special) \
2599 ( ( base + charid >= ucharcount \
2600 && base + charid < ubound \
2601 && state == trie->trans[ base - ucharcount + charid ].check \
2602 && trie->trans[ base - ucharcount + charid ].next ) \
2603 ? trie->trans[ base - ucharcount + charid ].next \
2604 : ( state==1 ? special : 0 ) \
2605 )
3dab1dad 2606
8bcafbf4
YO
2607#define TRIE_BITMAP_SET_FOLDED(trie, uvc, folder) \
2608STMT_START { \
2609 TRIE_BITMAP_SET(trie, uvc); \
2610 /* store the folded codepoint */ \
2611 if ( folder ) \
2612 TRIE_BITMAP_SET(trie, folder[(U8) uvc ]); \
2613 \
2614 if ( !UTF ) { \
2615 /* store first byte of utf8 representation of */ \
2616 /* variant codepoints */ \
2617 if (! UVCHR_IS_INVARIANT(uvc)) { \
2618 TRIE_BITMAP_SET(trie, UTF8_TWO_BYTE_HI(uvc)); \
2619 } \
2620 } \
2621} STMT_END
786e8c11
YO
2622#define MADE_TRIE 1
2623#define MADE_JUMP_TRIE 2
2624#define MADE_EXACT_TRIE 4
3dab1dad 2625
a3621e74 2626STATIC I32
538e84ed
KW
2627S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch,
2628 regnode *first, regnode *last, regnode *tail,
2629 U32 word_count, U32 flags, U32 depth)
a3621e74
YO
2630{
2631 /* first pass, loop through and scan words */
2632 reg_trie_data *trie;
55eed653 2633 HV *widecharmap = NULL;
2b8b4781 2634 AV *revcharmap = newAV();
a3621e74 2635 regnode *cur;
a3621e74
YO
2636 STRLEN len = 0;
2637 UV uvc = 0;
2638 U16 curword = 0;
2639 U32 next_alloc = 0;
786e8c11
YO
2640 regnode *jumper = NULL;
2641 regnode *nextbranch = NULL;
7f69552c 2642 regnode *convert = NULL;
2e64971a 2643 U32 *prev_states; /* temp array mapping each state to previous one */
a3621e74 2644 /* we just use folder as a flag in utf8 */
1e696034 2645 const U8 * folder = NULL;
a3621e74 2646
3a611511
YO
2647 /* in the below add_data call we are storing either 'tu' or 'tuaa'
2648 * which stands for one trie structure, one hash, optionally followed
2649 * by two arrays */
2b8b4781 2650#ifdef DEBUGGING
3a611511 2651 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("tuaa"));
2b8b4781
NC
2652 AV *trie_words = NULL;
2653 /* along with revcharmap, this only used during construction but both are
2654 * useful during debugging so we store them in the struct when debugging.
8e11feef 2655 */
2b8b4781 2656#else
cf78de0b 2657 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("tu"));
3dab1dad 2658 STRLEN trie_charcount=0;
3dab1dad 2659#endif
2b8b4781 2660 SV *re_trie_maxbuff;
a3621e74 2661 GET_RE_DEBUG_FLAGS_DECL;
7918f24d
NC
2662
2663 PERL_ARGS_ASSERT_MAKE_TRIE;
72f13be8
YO
2664#ifndef DEBUGGING
2665 PERL_UNUSED_ARG(depth);
2666#endif
a3621e74 2667
1e696034 2668 switch (flags) {
f6b4b99d 2669 case EXACT: case EXACT_ONLY8: case EXACTL: break;
89829bb5 2670 case EXACTFAA:
0ea669f4 2671 case EXACTFUP:
a4525e78
KW
2672 case EXACTFU:
2673 case EXACTFLU8: folder = PL_fold_latin1; break;
1e696034 2674 case EXACTF: folder = PL_fold; break;
fab2782b 2675 default: Perl_croak( aTHX_ "panic! In trie construction, unknown node type %u %s", (unsigned) flags, PL_reg_name[flags] );
1e696034
KW
2676 }
2677
c944940b 2678 trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
a3621e74 2679 trie->refcount = 1;
3dab1dad 2680 trie->startstate = 1;
786e8c11 2681 trie->wordcount = word_count;
f8fc2ecf 2682 RExC_rxi->data->data[ data_slot ] = (void*)trie;
c944940b 2683 trie->charmap = (U16 *) PerlMemShared_calloc( 256, sizeof(U16) );
f6b4b99d 2684 if (flags == EXACT || flags == EXACT_ONLY8 || flags == EXACTL)
c944940b 2685 trie->bitmap = (char *) PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 );
2e64971a
DM
2686 trie->wordinfo = (reg_trie_wordinfo *) PerlMemShared_calloc(
2687 trie->wordcount+1, sizeof(reg_trie_wordinfo));
2688
a3621e74 2689 DEBUG_r({
2b8b4781 2690 trie_words = newAV();
a3621e74 2691 });
a3621e74 2692
0111c4fd 2693 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
316ebaf2 2694 assert(re_trie_maxbuff);
a3621e74 2695 if (!SvIOK(re_trie_maxbuff)) {
0111c4fd 2696 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
a3621e74 2697 }
df826430 2698 DEBUG_TRIE_COMPILE_r({
6ad9a8ab 2699 Perl_re_indentf( aTHX_
cb41e5d6
YO
2700 "make_trie start==%d, first==%d, last==%d, tail==%d depth=%d\n",
2701 depth+1,
88f063b4 2702 REG_NODE_NUM(startbranch), REG_NODE_NUM(first),
538e84ed 2703 REG_NODE_NUM(last), REG_NODE_NUM(tail), (int)depth);
3dab1dad 2704 });
538e84ed 2705
7f69552c
YO
2706 /* Find the node we are going to overwrite */
2707 if ( first == startbranch && OP( last ) != BRANCH ) {
2708 /* whole branch chain */
2709 convert = first;
2710 } else {
2711 /* branch sub-chain */
2712 convert = NEXTOPER( first );
2713 }
538e84ed 2714
a3621e74
YO
2715 /* -- First loop and Setup --
2716
2717 We first traverse the branches and scan each word to determine if it
2718 contains widechars, and how many unique chars there are, this is
2719 important as we have to build a table with at least as many columns as we
2720 have unique chars.
