Commit | Line | Data |
---|---|---|
a0d0e21e LW |
1 | /* regcomp.c |
2 | */ | |
3 | ||
4 | /* | |
5 | * "A fair jaw-cracker dwarf-language must be." --Samwise Gamgee | |
6 | */ | |
7 | ||
61296642 DM |
8 | /* This file contains functions for compiling a regular expression. See |
9 | * also regexec.c which funnily enough, contains functions for executing | |
166f8a29 | 10 | * a regular expression. |
e4a054ea DM |
11 | * |
12 | * This file is also copied at build time to ext/re/re_comp.c, where | |
13 | * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT. | |
14 | * This causes the main functions to be compiled under new names and with | |
15 | * debugging support added, which makes "use re 'debug'" work. | |
166f8a29 DM |
16 | */ |
17 | ||
a687059c LW |
18 | /* NOTE: this is derived from Henry Spencer's regexp code, and should not |
19 | * confused with the original package (see point 3 below). Thanks, Henry! | |
20 | */ | |
21 | ||
22 | /* Additional note: this code is very heavily munged from Henry's version | |
23 | * in places. In some spots I've traded clarity for efficiency, so don't | |
24 | * blame Henry for some of the lack of readability. | |
25 | */ | |
26 | ||
e50aee73 AD |
27 | /* The names of the functions have been changed from regcomp and |
28 | * regexec to pregcomp and pregexec in order to avoid conflicts | |
29 | * with the POSIX routines of the same names. | |
30 | */ | |
31 | ||
b9d5759e | 32 | #ifdef PERL_EXT_RE_BUILD |
54df2634 | 33 | #include "re_top.h" |
b81d288d | 34 | #endif |
56953603 | 35 | |
a687059c | 36 | /* |
e50aee73 | 37 | * pregcomp and pregexec -- regsub and regerror are not used in perl |
a687059c LW |
38 | * |
39 | * Copyright (c) 1986 by University of Toronto. | |
40 | * Written by Henry Spencer. Not derived from licensed software. | |
41 | * | |
42 | * Permission is granted to anyone to use this software for any | |
43 | * purpose on any computer system, and to redistribute it freely, | |
44 | * subject to the following restrictions: | |
45 | * | |
46 | * 1. The author is not responsible for the consequences of use of | |
47 | * this software, no matter how awful, even if they arise | |
48 | * from defects in it. | |
49 | * | |
50 | * 2. The origin of this software must not be misrepresented, either | |
51 | * by explicit claim or by omission. | |
52 | * | |
53 | * 3. Altered versions must be plainly marked as such, and must not | |
54 | * be misrepresented as being the original software. | |
55 | * | |
56 | * | |
57 | **** Alterations to Henry's code are... | |
58 | **** | |
4bb101f2 | 59 | **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, |
fdf8c088 | 60 | **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007 by Larry Wall and others |
a687059c | 61 | **** |
9ef589d8 LW |
62 | **** You may distribute under the terms of either the GNU General Public |
63 | **** License or the Artistic License, as specified in the README file. | |
64 | ||
a687059c LW |
65 | * |
66 | * Beware that some of this code is subtly aware of the way operator | |
67 | * precedence is structured in regular expressions. Serious changes in | |
68 | * regular-expression syntax might require a total rethink. | |
69 | */ | |
70 | #include "EXTERN.h" | |
864dbfa3 | 71 | #define PERL_IN_REGCOMP_C |
a687059c | 72 | #include "perl.h" |
d06ea78c | 73 | |
acfe0abc | 74 | #ifndef PERL_IN_XSUB_RE |
d06ea78c GS |
75 | # include "INTERN.h" |
76 | #endif | |
c277df42 IZ |
77 | |
78 | #define REG_COMP_C | |
54df2634 NC |
79 | #ifdef PERL_IN_XSUB_RE |
80 | # include "re_comp.h" | |
81 | #else | |
82 | # include "regcomp.h" | |
83 | #endif | |
a687059c | 84 | |
d4cce5f1 | 85 | #ifdef op |
11343788 | 86 | #undef op |
d4cce5f1 | 87 | #endif /* op */ |
11343788 | 88 | |
fe14fcc3 | 89 | #ifdef MSDOS |
7e4e8c89 | 90 | # if defined(BUGGY_MSC6) |
fe14fcc3 | 91 | /* MSC 6.00A breaks on op/regexp.t test 85 unless we turn this off */ |
7e4e8c89 | 92 | # pragma optimize("a",off) |
fe14fcc3 | 93 | /* But MSC 6.00A is happy with 'w', for aliases only across function calls*/ |
7e4e8c89 NC |
94 | # pragma optimize("w",on ) |
95 | # endif /* BUGGY_MSC6 */ | |
fe14fcc3 LW |
96 | #endif /* MSDOS */ |
97 | ||
a687059c LW |
98 | #ifndef STATIC |
99 | #define STATIC static | |
100 | #endif | |
101 | ||
830247a4 | 102 | typedef struct RExC_state_t { |
e2509266 | 103 | U32 flags; /* are we folding, multilining? */ |
830247a4 | 104 | char *precomp; /* uncompiled string. */ |
f8fc2ecf YO |
105 | regexp *rx; /* perl core regexp structure */ |
106 | regexp_internal *rxi; /* internal data for regexp object pprivate field */ | |
fac92740 | 107 | char *start; /* Start of input for compile */ |
830247a4 IZ |
108 | char *end; /* End of input for compile */ |
109 | char *parse; /* Input-scan pointer. */ | |
110 | I32 whilem_seen; /* number of WHILEM in this expr */ | |
fac92740 | 111 | regnode *emit_start; /* Start of emitted-code area */ |
3b57cd43 | 112 | regnode *emit_bound; /* First regnode outside of the allocated space */ |
ffc61ed2 | 113 | regnode *emit; /* Code-emit pointer; ®dummy = don't = compiling */ |
830247a4 IZ |
114 | I32 naughty; /* How bad is this pattern? */ |
115 | I32 sawback; /* Did we see \1, ...? */ | |
116 | U32 seen; | |
117 | I32 size; /* Code size. */ | |
c74340f9 YO |
118 | I32 npar; /* Capture buffer count, (OPEN). */ |
119 | I32 cpar; /* Capture buffer count, (CLOSE). */ | |
e2e6a0f1 | 120 | I32 nestroot; /* root parens we are in - used by accept */ |
830247a4 IZ |
121 | I32 extralen; |
122 | I32 seen_zerolen; | |
123 | I32 seen_evals; | |
40d049e4 YO |
124 | regnode **open_parens; /* pointers to open parens */ |
125 | regnode **close_parens; /* pointers to close parens */ | |
126 | regnode *opend; /* END node in program */ | |
02daf0ab YO |
127 | I32 utf8; /* whether the pattern is utf8 or not */ |
128 | I32 orig_utf8; /* whether the pattern was originally in utf8 */ | |
129 | /* XXX use this for future optimisation of case | |
130 | * where pattern must be upgraded to utf8. */ | |
6bda09f9 | 131 | HV *charnames; /* cache of named sequences */ |
81714fb9 | 132 | HV *paren_names; /* Paren names */ |
1f1031fe | 133 | |
40d049e4 YO |
134 | regnode **recurse; /* Recurse regops */ |
135 | I32 recurse_count; /* Number of recurse regops */ | |
830247a4 IZ |
136 | #if ADD_TO_REGEXEC |
137 | char *starttry; /* -Dr: where regtry was called. */ | |
138 | #define RExC_starttry (pRExC_state->starttry) | |
139 | #endif | |
3dab1dad | 140 | #ifdef DEBUGGING |
be8e71aa | 141 | const char *lastparse; |
3dab1dad | 142 | I32 lastnum; |
1f1031fe | 143 | AV *paren_name_list; /* idx -> name */ |
3dab1dad YO |
144 | #define RExC_lastparse (pRExC_state->lastparse) |
145 | #define RExC_lastnum (pRExC_state->lastnum) | |
1f1031fe | 146 | #define RExC_paren_name_list (pRExC_state->paren_name_list) |
3dab1dad | 147 | #endif |
830247a4 IZ |
148 | } RExC_state_t; |
149 | ||
e2509266 | 150 | #define RExC_flags (pRExC_state->flags) |
830247a4 IZ |
151 | #define RExC_precomp (pRExC_state->precomp) |
152 | #define RExC_rx (pRExC_state->rx) | |
f8fc2ecf | 153 | #define RExC_rxi (pRExC_state->rxi) |
fac92740 | 154 | #define RExC_start (pRExC_state->start) |
830247a4 IZ |
155 | #define RExC_end (pRExC_state->end) |
156 | #define RExC_parse (pRExC_state->parse) | |
157 | #define RExC_whilem_seen (pRExC_state->whilem_seen) | |
7122b237 YO |
158 | #ifdef RE_TRACK_PATTERN_OFFSETS |
159 | #define RExC_offsets (pRExC_state->rxi->u.offsets) /* I am not like the others */ | |
160 | #endif | |
830247a4 | 161 | #define RExC_emit (pRExC_state->emit) |
fac92740 | 162 | #define RExC_emit_start (pRExC_state->emit_start) |
3b57cd43 | 163 | #define RExC_emit_bound (pRExC_state->emit_bound) |
830247a4 IZ |
164 | #define RExC_naughty (pRExC_state->naughty) |
165 | #define RExC_sawback (pRExC_state->sawback) | |
166 | #define RExC_seen (pRExC_state->seen) | |
167 | #define RExC_size (pRExC_state->size) | |
168 | #define RExC_npar (pRExC_state->npar) | |
e2e6a0f1 | 169 | #define RExC_nestroot (pRExC_state->nestroot) |
830247a4 IZ |
170 | #define RExC_extralen (pRExC_state->extralen) |
171 | #define RExC_seen_zerolen (pRExC_state->seen_zerolen) | |
172 | #define RExC_seen_evals (pRExC_state->seen_evals) | |
1aa99e6b | 173 | #define RExC_utf8 (pRExC_state->utf8) |
02daf0ab | 174 | #define RExC_orig_utf8 (pRExC_state->orig_utf8) |
fc8cd66c | 175 | #define RExC_charnames (pRExC_state->charnames) |
40d049e4 YO |
176 | #define RExC_open_parens (pRExC_state->open_parens) |
177 | #define RExC_close_parens (pRExC_state->close_parens) | |
178 | #define RExC_opend (pRExC_state->opend) | |
81714fb9 | 179 | #define RExC_paren_names (pRExC_state->paren_names) |
40d049e4 YO |
180 | #define RExC_recurse (pRExC_state->recurse) |
181 | #define RExC_recurse_count (pRExC_state->recurse_count) | |
830247a4 | 182 | |
cde0cee5 | 183 | |
a687059c LW |
184 | #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?') |
185 | #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \ | |
186 | ((*s) == '{' && regcurly(s))) | |
a687059c | 187 | |
35c8bce7 LW |
188 | #ifdef SPSTART |
189 | #undef SPSTART /* dratted cpp namespace... */ | |
190 | #endif | |
a687059c LW |
191 | /* |
192 | * Flags to be passed up and down. | |
193 | */ | |
a687059c | 194 | #define WORST 0 /* Worst case. */ |
a3b492c3 YO |
195 | #define HASWIDTH 0x01 /* Known to match non-null strings. */ |
196 | #define SIMPLE 0x02 /* Simple enough to be STAR/PLUS operand. */ | |
197 | #define SPSTART 0x04 /* Starts with * or +. */ | |
198 | #define TRYAGAIN 0x08 /* Weeded out a declaration. */ | |
199 | #define POSTPONED 0x10 /* (?1),(?&name), (??{...}) or similar */ | |
a687059c | 200 | |
3dab1dad YO |
201 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1) |
202 | ||
07be1b83 YO |
203 | /* whether trie related optimizations are enabled */ |
204 | #if PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION | |
205 | #define TRIE_STUDY_OPT | |
786e8c11 | 206 | #define FULL_TRIE_STUDY |
07be1b83 YO |
207 | #define TRIE_STCLASS |
208 | #endif | |
1de06328 YO |
209 | |
210 | ||
40d049e4 YO |
211 | |
212 | #define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3] | |
213 | #define PBITVAL(paren) (1 << ((paren) & 7)) | |
214 | #define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren)) | |
215 | #define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren) | |
216 | #define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren)) | |
217 | ||
218 | ||
1de06328 YO |
219 | /* About scan_data_t. |
220 | ||
221 | During optimisation we recurse through the regexp program performing | |
222 | various inplace (keyhole style) optimisations. In addition study_chunk | |
223 | and scan_commit populate this data structure with information about | |
224 | what strings MUST appear in the pattern. We look for the longest | |
225 | string that must appear for at a fixed location, and we look for the | |
226 | longest string that may appear at a floating location. So for instance | |
227 | in the pattern: | |
228 | ||
229 | /FOO[xX]A.*B[xX]BAR/ | |
230 | ||
231 | Both 'FOO' and 'A' are fixed strings. Both 'B' and 'BAR' are floating | |
232 | strings (because they follow a .* construct). study_chunk will identify | |
233 | both FOO and BAR as being the longest fixed and floating strings respectively. | |
234 | ||
235 | The strings can be composites, for instance | |
236 | ||
237 | /(f)(o)(o)/ | |
238 | ||
239 | will result in a composite fixed substring 'foo'. | |
240 | ||
241 | For each string some basic information is maintained: | |
242 | ||
243 | - offset or min_offset | |
244 | This is the position the string must appear at, or not before. | |
245 | It also implicitly (when combined with minlenp) tells us how many | |
246 | character must match before the string we are searching. | |
247 | Likewise when combined with minlenp and the length of the string | |
248 | tells us how many characters must appear after the string we have | |
249 | found. | |
250 | ||
251 | - max_offset | |
252 | Only used for floating strings. This is the rightmost point that | |
253 | the string can appear at. Ifset to I32 max it indicates that the | |
254 | string can occur infinitely far to the right. | |
255 | ||
256 | - minlenp | |
257 | A pointer to the minimum length of the pattern that the string | |
258 | was found inside. This is important as in the case of positive | |
259 | lookahead or positive lookbehind we can have multiple patterns | |
260 | involved. Consider | |
261 | ||
262 | /(?=FOO).*F/ | |
263 | ||
264 | The minimum length of the pattern overall is 3, the minimum length | |
265 | of the lookahead part is 3, but the minimum length of the part that | |
266 | will actually match is 1. So 'FOO's minimum length is 3, but the | |
267 | minimum length for the F is 1. This is important as the minimum length | |
268 | is used to determine offsets in front of and behind the string being | |
269 | looked for. Since strings can be composites this is the length of the | |
270 | pattern at the time it was commited with a scan_commit. Note that | |
271 | the length is calculated by study_chunk, so that the minimum lengths | |
272 | are not known until the full pattern has been compiled, thus the | |
273 | pointer to the value. | |
274 | ||
275 | - lookbehind | |
276 | ||
277 | In the case of lookbehind the string being searched for can be | |
278 | offset past the start point of the final matching string. | |
279 | If this value was just blithely removed from the min_offset it would | |
280 | invalidate some of the calculations for how many chars must match | |
281 | before or after (as they are derived from min_offset and minlen and | |
282 | the length of the string being searched for). | |
283 | When the final pattern is compiled and the data is moved from the | |
284 | scan_data_t structure into the regexp structure the information | |
285 | about lookbehind is factored in, with the information that would | |
286 | have been lost precalculated in the end_shift field for the | |
287 | associated string. | |
288 | ||
289 | The fields pos_min and pos_delta are used to store the minimum offset | |
290 | and the delta to the maximum offset at the current point in the pattern. | |
291 | ||
292 | */ | |
2c2d71f5 JH |
293 | |
294 | typedef struct scan_data_t { | |
1de06328 YO |
295 | /*I32 len_min; unused */ |
296 | /*I32 len_delta; unused */ | |
2c2d71f5 JH |
297 | I32 pos_min; |
298 | I32 pos_delta; | |
299 | SV *last_found; | |
1de06328 | 300 | I32 last_end; /* min value, <0 unless valid. */ |
2c2d71f5 JH |
301 | I32 last_start_min; |
302 | I32 last_start_max; | |
1de06328 YO |
303 | SV **longest; /* Either &l_fixed, or &l_float. */ |
304 | SV *longest_fixed; /* longest fixed string found in pattern */ | |
305 | I32 offset_fixed; /* offset where it starts */ | |
306 | I32 *minlen_fixed; /* pointer to the minlen relevent to the string */ | |
307 | I32 lookbehind_fixed; /* is the position of the string modfied by LB */ | |
308 | SV *longest_float; /* longest floating string found in pattern */ | |
309 | I32 offset_float_min; /* earliest point in string it can appear */ | |
310 | I32 offset_float_max; /* latest point in string it can appear */ | |
311 | I32 *minlen_float; /* pointer to the minlen relevent to the string */ | |
312 | I32 lookbehind_float; /* is the position of the string modified by LB */ | |
2c2d71f5 JH |
313 | I32 flags; |
314 | I32 whilem_c; | |
cb434fcc | 315 | I32 *last_closep; |
653099ff | 316 | struct regnode_charclass_class *start_class; |
2c2d71f5 JH |
317 | } scan_data_t; |
318 | ||
a687059c | 319 | /* |
e50aee73 | 320 | * Forward declarations for pregcomp()'s friends. |
a687059c | 321 | */ |
a0d0e21e | 322 | |
27da23d5 | 323 | static const scan_data_t zero_scan_data = |
1de06328 | 324 | { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ,0}; |
c277df42 IZ |
325 | |
326 | #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL) | |
07be1b83 YO |
327 | #define SF_BEFORE_SEOL 0x0001 |
328 | #define SF_BEFORE_MEOL 0x0002 | |
c277df42 IZ |
329 | #define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL) |
330 | #define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL) | |
331 | ||
09b7f37c CB |
332 | #ifdef NO_UNARY_PLUS |
333 | # define SF_FIX_SHIFT_EOL (0+2) | |
334 | # define SF_FL_SHIFT_EOL (0+4) | |
335 | #else | |
336 | # define SF_FIX_SHIFT_EOL (+2) | |
337 | # define SF_FL_SHIFT_EOL (+4) | |
338 | #endif | |
c277df42 IZ |
339 | |
340 | #define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL) | |
341 | #define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL) | |
342 | ||
343 | #define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL) | |
344 | #define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */ | |
07be1b83 YO |
345 | #define SF_IS_INF 0x0040 |
346 | #define SF_HAS_PAR 0x0080 | |
347 | #define SF_IN_PAR 0x0100 | |
348 | #define SF_HAS_EVAL 0x0200 | |
349 | #define SCF_DO_SUBSTR 0x0400 | |
653099ff GS |
350 | #define SCF_DO_STCLASS_AND 0x0800 |
351 | #define SCF_DO_STCLASS_OR 0x1000 | |
352 | #define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR) | |
e1901655 | 353 | #define SCF_WHILEM_VISITED_POS 0x2000 |
c277df42 | 354 | |
786e8c11 | 355 | #define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */ |
e2e6a0f1 | 356 | #define SCF_SEEN_ACCEPT 0x8000 |
07be1b83 | 357 | |
eb160463 | 358 | #define UTF (RExC_utf8 != 0) |
bbe252da YO |
359 | #define LOC ((RExC_flags & RXf_PMf_LOCALE) != 0) |
360 | #define FOLD ((RExC_flags & RXf_PMf_FOLD) != 0) | |
a0ed51b3 | 361 | |
ffc61ed2 | 362 | #define OOB_UNICODE 12345678 |
93733859 | 363 | #define OOB_NAMEDCLASS -1 |
b8c5462f | 364 | |
a0ed51b3 LW |
365 | #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv)) |
366 | #define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b) | |
367 | ||
8615cb43 | 368 | |
b45f050a JF |
369 | /* length of regex to show in messages that don't mark a position within */ |
370 | #define RegexLengthToShowInErrorMessages 127 | |
371 | ||
372 | /* | |
373 | * If MARKER[12] are adjusted, be sure to adjust the constants at the top | |
374 | * of t/op/regmesg.t, the tests in t/op/re_tests, and those in | |
375 | * op/pragma/warn/regcomp. | |
376 | */ | |
7253e4e3 RK |
377 | #define MARKER1 "<-- HERE" /* marker as it appears in the description */ |
378 | #define MARKER2 " <-- HERE " /* marker as it appears within the regex */ | |
b81d288d | 379 | |
7253e4e3 | 380 | #define REPORT_LOCATION " in regex; marked by " MARKER1 " in m/%.*s" MARKER2 "%s/" |
b45f050a JF |
381 | |
382 | /* | |
383 | * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given | |
384 | * arg. Show regex, up to a maximum length. If it's too long, chop and add | |
385 | * "...". | |
386 | */ | |
58e23c8d | 387 | #define _FAIL(code) STMT_START { \ |
bfed75c6 | 388 | const char *ellipses = ""; \ |
ccb2c380 MP |
389 | IV len = RExC_end - RExC_precomp; \ |
390 | \ | |
391 | if (!SIZE_ONLY) \ | |
392 | SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \ | |
393 | if (len > RegexLengthToShowInErrorMessages) { \ | |
394 | /* chop 10 shorter than the max, to ensure meaning of "..." */ \ | |
395 | len = RegexLengthToShowInErrorMessages - 10; \ | |
396 | ellipses = "..."; \ | |
397 | } \ | |
58e23c8d | 398 | code; \ |
ccb2c380 | 399 | } STMT_END |
8615cb43 | 400 | |
58e23c8d YO |
401 | #define FAIL(msg) _FAIL( \ |
402 | Perl_croak(aTHX_ "%s in regex m/%.*s%s/", \ | |
403 | msg, (int)len, RExC_precomp, ellipses)) | |
404 | ||
405 | #define FAIL2(msg,arg) _FAIL( \ | |
406 | Perl_croak(aTHX_ msg " in regex m/%.*s%s/", \ | |
407 | arg, (int)len, RExC_precomp, ellipses)) | |
408 | ||
b45f050a | 409 | /* |
b45f050a JF |
410 | * Simple_vFAIL -- like FAIL, but marks the current location in the scan |
411 | */ | |
ccb2c380 | 412 | #define Simple_vFAIL(m) STMT_START { \ |
a28509cc | 413 | const IV offset = RExC_parse - RExC_precomp; \ |
ccb2c380 MP |
414 | Perl_croak(aTHX_ "%s" REPORT_LOCATION, \ |
415 | m, (int)offset, RExC_precomp, RExC_precomp + offset); \ | |
416 | } STMT_END | |
b45f050a JF |
417 | |
418 | /* | |
419 | * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL() | |
420 | */ | |
ccb2c380 MP |
421 | #define vFAIL(m) STMT_START { \ |
422 | if (!SIZE_ONLY) \ | |
423 | SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \ | |
424 | Simple_vFAIL(m); \ | |
425 | } STMT_END | |
b45f050a JF |
426 | |
427 | /* | |
428 | * Like Simple_vFAIL(), but accepts two arguments. | |
429 | */ | |
ccb2c380 | 430 | #define Simple_vFAIL2(m,a1) STMT_START { \ |
a28509cc | 431 | const IV offset = RExC_parse - RExC_precomp; \ |
ccb2c380 MP |
432 | S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, \ |
433 | (int)offset, RExC_precomp, RExC_precomp + offset); \ | |
434 | } STMT_END | |
b45f050a JF |
435 | |
436 | /* | |
437 | * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2(). | |
438 | */ | |
ccb2c380 MP |
439 | #define vFAIL2(m,a1) STMT_START { \ |
440 | if (!SIZE_ONLY) \ | |
441 | SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \ | |
442 | Simple_vFAIL2(m, a1); \ | |
443 | } STMT_END | |
b45f050a JF |
444 | |
445 | ||
446 | /* | |
447 | * Like Simple_vFAIL(), but accepts three arguments. | |
448 | */ | |
ccb2c380 | 449 | #define Simple_vFAIL3(m, a1, a2) STMT_START { \ |
a28509cc | 450 | const IV offset = RExC_parse - RExC_precomp; \ |
ccb2c380 MP |
451 | S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, \ |
452 | (int)offset, RExC_precomp, RExC_precomp + offset); \ | |
453 | } STMT_END | |
b45f050a JF |
454 | |
455 | /* | |
456 | * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3(). | |
457 | */ | |
ccb2c380 MP |
458 | #define vFAIL3(m,a1,a2) STMT_START { \ |
459 | if (!SIZE_ONLY) \ | |
460 | SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \ | |
461 | Simple_vFAIL3(m, a1, a2); \ | |
462 | } STMT_END | |
b45f050a JF |
463 | |
464 | /* | |
465 | * Like Simple_vFAIL(), but accepts four arguments. | |
466 | */ | |
ccb2c380 | 467 | #define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \ |
a28509cc | 468 | const IV offset = RExC_parse - RExC_precomp; \ |
ccb2c380 MP |
469 | S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3, \ |
470 | (int)offset, RExC_precomp, RExC_precomp + offset); \ | |
471 | } STMT_END | |
b45f050a | 472 | |
ccb2c380 | 473 | #define vWARN(loc,m) STMT_START { \ |
a28509cc | 474 | const IV offset = loc - RExC_precomp; \ |
ccb2c380 MP |
475 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s" REPORT_LOCATION, \ |
476 | m, (int)offset, RExC_precomp, RExC_precomp + offset); \ | |
477 | } STMT_END | |
478 | ||
479 | #define vWARNdep(loc,m) STMT_START { \ | |
a28509cc | 480 | const IV offset = loc - RExC_precomp; \ |
ccb2c380 MP |
481 | Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \ |
482 | "%s" REPORT_LOCATION, \ | |
483 | m, (int)offset, RExC_precomp, RExC_precomp + offset); \ | |
484 | } STMT_END | |
485 | ||
486 | ||
487 | #define vWARN2(loc, m, a1) STMT_START { \ | |
a28509cc | 488 | const IV offset = loc - RExC_precomp; \ |
ccb2c380 MP |
489 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \ |
490 | a1, (int)offset, RExC_precomp, RExC_precomp + offset); \ | |
491 | } STMT_END | |
492 | ||
493 | #define vWARN3(loc, m, a1, a2) STMT_START { \ | |
a28509cc | 494 | const IV offset = loc - RExC_precomp; \ |
ccb2c380 MP |
495 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \ |
496 | a1, a2, (int)offset, RExC_precomp, RExC_precomp + offset); \ | |
497 | } STMT_END | |
498 | ||
499 | #define vWARN4(loc, m, a1, a2, a3) STMT_START { \ | |
a28509cc | 500 | const IV offset = loc - RExC_precomp; \ |
ccb2c380 MP |
501 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \ |
502 | a1, a2, a3, (int)offset, RExC_precomp, RExC_precomp + offset); \ | |
503 | } STMT_END | |
504 | ||
505 | #define vWARN5(loc, m, a1, a2, a3, a4) STMT_START { \ | |
a28509cc | 506 | const IV offset = loc - RExC_precomp; \ |
ccb2c380 MP |
507 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \ |
508 | a1, a2, a3, a4, (int)offset, RExC_precomp, RExC_precomp + offset); \ | |
509 | } STMT_END | |
9d1d55b5 | 510 | |
8615cb43 | 511 | |
cd439c50 | 512 | /* Allow for side effects in s */ |
ccb2c380 MP |
513 | #define REGC(c,s) STMT_START { \ |
514 | if (!SIZE_ONLY) *(s) = (c); else (void)(s); \ | |
515 | } STMT_END | |
cd439c50 | 516 | |
fac92740 MJD |
517 | /* Macros for recording node offsets. 20001227 mjd@plover.com |
518 | * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in | |
519 | * element 2*n-1 of the array. Element #2n holds the byte length node #n. | |
520 | * Element 0 holds the number n. | |
07be1b83 | 521 | * Position is 1 indexed. |
fac92740 | 522 | */ |
7122b237 YO |
523 | #ifndef RE_TRACK_PATTERN_OFFSETS |
524 | #define Set_Node_Offset_To_R(node,byte) | |
525 | #define Set_Node_Offset(node,byte) | |
526 | #define Set_Cur_Node_Offset | |
527 | #define Set_Node_Length_To_R(node,len) | |
528 | #define Set_Node_Length(node,len) | |
529 | #define Set_Node_Cur_Length(node) | |
530 | #define Node_Offset(n) | |
531 | #define Node_Length(n) | |
532 | #define Set_Node_Offset_Length(node,offset,len) | |
533 | #define ProgLen(ri) ri->u.proglen | |
534 | #define SetProgLen(ri,x) ri->u.proglen = x | |
535 | #else | |
536 | #define ProgLen(ri) ri->u.offsets[0] | |
537 | #define SetProgLen(ri,x) ri->u.offsets[0] = x | |
ccb2c380 MP |
538 | #define Set_Node_Offset_To_R(node,byte) STMT_START { \ |
539 | if (! SIZE_ONLY) { \ | |
540 | MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \ | |
2a49f0f5 | 541 | __LINE__, (int)(node), (int)(byte))); \ |
ccb2c380 | 542 | if((node) < 0) { \ |
551405c4 | 543 | Perl_croak(aTHX_ "value of node is %d in Offset macro", (int)(node)); \ |
ccb2c380 MP |
544 | } else { \ |
545 | RExC_offsets[2*(node)-1] = (byte); \ | |
546 | } \ | |
547 | } \ | |
548 | } STMT_END | |
549 | ||
550 | #define Set_Node_Offset(node,byte) \ | |
551 | Set_Node_Offset_To_R((node)-RExC_emit_start, (byte)-RExC_start) | |
552 | #define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse) | |
553 | ||
554 | #define Set_Node_Length_To_R(node,len) STMT_START { \ | |
555 | if (! SIZE_ONLY) { \ | |
556 | MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \ | |
551405c4 | 557 | __LINE__, (int)(node), (int)(len))); \ |
ccb2c380 | 558 | if((node) < 0) { \ |
551405c4 | 559 | Perl_croak(aTHX_ "value of node is %d in Length macro", (int)(node)); \ |
ccb2c380 MP |
560 | } else { \ |
561 | RExC_offsets[2*(node)] = (len); \ | |
562 | } \ | |
563 | } \ | |
564 | } STMT_END | |
565 | ||
566 | #define Set_Node_Length(node,len) \ | |
567 | Set_Node_Length_To_R((node)-RExC_emit_start, len) | |
568 | #define Set_Cur_Node_Length(len) Set_Node_Length(RExC_emit, len) | |
569 | #define Set_Node_Cur_Length(node) \ | |
570 | Set_Node_Length(node, RExC_parse - parse_start) | |
fac92740 MJD |
571 | |
572 | /* Get offsets and lengths */ | |
573 | #define Node_Offset(n) (RExC_offsets[2*((n)-RExC_emit_start)-1]) | |
574 | #define Node_Length(n) (RExC_offsets[2*((n)-RExC_emit_start)]) | |
575 | ||
07be1b83 YO |
576 | #define Set_Node_Offset_Length(node,offset,len) STMT_START { \ |
577 | Set_Node_Offset_To_R((node)-RExC_emit_start, (offset)); \ | |
578 | Set_Node_Length_To_R((node)-RExC_emit_start, (len)); \ | |
579 | } STMT_END | |
7122b237 | 580 | #endif |
07be1b83 YO |
581 | |
582 | #if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS | |
583 | #define EXPERIMENTAL_INPLACESCAN | |
7122b237 | 584 | #endif /*RE_TRACK_PATTERN_OFFSETS*/ |
07be1b83 | 585 | |
304ee84b YO |
586 | #define DEBUG_STUDYDATA(str,data,depth) \ |
587 | DEBUG_OPTIMISE_MORE_r(if(data){ \ | |
1de06328 | 588 | PerlIO_printf(Perl_debug_log, \ |
304ee84b YO |
589 | "%*s" str "Pos:%"IVdf"/%"IVdf \ |
590 | " Flags: 0x%"UVXf" Whilem_c: %"IVdf" Lcp: %"IVdf" %s", \ | |
1de06328 YO |
591 | (int)(depth)*2, "", \ |
592 | (IV)((data)->pos_min), \ | |
593 | (IV)((data)->pos_delta), \ | |
304ee84b | 594 | (UV)((data)->flags), \ |
1de06328 | 595 | (IV)((data)->whilem_c), \ |
304ee84b YO |
596 | (IV)((data)->last_closep ? *((data)->last_closep) : -1), \ |
597 | is_inf ? "INF " : "" \ | |
1de06328 YO |
598 | ); \ |
599 | if ((data)->last_found) \ | |
600 | PerlIO_printf(Perl_debug_log, \ | |
601 | "Last:'%s' %"IVdf":%"IVdf"/%"IVdf" %sFixed:'%s' @ %"IVdf \ | |
602 | " %sFloat: '%s' @ %"IVdf"/%"IVdf"", \ | |
603 | SvPVX_const((data)->last_found), \ | |
604 | (IV)((data)->last_end), \ | |
605 | (IV)((data)->last_start_min), \ | |
606 | (IV)((data)->last_start_max), \ | |
607 | ((data)->longest && \ | |
608 | (data)->longest==&((data)->longest_fixed)) ? "*" : "", \ | |
609 | SvPVX_const((data)->longest_fixed), \ | |
610 | (IV)((data)->offset_fixed), \ | |
611 | ((data)->longest && \ | |
612 | (data)->longest==&((data)->longest_float)) ? "*" : "", \ | |
613 | SvPVX_const((data)->longest_float), \ | |
614 | (IV)((data)->offset_float_min), \ | |
615 | (IV)((data)->offset_float_max) \ | |
616 | ); \ | |
617 | PerlIO_printf(Perl_debug_log,"\n"); \ | |
618 | }); | |
619 | ||
acfe0abc | 620 | static void clear_re(pTHX_ void *r); |
4327152a | 621 | |
653099ff | 622 | /* Mark that we cannot extend a found fixed substring at this point. |
786e8c11 | 623 | Update the longest found anchored substring and the longest found |
653099ff GS |
624 | floating substrings if needed. */ |
625 | ||
4327152a | 626 | STATIC void |
304ee84b | 627 | S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data, I32 *minlenp, int is_inf) |
c277df42 | 628 | { |
e1ec3a88 AL |
629 | const STRLEN l = CHR_SVLEN(data->last_found); |
630 | const STRLEN old_l = CHR_SVLEN(*data->longest); | |
1de06328 | 631 | GET_RE_DEBUG_FLAGS_DECL; |
b81d288d | 632 | |
c277df42 | 633 | if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) { |
6b43b216 | 634 | SvSetMagicSV(*data->longest, data->last_found); |
c277df42 IZ |
635 | if (*data->longest == data->longest_fixed) { |
636 | data->offset_fixed = l ? data->last_start_min : data->pos_min; | |
637 | if (data->flags & SF_BEFORE_EOL) | |
b81d288d | 638 | data->flags |
c277df42 IZ |
639 | |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL); |
640 | else | |
641 | data->flags &= ~SF_FIX_BEFORE_EOL; | |
1de06328 YO |
642 | data->minlen_fixed=minlenp; |
643 | data->lookbehind_fixed=0; | |
a0ed51b3 | 644 | } |
304ee84b | 645 | else { /* *data->longest == data->longest_float */ |
c277df42 | 646 | data->offset_float_min = l ? data->last_start_min : data->pos_min; |
b81d288d AB |
647 | data->offset_float_max = (l |
648 | ? data->last_start_max | |
c277df42 | 649 | : data->pos_min + data->pos_delta); |
304ee84b | 650 | if (is_inf || (U32)data->offset_float_max > (U32)I32_MAX) |
9051bda5 | 651 | data->offset_float_max = I32_MAX; |
c277df42 | 652 | if (data->flags & SF_BEFORE_EOL) |
b81d288d | 653 | data->flags |
c277df42 IZ |
654 | |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL); |
655 | else | |
656 | data->flags &= ~SF_FL_BEFORE_EOL; | |
1de06328 YO |
657 | data->minlen_float=minlenp; |
658 | data->lookbehind_float=0; | |
c277df42 IZ |
659 | } |
660 | } | |
661 | SvCUR_set(data->last_found, 0); | |
0eda9292 | 662 | { |
a28509cc | 663 | SV * const sv = data->last_found; |
097eb12c AL |
664 | if (SvUTF8(sv) && SvMAGICAL(sv)) { |
665 | MAGIC * const mg = mg_find(sv, PERL_MAGIC_utf8); | |
666 | if (mg) | |
667 | mg->mg_len = 0; | |
668 | } | |
0eda9292 | 669 | } |
c277df42 IZ |
670 | data->last_end = -1; |
671 | data->flags &= ~SF_BEFORE_EOL; | |
bcdf7404 | 672 | DEBUG_STUDYDATA("commit: ",data,0); |
c277df42 IZ |
673 | } |
674 | ||
653099ff GS |
675 | /* Can match anything (initialization) */ |
676 | STATIC void | |
097eb12c | 677 | S_cl_anything(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl) |
653099ff | 678 | { |
653099ff | 679 | ANYOF_CLASS_ZERO(cl); |
f8bef550 | 680 | ANYOF_BITMAP_SETALL(cl); |
1aa99e6b | 681 | cl->flags = ANYOF_EOS|ANYOF_UNICODE_ALL; |
653099ff GS |
682 | if (LOC) |
683 | cl->flags |= ANYOF_LOCALE; | |
684 | } | |
685 | ||
686 | /* Can match anything (initialization) */ | |
687 | STATIC int | |
5f66b61c | 688 | S_cl_is_anything(const struct regnode_charclass_class *cl) |
653099ff GS |
689 | { |
690 | int value; | |
691 | ||
aaa51d5e | 692 | for (value = 0; value <= ANYOF_MAX; value += 2) |
653099ff GS |
693 | if (ANYOF_CLASS_TEST(cl, value) && ANYOF_CLASS_TEST(cl, value + 1)) |
694 | return 1; | |
1aa99e6b IH |
695 | if (!(cl->flags & ANYOF_UNICODE_ALL)) |
696 | return 0; | |
10edeb5d | 697 | if (!ANYOF_BITMAP_TESTALLSET((const void*)cl)) |
f8bef550 | 698 | return 0; |
653099ff GS |
699 | return 1; |
700 | } | |
701 | ||
702 | /* Can match anything (initialization) */ | |
703 | STATIC void | |
097eb12c | 704 | S_cl_init(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl) |
653099ff | 705 | { |
8ecf7187 | 706 | Zero(cl, 1, struct regnode_charclass_class); |
653099ff | 707 | cl->type = ANYOF; |
830247a4 | 708 | cl_anything(pRExC_state, cl); |
653099ff GS |
709 | } |
710 | ||
711 | STATIC void | |
097eb12c | 712 | S_cl_init_zero(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl) |
653099ff | 713 | { |
8ecf7187 | 714 | Zero(cl, 1, struct regnode_charclass_class); |
653099ff | 715 | cl->type = ANYOF; |
830247a4 | 716 | cl_anything(pRExC_state, cl); |
653099ff GS |
717 | if (LOC) |
718 | cl->flags |= ANYOF_LOCALE; | |
719 | } | |
720 | ||
721 | /* 'And' a given class with another one. Can create false positives */ | |
722 | /* We assume that cl is not inverted */ | |
723 | STATIC void | |
5f66b61c | 724 | S_cl_and(struct regnode_charclass_class *cl, |
a28509cc | 725 | const struct regnode_charclass_class *and_with) |
653099ff | 726 | { |
40d049e4 YO |
727 | |
728 | assert(and_with->type == ANYOF); | |
653099ff GS |
729 | if (!(and_with->flags & ANYOF_CLASS) |
730 | && !(cl->flags & ANYOF_CLASS) | |
731 | && (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE) | |
732 | && !(and_with->flags & ANYOF_FOLD) | |
733 | && !(cl->flags & ANYOF_FOLD)) { | |
734 | int i; | |
735 | ||
736 | if (and_with->flags & ANYOF_INVERT) | |
737 | for (i = 0; i < ANYOF_BITMAP_SIZE; i++) | |
738 | cl->bitmap[i] &= ~and_with->bitmap[i]; | |
739 | else | |
740 | for (i = 0; i < ANYOF_BITMAP_SIZE; i++) | |
741 | cl->bitmap[i] &= and_with->bitmap[i]; | |
742 | } /* XXXX: logic is complicated otherwise, leave it along for a moment. */ | |
743 | if (!(and_with->flags & ANYOF_EOS)) | |
744 | cl->flags &= ~ANYOF_EOS; | |
1aa99e6b | 745 | |
14ebb1a2 JH |
746 | if (cl->flags & ANYOF_UNICODE_ALL && and_with->flags & ANYOF_UNICODE && |
747 | !(and_with->flags & ANYOF_INVERT)) { | |
1aa99e6b IH |
748 | cl->flags &= ~ANYOF_UNICODE_ALL; |
749 | cl->flags |= ANYOF_UNICODE; | |
750 | ARG_SET(cl, ARG(and_with)); | |
751 | } | |
14ebb1a2 JH |
752 | if (!(and_with->flags & ANYOF_UNICODE_ALL) && |
753 | !(and_with->flags & ANYOF_INVERT)) | |
1aa99e6b | 754 | cl->flags &= ~ANYOF_UNICODE_ALL; |
14ebb1a2 JH |
755 | if (!(and_with->flags & (ANYOF_UNICODE|ANYOF_UNICODE_ALL)) && |
756 | !(and_with->flags & ANYOF_INVERT)) | |
1aa99e6b | 757 | cl->flags &= ~ANYOF_UNICODE; |
653099ff GS |
758 | } |
759 | ||
760 | /* 'OR' a given class with another one. Can create false positives */ | |
761 | /* We assume that cl is not inverted */ | |
762 | STATIC void | |
097eb12c | 763 | S_cl_or(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl, const struct regnode_charclass_class *or_with) |
653099ff | 764 | { |
653099ff GS |
765 | if (or_with->flags & ANYOF_INVERT) { |
766 | /* We do not use | |
767 | * (B1 | CL1) | (!B2 & !CL2) = (B1 | !B2 & !CL2) | (CL1 | (!B2 & !CL2)) | |
768 | * <= (B1 | !B2) | (CL1 | !CL2) | |
769 | * which is wasteful if CL2 is small, but we ignore CL2: | |
770 | * (B1 | CL1) | (!B2 & !CL2) <= (B1 | CL1) | !B2 = (B1 | !B2) | CL1 | |
771 | * XXXX Can we handle case-fold? Unclear: | |
772 | * (OK1(i) | OK1(i')) | !(OK1(i) | OK1(i')) = | |
773 | * (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i')) | |
774 | */ | |
775 | if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE) | |
776 | && !(or_with->flags & ANYOF_FOLD) | |
777 | && !(cl->flags & ANYOF_FOLD) ) { | |
778 | int i; | |
779 | ||
780 | for (i = 0; i < ANYOF_BITMAP_SIZE; i++) | |
781 | cl->bitmap[i] |= ~or_with->bitmap[i]; | |
782 | } /* XXXX: logic is complicated otherwise */ | |
783 | else { | |
830247a4 | 784 | cl_anything(pRExC_state, cl); |
653099ff GS |
785 | } |
786 | } else { | |
787 | /* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */ | |
788 | if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE) | |
b81d288d | 789 | && (!(or_with->flags & ANYOF_FOLD) |
653099ff GS |
790 | || (cl->flags & ANYOF_FOLD)) ) { |
791 | int i; | |
792 | ||
793 | /* OR char bitmap and class bitmap separately */ | |
794 | for (i = 0; i < ANYOF_BITMAP_SIZE; i++) | |
795 | cl->bitmap[i] |= or_with->bitmap[i]; | |
796 | if (or_with->flags & ANYOF_CLASS) { | |
797 | for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++) | |
798 | cl->classflags[i] |= or_with->classflags[i]; | |
799 | cl->flags |= ANYOF_CLASS; | |
800 | } | |
801 | } | |
802 | else { /* XXXX: logic is complicated, leave it along for a moment. */ | |
830247a4 | 803 | cl_anything(pRExC_state, cl); |
653099ff GS |
804 | } |
805 | } | |
806 | if (or_with->flags & ANYOF_EOS) | |
807 | cl->flags |= ANYOF_EOS; | |
1aa99e6b IH |
808 | |
809 | if (cl->flags & ANYOF_UNICODE && or_with->flags & ANYOF_UNICODE && | |
810 | ARG(cl) != ARG(or_with)) { | |
811 | cl->flags |= ANYOF_UNICODE_ALL; | |
812 | cl->flags &= ~ANYOF_UNICODE; | |
813 | } | |
814 | if (or_with->flags & ANYOF_UNICODE_ALL) { | |
815 | cl->flags |= ANYOF_UNICODE_ALL; | |
816 | cl->flags &= ~ANYOF_UNICODE; | |
817 | } | |
653099ff GS |
818 | } |
819 | ||
a3621e74 YO |
820 | #define TRIE_LIST_ITEM(state,idx) (trie->states[state].trans.list)[ idx ] |
821 | #define TRIE_LIST_CUR(state) ( TRIE_LIST_ITEM( state, 0 ).forid ) | |
822 | #define TRIE_LIST_LEN(state) ( TRIE_LIST_ITEM( state, 0 ).newstate ) | |
823 | #define TRIE_LIST_USED(idx) ( trie->states[state].trans.list ? (TRIE_LIST_CUR( idx ) - 1) : 0 ) | |
824 | ||
3dab1dad YO |
825 | |
826 | #ifdef DEBUGGING | |
07be1b83 | 827 | /* |
2b8b4781 NC |
828 | dump_trie(trie,widecharmap,revcharmap) |
829 | dump_trie_interim_list(trie,widecharmap,revcharmap,next_alloc) | |
830 | dump_trie_interim_table(trie,widecharmap,revcharmap,next_alloc) | |
3dab1dad YO |
831 | |
832 | These routines dump out a trie in a somewhat readable format. | |
07be1b83 YO |
833 | The _interim_ variants are used for debugging the interim |
834 | tables that are used to generate the final compressed | |
835 | representation which is what dump_trie expects. | |
836 | ||
3dab1dad YO |
837 | Part of the reason for their existance is to provide a form |
838 | of documentation as to how the different representations function. | |
07be1b83 YO |
839 | |
840 | */ | |
3dab1dad YO |
841 | |
842 | /* | |
3dab1dad YO |
843 | Dumps the final compressed table form of the trie to Perl_debug_log. |
844 | Used for debugging make_trie(). | |
845 | */ | |
846 | ||
847 | STATIC void | |
2b8b4781 NC |
848 | S_dump_trie(pTHX_ const struct _reg_trie_data *trie, HV *widecharmap, |
849 | AV *revcharmap, U32 depth) | |
3dab1dad YO |
850 | { |
851 | U32 state; | |
ab3bbdeb | 852 | SV *sv=sv_newmortal(); |
55eed653 | 853 | int colwidth= widecharmap ? 6 : 4; |
3dab1dad YO |
854 | GET_RE_DEBUG_FLAGS_DECL; |
855 | ||
ab3bbdeb | 856 | |
3dab1dad YO |
857 | PerlIO_printf( Perl_debug_log, "%*sChar : %-6s%-6s%-4s ", |
858 | (int)depth * 2 + 2,"", | |
859 | "Match","Base","Ofs" ); | |
860 | ||
861 | for( state = 0 ; state < trie->uniquecharcount ; state++ ) { | |
2b8b4781 | 862 | SV ** const tmp = av_fetch( revcharmap, state, 0); |
3dab1dad | 863 | if ( tmp ) { |
ab3bbdeb YO |
864 | PerlIO_printf( Perl_debug_log, "%*s", |
865 | colwidth, | |
ddc5bc0f | 866 | pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth, |
ab3bbdeb YO |
867 | PL_colors[0], PL_colors[1], |
868 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) | | |
869 | PERL_PV_ESCAPE_FIRSTCHAR | |
870 | ) | |
871 | ); | |
3dab1dad YO |
872 | } |
873 | } | |
874 | PerlIO_printf( Perl_debug_log, "\n%*sState|-----------------------", | |
875 | (int)depth * 2 + 2,""); | |
876 | ||
877 | for( state = 0 ; state < trie->uniquecharcount ; state++ ) | |
ab3bbdeb | 878 | PerlIO_printf( Perl_debug_log, "%.*s", colwidth, "--------"); |
3dab1dad YO |
879 | PerlIO_printf( Perl_debug_log, "\n"); |
880 | ||
1e2e3d02 | 881 | for( state = 1 ; state < trie->statecount ; state++ ) { |
be8e71aa | 882 | const U32 base = trie->states[ state ].trans.base; |
3dab1dad YO |
883 | |
884 | PerlIO_printf( Perl_debug_log, "%*s#%4"UVXf"|", (int)depth * 2 + 2,"", (UV)state); | |
885 | ||
886 | if ( trie->states[ state ].wordnum ) { | |
887 | PerlIO_printf( Perl_debug_log, " W%4X", trie->states[ state ].wordnum ); | |
888 | } else { | |
889 | PerlIO_printf( Perl_debug_log, "%6s", "" ); | |
890 | } | |
891 | ||
892 | PerlIO_printf( Perl_debug_log, " @%4"UVXf" ", (UV)base ); | |
893 | ||
894 | if ( base ) { | |
895 | U32 ofs = 0; | |
896 | ||
897 | while( ( base + ofs < trie->uniquecharcount ) || | |
898 | ( base + ofs - trie->uniquecharcount < trie->lasttrans | |
899 | && trie->trans[ base + ofs - trie->uniquecharcount ].check != state)) | |
900 | ofs++; | |
901 | ||
902 | PerlIO_printf( Perl_debug_log, "+%2"UVXf"[ ", (UV)ofs); | |
903 | ||
904 | for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) { | |
905 | if ( ( base + ofs >= trie->uniquecharcount ) && | |
906 | ( base + ofs - trie->uniquecharcount < trie->lasttrans ) && | |
907 | trie->trans[ base + ofs - trie->uniquecharcount ].check == state ) | |
908 | { | |
ab3bbdeb YO |
909 | PerlIO_printf( Perl_debug_log, "%*"UVXf, |
910 | colwidth, | |
3dab1dad YO |
911 | (UV)trie->trans[ base + ofs - trie->uniquecharcount ].next ); |
912 | } else { | |
ab3bbdeb | 913 | PerlIO_printf( Perl_debug_log, "%*s",colwidth," ." ); |
3dab1dad YO |
914 | } |
915 | } | |
916 | ||
917 | PerlIO_printf( Perl_debug_log, "]"); | |
918 | ||
919 | } | |
920 | PerlIO_printf( Perl_debug_log, "\n" ); | |
921 | } | |
922 | } | |
923 | /* | |
3dab1dad YO |
924 | Dumps a fully constructed but uncompressed trie in list form. |
925 | List tries normally only are used for construction when the number of | |
926 | possible chars (trie->uniquecharcount) is very high. | |
927 | Used for debugging make_trie(). | |
928 | */ | |
929 | STATIC void | |
55eed653 | 930 | S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie, |
2b8b4781 NC |
931 | HV *widecharmap, AV *revcharmap, U32 next_alloc, |
932 | U32 depth) | |
3dab1dad YO |
933 | { |
934 | U32 state; | |
ab3bbdeb | 935 | SV *sv=sv_newmortal(); |
55eed653 | 936 | int colwidth= widecharmap ? 6 : 4; |
3dab1dad YO |
937 | GET_RE_DEBUG_FLAGS_DECL; |
938 | /* print out the table precompression. */ | |
ab3bbdeb YO |
939 | PerlIO_printf( Perl_debug_log, "%*sState :Word | Transition Data\n%*s%s", |
940 | (int)depth * 2 + 2,"", (int)depth * 2 + 2,"", | |
941 | "------:-----+-----------------\n" ); | |
3dab1dad YO |
942 | |
943 | for( state=1 ; state < next_alloc ; state ++ ) { | |
944 | U16 charid; | |
945 | ||
ab3bbdeb | 946 | PerlIO_printf( Perl_debug_log, "%*s %4"UVXf" :", |
3dab1dad YO |
947 | (int)depth * 2 + 2,"", (UV)state ); |
948 | if ( ! trie->states[ state ].wordnum ) { | |
949 | PerlIO_printf( Perl_debug_log, "%5s| ",""); | |
950 | } else { | |
951 | PerlIO_printf( Perl_debug_log, "W%4x| ", | |
952 | trie->states[ state ].wordnum | |
953 | ); | |
954 | } | |
955 | for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) { | |
2b8b4781 | 956 | SV ** const tmp = av_fetch( revcharmap, TRIE_LIST_ITEM(state,charid).forid, 0); |
ab3bbdeb YO |
957 | if ( tmp ) { |
958 | PerlIO_printf( Perl_debug_log, "%*s:%3X=%4"UVXf" | ", | |
959 | colwidth, | |
ddc5bc0f | 960 | pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth, |
ab3bbdeb YO |
961 | PL_colors[0], PL_colors[1], |
962 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) | | |
963 | PERL_PV_ESCAPE_FIRSTCHAR | |
964 | ) , | |
1e2e3d02 YO |
965 | TRIE_LIST_ITEM(state,charid).forid, |
966 | (UV)TRIE_LIST_ITEM(state,charid).newstate | |
967 | ); | |
968 | if (!(charid % 10)) | |
664e119d RGS |
969 | PerlIO_printf(Perl_debug_log, "\n%*s| ", |
970 | (int)((depth * 2) + 14), ""); | |
1e2e3d02 | 971 | } |
ab3bbdeb YO |
972 | } |
973 | PerlIO_printf( Perl_debug_log, "\n"); | |
3dab1dad YO |
974 | } |
975 | } | |
976 | ||
977 | /* | |
3dab1dad YO |
978 | Dumps a fully constructed but uncompressed trie in table form. |
979 | This is the normal DFA style state transition table, with a few | |
980 | twists to facilitate compression later. | |
981 | Used for debugging make_trie(). | |
982 | */ | |
983 | STATIC void | |
55eed653 | 984 | S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie, |
2b8b4781 NC |
985 | HV *widecharmap, AV *revcharmap, U32 next_alloc, |
986 | U32 depth) | |
3dab1dad YO |
987 | { |
988 | U32 state; | |
989 | U16 charid; | |
ab3bbdeb | 990 | SV *sv=sv_newmortal(); |
55eed653 | 991 | int colwidth= widecharmap ? 6 : 4; |
3dab1dad YO |
992 | GET_RE_DEBUG_FLAGS_DECL; |
993 | ||
994 | /* | |
995 | print out the table precompression so that we can do a visual check | |
996 | that they are identical. | |
997 | */ | |
998 | ||
999 | PerlIO_printf( Perl_debug_log, "%*sChar : ",(int)depth * 2 + 2,"" ); | |
1000 | ||
1001 | for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) { | |
2b8b4781 | 1002 | SV ** const tmp = av_fetch( revcharmap, charid, 0); |
3dab1dad | 1003 | if ( tmp ) { |
ab3bbdeb YO |
1004 | PerlIO_printf( Perl_debug_log, "%*s", |
1005 | colwidth, | |
ddc5bc0f | 1006 | pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth, |
ab3bbdeb YO |
1007 | PL_colors[0], PL_colors[1], |
1008 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) | | |
1009 | PERL_PV_ESCAPE_FIRSTCHAR | |
1010 | ) | |
1011 | ); | |
3dab1dad YO |
1012 | } |
1013 | } | |
1014 | ||
1015 | PerlIO_printf( Perl_debug_log, "\n%*sState+-",(int)depth * 2 + 2,"" ); | |
1016 | ||
1017 | for( charid=0 ; charid < trie->uniquecharcount ; charid++ ) { | |
ab3bbdeb | 1018 | PerlIO_printf( Perl_debug_log, "%.*s", colwidth,"--------"); |
3dab1dad YO |
1019 | } |
1020 | ||
1021 | PerlIO_printf( Perl_debug_log, "\n" ); | |
1022 | ||
1023 | for( state=1 ; state < next_alloc ; state += trie->uniquecharcount ) { | |
1024 | ||
1025 | PerlIO_printf( Perl_debug_log, "%*s%4"UVXf" : ", | |
1026 | (int)depth * 2 + 2,"", | |
1027 | (UV)TRIE_NODENUM( state ) ); | |
1028 | ||
1029 | for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) { | |
ab3bbdeb YO |
1030 | UV v=(UV)SAFE_TRIE_NODENUM( trie->trans[ state + charid ].next ); |
1031 | if (v) | |
1032 | PerlIO_printf( Perl_debug_log, "%*"UVXf, colwidth, v ); | |
1033 | else | |
1034 | PerlIO_printf( Perl_debug_log, "%*s", colwidth, "." ); | |
3dab1dad YO |
1035 | } |
1036 | if ( ! trie->states[ TRIE_NODENUM( state ) ].wordnum ) { | |
1037 | PerlIO_printf( Perl_debug_log, " (%4"UVXf")\n", (UV)trie->trans[ state ].check ); | |
1038 | } else { | |
1039 | PerlIO_printf( Perl_debug_log, " (%4"UVXf") W%4X\n", (UV)trie->trans[ state ].check, | |
1040 | trie->states[ TRIE_NODENUM( state ) ].wordnum ); | |
1041 | } | |
1042 | } | |
07be1b83 | 1043 | } |
3dab1dad YO |
1044 | |
1045 | #endif | |
1046 | ||
786e8c11 YO |
1047 | /* make_trie(startbranch,first,last,tail,word_count,flags,depth) |
1048 | startbranch: the first branch in the whole branch sequence | |
1049 | first : start branch of sequence of branch-exact nodes. | |
1050 | May be the same as startbranch | |
1051 | last : Thing following the last branch. | |
1052 | May be the same as tail. | |
1053 | tail : item following the branch sequence | |
1054 | count : words in the sequence | |
1055 | flags : currently the OP() type we will be building one of /EXACT(|F|Fl)/ | |
1056 | depth : indent depth | |
3dab1dad | 1057 | |
786e8c11 | 1058 | Inplace optimizes a sequence of 2 or more Branch-Exact nodes into a TRIE node. |
07be1b83 | 1059 | |
786e8c11 YO |
1060 | A trie is an N'ary tree where the branches are determined by digital |
1061 | decomposition of the key. IE, at the root node you look up the 1st character and | |
1062 | follow that branch repeat until you find the end of the branches. Nodes can be | |
1063 | marked as "accepting" meaning they represent a complete word. Eg: | |
07be1b83 | 1064 | |
786e8c11 | 1065 | /he|she|his|hers/ |
72f13be8 | 1066 | |
786e8c11 YO |
1067 | would convert into the following structure. Numbers represent states, letters |
1068 | following numbers represent valid transitions on the letter from that state, if | |
1069 | the number is in square brackets it represents an accepting state, otherwise it | |
1070 | will be in parenthesis. | |
07be1b83 | 1071 | |
786e8c11 YO |
1072 | +-h->+-e->[3]-+-r->(8)-+-s->[9] |
1073 | | | | |
1074 | | (2) | |
1075 | | | | |
1076 | (1) +-i->(6)-+-s->[7] | |
1077 | | | |
1078 | +-s->(3)-+-h->(4)-+-e->[5] | |
07be1b83 | 1079 | |
786e8c11 YO |
1080 | Accept Word Mapping: 3=>1 (he),5=>2 (she), 7=>3 (his), 9=>4 (hers) |
1081 | ||
1082 | This shows that when matching against the string 'hers' we will begin at state 1 | |
1083 | read 'h' and move to state 2, read 'e' and move to state 3 which is accepting, | |
1084 | then read 'r' and go to state 8 followed by 's' which takes us to state 9 which | |
1085 | is also accepting. Thus we know that we can match both 'he' and 'hers' with a | |
1086 | single traverse. We store a mapping from accepting to state to which word was | |
1087 | matched, and then when we have multiple possibilities we try to complete the | |
1088 | rest of the regex in the order in which they occured in the alternation. | |
1089 | ||
1090 | The only prior NFA like behaviour that would be changed by the TRIE support is | |
1091 | the silent ignoring of duplicate alternations which are of the form: | |
1092 | ||
1093 | / (DUPE|DUPE) X? (?{ ... }) Y /x | |
1094 | ||
1095 | Thus EVAL blocks follwing a trie may be called a different number of times with | |
1096 | and without the optimisation. With the optimisations dupes will be silently | |
1097 | ignored. This inconsistant behaviour of EVAL type nodes is well established as | |
1098 | the following demonstrates: | |
1099 | ||
1100 | 'words'=~/(word|word|word)(?{ print $1 })[xyz]/ | |
