Commit | Line | Data |
---|---|---|
a0d0e21e LW |
1 | /* regexec.c |
2 | */ | |
3 | ||
4 | /* | |
4ac71550 TC |
5 | * One Ring to rule them all, One Ring to find them |
6 | & | |
7 | * [p.v of _The Lord of the Rings_, opening poem] | |
8 | * [p.50 of _The Lord of the Rings_, I/iii: "The Shadow of the Past"] | |
9 | * [p.254 of _The Lord of the Rings_, II/ii: "The Council of Elrond"] | |
a0d0e21e LW |
10 | */ |
11 | ||
61296642 DM |
12 | /* This file contains functions for executing a regular expression. See |
13 | * also regcomp.c which funnily enough, contains functions for compiling | |
166f8a29 | 14 | * a regular expression. |
e4a054ea DM |
15 | * |
16 | * This file is also copied at build time to ext/re/re_exec.c, where | |
17 | * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT. | |
18 | * This causes the main functions to be compiled under new names and with | |
19 | * debugging support added, which makes "use re 'debug'" work. | |
166f8a29 DM |
20 | */ |
21 | ||
a687059c LW |
22 | /* NOTE: this is derived from Henry Spencer's regexp code, and should not |
23 | * confused with the original package (see point 3 below). Thanks, Henry! | |
24 | */ | |
25 | ||
26 | /* Additional note: this code is very heavily munged from Henry's version | |
27 | * in places. In some spots I've traded clarity for efficiency, so don't | |
28 | * blame Henry for some of the lack of readability. | |
29 | */ | |
30 | ||
e50aee73 AD |
31 | /* The names of the functions have been changed from regcomp and |
32 | * regexec to pregcomp and pregexec in order to avoid conflicts | |
33 | * with the POSIX routines of the same names. | |
34 | */ | |
35 | ||
b9d5759e | 36 | #ifdef PERL_EXT_RE_BUILD |
54df2634 | 37 | #include "re_top.h" |
9041c2e3 | 38 | #endif |
56953603 | 39 | |
a687059c | 40 | /* |
e50aee73 | 41 | * pregcomp and pregexec -- regsub and regerror are not used in perl |
a687059c LW |
42 | * |
43 | * Copyright (c) 1986 by University of Toronto. | |
44 | * Written by Henry Spencer. Not derived from licensed software. | |
45 | * | |
46 | * Permission is granted to anyone to use this software for any | |
47 | * purpose on any computer system, and to redistribute it freely, | |
48 | * subject to the following restrictions: | |
49 | * | |
50 | * 1. The author is not responsible for the consequences of use of | |
51 | * this software, no matter how awful, even if they arise | |
52 | * from defects in it. | |
53 | * | |
54 | * 2. The origin of this software must not be misrepresented, either | |
55 | * by explicit claim or by omission. | |
56 | * | |
57 | * 3. Altered versions must be plainly marked as such, and must not | |
58 | * be misrepresented as being the original software. | |
59 | * | |
60 | **** Alterations to Henry's code are... | |
61 | **** | |
4bb101f2 | 62 | **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, |
1129b882 NC |
63 | **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 |
64 | **** by Larry Wall and others | |
a687059c | 65 | **** |
9ef589d8 LW |
66 | **** You may distribute under the terms of either the GNU General Public |
67 | **** License or the Artistic License, as specified in the README file. | |
a687059c LW |
68 | * |
69 | * Beware that some of this code is subtly aware of the way operator | |
70 | * precedence is structured in regular expressions. Serious changes in | |
71 | * regular-expression syntax might require a total rethink. | |
72 | */ | |
73 | #include "EXTERN.h" | |
864dbfa3 | 74 | #define PERL_IN_REGEXEC_C |
a687059c | 75 | #include "perl.h" |
0f5d15d6 | 76 | |
54df2634 NC |
77 | #ifdef PERL_IN_XSUB_RE |
78 | # include "re_comp.h" | |
79 | #else | |
80 | # include "regcomp.h" | |
81 | #endif | |
a687059c | 82 | |
81e983c1 | 83 | #include "inline_invlist.c" |
1b0f46bf | 84 | #include "unicode_constants.h" |
81e983c1 | 85 | |
e1cf74e3 CB |
86 | #ifdef DEBUGGING |
87 | /* At least one required character in the target string is expressible only in | |
88 | * UTF-8. */ | |
89 | static const char* const non_utf8_target_but_utf8_required | |
90 | = "Can't match, because target string needs to be in UTF-8\n"; | |
91 | #endif | |
92 | ||
93 | #define NON_UTF8_TARGET_BUT_UTF8_REQUIRED(target) STMT_START { \ | |
94 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s", non_utf8_target_but_utf8_required));\ | |
95 | goto target; \ | |
96 | } STMT_END | |
97 | ||
c74f6de9 KW |
98 | #define HAS_NONLATIN1_FOLD_CLOSURE(i) _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i) |
99 | ||
a687059c LW |
100 | #ifndef STATIC |
101 | #define STATIC static | |
102 | #endif | |
103 | ||
e0193e47 | 104 | /* Valid for non-utf8 strings: avoids the reginclass |
7e2509c1 KW |
105 | * call if there are no complications: i.e., if everything matchable is |
106 | * straight forward in the bitmap */ | |
635cd5d4 | 107 | #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0) \ |
af364d03 | 108 | : ANYOF_BITMAP_TEST(p,*(c))) |
7d3e948e | 109 | |
c277df42 IZ |
110 | /* |
111 | * Forwards. | |
112 | */ | |
113 | ||
f2ed9b32 | 114 | #define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv)) |
ba44c216 | 115 | #define CHR_DIST(a,b) (reginfo->is_utf8_target ? utf8_distance(a,b) : a - b) |
a0ed51b3 | 116 | |
3dab1dad | 117 | #define HOPc(pos,off) \ |
ba44c216 | 118 | (char *)(reginfo->is_utf8_target \ |
220db18a | 119 | ? reghop3((U8*)pos, off, \ |
9d9163fb | 120 | (U8*)(off >= 0 ? reginfo->strend : reginfo->strbeg)) \ |
3dab1dad YO |
121 | : (U8*)(pos + off)) |
122 | #define HOPBACKc(pos, off) \ | |
ba44c216 | 123 | (char*)(reginfo->is_utf8_target \ |
9d9163fb DM |
124 | ? reghopmaybe3((U8*)pos, -off, (U8*)(reginfo->strbeg)) \ |
125 | : (pos - off >= reginfo->strbeg) \ | |
8e11feef | 126 | ? (U8*)pos - off \ |
3dab1dad | 127 | : NULL) |
efb30f32 | 128 | |
ba44c216 | 129 | #define HOP3(pos,off,lim) (reginfo->is_utf8_target ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off)) |
1aa99e6b | 130 | #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim)) |
1aa99e6b | 131 | |
7016d6eb DM |
132 | |
133 | #define NEXTCHR_EOS -10 /* nextchr has fallen off the end */ | |
134 | #define NEXTCHR_IS_EOS (nextchr < 0) | |
135 | ||
136 | #define SET_nextchr \ | |
220db18a | 137 | nextchr = ((locinput < reginfo->strend) ? UCHARAT(locinput) : NEXTCHR_EOS) |
7016d6eb DM |
138 | |
139 | #define SET_locinput(p) \ | |
140 | locinput = (p); \ | |
141 | SET_nextchr | |
142 | ||
143 | ||
c7304fe2 KW |
144 | #define LOAD_UTF8_CHARCLASS(swash_ptr, property_name) STMT_START { \ |
145 | if (!swash_ptr) { \ | |
146 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; \ | |
c7304fe2 KW |
147 | swash_ptr = _core_swash_init("utf8", property_name, &PL_sv_undef, \ |
148 | 1, 0, NULL, &flags); \ | |
149 | assert(swash_ptr); \ | |
150 | } \ | |
151 | } STMT_END | |
152 | ||
153 | /* If in debug mode, we test that a known character properly matches */ | |
154 | #ifdef DEBUGGING | |
155 | # define LOAD_UTF8_CHARCLASS_DEBUG_TEST(swash_ptr, \ | |
156 | property_name, \ | |
157 | utf8_char_in_property) \ | |
158 | LOAD_UTF8_CHARCLASS(swash_ptr, property_name); \ | |
159 | assert(swash_fetch(swash_ptr, (U8 *) utf8_char_in_property, TRUE)); | |
160 | #else | |
161 | # define LOAD_UTF8_CHARCLASS_DEBUG_TEST(swash_ptr, \ | |
162 | property_name, \ | |
163 | utf8_char_in_property) \ | |
164 | LOAD_UTF8_CHARCLASS(swash_ptr, property_name) | |
165 | #endif | |
d1eb3177 | 166 | |
c7304fe2 KW |
167 | #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS_DEBUG_TEST( \ |
168 | PL_utf8_swash_ptrs[_CC_WORDCHAR], \ | |
169 | swash_property_names[_CC_WORDCHAR], \ | |
df38da56 | 170 | LATIN_CAPITAL_LETTER_SHARP_S_UTF8); |
c7304fe2 KW |
171 | |
172 | #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \ | |
173 | STMT_START { \ | |
174 | LOAD_UTF8_CHARCLASS_DEBUG_TEST(PL_utf8_X_regular_begin, \ | |
175 | "_X_regular_begin", \ | |
df38da56 | 176 | LATIN_CAPITAL_LETTER_SHARP_S_UTF8); \ |
c7304fe2 KW |
177 | LOAD_UTF8_CHARCLASS_DEBUG_TEST(PL_utf8_X_extend, \ |
178 | "_X_extend", \ | |
179 | COMBINING_GRAVE_ACCENT_UTF8); \ | |
180 | } STMT_END | |
d1eb3177 | 181 | |
c7304fe2 | 182 | #define PLACEHOLDER /* Something for the preprocessor to grab onto */ |
3dab1dad YO |
183 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
184 | ||
5f80c4cf | 185 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
c35dcbe2 YO |
186 | /* it would be nice to rework regcomp.sym to generate this stuff. sigh |
187 | * | |
188 | * NOTE that *nothing* that affects backtracking should be in here, specifically | |
189 | * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a | |
190 | * node that is in between two EXACT like nodes when ascertaining what the required | |
191 | * "follow" character is. This should probably be moved to regex compile time | |
192 | * although it may be done at run time beause of the REF possibility - more | |
193 | * investigation required. -- demerphq | |
194 | */ | |
3e901dc0 YO |
195 | #define JUMPABLE(rn) ( \ |
196 | OP(rn) == OPEN || \ | |
197 | (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \ | |
198 | OP(rn) == EVAL || \ | |
cca55fe3 JP |
199 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
200 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
d1c771f5 | 201 | OP(rn) == KEEPS || \ |
3dab1dad | 202 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 | 203 | ) |
ee9b8eae | 204 | #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT) |
e2d8ce26 | 205 | |
ee9b8eae YO |
206 | #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF ) |
207 | ||
208 | #if 0 | |
209 | /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so | |
210 | we don't need this definition. */ | |
211 | #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) | |
098b07d5 | 212 | #define IS_TEXTF(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFA || OP(rn)==EXACTFA_NO_TRIE || OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF ) |
ee9b8eae YO |
213 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) |
214 | ||
215 | #else | |
216 | /* ... so we use this as its faster. */ | |
217 | #define IS_TEXT(rn) ( OP(rn)==EXACT ) | |
098b07d5 | 218 | #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn) == EXACTFA || OP(rn) == EXACTFA_NO_TRIE) |
ee9b8eae YO |
219 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) |
220 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) | |
221 | ||
222 | #endif | |
e2d8ce26 | 223 | |
a84d97b6 HS |
224 | /* |
225 | Search for mandatory following text node; for lookahead, the text must | |
226 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
227 | */ | |
cca55fe3 | 228 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
229 | while (JUMPABLE(rn)) { \ |
230 | const OPCODE type = OP(rn); \ | |
231 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 232 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 233 | else if (type == PLUS) \ |
cca55fe3 | 234 | rn = NEXTOPER(rn); \ |
3dab1dad | 235 | else if (type == IFMATCH) \ |
a84d97b6 | 236 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 237 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 238 | } \ |
5f80c4cf | 239 | } STMT_END |
74750237 | 240 | |
22913b96 KW |
241 | /* These constants are for finding GCB=LV and GCB=LVT in the CLUMP regnode. |
242 | * These are for the pre-composed Hangul syllables, which are all in a | |
243 | * contiguous block and arranged there in such a way so as to facilitate | |
244 | * alorithmic determination of their characteristics. As such, they don't need | |
245 | * a swash, but can be determined by simple arithmetic. Almost all are | |
246 | * GCB=LVT, but every 28th one is a GCB=LV */ | |
247 | #define SBASE 0xAC00 /* Start of block */ | |
248 | #define SCount 11172 /* Length of block */ | |
249 | #define TCount 28 | |
c476f425 | 250 | |
006f26b2 DM |
251 | #define SLAB_FIRST(s) (&(s)->states[0]) |
252 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
253 | ||
a75351a1 | 254 | static void S_setup_eval_state(pTHX_ regmatch_info *const reginfo); |
bf2039a9 | 255 | static void S_cleanup_regmatch_info_aux(pTHX_ void *arg); |
bf2039a9 | 256 | static regmatch_state * S_push_slab(pTHX); |
51371543 | 257 | |
87c0511b | 258 | #define REGCP_PAREN_ELEMS 3 |
f067efbf | 259 | #define REGCP_OTHER_ELEMS 3 |
e0fa7e2b | 260 | #define REGCP_FRAME_ELEMS 1 |
620d5b66 NC |
261 | /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and |
262 | * are needed for the regexp context stack bookkeeping. */ | |
263 | ||
76e3520e | 264 | STATIC CHECKPOINT |
92da3157 | 265 | S_regcppush(pTHX_ const regexp *rex, I32 parenfloor, U32 maxopenparen) |
a0d0e21e | 266 | { |
97aff369 | 267 | dVAR; |
a3b680e6 | 268 | const int retval = PL_savestack_ix; |
92da3157 DM |
269 | const int paren_elems_to_push = |
270 | (maxopenparen - parenfloor) * REGCP_PAREN_ELEMS; | |
e0fa7e2b NC |
271 | const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS; |
272 | const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT; | |
87c0511b | 273 | I32 p; |
40a82448 | 274 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 275 | |
b93070ed DM |
276 | PERL_ARGS_ASSERT_REGCPPUSH; |
277 | ||
e49a9654 | 278 | if (paren_elems_to_push < 0) |
5637ef5b NC |
279 | Perl_croak(aTHX_ "panic: paren_elems_to_push, %i < 0", |
280 | paren_elems_to_push); | |
e49a9654 | 281 | |
e0fa7e2b NC |
282 | if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems) |
283 | Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf | |
5df417d0 | 284 | " out of range (%lu-%ld)", |
92da3157 DM |
285 | total_elems, |
286 | (unsigned long)maxopenparen, | |
287 | (long)parenfloor); | |
e0fa7e2b | 288 | |
620d5b66 | 289 | SSGROW(total_elems + REGCP_FRAME_ELEMS); |
7f69552c | 290 | |
495f47a5 | 291 | DEBUG_BUFFERS_r( |
92da3157 | 292 | if ((int)maxopenparen > (int)parenfloor) |
495f47a5 DM |
293 | PerlIO_printf(Perl_debug_log, |
294 | "rex=0x%"UVxf" offs=0x%"UVxf": saving capture indices:\n", | |
295 | PTR2UV(rex), | |
296 | PTR2UV(rex->offs) | |
297 | ); | |
298 | ); | |
92da3157 | 299 | for (p = parenfloor+1; p <= (I32)maxopenparen; p++) { |
b1ce53c5 | 300 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
99a90e59 FC |
301 | SSPUSHIV(rex->offs[p].end); |
302 | SSPUSHIV(rex->offs[p].start); | |
1ca2007e | 303 | SSPUSHINT(rex->offs[p].start_tmp); |
e7707071 | 304 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
495f47a5 DM |
305 | " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"\n", |
306 | (UV)p, | |
307 | (IV)rex->offs[p].start, | |
308 | (IV)rex->offs[p].start_tmp, | |
309 | (IV)rex->offs[p].end | |
40a82448 | 310 | )); |
a0d0e21e | 311 | } |
b1ce53c5 | 312 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
92da3157 | 313 | SSPUSHINT(maxopenparen); |
b93070ed DM |
314 | SSPUSHINT(rex->lastparen); |
315 | SSPUSHINT(rex->lastcloseparen); | |
e0fa7e2b | 316 | SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */ |
41123dfd | 317 | |
a0d0e21e LW |
318 | return retval; |
319 | } | |
320 | ||
c277df42 | 321 | /* These are needed since we do not localize EVAL nodes: */ |
ab3bbdeb YO |
322 | #define REGCP_SET(cp) \ |
323 | DEBUG_STATE_r( \ | |
ab3bbdeb | 324 | PerlIO_printf(Perl_debug_log, \ |
e4f74956 | 325 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
ab3bbdeb YO |
326 | (IV)PL_savestack_ix)); \ |
327 | cp = PL_savestack_ix | |
c3464db5 | 328 | |
ab3bbdeb | 329 | #define REGCP_UNWIND(cp) \ |
e4f74956 | 330 | DEBUG_STATE_r( \ |
ab3bbdeb | 331 | if (cp != PL_savestack_ix) \ |
e4f74956 YO |
332 | PerlIO_printf(Perl_debug_log, \ |
333 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ | |
ab3bbdeb YO |
334 | (IV)(cp), (IV)PL_savestack_ix)); \ |
335 | regcpblow(cp) | |
c277df42 | 336 | |
a8d1f4b4 DM |
337 | #define UNWIND_PAREN(lp, lcp) \ |
338 | for (n = rex->lastparen; n > lp; n--) \ | |
339 | rex->offs[n].end = -1; \ | |
340 | rex->lastparen = n; \ | |
341 | rex->lastcloseparen = lcp; | |
342 | ||
343 | ||
f067efbf | 344 | STATIC void |
92da3157 | 345 | S_regcppop(pTHX_ regexp *rex, U32 *maxopenparen_p) |
a0d0e21e | 346 | { |
97aff369 | 347 | dVAR; |
e0fa7e2b | 348 | UV i; |
87c0511b | 349 | U32 paren; |
a3621e74 YO |
350 | GET_RE_DEBUG_FLAGS_DECL; |
351 | ||
7918f24d NC |
352 | PERL_ARGS_ASSERT_REGCPPOP; |
353 | ||
b1ce53c5 | 354 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
c6bf6a65 | 355 | i = SSPOPUV; |
e0fa7e2b NC |
356 | assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
357 | i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */ | |
b93070ed DM |
358 | rex->lastcloseparen = SSPOPINT; |
359 | rex->lastparen = SSPOPINT; | |
92da3157 | 360 | *maxopenparen_p = SSPOPINT; |
b1ce53c5 | 361 | |
620d5b66 | 362 | i -= REGCP_OTHER_ELEMS; |
b1ce53c5 | 363 | /* Now restore the parentheses context. */ |
495f47a5 DM |
364 | DEBUG_BUFFERS_r( |
365 | if (i || rex->lastparen + 1 <= rex->nparens) | |
366 | PerlIO_printf(Perl_debug_log, | |
367 | "rex=0x%"UVxf" offs=0x%"UVxf": restoring capture indices to:\n", | |
368 | PTR2UV(rex), | |
369 | PTR2UV(rex->offs) | |
370 | ); | |
371 | ); | |
92da3157 | 372 | paren = *maxopenparen_p; |
620d5b66 | 373 | for ( ; i > 0; i -= REGCP_PAREN_ELEMS) { |
ea3daa5d | 374 | SSize_t tmps; |
1ca2007e | 375 | rex->offs[paren].start_tmp = SSPOPINT; |
99a90e59 FC |
376 | rex->offs[paren].start = SSPOPIV; |
377 | tmps = SSPOPIV; | |
b93070ed DM |
378 | if (paren <= rex->lastparen) |
379 | rex->offs[paren].end = tmps; | |
495f47a5 DM |
380 | DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log, |
381 | " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"%s\n", | |
382 | (UV)paren, | |
383 | (IV)rex->offs[paren].start, | |
384 | (IV)rex->offs[paren].start_tmp, | |
385 | (IV)rex->offs[paren].end, | |
386 | (paren > rex->lastparen ? "(skipped)" : "")); | |
c277df42 | 387 | ); |
87c0511b | 388 | paren--; |
a0d0e21e | 389 | } |
daf18116 | 390 | #if 1 |
dafc8851 JH |
391 | /* It would seem that the similar code in regtry() |
392 | * already takes care of this, and in fact it is in | |
393 | * a better location to since this code can #if 0-ed out | |
394 | * but the code in regtry() is needed or otherwise tests | |
395 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
396 | * (as of patchlevel 7877) will fail. Then again, |
397 | * this code seems to be necessary or otherwise | |
225593e1 DM |
398 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ |
399 | * --jhi updated by dapm */ | |
b93070ed | 400 | for (i = rex->lastparen + 1; i <= rex->nparens; i++) { |
92da3157 | 401 | if (i > *maxopenparen_p) |
b93070ed DM |
402 | rex->offs[i].start = -1; |
403 | rex->offs[i].end = -1; | |
495f47a5 DM |