2721
2722 We use an array of integers to represent the character codes 0..255
538e84ed
KW
2723 (trie->charmap) and we use a an HV* to store Unicode characters. We use
2724 the native representation of the character value as the key and IV's for
2725 the coded index.
a3621e74
YO
2726
2727 *TODO* If we keep track of how many times each character is used we can
2728 remap the columns so that the table compression later on is more
3b753521 2729 efficient in terms of memory by ensuring the most common value is in the
a3621e74
YO
2730 middle and the least common are on the outside. IMO this would be better
2731 than a most to least common mapping as theres a decent chance the most
2732 common letter will share a node with the least common, meaning the node
486ec47a 2733 will not be compressible. With a middle is most common approach the worst
a3621e74
YO
2734 case is when we have the least common nodes twice.
2735
2736 */
2737
a3621e74 2738 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
df826430 2739 regnode *noper = NEXTOPER( cur );
944e05e3
YO
2740 const U8 *uc;
2741 const U8 *e;
bc031a7d 2742 int foldlen = 0;
07be1b83 2743 U32 wordlen = 0; /* required init */
bc031a7d
KW
2744 STRLEN minchars = 0;
2745 STRLEN maxchars = 0;
538e84ed
KW
2746 bool set_bit = trie->bitmap ? 1 : 0; /*store the first char in the
2747 bitmap?*/
a3621e74 2748
3dab1dad 2749 if (OP(noper) == NOTHING) {
20ed8c88
YO
2750 /* skip past a NOTHING at the start of an alternation
2751 * eg, /(?:)a|(?:b)/ should be the same as /a|b/
2752 */
df826430 2753 regnode *noper_next= regnext(noper);
944e05e3
YO
2754 if (noper_next < tail)
2755 noper= noper_next;
2756 }
2757
f6b4b99d
KW
2758 if ( noper < tail
2759 && ( OP(noper) == flags
2760 || (flags == EXACT && OP(noper) == EXACT_ONLY8)
a9f8c7ac 2761 || (flags == EXACTFU && ( OP(noper) == EXACTFU_ONLY8
0ea669f4 2762 || OP(noper) == EXACTFUP))))
f6b4b99d 2763 {
944e05e3
YO
2764 uc= (U8*)STRING(noper);
2765 e= uc + STR_LEN(noper);
2766 } else {
2767 trie->minlen= 0;
2768 continue;
3dab1dad 2769 }
df826430 2770
944e05e3 2771
fab2782b 2772 if ( set_bit ) { /* bitmap only alloced when !(UTF&&Folding) */
02daf0ab
YO
2773 TRIE_BITMAP_SET(trie,*uc); /* store the raw first byte
2774 regardless of encoding */
0ea669f4 2775 if (OP( noper ) == EXACTFUP) {
fab2782b 2776 /* false positives are ok, so just set this */
0dc4a61d 2777 TRIE_BITMAP_SET(trie, LATIN_SMALL_LETTER_SHARP_S);
fab2782b
YO
2778 }
2779 }
dca5fc4c 2780
bc031a7d
KW
2781 for ( ; uc < e ; uc += len ) { /* Look at each char in the current
2782 branch */
3dab1dad 2783 TRIE_CHARCOUNT(trie)++;
a3621e74 2784 TRIE_READ_CHAR;
645de4ce 2785
bc031a7d
KW
2786 /* TRIE_READ_CHAR returns the current character, or its fold if /i
2787 * is in effect. Under /i, this character can match itself, or
2788 * anything that folds to it. If not under /i, it can match just
2789 * itself. Most folds are 1-1, for example k, K, and KELVIN SIGN
2790 * all fold to k, and all are single characters. But some folds
2791 * expand to more than one character, so for example LATIN SMALL
2792 * LIGATURE FFI folds to the three character sequence 'ffi'. If
2793 * the string beginning at 'uc' is 'ffi', it could be matched by
2794 * three characters, or just by the one ligature character. (It
2795 * could also be matched by two characters: LATIN SMALL LIGATURE FF
2796 * followed by 'i', or by 'f' followed by LATIN SMALL LIGATURE FI).