1101 | ||
1102 | which prints out 'word' three times, but | |
1103 | ||
1104 | 'words'=~/(word|word|word)(?{ print $1 })S/ | |
1105 | ||
1106 | which doesnt print it out at all. This is due to other optimisations kicking in. | |
1107 | ||
1108 | Example of what happens on a structural level: | |
1109 | ||
1110 | The regexp /(ac|ad|ab)+/ will produce the folowing debug output: | |
1111 | ||
1112 | 1: CURLYM[1] {1,32767}(18) | |
1113 | 5: BRANCH(8) | |
1114 | 6: EXACT <ac>(16) | |
1115 | 8: BRANCH(11) | |
1116 | 9: EXACT <ad>(16) | |
1117 | 11: BRANCH(14) | |
1118 | 12: EXACT <ab>(16) | |
1119 | 16: SUCCEED(0) | |
1120 | 17: NOTHING(18) | |
1121 | 18: END(0) | |
1122 | ||
1123 | This would be optimizable with startbranch=5, first=5, last=16, tail=16 | |
1124 | and should turn into: | |
1125 | ||
1126 | 1: CURLYM[1] {1,32767}(18) | |
1127 | 5: TRIE(16) | |
1128 | [Words:3 Chars Stored:6 Unique Chars:4 States:5 NCP:1] | |
1129 | <ac> | |
1130 | <ad> | |
1131 | <ab> | |
1132 | 16: SUCCEED(0) | |
1133 | 17: NOTHING(18) | |
1134 | 18: END(0) | |
1135 | ||
1136 | Cases where tail != last would be like /(?foo|bar)baz/: | |
1137 | ||
1138 | 1: BRANCH(4) | |
1139 | 2: EXACT <foo>(8) | |
1140 | 4: BRANCH(7) | |
1141 | 5: EXACT <bar>(8) | |
1142 | 7: TAIL(8) | |
1143 | 8: EXACT <baz>(10) | |
1144 | 10: END(0) | |
1145 | ||
1146 | which would be optimizable with startbranch=1, first=1, last=7, tail=8 | |
1147 | and would end up looking like: | |
1148 | ||
1149 | 1: TRIE(8) | |
1150 | [Words:2 Chars Stored:6 Unique Chars:5 States:7 NCP:1] | |
1151 | <foo> | |
1152 | <bar> | |
1153 | 7: TAIL(8) | |
1154 | 8: EXACT <baz>(10) | |
1155 | 10: END(0) | |
1156 | ||
1157 | d = uvuni_to_utf8_flags(d, uv, 0); | |
1158 | ||
1159 | is the recommended Unicode-aware way of saying | |
1160 | ||
1161 | *(d++) = uv; | |
1162 | */ | |
1163 | ||
1e2e3d02 | 1164 | #define TRIE_STORE_REVCHAR \ |
786e8c11 | 1165 | STMT_START { \ |
1e2e3d02 | 1166 | SV *tmp = newSVpvs(""); \ |
786e8c11 | 1167 | if (UTF) SvUTF8_on(tmp); \ |
1e2e3d02 | 1168 | Perl_sv_catpvf( aTHX_ tmp, "%c", (int)uvc ); \ |
2b8b4781 | 1169 | av_push( revcharmap, tmp ); \ |
786e8c11 YO |
1170 | } STMT_END |
1171 | ||
1172 | #define TRIE_READ_CHAR STMT_START { \ | |
1173 | wordlen++; \ | |
1174 | if ( UTF ) { \ | |
1175 | if ( folder ) { \ | |
1176 | if ( foldlen > 0 ) { \ | |
1177 | uvc = utf8n_to_uvuni( scan, UTF8_MAXLEN, &len, uniflags ); \ | |
1178 | foldlen -= len; \ | |
1179 | scan += len; \ | |
1180 | len = 0; \ | |
1181 | } else { \ | |
1182 | uvc = utf8n_to_uvuni( (const U8*)uc, UTF8_MAXLEN, &len, uniflags);\ | |
1183 | uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \ | |
1184 | foldlen -= UNISKIP( uvc ); \ | |
1185 | scan = foldbuf + UNISKIP( uvc ); \ | |
1186 | } \ | |
1187 | } else { \ | |
1188 | uvc = utf8n_to_uvuni( (const U8*)uc, UTF8_MAXLEN, &len, uniflags);\ | |
1189 | } \ | |
1190 | } else { \ | |
1191 | uvc = (U32)*uc; \ | |
1192 | len = 1; \ | |
1193 | } \ | |
1194 | } STMT_END | |
1195 | ||
1196 | ||
1197 | ||
1198 | #define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \ | |
1199 | if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \ | |
f9003953 NC |
1200 | U32 ging = TRIE_LIST_LEN( state ) *= 2; \ |
1201 | Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \ | |
786e8c11 YO |
1202 | } \ |
1203 | TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \ | |
1204 | TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \ | |
1205 | TRIE_LIST_CUR( state )++; \ | |
1206 | } STMT_END | |
07be1b83 | 1207 | |
786e8c11 YO |
1208 | #define TRIE_LIST_NEW(state) STMT_START { \ |
1209 | Newxz( trie->states[ state ].trans.list, \ | |
1210 | 4, reg_trie_trans_le ); \ | |
1211 | TRIE_LIST_CUR( state ) = 1; \ | |
1212 | TRIE_LIST_LEN( state ) = 4; \ | |
1213 | } STMT_END | |
07be1b83 | 1214 | |
786e8c11 YO |
1215 | #define TRIE_HANDLE_WORD(state) STMT_START { \ |
1216 | U16 dupe= trie->states[ state ].wordnum; \ | |
1217 | regnode * const noper_next = regnext( noper ); \ | |
1218 | \ | |
1219 | if (trie->wordlen) \ | |
1220 | trie->wordlen[ curword ] = wordlen; \ | |
1221 | DEBUG_r({ \ | |
1222 | /* store the word for dumping */ \ | |
1223 | SV* tmp; \ | |
1224 | if (OP(noper) != NOTHING) \ | |
1225 | tmp = newSVpvn(STRING(noper), STR_LEN(noper)); \ | |
1226 | else \ | |
1227 | tmp = newSVpvn( "", 0 ); \ | |
1228 | if ( UTF ) SvUTF8_on( tmp ); \ | |
2b8b4781 | 1229 | av_push( trie_words, tmp ); \ |
786e8c11 YO |
1230 | }); \ |
1231 | \ | |
1232 | curword++; \ | |
1233 | \ | |
1234 | if ( noper_next < tail ) { \ | |
1235 | if (!trie->jump) \ | |
c944940b | 1236 | trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, sizeof(U16) ); \ |
7f69552c | 1237 | trie->jump[curword] = (U16)(noper_next - convert); \ |
786e8c11 YO |
1238 | if (!jumper) \ |
1239 | jumper = noper_next; \ | |
1240 | if (!nextbranch) \ | |
1241 | nextbranch= regnext(cur); \ | |
1242 | } \ | |
1243 | \ | |
1244 | if ( dupe ) { \ | |
1245 | /* So it's a dupe. This means we need to maintain a */\ | |
1246 | /* linked-list from the first to the next. */\ | |
1247 | /* we only allocate the nextword buffer when there */\ | |
1248 | /* a dupe, so first time we have to do the allocation */\ | |
1249 | if (!trie->nextword) \ | |
c944940b | 1250 | trie->nextword = (U16 *) \ |
446bd890 | 1251 | PerlMemShared_calloc( word_count + 1, sizeof(U16)); \ |
786e8c11 YO |
1252 | while ( trie->nextword[dupe] ) \ |
1253 | dupe= trie->nextword[dupe]; \ | |
1254 | trie->nextword[dupe]= curword; \ | |
1255 | } else { \ | |
1256 | /* we haven't inserted this word yet. */ \ | |
1257 | trie->states[ state ].wordnum = curword; \ | |
1258 | } \ | |
1259 | } STMT_END | |
07be1b83 | 1260 | |
3dab1dad | 1261 | |
786e8c11 YO |
1262 | #define TRIE_TRANS_STATE(state,base,ucharcount,charid,special) \ |
1263 | ( ( base + charid >= ucharcount \ | |
1264 | && base + charid < ubound \ | |
1265 | && state == trie->trans[ base - ucharcount + charid ].check \ | |
1266 | && trie->trans[ base - ucharcount + charid ].next ) \ | |
1267 | ? trie->trans[ base - ucharcount + charid ].next \ | |
1268 | : ( state==1 ? special : 0 ) \ | |
1269 | ) | |
3dab1dad | 1270 | |
786e8c11 YO |
1271 | #define MADE_TRIE 1 |
1272 | #define MADE_JUMP_TRIE 2 | |
1273 | #define MADE_EXACT_TRIE 4 | |
3dab1dad | 1274 | |
a3621e74 | 1275 | STATIC I32 |
786e8c11 | 1276 | S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch, regnode *first, regnode *last, regnode *tail, U32 word_count, U32 flags, U32 depth) |
a3621e74 | 1277 | { |
27da23d5 | 1278 | dVAR; |
a3621e74 YO |
1279 | /* first pass, loop through and scan words */ |
1280 | reg_trie_data *trie; | |
55eed653 | 1281 | HV *widecharmap = NULL; |
2b8b4781 | 1282 | AV *revcharmap = newAV(); |
a3621e74 | 1283 | regnode *cur; |
9f7f3913 | 1284 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
a3621e74 YO |
1285 | STRLEN len = 0; |
1286 | UV uvc = 0; | |
1287 | U16 curword = 0; | |
1288 | U32 next_alloc = 0; | |
786e8c11 YO |
1289 | regnode *jumper = NULL; |
1290 | regnode *nextbranch = NULL; | |
7f69552c | 1291 | regnode *convert = NULL; |
a3621e74 | 1292 | /* we just use folder as a flag in utf8 */ |
e1ec3a88 | 1293 | const U8 * const folder = ( flags == EXACTF |
a3621e74 YO |
1294 | ? PL_fold |
1295 | : ( flags == EXACTFL | |
1296 | ? PL_fold_locale | |
1297 | : NULL | |
1298 | ) | |
1299 | ); | |
1300 | ||
2b8b4781 NC |
1301 | #ifdef DEBUGGING |
1302 | const U32 data_slot = add_data( pRExC_state, 4, "tuuu" ); | |
1303 | AV *trie_words = NULL; | |
1304 | /* along with revcharmap, this only used during construction but both are | |
1305 | * useful during debugging so we store them in the struct when debugging. | |
8e11feef | 1306 | */ |
2b8b4781 NC |
1307 | #else |
1308 | const U32 data_slot = add_data( pRExC_state, 2, "tu" ); | |
3dab1dad | 1309 | STRLEN trie_charcount=0; |
3dab1dad | 1310 | #endif |
2b8b4781 | 1311 | SV *re_trie_maxbuff; |
a3621e74 | 1312 | GET_RE_DEBUG_FLAGS_DECL; |
72f13be8 YO |
1313 | #ifndef DEBUGGING |
1314 | PERL_UNUSED_ARG(depth); | |
1315 | #endif | |
a3621e74 | 1316 | |
c944940b | 1317 | trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) ); |
a3621e74 | 1318 | trie->refcount = 1; |
3dab1dad | 1319 | trie->startstate = 1; |
786e8c11 | 1320 | trie->wordcount = word_count; |
f8fc2ecf | 1321 | RExC_rxi->data->data[ data_slot ] = (void*)trie; |
c944940b | 1322 | trie->charmap = (U16 *) PerlMemShared_calloc( 256, sizeof(U16) ); |
3dab1dad | 1323 | if (!(UTF && folder)) |
c944940b | 1324 | trie->bitmap = (char *) PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 ); |
a3621e74 | 1325 | DEBUG_r({ |
2b8b4781 | 1326 | trie_words = newAV(); |
a3621e74 | 1327 | }); |
a3621e74 | 1328 | |
0111c4fd | 1329 | re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1); |
a3621e74 | 1330 | if (!SvIOK(re_trie_maxbuff)) { |
0111c4fd | 1331 | sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT); |
a3621e74 | 1332 | } |
3dab1dad YO |
1333 | DEBUG_OPTIMISE_r({ |
1334 | PerlIO_printf( Perl_debug_log, | |
786e8c11 | 1335 | "%*smake_trie start==%d, first==%d, last==%d, tail==%d depth=%d\n", |
3dab1dad YO |
1336 | (int)depth * 2 + 2, "", |
1337 | REG_NODE_NUM(startbranch),REG_NODE_NUM(first), | |
786e8c11 | 1338 | REG_NODE_NUM(last), REG_NODE_NUM(tail), |
85c3142d | 1339 | (int)depth); |
3dab1dad | 1340 | }); |
7f69552c YO |
1341 | |
1342 | /* Find the node we are going to overwrite */ | |
1343 | if ( first == startbranch && OP( last ) != BRANCH ) { | |
1344 | /* whole branch chain */ | |
1345 | convert = first; | |
1346 | } else { | |
1347 | /* branch sub-chain */ | |
1348 | convert = NEXTOPER( first ); | |
1349 | } | |
1350 | ||
a3621e74 YO |
1351 | /* -- First loop and Setup -- |
1352 | ||
1353 | We first traverse the branches and scan each word to determine if it | |
1354 | contains widechars, and how many unique chars there are, this is | |
1355 | important as we have to build a table with at least as many columns as we | |
1356 | have unique chars. | |
1357 | ||
1358 | We use an array of integers to represent the character codes 0..255 | |
38a44b82 | 1359 | (trie->charmap) and we use a an HV* to store Unicode characters. We use the |
a3621e74 YO |
1360 | native representation of the character value as the key and IV's for the |
1361 | coded index. | |
1362 | ||
1363 | *TODO* If we keep track of how many times each character is used we can | |
1364 | remap the columns so that the table compression later on is more | |
1365 | efficient in terms of memory by ensuring most common value is in the | |
1366 | middle and the least common are on the outside. IMO this would be better | |
1367 | than a most to least common mapping as theres a decent chance the most | |
1368 | common letter will share a node with the least common, meaning the node | |
1369 | will not be compressable. With a middle is most common approach the worst | |
1370 | case is when we have the least common nodes twice. | |
1371 | ||
1372 | */ | |
1373 | ||
a3621e74 | 1374 | for ( cur = first ; cur < last ; cur = regnext( cur ) ) { |
c445ea15 | 1375 | regnode * const noper = NEXTOPER( cur ); |
e1ec3a88 | 1376 | const U8 *uc = (U8*)STRING( noper ); |
a28509cc | 1377 | const U8 * const e = uc + STR_LEN( noper ); |
a3621e74 YO |
1378 | STRLEN foldlen = 0; |
1379 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; | |
2af232bd | 1380 | const U8 *scan = (U8*)NULL; |
07be1b83 | 1381 | U32 wordlen = 0; /* required init */ |
02daf0ab YO |
1382 | STRLEN chars = 0; |
1383 | bool set_bit = trie->bitmap ? 1 : 0; /*store the first char in the bitmap?*/ | |
a3621e74 | 1384 | |
3dab1dad YO |
1385 | if (OP(noper) == NOTHING) { |
1386 | trie->minlen= 0; | |
1387 | continue; | |
1388 | } | |
02daf0ab YO |
1389 | if ( set_bit ) /* bitmap only alloced when !(UTF&&Folding) */ |
1390 | TRIE_BITMAP_SET(trie,*uc); /* store the raw first byte | |
1391 | regardless of encoding */ | |
1392 | ||
a3621e74 | 1393 | for ( ; uc < e ; uc += len ) { |
3dab1dad | 1394 | TRIE_CHARCOUNT(trie)++; |
a3621e74 | 1395 | TRIE_READ_CHAR; |
3dab1dad | 1396 | chars++; |
a3621e74 YO |
1397 | if ( uvc < 256 ) { |
1398 | if ( !trie->charmap[ uvc ] ) { | |
1399 | trie->charmap[ uvc ]=( ++trie->uniquecharcount ); | |
1400 | if ( folder ) | |
1401 | trie->charmap[ folder[ uvc ] ] = trie->charmap[ uvc ]; | |
3dab1dad | 1402 | TRIE_STORE_REVCHAR; |
a3621e74 | 1403 | } |
02daf0ab YO |
1404 | if ( set_bit ) { |
1405 | /* store the codepoint in the bitmap, and if its ascii | |
1406 | also store its folded equivelent. */ | |
1407 | TRIE_BITMAP_SET(trie,uvc); | |
0921ee73 T |
1408 | |
1409 | /* store the folded codepoint */ | |
1410 | if ( folder ) TRIE_BITMAP_SET(trie,folder[ uvc ]); | |
1411 | ||
1412 | if ( !UTF ) { | |
1413 | /* store first byte of utf8 representation of | |
1414 | codepoints in the 127 < uvc < 256 range */ | |
1415 | if (127 < uvc && uvc < 192) { | |
1416 | TRIE_BITMAP_SET(trie,194); | |
1417 | } else if (191 < uvc ) { | |
1418 | TRIE_BITMAP_SET(trie,195); | |
1419 | /* && uvc < 256 -- we know uvc is < 256 already */ | |
1420 | } | |
1421 | } | |
02daf0ab YO |
1422 | set_bit = 0; /* We've done our bit :-) */ |
1423 | } | |
a3621e74 YO |
1424 | } else { |
1425 | SV** svpp; | |
55eed653 NC |
1426 | if ( !widecharmap ) |
1427 | widecharmap = newHV(); | |
a3621e74 | 1428 | |
55eed653 | 1429 | svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 1 ); |
a3621e74 YO |
1430 | |
1431 | if ( !svpp ) | |
e4584336 | 1432 | Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%"UVXf, uvc ); |
a3621e74 YO |
1433 | |
1434 | if ( !SvTRUE( *svpp ) ) { | |
1435 | sv_setiv( *svpp, ++trie->uniquecharcount ); | |
3dab1dad | 1436 | TRIE_STORE_REVCHAR; |
a3621e74 YO |
1437 | } |
1438 | } | |
1439 | } | |
3dab1dad YO |
1440 | if( cur == first ) { |
1441 | trie->minlen=chars; | |
1442 | trie->maxlen=chars; | |
1443 | } else if (chars < trie->minlen) { | |
1444 | trie->minlen=chars; | |
1445 | } else if (chars > trie->maxlen) { | |
1446 | trie->maxlen=chars; | |
1447 | } | |
1448 | ||
a3621e74 YO |
1449 | } /* end first pass */ |
1450 | DEBUG_TRIE_COMPILE_r( | |
3dab1dad YO |
1451 | PerlIO_printf( Perl_debug_log, "%*sTRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n", |
1452 | (int)depth * 2 + 2,"", | |
55eed653 | 1453 | ( widecharmap ? "UTF8" : "NATIVE" ), (int)word_count, |
be8e71aa YO |
1454 | (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount, |
1455 | (int)trie->minlen, (int)trie->maxlen ) | |
a3621e74 | 1456 | ); |
c944940b | 1457 | trie->wordlen = (U32 *) PerlMemShared_calloc( word_count, sizeof(U32) ); |
a3621e74 YO |
1458 | |
1459 | /* | |
1460 | We now know what we are dealing with in terms of unique chars and | |
1461 | string sizes so we can calculate how much memory a naive | |
0111c4fd RGS |
1462 | representation using a flat table will take. If it's over a reasonable |
1463 | limit (as specified by ${^RE_TRIE_MAXBUF}) we use a more memory | |
a3621e74 YO |
1464 | conservative but potentially much slower representation using an array |
1465 | of lists. | |
1466 | ||
1467 | At the end we convert both representations into the same compressed | |
1468 | form that will be used in regexec.c for matching with. The latter | |
1469 | is a form that cannot be used to construct with but has memory | |
1470 | properties similar to the list form and access properties similar | |
1471 | to the table form making it both suitable for fast searches and | |
1472 | small enough that its feasable to store for the duration of a program. | |
1473 | ||
1474 | See the comment in the code where the compressed table is produced | |
1475 | inplace from the flat tabe representation for an explanation of how | |
1476 | the compression works. | |
1477 | ||
1478 | */ | |
1479 | ||
1480 | ||
3dab1dad | 1481 | if ( (IV)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1) > SvIV(re_trie_maxbuff) ) { |
a3621e74 YO |
1482 | /* |
1483 | Second Pass -- Array Of Lists Representation | |
1484 | ||
1485 | Each state will be represented by a list of charid:state records | |
1486 | (reg_trie_trans_le) the first such element holds the CUR and LEN | |
1487 | points of the allocated array. (See defines above). | |
1488 | ||
1489 | We build the initial structure using the lists, and then convert | |
1490 | it into the compressed table form which allows faster lookups | |
1491 | (but cant be modified once converted). | |
a3621e74 YO |
1492 | */ |
1493 | ||
a3621e74 YO |
1494 | STRLEN transcount = 1; |
1495 | ||
1e2e3d02 YO |
1496 | DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log, |
1497 | "%*sCompiling trie using list compiler\n", | |
1498 | (int)depth * 2 + 2, "")); | |
446bd890 | 1499 | |
c944940b JH |
1500 | trie->states = (reg_trie_state *) |
1501 | PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2, | |
1502 | sizeof(reg_trie_state) ); | |
a3621e74 YO |
1503 | TRIE_LIST_NEW(1); |
1504 | next_alloc = 2; | |
1505 | ||
1506 | for ( cur = first ; cur < last ; cur = regnext( cur ) ) { | |
1507 | ||
c445ea15 AL |
1508 | regnode * const noper = NEXTOPER( cur ); |
1509 | U8 *uc = (U8*)STRING( noper ); | |
1510 | const U8 * const e = uc + STR_LEN( noper ); | |
1511 | U32 state = 1; /* required init */ | |
1512 | U16 charid = 0; /* sanity init */ | |
1513 | U8 *scan = (U8*)NULL; /* sanity init */ | |
1514 | STRLEN foldlen = 0; /* required init */ | |
07be1b83 | 1515 | U32 wordlen = 0; /* required init */ |
c445ea15 AL |
1516 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
1517 | ||
3dab1dad | 1518 | if (OP(noper) != NOTHING) { |
786e8c11 | 1519 | for ( ; uc < e ; uc += len ) { |
c445ea15 | 1520 | |
786e8c11 | 1521 | TRIE_READ_CHAR; |
c445ea15 | 1522 | |
786e8c11 YO |
1523 | if ( uvc < 256 ) { |
1524 | charid = trie->charmap[ uvc ]; | |
c445ea15 | 1525 | } else { |
55eed653 | 1526 | SV** const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0); |
786e8c11 YO |
1527 | if ( !svpp ) { |
1528 | charid = 0; | |
1529 | } else { | |
1530 | charid=(U16)SvIV( *svpp ); | |
1531 | } | |
c445ea15 | 1532 | } |
786e8c11 YO |
1533 | /* charid is now 0 if we dont know the char read, or nonzero if we do */ |
1534 | if ( charid ) { | |
a3621e74 | 1535 | |
786e8c11 YO |
1536 | U16 check; |
1537 | U32 newstate = 0; | |
a3621e74 | 1538 | |
786e8c11 YO |
1539 | charid--; |
1540 | if ( !trie->states[ state ].trans.list ) { | |
1541 | TRIE_LIST_NEW( state ); | |
c445ea15 | 1542 | } |
786e8c11 YO |
1543 | for ( check = 1; check <= TRIE_LIST_USED( state ); check++ ) { |
1544 | if ( TRIE_LIST_ITEM( state, check ).forid == charid ) { | |
1545 | newstate = TRIE_LIST_ITEM( state, check ).newstate; | |
1546 | break; | |
1547 | } | |
1548 | } | |
1549 | if ( ! newstate ) { | |
1550 | newstate = next_alloc++; | |
1551 | TRIE_LIST_PUSH( state, charid, newstate ); | |
1552 | transcount++; | |
1553 | } | |
1554 | state = newstate; | |
1555 | } else { | |
1556 | Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc ); | |
c445ea15 | 1557 | } |
a28509cc | 1558 | } |
c445ea15 | 1559 | } |
3dab1dad | 1560 | TRIE_HANDLE_WORD(state); |
a3621e74 YO |
1561 | |
1562 | } /* end second pass */ | |
1563 | ||
1e2e3d02 YO |
1564 | /* next alloc is the NEXT state to be allocated */ |
1565 | trie->statecount = next_alloc; | |
c944940b JH |
1566 | trie->states = (reg_trie_state *) |
1567 | PerlMemShared_realloc( trie->states, | |
1568 | next_alloc | |
1569 | * sizeof(reg_trie_state) ); | |
a3621e74 | 1570 | |
3dab1dad | 1571 | /* and now dump it out before we compress it */ |
2b8b4781 NC |
1572 | DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_list(trie, widecharmap, |
1573 | revcharmap, next_alloc, | |
1574 | depth+1) | |
1e2e3d02 | 1575 | ); |
a3621e74 | 1576 | |
c944940b JH |
1577 | trie->trans = (reg_trie_trans *) |
1578 | PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) ); | |
a3621e74 YO |
1579 | { |
1580 | U32 state; | |
a3621e74 YO |
1581 | U32 tp = 0; |
1582 | U32 zp = 0; | |
1583 | ||
1584 | ||
1585 | for( state=1 ; state < next_alloc ; state ++ ) { | |
1586 | U32 base=0; | |
1587 | ||
1588 | /* | |
1589 | DEBUG_TRIE_COMPILE_MORE_r( | |
1590 | PerlIO_printf( Perl_debug_log, "tp: %d zp: %d ",tp,zp) | |
1591 | ); | |
1592 | */ | |
1593 | ||
1594 | if (trie->states[state].trans.list) { | |
1595 | U16 minid=TRIE_LIST_ITEM( state, 1).forid; | |
1596 | U16 maxid=minid; | |
a28509cc | 1597 | U16 idx; |
a3621e74 YO |
1598 | |
1599 | for( idx = 2 ; idx <= TRIE_LIST_USED( state ) ; idx++ ) { | |
c445ea15 AL |
1600 | const U16 forid = TRIE_LIST_ITEM( state, idx).forid; |
1601 | if ( forid < minid ) { | |
1602 | minid=forid; | |
1603 | } else if ( forid > maxid ) { | |
1604 | maxid=forid; | |
1605 | } | |
a3621e74 YO |
1606 | } |
1607 | if ( transcount < tp + maxid - minid + 1) { | |
1608 | transcount *= 2; | |
c944940b JH |
1609 | trie->trans = (reg_trie_trans *) |
1610 | PerlMemShared_realloc( trie->trans, | |
446bd890 NC |
1611 | transcount |
1612 | * sizeof(reg_trie_trans) ); | |
a3621e74 YO |
1613 | Zero( trie->trans + (transcount / 2), transcount / 2 , reg_trie_trans ); |
1614 | } | |
1615 | base = trie->uniquecharcount + tp - minid; | |
1616 | if ( maxid == minid ) { | |
1617 | U32 set = 0; | |
1618 | for ( ; zp < tp ; zp++ ) { | |
1619 | if ( ! trie->trans[ zp ].next ) { | |
1620 | base = trie->uniquecharcount + zp - minid; | |
1621 | trie->trans[ zp ].next = TRIE_LIST_ITEM( state, 1).newstate; | |
1622 | trie->trans[ zp ].check = state; | |
1623 | set = 1; | |
1624 | break; | |
1625 | } | |
1626 | } | |
1627 | if ( !set ) { | |
1628 | trie->trans[ tp ].next = TRIE_LIST_ITEM( state, 1).newstate; | |
1629 | trie->trans[ tp ].check = state; | |
1630 | tp++; | |
1631 | zp = tp; | |
1632 | } | |
1633 | } else { | |
1634 | for ( idx=1; idx <= TRIE_LIST_USED( state ) ; idx++ ) { | |
c445ea15 | 1635 | const U32 tid = base - trie->uniquecharcount + TRIE_LIST_ITEM( state, idx ).forid; |
a3621e74 YO |
1636 | trie->trans[ tid ].next = TRIE_LIST_ITEM( state, idx ).newstate; |
1637 | trie->trans[ tid ].check = state; | |
1638 | } | |
1639 | tp += ( maxid - minid + 1 ); | |
1640 | } | |
1641 | Safefree(trie->states[ state ].trans.list); | |
1642 | } | |
1643 | /* | |
1644 | DEBUG_TRIE_COMPILE_MORE_r( | |
1645 | PerlIO_printf( Perl_debug_log, " base: %d\n",base); | |
1646 | ); | |
1647 | */ | |
1648 | trie->states[ state ].trans.base=base; | |
1649 | } | |
cc601c31 | 1650 | trie->lasttrans = tp + 1; |
a3621e74 YO |
1651 | } |
1652 | } else { | |
1653 | /* | |
1654 | Second Pass -- Flat Table Representation. | |
1655 | ||
1656 | we dont use the 0 slot of either trans[] or states[] so we add 1 to each. | |
1657 | We know that we will need Charcount+1 trans at most to store the data | |
1658 | (one row per char at worst case) So we preallocate both structures | |
1659 | assuming worst case. | |
1660 | ||
1661 | We then construct the trie using only the .next slots of the entry | |
1662 | structs. | |
1663 | ||
1664 | We use the .check field of the first entry of the node temporarily to | |
1665 | make compression both faster and easier by keeping track of how many non | |
1666 | zero fields are in the node. | |
1667 | ||
1668 | Since trans are numbered from 1 any 0 pointer in the table is a FAIL | |
1669 | transition. | |
1670 | ||
1671 | There are two terms at use here: state as a TRIE_NODEIDX() which is a | |
1672 | number representing the first entry of the node, and state as a | |
1673 | TRIE_NODENUM() which is the trans number. state 1 is TRIE_NODEIDX(1) and | |
1674 | TRIE_NODENUM(1), state 2 is TRIE_NODEIDX(2) and TRIE_NODENUM(3) if there | |
1675 | are 2 entrys per node. eg: | |
1676 | ||
1677 | A B A B | |
1678 | 1. 2 4 1. 3 7 | |
1679 | 2. 0 3 3. 0 5 | |
1680 | 3. 0 0 5. 0 0 | |
1681 | 4. 0 0 7. 0 0 | |
1682 | ||
1683 | The table is internally in the right hand, idx form. However as we also | |
1684 | have to deal with the states array which is indexed by nodenum we have to | |
1685 | use TRIE_NODENUM() to convert. | |
1686 | ||
1687 | */ | |
1e2e3d02 YO |
1688 | DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log, |
1689 | "%*sCompiling trie using table compiler\n", | |
1690 | (int)depth * 2 + 2, "")); | |
3dab1dad | 1691 | |
c944940b JH |
1692 | trie->trans = (reg_trie_trans *) |
1693 | PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 ) | |
1694 | * trie->uniquecharcount + 1, | |
1695 | sizeof(reg_trie_trans) ); | |
1696 | trie->states = (reg_trie_state *) | |
1697 | PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2, | |
1698 | sizeof(reg_trie_state) ); | |
a3621e74 YO |
1699 | next_alloc = trie->uniquecharcount + 1; |
1700 | ||
3dab1dad | 1701 | |
a3621e74 YO |
1702 | for ( cur = first ; cur < last ; cur = regnext( cur ) ) { |
1703 | ||
c445ea15 | 1704 | regnode * const noper = NEXTOPER( cur ); |
a28509cc AL |
1705 | const U8 *uc = (U8*)STRING( noper ); |
1706 | const U8 * const e = uc + STR_LEN( noper ); | |
a3621e74 YO |
1707 | |
1708 | U32 state = 1; /* required init */ | |
1709 | ||
1710 | U16 charid = 0; /* sanity init */ | |
1711 | U32 accept_state = 0; /* sanity init */ | |
1712 | U8 *scan = (U8*)NULL; /* sanity init */ | |
1713 | ||
1714 | STRLEN foldlen = 0; /* required init */ | |
07be1b83 | 1715 | U32 wordlen = 0; /* required init */ |
a3621e74 YO |
1716 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
1717 | ||
3dab1dad | 1718 | if ( OP(noper) != NOTHING ) { |
786e8c11 | 1719 | for ( ; uc < e ; uc += len ) { |
a3621e74 | 1720 | |
786e8c11 | 1721 | TRIE_READ_CHAR; |
a3621e74 | 1722 | |
786e8c11 YO |
1723 | if ( uvc < 256 ) { |
1724 | charid = trie->charmap[ uvc ]; | |
1725 | } else { | |
55eed653 | 1726 | SV* const * const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0); |
786e8c11 | 1727 | charid = svpp ? (U16)SvIV(*svpp) : 0; |
a3621e74 | 1728 | } |
786e8c11 YO |
1729 | if ( charid ) { |
1730 | charid--; | |
1731 | if ( !trie->trans[ state + charid ].next ) { | |
1732 | trie->trans[ state + charid ].next = next_alloc; | |
1733 | trie->trans[ state ].check++; | |
1734 | next_alloc += trie->uniquecharcount; | |
1735 | } | |
1736 | state = trie->trans[ state + charid ].next; | |
1737 | } else { | |
1738 | Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc ); | |
1739 | } | |
1740 | /* charid is now 0 if we dont know the char read, or nonzero if we do */ | |
a3621e74 | 1741 | } |
a3621e74 | 1742 | } |
3dab1dad YO |
1743 | accept_state = TRIE_NODENUM( state ); |
1744 | TRIE_HANDLE_WORD(accept_state); | |
a3621e74 YO |
1745 | |
1746 | } /* end second pass */ | |
1747 | ||
3dab1dad | 1748 | /* and now dump it out before we compress it */ |
2b8b4781 NC |
1749 | DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_table(trie, widecharmap, |
1750 | revcharmap, | |
1751 | next_alloc, depth+1)); | |
a3621e74 | 1752 | |
a3621e74 YO |
1753 | { |
1754 | /* | |
1755 | * Inplace compress the table.* | |
1756 | ||
1757 | For sparse data sets the table constructed by the trie algorithm will | |
1758 | be mostly 0/FAIL transitions or to put it another way mostly empty. | |
1759 | (Note that leaf nodes will not contain any transitions.) | |
1760 | ||
1761 | This algorithm compresses the tables by eliminating most such | |
1762 | transitions, at the cost of a modest bit of extra work during lookup: | |
1763 | ||
1764 | - Each states[] entry contains a .base field which indicates the | |
1765 | index in the state[] array wheres its transition data is stored. | |
1766 | ||
1767 | - If .base is 0 there are no valid transitions from that node. | |
1768 | ||
1769 | - If .base is nonzero then charid is added to it to find an entry in | |
1770 | the trans array. | |
1771 | ||
1772 | -If trans[states[state].base+charid].check!=state then the | |
1773 | transition is taken to be a 0/Fail transition. Thus if there are fail | |
1774 | transitions at the front of the node then the .base offset will point | |
1775 | somewhere inside the previous nodes data (or maybe even into a node | |
1776 | even earlier), but the .check field determines if the transition is | |
1777 | valid. | |
1778 | ||
786e8c11 | 1779 | XXX - wrong maybe? |
a3621e74 YO |
1780 | The following process inplace converts the table to the compressed |
1781 | table: We first do not compress the root node 1,and mark its all its | |
1782 | .check pointers as 1 and set its .base pointer as 1 as well. This | |
1783 | allows to do a DFA construction from the compressed table later, and | |
1784 | ensures that any .base pointers we calculate later are greater than | |
1785 | 0. | |
1786 | ||
1787 | - We set 'pos' to indicate the first entry of the second node. | |
1788 | ||
1789 | - We then iterate over the columns of the node, finding the first and | |
1790 | last used entry at l and m. We then copy l..m into pos..(pos+m-l), | |
1791 | and set the .check pointers accordingly, and advance pos | |
1792 | appropriately and repreat for the next node. Note that when we copy | |
1793 | the next pointers we have to convert them from the original | |
1794 | NODEIDX form to NODENUM form as the former is not valid post | |
1795 | compression. | |
1796 | ||
1797 | - If a node has no transitions used we mark its base as 0 and do not | |
1798 | advance the pos pointer. | |
1799 | ||
1800 | - If a node only has one transition we use a second pointer into the | |
1801 | structure to fill in allocated fail transitions from other states. | |
1802 | This pointer is independent of the main pointer and scans forward | |
1803 | looking for null transitions that are allocated to a state. When it | |
1804 | finds one it writes the single transition into the "hole". If the | |
786e8c11 | 1805 | pointer doesnt find one the single transition is appended as normal. |
a3621e74 YO |
1806 | |
1807 | - Once compressed we can Renew/realloc the structures to release the | |
1808 | excess space. | |
1809 | ||
1810 | See "Table-Compression Methods" in sec 3.9 of the Red Dragon, | |
1811 | specifically Fig 3.47 and the associated pseudocode. | |
1812 | ||
1813 | demq | |
1814 | */ | |
a3b680e6 | 1815 | const U32 laststate = TRIE_NODENUM( next_alloc ); |
a28509cc | 1816 | U32 state, charid; |
a3621e74 | 1817 | U32 pos = 0, zp=0; |
1e2e3d02 | 1818 | trie->statecount = laststate; |
a3621e74 YO |
1819 | |
1820 | for ( state = 1 ; state < laststate ; state++ ) { | |
1821 | U8 flag = 0; | |
a28509cc AL |
1822 | const U32 stateidx = TRIE_NODEIDX( state ); |
1823 | const U32 o_used = trie->trans[ stateidx ].check; | |
1824 | U32 used = trie->trans[ stateidx ].check; | |
a3621e74 YO |
1825 | trie->trans[ stateidx ].check = 0; |
1826 | ||
1827 | for ( charid = 0 ; used && charid < trie->uniquecharcount ; charid++ ) { | |
1828 | if ( flag || trie->trans[ stateidx + charid ].next ) { | |
1829 | if ( trie->trans[ stateidx + charid ].next ) { | |
1830 | if (o_used == 1) { | |
1831 | for ( ; zp < pos ; zp++ ) { | |
1832 | if ( ! trie->trans[ zp ].next ) { | |
1833 | break; | |
1834 | } | |
1835 | } | |
1836 | trie->states[ state ].trans.base = zp + trie->uniquecharcount - charid ; | |
1837 | trie->trans[ zp ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next ); | |
1838 | trie->trans[ zp ].check = state; | |
1839 | if ( ++zp > pos ) pos = zp; | |
1840 | break; | |
1841 | } | |
1842 | used--; | |
1843 | } | |
1844 | if ( !flag ) { | |
1845 | flag = 1; | |
1846 | trie->states[ state ].trans.base = pos + trie->uniquecharcount - charid ; | |
1847 | } | |
1848 | trie->trans[ pos ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next ); | |
1849 | trie->trans[ pos ].check = state; | |
1850 | pos++; | |
1851 | } | |
1852 | } | |
1853 | } | |
cc601c31 | 1854 | trie->lasttrans = pos + 1; |
c944940b JH |
1855 | trie->states = (reg_trie_state *) |
1856 | PerlMemShared_realloc( trie->states, laststate | |
1857 | * sizeof(reg_trie_state) ); | |
a3621e74 | 1858 | DEBUG_TRIE_COMPILE_MORE_r( |
e4584336 | 1859 | PerlIO_printf( Perl_debug_log, |
3dab1dad YO |
1860 | "%*sAlloc: %d Orig: %"IVdf" elements, Final:%"IVdf". Savings of %%%5.2f\n", |
1861 | (int)depth * 2 + 2,"", | |
1862 | (int)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1 ), | |
5d7488b2 AL |
1863 | (IV)next_alloc, |
1864 | (IV)pos, | |
a3621e74 YO |
1865 | ( ( next_alloc - pos ) * 100 ) / (double)next_alloc ); |
1866 | ); | |
1867 | ||
1868 | } /* end table compress */ | |
1869 | } | |
1e2e3d02 YO |
1870 | DEBUG_TRIE_COMPILE_MORE_r( |
1871 | PerlIO_printf(Perl_debug_log, "%*sStatecount:%"UVxf" Lasttrans:%"UVxf"\n", | |
1872 | (int)depth * 2 + 2, "", | |
1873 | (UV)trie->statecount, | |
1874 | (UV)trie->lasttrans) | |
1875 | ); | |
cc601c31 | 1876 | /* resize the trans array to remove unused space */ |
c944940b JH |
1877 | trie->trans = (reg_trie_trans *) |
1878 | PerlMemShared_realloc( trie->trans, trie->lasttrans | |
1879 | * sizeof(reg_trie_trans) ); | |
a3621e74 | 1880 | |
3dab1dad | 1881 | /* and now dump out the compressed format */ |
2b8b4781 | 1882 | DEBUG_TRIE_COMPILE_r(dump_trie(trie, widecharmap, revcharmap, depth+1)); |
07be1b83 | 1883 | |
3dab1dad | 1884 | { /* Modify the program and insert the new TRIE node*/ |
3dab1dad YO |
1885 | U8 nodetype =(U8)(flags & 0xFF); |
1886 | char *str=NULL; | |
786e8c11 | 1887 | |
07be1b83 | 1888 | #ifdef DEBUGGING |
e62cc96a | 1889 | regnode *optimize = NULL; |
7122b237 YO |
1890 | #ifdef RE_TRACK_PATTERN_OFFSETS |
1891 | ||
b57a0404 JH |
1892 | U32 mjd_offset = 0; |
1893 | U32 mjd_nodelen = 0; | |
7122b237 YO |
1894 | #endif /* RE_TRACK_PATTERN_OFFSETS */ |
1895 | #endif /* DEBUGGING */ | |
a3621e74 | 1896 | /* |
3dab1dad YO |
1897 | This means we convert either the first branch or the first Exact, |
1898 | depending on whether the thing following (in 'last') is a branch | |
1899 | or not and whther first is the startbranch (ie is it a sub part of | |
1900 | the alternation or is it the whole thing.) | |
1901 | Assuming its a sub part we conver the EXACT otherwise we convert | |
1902 | the whole branch sequence, including the first. | |
a3621e74 | 1903 | */ |
3dab1dad | 1904 | /* Find the node we are going to overwrite */ |
7f69552c | 1905 | if ( first != startbranch || OP( last ) == BRANCH ) { |
07be1b83 | 1906 | /* branch sub-chain */ |
3dab1dad | 1907 | NEXT_OFF( first ) = (U16)(last - first); |
7122b237 | 1908 | #ifdef RE_TRACK_PATTERN_OFFSETS |
07be1b83 YO |
1909 | DEBUG_r({ |
1910 | mjd_offset= Node_Offset((convert)); | |
1911 | mjd_nodelen= Node_Length((convert)); | |
1912 | }); | |
7122b237 | 1913 | #endif |
7f69552c | 1914 | /* whole branch chain */ |
7122b237 YO |
1915 | } |
1916 | #ifdef RE_TRACK_PATTERN_OFFSETS | |
1917 | else { | |
7f69552c YO |
1918 | DEBUG_r({ |
1919 | const regnode *nop = NEXTOPER( convert ); | |
1920 | mjd_offset= Node_Offset((nop)); | |
1921 | mjd_nodelen= Node_Length((nop)); | |
1922 | }); | |
07be1b83 YO |
1923 | } |
1924 | DEBUG_OPTIMISE_r( | |
1925 | PerlIO_printf(Perl_debug_log, "%*sMJD offset:%"UVuf" MJD length:%"UVuf"\n", | |
1926 | (int)depth * 2 + 2, "", | |
786e8c11 | 1927 | (UV)mjd_offset, (UV)mjd_nodelen) |
07be1b83 | 1928 | ); |
7122b237 | 1929 | #endif |
3dab1dad YO |
1930 | /* But first we check to see if there is a common prefix we can |
1931 | split out as an EXACT and put in front of the TRIE node. */ | |
1932 | trie->startstate= 1; | |
55eed653 | 1933 | if ( trie->bitmap && !widecharmap && !trie->jump ) { |
3dab1dad | 1934 | U32 state; |
1e2e3d02 | 1935 | for ( state = 1 ; state < trie->statecount-1 ; state++ ) { |
a3621e74 | 1936 | U32 ofs = 0; |
8e11feef RGS |
1937 | I32 idx = -1; |
1938 | U32 count = 0; | |
1939 | const U32 base = trie->states[ state ].trans.base; | |
a3621e74 | 1940 | |
3dab1dad | 1941 | if ( trie->states[state].wordnum ) |
8e11feef | 1942 | count = 1; |
a3621e74 | 1943 | |
8e11feef | 1944 | for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) { |
cc601c31 YO |
1945 | if ( ( base + ofs >= trie->uniquecharcount ) && |
1946 | ( base + ofs - trie->uniquecharcount < trie->lasttrans ) && | |
a3621e74 YO |
1947 | trie->trans[ base + ofs - trie->uniquecharcount ].check == state ) |
1948 | { | |
3dab1dad | 1949 | if ( ++count > 1 ) { |
2b8b4781 | 1950 | SV **tmp = av_fetch( revcharmap, ofs, 0); |
07be1b83 | 1951 | const U8 *ch = (U8*)SvPV_nolen_const( *tmp ); |
8e11feef | 1952 | if ( state == 1 ) break; |
3dab1dad YO |
1953 | if ( count == 2 ) { |
1954 | Zero(trie->bitmap, ANYOF_BITMAP_SIZE, char); | |
1955 | DEBUG_OPTIMISE_r( | |
8e11feef RGS |
1956 | PerlIO_printf(Perl_debug_log, |
1957 | "%*sNew Start State=%"UVuf" Class: [", | |
1958 | (int)depth * 2 + 2, "", | |
786e8c11 | 1959 | (UV)state)); |
be8e71aa | 1960 | if (idx >= 0) { |
2b8b4781 | 1961 | SV ** const tmp = av_fetch( revcharmap, idx, 0); |
be8e71aa | 1962 | const U8 * const ch = (U8*)SvPV_nolen_const( *tmp ); |
8e11feef | 1963 | |
3dab1dad | 1964 | TRIE_BITMAP_SET(trie,*ch); |
8e11feef RGS |
1965 | if ( folder ) |
1966 | TRIE_BITMAP_SET(trie, folder[ *ch ]); | |
3dab1dad | 1967 | DEBUG_OPTIMISE_r( |
07be1b83 | 1968 | PerlIO_printf(Perl_debug_log, (char*)ch) |
3dab1dad | 1969 | ); |
8e11feef RGS |
1970 | } |
1971 | } | |
1972 | TRIE_BITMAP_SET(trie,*ch); | |
1973 | if ( folder ) | |
1974 | TRIE_BITMAP_SET(trie,folder[ *ch ]); | |
1975 | DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"%s", ch)); | |
1976 | } | |
1977 | idx = ofs; | |
1978 | } | |
3dab1dad YO |
1979 | } |
1980 | if ( count == 1 ) { | |
2b8b4781 | 1981 | SV **tmp = av_fetch( revcharmap, idx, 0); |
c490c714 YO |
1982 | STRLEN len; |
1983 | char *ch = SvPV( *tmp, len ); | |
de734bd5 A |
1984 | DEBUG_OPTIMISE_r({ |
1985 | SV *sv=sv_newmortal(); | |
8e11feef RGS |
1986 | PerlIO_printf( Perl_debug_log, |
1987 | "%*sPrefix State: %"UVuf" Idx:%"UVuf" Char='%s'\n", | |
1988 | (int)depth * 2 + 2, "", | |
de734bd5 A |
1989 | (UV)state, (UV)idx, |
1990 | pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 6, | |
1991 | PL_colors[0], PL_colors[1], | |
1992 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) | | |
1993 | PERL_PV_ESCAPE_FIRSTCHAR | |
1994 | ) | |
1995 | ); | |
1996 | }); | |
3dab1dad YO |
1997 | if ( state==1 ) { |
1998 | OP( convert ) = nodetype; | |
1999 | str=STRING(convert); | |
2000 | STR_LEN(convert)=0; | |
2001 | } | |
c490c714 YO |
2002 | STR_LEN(convert) += len; |
2003 | while (len--) | |
de734bd5 | 2004 | *str++ = *ch++; |
8e11feef | 2005 | } else { |
f9049ba1 | 2006 | #ifdef DEBUGGING |
8e11feef RGS |
2007 | if (state>1) |
2008 | DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"]\n")); | |
f9049ba1 | 2009 | #endif |
8e11feef RGS |
2010 | break; |
2011 | } | |
2012 | } | |
3dab1dad | 2013 | if (str) { |
8e11feef | 2014 | regnode *n = convert+NODE_SZ_STR(convert); |
07be1b83 | 2015 | NEXT_OFF(convert) = NODE_SZ_STR(convert); |
8e11feef | 2016 | trie->startstate = state; |
07be1b83 YO |
2017 | trie->minlen -= (state - 1); |
2018 | trie->maxlen -= (state - 1); | |
2019 | DEBUG_r({ | |
2020 | regnode *fix = convert; | |
de734bd5 | 2021 | U32 word = trie->wordcount; |
07be1b83 YO |
2022 | mjd_nodelen++; |
2023 | Set_Node_Offset_Length(convert, mjd_offset, state - 1); | |
2024 | while( ++fix < n ) { | |
2025 | Set_Node_Offset_Length(fix, 0, 0); | |
2026 | } | |
de734bd5 | 2027 | while (word--) { |
2b8b4781 | 2028 | SV ** const tmp = av_fetch( trie_words, word, 0 ); |
de734bd5 A |
2029 | if (tmp) { |
2030 | if ( STR_LEN(convert) <= SvCUR(*tmp) ) | |
2031 | sv_chop(*tmp, SvPV_nolen(*tmp) + STR_LEN(convert)); | |
2032 | else | |
2033 | sv_chop(*tmp, SvPV_nolen(*tmp) + SvCUR(*tmp)); | |
2034 | } | |
2035 | } | |
07be1b83 | 2036 | }); |
8e11feef RGS |
2037 | if (trie->maxlen) { |
2038 | convert = n; | |
2039 | } else { | |
3dab1dad | 2040 | NEXT_OFF(convert) = (U16)(tail - convert); |
a5ca303d | 2041 | DEBUG_r(optimize= n); |
3dab1dad YO |
2042 | } |
2043 | } | |
2044 | } | |
a5ca303d YO |
2045 | if (!jumper) |
2046 | jumper = last; | |
3dab1dad | 2047 | if ( trie->maxlen ) { |
8e11feef RGS |
2048 | NEXT_OFF( convert ) = (U16)(tail - convert); |
2049 | ARG_SET( convert, data_slot ); | |
786e8c11 YO |
2050 | /* Store the offset to the first unabsorbed branch in |
2051 | jump[0], which is otherwise unused by the jump logic. | |
2052 | We use this when dumping a trie and during optimisation. */ | |
2053 | if (trie->jump) | |
7f69552c | 2054 | trie->jump[0] = (U16)(nextbranch - convert); |
a5ca303d | 2055 | |
786e8c11 YO |
2056 | /* XXXX */ |
2057 | if ( !trie->states[trie->startstate].wordnum && trie->bitmap && | |
1de06328 | 2058 | ( (char *)jumper - (char *)convert) >= (int)sizeof(struct regnode_charclass) ) |
786e8c11 YO |
2059 | { |
2060 | OP( convert ) = TRIEC; | |
2061 | Copy(trie->bitmap, ((struct regnode_charclass *)convert)->bitmap, ANYOF_BITMAP_SIZE, char); | |
446bd890 | 2062 | PerlMemShared_free(trie->bitmap); |
786e8c11 YO |
2063 | trie->bitmap= NULL; |
2064 | } else | |
2065 | OP( convert ) = TRIE; | |
a3621e74 | 2066 | |
3dab1dad YO |
2067 | /* store the type in the flags */ |
2068 | convert->flags = nodetype; | |
a5ca303d YO |
2069 | DEBUG_r({ |
2070 | optimize = convert | |
2071 | + NODE_STEP_REGNODE | |
2072 | + regarglen[ OP( convert ) ]; | |
2073 | }); | |
2074 | /* XXX We really should free up the resource in trie now, | |
2075 | as we won't use them - (which resources?) dmq */ | |
3dab1dad | 2076 | } |
a3621e74 | 2077 | /* needed for dumping*/ |
e62cc96a | 2078 | DEBUG_r(if (optimize) { |
07be1b83 | 2079 | regnode *opt = convert; |
bcdf7404 | 2080 | |
e62cc96a | 2081 | while ( ++opt < optimize) { |
07be1b83 YO |
2082 | Set_Node_Offset_Length(opt,0,0); |
2083 | } | |
786e8c11 YO |
2084 | /* |
2085 | Try to clean up some of the debris left after the | |
2086 | optimisation. | |
a3621e74 | 2087 | */ |
786e8c11 | 2088 | while( optimize < jumper ) { |
07be1b83 | 2089 | mjd_nodelen += Node_Length((optimize)); |
a3621e74 | 2090 | OP( optimize ) = OPTIMIZED; |
07be1b83 | 2091 | Set_Node_Offset_Length(optimize,0,0); |
a3621e74 YO |
2092 | optimize++; |
2093 | } | |
07be1b83 | 2094 | Set_Node_Offset_Length(convert,mjd_offset,mjd_nodelen); |
a3621e74 YO |
2095 | }); |
2096 | } /* end node insert */ | |
55eed653 | 2097 | RExC_rxi->data->data[ data_slot + 1 ] = (void*)widecharmap; |
2b8b4781 NC |
2098 | #ifdef DEBUGGING |
2099 | RExC_rxi->data->data[ data_slot + TRIE_WORDS_OFFSET ] = (void*)trie_words; | |
2100 | RExC_rxi->data->data[ data_slot + 3 ] = (void*)revcharmap; | |
2101 | #else | |
2102 | SvREFCNT_dec(revcharmap); | |
07be1b83 | 2103 | #endif |
786e8c11 YO |
2104 | return trie->jump |
2105 | ? MADE_JUMP_TRIE | |
2106 | : trie->startstate>1 | |
2107 | ? MADE_EXACT_TRIE | |
2108 | : MADE_TRIE; | |
2109 | } | |
2110 | ||
2111 | STATIC void | |
2112 | S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode *stclass, U32 depth) | |
2113 | { | |
2114 | /* The Trie is constructed and compressed now so we can build a fail array now if its needed | |
2115 | ||
2116 | This is basically the Aho-Corasick algorithm. Its from exercise 3.31 and 3.32 in the | |
2117 | "Red Dragon" -- Compilers, principles, techniques, and tools. Aho, Sethi, Ullman 1985/88 | |
2118 | ISBN 0-201-10088-6 | |
2119 | ||
2120 | We find the fail state for each state in the trie, this state is the longest proper | |
2121 | suffix of the current states 'word' that is also a proper prefix of another word in our | |
2122 | trie. State 1 represents the word '' and is the thus the default fail state. This allows | |
2123 | the DFA not to have to restart after its tried and failed a word at a given point, it | |
2124 | simply continues as though it had been matching the other word in the first place. | |
2125 | Consider | |
2126 | 'abcdgu'=~/abcdefg|cdgu/ | |
2127 | When we get to 'd' we are still matching the first word, we would encounter 'g' which would | |
2128 | fail, which would bring use to the state representing 'd' in the second word where we would | |
2129 | try 'g' and succeed, prodceding to match 'cdgu'. | |
2130 | */ | |
2131 | /* add a fail transition */ | |
3251b653 NC |
2132 | const U32 trie_offset = ARG(source); |
2133 | reg_trie_data *trie=(reg_trie_data *)RExC_rxi->data->data[trie_offset]; | |
786e8c11 YO |
2134 | U32 *q; |
2135 | const U32 ucharcount = trie->uniquecharcount; | |
1e2e3d02 | 2136 | const U32 numstates = trie->statecount; |
786e8c11 YO |
2137 | const U32 ubound = trie->lasttrans + ucharcount; |
2138 | U32 q_read = 0; | |
2139 | U32 q_write = 0; | |
2140 | U32 charid; | |
2141 | U32 base = trie->states[ 1 ].trans.base; | |
2142 | U32 *fail; | |
2143 | reg_ac_data *aho; | |
2144 | const U32 data_slot = add_data( pRExC_state, 1, "T" ); | |
2145 | GET_RE_DEBUG_FLAGS_DECL; | |
2146 | #ifndef DEBUGGING | |
2147 | PERL_UNUSED_ARG(depth); | |
2148 | #endif | |
2149 | ||
2150 | ||
2151 | ARG_SET( stclass, data_slot ); | |
c944940b | 2152 | aho = (reg_ac_data *) PerlMemShared_calloc( 1, sizeof(reg_ac_data) ); |
f8fc2ecf | 2153 | RExC_rxi->data->data[ data_slot ] = (void*)aho; |
3251b653 | 2154 | aho->trie=trie_offset; |
446bd890 NC |
2155 | aho->states=(reg_trie_state *)PerlMemShared_malloc( numstates * sizeof(reg_trie_state) ); |
2156 | Copy( trie->states, aho->states, numstates, reg_trie_state ); | |
786e8c11 | 2157 | Newxz( q, numstates, U32); |
c944940b | 2158 | aho->fail = (U32 *) PerlMemShared_calloc( numstates, sizeof(U32) ); |
786e8c11 YO |
2159 | aho->refcount = 1; |
2160 | fail = aho->fail; | |
2161 | /* initialize fail[0..1] to be 1 so that we always have | |
2162 | a valid final fail state */ | |
2163 | fail[ 0 ] = fail[ 1 ] = 1; | |
2164 | ||
2165 | for ( charid = 0; charid < ucharcount ; charid++ ) { | |
2166 | const U32 newstate = TRIE_TRANS_STATE( 1, base, ucharcount, charid, 0 ); | |
2167 | if ( newstate ) { | |
2168 | q[ q_write ] = newstate; | |
2169 | /* set to point at the root */ | |
2170 | fail[ q[ q_write++ ] ]=1; | |
2171 | } | |
2172 | } | |
2173 | while ( q_read < q_write) { | |
2174 | const U32 cur = q[ q_read++ % numstates ]; | |
2175 | base = trie->states[ cur ].trans.base; | |
2176 | ||
2177 | for ( charid = 0 ; charid < ucharcount ; charid++ ) { | |
2178 | const U32 ch_state = TRIE_TRANS_STATE( cur, base, ucharcount, charid, 1 ); | |
2179 | if (ch_state) { | |
2180 | U32 fail_state = cur; | |
2181 | U32 fail_base; | |
2182 | do { | |
2183 | fail_state = fail[ fail_state ]; | |
2184 | fail_base = aho->states[ fail_state ].trans.base; | |
2185 | } while ( !TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 ) ); | |
2186 | ||
2187 | fail_state = TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 ); | |
2188 | fail[ ch_state ] = fail_state; | |
2189 | if ( !aho->states[ ch_state ].wordnum && aho->states[ fail_state ].wordnum ) | |
2190 | { | |
2191 | aho->states[ ch_state ].wordnum = aho->states[ fail_state ].wordnum; | |
2192 | } | |
2193 | q[ q_write++ % numstates] = ch_state; | |
2194 | } | |
2195 | } | |
2196 | } | |
2197 | /* restore fail[0..1] to 0 so that we "fall out" of the AC loop | |
2198 | when we fail in state 1, this allows us to use the | |
2199 | charclass scan to find a valid start char. This is based on the principle | |
2200 | that theres a good chance the string being searched contains lots of stuff | |
2201 | that cant be a start char. | |
2202 | */ | |
2203 | fail[ 0 ] = fail[ 1 ] = 0; | |
2204 | DEBUG_TRIE_COMPILE_r({ | |
6d99fb9b JH |
2205 | PerlIO_printf(Perl_debug_log, |
2206 | "%*sStclass Failtable (%"UVuf" states): 0", | |
2207 | (int)(depth * 2), "", (UV)numstates | |
1e2e3d02 | 2208 | ); |
786e8c11 YO |
2209 | for( q_read=1; q_read<numstates; q_read++ ) { |
2210 | PerlIO_printf(Perl_debug_log, ", %"UVuf, (UV)fail[q_read]); | |
2211 | } | |
2212 | PerlIO_printf(Perl_debug_log, "\n"); | |
2213 | }); | |
2214 | Safefree(q); | |
2215 | /*RExC_seen |= REG_SEEN_TRIEDFA;*/ | |
a3621e74 YO |
2216 | } |
2217 | ||
786e8c11 | 2218 | |
a3621e74 | 2219 | /* |
5d1c421c JH |
2220 | * There are strange code-generation bugs caused on sparc64 by gcc-2.95.2. |
2221 | * These need to be revisited when a newer toolchain becomes available. | |
2222 | */ | |
2223 | #if defined(__sparc64__) && defined(__GNUC__) | |
2224 | # if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 96) | |
2225 | # undef SPARC64_GCC_WORKAROUND | |
2226 | # define SPARC64_GCC_WORKAROUND 1 | |
2227 | # endif | |
2228 | #endif | |
2229 | ||
07be1b83 | 2230 | #define DEBUG_PEEP(str,scan,depth) \ |
b515a41d | 2231 | DEBUG_OPTIMISE_r({if (scan){ \ |
07be1b83 YO |
2232 | SV * const mysv=sv_newmortal(); \ |
2233 | regnode *Next = regnext(scan); \ | |
2234 | regprop(RExC_rx, mysv, scan); \ | |
7f69552c | 2235 | PerlIO_printf(Perl_debug_log, "%*s" str ">%3d: %s (%d)\n", \ |
07be1b83 YO |
2236 | (int)depth*2, "", REG_NODE_NUM(scan), SvPV_nolen_const(mysv),\ |
2237 | Next ? (REG_NODE_NUM(Next)) : 0 ); \ | |
b515a41d | 2238 | }}); |
07be1b83 | 2239 | |
1de06328 YO |
2240 | |
2241 | ||
2242 | ||
2243 | ||
07be1b83 YO |
2244 | #define JOIN_EXACT(scan,min,flags) \ |
2245 | if (PL_regkind[OP(scan)] == EXACT) \ | |
2246 | join_exact(pRExC_state,(scan),(min),(flags),NULL,depth+1) | |
2247 | ||
be8e71aa | 2248 | STATIC U32 |
07be1b83 YO |
2249 | S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan, I32 *min, U32 flags,regnode *val, U32 depth) { |
2250 | /* Merge several consecutive EXACTish nodes into one. */ | |
2251 | regnode *n = regnext(scan); | |
2252 | U32 stringok = 1; | |
2253 | regnode *next = scan + NODE_SZ_STR(scan); | |
2254 | U32 merged = 0; | |
2255 | U32 stopnow = 0; | |
2256 | #ifdef DEBUGGING | |
2257 | regnode *stop = scan; | |
72f13be8 | 2258 | GET_RE_DEBUG_FLAGS_DECL; |
f9049ba1 | 2259 | #else |
d47053eb RGS |
2260 | PERL_UNUSED_ARG(depth); |
2261 | #endif | |
2262 | #ifndef EXPERIMENTAL_INPLACESCAN | |
f9049ba1 SP |
2263 | PERL_UNUSED_ARG(flags); |
2264 | PERL_UNUSED_ARG(val); | |
07be1b83 | 2265 | #endif |
07be1b83 YO |
2266 | DEBUG_PEEP("join",scan,depth); |
2267 | ||
2268 | /* Skip NOTHING, merge EXACT*. */ | |
2269 | while (n && | |
2270 | ( PL_regkind[OP(n)] == NOTHING || | |
2271 | (stringok && (OP(n) == OP(scan)))) | |
2272 | && NEXT_OFF(n) | |
2273 | && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX) { | |
2274 | ||
2275 | if (OP(n) == TAIL || n > next) | |
2276 | stringok = 0; | |
2277 | if (PL_regkind[OP(n)] == NOTHING) { | |
07be1b83 YO |
2278 | DEBUG_PEEP("skip:",n,depth); |
2279 | NEXT_OFF(scan) += NEXT_OFF(n); | |
2280 | next = n + NODE_STEP_REGNODE; | |
2281 | #ifdef DEBUGGING | |
2282 | if (stringok) | |