404 | DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log, |
405 | " \\%"UVuf": %s ..-1 undeffing\n", | |
406 | (UV)i, | |
92da3157 | 407 | (i > *maxopenparen_p) ? "-1" : " " |
495f47a5 | 408 | )); |
a0d0e21e | 409 | } |
dafc8851 | 410 | #endif |
a0d0e21e LW |
411 | } |
412 | ||
74088413 DM |
413 | /* restore the parens and associated vars at savestack position ix, |
414 | * but without popping the stack */ | |
415 | ||
416 | STATIC void | |
92da3157 | 417 | S_regcp_restore(pTHX_ regexp *rex, I32 ix, U32 *maxopenparen_p) |
74088413 DM |
418 | { |
419 | I32 tmpix = PL_savestack_ix; | |
420 | PL_savestack_ix = ix; | |
92da3157 | 421 | regcppop(rex, maxopenparen_p); |
74088413 DM |
422 | PL_savestack_ix = tmpix; |
423 | } | |
424 | ||
02db2b7b | 425 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 426 | |
31c7f561 KW |
427 | STATIC bool |
428 | S_isFOO_lc(pTHX_ const U8 classnum, const U8 character) | |
429 | { | |
430 | /* Returns a boolean as to whether or not 'character' is a member of the | |
431 | * Posix character class given by 'classnum' that should be equivalent to a | |
432 | * value in the typedef '_char_class_number'. | |
433 | * | |
434 | * Ideally this could be replaced by a just an array of function pointers | |
435 | * to the C library functions that implement the macros this calls. | |
436 | * However, to compile, the precise function signatures are required, and | |
437 | * these may vary from platform to to platform. To avoid having to figure | |
438 | * out what those all are on each platform, I (khw) am using this method, | |
7aee35ff KW |
439 | * which adds an extra layer of function call overhead (unless the C |
440 | * optimizer strips it away). But we don't particularly care about | |
441 | * performance with locales anyway. */ | |
31c7f561 KW |
442 | |
443 | switch ((_char_class_number) classnum) { | |
15861f94 | 444 | case _CC_ENUM_ALPHANUMERIC: return isALPHANUMERIC_LC(character); |
31c7f561 | 445 | case _CC_ENUM_ALPHA: return isALPHA_LC(character); |
e8d596e0 KW |
446 | case _CC_ENUM_ASCII: return isASCII_LC(character); |
447 | case _CC_ENUM_BLANK: return isBLANK_LC(character); | |
b0d691b2 KW |
448 | case _CC_ENUM_CASED: return isLOWER_LC(character) |
449 | || isUPPER_LC(character); | |
e8d596e0 | 450 | case _CC_ENUM_CNTRL: return isCNTRL_LC(character); |
31c7f561 KW |
451 | case _CC_ENUM_DIGIT: return isDIGIT_LC(character); |
452 | case _CC_ENUM_GRAPH: return isGRAPH_LC(character); | |
453 | case _CC_ENUM_LOWER: return isLOWER_LC(character); | |
454 | case _CC_ENUM_PRINT: return isPRINT_LC(character); | |
e8d596e0 | 455 | case _CC_ENUM_PSXSPC: return isPSXSPC_LC(character); |
31c7f561 | 456 | case _CC_ENUM_PUNCT: return isPUNCT_LC(character); |
e8d596e0 | 457 | case _CC_ENUM_SPACE: return isSPACE_LC(character); |
31c7f561 KW |
458 | case _CC_ENUM_UPPER: return isUPPER_LC(character); |
459 | case _CC_ENUM_WORDCHAR: return isWORDCHAR_LC(character); | |
31c7f561 | 460 | case _CC_ENUM_XDIGIT: return isXDIGIT_LC(character); |
31c7f561 KW |
461 | default: /* VERTSPACE should never occur in locales */ |
462 | Perl_croak(aTHX_ "panic: isFOO_lc() has an unexpected character class '%d'", classnum); | |
463 | } | |
464 | ||
465 | assert(0); /* NOTREACHED */ | |
466 | return FALSE; | |
467 | } | |
468 | ||
3018b823 KW |
469 | STATIC bool |
470 | S_isFOO_utf8_lc(pTHX_ const U8 classnum, const U8* character) | |
471 | { | |
472 | /* Returns a boolean as to whether or not the (well-formed) UTF-8-encoded | |
473 | * 'character' is a member of the Posix character class given by 'classnum' | |
474 | * that should be equivalent to a value in the typedef | |
475 | * '_char_class_number'. | |
476 | * | |
477 | * This just calls isFOO_lc on the code point for the character if it is in | |
478 | * the range 0-255. Outside that range, all characters avoid Unicode | |
479 | * rules, ignoring any locale. So use the Unicode function if this class | |
480 | * requires a swash, and use the Unicode macro otherwise. */ | |
481 | ||
482 | PERL_ARGS_ASSERT_ISFOO_UTF8_LC; | |
483 | ||
484 | if (UTF8_IS_INVARIANT(*character)) { | |
485 | return isFOO_lc(classnum, *character); | |
486 | } | |
487 | else if (UTF8_IS_DOWNGRADEABLE_START(*character)) { | |
488 | return isFOO_lc(classnum, | |
94bb8c36 | 489 | TWO_BYTE_UTF8_TO_NATIVE(*character, *(character + 1))); |
3018b823 KW |
490 | } |
491 | ||
492 | if (classnum < _FIRST_NON_SWASH_CC) { | |
493 | ||
494 | /* Initialize the swash unless done already */ | |
495 | if (! PL_utf8_swash_ptrs[classnum]) { | |
496 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; | |
497 | PL_utf8_swash_ptrs[classnum] = _core_swash_init("utf8", | |
498 | swash_property_names[classnum], &PL_sv_undef, 1, 0, NULL, &flags); | |
499 | } | |
500 | ||
92a2046b KW |
501 | return cBOOL(swash_fetch(PL_utf8_swash_ptrs[classnum], (U8 *) |
502 | character, | |
503 | TRUE /* is UTF */ )); | |
3018b823 KW |
504 | } |
505 | ||
506 | switch ((_char_class_number) classnum) { | |
507 | case _CC_ENUM_SPACE: | |
508 | case _CC_ENUM_PSXSPC: return is_XPERLSPACE_high(character); | |
509 | ||
510 | case _CC_ENUM_BLANK: return is_HORIZWS_high(character); | |
511 | case _CC_ENUM_XDIGIT: return is_XDIGIT_high(character); | |
512 | case _CC_ENUM_VERTSPACE: return is_VERTWS_high(character); | |
513 | default: return 0; /* Things like CNTRL are always | |
514 | below 256 */ | |
515 | } | |
516 | ||
517 | assert(0); /* NOTREACHED */ | |
518 | return FALSE; | |
519 | } | |
520 | ||
a687059c | 521 | /* |
e50aee73 | 522 | * pregexec and friends |
a687059c LW |
523 | */ |
524 | ||
76234dfb | 525 | #ifndef PERL_IN_XSUB_RE |
a687059c | 526 | /* |
c277df42 | 527 | - pregexec - match a regexp against a string |
a687059c | 528 | */ |
c277df42 | 529 | I32 |
5aaab254 | 530 | Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, char *strend, |
ea3daa5d | 531 | char *strbeg, SSize_t minend, SV *screamer, U32 nosave) |
8fd1a950 DM |
532 | /* stringarg: the point in the string at which to begin matching */ |
533 | /* strend: pointer to null at end of string */ | |
534 | /* strbeg: real beginning of string */ | |
535 | /* minend: end of match must be >= minend bytes after stringarg. */ | |
536 | /* screamer: SV being matched: only used for utf8 flag, pos() etc; string | |
537 | * itself is accessed via the pointers above */ | |
538 | /* nosave: For optimizations. */ | |
c277df42 | 539 | { |
7918f24d NC |
540 | PERL_ARGS_ASSERT_PREGEXEC; |
541 | ||
c277df42 | 542 | return |
9041c2e3 | 543 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
544 | nosave ? 0 : REXEC_COPY_STR); |
545 | } | |
76234dfb | 546 | #endif |
22e551b9 | 547 | |
9041c2e3 | 548 | /* |
cad2e5aa JH |
549 | * Need to implement the following flags for reg_anch: |
550 | * | |
551 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
552 | * USE_INTUIT_ML | |
553 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
554 | * INTUIT_AUTORITATIVE_ML | |
555 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
556 | * INTUIT_ONCE_ML | |
557 | * | |
558 | * Another flag for this function: SECOND_TIME (so that float substrs | |
559 | * with giant delta may be not rechecked). | |
560 | */ | |
561 | ||
3f7c398e | 562 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
563 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
564 | ||
6eb5f6b9 JH |
565 | /* XXXX Some places assume that there is a fixed substring. |
566 | An update may be needed if optimizer marks as "INTUITable" | |
567 | RExen without fixed substrings. Similarly, it is assumed that | |
568 | lengths of all the strings are no more than minlen, thus they | |
569 | cannot come from lookahead. | |
40d049e4 YO |
570 | (Or minlen should take into account lookahead.) |
571 | NOTE: Some of this comment is not correct. minlen does now take account | |
572 | of lookahead/behind. Further research is required. -- demerphq | |
573 | ||
574 | */ | |
6eb5f6b9 | 575 | |
2c2d71f5 JH |
576 | /* A failure to find a constant substring means that there is no need to make |
577 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
d8da0584 | 578 | finding a substring too deep into the string means that fewer calls to |
30944b6d IZ |
579 | regtry() should be needed. |
580 | ||
581 | REx compiler's optimizer found 4 possible hints: | |
582 | a) Anchored substring; | |
583 | b) Fixed substring; | |
584 | c) Whether we are anchored (beginning-of-line or \G); | |
486ec47a | 585 | d) First node (of those at offset 0) which may distinguish positions; |
6eb5f6b9 | 586 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
587 | string which does not contradict any of them. |
588 | */ | |
2c2d71f5 | 589 | |
6eb5f6b9 JH |
590 | /* Most of decisions we do here should have been done at compile time. |
591 | The nodes of the REx which we used for the search should have been | |
592 | deleted from the finite automaton. */ | |
593 | ||
52a21eb3 DM |
594 | /* args: |
595 | * rx: the regex to match against | |
596 | * sv: the SV being matched: only used for utf8 flag; the string | |
597 | * itself is accessed via the pointers below. Note that on | |
598 | * something like an overloaded SV, SvPOK(sv) may be false | |
599 | * and the string pointers may point to something unrelated to | |
600 | * the SV itself. | |
601 | * strbeg: real beginning of string | |
602 | * strpos: the point in the string at which to begin matching | |
603 | * strend: pointer to the byte following the last char of the string | |
604 | * flags currently unused; set to 0 | |
605 | * data: currently unused; set to NULL | |
606 | */ | |
607 | ||
cad2e5aa | 608 | char * |
52a21eb3 DM |
609 | Perl_re_intuit_start(pTHX_ |
610 | REGEXP * const rx, | |
611 | SV *sv, | |
612 | const char * const strbeg, | |
613 | char *strpos, | |
614 | char *strend, | |
615 | const U32 flags, | |
616 | re_scream_pos_data *data) | |
cad2e5aa | 617 | { |
97aff369 | 618 | dVAR; |
8d919b0a | 619 | struct regexp *const prog = ReANY(rx); |
49f55535 | 620 | SSize_t start_shift = 0; |
cad2e5aa | 621 | /* Should be nonnegative! */ |
ea3daa5d | 622 | SSize_t end_shift = 0; |
eb578fdb KW |
623 | char *s; |
624 | SV *check; | |
cad2e5aa | 625 | char *t; |
f2ed9b32 | 626 | const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 627 | I32 ml_anch; |
eb578fdb | 628 | char *other_last = NULL; /* other substr checked before this */ |
bd61b366 | 629 | char *check_at = NULL; /* check substr found at this pos */ |
d8080198 | 630 | char *checked_upto = NULL; /* how far into the string we have already checked using find_byclass*/ |
bbe252da | 631 | const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; |
f8fc2ecf | 632 | RXi_GET_DECL(prog,progi); |
02d5137b DM |
633 | regmatch_info reginfo_buf; /* create some info to pass to find_byclass */ |
634 | regmatch_info *const reginfo = ®info_buf; | |
30944b6d | 635 | #ifdef DEBUGGING |
890ce7af | 636 | const char * const i_strpos = strpos; |
30944b6d | 637 | #endif |
a3621e74 YO |
638 | GET_RE_DEBUG_FLAGS_DECL; |
639 | ||
7918f24d | 640 | PERL_ARGS_ASSERT_RE_INTUIT_START; |
c33e64f0 FC |
641 | PERL_UNUSED_ARG(flags); |
642 | PERL_UNUSED_ARG(data); | |
7918f24d | 643 | |
c344f387 JH |
644 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
645 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 646 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 647 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 648 | goto fail; |
2c2d71f5 | 649 | } |
d8da0584 | 650 | |
6c3fea77 | 651 | reginfo->is_utf8_target = cBOOL(utf8_target); |
bf2039a9 | 652 | reginfo->info_aux = NULL; |
9d9163fb | 653 | reginfo->strbeg = strbeg; |
220db18a | 654 | reginfo->strend = strend; |
aed7b151 | 655 | reginfo->is_utf8_pat = cBOOL(RX_UTF8(rx)); |
02d5137b | 656 | reginfo->intuit = 1; |
1cb48e53 DM |
657 | /* not actually used within intuit, but zero for safety anyway */ |
658 | reginfo->poscache_maxiter = 0; | |
02d5137b | 659 | |
f2ed9b32 | 660 | if (utf8_target) { |
33b8afdf JH |
661 | if (!prog->check_utf8 && prog->check_substr) |
662 | to_utf8_substr(prog); | |
663 | check = prog->check_utf8; | |
664 | } else { | |
7e0d5ad7 KW |
665 | if (!prog->check_substr && prog->check_utf8) { |
666 | if (! to_byte_substr(prog)) { | |
6b54ddc5 | 667 | NON_UTF8_TARGET_BUT_UTF8_REQUIRED(fail); |
7e0d5ad7 KW |
668 | } |
669 | } | |
33b8afdf JH |
670 | check = prog->check_substr; |
671 | } | |
0b2c2a84 DM |
672 | if (prog->extflags & RXf_ANCH) { /* Match at \G, beg-of-str or after \n */ |
673 | ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) | |
bbe252da | 674 | || ( (prog->extflags & RXf_ANCH_BOL) |
7fba1cd6 | 675 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 676 | |
7e25d62c | 677 | if (!ml_anch) { |
0b2c2a84 DM |
678 | /* we are only allowed to match at BOS or \G */ |
679 | ||
680 | if (prog->extflags & RXf_ANCH_GPOS) { | |
681 | /* in this case, we hope(!) that the caller has already | |
682 | * set strpos to pos()-gofs, and will already have checked | |
683 | * that this anchor position is legal | |
684 | */ | |
685 | ; | |
686 | } | |
687 | else if (!(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ | |
688 | && (strpos != strbeg)) | |
689 | { | |
a3621e74 | 690 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
691 | goto fail; |
692 | } | |
d46b78c6 KW |
693 | if (prog->check_offset_min == prog->check_offset_max |
694 | && !(prog->extflags & RXf_CANY_SEEN) | |
695 | && ! multiline) /* /m can cause \n's to match that aren't | |
696 | accounted for in the string max length. | |
697 | See [perl #115242] */ | |
698 | { | |
2c2d71f5 | 699 | /* Substring at constant offset from beg-of-str... */ |
ea3daa5d | 700 | SSize_t slen; |
cad2e5aa | 701 | |
1aa99e6b | 702 | s = HOP3c(strpos, prog->check_offset_min, strend); |
1de06328 | 703 | |
653099ff GS |
704 | if (SvTAIL(check)) { |
705 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 706 | |
9041c2e3 | 707 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 708 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 709 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 710 | goto fail_finish; |
cad2e5aa JH |
711 | } |
712 | /* Now should match s[0..slen-2] */ | |
713 | slen--; | |
3f7c398e | 714 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 715 | || (slen > 1 |
3f7c398e | 716 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 717 | report_neq: |
a3621e74 | 718 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
719 | goto fail_finish; |
720 | } | |
cad2e5aa | 721 | } |
3f7c398e | 722 | else if (*SvPVX_const(check) != *s |
653099ff | 723 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 724 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 725 | goto report_neq; |
c315bfe8 | 726 | check_at = s; |
2c2d71f5 | 727 | goto success_at_start; |
7e25d62c | 728 | } |
cad2e5aa | 729 | } |
2c2d71f5 | 730 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 731 | s = strpos; |
2c2d71f5 | 732 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
1de06328 YO |
733 | end_shift = prog->check_end_shift; |
734 | ||
2c2d71f5 | 735 | if (!ml_anch) { |
ea3daa5d | 736 | const SSize_t end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 737 | - (SvTAIL(check) != 0); |
ea3daa5d | 738 | const SSize_t eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
739 | |
740 | if (end_shift < eshift) | |
741 | end_shift = eshift; | |
742 | } | |
cad2e5aa | 743 | } |
2c2d71f5 | 744 | else { /* Can match at random position */ |
cad2e5aa JH |
745 | ml_anch = 0; |
746 | s = strpos; | |
1de06328 YO |
747 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
748 | end_shift = prog->check_end_shift; | |
749 | ||
750 | /* end shift should be non negative here */ | |
cad2e5aa JH |
751 | } |
752 | ||
19188028 | 753 | #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 754 | if (end_shift < 0) |
1de06328 | 755 | Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", |
220fc49f | 756 | (IV)end_shift, RX_PRECOMP(prog)); |
2c2d71f5 JH |
757 | #endif |
758 | ||
2c2d71f5 JH |
759 | restart: |
760 | /* Find a possible match in the region s..strend by looking for | |
761 | the "check" substring in the region corrected by start/end_shift. */ | |
1de06328 YO |
762 | |
763 | { | |
49f55535 | 764 | SSize_t srch_start_shift = start_shift; |
ea3daa5d | 765 | SSize_t srch_end_shift = end_shift; |
c33e64f0 FC |
766 | U8* start_point; |
767 | U8* end_point; | |
1de06328 YO |
768 | if (srch_start_shift < 0 && strbeg - s > srch_start_shift) { |
769 | srch_end_shift -= ((strbeg - s) - srch_start_shift); | |
770 | srch_start_shift = strbeg - s; | |
771 | } | |
6bda09f9 | 772 | DEBUG_OPTIMISE_MORE_r({ |
1de06328 YO |
773 | PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n", |
774 | (IV)prog->check_offset_min, | |
775 | (IV)srch_start_shift, | |
776 | (IV)srch_end_shift, | |
777 | (IV)prog->check_end_shift); | |
778 | }); | |
779 | ||
bbe252da | 780 | if (prog->extflags & RXf_CANY_SEEN) { |
1de06328 YO |
781 | start_point= (U8*)(s + srch_start_shift); |
782 | end_point= (U8*)(strend - srch_end_shift); | |
783 | } else { | |
784 | start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); | |
785 | end_point= HOP3(strend, -srch_end_shift, strbeg); | |
786 | } | |
6bda09f9 | 787 | DEBUG_OPTIMISE_MORE_r({ |
56570a2c | 788 | PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", |
1de06328 | 789 | (int)(end_point - start_point), |
fc8cd66c | 790 | (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), |
1de06328 YO |
791 | start_point); |
792 | }); | |
793 | ||
794 | s = fbm_instr( start_point, end_point, | |
7fba1cd6 | 795 | check, multiline ? FBMrf_MULTILINE : 0); |
1de06328 | 796 | } |
cad2e5aa JH |
797 | /* Update the count-of-usability, remove useless subpatterns, |
798 | unshift s. */ | |
2c2d71f5 | 799 | |
ab3bbdeb | 800 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 801 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
802 | SvPVX_const(check), RE_SV_DUMPLEN(check), 30); |
803 | PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", | |
2c2d71f5 | 804 | (s ? "Found" : "Did not find"), |
f2ed9b32 | 805 | (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) |
ab3bbdeb YO |
806 | ? "anchored" : "floating"), |
807 | quoted, | |
808 | RE_SV_TAIL(check), | |
809 | (s ? " at offset " : "...\n") ); | |
810 | }); | |
2c2d71f5 JH |
811 | |
812 | if (!s) | |
813 | goto fail_finish; | |
2c2d71f5 | 814 | /* Finish the diagnostic message */ |
a3621e74 | 815 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 | 816 | |
1de06328 YO |
817 | /* XXX dmq: first branch is for positive lookbehind... |