2797 * (Of course 'I' and/or 'F' instead of 'i' and 'f' can also
2798 * match.) The trie needs to know the minimum and maximum number
2799 * of characters that could match so that it can use size alone to
2800 * quickly reject many match attempts. The max is simple: it is
2801 * the number of folded characters in this branch (since a fold is
2802 * never shorter than what folds to it. */
2803
2804 maxchars++;
2805
2806 /* And the min is equal to the max if not under /i (indicated by
2807 * 'folder' being NULL), or there are no multi-character folds. If
2808 * there is a multi-character fold, the min is incremented just
2809 * once, for the character that folds to the sequence. Each
2810 * character in the sequence needs to be added to the list below of
2811 * characters in the trie, but we count only the first towards the
2812 * min number of characters needed. This is done through the
2813 * variable 'foldlen', which is returned by the macros that look
2814 * for these sequences as the number of bytes the sequence
2815 * occupies. Each time through the loop, we decrement 'foldlen' by
2816 * how many bytes the current char occupies. Only when it reaches
2817 * 0 do we increment 'minchars' or look for another multi-character
2818 * sequence. */
2819 if (folder == NULL) {
2820 minchars++;
2821 }
2822 else if (foldlen > 0) {
2823 foldlen -= (UTF) ? UTF8SKIP(uc) : 1;
645de4ce
KW
2824 }
2825 else {
bc031a7d
KW
2826 minchars++;
2827
2828 /* See if *uc is the beginning of a multi-character fold. If
2829 * so, we decrement the length remaining to look at, to account
2830 * for the current character this iteration. (We can use 'uc'
2831 * instead of the fold returned by TRIE_READ_CHAR because for
2832 * non-UTF, the latin1_safe macro is smart enough to account
2833 * for all the unfolded characters, and because for UTF, the
2834 * string will already have been folded earlier in the
2835 * compilation process */
2836 if (UTF) {
2837 if ((foldlen = is_MULTI_CHAR_FOLD_utf8_safe(uc, e))) {
2838 foldlen -= UTF8SKIP(uc);
645de4ce
KW
2839 }
2840 }
bc031a7d
KW
2841 else if ((foldlen = is_MULTI_CHAR_FOLD_latin1_safe(uc, e))) {
2842 foldlen--;
2843 }
645de4ce 2844 }
bc031a7d
KW
2845
2846 /* The current character (and any potential folds) should be added
2847 * to the possible matching characters for this position in this
2848 * branch */
a3621e74 2849 if ( uvc < 256 ) {
fab2782b
YO
2850 if ( folder ) {
2851 U8 folded= folder[ (U8) uvc ];
2852 if ( !trie->charmap[ folded ] ) {
2853 trie->charmap[ folded ]=( ++trie->uniquecharcount );
2854 TRIE_STORE_REVCHAR( folded );
2855 }
2856 }
a3621e74
YO
2857 if ( !trie->charmap[ uvc ] ) {
2858 trie->charmap[ uvc ]=( ++trie->uniquecharcount );
fab2782b 2859 TRIE_STORE_REVCHAR( uvc );
a3621e74 2860 }
02daf0ab 2861 if ( set_bit ) {
62012aee
KW
2862 /* store the codepoint in the bitmap, and its folded
2863 * equivalent. */
8bcafbf4 2864 TRIE_BITMAP_SET_FOLDED(trie, uvc, folder);
02daf0ab
YO
2865 set_bit = 0; /* We've done our bit :-) */
2866 }
a3621e74 2867 } else {
bc031a7d
KW
2868
2869 /* XXX We could come up with the list of code points that fold
2870 * to this using PL_utf8_foldclosures, except not for
2871 * multi-char folds, as there may be multiple combinations
2872 * there that could work, which needs to wait until runtime to
2873 * resolve (The comment about LIGATURE FFI above is such an
2874 * example */
2875
a3621e74 2876 SV** svpp;
55eed653
NC
2877 if ( !widecharmap )
2878 widecharmap = newHV();
a3621e74 2879
55eed653 2880 svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 1 );
a3621e74
YO
2881
2882 if ( !svpp )
147e3846 2883 Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%" UVXf, uvc );
a3621e74
YO
2884
2885 if ( !SvTRUE( *svpp ) ) {
2886 sv_setiv( *svpp, ++trie->uniquecharcount );
fab2782b 2887 TRIE_STORE_REVCHAR(uvc);
a3621e74
YO
2888 }
2889 }
bc031a7d
KW
2890 } /* end loop through characters in this branch of the trie */
2891
2892 /* We take the min and max for this branch and combine to find the min