2283 | stop = n; | |
2284 | #endif | |
2285 | n = regnext(n); | |
2286 | } | |
2287 | else if (stringok) { | |
786e8c11 | 2288 | const unsigned int oldl = STR_LEN(scan); |
07be1b83 YO |
2289 | regnode * const nnext = regnext(n); |
2290 | ||
2291 | DEBUG_PEEP("merg",n,depth); | |
2292 | ||
2293 | merged++; | |
2294 | if (oldl + STR_LEN(n) > U8_MAX) | |
2295 | break; | |
2296 | NEXT_OFF(scan) += NEXT_OFF(n); | |
2297 | STR_LEN(scan) += STR_LEN(n); | |
2298 | next = n + NODE_SZ_STR(n); | |
2299 | /* Now we can overwrite *n : */ | |
2300 | Move(STRING(n), STRING(scan) + oldl, STR_LEN(n), char); | |
2301 | #ifdef DEBUGGING | |
2302 | stop = next - 1; | |
2303 | #endif | |
2304 | n = nnext; | |
2305 | if (stopnow) break; | |
2306 | } | |
2307 | ||
d47053eb RGS |
2308 | #ifdef EXPERIMENTAL_INPLACESCAN |
2309 | if (flags && !NEXT_OFF(n)) { | |
2310 | DEBUG_PEEP("atch", val, depth); | |
2311 | if (reg_off_by_arg[OP(n)]) { | |
2312 | ARG_SET(n, val - n); | |
2313 | } | |
2314 | else { | |
2315 | NEXT_OFF(n) = val - n; | |
2316 | } | |
2317 | stopnow = 1; | |
2318 | } | |
07be1b83 YO |
2319 | #endif |
2320 | } | |
2321 | ||
2322 | if (UTF && ( OP(scan) == EXACTF ) && ( STR_LEN(scan) >= 6 ) ) { | |
2323 | /* | |
2324 | Two problematic code points in Unicode casefolding of EXACT nodes: | |
2325 | ||
2326 | U+0390 - GREEK SMALL LETTER IOTA WITH DIALYTIKA AND TONOS | |
2327 | U+03B0 - GREEK SMALL LETTER UPSILON WITH DIALYTIKA AND TONOS | |
2328 | ||
2329 | which casefold to | |
2330 | ||
2331 | Unicode UTF-8 | |
2332 | ||
2333 | U+03B9 U+0308 U+0301 0xCE 0xB9 0xCC 0x88 0xCC 0x81 | |
2334 | U+03C5 U+0308 U+0301 0xCF 0x85 0xCC 0x88 0xCC 0x81 | |
2335 | ||
2336 | This means that in case-insensitive matching (or "loose matching", | |
2337 | as Unicode calls it), an EXACTF of length six (the UTF-8 encoded byte | |
2338 | length of the above casefolded versions) can match a target string | |
2339 | of length two (the byte length of UTF-8 encoded U+0390 or U+03B0). | |
2340 | This would rather mess up the minimum length computation. | |
2341 | ||
2342 | What we'll do is to look for the tail four bytes, and then peek | |
2343 | at the preceding two bytes to see whether we need to decrease | |
2344 | the minimum length by four (six minus two). | |
2345 | ||
2346 | Thanks to the design of UTF-8, there cannot be false matches: | |
2347 | A sequence of valid UTF-8 bytes cannot be a subsequence of | |
2348 | another valid sequence of UTF-8 bytes. | |
2349 | ||
2350 | */ | |
2351 | char * const s0 = STRING(scan), *s, *t; | |
2352 | char * const s1 = s0 + STR_LEN(scan) - 1; | |
2353 | char * const s2 = s1 - 4; | |
e294cc5d JH |
2354 | #ifdef EBCDIC /* RD tunifold greek 0390 and 03B0 */ |
2355 | const char t0[] = "\xaf\x49\xaf\x42"; | |
2356 | #else | |
07be1b83 | 2357 | const char t0[] = "\xcc\x88\xcc\x81"; |
e294cc5d | 2358 | #endif |
07be1b83 YO |
2359 | const char * const t1 = t0 + 3; |
2360 | ||
2361 | for (s = s0 + 2; | |
2362 | s < s2 && (t = ninstr(s, s1, t0, t1)); | |
2363 | s = t + 4) { | |
e294cc5d JH |
2364 | #ifdef EBCDIC |
2365 | if (((U8)t[-1] == 0x68 && (U8)t[-2] == 0xB4) || | |
2366 | ((U8)t[-1] == 0x46 && (U8)t[-2] == 0xB5)) | |
2367 | #else | |
07be1b83 YO |
2368 | if (((U8)t[-1] == 0xB9 && (U8)t[-2] == 0xCE) || |
2369 | ((U8)t[-1] == 0x85 && (U8)t[-2] == 0xCF)) | |
e294cc5d | 2370 | #endif |
07be1b83 YO |
2371 | *min -= 4; |
2372 | } | |
2373 | } | |
2374 | ||
2375 | #ifdef DEBUGGING | |
2376 | /* Allow dumping */ | |
2377 | n = scan + NODE_SZ_STR(scan); | |
2378 | while (n <= stop) { | |
2379 | if (PL_regkind[OP(n)] != NOTHING || OP(n) == NOTHING) { | |
2380 | OP(n) = OPTIMIZED; | |
2381 | NEXT_OFF(n) = 0; | |
2382 | } | |
2383 | n++; | |
2384 | } | |
2385 | #endif | |
2386 | DEBUG_OPTIMISE_r(if (merged){DEBUG_PEEP("finl",scan,depth)}); | |
2387 | return stopnow; | |
2388 | } | |
2389 | ||
653099ff GS |
2390 | /* REx optimizer. Converts nodes into quickier variants "in place". |
2391 | Finds fixed substrings. */ | |
2392 | ||
a0288114 | 2393 | /* Stops at toplevel WHILEM as well as at "last". At end *scanp is set |
c277df42 IZ |
2394 | to the position after last scanned or to NULL. */ |
2395 | ||
40d049e4 YO |
2396 | #define INIT_AND_WITHP \ |
2397 | assert(!and_withp); \ | |
2398 | Newx(and_withp,1,struct regnode_charclass_class); \ | |
2399 | SAVEFREEPV(and_withp) | |
07be1b83 | 2400 | |
b515a41d YO |
2401 | /* this is a chain of data about sub patterns we are processing that |
2402 | need to be handled seperately/specially in study_chunk. Its so | |
2403 | we can simulate recursion without losing state. */ | |
2404 | struct scan_frame; | |
2405 | typedef struct scan_frame { | |
2406 | regnode *last; /* last node to process in this frame */ | |
2407 | regnode *next; /* next node to process when last is reached */ | |
2408 | struct scan_frame *prev; /*previous frame*/ | |
2409 | I32 stop; /* what stopparen do we use */ | |
2410 | } scan_frame; | |
2411 | ||
304ee84b YO |
2412 | |
2413 | #define SCAN_COMMIT(s, data, m) scan_commit(s, data, m, is_inf) | |
2414 | ||
e1d1eefb YO |
2415 | #define CASE_SYNST_FNC(nAmE) \ |
2416 | case nAmE: \ | |
2417 | if (flags & SCF_DO_STCLASS_AND) { \ | |
2418 | for (value = 0; value < 256; value++) \ | |
2419 | if (!is_ ## nAmE ## _cp(value)) \ | |
2420 | ANYOF_BITMAP_CLEAR(data->start_class, value); \ | |
2421 | } \ | |
2422 | else { \ | |
2423 | for (value = 0; value < 256; value++) \ | |
2424 | if (is_ ## nAmE ## _cp(value)) \ | |
2425 | ANYOF_BITMAP_SET(data->start_class, value); \ | |
2426 | } \ | |
2427 | break; \ | |
2428 | case N ## nAmE: \ | |
2429 | if (flags & SCF_DO_STCLASS_AND) { \ | |
2430 | for (value = 0; value < 256; value++) \ | |
2431 | if (is_ ## nAmE ## _cp(value)) \ | |
2432 | ANYOF_BITMAP_CLEAR(data->start_class, value); \ | |
2433 | } \ | |
2434 | else { \ | |
2435 | for (value = 0; value < 256; value++) \ | |
2436 | if (!is_ ## nAmE ## _cp(value)) \ | |
2437 | ANYOF_BITMAP_SET(data->start_class, value); \ | |
2438 | } \ | |
2439 | break | |
2440 | ||
2441 | ||
2442 | ||
76e3520e | 2443 | STATIC I32 |
40d049e4 | 2444 | S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp, |
1de06328 | 2445 | I32 *minlenp, I32 *deltap, |
40d049e4 YO |
2446 | regnode *last, |
2447 | scan_data_t *data, | |
2448 | I32 stopparen, | |
2449 | U8* recursed, | |
2450 | struct regnode_charclass_class *and_withp, | |
2451 | U32 flags, U32 depth) | |
c277df42 IZ |
2452 | /* scanp: Start here (read-write). */ |
2453 | /* deltap: Write maxlen-minlen here. */ | |
2454 | /* last: Stop before this one. */ | |
40d049e4 YO |
2455 | /* data: string data about the pattern */ |
2456 | /* stopparen: treat close N as END */ | |
2457 | /* recursed: which subroutines have we recursed into */ | |
2458 | /* and_withp: Valid if flags & SCF_DO_STCLASS_OR */ | |
c277df42 | 2459 | { |
97aff369 | 2460 | dVAR; |
c277df42 IZ |
2461 | I32 min = 0, pars = 0, code; |
2462 | regnode *scan = *scanp, *next; | |
2463 | I32 delta = 0; | |
2464 | int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF); | |
aca2d497 | 2465 | int is_inf_internal = 0; /* The studied chunk is infinite */ |
c277df42 IZ |
2466 | I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0; |
2467 | scan_data_t data_fake; | |
a3621e74 | 2468 | SV *re_trie_maxbuff = NULL; |
786e8c11 | 2469 | regnode *first_non_open = scan; |
e2e6a0f1 | 2470 | I32 stopmin = I32_MAX; |
8aa23a47 YO |
2471 | scan_frame *frame = NULL; |
2472 | ||
a3621e74 | 2473 | GET_RE_DEBUG_FLAGS_DECL; |
8aa23a47 | 2474 | |
13a24bad | 2475 | #ifdef DEBUGGING |
40d049e4 | 2476 | StructCopy(&zero_scan_data, &data_fake, scan_data_t); |
13a24bad | 2477 | #endif |
40d049e4 | 2478 | |
786e8c11 | 2479 | if ( depth == 0 ) { |
40d049e4 | 2480 | while (first_non_open && OP(first_non_open) == OPEN) |
786e8c11 YO |
2481 | first_non_open=regnext(first_non_open); |
2482 | } | |
2483 | ||
b81d288d | 2484 | |
8aa23a47 YO |
2485 | fake_study_recurse: |
2486 | while ( scan && OP(scan) != END && scan < last ){ | |
2487 | /* Peephole optimizer: */ | |
304ee84b | 2488 | DEBUG_STUDYDATA("Peep:", data,depth); |
8aa23a47 YO |
2489 | DEBUG_PEEP("Peep",scan,depth); |
2490 | JOIN_EXACT(scan,&min,0); | |
2491 | ||
2492 | /* Follow the next-chain of the current node and optimize | |
2493 | away all the NOTHINGs from it. */ | |
2494 | if (OP(scan) != CURLYX) { | |
2495 | const int max = (reg_off_by_arg[OP(scan)] | |
2496 | ? I32_MAX | |
2497 | /* I32 may be smaller than U16 on CRAYs! */ | |
2498 | : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX)); | |
2499 | int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan)); | |
2500 | int noff; | |
2501 | regnode *n = scan; | |
2502 | ||
2503 | /* Skip NOTHING and LONGJMP. */ | |
2504 | while ((n = regnext(n)) | |
2505 | && ((PL_regkind[OP(n)] == NOTHING && (noff = NEXT_OFF(n))) | |
2506 | || ((OP(n) == LONGJMP) && (noff = ARG(n)))) | |
2507 | && off + noff < max) | |
2508 | off += noff; | |
2509 | if (reg_off_by_arg[OP(scan)]) | |
2510 | ARG(scan) = off; | |
2511 | else | |
2512 | NEXT_OFF(scan) = off; | |
2513 | } | |
a3621e74 | 2514 | |
c277df42 | 2515 | |
8aa23a47 YO |
2516 | |
2517 | /* The principal pseudo-switch. Cannot be a switch, since we | |
2518 | look into several different things. */ | |
2519 | if (OP(scan) == BRANCH || OP(scan) == BRANCHJ | |
2520 | || OP(scan) == IFTHEN) { | |
2521 | next = regnext(scan); | |
2522 | code = OP(scan); | |
2523 | /* demq: the op(next)==code check is to see if we have "branch-branch" AFAICT */ | |
2524 | ||
2525 | if (OP(next) == code || code == IFTHEN) { | |
2526 | /* NOTE - There is similar code to this block below for handling | |
2527 | TRIE nodes on a re-study. If you change stuff here check there | |
2528 | too. */ | |
2529 | I32 max1 = 0, min1 = I32_MAX, num = 0; | |
2530 | struct regnode_charclass_class accum; | |
2531 | regnode * const startbranch=scan; | |
2532 | ||
2533 | if (flags & SCF_DO_SUBSTR) | |
304ee84b | 2534 | SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot merge strings after this. */ |
8aa23a47 YO |
2535 | if (flags & SCF_DO_STCLASS) |
2536 | cl_init_zero(pRExC_state, &accum); | |
2537 | ||
2538 | while (OP(scan) == code) { | |
2539 | I32 deltanext, minnext, f = 0, fake; | |
2540 | struct regnode_charclass_class this_class; | |
2541 | ||
2542 | num++; | |
2543 | data_fake.flags = 0; | |
2544 | if (data) { | |
2545 | data_fake.whilem_c = data->whilem_c; | |
2546 | data_fake.last_closep = data->last_closep; | |
2547 | } | |
2548 | else | |
2549 | data_fake.last_closep = &fake; | |
58e23c8d YO |
2550 | |
2551 | data_fake.pos_delta = delta; | |
8aa23a47 YO |
2552 | next = regnext(scan); |
2553 | scan = NEXTOPER(scan); | |
2554 | if (code != BRANCH) | |
c277df42 | 2555 | scan = NEXTOPER(scan); |
8aa23a47 YO |
2556 | if (flags & SCF_DO_STCLASS) { |
2557 | cl_init(pRExC_state, &this_class); | |
2558 | data_fake.start_class = &this_class; | |
2559 | f = SCF_DO_STCLASS_AND; | |
58e23c8d | 2560 | } |
8aa23a47 YO |
2561 | if (flags & SCF_WHILEM_VISITED_POS) |
2562 | f |= SCF_WHILEM_VISITED_POS; | |
2563 | ||
2564 | /* we suppose the run is continuous, last=next...*/ | |
2565 | minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext, | |
2566 | next, &data_fake, | |
2567 | stopparen, recursed, NULL, f,depth+1); | |
2568 | if (min1 > minnext) | |
2569 | min1 = minnext; | |
2570 | if (max1 < minnext + deltanext) | |
2571 | max1 = minnext + deltanext; | |
2572 | if (deltanext == I32_MAX) | |
2573 | is_inf = is_inf_internal = 1; | |
2574 | scan = next; | |
2575 | if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR)) | |
2576 | pars++; | |
2577 | if (data_fake.flags & SCF_SEEN_ACCEPT) { | |
2578 | if ( stopmin > minnext) | |
2579 | stopmin = min + min1; | |
2580 | flags &= ~SCF_DO_SUBSTR; | |
2581 | if (data) | |
2582 | data->flags |= SCF_SEEN_ACCEPT; | |
2583 | } | |
2584 | if (data) { | |
2585 | if (data_fake.flags & SF_HAS_EVAL) | |
2586 | data->flags |= SF_HAS_EVAL; | |
2587 | data->whilem_c = data_fake.whilem_c; | |
3dab1dad | 2588 | } |
8aa23a47 YO |
2589 | if (flags & SCF_DO_STCLASS) |
2590 | cl_or(pRExC_state, &accum, &this_class); | |
2591 | } | |
2592 | if (code == IFTHEN && num < 2) /* Empty ELSE branch */ | |
2593 | min1 = 0; | |
2594 | if (flags & SCF_DO_SUBSTR) { | |
2595 | data->pos_min += min1; | |
2596 | data->pos_delta += max1 - min1; | |
2597 | if (max1 != min1 || is_inf) | |
2598 | data->longest = &(data->longest_float); | |
2599 | } | |
2600 | min += min1; | |
2601 | delta += max1 - min1; | |
2602 | if (flags & SCF_DO_STCLASS_OR) { | |
2603 | cl_or(pRExC_state, data->start_class, &accum); | |
2604 | if (min1) { | |
2605 | cl_and(data->start_class, and_withp); | |
2606 | flags &= ~SCF_DO_STCLASS; | |
653099ff | 2607 | } |
8aa23a47 YO |
2608 | } |
2609 | else if (flags & SCF_DO_STCLASS_AND) { | |
2610 | if (min1) { | |
2611 | cl_and(data->start_class, &accum); | |
2612 | flags &= ~SCF_DO_STCLASS; | |
de0c8cb8 | 2613 | } |
8aa23a47 YO |
2614 | else { |
2615 | /* Switch to OR mode: cache the old value of | |
2616 | * data->start_class */ | |
2617 | INIT_AND_WITHP; | |
2618 | StructCopy(data->start_class, and_withp, | |
2619 | struct regnode_charclass_class); | |
2620 | flags &= ~SCF_DO_STCLASS_AND; | |
2621 | StructCopy(&accum, data->start_class, | |
2622 | struct regnode_charclass_class); | |
2623 | flags |= SCF_DO_STCLASS_OR; | |
2624 | data->start_class->flags |= ANYOF_EOS; | |
de0c8cb8 | 2625 | } |
8aa23a47 | 2626 | } |
a3621e74 | 2627 | |
8aa23a47 YO |
2628 | if (PERL_ENABLE_TRIE_OPTIMISATION && OP( startbranch ) == BRANCH ) { |
2629 | /* demq. | |
a3621e74 | 2630 | |
8aa23a47 YO |
2631 | Assuming this was/is a branch we are dealing with: 'scan' now |
2632 | points at the item that follows the branch sequence, whatever | |
2633 | it is. We now start at the beginning of the sequence and look | |
2634 | for subsequences of | |
a3621e74 | 2635 | |
8aa23a47 YO |
2636 | BRANCH->EXACT=>x1 |
2637 | BRANCH->EXACT=>x2 | |
2638 | tail | |
a3621e74 | 2639 | |
8aa23a47 | 2640 | which would be constructed from a pattern like /A|LIST|OF|WORDS/ |
a3621e74 | 2641 | |
8aa23a47 YO |
2642 | If we can find such a subseqence we need to turn the first |
2643 | element into a trie and then add the subsequent branch exact | |
2644 | strings to the trie. | |
a3621e74 | 2645 | |
8aa23a47 | 2646 | We have two cases |
a3621e74 | 2647 | |
8aa23a47 | 2648 | 1. patterns where the whole set of branch can be converted. |
a3621e74 | 2649 | |
8aa23a47 | 2650 | 2. patterns where only a subset can be converted. |
a3621e74 | 2651 | |
8aa23a47 YO |
2652 | In case 1 we can replace the whole set with a single regop |
2653 | for the trie. In case 2 we need to keep the start and end | |
2654 | branchs so | |
a3621e74 | 2655 | |
8aa23a47 YO |
2656 | 'BRANCH EXACT; BRANCH EXACT; BRANCH X' |
2657 | becomes BRANCH TRIE; BRANCH X; | |
786e8c11 | 2658 | |
8aa23a47 YO |
2659 | There is an additional case, that being where there is a |
2660 | common prefix, which gets split out into an EXACT like node | |
2661 | preceding the TRIE node. | |
a3621e74 | 2662 | |
8aa23a47 YO |
2663 | If x(1..n)==tail then we can do a simple trie, if not we make |
2664 | a "jump" trie, such that when we match the appropriate word | |
2665 | we "jump" to the appopriate tail node. Essentailly we turn | |
2666 | a nested if into a case structure of sorts. | |
b515a41d | 2667 | |
8aa23a47 YO |
2668 | */ |
2669 | ||
2670 | int made=0; | |
2671 | if (!re_trie_maxbuff) { | |
2672 | re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1); | |
2673 | if (!SvIOK(re_trie_maxbuff)) | |
2674 | sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT); | |
2675 | } | |
2676 | if ( SvIV(re_trie_maxbuff)>=0 ) { | |
2677 | regnode *cur; | |
2678 | regnode *first = (regnode *)NULL; | |
2679 | regnode *last = (regnode *)NULL; | |
2680 | regnode *tail = scan; | |
2681 | U8 optype = 0; | |
2682 | U32 count=0; | |
a3621e74 YO |
2683 | |
2684 | #ifdef DEBUGGING | |
8aa23a47 | 2685 | SV * const mysv = sv_newmortal(); /* for dumping */ |
a3621e74 | 2686 | #endif |
8aa23a47 YO |
2687 | /* var tail is used because there may be a TAIL |
2688 | regop in the way. Ie, the exacts will point to the | |
2689 | thing following the TAIL, but the last branch will | |
2690 | point at the TAIL. So we advance tail. If we | |
2691 | have nested (?:) we may have to move through several | |
2692 | tails. | |
2693 | */ | |
2694 | ||
2695 | while ( OP( tail ) == TAIL ) { | |
2696 | /* this is the TAIL generated by (?:) */ | |
2697 | tail = regnext( tail ); | |
2698 | } | |
a3621e74 | 2699 | |
8aa23a47 YO |
2700 | |
2701 | DEBUG_OPTIMISE_r({ | |
2702 | regprop(RExC_rx, mysv, tail ); | |
2703 | PerlIO_printf( Perl_debug_log, "%*s%s%s\n", | |
2704 | (int)depth * 2 + 2, "", | |
2705 | "Looking for TRIE'able sequences. Tail node is: ", | |
2706 | SvPV_nolen_const( mysv ) | |
2707 | ); | |
2708 | }); | |
2709 | ||
2710 | /* | |
2711 | ||
2712 | step through the branches, cur represents each | |
2713 | branch, noper is the first thing to be matched | |
2714 | as part of that branch and noper_next is the | |
2715 | regnext() of that node. if noper is an EXACT | |
2716 | and noper_next is the same as scan (our current | |
2717 | position in the regex) then the EXACT branch is | |
2718 | a possible optimization target. Once we have | |
2719 | two or more consequetive such branches we can | |
2720 | create a trie of the EXACT's contents and stich | |
2721 | it in place. If the sequence represents all of | |
2722 | the branches we eliminate the whole thing and | |
2723 | replace it with a single TRIE. If it is a | |
2724 | subsequence then we need to stitch it in. This | |
2725 | means the first branch has to remain, and needs | |
2726 | to be repointed at the item on the branch chain | |
2727 | following the last branch optimized. This could | |
2728 | be either a BRANCH, in which case the | |
2729 | subsequence is internal, or it could be the | |
2730 | item following the branch sequence in which | |
2731 | case the subsequence is at the end. | |
2732 | ||
2733 | */ | |
2734 | ||
2735 | /* dont use tail as the end marker for this traverse */ | |
2736 | for ( cur = startbranch ; cur != scan ; cur = regnext( cur ) ) { | |
2737 | regnode * const noper = NEXTOPER( cur ); | |
b515a41d | 2738 | #if defined(DEBUGGING) || defined(NOJUMPTRIE) |
8aa23a47 | 2739 | regnode * const noper_next = regnext( noper ); |
b515a41d YO |
2740 | #endif |
2741 | ||
8aa23a47 YO |
2742 | DEBUG_OPTIMISE_r({ |
2743 | regprop(RExC_rx, mysv, cur); | |
2744 | PerlIO_printf( Perl_debug_log, "%*s- %s (%d)", | |
2745 | (int)depth * 2 + 2,"", SvPV_nolen_const( mysv ), REG_NODE_NUM(cur) ); | |
2746 | ||
2747 | regprop(RExC_rx, mysv, noper); | |
2748 | PerlIO_printf( Perl_debug_log, " -> %s", | |
2749 | SvPV_nolen_const(mysv)); | |
2750 | ||
2751 | if ( noper_next ) { | |
2752 | regprop(RExC_rx, mysv, noper_next ); | |
2753 | PerlIO_printf( Perl_debug_log,"\t=> %s\t", | |
2754 | SvPV_nolen_const(mysv)); | |
2755 | } | |
2756 | PerlIO_printf( Perl_debug_log, "(First==%d,Last==%d,Cur==%d)\n", | |
2757 | REG_NODE_NUM(first), REG_NODE_NUM(last), REG_NODE_NUM(cur) ); | |
2758 | }); | |
2759 | if ( (((first && optype!=NOTHING) ? OP( noper ) == optype | |
2760 | : PL_regkind[ OP( noper ) ] == EXACT ) | |
2761 | || OP(noper) == NOTHING ) | |
786e8c11 | 2762 | #ifdef NOJUMPTRIE |
8aa23a47 | 2763 | && noper_next == tail |
786e8c11 | 2764 | #endif |
8aa23a47 YO |
2765 | && count < U16_MAX) |
2766 | { | |
2767 | count++; | |
2768 | if ( !first || optype == NOTHING ) { | |
2769 | if (!first) first = cur; | |
2770 | optype = OP( noper ); | |
2771 | } else { | |
2772 | last = cur; | |
2773 | } | |
2774 | } else { | |
2775 | if ( last ) { | |
2776 | make_trie( pRExC_state, | |
2777 | startbranch, first, cur, tail, count, | |
2778 | optype, depth+1 ); | |
2779 | } | |
2780 | if ( PL_regkind[ OP( noper ) ] == EXACT | |
786e8c11 | 2781 | #ifdef NOJUMPTRIE |
8aa23a47 | 2782 | && noper_next == tail |
786e8c11 | 2783 | #endif |
8aa23a47 YO |
2784 | ){ |
2785 | count = 1; | |
2786 | first = cur; | |
2787 | optype = OP( noper ); | |
2788 | } else { | |
2789 | count = 0; | |
2790 | first = NULL; | |
2791 | optype = 0; | |
2792 | } | |
2793 | last = NULL; | |
2794 | } | |
2795 | } | |
2796 | DEBUG_OPTIMISE_r({ | |
2797 | regprop(RExC_rx, mysv, cur); | |
2798 | PerlIO_printf( Perl_debug_log, | |
2799 | "%*s- %s (%d) <SCAN FINISHED>\n", (int)depth * 2 + 2, | |
2800 | "", SvPV_nolen_const( mysv ),REG_NODE_NUM(cur)); | |
2801 | ||
2802 | }); | |
2803 | if ( last ) { | |
2804 | made= make_trie( pRExC_state, startbranch, first, scan, tail, count, optype, depth+1 ); | |
3dab1dad | 2805 | #ifdef TRIE_STUDY_OPT |
8aa23a47 YO |
2806 | if ( ((made == MADE_EXACT_TRIE && |
2807 | startbranch == first) | |
2808 | || ( first_non_open == first )) && | |
2809 | depth==0 ) { | |
2810 | flags |= SCF_TRIE_RESTUDY; | |
2811 | if ( startbranch == first | |
2812 | && scan == tail ) | |
2813 | { | |
2814 | RExC_seen &=~REG_TOP_LEVEL_BRANCHES; | |
2815 | } | |
2816 | } | |
3dab1dad | 2817 | #endif |
8aa23a47 YO |
2818 | } |
2819 | } | |
2820 | ||
2821 | } /* do trie */ | |
2822 | ||
653099ff | 2823 | } |
8aa23a47 YO |
2824 | else if ( code == BRANCHJ ) { /* single branch is optimized. */ |
2825 | scan = NEXTOPER(NEXTOPER(scan)); | |
2826 | } else /* single branch is optimized. */ | |
2827 | scan = NEXTOPER(scan); | |
2828 | continue; | |
2829 | } else if (OP(scan) == SUSPEND || OP(scan) == GOSUB || OP(scan) == GOSTART) { | |
2830 | scan_frame *newframe = NULL; | |
2831 | I32 paren; | |
2832 | regnode *start; | |
2833 | regnode *end; | |
2834 | ||
2835 | if (OP(scan) != SUSPEND) { | |
2836 | /* set the pointer */ | |
2837 | if (OP(scan) == GOSUB) { | |
2838 | paren = ARG(scan); | |
2839 | RExC_recurse[ARG2L(scan)] = scan; | |
2840 | start = RExC_open_parens[paren-1]; | |
2841 | end = RExC_close_parens[paren-1]; | |
2842 | } else { | |
2843 | paren = 0; | |
f8fc2ecf | 2844 | start = RExC_rxi->program + 1; |
8aa23a47 YO |
2845 | end = RExC_opend; |
2846 | } | |
2847 | if (!recursed) { | |
2848 | Newxz(recursed, (((RExC_npar)>>3) +1), U8); | |
2849 | SAVEFREEPV(recursed); | |
2850 | } | |
2851 | if (!PAREN_TEST(recursed,paren+1)) { | |
2852 | PAREN_SET(recursed,paren+1); | |
2853 | Newx(newframe,1,scan_frame); | |
2854 | } else { | |
2855 | if (flags & SCF_DO_SUBSTR) { | |
304ee84b | 2856 | SCAN_COMMIT(pRExC_state,data,minlenp); |
8aa23a47 YO |
2857 | data->longest = &(data->longest_float); |
2858 | } | |
2859 | is_inf = is_inf_internal = 1; | |
2860 | if (flags & SCF_DO_STCLASS_OR) /* Allow everything */ | |
2861 | cl_anything(pRExC_state, data->start_class); | |
2862 | flags &= ~SCF_DO_STCLASS; | |
2863 | } | |
2864 | } else { | |
2865 | Newx(newframe,1,scan_frame); | |
2866 | paren = stopparen; | |
2867 | start = scan+2; | |
2868 | end = regnext(scan); | |
2869 | } | |
2870 | if (newframe) { | |
2871 | assert(start); | |
2872 | assert(end); | |
2873 | SAVEFREEPV(newframe); | |
2874 | newframe->next = regnext(scan); | |
2875 | newframe->last = last; | |
2876 | newframe->stop = stopparen; | |
2877 | newframe->prev = frame; | |
2878 | ||
2879 | frame = newframe; | |
2880 | scan = start; | |
2881 | stopparen = paren; | |
2882 | last = end; | |
2883 | ||
2884 | continue; | |
2885 | } | |
2886 | } | |
2887 | else if (OP(scan) == EXACT) { | |
2888 | I32 l = STR_LEN(scan); | |
2889 | UV uc; | |
2890 | if (UTF) { | |
2891 | const U8 * const s = (U8*)STRING(scan); | |
2892 | l = utf8_length(s, s + l); | |
2893 | uc = utf8_to_uvchr(s, NULL); | |
2894 | } else { | |
2895 | uc = *((U8*)STRING(scan)); | |
2896 | } | |
2897 | min += l; | |
2898 | if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */ | |
2899 | /* The code below prefers earlier match for fixed | |
2900 | offset, later match for variable offset. */ | |
2901 | if (data->last_end == -1) { /* Update the start info. */ | |
2902 | data->last_start_min = data->pos_min; | |
2903 | data->last_start_max = is_inf | |
2904 | ? I32_MAX : data->pos_min + data->pos_delta; | |
b515a41d | 2905 | } |
8aa23a47 YO |
2906 | sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan)); |
2907 | if (UTF) | |
2908 | SvUTF8_on(data->last_found); | |
2909 | { | |
2910 | SV * const sv = data->last_found; | |
2911 | MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ? | |
2912 | mg_find(sv, PERL_MAGIC_utf8) : NULL; | |
2913 | if (mg && mg->mg_len >= 0) | |
2914 | mg->mg_len += utf8_length((U8*)STRING(scan), | |
2915 | (U8*)STRING(scan)+STR_LEN(scan)); | |
b515a41d | 2916 | } |
8aa23a47 YO |
2917 | data->last_end = data->pos_min + l; |
2918 | data->pos_min += l; /* As in the first entry. */ | |
2919 | data->flags &= ~SF_BEFORE_EOL; | |
2920 | } | |
2921 | if (flags & SCF_DO_STCLASS_AND) { | |
2922 | /* Check whether it is compatible with what we know already! */ | |
2923 | int compat = 1; | |
2924 | ||
2925 | if (uc >= 0x100 || | |
2926 | (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE)) | |
2927 | && !ANYOF_BITMAP_TEST(data->start_class, uc) | |
2928 | && (!(data->start_class->flags & ANYOF_FOLD) | |
2929 | || !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc]))) | |
2930 | ) | |
2931 | compat = 0; | |
2932 | ANYOF_CLASS_ZERO(data->start_class); | |
2933 | ANYOF_BITMAP_ZERO(data->start_class); | |
2934 | if (compat) | |
2935 | ANYOF_BITMAP_SET(data->start_class, uc); | |
2936 | data->start_class->flags &= ~ANYOF_EOS; | |
2937 | if (uc < 0x100) | |
2938 | data->start_class->flags &= ~ANYOF_UNICODE_ALL; | |
2939 | } | |
2940 | else if (flags & SCF_DO_STCLASS_OR) { | |
2941 | /* false positive possible if the class is case-folded */ | |
2942 | if (uc < 0x100) | |
2943 | ANYOF_BITMAP_SET(data->start_class, uc); | |
2944 | else | |
2945 | data->start_class->flags |= ANYOF_UNICODE_ALL; | |
2946 | data->start_class->flags &= ~ANYOF_EOS; | |
2947 | cl_and(data->start_class, and_withp); | |
2948 | } | |
2949 | flags &= ~SCF_DO_STCLASS; | |
2950 | } | |
2951 | else if (PL_regkind[OP(scan)] == EXACT) { /* But OP != EXACT! */ | |
2952 | I32 l = STR_LEN(scan); | |
2953 | UV uc = *((U8*)STRING(scan)); | |
2954 | ||
2955 | /* Search for fixed substrings supports EXACT only. */ | |
2956 | if (flags & SCF_DO_SUBSTR) { | |
2957 | assert(data); | |
304ee84b | 2958 | SCAN_COMMIT(pRExC_state, data, minlenp); |
8aa23a47 YO |
2959 | } |
2960 | if (UTF) { | |
2961 | const U8 * const s = (U8 *)STRING(scan); | |
2962 | l = utf8_length(s, s + l); | |
2963 | uc = utf8_to_uvchr(s, NULL); | |
2964 | } | |
2965 | min += l; | |
2966 | if (flags & SCF_DO_SUBSTR) | |
2967 | data->pos_min += l; | |
2968 | if (flags & SCF_DO_STCLASS_AND) { | |
2969 | /* Check whether it is compatible with what we know already! */ | |
2970 | int compat = 1; | |
2971 | ||
2972 | if (uc >= 0x100 || | |
2973 | (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE)) | |
2974 | && !ANYOF_BITMAP_TEST(data->start_class, uc) | |
2975 | && !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc]))) | |
2976 | compat = 0; | |
2977 | ANYOF_CLASS_ZERO(data->start_class); | |
2978 | ANYOF_BITMAP_ZERO(data->start_class); | |
2979 | if (compat) { | |
2980 | ANYOF_BITMAP_SET(data->start_class, uc); | |
653099ff | 2981 | data->start_class->flags &= ~ANYOF_EOS; |
8aa23a47 YO |
2982 | data->start_class->flags |= ANYOF_FOLD; |
2983 | if (OP(scan) == EXACTFL) | |
2984 | data->start_class->flags |= ANYOF_LOCALE; | |
653099ff | 2985 | } |
8aa23a47 YO |
2986 | } |
2987 | else if (flags & SCF_DO_STCLASS_OR) { | |
2988 | if (data->start_class->flags & ANYOF_FOLD) { | |
2989 | /* false positive possible if the class is case-folded. | |
2990 | Assume that the locale settings are the same... */ | |
1aa99e6b IH |
2991 | if (uc < 0x100) |
2992 | ANYOF_BITMAP_SET(data->start_class, uc); | |
653099ff GS |
2993 | data->start_class->flags &= ~ANYOF_EOS; |
2994 | } | |
8aa23a47 | 2995 | cl_and(data->start_class, and_withp); |
653099ff | 2996 | } |
8aa23a47 YO |
2997 | flags &= ~SCF_DO_STCLASS; |
2998 | } | |
2999 | else if (strchr((const char*)PL_varies,OP(scan))) { | |
3000 | I32 mincount, maxcount, minnext, deltanext, fl = 0; | |
3001 | I32 f = flags, pos_before = 0; | |
3002 | regnode * const oscan = scan; | |
3003 | struct regnode_charclass_class this_class; | |
3004 | struct regnode_charclass_class *oclass = NULL; | |
3005 | I32 next_is_eval = 0; | |
3006 | ||
3007 | switch (PL_regkind[OP(scan)]) { | |
3008 | case WHILEM: /* End of (?:...)* . */ | |
3009 | scan = NEXTOPER(scan); | |
3010 | goto finish; | |
3011 | case PLUS: | |
3012 | if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) { | |
3013 | next = NEXTOPER(scan); | |
3014 | if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) { | |
3015 | mincount = 1; | |
3016 | maxcount = REG_INFTY; | |
3017 | next = regnext(scan); | |
3018 | scan = NEXTOPER(scan); | |
3019 | goto do_curly; | |
3020 | } | |
3021 | } | |
3022 | if (flags & SCF_DO_SUBSTR) | |
3023 | data->pos_min++; | |
3024 | min++; | |
3025 | /* Fall through. */ | |
3026 | case STAR: | |
3027 | if (flags & SCF_DO_STCLASS) { | |
3028 | mincount = 0; | |
3029 | maxcount = REG_INFTY; | |
3030 | next = regnext(scan); | |
3031 | scan = NEXTOPER(scan); | |
3032 | goto do_curly; | |
3033 | } | |
3034 | is_inf = is_inf_internal = 1; | |
3035 | scan = regnext(scan); | |
c277df42 | 3036 | if (flags & SCF_DO_SUBSTR) { |
304ee84b | 3037 | SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot extend fixed substrings */ |
8aa23a47 | 3038 | data->longest = &(data->longest_float); |
c277df42 | 3039 | } |
8aa23a47 YO |
3040 | goto optimize_curly_tail; |
3041 | case CURLY: | |
3042 | if (stopparen>0 && (OP(scan)==CURLYN || OP(scan)==CURLYM) | |
3043 | && (scan->flags == stopparen)) | |
3044 | { | |
3045 | mincount = 1; | |
3046 | maxcount = 1; | |
3047 | } else { | |
3048 | mincount = ARG1(scan); | |
3049 | maxcount = ARG2(scan); | |
653099ff | 3050 | } |
8aa23a47 YO |
3051 | next = regnext(scan); |
3052 | if (OP(scan) == CURLYX) { | |
3053 | I32 lp = (data ? *(data->last_closep) : 0); | |
3054 | scan->flags = ((lp <= (I32)U8_MAX) ? (U8)lp : U8_MAX); | |
653099ff | 3055 | } |
8aa23a47 YO |
3056 | scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS; |
3057 | next_is_eval = (OP(scan) == EVAL); | |
3058 | do_curly: | |
3059 | if (flags & SCF_DO_SUBSTR) { | |
304ee84b | 3060 | if (mincount == 0) SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot extend fixed substrings */ |
8aa23a47 | 3061 | pos_before = data->pos_min; |
b45f050a | 3062 | } |
8aa23a47 YO |
3063 | if (data) { |
3064 | fl = data->flags; | |
3065 | data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL); | |
3066 | if (is_inf) | |
3067 | data->flags |= SF_IS_INF; | |
3068 | } | |
3069 | if (flags & SCF_DO_STCLASS) { | |
3070 | cl_init(pRExC_state, &this_class); | |
3071 | oclass = data->start_class; | |
3072 | data->start_class = &this_class; | |
3073 | f |= SCF_DO_STCLASS_AND; | |
3074 | f &= ~SCF_DO_STCLASS_OR; | |
3075 | } | |
3076 | /* These are the cases when once a subexpression | |
3077 | fails at a particular position, it cannot succeed | |
3078 | even after backtracking at the enclosing scope. | |
3079 | ||
3080 | XXXX what if minimal match and we are at the | |
3081 | initial run of {n,m}? */ | |
3082 | if ((mincount != maxcount - 1) && (maxcount != REG_INFTY)) | |
3083 | f &= ~SCF_WHILEM_VISITED_POS; | |
b45f050a | 3084 | |
8aa23a47 YO |
3085 | /* This will finish on WHILEM, setting scan, or on NULL: */ |
3086 | minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext, | |
3087 | last, data, stopparen, recursed, NULL, | |
3088 | (mincount == 0 | |
3089 | ? (f & ~SCF_DO_SUBSTR) : f),depth+1); | |
b515a41d | 3090 | |
8aa23a47 YO |
3091 | if (flags & SCF_DO_STCLASS) |
3092 | data->start_class = oclass; | |
3093 | if (mincount == 0 || minnext == 0) { | |
3094 | if (flags & SCF_DO_STCLASS_OR) { | |
3095 | cl_or(pRExC_state, data->start_class, &this_class); | |
3096 | } | |
3097 | else if (flags & SCF_DO_STCLASS_AND) { | |
3098 | /* Switch to OR mode: cache the old value of | |
3099 | * data->start_class */ | |
3100 | INIT_AND_WITHP; | |
3101 | StructCopy(data->start_class, and_withp, | |
3102 | struct regnode_charclass_class); | |
3103 | flags &= ~SCF_DO_STCLASS_AND; | |
3104 | StructCopy(&this_class, data->start_class, | |
3105 | struct regnode_charclass_class); | |
3106 | flags |= SCF_DO_STCLASS_OR; | |
3107 | data->start_class->flags |= ANYOF_EOS; | |
3108 | } | |
3109 | } else { /* Non-zero len */ | |
3110 | if (flags & SCF_DO_STCLASS_OR) { | |
3111 | cl_or(pRExC_state, data->start_class, &this_class); | |
3112 | cl_and(data->start_class, and_withp); | |
3113 | } | |
3114 | else if (flags & SCF_DO_STCLASS_AND) | |
3115 | cl_and(data->start_class, &this_class); | |
3116 | flags &= ~SCF_DO_STCLASS; | |
3117 | } | |
3118 | if (!scan) /* It was not CURLYX, but CURLY. */ | |
3119 | scan = next; | |
3120 | if ( /* ? quantifier ok, except for (?{ ... }) */ | |
3121 | (next_is_eval || !(mincount == 0 && maxcount == 1)) | |
3122 | && (minnext == 0) && (deltanext == 0) | |
3123 | && data && !(data->flags & (SF_HAS_PAR|SF_IN_PAR)) | |
3124 | && maxcount <= REG_INFTY/3 /* Complement check for big count */ | |
3125 | && ckWARN(WARN_REGEXP)) | |
3126 | { | |
3127 | vWARN(RExC_parse, | |
3128 | "Quantifier unexpected on zero-length expression"); | |
3129 | } | |
3130 | ||
3131 | min += minnext * mincount; | |
3132 | is_inf_internal |= ((maxcount == REG_INFTY | |
3133 | && (minnext + deltanext) > 0) | |
3134 | || deltanext == I32_MAX); | |
3135 | is_inf |= is_inf_internal; | |
3136 | delta += (minnext + deltanext) * maxcount - minnext * mincount; | |
3137 | ||
3138 | /* Try powerful optimization CURLYX => CURLYN. */ | |
3139 | if ( OP(oscan) == CURLYX && data | |
3140 | && data->flags & SF_IN_PAR | |
3141 | && !(data->flags & SF_HAS_EVAL) | |
3142 | && !deltanext && minnext == 1 ) { | |
3143 | /* Try to optimize to CURLYN. */ | |
3144 | regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; | |
3145 | regnode * const nxt1 = nxt; | |
497b47a8 | 3146 | #ifdef DEBUGGING |
8aa23a47 | 3147 | regnode *nxt2; |
497b47a8 | 3148 | #endif |
c277df42 | 3149 | |
8aa23a47 YO |
3150 | /* Skip open. */ |
3151 | nxt = regnext(nxt); | |
3152 | if (!strchr((const char*)PL_simple,OP(nxt)) | |
3153 | && !(PL_regkind[OP(nxt)] == EXACT | |
3154 | && STR_LEN(nxt) == 1)) | |
3155 | goto nogo; | |
497b47a8 | 3156 | #ifdef DEBUGGING |
8aa23a47 | 3157 | nxt2 = nxt; |
497b47a8 | 3158 | #endif |
8aa23a47 YO |
3159 | nxt = regnext(nxt); |
3160 | if (OP(nxt) != CLOSE) | |
3161 | goto nogo; | |
3162 | if (RExC_open_parens) { | |
3163 | RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/ | |
3164 | RExC_close_parens[ARG(nxt1)-1]=nxt+2; /*close->while*/ | |
3165 | } | |
3166 | /* Now we know that nxt2 is the only contents: */ | |
3167 | oscan->flags = (U8)ARG(nxt); | |
3168 | OP(oscan) = CURLYN; | |
3169 | OP(nxt1) = NOTHING; /* was OPEN. */ | |
40d049e4 | 3170 | |
c277df42 | 3171 | #ifdef DEBUGGING |
8aa23a47 YO |
3172 | OP(nxt1 + 1) = OPTIMIZED; /* was count. */ |
3173 | NEXT_OFF(nxt1+ 1) = 0; /* just for consistancy. */ | |
3174 | NEXT_OFF(nxt2) = 0; /* just for consistancy with CURLY. */ | |
3175 | OP(nxt) = OPTIMIZED; /* was CLOSE. */ | |
3176 | OP(nxt + 1) = OPTIMIZED; /* was count. */ | |
3177 | NEXT_OFF(nxt+ 1) = 0; /* just for consistancy. */ | |
b81d288d | 3178 | #endif |
8aa23a47 YO |
3179 | } |
3180 | nogo: | |
3181 | ||
3182 | /* Try optimization CURLYX => CURLYM. */ | |
3183 | if ( OP(oscan) == CURLYX && data | |
3184 | && !(data->flags & SF_HAS_PAR) | |
3185 | && !(data->flags & SF_HAS_EVAL) | |
3186 | && !deltanext /* atom is fixed width */ | |
3187 | && minnext != 0 /* CURLYM can't handle zero width */ | |
3188 | ) { | |
3189 | /* XXXX How to optimize if data == 0? */ | |
3190 | /* Optimize to a simpler form. */ | |
3191 | regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */ | |
3192 | regnode *nxt2; | |
3193 | ||
3194 | OP(oscan) = CURLYM; | |
3195 | while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/ | |
3196 | && (OP(nxt2) != WHILEM)) | |
3197 | nxt = nxt2; | |
3198 | OP(nxt2) = SUCCEED; /* Whas WHILEM */ | |
3199 | /* Need to optimize away parenths. */ | |
3200 | if (data->flags & SF_IN_PAR) { | |
3201 | /* Set the parenth number. */ | |
3202 | regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/ | |
3203 | ||
3204 | if (OP(nxt) != CLOSE) | |
3205 | FAIL("Panic opt close"); | |
3206 | oscan->flags = (U8)ARG(nxt); | |
3207 | if (RExC_open_parens) { | |
3208 | RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/ | |
3209 | RExC_close_parens[ARG(nxt1)-1]=nxt2+1; /*close->NOTHING*/ | |
40d049e4 | 3210 | } |
8aa23a47 YO |
3211 | OP(nxt1) = OPTIMIZED; /* was OPEN. */ |
3212 | OP(nxt) = OPTIMIZED; /* was CLOSE. */ | |
40d049e4 | 3213 | |
c277df42 | 3214 | #ifdef DEBUGGING |
8aa23a47 YO |
3215 | OP(nxt1 + 1) = OPTIMIZED; /* was count. */ |
3216 | OP(nxt + 1) = OPTIMIZED; /* was count. */ | |
3217 | NEXT_OFF(nxt1 + 1) = 0; /* just for consistancy. */ | |
3218 | NEXT_OFF(nxt + 1) = 0; /* just for consistancy. */ | |
b81d288d | 3219 | #endif |
c277df42 | 3220 | #if 0 |
8aa23a47 YO |
3221 | while ( nxt1 && (OP(nxt1) != WHILEM)) { |
3222 | regnode *nnxt = regnext(nxt1); | |
3223 | ||
3224 | if (nnxt == nxt) { | |
3225 | if (reg_off_by_arg[OP(nxt1)]) | |
3226 | ARG_SET(nxt1, nxt2 - nxt1); | |
3227 | else if (nxt2 - nxt1 < U16_MAX) | |
3228 | NEXT_OFF(nxt1) = nxt2 - nxt1; | |
3229 | else | |
3230 | OP(nxt) = NOTHING; /* Cannot beautify */ | |
c277df42 | 3231 | } |
8aa23a47 | 3232 | nxt1 = nnxt; |
c277df42 | 3233 | } |
5d1c421c | 3234 | #endif |
8aa23a47 YO |
3235 | /* Optimize again: */ |
3236 | study_chunk(pRExC_state, &nxt1, minlenp, &deltanext, nxt, | |
3237 | NULL, stopparen, recursed, NULL, 0,depth+1); | |
3238 | } | |
3239 | else | |
3240 | oscan->flags = 0; | |
3241 | } | |
3242 | else if ((OP(oscan) == CURLYX) | |
3243 | && (flags & SCF_WHILEM_VISITED_POS) | |
3244 | /* See the comment on a similar expression above. | |
3245 | However, this time it not a subexpression | |
3246 | we care about, but the expression itself. */ | |
3247 | && (maxcount == REG_INFTY) | |
3248 | && data && ++data->whilem_c < 16) { | |
3249 | /* This stays as CURLYX, we can put the count/of pair. */ | |
3250 | /* Find WHILEM (as in regexec.c) */ | |
3251 | regnode *nxt = oscan + NEXT_OFF(oscan); | |
3252 | ||
3253 | if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */ | |
3254 | nxt += ARG(nxt); | |
3255 | PREVOPER(nxt)->flags = (U8)(data->whilem_c | |
3256 | | (RExC_whilem_seen << 4)); /* On WHILEM */ | |
3257 | } | |
3258 | if (data && fl & (SF_HAS_PAR|SF_IN_PAR)) | |
3259 | pars++; | |
3260 | if (flags & SCF_DO_SUBSTR) { | |
3261 | SV *last_str = NULL; | |
3262 | int counted = mincount != 0; | |
a0ed51b3 | 3263 | |
8aa23a47 YO |
3264 | if (data->last_end > 0 && mincount != 0) { /* Ends with a string. */ |
3265 | #if defined(SPARC64_GCC_WORKAROUND) | |
3266 | I32 b = 0; | |
3267 | STRLEN l = 0; | |
3268 | const char *s = NULL; | |
3269 | I32 old = 0; | |
b515a41d | 3270 | |
8aa23a47 YO |
3271 | if (pos_before >= data->last_start_min) |
3272 | b = pos_before; | |
3273 | else | |
3274 | b = data->last_start_min; | |
b515a41d | 3275 | |
8aa23a47 YO |
3276 | l = 0; |
3277 | s = SvPV_const(data->last_found, l); | |
3278 | old = b - data->last_start_min; | |
3279 | ||
3280 | #else | |
3281 | I32 b = pos_before >= data->last_start_min | |
3282 | ? pos_before : data->last_start_min; | |
3283 | STRLEN l; | |
3284 | const char * const s = SvPV_const(data->last_found, l); | |
3285 | I32 old = b - data->last_start_min; | |
3286 | #endif | |
3287 | ||
3288 | if (UTF) | |
3289 | old = utf8_hop((U8*)s, old) - (U8*)s; | |
3290 | ||
3291 | l -= old; | |
3292 | /* Get the added string: */ | |
3293 | last_str = newSVpvn(s + old, l); | |
3294 | if (UTF) | |
3295 | SvUTF8_on(last_str); | |
3296 | if (deltanext == 0 && pos_before == b) { | |
3297 | /* What was added is a constant string */ | |
3298 | if (mincount > 1) { | |
3299 | SvGROW(last_str, (mincount * l) + 1); | |
3300 | repeatcpy(SvPVX(last_str) + l, | |
3301 | SvPVX_const(last_str), l, mincount - 1); | |
3302 | SvCUR_set(last_str, SvCUR(last_str) * mincount); | |
3303 | /* Add additional parts. */ | |
3304 | SvCUR_set(data->last_found, | |
3305 | SvCUR(data->last_found) - l); | |
3306 | sv_catsv(data->last_found, last_str); | |
3307 | { | |
3308 | SV * sv = data->last_found; | |
3309 | MAGIC *mg = | |
3310 | SvUTF8(sv) && SvMAGICAL(sv) ? | |
3311 | mg_find(sv, PERL_MAGIC_utf8) : NULL; | |
3312 | if (mg && mg->mg_len >= 0) | |
bd94e887 | 3313 | mg->mg_len += CHR_SVLEN(last_str) - l; |
b515a41d | 3314 | } |
8aa23a47 | 3315 | data->last_end += l * (mincount - 1); |
b515a41d | 3316 | } |
8aa23a47 YO |
3317 | } else { |
3318 | /* start offset must point into the last copy */ | |
3319 | data->last_start_min += minnext * (mincount - 1); | |
3320 | data->last_start_max += is_inf ? I32_MAX | |
3321 | : (maxcount - 1) * (minnext + data->pos_delta); | |
3322 | } | |
c277df42 | 3323 | } |
8aa23a47 YO |
3324 | /* It is counted once already... */ |
3325 | data->pos_min += minnext * (mincount - counted); | |
3326 | data->pos_delta += - counted * deltanext + | |
3327 | (minnext + deltanext) * maxcount - minnext * mincount; | |
3328 | if (mincount != maxcount) { | |
3329 | /* Cannot extend fixed substrings found inside | |
3330 | the group. */ | |
304ee84b | 3331 | SCAN_COMMIT(pRExC_state,data,minlenp); |
8aa23a47 YO |
3332 | if (mincount && last_str) { |
3333 | SV * const sv = data->last_found; | |
3334 | MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ? | |
3335 | mg_find(sv, PERL_MAGIC_utf8) : NULL; | |
3336 | ||
3337 | if (mg) | |
3338 | mg->mg_len = -1; | |
3339 | sv_setsv(sv, last_str); | |
3340 | data->last_end = data->pos_min; | |
3341 | data->last_start_min = | |
3342 | data->pos_min - CHR_SVLEN(last_str); | |
3343 | data->last_start_max = is_inf | |
3344 | ? I32_MAX | |
3345 | : data->pos_min + data->pos_delta | |
3346 | - CHR_SVLEN(last_str); | |
3347 | } | |
3348 | data->longest = &(data->longest_float); | |
3349 | } | |
3350 | SvREFCNT_dec(last_str); | |
c277df42 | 3351 | } |
8aa23a47 YO |
3352 | if (data && (fl & SF_HAS_EVAL)) |
3353 | data->flags |= SF_HAS_EVAL; | |
3354 | optimize_curly_tail: | |
3355 | if (OP(oscan) != CURLYX) { | |
3356 | while (PL_regkind[OP(next = regnext(oscan))] == NOTHING | |
3357 | && NEXT_OFF(next)) | |
3358 | NEXT_OFF(oscan) += NEXT_OFF(next); | |
3359 | } | |
3360 | continue; | |
3361 | default: /* REF and CLUMP only? */ | |
3362 | if (flags & SCF_DO_SUBSTR) { | |
304ee84b | 3363 | SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */ |
8aa23a47 YO |
3364 | data->longest = &(data->longest_float); |
3365 | } | |
3366 | is_inf = is_inf_internal = 1; | |
3367 | if (flags & SCF_DO_STCLASS_OR) | |
3368 | cl_anything(pRExC_state, data->start_class); | |
3369 | flags &= ~SCF_DO_STCLASS; | |
3370 | break; | |
c277df42 | 3371 | } |
8aa23a47 | 3372 | } |
e1d1eefb YO |
3373 | else if (OP(scan) == LNBREAK) { |
3374 | if (flags & SCF_DO_STCLASS) { | |
3375 | int value = 0; | |
3376 | data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */ | |
3377 | if (flags & SCF_DO_STCLASS_AND) { | |
3378 | for (value = 0; value < 256; value++) | |
e64b1bd1 | 3379 | if (!is_VERTWS_cp(value)) |
e1d1eefb YO |
3380 | ANYOF_BITMAP_CLEAR(data->start_class, value); |
3381 | } | |
3382 | else { | |
3383 | for (value = 0; value < 256; value++) | |
e64b1bd1 | 3384 | if (is_VERTWS_cp(value)) |
e1d1eefb YO |
3385 | ANYOF_BITMAP_SET(data->start_class, value); |
3386 | } | |
3387 | if (flags & SCF_DO_STCLASS_OR) | |
3388 | cl_and(data->start_class, and_withp); | |
3389 | flags &= ~SCF_DO_STCLASS; | |
3390 | } | |
3391 | min += 1; | |
f9a79580 | 3392 | delta += 1; |
e1d1eefb YO |
3393 | if (flags & SCF_DO_SUBSTR) { |
3394 | SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */ | |
3395 | data->pos_min += 1; | |
f9a79580 | 3396 | data->pos_delta += 1; |
e1d1eefb YO |
3397 | data->longest = &(data->longest_float); |
3398 | } | |
3399 | ||
3400 | } | |
f9a79580 RGS |
3401 | else if (OP(scan) == FOLDCHAR) { |
3402 | int d = ARG(scan)==0xDF ? 1 : 2; | |
3403 | flags &= ~SCF_DO_STCLASS; | |
3404 | min += 1; | |
3405 | delta += d; | |
3406 | if (flags & SCF_DO_SUBSTR) { | |
3407 | SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */ | |
3408 | data->pos_min += 1; | |
3409 | data->pos_delta += d; | |
3410 | data->longest = &(data->longest_float); | |
3411 | } | |
3412 | } | |
8aa23a47 YO |
3413 | else if (strchr((const char*)PL_simple,OP(scan))) { |
3414 | int value = 0; | |
653099ff | 3415 | |
8aa23a47 | 3416 | if (flags & SCF_DO_SUBSTR) { |
304ee84b | 3417 | SCAN_COMMIT(pRExC_state,data,minlenp); |
8aa23a47 YO |
3418 | data->pos_min++; |
3419 | } | |
3420 | min++; | |
3421 | if (flags & SCF_DO_STCLASS) { | |
3422 | data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */ | |
b515a41d | 3423 | |
8aa23a47 YO |
3424 | /* Some of the logic below assumes that switching |
3425 | locale on will only add false positives. */ | |
3426 | switch (PL_regkind[OP(scan)]) { | |
3427 | case SANY: | |
3428 | default: | |
3429 | do_default: | |
3430 | /* Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d", OP(scan)); */ | |
3431 | if (flags & SCF_DO_STCLASS_OR) /* Allow everything */ | |
3432 | cl_anything(pRExC_state, data->start_class); | |
3433 | break; | |
3434 | case REG_ANY: | |
3435 | if (OP(scan) == SANY) | |
3436 | goto do_default; | |
3437 | if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */ | |
3438 | value = (ANYOF_BITMAP_TEST(data->start_class,'\n') | |
3439 | || (data->start_class->flags & ANYOF_CLASS)); | |
3440 | cl_anything(pRExC_state, data->start_class); | |
653099ff | 3441 | } |
8aa23a47 YO |
3442 | if (flags & SCF_DO_STCLASS_AND || !value) |
3443 | ANYOF_BITMAP_CLEAR(data->start_class,'\n'); | |
3444 | break; | |
3445 | case ANYOF: | |
3446 | if (flags & SCF_DO_STCLASS_AND) | |
3447 | cl_and(data->start_class, | |
3448 | (struct regnode_charclass_class*)scan); | |
653099ff | 3449 | else |
8aa23a47 YO |
3450 | cl_or(pRExC_state, data->start_class, |
3451 | (struct regnode_charclass_class*)scan); | |
3452 | break; | |
3453 | case ALNUM: | |
3454 | if (flags & SCF_DO_STCLASS_AND) { | |
3455 | if (!(data->start_class->flags & ANYOF_LOCALE)) { | |
3456 | ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM); | |
3457 | for (value = 0; value < 256; value++) | |
3458 | if (!isALNUM(value)) | |
3459 | ANYOF_BITMAP_CLEAR(data->start_class, value); | |
3460 | } | |
653099ff | 3461 | } |
8aa23a47 YO |
3462 | else { |
3463 | if (data->start_class->flags & ANYOF_LOCALE) | |
3464 | ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM); | |
3465 | else { | |
3466 | for (value = 0; value < 256; value++) | |
3467 | if (isALNUM(value)) | |
3468 | ANYOF_BITMAP_SET(data->start_class, value); | |
653099ff | 3469 | } |
8aa23a47 YO |
3470 | } |
3471 | break; | |
3472 | case ALNUML: | |
3473 | if (flags & SCF_DO_STCLASS_AND) { | |
3474 | if (data->start_class->flags & ANYOF_LOCALE) | |
3475 | ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM); | |
3476 | } | |
3477 | else { | |
3478 | ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM); | |
3479 | data->start_class->flags |= ANYOF_LOCALE; | |
3480 | } | |
3481 | break; | |
3482 | case NALNUM: | |
3483 | if (flags & SCF_DO_STCLASS_AND) { | |
3484 | if (!(data->start_class->flags & ANYOF_LOCALE)) { | |
3485 | ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM); | |
3486 | for (value = 0; value < 256; value++) | |
3487 | if (isALNUM(value)) | |
3488 | ANYOF_BITMAP_CLEAR(data->start_class, value); | |
653099ff GS |
3489 | } |
3490 | } | |
8aa23a47 YO |
3491 | else { |
3492 | if (data->start_class->flags & ANYOF_LOCALE) | |
3493 | ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM); | |
3494 | else { | |
3495 | for (value = 0; value < 256; value++) | |
3496 | if (!isALNUM(value)) | |
3497 | ANYOF_BITMAP_SET(data->start_class, value); | |
3498 | } | |
653099ff | 3499 | } |
8aa23a47 YO |
3500 | break; |
3501 | case NALNUML: | |
3502 | if (flags & SCF_DO_STCLASS_AND) { | |
3503 | if (data->start_class->flags & ANYOF_LOCALE) | |
3504 | ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM); | |
653099ff | 3505 | } |
8aa23a47 YO |
3506 | else { |
3507 | data->start_class->flags |= ANYOF_LOCALE; | |
3508 | ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM); | |
3509 | } | |
3510 | break; | |
3511 | case SPACE: | |
3512 | if (flags & SCF_DO_STCLASS_AND) { | |
3513 | if (!(data->start_class->flags & ANYOF_LOCALE)) { | |
3514 | ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE); | |
3515 | for (value = 0; value < 256; value++) | |
3516 | if (!isSPACE(value)) | |
3517 | ANYOF_BITMAP_CLEAR(data->start_class, value); | |
653099ff GS |
3518 | } |
3519 | } | |
8aa23a47 YO |
3520 | else { |
3521 | if (data->start_class->flags & ANYOF_LOCALE) | |
3522 | ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE); | |
3523 | else { | |
3524 | for (value = 0; value < 256; value++) | |
3525 | if (isSPACE(value)) | |
3526 | ANYOF_BITMAP_SET(data->start_class, value); | |
3527 | } | |
653099ff | 3528 | } |
8aa23a47 YO |
3529 | break; |
3530 | case SPACEL: | |
3531 | if (flags & SCF_DO_STCLASS_AND) { | |
3532 | if (data->start_class->flags & ANYOF_LOCALE) | |
3533 | ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE); | |
3534 | } | |
3535 | else { | |
3536 | data->start_class->flags |= ANYOF_LOCALE; | |
3537 | ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE); | |
3538 | } | |
3539 | break; | |
3540 | case NSPACE: | |
3541 | if (flags & SCF_DO_STCLASS_AND) { | |
3542 | if (!(data->start_class->flags & ANYOF_LOCALE)) { | |
3543 | ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE); | |
3544 | for (value = 0; value < 256; value++) | |
3545 | if (isSPACE(value)) | |
3546 | ANYOF_BITMAP_CLEAR(data->start_class, value); | |
653099ff | 3547 | } |
8aa23a47 YO |
3548 | } |
3549 | else { | |
3550 | if (data->start_class->flags & ANYOF_LOCALE) | |
3551 | ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE); | |
3552 | else { | |
3553 | for (value = 0; value < 256; value++) | |
3554 | if (!isSPACE(value)) | |
3555 | ANYOF_BITMAP_SET(data->start_class, value); | |
653099ff GS |