818 | Our check string is offset from the beginning of the pattern. | |
819 | So we need to do any stclass tests offset forward from that | |
820 | point. I think. :-( | |
821 | */ | |
822 | ||
823 | ||
824 | ||
825 | check_at=s; | |
826 | ||
827 | ||
2c2d71f5 JH |
828 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. |
829 | Start with the other substr. | |
830 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 831 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
832 | *always* match. Probably should be marked during compile... |
833 | Probably it is right to do no SCREAM here... | |
834 | */ | |
835 | ||
f2ed9b32 | 836 | if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8) |
1de06328 YO |
837 | : (prog->float_substr && prog->anchored_substr)) |
838 | { | |
30944b6d | 839 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
840 | /* XXXX May be hopelessly wrong for UTF... */ |
841 | if (!other_last) | |
6eb5f6b9 | 842 | other_last = strpos; |
f2ed9b32 | 843 | if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
844 | do_other_anchored: |
845 | { | |
890ce7af AL |
846 | char * const last = HOP3c(s, -start_shift, strbeg); |
847 | char *last1, *last2; | |
be8e71aa | 848 | char * const saved_s = s; |
33b8afdf | 849 | SV* must; |
2c2d71f5 | 850 | |
2c2d71f5 JH |
851 | t = s - prog->check_offset_max; |
852 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
f2ed9b32 | 853 | && (!utf8_target |
0ce71af7 | 854 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 855 | && t > strpos))) |
6f207bd3 | 856 | NOOP; |
2c2d71f5 JH |
857 | else |
858 | t = strpos; | |
1aa99e6b | 859 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
860 | if (t < other_last) /* These positions already checked */ |
861 | t = other_last; | |
1aa99e6b | 862 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
863 | if (last < last1) |
864 | last1 = last; | |
1de06328 YO |
865 | /* XXXX It is not documented what units *_offsets are in. |
866 | We assume bytes, but this is clearly wrong. | |
867 | Meaning this code needs to be carefully reviewed for errors. | |
868 | dmq. | |
869 | */ | |
870 | ||
2c2d71f5 | 871 | /* On end-of-str: see comment below. */ |
f2ed9b32 | 872 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; |
33b8afdf JH |
873 | if (must == &PL_sv_undef) { |
874 | s = (char*)NULL; | |
1de06328 | 875 | DEBUG_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
876 | } |
877 | else | |
878 | s = fbm_instr( | |
879 | (unsigned char*)t, | |
880 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
881 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
882 | must, | |
7fba1cd6 | 883 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 884 | ); |
ab3bbdeb | 885 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 886 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
887 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
888 | PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", | |
2c2d71f5 | 889 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
890 | quoted, RE_SV_TAIL(must)); |
891 | }); | |
892 | ||
893 | ||
2c2d71f5 JH |
894 | if (!s) { |
895 | if (last1 >= last2) { | |
a3621e74 | 896 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
897 | ", giving up...\n")); |
898 | goto fail_finish; | |
899 | } | |
a3621e74 | 900 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 901 | ", trying floating at offset %ld...\n", |
be8e71aa | 902 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
903 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
904 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
905 | goto restart; |
906 | } | |
907 | else { | |
a3621e74 | 908 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 909 | (long)(s - i_strpos))); |
1aa99e6b IH |
910 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
911 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 912 | s = saved_s; |
2c2d71f5 JH |
913 | if (t == strpos) |
914 | goto try_at_start; | |
2c2d71f5 JH |
915 | goto try_at_offset; |
916 | } | |
30944b6d | 917 | } |
2c2d71f5 JH |
918 | } |
919 | else { /* Take into account the floating substring. */ | |
33b8afdf | 920 | char *last, *last1; |
be8e71aa | 921 | char * const saved_s = s; |
33b8afdf JH |
922 | SV* must; |
923 | ||
924 | t = HOP3c(s, -start_shift, strbeg); | |
925 | last1 = last = | |
926 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
927 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
928 | last = HOP3c(t, prog->float_max_offset, strend); | |
929 | s = HOP3c(t, prog->float_min_offset, strend); | |
930 | if (s < other_last) | |
931 | s = other_last; | |
2c2d71f5 | 932 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
f2ed9b32 | 933 | must = utf8_target ? prog->float_utf8 : prog->float_substr; |
33b8afdf JH |
934 | /* fbm_instr() takes into account exact value of end-of-str |
935 | if the check is SvTAIL(ed). Since false positives are OK, | |
936 | and end-of-str is not later than strend we are OK. */ | |
937 | if (must == &PL_sv_undef) { | |
938 | s = (char*)NULL; | |
1de06328 | 939 | DEBUG_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
940 | } |
941 | else | |
2c2d71f5 | 942 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
943 | (unsigned char*)last + SvCUR(must) |
944 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 945 | must, multiline ? FBMrf_MULTILINE : 0); |
ab3bbdeb | 946 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 947 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
948 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
949 | PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", | |
33b8afdf | 950 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
951 | quoted, RE_SV_TAIL(must)); |
952 | }); | |
33b8afdf JH |
953 | if (!s) { |
954 | if (last1 == last) { | |
a3621e74 | 955 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
956 | ", giving up...\n")); |
957 | goto fail_finish; | |
2c2d71f5 | 958 | } |
a3621e74 | 959 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 960 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 961 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
962 | other_last = last; |
963 | s = HOP3c(t, 1, strend); | |
964 | goto restart; | |
965 | } | |
966 | else { | |
a3621e74 | 967 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
968 | (long)(s - i_strpos))); |
969 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 970 | s = saved_s; |
33b8afdf JH |
971 | if (t == strpos) |
972 | goto try_at_start; | |
973 | goto try_at_offset; | |
974 | } | |
2c2d71f5 | 975 | } |
cad2e5aa | 976 | } |
2c2d71f5 | 977 | |
1de06328 | 978 | |
9ef43ace | 979 | t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); |
1de06328 | 980 | |
6bda09f9 | 981 | DEBUG_OPTIMISE_MORE_r( |
1de06328 YO |
982 | PerlIO_printf(Perl_debug_log, |
983 | "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", | |
984 | (IV)prog->check_offset_min, | |
985 | (IV)prog->check_offset_max, | |
986 | (IV)(s-strpos), | |
987 | (IV)(t-strpos), | |
988 | (IV)(t-s), | |
989 | (IV)(strend-strpos) | |
990 | ) | |
991 | ); | |
992 | ||
2c2d71f5 | 993 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
f2ed9b32 | 994 | && (!utf8_target |
9ef43ace | 995 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) |
1de06328 YO |
996 | && t > strpos))) |
997 | { | |
2c2d71f5 JH |
998 | /* Fixed substring is found far enough so that the match |
999 | cannot start at strpos. */ | |
1000 | try_at_offset: | |
cad2e5aa | 1001 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
1002 | /* Eventually fbm_*() should handle this, but often |
1003 | anchored_offset is not 0, so this check will not be wasted. */ | |
1004 | /* XXXX In the code below we prefer to look for "^" even in | |
1005 | presence of anchored substrings. And we search even | |
1006 | beyond the found float position. These pessimizations | |
1007 | are historical artefacts only. */ | |
1008 | find_anchor: | |
2c2d71f5 | 1009 | while (t < strend - prog->minlen) { |
cad2e5aa | 1010 | if (*t == '\n') { |
4ee3650e | 1011 | if (t < check_at - prog->check_offset_min) { |
f2ed9b32 | 1012 | if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
1013 | /* Since we moved from the found position, |
1014 | we definitely contradict the found anchored | |
30944b6d IZ |
1015 | substr. Due to the above check we do not |
1016 | contradict "check" substr. | |
1017 | Thus we can arrive here only if check substr | |
1018 | is float. Redo checking for "other"=="fixed". | |
1019 | */ | |
9041c2e3 | 1020 | strpos = t + 1; |
a3621e74 | 1021 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 1022 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
1023 | goto do_other_anchored; |
1024 | } | |
4ee3650e GS |
1025 | /* We don't contradict the found floating substring. */ |
1026 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 1027 | s = t + 1; |
a3621e74 | 1028 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 1029 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
1030 | goto set_useful; |
1031 | } | |
4ee3650e GS |
1032 | /* Position contradicts check-string */ |
1033 | /* XXXX probably better to look for check-string | |
1034 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 1035 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 1036 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 1037 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
1038 | goto restart; |
1039 | } | |
1040 | t++; | |
1041 | } | |
a3621e74 | 1042 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 1043 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 1044 | goto fail_finish; |
cad2e5aa | 1045 | } |
f5952150 | 1046 | else { |
a3621e74 | 1047 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 1048 | PL_colors[0], PL_colors[1])); |
f5952150 | 1049 | } |
cad2e5aa JH |
1050 | s = t; |
1051 | set_useful: | |
f2ed9b32 | 1052 | ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
1053 | } |
1054 | else { | |
f5952150 | 1055 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 1056 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
1057 | unless it was an MBOL and we are not after MBOL, |
1058 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
1059 | try_at_start: |
1060 | /* Even in this situation we may use MBOL flag if strpos is offset | |
1061 | wrt the start of the string. */ | |
52a21eb3 | 1062 | if (ml_anch && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d | 1063 | /* May be due to an implicit anchor of m{.*foo} */ |
bbe252da | 1064 | && !(prog->intflags & PREGf_IMPLICIT)) |
d506a20d | 1065 | { |
cad2e5aa JH |
1066 | t = strpos; |
1067 | goto find_anchor; | |
1068 | } | |
a3621e74 | 1069 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 1070 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
70685ca0 | 1071 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 1072 | ); |
2c2d71f5 | 1073 | success_at_start: |
bbe252da | 1074 | if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ |
f2ed9b32 | 1075 | && (utf8_target ? ( |
33b8afdf JH |
1076 | prog->check_utf8 /* Could be deleted already */ |
1077 | && --BmUSEFUL(prog->check_utf8) < 0 | |
1078 | && (prog->check_utf8 == prog->float_utf8) | |
1079 | ) : ( | |
1080 | prog->check_substr /* Could be deleted already */ | |
1081 | && --BmUSEFUL(prog->check_substr) < 0 | |
1082 | && (prog->check_substr == prog->float_substr) | |
1083 | ))) | |
66e933ab | 1084 | { |
cad2e5aa | 1085 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 1086 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
f2ed9b32 KW |
1087 | /* XXX Does the destruction order has to change with utf8_target? */ |
1088 | SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr); | |
1089 | SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
1090 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
1091 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
1092 | check = NULL; /* abort */ | |
cad2e5aa | 1093 | s = strpos; |
486ec47a | 1094 | /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag |
c9415951 YO |
1095 | see http://bugs.activestate.com/show_bug.cgi?id=87173 */ |
1096 | if (prog->intflags & PREGf_IMPLICIT) | |
1097 | prog->extflags &= ~RXf_ANCH_MBOL; | |
3cf5c195 IZ |
1098 | /* XXXX This is a remnant of the old implementation. It |
1099 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 1100 | other heuristics. */ |
bbe252da | 1101 | prog->extflags &= ~RXf_USE_INTUIT; |
c9415951 | 1102 | /* XXXX What other flags might need to be cleared in this branch? */ |
cad2e5aa JH |
1103 | } |
1104 | else | |
1105 | s = strpos; | |
1106 | } | |
1107 | ||
6eb5f6b9 JH |
1108 | /* Last resort... */ |
1109 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
1de06328 YO |
1110 | /* trie stclasses are too expensive to use here, we are better off to |
1111 | leave it to regmatch itself */ | |
f8fc2ecf | 1112 | if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { |
6eb5f6b9 JH |
1113 | /* minlen == 0 is possible if regstclass is \b or \B, |
1114 | and the fixed substr is ''$. | |
1115 | Since minlen is already taken into account, s+1 is before strend; | |
1116 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
1117 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
1118 | regstclass does not come from lookahead... */ | |
1119 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
af944926 | 1120 | This leaves EXACTF-ish only, which are dealt with in find_byclass(). */ |
f8fc2ecf | 1121 | const U8* const str = (U8*)STRING(progi->regstclass); |
2c75e362 | 1122 | /* XXX this value could be pre-computed */ |
f8fc2ecf | 1123 | const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT |
2c75e362 DM |
1124 | ? (reginfo->is_utf8_pat |
1125 | ? utf8_distance(str + STR_LEN(progi->regstclass), str) | |
1126 | : STR_LEN(progi->regstclass)) | |
66e933ab | 1127 | : 1); |
1de06328 YO |
1128 | char * endpos; |
1129 | if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) | |
1130 | endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); | |
1131 | else if (prog->float_substr || prog->float_utf8) | |
1132 | endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); | |
1133 | else | |
1134 | endpos= strend; | |
1135 | ||
d8080198 YO |
1136 | if (checked_upto < s) |
1137 | checked_upto = s; | |
1138 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n", | |
1139 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg))); | |
1140 | ||
6eb5f6b9 | 1141 | t = s; |
984e6dd1 | 1142 | s = find_byclass(prog, progi->regstclass, checked_upto, endpos, |
f9176b44 | 1143 | reginfo); |
d8080198 YO |
1144 | if (s) { |
1145 | checked_upto = s; | |
1146 | } else { | |
6eb5f6b9 | 1147 | #ifdef DEBUGGING |
cbbf8932 | 1148 | const char *what = NULL; |
6eb5f6b9 JH |
1149 | #endif |
1150 | if (endpos == strend) { | |
a3621e74 | 1151 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
1152 | "Could not match STCLASS...\n") ); |
1153 | goto fail; | |
1154 | } | |
a3621e74 | 1155 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 1156 | "This position contradicts STCLASS...\n") ); |
bbe252da | 1157 | if ((prog->extflags & RXf_ANCH) && !ml_anch) |
653099ff | 1158 | goto fail; |
d8080198 YO |
1159 | checked_upto = HOPBACKc(endpos, start_shift); |
1160 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n", | |
1161 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg))); | |
6eb5f6b9 | 1162 | /* Contradict one of substrings */ |
33b8afdf | 1163 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 | 1164 | if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) { |
a3621e74 | 1165 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 1166 | hop_and_restart: |
1aa99e6b | 1167 | s = HOP3c(t, 1, strend); |
66e933ab GS |
1168 | if (s + start_shift + end_shift > strend) { |
1169 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 1170 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
1171 | "Could not match STCLASS...\n") ); |
1172 | goto fail; | |
1173 | } | |
5e39e1e5 HS |
1174 | if (!check) |
1175 | goto giveup; | |
a3621e74 | 1176 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1177 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1178 | what, (long)(s + start_shift - i_strpos)) ); |
1179 | goto restart; | |
1180 | } | |
66e933ab | 1181 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
1182 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
1183 | goto retry_floating_check; | |
1184 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 1185 | s = check_at; |
5e39e1e5 HS |
1186 | if (!check) |
1187 | goto giveup; | |
a3621e74 | 1188 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1189 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1190 | (long)(other_last - i_strpos)) ); |
1191 | goto do_other_anchored; | |
1192 | } | |
60e71179 GS |
1193 | /* Another way we could have checked stclass at the |
1194 | current position only: */ | |
1195 | if (ml_anch) { | |
1196 | s = t = t + 1; | |
5e39e1e5 HS |
1197 | if (!check) |
1198 | goto giveup; | |
a3621e74 | 1199 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1200 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 1201 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 1202 | goto try_at_offset; |
66e933ab | 1203 | } |
f2ed9b32 | 1204 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 1205 | goto fail; |
486ec47a | 1206 | /* Check is floating substring. */ |
6eb5f6b9 JH |
1207 | retry_floating_check: |
1208 | t = check_at - start_shift; | |
a3621e74 | 1209 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
1210 | goto hop_and_restart; |
1211 | } | |
b7953727 | 1212 | if (t != s) { |
a3621e74 | 1213 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 1214 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
1215 | (long)(t - i_strpos), (long)(s - i_strpos)) |
1216 | ); | |
1217 | } | |
1218 | else { | |
a3621e74 | 1219 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1220 | "Does not contradict STCLASS...\n"); |
1221 | ); | |
1222 | } | |
6eb5f6b9 | 1223 | } |
5e39e1e5 | 1224 | giveup: |
a3621e74 | 1225 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
1226 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
1227 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 1228 | return s; |
2c2d71f5 JH |
1229 | |
1230 | fail_finish: /* Substring not found */ | |
33b8afdf | 1231 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
f2ed9b32 | 1232 | BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ |
cad2e5aa | 1233 | fail: |
a3621e74 | 1234 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 1235 | PL_colors[4], PL_colors[5])); |
bd61b366 | 1236 | return NULL; |