2893 * and max for all branches processed so far */
3dab1dad 2894 if( cur == first ) {
bc031a7d
KW
2895 trie->minlen = minchars;
2896 trie->maxlen = maxchars;
2897 } else if (minchars < trie->minlen) {
2898 trie->minlen = minchars;
2899 } else if (maxchars > trie->maxlen) {
2900 trie->maxlen = maxchars;
fab2782b 2901 }
a3621e74
YO
2902 } /* end first pass */
2903 DEBUG_TRIE_COMPILE_r(
6ad9a8ab 2904 Perl_re_indentf( aTHX_
cb41e5d6
YO
2905 "TRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n",
2906 depth+1,
55eed653 2907 ( widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
be8e71aa
YO
2908 (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount,
2909 (int)trie->minlen, (int)trie->maxlen )
a3621e74 2910 );
a3621e74
YO
2911
2912 /*
2913 We now know what we are dealing with in terms of unique chars and
2914 string sizes so we can calculate how much memory a naive
0111c4fd
RGS
2915 representation using a flat table will take. If it's over a reasonable
2916 limit (as specified by ${^RE_TRIE_MAXBUF}) we use a more memory
a3621e74
YO
2917 conservative but potentially much slower representation using an array
2918 of lists.
2919
2920 At the end we convert both representations into the same compressed
2921 form that will be used in regexec.c for matching with. The latter
2922 is a form that cannot be used to construct with but has memory
2923 properties similar to the list form and access properties similar
2924 to the table form making it both suitable for fast searches and
2925 small enough that its feasable to store for the duration of a program.
2926
2927 See the comment in the code where the compressed table is produced
2928 inplace from the flat tabe representation for an explanation of how
2929 the compression works.
2930
2931 */
2932
2933
2e64971a
DM
2934 Newx(prev_states, TRIE_CHARCOUNT(trie) + 2, U32);
2935 prev_states[1] = 0;
2936
538e84ed
KW
2937 if ( (IV)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1)
2938 > SvIV(re_trie_maxbuff) )
2939 {
a3621e74
YO
2940 /*
2941 Second Pass -- Array Of Lists Representation
2942
2943 Each state will be represented by a list of charid:state records
2944 (reg_trie_trans_le) the first such element holds the CUR and LEN
2945 points of the allocated array. (See defines above).
2946
2947 We build the initial structure using the lists, and then convert
2948 it into the compressed table form which allows faster lookups
2949 (but cant be modified once converted).
a3621e74
YO
2950 */
2951
a3621e74
YO
2952 STRLEN transcount = 1;
2953
6ad9a8ab 2954 DEBUG_TRIE_COMPILE_MORE_r( Perl_re_indentf( aTHX_ "Compiling trie using list compiler\n",
cb41e5d6 2955 depth+1));
686b73d4 2956
c944940b
JH
2957 trie->states = (reg_trie_state *)
2958 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
2959 sizeof(reg_trie_state) );
a3621e74
YO
2960 TRIE_LIST_NEW(1);
2961 next_alloc = 2;
2962
2963 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
2964
df826430 2965 regnode *noper = NEXTOPER( cur );
c445ea15
AL
2966 U32 state = 1; /* required init */
2967 U16 charid = 0; /* sanity init */
07be1b83 2968 U32 wordlen = 0; /* required init */
c445ea15 2969
df826430
YO
2970 if (OP(noper) == NOTHING) {
2971 regnode *noper_next= regnext(noper);
944e05e3
YO
2972 if (noper_next < tail)
2973 noper= noper_next;
df826430
YO
2974 }
2975
f6b4b99d
KW
2976 if ( noper < tail
2977 && ( OP(noper) == flags
2978 || (flags == EXACT && OP(noper) == EXACT_ONLY8)
a9f8c7ac 2979 || (flags == EXACTFU && ( OP(noper) == EXACTFU_ONLY8
0ea669f4 2980 || OP(noper) == EXACTFUP))))
f6b4b99d 2981 {
944e05e3
YO
2982 const U8 *uc= (U8*)STRING(noper);
2983 const U8 *e= uc + STR_LEN(noper);
2984
786e8c11 2985 for ( ; uc < e ; uc += len ) {
c445ea15 2986
786e8c11 2987 TRIE_READ_CHAR;
c445ea15 2988
786e8c11
YO
2989 if ( uvc < 256 ) {
2990 charid = trie->charmap[ uvc ];
c445ea15 2991 } else {
538e84ed
KW
2992 SV** const svpp = hv_fetch( widecharmap,
2993 (char*)&uvc,
2994 sizeof( UV ),
2995 0);
786e8c11
YO
2996 if ( !svpp ) {