3556 | } |
3557 | } | |
8aa23a47 YO |
3558 | break; |
3559 | case NSPACEL: | |
3560 | if (flags & SCF_DO_STCLASS_AND) { | |
3561 | if (data->start_class->flags & ANYOF_LOCALE) { | |
3562 | ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE); | |
3563 | for (value = 0; value < 256; value++) | |
3564 | if (!isSPACE(value)) | |
3565 | ANYOF_BITMAP_CLEAR(data->start_class, value); | |
3566 | } | |
653099ff | 3567 | } |
8aa23a47 YO |
3568 | else { |
3569 | data->start_class->flags |= ANYOF_LOCALE; | |
3570 | ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE); | |
3571 | } | |
3572 | break; | |
3573 | case DIGIT: | |
3574 | if (flags & SCF_DO_STCLASS_AND) { | |
3575 | ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT); | |
3576 | for (value = 0; value < 256; value++) | |
3577 | if (!isDIGIT(value)) | |
3578 | ANYOF_BITMAP_CLEAR(data->start_class, value); | |
3579 | } | |
3580 | else { | |
3581 | if (data->start_class->flags & ANYOF_LOCALE) | |
3582 | ANYOF_CLASS_SET(data->start_class,ANYOF_DIGIT); | |
3583 | else { | |
3584 | for (value = 0; value < 256; value++) | |
3585 | if (isDIGIT(value)) | |
3586 | ANYOF_BITMAP_SET(data->start_class, value); | |
3587 | } | |
3588 | } | |
3589 | break; | |
3590 | case NDIGIT: | |
3591 | if (flags & SCF_DO_STCLASS_AND) { | |
3592 | ANYOF_CLASS_CLEAR(data->start_class,ANYOF_DIGIT); | |
3593 | for (value = 0; value < 256; value++) | |
3594 | if (isDIGIT(value)) | |
3595 | ANYOF_BITMAP_CLEAR(data->start_class, value); | |
3596 | } | |
3597 | else { | |
3598 | if (data->start_class->flags & ANYOF_LOCALE) | |
3599 | ANYOF_CLASS_SET(data->start_class,ANYOF_NDIGIT); | |
3600 | else { | |
3601 | for (value = 0; value < 256; value++) | |
3602 | if (!isDIGIT(value)) | |
3603 | ANYOF_BITMAP_SET(data->start_class, value); | |
653099ff GS |
3604 | } |
3605 | } | |
8aa23a47 | 3606 | break; |
e1d1eefb YO |
3607 | CASE_SYNST_FNC(VERTWS); |
3608 | CASE_SYNST_FNC(HORIZWS); | |
3609 | ||
8aa23a47 YO |
3610 | } |
3611 | if (flags & SCF_DO_STCLASS_OR) | |
3612 | cl_and(data->start_class, and_withp); | |
3613 | flags &= ~SCF_DO_STCLASS; | |
3614 | } | |
3615 | } | |
3616 | else if (PL_regkind[OP(scan)] == EOL && flags & SCF_DO_SUBSTR) { | |
3617 | data->flags |= (OP(scan) == MEOL | |
3618 | ? SF_BEFORE_MEOL | |
3619 | : SF_BEFORE_SEOL); | |
3620 | } | |
3621 | else if ( PL_regkind[OP(scan)] == BRANCHJ | |
3622 | /* Lookbehind, or need to calculate parens/evals/stclass: */ | |
3623 | && (scan->flags || data || (flags & SCF_DO_STCLASS)) | |
3624 | && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) { | |
3625 | if ( !PERL_ENABLE_POSITIVE_ASSERTION_STUDY | |
3626 | || OP(scan) == UNLESSM ) | |
3627 | { | |
3628 | /* Negative Lookahead/lookbehind | |
3629 | In this case we can't do fixed string optimisation. | |
3630 | */ | |
1de06328 | 3631 | |
8aa23a47 YO |
3632 | I32 deltanext, minnext, fake = 0; |
3633 | regnode *nscan; | |
3634 | struct regnode_charclass_class intrnl; | |
3635 | int f = 0; | |
1de06328 | 3636 | |
8aa23a47 YO |
3637 | data_fake.flags = 0; |
3638 | if (data) { | |
3639 | data_fake.whilem_c = data->whilem_c; | |
3640 | data_fake.last_closep = data->last_closep; | |
c277df42 | 3641 | } |
8aa23a47 YO |
3642 | else |
3643 | data_fake.last_closep = &fake; | |
58e23c8d | 3644 | data_fake.pos_delta = delta; |
8aa23a47 YO |
3645 | if ( flags & SCF_DO_STCLASS && !scan->flags |
3646 | && OP(scan) == IFMATCH ) { /* Lookahead */ | |
3647 | cl_init(pRExC_state, &intrnl); | |
3648 | data_fake.start_class = &intrnl; | |
3649 | f |= SCF_DO_STCLASS_AND; | |
3650 | } | |
3651 | if (flags & SCF_WHILEM_VISITED_POS) | |
3652 | f |= SCF_WHILEM_VISITED_POS; | |
3653 | next = regnext(scan); | |
3654 | nscan = NEXTOPER(NEXTOPER(scan)); | |
3655 | minnext = study_chunk(pRExC_state, &nscan, minlenp, &deltanext, | |
3656 | last, &data_fake, stopparen, recursed, NULL, f, depth+1); | |
3657 | if (scan->flags) { | |
3658 | if (deltanext) { | |
58e23c8d | 3659 | FAIL("Variable length lookbehind not implemented"); |
8aa23a47 YO |
3660 | } |
3661 | else if (minnext > (I32)U8_MAX) { | |
58e23c8d | 3662 | FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX); |
8aa23a47 YO |
3663 | } |
3664 | scan->flags = (U8)minnext; | |
3665 | } | |
3666 | if (data) { | |
3667 | if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR)) | |
3668 | pars++; | |
3669 | if (data_fake.flags & SF_HAS_EVAL) | |
3670 | data->flags |= SF_HAS_EVAL; | |
3671 | data->whilem_c = data_fake.whilem_c; | |
3672 | } | |
3673 | if (f & SCF_DO_STCLASS_AND) { | |
3674 | const int was = (data->start_class->flags & ANYOF_EOS); | |
3675 | ||
3676 | cl_and(data->start_class, &intrnl); | |
3677 | if (was) | |
3678 | data->start_class->flags |= ANYOF_EOS; | |
3679 | } | |
cb434fcc | 3680 | } |
8aa23a47 YO |
3681 | #if PERL_ENABLE_POSITIVE_ASSERTION_STUDY |
3682 | else { | |
3683 | /* Positive Lookahead/lookbehind | |
3684 | In this case we can do fixed string optimisation, | |
3685 | but we must be careful about it. Note in the case of | |
3686 | lookbehind the positions will be offset by the minimum | |
3687 | length of the pattern, something we won't know about | |
3688 | until after the recurse. | |
3689 | */ | |
3690 | I32 deltanext, fake = 0; | |
3691 | regnode *nscan; | |
3692 | struct regnode_charclass_class intrnl; | |
3693 | int f = 0; | |
3694 | /* We use SAVEFREEPV so that when the full compile | |
3695 | is finished perl will clean up the allocated | |
3696 | minlens when its all done. This was we don't | |
3697 | have to worry about freeing them when we know | |
3698 | they wont be used, which would be a pain. | |
3699 | */ | |
3700 | I32 *minnextp; | |
3701 | Newx( minnextp, 1, I32 ); | |
3702 | SAVEFREEPV(minnextp); | |
3703 | ||
3704 | if (data) { | |
3705 | StructCopy(data, &data_fake, scan_data_t); | |
3706 | if ((flags & SCF_DO_SUBSTR) && data->last_found) { | |
3707 | f |= SCF_DO_SUBSTR; | |
3708 | if (scan->flags) | |
304ee84b | 3709 | SCAN_COMMIT(pRExC_state, &data_fake,minlenp); |
8aa23a47 YO |
3710 | data_fake.last_found=newSVsv(data->last_found); |
3711 | } | |
3712 | } | |
3713 | else | |
3714 | data_fake.last_closep = &fake; | |
3715 | data_fake.flags = 0; | |
58e23c8d | 3716 | data_fake.pos_delta = delta; |
8aa23a47 YO |
3717 | if (is_inf) |
3718 | data_fake.flags |= SF_IS_INF; | |
3719 | if ( flags & SCF_DO_STCLASS && !scan->flags | |
3720 | && OP(scan) == IFMATCH ) { /* Lookahead */ | |
3721 | cl_init(pRExC_state, &intrnl); | |
3722 | data_fake.start_class = &intrnl; | |
3723 | f |= SCF_DO_STCLASS_AND; | |
3724 | } | |
3725 | if (flags & SCF_WHILEM_VISITED_POS) | |
3726 | f |= SCF_WHILEM_VISITED_POS; | |
3727 | next = regnext(scan); | |
3728 | nscan = NEXTOPER(NEXTOPER(scan)); | |
3729 | ||
3730 | *minnextp = study_chunk(pRExC_state, &nscan, minnextp, &deltanext, | |
3731 | last, &data_fake, stopparen, recursed, NULL, f,depth+1); | |
3732 | if (scan->flags) { | |
3733 | if (deltanext) { | |
58e23c8d | 3734 | FAIL("Variable length lookbehind not implemented"); |
8aa23a47 YO |
3735 | } |
3736 | else if (*minnextp > (I32)U8_MAX) { | |
58e23c8d | 3737 | FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX); |
8aa23a47 YO |
3738 | } |
3739 | scan->flags = (U8)*minnextp; | |
3740 | } | |
3741 | ||
3742 | *minnextp += min; | |
3743 | ||
3744 | if (f & SCF_DO_STCLASS_AND) { | |
3745 | const int was = (data->start_class->flags & ANYOF_EOS); | |
3746 | ||
3747 | cl_and(data->start_class, &intrnl); | |
3748 | if (was) | |
3749 | data->start_class->flags |= ANYOF_EOS; | |
3750 | } | |
3751 | if (data) { | |
3752 | if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR)) | |
3753 | pars++; | |
3754 | if (data_fake.flags & SF_HAS_EVAL) | |
3755 | data->flags |= SF_HAS_EVAL; | |
3756 | data->whilem_c = data_fake.whilem_c; | |
3757 | if ((flags & SCF_DO_SUBSTR) && data_fake.last_found) { | |
3758 | if (RExC_rx->minlen<*minnextp) | |
3759 | RExC_rx->minlen=*minnextp; | |
304ee84b | 3760 | SCAN_COMMIT(pRExC_state, &data_fake, minnextp); |
8aa23a47 YO |
3761 | SvREFCNT_dec(data_fake.last_found); |
3762 | ||
3763 | if ( data_fake.minlen_fixed != minlenp ) | |
3764 | { | |
3765 | data->offset_fixed= data_fake.offset_fixed; | |
3766 | data->minlen_fixed= data_fake.minlen_fixed; | |
3767 | data->lookbehind_fixed+= scan->flags; | |
3768 | } | |
3769 | if ( data_fake.minlen_float != minlenp ) | |
3770 | { | |
3771 | data->minlen_float= data_fake.minlen_float; | |
3772 | data->offset_float_min=data_fake.offset_float_min; | |
3773 | data->offset_float_max=data_fake.offset_float_max; | |
3774 | data->lookbehind_float+= scan->flags; | |
3775 | } | |
3776 | } | |
3777 | } | |
3778 | ||
3779 | ||
40d049e4 | 3780 | } |
8aa23a47 YO |
3781 | #endif |
3782 | } | |
3783 | else if (OP(scan) == OPEN) { | |
3784 | if (stopparen != (I32)ARG(scan)) | |
3785 | pars++; | |
3786 | } | |
3787 | else if (OP(scan) == CLOSE) { | |
3788 | if (stopparen == (I32)ARG(scan)) { | |
3789 | break; | |
3790 | } | |
3791 | if ((I32)ARG(scan) == is_par) { | |
3792 | next = regnext(scan); | |
b515a41d | 3793 | |
8aa23a47 YO |
3794 | if ( next && (OP(next) != WHILEM) && next < last) |
3795 | is_par = 0; /* Disable optimization */ | |
40d049e4 | 3796 | } |
8aa23a47 YO |
3797 | if (data) |
3798 | *(data->last_closep) = ARG(scan); | |
3799 | } | |
3800 | else if (OP(scan) == EVAL) { | |
c277df42 IZ |
3801 | if (data) |
3802 | data->flags |= SF_HAS_EVAL; | |
8aa23a47 YO |
3803 | } |
3804 | else if ( PL_regkind[OP(scan)] == ENDLIKE ) { | |
3805 | if (flags & SCF_DO_SUBSTR) { | |
304ee84b | 3806 | SCAN_COMMIT(pRExC_state,data,minlenp); |
8aa23a47 | 3807 | flags &= ~SCF_DO_SUBSTR; |
40d049e4 | 3808 | } |
8aa23a47 YO |
3809 | if (data && OP(scan)==ACCEPT) { |
3810 | data->flags |= SCF_SEEN_ACCEPT; | |
3811 | if (stopmin > min) | |
3812 | stopmin = min; | |
e2e6a0f1 | 3813 | } |
8aa23a47 YO |
3814 | } |
3815 | else if (OP(scan) == LOGICAL && scan->flags == 2) /* Embedded follows */ | |
3816 | { | |
0f5d15d6 | 3817 | if (flags & SCF_DO_SUBSTR) { |
304ee84b | 3818 | SCAN_COMMIT(pRExC_state,data,minlenp); |
0f5d15d6 IZ |
3819 | data->longest = &(data->longest_float); |
3820 | } | |
3821 | is_inf = is_inf_internal = 1; | |
653099ff | 3822 | if (flags & SCF_DO_STCLASS_OR) /* Allow everything */ |
830247a4 | 3823 | cl_anything(pRExC_state, data->start_class); |
96776eda | 3824 | flags &= ~SCF_DO_STCLASS; |
8aa23a47 | 3825 | } |
58e23c8d | 3826 | else if (OP(scan) == GPOS) { |
bbe252da | 3827 | if (!(RExC_rx->extflags & RXf_GPOS_FLOAT) && |
58e23c8d YO |
3828 | !(delta || is_inf || (data && data->pos_delta))) |
3829 | { | |
bbe252da YO |
3830 | if (!(RExC_rx->extflags & RXf_ANCH) && (flags & SCF_DO_SUBSTR)) |
3831 | RExC_rx->extflags |= RXf_ANCH_GPOS; | |
58e23c8d YO |
3832 | if (RExC_rx->gofs < (U32)min) |
3833 | RExC_rx->gofs = min; | |
3834 | } else { | |
bbe252da | 3835 | RExC_rx->extflags |= RXf_GPOS_FLOAT; |
58e23c8d YO |
3836 | RExC_rx->gofs = 0; |
3837 | } | |
3838 | } | |
786e8c11 | 3839 | #ifdef TRIE_STUDY_OPT |
40d049e4 | 3840 | #ifdef FULL_TRIE_STUDY |
8aa23a47 YO |
3841 | else if (PL_regkind[OP(scan)] == TRIE) { |
3842 | /* NOTE - There is similar code to this block above for handling | |
3843 | BRANCH nodes on the initial study. If you change stuff here | |
3844 | check there too. */ | |
3845 | regnode *trie_node= scan; | |
3846 | regnode *tail= regnext(scan); | |
f8fc2ecf | 3847 | reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ]; |
8aa23a47 YO |
3848 | I32 max1 = 0, min1 = I32_MAX; |
3849 | struct regnode_charclass_class accum; | |
3850 | ||
3851 | if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */ | |
304ee84b | 3852 | SCAN_COMMIT(pRExC_state, data,minlenp); /* Cannot merge strings after this. */ |
8aa23a47 YO |
3853 | if (flags & SCF_DO_STCLASS) |
3854 | cl_init_zero(pRExC_state, &accum); | |
3855 | ||
3856 | if (!trie->jump) { | |
3857 | min1= trie->minlen; | |
3858 | max1= trie->maxlen; | |
3859 | } else { | |
3860 | const regnode *nextbranch= NULL; | |
3861 | U32 word; | |
3862 | ||
3863 | for ( word=1 ; word <= trie->wordcount ; word++) | |
3864 | { | |
3865 | I32 deltanext=0, minnext=0, f = 0, fake; | |
3866 | struct regnode_charclass_class this_class; | |
3867 | ||
3868 | data_fake.flags = 0; | |
3869 | if (data) { | |
3870 | data_fake.whilem_c = data->whilem_c; | |
3871 | data_fake.last_closep = data->last_closep; | |
3872 | } | |
3873 | else | |
3874 | data_fake.last_closep = &fake; | |
58e23c8d | 3875 | data_fake.pos_delta = delta; |
8aa23a47 YO |
3876 | if (flags & SCF_DO_STCLASS) { |
3877 | cl_init(pRExC_state, &this_class); | |
3878 | data_fake.start_class = &this_class; | |
3879 | f = SCF_DO_STCLASS_AND; | |
3880 | } | |
3881 | if (flags & SCF_WHILEM_VISITED_POS) | |
3882 | f |= SCF_WHILEM_VISITED_POS; | |
3883 | ||
3884 | if (trie->jump[word]) { | |
3885 | if (!nextbranch) | |
3886 | nextbranch = trie_node + trie->jump[0]; | |
3887 | scan= trie_node + trie->jump[word]; | |
3888 | /* We go from the jump point to the branch that follows | |
3889 | it. Note this means we need the vestigal unused branches | |
3890 | even though they arent otherwise used. | |
3891 | */ | |
3892 | minnext = study_chunk(pRExC_state, &scan, minlenp, | |
3893 | &deltanext, (regnode *)nextbranch, &data_fake, | |
3894 | stopparen, recursed, NULL, f,depth+1); | |
3895 | } | |
3896 | if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH) | |
3897 | nextbranch= regnext((regnode*)nextbranch); | |
3898 | ||
3899 | if (min1 > (I32)(minnext + trie->minlen)) | |
3900 | min1 = minnext + trie->minlen; | |
3901 | if (max1 < (I32)(minnext + deltanext + trie->maxlen)) | |
3902 | max1 = minnext + deltanext + trie->maxlen; | |
3903 | if (deltanext == I32_MAX) | |
3904 | is_inf = is_inf_internal = 1; | |
3905 | ||
3906 | if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR)) | |
3907 | pars++; | |
3908 | if (data_fake.flags & SCF_SEEN_ACCEPT) { | |
3909 | if ( stopmin > min + min1) | |
3910 | stopmin = min + min1; | |
3911 | flags &= ~SCF_DO_SUBSTR; | |
3912 | if (data) | |
3913 | data->flags |= SCF_SEEN_ACCEPT; | |
3914 | } | |
3915 | if (data) { | |
3916 | if (data_fake.flags & SF_HAS_EVAL) | |
3917 | data->flags |= SF_HAS_EVAL; | |
3918 | data->whilem_c = data_fake.whilem_c; | |
3919 | } | |
3920 | if (flags & SCF_DO_STCLASS) | |
3921 | cl_or(pRExC_state, &accum, &this_class); | |
3922 | } | |
3923 | } | |
3924 | if (flags & SCF_DO_SUBSTR) { | |
3925 | data->pos_min += min1; | |
3926 | data->pos_delta += max1 - min1; | |
3927 | if (max1 != min1 || is_inf) | |
3928 | data->longest = &(data->longest_float); | |
3929 | } | |
3930 | min += min1; | |
3931 | delta += max1 - min1; | |
3932 | if (flags & SCF_DO_STCLASS_OR) { | |
3933 | cl_or(pRExC_state, data->start_class, &accum); | |
3934 | if (min1) { | |
3935 | cl_and(data->start_class, and_withp); | |
3936 | flags &= ~SCF_DO_STCLASS; | |
3937 | } | |
3938 | } | |
3939 | else if (flags & SCF_DO_STCLASS_AND) { | |
3940 | if (min1) { | |
3941 | cl_and(data->start_class, &accum); | |
3942 | flags &= ~SCF_DO_STCLASS; | |
3943 | } | |
3944 | else { | |
3945 | /* Switch to OR mode: cache the old value of | |
3946 | * data->start_class */ | |
3947 | INIT_AND_WITHP; | |
3948 | StructCopy(data->start_class, and_withp, | |
3949 | struct regnode_charclass_class); | |
3950 | flags &= ~SCF_DO_STCLASS_AND; | |
3951 | StructCopy(&accum, data->start_class, | |
3952 | struct regnode_charclass_class); | |
3953 | flags |= SCF_DO_STCLASS_OR; | |
3954 | data->start_class->flags |= ANYOF_EOS; | |
3955 | } | |
3956 | } | |
3957 | scan= tail; | |
3958 | continue; | |
3959 | } | |
786e8c11 | 3960 | #else |
8aa23a47 | 3961 | else if (PL_regkind[OP(scan)] == TRIE) { |
f8fc2ecf | 3962 | reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ]; |
8aa23a47 YO |
3963 | U8*bang=NULL; |
3964 | ||
3965 | min += trie->minlen; | |
3966 | delta += (trie->maxlen - trie->minlen); | |
3967 | flags &= ~SCF_DO_STCLASS; /* xxx */ | |
3968 | if (flags & SCF_DO_SUBSTR) { | |
304ee84b | 3969 | SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */ |
8aa23a47 YO |
3970 | data->pos_min += trie->minlen; |
3971 | data->pos_delta += (trie->maxlen - trie->minlen); | |
3972 | if (trie->maxlen != trie->minlen) | |
3973 | data->longest = &(data->longest_float); | |
3974 | } | |
3975 | if (trie->jump) /* no more substrings -- for now /grr*/ | |
3976 | flags &= ~SCF_DO_SUBSTR; | |
b515a41d | 3977 | } |
8aa23a47 YO |
3978 | #endif /* old or new */ |
3979 | #endif /* TRIE_STUDY_OPT */ | |
e1d1eefb | 3980 | |
8aa23a47 YO |
3981 | /* Else: zero-length, ignore. */ |
3982 | scan = regnext(scan); | |
3983 | } | |
3984 | if (frame) { | |
3985 | last = frame->last; | |
3986 | scan = frame->next; | |
3987 | stopparen = frame->stop; | |
3988 | frame = frame->prev; | |
3989 | goto fake_study_recurse; | |
c277df42 IZ |
3990 | } |
3991 | ||
3992 | finish: | |
8aa23a47 | 3993 | assert(!frame); |
304ee84b | 3994 | DEBUG_STUDYDATA("pre-fin:",data,depth); |
8aa23a47 | 3995 | |
c277df42 | 3996 | *scanp = scan; |
aca2d497 | 3997 | *deltap = is_inf_internal ? I32_MAX : delta; |
b81d288d | 3998 | if (flags & SCF_DO_SUBSTR && is_inf) |
c277df42 | 3999 | data->pos_delta = I32_MAX - data->pos_min; |
786e8c11 | 4000 | if (is_par > (I32)U8_MAX) |
c277df42 IZ |
4001 | is_par = 0; |
4002 | if (is_par && pars==1 && data) { | |
4003 | data->flags |= SF_IN_PAR; | |
4004 | data->flags &= ~SF_HAS_PAR; | |
a0ed51b3 LW |
4005 | } |
4006 | else if (pars && data) { | |
c277df42 IZ |
4007 | data->flags |= SF_HAS_PAR; |
4008 | data->flags &= ~SF_IN_PAR; | |
4009 | } | |
653099ff | 4010 | if (flags & SCF_DO_STCLASS_OR) |
40d049e4 | 4011 | cl_and(data->start_class, and_withp); |
786e8c11 YO |
4012 | if (flags & SCF_TRIE_RESTUDY) |
4013 | data->flags |= SCF_TRIE_RESTUDY; | |
1de06328 | 4014 | |
304ee84b | 4015 | DEBUG_STUDYDATA("post-fin:",data,depth); |
1de06328 | 4016 | |
e2e6a0f1 | 4017 | return min < stopmin ? min : stopmin; |
c277df42 IZ |
4018 | } |
4019 | ||
2eccd3b2 NC |
4020 | STATIC U32 |
4021 | S_add_data(RExC_state_t *pRExC_state, U32 n, const char *s) | |
c277df42 | 4022 | { |
4a4e7719 NC |
4023 | U32 count = RExC_rxi->data ? RExC_rxi->data->count : 0; |
4024 | ||
4025 | Renewc(RExC_rxi->data, | |
4026 | sizeof(*RExC_rxi->data) + sizeof(void*) * (count + n - 1), | |
4027 | char, struct reg_data); | |
4028 | if(count) | |
f8fc2ecf | 4029 | Renew(RExC_rxi->data->what, count + n, U8); |
4a4e7719 | 4030 | else |
f8fc2ecf | 4031 | Newx(RExC_rxi->data->what, n, U8); |
4a4e7719 NC |
4032 | RExC_rxi->data->count = count + n; |
4033 | Copy(s, RExC_rxi->data->what + count, n, U8); | |
4034 | return count; | |
c277df42 IZ |
4035 | } |
4036 | ||
f8149455 | 4037 | /*XXX: todo make this not included in a non debugging perl */ |
76234dfb | 4038 | #ifndef PERL_IN_XSUB_RE |
d88dccdf | 4039 | void |
864dbfa3 | 4040 | Perl_reginitcolors(pTHX) |
d88dccdf | 4041 | { |
97aff369 | 4042 | dVAR; |
1df70142 | 4043 | const char * const s = PerlEnv_getenv("PERL_RE_COLORS"); |
d88dccdf | 4044 | if (s) { |
1df70142 AL |
4045 | char *t = savepv(s); |
4046 | int i = 0; | |
4047 | PL_colors[0] = t; | |
d88dccdf | 4048 | while (++i < 6) { |
1df70142 AL |
4049 | t = strchr(t, '\t'); |
4050 | if (t) { | |
4051 | *t = '\0'; | |
4052 | PL_colors[i] = ++t; | |
d88dccdf IZ |
4053 | } |
4054 | else | |
1df70142 | 4055 | PL_colors[i] = t = (char *)""; |
d88dccdf IZ |
4056 | } |
4057 | } else { | |
1df70142 | 4058 | int i = 0; |
b81d288d | 4059 | while (i < 6) |
06b5626a | 4060 | PL_colors[i++] = (char *)""; |
d88dccdf IZ |
4061 | } |
4062 | PL_colorset = 1; | |
4063 | } | |
76234dfb | 4064 | #endif |
8615cb43 | 4065 | |
07be1b83 | 4066 | |
786e8c11 YO |
4067 | #ifdef TRIE_STUDY_OPT |
4068 | #define CHECK_RESTUDY_GOTO \ | |
4069 | if ( \ | |
4070 | (data.flags & SCF_TRIE_RESTUDY) \ | |
4071 | && ! restudied++ \ | |
4072 | ) goto reStudy | |
4073 | #else | |
4074 | #define CHECK_RESTUDY_GOTO | |
4075 | #endif | |
f9f4320a | 4076 | |
a687059c | 4077 | /* |
e50aee73 | 4078 | - pregcomp - compile a regular expression into internal code |
a687059c LW |
4079 | * |
4080 | * We can't allocate space until we know how big the compiled form will be, | |
4081 | * but we can't compile it (and thus know how big it is) until we've got a | |
4082 | * place to put the code. So we cheat: we compile it twice, once with code | |
4083 | * generation turned off and size counting turned on, and once "for real". | |
4084 | * This also means that we don't allocate space until we are sure that the | |
4085 | * thing really will compile successfully, and we never have to move the | |
4086 | * code and thus invalidate pointers into it. (Note that it has to be in | |
4087 | * one piece because free() must be able to free it all.) [NB: not true in perl] | |
4088 | * | |
4089 | * Beware that the optimization-preparation code in here knows about some | |
4090 | * of the structure of the compiled regexp. [I'll say.] | |
4091 | */ | |
b9b4dddf YO |
4092 | |
4093 | ||
4094 | ||
f9f4320a | 4095 | #ifndef PERL_IN_XSUB_RE |
f9f4320a YO |
4096 | #define RE_ENGINE_PTR &PL_core_reg_engine |
4097 | #else | |
f9f4320a YO |
4098 | extern const struct regexp_engine my_reg_engine; |
4099 | #define RE_ENGINE_PTR &my_reg_engine | |
4100 | #endif | |
6d5c990f RGS |
4101 | |
4102 | #ifndef PERL_IN_XSUB_RE | |
3ab4a224 AB |
4103 | REGEXP * |
4104 | Perl_pregcomp(pTHX_ const SV * const pattern, const U32 flags) | |
a687059c | 4105 | { |
97aff369 | 4106 | dVAR; |
6d5c990f | 4107 | HV * const table = GvHV(PL_hintgv); |
f9f4320a YO |
4108 | /* Dispatch a request to compile a regexp to correct |
4109 | regexp engine. */ | |
f9f4320a YO |
4110 | if (table) { |
4111 | SV **ptr= hv_fetchs(table, "regcomp", FALSE); | |
6d5c990f | 4112 | GET_RE_DEBUG_FLAGS_DECL; |
1e2e3d02 | 4113 | if (ptr && SvIOK(*ptr) && SvIV(*ptr)) { |
f9f4320a YO |
4114 | const regexp_engine *eng=INT2PTR(regexp_engine*,SvIV(*ptr)); |
4115 | DEBUG_COMPILE_r({ | |
8d8756e7 | 4116 | PerlIO_printf(Perl_debug_log, "Using engine %"UVxf"\n", |
f9f4320a YO |
4117 | SvIV(*ptr)); |
4118 | }); | |
3ab4a224 | 4119 | return CALLREGCOMP_ENG(eng, pattern, flags); |
f9f4320a | 4120 | } |
b9b4dddf | 4121 | } |
3ab4a224 | 4122 | return Perl_re_compile(aTHX_ pattern, flags); |
2a5d9b1d | 4123 | } |
6d5c990f | 4124 | #endif |
2a5d9b1d | 4125 | |
3ab4a224 AB |
4126 | REGEXP * |
4127 | Perl_re_compile(pTHX_ const SV * const pattern, const U32 pm_flags) | |
2a5d9b1d RGS |
4128 | { |
4129 | dVAR; | |
3ab4a224 | 4130 | register REGEXP *r; |
f8fc2ecf | 4131 | register regexp_internal *ri; |
3ab4a224 AB |
4132 | STRLEN plen; |
4133 | char* exp = SvPV((SV*)pattern, plen); | |
4134 | char* xend = exp + plen; | |
c277df42 | 4135 | regnode *scan; |
a0d0e21e | 4136 | I32 flags; |
a0d0e21e LW |
4137 | I32 minlen = 0; |
4138 | I32 sawplus = 0; | |
4139 | I32 sawopen = 0; | |
2c2d71f5 | 4140 | scan_data_t data; |
830247a4 | 4141 | RExC_state_t RExC_state; |
be8e71aa | 4142 | RExC_state_t * const pRExC_state = &RExC_state; |
07be1b83 YO |
4143 | #ifdef TRIE_STUDY_OPT |
4144 | int restudied= 0; | |
4145 | RExC_state_t copyRExC_state; | |
4146 | #endif | |
2a5d9b1d | 4147 | GET_RE_DEBUG_FLAGS_DECL; |
6d5c990f | 4148 | DEBUG_r(if (!PL_colorset) reginitcolors()); |
a0d0e21e | 4149 | |
c737faaf | 4150 | RExC_utf8 = RExC_orig_utf8 = pm_flags & RXf_UTF8; |
a0ed51b3 | 4151 | |
a3621e74 | 4152 | DEBUG_COMPILE_r({ |
ab3bbdeb YO |
4153 | SV *dsv= sv_newmortal(); |
4154 | RE_PV_QUOTED_DECL(s, RExC_utf8, | |
3ab4a224 | 4155 | dsv, exp, plen, 60); |
ab3bbdeb YO |
4156 | PerlIO_printf(Perl_debug_log, "%sCompiling REx%s %s\n", |
4157 | PL_colors[4],PL_colors[5],s); | |
a5961de5 | 4158 | }); |
02daf0ab YO |
4159 | |
4160 | redo_first_pass: | |
4161 | RExC_precomp = exp; | |
c737faaf | 4162 | RExC_flags = pm_flags; |
830247a4 | 4163 | RExC_sawback = 0; |
bbce6d69 | 4164 | |
830247a4 IZ |
4165 | RExC_seen = 0; |
4166 | RExC_seen_zerolen = *exp == '^' ? -1 : 0; | |
4167 | RExC_seen_evals = 0; | |
4168 | RExC_extralen = 0; | |
c277df42 | 4169 | |
bbce6d69 | 4170 | /* First pass: determine size, legality. */ |
830247a4 | 4171 | RExC_parse = exp; |
fac92740 | 4172 | RExC_start = exp; |
830247a4 IZ |
4173 | RExC_end = xend; |
4174 | RExC_naughty = 0; | |
4175 | RExC_npar = 1; | |
e2e6a0f1 | 4176 | RExC_nestroot = 0; |
830247a4 IZ |
4177 | RExC_size = 0L; |
4178 | RExC_emit = &PL_regdummy; | |
4179 | RExC_whilem_seen = 0; | |
fc8cd66c | 4180 | RExC_charnames = NULL; |
40d049e4 YO |
4181 | RExC_open_parens = NULL; |
4182 | RExC_close_parens = NULL; | |
4183 | RExC_opend = NULL; | |
81714fb9 | 4184 | RExC_paren_names = NULL; |
1f1031fe YO |
4185 | #ifdef DEBUGGING |
4186 | RExC_paren_name_list = NULL; | |
4187 | #endif | |
40d049e4 YO |
4188 | RExC_recurse = NULL; |
4189 | RExC_recurse_count = 0; | |
81714fb9 | 4190 | |
85ddcde9 JH |
4191 | #if 0 /* REGC() is (currently) a NOP at the first pass. |
4192 | * Clever compilers notice this and complain. --jhi */ | |
830247a4 | 4193 | REGC((U8)REG_MAGIC, (char*)RExC_emit); |
85ddcde9 | 4194 | #endif |
3dab1dad YO |
4195 | DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "Starting first pass (sizing)\n")); |
4196 | if (reg(pRExC_state, 0, &flags,1) == NULL) { | |
c445ea15 | 4197 | RExC_precomp = NULL; |
a0d0e21e LW |
4198 | return(NULL); |
4199 | } | |
02daf0ab | 4200 | if (RExC_utf8 && !RExC_orig_utf8) { |
38a44b82 | 4201 | /* It's possible to write a regexp in ascii that represents Unicode |
02daf0ab YO |
4202 | codepoints outside of the byte range, such as via \x{100}. If we |
4203 | detect such a sequence we have to convert the entire pattern to utf8 | |
4204 | and then recompile, as our sizing calculation will have been based | |
4205 | on 1 byte == 1 character, but we will need to use utf8 to encode | |
4206 | at least some part of the pattern, and therefore must convert the whole | |
4207 | thing. | |
4208 | XXX: somehow figure out how to make this less expensive... | |
4209 | -- dmq */ | |
3ab4a224 | 4210 | STRLEN len = plen; |
02daf0ab YO |
4211 | DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, |
4212 | "UTF8 mismatch! Converting to utf8 for resizing and compile\n")); | |
4213 | exp = (char*)Perl_bytes_to_utf8(aTHX_ (U8*)exp, &len); | |
4214 | xend = exp + len; | |
4215 | RExC_orig_utf8 = RExC_utf8; | |
4216 | SAVEFREEPV(exp); | |
4217 | goto redo_first_pass; | |
4218 | } | |
07be1b83 | 4219 | DEBUG_PARSE_r({ |
81714fb9 YO |
4220 | PerlIO_printf(Perl_debug_log, |
4221 | "Required size %"IVdf" nodes\n" | |
4222 | "Starting second pass (creation)\n", | |
4223 | (IV)RExC_size); | |
07be1b83 YO |
4224 | RExC_lastnum=0; |
4225 | RExC_lastparse=NULL; | |
4226 | }); | |
c277df42 IZ |
4227 | /* Small enough for pointer-storage convention? |
4228 | If extralen==0, this means that we will not need long jumps. */ | |
830247a4 IZ |
4229 | if (RExC_size >= 0x10000L && RExC_extralen) |
4230 | RExC_size += RExC_extralen; | |
c277df42 | 4231 | else |
830247a4 IZ |
4232 | RExC_extralen = 0; |
4233 | if (RExC_whilem_seen > 15) | |
4234 | RExC_whilem_seen = 15; | |
a0d0e21e | 4235 | |
f9f4320a YO |
4236 | /* Allocate space and zero-initialize. Note, the two step process |
4237 | of zeroing when in debug mode, thus anything assigned has to | |
4238 | happen after that */ | |
f8fc2ecf YO |
4239 | Newxz(r, 1, regexp); |
4240 | Newxc(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode), | |
4241 | char, regexp_internal); | |
4242 | if ( r == NULL || ri == NULL ) | |
b45f050a | 4243 | FAIL("Regexp out of space"); |
0f79a09d GS |
4244 | #ifdef DEBUGGING |
4245 | /* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */ | |
f8fc2ecf | 4246 | Zero(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode), char); |
58e23c8d | 4247 | #else |
f8fc2ecf YO |
4248 | /* bulk initialize base fields with 0. */ |
4249 | Zero(ri, sizeof(regexp_internal), char); | |
0f79a09d | 4250 | #endif |
58e23c8d YO |
4251 | |
4252 | /* non-zero initialization begins here */ | |
f8fc2ecf | 4253 | RXi_SET( r, ri ); |
f9f4320a | 4254 | r->engine= RE_ENGINE_PTR; |
c277df42 | 4255 | r->refcnt = 1; |
3ab4a224 | 4256 | r->prelen = plen; |
c737faaf | 4257 | r->extflags = pm_flags; |
bcdf7404 | 4258 | { |
f7819f85 | 4259 | bool has_p = ((r->extflags & RXf_PMf_KEEPCOPY) == RXf_PMf_KEEPCOPY); |
bcdf7404 YO |
4260 | bool has_minus = ((r->extflags & RXf_PMf_STD_PMMOD) != RXf_PMf_STD_PMMOD); |
4261 | bool has_runon = ((RExC_seen & REG_SEEN_RUN_ON_COMMENT)==REG_SEEN_RUN_ON_COMMENT); | |
4262 | U16 reganch = (U16)((r->extflags & RXf_PMf_STD_PMMOD) >> 12); | |
4263 | const char *fptr = STD_PAT_MODS; /*"msix"*/ | |
4264 | char *p; | |
f7819f85 | 4265 | r->wraplen = r->prelen + has_minus + has_p + has_runon |
bcdf7404 YO |
4266 | + (sizeof(STD_PAT_MODS) - 1) |
4267 | + (sizeof("(?:)") - 1); | |
4268 | ||
28d8d7f4 | 4269 | Newx(r->wrapped, r->wraplen + 1, char ); |
bcdf7404 YO |
4270 | p = r->wrapped; |
4271 | *p++='('; *p++='?'; | |
f7819f85 A |
4272 | if (has_p) |
4273 | *p++ = KEEPCOPY_PAT_MOD; /*'p'*/ | |
bcdf7404 YO |
4274 | { |
4275 | char *r = p + (sizeof(STD_PAT_MODS) - 1) + has_minus - 1; | |
4276 | char *colon = r + 1; | |
4277 | char ch; | |
4278 | ||
4279 | while((ch = *fptr++)) { | |
4280 | if(reganch & 1) | |
4281 | *p++ = ch; | |
4282 | else | |
4283 | *r-- = ch; | |
4284 | reganch >>= 1; | |
4285 | } | |
4286 | if(has_minus) { | |
4287 | *r = '-'; | |
4288 | p = colon; | |
4289 | } | |
4290 | } | |
4291 | ||
28d8d7f4 | 4292 | *p++ = ':'; |
bcdf7404 YO |
4293 | Copy(RExC_precomp, p, r->prelen, char); |
4294 | r->precomp = p; | |
4295 | p += r->prelen; | |
4296 | if (has_runon) | |
28d8d7f4 YO |
4297 | *p++ = '\n'; |
4298 | *p++ = ')'; | |
4299 | *p = 0; | |
bcdf7404 YO |
4300 | } |
4301 | ||
bbe252da | 4302 | r->intflags = 0; |
830247a4 | 4303 | r->nparens = RExC_npar - 1; /* set early to validate backrefs */ |
81714fb9 | 4304 | |
6bda09f9 | 4305 | if (RExC_seen & REG_SEEN_RECURSE) { |
40d049e4 YO |
4306 | Newxz(RExC_open_parens, RExC_npar,regnode *); |
4307 | SAVEFREEPV(RExC_open_parens); | |
4308 | Newxz(RExC_close_parens,RExC_npar,regnode *); | |
4309 | SAVEFREEPV(RExC_close_parens); | |
6bda09f9 YO |
4310 | } |
4311 | ||
4312 | /* Useful during FAIL. */ | |
7122b237 YO |
4313 | #ifdef RE_TRACK_PATTERN_OFFSETS |
4314 | Newxz(ri->u.offsets, 2*RExC_size+1, U32); /* MJD 20001228 */ | |
a3621e74 | 4315 | DEBUG_OFFSETS_r(PerlIO_printf(Perl_debug_log, |
2af232bd | 4316 | "%s %"UVuf" bytes for offset annotations.\n", |
7122b237 | 4317 | ri->u.offsets ? "Got" : "Couldn't get", |
392fbf5d | 4318 | (UV)((2*RExC_size+1) * sizeof(U32)))); |
7122b237 YO |
4319 | #endif |
4320 | SetProgLen(ri,RExC_size); | |
830247a4 | 4321 | RExC_rx = r; |
f8fc2ecf | 4322 | RExC_rxi = ri; |
bbce6d69 | 4323 | |
4324 | /* Second pass: emit code. */ | |
c737faaf | 4325 | RExC_flags = pm_flags; /* don't let top level (?i) bleed */ |
830247a4 IZ |
4326 | RExC_parse = exp; |
4327 | RExC_end = xend; | |
4328 | RExC_naughty = 0; | |
4329 | RExC_npar = 1; | |
f8fc2ecf YO |
4330 | RExC_emit_start = ri->program; |
4331 | RExC_emit = ri->program; | |
3b57cd43 YO |
4332 | RExC_emit_bound = ri->program + RExC_size + 1; |
4333 | ||
2cd61cdb | 4334 | /* Store the count of eval-groups for security checks: */ |
f8149455 | 4335 | RExC_rx->seen_evals = RExC_seen_evals; |
830247a4 | 4336 | REGC((U8)REG_MAGIC, (char*) RExC_emit++); |
80757612 YO |
4337 | if (reg(pRExC_state, 0, &flags,1) == NULL) { |
4338 | ReREFCNT_dec(r); | |
a0d0e21e | 4339 | return(NULL); |
80757612 | 4340 | } |
07be1b83 YO |
4341 | /* XXXX To minimize changes to RE engine we always allocate |
4342 | 3-units-long substrs field. */ | |
4343 | Newx(r->substrs, 1, struct reg_substr_data); | |
40d049e4 YO |
4344 | if (RExC_recurse_count) { |
4345 | Newxz(RExC_recurse,RExC_recurse_count,regnode *); | |
4346 | SAVEFREEPV(RExC_recurse); | |
4347 | } | |
a0d0e21e | 4348 | |
07be1b83 | 4349 | reStudy: |
1de06328 | 4350 | r->minlen = minlen = sawplus = sawopen = 0; |
07be1b83 | 4351 | Zero(r->substrs, 1, struct reg_substr_data); |
a3621e74 | 4352 | |
07be1b83 YO |
4353 | #ifdef TRIE_STUDY_OPT |
4354 | if ( restudied ) { | |
5d458dd8 | 4355 | U32 seen=RExC_seen; |
07be1b83 | 4356 | DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log,"Restudying\n")); |
5d458dd8 YO |
4357 | |
4358 | RExC_state = copyRExC_state; | |
4359 | if (seen & REG_TOP_LEVEL_BRANCHES) | |
4360 | RExC_seen |= REG_TOP_LEVEL_BRANCHES; | |
4361 | else | |
4362 | RExC_seen &= ~REG_TOP_LEVEL_BRANCHES; | |
1de06328 | 4363 | if (data.last_found) { |
07be1b83 | 4364 | SvREFCNT_dec(data.longest_fixed); |
07be1b83 | 4365 | SvREFCNT_dec(data.longest_float); |
07be1b83 | 4366 | SvREFCNT_dec(data.last_found); |
1de06328 | 4367 | } |
40d049e4 | 4368 | StructCopy(&zero_scan_data, &data, scan_data_t); |
07be1b83 | 4369 | } else { |
40d049e4 | 4370 | StructCopy(&zero_scan_data, &data, scan_data_t); |
5d458dd8 | 4371 | copyRExC_state = RExC_state; |
07be1b83 | 4372 | } |
40d049e4 YO |
4373 | #else |
4374 | StructCopy(&zero_scan_data, &data, scan_data_t); | |
07be1b83 | 4375 | #endif |
fc8cd66c | 4376 | |
a0d0e21e | 4377 | /* Dig out information for optimizations. */ |
f7819f85 | 4378 | r->extflags = RExC_flags; /* was pm_op */ |
c737faaf YO |
4379 | /*dmq: removed as part of de-PMOP: pm->op_pmflags = RExC_flags; */ |
4380 | ||
a0ed51b3 | 4381 | if (UTF) |
bbe252da | 4382 | r->extflags |= RXf_UTF8; /* Unicode in it? */ |
f8fc2ecf | 4383 | ri->regstclass = NULL; |
830247a4 | 4384 | if (RExC_naughty >= 10) /* Probably an expensive pattern. */ |
bbe252da | 4385 | r->intflags |= PREGf_NAUGHTY; |
f8fc2ecf | 4386 | scan = ri->program + 1; /* First BRANCH. */ |
2779dcf1 | 4387 | |
1de06328 YO |
4388 | /* testing for BRANCH here tells us whether there is "must appear" |
4389 | data in the pattern. If there is then we can use it for optimisations */ | |
eaf3ca90 | 4390 | if (!(RExC_seen & REG_TOP_LEVEL_BRANCHES)) { /* Only one top-level choice. */ |
c277df42 | 4391 | I32 fake; |
c5254dd6 | 4392 | STRLEN longest_float_length, longest_fixed_length; |
07be1b83 | 4393 | struct regnode_charclass_class ch_class; /* pointed to by data */ |
653099ff | 4394 | int stclass_flag; |
07be1b83 | 4395 | I32 last_close = 0; /* pointed to by data */ |
5339e136 YO |
4396 | regnode *first= scan; |
4397 | regnode *first_next= regnext(first); | |
4398 | ||
c277df42 | 4399 | /* Skip introductions and multiplicators >= 1. */ |
a0d0e21e | 4400 | while ((OP(first) == OPEN && (sawopen = 1)) || |
653099ff | 4401 | /* An OR of *one* alternative - should not happen now. */ |
5339e136 | 4402 | (OP(first) == BRANCH && OP(first_next) != BRANCH) || |
07be1b83 YO |
4403 | /* for now we can't handle lookbehind IFMATCH*/ |
4404 | (OP(first) == IFMATCH && !first->flags) || | |
a0d0e21e LW |
4405 | (OP(first) == PLUS) || |
4406 | (OP(first) == MINMOD) || | |
653099ff | 4407 | /* An {n,m} with n>0 */ |
5339e136 YO |
4408 | (PL_regkind[OP(first)] == CURLY && ARG1(first) > 0) || |
4409 | (OP(first) == NOTHING && PL_regkind[OP(first_next)] != END )) | |
07be1b83 | 4410 | { |
786e8c11 | 4411 | |
a0d0e21e LW |
4412 | if (OP(first) == PLUS) |
4413 | sawplus = 1; | |
4414 | else | |
3dab1dad | 4415 | first += regarglen[OP(first)]; |
07be1b83 YO |
4416 | if (OP(first) == IFMATCH) { |
4417 | first = NEXTOPER(first); | |
4418 | first += EXTRA_STEP_2ARGS; | |
7c167cea | 4419 | } else /* XXX possible optimisation for /(?=)/ */ |
07be1b83 | 4420 | first = NEXTOPER(first); |
5339e136 | 4421 | first_next= regnext(first); |
a687059c LW |
4422 | } |
4423 | ||
a0d0e21e LW |
4424 | /* Starting-point info. */ |
4425 | again: | |
786e8c11 | 4426 | DEBUG_PEEP("first:",first,0); |
07be1b83 | 4427 | /* Ignore EXACT as we deal with it later. */ |
3dab1dad | 4428 | if (PL_regkind[OP(first)] == EXACT) { |
1aa99e6b | 4429 | if (OP(first) == EXACT) |
6f207bd3 | 4430 | NOOP; /* Empty, get anchored substr later. */ |
1aa99e6b | 4431 | else if ((OP(first) == EXACTF || OP(first) == EXACTFL)) |
f8fc2ecf | 4432 | ri->regstclass = first; |
b3c9acc1 | 4433 | } |
07be1b83 | 4434 | #ifdef TRIE_STCLASS |
786e8c11 | 4435 | else if (PL_regkind[OP(first)] == TRIE && |
f8fc2ecf | 4436 | ((reg_trie_data *)ri->data->data[ ARG(first) ])->minlen>0) |
07be1b83 | 4437 | { |
786e8c11 | 4438 | regnode *trie_op; |
07be1b83 | 4439 | /* this can happen only on restudy */ |
786e8c11 | 4440 | if ( OP(first) == TRIE ) { |
c944940b | 4441 | struct regnode_1 *trieop = (struct regnode_1 *) |
446bd890 | 4442 | PerlMemShared_calloc(1, sizeof(struct regnode_1)); |
786e8c11 YO |
4443 | StructCopy(first,trieop,struct regnode_1); |
4444 | trie_op=(regnode *)trieop; | |
4445 | } else { | |
c944940b | 4446 | struct regnode_charclass *trieop = (struct regnode_charclass *) |
446bd890 | 4447 | PerlMemShared_calloc(1, sizeof(struct regnode_charclass)); |
786e8c11 YO |
4448 | StructCopy(first,trieop,struct regnode_charclass); |
4449 | trie_op=(regnode *)trieop; | |
4450 | } | |
1de06328 | 4451 | OP(trie_op)+=2; |
786e8c11 | 4452 | make_trie_failtable(pRExC_state, (regnode *)first, trie_op, 0); |
f8fc2ecf | 4453 | ri->regstclass = trie_op; |
07be1b83 YO |
4454 | } |
4455 | #endif | |
bfed75c6 | 4456 | else if (strchr((const char*)PL_simple,OP(first))) |
f8fc2ecf | 4457 | ri->regstclass = first; |
3dab1dad YO |
4458 | else if (PL_regkind[OP(first)] == BOUND || |
4459 | PL_regkind[OP(first)] == NBOUND) | |
f8fc2ecf | 4460 | ri->regstclass = first; |
3dab1dad | 4461 | else if (PL_regkind[OP(first)] == BOL) { |
bbe252da YO |
4462 | r->extflags |= (OP(first) == MBOL |
4463 | ? RXf_ANCH_MBOL | |
cad2e5aa | 4464 | : (OP(first) == SBOL |
bbe252da YO |
4465 | ? RXf_ANCH_SBOL |
4466 | : RXf_ANCH_BOL)); | |
a0d0e21e | 4467 | first = NEXTOPER(first); |
774d564b | 4468 | goto again; |
4469 | } | |
4470 | else if (OP(first) == GPOS) { | |
bbe252da | 4471 | r->extflags |= RXf_ANCH_GPOS; |
774d564b | 4472 | first = NEXTOPER(first); |
4473 | goto again; | |
a0d0e21e | 4474 | } |
cf2a2b69 YO |
4475 | else if ((!sawopen || !RExC_sawback) && |
4476 | (OP(first) == STAR && | |
3dab1dad | 4477 | PL_regkind[OP(NEXTOPER(first))] == REG_ANY) && |
bbe252da | 4478 | !(r->extflags & RXf_ANCH) && !(RExC_seen & REG_SEEN_EVAL)) |
a0d0e21e LW |
4479 | { |
4480 | /* turn .* into ^.* with an implied $*=1 */ | |
1df70142 AL |
4481 | const int type = |
4482 | (OP(NEXTOPER(first)) == REG_ANY) | |
bbe252da YO |
4483 | ? RXf_ANCH_MBOL |
4484 | : RXf_ANCH_SBOL; | |
4485 | r->extflags |= type; | |
4486 | r->intflags |= PREGf_IMPLICIT; | |
a0d0e21e | 4487 | first = NEXTOPER(first); |
774d564b | 4488 | goto again; |
a0d0e21e | 4489 | } |
b81d288d | 4490 | if (sawplus && (!sawopen || !RExC_sawback) |
830247a4 | 4491 | && !(RExC_seen & REG_SEEN_EVAL)) /* May examine pos and $& */ |
cad2e5aa | 4492 | /* x+ must match at the 1st pos of run of x's */ |
bbe252da | 4493 | r->intflags |= PREGf_SKIP; |
a0d0e21e | 4494 | |
c277df42 | 4495 | /* Scan is after the zeroth branch, first is atomic matcher. */ |
be8e71aa | 4496 | #ifdef TRIE_STUDY_OPT |
81714fb9 | 4497 | DEBUG_PARSE_r( |
be8e71aa YO |
4498 | if (!restudied) |
4499 | PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n", | |
4500 | (IV)(first - scan + 1)) | |
4501 | ); | |
4502 | #else | |
81714fb9 | 4503 | DEBUG_PARSE_r( |
be8e71aa YO |
4504 | PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n", |
4505 | (IV)(first - scan + 1)) | |
4506 | ); | |
4507 | #endif | |
4508 | ||
4509 | ||
a0d0e21e LW |
4510 | /* |
4511 | * If there's something expensive in the r.e., find the | |
4512 | * longest literal string that must appear and make it the | |
4513 | * regmust. Resolve ties in favor of later strings, since | |
4514 | * the regstart check works with the beginning of the r.e. | |
4515 | * and avoiding duplication strengthens checking. Not a | |
4516 | * strong reason, but sufficient in the absence of others. | |
4517 | * [Now we resolve ties in favor of the earlier string if | |
c277df42 | 4518 | * it happens that c_offset_min has been invalidated, since the |
a0d0e21e LW |
4519 | * earlier string may buy us something the later one won't.] |
4520 | */ | |
de8c5301 | 4521 | |
396482e1 GA |
4522 | data.longest_fixed = newSVpvs(""); |
4523 | data.longest_float = newSVpvs(""); | |
4524 | data.last_found = newSVpvs(""); | |
c277df42 IZ |
4525 | data.longest = &(data.longest_fixed); |
4526 | first = scan; | |
f8fc2ecf | 4527 | if (!ri->regstclass) { |
830247a4 | 4528 | cl_init(pRExC_state, &ch_class); |
653099ff GS |
4529 | data.start_class = &ch_class; |
4530 | stclass_flag = SCF_DO_STCLASS_AND; | |
4531 | } else /* XXXX Check for BOUND? */ | |
4532 | stclass_flag = 0; | |
cb434fcc | 4533 | data.last_closep = &last_close; |
de8c5301 | 4534 | |
1de06328 | 4535 | minlen = study_chunk(pRExC_state, &first, &minlen, &fake, scan + RExC_size, /* Up to end */ |
40d049e4 YO |
4536 | &data, -1, NULL, NULL, |
4537 | SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag,0); | |
07be1b83 | 4538 | |
07be1b83 | 4539 | |
786e8c11 YO |
4540 | CHECK_RESTUDY_GOTO; |
4541 | ||
4542 | ||
830247a4 | 4543 | if ( RExC_npar == 1 && data.longest == &(data.longest_fixed) |
b81d288d | 4544 | && data.last_start_min == 0 && data.last_end > 0 |
830247a4 | 4545 | && !RExC_seen_zerolen |
2bf803e2 | 4546 | && !(RExC_seen & REG_SEEN_VERBARG) |
bbe252da YO |
4547 | && (!(RExC_seen & REG_SEEN_GPOS) || (r->extflags & RXf_ANCH_GPOS))) |
4548 | r->extflags |= RXf_CHECK_ALL; | |
304ee84b | 4549 | scan_commit(pRExC_state, &data,&minlen,0); |
c277df42 IZ |
4550 | SvREFCNT_dec(data.last_found); |
4551 | ||
1de06328 YO |
4552 | /* Note that code very similar to this but for anchored string |
4553 | follows immediately below, changes may need to be made to both. | |
4554 | Be careful. | |
4555 | */ | |
a0ed51b3 | 4556 | longest_float_length = CHR_SVLEN(data.longest_float); |
c5254dd6 | 4557 | if (longest_float_length |
c277df42 IZ |
4558 | || (data.flags & SF_FL_BEFORE_EOL |
4559 | && (!(data.flags & SF_FL_BEFORE_MEOL) | |
bbe252da | 4560 | || (RExC_flags & RXf_PMf_MULTILINE)))) |
1de06328 | 4561 | { |
1182767e | 4562 | I32 t,ml; |
cf93c79d | 4563 | |
1de06328 | 4564 | if (SvCUR(data.longest_fixed) /* ok to leave SvCUR */ |
aca2d497 IZ |
4565 | && data.offset_fixed == data.offset_float_min |
4566 | && SvCUR(data.longest_fixed) == SvCUR(data.longest_float)) | |
4567 | goto remove_float; /* As in (a)+. */ | |
4568 | ||
1de06328 YO |
4569 | /* copy the information about the longest float from the reg_scan_data |
4570 | over to the program. */ | |
33b8afdf JH |
4571 | if (SvUTF8(data.longest_float)) { |
4572 | r->float_utf8 = data.longest_float; | |
c445ea15 | 4573 | r->float_substr = NULL; |
33b8afdf JH |
4574 | } else { |
4575 | r->float_substr = data.longest_float; | |
c445ea15 | 4576 | r->float_utf8 = NULL; |
33b8afdf | 4577 | } |
1de06328 YO |
4578 | /* float_end_shift is how many chars that must be matched that |
4579 | follow this item. We calculate it ahead of time as once the | |
4580 | lookbehind offset is added in we lose the ability to correctly | |
4581 | calculate it.*/ | |
4582 | ml = data.minlen_float ? *(data.minlen_float) | |
1182767e | 4583 | : (I32)longest_float_length; |
1de06328 YO |
4584 | r->float_end_shift = ml - data.offset_float_min |
4585 | - longest_float_length + (SvTAIL(data.longest_float) != 0) | |
4586 | + data.lookbehind_float; | |
4587 | r->float_min_offset = data.offset_float_min - data.lookbehind_float; | |
c277df42 | 4588 | r->float_max_offset = data.offset_float_max; |
1182767e | 4589 | if (data.offset_float_max < I32_MAX) /* Don't offset infinity */ |
1de06328 YO |
4590 | r->float_max_offset -= data.lookbehind_float; |
4591 | ||
cf93c79d IZ |
4592 | t = (data.flags & SF_FL_BEFORE_EOL /* Can't have SEOL and MULTI */ |
4593 | && (!(data.flags & SF_FL_BEFORE_MEOL) | |
bbe252da | 4594 | || (RExC_flags & RXf_PMf_MULTILINE))); |
33b8afdf | 4595 | fbm_compile(data.longest_float, t ? FBMcf_TAIL : 0); |
a0ed51b3 LW |
4596 | } |
4597 | else { | |
aca2d497 | 4598 | remove_float: |
c445ea15 | 4599 | r->float_substr = r->float_utf8 = NULL; |
c277df42 | 4600 | SvREFCNT_dec(data.longest_float); |
c5254dd6 | 4601 | longest_float_length = 0; |
a0d0e21e | 4602 | } |
c277df42 | 4603 | |
1de06328 YO |
4604 | /* Note that code very similar to this but for floating string |
4605 | is immediately above, changes may need to be made to both. | |
4606 | Be careful. | |
4607 | */ | |
a0ed51b3 | 4608 | longest_fixed_length = CHR_SVLEN(data.longest_fixed); |
c5254dd6 | 4609 | if (longest_fixed_length |
c277df42 IZ |
4610 | || (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */ |
4611 | && (!(data.flags & SF_FIX_BEFORE_MEOL) | |
bbe252da | 4612 | || (RExC_flags & RXf_PMf_MULTILINE)))) |
1de06328 | 4613 | { |
1182767e | 4614 | I32 t,ml; |
cf93c79d | 4615 | |
1de06328 YO |
4616 | /* copy the information about the longest fixed |
4617 | from the reg_scan_data over to the program. */ | |
33b8afdf JH |
4618 | if (SvUTF8(data.longest_fixed)) { |
4619 | r->anchored_utf8 = data.longest_fixed; | |
c445ea15 | 4620 | r->anchored_substr = NULL; |
33b8afdf JH |
4621 | } else { |
4622 | r->anchored_substr = data.longest_fixed; | |
c445ea15 | 4623 | r->anchored_utf8 = NULL; |
33b8afdf | 4624 | } |
1de06328 YO |
4625 | /* fixed_end_shift is how many chars that must be matched that |
4626 | follow this item. We calculate it ahead of time as once the | |
4627 | lookbehind offset is added in we lose the ability to correctly | |
4628 | calculate it.*/ | |
4629 | ml = data.minlen_fixed ? *(data.minlen_fixed) | |
1182767e | 4630 | : (I32)longest_fixed_length; |
1de06328 YO |
4631 | r->anchored_end_shift = ml - data.offset_fixed |
4632 | - longest_fixed_length + (SvTAIL(data.longest_fixed) != 0) | |
4633 | + data.lookbehind_fixed; | |
4634 | r->anchored_offset = data.offset_fixed - data.lookbehind_fixed; | |
4635 | ||
cf93c79d IZ |
4636 | t = (data.flags & SF_FIX_BEFORE_EOL /* Can't have SEOL and MULTI */ |
4637 | && (!(data.flags & SF_FIX_BEFORE_MEOL) | |
bbe252da | 4638 | || (RExC_flags & RXf_PMf_MULTILINE))); |
33b8afdf | 4639 | fbm_compile(data.longest_fixed, t ? FBMcf_TAIL : 0); |
a0ed51b3 LW |
4640 | } |
4641 | else { | |
c445ea15 | 4642 | r->anchored_substr = r->anchored_utf8 = NULL; |
c277df42 | 4643 | SvREFCNT_dec(data.longest_fixed); |
c5254dd6 | 4644 | longest_fixed_length = 0; |
a0d0e21e | 4645 | } |
f8fc2ecf YO |
4646 | if (ri->regstclass |
4647 | && (OP(ri->regstclass) == REG_ANY || OP(ri->regstclass) == SANY)) | |
4648 | ri->regstclass = NULL; | |
33b8afdf JH |
4649 | if ((!(r->anchored_substr || r->anchored_utf8) || r->anchored_offset) |
4650 | && stclass_flag | |
653099ff | 4651 | && !(data.start_class->flags & ANYOF_EOS) |
eb160463 GS |
4652 | && !cl_is_anything(data.start_class)) |
4653 | { | |
2eccd3b2 | 4654 | const U32 n = add_data(pRExC_state, 1, "f"); |
653099ff | 4655 | |
f8fc2ecf | 4656 | Newx(RExC_rxi->data->data[n], 1, |
653099ff GS |
4657 | struct regnode_charclass_class); |
4658 | StructCopy(data.start_class, | |
f8fc2ecf | 4659 | (struct regnode_charclass_class*)RExC_rxi->data->data[n], |
653099ff | 4660 | struct regnode_charclass_class); |
f8fc2ecf | 4661 | ri->regstclass = (regnode*)RExC_rxi->data->data[n]; |
bbe252da | 4662 | r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */ |
a3621e74 | 4663 | DEBUG_COMPILE_r({ SV *sv = sv_newmortal(); |
32fc9b6a | 4664 | regprop(r, sv, (regnode*)data.start_class); |
9c5ffd7c | 4665 | PerlIO_printf(Perl_debug_log, |
a0288114 | 4666 | "synthetic stclass \"%s\".\n", |
3f7c398e | 4667 | SvPVX_const(sv));}); |
653099ff | 4668 | } |
c277df42 IZ |
4669 | |
4670 | /* A temporary algorithm prefers floated substr to fixed one to dig more info. */ | |
c5254dd6 | 4671 | if (longest_fixed_length > longest_float_length) { |
1de06328 | 4672 | r->check_end_shift = r->anchored_end_shift; |
c277df42 | 4673 | r->check_substr = r->anchored_substr; |
33b8afdf | 4674 | r->check_utf8 = r->anchored_utf8; |
c277df42 | 4675 | r->check_offset_min = r->check_offset_max = r->anchored_offset; |
bbe252da YO |
4676 | if (r->extflags & RXf_ANCH_SINGLE) |
4677 | r->extflags |= RXf_NOSCAN; | |
a0ed51b3 LW |
4678 | } |
4679 | else { | |
1de06328 | 4680 | r->check_end_shift = r->float_end_shift; |
c277df42 | 4681 | r->check_substr = r->float_substr; |
33b8afdf | 4682 | r->check_utf8 = r->float_utf8; |
1de06328 YO |
4683 | r->check_offset_min = r->float_min_offset; |
4684 | r->check_offset_max = r->float_max_offset; | |
a0d0e21e | 4685 | } |
30382c73 IZ |
4686 | /* XXXX Currently intuiting is not compatible with ANCH_GPOS. |
4687 | This should be changed ASAP! */ | |
bbe252da YO |
4688 | if ((r->check_substr || r->check_utf8) && !(r->extflags & RXf_ANCH_GPOS)) { |
4689 | r->extflags |= RXf_USE_INTUIT; | |
33b8afdf | 4690 | if (SvTAIL(r->check_substr ? r->check_substr : r->check_utf8)) |
bbe252da | 4691 | r->extflags |= RXf_INTUIT_TAIL; |
cad2e5aa | 4692 | } |
1de06328 YO |
4693 | /* XXX Unneeded? dmq (shouldn't as this is handled elsewhere) |
4694 | if ( (STRLEN)minlen < longest_float_length ) | |
4695 | minlen= longest_float_length; | |
4696 | if ( (STRLEN)minlen < longest_fixed_length ) | |
4697 | minlen= longest_fixed_length; | |
4698 | */ | |
a0ed51b3 LW |
4699 | } |
4700 | else { | |
c277df42 IZ |
4701 | /* Several toplevels. Best we can is to set minlen. */ |
4702 | I32 fake; | |
653099ff | 4703 | struct regnode_charclass_class ch_class; |
cb434fcc | 4704 | I32 last_close = 0; |
c277df42 | 4705 | |
5d458dd8 | 4706 | DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "\nMulti Top Level\n")); |
07be1b83 | 4707 | |
f8fc2ecf | 4708 | scan = ri->program + 1; |
830247a4 | 4709 | cl_init(pRExC_state, &ch_class); |
653099ff | 4710 | data.start_class = &ch_class; |
cb434fcc | 4711 | data.last_closep = &last_close; |
07be1b83 | 4712 | |
de8c5301 | 4713 | |
1de06328 | 4714 | minlen = study_chunk(pRExC_state, &scan, &minlen, &fake, scan + RExC_size, |
40d049e4 | 4715 | &data, -1, NULL, NULL, SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS,0); |
de8c5301 | 4716 | |
786e8c11 | 4717 | CHECK_RESTUDY_GOTO; |
07be1b83 | 4718 | |
33b8afdf | 4719 | r->check_substr = r->check_utf8 = r->anchored_substr = r->anchored_utf8 |
c445ea15 | 4720 | = r->float_substr = r->float_utf8 = NULL; |
653099ff | 4721 | if (!(data.start_class->flags & ANYOF_EOS) |
eb160463 GS |
4722 | && !cl_is_anything(data.start_class)) |
4723 | { | |