cad2e5aa | 1237 | } |
9661b544 | 1238 | |
a0a388a1 | 1239 | #define DECL_TRIE_TYPE(scan) \ |
098b07d5 KW |
1240 | const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold, \ |
1241 | trie_utf8_exactfa_fold, trie_latin_utf8_exactfa_fold } \ | |
fab2782b YO |
1242 | trie_type = ((scan->flags == EXACT) \ |
1243 | ? (utf8_target ? trie_utf8 : trie_plain) \ | |
098b07d5 KW |
1244 | : (scan->flags == EXACTFA) \ |
1245 | ? (utf8_target ? trie_utf8_exactfa_fold : trie_latin_utf8_exactfa_fold) \ | |
1246 | : (utf8_target ? trie_utf8_fold : trie_latin_utf8_fold)) | |
fab2782b | 1247 | |
fd3249ee YO |
1248 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, uvc, charid, foldlen, foldbuf, uniflags) \ |
1249 | STMT_START { \ | |
fab2782b | 1250 | STRLEN skiplen; \ |
098b07d5 | 1251 | U8 flags = FOLD_FLAGS_FULL; \ |
fab2782b | 1252 | switch (trie_type) { \ |
098b07d5 KW |
1253 | case trie_utf8_exactfa_fold: \ |
1254 | flags |= FOLD_FLAGS_NOMIX_ASCII; \ | |
1255 | /* FALL THROUGH */ \ | |
fab2782b YO |
1256 | case trie_utf8_fold: \ |
1257 | if ( foldlen>0 ) { \ | |
c80e42f3 | 1258 | uvc = utf8n_to_uvchr( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ |
fab2782b YO |
1259 | foldlen -= len; \ |
1260 | uscan += len; \ | |
1261 | len=0; \ | |
1262 | } else { \ | |
098b07d5 | 1263 | uvc = _to_utf8_fold_flags( (const U8*) uc, foldbuf, &foldlen, flags, NULL); \ |
fab2782b YO |
1264 | len = UTF8SKIP(uc); \ |
1265 | skiplen = UNISKIP( uvc ); \ | |
1266 | foldlen -= skiplen; \ | |
1267 | uscan = foldbuf + skiplen; \ | |
1268 | } \ | |
1269 | break; \ | |
098b07d5 KW |
1270 | case trie_latin_utf8_exactfa_fold: \ |
1271 | flags |= FOLD_FLAGS_NOMIX_ASCII; \ | |
1272 | /* FALL THROUGH */ \ | |
fab2782b YO |
1273 | case trie_latin_utf8_fold: \ |
1274 | if ( foldlen>0 ) { \ | |
c80e42f3 | 1275 | uvc = utf8n_to_uvchr( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ |
fab2782b YO |
1276 | foldlen -= len; \ |
1277 | uscan += len; \ | |
1278 | len=0; \ | |
1279 | } else { \ | |
1280 | len = 1; \ | |
098b07d5 | 1281 | uvc = _to_fold_latin1( (U8) *uc, foldbuf, &foldlen, flags); \ |
fab2782b YO |
1282 | skiplen = UNISKIP( uvc ); \ |
1283 | foldlen -= skiplen; \ | |
1284 | uscan = foldbuf + skiplen; \ | |
1285 | } \ | |
1286 | break; \ | |
1287 | case trie_utf8: \ | |
c80e42f3 | 1288 | uvc = utf8n_to_uvchr( (const U8*) uc, UTF8_MAXLEN, &len, uniflags ); \ |
fab2782b YO |
1289 | break; \ |
1290 | case trie_plain: \ | |
1291 | uvc = (UV)*uc; \ | |
1292 | len = 1; \ | |
1293 | } \ | |
1294 | if (uvc < 256) { \ | |
1295 | charid = trie->charmap[ uvc ]; \ | |
1296 | } \ | |
1297 | else { \ | |
1298 | charid = 0; \ | |
1299 | if (widecharmap) { \ | |
1300 | SV** const svpp = hv_fetch(widecharmap, \ | |
1301 | (char*)&uvc, sizeof(UV), 0); \ | |
1302 | if (svpp) \ | |
1303 | charid = (U16)SvIV(*svpp); \ | |
1304 | } \ | |
1305 | } \ | |
4cadc6a9 YO |
1306 | } STMT_END |
1307 | ||
4cadc6a9 YO |
1308 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ |
1309 | STMT_START { \ | |
1310 | while (s <= e) { \ | |
1311 | if ( (CoNd) \ | |
fac1af77 | 1312 | && (ln == 1 || folder(s, pat_string, ln)) \ |
02d5137b | 1313 | && (reginfo->intuit || regtry(reginfo, &s)) )\ |
4cadc6a9 YO |
1314 | goto got_it; \ |
1315 | s++; \ | |
1316 | } \ | |
1317 | } STMT_END | |
1318 | ||
1319 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
1320 | STMT_START { \ | |
9a902117 | 1321 | while (s < strend) { \ |
4cadc6a9 | 1322 | CoDe \ |
9a902117 | 1323 | s += UTF8SKIP(s); \ |
4cadc6a9 YO |
1324 | } \ |
1325 | } STMT_END | |
1326 | ||
1327 | #define REXEC_FBC_SCAN(CoDe) \ | |
1328 | STMT_START { \ | |
1329 | while (s < strend) { \ | |
1330 | CoDe \ | |
1331 | s++; \ | |
1332 | } \ | |
1333 | } STMT_END | |
1334 | ||
1335 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1336 | REXEC_FBC_UTF8_SCAN( \ | |
1337 | if (CoNd) { \ | |
02d5137b | 1338 | if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1339 | goto got_it; \ |
1340 | else \ | |
1341 | tmp = doevery; \ | |
1342 | } \ | |
1343 | else \ | |
1344 | tmp = 1; \ | |
1345 | ) | |
1346 | ||
1347 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1348 | REXEC_FBC_SCAN( \ | |
1349 | if (CoNd) { \ | |
02d5137b | 1350 | if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1351 | goto got_it; \ |
1352 | else \ | |
1353 | tmp = doevery; \ | |
1354 | } \ | |
1355 | else \ | |
1356 | tmp = 1; \ | |
1357 | ) | |
1358 | ||
1359 | #define REXEC_FBC_TRYIT \ | |
02d5137b | 1360 | if ((reginfo->intuit || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1361 | goto got_it |
1362 | ||
e1d1eefb | 1363 | #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \ |
f2ed9b32 | 1364 | if (utf8_target) { \ |
e1d1eefb YO |
1365 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1366 | } \ | |
1367 | else { \ | |
1368 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1369 | } |
e1d1eefb | 1370 | |
786e8c11 | 1371 | #define DUMP_EXEC_POS(li,s,doutf8) \ |
9d9163fb | 1372 | dump_exec_pos(li,s,(reginfo->strend),(reginfo->strbeg), \ |
6d59b646 | 1373 | startpos, doutf8) |
786e8c11 | 1374 | |
cfaf538b KW |
1375 | |
1376 | #define UTF8_NOLOAD(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ | |
9d9163fb | 1377 | tmp = (s != reginfo->strbeg) ? UCHARAT(s - 1) : '\n'; \ |
cfaf538b KW |
1378 | tmp = TEST_NON_UTF8(tmp); \ |
1379 | REXEC_FBC_UTF8_SCAN( \ | |
1380 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1381 | tmp = !tmp; \ | |
1382 | IF_SUCCESS; \ | |
1383 | } \ | |
1384 | else { \ | |
1385 | IF_FAIL; \ | |
1386 | } \ | |
1387 | ); \ | |
1388 | ||
1389 | #define UTF8_LOAD(TeSt1_UtF8, TeSt2_UtF8, IF_SUCCESS, IF_FAIL) \ | |
9d9163fb | 1390 | if (s == reginfo->strbeg) { \ |
cfaf538b KW |
1391 | tmp = '\n'; \ |
1392 | } \ | |
1393 | else { \ | |
9d9163fb | 1394 | U8 * const r = reghop3((U8*)s, -1, (U8*)reginfo->strbeg); \ |
cfaf538b KW |
1395 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); \ |
1396 | } \ | |
1397 | tmp = TeSt1_UtF8; \ | |
1398 | LOAD_UTF8_CHARCLASS_ALNUM(); \ | |
1399 | REXEC_FBC_UTF8_SCAN( \ | |
1400 | if (tmp == ! (TeSt2_UtF8)) { \ | |
1401 | tmp = !tmp; \ | |
1402 | IF_SUCCESS; \ | |
1403 | } \ | |
1404 | else { \ | |
1405 | IF_FAIL; \ | |
1406 | } \ | |
1407 | ); \ | |
1408 | ||
63ac0dad KW |
1409 | /* The only difference between the BOUND and NBOUND cases is that |
1410 | * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in | |
1411 | * NBOUND. This is accomplished by passing it in either the if or else clause, | |
1412 | * with the other one being empty */ | |
1413 | #define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1414 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
cfaf538b KW |
1415 | |
1416 | #define FBC_BOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1417 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
63ac0dad KW |
1418 | |
1419 | #define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1420 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b KW |
1421 | |
1422 | #define FBC_NBOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1423 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b | 1424 | |
63ac0dad KW |
1425 | |
1426 | /* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to | |
1427 | * be passed in completely with the variable name being tested, which isn't | |
1428 | * such a clean interface, but this is easier to read than it was before. We | |
1429 | * are looking for the boundary (or non-boundary between a word and non-word | |
1430 | * character. The utf8 and non-utf8 cases have the same logic, but the details | |
1431 | * must be different. Find the "wordness" of the character just prior to this | |
1432 | * one, and compare it with the wordness of this one. If they differ, we have | |
1433 | * a boundary. At the beginning of the string, pretend that the previous | |
1434 | * character was a new-line */ | |
cfaf538b | 1435 | #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ |
63ac0dad | 1436 | if (utf8_target) { \ |
cfaf538b | 1437 | UTF8_CODE \ |
63ac0dad KW |
1438 | } \ |
1439 | else { /* Not utf8 */ \ | |
9d9163fb | 1440 | tmp = (s != reginfo->strbeg) ? UCHARAT(s - 1) : '\n'; \ |
63ac0dad KW |
1441 | tmp = TEST_NON_UTF8(tmp); \ |
1442 | REXEC_FBC_SCAN( \ | |
1443 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1444 | tmp = !tmp; \ | |
1445 | IF_SUCCESS; \ | |
1446 | } \ | |
1447 | else { \ | |
1448 | IF_FAIL; \ | |
1449 | } \ | |
1450 | ); \ | |
1451 | } \ | |
02d5137b | 1452 | if ((!prog->minlen && tmp) && (reginfo->intuit || regtry(reginfo, &s))) \ |
63ac0dad KW |
1453 | goto got_it; |
1454 | ||
786e8c11 | 1455 | /* We know what class REx starts with. Try to find this position... */ |
02d5137b | 1456 | /* if reginfo->intuit, its a dryrun */ |
786e8c11 YO |
1457 | /* annoyingly all the vars in this routine have different names from their counterparts |
1458 | in regmatch. /grrr */ | |
1459 | ||
3c3eec57 | 1460 | STATIC char * |
07be1b83 | 1461 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
f9176b44 | 1462 | const char *strend, regmatch_info *reginfo) |
a687059c | 1463 | { |
73104a1b KW |
1464 | dVAR; |
1465 | const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; | |
1466 | char *pat_string; /* The pattern's exactish string */ | |
1467 | char *pat_end; /* ptr to end char of pat_string */ | |
1468 | re_fold_t folder; /* Function for computing non-utf8 folds */ | |
1469 | const U8 *fold_array; /* array for folding ords < 256 */ | |
1470 | STRLEN ln; | |
1471 | STRLEN lnc; | |
73104a1b KW |
1472 | U8 c1; |
1473 | U8 c2; | |
1474 | char *e; | |
1475 | I32 tmp = 1; /* Scratch variable? */ | |
ba44c216 | 1476 | const bool utf8_target = reginfo->is_utf8_target; |
73104a1b | 1477 | UV utf8_fold_flags = 0; |
f9176b44 | 1478 | const bool is_utf8_pat = reginfo->is_utf8_pat; |
3018b823 KW |
1479 | bool to_complement = FALSE; /* Invert the result? Taking the xor of this |
1480 | with a result inverts that result, as 0^1 = | |
1481 | 1 and 1^1 = 0 */ | |
1482 | _char_class_number classnum; | |
1483 | ||
73104a1b | 1484 | RXi_GET_DECL(prog,progi); |
2f7f8cb1 | 1485 | |
73104a1b | 1486 | PERL_ARGS_ASSERT_FIND_BYCLASS; |
2f7f8cb1 | 1487 | |
73104a1b KW |
1488 | /* We know what class it must start with. */ |
1489 | switch (OP(c)) { | |
1490 | case ANYOF: | |
9aa1e39f | 1491 | case ANYOF_SYNTHETIC: |
954a2af6 | 1492 | case ANYOF_WARN_SUPER: |
73104a1b KW |
1493 | if (utf8_target) { |
1494 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1495 | reginclass(prog, c, (U8*)s, utf8_target)); | |
1496 | } | |
1497 | else { | |
1498 | REXEC_FBC_CLASS_SCAN(REGINCLASS(prog, c, (U8*)s)); | |
1499 | } | |
1500 | break; | |
1501 | case CANY: | |
1502 | REXEC_FBC_SCAN( | |
02d5137b | 1503 | if (tmp && (reginfo->intuit || regtry(reginfo, &s))) |
73104a1b KW |
1504 | goto got_it; |
1505 | else | |
1506 | tmp = doevery; | |
1507 | ); | |
1508 | break; | |
1509 | ||
098b07d5 KW |
1510 | case EXACTFA_NO_TRIE: /* This node only generated for non-utf8 patterns */ |
1511 | assert(! is_utf8_pat); | |
1512 | /* FALL THROUGH */ | |
73104a1b | 1513 | case EXACTFA: |
984e6dd1 | 1514 | if (is_utf8_pat || utf8_target) { |
73104a1b KW |
1515 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; |
1516 | goto do_exactf_utf8; | |
1517 | } | |
1518 | fold_array = PL_fold_latin1; /* Latin1 folds are not affected by */ | |
1519 | folder = foldEQ_latin1; /* /a, except the sharp s one which */ | |
1520 | goto do_exactf_non_utf8; /* isn't dealt with by these */ | |
77a6d856 | 1521 | |
2fdb7295 KW |
1522 | case EXACTF: /* This node only generated for non-utf8 patterns */ |
1523 | assert(! is_utf8_pat); | |
73104a1b | 1524 | if (utf8_target) { |
73104a1b KW |
1525 | utf8_fold_flags = 0; |
1526 | goto do_exactf_utf8; | |
1527 | } | |
1528 | fold_array = PL_fold; | |
1529 | folder = foldEQ; | |
1530 | goto do_exactf_non_utf8; | |
1531 | ||
1532 | case EXACTFL: | |
984e6dd1 | 1533 | if (is_utf8_pat || utf8_target) { |
73104a1b KW |
1534 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
1535 | goto do_exactf_utf8; | |
1536 | } | |
1537 | fold_array = PL_fold_locale; | |
1538 | folder = foldEQ_locale; | |
1539 | goto do_exactf_non_utf8; | |
3c760661 | 1540 | |
73104a1b | 1541 | case EXACTFU_SS: |
984e6dd1 | 1542 | if (is_utf8_pat) { |
73104a1b KW |
1543 | utf8_fold_flags = FOLDEQ_S2_ALREADY_FOLDED; |
1544 | } | |
1545 | goto do_exactf_utf8; | |
16d951b7 | 1546 | |
73104a1b | 1547 | case EXACTFU: |
984e6dd1 DM |
1548 | if (is_utf8_pat || utf8_target) { |
1549 | utf8_fold_flags = is_utf8_pat ? FOLDEQ_S2_ALREADY_FOLDED : 0; | |
73104a1b KW |
1550 | goto do_exactf_utf8; |
1551 | } | |
fac1af77 | 1552 | |
73104a1b KW |
1553 | /* Any 'ss' in the pattern should have been replaced by regcomp, |
1554 | * so we don't have to worry here about this single special case | |
1555 | * in the Latin1 range */ | |
1556 | fold_array = PL_fold_latin1; | |
1557 | folder = foldEQ_latin1; | |
1558 | ||
1559 | /* FALL THROUGH */ | |
1560 | ||
1561 | do_exactf_non_utf8: /* Neither pattern nor string are UTF8, and there | |
1562 | are no glitches with fold-length differences | |
1563 | between the target string and pattern */ | |
1564 | ||
1565 | /* The idea in the non-utf8 EXACTF* cases is to first find the | |
1566 | * first character of the EXACTF* node and then, if necessary, | |
1567 | * case-insensitively compare the full text of the node. c1 is the | |
1568 | * first character. c2 is its fold. This logic will not work for | |
1569 | * Unicode semantics and the german sharp ss, which hence should | |
1570 | * not be compiled into a node that gets here. */ | |
1571 | pat_string = STRING(c); | |
1572 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1573 | ||
1574 | /* We know that we have to match at least 'ln' bytes (which is the | |
1575 | * same as characters, since not utf8). If we have to match 3 | |
1576 | * characters, and there are only 2 availabe, we know without | |
1577 | * trying that it will fail; so don't start a match past the | |
1578 | * required minimum number from the far end */ | |
ea3daa5d | 1579 | e = HOP3c(strend, -((SSize_t)ln), s); |
73104a1b | 1580 | |
02d5137b | 1581 | if (reginfo->intuit && e < s) { |
73104a1b KW |
1582 | e = s; /* Due to minlen logic of intuit() */ |
1583 | } | |
fac1af77 | 1584 | |
73104a1b KW |
1585 | c1 = *pat_string; |
1586 | c2 = fold_array[c1]; | |
1587 | if (c1 == c2) { /* If char and fold are the same */ | |
1588 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); | |
1589 | } | |
1590 | else { | |
1591 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); | |
1592 | } | |
1593 | break; | |
fac1af77 | 1594 | |
73104a1b KW |
1595 | do_exactf_utf8: |
1596 | { | |
1597 | unsigned expansion; | |
1598 | ||
1599 | /* If one of the operands is in utf8, we can't use the simpler folding | |
1600 | * above, due to the fact that many different characters can have the | |
1601 | * same fold, or portion of a fold, or different- length fold */ | |
1602 | pat_string = STRING(c); | |
1603 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1604 | pat_end = pat_string + ln; | |
984e6dd1 | 1605 | lnc = is_utf8_pat /* length to match in characters */ |
73104a1b KW |
1606 | ? utf8_length((U8 *) pat_string, (U8 *) pat_end) |
1607 | : ln; | |
1608 | ||
1609 | /* We have 'lnc' characters to match in the pattern, but because of | |
1610 | * multi-character folding, each character in the target can match | |
1611 | * up to 3 characters (Unicode guarantees it will never exceed | |
1612 | * this) if it is utf8-encoded; and up to 2 if not (based on the | |
1613 | * fact that the Latin 1 folds are already determined, and the | |
1614 | * only multi-char fold in that range is the sharp-s folding to | |
1615 | * 'ss'. Thus, a pattern character can match as little as 1/3 of a | |
1616 | * string character. Adjust lnc accordingly, rounding up, so that | |
1617 | * if we need to match at least 4+1/3 chars, that really is 5. */ | |
1618 | expansion = (utf8_target) ? UTF8_MAX_FOLD_CHAR_EXPAND : 2; | |
1619 | lnc = (lnc + expansion - 1) / expansion; | |
1620 | ||
1621 | /* As in the non-UTF8 case, if we have to match 3 characters, and | |
1622 | * only 2 are left, it's guaranteed to fail, so don't start a | |
1623 | * match that would require us to go beyond the end of the string | |
1624 | */ | |
ea3daa5d | 1625 | e = HOP3c(strend, -((SSize_t)lnc), s); |
73104a1b | 1626 | |
02d5137b | 1627 | if (reginfo->intuit && e < s) { |
73104a1b KW |
1628 | e = s; /* Due to minlen logic of intuit() */ |
1629 | } | |
0658cdde | 1630 | |
73104a1b KW |
1631 | /* XXX Note that we could recalculate e to stop the loop earlier, |
1632 | * as the worst case expansion above will rarely be met, and as we | |
1633 | * go along we would usually find that e moves further to the left. | |
1634 | * This would happen only after we reached the point in the loop | |
1635 | * where if there were no expansion we should fail. Unclear if | |
1636 | * worth the expense */ | |
1637 | ||
1638 | while (s <= e) { | |
1639 | char *my_strend= (char *)strend; | |
1640 | if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target, | |
984e6dd1 | 1641 | pat_string, NULL, ln, is_utf8_pat, utf8_fold_flags) |
02d5137b | 1642 | && (reginfo->intuit || regtry(reginfo, &s)) ) |
73104a1b KW |
1643 | { |
1644 | goto got_it; | |
1645 | } | |
1646 | s += (utf8_target) ? UTF8SKIP(s) : 1; | |
1647 | } | |
1648 | break; | |
1649 | } | |
1650 | case BOUNDL: | |
272d35c9 | 1651 | RXp_MATCH_TAINTED_on(prog); |
0eb30aeb | 1652 | FBC_BOUND(isWORDCHAR_LC, |
f4cd282c | 1653 | isWORDCHAR_LC_uvchr(tmp), |
0eb30aeb | 1654 | isWORDCHAR_LC_utf8((U8*)s)); |
73104a1b KW |
1655 | break; |
1656 | case NBOUNDL: | |
272d35c9 | 1657 | RXp_MATCH_TAINTED_on(prog); |
0eb30aeb | 1658 | FBC_NBOUND(isWORDCHAR_LC, |
f4cd282c | 1659 | isWORDCHAR_LC_uvchr(tmp), |
0eb30aeb | 1660 | isWORDCHAR_LC_utf8((U8*)s)); |
73104a1b KW |
1661 | break; |
1662 | case BOUND: | |
1663 | FBC_BOUND(isWORDCHAR, | |
0eb30aeb | 1664 | isWORDCHAR_uni(tmp), |
03940dc2 | 1665 | cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target))); |
73104a1b KW |
1666 | break; |
1667 | case BOUNDA: | |
1668 | FBC_BOUND_NOLOAD(isWORDCHAR_A, | |
1669 | isWORDCHAR_A(tmp), | |
1670 | isWORDCHAR_A((U8*)s)); | |
1671 | break; | |
1672 | case NBOUND: | |
1673 | FBC_NBOUND(isWORDCHAR, | |
0eb30aeb | 1674 | isWORDCHAR_uni(tmp), |
03940dc2 | 1675 | cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target))); |
73104a1b KW |
1676 | break; |
1677 | case NBOUNDA: | |
1678 | FBC_NBOUND_NOLOAD(isWORDCHAR_A, | |
1679 | isWORDCHAR_A(tmp), | |