2997 charid = 0;
2998 } else {
2999 charid=(U16)SvIV( *svpp );
3000 }
c445ea15 3001 }
538e84ed
KW
3002 /* charid is now 0 if we dont know the char read, or
3003 * nonzero if we do */
786e8c11 3004 if ( charid ) {
a3621e74 3005
786e8c11
YO
3006 U16 check;
3007 U32 newstate = 0;
a3621e74 3008
786e8c11
YO
3009 charid--;
3010 if ( !trie->states[ state ].trans.list ) {
3011 TRIE_LIST_NEW( state );
c445ea15 3012 }
538e84ed
KW
3013 for ( check = 1;
3014 check <= TRIE_LIST_USED( state );
3015 check++ )
3016 {
3017 if ( TRIE_LIST_ITEM( state, check ).forid
3018 == charid )
3019 {
786e8c11
YO
3020 newstate = TRIE_LIST_ITEM( state, check ).newstate;
3021 break;
3022 }
3023 }
3024 if ( ! newstate ) {
3025 newstate = next_alloc++;
2e64971a 3026 prev_states[newstate] = state;
786e8c11
YO
3027 TRIE_LIST_PUSH( state, charid, newstate );
3028 transcount++;
3029 }
3030 state = newstate;
3031 } else {
147e3846 3032 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %" IVdf, uvc );
c445ea15 3033 }
a28509cc 3034 }
c445ea15 3035 }
3dab1dad 3036 TRIE_HANDLE_WORD(state);
a3621e74
YO
3037
3038 } /* end second pass */
3039
1e2e3d02 3040 /* next alloc is the NEXT state to be allocated */
538e84ed 3041 trie->statecount = next_alloc;
c944940b
JH
3042 trie->states = (reg_trie_state *)
3043 PerlMemShared_realloc( trie->states,
3044 next_alloc
3045 * sizeof(reg_trie_state) );
a3621e74 3046
3dab1dad 3047 /* and now dump it out before we compress it */
2b8b4781
NC
3048 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_list(trie, widecharmap,
3049 revcharmap, next_alloc,
3050 depth+1)
1e2e3d02 3051 );
a3621e74 3052
c944940b
JH
3053 trie->trans = (reg_trie_trans *)
3054 PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) );
a3621e74
YO
3055 {
3056 U32 state;
a3621e74
YO
3057 U32 tp = 0;
3058 U32 zp = 0;
3059
3060
3061 for( state=1 ; state < next_alloc ; state ++ ) {
3062 U32 base=0;
3063
3064 /*
3065 DEBUG_TRIE_COMPILE_MORE_r(
6ad9a8ab 3066 Perl_re_printf( aTHX_ "tp: %d zp: %d ",tp,zp)
a3621e74
YO
3067 );
3068 */
3069
3070 if (trie->states[state].trans.list) {
3071 U16 minid=TRIE_LIST_ITEM( state, 1).forid;
3072 U16 maxid=minid;
a28509cc 3073 U16 idx;
a3621e74
YO
3074
3075 for( idx = 2 ; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
c445ea15
AL
3076 const U16 forid = TRIE_LIST_ITEM( state, idx).forid;
3077 if ( forid < minid ) {
3078 minid=forid;
3079 } else if ( forid > maxid ) {
3080 maxid=forid;
3081 }
a3621e74
YO
3082 }
3083 if ( transcount < tp + maxid - minid + 1) {
3084 transcount *= 2;
c944940b
JH
3085 trie->trans = (reg_trie_trans *)
3086 PerlMemShared_realloc( trie->trans,
446bd890
NC
3087 transcount
3088 * sizeof(reg_trie_trans) );
538e84ed
KW
3089 Zero( trie->trans + (transcount / 2),
3090 transcount / 2,
3091 reg_trie_trans );
a3621e74
YO
3092 }
3093 base = trie->uniquecharcount + tp - minid;
3094 if ( maxid == minid ) {
3095 U32 set = 0;
3096 for ( ; zp < tp ; zp++ ) {
3097 if ( ! trie->trans[ zp ].next ) {
3098 base = trie->uniquecharcount + zp - minid;
538e84ed
KW
3099 trie->trans[ zp ].next = TRIE_LIST_ITEM( state,
3100 1).newstate;
a3621e74
YO
3101 trie->trans[ zp ].check = state;
3102 set = 1;
3103 break;
3104 }
3105 }
3106 if ( !set ) {
538e84ed
KW
3107 trie->trans[ tp ].next = TRIE_LIST_ITEM( state,
3108 1).newstate;
a3621e74
YO
3109 trie->trans[ tp ].check = state;
3110 tp++;
3111 zp = tp;
3112 }
3113 } else {
3114 for ( idx=1; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
538e84ed
KW
3115 const U32 tid = base
3116 - trie->uniquecharcount
3117 + TRIE_LIST_ITEM( state, idx ).forid;
3118 trie->trans[ tid ].next = TRIE_LIST_ITEM( state,
3119 idx ).newstate;
a3621e74
YO
3120 trie->trans[ tid ].check = state;
3121 }
3122 tp += ( maxid - minid + 1 );
3123 }
3124 Safefree(trie->states[ state ].trans.list);
3125 }
3126 /*
3127 DEBUG_TRIE_COMPILE_MORE_r(
6ad9a8ab 3128 Perl_re_printf( aTHX_ " base: %d\n",base);
a3621e74
YO
3129 );
3130 */
3131 trie->states[ state ].trans.base=base;
3132 }
cc601c31 3133 trie->lasttrans = tp + 1;