2eccd3b2 | 4724 | const U32 n = add_data(pRExC_state, 1, "f"); |
653099ff | 4725 | |
f8fc2ecf | 4726 | Newx(RExC_rxi->data->data[n], 1, |
653099ff GS |
4727 | struct regnode_charclass_class); |
4728 | StructCopy(data.start_class, | |
f8fc2ecf | 4729 | (struct regnode_charclass_class*)RExC_rxi->data->data[n], |
653099ff | 4730 | struct regnode_charclass_class); |
f8fc2ecf | 4731 | ri->regstclass = (regnode*)RExC_rxi->data->data[n]; |
bbe252da | 4732 | r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */ |
a3621e74 | 4733 | DEBUG_COMPILE_r({ SV* sv = sv_newmortal(); |
32fc9b6a | 4734 | regprop(r, sv, (regnode*)data.start_class); |
9c5ffd7c | 4735 | PerlIO_printf(Perl_debug_log, |
a0288114 | 4736 | "synthetic stclass \"%s\".\n", |
3f7c398e | 4737 | SvPVX_const(sv));}); |
653099ff | 4738 | } |
a0d0e21e LW |
4739 | } |
4740 | ||
1de06328 YO |
4741 | /* Guard against an embedded (?=) or (?<=) with a longer minlen than |
4742 | the "real" pattern. */ | |
cf9788e3 RGS |
4743 | DEBUG_OPTIMISE_r({ |
4744 | PerlIO_printf(Perl_debug_log,"minlen: %"IVdf" r->minlen:%"IVdf"\n", | |
70685ca0 | 4745 | (IV)minlen, (IV)r->minlen); |
cf9788e3 | 4746 | }); |
de8c5301 | 4747 | r->minlenret = minlen; |
1de06328 YO |
4748 | if (r->minlen < minlen) |
4749 | r->minlen = minlen; | |
4750 | ||
b81d288d | 4751 | if (RExC_seen & REG_SEEN_GPOS) |
bbe252da | 4752 | r->extflags |= RXf_GPOS_SEEN; |
830247a4 | 4753 | if (RExC_seen & REG_SEEN_LOOKBEHIND) |
bbe252da | 4754 | r->extflags |= RXf_LOOKBEHIND_SEEN; |
830247a4 | 4755 | if (RExC_seen & REG_SEEN_EVAL) |
bbe252da | 4756 | r->extflags |= RXf_EVAL_SEEN; |
f33976b4 | 4757 | if (RExC_seen & REG_SEEN_CANY) |
bbe252da | 4758 | r->extflags |= RXf_CANY_SEEN; |
e2e6a0f1 | 4759 | if (RExC_seen & REG_SEEN_VERBARG) |
bbe252da | 4760 | r->intflags |= PREGf_VERBARG_SEEN; |
5d458dd8 | 4761 | if (RExC_seen & REG_SEEN_CUTGROUP) |
bbe252da | 4762 | r->intflags |= PREGf_CUTGROUP_SEEN; |
81714fb9 YO |
4763 | if (RExC_paren_names) |
4764 | r->paren_names = (HV*)SvREFCNT_inc(RExC_paren_names); | |
4765 | else | |
4766 | r->paren_names = NULL; | |
0ac6acae | 4767 | |
7bd1e614 | 4768 | #ifdef STUPID_PATTERN_CHECKS |
640f820d AB |
4769 | if (r->prelen == 0) |
4770 | r->extflags |= RXf_NULL; | |
0ac6acae AB |
4771 | if (r->extflags & RXf_SPLIT && r->prelen == 1 && r->precomp[0] == ' ') |
4772 | /* XXX: this should happen BEFORE we compile */ | |
4773 | r->extflags |= (RXf_SKIPWHITE|RXf_WHITE); | |
4774 | else if (r->prelen == 3 && memEQ("\\s+", r->precomp, 3)) | |
4775 | r->extflags |= RXf_WHITE; | |
e357fc67 YO |
4776 | else if (r->prelen == 1 && r->precomp[0] == '^') |
4777 | r->extflags |= RXf_START_ONLY; | |
f1b875a0 | 4778 | #else |
7bd1e614 YO |
4779 | if (r->extflags & RXf_SPLIT && r->prelen == 1 && r->precomp[0] == ' ') |
4780 | /* XXX: this should happen BEFORE we compile */ | |
4781 | r->extflags |= (RXf_SKIPWHITE|RXf_WHITE); | |
4782 | else { | |
4783 | regnode *first = ri->program + 1; | |
39aa8307 JH |
4784 | U8 fop = OP(first); |
4785 | U8 nop = OP(NEXTOPER(first)); | |
7bd1e614 | 4786 | |
640f820d AB |
4787 | if (PL_regkind[fop] == NOTHING && nop == END) |
4788 | r->extflags |= RXf_NULL; | |
4789 | else if (PL_regkind[fop] == BOL && nop == END) | |
7bd1e614 YO |
4790 | r->extflags |= RXf_START_ONLY; |
4791 | else if (fop == PLUS && nop ==SPACE && OP(regnext(first))==END) | |
4792 | r->extflags |= RXf_WHITE; | |
4793 | } | |
f1b875a0 | 4794 | #endif |
1f1031fe YO |
4795 | #ifdef DEBUGGING |
4796 | if (RExC_paren_names) { | |
4797 | ri->name_list_idx = add_data( pRExC_state, 1, "p" ); | |
4798 | ri->data->data[ri->name_list_idx] = (void*)SvREFCNT_inc(RExC_paren_name_list); | |
4799 | } else | |
1f1031fe | 4800 | #endif |
cde0cee5 | 4801 | ri->name_list_idx = 0; |
1f1031fe | 4802 | |
40d049e4 YO |
4803 | if (RExC_recurse_count) { |
4804 | for ( ; RExC_recurse_count ; RExC_recurse_count-- ) { | |
4805 | const regnode *scan = RExC_recurse[RExC_recurse_count-1]; | |
4806 | ARG2L_SET( scan, RExC_open_parens[ARG(scan)-1] - scan ); | |
4807 | } | |
4808 | } | |
f0ab9afb | 4809 | Newxz(r->offs, RExC_npar, regexp_paren_pair); |
c74340f9 YO |
4810 | /* assume we don't need to swap parens around before we match */ |
4811 | ||
be8e71aa YO |
4812 | DEBUG_DUMP_r({ |
4813 | PerlIO_printf(Perl_debug_log,"Final program:\n"); | |
3dab1dad YO |
4814 | regdump(r); |
4815 | }); | |
7122b237 YO |
4816 | #ifdef RE_TRACK_PATTERN_OFFSETS |
4817 | DEBUG_OFFSETS_r(if (ri->u.offsets) { | |
4818 | const U32 len = ri->u.offsets[0]; | |
8e9a8a48 YO |
4819 | U32 i; |
4820 | GET_RE_DEBUG_FLAGS_DECL; | |
7122b237 | 4821 | PerlIO_printf(Perl_debug_log, "Offsets: [%"UVuf"]\n\t", (UV)ri->u.offsets[0]); |
8e9a8a48 | 4822 | for (i = 1; i <= len; i++) { |
7122b237 | 4823 | if (ri->u.offsets[i*2-1] || ri->u.offsets[i*2]) |
8e9a8a48 | 4824 | PerlIO_printf(Perl_debug_log, "%"UVuf":%"UVuf"[%"UVuf"] ", |
7122b237 | 4825 | (UV)i, (UV)ri->u.offsets[i*2-1], (UV)ri->u.offsets[i*2]); |
8e9a8a48 YO |
4826 | } |
4827 | PerlIO_printf(Perl_debug_log, "\n"); | |
4828 | }); | |
7122b237 | 4829 | #endif |
a0d0e21e | 4830 | return(r); |
a687059c LW |
4831 | } |
4832 | ||
f9f4320a | 4833 | #undef RE_ENGINE_PTR |
3dab1dad | 4834 | |
93b32b6d | 4835 | |
81714fb9 | 4836 | SV* |
192b9cd1 AB |
4837 | Perl_reg_named_buff(pTHX_ REGEXP * const rx, SV * const key, SV * const value, |
4838 | const U32 flags) | |
4839 | { | |
4840 | PERL_UNUSED_ARG(value); | |
4841 | ||
f1b875a0 | 4842 | if (flags & RXapif_FETCH) { |
192b9cd1 | 4843 | return reg_named_buff_fetch(rx, key, flags); |
f1b875a0 | 4844 | } else if (flags & (RXapif_STORE | RXapif_DELETE | RXapif_CLEAR)) { |
192b9cd1 AB |
4845 | Perl_croak(aTHX_ PL_no_modify); |
4846 | return NULL; | |
f1b875a0 | 4847 | } else if (flags & RXapif_EXISTS) { |
192b9cd1 AB |
4848 | return reg_named_buff_exists(rx, key, flags) |
4849 | ? &PL_sv_yes | |
4850 | : &PL_sv_no; | |
f1b875a0 | 4851 | } else if (flags & RXapif_REGNAMES) { |
192b9cd1 | 4852 | return reg_named_buff_all(rx, flags); |
f1b875a0 | 4853 | } else if (flags & (RXapif_SCALAR | RXapif_REGNAMES_COUNT)) { |
192b9cd1 AB |
4854 | return reg_named_buff_scalar(rx, flags); |
4855 | } else { | |
4856 | Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff", (int)flags); | |
4857 | return NULL; | |
4858 | } | |
4859 | } | |
4860 | ||
4861 | SV* | |
4862 | Perl_reg_named_buff_iter(pTHX_ REGEXP * const rx, const SV * const lastkey, | |
4863 | const U32 flags) | |
4864 | { | |
4865 | PERL_UNUSED_ARG(lastkey); | |
4866 | ||
f1b875a0 | 4867 | if (flags & RXapif_FIRSTKEY) |
192b9cd1 | 4868 | return reg_named_buff_firstkey(rx, flags); |
f1b875a0 | 4869 | else if (flags & RXapif_NEXTKEY) |
192b9cd1 AB |
4870 | return reg_named_buff_nextkey(rx, flags); |
4871 | else { | |
4872 | Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_iter", (int)flags); | |
4873 | return NULL; | |
4874 | } | |
4875 | } | |
4876 | ||
4877 | SV* | |
2fdbfb4d | 4878 | Perl_reg_named_buff_fetch(pTHX_ REGEXP * const rx, SV * const namesv, const U32 flags) |
81714fb9 | 4879 | { |
44a2ac75 YO |
4880 | AV *retarray = NULL; |
4881 | SV *ret; | |
f1b875a0 | 4882 | if (flags & RXapif_ALL) |
44a2ac75 | 4883 | retarray=newAV(); |
93b32b6d YO |
4884 | |
4885 | if (rx && rx->paren_names) { | |
4886 | HE *he_str = hv_fetch_ent( rx->paren_names, namesv, 0, 0 ); | |
4887 | if (he_str) { | |
4888 | IV i; | |
4889 | SV* sv_dat=HeVAL(he_str); | |
4890 | I32 *nums=(I32*)SvPVX(sv_dat); | |
4891 | for ( i=0; i<SvIVX(sv_dat); i++ ) { | |
192b9cd1 AB |
4892 | if ((I32)(rx->nparens) >= nums[i] |
4893 | && rx->offs[nums[i]].start != -1 | |
4894 | && rx->offs[nums[i]].end != -1) | |
93b32b6d | 4895 | { |
49d7dfbc | 4896 | ret = newSVpvs(""); |
2fdbfb4d | 4897 | CALLREG_NUMBUF_FETCH(rx,nums[i],ret); |
93b32b6d YO |
4898 | if (!retarray) |
4899 | return ret; | |
4900 | } else { | |
4901 | ret = newSVsv(&PL_sv_undef); | |
4902 | } | |
4903 | if (retarray) { | |
5a5094bd | 4904 | SvREFCNT_inc_simple_void(ret); |
93b32b6d | 4905 | av_push(retarray, ret); |
81714fb9 YO |
4906 | } |
4907 | } | |
93b32b6d | 4908 | if (retarray) |
192b9cd1 AB |
4909 | return newRV((SV*)retarray); |
4910 | } | |
4911 | } | |
4912 | return NULL; | |
4913 | } | |
4914 | ||
4915 | bool | |
4916 | Perl_reg_named_buff_exists(pTHX_ REGEXP * const rx, SV * const key, | |
4917 | const U32 flags) | |
4918 | { | |
4919 | if (rx && rx->paren_names) { | |
f1b875a0 | 4920 | if (flags & RXapif_ALL) { |
192b9cd1 AB |
4921 | return hv_exists_ent(rx->paren_names, key, 0); |
4922 | } else { | |
6499cc01 RGS |
4923 | SV *sv = CALLREG_NAMED_BUFF_FETCH(rx, key, flags); |
4924 | if (sv) { | |
4925 | SvREFCNT_dec(sv); | |
192b9cd1 AB |
4926 | return TRUE; |
4927 | } else { | |
4928 | return FALSE; | |
4929 | } | |
4930 | } | |
4931 | } else { | |
4932 | return FALSE; | |
4933 | } | |
4934 | } | |
4935 | ||
4936 | SV* | |
4937 | Perl_reg_named_buff_firstkey(pTHX_ REGEXP * const rx, const U32 flags) | |
4938 | { | |
4939 | (void)hv_iterinit(rx->paren_names); | |
4940 | ||
f1b875a0 | 4941 | return CALLREG_NAMED_BUFF_NEXTKEY(rx, NULL, flags & ~RXapif_FIRSTKEY); |
192b9cd1 AB |
4942 | } |
4943 | ||
4944 | SV* | |
4945 | Perl_reg_named_buff_nextkey(pTHX_ REGEXP * const rx, const U32 flags) | |
4946 | { | |
4947 | if (rx && rx->paren_names) { | |
4948 | HV *hv = rx->paren_names; | |
4949 | HE *temphe; | |
4950 | while ( (temphe = hv_iternext_flags(hv,0)) ) { | |
4951 | IV i; | |
4952 | IV parno = 0; | |
4953 | SV* sv_dat = HeVAL(temphe); | |
4954 | I32 *nums = (I32*)SvPVX(sv_dat); | |
4955 | for ( i = 0; i < SvIVX(sv_dat); i++ ) { | |
4956 | if ((I32)(rx->lastcloseparen) >= nums[i] && | |
4957 | rx->offs[nums[i]].start != -1 && | |
4958 | rx->offs[nums[i]].end != -1) | |
4959 | { | |
4960 | parno = nums[i]; | |
4961 | break; | |
4962 | } | |
4963 | } | |
f1b875a0 | 4964 | if (parno || flags & RXapif_ALL) { |
192b9cd1 AB |
4965 | STRLEN len; |
4966 | char *pv = HePV(temphe, len); | |
4967 | return newSVpvn(pv,len); | |
4968 | } | |
81714fb9 YO |
4969 | } |
4970 | } | |
44a2ac75 YO |
4971 | return NULL; |
4972 | } | |
4973 | ||
192b9cd1 AB |
4974 | SV* |
4975 | Perl_reg_named_buff_scalar(pTHX_ REGEXP * const rx, const U32 flags) | |
4976 | { | |
4977 | SV *ret; | |
4978 | AV *av; | |
4979 | I32 length; | |
4980 | ||
4981 | if (rx && rx->paren_names) { | |
f1b875a0 | 4982 | if (flags & (RXapif_ALL | RXapif_REGNAMES_COUNT)) { |
192b9cd1 | 4983 | return newSViv(HvTOTALKEYS(rx->paren_names)); |
f1b875a0 YO |
4984 | } else if (flags & RXapif_ONE) { |
4985 | ret = CALLREG_NAMED_BUFF_ALL(rx, (flags | RXapif_REGNAMES)); | |
192b9cd1 AB |
4986 | av = (AV*)SvRV(ret); |
4987 | length = av_len(av); | |
4988 | return newSViv(length + 1); | |
4989 | } else { | |
4990 | Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_scalar", (int)flags); | |
4991 | return NULL; | |
4992 | } | |
4993 | } | |
4994 | return &PL_sv_undef; | |
4995 | } | |
4996 | ||
4997 | SV* | |
4998 | Perl_reg_named_buff_all(pTHX_ REGEXP * const rx, const U32 flags) | |
4999 | { | |
5000 | AV *av = newAV(); | |
5001 | ||
5002 | if (rx && rx->paren_names) { | |
5003 | HV *hv= rx->paren_names; | |
5004 | HE *temphe; | |
5005 | (void)hv_iterinit(hv); | |
5006 | while ( (temphe = hv_iternext_flags(hv,0)) ) { | |
5007 | IV i; | |
5008 | IV parno = 0; | |
5009 | SV* sv_dat = HeVAL(temphe); | |
5010 | I32 *nums = (I32*)SvPVX(sv_dat); | |
5011 | for ( i = 0; i < SvIVX(sv_dat); i++ ) { | |
5012 | if ((I32)(rx->lastcloseparen) >= nums[i] && | |
5013 | rx->offs[nums[i]].start != -1 && | |
5014 | rx->offs[nums[i]].end != -1) | |
5015 | { | |
5016 | parno = nums[i]; | |
5017 | break; | |
5018 | } | |
5019 | } | |
f1b875a0 | 5020 | if (parno || flags & RXapif_ALL) { |
192b9cd1 AB |
5021 | STRLEN len; |
5022 | char *pv = HePV(temphe, len); | |
5023 | av_push(av, newSVpvn(pv,len)); | |
5024 | } | |
5025 | } | |
5026 | } | |
5027 | ||
5028 | return newRV((SV*)av); | |
5029 | } | |
5030 | ||
49d7dfbc | 5031 | void |
2fdbfb4d | 5032 | Perl_reg_numbered_buff_fetch(pTHX_ REGEXP * const rx, const I32 paren, SV * const sv) |
44a2ac75 YO |
5033 | { |
5034 | char *s = NULL; | |
a9d504c3 | 5035 | I32 i = 0; |
44a2ac75 | 5036 | I32 s1, t1; |
44a2ac75 | 5037 | |
cde0cee5 YO |
5038 | if (!rx->subbeg) { |
5039 | sv_setsv(sv,&PL_sv_undef); | |
49d7dfbc | 5040 | return; |
cde0cee5 YO |
5041 | } |
5042 | else | |
f1b875a0 | 5043 | if (paren == RX_BUFF_IDX_PREMATCH && rx->offs[0].start != -1) { |
44a2ac75 | 5044 | /* $` */ |
f0ab9afb | 5045 | i = rx->offs[0].start; |
cde0cee5 | 5046 | s = rx->subbeg; |
44a2ac75 YO |
5047 | } |
5048 | else | |
f1b875a0 | 5049 | if (paren == RX_BUFF_IDX_POSTMATCH && rx->offs[0].end != -1) { |
44a2ac75 | 5050 | /* $' */ |
f0ab9afb NC |
5051 | s = rx->subbeg + rx->offs[0].end; |
5052 | i = rx->sublen - rx->offs[0].end; | |
44a2ac75 YO |
5053 | } |
5054 | else | |
5055 | if ( 0 <= paren && paren <= (I32)rx->nparens && | |
f0ab9afb NC |
5056 | (s1 = rx->offs[paren].start) != -1 && |
5057 | (t1 = rx->offs[paren].end) != -1) | |
44a2ac75 YO |
5058 | { |
5059 | /* $& $1 ... */ | |
5060 | i = t1 - s1; | |
5061 | s = rx->subbeg + s1; | |
cde0cee5 YO |
5062 | } else { |
5063 | sv_setsv(sv,&PL_sv_undef); | |
49d7dfbc | 5064 | return; |
cde0cee5 YO |
5065 | } |
5066 | assert(rx->sublen >= (s - rx->subbeg) + i ); | |
5067 | if (i >= 0) { | |
5068 | const int oldtainted = PL_tainted; | |
5069 | TAINT_NOT; | |
5070 | sv_setpvn(sv, s, i); | |
5071 | PL_tainted = oldtainted; | |
5072 | if ( (rx->extflags & RXf_CANY_SEEN) | |
5073 | ? (RX_MATCH_UTF8(rx) | |
5074 | && (!i || is_utf8_string((U8*)s, i))) | |
5075 | : (RX_MATCH_UTF8(rx)) ) | |
5076 | { | |
5077 | SvUTF8_on(sv); | |
5078 | } | |
5079 | else | |
5080 | SvUTF8_off(sv); | |
5081 | if (PL_tainting) { | |
5082 | if (RX_MATCH_TAINTED(rx)) { | |
5083 | if (SvTYPE(sv) >= SVt_PVMG) { | |
5084 | MAGIC* const mg = SvMAGIC(sv); | |
5085 | MAGIC* mgt; | |
5086 | PL_tainted = 1; | |
5087 | SvMAGIC_set(sv, mg->mg_moremagic); | |
5088 | SvTAINT(sv); | |
5089 | if ((mgt = SvMAGIC(sv))) { | |
5090 | mg->mg_moremagic = mgt; | |
5091 | SvMAGIC_set(sv, mg); | |
44a2ac75 | 5092 | } |
cde0cee5 YO |
5093 | } else { |
5094 | PL_tainted = 1; | |
5095 | SvTAINT(sv); | |
5096 | } | |
5097 | } else | |
5098 | SvTAINTED_off(sv); | |
44a2ac75 | 5099 | } |
81714fb9 | 5100 | } else { |
44a2ac75 | 5101 | sv_setsv(sv,&PL_sv_undef); |
49d7dfbc | 5102 | return; |
81714fb9 YO |
5103 | } |
5104 | } | |
93b32b6d | 5105 | |
2fdbfb4d AB |
5106 | void |
5107 | Perl_reg_numbered_buff_store(pTHX_ REGEXP * const rx, const I32 paren, | |
5108 | SV const * const value) | |
5109 | { | |
5110 | PERL_UNUSED_ARG(rx); | |
5111 | PERL_UNUSED_ARG(paren); | |
5112 | PERL_UNUSED_ARG(value); | |
5113 | ||
5114 | if (!PL_localizing) | |
5115 | Perl_croak(aTHX_ PL_no_modify); | |
5116 | } | |
5117 | ||
5118 | I32 | |
5119 | Perl_reg_numbered_buff_length(pTHX_ REGEXP * const rx, const SV * const sv, | |
5120 | const I32 paren) | |
5121 | { | |
5122 | I32 i; | |
5123 | I32 s1, t1; | |
5124 | ||
5125 | /* Some of this code was originally in C<Perl_magic_len> in F<mg.c> */ | |
5126 | switch (paren) { | |
192b9cd1 | 5127 | /* $` / ${^PREMATCH} */ |
f1b875a0 | 5128 | case RX_BUFF_IDX_PREMATCH: |
2fdbfb4d AB |
5129 | if (rx->offs[0].start != -1) { |
5130 | i = rx->offs[0].start; | |
5131 | if (i > 0) { | |
5132 | s1 = 0; | |
5133 | t1 = i; | |
5134 | goto getlen; | |
5135 | } | |
5136 | } | |
5137 | return 0; | |
192b9cd1 | 5138 | /* $' / ${^POSTMATCH} */ |
f1b875a0 | 5139 | case RX_BUFF_IDX_POSTMATCH: |
2fdbfb4d AB |
5140 | if (rx->offs[0].end != -1) { |
5141 | i = rx->sublen - rx->offs[0].end; | |
5142 | if (i > 0) { | |
5143 | s1 = rx->offs[0].end; | |
5144 | t1 = rx->sublen; | |
5145 | goto getlen; | |
5146 | } | |
5147 | } | |
5148 | return 0; | |
192b9cd1 AB |
5149 | /* $& / ${^MATCH}, $1, $2, ... */ |
5150 | default: | |
2fdbfb4d AB |
5151 | if (paren <= (I32)rx->nparens && |
5152 | (s1 = rx->offs[paren].start) != -1 && | |
5153 | (t1 = rx->offs[paren].end) != -1) | |
5154 | { | |
5155 | i = t1 - s1; | |
5156 | goto getlen; | |
5157 | } else { | |
5158 | if (ckWARN(WARN_UNINITIALIZED)) | |
5159 | report_uninit((SV*)sv); | |
5160 | return 0; | |
5161 | } | |
5162 | } | |
5163 | getlen: | |
5164 | if (i > 0 && RX_MATCH_UTF8(rx)) { | |
5165 | const char * const s = rx->subbeg + s1; | |
5166 | const U8 *ep; | |
5167 | STRLEN el; | |
5168 | ||
5169 | i = t1 - s1; | |
5170 | if (is_utf8_string_loclen((U8*)s, i, &ep, &el)) | |
5171 | i = el; | |
5172 | } | |
5173 | return i; | |
5174 | } | |
5175 | ||
fe578d7f | 5176 | SV* |
49d7dfbc | 5177 | Perl_reg_qr_package(pTHX_ REGEXP * const rx) |
fe578d7f AB |
5178 | { |
5179 | PERL_UNUSED_ARG(rx); | |
5180 | return newSVpvs("Regexp"); | |
5181 | } | |
0a4db386 | 5182 | |
894be9b7 | 5183 | /* Scans the name of a named buffer from the pattern. |
0a4db386 YO |
5184 | * If flags is REG_RSN_RETURN_NULL returns null. |
5185 | * If flags is REG_RSN_RETURN_NAME returns an SV* containing the name | |
5186 | * If flags is REG_RSN_RETURN_DATA returns the data SV* corresponding | |
5187 | * to the parsed name as looked up in the RExC_paren_names hash. | |
5188 | * If there is an error throws a vFAIL().. type exception. | |
894be9b7 | 5189 | */ |
0a4db386 YO |
5190 | |
5191 | #define REG_RSN_RETURN_NULL 0 | |
5192 | #define REG_RSN_RETURN_NAME 1 | |
5193 | #define REG_RSN_RETURN_DATA 2 | |
5194 | ||
894be9b7 YO |
5195 | STATIC SV* |
5196 | S_reg_scan_name(pTHX_ RExC_state_t *pRExC_state, U32 flags) { | |
5197 | char *name_start = RExC_parse; | |
1f1031fe YO |
5198 | |
5199 | if (isIDFIRST_lazy_if(RExC_parse, UTF)) { | |
5200 | /* skip IDFIRST by using do...while */ | |
5201 | if (UTF) | |
5202 | do { | |
5203 | RExC_parse += UTF8SKIP(RExC_parse); | |
5204 | } while (isALNUM_utf8((U8*)RExC_parse)); | |
5205 | else | |
5206 | do { | |
5207 | RExC_parse++; | |
5208 | } while (isALNUM(*RExC_parse)); | |
894be9b7 | 5209 | } |
1f1031fe | 5210 | |
0a4db386 YO |
5211 | if ( flags ) { |
5212 | SV* sv_name = sv_2mortal(Perl_newSVpvn(aTHX_ name_start, | |
5213 | (int)(RExC_parse - name_start))); | |
894be9b7 | 5214 | if (UTF) |
0a4db386 YO |
5215 | SvUTF8_on(sv_name); |
5216 | if ( flags == REG_RSN_RETURN_NAME) | |
5217 | return sv_name; | |
5218 | else if (flags==REG_RSN_RETURN_DATA) { | |
5219 | HE *he_str = NULL; | |
5220 | SV *sv_dat = NULL; | |
5221 | if ( ! sv_name ) /* should not happen*/ | |
5222 | Perl_croak(aTHX_ "panic: no svname in reg_scan_name"); | |
5223 | if (RExC_paren_names) | |
5224 | he_str = hv_fetch_ent( RExC_paren_names, sv_name, 0, 0 ); | |
5225 | if ( he_str ) | |
5226 | sv_dat = HeVAL(he_str); | |
5227 | if ( ! sv_dat ) | |
5228 | vFAIL("Reference to nonexistent named group"); | |
5229 | return sv_dat; | |
5230 | } | |
5231 | else { | |
5232 | Perl_croak(aTHX_ "panic: bad flag in reg_scan_name"); | |
5233 | } | |
5234 | /* NOT REACHED */ | |
894be9b7 | 5235 | } |
0a4db386 | 5236 | return NULL; |
894be9b7 YO |
5237 | } |
5238 | ||
3dab1dad YO |
5239 | #define DEBUG_PARSE_MSG(funcname) DEBUG_PARSE_r({ \ |
5240 | int rem=(int)(RExC_end - RExC_parse); \ | |
5241 | int cut; \ | |
5242 | int num; \ | |
5243 | int iscut=0; \ | |
5244 | if (rem>10) { \ | |
5245 | rem=10; \ | |
5246 | iscut=1; \ | |
5247 | } \ | |
5248 | cut=10-rem; \ | |
5249 | if (RExC_lastparse!=RExC_parse) \ | |
5250 | PerlIO_printf(Perl_debug_log," >%.*s%-*s", \ | |
5251 | rem, RExC_parse, \ | |
5252 | cut + 4, \ | |
5253 | iscut ? "..." : "<" \ | |
5254 | ); \ | |
5255 | else \ | |
5256 | PerlIO_printf(Perl_debug_log,"%16s",""); \ | |
5257 | \ | |
5258 | if (SIZE_ONLY) \ | |
3b57cd43 | 5259 | num = RExC_size + 1; \ |
3dab1dad YO |
5260 | else \ |
5261 | num=REG_NODE_NUM(RExC_emit); \ | |
5262 | if (RExC_lastnum!=num) \ | |
0a4db386 | 5263 | PerlIO_printf(Perl_debug_log,"|%4d",num); \ |
3dab1dad | 5264 | else \ |
0a4db386 | 5265 | PerlIO_printf(Perl_debug_log,"|%4s",""); \ |
be8e71aa YO |
5266 | PerlIO_printf(Perl_debug_log,"|%*s%-4s", \ |
5267 | (int)((depth*2)), "", \ | |
3dab1dad YO |
5268 | (funcname) \ |
5269 | ); \ | |
5270 | RExC_lastnum=num; \ | |
5271 | RExC_lastparse=RExC_parse; \ | |
5272 | }) | |
5273 | ||
07be1b83 YO |
5274 | |
5275 | ||
3dab1dad YO |
5276 | #define DEBUG_PARSE(funcname) DEBUG_PARSE_r({ \ |
5277 | DEBUG_PARSE_MSG((funcname)); \ | |
5278 | PerlIO_printf(Perl_debug_log,"%4s","\n"); \ | |
5279 | }) | |
6bda09f9 YO |
5280 | #define DEBUG_PARSE_FMT(funcname,fmt,args) DEBUG_PARSE_r({ \ |
5281 | DEBUG_PARSE_MSG((funcname)); \ | |
5282 | PerlIO_printf(Perl_debug_log,fmt "\n",args); \ | |
5283 | }) | |
a687059c LW |
5284 | /* |
5285 | - reg - regular expression, i.e. main body or parenthesized thing | |
5286 | * | |
5287 | * Caller must absorb opening parenthesis. | |
5288 | * | |
5289 | * Combining parenthesis handling with the base level of regular expression | |
5290 | * is a trifle forced, but the need to tie the tails of the branches to what | |
5291 | * follows makes it hard to avoid. | |
5292 | */ | |
07be1b83 YO |
5293 | #define REGTAIL(x,y,z) regtail((x),(y),(z),depth+1) |
5294 | #ifdef DEBUGGING | |
5295 | #define REGTAIL_STUDY(x,y,z) regtail_study((x),(y),(z),depth+1) | |
5296 | #else | |
5297 | #define REGTAIL_STUDY(x,y,z) regtail((x),(y),(z),depth+1) | |
5298 | #endif | |
3dab1dad | 5299 | |
76e3520e | 5300 | STATIC regnode * |
3dab1dad | 5301 | S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp,U32 depth) |
c277df42 | 5302 | /* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */ |
a687059c | 5303 | { |
27da23d5 | 5304 | dVAR; |
c277df42 IZ |
5305 | register regnode *ret; /* Will be the head of the group. */ |
5306 | register regnode *br; | |
5307 | register regnode *lastbr; | |
cbbf8932 | 5308 | register regnode *ender = NULL; |
a0d0e21e | 5309 | register I32 parno = 0; |
cbbf8932 | 5310 | I32 flags; |
f7819f85 | 5311 | U32 oregflags = RExC_flags; |
6136c704 AL |
5312 | bool have_branch = 0; |
5313 | bool is_open = 0; | |
594d7033 YO |
5314 | I32 freeze_paren = 0; |
5315 | I32 after_freeze = 0; | |
9d1d55b5 JP |
5316 | |
5317 | /* for (?g), (?gc), and (?o) warnings; warning | |
5318 | about (?c) will warn about (?g) -- japhy */ | |
5319 | ||
6136c704 AL |
5320 | #define WASTED_O 0x01 |
5321 | #define WASTED_G 0x02 | |
5322 | #define WASTED_C 0x04 | |
5323 | #define WASTED_GC (0x02|0x04) | |
cbbf8932 | 5324 | I32 wastedflags = 0x00; |
9d1d55b5 | 5325 | |
fac92740 | 5326 | char * parse_start = RExC_parse; /* MJD */ |
a28509cc | 5327 | char * const oregcomp_parse = RExC_parse; |
a0d0e21e | 5328 | |
3dab1dad YO |
5329 | GET_RE_DEBUG_FLAGS_DECL; |
5330 | DEBUG_PARSE("reg "); | |
5331 | ||
821b33a5 | 5332 | *flagp = 0; /* Tentatively. */ |
a0d0e21e | 5333 | |
9d1d55b5 | 5334 | |
a0d0e21e LW |
5335 | /* Make an OPEN node, if parenthesized. */ |
5336 | if (paren) { | |
e2e6a0f1 YO |
5337 | if ( *RExC_parse == '*') { /* (*VERB:ARG) */ |
5338 | char *start_verb = RExC_parse; | |
5339 | STRLEN verb_len = 0; | |
5340 | char *start_arg = NULL; | |
5341 | unsigned char op = 0; | |
5342 | int argok = 1; | |
5343 | int internal_argval = 0; /* internal_argval is only useful if !argok */ | |
5344 | while ( *RExC_parse && *RExC_parse != ')' ) { | |
5345 | if ( *RExC_parse == ':' ) { | |
5346 | start_arg = RExC_parse + 1; | |
5347 | break; | |
5348 | } | |
5349 | RExC_parse++; | |
5350 | } | |
5351 | ++start_verb; | |
5352 | verb_len = RExC_parse - start_verb; | |
5353 | if ( start_arg ) { | |
5354 | RExC_parse++; | |
5355 | while ( *RExC_parse && *RExC_parse != ')' ) | |
5356 | RExC_parse++; | |
5357 | if ( *RExC_parse != ')' ) | |
5358 | vFAIL("Unterminated verb pattern argument"); | |
5359 | if ( RExC_parse == start_arg ) | |
5360 | start_arg = NULL; | |
5361 | } else { | |
5362 | if ( *RExC_parse != ')' ) | |
5363 | vFAIL("Unterminated verb pattern"); | |
5364 | } | |
5d458dd8 | 5365 | |
e2e6a0f1 YO |
5366 | switch ( *start_verb ) { |
5367 | case 'A': /* (*ACCEPT) */ | |
568a785a | 5368 | if ( memEQs(start_verb,verb_len,"ACCEPT") ) { |
e2e6a0f1 YO |
5369 | op = ACCEPT; |
5370 | internal_argval = RExC_nestroot; | |
5371 | } | |
5372 | break; | |
5373 | case 'C': /* (*COMMIT) */ | |
568a785a | 5374 | if ( memEQs(start_verb,verb_len,"COMMIT") ) |
e2e6a0f1 | 5375 | op = COMMIT; |
e2e6a0f1 YO |
5376 | break; |
5377 | case 'F': /* (*FAIL) */ | |
568a785a | 5378 | if ( verb_len==1 || memEQs(start_verb,verb_len,"FAIL") ) { |
e2e6a0f1 YO |
5379 | op = OPFAIL; |
5380 | argok = 0; | |
5381 | } | |
5382 | break; | |
5d458dd8 YO |
5383 | case ':': /* (*:NAME) */ |
5384 | case 'M': /* (*MARK:NAME) */ | |
568a785a | 5385 | if ( verb_len==0 || memEQs(start_verb,verb_len,"MARK") ) { |
e2e6a0f1 | 5386 | op = MARKPOINT; |
5d458dd8 YO |
5387 | argok = -1; |
5388 | } | |
5389 | break; | |
5390 | case 'P': /* (*PRUNE) */ | |
568a785a | 5391 | if ( memEQs(start_verb,verb_len,"PRUNE") ) |
5d458dd8 | 5392 | op = PRUNE; |
e2e6a0f1 | 5393 | break; |
5d458dd8 | 5394 | case 'S': /* (*SKIP) */ |
568a785a | 5395 | if ( memEQs(start_verb,verb_len,"SKIP") ) |
5d458dd8 YO |
5396 | op = SKIP; |
5397 | break; | |
5398 | case 'T': /* (*THEN) */ | |
5399 | /* [19:06] <TimToady> :: is then */ | |
568a785a | 5400 | if ( memEQs(start_verb,verb_len,"THEN") ) { |
5d458dd8 YO |
5401 | op = CUTGROUP; |
5402 | RExC_seen |= REG_SEEN_CUTGROUP; | |
5403 | } | |
e2e6a0f1 YO |
5404 | break; |
5405 | } | |
5406 | if ( ! op ) { | |
5407 | RExC_parse++; | |
5408 | vFAIL3("Unknown verb pattern '%.*s'", | |
5409 | verb_len, start_verb); | |
5410 | } | |
5411 | if ( argok ) { | |
5412 | if ( start_arg && internal_argval ) { | |
5413 | vFAIL3("Verb pattern '%.*s' may not have an argument", | |
5414 | verb_len, start_verb); | |
5415 | } else if ( argok < 0 && !start_arg ) { | |
5416 | vFAIL3("Verb pattern '%.*s' has a mandatory argument", | |
5417 | verb_len, start_verb); | |
5418 | } else { | |
5419 | ret = reganode(pRExC_state, op, internal_argval); | |
5420 | if ( ! internal_argval && ! SIZE_ONLY ) { | |
5421 | if (start_arg) { | |
5422 | SV *sv = newSVpvn( start_arg, RExC_parse - start_arg); | |
5423 | ARG(ret) = add_data( pRExC_state, 1, "S" ); | |
f8fc2ecf | 5424 | RExC_rxi->data->data[ARG(ret)]=(void*)sv; |
e2e6a0f1 YO |
5425 | ret->flags = 0; |
5426 | } else { | |
5427 | ret->flags = 1; | |
5428 | } | |
5429 | } | |
5430 | } | |
5431 | if (!internal_argval) | |
5432 | RExC_seen |= REG_SEEN_VERBARG; | |
5433 | } else if ( start_arg ) { | |
5434 | vFAIL3("Verb pattern '%.*s' may not have an argument", | |
5435 | verb_len, start_verb); | |
5436 | } else { | |
5437 | ret = reg_node(pRExC_state, op); | |
5438 | } | |
5439 | nextchar(pRExC_state); | |
5440 | return ret; | |
5441 | } else | |
fac92740 | 5442 | if (*RExC_parse == '?') { /* (?...) */ |
6136c704 | 5443 | bool is_logical = 0; |
a28509cc | 5444 | const char * const seqstart = RExC_parse; |
ca9dfc88 | 5445 | |
830247a4 IZ |
5446 | RExC_parse++; |
5447 | paren = *RExC_parse++; | |
c277df42 | 5448 | ret = NULL; /* For look-ahead/behind. */ |
a0d0e21e | 5449 | switch (paren) { |
894be9b7 | 5450 | |
1f1031fe YO |
5451 | case 'P': /* (?P...) variants for those used to PCRE/Python */ |
5452 | paren = *RExC_parse++; | |
5453 | if ( paren == '<') /* (?P<...>) named capture */ | |
5454 | goto named_capture; | |
5455 | else if (paren == '>') { /* (?P>name) named recursion */ | |
5456 | goto named_recursion; | |
5457 | } | |
5458 | else if (paren == '=') { /* (?P=...) named backref */ | |
5459 | /* this pretty much dupes the code for \k<NAME> in regatom(), if | |
5460 | you change this make sure you change that */ | |
5461 | char* name_start = RExC_parse; | |
5462 | U32 num = 0; | |
5463 | SV *sv_dat = reg_scan_name(pRExC_state, | |
5464 | SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA); | |
5465 | if (RExC_parse == name_start || *RExC_parse != ')') | |
5466 | vFAIL2("Sequence %.3s... not terminated",parse_start); | |
5467 | ||
5468 | if (!SIZE_ONLY) { | |
5469 | num = add_data( pRExC_state, 1, "S" ); | |
5470 | RExC_rxi->data->data[num]=(void*)sv_dat; | |
5a5094bd | 5471 | SvREFCNT_inc_simple_void(sv_dat); |
1f1031fe YO |
5472 | } |
5473 | RExC_sawback = 1; | |
5474 | ret = reganode(pRExC_state, | |
5475 | (U8)(FOLD ? (LOC ? NREFFL : NREFF) : NREF), | |
5476 | num); | |
5477 | *flagp |= HASWIDTH; | |
5478 | ||
5479 | Set_Node_Offset(ret, parse_start+1); | |
5480 | Set_Node_Cur_Length(ret); /* MJD */ | |
5481 | ||
5482 | nextchar(pRExC_state); | |
5483 | return ret; | |
5484 | } | |
57b84237 YO |
5485 | RExC_parse++; |
5486 | vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart); | |
5487 | /*NOTREACHED*/ | |
5488 | case '<': /* (?<...) */ | |
b81d288d | 5489 | if (*RExC_parse == '!') |
c277df42 | 5490 | paren = ','; |
0a4db386 | 5491 | else if (*RExC_parse != '=') |
1f1031fe | 5492 | named_capture: |
0a4db386 | 5493 | { /* (?<...>) */ |
81714fb9 | 5494 | char *name_start; |
894be9b7 | 5495 | SV *svname; |
81714fb9 YO |
5496 | paren= '>'; |
5497 | case '\'': /* (?'...') */ | |
5498 | name_start= RExC_parse; | |
0a4db386 YO |
5499 | svname = reg_scan_name(pRExC_state, |
5500 | SIZE_ONLY ? /* reverse test from the others */ | |
5501 | REG_RSN_RETURN_NAME : | |
5502 | REG_RSN_RETURN_NULL); | |
57b84237 YO |
5503 | if (RExC_parse == name_start) { |
5504 | RExC_parse++; | |
5505 | vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart); | |
5506 | /*NOTREACHED*/ | |
5507 | } | |
81714fb9 YO |
5508 | if (*RExC_parse != paren) |
5509 | vFAIL2("Sequence (?%c... not terminated", | |
5510 | paren=='>' ? '<' : paren); | |
5511 | if (SIZE_ONLY) { | |
e62cc96a YO |
5512 | HE *he_str; |
5513 | SV *sv_dat = NULL; | |
894be9b7 YO |
5514 | if (!svname) /* shouldnt happen */ |
5515 | Perl_croak(aTHX_ | |
5516 | "panic: reg_scan_name returned NULL"); | |
81714fb9 YO |
5517 | if (!RExC_paren_names) { |
5518 | RExC_paren_names= newHV(); | |
5519 | sv_2mortal((SV*)RExC_paren_names); | |
1f1031fe YO |
5520 | #ifdef DEBUGGING |
5521 | RExC_paren_name_list= newAV(); | |
5522 | sv_2mortal((SV*)RExC_paren_name_list); | |
5523 | #endif | |
81714fb9 YO |
5524 | } |
5525 | he_str = hv_fetch_ent( RExC_paren_names, svname, 1, 0 ); | |
e62cc96a | 5526 | if ( he_str ) |
81714fb9 | 5527 | sv_dat = HeVAL(he_str); |
e62cc96a | 5528 | if ( ! sv_dat ) { |
81714fb9 | 5529 | /* croak baby croak */ |
e62cc96a YO |
5530 | Perl_croak(aTHX_ |
5531 | "panic: paren_name hash element allocation failed"); | |
5532 | } else if ( SvPOK(sv_dat) ) { | |
76a476f9 YO |
5533 | /* (?|...) can mean we have dupes so scan to check |
5534 | its already been stored. Maybe a flag indicating | |
5535 | we are inside such a construct would be useful, | |
5536 | but the arrays are likely to be quite small, so | |
5537 | for now we punt -- dmq */ | |
5538 | IV count = SvIV(sv_dat); | |
5539 | I32 *pv = (I32*)SvPVX(sv_dat); | |
5540 | IV i; | |
5541 | for ( i = 0 ; i < count ; i++ ) { | |
5542 | if ( pv[i] == RExC_npar ) { | |
5543 | count = 0; | |
5544 | break; | |
5545 | } | |
5546 | } | |
5547 | if ( count ) { | |
5548 | pv = (I32*)SvGROW(sv_dat, SvCUR(sv_dat) + sizeof(I32)+1); | |
5549 | SvCUR_set(sv_dat, SvCUR(sv_dat) + sizeof(I32)); | |
5550 | pv[count] = RExC_npar; | |
5551 | SvIVX(sv_dat)++; | |
5552 | } | |
81714fb9 YO |
5553 | } else { |
5554 | (void)SvUPGRADE(sv_dat,SVt_PVNV); | |
5555 | sv_setpvn(sv_dat, (char *)&(RExC_npar), sizeof(I32)); | |
5556 | SvIOK_on(sv_dat); | |
5557 | SvIVX(sv_dat)= 1; | |
e62cc96a | 5558 | } |
1f1031fe YO |
5559 | #ifdef DEBUGGING |
5560 | if (!av_store(RExC_paren_name_list, RExC_npar, SvREFCNT_inc(svname))) | |
5561 | SvREFCNT_dec(svname); | |
5562 | #endif | |
e62cc96a | 5563 | |
81714fb9 YO |
5564 | /*sv_dump(sv_dat);*/ |
5565 | } | |
5566 | nextchar(pRExC_state); | |
5567 | paren = 1; | |
5568 | goto capturing_parens; | |
5569 | } | |
5570 | RExC_seen |= REG_SEEN_LOOKBEHIND; | |
830247a4 | 5571 | RExC_parse++; |
fac92740 MJD |
5572 | case '=': /* (?=...) */ |
5573 | case '!': /* (?!...) */ | |
830247a4 | 5574 | RExC_seen_zerolen++; |
e2e6a0f1 YO |
5575 | if (*RExC_parse == ')') { |
5576 | ret=reg_node(pRExC_state, OPFAIL); | |
5577 | nextchar(pRExC_state); | |
5578 | return ret; | |
5579 | } | |
594d7033 YO |
5580 | break; |
5581 | case '|': /* (?|...) */ | |
5582 | /* branch reset, behave like a (?:...) except that | |
5583 | buffers in alternations share the same numbers */ | |
5584 | paren = ':'; | |
5585 | after_freeze = freeze_paren = RExC_npar; | |
5586 | break; | |
fac92740 MJD |
5587 | case ':': /* (?:...) */ |
5588 | case '>': /* (?>...) */ | |
a0d0e21e | 5589 | break; |
fac92740 MJD |
5590 | case '$': /* (?$...) */ |
5591 | case '@': /* (?@...) */ | |
8615cb43 | 5592 | vFAIL2("Sequence (?%c...) not implemented", (int)paren); |
a0d0e21e | 5593 | break; |
fac92740 | 5594 | case '#': /* (?#...) */ |
830247a4 IZ |
5595 | while (*RExC_parse && *RExC_parse != ')') |
5596 | RExC_parse++; | |
5597 | if (*RExC_parse != ')') | |
c277df42 | 5598 | FAIL("Sequence (?#... not terminated"); |
830247a4 | 5599 | nextchar(pRExC_state); |
a0d0e21e LW |
5600 | *flagp = TRYAGAIN; |
5601 | return NULL; | |
894be9b7 YO |
5602 | case '0' : /* (?0) */ |
5603 | case 'R' : /* (?R) */ | |
5604 | if (*RExC_parse != ')') | |
6bda09f9 | 5605 | FAIL("Sequence (?R) not terminated"); |
1a147d38 | 5606 | ret = reg_node(pRExC_state, GOSTART); |
a3b492c3 | 5607 | *flagp |= POSTPONED; |
7f69552c YO |
5608 | nextchar(pRExC_state); |
5609 | return ret; | |
5610 | /*notreached*/ | |
894be9b7 YO |
5611 | { /* named and numeric backreferences */ |
5612 | I32 num; | |
894be9b7 YO |
5613 | case '&': /* (?&NAME) */ |
5614 | parse_start = RExC_parse - 1; | |
1f1031fe | 5615 | named_recursion: |
894be9b7 | 5616 | { |
0a4db386 YO |
5617 | SV *sv_dat = reg_scan_name(pRExC_state, |
5618 | SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA); | |
5619 | num = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0; | |
894be9b7 YO |
5620 | } |
5621 | goto gen_recurse_regop; | |
5622 | /* NOT REACHED */ | |
542fa716 YO |
5623 | case '+': |
5624 | if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) { | |
5625 | RExC_parse++; | |
5626 | vFAIL("Illegal pattern"); | |
5627 | } | |
5628 | goto parse_recursion; | |
5629 | /* NOT REACHED*/ | |
5630 | case '-': /* (?-1) */ | |
5631 | if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) { | |
5632 | RExC_parse--; /* rewind to let it be handled later */ | |
5633 | goto parse_flags; | |
5634 | } | |
5635 | /*FALLTHROUGH */ | |
6bda09f9 YO |
5636 | case '1': case '2': case '3': case '4': /* (?1) */ |
5637 | case '5': case '6': case '7': case '8': case '9': | |
5638 | RExC_parse--; | |
542fa716 | 5639 | parse_recursion: |
894be9b7 YO |
5640 | num = atoi(RExC_parse); |
5641 | parse_start = RExC_parse - 1; /* MJD */ | |
542fa716 YO |
5642 | if (*RExC_parse == '-') |
5643 | RExC_parse++; | |
6bda09f9 YO |
5644 | while (isDIGIT(*RExC_parse)) |
5645 | RExC_parse++; | |
5646 | if (*RExC_parse!=')') | |
5647 | vFAIL("Expecting close bracket"); | |
894be9b7 YO |
5648 | |
5649 | gen_recurse_regop: | |
542fa716 YO |
5650 | if ( paren == '-' ) { |
5651 | /* | |
5652 | Diagram of capture buffer numbering. | |
5653 | Top line is the normal capture buffer numbers | |
5654 | Botton line is the negative indexing as from | |
5655 | the X (the (?-2)) | |
5656 | ||
5657 | + 1 2 3 4 5 X 6 7 | |
5658 | /(a(x)y)(a(b(c(?-2)d)e)f)(g(h))/ | |
5659 | - 5 4 3 2 1 X x x | |
5660 | ||
5661 | */ | |
5662 | num = RExC_npar + num; | |
5663 | if (num < 1) { | |
5664 | RExC_parse++; | |
5665 | vFAIL("Reference to nonexistent group"); | |
5666 | } | |
5667 | } else if ( paren == '+' ) { | |
5668 | num = RExC_npar + num - 1; | |
5669 | } | |
5670 | ||
1a147d38 | 5671 | ret = reganode(pRExC_state, GOSUB, num); |
6bda09f9 YO |
5672 | if (!SIZE_ONLY) { |
5673 | if (num > (I32)RExC_rx->nparens) { | |
5674 | RExC_parse++; | |
5675 | vFAIL("Reference to nonexistent group"); | |
5676 | } | |
40d049e4 | 5677 | ARG2L_SET( ret, RExC_recurse_count++); |
6bda09f9 | 5678 | RExC_emit++; |
226de585 | 5679 | DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log, |
acff02b8 | 5680 | "Recurse #%"UVuf" to %"IVdf"\n", (UV)ARG(ret), (IV)ARG2L(ret))); |
894be9b7 | 5681 | } else { |
6bda09f9 | 5682 | RExC_size++; |
6bda09f9 | 5683 | } |
0a4db386 | 5684 | RExC_seen |= REG_SEEN_RECURSE; |
6bda09f9 | 5685 | Set_Node_Length(ret, 1 + regarglen[OP(ret)]); /* MJD */ |
58663417 RGS |
5686 | Set_Node_Offset(ret, parse_start); /* MJD */ |
5687 | ||
a3b492c3 | 5688 | *flagp |= POSTPONED; |
6bda09f9 YO |
5689 | nextchar(pRExC_state); |
5690 | return ret; | |
894be9b7 YO |
5691 | } /* named and numeric backreferences */ |
5692 | /* NOT REACHED */ | |
5693 | ||
fac92740 | 5694 | case '?': /* (??...) */ |
6136c704 | 5695 | is_logical = 1; |
57b84237 YO |
5696 | if (*RExC_parse != '{') { |
5697 | RExC_parse++; | |
5698 | vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart); | |
5699 | /*NOTREACHED*/ | |
5700 | } | |
a3b492c3 | 5701 | *flagp |= POSTPONED; |
830247a4 | 5702 | paren = *RExC_parse++; |
0f5d15d6 | 5703 | /* FALL THROUGH */ |
fac92740 | 5704 | case '{': /* (?{...}) */ |
c277df42 | 5705 | { |
2eccd3b2 NC |
5706 | I32 count = 1; |
5707 | U32 n = 0; | |
c277df42 | 5708 | char c; |
830247a4 | 5709 | char *s = RExC_parse; |
c277df42 | 5710 | |
830247a4 IZ |
5711 | RExC_seen_zerolen++; |
5712 | RExC_seen |= REG_SEEN_EVAL; | |
5713 | while (count && (c = *RExC_parse)) { | |
6136c704 AL |
5714 | if (c == '\\') { |
5715 | if (RExC_parse[1]) | |
5716 | RExC_parse++; | |
5717 | } | |
b81d288d | 5718 | else if (c == '{') |
c277df42 | 5719 | count++; |
b81d288d | 5720 | else if (c == '}') |
c277df42 | 5721 | count--; |
830247a4 | 5722 | RExC_parse++; |
c277df42 | 5723 | } |
6136c704 | 5724 | if (*RExC_parse != ')') { |
b81d288d | 5725 | RExC_parse = s; |
b45f050a JF |
5726 | vFAIL("Sequence (?{...}) not terminated or not {}-balanced"); |
5727 | } | |
c277df42 | 5728 | if (!SIZE_ONLY) { |
f3548bdc | 5729 | PAD *pad; |
6136c704 AL |
5730 | OP_4tree *sop, *rop; |
5731 | SV * const sv = newSVpvn(s, RExC_parse - 1 - s); | |
c277df42 | 5732 | |
569233ed SB |
5733 | ENTER; |
5734 | Perl_save_re_context(aTHX); | |
f3548bdc | 5735 | rop = sv_compile_2op(sv, &sop, "re", &pad); |
9b978d73 DM |
5736 | sop->op_private |= OPpREFCOUNTED; |
5737 | /* re_dup will OpREFCNT_inc */ | |
5738 | OpREFCNT_set(sop, 1); | |
569233ed | 5739 | LEAVE; |
c277df42 | 5740 | |
830247a4 | 5741 | n = add_data(pRExC_state, 3, "nop"); |
f8fc2ecf YO |
5742 | RExC_rxi->data->data[n] = (void*)rop; |
5743 | RExC_rxi->data->data[n+1] = (void*)sop; | |
5744 | RExC_rxi->data->data[n+2] = (void*)pad; | |
c277df42 | 5745 | SvREFCNT_dec(sv); |
a0ed51b3 | 5746 | } |
e24b16f9 | 5747 | else { /* First pass */ |
830247a4 | 5748 | if (PL_reginterp_cnt < ++RExC_seen_evals |
923e4eb5 | 5749 | && IN_PERL_RUNTIME) |
2cd61cdb IZ |
5750 | /* No compiled RE interpolated, has runtime |
5751 | components ===> unsafe. */ | |
5752 | FAIL("Eval-group not allowed at runtime, use re 'eval'"); | |
5b61d3f7 | 5753 | if (PL_tainting && PL_tainted) |
cc6b7395 | 5754 | FAIL("Eval-group in insecure regular expression"); |
54df2634 | 5755 | #if PERL_VERSION > 8 |
923e4eb5 | 5756 | if (IN_PERL_COMPILETIME) |
b5c19bd7 | 5757 | PL_cv_has_eval = 1; |
54df2634 | 5758 | #endif |
c277df42 | 5759 | } |
b5c19bd7 | 5760 | |
830247a4 | 5761 | nextchar(pRExC_state); |
6136c704 | 5762 | if (is_logical) { |
830247a4 | 5763 | ret = reg_node(pRExC_state, LOGICAL); |
0f5d15d6 IZ |
5764 | if (!SIZE_ONLY) |
5765 | ret->flags = 2; | |
3dab1dad | 5766 | REGTAIL(pRExC_state, ret, reganode(pRExC_state, EVAL, n)); |
fac92740 | 5767 | /* deal with the length of this later - MJD */ |
0f5d15d6 IZ |
5768 | return ret; |
5769 | } | |
ccb2c380 MP |
5770 | ret = reganode(pRExC_state, EVAL, n); |
5771 | Set_Node_Length(ret, RExC_parse - parse_start + 1); | |
5772 | Set_Node_Offset(ret, parse_start); | |
5773 | return ret; | |
c277df42 | 5774 | } |
fac92740 | 5775 | case '(': /* (?(?{...})...) and (?(?=...)...) */ |
c277df42 | 5776 | { |
0a4db386 | 5777 | int is_define= 0; |
fac92740 | 5778 | if (RExC_parse[0] == '?') { /* (?(?...)) */ |
b81d288d AB |
5779 | if (RExC_parse[1] == '=' || RExC_parse[1] == '!' |
5780 | || RExC_parse[1] == '<' | |
830247a4 | 5781 | || RExC_parse[1] == '{') { /* Lookahead or eval. */ |
c277df42 IZ |
5782 | I32 flag; |
5783 | ||
830247a4 | 5784 | ret = reg_node(pRExC_state, LOGICAL); |
0f5d15d6 IZ |
5785 | if (!SIZE_ONLY) |
5786 | ret->flags = 1; | |
3dab1dad | 5787 | REGTAIL(pRExC_state, ret, reg(pRExC_state, 1, &flag,depth+1)); |
c277df42 | 5788 | goto insert_if; |
b81d288d | 5789 | } |
a0ed51b3 | 5790 | } |
0a4db386 YO |
5791 | else if ( RExC_parse[0] == '<' /* (?(<NAME>)...) */ |
5792 | || RExC_parse[0] == '\'' ) /* (?('NAME')...) */ | |
5793 | { | |
5794 | char ch = RExC_parse[0] == '<' ? '>' : '\''; | |
5795 | char *name_start= RExC_parse++; | |
2eccd3b2 | 5796 | U32 num = 0; |
0a4db386 YO |
5797 | SV *sv_dat=reg_scan_name(pRExC_state, |
5798 | SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA); | |
5799 | if (RExC_parse == name_start || *RExC_parse != ch) | |
5800 | vFAIL2("Sequence (?(%c... not terminated", | |
5801 | (ch == '>' ? '<' : ch)); | |
5802 | RExC_parse++; | |
5803 | if (!SIZE_ONLY) { | |
5804 | num = add_data( pRExC_state, 1, "S" ); | |
f8fc2ecf | 5805 | RExC_rxi->data->data[num]=(void*)sv_dat; |
5a5094bd | 5806 | SvREFCNT_inc_simple_void(sv_dat); |
0a4db386 YO |
5807 | } |
5808 | ret = reganode(pRExC_state,NGROUPP,num); | |
5809 | goto insert_if_check_paren; | |
5810 | } | |
5811 | else if (RExC_parse[0] == 'D' && | |
5812 | RExC_parse[1] == 'E' && | |
5813 | RExC_parse[2] == 'F' && | |
5814 | RExC_parse[3] == 'I' && | |
5815 | RExC_parse[4] == 'N' && | |
5816 | RExC_parse[5] == 'E') | |
5817 | { | |
5818 | ret = reganode(pRExC_state,DEFINEP,0); | |
5819 | RExC_parse +=6 ; | |
5820 | is_define = 1; | |
5821 | goto insert_if_check_paren; | |
5822 | } | |
5823 | else if (RExC_parse[0] == 'R') { | |
5824 | RExC_parse++; | |
5825 | parno = 0; | |
5826 | if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) { | |
5827 | parno = atoi(RExC_parse++); | |
5828 | while (isDIGIT(*RExC_parse)) | |
5829 | RExC_parse++; | |
5830 | } else if (RExC_parse[0] == '&') { | |
5831 | SV *sv_dat; | |
5832 | RExC_parse++; | |
5833 | sv_dat = reg_scan_name(pRExC_state, | |
5834 | SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA); | |
5835 | parno = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0; | |
5836 | } | |
1a147d38 | 5837 | ret = reganode(pRExC_state,INSUBP,parno); |
0a4db386 YO |
5838 | goto insert_if_check_paren; |
5839 | } | |
830247a4 | 5840 | else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) { |
fac92740 | 5841 | /* (?(1)...) */ |
6136c704 | 5842 | char c; |
830247a4 | 5843 | parno = atoi(RExC_parse++); |
c277df42 | 5844 | |
830247a4 IZ |
5845 | while (isDIGIT(*RExC_parse)) |
5846 | RExC_parse++; | |
fac92740 | 5847 | ret = reganode(pRExC_state, GROUPP, parno); |
2af232bd | 5848 | |
0a4db386 | 5849 | insert_if_check_paren: |
830247a4 | 5850 | if ((c = *nextchar(pRExC_state)) != ')') |
b45f050a | 5851 | vFAIL("Switch condition not recognized"); |
c277df42 | 5852 | insert_if: |
3dab1dad YO |
5853 | REGTAIL(pRExC_state, ret, reganode(pRExC_state, IFTHEN, 0)); |
5854 | br = regbranch(pRExC_state, &flags, 1,depth+1); | |
c277df42 | 5855 | if (br == NULL) |
830247a4 | 5856 | br = reganode(pRExC_state, LONGJMP, 0); |
c277df42 | 5857 | else |
3dab1dad | 5858 | REGTAIL(pRExC_state, br, reganode(pRExC_state, LONGJMP, 0)); |
830247a4 | 5859 | c = *nextchar(pRExC_state); |
d1b80229 IZ |
5860 | if (flags&HASWIDTH) |
5861 | *flagp |= HASWIDTH; | |
c277df42 | 5862 | if (c == '|') { |
0a4db386 YO |
5863 | if (is_define) |
5864 | vFAIL("(?(DEFINE)....) does not allow branches"); | |
830247a4 | 5865 | lastbr = reganode(pRExC_state, IFTHEN, 0); /* Fake one for optimizer. */ |
3dab1dad YO |
5866 | regbranch(pRExC_state, &flags, 1,depth+1); |
5867 | REGTAIL(pRExC_state, ret, lastbr); | |
d1b80229 IZ |
5868 | if (flags&HASWIDTH) |
5869 | *flagp |= HASWIDTH; | |
830247a4 | 5870 | c = *nextchar(pRExC_state); |
a0ed51b3 LW |
5871 | } |
5872 | else | |
c277df42 IZ |
5873 | lastbr = NULL; |
5874 | if (c != ')') | |
8615cb43 | 5875 | vFAIL("Switch (?(condition)... contains too many branches"); |
830247a4 | 5876 | ender = reg_node(pRExC_state, TAIL); |
3dab1dad | 5877 | REGTAIL(pRExC_state, br, ender); |
c277df42 | 5878 | if (lastbr) { |
3dab1dad YO |
5879 | REGTAIL(pRExC_state, lastbr, ender); |
5880 | REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender); | |
a0ed51b3 LW |
5881 | } |
5882 | else | |
3dab1dad | 5883 | REGTAIL(pRExC_state, ret, ender); |
3b57cd43 YO |
5884 | RExC_size++; /* XXX WHY do we need this?!! |
5885 | For large programs it seems to be required | |
5886 | but I can't figure out why. -- dmq*/ | |
c277df42 | 5887 | return ret; |
a0ed51b3 LW |
5888 | } |
5889 | else { | |
830247a4 | 5890 | vFAIL2("Unknown switch condition (?(%.2s", RExC_parse); |
c277df42 IZ |
5891 | } |
5892 | } | |
1b1626e4 | 5893 | case 0: |
830247a4 | 5894 | RExC_parse--; /* for vFAIL to print correctly */ |
8615cb43 | 5895 | vFAIL("Sequence (? incomplete"); |
1b1626e4 | 5896 | break; |
a0d0e21e | 5897 | default: |
cde0cee5 YO |
5898 | --RExC_parse; |
5899 | parse_flags: /* (?i) */ | |
5900 | { | |
5901 | U32 posflags = 0, negflags = 0; | |
5902 | U32 *flagsp = &posflags; | |
5903 | ||
5904 | while (*RExC_parse) { | |
5905 | /* && strchr("iogcmsx", *RExC_parse) */ | |
9d1d55b5 JP |
5906 | /* (?g), (?gc) and (?o) are useless here |
5907 | and must be globally applied -- japhy */ | |
cde0cee5 YO |
5908 | switch (*RExC_parse) { |
5909 | CASE_STD_PMMOD_FLAGS_PARSE_SET(flagsp); | |
f7819f85 A |
5910 | case ONCE_PAT_MOD: /* 'o' */ |
5911 | case GLOBAL_PAT_MOD: /* 'g' */ | |
9d1d55b5 | 5912 | if (SIZE_ONLY && ckWARN(WARN_REGEXP)) { |
6136c704 | 5913 | const I32 wflagbit = *RExC_parse == 'o' ? WASTED_O : WASTED_G; |
9d1d55b5 JP |
5914 | if (! (wastedflags & wflagbit) ) { |
5915 | wastedflags |= wflagbit; | |
5916 | vWARN5( | |
5917 | RExC_parse + 1, | |
5918 | "Useless (%s%c) - %suse /%c modifier", | |
5919 | flagsp == &negflags ? "?-" : "?", | |
5920 | *RExC_parse, | |
5921 | flagsp == &negflags ? "don't " : "", | |
5922 | *RExC_parse | |
5923 | ); | |
5924 | } | |
5925 | } | |
cde0cee5 YO |
5926 | break; |
5927 | ||
f7819f85 | 5928 | case CONTINUE_PAT_MOD: /* 'c' */ |
9d1d55b5 | 5929 | if (SIZE_ONLY && ckWARN(WARN_REGEXP)) { |
6136c704 AL |
5930 | if (! (wastedflags & WASTED_C) ) { |
5931 | wastedflags |= WASTED_GC; | |
9d1d55b5 JP |
5932 | vWARN3( |
5933 | RExC_parse + 1, | |
5934 | "Useless (%sc) - %suse /gc modifier", | |
5935 | flagsp == &negflags ? "?-" : "?", | |
5936 | flagsp == &negflags ? "don't " : "" | |
5937 | ); | |
5938 | } | |
5939 | } | |
cde0cee5 | 5940 | break; |
f7819f85 | 5941 | case KEEPCOPY_PAT_MOD: /* 'p' */ |
cde0cee5 YO |
5942 | if (flagsp == &negflags) { |
5943 | if (SIZE_ONLY && ckWARN(WARN_REGEXP)) | |
f7819f85 | 5944 | vWARN(RExC_parse + 1,"Useless use of (?-p)"); |
cde0cee5 YO |
5945 | } else { |
5946 | *flagsp |= RXf_PMf_KEEPCOPY; | |
5947 | } | |
5948 | break; | |
5949 | case '-': | |
57b84237 YO |
5950 | if (flagsp == &negflags) { |
5951 | RExC_parse++; | |
5952 | vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart); | |
5953 | /*NOTREACHED*/ | |
5954 | } | |
cde0cee5 YO |
5955 | flagsp = &negflags; |
5956 | wastedflags = 0; /* reset so (?g-c) warns twice */ | |
5957 | break; | |
5958 | case ':': | |
5959 | paren = ':'; | |
5960 | /*FALLTHROUGH*/ | |
5961 | case ')': | |
5962 | RExC_flags |= posflags; | |
5963 | RExC_flags &= ~negflags; | |
f7819f85 A |
5964 | if (paren != ':') { |
5965 | oregflags |= posflags; | |
5966 | oregflags &= ~negflags; | |
5967 | } | |
cde0cee5 YO |
5968 | nextchar(pRExC_state); |
5969 | if (paren != ':') { | |
5970 | *flagp = TRYAGAIN; | |
5971 | return NULL; | |
5972 | } else { | |
5973 | ret = NULL; | |
5974 | goto parse_rest; | |
5975 | } | |
5976 | /*NOTREACHED*/ | |
5977 | default: | |
cde0cee5 YO |
5978 | RExC_parse++; |
5979 | vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart); | |
5980 | /*NOTREACHED*/ | |
5981 | } | |
830247a4 | 5982 | ++RExC_parse; |
48c036b1 | 5983 | } |
cde0cee5 | 5984 | }} /* one for the default block, one for the switch */ |
a0d0e21e | 5985 | } |
fac92740 | 5986 | else { /* (...) */ |
81714fb9 | 5987 | capturing_parens: |
830247a4 IZ |
5988 | parno = RExC_npar; |
5989 | RExC_npar++; | |
e2e6a0f1 | 5990 | |
830247a4 | 5991 | ret = reganode(pRExC_state, OPEN, parno); |
e2e6a0f1 YO |
5992 | if (!SIZE_ONLY ){ |
5993 | if (!RExC_nestroot) | |
5994 | RExC_nestroot = parno; | |
c009da3d YO |
5995 | if (RExC_seen & REG_SEEN_RECURSE |
5996 | && !RExC_open_parens[parno-1]) | |
5997 | { | |
e2e6a0f1 | 5998 | DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log, |
40d049e4 YO |
5999 | "Setting open paren #%"IVdf" to %d\n", |
6000 | (IV)parno, REG_NODE_NUM(ret))); | |
e2e6a0f1 YO |
6001 | RExC_open_parens[parno-1]= ret; |
6002 | } | |
6bda09f9 | 6003 | } |
fac92740 MJD |
6004 | Set_Node_Length(ret, 1); /* MJD */ |
6005 | Set_Node_Offset(ret, RExC_parse); /* MJD */ | |
6136c704 | 6006 | is_open = 1; |
a0d0e21e | 6007 | } |
a0ed51b3 | 6008 | } |
fac92740 | 6009 | else /* ! paren */ |
a0d0e21e | 6010 | ret = NULL; |
cde0cee5 YO |
6011 | |
6012 | parse_rest: | |
a0d0e21e | 6013 | /* Pick up the branches, linking them together. */ |
fac92740 | 6014 | parse_start = RExC_parse; /* MJD */ |
3dab1dad | 6015 | br = regbranch(pRExC_state, &flags, 1,depth+1); |
fac92740 | 6016 | /* branch_len = (paren != 0); */ |
2af232bd | 6017 | |
a0d0e21e LW |
6018 | if (br == NULL) |
6019 | return(NULL); | |
830247a4 IZ |
6020 | if (*RExC_parse == '|') { |
6021 | if (!SIZE_ONLY && RExC_extralen) { | |
6bda09f9 | 6022 | reginsert(pRExC_state, BRANCHJ, br, depth+1); |
a0ed51b3 | 6023 | } |
fac92740 | 6024 | else { /* MJD */ |
6bda09f9 | 6025 | reginsert(pRExC_state, BRANCH, br, depth+1); |
fac92740 MJD |
6026 | Set_Node_Length(br, paren != 0); |
6027 | Set_Node_Offset_To_R(br-RExC_emit_start, parse_start-RExC_start); | |
6028 | } | |
c277df42 IZ |
6029 | have_branch = 1; |
6030 | if (SIZE_ONLY) | |
830247a4 | 6031 | RExC_extralen += 1; /* For BRANCHJ-BRANCH. */ |
a0ed51b3 LW |
6032 | } |
6033 | else if (paren == ':') { | |
c277df42 IZ |
6034 | *flagp |= flags&SIMPLE; |
6035 | } | |
6136c704 | 6036 | if (is_open) { /* Starts with OPEN. */ |