1680 | isWORDCHAR_A((U8*)s)); | |
1681 | break; | |
1682 | case BOUNDU: | |
1683 | FBC_BOUND(isWORDCHAR_L1, | |
0eb30aeb | 1684 | isWORDCHAR_uni(tmp), |
03940dc2 | 1685 | cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target))); |
73104a1b KW |
1686 | break; |
1687 | case NBOUNDU: | |
1688 | FBC_NBOUND(isWORDCHAR_L1, | |
0eb30aeb | 1689 | isWORDCHAR_uni(tmp), |
03940dc2 | 1690 | cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target))); |
73104a1b | 1691 | break; |
73104a1b KW |
1692 | case LNBREAK: |
1693 | REXEC_FBC_CSCAN(is_LNBREAK_utf8_safe(s, strend), | |
1694 | is_LNBREAK_latin1_safe(s, strend) | |
1695 | ); | |
1696 | break; | |
3018b823 KW |
1697 | |
1698 | /* The argument to all the POSIX node types is the class number to pass to | |
1699 | * _generic_isCC() to build a mask for searching in PL_charclass[] */ | |
1700 | ||
1701 | case NPOSIXL: | |
1702 | to_complement = 1; | |
1703 | /* FALLTHROUGH */ | |
1704 | ||
1705 | case POSIXL: | |
272d35c9 | 1706 | RXp_MATCH_TAINTED_on(prog); |
3018b823 KW |
1707 | REXEC_FBC_CSCAN(to_complement ^ cBOOL(isFOO_utf8_lc(FLAGS(c), (U8 *) s)), |
1708 | to_complement ^ cBOOL(isFOO_lc(FLAGS(c), *s))); | |
73104a1b | 1709 | break; |
3018b823 KW |
1710 | |
1711 | case NPOSIXD: | |
1712 | to_complement = 1; | |
1713 | /* FALLTHROUGH */ | |
1714 | ||
1715 | case POSIXD: | |
1716 | if (utf8_target) { | |
1717 | goto posix_utf8; | |
1718 | } | |
1719 | goto posixa; | |
1720 | ||
1721 | case NPOSIXA: | |
1722 | if (utf8_target) { | |
1723 | /* The complement of something that matches only ASCII matches all | |
1724 | * UTF-8 variant code points, plus everything in ASCII that isn't | |
1725 | * in the class */ | |
1726 | REXEC_FBC_UTF8_CLASS_SCAN(! UTF8_IS_INVARIANT(*s) | |
1727 | || ! _generic_isCC_A(*s, FLAGS(c))); | |
1728 | break; | |
1729 | } | |
1730 | ||
1731 | to_complement = 1; | |
1732 | /* FALLTHROUGH */ | |
1733 | ||
73104a1b | 1734 | case POSIXA: |
3018b823 | 1735 | posixa: |
73104a1b | 1736 | /* Don't need to worry about utf8, as it can match only a single |
3018b823 KW |
1737 | * byte invariant character. */ |
1738 | REXEC_FBC_CLASS_SCAN( | |
1739 | to_complement ^ cBOOL(_generic_isCC_A(*s, FLAGS(c)))); | |
73104a1b | 1740 | break; |
3018b823 KW |
1741 | |
1742 | case NPOSIXU: | |
1743 | to_complement = 1; | |
1744 | /* FALLTHROUGH */ | |
1745 | ||
1746 | case POSIXU: | |
1747 | if (! utf8_target) { | |
1748 | REXEC_FBC_CLASS_SCAN(to_complement ^ cBOOL(_generic_isCC(*s, | |
1749 | FLAGS(c)))); | |
1750 | } | |
1751 | else { | |
1752 | ||
1753 | posix_utf8: | |
1754 | classnum = (_char_class_number) FLAGS(c); | |
1755 | if (classnum < _FIRST_NON_SWASH_CC) { | |
1756 | while (s < strend) { | |
1757 | ||
1758 | /* We avoid loading in the swash as long as possible, but | |
1759 | * should we have to, we jump to a separate loop. This | |
1760 | * extra 'if' statement is what keeps this code from being | |
1761 | * just a call to REXEC_FBC_UTF8_CLASS_SCAN() */ | |
1762 | if (UTF8_IS_ABOVE_LATIN1(*s)) { | |
1763 | goto found_above_latin1; | |
1764 | } | |
1765 | if ((UTF8_IS_INVARIANT(*s) | |
1766 | && to_complement ^ cBOOL(_generic_isCC((U8) *s, | |
1767 | classnum))) | |
1768 | || (UTF8_IS_DOWNGRADEABLE_START(*s) | |
1769 | && to_complement ^ cBOOL( | |
94bb8c36 KW |
1770 | _generic_isCC(TWO_BYTE_UTF8_TO_NATIVE(*s, |
1771 | *(s + 1)), | |
3018b823 KW |
1772 | classnum)))) |
1773 | { | |
02d5137b | 1774 | if (tmp && (reginfo->intuit || regtry(reginfo, &s))) |
3018b823 KW |
1775 | goto got_it; |
1776 | else { | |
1777 | tmp = doevery; | |
1778 | } | |
1779 | } | |
1780 | else { | |
1781 | tmp = 1; | |
1782 | } | |
1783 | s += UTF8SKIP(s); | |
1784 | } | |
1785 | } | |
1786 | else switch (classnum) { /* These classes are implemented as | |
1787 | macros */ | |
1788 | case _CC_ENUM_SPACE: /* XXX would require separate code if we | |
1789 | revert the change of \v matching this */ | |
1790 | /* FALL THROUGH */ | |
1791 | ||
1792 | case _CC_ENUM_PSXSPC: | |
1793 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1794 | to_complement ^ cBOOL(isSPACE_utf8(s))); | |
1795 | break; | |
1796 | ||
1797 | case _CC_ENUM_BLANK: | |
1798 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1799 | to_complement ^ cBOOL(isBLANK_utf8(s))); | |
1800 | break; | |
1801 | ||
1802 | case _CC_ENUM_XDIGIT: | |
1803 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1804 | to_complement ^ cBOOL(isXDIGIT_utf8(s))); | |
1805 | break; | |
1806 | ||
1807 | case _CC_ENUM_VERTSPACE: | |
1808 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1809 | to_complement ^ cBOOL(isVERTWS_utf8(s))); | |
1810 | break; | |
1811 | ||
1812 | case _CC_ENUM_CNTRL: | |
1813 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1814 | to_complement ^ cBOOL(isCNTRL_utf8(s))); | |
1815 | break; | |
1816 | ||
1817 | default: | |
1818 | Perl_croak(aTHX_ "panic: find_byclass() node %d='%s' has an unexpected character class '%d'", OP(c), PL_reg_name[OP(c)], classnum); | |
1819 | assert(0); /* NOTREACHED */ | |
1820 | } | |
1821 | } | |
1822 | break; | |
1823 | ||
1824 | found_above_latin1: /* Here we have to load a swash to get the result | |
1825 | for the current code point */ | |
1826 | if (! PL_utf8_swash_ptrs[classnum]) { | |
1827 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; | |
1828 | PL_utf8_swash_ptrs[classnum] = | |
1829 | _core_swash_init("utf8", swash_property_names[classnum], | |
1830 | &PL_sv_undef, 1, 0, NULL, &flags); | |
1831 | } | |
1832 | ||
1833 | /* This is a copy of the loop above for swash classes, though using the | |
1834 | * FBC macro instead of being expanded out. Since we've loaded the | |
1835 | * swash, we don't have to check for that each time through the loop */ | |
1836 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1837 | to_complement ^ cBOOL(_generic_utf8( | |
1838 | classnum, | |
1839 | s, | |
1840 | swash_fetch(PL_utf8_swash_ptrs[classnum], | |
1841 | (U8 *) s, TRUE)))); | |
73104a1b KW |
1842 | break; |
1843 | ||
1844 | case AHOCORASICKC: | |
1845 | case AHOCORASICK: | |
1846 | { | |
1847 | DECL_TRIE_TYPE(c); | |
1848 | /* what trie are we using right now */ | |
1849 | reg_ac_data *aho = (reg_ac_data*)progi->data->data[ ARG( c ) ]; | |
1850 | reg_trie_data *trie = (reg_trie_data*)progi->data->data[ aho->trie ]; | |
1851 | HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); | |
1852 | ||
1853 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1854 | #ifdef DEBUGGING |
73104a1b | 1855 | const char *real_start = s; |
6148ee25 | 1856 | #endif |
73104a1b KW |
1857 | STRLEN maxlen = trie->maxlen; |
1858 | SV *sv_points; | |
1859 | U8 **points; /* map of where we were in the input string | |
1860 | when reading a given char. For ASCII this | |
1861 | is unnecessary overhead as the relationship | |
1862 | is always 1:1, but for Unicode, especially | |
1863 | case folded Unicode this is not true. */ | |
1864 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; | |
1865 | U8 *bitmap=NULL; | |
1866 | ||
1867 | ||
1868 | GET_RE_DEBUG_FLAGS_DECL; | |
1869 | ||
1870 | /* We can't just allocate points here. We need to wrap it in | |
1871 | * an SV so it gets freed properly if there is a croak while | |
1872 | * running the match */ | |
1873 | ENTER; | |
1874 | SAVETMPS; | |
1875 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1876 | SvCUR_set(sv_points, | |
1877 | maxlen * sizeof(U8 *)); | |
1878 | SvPOK_on(sv_points); | |
1879 | sv_2mortal(sv_points); | |
1880 | points=(U8**)SvPV_nolen(sv_points ); | |
1881 | if ( trie_type != trie_utf8_fold | |
1882 | && (trie->bitmap || OP(c)==AHOCORASICKC) ) | |
1883 | { | |
1884 | if (trie->bitmap) | |
1885 | bitmap=(U8*)trie->bitmap; | |
1886 | else | |
1887 | bitmap=(U8*)ANYOF_BITMAP(c); | |
1888 | } | |
1889 | /* this is the Aho-Corasick algorithm modified a touch | |
1890 | to include special handling for long "unknown char" sequences. | |
1891 | The basic idea being that we use AC as long as we are dealing | |
1892 | with a possible matching char, when we encounter an unknown char | |
1893 | (and we have not encountered an accepting state) we scan forward | |
1894 | until we find a legal starting char. | |
1895 | AC matching is basically that of trie matching, except that when | |
1896 | we encounter a failing transition, we fall back to the current | |
1897 | states "fail state", and try the current char again, a process | |
1898 | we repeat until we reach the root state, state 1, or a legal | |
1899 | transition. If we fail on the root state then we can either | |
1900 | terminate if we have reached an accepting state previously, or | |
1901 | restart the entire process from the beginning if we have not. | |
1902 | ||
1903 | */ | |
1904 | while (s <= last_start) { | |
1905 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1906 | U8 *uc = (U8*)s; | |
1907 | U16 charid = 0; | |
1908 | U32 base = 1; | |
1909 | U32 state = 1; | |
1910 | UV uvc = 0; | |
1911 | STRLEN len = 0; | |
1912 | STRLEN foldlen = 0; | |
1913 | U8 *uscan = (U8*)NULL; | |
1914 | U8 *leftmost = NULL; | |
1915 | #ifdef DEBUGGING | |
1916 | U32 accepted_word= 0; | |
786e8c11 | 1917 | #endif |
73104a1b KW |
1918 | U32 pointpos = 0; |
1919 | ||
1920 | while ( state && uc <= (U8*)strend ) { | |
1921 | int failed=0; | |
1922 | U32 word = aho->states[ state ].wordnum; | |
1923 | ||
1924 | if( state==1 ) { | |
1925 | if ( bitmap ) { | |
1926 | DEBUG_TRIE_EXECUTE_r( | |
1927 | if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1928 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
1929 | (char *)uc, utf8_target ); | |
1930 | PerlIO_printf( Perl_debug_log, | |
1931 | " Scanning for legal start char...\n"); | |
1932 | } | |
1933 | ); | |
1934 | if (utf8_target) { | |
1935 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1936 | uc += UTF8SKIP(uc); | |
1937 | } | |
1938 | } else { | |
1939 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1940 | uc++; | |
1941 | } | |
786e8c11 | 1942 | } |
73104a1b | 1943 | s= (char *)uc; |
07be1b83 | 1944 | } |
73104a1b KW |
1945 | if (uc >(U8*)last_start) break; |
1946 | } | |
1947 | ||
1948 | if ( word ) { | |
1949 | U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ]; | |
1950 | if (!leftmost || lpos < leftmost) { | |
1951 | DEBUG_r(accepted_word=word); | |
1952 | leftmost= lpos; | |
7016d6eb | 1953 | } |
73104a1b | 1954 | if (base==0) break; |
7016d6eb | 1955 | |
73104a1b KW |
1956 | } |
1957 | points[pointpos++ % maxlen]= uc; | |
1958 | if (foldlen || uc < (U8*)strend) { | |
1959 | REXEC_TRIE_READ_CHAR(trie_type, trie, | |
1960 | widecharmap, uc, | |
1961 | uscan, len, uvc, charid, foldlen, | |
1962 | foldbuf, uniflags); | |
1963 | DEBUG_TRIE_EXECUTE_r({ | |
1964 | dump_exec_pos( (char *)uc, c, strend, | |
1965 | real_start, s, utf8_target); | |
1966 | PerlIO_printf(Perl_debug_log, | |
1967 | " Charid:%3u CP:%4"UVxf" ", | |
1968 | charid, uvc); | |
1969 | }); | |
1970 | } | |
1971 | else { | |
1972 | len = 0; | |
1973 | charid = 0; | |
1974 | } | |
07be1b83 | 1975 | |
73104a1b KW |
1976 | |
1977 | do { | |
6148ee25 | 1978 | #ifdef DEBUGGING |
73104a1b | 1979 | word = aho->states[ state ].wordnum; |
6148ee25 | 1980 | #endif |
73104a1b KW |
1981 | base = aho->states[ state ].trans.base; |
1982 | ||
1983 | DEBUG_TRIE_EXECUTE_r({ | |
1984 | if (failed) | |
1985 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
1986 | s, utf8_target ); | |
1987 | PerlIO_printf( Perl_debug_log, | |
1988 | "%sState: %4"UVxf", word=%"UVxf, | |
1989 | failed ? " Fail transition to " : "", | |
1990 | (UV)state, (UV)word); | |
1991 | }); | |
1992 | if ( base ) { | |
1993 | U32 tmp; | |
1994 | I32 offset; | |
1995 | if (charid && | |
1996 | ( ((offset = base + charid | |
1997 | - 1 - trie->uniquecharcount)) >= 0) | |
1998 | && ((U32)offset < trie->lasttrans) | |
1999 | && trie->trans[offset].check == state | |
2000 | && (tmp=trie->trans[offset].next)) | |
2001 | { | |
2002 | DEBUG_TRIE_EXECUTE_r( | |
2003 | PerlIO_printf( Perl_debug_log," - legal\n")); | |
2004 | state = tmp; | |
2005 | break; | |
07be1b83 YO |
2006 | } |
2007 | else { | |
786e8c11 | 2008 | DEBUG_TRIE_EXECUTE_r( |
73104a1b | 2009 | PerlIO_printf( Perl_debug_log," - fail\n")); |
786e8c11 | 2010 | failed = 1; |
73104a1b | 2011 | state = aho->fail[state]; |
07be1b83 | 2012 | } |
07be1b83 | 2013 | } |
73104a1b KW |
2014 | else { |
2015 | /* we must be accepting here */ | |
2016 | DEBUG_TRIE_EXECUTE_r( | |
2017 | PerlIO_printf( Perl_debug_log," - accepting\n")); | |
2018 | failed = 1; | |
2019 | break; | |
786e8c11 | 2020 | } |
73104a1b KW |
2021 | } while(state); |
2022 | uc += len; | |
2023 | if (failed) { | |
2024 | if (leftmost) | |
2025 | break; | |
2026 | if (!state) state = 1; | |
07be1b83 | 2027 | } |
73104a1b KW |
2028 | } |
2029 | if ( aho->states[ state ].wordnum ) { | |
2030 | U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ]; | |
2031 | if (!leftmost || lpos < leftmost) { | |
2032 | DEBUG_r(accepted_word=aho->states[ state ].wordnum); | |
2033 | leftmost = lpos; | |
07be1b83 YO |
2034 | } |
2035 | } | |
73104a1b KW |
2036 | if (leftmost) { |
2037 | s = (char*)leftmost; | |
2038 | DEBUG_TRIE_EXECUTE_r({ | |
2039 | PerlIO_printf( | |
2040 | Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", | |
2041 | (UV)accepted_word, (IV)(s - real_start) | |
2042 | ); | |
2043 | }); | |
02d5137b | 2044 | if (reginfo->intuit || regtry(reginfo, &s)) { |
73104a1b KW |
2045 | FREETMPS; |
2046 | LEAVE; | |
2047 | goto got_it; | |
2048 | } | |
2049 | s = HOPc(s,1); | |
2050 | DEBUG_TRIE_EXECUTE_r({ | |
2051 | PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); | |
2052 | }); | |
2053 | } else { | |
2054 | DEBUG_TRIE_EXECUTE_r( | |
2055 | PerlIO_printf( Perl_debug_log,"No match.\n")); | |
2056 | break; | |
2057 | } | |
2058 | } | |
2059 | FREETMPS; | |
2060 | LEAVE; | |
2061 | } | |
2062 | break; | |
2063 | default: | |
2064 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); | |
2065 | break; | |
2066 | } | |
2067 | return 0; | |
2068 | got_it: | |
2069 | return s; | |
6eb5f6b9 JH |
2070 | } |
2071 | ||
60165aa4 DM |
2072 | /* set RX_SAVED_COPY, RX_SUBBEG etc. |
2073 | * flags have same meanings as with regexec_flags() */ | |
2074 | ||
749f4950 DM |
2075 | static void |
2076 | S_reg_set_capture_string(pTHX_ REGEXP * const rx, | |
60165aa4 DM |
2077 | char *strbeg, |
2078 | char *strend, | |
2079 | SV *sv, | |
2080 | U32 flags, | |
2081 | bool utf8_target) | |
2082 | { | |
2083 | struct regexp *const prog = ReANY(rx); | |
2084 | ||
60165aa4 DM |
2085 | if (flags & REXEC_COPY_STR) { |
2086 | #ifdef PERL_ANY_COW | |
2087 | if (SvCANCOW(sv)) { | |
2088 | if (DEBUG_C_TEST) { | |
2089 | PerlIO_printf(Perl_debug_log, | |
2090 | "Copy on write: regexp capture, type %d\n", | |
2091 | (int) SvTYPE(sv)); | |
2092 | } | |
5411a0e5 DM |
2093 | /* Create a new COW SV to share the match string and store |
2094 | * in saved_copy, unless the current COW SV in saved_copy | |
2095 | * is valid and suitable for our purpose */ | |
2096 | if (( prog->saved_copy | |
2097 | && SvIsCOW(prog->saved_copy) | |
2098 | && SvPOKp(prog->saved_copy) | |
2099 | && SvIsCOW(sv) | |
2100 | && SvPOKp(sv) | |
2101 | && SvPVX(sv) == SvPVX(prog->saved_copy))) | |
a76b0e90 | 2102 | { |
5411a0e5 DM |
2103 | /* just reuse saved_copy SV */ |
2104 | if (RXp_MATCH_COPIED(prog)) { | |
2105 | Safefree(prog->subbeg); | |
2106 | RXp_MATCH_COPIED_off(prog); | |
2107 | } | |
2108 | } | |
2109 | else { | |
2110 | /* create new COW SV to share string */ | |
a76b0e90 DM |
2111 | RX_MATCH_COPY_FREE(rx); |
2112 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); | |
a76b0e90 | 2113 | } |
5411a0e5 DM |
2114 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
2115 | assert (SvPOKp(prog->saved_copy)); | |
60165aa4 DM |
2116 | prog->sublen = strend - strbeg; |
2117 | prog->suboffset = 0; | |
2118 | prog->subcoffset = 0; | |
2119 | } else | |
2120 | #endif | |
2121 | { | |
99a90e59 FC |
2122 | SSize_t min = 0; |
2123 | SSize_t max = strend - strbeg; | |
ea3daa5d | 2124 | SSize_t sublen; |
60165aa4 DM |
2125 | |
2126 | if ( (flags & REXEC_COPY_SKIP_POST) | |
e322109a | 2127 | && !(prog->extflags & RXf_PMf_KEEPCOPY) /* //p */ |
60165aa4 DM |
2128 | && !(PL_sawampersand & SAWAMPERSAND_RIGHT) |
2129 | ) { /* don't copy $' part of string */ | |
2130 | U32 n = 0; | |
2131 | max = -1; | |
2132 | /* calculate the right-most part of the string covered | |
2133 | * by a capture. Due to look-ahead, this may be to | |
2134 | * the right of $&, so we have to scan all captures */ | |
2135 | while (n <= prog->lastparen) { | |
2136 | if (prog->offs[n].end > max) | |
2137 | max = prog->offs[n].end; | |
2138 | n++; | |
2139 | } | |
2140 | if (max == -1) | |
2141 | max = (PL_sawampersand & SAWAMPERSAND_LEFT) | |
2142 | ? prog->offs[0].start | |
2143 | : 0; | |
2144 | assert(max >= 0 && max <= strend - strbeg); | |
2145 | } | |
2146 | ||
2147 | if ( (flags & REXEC_COPY_SKIP_PRE) | |
e322109a | 2148 | && !(prog->extflags & RXf_PMf_KEEPCOPY) /* //p */ |
60165aa4 DM |
2149 | && !(PL_sawampersand & SAWAMPERSAND_LEFT) |
2150 | ) { /* don't copy $` part of string */ | |
2151 | U32 n = 0; | |
2152 | min = max; | |
2153 | /* calculate the left-most part of the string covered | |
2154 | * by a capture. Due to look-behind, this may be to | |
2155 | * the left of $&, so we have to scan all captures */ | |
2156 | while (min && n <= prog->lastparen) { | |
2157 | if ( prog->offs[n].start != -1 | |
2158 | && prog->offs[n].start < min) | |
2159 | { | |
2160 | min = prog->offs[n].start; | |
2161 | } | |
2162 | n++; | |
2163 | } | |
2164 | if ((PL_sawampersand & SAWAMPERSAND_RIGHT) | |
2165 | && min > prog->offs[0].end | |
2166 | ) | |
2167 | min = prog->offs[0].end; | |
2168 | ||
2169 | } | |
2170 | ||
2171 | assert(min >= 0 && min <= max && min <= strend - strbeg); | |
2172 | sublen = max - min; | |
2173 | ||
2174 | if (RX_MATCH_COPIED(rx)) { | |
2175 | if (sublen > prog->sublen) | |
2176 | prog->subbeg = | |
2177 | (char*)saferealloc(prog->subbeg, sublen+1); | |
2178 | } | |
2179 | else | |
2180 | prog->subbeg = (char*)safemalloc(sublen+1); | |
2181 | Copy(strbeg + min, prog->subbeg, sublen, char); | |
2182 | prog->subbeg[sublen] = '\0'; | |
2183 | prog->suboffset = min; | |
2184 | prog->sublen = sublen; | |
2185 | RX_MATCH_COPIED_on(rx); | |
2186 | } | |
2187 | prog->subcoffset = prog->suboffset; | |
2188 | if (prog->suboffset && utf8_target) { | |
2189 | /* Convert byte offset to chars. | |
2190 | * XXX ideally should only compute this if @-/@+ | |
2191 | * has been seen, a la PL_sawampersand ??? */ | |
2192 | ||