a3621e74
YO
3134 }
3135 } else {
3136 /*
3137 Second Pass -- Flat Table Representation.
3138
b423522f
KW
3139 we dont use the 0 slot of either trans[] or states[] so we add 1 to
3140 each. We know that we will need Charcount+1 trans at most to store
3141 the data (one row per char at worst case) So we preallocate both
3142 structures assuming worst case.
a3621e74
YO
3143
3144 We then construct the trie using only the .next slots of the entry
3145 structs.
3146
b423522f
KW
3147 We use the .check field of the first entry of the node temporarily
3148 to make compression both faster and easier by keeping track of how
3149 many non zero fields are in the node.
a3621e74
YO
3150
3151 Since trans are numbered from 1 any 0 pointer in the table is a FAIL
3152 transition.
3153
b423522f
KW
3154 There are two terms at use here: state as a TRIE_NODEIDX() which is
3155 a number representing the first entry of the node, and state as a
3156 TRIE_NODENUM() which is the trans number. state 1 is TRIE_NODEIDX(1)
3157 and TRIE_NODENUM(1), state 2 is TRIE_NODEIDX(2) and TRIE_NODENUM(3)
3158 if there are 2 entrys per node. eg:
a3621e74
YO
3159
3160 A B A B
3161 1. 2 4 1. 3 7
3162 2. 0 3 3. 0 5
3163 3. 0 0 5. 0 0
3164 4. 0 0 7. 0 0
3165
b423522f
KW
3166 The table is internally in the right hand, idx form. However as we
3167 also have to deal with the states array which is indexed by nodenum
3168 we have to use TRIE_NODENUM() to convert.
a3621e74
YO
3169
3170 */
6ad9a8ab 3171 DEBUG_TRIE_COMPILE_MORE_r( Perl_re_indentf( aTHX_ "Compiling trie using table compiler\n",
cb41e5d6 3172 depth+1));
3dab1dad 3173
c944940b
JH
3174 trie->trans = (reg_trie_trans *)
3175 PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 )
3176 * trie->uniquecharcount + 1,
3177 sizeof(reg_trie_trans) );
3178 trie->states = (reg_trie_state *)
3179 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
3180 sizeof(reg_trie_state) );
a3621e74
YO
3181 next_alloc = trie->uniquecharcount + 1;
3182
3dab1dad 3183
a3621e74
YO
3184 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
3185
df826430 3186 regnode *noper = NEXTOPER( cur );
a3621e74
YO
3187
3188 U32 state = 1; /* required init */
3189
3190 U16 charid = 0; /* sanity init */
3191 U32 accept_state = 0; /* sanity init */
a3621e74 3192
07be1b83 3193 U32 wordlen = 0; /* required init */
a3621e74 3194
df826430
YO
3195 if (OP(noper) == NOTHING) {
3196 regnode *noper_next= regnext(noper);
944e05e3
YO
3197 if (noper_next < tail)
3198 noper= noper_next;
df826430 3199 }
fab2782b 3200
f6b4b99d
KW
3201 if ( noper < tail
3202 && ( OP(noper) == flags
3203 || (flags == EXACT && OP(noper) == EXACT_ONLY8)
a9f8c7ac 3204 || (flags == EXACTFU && ( OP(noper) == EXACTFU_ONLY8
0ea669f4 3205 || OP(noper) == EXACTFUP))))
f6b4b99d 3206 {
944e05e3
YO
3207 const U8 *uc= (U8*)STRING(noper);
3208 const U8 *e= uc + STR_LEN(noper);
3209
786e8c11 3210 for ( ; uc < e ; uc += len ) {
a3621e74 3211
786e8c11 3212 TRIE_READ_CHAR;
a3621e74 3213
786e8c11
YO
3214 if ( uvc < 256 ) {
3215 charid = trie->charmap[ uvc ];
3216 } else {
538e84ed
KW
3217 SV* const * const svpp = hv_fetch( widecharmap,
3218 (char*)&uvc,
3219 sizeof( UV ),
3220 0);
786e8c11 3221 charid = svpp ? (U16)SvIV(*svpp) : 0;
a3621e74 3222 }
786e8c11
YO
3223 if ( charid ) {
3224 charid--;
3225 if ( !trie->trans[ state + charid ].next ) {
3226 trie->trans[ state + charid ].next = next_alloc;
3227 trie->trans[ state ].check++;
2e64971a
DM
3228 prev_states[TRIE_NODENUM(next_alloc)]
3229 = TRIE_NODENUM(state);
786e8c11
YO
3230 next_alloc += trie->uniquecharcount;
3231 }
3232 state = trie->trans[ state + charid ].next;
3233 } else {
147e3846 3234 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %" IVdf, uvc );
786e8c11 3235 }
538e84ed
KW
3236 /* charid is now 0 if we dont know the char read, or
3237 * nonzero if we do */
a3621e74 3238 }
a3621e74 3239 }
3dab1dad
YO
3240 accept_state = TRIE_NODENUM( state );
3241 TRIE_HANDLE_WORD(accept_state);
a3621e74
YO
3242
3243 } /* end second pass */
3244
3dab1dad 3245 /* and now dump it out before we compress it */
2b8b4781
NC
3246 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_table(trie, widecharmap,
3247 revcharmap,
3248 next_alloc, depth+1));
a3621e74 3249
a3621e74
YO
3250 {
3251 /*
3252 * Inplace compress the table.*
3253
3254 For sparse data sets the table constructed by the trie algorithm will
3255 be mostly 0/FAIL transitions or to put it another way mostly empty.