3dab1dad | 6037 | REGTAIL(pRExC_state, ret, br); /* OPEN -> first. */ |
a0ed51b3 LW |
6038 | } |
6039 | else if (paren != '?') /* Not Conditional */ | |
a0d0e21e | 6040 | ret = br; |
8ae10a67 | 6041 | *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED); |
c277df42 | 6042 | lastbr = br; |
830247a4 IZ |
6043 | while (*RExC_parse == '|') { |
6044 | if (!SIZE_ONLY && RExC_extralen) { | |
6045 | ender = reganode(pRExC_state, LONGJMP,0); | |
3dab1dad | 6046 | REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender); /* Append to the previous. */ |
c277df42 IZ |
6047 | } |
6048 | if (SIZE_ONLY) | |
830247a4 IZ |
6049 | RExC_extralen += 2; /* Account for LONGJMP. */ |
6050 | nextchar(pRExC_state); | |
594d7033 YO |
6051 | if (freeze_paren) { |
6052 | if (RExC_npar > after_freeze) | |
6053 | after_freeze = RExC_npar; | |
6054 | RExC_npar = freeze_paren; | |
6055 | } | |
3dab1dad | 6056 | br = regbranch(pRExC_state, &flags, 0, depth+1); |
2af232bd | 6057 | |
a687059c | 6058 | if (br == NULL) |
a0d0e21e | 6059 | return(NULL); |
3dab1dad | 6060 | REGTAIL(pRExC_state, lastbr, br); /* BRANCH -> BRANCH. */ |
c277df42 | 6061 | lastbr = br; |
8ae10a67 | 6062 | *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED); |
a0d0e21e LW |
6063 | } |
6064 | ||
c277df42 IZ |
6065 | if (have_branch || paren != ':') { |
6066 | /* Make a closing node, and hook it on the end. */ | |
6067 | switch (paren) { | |
6068 | case ':': | |
830247a4 | 6069 | ender = reg_node(pRExC_state, TAIL); |
c277df42 IZ |
6070 | break; |
6071 | case 1: | |
830247a4 | 6072 | ender = reganode(pRExC_state, CLOSE, parno); |
40d049e4 YO |
6073 | if (!SIZE_ONLY && RExC_seen & REG_SEEN_RECURSE) { |
6074 | DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log, | |
6075 | "Setting close paren #%"IVdf" to %d\n", | |
6076 | (IV)parno, REG_NODE_NUM(ender))); | |
6077 | RExC_close_parens[parno-1]= ender; | |
e2e6a0f1 YO |
6078 | if (RExC_nestroot == parno) |
6079 | RExC_nestroot = 0; | |
40d049e4 | 6080 | } |
fac92740 MJD |
6081 | Set_Node_Offset(ender,RExC_parse+1); /* MJD */ |
6082 | Set_Node_Length(ender,1); /* MJD */ | |
c277df42 IZ |
6083 | break; |
6084 | case '<': | |
c277df42 IZ |
6085 | case ',': |
6086 | case '=': | |
6087 | case '!': | |
c277df42 | 6088 | *flagp &= ~HASWIDTH; |
821b33a5 IZ |
6089 | /* FALL THROUGH */ |
6090 | case '>': | |
830247a4 | 6091 | ender = reg_node(pRExC_state, SUCCEED); |
c277df42 IZ |
6092 | break; |
6093 | case 0: | |
830247a4 | 6094 | ender = reg_node(pRExC_state, END); |
40d049e4 YO |
6095 | if (!SIZE_ONLY) { |
6096 | assert(!RExC_opend); /* there can only be one! */ | |
6097 | RExC_opend = ender; | |
6098 | } | |
c277df42 IZ |
6099 | break; |
6100 | } | |
eaf3ca90 | 6101 | REGTAIL(pRExC_state, lastbr, ender); |
a0d0e21e | 6102 | |
9674d46a | 6103 | if (have_branch && !SIZE_ONLY) { |
eaf3ca90 YO |
6104 | if (depth==1) |
6105 | RExC_seen |= REG_TOP_LEVEL_BRANCHES; | |
6106 | ||
c277df42 | 6107 | /* Hook the tails of the branches to the closing node. */ |
9674d46a AL |
6108 | for (br = ret; br; br = regnext(br)) { |
6109 | const U8 op = PL_regkind[OP(br)]; | |
6110 | if (op == BRANCH) { | |
07be1b83 | 6111 | REGTAIL_STUDY(pRExC_state, NEXTOPER(br), ender); |
9674d46a AL |
6112 | } |
6113 | else if (op == BRANCHJ) { | |
07be1b83 | 6114 | REGTAIL_STUDY(pRExC_state, NEXTOPER(NEXTOPER(br)), ender); |
9674d46a | 6115 | } |
c277df42 IZ |
6116 | } |
6117 | } | |
a0d0e21e | 6118 | } |
c277df42 IZ |
6119 | |
6120 | { | |
e1ec3a88 AL |
6121 | const char *p; |
6122 | static const char parens[] = "=!<,>"; | |
c277df42 IZ |
6123 | |
6124 | if (paren && (p = strchr(parens, paren))) { | |
eb160463 | 6125 | U8 node = ((p - parens) % 2) ? UNLESSM : IFMATCH; |
c277df42 IZ |
6126 | int flag = (p - parens) > 1; |
6127 | ||
6128 | if (paren == '>') | |
6129 | node = SUSPEND, flag = 0; | |
6bda09f9 | 6130 | reginsert(pRExC_state, node,ret, depth+1); |
45948336 EP |
6131 | Set_Node_Cur_Length(ret); |
6132 | Set_Node_Offset(ret, parse_start + 1); | |
c277df42 | 6133 | ret->flags = flag; |
07be1b83 | 6134 | REGTAIL_STUDY(pRExC_state, ret, reg_node(pRExC_state, TAIL)); |
c277df42 | 6135 | } |
a0d0e21e LW |
6136 | } |
6137 | ||
6138 | /* Check for proper termination. */ | |
ce3e6498 | 6139 | if (paren) { |
e2509266 | 6140 | RExC_flags = oregflags; |
830247a4 IZ |
6141 | if (RExC_parse >= RExC_end || *nextchar(pRExC_state) != ')') { |
6142 | RExC_parse = oregcomp_parse; | |
380a0633 | 6143 | vFAIL("Unmatched ("); |
ce3e6498 | 6144 | } |
a0ed51b3 | 6145 | } |
830247a4 IZ |
6146 | else if (!paren && RExC_parse < RExC_end) { |
6147 | if (*RExC_parse == ')') { | |
6148 | RExC_parse++; | |
380a0633 | 6149 | vFAIL("Unmatched )"); |
a0ed51b3 LW |
6150 | } |
6151 | else | |
b45f050a | 6152 | FAIL("Junk on end of regexp"); /* "Can't happen". */ |
a0d0e21e LW |
6153 | /* NOTREACHED */ |
6154 | } | |
594d7033 YO |
6155 | if (after_freeze) |
6156 | RExC_npar = after_freeze; | |
a0d0e21e | 6157 | return(ret); |
a687059c LW |
6158 | } |
6159 | ||
6160 | /* | |
6161 | - regbranch - one alternative of an | operator | |
6162 | * | |
6163 | * Implements the concatenation operator. | |
6164 | */ | |
76e3520e | 6165 | STATIC regnode * |
3dab1dad | 6166 | S_regbranch(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, I32 first, U32 depth) |
a687059c | 6167 | { |
97aff369 | 6168 | dVAR; |
c277df42 IZ |
6169 | register regnode *ret; |
6170 | register regnode *chain = NULL; | |
6171 | register regnode *latest; | |
6172 | I32 flags = 0, c = 0; | |
3dab1dad YO |
6173 | GET_RE_DEBUG_FLAGS_DECL; |
6174 | DEBUG_PARSE("brnc"); | |
02daf0ab | 6175 | |
b81d288d | 6176 | if (first) |
c277df42 IZ |
6177 | ret = NULL; |
6178 | else { | |
b81d288d | 6179 | if (!SIZE_ONLY && RExC_extralen) |
830247a4 | 6180 | ret = reganode(pRExC_state, BRANCHJ,0); |
fac92740 | 6181 | else { |
830247a4 | 6182 | ret = reg_node(pRExC_state, BRANCH); |
fac92740 MJD |
6183 | Set_Node_Length(ret, 1); |
6184 | } | |
c277df42 IZ |
6185 | } |
6186 | ||
b81d288d | 6187 | if (!first && SIZE_ONLY) |
830247a4 | 6188 | RExC_extralen += 1; /* BRANCHJ */ |
b81d288d | 6189 | |
c277df42 | 6190 | *flagp = WORST; /* Tentatively. */ |
a0d0e21e | 6191 | |
830247a4 IZ |
6192 | RExC_parse--; |
6193 | nextchar(pRExC_state); | |
6194 | while (RExC_parse < RExC_end && *RExC_parse != '|' && *RExC_parse != ')') { | |
a0d0e21e | 6195 | flags &= ~TRYAGAIN; |
3dab1dad | 6196 | latest = regpiece(pRExC_state, &flags,depth+1); |
a0d0e21e LW |
6197 | if (latest == NULL) { |
6198 | if (flags & TRYAGAIN) | |
6199 | continue; | |
6200 | return(NULL); | |
a0ed51b3 LW |
6201 | } |
6202 | else if (ret == NULL) | |
c277df42 | 6203 | ret = latest; |
8ae10a67 | 6204 | *flagp |= flags&(HASWIDTH|POSTPONED); |
c277df42 | 6205 | if (chain == NULL) /* First piece. */ |
a0d0e21e LW |
6206 | *flagp |= flags&SPSTART; |
6207 | else { | |
830247a4 | 6208 | RExC_naughty++; |
3dab1dad | 6209 | REGTAIL(pRExC_state, chain, latest); |
a687059c | 6210 | } |
a0d0e21e | 6211 | chain = latest; |
c277df42 IZ |
6212 | c++; |
6213 | } | |
6214 | if (chain == NULL) { /* Loop ran zero times. */ | |
830247a4 | 6215 | chain = reg_node(pRExC_state, NOTHING); |
c277df42 IZ |
6216 | if (ret == NULL) |
6217 | ret = chain; | |
6218 | } | |
6219 | if (c == 1) { | |
6220 | *flagp |= flags&SIMPLE; | |
a0d0e21e | 6221 | } |
a687059c | 6222 | |
d4c19fe8 | 6223 | return ret; |
a687059c LW |
6224 | } |
6225 | ||
6226 | /* | |
6227 | - regpiece - something followed by possible [*+?] | |
6228 | * | |
6229 | * Note that the branching code sequences used for ? and the general cases | |
6230 | * of * and + are somewhat optimized: they use the same NOTHING node as | |
6231 | * both the endmarker for their branch list and the body of the last branch. | |
6232 | * It might seem that this node could be dispensed with entirely, but the | |
6233 | * endmarker role is not redundant. | |
6234 | */ | |
76e3520e | 6235 | STATIC regnode * |
3dab1dad | 6236 | S_regpiece(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth) |
a687059c | 6237 | { |
97aff369 | 6238 | dVAR; |
c277df42 | 6239 | register regnode *ret; |
a0d0e21e LW |
6240 | register char op; |
6241 | register char *next; | |
6242 | I32 flags; | |
1df70142 | 6243 | const char * const origparse = RExC_parse; |
a0d0e21e | 6244 | I32 min; |
c277df42 | 6245 | I32 max = REG_INFTY; |
fac92740 | 6246 | char *parse_start; |
10edeb5d | 6247 | const char *maxpos = NULL; |
3dab1dad YO |
6248 | GET_RE_DEBUG_FLAGS_DECL; |
6249 | DEBUG_PARSE("piec"); | |
a0d0e21e | 6250 | |
3dab1dad | 6251 | ret = regatom(pRExC_state, &flags,depth+1); |
a0d0e21e LW |
6252 | if (ret == NULL) { |
6253 | if (flags & TRYAGAIN) | |
6254 | *flagp |= TRYAGAIN; | |
6255 | return(NULL); | |
6256 | } | |
6257 | ||
830247a4 | 6258 | op = *RExC_parse; |
a0d0e21e | 6259 | |
830247a4 | 6260 | if (op == '{' && regcurly(RExC_parse)) { |
10edeb5d | 6261 | maxpos = NULL; |
fac92740 | 6262 | parse_start = RExC_parse; /* MJD */ |
830247a4 | 6263 | next = RExC_parse + 1; |
a0d0e21e LW |
6264 | while (isDIGIT(*next) || *next == ',') { |
6265 | if (*next == ',') { | |
6266 | if (maxpos) | |
6267 | break; | |
6268 | else | |
6269 | maxpos = next; | |
a687059c | 6270 | } |
a0d0e21e LW |
6271 | next++; |
6272 | } | |
6273 | if (*next == '}') { /* got one */ | |
6274 | if (!maxpos) | |
6275 | maxpos = next; | |
830247a4 IZ |
6276 | RExC_parse++; |
6277 | min = atoi(RExC_parse); | |
a0d0e21e LW |
6278 | if (*maxpos == ',') |
6279 | maxpos++; | |
6280 | else | |
830247a4 | 6281 | maxpos = RExC_parse; |
a0d0e21e LW |
6282 | max = atoi(maxpos); |
6283 | if (!max && *maxpos != '0') | |
c277df42 IZ |
6284 | max = REG_INFTY; /* meaning "infinity" */ |
6285 | else if (max >= REG_INFTY) | |
8615cb43 | 6286 | vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1); |
830247a4 IZ |
6287 | RExC_parse = next; |
6288 | nextchar(pRExC_state); | |
a0d0e21e LW |
6289 | |
6290 | do_curly: | |
6291 | if ((flags&SIMPLE)) { | |
830247a4 | 6292 | RExC_naughty += 2 + RExC_naughty / 2; |
6bda09f9 | 6293 | reginsert(pRExC_state, CURLY, ret, depth+1); |
fac92740 MJD |
6294 | Set_Node_Offset(ret, parse_start+1); /* MJD */ |
6295 | Set_Node_Cur_Length(ret); | |
a0d0e21e LW |
6296 | } |
6297 | else { | |
3dab1dad | 6298 | regnode * const w = reg_node(pRExC_state, WHILEM); |
2c2d71f5 JH |
6299 | |
6300 | w->flags = 0; | |
3dab1dad | 6301 | REGTAIL(pRExC_state, ret, w); |
830247a4 | 6302 | if (!SIZE_ONLY && RExC_extralen) { |
6bda09f9 YO |
6303 | reginsert(pRExC_state, LONGJMP,ret, depth+1); |
6304 | reginsert(pRExC_state, NOTHING,ret, depth+1); | |
c277df42 IZ |
6305 | NEXT_OFF(ret) = 3; /* Go over LONGJMP. */ |
6306 | } | |
6bda09f9 | 6307 | reginsert(pRExC_state, CURLYX,ret, depth+1); |
fac92740 MJD |
6308 | /* MJD hk */ |
6309 | Set_Node_Offset(ret, parse_start+1); | |
2af232bd | 6310 | Set_Node_Length(ret, |
fac92740 | 6311 | op == '{' ? (RExC_parse - parse_start) : 1); |
2af232bd | 6312 | |
830247a4 | 6313 | if (!SIZE_ONLY && RExC_extralen) |
c277df42 | 6314 | NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */ |
3dab1dad | 6315 | REGTAIL(pRExC_state, ret, reg_node(pRExC_state, NOTHING)); |
c277df42 | 6316 | if (SIZE_ONLY) |
830247a4 IZ |
6317 | RExC_whilem_seen++, RExC_extralen += 3; |
6318 | RExC_naughty += 4 + RExC_naughty; /* compound interest */ | |
a0d0e21e | 6319 | } |
c277df42 | 6320 | ret->flags = 0; |
a0d0e21e LW |
6321 | |
6322 | if (min > 0) | |
821b33a5 IZ |
6323 | *flagp = WORST; |
6324 | if (max > 0) | |
6325 | *flagp |= HASWIDTH; | |
a0d0e21e | 6326 | if (max && max < min) |
8615cb43 | 6327 | vFAIL("Can't do {n,m} with n > m"); |
c277df42 | 6328 | if (!SIZE_ONLY) { |
eb160463 GS |
6329 | ARG1_SET(ret, (U16)min); |
6330 | ARG2_SET(ret, (U16)max); | |
a687059c | 6331 | } |
a687059c | 6332 | |
a0d0e21e | 6333 | goto nest_check; |
a687059c | 6334 | } |
a0d0e21e | 6335 | } |
a687059c | 6336 | |
a0d0e21e LW |
6337 | if (!ISMULT1(op)) { |
6338 | *flagp = flags; | |
a687059c | 6339 | return(ret); |
a0d0e21e | 6340 | } |
bb20fd44 | 6341 | |
c277df42 | 6342 | #if 0 /* Now runtime fix should be reliable. */ |
b45f050a JF |
6343 | |
6344 | /* if this is reinstated, don't forget to put this back into perldiag: | |
6345 | ||
6346 | =item Regexp *+ operand could be empty at {#} in regex m/%s/ | |
6347 | ||
6348 | (F) The part of the regexp subject to either the * or + quantifier | |
6349 | could match an empty string. The {#} shows in the regular | |
6350 | expression about where the problem was discovered. | |
6351 | ||
6352 | */ | |
6353 | ||
bb20fd44 | 6354 | if (!(flags&HASWIDTH) && op != '?') |
b45f050a | 6355 | vFAIL("Regexp *+ operand could be empty"); |
b81d288d | 6356 | #endif |
bb20fd44 | 6357 | |
fac92740 | 6358 | parse_start = RExC_parse; |
830247a4 | 6359 | nextchar(pRExC_state); |
a0d0e21e | 6360 | |
821b33a5 | 6361 | *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH); |
a0d0e21e LW |
6362 | |
6363 | if (op == '*' && (flags&SIMPLE)) { | |
6bda09f9 | 6364 | reginsert(pRExC_state, STAR, ret, depth+1); |
c277df42 | 6365 | ret->flags = 0; |
830247a4 | 6366 | RExC_naughty += 4; |
a0d0e21e LW |
6367 | } |
6368 | else if (op == '*') { | |
6369 | min = 0; | |
6370 | goto do_curly; | |
a0ed51b3 LW |
6371 | } |
6372 | else if (op == '+' && (flags&SIMPLE)) { | |
6bda09f9 | 6373 | reginsert(pRExC_state, PLUS, ret, depth+1); |
c277df42 | 6374 | ret->flags = 0; |
830247a4 | 6375 | RExC_naughty += 3; |
a0d0e21e LW |
6376 | } |
6377 | else if (op == '+') { | |
6378 | min = 1; | |
6379 | goto do_curly; | |
a0ed51b3 LW |
6380 | } |
6381 | else if (op == '?') { | |
a0d0e21e LW |
6382 | min = 0; max = 1; |
6383 | goto do_curly; | |
6384 | } | |
6385 | nest_check: | |
a3b492c3 | 6386 | if (!SIZE_ONLY && !(flags&(HASWIDTH|POSTPONED)) && max > REG_INFTY/3 && ckWARN(WARN_REGEXP)) { |
830247a4 | 6387 | vWARN3(RExC_parse, |
b45f050a | 6388 | "%.*s matches null string many times", |
afd78fd5 | 6389 | (int)(RExC_parse >= origparse ? RExC_parse - origparse : 0), |
b45f050a | 6390 | origparse); |
a0d0e21e LW |
6391 | } |
6392 | ||
b9b4dddf | 6393 | if (RExC_parse < RExC_end && *RExC_parse == '?') { |
830247a4 | 6394 | nextchar(pRExC_state); |
6bda09f9 | 6395 | reginsert(pRExC_state, MINMOD, ret, depth+1); |
3dab1dad | 6396 | REGTAIL(pRExC_state, ret, ret + NODE_STEP_REGNODE); |
a0d0e21e | 6397 | } |
b9b4dddf YO |
6398 | #ifndef REG_ALLOW_MINMOD_SUSPEND |
6399 | else | |
6400 | #endif | |
6401 | if (RExC_parse < RExC_end && *RExC_parse == '+') { | |
6402 | regnode *ender; | |
6403 | nextchar(pRExC_state); | |
6404 | ender = reg_node(pRExC_state, SUCCEED); | |
6405 | REGTAIL(pRExC_state, ret, ender); | |
6406 | reginsert(pRExC_state, SUSPEND, ret, depth+1); | |
6407 | ret->flags = 0; | |
6408 | ender = reg_node(pRExC_state, TAIL); | |
6409 | REGTAIL(pRExC_state, ret, ender); | |
6410 | /*ret= ender;*/ | |
6411 | } | |
6412 | ||
6413 | if (RExC_parse < RExC_end && ISMULT2(RExC_parse)) { | |
830247a4 | 6414 | RExC_parse++; |
b45f050a JF |
6415 | vFAIL("Nested quantifiers"); |
6416 | } | |
a0d0e21e LW |
6417 | |
6418 | return(ret); | |
a687059c LW |
6419 | } |
6420 | ||
fc8cd66c YO |
6421 | |
6422 | /* reg_namedseq(pRExC_state,UVp) | |
6423 | ||
6424 | This is expected to be called by a parser routine that has | |
6425 | recognized'\N' and needs to handle the rest. RExC_parse is | |
6426 | expected to point at the first char following the N at the time | |
6427 | of the call. | |
6428 | ||
6429 | If valuep is non-null then it is assumed that we are parsing inside | |
6430 | of a charclass definition and the first codepoint in the resolved | |
6431 | string is returned via *valuep and the routine will return NULL. | |
6432 | In this mode if a multichar string is returned from the charnames | |
6433 | handler a warning will be issued, and only the first char in the | |
6434 | sequence will be examined. If the string returned is zero length | |
6435 | then the value of *valuep is undefined and NON-NULL will | |
6436 | be returned to indicate failure. (This will NOT be a valid pointer | |
6437 | to a regnode.) | |
6438 | ||
6439 | If value is null then it is assumed that we are parsing normal text | |
6440 | and inserts a new EXACT node into the program containing the resolved | |
6441 | string and returns a pointer to the new node. If the string is | |
6442 | zerolength a NOTHING node is emitted. | |
6443 | ||
6444 | On success RExC_parse is set to the char following the endbrace. | |
6445 | Parsing failures will generate a fatal errorvia vFAIL(...) | |
6446 | ||
6447 | NOTE: We cache all results from the charnames handler locally in | |
6448 | the RExC_charnames hash (created on first use) to prevent a charnames | |
6449 | handler from playing silly-buggers and returning a short string and | |
6450 | then a long string for a given pattern. Since the regexp program | |
6451 | size is calculated during an initial parse this would result | |
6452 | in a buffer overrun so we cache to prevent the charname result from | |
6453 | changing during the course of the parse. | |
6454 | ||
6455 | */ | |
6456 | STATIC regnode * | |
6457 | S_reg_namedseq(pTHX_ RExC_state_t *pRExC_state, UV *valuep) | |
6458 | { | |
6459 | char * name; /* start of the content of the name */ | |
6460 | char * endbrace; /* endbrace following the name */ | |
6461 | SV *sv_str = NULL; | |
6462 | SV *sv_name = NULL; | |
6463 | STRLEN len; /* this has various purposes throughout the code */ | |
6464 | bool cached = 0; /* if this is true then we shouldn't refcount dev sv_str */ | |
6465 | regnode *ret = NULL; | |
6466 | ||
6467 | if (*RExC_parse != '{') { | |
6468 | vFAIL("Missing braces on \\N{}"); | |
6469 | } | |
6470 | name = RExC_parse+1; | |
6471 | endbrace = strchr(RExC_parse, '}'); | |
6472 | if ( ! endbrace ) { | |
6473 | RExC_parse++; | |
6474 | vFAIL("Missing right brace on \\N{}"); | |
6475 | } | |
6476 | RExC_parse = endbrace + 1; | |
6477 | ||
6478 | ||
6479 | /* RExC_parse points at the beginning brace, | |
6480 | endbrace points at the last */ | |
6481 | if ( name[0]=='U' && name[1]=='+' ) { | |
38a44b82 | 6482 | /* its a "Unicode hex" notation {U+89AB} */ |
fc8cd66c YO |
6483 | I32 fl = PERL_SCAN_ALLOW_UNDERSCORES |
6484 | | PERL_SCAN_DISALLOW_PREFIX | |
6485 | | (SIZE_ONLY ? PERL_SCAN_SILENT_ILLDIGIT : 0); | |
6486 | UV cp; | |
196f1508 | 6487 | len = (STRLEN)(endbrace - name - 2); |
fc8cd66c | 6488 | cp = grok_hex(name + 2, &len, &fl, NULL); |
196f1508 | 6489 | if ( len != (STRLEN)(endbrace - name - 2) ) { |
fc8cd66c YO |
6490 | cp = 0xFFFD; |
6491 | } | |
6492 | if (cp > 0xff) | |
6493 | RExC_utf8 = 1; | |
6494 | if ( valuep ) { | |
6495 | *valuep = cp; | |
6496 | return NULL; | |
6497 | } | |
6498 | sv_str= Perl_newSVpvf_nocontext("%c",(int)cp); | |
6499 | } else { | |
6500 | /* fetch the charnames handler for this scope */ | |
6501 | HV * const table = GvHV(PL_hintgv); | |
6502 | SV **cvp= table ? | |
6503 | hv_fetchs(table, "charnames", FALSE) : | |
6504 | NULL; | |
6505 | SV *cv= cvp ? *cvp : NULL; | |
6506 | HE *he_str; | |
6507 | int count; | |
6508 | /* create an SV with the name as argument */ | |
6509 | sv_name = newSVpvn(name, endbrace - name); | |
6510 | ||
6511 | if (!table || !(PL_hints & HINT_LOCALIZE_HH)) { | |
6512 | vFAIL2("Constant(\\N{%s}) unknown: " | |
6513 | "(possibly a missing \"use charnames ...\")", | |
6514 | SvPVX(sv_name)); | |
6515 | } | |
6516 | if (!cvp || !SvOK(*cvp)) { /* when $^H{charnames} = undef; */ | |
6517 | vFAIL2("Constant(\\N{%s}): " | |
6518 | "$^H{charnames} is not defined",SvPVX(sv_name)); | |
6519 | } | |
6520 | ||
6521 | ||
6522 | ||
6523 | if (!RExC_charnames) { | |
6524 | /* make sure our cache is allocated */ | |
6525 | RExC_charnames = newHV(); | |
6bda09f9 | 6526 | sv_2mortal((SV*)RExC_charnames); |
fc8cd66c YO |
6527 | } |
6528 | /* see if we have looked this one up before */ | |
6529 | he_str = hv_fetch_ent( RExC_charnames, sv_name, 0, 0 ); | |
6530 | if ( he_str ) { | |
6531 | sv_str = HeVAL(he_str); | |
6532 | cached = 1; | |
6533 | } else { | |
6534 | dSP ; | |
6535 | ||
6536 | ENTER ; | |
6537 | SAVETMPS ; | |
6538 | PUSHMARK(SP) ; | |
6539 | ||
6540 | XPUSHs(sv_name); | |
6541 | ||
6542 | PUTBACK ; | |
6543 | ||
6544 | count= call_sv(cv, G_SCALAR); | |
6545 | ||
6546 | if (count == 1) { /* XXXX is this right? dmq */ | |
6547 | sv_str = POPs; | |
6548 | SvREFCNT_inc_simple_void(sv_str); | |
6549 | } | |
6550 | ||
6551 | SPAGAIN ; | |
6552 | PUTBACK ; | |
6553 | FREETMPS ; | |
6554 | LEAVE ; | |
6555 | ||
6556 | if ( !sv_str || !SvOK(sv_str) ) { | |
6557 | vFAIL2("Constant(\\N{%s}): Call to &{$^H{charnames}} " | |
6558 | "did not return a defined value",SvPVX(sv_name)); | |
6559 | } | |
6560 | if (hv_store_ent( RExC_charnames, sv_name, sv_str, 0)) | |
6561 | cached = 1; | |
6562 | } | |
6563 | } | |
6564 | if (valuep) { | |
6565 | char *p = SvPV(sv_str, len); | |
6566 | if (len) { | |
6567 | STRLEN numlen = 1; | |
6568 | if ( SvUTF8(sv_str) ) { | |
196f1508 | 6569 | *valuep = utf8_to_uvchr((U8*)p, &numlen); |
fc8cd66c YO |
6570 | if (*valuep > 0x7F) |
6571 | RExC_utf8 = 1; | |
6572 | /* XXXX | |
6573 | We have to turn on utf8 for high bit chars otherwise | |
6574 | we get failures with | |
6575 | ||
6576 | "ss" =~ /[\N{LATIN SMALL LETTER SHARP S}]/i | |
6577 | "SS" =~ /[\N{LATIN SMALL LETTER SHARP S}]/i | |
6578 | ||
6579 | This is different from what \x{} would do with the same | |
6580 | codepoint, where the condition is > 0xFF. | |
6581 | - dmq | |
6582 | */ | |
6583 | ||
6584 | ||
6585 | } else { | |
6586 | *valuep = (UV)*p; | |
6587 | /* warn if we havent used the whole string? */ | |
6588 | } | |
6589 | if (numlen<len && SIZE_ONLY && ckWARN(WARN_REGEXP)) { | |
6590 | vWARN2(RExC_parse, | |
6591 | "Ignoring excess chars from \\N{%s} in character class", | |
6592 | SvPVX(sv_name) | |
6593 | ); | |
6594 | } | |
6595 | } else if (SIZE_ONLY && ckWARN(WARN_REGEXP)) { | |
6596 | vWARN2(RExC_parse, | |
6597 | "Ignoring zero length \\N{%s} in character class", | |
6598 | SvPVX(sv_name) | |
6599 | ); | |
6600 | } | |
6601 | if (sv_name) | |
6602 | SvREFCNT_dec(sv_name); | |
6603 | if (!cached) | |
6604 | SvREFCNT_dec(sv_str); | |
6605 | return len ? NULL : (regnode *)&len; | |
6606 | } else if(SvCUR(sv_str)) { | |
6607 | ||
6608 | char *s; | |
6609 | char *p, *pend; | |
6610 | STRLEN charlen = 1; | |
d008bc60 | 6611 | #ifdef DEBUGGING |
fc8cd66c | 6612 | char * parse_start = name-3; /* needed for the offsets */ |
d008bc60 | 6613 | #endif |
fc8cd66c YO |
6614 | GET_RE_DEBUG_FLAGS_DECL; /* needed for the offsets */ |
6615 | ||
6616 | ret = reg_node(pRExC_state, | |
6617 | (U8)(FOLD ? (LOC ? EXACTFL : EXACTF) : EXACT)); | |
6618 | s= STRING(ret); | |
6619 | ||
6620 | if ( RExC_utf8 && !SvUTF8(sv_str) ) { | |
6621 | sv_utf8_upgrade(sv_str); | |
6622 | } else if ( !RExC_utf8 && SvUTF8(sv_str) ) { | |
6623 | RExC_utf8= 1; | |
6624 | } | |
6625 | ||
6626 | p = SvPV(sv_str, len); | |
6627 | pend = p + len; | |
6628 | /* len is the length written, charlen is the size the char read */ | |
6629 | for ( len = 0; p < pend; p += charlen ) { | |
6630 | if (UTF) { | |
196f1508 | 6631 | UV uvc = utf8_to_uvchr((U8*)p, &charlen); |
fc8cd66c YO |
6632 | if (FOLD) { |
6633 | STRLEN foldlen,numlen; | |
6634 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1], *foldbuf; | |
6635 | uvc = toFOLD_uni(uvc, tmpbuf, &foldlen); | |
6636 | /* Emit all the Unicode characters. */ | |
6637 | ||
6638 | for (foldbuf = tmpbuf; | |
6639 | foldlen; | |
6640 | foldlen -= numlen) | |
6641 | { | |
6642 | uvc = utf8_to_uvchr(foldbuf, &numlen); | |
6643 | if (numlen > 0) { | |
6644 | const STRLEN unilen = reguni(pRExC_state, uvc, s); | |
6645 | s += unilen; | |
6646 | len += unilen; | |
6647 | /* In EBCDIC the numlen | |
6648 | * and unilen can differ. */ | |
6649 | foldbuf += numlen; | |
6650 | if (numlen >= foldlen) | |
6651 | break; | |
6652 | } | |
6653 | else | |
6654 | break; /* "Can't happen." */ | |
6655 | } | |
6656 | } else { | |
6657 | const STRLEN unilen = reguni(pRExC_state, uvc, s); | |
6658 | if (unilen > 0) { | |
6659 | s += unilen; | |
6660 | len += unilen; | |
6661 | } | |
6662 | } | |
6663 | } else { | |
6664 | len++; | |
6665 | REGC(*p, s++); | |
6666 | } | |
6667 | } | |
6668 | if (SIZE_ONLY) { | |
6669 | RExC_size += STR_SZ(len); | |
6670 | } else { | |
6671 | STR_LEN(ret) = len; | |
6672 | RExC_emit += STR_SZ(len); | |
6673 | } | |
6674 | Set_Node_Cur_Length(ret); /* MJD */ | |
6675 | RExC_parse--; | |
6676 | nextchar(pRExC_state); | |
6677 | } else { | |
6678 | ret = reg_node(pRExC_state,NOTHING); | |
6679 | } | |
6680 | if (!cached) { | |
6681 | SvREFCNT_dec(sv_str); | |
6682 | } | |
6683 | if (sv_name) { | |
6684 | SvREFCNT_dec(sv_name); | |
6685 | } | |
6686 | return ret; | |
6687 | ||
6688 | } | |
6689 | ||
6690 | ||
9e08bc66 TS |
6691 | /* |
6692 | * reg_recode | |
6693 | * | |
6694 | * It returns the code point in utf8 for the value in *encp. | |
6695 | * value: a code value in the source encoding | |
6696 | * encp: a pointer to an Encode object | |
6697 | * | |
6698 | * If the result from Encode is not a single character, | |
6699 | * it returns U+FFFD (Replacement character) and sets *encp to NULL. | |
6700 | */ | |
6701 | STATIC UV | |
6702 | S_reg_recode(pTHX_ const char value, SV **encp) | |
6703 | { | |
6704 | STRLEN numlen = 1; | |
6705 | SV * const sv = sv_2mortal(newSVpvn(&value, numlen)); | |
c86f7df5 | 6706 | const char * const s = *encp ? sv_recode_to_utf8(sv, *encp) : SvPVX(sv); |
9e08bc66 TS |
6707 | const STRLEN newlen = SvCUR(sv); |
6708 | UV uv = UNICODE_REPLACEMENT; | |
6709 | ||
6710 | if (newlen) | |
6711 | uv = SvUTF8(sv) | |
6712 | ? utf8n_to_uvchr((U8*)s, newlen, &numlen, UTF8_ALLOW_DEFAULT) | |
6713 | : *(U8*)s; | |
6714 | ||
6715 | if (!newlen || numlen != newlen) { | |
6716 | uv = UNICODE_REPLACEMENT; | |
c86f7df5 | 6717 | *encp = NULL; |
9e08bc66 TS |
6718 | } |
6719 | return uv; | |
6720 | } | |
6721 | ||
fc8cd66c | 6722 | |
a687059c LW |
6723 | /* |
6724 | - regatom - the lowest level | |
ee9b8eae YO |
6725 | |
6726 | Try to identify anything special at the start of the pattern. If there | |
6727 | is, then handle it as required. This may involve generating a single regop, | |
6728 | such as for an assertion; or it may involve recursing, such as to | |
6729 | handle a () structure. | |
6730 | ||
6731 | If the string doesn't start with something special then we gobble up | |
6732 | as much literal text as we can. | |
6733 | ||
6734 | Once we have been able to handle whatever type of thing started the | |
6735 | sequence, we return. | |
6736 | ||
6737 | Note: we have to be careful with escapes, as they can be both literal | |
6738 | and special, and in the case of \10 and friends can either, depending | |
6739 | on context. Specifically there are two seperate switches for handling | |
6740 | escape sequences, with the one for handling literal escapes requiring | |
6741 | a dummy entry for all of the special escapes that are actually handled | |
6742 | by the other. | |
6743 | */ | |
6744 | ||
76e3520e | 6745 | STATIC regnode * |
3dab1dad | 6746 | S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth) |
a687059c | 6747 | { |
97aff369 | 6748 | dVAR; |
cbbf8932 | 6749 | register regnode *ret = NULL; |
a0d0e21e | 6750 | I32 flags; |
45948336 | 6751 | char *parse_start = RExC_parse; |
3dab1dad YO |
6752 | GET_RE_DEBUG_FLAGS_DECL; |
6753 | DEBUG_PARSE("atom"); | |
a0d0e21e LW |
6754 | *flagp = WORST; /* Tentatively. */ |
6755 | ||
ee9b8eae | 6756 | |
a0d0e21e | 6757 | tryagain: |
f9a79580 | 6758 | switch ((U8)*RExC_parse) { |
a0d0e21e | 6759 | case '^': |
830247a4 IZ |
6760 | RExC_seen_zerolen++; |
6761 | nextchar(pRExC_state); | |
bbe252da | 6762 | if (RExC_flags & RXf_PMf_MULTILINE) |
830247a4 | 6763 | ret = reg_node(pRExC_state, MBOL); |
bbe252da | 6764 | else if (RExC_flags & RXf_PMf_SINGLELINE) |
830247a4 | 6765 | ret = reg_node(pRExC_state, SBOL); |
a0d0e21e | 6766 | else |
830247a4 | 6767 | ret = reg_node(pRExC_state, BOL); |
fac92740 | 6768 | Set_Node_Length(ret, 1); /* MJD */ |
a0d0e21e LW |
6769 | break; |
6770 | case '$': | |
830247a4 | 6771 | nextchar(pRExC_state); |
b81d288d | 6772 | if (*RExC_parse) |
830247a4 | 6773 | RExC_seen_zerolen++; |
bbe252da | 6774 | if (RExC_flags & RXf_PMf_MULTILINE) |
830247a4 | 6775 | ret = reg_node(pRExC_state, MEOL); |
bbe252da | 6776 | else if (RExC_flags & RXf_PMf_SINGLELINE) |
830247a4 | 6777 | ret = reg_node(pRExC_state, SEOL); |
a0d0e21e | 6778 | else |
830247a4 | 6779 | ret = reg_node(pRExC_state, EOL); |
fac92740 | 6780 | Set_Node_Length(ret, 1); /* MJD */ |
a0d0e21e LW |
6781 | break; |
6782 | case '.': | |
830247a4 | 6783 | nextchar(pRExC_state); |
bbe252da | 6784 | if (RExC_flags & RXf_PMf_SINGLELINE) |
ffc61ed2 JH |
6785 | ret = reg_node(pRExC_state, SANY); |
6786 | else | |
6787 | ret = reg_node(pRExC_state, REG_ANY); | |
6788 | *flagp |= HASWIDTH|SIMPLE; | |
830247a4 | 6789 | RExC_naughty++; |
fac92740 | 6790 | Set_Node_Length(ret, 1); /* MJD */ |
a0d0e21e LW |
6791 | break; |
6792 | case '[': | |
b45f050a | 6793 | { |
3dab1dad YO |
6794 | char * const oregcomp_parse = ++RExC_parse; |
6795 | ret = regclass(pRExC_state,depth+1); | |
830247a4 IZ |
6796 | if (*RExC_parse != ']') { |
6797 | RExC_parse = oregcomp_parse; | |
b45f050a JF |
6798 | vFAIL("Unmatched ["); |
6799 | } | |
830247a4 | 6800 | nextchar(pRExC_state); |
a0d0e21e | 6801 | *flagp |= HASWIDTH|SIMPLE; |
fac92740 | 6802 | Set_Node_Length(ret, RExC_parse - oregcomp_parse + 1); /* MJD */ |
a0d0e21e | 6803 | break; |
b45f050a | 6804 | } |
a0d0e21e | 6805 | case '(': |
830247a4 | 6806 | nextchar(pRExC_state); |
3dab1dad | 6807 | ret = reg(pRExC_state, 1, &flags,depth+1); |
a0d0e21e | 6808 | if (ret == NULL) { |
bf93d4cc | 6809 | if (flags & TRYAGAIN) { |
830247a4 | 6810 | if (RExC_parse == RExC_end) { |
bf93d4cc GS |
6811 | /* Make parent create an empty node if needed. */ |
6812 | *flagp |= TRYAGAIN; | |
6813 | return(NULL); | |
6814 | } | |
a0d0e21e | 6815 | goto tryagain; |
bf93d4cc | 6816 | } |
a0d0e21e LW |
6817 | return(NULL); |
6818 | } | |
a3b492c3 | 6819 | *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE|POSTPONED); |
a0d0e21e LW |
6820 | break; |
6821 | case '|': | |
6822 | case ')': | |
6823 | if (flags & TRYAGAIN) { | |
6824 | *flagp |= TRYAGAIN; | |
6825 | return NULL; | |
6826 | } | |
b45f050a | 6827 | vFAIL("Internal urp"); |
a0d0e21e LW |
6828 | /* Supposed to be caught earlier. */ |
6829 | break; | |
85afd4ae | 6830 | case '{': |
830247a4 IZ |
6831 | if (!regcurly(RExC_parse)) { |
6832 | RExC_parse++; | |
85afd4ae CS |
6833 | goto defchar; |
6834 | } | |
6835 | /* FALL THROUGH */ | |
a0d0e21e LW |
6836 | case '?': |
6837 | case '+': | |
6838 | case '*': | |
830247a4 | 6839 | RExC_parse++; |
b45f050a | 6840 | vFAIL("Quantifier follows nothing"); |
a0d0e21e | 6841 | break; |
f9a79580 RGS |
6842 | case 0xDF: |
6843 | case 0xC3: | |
6844 | case 0xCE: | |
56d400ed | 6845 | if (!LOC && FOLD) { |
e64b1bd1 | 6846 | U32 len,cp; |
7cf3a6a3 | 6847 | len=0; /* silence a spurious compiler warning */ |
56d400ed | 6848 | if ((cp = what_len_TRICKYFOLD_safe(RExC_parse,RExC_end,UTF,len))) { |
e64b1bd1 YO |
6849 | *flagp |= HASWIDTH; /* could be SIMPLE too, but needs a handler in regexec.regrepeat */ |
6850 | RExC_parse+=len-1; /* we get one from nextchar() as well. :-( */ | |
6851 | ret = reganode(pRExC_state, FOLDCHAR, cp); | |
6852 | Set_Node_Length(ret, 1); /* MJD */ | |
6853 | nextchar(pRExC_state); /* kill whitespace under /x */ | |
6854 | return ret; | |
6855 | } | |
6856 | } | |
6857 | goto outer_default; | |
a0d0e21e | 6858 | case '\\': |
ee9b8eae YO |
6859 | /* Special Escapes |
6860 | ||
6861 | This switch handles escape sequences that resolve to some kind | |
6862 | of special regop and not to literal text. Escape sequnces that | |
6863 | resolve to literal text are handled below in the switch marked | |
6864 | "Literal Escapes". | |
6865 | ||
6866 | Every entry in this switch *must* have a corresponding entry | |
6867 | in the literal escape switch. However, the opposite is not | |
6868 | required, as the default for this switch is to jump to the | |
6869 | literal text handling code. | |
6870 | */ | |
830247a4 | 6871 | switch (*++RExC_parse) { |
ee9b8eae | 6872 | /* Special Escapes */ |
a0d0e21e | 6873 | case 'A': |
830247a4 IZ |
6874 | RExC_seen_zerolen++; |
6875 | ret = reg_node(pRExC_state, SBOL); | |
a0d0e21e | 6876 | *flagp |= SIMPLE; |
ee9b8eae | 6877 | goto finish_meta_pat; |
a0d0e21e | 6878 | case 'G': |
830247a4 IZ |
6879 | ret = reg_node(pRExC_state, GPOS); |
6880 | RExC_seen |= REG_SEEN_GPOS; | |
a0d0e21e | 6881 | *flagp |= SIMPLE; |
ee9b8eae YO |
6882 | goto finish_meta_pat; |
6883 | case 'K': | |
6884 | RExC_seen_zerolen++; | |
6885 | ret = reg_node(pRExC_state, KEEPS); | |
6886 | *flagp |= SIMPLE; | |
6887 | goto finish_meta_pat; | |
a0d0e21e | 6888 | case 'Z': |
830247a4 | 6889 | ret = reg_node(pRExC_state, SEOL); |
a0d0e21e | 6890 | *flagp |= SIMPLE; |
a1917ab9 | 6891 | RExC_seen_zerolen++; /* Do not optimize RE away */ |
ee9b8eae | 6892 | goto finish_meta_pat; |
b85d18e9 | 6893 | case 'z': |
830247a4 | 6894 | ret = reg_node(pRExC_state, EOS); |
b85d18e9 | 6895 | *flagp |= SIMPLE; |
830247a4 | 6896 | RExC_seen_zerolen++; /* Do not optimize RE away */ |
ee9b8eae | 6897 | goto finish_meta_pat; |
4a2d328f | 6898 | case 'C': |
f33976b4 DB |
6899 | ret = reg_node(pRExC_state, CANY); |
6900 | RExC_seen |= REG_SEEN_CANY; | |
a0ed51b3 | 6901 | *flagp |= HASWIDTH|SIMPLE; |
ee9b8eae | 6902 | goto finish_meta_pat; |
a0ed51b3 | 6903 | case 'X': |
830247a4 | 6904 | ret = reg_node(pRExC_state, CLUMP); |
a0ed51b3 | 6905 | *flagp |= HASWIDTH; |
ee9b8eae | 6906 | goto finish_meta_pat; |
a0d0e21e | 6907 | case 'w': |
eb160463 | 6908 | ret = reg_node(pRExC_state, (U8)(LOC ? ALNUML : ALNUM)); |
a0d0e21e | 6909 | *flagp |= HASWIDTH|SIMPLE; |
ee9b8eae | 6910 | goto finish_meta_pat; |
a0d0e21e | 6911 | case 'W': |
eb160463 | 6912 | ret = reg_node(pRExC_state, (U8)(LOC ? NALNUML : NALNUM)); |
a0d0e21e | 6913 | *flagp |= HASWIDTH|SIMPLE; |
ee9b8eae | 6914 | goto finish_meta_pat; |
a0d0e21e | 6915 | case 'b': |
830247a4 IZ |
6916 | RExC_seen_zerolen++; |
6917 | RExC_seen |= REG_SEEN_LOOKBEHIND; | |
eb160463 | 6918 | ret = reg_node(pRExC_state, (U8)(LOC ? BOUNDL : BOUND)); |
a0d0e21e | 6919 | *flagp |= SIMPLE; |
ee9b8eae | 6920 | goto finish_meta_pat; |
a0d0e21e | 6921 | case 'B': |
830247a4 IZ |
6922 | RExC_seen_zerolen++; |
6923 | RExC_seen |= REG_SEEN_LOOKBEHIND; | |
eb160463 | 6924 | ret = reg_node(pRExC_state, (U8)(LOC ? NBOUNDL : NBOUND)); |
a0d0e21e | 6925 | *flagp |= SIMPLE; |
ee9b8eae | 6926 | goto finish_meta_pat; |
a0d0e21e | 6927 | case 's': |
eb160463 | 6928 | ret = reg_node(pRExC_state, (U8)(LOC ? SPACEL : SPACE)); |
a0d0e21e | 6929 | *flagp |= HASWIDTH|SIMPLE; |
ee9b8eae | 6930 | goto finish_meta_pat; |
a0d0e21e | 6931 | case 'S': |
eb160463 | 6932 | ret = reg_node(pRExC_state, (U8)(LOC ? NSPACEL : NSPACE)); |
a0d0e21e | 6933 | *flagp |= HASWIDTH|SIMPLE; |
ee9b8eae | 6934 | goto finish_meta_pat; |
a0d0e21e | 6935 | case 'd': |
ffc61ed2 | 6936 | ret = reg_node(pRExC_state, DIGIT); |
a0d0e21e | 6937 | *flagp |= HASWIDTH|SIMPLE; |
ee9b8eae | 6938 | goto finish_meta_pat; |
a0d0e21e | 6939 | case 'D': |
ffc61ed2 | 6940 | ret = reg_node(pRExC_state, NDIGIT); |
a0d0e21e | 6941 | *flagp |= HASWIDTH|SIMPLE; |
ee9b8eae | 6942 | goto finish_meta_pat; |
e1d1eefb YO |
6943 | case 'R': |
6944 | ret = reg_node(pRExC_state, LNBREAK); | |
6945 | *flagp |= HASWIDTH|SIMPLE; | |
6946 | goto finish_meta_pat; | |
6947 | case 'h': | |
6948 | ret = reg_node(pRExC_state, HORIZWS); | |
6949 | *flagp |= HASWIDTH|SIMPLE; | |
6950 | goto finish_meta_pat; | |
6951 | case 'H': | |
6952 | ret = reg_node(pRExC_state, NHORIZWS); | |
6953 | *flagp |= HASWIDTH|SIMPLE; | |
6954 | goto finish_meta_pat; | |
ee9b8eae | 6955 | case 'v': |
e1d1eefb YO |
6956 | ret = reg_node(pRExC_state, VERTWS); |
6957 | *flagp |= HASWIDTH|SIMPLE; | |
ee9b8eae YO |
6958 | goto finish_meta_pat; |
6959 | case 'V': | |
e1d1eefb YO |
6960 | ret = reg_node(pRExC_state, NVERTWS); |
6961 | *flagp |= HASWIDTH|SIMPLE; | |
ee9b8eae | 6962 | finish_meta_pat: |
830247a4 | 6963 | nextchar(pRExC_state); |
fac92740 | 6964 | Set_Node_Length(ret, 2); /* MJD */ |
ee9b8eae | 6965 | break; |
a14b48bc LW |
6966 | case 'p': |
6967 | case 'P': | |
3568d838 | 6968 | { |
3dab1dad | 6969 | char* const oldregxend = RExC_end; |
d008bc60 | 6970 | #ifdef DEBUGGING |
ccb2c380 | 6971 | char* parse_start = RExC_parse - 2; |
d008bc60 | 6972 | #endif |
a14b48bc | 6973 | |
830247a4 | 6974 | if (RExC_parse[1] == '{') { |
3568d838 | 6975 | /* a lovely hack--pretend we saw [\pX] instead */ |
830247a4 IZ |
6976 | RExC_end = strchr(RExC_parse, '}'); |
6977 | if (!RExC_end) { | |
3dab1dad | 6978 | const U8 c = (U8)*RExC_parse; |
830247a4 IZ |
6979 | RExC_parse += 2; |
6980 | RExC_end = oldregxend; | |
0da60cf5 | 6981 | vFAIL2("Missing right brace on \\%c{}", c); |
b45f050a | 6982 | } |
830247a4 | 6983 | RExC_end++; |
a14b48bc | 6984 | } |
af6f566e | 6985 | else { |
830247a4 | 6986 | RExC_end = RExC_parse + 2; |
af6f566e HS |
6987 | if (RExC_end > oldregxend) |
6988 | RExC_end = oldregxend; | |
6989 | } | |
830247a4 | 6990 | RExC_parse--; |
a14b48bc | 6991 | |
3dab1dad | 6992 | ret = regclass(pRExC_state,depth+1); |
a14b48bc | 6993 | |
830247a4 IZ |
6994 | RExC_end = oldregxend; |
6995 | RExC_parse--; | |
ccb2c380 MP |
6996 | |
6997 | Set_Node_Offset(ret, parse_start + 2); | |
6998 | Set_Node_Cur_Length(ret); | |
830247a4 | 6999 | nextchar(pRExC_state); |
a14b48bc LW |
7000 | *flagp |= HASWIDTH|SIMPLE; |
7001 | } | |
7002 | break; | |
fc8cd66c YO |
7003 | case 'N': |
7004 | /* Handle \N{NAME} here and not below because it can be | |
7005 | multicharacter. join_exact() will join them up later on. | |
7006 | Also this makes sure that things like /\N{BLAH}+/ and | |
7007 | \N{BLAH} being multi char Just Happen. dmq*/ | |
7008 | ++RExC_parse; | |
7009 | ret= reg_namedseq(pRExC_state, NULL); | |
7010 | break; | |
0a4db386 | 7011 | case 'k': /* Handle \k<NAME> and \k'NAME' */ |
1f1031fe | 7012 | parse_named_seq: |
81714fb9 YO |
7013 | { |
7014 | char ch= RExC_parse[1]; | |
1f1031fe YO |
7015 | if (ch != '<' && ch != '\'' && ch != '{') { |
7016 | RExC_parse++; | |
7017 | vFAIL2("Sequence %.2s... not terminated",parse_start); | |
81714fb9 | 7018 | } else { |
1f1031fe YO |
7019 | /* this pretty much dupes the code for (?P=...) in reg(), if |
7020 | you change this make sure you change that */ | |
81714fb9 | 7021 | char* name_start = (RExC_parse += 2); |
2eccd3b2 | 7022 | U32 num = 0; |
0a4db386 YO |
7023 | SV *sv_dat = reg_scan_name(pRExC_state, |
7024 | SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA); | |
1f1031fe | 7025 | ch= (ch == '<') ? '>' : (ch == '{') ? '}' : '\''; |
81714fb9 | 7026 | if (RExC_parse == name_start || *RExC_parse != ch) |
1f1031fe YO |
7027 | vFAIL2("Sequence %.3s... not terminated",parse_start); |
7028 | ||
7029 | if (!SIZE_ONLY) { | |
7030 | num = add_data( pRExC_state, 1, "S" ); | |
7031 | RExC_rxi->data->data[num]=(void*)sv_dat; | |
5a5094bd | 7032 | SvREFCNT_inc_simple_void(sv_dat); |
1f1031fe YO |
7033 | } |
7034 | ||
81714fb9 YO |
7035 | RExC_sawback = 1; |
7036 | ret = reganode(pRExC_state, | |
7037 | (U8)(FOLD ? (LOC ? NREFFL : NREFF) : NREF), | |
7038 | num); | |
7039 | *flagp |= HASWIDTH; | |
1f1031fe | 7040 | |
81714fb9 YO |
7041 | /* override incorrect value set in reganode MJD */ |
7042 | Set_Node_Offset(ret, parse_start+1); | |
7043 | Set_Node_Cur_Length(ret); /* MJD */ | |
7044 | nextchar(pRExC_state); | |
1f1031fe | 7045 | |
81714fb9 YO |
7046 | } |
7047 | break; | |
1f1031fe | 7048 | } |
2bf803e2 | 7049 | case 'g': |
a0d0e21e LW |
7050 | case '1': case '2': case '3': case '4': |
7051 | case '5': case '6': case '7': case '8': case '9': | |
7052 | { | |
c74340f9 | 7053 | I32 num; |
2bf803e2 YO |
7054 | bool isg = *RExC_parse == 'g'; |
7055 | bool isrel = 0; | |
7056 | bool hasbrace = 0; | |
7057 | if (isg) { | |
c74340f9 | 7058 | RExC_parse++; |
2bf803e2 YO |
7059 | if (*RExC_parse == '{') { |
7060 | RExC_parse++; | |
7061 | hasbrace = 1; | |
7062 | } | |
7063 | if (*RExC_parse == '-') { | |
7064 | RExC_parse++; | |
7065 | isrel = 1; | |
7066 | } | |
1f1031fe YO |
7067 | if (hasbrace && !isDIGIT(*RExC_parse)) { |
7068 | if (isrel) RExC_parse--; | |
7069 | RExC_parse -= 2; | |
7070 | goto parse_named_seq; | |
7071 | } } | |
c74340f9 | 7072 | num = atoi(RExC_parse); |
b72d83b2 RGS |
7073 | if (isg && num == 0) |
7074 | vFAIL("Reference to invalid group 0"); | |
c74340f9 | 7075 | if (isrel) { |
5624f11d | 7076 | num = RExC_npar - num; |
c74340f9 YO |
7077 | if (num < 1) |
7078 | vFAIL("Reference to nonexistent or unclosed group"); | |
7079 | } | |
2bf803e2 | 7080 | if (!isg && num > 9 && num >= RExC_npar) |
a0d0e21e LW |
7081 | goto defchar; |
7082 | else { | |
3dab1dad | 7083 | char * const parse_start = RExC_parse - 1; /* MJD */ |
830247a4 IZ |
7084 | while (isDIGIT(*RExC_parse)) |
7085 | RExC_parse++; | |
1f1031fe YO |
7086 | if (parse_start == RExC_parse - 1) |
7087 | vFAIL("Unterminated \\g... pattern"); | |
2bf803e2 YO |
7088 | if (hasbrace) { |
7089 | if (*RExC_parse != '}') | |
7090 | vFAIL("Unterminated \\g{...} pattern"); | |
7091 | RExC_parse++; | |
7092 | } | |
c74340f9 YO |
7093 | if (!SIZE_ONLY) { |
7094 | if (num > (I32)RExC_rx->nparens) | |
7095 | vFAIL("Reference to nonexistent group"); | |
c74340f9 | 7096 | } |
830247a4 | 7097 | RExC_sawback = 1; |
eb160463 GS |
7098 | ret = reganode(pRExC_state, |
7099 | (U8)(FOLD ? (LOC ? REFFL : REFF) : REF), | |
7100 | num); | |
a0d0e21e | 7101 | *flagp |= HASWIDTH; |
2af232bd | 7102 | |
fac92740 | 7103 | /* override incorrect value set in reganode MJD */ |
2af232bd | 7104 | Set_Node_Offset(ret, parse_start+1); |
fac92740 | 7105 | Set_Node_Cur_Length(ret); /* MJD */ |
830247a4 IZ |
7106 | RExC_parse--; |
7107 | nextchar(pRExC_state); | |
a0d0e21e LW |
7108 | } |
7109 | } | |
7110 | break; | |
7111 | case '\0': | |
830247a4 | 7112 | if (RExC_parse >= RExC_end) |
b45f050a | 7113 | FAIL("Trailing \\"); |
a0d0e21e LW |
7114 | /* FALL THROUGH */ |
7115 | default: | |
a0288114 | 7116 | /* Do not generate "unrecognized" warnings here, we fall |
c9f97d15 | 7117 | back into the quick-grab loop below */ |
45948336 | 7118 | parse_start--; |
a0d0e21e LW |
7119 | goto defchar; |
7120 | } | |
7121 | break; | |
4633a7c4 LW |
7122 | |
7123 | case '#': | |
bbe252da | 7124 | if (RExC_flags & RXf_PMf_EXTENDED) { |
bcdf7404 | 7125 | if ( reg_skipcomment( pRExC_state ) ) |
4633a7c4 LW |
7126 | goto tryagain; |
7127 | } | |
7128 | /* FALL THROUGH */ | |
7129 | ||
f9a79580 RGS |
7130 | default: |
7131 | outer_default:{ | |
ba210ebe | 7132 | register STRLEN len; |
58ae7d3f | 7133 | register UV ender; |
a0d0e21e | 7134 | register char *p; |
3dab1dad | 7135 | char *s; |
80aecb99 | 7136 | STRLEN foldlen; |
89ebb4a3 | 7137 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1], *foldbuf; |
f06dbbb7 JH |
7138 | |
7139 | parse_start = RExC_parse - 1; | |
a0d0e21e | 7140 | |
830247a4 | 7141 | RExC_parse++; |
a0d0e21e LW |
7142 | |
7143 | defchar: | |
58ae7d3f | 7144 | ender = 0; |
eb160463 GS |
7145 | ret = reg_node(pRExC_state, |
7146 | (U8)(FOLD ? (LOC ? EXACTFL : EXACTF) : EXACT)); | |
cd439c50 | 7147 | s = STRING(ret); |
830247a4 IZ |
7148 | for (len = 0, p = RExC_parse - 1; |
7149 | len < 127 && p < RExC_end; | |
a0d0e21e LW |
7150 | len++) |
7151 | { | |
3dab1dad | 7152 | char * const oldp = p; |
5b5a24f7 | 7153 | |
bbe252da | 7154 | if (RExC_flags & RXf_PMf_EXTENDED) |
bcdf7404 | 7155 | p = regwhite( pRExC_state, p ); |
f9a79580 RGS |
7156 | switch ((U8)*p) { |
7157 | case 0xDF: | |
7158 | case 0xC3: | |
7159 | case 0xCE: | |
56d400ed | 7160 | if (LOC || !FOLD || !is_TRICKYFOLD_safe(p,RExC_end,UTF)) |
f9a79580 | 7161 | goto normal_default; |
a0d0e21e LW |
7162 | case '^': |
7163 | case '$': | |
7164 | case '.': | |
7165 | case '[': | |
7166 | case '(': | |
7167 | case ')': | |
7168 | case '|': | |
7169 | goto loopdone; | |
7170 | case '\\': | |
ee9b8eae YO |
7171 | /* Literal Escapes Switch |
7172 | ||
7173 | This switch is meant to handle escape sequences that | |
7174 | resolve to a literal character. | |
7175 | ||
7176 | Every escape sequence that represents something | |
7177 | else, like an assertion or a char class, is handled | |
7178 | in the switch marked 'Special Escapes' above in this | |
7179 | routine, but also has an entry here as anything that | |
7180 | isn't explicitly mentioned here will be treated as | |
7181 | an unescaped equivalent literal. | |
7182 | */ | |
7183 | ||
a0d0e21e | 7184 | switch (*++p) { |
ee9b8eae YO |
7185 | /* These are all the special escapes. */ |
7186 | case 'A': /* Start assertion */ | |
7187 | case 'b': case 'B': /* Word-boundary assertion*/ | |
7188 | case 'C': /* Single char !DANGEROUS! */ | |
7189 | case 'd': case 'D': /* digit class */ | |
7190 | case 'g': case 'G': /* generic-backref, pos assertion */ | |
e1d1eefb | 7191 | case 'h': case 'H': /* HORIZWS */ |
ee9b8eae YO |
7192 | case 'k': case 'K': /* named backref, keep marker */ |
7193 | case 'N': /* named char sequence */ | |
38a44b82 | 7194 | case 'p': case 'P': /* Unicode property */ |
e1d1eefb | 7195 | case 'R': /* LNBREAK */ |
ee9b8eae | 7196 | case 's': case 'S': /* space class */ |
e1d1eefb | 7197 | case 'v': case 'V': /* VERTWS */ |
ee9b8eae YO |
7198 | case 'w': case 'W': /* word class */ |
7199 | case 'X': /* eXtended Unicode "combining character sequence" */ | |
7200 | case 'z': case 'Z': /* End of line/string assertion */ | |
a0d0e21e LW |
7201 | --p; |
7202 | goto loopdone; | |
ee9b8eae YO |
7203 | |
7204 | /* Anything after here is an escape that resolves to a | |
7205 | literal. (Except digits, which may or may not) | |
7206 | */ | |
a0d0e21e LW |
7207 | case 'n': |
7208 | ender = '\n'; | |
7209 | p++; | |
a687059c | 7210 | break; |
a0d0e21e LW |
7211 | case 'r': |
7212 | ender = '\r'; | |
7213 | p++; | |
a687059c | 7214 | break; |
a0d0e21e LW |
7215 | case 't': |
7216 | ender = '\t'; | |
7217 | p++; | |
a687059c | 7218 | break; |
a0d0e21e LW |
7219 | case 'f': |
7220 | ender = '\f'; | |
7221 | p++; | |
a687059c | 7222 | break; |
a0d0e21e | 7223 | case 'e': |
c7f1f016 | 7224 | ender = ASCII_TO_NATIVE('\033'); |
a0d0e21e | 7225 | p++; |
a687059c | 7226 | break; |
a0d0e21e | 7227 | case 'a': |
c7f1f016 | 7228 | ender = ASCII_TO_NATIVE('\007'); |
a0d0e21e | 7229 | p++; |
a687059c | 7230 | break; |
a0d0e21e | 7231 | case 'x': |
a0ed51b3 | 7232 | if (*++p == '{') { |
1df70142 | 7233 | char* const e = strchr(p, '}'); |
b81d288d | 7234 | |
b45f050a | 7235 | if (!e) { |
830247a4 | 7236 | RExC_parse = p + 1; |
b45f050a JF |
7237 | vFAIL("Missing right brace on \\x{}"); |
7238 | } | |
de5f0749 | 7239 | else { |
a4c04bdc NC |
7240 | I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
7241 | | PERL_SCAN_DISALLOW_PREFIX; | |
1df70142 | 7242 | STRLEN numlen = e - p - 1; |
53305cf1 | 7243 | ender = grok_hex(p + 1, &numlen, &flags, NULL); |
aaa80028 JH |
7244 | if (ender > 0xff) |
7245 | RExC_utf8 = 1; | |
a0ed51b3 LW |
7246 | p = e + 1; |
7247 | } | |
a0ed51b3 LW |
7248 | } |
7249 | else { | |
a4c04bdc | 7250 | I32 flags = PERL_SCAN_DISALLOW_PREFIX; |
1df70142 | 7251 | STRLEN numlen = 2; |
53305cf1 | 7252 | ender = grok_hex(p, &numlen, &flags, NULL); |
a0ed51b3 LW |
7253 | p += numlen; |
7254 | } | |
9e08bc66 TS |
7255 | if (PL_encoding && ender < 0x100) |
7256 | goto recode_encoding; | |
a687059c | 7257 | break; |
a0d0e21e LW |
7258 | case 'c': |
7259 | p++; | |
bbce6d69 | 7260 | ender = UCHARAT(p++); |
7261 | ender = toCTRL(ender); | |
a687059c | 7262 | break; |
a0d0e21e LW |
7263 | case '0': case '1': case '2': case '3':case '4': |
7264 | case '5': case '6': case '7': case '8':case '9': | |
7265 | if (*p == '0' || | |
830247a4 | 7266 | (isDIGIT(p[1]) && atoi(p) >= RExC_npar) ) { |
53305cf1 | 7267 | I32 flags = 0; |
1df70142 | 7268 | STRLEN numlen = 3; |
53305cf1 | 7269 | ender = grok_oct(p, &numlen, &flags, NULL); |
a0d0e21e LW |
7270 | p += numlen; |
7271 | } | |
7272 | else { | |
7273 | --p; | |
7274 | goto loopdone; | |
a687059c | 7275 | } |
9e08bc66 TS |
7276 | if (PL_encoding && ender < 0x100) |
7277 | goto recode_encoding; | |
7278 | break; | |
7279 | recode_encoding: | |
7280 | { | |
7281 | SV* enc = PL_encoding; | |
7282 | ender = reg_recode((const char)(U8)ender, &enc); | |
7283 | if (!enc && SIZE_ONLY && ckWARN(WARN_REGEXP)) | |
7284 | vWARN(p, "Invalid escape in the specified encoding"); | |
7285 | RExC_utf8 = 1; | |
7286 | } | |
a687059c | 7287 | break; |
a0d0e21e | 7288 | case '\0': |
830247a4 | 7289 | if (p >= RExC_end) |
b45f050a | 7290 | FAIL("Trailing \\"); |
a687059c | 7291 | /* FALL THROUGH */ |
a0d0e21e | 7292 | default: |
041457d9 | 7293 | if (!SIZE_ONLY&& isALPHA(*p) && ckWARN(WARN_REGEXP)) |
4193bef7 | 7294 | vWARN2(p + 1, "Unrecognized escape \\%c passed through", UCHARAT(p)); |
a0ed51b3 | 7295 | goto normal_default; |
a0d0e21e LW |
7296 | } |
7297 | break; | |
a687059c | 7298 | default: |
a0ed51b3 | 7299 | normal_default: |
fd400ab9 | 7300 | if (UTF8_IS_START(*p) && UTF) { |
1df70142 | 7301 | STRLEN numlen; |
5e12f4fb | 7302 | ender = utf8n_to_uvchr((U8*)p, RExC_end - p, |
9f7f3913 | 7303 | &numlen, UTF8_ALLOW_DEFAULT); |
a0ed51b3 LW |
7304 | p += numlen; |
7305 | } | |
7306 | else | |
7307 | ender = *p++; | |
a0d0e21e | 7308 | break; |
a687059c | 7309 | } |
bcdf7404 YO |
7310 | if ( RExC_flags & RXf_PMf_EXTENDED) |
7311 | p = regwhite( pRExC_state, p ); | |
60a8b682 JH |
7312 | if (UTF && FOLD) { |
7313 | /* Prime the casefolded buffer. */ | |
ac7e0132 | 7314 | ender = toFOLD_uni(ender, tmpbuf, &foldlen); |
60a8b682 | 7315 | } |
bcdf7404 | 7316 | if (p < RExC_end && ISMULT2(p)) { /* Back off on ?+*. */ |
a0d0e21e LW |
7317 | if (len) |
7318 | p = oldp; | |
16ea2a2e | 7319 | else if (UTF) { |
80aecb99 | 7320 | if (FOLD) { |
60a8b682 | 7321 | /* Emit all the Unicode characters. */ |
1df70142 | 7322 | STRLEN numlen; |
80aecb99 JH |
7323 | for (foldbuf = tmpbuf; |
7324 | foldlen; | |
7325 | foldlen -= numlen) { | |
7326 | ender = utf8_to_uvchr(foldbuf, &numlen); | |
9dc45d57 | 7327 | if (numlen > 0) { |
71207a34 | 7328 | const STRLEN unilen = reguni(pRExC_state, ender, s); |
0ebc6274 JH |
7329 | s += unilen; |