2193 | /* If there's a direct correspondence between the | |
2194 | * string which we're matching and the original SV, | |
2195 | * then we can use the utf8 len cache associated with | |
2196 | * the SV. In particular, it means that under //g, | |
2197 | * sv_pos_b2u() will use the previously cached | |
2198 | * position to speed up working out the new length of | |
2199 | * subcoffset, rather than counting from the start of | |
2200 | * the string each time. This stops | |
2201 | * $x = "\x{100}" x 1E6; 1 while $x =~ /(.)/g; | |
2202 | * from going quadratic */ | |
2203 | if (SvPOKp(sv) && SvPVX(sv) == strbeg) | |
ea3daa5d FC |
2204 | prog->subcoffset = sv_pos_b2u_flags(sv, prog->subcoffset, |
2205 | SV_GMAGIC|SV_CONST_RETURN); | |
60165aa4 DM |
2206 | else |
2207 | prog->subcoffset = utf8_length((U8*)strbeg, | |
2208 | (U8*)(strbeg+prog->suboffset)); | |
2209 | } | |
2210 | } | |
2211 | else { | |
2212 | RX_MATCH_COPY_FREE(rx); | |
2213 | prog->subbeg = strbeg; | |
2214 | prog->suboffset = 0; | |
2215 | prog->subcoffset = 0; | |
2216 | prog->sublen = strend - strbeg; | |
2217 | } | |
2218 | } | |
2219 | ||
2220 | ||
2221 | ||
fae667d5 | 2222 | |
6eb5f6b9 JH |
2223 | /* |
2224 | - regexec_flags - match a regexp against a string | |
2225 | */ | |
2226 | I32 | |
5aaab254 | 2227 | Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, char *strend, |
ea3daa5d | 2228 | char *strbeg, SSize_t minend, SV *sv, void *data, U32 flags) |
8fd1a950 DM |
2229 | /* stringarg: the point in the string at which to begin matching */ |
2230 | /* strend: pointer to null at end of string */ | |
2231 | /* strbeg: real beginning of string */ | |
2232 | /* minend: end of match must be >= minend bytes after stringarg. */ | |
2233 | /* sv: SV being matched: only used for utf8 flag, pos() etc; string | |
2234 | * itself is accessed via the pointers above */ | |
2235 | /* data: May be used for some additional optimizations. | |
d058ec57 | 2236 | Currently unused. */ |
a340edde | 2237 | /* flags: For optimizations. See REXEC_* in regexp.h */ |
8fd1a950 | 2238 | |
6eb5f6b9 | 2239 | { |
97aff369 | 2240 | dVAR; |
8d919b0a | 2241 | struct regexp *const prog = ReANY(rx); |
5aaab254 | 2242 | char *s; |
eb578fdb | 2243 | regnode *c; |
03c83e26 | 2244 | char *startpos; |
ea3daa5d FC |
2245 | SSize_t minlen; /* must match at least this many chars */ |
2246 | SSize_t dontbother = 0; /* how many characters not to try at end */ | |
f2ed9b32 | 2247 | const bool utf8_target = cBOOL(DO_UTF8(sv)); |
2757e526 | 2248 | I32 multiline; |
f8fc2ecf | 2249 | RXi_GET_DECL(prog,progi); |
02d5137b DM |
2250 | regmatch_info reginfo_buf; /* create some info to pass to regtry etc */ |
2251 | regmatch_info *const reginfo = ®info_buf; | |
e9105d30 | 2252 | regexp_paren_pair *swap = NULL; |
006f26b2 | 2253 | I32 oldsave; |
a3621e74 YO |
2254 | GET_RE_DEBUG_FLAGS_DECL; |
2255 | ||
7918f24d | 2256 | PERL_ARGS_ASSERT_REGEXEC_FLAGS; |
9d4ba2ae | 2257 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
2258 | |
2259 | /* Be paranoid... */ | |
03c83e26 | 2260 | if (prog == NULL || stringarg == NULL) { |
6eb5f6b9 JH |
2261 | Perl_croak(aTHX_ "NULL regexp parameter"); |
2262 | return 0; | |
2263 | } | |
2264 | ||
6c3fea77 | 2265 | DEBUG_EXECUTE_r( |
03c83e26 | 2266 | debug_start_match(rx, utf8_target, stringarg, strend, |
6c3fea77 DM |
2267 | "Matching"); |
2268 | ); | |
8adc0f72 | 2269 | |
b342a604 DM |
2270 | startpos = stringarg; |
2271 | ||
03c83e26 | 2272 | if (prog->extflags & RXf_GPOS_SEEN) { |
d307c076 DM |
2273 | MAGIC *mg; |
2274 | ||
fef7148b DM |
2275 | /* set reginfo->ganch, the position where \G can match */ |
2276 | ||
2277 | reginfo->ganch = | |
2278 | (flags & REXEC_IGNOREPOS) | |
2279 | ? stringarg /* use start pos rather than pos() */ | |
2280 | : (sv && (mg = mg_find_mglob(sv)) && mg->mg_len >= 0) | |
25fdce4a FC |
2281 | /* Defined pos(): */ |
2282 | ? strbeg + MgBYTEPOS(mg, sv, strbeg, strend-strbeg) | |
fef7148b DM |
2283 | : strbeg; /* pos() not defined; use start of string */ |
2284 | ||
2285 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, | |
7b0eb0b8 | 2286 | "GPOS ganch set to strbeg[%"IVdf"]\n", (IV)(reginfo->ganch - strbeg))); |
fef7148b | 2287 | |
03c83e26 DM |
2288 | /* in the presence of \G, we may need to start looking earlier in |
2289 | * the string than the suggested start point of stringarg: | |
0b2c2a84 | 2290 | * if prog->gofs is set, then that's a known, fixed minimum |
03c83e26 DM |
2291 | * offset, such as |
2292 | * /..\G/: gofs = 2 | |
2293 | * /ab|c\G/: gofs = 1 | |
2294 | * or if the minimum offset isn't known, then we have to go back | |
2295 | * to the start of the string, e.g. /w+\G/ | |
2296 | */ | |
2bfbe302 DM |
2297 | |
2298 | if (prog->extflags & RXf_ANCH_GPOS) { | |
2299 | startpos = reginfo->ganch - prog->gofs; | |
2300 | if (startpos < | |
2301 | ((flags & REXEC_FAIL_ON_UNDERFLOW) ? stringarg : strbeg)) | |
2302 | { | |
2303 | DEBUG_r(PerlIO_printf(Perl_debug_log, | |
2304 | "fail: ganch-gofs before earliest possible start\n")); | |
2305 | return 0; | |
2306 | } | |
2307 | } | |
2308 | else if (prog->gofs) { | |
b342a604 DM |
2309 | if (startpos - prog->gofs < strbeg) |
2310 | startpos = strbeg; | |
2311 | else | |
2312 | startpos -= prog->gofs; | |
03c83e26 | 2313 | } |
b342a604 DM |
2314 | else if (prog->extflags & RXf_GPOS_FLOAT) |
2315 | startpos = strbeg; | |
03c83e26 DM |
2316 | } |
2317 | ||
2318 | minlen = prog->minlen; | |
b342a604 | 2319 | if ((startpos + minlen) > strend || startpos < strbeg) { |
03c83e26 DM |
2320 | DEBUG_r(PerlIO_printf(Perl_debug_log, |
2321 | "Regex match can't succeed, so not even tried\n")); | |
2322 | return 0; | |
2323 | } | |
2324 | ||
63a3746a DM |
2325 | /* at the end of this function, we'll do a LEAVE_SCOPE(oldsave), |
2326 | * which will call destuctors to reset PL_regmatch_state, free higher | |
2327 | * PL_regmatch_slabs, and clean up regmatch_info_aux and | |
2328 | * regmatch_info_aux_eval */ | |
2329 | ||
2330 | oldsave = PL_savestack_ix; | |
2331 | ||
dfa77d06 DM |
2332 | s = startpos; |
2333 | ||
e322109a | 2334 | if ((prog->extflags & RXf_USE_INTUIT) |
7fadf4a7 DM |
2335 | && !(flags & REXEC_CHECKED)) |
2336 | { | |
dfa77d06 | 2337 | s = re_intuit_start(rx, sv, strbeg, startpos, strend, |
7fadf4a7 | 2338 | flags, NULL); |
dfa77d06 | 2339 | if (!s) |
7fadf4a7 DM |
2340 | return 0; |
2341 | ||
e322109a | 2342 | if (prog->extflags & RXf_CHECK_ALL) { |
7fadf4a7 DM |
2343 | /* we can match based purely on the result of INTUIT. |
2344 | * Set up captures etc just for $& and $-[0] | |
2345 | * (an intuit-only match wont have $1,$2,..) */ | |
2346 | assert(!prog->nparens); | |
d5e7783a DM |
2347 | |
2348 | /* s/// doesn't like it if $& is earlier than where we asked it to | |
2349 | * start searching (which can happen on something like /.\G/) */ | |
2350 | if ( (flags & REXEC_FAIL_ON_UNDERFLOW) | |
2351 | && (s < stringarg)) | |
2352 | { | |
2353 | /* this should only be possible under \G */ | |
2354 | assert(prog->extflags & RXf_GPOS_SEEN); | |
2355 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
2356 | "matched, but failing for REXEC_FAIL_ON_UNDERFLOW\n")); | |
2357 | goto phooey; | |
2358 | } | |
2359 | ||
7fadf4a7 DM |
2360 | /* match via INTUIT shouldn't have any captures. |
2361 | * Let @-, @+, $^N know */ | |
2362 | prog->lastparen = prog->lastcloseparen = 0; | |
2363 | RX_MATCH_UTF8_set(rx, utf8_target); | |
3ff69bd6 DM |
2364 | prog->offs[0].start = s - strbeg; |
2365 | prog->offs[0].end = utf8_target | |
2366 | ? (char*)utf8_hop((U8*)s, prog->minlenret) - strbeg | |
2367 | : s - strbeg + prog->minlenret; | |
7fadf4a7 | 2368 | if ( !(flags & REXEC_NOT_FIRST) ) |
749f4950 | 2369 | S_reg_set_capture_string(aTHX_ rx, |
7fadf4a7 DM |
2370 | strbeg, strend, |
2371 | sv, flags, utf8_target); | |
2372 | ||
7fadf4a7 DM |
2373 | return 1; |
2374 | } | |
2375 | } | |
2376 | ||
6c3fea77 | 2377 | multiline = prog->extflags & RXf_PMf_MULTILINE; |
1de06328 | 2378 | |
dfa77d06 | 2379 | if (strend - s < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { |
a3621e74 | 2380 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
2381 | "String too short [regexec_flags]...\n")); |
2382 | goto phooey; | |
1aa99e6b | 2383 | } |
1de06328 | 2384 | |
6eb5f6b9 | 2385 | /* Check validity of program. */ |
f8fc2ecf | 2386 | if (UCHARAT(progi->program) != REG_MAGIC) { |
6eb5f6b9 JH |
2387 | Perl_croak(aTHX_ "corrupted regexp program"); |
2388 | } | |
2389 | ||
272d35c9 | 2390 | RX_MATCH_TAINTED_off(rx); |
6eb5f6b9 | 2391 | |
6c3fea77 DM |
2392 | reginfo->prog = rx; /* Yes, sorry that this is confusing. */ |
2393 | reginfo->intuit = 0; | |
2394 | reginfo->is_utf8_target = cBOOL(utf8_target); | |
02d5137b DM |
2395 | reginfo->is_utf8_pat = cBOOL(RX_UTF8(rx)); |
2396 | reginfo->warned = FALSE; | |
9d9163fb | 2397 | reginfo->strbeg = strbeg; |
02d5137b | 2398 | reginfo->sv = sv; |
1cb48e53 | 2399 | reginfo->poscache_maxiter = 0; /* not yet started a countdown */ |
220db18a | 2400 | reginfo->strend = strend; |
6eb5f6b9 | 2401 | /* see how far we have to get to not match where we matched before */ |
fe3974be | 2402 | reginfo->till = stringarg + minend; |
6eb5f6b9 | 2403 | |
82c23608 FC |
2404 | if (prog->extflags & RXf_EVAL_SEEN && SvPADTMP(sv) && !IS_PADGV(sv)) { |
2405 | /* SAVEFREESV, not sv_mortalcopy, as this SV must last until after | |
2406 | S_cleanup_regmatch_info_aux has executed (registered by | |
2407 | SAVEDESTRUCTOR_X below). S_cleanup_regmatch_info_aux modifies | |
2408 | magic belonging to this SV. | |
2409 | Not newSVsv, either, as it does not COW. | |
2410 | */ | |
2411 | reginfo->sv = newSV(0); | |
2412 | sv_setsv(reginfo->sv, sv); | |
2413 | SAVEFREESV(reginfo->sv); | |
2414 | } | |
2415 | ||
331b2dcc DM |
2416 | /* reserve next 2 or 3 slots in PL_regmatch_state: |
2417 | * slot N+0: may currently be in use: skip it | |
2418 | * slot N+1: use for regmatch_info_aux struct | |
2419 | * slot N+2: use for regmatch_info_aux_eval struct if we have (?{})'s | |
2420 | * slot N+3: ready for use by regmatch() | |
2421 | */ | |
bf2039a9 | 2422 | |
331b2dcc DM |
2423 | { |
2424 | regmatch_state *old_regmatch_state; | |
2425 | regmatch_slab *old_regmatch_slab; | |
2426 | int i, max = (prog->extflags & RXf_EVAL_SEEN) ? 2 : 1; | |
2427 | ||
2428 | /* on first ever match, allocate first slab */ | |
2429 | if (!PL_regmatch_slab) { | |
2430 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
2431 | PL_regmatch_slab->prev = NULL; | |
2432 | PL_regmatch_slab->next = NULL; | |
2433 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); | |
2434 | } | |
bf2039a9 | 2435 | |
331b2dcc DM |
2436 | old_regmatch_state = PL_regmatch_state; |
2437 | old_regmatch_slab = PL_regmatch_slab; | |
bf2039a9 | 2438 | |
331b2dcc DM |
2439 | for (i=0; i <= max; i++) { |
2440 | if (i == 1) | |
2441 | reginfo->info_aux = &(PL_regmatch_state->u.info_aux); | |
2442 | else if (i ==2) | |
2443 | reginfo->info_aux_eval = | |
2444 | reginfo->info_aux->info_aux_eval = | |
2445 | &(PL_regmatch_state->u.info_aux_eval); | |
bf2039a9 | 2446 | |
331b2dcc DM |
2447 | if (++PL_regmatch_state > SLAB_LAST(PL_regmatch_slab)) |
2448 | PL_regmatch_state = S_push_slab(aTHX); | |
2449 | } | |
bf2039a9 | 2450 | |
331b2dcc DM |
2451 | /* note initial PL_regmatch_state position; at end of match we'll |
2452 | * pop back to there and free any higher slabs */ | |
bf2039a9 | 2453 | |
331b2dcc DM |
2454 | reginfo->info_aux->old_regmatch_state = old_regmatch_state; |
2455 | reginfo->info_aux->old_regmatch_slab = old_regmatch_slab; | |
2ac8ff4b | 2456 | reginfo->info_aux->poscache = NULL; |
bf2039a9 | 2457 | |
331b2dcc | 2458 | SAVEDESTRUCTOR_X(S_cleanup_regmatch_info_aux, reginfo->info_aux); |
bf2039a9 | 2459 | |
331b2dcc DM |
2460 | if ((prog->extflags & RXf_EVAL_SEEN)) |
2461 | S_setup_eval_state(aTHX_ reginfo); | |
2462 | else | |
2463 | reginfo->info_aux_eval = reginfo->info_aux->info_aux_eval = NULL; | |
bf2039a9 | 2464 | } |
d3aa529c | 2465 | |
6eb5f6b9 | 2466 | /* If there is a "must appear" string, look for it. */ |
6eb5f6b9 | 2467 | |
288b8c02 | 2468 | if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { |
e9105d30 GG |
2469 | /* We have to be careful. If the previous successful match |
2470 | was from this regex we don't want a subsequent partially | |
2471 | successful match to clobber the old results. | |
2472 | So when we detect this possibility we add a swap buffer | |
d8da0584 KW |
2473 | to the re, and switch the buffer each match. If we fail, |
2474 | we switch it back; otherwise we leave it swapped. | |
e9105d30 GG |
2475 | */ |
2476 | swap = prog->offs; | |
2477 | /* do we need a save destructor here for eval dies? */ | |
2478 | Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); | |
495f47a5 DM |
2479 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
2480 | "rex=0x%"UVxf" saving offs: orig=0x%"UVxf" new=0x%"UVxf"\n", | |
2481 | PTR2UV(prog), | |
2482 | PTR2UV(swap), | |
2483 | PTR2UV(prog->offs) | |
2484 | )); | |
c74340f9 | 2485 | } |
6eb5f6b9 JH |
2486 | |
2487 | /* Simplest case: anchored match need be tried only once. */ | |
2488 | /* [unless only anchor is BOL and multiline is set] */ | |
bbe252da | 2489 | if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { |
3542935d | 2490 | if (s == startpos && regtry(reginfo, &s)) |
6eb5f6b9 | 2491 | goto got_it; |
bbe252da YO |
2492 | else if (multiline || (prog->intflags & PREGf_IMPLICIT) |
2493 | || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ | |
6eb5f6b9 JH |
2494 | { |
2495 | char *end; | |
2496 | ||
2497 | if (minlen) | |
2498 | dontbother = minlen - 1; | |
1aa99e6b | 2499 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 2500 | /* for multiline we only have to try after newlines */ |
33b8afdf | 2501 | if (prog->check_substr || prog->check_utf8) { |
92f3d482 YO |
2502 | /* because of the goto we can not easily reuse the macros for bifurcating the |
2503 | unicode/non-unicode match modes here like we do elsewhere - demerphq */ | |
2504 | if (utf8_target) { | |
2505 | if (s == startpos) | |
2506 | goto after_try_utf8; | |
2507 | while (1) { | |
02d5137b | 2508 | if (regtry(reginfo, &s)) { |
92f3d482 YO |
2509 | goto got_it; |
2510 | } | |
2511 | after_try_utf8: | |
2512 | if (s > end) { | |
2513 | goto phooey; | |
2514 | } | |
2515 | if (prog->extflags & RXf_USE_INTUIT) { | |
52a21eb3 DM |
2516 | s = re_intuit_start(rx, sv, strbeg, |
2517 | s + UTF8SKIP(s), strend, flags, NULL); | |
92f3d482 YO |
2518 | if (!s) { |
2519 | goto phooey; | |
2520 | } | |
2521 | } | |
2522 | else { | |
2523 | s += UTF8SKIP(s); | |
2524 | } | |
2525 | } | |
2526 | } /* end search for check string in unicode */ | |
2527 | else { | |
2528 | if (s == startpos) { | |
2529 | goto after_try_latin; | |
2530 | } | |
2531 | while (1) { | |
02d5137b | 2532 | if (regtry(reginfo, &s)) { |
92f3d482 YO |
2533 | goto got_it; |
2534 | } | |
2535 | after_try_latin: | |
2536 | if (s > end) { | |
2537 | goto phooey; | |
2538 | } | |
2539 | if (prog->extflags & RXf_USE_INTUIT) { | |
52a21eb3 DM |
2540 | s = re_intuit_start(rx, sv, strbeg, |
2541 | s + 1, strend, flags, NULL); | |
92f3d482 YO |
2542 | if (!s) { |
2543 | goto phooey; | |
2544 | } | |
2545 | } | |
2546 | else { | |
2547 | s++; | |
2548 | } | |
2549 | } | |
2550 | } /* end search for check string in latin*/ | |
2551 | } /* end search for check string */ | |
2552 | else { /* search for newline */ | |
2553 | if (s > startpos) { | |
2554 | /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/ | |
6eb5f6b9 | 2555 | s--; |
92f3d482 | 2556 | } |
21eede78 YO |
2557 | /* We can use a more efficient search as newlines are the same in unicode as they are in latin */ |
2558 | while (s <= end) { /* note it could be possible to match at the end of the string */ | |
6eb5f6b9 | 2559 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ |
02d5137b | 2560 | if (regtry(reginfo, &s)) |
6eb5f6b9 JH |
2561 | goto got_it; |
2562 | } | |
92f3d482 YO |
2563 | } |
2564 | } /* end search for newline */ | |
2565 | } /* end anchored/multiline check string search */ | |
6eb5f6b9 | 2566 | goto phooey; |
bbe252da | 2567 | } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) |
f9f4320a | 2568 | { |
2bfbe302 DM |
2569 | /* For anchored \G, the only position it can match from is |
2570 | * (ganch-gofs); we already set startpos to this above; if intuit | |
2571 | * moved us on from there, we can't possibly succeed */ | |
2572 | assert(startpos == reginfo->ganch - prog->gofs); | |
2573 | if (s == startpos && regtry(reginfo, &s)) | |
6eb5f6b9 JH |
2574 | goto got_it; |
2575 | goto phooey; | |
2576 | } | |
2577 | ||
2578 | /* Messy cases: unanchored match. */ | |
bbe252da | 2579 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { |
6eb5f6b9 | 2580 | /* we have /x+whatever/ */ |
984e6dd1 | 2581 | /* it must be a one character string (XXXX Except is_utf8_pat?) */ |
33b8afdf | 2582 | char ch; |
bf93d4cc GS |
2583 | #ifdef DEBUGGING |
2584 | int did_match = 0; | |
2585 | #endif | |
f2ed9b32 | 2586 | if (utf8_target) { |
7e0d5ad7 KW |
2587 | if (! prog->anchored_utf8) { |
2588 | to_utf8_substr(prog); | |
2589 | } | |
2590 | ch = SvPVX_const(prog->anchored_utf8)[0]; | |
4cadc6a9 | 2591 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2592 | if (*s == ch) { |
a3621e74 | 2593 | DEBUG_EXECUTE_r( did_match = 1 ); |
02d5137b | 2594 | if (regtry(reginfo, &s)) goto got_it; |
6eb5f6b9 JH |
2595 | s += UTF8SKIP(s); |
2596 | while (s < strend && *s == ch) | |
2597 | s += UTF8SKIP(s); | |
2598 | } | |
4cadc6a9 | 2599 | ); |
7e0d5ad7 | 2600 | |
6eb5f6b9 JH |
2601 | } |
2602 | else { | |
7e0d5ad7 KW |
2603 | if (! prog->anchored_substr) { |
2604 | if (! to_byte_substr(prog)) { | |
6b54ddc5 | 2605 | NON_UTF8_TARGET_BUT_UTF8_REQUIRED(phooey); |
7e0d5ad7 KW |
2606 | } |
2607 | } | |
2608 | ch = SvPVX_const(prog->anchored_substr)[0]; | |
4cadc6a9 | 2609 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2610 | if (*s == ch) { |
a3621e74 | 2611 | DEBUG_EXECUTE_r( did_match = 1 ); |
02d5137b | 2612 | if (regtry(reginfo, &s)) goto got_it; |
6eb5f6b9 JH |
2613 | s++; |
2614 | while (s < strend && *s == ch) | |
2615 | s++; | |
2616 | } | |
4cadc6a9 | 2617 | ); |
6eb5f6b9 | 2618 | } |
a3621e74 | 2619 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 2620 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
2621 | "Did not find anchored character...\n") |
2622 | ); | |
6eb5f6b9 | 2623 | } |
a0714e2c SS |
2624 | else if (prog->anchored_substr != NULL |
2625 | || prog->anchored_utf8 != NULL | |
2626 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