3256 (Note that leaf nodes will not contain any transitions.)
3257
3258 This algorithm compresses the tables by eliminating most such
3259 transitions, at the cost of a modest bit of extra work during lookup:
3260
3261 - Each states[] entry contains a .base field which indicates the
3262 index in the state[] array wheres its transition data is stored.
3263
3b753521 3264 - If .base is 0 there are no valid transitions from that node.
a3621e74
YO
3265
3266 - If .base is nonzero then charid is added to it to find an entry in
3267 the trans array.
3268
3269 -If trans[states[state].base+charid].check!=state then the
3270 transition is taken to be a 0/Fail transition. Thus if there are fail
3271 transitions at the front of the node then the .base offset will point
3272 somewhere inside the previous nodes data (or maybe even into a node
3273 even earlier), but the .check field determines if the transition is
3274 valid.
3275
786e8c11 3276 XXX - wrong maybe?
a3621e74 3277 The following process inplace converts the table to the compressed
3b753521 3278 table: We first do not compress the root node 1,and mark all its
a3621e74 3279 .check pointers as 1 and set its .base pointer as 1 as well. This
3b753521
FN
3280 allows us to do a DFA construction from the compressed table later,
3281 and ensures that any .base pointers we calculate later are greater
3282 than 0.
a3621e74
YO
3283
3284 - We set 'pos' to indicate the first entry of the second node.
3285
3286 - We then iterate over the columns of the node, finding the first and
3287 last used entry at l and m. We then copy l..m into pos..(pos+m-l),
3288 and set the .check pointers accordingly, and advance pos
3289 appropriately and repreat for the next node. Note that when we copy
3290 the next pointers we have to convert them from the original
3291 NODEIDX form to NODENUM form as the former is not valid post
3292 compression.
3293
3294 - If a node has no transitions used we mark its base as 0 and do not
3295 advance the pos pointer.
3296
3297 - If a node only has one transition we use a second pointer into the
3298 structure to fill in allocated fail transitions from other states.
3299 This pointer is independent of the main pointer and scans forward
3300 looking for null transitions that are allocated to a state. When it
3301 finds one it writes the single transition into the "hole". If the
786e8c11 3302 pointer doesnt find one the single transition is appended as normal.
a3621e74
YO
3303
3304 - Once compressed we can Renew/realloc the structures to release the
3305 excess space.
3306
3307 See "Table-Compression Methods" in sec 3.9 of the Red Dragon,
3308 specifically Fig 3.47 and the associated pseudocode.