7330 | len += unilen; | |
7331 | /* In EBCDIC the numlen | |
7332 | * and unilen can differ. */ | |
9dc45d57 | 7333 | foldbuf += numlen; |
47654450 JH |
7334 | if (numlen >= foldlen) |
7335 | break; | |
9dc45d57 JH |
7336 | } |
7337 | else | |
7338 | break; /* "Can't happen." */ | |
80aecb99 JH |
7339 | } |
7340 | } | |
7341 | else { | |
71207a34 | 7342 | const STRLEN unilen = reguni(pRExC_state, ender, s); |
9ede7db1 | 7343 | if (unilen > 0) { |
0ebc6274 JH |
7344 | s += unilen; |
7345 | len += unilen; | |
9dc45d57 | 7346 | } |
80aecb99 | 7347 | } |
a0ed51b3 | 7348 | } |
a0d0e21e LW |
7349 | else { |
7350 | len++; | |
eb160463 | 7351 | REGC((char)ender, s++); |
a0d0e21e LW |
7352 | } |
7353 | break; | |
a687059c | 7354 | } |
16ea2a2e | 7355 | if (UTF) { |
80aecb99 | 7356 | if (FOLD) { |
60a8b682 | 7357 | /* Emit all the Unicode characters. */ |
1df70142 | 7358 | STRLEN numlen; |
80aecb99 JH |
7359 | for (foldbuf = tmpbuf; |
7360 | foldlen; | |
7361 | foldlen -= numlen) { | |
7362 | ender = utf8_to_uvchr(foldbuf, &numlen); | |
9dc45d57 | 7363 | if (numlen > 0) { |
71207a34 | 7364 | const STRLEN unilen = reguni(pRExC_state, ender, s); |
0ebc6274 JH |
7365 | len += unilen; |
7366 | s += unilen; | |
7367 | /* In EBCDIC the numlen | |
7368 | * and unilen can differ. */ | |
9dc45d57 | 7369 | foldbuf += numlen; |
47654450 JH |
7370 | if (numlen >= foldlen) |
7371 | break; | |
9dc45d57 JH |
7372 | } |
7373 | else | |
7374 | break; | |
80aecb99 JH |
7375 | } |
7376 | } | |
7377 | else { | |
71207a34 | 7378 | const STRLEN unilen = reguni(pRExC_state, ender, s); |
9ede7db1 | 7379 | if (unilen > 0) { |
0ebc6274 JH |
7380 | s += unilen; |
7381 | len += unilen; | |
9dc45d57 | 7382 | } |
80aecb99 JH |
7383 | } |
7384 | len--; | |
a0ed51b3 LW |
7385 | } |
7386 | else | |
eb160463 | 7387 | REGC((char)ender, s++); |
a0d0e21e LW |
7388 | } |
7389 | loopdone: | |
830247a4 | 7390 | RExC_parse = p - 1; |
fac92740 | 7391 | Set_Node_Cur_Length(ret); /* MJD */ |
830247a4 | 7392 | nextchar(pRExC_state); |
793db0cb JH |
7393 | { |
7394 | /* len is STRLEN which is unsigned, need to copy to signed */ | |
7395 | IV iv = len; | |
7396 | if (iv < 0) | |
7397 | vFAIL("Internal disaster"); | |
7398 | } | |
a0d0e21e LW |
7399 | if (len > 0) |
7400 | *flagp |= HASWIDTH; | |
090f7165 | 7401 | if (len == 1 && UNI_IS_INVARIANT(ender)) |
a0d0e21e | 7402 | *flagp |= SIMPLE; |
3dab1dad | 7403 | |
cd439c50 | 7404 | if (SIZE_ONLY) |
830247a4 | 7405 | RExC_size += STR_SZ(len); |
3dab1dad YO |
7406 | else { |
7407 | STR_LEN(ret) = len; | |
830247a4 | 7408 | RExC_emit += STR_SZ(len); |
07be1b83 | 7409 | } |
3dab1dad | 7410 | } |
a0d0e21e LW |
7411 | break; |
7412 | } | |
a687059c | 7413 | |
a0d0e21e | 7414 | return(ret); |
a687059c LW |
7415 | } |
7416 | ||
873ef191 | 7417 | STATIC char * |
bcdf7404 | 7418 | S_regwhite( RExC_state_t *pRExC_state, char *p ) |
5b5a24f7 | 7419 | { |
bcdf7404 | 7420 | const char *e = RExC_end; |
5b5a24f7 CS |
7421 | while (p < e) { |
7422 | if (isSPACE(*p)) | |
7423 | ++p; | |
7424 | else if (*p == '#') { | |
bcdf7404 | 7425 | bool ended = 0; |
5b5a24f7 | 7426 | do { |
bcdf7404 YO |
7427 | if (*p++ == '\n') { |
7428 | ended = 1; | |
7429 | break; | |
7430 | } | |
7431 | } while (p < e); | |
7432 | if (!ended) | |
7433 | RExC_seen |= REG_SEEN_RUN_ON_COMMENT; | |
5b5a24f7 CS |
7434 | } |
7435 | else | |
7436 | break; | |
7437 | } | |
7438 | return p; | |
7439 | } | |
7440 | ||
b8c5462f JH |
7441 | /* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]]. |
7442 | Character classes ([:foo:]) can also be negated ([:^foo:]). | |
7443 | Returns a named class id (ANYOF_XXX) if successful, -1 otherwise. | |
7444 | Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed, | |
beeb77fc | 7445 | but trigger failures because they are currently unimplemented. */ |
9a86a77b JH |
7446 | |
7447 | #define POSIXCC_DONE(c) ((c) == ':') | |
7448 | #define POSIXCC_NOTYET(c) ((c) == '=' || (c) == '.') | |
7449 | #define POSIXCC(c) (POSIXCC_DONE(c) || POSIXCC_NOTYET(c)) | |
7450 | ||
b8c5462f | 7451 | STATIC I32 |
830247a4 | 7452 | S_regpposixcc(pTHX_ RExC_state_t *pRExC_state, I32 value) |
620e46c5 | 7453 | { |
97aff369 | 7454 | dVAR; |
936ed897 | 7455 | I32 namedclass = OOB_NAMEDCLASS; |
620e46c5 | 7456 | |
830247a4 | 7457 | if (value == '[' && RExC_parse + 1 < RExC_end && |
620e46c5 | 7458 | /* I smell either [: or [= or [. -- POSIX has been here, right? */ |
9a86a77b | 7459 | POSIXCC(UCHARAT(RExC_parse))) { |
1df70142 | 7460 | const char c = UCHARAT(RExC_parse); |
097eb12c | 7461 | char* const s = RExC_parse++; |
b81d288d | 7462 | |
9a86a77b | 7463 | while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != c) |
830247a4 IZ |
7464 | RExC_parse++; |
7465 | if (RExC_parse == RExC_end) | |
620e46c5 | 7466 | /* Grandfather lone [:, [=, [. */ |
830247a4 | 7467 | RExC_parse = s; |
620e46c5 | 7468 | else { |
3dab1dad | 7469 | const char* const t = RExC_parse++; /* skip over the c */ |
80916619 NC |
7470 | assert(*t == c); |
7471 | ||
9a86a77b | 7472 | if (UCHARAT(RExC_parse) == ']') { |
3dab1dad | 7473 | const char *posixcc = s + 1; |
830247a4 | 7474 | RExC_parse++; /* skip over the ending ] */ |
3dab1dad | 7475 | |
b8c5462f | 7476 | if (*s == ':') { |
1df70142 AL |
7477 | const I32 complement = *posixcc == '^' ? *posixcc++ : 0; |
7478 | const I32 skip = t - posixcc; | |
80916619 NC |
7479 | |
7480 | /* Initially switch on the length of the name. */ | |
7481 | switch (skip) { | |
7482 | case 4: | |
3dab1dad YO |
7483 | if (memEQ(posixcc, "word", 4)) /* this is not POSIX, this is the Perl \w */ |
7484 | namedclass = complement ? ANYOF_NALNUM : ANYOF_ALNUM; | |
cc4319de | 7485 | break; |
80916619 NC |
7486 | case 5: |
7487 | /* Names all of length 5. */ | |
7488 | /* alnum alpha ascii blank cntrl digit graph lower | |
7489 | print punct space upper */ | |
7490 | /* Offset 4 gives the best switch position. */ | |
7491 | switch (posixcc[4]) { | |
7492 | case 'a': | |
3dab1dad YO |
7493 | if (memEQ(posixcc, "alph", 4)) /* alpha */ |
7494 | namedclass = complement ? ANYOF_NALPHA : ANYOF_ALPHA; | |
80916619 NC |
7495 | break; |
7496 | case 'e': | |
3dab1dad YO |
7497 | if (memEQ(posixcc, "spac", 4)) /* space */ |
7498 | namedclass = complement ? ANYOF_NPSXSPC : ANYOF_PSXSPC; | |
80916619 NC |
7499 | break; |
7500 | case 'h': | |
3dab1dad YO |
7501 | if (memEQ(posixcc, "grap", 4)) /* graph */ |
7502 | namedclass = complement ? ANYOF_NGRAPH : ANYOF_GRAPH; | |
80916619 NC |
7503 | break; |
7504 | case 'i': | |
3dab1dad YO |
7505 | if (memEQ(posixcc, "asci", 4)) /* ascii */ |
7506 | namedclass = complement ? ANYOF_NASCII : ANYOF_ASCII; | |
80916619 NC |
7507 | break; |
7508 | case 'k': | |
3dab1dad YO |
7509 | if (memEQ(posixcc, "blan", 4)) /* blank */ |
7510 | namedclass = complement ? ANYOF_NBLANK : ANYOF_BLANK; | |
80916619 NC |
7511 | break; |
7512 | case 'l': | |
3dab1dad YO |
7513 | if (memEQ(posixcc, "cntr", 4)) /* cntrl */ |
7514 | namedclass = complement ? ANYOF_NCNTRL : ANYOF_CNTRL; | |
80916619 NC |
7515 | break; |
7516 | case 'm': | |
3dab1dad YO |
7517 | if (memEQ(posixcc, "alnu", 4)) /* alnum */ |
7518 | namedclass = complement ? ANYOF_NALNUMC : ANYOF_ALNUMC; | |
80916619 NC |
7519 | break; |
7520 | case 'r': | |
3dab1dad YO |
7521 | if (memEQ(posixcc, "lowe", 4)) /* lower */ |
7522 | namedclass = complement ? ANYOF_NLOWER : ANYOF_LOWER; | |
7523 | else if (memEQ(posixcc, "uppe", 4)) /* upper */ | |
7524 | namedclass = complement ? ANYOF_NUPPER : ANYOF_UPPER; | |
80916619 NC |
7525 | break; |
7526 | case 't': | |
3dab1dad YO |
7527 | if (memEQ(posixcc, "digi", 4)) /* digit */ |
7528 | namedclass = complement ? ANYOF_NDIGIT : ANYOF_DIGIT; | |
7529 | else if (memEQ(posixcc, "prin", 4)) /* print */ | |
7530 | namedclass = complement ? ANYOF_NPRINT : ANYOF_PRINT; | |
7531 | else if (memEQ(posixcc, "punc", 4)) /* punct */ | |
7532 | namedclass = complement ? ANYOF_NPUNCT : ANYOF_PUNCT; | |
80916619 | 7533 | break; |
b8c5462f JH |
7534 | } |
7535 | break; | |
80916619 | 7536 | case 6: |
3dab1dad YO |
7537 | if (memEQ(posixcc, "xdigit", 6)) |
7538 | namedclass = complement ? ANYOF_NXDIGIT : ANYOF_XDIGIT; | |
b8c5462f JH |
7539 | break; |
7540 | } | |
80916619 NC |
7541 | |
7542 | if (namedclass == OOB_NAMEDCLASS) | |
b45f050a JF |
7543 | Simple_vFAIL3("POSIX class [:%.*s:] unknown", |
7544 | t - s - 1, s + 1); | |
80916619 NC |
7545 | assert (posixcc[skip] == ':'); |
7546 | assert (posixcc[skip+1] == ']'); | |
b45f050a | 7547 | } else if (!SIZE_ONLY) { |
b8c5462f | 7548 | /* [[=foo=]] and [[.foo.]] are still future. */ |
b45f050a | 7549 | |
830247a4 | 7550 | /* adjust RExC_parse so the warning shows after |
b45f050a | 7551 | the class closes */ |
9a86a77b | 7552 | while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse) != ']') |
830247a4 | 7553 | RExC_parse++; |
b45f050a JF |
7554 | Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c); |
7555 | } | |
b8c5462f JH |
7556 | } else { |
7557 | /* Maternal grandfather: | |
7558 | * "[:" ending in ":" but not in ":]" */ | |
830247a4 | 7559 | RExC_parse = s; |
767d463e | 7560 | } |
620e46c5 JH |
7561 | } |
7562 | } | |
7563 | ||
b8c5462f JH |
7564 | return namedclass; |
7565 | } | |
7566 | ||
7567 | STATIC void | |
830247a4 | 7568 | S_checkposixcc(pTHX_ RExC_state_t *pRExC_state) |
b8c5462f | 7569 | { |
97aff369 | 7570 | dVAR; |
3dab1dad | 7571 | if (POSIXCC(UCHARAT(RExC_parse))) { |
1df70142 AL |
7572 | const char *s = RExC_parse; |
7573 | const char c = *s++; | |
b8c5462f | 7574 | |
3dab1dad | 7575 | while (isALNUM(*s)) |
b8c5462f JH |
7576 | s++; |
7577 | if (*s && c == *s && s[1] == ']') { | |
cd84f5b2 RGS |
7578 | if (ckWARN(WARN_REGEXP)) |
7579 | vWARN3(s+2, | |
7580 | "POSIX syntax [%c %c] belongs inside character classes", | |
7581 | c, c); | |
b45f050a JF |
7582 | |
7583 | /* [[=foo=]] and [[.foo.]] are still future. */ | |
9a86a77b | 7584 | if (POSIXCC_NOTYET(c)) { |
830247a4 | 7585 | /* adjust RExC_parse so the error shows after |
b45f050a | 7586 | the class closes */ |
9a86a77b | 7587 | while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse++) != ']') |
3dab1dad | 7588 | NOOP; |
b45f050a JF |
7589 | Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c); |
7590 | } | |
b8c5462f JH |
7591 | } |
7592 | } | |
620e46c5 JH |
7593 | } |
7594 | ||
7f6f358c | 7595 | |
89836f1f YO |
7596 | #define _C_C_T_(NAME,TEST,WORD) \ |
7597 | ANYOF_##NAME: \ | |
7598 | if (LOC) \ | |
7599 | ANYOF_CLASS_SET(ret, ANYOF_##NAME); \ | |
7600 | else { \ | |
7601 | for (value = 0; value < 256; value++) \ | |
7602 | if (TEST) \ | |
7603 | ANYOF_BITMAP_SET(ret, value); \ | |
7604 | } \ | |
7605 | yesno = '+'; \ | |
7606 | what = WORD; \ | |
7607 | break; \ | |
7608 | case ANYOF_N##NAME: \ | |
7609 | if (LOC) \ | |
7610 | ANYOF_CLASS_SET(ret, ANYOF_N##NAME); \ | |
7611 | else { \ | |
7612 | for (value = 0; value < 256; value++) \ | |
7613 | if (!TEST) \ | |
7614 | ANYOF_BITMAP_SET(ret, value); \ | |
7615 | } \ | |
7616 | yesno = '!'; \ | |
7617 | what = WORD; \ | |
7618 | break | |
7619 | ||
e1d1eefb YO |
7620 | #define _C_C_T_NOLOC_(NAME,TEST,WORD) \ |
7621 | ANYOF_##NAME: \ | |
7622 | for (value = 0; value < 256; value++) \ | |
7623 | if (TEST) \ | |
7624 | ANYOF_BITMAP_SET(ret, value); \ | |
7625 | yesno = '+'; \ | |
7626 | what = WORD; \ | |
7627 | break; \ | |
7628 | case ANYOF_N##NAME: \ | |
7629 | for (value = 0; value < 256; value++) \ | |
7630 | if (!TEST) \ | |
7631 | ANYOF_BITMAP_SET(ret, value); \ | |
7632 | yesno = '!'; \ | |
7633 | what = WORD; \ | |
7634 | break | |
89836f1f | 7635 | |
7f6f358c YO |
7636 | /* |
7637 | parse a class specification and produce either an ANYOF node that | |
89836f1f YO |
7638 | matches the pattern or if the pattern matches a single char only and |
7639 | that char is < 256 and we are case insensitive then we produce an | |
7640 | EXACT node instead. | |
7f6f358c | 7641 | */ |
89836f1f | 7642 | |
76e3520e | 7643 | STATIC regnode * |
3dab1dad | 7644 | S_regclass(pTHX_ RExC_state_t *pRExC_state, U32 depth) |
a687059c | 7645 | { |
97aff369 | 7646 | dVAR; |
9a86a77b | 7647 | register UV nextvalue; |
3568d838 | 7648 | register IV prevvalue = OOB_UNICODE; |
ffc61ed2 | 7649 | register IV range = 0; |
e1d1eefb | 7650 | UV value = 0; /* XXX:dmq: needs to be referenceable (unfortunately) */ |
c277df42 | 7651 | register regnode *ret; |
ba210ebe | 7652 | STRLEN numlen; |
ffc61ed2 | 7653 | IV namedclass; |
cbbf8932 | 7654 | char *rangebegin = NULL; |
936ed897 | 7655 | bool need_class = 0; |
c445ea15 | 7656 | SV *listsv = NULL; |
ffc61ed2 | 7657 | UV n; |
9e55ce06 | 7658 | bool optimize_invert = TRUE; |
cbbf8932 | 7659 | AV* unicode_alternate = NULL; |
1b2d223b JH |
7660 | #ifdef EBCDIC |
7661 | UV literal_endpoint = 0; | |
7662 | #endif | |
7f6f358c | 7663 | UV stored = 0; /* number of chars stored in the class */ |
ffc61ed2 | 7664 | |
3dab1dad | 7665 | regnode * const orig_emit = RExC_emit; /* Save the original RExC_emit in |
7f6f358c | 7666 | case we need to change the emitted regop to an EXACT. */ |
07be1b83 | 7667 | const char * orig_parse = RExC_parse; |
72f13be8 | 7668 | GET_RE_DEBUG_FLAGS_DECL; |
76e84362 SH |
7669 | #ifndef DEBUGGING |
7670 | PERL_UNUSED_ARG(depth); | |
7671 | #endif | |
72f13be8 | 7672 | |
3dab1dad | 7673 | DEBUG_PARSE("clas"); |
7f6f358c YO |
7674 | |
7675 | /* Assume we are going to generate an ANYOF node. */ | |
ffc61ed2 JH |
7676 | ret = reganode(pRExC_state, ANYOF, 0); |
7677 | ||
7678 | if (!SIZE_ONLY) | |
7679 | ANYOF_FLAGS(ret) = 0; | |
7680 | ||
9a86a77b | 7681 | if (UCHARAT(RExC_parse) == '^') { /* Complement of range. */ |
ffc61ed2 JH |
7682 | RExC_naughty++; |
7683 | RExC_parse++; | |
7684 | if (!SIZE_ONLY) | |
7685 | ANYOF_FLAGS(ret) |= ANYOF_INVERT; | |
7686 | } | |
a0d0e21e | 7687 | |
73060fc4 | 7688 | if (SIZE_ONLY) { |
830247a4 | 7689 | RExC_size += ANYOF_SKIP; |
73060fc4 JH |
7690 | listsv = &PL_sv_undef; /* For code scanners: listsv always non-NULL. */ |
7691 | } | |
936ed897 | 7692 | else { |
830247a4 | 7693 | RExC_emit += ANYOF_SKIP; |
936ed897 IZ |
7694 | if (FOLD) |
7695 | ANYOF_FLAGS(ret) |= ANYOF_FOLD; | |
7696 | if (LOC) | |
7697 | ANYOF_FLAGS(ret) |= ANYOF_LOCALE; | |
ffc61ed2 | 7698 | ANYOF_BITMAP_ZERO(ret); |
396482e1 | 7699 | listsv = newSVpvs("# comment\n"); |
a0d0e21e | 7700 | } |
b8c5462f | 7701 | |
9a86a77b JH |
7702 | nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0; |
7703 | ||
b938889d | 7704 | if (!SIZE_ONLY && POSIXCC(nextvalue)) |
830247a4 | 7705 | checkposixcc(pRExC_state); |
b8c5462f | 7706 | |
f064b6ad HS |
7707 | /* allow 1st char to be ] (allowing it to be - is dealt with later) */ |
7708 | if (UCHARAT(RExC_parse) == ']') | |
7709 | goto charclassloop; | |
ffc61ed2 | 7710 | |
fc8cd66c | 7711 | parseit: |
9a86a77b | 7712 | while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != ']') { |
ffc61ed2 JH |
7713 | |
7714 | charclassloop: | |
7715 | ||
7716 | namedclass = OOB_NAMEDCLASS; /* initialize as illegal */ | |
7717 | ||
73b437c8 | 7718 | if (!range) |
830247a4 | 7719 | rangebegin = RExC_parse; |
ffc61ed2 | 7720 | if (UTF) { |
5e12f4fb | 7721 | value = utf8n_to_uvchr((U8*)RExC_parse, |
3568d838 | 7722 | RExC_end - RExC_parse, |
9f7f3913 | 7723 | &numlen, UTF8_ALLOW_DEFAULT); |
ffc61ed2 JH |
7724 | RExC_parse += numlen; |
7725 | } | |
7726 | else | |
7727 | value = UCHARAT(RExC_parse++); | |
7f6f358c | 7728 | |
9a86a77b JH |
7729 | nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0; |
7730 | if (value == '[' && POSIXCC(nextvalue)) | |
830247a4 | 7731 | namedclass = regpposixcc(pRExC_state, value); |
620e46c5 | 7732 | else if (value == '\\') { |
ffc61ed2 | 7733 | if (UTF) { |
5e12f4fb | 7734 | value = utf8n_to_uvchr((U8*)RExC_parse, |
ffc61ed2 | 7735 | RExC_end - RExC_parse, |
9f7f3913 | 7736 | &numlen, UTF8_ALLOW_DEFAULT); |
ffc61ed2 JH |
7737 | RExC_parse += numlen; |
7738 | } | |
7739 | else | |
7740 | value = UCHARAT(RExC_parse++); | |
470c3474 | 7741 | /* Some compilers cannot handle switching on 64-bit integer |
ffc61ed2 | 7742 | * values, therefore value cannot be an UV. Yes, this will |
e2962f66 JH |
7743 | * be a problem later if we want switch on Unicode. |
7744 | * A similar issue a little bit later when switching on | |
7745 | * namedclass. --jhi */ | |
ffc61ed2 | 7746 | switch ((I32)value) { |
b8c5462f JH |
7747 | case 'w': namedclass = ANYOF_ALNUM; break; |
7748 | case 'W': namedclass = ANYOF_NALNUM; break; | |
7749 | case 's': namedclass = ANYOF_SPACE; break; | |
7750 | case 'S': namedclass = ANYOF_NSPACE; break; | |
7751 | case 'd': namedclass = ANYOF_DIGIT; break; | |
7752 | case 'D': namedclass = ANYOF_NDIGIT; break; | |
e1d1eefb YO |
7753 | case 'v': namedclass = ANYOF_VERTWS; break; |
7754 | case 'V': namedclass = ANYOF_NVERTWS; break; | |
7755 | case 'h': namedclass = ANYOF_HORIZWS; break; | |
7756 | case 'H': namedclass = ANYOF_NHORIZWS; break; | |
fc8cd66c YO |
7757 | case 'N': /* Handle \N{NAME} in class */ |
7758 | { | |
7759 | /* We only pay attention to the first char of | |
7760 | multichar strings being returned. I kinda wonder | |
7761 | if this makes sense as it does change the behaviour | |
7762 | from earlier versions, OTOH that behaviour was broken | |
7763 | as well. */ | |
7764 | UV v; /* value is register so we cant & it /grrr */ | |
7765 | if (reg_namedseq(pRExC_state, &v)) { | |
7766 | goto parseit; | |
7767 | } | |
7768 | value= v; | |
7769 | } | |
7770 | break; | |
ffc61ed2 JH |
7771 | case 'p': |
7772 | case 'P': | |
3dab1dad YO |
7773 | { |
7774 | char *e; | |
af6f566e | 7775 | if (RExC_parse >= RExC_end) |
2a4859cd | 7776 | vFAIL2("Empty \\%c{}", (U8)value); |
ffc61ed2 | 7777 | if (*RExC_parse == '{') { |
1df70142 | 7778 | const U8 c = (U8)value; |
ffc61ed2 JH |
7779 | e = strchr(RExC_parse++, '}'); |
7780 | if (!e) | |
0da60cf5 | 7781 | vFAIL2("Missing right brace on \\%c{}", c); |
ab13f0c7 JH |
7782 | while (isSPACE(UCHARAT(RExC_parse))) |
7783 | RExC_parse++; | |
7784 | if (e == RExC_parse) | |
0da60cf5 | 7785 | vFAIL2("Empty \\%c{}", c); |
ffc61ed2 | 7786 | n = e - RExC_parse; |
ab13f0c7 JH |
7787 | while (isSPACE(UCHARAT(RExC_parse + n - 1))) |
7788 | n--; | |
ffc61ed2 JH |
7789 | } |
7790 | else { | |
7791 | e = RExC_parse; | |
7792 | n = 1; | |
7793 | } | |
7794 | if (!SIZE_ONLY) { | |
ab13f0c7 JH |
7795 | if (UCHARAT(RExC_parse) == '^') { |
7796 | RExC_parse++; | |
7797 | n--; | |
7798 | value = value == 'p' ? 'P' : 'p'; /* toggle */ | |
7799 | while (isSPACE(UCHARAT(RExC_parse))) { | |
7800 | RExC_parse++; | |
7801 | n--; | |
7802 | } | |
7803 | } | |
097eb12c AL |
7804 | Perl_sv_catpvf(aTHX_ listsv, "%cutf8::%.*s\n", |
7805 | (value=='p' ? '+' : '!'), (int)n, RExC_parse); | |
ffc61ed2 JH |
7806 | } |
7807 | RExC_parse = e + 1; | |
7808 | ANYOF_FLAGS(ret) |= ANYOF_UNICODE; | |
f81125e2 | 7809 | namedclass = ANYOF_MAX; /* no official name, but it's named */ |
3dab1dad | 7810 | } |
f81125e2 | 7811 | break; |
b8c5462f JH |
7812 | case 'n': value = '\n'; break; |
7813 | case 'r': value = '\r'; break; | |
7814 | case 't': value = '\t'; break; | |
7815 | case 'f': value = '\f'; break; | |
7816 | case 'b': value = '\b'; break; | |
c7f1f016 NIS |
7817 | case 'e': value = ASCII_TO_NATIVE('\033');break; |
7818 | case 'a': value = ASCII_TO_NATIVE('\007');break; | |
b8c5462f | 7819 | case 'x': |
ffc61ed2 | 7820 | if (*RExC_parse == '{') { |
a4c04bdc NC |
7821 | I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
7822 | | PERL_SCAN_DISALLOW_PREFIX; | |
3dab1dad | 7823 | char * const e = strchr(RExC_parse++, '}'); |
b81d288d | 7824 | if (!e) |
ffc61ed2 | 7825 | vFAIL("Missing right brace on \\x{}"); |
53305cf1 NC |
7826 | |
7827 | numlen = e - RExC_parse; | |
7828 | value = grok_hex(RExC_parse, &numlen, &flags, NULL); | |
ffc61ed2 JH |
7829 | RExC_parse = e + 1; |
7830 | } | |
7831 | else { | |
a4c04bdc | 7832 | I32 flags = PERL_SCAN_DISALLOW_PREFIX; |
53305cf1 NC |
7833 | numlen = 2; |
7834 | value = grok_hex(RExC_parse, &numlen, &flags, NULL); | |
ffc61ed2 JH |
7835 | RExC_parse += numlen; |
7836 | } | |
9e08bc66 TS |
7837 | if (PL_encoding && value < 0x100) |
7838 | goto recode_encoding; | |
b8c5462f JH |
7839 | break; |
7840 | case 'c': | |
830247a4 | 7841 | value = UCHARAT(RExC_parse++); |
b8c5462f JH |
7842 | value = toCTRL(value); |
7843 | break; | |
7844 | case '0': case '1': case '2': case '3': case '4': | |
7845 | case '5': case '6': case '7': case '8': case '9': | |
9e08bc66 TS |
7846 | { |
7847 | I32 flags = 0; | |
7848 | numlen = 3; | |
7849 | value = grok_oct(--RExC_parse, &numlen, &flags, NULL); | |
7850 | RExC_parse += numlen; | |
7851 | if (PL_encoding && value < 0x100) | |
7852 | goto recode_encoding; | |
7853 | break; | |
7854 | } | |
7855 | recode_encoding: | |
7856 | { | |
7857 | SV* enc = PL_encoding; | |
7858 | value = reg_recode((const char)(U8)value, &enc); | |
7859 | if (!enc && SIZE_ONLY && ckWARN(WARN_REGEXP)) | |
7860 | vWARN(RExC_parse, | |
7861 | "Invalid escape in the specified encoding"); | |
7862 | break; | |
7863 | } | |
1028017a | 7864 | default: |
041457d9 | 7865 | if (!SIZE_ONLY && isALPHA(value) && ckWARN(WARN_REGEXP)) |
ffc61ed2 JH |
7866 | vWARN2(RExC_parse, |
7867 | "Unrecognized escape \\%c in character class passed through", | |
7868 | (int)value); | |
1028017a | 7869 | break; |
b8c5462f | 7870 | } |
ffc61ed2 | 7871 | } /* end of \blah */ |
1b2d223b JH |
7872 | #ifdef EBCDIC |
7873 | else | |
7874 | literal_endpoint++; | |
7875 | #endif | |
ffc61ed2 JH |
7876 | |
7877 | if (namedclass > OOB_NAMEDCLASS) { /* this is a named class \blah */ | |
7878 | ||
7879 | if (!SIZE_ONLY && !need_class) | |
936ed897 | 7880 | ANYOF_CLASS_ZERO(ret); |
ffc61ed2 | 7881 | |
936ed897 | 7882 | need_class = 1; |
ffc61ed2 JH |
7883 | |
7884 | /* a bad range like a-\d, a-[:digit:] ? */ | |
7885 | if (range) { | |
73b437c8 | 7886 | if (!SIZE_ONLY) { |
afd78fd5 | 7887 | if (ckWARN(WARN_REGEXP)) { |
097eb12c | 7888 | const int w = |
afd78fd5 JH |
7889 | RExC_parse >= rangebegin ? |
7890 | RExC_parse - rangebegin : 0; | |
830247a4 | 7891 | vWARN4(RExC_parse, |
b45f050a | 7892 | "False [] range \"%*.*s\"", |
097eb12c | 7893 | w, w, rangebegin); |
afd78fd5 | 7894 | } |
3568d838 JH |
7895 | if (prevvalue < 256) { |
7896 | ANYOF_BITMAP_SET(ret, prevvalue); | |
ffc61ed2 JH |
7897 | ANYOF_BITMAP_SET(ret, '-'); |
7898 | } | |
7899 | else { | |
7900 | ANYOF_FLAGS(ret) |= ANYOF_UNICODE; | |
7901 | Perl_sv_catpvf(aTHX_ listsv, | |
3568d838 | 7902 | "%04"UVxf"\n%04"UVxf"\n", (UV)prevvalue, (UV) '-'); |
ffc61ed2 | 7903 | } |
b8c5462f | 7904 | } |
ffc61ed2 JH |
7905 | |
7906 | range = 0; /* this was not a true range */ | |
73b437c8 | 7907 | } |
ffc61ed2 | 7908 | |
89836f1f YO |
7909 | |
7910 | ||
73b437c8 | 7911 | if (!SIZE_ONLY) { |
c49a72a9 NC |
7912 | const char *what = NULL; |
7913 | char yesno = 0; | |
7914 | ||
3568d838 JH |
7915 | if (namedclass > OOB_NAMEDCLASS) |
7916 | optimize_invert = FALSE; | |
e2962f66 JH |
7917 | /* Possible truncation here but in some 64-bit environments |
7918 | * the compiler gets heartburn about switch on 64-bit values. | |
7919 | * A similar issue a little earlier when switching on value. | |
98f323fa | 7920 | * --jhi */ |
e2962f66 | 7921 | switch ((I32)namedclass) { |
89836f1f YO |
7922 | case _C_C_T_(ALNUM, isALNUM(value), "Word"); |
7923 | case _C_C_T_(ALNUMC, isALNUMC(value), "Alnum"); | |
7924 | case _C_C_T_(ALPHA, isALPHA(value), "Alpha"); | |
7925 | case _C_C_T_(BLANK, isBLANK(value), "Blank"); | |
7926 | case _C_C_T_(CNTRL, isCNTRL(value), "Cntrl"); | |
7927 | case _C_C_T_(GRAPH, isGRAPH(value), "Graph"); | |
7928 | case _C_C_T_(LOWER, isLOWER(value), "Lower"); | |
7929 | case _C_C_T_(PRINT, isPRINT(value), "Print"); | |
7930 | case _C_C_T_(PSXSPC, isPSXSPC(value), "Space"); | |
7931 | case _C_C_T_(PUNCT, isPUNCT(value), "Punct"); | |
7932 | case _C_C_T_(SPACE, isSPACE(value), "SpacePerl"); | |
7933 | case _C_C_T_(UPPER, isUPPER(value), "Upper"); | |
7934 | case _C_C_T_(XDIGIT, isXDIGIT(value), "XDigit"); | |
e1d1eefb YO |
7935 | case _C_C_T_NOLOC_(VERTWS, is_VERTWS_latin1(&value), "VertSpace"); |
7936 | case _C_C_T_NOLOC_(HORIZWS, is_HORIZWS_latin1(&value), "HorizSpace"); | |
73b437c8 JH |
7937 | case ANYOF_ASCII: |
7938 | if (LOC) | |
936ed897 | 7939 | ANYOF_CLASS_SET(ret, ANYOF_ASCII); |
73b437c8 | 7940 | else { |
c7f1f016 | 7941 | #ifndef EBCDIC |
1ba5c669 JH |
7942 | for (value = 0; value < 128; value++) |
7943 | ANYOF_BITMAP_SET(ret, value); | |
7944 | #else /* EBCDIC */ | |
ffbc6a93 | 7945 | for (value = 0; value < 256; value++) { |
3a3c4447 JH |
7946 | if (isASCII(value)) |
7947 | ANYOF_BITMAP_SET(ret, value); | |
ffbc6a93 | 7948 | } |
1ba5c669 | 7949 | #endif /* EBCDIC */ |
73b437c8 | 7950 | } |
c49a72a9 NC |
7951 | yesno = '+'; |
7952 | what = "ASCII"; | |
73b437c8 JH |
7953 | break; |
7954 | case ANYOF_NASCII: | |
7955 | if (LOC) | |
936ed897 | 7956 | ANYOF_CLASS_SET(ret, ANYOF_NASCII); |
73b437c8 | 7957 | else { |
c7f1f016 | 7958 | #ifndef EBCDIC |
1ba5c669 JH |
7959 | for (value = 128; value < 256; value++) |
7960 | ANYOF_BITMAP_SET(ret, value); | |
7961 | #else /* EBCDIC */ | |
ffbc6a93 | 7962 | for (value = 0; value < 256; value++) { |
3a3c4447 JH |
7963 | if (!isASCII(value)) |
7964 | ANYOF_BITMAP_SET(ret, value); | |
ffbc6a93 | 7965 | } |
1ba5c669 | 7966 | #endif /* EBCDIC */ |
73b437c8 | 7967 | } |
c49a72a9 NC |
7968 | yesno = '!'; |
7969 | what = "ASCII"; | |
89836f1f | 7970 | break; |
ffc61ed2 JH |
7971 | case ANYOF_DIGIT: |
7972 | if (LOC) | |
7973 | ANYOF_CLASS_SET(ret, ANYOF_DIGIT); | |
7974 | else { | |
7975 | /* consecutive digits assumed */ | |
7976 | for (value = '0'; value <= '9'; value++) | |
7977 | ANYOF_BITMAP_SET(ret, value); | |
7978 | } | |
c49a72a9 NC |
7979 | yesno = '+'; |
7980 | what = "Digit"; | |
ffc61ed2 JH |
7981 | break; |
7982 | case ANYOF_NDIGIT: | |
7983 | if (LOC) | |
7984 | ANYOF_CLASS_SET(ret, ANYOF_NDIGIT); | |
7985 | else { | |
7986 | /* consecutive digits assumed */ | |
7987 | for (value = 0; value < '0'; value++) | |
7988 | ANYOF_BITMAP_SET(ret, value); | |
7989 | for (value = '9' + 1; value < 256; value++) | |
7990 | ANYOF_BITMAP_SET(ret, value); | |
7991 | } | |
c49a72a9 NC |
7992 | yesno = '!'; |
7993 | what = "Digit"; | |
89836f1f | 7994 | break; |
f81125e2 JP |
7995 | case ANYOF_MAX: |
7996 | /* this is to handle \p and \P */ | |
7997 | break; | |
73b437c8 | 7998 | default: |
b45f050a | 7999 | vFAIL("Invalid [::] class"); |
73b437c8 | 8000 | break; |
b8c5462f | 8001 | } |
c49a72a9 NC |
8002 | if (what) { |
8003 | /* Strings such as "+utf8::isWord\n" */ | |
8004 | Perl_sv_catpvf(aTHX_ listsv, "%cutf8::Is%s\n", yesno, what); | |
8005 | } | |
b8c5462f | 8006 | if (LOC) |
936ed897 | 8007 | ANYOF_FLAGS(ret) |= ANYOF_CLASS; |
73b437c8 | 8008 | continue; |
a0d0e21e | 8009 | } |
ffc61ed2 JH |
8010 | } /* end of namedclass \blah */ |
8011 | ||
a0d0e21e | 8012 | if (range) { |
eb160463 | 8013 | if (prevvalue > (IV)value) /* b-a */ { |
d4c19fe8 AL |
8014 | const int w = RExC_parse - rangebegin; |
8015 | Simple_vFAIL4("Invalid [] range \"%*.*s\"", w, w, rangebegin); | |
3568d838 | 8016 | range = 0; /* not a valid range */ |
73b437c8 | 8017 | } |
a0d0e21e LW |
8018 | } |
8019 | else { | |
3568d838 | 8020 | prevvalue = value; /* save the beginning of the range */ |
830247a4 IZ |
8021 | if (*RExC_parse == '-' && RExC_parse+1 < RExC_end && |
8022 | RExC_parse[1] != ']') { | |
8023 | RExC_parse++; | |
ffc61ed2 JH |
8024 | |
8025 | /* a bad range like \w-, [:word:]- ? */ | |
8026 | if (namedclass > OOB_NAMEDCLASS) { | |
afd78fd5 | 8027 | if (ckWARN(WARN_REGEXP)) { |
d4c19fe8 | 8028 | const int w = |
afd78fd5 JH |
8029 | RExC_parse >= rangebegin ? |
8030 | RExC_parse - rangebegin : 0; | |
830247a4 | 8031 | vWARN4(RExC_parse, |
b45f050a | 8032 | "False [] range \"%*.*s\"", |
097eb12c | 8033 | w, w, rangebegin); |
afd78fd5 | 8034 | } |
73b437c8 | 8035 | if (!SIZE_ONLY) |
936ed897 | 8036 | ANYOF_BITMAP_SET(ret, '-'); |
73b437c8 | 8037 | } else |
ffc61ed2 JH |
8038 | range = 1; /* yeah, it's a range! */ |
8039 | continue; /* but do it the next time */ | |
a0d0e21e | 8040 | } |
a687059c | 8041 | } |
ffc61ed2 | 8042 | |
93733859 | 8043 | /* now is the next time */ |
07be1b83 | 8044 | /*stored += (value - prevvalue + 1);*/ |
ae5c130c | 8045 | if (!SIZE_ONLY) { |
3568d838 | 8046 | if (prevvalue < 256) { |
1df70142 | 8047 | const IV ceilvalue = value < 256 ? value : 255; |
3dab1dad | 8048 | IV i; |
3568d838 | 8049 | #ifdef EBCDIC |
1b2d223b JH |
8050 | /* In EBCDIC [\x89-\x91] should include |
8051 | * the \x8e but [i-j] should not. */ | |
8052 | if (literal_endpoint == 2 && | |
8053 | ((isLOWER(prevvalue) && isLOWER(ceilvalue)) || | |
8054 | (isUPPER(prevvalue) && isUPPER(ceilvalue)))) | |
ffc61ed2 | 8055 | { |
3568d838 JH |
8056 | if (isLOWER(prevvalue)) { |
8057 | for (i = prevvalue; i <= ceilvalue; i++) | |
2670d666 BC |
8058 | if (isLOWER(i) && !ANYOF_BITMAP_TEST(ret,i)) { |
8059 | stored++; | |
ffc61ed2 | 8060 | ANYOF_BITMAP_SET(ret, i); |
2670d666 | 8061 | } |
ffc61ed2 | 8062 | } else { |
3568d838 | 8063 | for (i = prevvalue; i <= ceilvalue; i++) |
2670d666 BC |
8064 | if (isUPPER(i) && !ANYOF_BITMAP_TEST(ret,i)) { |
8065 | stored++; | |
ffc61ed2 | 8066 | ANYOF_BITMAP_SET(ret, i); |
2670d666 | 8067 | } |
ffc61ed2 | 8068 | } |
8ada0baa | 8069 | } |
ffc61ed2 | 8070 | else |
8ada0baa | 8071 | #endif |
07be1b83 YO |
8072 | for (i = prevvalue; i <= ceilvalue; i++) { |
8073 | if (!ANYOF_BITMAP_TEST(ret,i)) { | |
8074 | stored++; | |
8075 | ANYOF_BITMAP_SET(ret, i); | |
8076 | } | |
8077 | } | |
3568d838 | 8078 | } |
a5961de5 | 8079 | if (value > 255 || UTF) { |
1df70142 AL |
8080 | const UV prevnatvalue = NATIVE_TO_UNI(prevvalue); |
8081 | const UV natvalue = NATIVE_TO_UNI(value); | |
07be1b83 | 8082 | stored+=2; /* can't optimize this class */ |
ffc61ed2 | 8083 | ANYOF_FLAGS(ret) |= ANYOF_UNICODE; |
b08decb7 | 8084 | if (prevnatvalue < natvalue) { /* what about > ? */ |
ffc61ed2 | 8085 | Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\t%04"UVxf"\n", |
b08decb7 JH |
8086 | prevnatvalue, natvalue); |
8087 | } | |
8088 | else if (prevnatvalue == natvalue) { | |
8089 | Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n", natvalue); | |
09091399 | 8090 | if (FOLD) { |
89ebb4a3 | 8091 | U8 foldbuf[UTF8_MAXBYTES_CASE+1]; |
254ba52a | 8092 | STRLEN foldlen; |
1df70142 | 8093 | const UV f = to_uni_fold(natvalue, foldbuf, &foldlen); |
254ba52a | 8094 | |
e294cc5d JH |
8095 | #ifdef EBCDIC /* RD t/uni/fold ff and 6b */ |
8096 | if (RExC_precomp[0] == ':' && | |
8097 | RExC_precomp[1] == '[' && | |
8098 | (f == 0xDF || f == 0x92)) { | |
8099 | f = NATIVE_TO_UNI(f); | |
8100 | } | |
8101 | #endif | |
c840d2a2 JH |
8102 | /* If folding and foldable and a single |
8103 | * character, insert also the folded version | |
8104 | * to the charclass. */ | |
9e55ce06 | 8105 | if (f != value) { |
e294cc5d JH |
8106 | #ifdef EBCDIC /* RD tunifold ligatures s,t fb05, fb06 */ |
8107 | if ((RExC_precomp[0] == ':' && | |
8108 | RExC_precomp[1] == '[' && | |
8109 | (f == 0xA2 && | |
8110 | (value == 0xFB05 || value == 0xFB06))) ? | |
8111 | foldlen == ((STRLEN)UNISKIP(f) - 1) : | |
8112 | foldlen == (STRLEN)UNISKIP(f) ) | |
8113 | #else | |
eb160463 | 8114 | if (foldlen == (STRLEN)UNISKIP(f)) |
e294cc5d | 8115 | #endif |
9e55ce06 JH |
8116 | Perl_sv_catpvf(aTHX_ listsv, |
8117 | "%04"UVxf"\n", f); | |
8118 | else { | |
8119 | /* Any multicharacter foldings | |
8120 | * require the following transform: | |
8121 | * [ABCDEF] -> (?:[ABCabcDEFd]|pq|rst) | |
8122 | * where E folds into "pq" and F folds | |
8123 | * into "rst", all other characters | |
8124 | * fold to single characters. We save | |
8125 | * away these multicharacter foldings, | |
8126 | * to be later saved as part of the | |
8127 | * additional "s" data. */ | |
8128 | SV *sv; | |
8129 | ||
8130 | if (!unicode_alternate) | |
8131 | unicode_alternate = newAV(); | |
8132 | sv = newSVpvn((char*)foldbuf, foldlen); | |
8133 | SvUTF8_on(sv); | |
8134 | av_push(unicode_alternate, sv); | |
8135 | } | |
8136 | } | |
254ba52a | 8137 | |
60a8b682 JH |
8138 | /* If folding and the value is one of the Greek |
8139 | * sigmas insert a few more sigmas to make the | |
8140 | * folding rules of the sigmas to work right. | |
8141 | * Note that not all the possible combinations | |
8142 | * are handled here: some of them are handled | |
9e55ce06 JH |
8143 | * by the standard folding rules, and some of |
8144 | * them (literal or EXACTF cases) are handled | |
8145 | * during runtime in regexec.c:S_find_byclass(). */ | |
09091399 JH |
8146 | if (value == UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA) { |
8147 | Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n", | |
eb050b28 | 8148 | (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA); |
09091399 | 8149 | Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n", |
eb050b28 | 8150 | (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA); |
09091399 JH |
8151 | } |
8152 | else if (value == UNICODE_GREEK_CAPITAL_LETTER_SIGMA) | |
8153 | Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n", | |
eb050b28 | 8154 | (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA); |
09091399 JH |
8155 | } |
8156 | } | |
ffc61ed2 | 8157 | } |
1b2d223b JH |
8158 | #ifdef EBCDIC |
8159 | literal_endpoint = 0; | |
8160 | #endif | |
8ada0baa | 8161 | } |
ffc61ed2 JH |
8162 | |
8163 | range = 0; /* this range (if it was one) is done now */ | |
a0d0e21e | 8164 | } |
ffc61ed2 | 8165 | |
936ed897 | 8166 | if (need_class) { |
4f66b38d | 8167 | ANYOF_FLAGS(ret) |= ANYOF_LARGE; |
936ed897 | 8168 | if (SIZE_ONLY) |
830247a4 | 8169 | RExC_size += ANYOF_CLASS_ADD_SKIP; |
936ed897 | 8170 | else |
830247a4 | 8171 | RExC_emit += ANYOF_CLASS_ADD_SKIP; |
936ed897 | 8172 | } |
ffc61ed2 | 8173 | |
7f6f358c YO |
8174 | |
8175 | if (SIZE_ONLY) | |
8176 | return ret; | |
8177 | /****** !SIZE_ONLY AFTER HERE *********/ | |
8178 | ||
3b57cd43 | 8179 | if( stored == 1 && (value < 128 || (value < 256 && !UTF)) |
7f6f358c YO |
8180 | && !( ANYOF_FLAGS(ret) & ( ANYOF_FLAGS_ALL ^ ANYOF_FOLD ) ) |
8181 | ) { | |
8182 | /* optimize single char class to an EXACT node | |
8183 | but *only* when its not a UTF/high char */ | |
07be1b83 YO |
8184 | const char * cur_parse= RExC_parse; |
8185 | RExC_emit = (regnode *)orig_emit; | |
8186 | RExC_parse = (char *)orig_parse; | |
7f6f358c YO |
8187 | ret = reg_node(pRExC_state, |
8188 | (U8)((ANYOF_FLAGS(ret) & ANYOF_FOLD) ? EXACTF : EXACT)); | |
07be1b83 | 8189 | RExC_parse = (char *)cur_parse; |
7f6f358c YO |
8190 | *STRING(ret)= (char)value; |
8191 | STR_LEN(ret)= 1; | |
8192 | RExC_emit += STR_SZ(1); | |
8193 | return ret; | |
8194 | } | |
ae5c130c | 8195 | /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */ |
7f6f358c | 8196 | if ( /* If the only flag is folding (plus possibly inversion). */ |
516a5887 JH |
8197 | ((ANYOF_FLAGS(ret) & (ANYOF_FLAGS_ALL ^ ANYOF_INVERT)) == ANYOF_FOLD) |
8198 | ) { | |
a0ed51b3 | 8199 | for (value = 0; value < 256; ++value) { |
936ed897 | 8200 | if (ANYOF_BITMAP_TEST(ret, value)) { |
eb160463 | 8201 | UV fold = PL_fold[value]; |
ffc61ed2 JH |
8202 | |
8203 | if (fold != value) | |
8204 | ANYOF_BITMAP_SET(ret, fold); | |
ae5c130c GS |
8205 | } |
8206 | } | |
936ed897 | 8207 | ANYOF_FLAGS(ret) &= ~ANYOF_FOLD; |
ae5c130c | 8208 | } |
ffc61ed2 | 8209 | |
ae5c130c | 8210 | /* optimize inverted simple patterns (e.g. [^a-z]) */ |
7f6f358c | 8211 | if (optimize_invert && |
ffc61ed2 JH |
8212 | /* If the only flag is inversion. */ |
8213 | (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) { | |
b8c5462f | 8214 | for (value = 0; value < ANYOF_BITMAP_SIZE; ++value) |
936ed897 | 8215 | ANYOF_BITMAP(ret)[value] ^= ANYOF_FLAGS_ALL; |
1aa99e6b | 8216 | ANYOF_FLAGS(ret) = ANYOF_UNICODE_ALL; |
ae5c130c | 8217 | } |
7f6f358c | 8218 | { |
097eb12c | 8219 | AV * const av = newAV(); |
ffc61ed2 | 8220 | SV *rv; |
9e55ce06 | 8221 | /* The 0th element stores the character class description |
6a0407ee | 8222 | * in its textual form: used later (regexec.c:Perl_regclass_swash()) |
9e55ce06 JH |
8223 | * to initialize the appropriate swash (which gets stored in |
8224 | * the 1st element), and also useful for dumping the regnode. | |
8225 | * The 2nd element stores the multicharacter foldings, | |
6a0407ee | 8226 | * used later (regexec.c:S_reginclass()). */ |
ffc61ed2 JH |
8227 | av_store(av, 0, listsv); |
8228 | av_store(av, 1, NULL); | |
9e55ce06 | 8229 | av_store(av, 2, (SV*)unicode_alternate); |
ffc61ed2 | 8230 | rv = newRV_noinc((SV*)av); |
19860706 | 8231 | n = add_data(pRExC_state, 1, "s"); |
f8fc2ecf | 8232 | RExC_rxi->data->data[n] = (void*)rv; |
ffc61ed2 | 8233 | ARG_SET(ret, n); |
a0ed51b3 | 8234 | } |
a0ed51b3 LW |
8235 | return ret; |
8236 | } | |
89836f1f YO |
8237 | #undef _C_C_T_ |
8238 | ||
a0ed51b3 | 8239 | |
bcdf7404 YO |
8240 | /* reg_skipcomment() |
8241 | ||
8242 | Absorbs an /x style # comments from the input stream. | |
8243 | Returns true if there is more text remaining in the stream. | |
8244 | Will set the REG_SEEN_RUN_ON_COMMENT flag if the comment | |
8245 | terminates the pattern without including a newline. | |
8246 | ||
8247 | Note its the callers responsibility to ensure that we are | |
8248 | actually in /x mode | |
8249 | ||
8250 | */ | |
8251 | ||
8252 | STATIC bool | |
8253 | S_reg_skipcomment(pTHX_ RExC_state_t *pRExC_state) | |
8254 | { | |
8255 | bool ended = 0; | |
8256 | while (RExC_parse < RExC_end) | |
8257 | if (*RExC_parse++ == '\n') { | |
8258 | ended = 1; | |
8259 | break; | |
8260 | } | |
8261 | if (!ended) { | |
8262 | /* we ran off the end of the pattern without ending | |
8263 | the comment, so we have to add an \n when wrapping */ | |
8264 | RExC_seen |= REG_SEEN_RUN_ON_COMMENT; | |
8265 | return 0; | |
8266 | } else | |
8267 | return 1; | |
8268 | } | |
8269 | ||
8270 | /* nextchar() | |
8271 | ||
8272 | Advance that parse position, and optionall absorbs | |
8273 | "whitespace" from the inputstream. | |
8274 | ||
8275 | Without /x "whitespace" means (?#...) style comments only, | |
8276 | with /x this means (?#...) and # comments and whitespace proper. | |
8277 | ||
8278 | Returns the RExC_parse point from BEFORE the scan occurs. | |
8279 | ||
8280 | This is the /x friendly way of saying RExC_parse++. | |
8281 | */ | |
8282 | ||
76e3520e | 8283 | STATIC char* |
830247a4 | 8284 | S_nextchar(pTHX_ RExC_state_t *pRExC_state) |
a0d0e21e | 8285 | { |
097eb12c | 8286 | char* const retval = RExC_parse++; |
a0d0e21e | 8287 | |
4633a7c4 | 8288 | for (;;) { |
830247a4 IZ |
8289 | if (*RExC_parse == '(' && RExC_parse[1] == '?' && |
8290 | RExC_parse[2] == '#') { | |
e994fd66 AE |
8291 | while (*RExC_parse != ')') { |
8292 | if (RExC_parse == RExC_end) | |
8293 | FAIL("Sequence (?#... not terminated"); | |
830247a4 | 8294 | RExC_parse++; |
e994fd66 | 8295 | } |
830247a4 | 8296 | RExC_parse++; |
4633a7c4 LW |
8297 | continue; |
8298 | } | |
bbe252da | 8299 | if (RExC_flags & RXf_PMf_EXTENDED) { |
830247a4 IZ |
8300 | if (isSPACE(*RExC_parse)) { |
8301 | RExC_parse++; | |
748a9306 LW |
8302 | continue; |
8303 | } | |
830247a4 | 8304 | else if (*RExC_parse == '#') { |
bcdf7404 YO |
8305 | if ( reg_skipcomment( pRExC_state ) ) |
8306 | continue; | |
748a9306 | 8307 | } |
748a9306 | 8308 | } |
4633a7c4 | 8309 | return retval; |
a0d0e21e | 8310 | } |
a687059c LW |
8311 | } |
8312 | ||
8313 | /* | |
c277df42 | 8314 | - reg_node - emit a node |
a0d0e21e | 8315 | */ |
76e3520e | 8316 | STATIC regnode * /* Location. */ |
830247a4 | 8317 | S_reg_node(pTHX_ RExC_state_t *pRExC_state, U8 op) |
a687059c | 8318 | { |
97aff369 | 8319 | dVAR; |
c277df42 | 8320 | register regnode *ptr; |
504618e9 | 8321 | regnode * const ret = RExC_emit; |
07be1b83 | 8322 | GET_RE_DEBUG_FLAGS_DECL; |
a687059c | 8323 | |
c277df42 | 8324 | if (SIZE_ONLY) { |
830247a4 IZ |
8325 | SIZE_ALIGN(RExC_size); |
8326 | RExC_size += 1; | |
a0d0e21e LW |
8327 | return(ret); |
8328 | } | |
3b57cd43 YO |
8329 | if (RExC_emit >= RExC_emit_bound) |
8330 | Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %d", op); | |
8331 | ||
c277df42 | 8332 | NODE_ALIGN_FILL(ret); |
a0d0e21e | 8333 | ptr = ret; |
c277df42 | 8334 | FILL_ADVANCE_NODE(ptr, op); |
7122b237 | 8335 | #ifdef RE_TRACK_PATTERN_OFFSETS |
fac92740 | 8336 | if (RExC_offsets) { /* MJD */ |
07be1b83 | 8337 | MJD_OFFSET_DEBUG(("%s:%d: (op %s) %s %"UVuf" (len %"UVuf") (max %"UVuf").\n", |
fac92740 | 8338 | "reg_node", __LINE__, |
13d6edb4 | 8339 | PL_reg_name[op], |
07be1b83 YO |
8340 | (UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0] |
8341 | ? "Overwriting end of array!\n" : "OK", | |
8342 | (UV)(RExC_emit - RExC_emit_start), | |
8343 | (UV)(RExC_parse - RExC_start), | |
8344 | (UV)RExC_offsets[0])); | |
ccb2c380 | 8345 | Set_Node_Offset(RExC_emit, RExC_parse + (op == END)); |
fac92740 | 8346 | } |
7122b237 | 8347 | #endif |
830247a4 | 8348 | RExC_emit = ptr; |
a0d0e21e | 8349 | return(ret); |
a687059c LW |
8350 | } |
8351 | ||
8352 | /* | |
a0d0e21e LW |
8353 | - reganode - emit a node with an argument |
8354 | */ | |
76e3520e | 8355 | STATIC regnode * /* Location. */ |
830247a4 | 8356 | S_reganode(pTHX_ RExC_state_t *pRExC_state, U8 op, U32 arg) |
fe14fcc3 | 8357 | { |
97aff369 | 8358 | dVAR; |
c277df42 | 8359 | register regnode *ptr; |
504618e9 | 8360 | regnode * const ret = RExC_emit; |
07be1b83 | 8361 | GET_RE_DEBUG_FLAGS_DECL; |
fe14fcc3 | 8362 | |
c277df42 | 8363 | if (SIZE_ONLY) { |
830247a4 IZ |
8364 | SIZE_ALIGN(RExC_size); |
8365 | RExC_size += 2; | |
6bda09f9 YO |
8366 | /* |
8367 | We can't do this: | |
8368 | ||
8369 | assert(2==regarglen[op]+1); | |
8370 | ||
8371 | Anything larger than this has to allocate the extra amount. | |
8372 | If we changed this to be: | |
8373 | ||
8374 | RExC_size += (1 + regarglen[op]); | |
8375 | ||
8376 | then it wouldn't matter. Its not clear what side effect | |
8377 | might come from that so its not done so far. | |
8378 | -- dmq | |
8379 | */ | |
a0d0e21e LW |
8380 | return(ret); |
8381 | } | |
3b57cd43 YO |
8382 | if (RExC_emit >= RExC_emit_bound) |
8383 | Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %d", op); | |
8384 | ||
c277df42 | 8385 | NODE_ALIGN_FILL(ret); |
a0d0e21e | 8386 | ptr = ret; |
c277df42 | 8387 | FILL_ADVANCE_NODE_ARG(ptr, op, arg); |
7122b237 | 8388 | #ifdef RE_TRACK_PATTERN_OFFSETS |
fac92740 | 8389 | if (RExC_offsets) { /* MJD */ |
07be1b83 | 8390 | MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n", |
fac92740 | 8391 | "reganode", |
ccb2c380 | 8392 | __LINE__, |
13d6edb4 | 8393 | PL_reg_name[op], |
07be1b83 | 8394 | (UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0] ? |
fac92740 | 8395 | "Overwriting end of array!\n" : "OK", |
07be1b83 YO |
8396 | (UV)(RExC_emit - RExC_emit_start), |
8397 | (UV)(RExC_parse - RExC_start), | |
8398 | (UV)RExC_offsets[0])); | |
ccb2c380 | 8399 | Set_Cur_Node_Offset; |
fac92740 | 8400 | } |
7122b237 | 8401 | #endif |
830247a4 | 8402 | RExC_emit = ptr; |
a0d0e21e | 8403 | return(ret); |
fe14fcc3 LW |
8404 | } |
8405 | ||
8406 | /* | |
cd439c50 | 8407 | - reguni - emit (if appropriate) a Unicode character |
a0ed51b3 | 8408 | */ |
71207a34 AL |
8409 | STATIC STRLEN |
8410 | S_reguni(pTHX_ const RExC_state_t *pRExC_state, UV uv, char* s) | |
a0ed51b3 | 8411 | { |
97aff369 | 8412 | dVAR; |
71207a34 | 8413 | return SIZE_ONLY ? UNISKIP(uv) : (uvchr_to_utf8((U8*)s, uv) - (U8*)s); |
a0ed51b3 LW |
8414 | } |
8415 | ||
8416 | /* | |
a0d0e21e LW |
8417 | - reginsert - insert an operator in front of already-emitted operand |
8418 | * | |
8419 | * Means relocating the operand. | |
8420 | */ | |
76e3520e | 8421 | STATIC void |
6bda09f9 | 8422 | S_reginsert(pTHX_ RExC_state_t *pRExC_state, U8 op, regnode *opnd, U32 depth) |
a687059c | 8423 | { |
97aff369 | 8424 | dVAR; |
c277df42 IZ |
8425 | register regnode *src; |
8426 | register regnode *dst; | |
8427 | register regnode *place; | |
504618e9 | 8428 | const int offset = regarglen[(U8)op]; |
6bda09f9 | 8429 | const int size = NODE_STEP_REGNODE + offset; |
07be1b83 | 8430 | GET_RE_DEBUG_FLAGS_DECL; |
def51078 | 8431 | PERL_UNUSED_ARG(depth); |
22c35a8c | 8432 | /* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */ |
13d6edb4 | 8433 | DEBUG_PARSE_FMT("inst"," - %s",PL_reg_name[op]); |
c277df42 | 8434 | if (SIZE_ONLY) { |
6bda09f9 | 8435 | RExC_size += size; |
a0d0e21e LW |
8436 | return; |
8437 | } | |
a687059c | 8438 | |
830247a4 | 8439 | src = RExC_emit; |
6bda09f9 | 8440 | RExC_emit += size; |
830247a4 | 8441 | dst = RExC_emit; |
40d049e4 | 8442 | if (RExC_open_parens) { |
6bda09f9 | 8443 | int paren; |
3b57cd43 | 8444 | /*DEBUG_PARSE_FMT("inst"," - %"IVdf, (IV)RExC_npar);*/ |
6bda09f9 | 8445 | for ( paren=0 ; paren < RExC_npar ; paren++ ) { |
40d049e4 | 8446 | if ( RExC_open_parens[paren] >= opnd ) { |
3b57cd43 | 8447 | /*DEBUG_PARSE_FMT("open"," - %d",size);*/ |
40d049e4 YO |
8448 | RExC_open_parens[paren] += size; |
8449 | } else { | |
3b57cd43 | 8450 | /*DEBUG_PARSE_FMT("open"," - %s","ok");*/ |
40d049e4 YO |
8451 | } |
8452 | if ( RExC_close_parens[paren] >= opnd ) { | |
3b57cd43 | 8453 | /*DEBUG_PARSE_FMT("close"," - %d",size);*/ |
40d049e4 YO |
8454 | RExC_close_parens[paren] += size; |
8455 | } else { | |
3b57cd43 | 8456 | /*DEBUG_PARSE_FMT("close"," - %s","ok");*/ |
40d049e4 YO |
8457 | } |
8458 | } | |
6bda09f9 | 8459 | } |
40d049e4 | 8460 | |
fac92740 | 8461 | while (src > opnd) { |
c277df42 | 8462 | StructCopy(--src, --dst, regnode); |
7122b237 | 8463 | #ifdef RE_TRACK_PATTERN_OFFSETS |
fac92740 | 8464 | if (RExC_offsets) { /* MJD 20010112 */ |
07be1b83 | 8465 | MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s copy %"UVuf" -> %"UVuf" (max %"UVuf").\n", |
fac92740 | 8466 | "reg_insert", |
ccb2c380 | 8467 | __LINE__, |
13d6edb4 | 8468 | PL_reg_name[op], |
07be1b83 YO |
8469 | (UV)(dst - RExC_emit_start) > RExC_offsets[0] |
8470 | ? "Overwriting end of array!\n" : "OK", | |
8471 | (UV)(src - RExC_emit_start), | |
8472 | (UV)(dst - RExC_emit_start), | |
8473 | (UV)RExC_offsets[0])); | |
ccb2c380 MP |
8474 | Set_Node_Offset_To_R(dst-RExC_emit_start, Node_Offset(src)); |
8475 | Set_Node_Length_To_R(dst-RExC_emit_start, Node_Length(src)); | |
fac92740 | 8476 | } |
7122b237 | 8477 | #endif |
fac92740 MJD |
8478 | } |
8479 | ||
a0d0e21e LW |
8480 | |
8481 | place = opnd; /* Op node, where operand used to be. */ | |
7122b237 | 8482 | #ifdef RE_TRACK_PATTERN_OFFSETS |
fac92740 | 8483 | if (RExC_offsets) { /* MJD */ |
07be1b83 | 8484 | MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n", |
fac92740 | 8485 | "reginsert", |
ccb2c380 | 8486 | __LINE__, |
13d6edb4 | 8487 | PL_reg_name[op], |
07be1b83 | 8488 | (UV)(place - RExC_emit_start) > RExC_offsets[0] |
fac92740 | 8489 | ? "Overwriting end of array!\n" : "OK", |
07be1b83 YO |
8490 | (UV)(place - RExC_emit_start), |
8491 | (UV)(RExC_parse - RExC_start), | |
786e8c11 | 8492 | (UV)RExC_offsets[0])); |
ccb2c380 | 8493 | Set_Node_Offset(place, RExC_parse); |
45948336 | 8494 | Set_Node_Length(place, 1); |
fac92740 | 8495 | } |
7122b237 | 8496 | #endif |
c277df42 IZ |
8497 | src = NEXTOPER(place); |
8498 | FILL_ADVANCE_NODE(place, op); | |
8499 | Zero(src, offset, regnode); | |
a687059c LW |
8500 | } |
8501 | ||
8502 | /* | |
c277df42 | 8503 | - regtail - set the next-pointer at the end of a node chain of p to val. |
3dab1dad | 8504 | - SEE ALSO: regtail_study |
a0d0e21e | 8505 | */ |
097eb12c | 8506 | /* TODO: All three parms should be const */ |
76e3520e | 8507 | STATIC void |
3dab1dad | 8508 | S_regtail(pTHX_ RExC_state_t *pRExC_state, regnode *p, const regnode *val,U32 depth) |
a687059c | 8509 | { |
97aff369 | 8510 | dVAR; |
c277df42 | 8511 | register regnode *scan; |
72f13be8 | 8512 | GET_RE_DEBUG_FLAGS_DECL; |
f9049ba1 SP |
8513 | #ifndef DEBUGGING |
8514 | PERL_UNUSED_ARG(depth); | |
8515 | #endif | |
a0d0e21e | 8516 | |
c277df42 | 8517 | if (SIZE_ONLY) |
a0d0e21e LW |
8518 | return; |
8519 | ||
8520 | /* Find last node. */ | |
8521 | scan = p; | |
8522 | for (;;) { | |
504618e9 | 8523 | regnode * const temp = regnext(scan); |
3dab1dad YO |
8524 | DEBUG_PARSE_r({ |
8525 | SV * const mysv=sv_newmortal(); | |
8526 | DEBUG_PARSE_MSG((scan==p ? "tail" : "")); | |
8527 | regprop(RExC_rx, mysv, scan); | |
eaf3ca90 YO |
8528 | PerlIO_printf(Perl_debug_log, "~ %s (%d) %s %s\n", |
8529 | SvPV_nolen_const(mysv), REG_NODE_NUM(scan), | |
8530 | (temp == NULL ? "->" : ""), | |
13d6edb4 | 8531 | (temp == NULL ? PL_reg_name[OP(val)] : "") |
eaf3ca90 | 8532 | ); |
3dab1dad YO |
8533 | }); |
8534 | if (temp == NULL) | |
8535 | break; | |
8536 | scan = temp; | |
8537 | } | |
8538 | ||
8539 | if (reg_off_by_arg[OP(scan)]) { | |
8540 | ARG_SET(scan, val - scan); | |
8541 | } | |
8542 | else { | |
8543 | NEXT_OFF(scan) = val - scan; | |
8544 | } | |
8545 | } | |
8546 | ||
07be1b83 | 8547 | #ifdef DEBUGGING |
3dab1dad YO |
8548 | /* |
8549 | - regtail_study - set the next-pointer at the end of a node chain of p to val. | |
8550 | - Look for optimizable sequences at the same time. | |
8551 | - currently only looks for EXACT chains. | |
07be1b83 YO |
8552 | |
8553 | This is expermental code. The idea is to use this routine to perform | |
8554 | in place optimizations on branches and groups as they are constructed, | |
8555 | with the long term intention of removing optimization from study_chunk so | |
8556 | that it is purely analytical. | |
8557 | ||
8558 | Currently only used when in DEBUG mode. The macro REGTAIL_STUDY() is used | |
8559 | to control which is which. | |
8560 | ||
3dab1dad YO |
8561 | */ |
8562 | /* TODO: All four parms should be const */ | |
07be1b83 | 8563 | |
3dab1dad YO |
8564 | STATIC U8 |
8565 | S_regtail_study(pTHX_ RExC_state_t *pRExC_state, regnode *p, const regnode *val,U32 depth) | |
8566 | { | |
8567 | dVAR; | |
8568 | register regnode *scan; | |
07be1b83 YO |
8569 | U8 exact = PSEUDO; |
8570 | #ifdef EXPERIMENTAL_INPLACESCAN | |
8571 | I32 min = 0; | |
8572 | #endif | |
8573 | ||
3dab1dad YO |
8574 | GET_RE_DEBUG_FLAGS_DECL; |
8575 | ||
07be1b83 | 8576 | |
3dab1dad YO |
8577 | if (SIZE_ONLY) |
8578 | return exact; | |
8579 | ||
8580 | /* Find last node. */ | |
8581 | ||
8582 | scan = p; | |
8583 | for (;;) { | |
8584 | regnode * const temp = regnext(scan); | |
07be1b83 YO |
8585 | #ifdef EXPERIMENTAL_INPLACESCAN |
8586 | if (PL_regkind[OP(scan)] == EXACT) | |
8587 | if (join_exact(pRExC_state,scan,&min,1,val,depth+1)) | |