2627 | && prog->float_max_offset < strend - s)) { |
2628 | SV *must; | |
ea3daa5d FC |
2629 | SSize_t back_max; |
2630 | SSize_t back_min; | |
33b8afdf | 2631 | char *last; |
6eb5f6b9 | 2632 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
2633 | #ifdef DEBUGGING |
2634 | int did_match = 0; | |
2635 | #endif | |
33b8afdf | 2636 | if (prog->anchored_substr || prog->anchored_utf8) { |
7e0d5ad7 KW |
2637 | if (utf8_target) { |
2638 | if (! prog->anchored_utf8) { | |
2639 | to_utf8_substr(prog); | |
2640 | } | |
2641 | must = prog->anchored_utf8; | |
2642 | } | |
2643 | else { | |
2644 | if (! prog->anchored_substr) { | |
2645 | if (! to_byte_substr(prog)) { | |
6b54ddc5 | 2646 | NON_UTF8_TARGET_BUT_UTF8_REQUIRED(phooey); |
7e0d5ad7 KW |
2647 | } |
2648 | } | |
2649 | must = prog->anchored_substr; | |
2650 | } | |
33b8afdf JH |
2651 | back_max = back_min = prog->anchored_offset; |
2652 | } else { | |
7e0d5ad7 KW |
2653 | if (utf8_target) { |
2654 | if (! prog->float_utf8) { | |
2655 | to_utf8_substr(prog); | |
2656 | } | |
2657 | must = prog->float_utf8; | |
2658 | } | |
2659 | else { | |
2660 | if (! prog->float_substr) { | |
2661 | if (! to_byte_substr(prog)) { | |
6b54ddc5 | 2662 | NON_UTF8_TARGET_BUT_UTF8_REQUIRED(phooey); |
7e0d5ad7 KW |
2663 | } |
2664 | } | |
2665 | must = prog->float_substr; | |
2666 | } | |
33b8afdf JH |
2667 | back_max = prog->float_max_offset; |
2668 | back_min = prog->float_min_offset; | |
2669 | } | |
1de06328 | 2670 | |
1de06328 YO |
2671 | if (back_min<0) { |
2672 | last = strend; | |
2673 | } else { | |
2674 | last = HOP3c(strend, /* Cannot start after this */ | |
ea3daa5d | 2675 | -(SSize_t)(CHR_SVLEN(must) |
1de06328 YO |
2676 | - (SvTAIL(must) != 0) + back_min), strbeg); |
2677 | } | |
9d9163fb | 2678 | if (s > reginfo->strbeg) |
6eb5f6b9 JH |
2679 | last1 = HOPc(s, -1); |
2680 | else | |
2681 | last1 = s - 1; /* bogus */ | |
2682 | ||
a0288114 | 2683 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 | 2684 | check_substr==must. */ |
bf05793b | 2685 | dontbother = 0; |
6eb5f6b9 JH |
2686 | strend = HOPc(strend, -dontbother); |
2687 | while ( (s <= last) && | |
c33e64f0 | 2688 | (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), |
9041c2e3 | 2689 | (unsigned char*)strend, must, |
c33e64f0 | 2690 | multiline ? FBMrf_MULTILINE : 0)) ) { |
a3621e74 | 2691 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
2692 | if (HOPc(s, -back_max) > last1) { |
2693 | last1 = HOPc(s, -back_min); | |
2694 | s = HOPc(s, -back_max); | |
2695 | } | |
2696 | else { | |
9d9163fb DM |
2697 | char * const t = (last1 >= reginfo->strbeg) |
2698 | ? HOPc(last1, 1) : last1 + 1; | |
6eb5f6b9 JH |
2699 | |
2700 | last1 = HOPc(s, -back_min); | |
52657f30 | 2701 | s = t; |
6eb5f6b9 | 2702 | } |
f2ed9b32 | 2703 | if (utf8_target) { |
6eb5f6b9 | 2704 | while (s <= last1) { |
02d5137b | 2705 | if (regtry(reginfo, &s)) |
6eb5f6b9 | 2706 | goto got_it; |
7016d6eb DM |
2707 | if (s >= last1) { |
2708 | s++; /* to break out of outer loop */ | |
2709 | break; | |
2710 | } | |
2711 | s += UTF8SKIP(s); | |
6eb5f6b9 JH |
2712 | } |
2713 | } | |
2714 | else { | |
2715 | while (s <= last1) { | |
02d5137b | 2716 | if (regtry(reginfo, &s)) |
6eb5f6b9 JH |
2717 | goto got_it; |
2718 | s++; | |
2719 | } | |
2720 | } | |
2721 | } | |
ab3bbdeb | 2722 | DEBUG_EXECUTE_r(if (!did_match) { |
f2ed9b32 | 2723 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
2724 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
2725 | PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", | |
33b8afdf | 2726 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc | 2727 | ? "anchored" : "floating"), |
ab3bbdeb YO |
2728 | quoted, RE_SV_TAIL(must)); |
2729 | }); | |
6eb5f6b9 JH |
2730 | goto phooey; |
2731 | } | |
f8fc2ecf | 2732 | else if ( (c = progi->regstclass) ) { |
f14c76ed | 2733 | if (minlen) { |
f8fc2ecf | 2734 | const OPCODE op = OP(progi->regstclass); |
66e933ab | 2735 | /* don't bother with what can't match */ |
786e8c11 | 2736 | if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) |
f14c76ed RGS |
2737 | strend = HOPc(strend, -(minlen - 1)); |
2738 | } | |
a3621e74 | 2739 | DEBUG_EXECUTE_r({ |
be8e71aa | 2740 | SV * const prop = sv_newmortal(); |
32fc9b6a | 2741 | regprop(prog, prop, c); |
0df25f3d | 2742 | { |
f2ed9b32 | 2743 | RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb | 2744 | s,strend-s,60); |
0df25f3d | 2745 | PerlIO_printf(Perl_debug_log, |
1c8f8eb1 | 2746 | "Matching stclass %.*s against %s (%d bytes)\n", |
e4f74956 | 2747 | (int)SvCUR(prop), SvPVX_const(prop), |
ab3bbdeb | 2748 | quoted, (int)(strend - s)); |
0df25f3d | 2749 | } |
ffc61ed2 | 2750 | }); |
f9176b44 | 2751 | if (find_byclass(prog, c, s, strend, reginfo)) |
6eb5f6b9 | 2752 | goto got_it; |
07be1b83 | 2753 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
2754 | } |
2755 | else { | |
2756 | dontbother = 0; | |
a0714e2c | 2757 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 2758 | /* Trim the end. */ |
6af40bd7 | 2759 | char *last= NULL; |
33b8afdf | 2760 | SV* float_real; |
c33e64f0 FC |
2761 | STRLEN len; |
2762 | const char *little; | |
33b8afdf | 2763 | |
7e0d5ad7 KW |
2764 | if (utf8_target) { |
2765 | if (! prog->float_utf8) { | |
2766 | to_utf8_substr(prog); | |
2767 | } | |
2768 | float_real = prog->float_utf8; | |
2769 | } | |
2770 | else { | |
2771 | if (! prog->float_substr) { | |
2772 | if (! to_byte_substr(prog)) { | |
6b54ddc5 | 2773 | NON_UTF8_TARGET_BUT_UTF8_REQUIRED(phooey); |
7e0d5ad7 KW |
2774 | } |
2775 | } | |
2776 | float_real = prog->float_substr; | |
2777 | } | |
d6a28714 | 2778 | |
c33e64f0 FC |
2779 | little = SvPV_const(float_real, len); |
2780 | if (SvTAIL(float_real)) { | |
7f18ad16 KW |
2781 | /* This means that float_real contains an artificial \n on |
2782 | * the end due to the presence of something like this: | |
2783 | * /foo$/ where we can match both "foo" and "foo\n" at the | |
2784 | * end of the string. So we have to compare the end of the | |
2785 | * string first against the float_real without the \n and | |
2786 | * then against the full float_real with the string. We | |
2787 | * have to watch out for cases where the string might be | |
2788 | * smaller than the float_real or the float_real without | |
2789 | * the \n. */ | |
1a13b075 YO |
2790 | char *checkpos= strend - len; |
2791 | DEBUG_OPTIMISE_r( | |
2792 | PerlIO_printf(Perl_debug_log, | |
2793 | "%sChecking for float_real.%s\n", | |
2794 | PL_colors[4], PL_colors[5])); | |
2795 | if (checkpos + 1 < strbeg) { | |
7f18ad16 KW |
2796 | /* can't match, even if we remove the trailing \n |
2797 | * string is too short to match */ | |
1a13b075 YO |
2798 | DEBUG_EXECUTE_r( |
2799 | PerlIO_printf(Perl_debug_log, | |
2800 | "%sString shorter than required trailing substring, cannot match.%s\n", | |
2801 | PL_colors[4], PL_colors[5])); | |
2802 | goto phooey; | |
2803 | } else if (memEQ(checkpos + 1, little, len - 1)) { | |
7f18ad16 KW |
2804 | /* can match, the end of the string matches without the |
2805 | * "\n" */ | |
1a13b075 YO |
2806 | last = checkpos + 1; |
2807 | } else if (checkpos < strbeg) { | |
7f18ad16 KW |
2808 | /* cant match, string is too short when the "\n" is |
2809 | * included */ | |
1a13b075 YO |
2810 | DEBUG_EXECUTE_r( |
2811 | PerlIO_printf(Perl_debug_log, | |
2812 | "%sString does not contain required trailing substring, cannot match.%s\n", | |
2813 | PL_colors[4], PL_colors[5])); | |
2814 | goto phooey; | |
2815 | } else if (!multiline) { | |
7f18ad16 KW |
2816 | /* non multiline match, so compare with the "\n" at the |
2817 | * end of the string */ | |
1a13b075 YO |
2818 | if (memEQ(checkpos, little, len)) { |
2819 | last= checkpos; | |
2820 | } else { | |
2821 | DEBUG_EXECUTE_r( | |
2822 | PerlIO_printf(Perl_debug_log, | |
2823 | "%sString does not contain required trailing substring, cannot match.%s\n", | |
2824 | PL_colors[4], PL_colors[5])); | |
2825 | goto phooey; | |
2826 | } | |
2827 | } else { | |
7f18ad16 KW |
2828 | /* multiline match, so we have to search for a place |
2829 | * where the full string is located */ | |
d6a28714 | 2830 | goto find_last; |
1a13b075 | 2831 | } |
c33e64f0 | 2832 | } else { |
d6a28714 | 2833 | find_last: |
9041c2e3 | 2834 | if (len) |
d6a28714 | 2835 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 2836 | else |
a0288114 | 2837 | last = strend; /* matching "$" */ |
b8c5462f | 2838 | } |
6af40bd7 | 2839 | if (!last) { |
7f18ad16 KW |
2840 | /* at one point this block contained a comment which was |
2841 | * probably incorrect, which said that this was a "should not | |
2842 | * happen" case. Even if it was true when it was written I am | |
2843 | * pretty sure it is not anymore, so I have removed the comment | |
2844 | * and replaced it with this one. Yves */ | |
6bda09f9 YO |
2845 | DEBUG_EXECUTE_r( |
2846 | PerlIO_printf(Perl_debug_log, | |
6af40bd7 YO |
2847 | "String does not contain required substring, cannot match.\n" |
2848 | )); | |
2849 | goto phooey; | |
bf93d4cc | 2850 | } |
d6a28714 JH |
2851 | dontbother = strend - last + prog->float_min_offset; |
2852 | } | |
2853 | if (minlen && (dontbother < minlen)) | |
2854 | dontbother = minlen - 1; | |
2855 | strend -= dontbother; /* this one's always in bytes! */ | |
2856 | /* We don't know much -- general case. */ | |
f2ed9b32 | 2857 | if (utf8_target) { |
d6a28714 | 2858 | for (;;) { |
02d5137b | 2859 | if (regtry(reginfo, &s)) |
d6a28714 JH |
2860 | goto got_it; |
2861 | if (s >= strend) | |
2862 | break; | |
b8c5462f | 2863 | s += UTF8SKIP(s); |
d6a28714 JH |
2864 | }; |
2865 | } | |
2866 | else { | |
2867 | do { | |
02d5137b | 2868 | if (regtry(reginfo, &s)) |
d6a28714 JH |
2869 | goto got_it; |
2870 | } while (s++ < strend); | |
2871 | } | |
2872 | } | |
2873 | ||
2874 | /* Failure. */ | |
2875 | goto phooey; | |
2876 | ||
2877 | got_it: | |
d5e7783a DM |
2878 | /* s/// doesn't like it if $& is earlier than where we asked it to |
2879 | * start searching (which can happen on something like /.\G/) */ | |
2880 | if ( (flags & REXEC_FAIL_ON_UNDERFLOW) | |
2881 | && (prog->offs[0].start < stringarg - strbeg)) | |
2882 | { | |
2883 | /* this should only be possible under \G */ | |
2884 | assert(prog->extflags & RXf_GPOS_SEEN); | |
2885 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
2886 | "matched, but failing for REXEC_FAIL_ON_UNDERFLOW\n")); | |
2887 | goto phooey; | |
2888 | } | |
2889 | ||
495f47a5 DM |
2890 | DEBUG_BUFFERS_r( |
2891 | if (swap) | |
2892 | PerlIO_printf(Perl_debug_log, | |
2893 | "rex=0x%"UVxf" freeing offs: 0x%"UVxf"\n", | |
2894 | PTR2UV(prog), | |
2895 | PTR2UV(swap) | |
2896 | ); | |
2897 | ); | |
e9105d30 | 2898 | Safefree(swap); |
d6a28714 | 2899 | |
bf2039a9 DM |
2900 | /* clean up; this will trigger destructors that will free all slabs |
2901 | * above the current one, and cleanup the regmatch_info_aux | |
2902 | * and regmatch_info_aux_eval sructs */ | |
8adc0f72 | 2903 | |
006f26b2 DM |
2904 | LEAVE_SCOPE(oldsave); |
2905 | ||
5daac39c NC |
2906 | if (RXp_PAREN_NAMES(prog)) |
2907 | (void)hv_iterinit(RXp_PAREN_NAMES(prog)); | |
d6a28714 | 2908 | |
0254aed9 DM |
2909 | RX_MATCH_UTF8_set(rx, utf8_target); |
2910 | ||
d6a28714 | 2911 | /* make sure $`, $&, $', and $digit will work later */ |
60165aa4 | 2912 | if ( !(flags & REXEC_NOT_FIRST) ) |
749f4950 | 2913 | S_reg_set_capture_string(aTHX_ rx, |
60165aa4 DM |
2914 | strbeg, reginfo->strend, |
2915 | sv, flags, utf8_target); | |
9041c2e3 | 2916 | |
d6a28714 JH |
2917 | return 1; |
2918 | ||
2919 | phooey: | |
a3621e74 | 2920 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 2921 | PL_colors[4], PL_colors[5])); |
8adc0f72 | 2922 | |
bf2039a9 DM |
2923 | /* clean up; this will trigger destructors that will free all slabs |
2924 | * above the current one, and cleanup the regmatch_info_aux | |
2925 | * and regmatch_info_aux_eval sructs */ | |
8adc0f72 | 2926 | |
006f26b2 DM |
2927 | LEAVE_SCOPE(oldsave); |
2928 | ||
e9105d30 | 2929 | if (swap) { |
c74340f9 | 2930 | /* we failed :-( roll it back */ |
495f47a5 DM |
2931 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
2932 | "rex=0x%"UVxf" rolling back offs: freeing=0x%"UVxf" restoring=0x%"UVxf"\n", | |
2933 | PTR2UV(prog), | |
2934 | PTR2UV(prog->offs), | |
2935 | PTR2UV(swap) | |
2936 | )); | |
e9105d30 GG |
2937 | Safefree(prog->offs); |
2938 | prog->offs = swap; | |
2939 | } | |
d6a28714 JH |
2940 | return 0; |
2941 | } | |
2942 | ||
6bda09f9 | 2943 | |
b3d298be | 2944 | /* Set which rex is pointed to by PL_reg_curpm, handling ref counting. |
ec43f78b DM |
2945 | * Do inc before dec, in case old and new rex are the same */ |
2946 | #define SET_reg_curpm(Re2) \ | |
bf2039a9 | 2947 | if (reginfo->info_aux_eval) { \ |
ec43f78b DM |
2948 | (void)ReREFCNT_inc(Re2); \ |
2949 | ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); \ | |
2950 | PM_SETRE((PL_reg_curpm), (Re2)); \ | |
2951 | } | |
2952 | ||
2953 | ||
d6a28714 JH |
2954 | /* |
2955 | - regtry - try match at specific point | |
2956 | */ | |
2957 | STATIC I32 /* 0 failure, 1 success */ | |
f73aaa43 | 2958 | S_regtry(pTHX_ regmatch_info *reginfo, char **startposp) |
d6a28714 | 2959 | { |
97aff369 | 2960 | dVAR; |
d6a28714 | 2961 | CHECKPOINT lastcp; |
288b8c02 | 2962 | REGEXP *const rx = reginfo->prog; |
8d919b0a | 2963 | regexp *const prog = ReANY(rx); |
99a90e59 | 2964 | SSize_t result; |
f8fc2ecf | 2965 | RXi_GET_DECL(prog,progi); |
a3621e74 | 2966 | GET_RE_DEBUG_FLAGS_DECL; |
7918f24d NC |
2967 | |
2968 | PERL_ARGS_ASSERT_REGTRY; | |
2969 | ||
24b23f37 | 2970 | reginfo->cutpoint=NULL; |
d6a28714 | 2971 | |
9d9163fb | 2972 | prog->offs[0].start = *startposp - reginfo->strbeg; |
d6a28714 | 2973 | prog->lastparen = 0; |
03994de8 | 2974 | prog->lastcloseparen = 0; |
d6a28714 JH |
2975 | |
2976 | /* XXXX What this code is doing here?!!! There should be no need | |
b93070ed | 2977 | to do this again and again, prog->lastparen should take care of |
3dd2943c | 2978 | this! --ilya*/ |
dafc8851 JH |
2979 | |
2980 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2981 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
b93070ed | 2982 | * prog->lastparen), is not needed at all by the test suite |
225593e1 DM |
2983 | * (op/regexp, op/pat, op/split), but that code is needed otherwise |
2984 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ | |
2985 | * Meanwhile, this code *is* needed for the | |
daf18116 JH |
2986 | * above-mentioned test suite tests to succeed. The common theme |
2987 | * on those tests seems to be returning null fields from matches. | |
225593e1 | 2988 | * --jhi updated by dapm */ |
dafc8851 | 2989 | #if 1 |
d6a28714 | 2990 | if (prog->nparens) { |
b93070ed | 2991 | regexp_paren_pair *pp = prog->offs; |
eb578fdb | 2992 | I32 i; |
b93070ed | 2993 | for (i = prog->nparens; i > (I32)prog->lastparen; i--) { |
f0ab9afb NC |
2994 | ++pp; |
2995 | pp->start = -1; | |
2996 | pp->end = -1; | |
d6a28714 JH |
2997 | } |
2998 | } | |
dafc8851 | 2999 | #endif |
02db2b7b | 3000 | REGCP_SET(lastcp); |
f73aaa43 DM |
3001 | result = regmatch(reginfo, *startposp, progi->program + 1); |
3002 | if (result != -1) { | |
3003 | prog->offs[0].end = result; | |
d6a28714 JH |
3004 | return 1; |
3005 | } | |
24b23f37 | 3006 | if (reginfo->cutpoint) |
f73aaa43 | 3007 | *startposp= reginfo->cutpoint; |
02db2b7b | 3008 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
3009 | return 0; |
3010 | } | |
3011 | ||
02db2b7b | 3012 | |
8ba1375e MJD |
3013 | #define sayYES goto yes |
3014 | #define sayNO goto no | |
262b90c4 | 3015 | #define sayNO_SILENT goto no_silent |
8ba1375e | 3016 | |
f9f4320a YO |
3017 | /* we dont use STMT_START/END here because it leads to |
3018 | "unreachable code" warnings, which are bogus, but distracting. */ | |
3019 | #define CACHEsayNO \ | |
c476f425 | 3020 | if (ST.cache_mask) \ |
2ac8ff4b | 3021 | reginfo->info_aux->poscache[ST.cache_offset] |= ST.cache_mask; \ |
f9f4320a | 3022 | sayNO |
3298f257 | 3023 | |
a3621e74 | 3024 | /* this is used to determine how far from the left messages like |
265c4333 YO |
3025 | 'failed...' are printed. It should be set such that messages |
3026 | are inline with the regop output that created them. | |
a3621e74 | 3027 | */ |
265c4333 | 3028 | #define REPORT_CODE_OFF 32 |
a3621e74 YO |
3029 | |
3030 | ||
40a82448 DM |
3031 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
3032 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
79a2a0e8 KW |
3033 | #define CHRTEST_NOT_A_CP_1 -999 |
3034 | #define CHRTEST_NOT_A_CP_2 -998 | |
9e137952 | 3035 | |
5d9a96ca DM |
3036 | /* grab a new slab and return the first slot in it */ |
3037 | ||
3038 | STATIC regmatch_state * | |
3039 | S_push_slab(pTHX) | |
3040 | { | |
a35a87e7 | 3041 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
3042 | dMY_CXT; |
3043 | #endif | |
5d9a96ca DM |
3044 | regmatch_slab *s = PL_regmatch_slab->next; |
3045 | if (!s) { | |
3046 | Newx(s, 1, regmatch_slab); | |
3047 | s->prev = PL_regmatch_slab; | |
3048 | s->next = NULL; | |
3049 | PL_regmatch_slab->next = s; | |
3050 | } | |
3051 | PL_regmatch_slab = s; | |
86545054 | 3052 | return SLAB_FIRST(s); |
5d9a96ca | 3053 | } |
5b47454d | 3054 | |
95b24440 | 3055 | |
40a82448 DM |
3056 | /* push a new state then goto it */ |
3057 | ||
4d5016e5 DM |
3058 | #define PUSH_STATE_GOTO(state, node, input) \ |
3059 | pushinput = input; \ | |
40a82448 DM |
3060 | scan = node; \ |
3061 | st->resume_state = state; \ | |
3062 | goto push_state; | |
3063 | ||
3064 | /* push a new state with success backtracking, then goto it */ | |
3065 | ||
4d5016e5 DM |
3066 | #define PUSH_YES_STATE_GOTO(state, node, input) \ |
3067 | pushinput = input; \ | |
40a82448 DM |
3068 | scan = node; \ |
3069 | st->resume_state = state; \ | |
3070 | goto push_yes_state; | |
3071 | ||
aa283a38 | 3072 | |
aa283a38 | 3073 | |
4d5016e5 | 3074 | |
d6a28714 | 3075 | /* |
95b24440 | 3076 | |
bf1f174e DM |
3077 | regmatch() - main matching routine |
3078 | ||
3079 | This is basically one big switch statement in a loop. We execute an op, | |
3080 | set 'next' to point the next op, and continue. If we come to a point which | |
3081 | we may need to backtrack to on failure such as (A|B|C), we push a | |
3082 | backtrack state onto the backtrack stack. On failure, we pop the top | |
3083 | state, and re-enter the loop at the state indicated. If there are no more | |
3084 | states to pop, we return failure. | |
3085 | ||
3086 | Sometimes we also need to backtrack on success; for example /A+/, where | |
3087 | after successfully matching one A, we need to go back and try to | |