3309
3310 demq
3311 */
a3b680e6 3312 const U32 laststate = TRIE_NODENUM( next_alloc );
a28509cc 3313 U32 state, charid;
a3621e74 3314 U32 pos = 0, zp=0;
1e2e3d02 3315 trie->statecount = laststate;
a3621e74
YO
3316
3317 for ( state = 1 ; state < laststate ; state++ ) {
3318 U8 flag = 0;
a28509cc
AL
3319 const U32 stateidx = TRIE_NODEIDX( state );
3320 const U32 o_used = trie->trans[ stateidx ].check;
3321 U32 used = trie->trans[ stateidx ].check;
a3621e74
YO
3322 trie->trans[ stateidx ].check = 0;
3323
538e84ed
KW
3324 for ( charid = 0;
3325 used && charid < trie->uniquecharcount;
3326 charid++ )
3327 {
a3621e74
YO
3328 if ( flag || trie->trans[ stateidx + charid ].next ) {
3329 if ( trie->trans[ stateidx + charid ].next ) {
3330 if (o_used == 1) {
3331 for ( ; zp < pos ; zp++ ) {
3332 if ( ! trie->trans[ zp ].next ) {
3333 break;
3334 }
3335 }
538e84ed
KW
3336 trie->states[ state ].trans.base
3337 = zp
3338 + trie->uniquecharcount
3339 - charid ;
3340 trie->trans[ zp ].next
3341 = SAFE_TRIE_NODENUM( trie->trans[ stateidx
3342 + charid ].next );
a3621e74
YO
3343 trie->trans[ zp ].check = state;
3344 if ( ++zp > pos ) pos = zp;
3345 break;
3346 }
3347 used--;
3348 }
3349 if ( !flag ) {
3350 flag = 1;
538e84ed
KW
3351 trie->states[ state ].trans.base
3352 = pos + trie->uniquecharcount - charid ;
a3621e74 3353 }
538e84ed
KW
3354 trie->trans[ pos ].next
3355 = SAFE_TRIE_NODENUM(
3356 trie->trans[ stateidx + charid ].next );
a3621e74
YO
3357 trie->trans[ pos ].check = state;
3358 pos++;
3359 }
3360 }
3361 }
cc601c31 3362 trie->lasttrans = pos + 1;
c944940b
JH
3363 trie->states = (reg_trie_state *)
3364 PerlMemShared_realloc( trie->states, laststate
3365 * sizeof(reg_trie_state) );
a3621e74 3366 DEBUG_TRIE_COMPILE_MORE_r(
147e3846 3367 Perl_re_indentf( aTHX_ "Alloc: %d Orig: %" IVdf " elements, Final:%" IVdf ". Savings of %%%5.2f\n",
cb41e5d6 3368 depth+1,
538e84ed
KW
3369 (int)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount
3370 + 1 ),
3371 (IV)next_alloc,
3372 (IV)pos,
3373 ( ( next_alloc - pos ) * 100 ) / (double)next_alloc );
a3621e74
YO
3374 );
3375
3376 } /* end table compress */
3377 }
1e2e3d02 3378 DEBUG_TRIE_COMPILE_MORE_r(
147e3846 3379 Perl_re_indentf( aTHX_ "Statecount:%" UVxf " Lasttrans:%" UVxf "\n",
cb41e5d6 3380 depth+1,
1e2e3d02
YO
3381 (UV)trie->statecount,
3382 (UV)trie->lasttrans)
3383 );
cc601c31 3384 /* resize the trans array to remove unused space */
c944940b
JH
3385 trie->trans = (reg_trie_trans *)
3386 PerlMemShared_realloc( trie->trans, trie->lasttrans
3387 * sizeof(reg_trie_trans) );
a3621e74 3388
538e84ed 3389 { /* Modify the program and insert the new TRIE node */
3dab1dad
YO
3390 U8 nodetype =(U8)(flags & 0xFF);
3391 char *str=NULL;
538e84ed 3392
07be1b83 3393#ifdef DEBUGGING
e62cc96a 3394 regnode *optimize = NULL;
7122b237
YO
3395#ifdef RE_TRACK_PATTERN_OFFSETS
3396
b57a0404
JH
3397 U32 mjd_offset = 0;
3398 U32 mjd_nodelen = 0;
7122b237
YO
3399#endif /* RE_TRACK_PATTERN_OFFSETS */
3400#endif /* DEBUGGING */
a3621e74 3401 /*
3dab1dad
YO
3402 This means we convert either the first branch or the first Exact,
3403 depending on whether the thing following (in 'last') is a branch
3404 or not and whther first is the startbranch (ie is it a sub part of
3405 the alternation or is it the whole thing.)
3b753521 3406 Assuming its a sub part we convert the EXACT otherwise we convert
3dab1dad 3407 the whole branch sequence, including the first.
a3621e74 3408 */
3dab1dad 3409 /* Find the node we are going to overwrite */
7f69552c 3410 if ( first != startbranch || OP( last ) == BRANCH ) {
07be1b83 3411 /* branch sub-chain */
3dab1dad 3412 NEXT_OFF( first ) = (U16)(last - first);
7122b237 3413#ifdef RE_TRACK_PATTERN_OFFSETS
07be1b83
YO
3414 DEBUG_r({
3415 mjd_offset= Node_Offset((convert));
3416 mjd_nodelen= Node_Length((convert));
3417 });
7122b237 3418#endif
7f69552c 3419 /* whole branch chain */
7122b237
YO
3420 }
3421#ifdef RE_TRACK_PATTERN_OFFSETS
3422 else {
7f69552c
YO
3423 DEBUG_r({
3424 const regnode *nop = NEXTOPER( convert );
3425 mjd_offset= Node_Offset((nop));
3426 mjd_nodelen= Node_Length((nop));
3427 });
07be1b83
YO
3428 }
3429 DEBUG_OPTIMISE_r(
147e3846 3430 Perl_re_indentf( aTHX_ "MJD offset:%" UVuf " MJD length:%" UVuf "\n",