8588 | return EXACT; | |
8589 | #endif | |
3dab1dad YO |
8590 | if ( exact ) { |
8591 | switch (OP(scan)) { | |
8592 | case EXACT: | |
8593 | case EXACTF: | |
8594 | case EXACTFL: | |
8595 | if( exact == PSEUDO ) | |
8596 | exact= OP(scan); | |
07be1b83 YO |
8597 | else if ( exact != OP(scan) ) |
8598 | exact= 0; | |
3dab1dad YO |
8599 | case NOTHING: |
8600 | break; | |
8601 | default: | |
8602 | exact= 0; | |
8603 | } | |
8604 | } | |
8605 | DEBUG_PARSE_r({ | |
8606 | SV * const mysv=sv_newmortal(); | |
8607 | DEBUG_PARSE_MSG((scan==p ? "tsdy" : "")); | |
8608 | regprop(RExC_rx, mysv, scan); | |
eaf3ca90 | 8609 | PerlIO_printf(Perl_debug_log, "~ %s (%d) -> %s\n", |
3dab1dad | 8610 | SvPV_nolen_const(mysv), |
eaf3ca90 | 8611 | REG_NODE_NUM(scan), |
13d6edb4 | 8612 | PL_reg_name[exact]); |
3dab1dad | 8613 | }); |
a0d0e21e LW |
8614 | if (temp == NULL) |
8615 | break; | |
8616 | scan = temp; | |
8617 | } | |
07be1b83 YO |
8618 | DEBUG_PARSE_r({ |
8619 | SV * const mysv_val=sv_newmortal(); | |
8620 | DEBUG_PARSE_MSG(""); | |
8621 | regprop(RExC_rx, mysv_val, val); | |
70685ca0 JH |
8622 | PerlIO_printf(Perl_debug_log, "~ attach to %s (%"IVdf") offset to %"IVdf"\n", |
8623 | SvPV_nolen_const(mysv_val), | |
8624 | (IV)REG_NODE_NUM(val), | |
8625 | (IV)(val - scan) | |
07be1b83 YO |
8626 | ); |
8627 | }); | |
c277df42 IZ |
8628 | if (reg_off_by_arg[OP(scan)]) { |
8629 | ARG_SET(scan, val - scan); | |
a0ed51b3 LW |
8630 | } |
8631 | else { | |
c277df42 IZ |
8632 | NEXT_OFF(scan) = val - scan; |
8633 | } | |
3dab1dad YO |
8634 | |
8635 | return exact; | |
a687059c | 8636 | } |
07be1b83 | 8637 | #endif |
a687059c LW |
8638 | |
8639 | /* | |
a687059c LW |
8640 | - regcurly - a little FSA that accepts {\d+,?\d*} |
8641 | */ | |
79072805 | 8642 | STATIC I32 |
5f66b61c | 8643 | S_regcurly(register const char *s) |
a687059c LW |
8644 | { |
8645 | if (*s++ != '{') | |
8646 | return FALSE; | |
f0fcb552 | 8647 | if (!isDIGIT(*s)) |
a687059c | 8648 | return FALSE; |
f0fcb552 | 8649 | while (isDIGIT(*s)) |
a687059c LW |
8650 | s++; |
8651 | if (*s == ',') | |
8652 | s++; | |
f0fcb552 | 8653 | while (isDIGIT(*s)) |
a687059c LW |
8654 | s++; |
8655 | if (*s != '}') | |
8656 | return FALSE; | |
8657 | return TRUE; | |
8658 | } | |
8659 | ||
a687059c LW |
8660 | |
8661 | /* | |
fd181c75 | 8662 | - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form |
a687059c | 8663 | */ |
f7819f85 A |
8664 | #ifdef DEBUGGING |
8665 | void | |
8666 | S_regdump_extflags(pTHX_ const char *lead, const U32 flags) { | |
8667 | int bit; | |
8668 | int set=0; | |
8669 | for (bit=0; bit<32; bit++) { | |
8670 | if (flags & (1<<bit)) { | |
8671 | if (!set++ && lead) | |
8672 | PerlIO_printf(Perl_debug_log, "%s",lead); | |
8673 | PerlIO_printf(Perl_debug_log, "%s ",PL_reg_extflags_name[bit]); | |
8674 | } | |
8675 | } | |
8676 | if (lead) { | |
8677 | if (set) | |
8678 | PerlIO_printf(Perl_debug_log, "\n"); | |
8679 | else | |
8680 | PerlIO_printf(Perl_debug_log, "%s[none-set]\n",lead); | |
8681 | } | |
8682 | } | |
8683 | #endif | |
8684 | ||
a687059c | 8685 | void |
097eb12c | 8686 | Perl_regdump(pTHX_ const regexp *r) |
a687059c | 8687 | { |
35ff7856 | 8688 | #ifdef DEBUGGING |
97aff369 | 8689 | dVAR; |
c445ea15 | 8690 | SV * const sv = sv_newmortal(); |
ab3bbdeb | 8691 | SV *dsv= sv_newmortal(); |
f8fc2ecf | 8692 | RXi_GET_DECL(r,ri); |
f7819f85 | 8693 | GET_RE_DEBUG_FLAGS_DECL; |
a687059c | 8694 | |
f8fc2ecf | 8695 | (void)dumpuntil(r, ri->program, ri->program + 1, NULL, NULL, sv, 0, 0); |
a0d0e21e LW |
8696 | |
8697 | /* Header fields of interest. */ | |
ab3bbdeb YO |
8698 | if (r->anchored_substr) { |
8699 | RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->anchored_substr), | |
8700 | RE_SV_DUMPLEN(r->anchored_substr), 30); | |
7b0972df | 8701 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb YO |
8702 | "anchored %s%s at %"IVdf" ", |
8703 | s, RE_SV_TAIL(r->anchored_substr), | |
7b0972df | 8704 | (IV)r->anchored_offset); |
ab3bbdeb YO |
8705 | } else if (r->anchored_utf8) { |
8706 | RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->anchored_utf8), | |
8707 | RE_SV_DUMPLEN(r->anchored_utf8), 30); | |
33b8afdf | 8708 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb YO |
8709 | "anchored utf8 %s%s at %"IVdf" ", |
8710 | s, RE_SV_TAIL(r->anchored_utf8), | |
33b8afdf | 8711 | (IV)r->anchored_offset); |
ab3bbdeb YO |
8712 | } |
8713 | if (r->float_substr) { | |
8714 | RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->float_substr), | |
8715 | RE_SV_DUMPLEN(r->float_substr), 30); | |
7b0972df | 8716 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb YO |
8717 | "floating %s%s at %"IVdf"..%"UVuf" ", |
8718 | s, RE_SV_TAIL(r->float_substr), | |
7b0972df | 8719 | (IV)r->float_min_offset, (UV)r->float_max_offset); |
ab3bbdeb YO |
8720 | } else if (r->float_utf8) { |
8721 | RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->float_utf8), | |
8722 | RE_SV_DUMPLEN(r->float_utf8), 30); | |
33b8afdf | 8723 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb YO |
8724 | "floating utf8 %s%s at %"IVdf"..%"UVuf" ", |
8725 | s, RE_SV_TAIL(r->float_utf8), | |
33b8afdf | 8726 | (IV)r->float_min_offset, (UV)r->float_max_offset); |
ab3bbdeb | 8727 | } |
33b8afdf | 8728 | if (r->check_substr || r->check_utf8) |
b81d288d | 8729 | PerlIO_printf(Perl_debug_log, |
10edeb5d JH |
8730 | (const char *) |
8731 | (r->check_substr == r->float_substr | |
8732 | && r->check_utf8 == r->float_utf8 | |
8733 | ? "(checking floating" : "(checking anchored")); | |
bbe252da | 8734 | if (r->extflags & RXf_NOSCAN) |
c277df42 | 8735 | PerlIO_printf(Perl_debug_log, " noscan"); |
bbe252da | 8736 | if (r->extflags & RXf_CHECK_ALL) |
c277df42 | 8737 | PerlIO_printf(Perl_debug_log, " isall"); |
33b8afdf | 8738 | if (r->check_substr || r->check_utf8) |
c277df42 IZ |
8739 | PerlIO_printf(Perl_debug_log, ") "); |
8740 | ||
f8fc2ecf YO |
8741 | if (ri->regstclass) { |
8742 | regprop(r, sv, ri->regstclass); | |
1de06328 | 8743 | PerlIO_printf(Perl_debug_log, "stclass %s ", SvPVX_const(sv)); |
46fc3d4c | 8744 | } |
bbe252da | 8745 | if (r->extflags & RXf_ANCH) { |
774d564b | 8746 | PerlIO_printf(Perl_debug_log, "anchored"); |
bbe252da | 8747 | if (r->extflags & RXf_ANCH_BOL) |
774d564b | 8748 | PerlIO_printf(Perl_debug_log, "(BOL)"); |
bbe252da | 8749 | if (r->extflags & RXf_ANCH_MBOL) |
c277df42 | 8750 | PerlIO_printf(Perl_debug_log, "(MBOL)"); |
bbe252da | 8751 | if (r->extflags & RXf_ANCH_SBOL) |
cad2e5aa | 8752 | PerlIO_printf(Perl_debug_log, "(SBOL)"); |
bbe252da | 8753 | if (r->extflags & RXf_ANCH_GPOS) |
774d564b | 8754 | PerlIO_printf(Perl_debug_log, "(GPOS)"); |
8755 | PerlIO_putc(Perl_debug_log, ' '); | |
8756 | } | |
bbe252da | 8757 | if (r->extflags & RXf_GPOS_SEEN) |
70685ca0 | 8758 | PerlIO_printf(Perl_debug_log, "GPOS:%"UVuf" ", (UV)r->gofs); |
bbe252da | 8759 | if (r->intflags & PREGf_SKIP) |
760ac839 | 8760 | PerlIO_printf(Perl_debug_log, "plus "); |
bbe252da | 8761 | if (r->intflags & PREGf_IMPLICIT) |
760ac839 | 8762 | PerlIO_printf(Perl_debug_log, "implicit "); |
70685ca0 | 8763 | PerlIO_printf(Perl_debug_log, "minlen %"IVdf" ", (IV)r->minlen); |
bbe252da | 8764 | if (r->extflags & RXf_EVAL_SEEN) |
ce862d02 | 8765 | PerlIO_printf(Perl_debug_log, "with eval "); |
760ac839 | 8766 | PerlIO_printf(Perl_debug_log, "\n"); |
f7819f85 | 8767 | DEBUG_FLAGS_r(regdump_extflags("r->extflags: ",r->extflags)); |
65e66c80 | 8768 | #else |
96a5add6 | 8769 | PERL_UNUSED_CONTEXT; |
65e66c80 | 8770 | PERL_UNUSED_ARG(r); |
17c3b450 | 8771 | #endif /* DEBUGGING */ |
a687059c LW |
8772 | } |
8773 | ||
8774 | /* | |
a0d0e21e LW |
8775 | - regprop - printable representation of opcode |
8776 | */ | |
46fc3d4c | 8777 | void |
32fc9b6a | 8778 | Perl_regprop(pTHX_ const regexp *prog, SV *sv, const regnode *o) |
a687059c | 8779 | { |
35ff7856 | 8780 | #ifdef DEBUGGING |
97aff369 | 8781 | dVAR; |
9b155405 | 8782 | register int k; |
f8fc2ecf | 8783 | RXi_GET_DECL(prog,progi); |
1de06328 | 8784 | GET_RE_DEBUG_FLAGS_DECL; |
f8fc2ecf | 8785 | |
a0d0e21e | 8786 | |
54dc92de | 8787 | sv_setpvn(sv, "", 0); |
8aa23a47 | 8788 | |
03363afd | 8789 | if (OP(o) > REGNODE_MAX) /* regnode.type is unsigned */ |
830247a4 IZ |
8790 | /* It would be nice to FAIL() here, but this may be called from |
8791 | regexec.c, and it would be hard to supply pRExC_state. */ | |
a5ca303d | 8792 | Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d", (int)OP(o), (int)REGNODE_MAX); |
13d6edb4 | 8793 | sv_catpv(sv, PL_reg_name[OP(o)]); /* Take off const! */ |
9b155405 | 8794 | |
3dab1dad | 8795 | k = PL_regkind[OP(o)]; |
9b155405 | 8796 | |
2a782b5b | 8797 | if (k == EXACT) { |
396482e1 | 8798 | SV * const dsv = sv_2mortal(newSVpvs("")); |
ab3bbdeb YO |
8799 | /* Using is_utf8_string() (via PERL_PV_UNI_DETECT) |
8800 | * is a crude hack but it may be the best for now since | |
8801 | * we have no flag "this EXACTish node was UTF-8" | |
8802 | * --jhi */ | |
8803 | const char * const s = | |
ddc5bc0f | 8804 | pv_pretty(dsv, STRING(o), STR_LEN(o), 60, |
ab3bbdeb YO |
8805 | PL_colors[0], PL_colors[1], |
8806 | PERL_PV_ESCAPE_UNI_DETECT | | |
95b611b0 | 8807 | PERL_PV_PRETTY_ELLIPSES | |
ab3bbdeb YO |
8808 | PERL_PV_PRETTY_LTGT |
8809 | ); | |
8810 | Perl_sv_catpvf(aTHX_ sv, " %s", s ); | |
bb263b4e | 8811 | } else if (k == TRIE) { |
3dab1dad | 8812 | /* print the details of the trie in dumpuntil instead, as |
f8fc2ecf | 8813 | * progi->data isn't available here */ |
1de06328 | 8814 | const char op = OP(o); |
647f639f | 8815 | const U32 n = ARG(o); |
1de06328 | 8816 | const reg_ac_data * const ac = IS_TRIE_AC(op) ? |
f8fc2ecf | 8817 | (reg_ac_data *)progi->data->data[n] : |
1de06328 | 8818 | NULL; |
3251b653 NC |
8819 | const reg_trie_data * const trie |
8820 | = (reg_trie_data*)progi->data->data[!IS_TRIE_AC(op) ? n : ac->trie]; | |
1de06328 | 8821 | |
13d6edb4 | 8822 | Perl_sv_catpvf(aTHX_ sv, "-%s",PL_reg_name[o->flags]); |
1de06328 YO |
8823 | DEBUG_TRIE_COMPILE_r( |
8824 | Perl_sv_catpvf(aTHX_ sv, | |
8825 | "<S:%"UVuf"/%"IVdf" W:%"UVuf" L:%"UVuf"/%"UVuf" C:%"UVuf"/%"UVuf">", | |
8826 | (UV)trie->startstate, | |
1e2e3d02 | 8827 | (IV)trie->statecount-1, /* -1 because of the unused 0 element */ |
1de06328 YO |
8828 | (UV)trie->wordcount, |
8829 | (UV)trie->minlen, | |
8830 | (UV)trie->maxlen, | |
8831 | (UV)TRIE_CHARCOUNT(trie), | |
8832 | (UV)trie->uniquecharcount | |
8833 | ) | |
8834 | ); | |
8835 | if ( IS_ANYOF_TRIE(op) || trie->bitmap ) { | |
8836 | int i; | |
8837 | int rangestart = -1; | |
f46cb337 | 8838 | U8* bitmap = IS_ANYOF_TRIE(op) ? (U8*)ANYOF_BITMAP(o) : (U8*)TRIE_BITMAP(trie); |
1de06328 YO |
8839 | Perl_sv_catpvf(aTHX_ sv, "["); |
8840 | for (i = 0; i <= 256; i++) { | |
8841 | if (i < 256 && BITMAP_TEST(bitmap,i)) { | |
8842 | if (rangestart == -1) | |
8843 | rangestart = i; | |
8844 | } else if (rangestart != -1) { | |
8845 | if (i <= rangestart + 3) | |
8846 | for (; rangestart < i; rangestart++) | |
8847 | put_byte(sv, rangestart); | |
8848 | else { | |
8849 | put_byte(sv, rangestart); | |
8850 | sv_catpvs(sv, "-"); | |
8851 | put_byte(sv, i - 1); | |
8852 | } | |
8853 | rangestart = -1; | |
8854 | } | |
8855 | } | |
8856 | Perl_sv_catpvf(aTHX_ sv, "]"); | |
8857 | } | |
8858 | ||
a3621e74 | 8859 | } else if (k == CURLY) { |
cb434fcc | 8860 | if (OP(o) == CURLYM || OP(o) == CURLYN || OP(o) == CURLYX) |
cea2e8a9 GS |
8861 | Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */ |
8862 | Perl_sv_catpvf(aTHX_ sv, " {%d,%d}", ARG1(o), ARG2(o)); | |
a0d0e21e | 8863 | } |
2c2d71f5 JH |
8864 | else if (k == WHILEM && o->flags) /* Ordinal/of */ |
8865 | Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4); | |
1f1031fe | 8866 | else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP || OP(o)==ACCEPT) { |
894356b3 | 8867 | Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o)); /* Parenth number */ |
1f1031fe | 8868 | if ( prog->paren_names ) { |
ee9b8eae YO |
8869 | if ( k != REF || OP(o) < NREF) { |
8870 | AV *list= (AV *)progi->data->data[progi->name_list_idx]; | |
8871 | SV **name= av_fetch(list, ARG(o), 0 ); | |
8872 | if (name) | |
8873 | Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name)); | |
8874 | } | |
8875 | else { | |
8876 | AV *list= (AV *)progi->data->data[ progi->name_list_idx ]; | |
8877 | SV *sv_dat=(SV*)progi->data->data[ ARG( o ) ]; | |
8878 | I32 *nums=(I32*)SvPVX(sv_dat); | |
8879 | SV **name= av_fetch(list, nums[0], 0 ); | |
8880 | I32 n; | |
8881 | if (name) { | |
8882 | for ( n=0; n<SvIVX(sv_dat); n++ ) { | |
8883 | Perl_sv_catpvf(aTHX_ sv, "%s%"IVdf, | |
8884 | (n ? "," : ""), (IV)nums[n]); | |
8885 | } | |
8886 | Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name)); | |
1f1031fe | 8887 | } |
1f1031fe | 8888 | } |
ee9b8eae | 8889 | } |
1f1031fe | 8890 | } else if (k == GOSUB) |
6bda09f9 | 8891 | Perl_sv_catpvf(aTHX_ sv, "%d[%+d]", (int)ARG(o),(int)ARG2L(o)); /* Paren and offset */ |
e2e6a0f1 YO |
8892 | else if (k == VERB) { |
8893 | if (!o->flags) | |
8894 | Perl_sv_catpvf(aTHX_ sv, ":%"SVf, | |
be2597df | 8895 | SVfARG((SV*)progi->data->data[ ARG( o ) ])); |
e2e6a0f1 | 8896 | } else if (k == LOGICAL) |
04ebc1ab | 8897 | Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* 2: embedded, otherwise 1 */ |
f9a79580 | 8898 | else if (k == FOLDCHAR) |
df44d732 | 8899 | Perl_sv_catpvf(aTHX_ sv, "[0x%"UVXf"]", PTR2UV(ARG(o)) ); |
653099ff GS |
8900 | else if (k == ANYOF) { |
8901 | int i, rangestart = -1; | |
2d03de9c | 8902 | const U8 flags = ANYOF_FLAGS(o); |
0bd48802 AL |
8903 | |
8904 | /* Should be synchronized with * ANYOF_ #xdefines in regcomp.h */ | |
8905 | static const char * const anyofs[] = { | |
653099ff GS |
8906 | "\\w", |
8907 | "\\W", | |
8908 | "\\s", | |
8909 | "\\S", | |
8910 | "\\d", | |
8911 | "\\D", | |
8912 | "[:alnum:]", | |
8913 | "[:^alnum:]", | |
8914 | "[:alpha:]", | |
8915 | "[:^alpha:]", | |
8916 | "[:ascii:]", | |
8917 | "[:^ascii:]", | |
8918 | "[:ctrl:]", | |
8919 | "[:^ctrl:]", | |
8920 | "[:graph:]", | |
8921 | "[:^graph:]", | |
8922 | "[:lower:]", | |
8923 | "[:^lower:]", | |
8924 | "[:print:]", | |
8925 | "[:^print:]", | |
8926 | "[:punct:]", | |
8927 | "[:^punct:]", | |
8928 | "[:upper:]", | |
aaa51d5e | 8929 | "[:^upper:]", |
653099ff | 8930 | "[:xdigit:]", |
aaa51d5e JF |
8931 | "[:^xdigit:]", |
8932 | "[:space:]", | |
8933 | "[:^space:]", | |
8934 | "[:blank:]", | |
8935 | "[:^blank:]" | |
653099ff GS |
8936 | }; |
8937 | ||
19860706 | 8938 | if (flags & ANYOF_LOCALE) |
396482e1 | 8939 | sv_catpvs(sv, "{loc}"); |
19860706 | 8940 | if (flags & ANYOF_FOLD) |
396482e1 | 8941 | sv_catpvs(sv, "{i}"); |
653099ff | 8942 | Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]); |
19860706 | 8943 | if (flags & ANYOF_INVERT) |
396482e1 | 8944 | sv_catpvs(sv, "^"); |
ffc61ed2 JH |
8945 | for (i = 0; i <= 256; i++) { |
8946 | if (i < 256 && ANYOF_BITMAP_TEST(o,i)) { | |
8947 | if (rangestart == -1) | |
8948 | rangestart = i; | |
8949 | } else if (rangestart != -1) { | |
8950 | if (i <= rangestart + 3) | |
8951 | for (; rangestart < i; rangestart++) | |
653099ff | 8952 | put_byte(sv, rangestart); |
ffc61ed2 JH |
8953 | else { |
8954 | put_byte(sv, rangestart); | |
396482e1 | 8955 | sv_catpvs(sv, "-"); |
ffc61ed2 | 8956 | put_byte(sv, i - 1); |
653099ff | 8957 | } |
ffc61ed2 | 8958 | rangestart = -1; |
653099ff | 8959 | } |
847a199f | 8960 | } |
ffc61ed2 JH |
8961 | |
8962 | if (o->flags & ANYOF_CLASS) | |
bb7a0f54 | 8963 | for (i = 0; i < (int)(sizeof(anyofs)/sizeof(char*)); i++) |
ffc61ed2 JH |
8964 | if (ANYOF_CLASS_TEST(o,i)) |
8965 | sv_catpv(sv, anyofs[i]); | |
8966 | ||
8967 | if (flags & ANYOF_UNICODE) | |
396482e1 | 8968 | sv_catpvs(sv, "{unicode}"); |
1aa99e6b | 8969 | else if (flags & ANYOF_UNICODE_ALL) |
396482e1 | 8970 | sv_catpvs(sv, "{unicode_all}"); |
ffc61ed2 JH |
8971 | |
8972 | { | |
8973 | SV *lv; | |
32fc9b6a | 8974 | SV * const sw = regclass_swash(prog, o, FALSE, &lv, 0); |
b81d288d | 8975 | |
ffc61ed2 JH |
8976 | if (lv) { |
8977 | if (sw) { | |
89ebb4a3 | 8978 | U8 s[UTF8_MAXBYTES_CASE+1]; |
b81d288d | 8979 | |
ffc61ed2 | 8980 | for (i = 0; i <= 256; i++) { /* just the first 256 */ |
1df70142 | 8981 | uvchr_to_utf8(s, i); |
ffc61ed2 | 8982 | |
3568d838 | 8983 | if (i < 256 && swash_fetch(sw, s, TRUE)) { |
ffc61ed2 JH |
8984 | if (rangestart == -1) |
8985 | rangestart = i; | |
8986 | } else if (rangestart != -1) { | |
ffc61ed2 JH |
8987 | if (i <= rangestart + 3) |
8988 | for (; rangestart < i; rangestart++) { | |
2d03de9c AL |
8989 | const U8 * const e = uvchr_to_utf8(s,rangestart); |
8990 | U8 *p; | |
8991 | for(p = s; p < e; p++) | |
ffc61ed2 JH |
8992 | put_byte(sv, *p); |
8993 | } | |
8994 | else { | |
2d03de9c AL |
8995 | const U8 *e = uvchr_to_utf8(s,rangestart); |
8996 | U8 *p; | |
8997 | for (p = s; p < e; p++) | |
ffc61ed2 | 8998 | put_byte(sv, *p); |
396482e1 | 8999 | sv_catpvs(sv, "-"); |
2d03de9c AL |
9000 | e = uvchr_to_utf8(s, i-1); |
9001 | for (p = s; p < e; p++) | |
1df70142 | 9002 | put_byte(sv, *p); |
ffc61ed2 JH |
9003 | } |
9004 | rangestart = -1; | |
9005 | } | |
19860706 | 9006 | } |
ffc61ed2 | 9007 | |
396482e1 | 9008 | sv_catpvs(sv, "..."); /* et cetera */ |
19860706 | 9009 | } |
fde631ed | 9010 | |
ffc61ed2 | 9011 | { |
2e0de35c | 9012 | char *s = savesvpv(lv); |
c445ea15 | 9013 | char * const origs = s; |
b81d288d | 9014 | |
3dab1dad YO |
9015 | while (*s && *s != '\n') |
9016 | s++; | |
b81d288d | 9017 | |
ffc61ed2 | 9018 | if (*s == '\n') { |
2d03de9c | 9019 | const char * const t = ++s; |
ffc61ed2 JH |
9020 | |
9021 | while (*s) { | |
9022 | if (*s == '\n') | |
9023 | *s = ' '; | |
9024 | s++; | |
9025 | } | |
9026 | if (s[-1] == ' ') | |
9027 | s[-1] = 0; | |
9028 | ||
9029 | sv_catpv(sv, t); | |
fde631ed | 9030 | } |
b81d288d | 9031 | |
ffc61ed2 | 9032 | Safefree(origs); |
fde631ed JH |
9033 | } |
9034 | } | |
653099ff | 9035 | } |
ffc61ed2 | 9036 | |
653099ff GS |
9037 | Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]); |
9038 | } | |
9b155405 | 9039 | else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH)) |
07be1b83 | 9040 | Perl_sv_catpvf(aTHX_ sv, "[%d]", -(o->flags)); |
65e66c80 | 9041 | #else |
96a5add6 | 9042 | PERL_UNUSED_CONTEXT; |
65e66c80 SP |
9043 | PERL_UNUSED_ARG(sv); |
9044 | PERL_UNUSED_ARG(o); | |
f9049ba1 | 9045 | PERL_UNUSED_ARG(prog); |
17c3b450 | 9046 | #endif /* DEBUGGING */ |
35ff7856 | 9047 | } |
a687059c | 9048 | |
cad2e5aa | 9049 | SV * |
49d7dfbc | 9050 | Perl_re_intuit_string(pTHX_ REGEXP * const prog) |
cad2e5aa | 9051 | { /* Assume that RE_INTUIT is set */ |
97aff369 | 9052 | dVAR; |
a3621e74 | 9053 | GET_RE_DEBUG_FLAGS_DECL; |
96a5add6 AL |
9054 | PERL_UNUSED_CONTEXT; |
9055 | ||
a3621e74 | 9056 | DEBUG_COMPILE_r( |
cfd0369c | 9057 | { |
2d03de9c | 9058 | const char * const s = SvPV_nolen_const(prog->check_substr |
cfd0369c | 9059 | ? prog->check_substr : prog->check_utf8); |
cad2e5aa JH |
9060 | |
9061 | if (!PL_colorset) reginitcolors(); | |
9062 | PerlIO_printf(Perl_debug_log, | |
a0288114 | 9063 | "%sUsing REx %ssubstr:%s \"%s%.60s%s%s\"\n", |
33b8afdf JH |
9064 | PL_colors[4], |
9065 | prog->check_substr ? "" : "utf8 ", | |
9066 | PL_colors[5],PL_colors[0], | |
cad2e5aa JH |
9067 | s, |
9068 | PL_colors[1], | |
9069 | (strlen(s) > 60 ? "..." : "")); | |
9070 | } ); | |
9071 | ||
33b8afdf | 9072 | return prog->check_substr ? prog->check_substr : prog->check_utf8; |
cad2e5aa JH |
9073 | } |
9074 | ||
84da74a7 | 9075 | /* |
f8149455 | 9076 | pregfree() |
84da74a7 | 9077 | |
f8149455 YO |
9078 | handles refcounting and freeing the perl core regexp structure. When |
9079 | it is necessary to actually free the structure the first thing it | |
9080 | does is call the 'free' method of the regexp_engine associated to to | |
9081 | the regexp, allowing the handling of the void *pprivate; member | |
9082 | first. (This routine is not overridable by extensions, which is why | |
9083 | the extensions free is called first.) | |
9084 | ||
9085 | See regdupe and regdupe_internal if you change anything here. | |
84da74a7 | 9086 | */ |
f8149455 | 9087 | #ifndef PERL_IN_XSUB_RE |
2b69d0c2 | 9088 | void |
864dbfa3 | 9089 | Perl_pregfree(pTHX_ struct regexp *r) |
a687059c | 9090 | { |
27da23d5 | 9091 | dVAR; |
fc32ee4a | 9092 | GET_RE_DEBUG_FLAGS_DECL; |
a3621e74 | 9093 | |
7821416a IZ |
9094 | if (!r || (--r->refcnt > 0)) |
9095 | return; | |
28d8d7f4 YO |
9096 | if (r->mother_re) { |
9097 | ReREFCNT_dec(r->mother_re); | |
9098 | } else { | |
9099 | CALLREGFREE_PVT(r); /* free the private data */ | |
9100 | if (r->paren_names) | |
9101 | SvREFCNT_dec(r->paren_names); | |
9102 | Safefree(r->wrapped); | |
9103 | } | |
9104 | if (r->substrs) { | |
9105 | if (r->anchored_substr) | |
9106 | SvREFCNT_dec(r->anchored_substr); | |
9107 | if (r->anchored_utf8) | |
9108 | SvREFCNT_dec(r->anchored_utf8); | |
9109 | if (r->float_substr) | |
9110 | SvREFCNT_dec(r->float_substr); | |
9111 | if (r->float_utf8) | |
9112 | SvREFCNT_dec(r->float_utf8); | |
9113 | Safefree(r->substrs); | |
9114 | } | |
ed252734 | 9115 | RX_MATCH_COPY_FREE(r); |
f8c7b90f | 9116 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 | 9117 | if (r->saved_copy) |
28d8d7f4 | 9118 | SvREFCNT_dec(r->saved_copy); |
ed252734 | 9119 | #endif |
f0ab9afb NC |
9120 | Safefree(r->swap); |
9121 | Safefree(r->offs); | |
f8149455 YO |
9122 | Safefree(r); |
9123 | } | |
28d8d7f4 YO |
9124 | |
9125 | /* reg_temp_copy() | |
9126 | ||
9127 | This is a hacky workaround to the structural issue of match results | |
9128 | being stored in the regexp structure which is in turn stored in | |
9129 | PL_curpm/PL_reg_curpm. The problem is that due to qr// the pattern | |
9130 | could be PL_curpm in multiple contexts, and could require multiple | |
9131 | result sets being associated with the pattern simultaneously, such | |
9132 | as when doing a recursive match with (??{$qr}) | |
9133 | ||
9134 | The solution is to make a lightweight copy of the regexp structure | |
9135 | when a qr// is returned from the code executed by (??{$qr}) this | |
9136 | lightweight copy doesnt actually own any of its data except for | |
9137 | the starp/end and the actual regexp structure itself. | |
9138 | ||
9139 | */ | |
9140 | ||
9141 | ||
9142 | regexp * | |
9143 | Perl_reg_temp_copy (pTHX_ struct regexp *r) { | |
9144 | regexp *ret; | |
9145 | register const I32 npar = r->nparens+1; | |
9146 | (void)ReREFCNT_inc(r); | |
9147 | Newx(ret, 1, regexp); | |
9148 | StructCopy(r, ret, regexp); | |
f0ab9afb NC |
9149 | Newx(ret->offs, npar, regexp_paren_pair); |
9150 | Copy(r->offs, ret->offs, npar, regexp_paren_pair); | |
28d8d7f4 YO |
9151 | ret->refcnt = 1; |
9152 | if (r->substrs) { | |
28d8d7f4 | 9153 | Newx(ret->substrs, 1, struct reg_substr_data); |
6ab65676 NC |
9154 | StructCopy(r->substrs, ret->substrs, struct reg_substr_data); |
9155 | ||
9156 | SvREFCNT_inc_void(ret->anchored_substr); | |
9157 | SvREFCNT_inc_void(ret->anchored_utf8); | |
9158 | SvREFCNT_inc_void(ret->float_substr); | |
9159 | SvREFCNT_inc_void(ret->float_utf8); | |
9160 | ||
9161 | /* check_substr and check_utf8, if non-NULL, point to either their | |
9162 | anchored or float namesakes, and don't hold a second reference. */ | |
486913e4 | 9163 | } |
28d8d7f4 YO |
9164 | RX_MATCH_COPIED_off(ret); |
9165 | #ifdef PERL_OLD_COPY_ON_WRITE | |
b89b0c6f | 9166 | ret->saved_copy = NULL; |
28d8d7f4 YO |
9167 | #endif |
9168 | ret->mother_re = r; | |
9169 | ret->swap = NULL; | |
9170 | ||
9171 | return ret; | |
9172 | } | |
f8149455 YO |
9173 | #endif |
9174 | ||
9175 | /* regfree_internal() | |
9176 | ||
9177 | Free the private data in a regexp. This is overloadable by | |
9178 | extensions. Perl takes care of the regexp structure in pregfree(), | |
9179 | this covers the *pprivate pointer which technically perldoesnt | |
9180 | know about, however of course we have to handle the | |
9181 | regexp_internal structure when no extension is in use. | |
9182 | ||
9183 | Note this is called before freeing anything in the regexp | |
9184 | structure. | |
9185 | */ | |
9186 | ||
9187 | void | |
973f7e2e | 9188 | Perl_regfree_internal(pTHX_ REGEXP * const r) |
f8149455 YO |
9189 | { |
9190 | dVAR; | |
9191 | RXi_GET_DECL(r,ri); | |
9192 | GET_RE_DEBUG_FLAGS_DECL; | |
9193 | ||
9194 | DEBUG_COMPILE_r({ | |
9195 | if (!PL_colorset) | |
9196 | reginitcolors(); | |
9197 | { | |
9198 | SV *dsv= sv_newmortal(); | |
9199 | RE_PV_QUOTED_DECL(s, (r->extflags & RXf_UTF8), | |
9200 | dsv, r->precomp, r->prelen, 60); | |
9201 | PerlIO_printf(Perl_debug_log,"%sFreeing REx:%s %s\n", | |
9202 | PL_colors[4],PL_colors[5],s); | |
9203 | } | |
9204 | }); | |
7122b237 YO |
9205 | #ifdef RE_TRACK_PATTERN_OFFSETS |
9206 | if (ri->u.offsets) | |
9207 | Safefree(ri->u.offsets); /* 20010421 MJD */ | |
9208 | #endif | |
f8fc2ecf YO |
9209 | if (ri->data) { |
9210 | int n = ri->data->count; | |
f3548bdc DM |
9211 | PAD* new_comppad = NULL; |
9212 | PAD* old_comppad; | |
4026c95a | 9213 | PADOFFSET refcnt; |
dfad63ad | 9214 | |
c277df42 | 9215 | while (--n >= 0) { |
261faec3 | 9216 | /* If you add a ->what type here, update the comment in regcomp.h */ |
f8fc2ecf | 9217 | switch (ri->data->what[n]) { |
c277df42 | 9218 | case 's': |
81714fb9 | 9219 | case 'S': |
55eed653 | 9220 | case 'u': |
f8fc2ecf | 9221 | SvREFCNT_dec((SV*)ri->data->data[n]); |
c277df42 | 9222 | break; |
653099ff | 9223 | case 'f': |
f8fc2ecf | 9224 | Safefree(ri->data->data[n]); |
653099ff | 9225 | break; |
dfad63ad | 9226 | case 'p': |
f8fc2ecf | 9227 | new_comppad = (AV*)ri->data->data[n]; |
dfad63ad | 9228 | break; |
c277df42 | 9229 | case 'o': |
dfad63ad | 9230 | if (new_comppad == NULL) |
cea2e8a9 | 9231 | Perl_croak(aTHX_ "panic: pregfree comppad"); |
f3548bdc DM |
9232 | PAD_SAVE_LOCAL(old_comppad, |
9233 | /* Watch out for global destruction's random ordering. */ | |
c445ea15 | 9234 | (SvTYPE(new_comppad) == SVt_PVAV) ? new_comppad : NULL |
f3548bdc | 9235 | ); |
b34c0dd4 | 9236 | OP_REFCNT_LOCK; |
f8fc2ecf | 9237 | refcnt = OpREFCNT_dec((OP_4tree*)ri->data->data[n]); |
4026c95a SH |
9238 | OP_REFCNT_UNLOCK; |
9239 | if (!refcnt) | |
f8fc2ecf | 9240 | op_free((OP_4tree*)ri->data->data[n]); |
9b978d73 | 9241 | |
f3548bdc | 9242 | PAD_RESTORE_LOCAL(old_comppad); |
dfad63ad HS |
9243 | SvREFCNT_dec((SV*)new_comppad); |
9244 | new_comppad = NULL; | |
c277df42 IZ |
9245 | break; |
9246 | case 'n': | |
9e55ce06 | 9247 | break; |
07be1b83 | 9248 | case 'T': |
be8e71aa YO |
9249 | { /* Aho Corasick add-on structure for a trie node. |
9250 | Used in stclass optimization only */ | |
07be1b83 | 9251 | U32 refcount; |
f8fc2ecf | 9252 | reg_ac_data *aho=(reg_ac_data*)ri->data->data[n]; |
07be1b83 YO |
9253 | OP_REFCNT_LOCK; |
9254 | refcount = --aho->refcount; | |
9255 | OP_REFCNT_UNLOCK; | |
9256 | if ( !refcount ) { | |
446bd890 NC |
9257 | PerlMemShared_free(aho->states); |
9258 | PerlMemShared_free(aho->fail); | |
446bd890 NC |
9259 | /* do this last!!!! */ |
9260 | PerlMemShared_free(ri->data->data[n]); | |
9261 | PerlMemShared_free(ri->regstclass); | |
07be1b83 YO |
9262 | } |
9263 | } | |
9264 | break; | |
a3621e74 | 9265 | case 't': |
07be1b83 | 9266 | { |
be8e71aa | 9267 | /* trie structure. */ |
07be1b83 | 9268 | U32 refcount; |
f8fc2ecf | 9269 | reg_trie_data *trie=(reg_trie_data*)ri->data->data[n]; |
07be1b83 YO |
9270 | OP_REFCNT_LOCK; |
9271 | refcount = --trie->refcount; | |
9272 | OP_REFCNT_UNLOCK; | |
9273 | if ( !refcount ) { | |
446bd890 | 9274 | PerlMemShared_free(trie->charmap); |
446bd890 NC |
9275 | PerlMemShared_free(trie->states); |
9276 | PerlMemShared_free(trie->trans); | |
07be1b83 | 9277 | if (trie->bitmap) |
446bd890 | 9278 | PerlMemShared_free(trie->bitmap); |
07be1b83 | 9279 | if (trie->wordlen) |
446bd890 | 9280 | PerlMemShared_free(trie->wordlen); |
786e8c11 | 9281 | if (trie->jump) |
446bd890 | 9282 | PerlMemShared_free(trie->jump); |
786e8c11 | 9283 | if (trie->nextword) |
446bd890 | 9284 | PerlMemShared_free(trie->nextword); |
446bd890 NC |
9285 | /* do this last!!!! */ |
9286 | PerlMemShared_free(ri->data->data[n]); | |
a3621e74 | 9287 | } |
07be1b83 YO |
9288 | } |
9289 | break; | |
c277df42 | 9290 | default: |
f8fc2ecf | 9291 | Perl_croak(aTHX_ "panic: regfree data code '%c'", ri->data->what[n]); |
c277df42 IZ |
9292 | } |
9293 | } | |
f8fc2ecf YO |
9294 | Safefree(ri->data->what); |
9295 | Safefree(ri->data); | |
a0d0e21e | 9296 | } |
28d8d7f4 | 9297 | |
f8fc2ecf | 9298 | Safefree(ri); |
a687059c | 9299 | } |
c277df42 | 9300 | |
84da74a7 YO |
9301 | #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t)) |
9302 | #define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t)) | |
81714fb9 | 9303 | #define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t)) |
84da74a7 YO |
9304 | #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL) |
9305 | ||
9306 | /* | |
32cd70f6 | 9307 | re_dup - duplicate a regexp. |
84da74a7 | 9308 | |
32cd70f6 NC |
9309 | This routine is expected to clone a given regexp structure. It is not |
9310 | compiler under USE_ITHREADS. | |
9311 | ||
f8149455 YO |
9312 | After all of the core data stored in struct regexp is duplicated |
9313 | the regexp_engine.dupe method is used to copy any private data | |
9314 | stored in the *pprivate pointer. This allows extensions to handle | |
9315 | any duplication it needs to do. | |
9316 | ||
9317 | See pregfree() and regfree_internal() if you change anything here. | |
84da74a7 | 9318 | */ |
a3c0e9ca | 9319 | #if defined(USE_ITHREADS) |
f8149455 | 9320 | #ifndef PERL_IN_XSUB_RE |
84da74a7 | 9321 | regexp * |
f8149455 | 9322 | Perl_re_dup(pTHX_ const regexp *r, CLONE_PARAMS *param) |
84da74a7 | 9323 | { |
84da74a7 | 9324 | dVAR; |
f8fc2ecf | 9325 | regexp *ret; |
a86a1ca7 | 9326 | I32 npar; |
644c02aa | 9327 | |
84da74a7 YO |
9328 | if (!r) |
9329 | return (REGEXP *)NULL; | |
9330 | ||
9331 | if ((ret = (REGEXP *)ptr_table_fetch(PL_ptr_table, r))) | |
9332 | return ret; | |
9333 | ||
f8149455 | 9334 | |
84da74a7 | 9335 | npar = r->nparens+1; |
6057429f NC |
9336 | Newx(ret, 1, regexp); |
9337 | StructCopy(r, ret, regexp); | |
f0ab9afb NC |
9338 | Newx(ret->offs, npar, regexp_paren_pair); |
9339 | Copy(r->offs, ret->offs, npar, regexp_paren_pair); | |
6057429f | 9340 | if(ret->swap) { |
28d8d7f4 | 9341 | /* no need to copy these */ |
f0ab9afb | 9342 | Newx(ret->swap, npar, regexp_paren_pair); |
28d8d7f4 | 9343 | } |
84da74a7 | 9344 | |
6057429f | 9345 | if (ret->substrs) { |
32cd70f6 NC |
9346 | /* Do it this way to avoid reading from *r after the StructCopy(). |
9347 | That way, if any of the sv_dup_inc()s dislodge *r from the L1 | |
9348 | cache, it doesn't matter. */ | |
9349 | const bool anchored = r->check_substr == r->anchored_substr; | |
785a26d5 | 9350 | Newx(ret->substrs, 1, struct reg_substr_data); |
a86a1ca7 NC |
9351 | StructCopy(r->substrs, ret->substrs, struct reg_substr_data); |
9352 | ||
32cd70f6 NC |
9353 | ret->anchored_substr = sv_dup_inc(ret->anchored_substr, param); |
9354 | ret->anchored_utf8 = sv_dup_inc(ret->anchored_utf8, param); | |
9355 | ret->float_substr = sv_dup_inc(ret->float_substr, param); | |
9356 | ret->float_utf8 = sv_dup_inc(ret->float_utf8, param); | |
a86a1ca7 | 9357 | |
32cd70f6 NC |
9358 | /* check_substr and check_utf8, if non-NULL, point to either their |
9359 | anchored or float namesakes, and don't hold a second reference. */ | |
9360 | ||
9361 | if (ret->check_substr) { | |
9362 | if (anchored) { | |
9363 | assert(r->check_utf8 == r->anchored_utf8); | |
9364 | ret->check_substr = ret->anchored_substr; | |
9365 | ret->check_utf8 = ret->anchored_utf8; | |
9366 | } else { | |
9367 | assert(r->check_substr == r->float_substr); | |
9368 | assert(r->check_utf8 == r->float_utf8); | |
9369 | ret->check_substr = ret->float_substr; | |
9370 | ret->check_utf8 = ret->float_utf8; | |
9371 | } | |
9372 | } | |
6057429f | 9373 | } |
f8149455 | 9374 | |
6057429f NC |
9375 | ret->wrapped = SAVEPVN(ret->wrapped, ret->wraplen+1); |
9376 | ret->precomp = ret->wrapped + (ret->precomp - ret->wrapped); | |
9377 | ret->paren_names = hv_dup_inc(ret->paren_names, param); | |
bcdf7404 | 9378 | |
6057429f NC |
9379 | if (ret->pprivate) |
9380 | RXi_SET(ret,CALLREGDUPE_PVT(ret,param)); | |
f8149455 YO |
9381 | |
9382 | if (RX_MATCH_COPIED(ret)) | |
6057429f | 9383 | ret->subbeg = SAVEPVN(ret->subbeg, ret->sublen); |
f8149455 YO |
9384 | else |
9385 | ret->subbeg = NULL; | |
9386 | #ifdef PERL_OLD_COPY_ON_WRITE | |
9387 | ret->saved_copy = NULL; | |
9388 | #endif | |
6057429f NC |
9389 | |
9390 | ret->mother_re = NULL; | |
9391 | ret->gofs = 0; | |
9392 | ret->seen_evals = 0; | |
f8149455 YO |
9393 | |
9394 | ptr_table_store(PL_ptr_table, r, ret); | |
9395 | return ret; | |
9396 | } | |
9397 | #endif /* PERL_IN_XSUB_RE */ | |
9398 | ||
9399 | /* | |
9400 | regdupe_internal() | |
9401 | ||
9402 | This is the internal complement to regdupe() which is used to copy | |
9403 | the structure pointed to by the *pprivate pointer in the regexp. | |
9404 | This is the core version of the extension overridable cloning hook. | |
9405 | The regexp structure being duplicated will be copied by perl prior | |
9406 | to this and will be provided as the regexp *r argument, however | |
9407 | with the /old/ structures pprivate pointer value. Thus this routine | |
9408 | may override any copying normally done by perl. | |
9409 | ||
9410 | It returns a pointer to the new regexp_internal structure. | |
9411 | */ | |
9412 | ||
9413 | void * | |
49d7dfbc | 9414 | Perl_regdupe_internal(pTHX_ REGEXP * const r, CLONE_PARAMS *param) |
f8149455 YO |
9415 | { |
9416 | dVAR; | |
9417 | regexp_internal *reti; | |
9418 | int len, npar; | |
9419 | RXi_GET_DECL(r,ri); | |
9420 | ||
9421 | npar = r->nparens+1; | |
7122b237 | 9422 | len = ProgLen(ri); |
f8149455 YO |
9423 | |
9424 | Newxc(reti, sizeof(regexp_internal) + (len+1)*sizeof(regnode), char, regexp_internal); | |
9425 | Copy(ri->program, reti->program, len+1, regnode); | |
9426 | ||
f8149455 | 9427 | |
f8fc2ecf | 9428 | reti->regstclass = NULL; |
bcdf7404 | 9429 | |
f8fc2ecf | 9430 | if (ri->data) { |
84da74a7 | 9431 | struct reg_data *d; |
f8fc2ecf | 9432 | const int count = ri->data->count; |
84da74a7 YO |
9433 | int i; |
9434 | ||
9435 | Newxc(d, sizeof(struct reg_data) + count*sizeof(void *), | |
9436 | char, struct reg_data); | |
9437 | Newx(d->what, count, U8); | |
9438 | ||
9439 | d->count = count; | |
9440 | for (i = 0; i < count; i++) { | |
f8fc2ecf | 9441 | d->what[i] = ri->data->what[i]; |
84da74a7 | 9442 | switch (d->what[i]) { |
55eed653 | 9443 | /* legal options are one of: sSfpontTu |
84da74a7 YO |
9444 | see also regcomp.h and pregfree() */ |
9445 | case 's': | |
81714fb9 | 9446 | case 'S': |
0536c0a7 | 9447 | case 'p': /* actually an AV, but the dup function is identical. */ |
55eed653 | 9448 | case 'u': /* actually an HV, but the dup function is identical. */ |
f8fc2ecf | 9449 | d->data[i] = sv_dup_inc((SV *)ri->data->data[i], param); |
84da74a7 | 9450 | break; |
84da74a7 YO |
9451 | case 'f': |
9452 | /* This is cheating. */ | |
9453 | Newx(d->data[i], 1, struct regnode_charclass_class); | |
f8fc2ecf | 9454 | StructCopy(ri->data->data[i], d->data[i], |
84da74a7 | 9455 | struct regnode_charclass_class); |
f8fc2ecf | 9456 | reti->regstclass = (regnode*)d->data[i]; |
84da74a7 YO |
9457 | break; |
9458 | case 'o': | |
bbe252da YO |
9459 | /* Compiled op trees are readonly and in shared memory, |
9460 | and can thus be shared without duplication. */ | |
84da74a7 | 9461 | OP_REFCNT_LOCK; |
f8fc2ecf | 9462 | d->data[i] = (void*)OpREFCNT_inc((OP*)ri->data->data[i]); |
84da74a7 YO |
9463 | OP_REFCNT_UNLOCK; |
9464 | break; | |
23eab42c NC |
9465 | case 'T': |
9466 | /* Trie stclasses are readonly and can thus be shared | |
9467 | * without duplication. We free the stclass in pregfree | |
9468 | * when the corresponding reg_ac_data struct is freed. | |
9469 | */ | |
9470 | reti->regstclass= ri->regstclass; | |
9471 | /* Fall through */ | |
84da74a7 | 9472 | case 't': |
84da74a7 | 9473 | OP_REFCNT_LOCK; |
0536c0a7 | 9474 | ((reg_trie_data*)ri->data->data[i])->refcount++; |
84da74a7 | 9475 | OP_REFCNT_UNLOCK; |
0536c0a7 NC |
9476 | /* Fall through */ |
9477 | case 'n': | |
9478 | d->data[i] = ri->data->data[i]; | |
84da74a7 | 9479 | break; |
84da74a7 | 9480 | default: |
f8fc2ecf | 9481 | Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", ri->data->what[i]); |
84da74a7 YO |
9482 | } |
9483 | } | |
9484 | ||
f8fc2ecf | 9485 | reti->data = d; |
84da74a7 YO |
9486 | } |
9487 | else | |
f8fc2ecf | 9488 | reti->data = NULL; |
84da74a7 | 9489 | |
cde0cee5 YO |
9490 | reti->name_list_idx = ri->name_list_idx; |
9491 | ||
7122b237 YO |
9492 | #ifdef RE_TRACK_PATTERN_OFFSETS |
9493 | if (ri->u.offsets) { | |
9494 | Newx(reti->u.offsets, 2*len+1, U32); | |
9495 | Copy(ri->u.offsets, reti->u.offsets, 2*len+1, U32); | |
9496 | } | |
9497 | #else | |
9498 | SetProgLen(reti,len); | |
9499 | #endif | |
9500 | ||
f8149455 | 9501 | return (void*)reti; |
84da74a7 | 9502 | } |
f8149455 YO |
9503 | |
9504 | #endif /* USE_ITHREADS */ | |
84da74a7 | 9505 | |
de8c5301 YO |
9506 | /* |
9507 | reg_stringify() | |
9508 | ||
9509 | converts a regexp embedded in a MAGIC struct to its stringified form, | |
9510 | caching the converted form in the struct and returns the cached | |
9511 | string. | |
9512 | ||
9513 | If lp is nonnull then it is used to return the length of the | |
9514 | resulting string | |
9515 | ||
9516 | If flags is nonnull and the returned string contains UTF8 then | |
f8149455 | 9517 | (*flags & 1) will be true. |
de8c5301 YO |
9518 | |
9519 | If haseval is nonnull then it is used to return whether the pattern | |
9520 | contains evals. | |
9521 | ||
9522 | Normally called via macro: | |
9523 | ||
f8149455 | 9524 | CALLREG_STRINGIFY(mg,&len,&utf8); |
de8c5301 YO |
9525 | |
9526 | And internally with | |
9527 | ||
f8149455 | 9528 | CALLREG_AS_STR(mg,&lp,&flags,&haseval) |
de8c5301 YO |
9529 | |
9530 | See sv_2pv_flags() in sv.c for an example of internal usage. | |
9531 | ||
9532 | */ | |
f8149455 | 9533 | #ifndef PERL_IN_XSUB_RE |
bcdf7404 | 9534 | |
de8c5301 YO |
9535 | char * |
9536 | Perl_reg_stringify(pTHX_ MAGIC *mg, STRLEN *lp, U32 *flags, I32 *haseval ) { | |
9537 | dVAR; | |
9538 | const regexp * const re = (regexp *)mg->mg_obj; | |
de8c5301 | 9539 | if (haseval) |
f8149455 | 9540 | *haseval = re->seen_evals; |
de8c5301 | 9541 | if (flags) |
bbe252da | 9542 | *flags = ((re->extflags & RXf_UTF8) ? 1 : 0); |
de8c5301 | 9543 | if (lp) |
bcdf7404 YO |
9544 | *lp = re->wraplen; |
9545 | return re->wrapped; | |
de8c5301 YO |
9546 | } |
9547 | ||
c277df42 IZ |
9548 | /* |
9549 | - regnext - dig the "next" pointer out of a node | |
c277df42 IZ |
9550 | */ |
9551 | regnode * | |
864dbfa3 | 9552 | Perl_regnext(pTHX_ register regnode *p) |
c277df42 | 9553 | { |
97aff369 | 9554 | dVAR; |
c277df42 IZ |
9555 | register I32 offset; |
9556 | ||
f8fc2ecf | 9557 | if (!p) |
c277df42 IZ |
9558 | return(NULL); |
9559 | ||
9560 | offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p)); | |
9561 | if (offset == 0) | |
9562 | return(NULL); | |
9563 | ||
c277df42 | 9564 | return(p+offset); |
c277df42 | 9565 | } |
76234dfb | 9566 | #endif |
c277df42 | 9567 | |
01f988be | 9568 | STATIC void |
cea2e8a9 | 9569 | S_re_croak2(pTHX_ const char* pat1,const char* pat2,...) |
c277df42 IZ |
9570 | { |
9571 | va_list args; | |
9572 | STRLEN l1 = strlen(pat1); | |
9573 | STRLEN l2 = strlen(pat2); | |
9574 | char buf[512]; | |
06bf62c7 | 9575 | SV *msv; |
73d840c0 | 9576 | const char *message; |
c277df42 IZ |
9577 | |
9578 | if (l1 > 510) | |
9579 | l1 = 510; | |
9580 | if (l1 + l2 > 510) | |
9581 | l2 = 510 - l1; | |
9582 | Copy(pat1, buf, l1 , char); | |
9583 | Copy(pat2, buf + l1, l2 , char); | |
3b818b81 GS |
9584 | buf[l1 + l2] = '\n'; |
9585 | buf[l1 + l2 + 1] = '\0'; | |
8736538c AS |
9586 | #ifdef I_STDARG |
9587 | /* ANSI variant takes additional second argument */ | |
c277df42 | 9588 | va_start(args, pat2); |
8736538c AS |
9589 | #else |
9590 | va_start(args); | |
9591 | #endif | |
5a844595 | 9592 | msv = vmess(buf, &args); |
c277df42 | 9593 | va_end(args); |
cfd0369c | 9594 | message = SvPV_const(msv,l1); |
c277df42 IZ |
9595 | if (l1 > 512) |
9596 | l1 = 512; | |
9597 | Copy(message, buf, l1 , char); | |
197cf9b9 | 9598 | buf[l1-1] = '\0'; /* Overwrite \n */ |
cea2e8a9 | 9599 | Perl_croak(aTHX_ "%s", buf); |
c277df42 | 9600 | } |
a0ed51b3 LW |
9601 | |
9602 | /* XXX Here's a total kludge. But we need to re-enter for swash routines. */ | |
9603 | ||
76234dfb | 9604 | #ifndef PERL_IN_XSUB_RE |
a0ed51b3 | 9605 | void |
864dbfa3 | 9606 | Perl_save_re_context(pTHX) |
b81d288d | 9607 | { |
97aff369 | 9608 | dVAR; |
1ade1aa1 NC |
9609 | |
9610 | struct re_save_state *state; | |
9611 | ||
9612 | SAVEVPTR(PL_curcop); | |
9613 | SSGROW(SAVESTACK_ALLOC_FOR_RE_SAVE_STATE + 1); | |
9614 | ||
9615 | state = (struct re_save_state *)(PL_savestack + PL_savestack_ix); | |
9616 | PL_savestack_ix += SAVESTACK_ALLOC_FOR_RE_SAVE_STATE; | |
9617 | SSPUSHINT(SAVEt_RE_STATE); | |
9618 | ||
46ab3289 | 9619 | Copy(&PL_reg_state, state, 1, struct re_save_state); |
1ade1aa1 | 9620 | |
a0ed51b3 | 9621 | PL_reg_start_tmp = 0; |
a0ed51b3 | 9622 | PL_reg_start_tmpl = 0; |
c445ea15 | 9623 | PL_reg_oldsaved = NULL; |
a5db57d6 | 9624 | PL_reg_oldsavedlen = 0; |
a5db57d6 | 9625 | PL_reg_maxiter = 0; |
a5db57d6 | 9626 | PL_reg_leftiter = 0; |
c445ea15 | 9627 | PL_reg_poscache = NULL; |
a5db57d6 | 9628 | PL_reg_poscache_size = 0; |
1ade1aa1 NC |
9629 | #ifdef PERL_OLD_COPY_ON_WRITE |
9630 | PL_nrs = NULL; | |
9631 | #endif | |
ada6e8a9 | 9632 | |
c445ea15 AL |
9633 | /* Save $1..$n (#18107: UTF-8 s/(\w+)/uc($1)/e); AMS 20021106. */ |
9634 | if (PL_curpm) { | |
9635 | const REGEXP * const rx = PM_GETRE(PL_curpm); | |
9636 | if (rx) { | |
1df70142 | 9637 | U32 i; |
ada6e8a9 | 9638 | for (i = 1; i <= rx->nparens; i++) { |
1df70142 | 9639 | char digits[TYPE_CHARS(long)]; |
d9fad198 | 9640 | const STRLEN len = my_snprintf(digits, sizeof(digits), "%lu", (long)i); |
49f27e4b NC |
9641 | GV *const *const gvp |
9642 | = (GV**)hv_fetch(PL_defstash, digits, len, 0); | |
9643 | ||
b37c2d43 AL |
9644 | if (gvp) { |
9645 | GV * const gv = *gvp; | |
9646 | if (SvTYPE(gv) == SVt_PVGV && GvSV(gv)) | |
9647 | save_scalar(gv); | |
49f27e4b | 9648 | } |
ada6e8a9 AMS |
9649 | } |
9650 | } | |
9651 | } | |
a0ed51b3 | 9652 | } |
76234dfb | 9653 | #endif |
51371543 | 9654 | |
51371543 | 9655 | static void |
acfe0abc | 9656 | clear_re(pTHX_ void *r) |
51371543 | 9657 | { |
97aff369 | 9658 | dVAR; |
51371543 GS |
9659 | ReREFCNT_dec((regexp *)r); |
9660 | } | |
ffbc6a93 | 9661 | |
a28509cc AL |
9662 | #ifdef DEBUGGING |
9663 | ||
9664 | STATIC void | |
9665 | S_put_byte(pTHX_ SV *sv, int c) | |
9666 | { | |
9667 | if (isCNTRL(c) || c == 255 || !isPRINT(c)) | |
9668 | Perl_sv_catpvf(aTHX_ sv, "\\%o", c); | |
9669 | else if (c == '-' || c == ']' || c == '\\' || c == '^') | |
9670 | Perl_sv_catpvf(aTHX_ sv, "\\%c", c); | |
9671 | else | |
9672 | Perl_sv_catpvf(aTHX_ sv, "%c", c); | |
9673 | } | |
9674 | ||
786e8c11 | 9675 | |
3dab1dad YO |
9676 | #define CLEAR_OPTSTART \ |
9677 | if (optstart) STMT_START { \ | |
70685ca0 | 9678 | DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log, " (%"IVdf" nodes)\n", (IV)(node - optstart))); \ |
3dab1dad YO |
9679 | optstart=NULL; \ |
9680 | } STMT_END | |
9681 | ||
786e8c11 | 9682 | #define DUMPUNTIL(b,e) CLEAR_OPTSTART; node=dumpuntil(r,start,(b),(e),last,sv,indent+1,depth+1); |
3dab1dad | 9683 | |
b5a2f8d8 NC |
9684 | STATIC const regnode * |
9685 | S_dumpuntil(pTHX_ const regexp *r, const regnode *start, const regnode *node, | |
786e8c11 YO |
9686 | const regnode *last, const regnode *plast, |
9687 | SV* sv, I32 indent, U32 depth) | |
a28509cc | 9688 | { |
97aff369 | 9689 | dVAR; |
786e8c11 | 9690 | register U8 op = PSEUDO; /* Arbitrary non-END op. */ |
b5a2f8d8 | 9691 | register const regnode *next; |
3dab1dad | 9692 | const regnode *optstart= NULL; |
1f1031fe | 9693 | |
f8fc2ecf | 9694 | RXi_GET_DECL(r,ri); |
3dab1dad | 9695 | GET_RE_DEBUG_FLAGS_DECL; |
1f1031fe | 9696 | |
786e8c11 YO |
9697 | #ifdef DEBUG_DUMPUNTIL |
9698 | PerlIO_printf(Perl_debug_log, "--- %d : %d - %d - %d\n",indent,node-start, | |
9699 | last ? last-start : 0,plast ? plast-start : 0); | |
9700 | #endif | |
9701 | ||
9702 | if (plast && plast < last) | |
9703 | last= plast; | |
9704 | ||
9705 | while (PL_regkind[op] != END && (!last || node < last)) { | |
a28509cc | 9706 | /* While that wasn't END last time... */ |
a28509cc AL |
9707 | NODE_ALIGN(node); |
9708 | op = OP(node); | |
de734bd5 | 9709 | if (op == CLOSE || op == WHILEM) |
786e8c11 | 9710 | indent--; |
b5a2f8d8 | 9711 | next = regnext((regnode *)node); |
1f1031fe | 9712 | |
a28509cc | 9713 | /* Where, what. */ |
8e11feef | 9714 | if (OP(node) == OPTIMIZED) { |
e68ec53f | 9715 | if (!optstart && RE_DEBUG_FLAG(RE_DEBUG_COMPILE_OPTIMISE)) |
8e11feef | 9716 | optstart = node; |
3dab1dad | 9717 | else |
8e11feef | 9718 | goto after_print; |
3dab1dad YO |
9719 | } else |
9720 | CLEAR_OPTSTART; | |
1f1031fe | 9721 | |
32fc9b6a | 9722 | regprop(r, sv, node); |
a28509cc | 9723 | PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start), |
786e8c11 | 9724 | (int)(2*indent + 1), "", SvPVX_const(sv)); |
1f1031fe YO |
9725 | |
9726 | if (OP(node) != OPTIMIZED) { | |
9727 | if (next == NULL) /* Next ptr. */ | |
9728 | PerlIO_printf(Perl_debug_log, " (0)"); | |
9729 | else if (PL_regkind[(U8)op] == BRANCH && PL_regkind[OP(next)] != BRANCH ) | |
9730 | PerlIO_printf(Perl_debug_log, " (FAIL)"); | |
9731 | else | |
9732 | PerlIO_printf(Perl_debug_log, " (%"IVdf")", (IV)(next - start)); | |
9733 | (void)PerlIO_putc(Perl_debug_log, '\n'); | |
9734 | } | |
9735 | ||
a28509cc AL |
9736 | after_print: |
9737 | if (PL_regkind[(U8)op] == BRANCHJ) { | |
be8e71aa YO |
9738 | assert(next); |
9739 | { | |
9740 | register const regnode *nnode = (OP(next) == LONGJMP | |
b5a2f8d8 NC |
9741 | ? regnext((regnode *)next) |
9742 | : next); | |
be8e71aa YO |
9743 | if (last && nnode > last) |
9744 | nnode = last; | |
786e8c11 | 9745 | DUMPUNTIL(NEXTOPER(NEXTOPER(node)), nnode); |
be8e71aa | 9746 | } |
a28509cc AL |
9747 | } |
9748 | else if (PL_regkind[(U8)op] == BRANCH) { | |
be8e71aa | 9749 | assert(next); |
786e8c11 | 9750 | DUMPUNTIL(NEXTOPER(node), next); |
a28509cc AL |
9751 | } |
9752 | else if ( PL_regkind[(U8)op] == TRIE ) { | |
7f69552c | 9753 | const regnode *this_trie = node; |
1de06328 | 9754 | const char op = OP(node); |
647f639f | 9755 | const U32 n = ARG(node); |
1de06328 | 9756 | const reg_ac_data * const ac = op>=AHOCORASICK ? |
f8fc2ecf | 9757 | (reg_ac_data *)ri->data->data[n] : |
1de06328 | 9758 | NULL; |
3251b653 NC |
9759 | const reg_trie_data * const trie = |
9760 | (reg_trie_data*)ri->data->data[op<AHOCORASICK ? n : ac->trie]; | |
2b8b4781 NC |
9761 | #ifdef DEBUGGING |
9762 | AV *const trie_words = (AV *) ri->data->data[n + TRIE_WORDS_OFFSET]; | |
9763 | #endif | |
786e8c11 | 9764 | const regnode *nextbranch= NULL; |
a28509cc | 9765 | I32 word_idx; |
1de06328 | 9766 | sv_setpvn(sv, "", 0); |
786e8c11 | 9767 | for (word_idx= 0; word_idx < (I32)trie->wordcount; word_idx++) { |
2b8b4781 | 9768 | SV ** const elem_ptr = av_fetch(trie_words,word_idx,0); |
786e8c11 YO |
9769 | |
9770 | PerlIO_printf(Perl_debug_log, "%*s%s ", | |
9771 | (int)(2*(indent+3)), "", | |
9772 | elem_ptr ? pv_pretty(sv, SvPV_nolen_const(*elem_ptr), SvCUR(*elem_ptr), 60, | |
ab3bbdeb YO |
9773 | PL_colors[0], PL_colors[1], |
9774 | (SvUTF8(*elem_ptr) ? PERL_PV_ESCAPE_UNI : 0) | | |
95b611b0 | 9775 | PERL_PV_PRETTY_ELLIPSES | |
7f69552c | 9776 | PERL_PV_PRETTY_LTGT |
786e8c11 YO |
9777 | ) |
9778 | : "???" | |
9779 | ); | |
9780 | if (trie->jump) { | |
40d049e4 | 9781 | U16 dist= trie->jump[word_idx+1]; |
70685ca0 JH |
9782 | PerlIO_printf(Perl_debug_log, "(%"UVuf")\n", |
9783 | (UV)((dist ? this_trie + dist : next) - start)); | |
786e8c11 YO |
9784 | if (dist) { |
9785 | if (!nextbranch) | |
24b23f37 | 9786 | nextbranch= this_trie + trie->jump[0]; |
7f69552c YO |
9787 | DUMPUNTIL(this_trie + dist, nextbranch); |
9788 | } | |
786e8c11 YO |
9789 | if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH) |
9790 | nextbranch= regnext((regnode *)nextbranch); | |
9791 | } else { | |
9792 | PerlIO_printf(Perl_debug_log, "\n"); | |
a28509cc | 9793 | } |
786e8c11 YO |
9794 | } |
9795 | if (last && next > last) | |
9796 | node= last; | |
9797 | else | |
9798 | node= next; | |
a28509cc | 9799 | } |
786e8c11 YO |
9800 | else if ( op == CURLY ) { /* "next" might be very big: optimizer */ |
9801 | DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS, | |
9802 | NEXTOPER(node) + EXTRA_STEP_2ARGS + 1); | |
a28509cc AL |
9803 | } |
9804 | else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) { | |
be8e71aa | 9805 | assert(next); |
786e8c11 | 9806 | DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS, next); |
a28509cc AL |
9807 | } |
9808 | else if ( op == PLUS || op == STAR) { | |
786e8c11 | 9809 | DUMPUNTIL(NEXTOPER(node), NEXTOPER(node) + 1); |
a28509cc AL |
9810 | } |
9811 | else if (op == ANYOF) { | |
9812 | /* arglen 1 + class block */ | |
9813 | node += 1 + ((ANYOF_FLAGS(node) & ANYOF_LARGE) | |
9814 | ? ANYOF_CLASS_SKIP : ANYOF_SKIP); | |
9815 | node = NEXTOPER(node); | |
9816 | } | |
9817 | else if (PL_regkind[(U8)op] == EXACT) { | |
9818 | /* Literal string, where present. */ | |
9819 | node += NODE_SZ_STR(node) - 1; | |
9820 | node = NEXTOPER(node); | |
9821 | } | |
9822 | else { | |
9823 | node = NEXTOPER(node); | |
9824 | node += regarglen[(U8)op]; | |
9825 | } | |
9826 | if (op == CURLYX || op == OPEN) | |
786e8c11 | 9827 | indent++; |
a28509cc | 9828 | } |
3dab1dad | 9829 | CLEAR_OPTSTART; |
786e8c11 | 9830 | #ifdef DEBUG_DUMPUNTIL |
70685ca0 | 9831 | PerlIO_printf(Perl_debug_log, "--- %d\n", (int)indent); |
786e8c11 | 9832 | #endif |
1de06328 | 9833 | return node; |
a28509cc AL |
9834 | } |
9835 | ||
9836 | #endif /* DEBUGGING */ | |
9837 | ||
241d1a3b NC |
9838 | /* |
9839 | * Local variables: | |
9840 | * c-indentation-style: bsd | |
9841 | * c-basic-offset: 4 | |
9842 | * indent-tabs-mode: t | |
9843 | * End: | |
9844 | * | |
37442d52 RGS |
9845 | * ex: set ts=8 sts=4 sw=4 noet: |
9846 | */ |