3088 | match another one; similarly for lookahead assertions: if the assertion | |
3089 | completes successfully, we backtrack to the state just before the assertion | |
3090 | and then carry on. In these cases, the pushed state is marked as | |
3091 | 'backtrack on success too'. This marking is in fact done by a chain of | |
3092 | pointers, each pointing to the previous 'yes' state. On success, we pop to | |
3093 | the nearest yes state, discarding any intermediate failure-only states. | |
3094 | Sometimes a yes state is pushed just to force some cleanup code to be | |
3095 | called at the end of a successful match or submatch; e.g. (??{$re}) uses | |
3096 | it to free the inner regex. | |
3097 | ||
3098 | Note that failure backtracking rewinds the cursor position, while | |
3099 | success backtracking leaves it alone. | |
3100 | ||
3101 | A pattern is complete when the END op is executed, while a subpattern | |
3102 | such as (?=foo) is complete when the SUCCESS op is executed. Both of these | |
3103 | ops trigger the "pop to last yes state if any, otherwise return true" | |
3104 | behaviour. | |
3105 | ||
3106 | A common convention in this function is to use A and B to refer to the two | |
3107 | subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is | |
3108 | the subpattern to be matched possibly multiple times, while B is the entire | |
3109 | rest of the pattern. Variable and state names reflect this convention. | |
3110 | ||
3111 | The states in the main switch are the union of ops and failure/success of | |
3112 | substates associated with with that op. For example, IFMATCH is the op | |
3113 | that does lookahead assertions /(?=A)B/ and so the IFMATCH state means | |
3114 | 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just | |
3115 | successfully matched A and IFMATCH_A_fail is a state saying that we have | |
3116 | just failed to match A. Resume states always come in pairs. The backtrack | |
3117 | state we push is marked as 'IFMATCH_A', but when that is popped, we resume | |
3118 | at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking | |
3119 | on success or failure. | |
3120 | ||
3121 | The struct that holds a backtracking state is actually a big union, with | |
3122 | one variant for each major type of op. The variable st points to the | |
3123 | top-most backtrack struct. To make the code clearer, within each | |
3124 | block of code we #define ST to alias the relevant union. | |
3125 | ||
3126 | Here's a concrete example of a (vastly oversimplified) IFMATCH | |
3127 | implementation: | |
3128 | ||
3129 | switch (state) { | |
3130 | .... | |
3131 | ||
3132 | #define ST st->u.ifmatch | |
3133 | ||
3134 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B | |
3135 | ST.foo = ...; // some state we wish to save | |
95b24440 | 3136 | ... |
bf1f174e DM |
3137 | // push a yes backtrack state with a resume value of |
3138 | // IFMATCH_A/IFMATCH_A_fail, then continue execution at the | |
3139 | // first node of A: | |
4d5016e5 | 3140 | PUSH_YES_STATE_GOTO(IFMATCH_A, A, newinput); |
bf1f174e DM |
3141 | // NOTREACHED |
3142 | ||
3143 | case IFMATCH_A: // we have successfully executed A; now continue with B | |
3144 | next = B; | |
3145 | bar = ST.foo; // do something with the preserved value | |
3146 | break; | |
3147 | ||
3148 | case IFMATCH_A_fail: // A failed, so the assertion failed | |
3149 | ...; // do some housekeeping, then ... | |
3150 | sayNO; // propagate the failure | |
3151 | ||
3152 | #undef ST | |
95b24440 | 3153 | |
bf1f174e DM |
3154 | ... |
3155 | } | |
95b24440 | 3156 | |
bf1f174e DM |
3157 | For any old-timers reading this who are familiar with the old recursive |
3158 | approach, the code above is equivalent to: | |
95b24440 | 3159 | |
bf1f174e DM |
3160 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B |
3161 | { | |
3162 | int foo = ... | |
95b24440 | 3163 | ... |
bf1f174e DM |
3164 | if (regmatch(A)) { |
3165 | next = B; | |
3166 | bar = foo; | |
3167 | break; | |
95b24440 | 3168 | } |
bf1f174e DM |
3169 | ...; // do some housekeeping, then ... |
3170 | sayNO; // propagate the failure | |
95b24440 | 3171 | } |
bf1f174e DM |
3172 | |
3173 | The topmost backtrack state, pointed to by st, is usually free. If you | |
3174 | want to claim it, populate any ST.foo fields in it with values you wish to | |
3175 | save, then do one of | |
3176 | ||
4d5016e5 DM |
3177 | PUSH_STATE_GOTO(resume_state, node, newinput); |
3178 | PUSH_YES_STATE_GOTO(resume_state, node, newinput); | |
bf1f174e DM |
3179 | |
3180 | which sets that backtrack state's resume value to 'resume_state', pushes a | |
3181 | new free entry to the top of the backtrack stack, then goes to 'node'. | |
3182 | On backtracking, the free slot is popped, and the saved state becomes the | |
3183 | new free state. An ST.foo field in this new top state can be temporarily | |
3184 | accessed to retrieve values, but once the main loop is re-entered, it | |
3185 | becomes available for reuse. | |
3186 | ||
3187 | Note that the depth of the backtrack stack constantly increases during the | |
3188 | left-to-right execution of the pattern, rather than going up and down with | |
3189 | the pattern nesting. For example the stack is at its maximum at Z at the | |
3190 | end of the pattern, rather than at X in the following: | |
3191 | ||
3192 | /(((X)+)+)+....(Y)+....Z/ | |
3193 | ||
3194 | The only exceptions to this are lookahead/behind assertions and the cut, | |
3195 | (?>A), which pop all the backtrack states associated with A before | |
3196 | continuing. | |
3197 | ||
486ec47a | 3198 | Backtrack state structs are allocated in slabs of about 4K in size. |
bf1f174e DM |
3199 | PL_regmatch_state and st always point to the currently active state, |
3200 | and PL_regmatch_slab points to the slab currently containing | |
3201 | PL_regmatch_state. The first time regmatch() is called, the first slab is | |
3202 | allocated, and is never freed until interpreter destruction. When the slab | |
3203 | is full, a new one is allocated and chained to the end. At exit from | |
3204 | regmatch(), slabs allocated since entry are freed. | |
3205 | ||
3206 | */ | |
95b24440 | 3207 | |
40a82448 | 3208 | |
5bc10b2c | 3209 | #define DEBUG_STATE_pp(pp) \ |
265c4333 | 3210 | DEBUG_STATE_r({ \ |
f2ed9b32 | 3211 | DUMP_EXEC_POS(locinput, scan, utf8_target); \ |
5bc10b2c | 3212 | PerlIO_printf(Perl_debug_log, \ |
5d458dd8 | 3213 | " %*s"pp" %s%s%s%s%s\n", \ |
5bc10b2c | 3214 | depth*2, "", \ |
13d6edb4 | 3215 | PL_reg_name[st->resume_state], \ |
5d458dd8 YO |
3216 | ((st==yes_state||st==mark_state) ? "[" : ""), \ |
3217 | ((st==yes_state) ? "Y" : ""), \ | |
3218 | ((st==mark_state) ? "M" : ""), \ | |
3219 | ((st==yes_state||st==mark_state) ? "]" : "") \ | |
3220 | ); \ | |
265c4333 | 3221 | }); |
5bc10b2c | 3222 | |
40a82448 | 3223 | |
3dab1dad | 3224 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 3225 | |
3df15adc | 3226 | #ifdef DEBUGGING |
5bc10b2c | 3227 | |
ab3bbdeb | 3228 | STATIC void |
f2ed9b32 | 3229 | S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target, |
ab3bbdeb YO |
3230 | const char *start, const char *end, const char *blurb) |
3231 | { | |
efd26800 | 3232 | const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; |
7918f24d NC |
3233 | |
3234 | PERL_ARGS_ASSERT_DEBUG_START_MATCH; | |
3235 | ||
ab3bbdeb YO |
3236 | if (!PL_colorset) |
3237 | reginitcolors(); | |
3238 | { | |
3239 | RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), | |
d2c6dc5e | 3240 | RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); |
ab3bbdeb | 3241 | |
f2ed9b32 | 3242 | RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb YO |
3243 | start, end - start, 60); |
3244 | ||
3245 | PerlIO_printf(Perl_debug_log, | |
3246 | "%s%s REx%s %s against %s\n", | |
3247 | PL_colors[4], blurb, PL_colors[5], s0, s1); | |
3248 | ||
f2ed9b32 | 3249 | if (utf8_target||utf8_pat) |
1de06328 YO |
3250 | PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", |
3251 | utf8_pat ? "pattern" : "", | |
f2ed9b32 KW |
3252 | utf8_pat && utf8_target ? " and " : "", |
3253 | utf8_target ? "string" : "" | |
ab3bbdeb YO |
3254 | ); |
3255 | } | |
3256 | } | |
3df15adc YO |
3257 | |
3258 | STATIC void | |
786e8c11 YO |
3259 | S_dump_exec_pos(pTHX_ const char *locinput, |
3260 | const regnode *scan, | |
3261 | const char *loc_regeol, | |
3262 | const char *loc_bostr, | |
3263 | const char *loc_reg_starttry, | |
f2ed9b32 | 3264 | const bool utf8_target) |
07be1b83 | 3265 | { |
786e8c11 | 3266 | const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; |
07be1b83 | 3267 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ |
786e8c11 | 3268 | int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); |
07be1b83 YO |
3269 | /* The part of the string before starttry has one color |
3270 | (pref0_len chars), between starttry and current | |
3271 | position another one (pref_len - pref0_len chars), | |
3272 | after the current position the third one. | |
3273 | We assume that pref0_len <= pref_len, otherwise we | |
3274 | decrease pref0_len. */ | |
786e8c11 YO |
3275 | int pref_len = (locinput - loc_bostr) > (5 + taill) - l |
3276 | ? (5 + taill) - l : locinput - loc_bostr; | |
07be1b83 YO |
3277 | int pref0_len; |
3278 | ||
7918f24d NC |
3279 | PERL_ARGS_ASSERT_DUMP_EXEC_POS; |
3280 | ||
f2ed9b32 | 3281 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) |
07be1b83 | 3282 | pref_len++; |
786e8c11 YO |
3283 | pref0_len = pref_len - (locinput - loc_reg_starttry); |
3284 | if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) | |
3285 | l = ( loc_regeol - locinput > (5 + taill) - pref_len | |
3286 | ? (5 + taill) - pref_len : loc_regeol - locinput); | |
f2ed9b32 | 3287 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) |
07be1b83 YO |
3288 | l--; |
3289 | if (pref0_len < 0) | |
3290 | pref0_len = 0; | |
3291 | if (pref0_len > pref_len) | |
3292 | pref0_len = pref_len; | |
3293 | { | |
f2ed9b32 | 3294 | const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0; |
0df25f3d | 3295 | |
ab3bbdeb | 3296 | RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), |
1de06328 | 3297 | (locinput - pref_len),pref0_len, 60, 4, 5); |
0df25f3d | 3298 | |
ab3bbdeb | 3299 | RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), |
3df15adc | 3300 | (locinput - pref_len + pref0_len), |
1de06328 | 3301 | pref_len - pref0_len, 60, 2, 3); |
0df25f3d | 3302 | |
ab3bbdeb | 3303 | RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), |
1de06328 | 3304 | locinput, loc_regeol - locinput, 10, 0, 1); |
0df25f3d | 3305 | |
1de06328 | 3306 | const STRLEN tlen=len0+len1+len2; |
3df15adc | 3307 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb | 3308 | "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", |
786e8c11 | 3309 | (IV)(locinput - loc_bostr), |
07be1b83 | 3310 | len0, s0, |
07be1b83 | 3311 | len1, s1, |
07be1b83 | 3312 | (docolor ? "" : "> <"), |
07be1b83 | 3313 | len2, s2, |
f9f4320a | 3314 | (int)(tlen > 19 ? 0 : 19 - tlen), |
07be1b83 YO |
3315 | ""); |
3316 | } | |
3317 | } | |
3df15adc | 3318 | |
07be1b83 YO |
3319 | #endif |
3320 | ||
0a4db386 YO |
3321 | /* reg_check_named_buff_matched() |
3322 | * Checks to see if a named buffer has matched. The data array of | |
3323 | * buffer numbers corresponding to the buffer is expected to reside | |
3324 | * in the regexp->data->data array in the slot stored in the ARG() of | |
3325 | * node involved. Note that this routine doesn't actually care about the | |
3326 | * name, that information is not preserved from compilation to execution. | |
3327 | * Returns the index of the leftmost defined buffer with the given name | |
3328 | * or 0 if non of the buffers matched. | |
3329 | */ | |
3330 | STATIC I32 | |
7918f24d NC |
3331 | S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) |
3332 | { | |
0a4db386 | 3333 | I32 n; |
f8fc2ecf | 3334 | RXi_GET_DECL(rex,rexi); |
ad64d0ec | 3335 | SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
0a4db386 | 3336 | I32 *nums=(I32*)SvPVX(sv_dat); |
7918f24d NC |
3337 | |
3338 | PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; | |
3339 | ||
0a4db386 | 3340 | for ( n=0; n<SvIVX(sv_dat); n++ ) { |
b93070ed DM |
3341 | if ((I32)rex->lastparen >= nums[n] && |
3342 | rex->offs[nums[n]].end != -1) | |
0a4db386 YO |
3343 | { |
3344 | return nums[n]; | |
3345 | } | |
3346 | } | |
3347 | return 0; | |
3348 | } | |
3349 | ||
2f554ef7 | 3350 | |
c74f6de9 | 3351 | static bool |
984e6dd1 | 3352 | S_setup_EXACTISH_ST_c1_c2(pTHX_ const regnode * const text_node, int *c1p, |
aed7b151 | 3353 | U8* c1_utf8, int *c2p, U8* c2_utf8, regmatch_info *reginfo) |
c74f6de9 | 3354 | { |
79a2a0e8 KW |
3355 | /* This function determines if there are one or two characters that match |
3356 | * the first character of the passed-in EXACTish node <text_node>, and if | |
3357 | * so, returns them in the passed-in pointers. | |
c74f6de9 | 3358 | * |
79a2a0e8 KW |
3359 | * If it determines that no possible character in the target string can |
3360 | * match, it returns FALSE; otherwise TRUE. (The FALSE situation occurs if | |
3361 | * the first character in <text_node> requires UTF-8 to represent, and the | |
3362 | * target string isn't in UTF-8.) | |
c74f6de9 | 3363 | * |
79a2a0e8 KW |
3364 | * If there are more than two characters that could match the beginning of |
3365 | * <text_node>, or if more context is required to determine a match or not, | |
3366 | * it sets both *<c1p> and *<c2p> to CHRTEST_VOID. | |
3367 | * | |
3368 | * The motiviation behind this function is to allow the caller to set up | |
3369 | * tight loops for matching. If <text_node> is of type EXACT, there is | |
3370 | * only one possible character that can match its first character, and so | |
3371 | * the situation is quite simple. But things get much more complicated if | |
3372 | * folding is involved. It may be that the first character of an EXACTFish | |
3373 | * node doesn't participate in any possible fold, e.g., punctuation, so it | |
3374 | * can be matched only by itself. The vast majority of characters that are | |
3375 | * in folds match just two things, their lower and upper-case equivalents. | |
3376 | * But not all are like that; some have multiple possible matches, or match | |
3377 | * sequences of more than one character. This function sorts all that out. | |
3378 | * | |
3379 | * Consider the patterns A*B or A*?B where A and B are arbitrary. In a | |
3380 | * loop of trying to match A*, we know we can't exit where the thing | |
3381 | * following it isn't a B. And something can't be a B unless it is the | |
3382 | * beginning of B. By putting a quick test for that beginning in a tight | |
3383 | * loop, we can rule out things that can't possibly be B without having to | |
3384 | * break out of the loop, thus avoiding work. Similarly, if A is a single | |
3385 | * character, we can make a tight loop matching A*, using the outputs of | |
3386 | * this function. | |
3387 | * | |
3388 | * If the target string to match isn't in UTF-8, and there aren't | |
3389 | * complications which require CHRTEST_VOID, *<c1p> and *<c2p> are set to | |
3390 | * the one or two possible octets (which are characters in this situation) | |
3391 | * that can match. In all cases, if there is only one character that can | |
3392 | * match, *<c1p> and *<c2p> will be identical. | |
3393 | * | |
3394 | * If the target string is in UTF-8, the buffers pointed to by <c1_utf8> | |
3395 | * and <c2_utf8> will contain the one or two UTF-8 sequences of bytes that | |
3396 | * can match the beginning of <text_node>. They should be declared with at | |
3397 | * least length UTF8_MAXBYTES+1. (If the target string isn't in UTF-8, it is | |
3398 | * undefined what these contain.) If one or both of the buffers are | |
3399 | * invariant under UTF-8, *<c1p>, and *<c2p> will also be set to the | |
3400 | * corresponding invariant. If variant, the corresponding *<c1p> and/or | |
3401 | * *<c2p> will be set to a negative number(s) that shouldn't match any code | |
3402 | * point (unless inappropriately coerced to unsigned). *<c1p> will equal | |
3403 | * *<c2p> if and only if <c1_utf8> and <c2_utf8> are the same. */ | |
c74f6de9 | 3404 | |
ba44c216 | 3405 | const bool utf8_target = reginfo->is_utf8_target; |
79a2a0e8 | 3406 | |
ddb0d839 KW |
3407 | UV c1 = CHRTEST_NOT_A_CP_1; |
3408 | UV c2 = CHRTEST_NOT_A_CP_2; | |
79a2a0e8 | 3409 | bool use_chrtest_void = FALSE; |
aed7b151 | 3410 | const bool is_utf8_pat = reginfo->is_utf8_pat; |
79a2a0e8 KW |
3411 | |
3412 | /* Used when we have both utf8 input and utf8 output, to avoid converting | |
3413 | * to/from code points */ | |
3414 | bool utf8_has_been_setup = FALSE; | |
3415 | ||
c74f6de9 KW |
3416 | dVAR; |
3417 | ||
b4291290 | 3418 | U8 *pat = (U8*)STRING(text_node); |
c74f6de9 | 3419 | |
79a2a0e8 KW |
3420 | if (OP(text_node) == EXACT) { |
3421 | ||
3422 | /* In an exact node, only one thing can be matched, that first | |
3423 | * character. If both the pat and the target are UTF-8, we can just | |
3424 | * copy the input to the output, avoiding finding the code point of | |
3425 | * that character */ | |
984e6dd1 | 3426 | if (!is_utf8_pat) { |
79a2a0e8 KW |
3427 | c2 = c1 = *pat; |
3428 | } | |
3429 | else if (utf8_target) { | |
3430 | Copy(pat, c1_utf8, UTF8SKIP(pat), U8); | |
3431 | Copy(pat, c2_utf8, UTF8SKIP(pat), U8); | |
3432 | utf8_has_been_setup = TRUE; | |
3433 | } | |
3434 | else { | |
3435 | c2 = c1 = valid_utf8_to_uvchr(pat, NULL); | |
c74f6de9 | 3436 | } |
79a2a0e8 KW |
3437 | } |
3438 | else /* an EXACTFish node */ | |
984e6dd1 | 3439 | if ((is_utf8_pat |
79a2a0e8 KW |
3440 | && is_MULTI_CHAR_FOLD_utf8_safe(pat, |
3441 | pat + STR_LEN(text_node))) | |
984e6dd1 | 3442 | || (!is_utf8_pat |
79a2a0e8 KW |
3443 | && is_MULTI_CHAR_FOLD_latin1_safe(pat, |
3444 | pat + STR_LEN(text_node)))) | |
3445 | { | |
3446 | /* Multi-character folds require more context to sort out. Also | |
3447 | * PL_utf8_foldclosures used below doesn't handle them, so have to be | |
3448 | * handled outside this routine */ | |
3449 | use_chrtest_void = TRUE; | |
3450 | } | |
3451 | else { /* an EXACTFish node which doesn't begin with a multi-char fold */ | |
984e6dd1 | 3452 | c1 = is_utf8_pat ? valid_utf8_to_uvchr(pat, NULL) : *pat; |
79a2a0e8 KW |
3453 | if (c1 > 256) { |
3454 | /* Load the folds hash, if not already done */ | |
3455 | SV** listp; | |
3456 | if (! PL_utf8_foldclosures) { | |
3457 | if (! PL_utf8_tofold) { | |
632ccf21 | 3458 | U8 dummy[UTF8_MAXBYTES_CASE+1]; |
79a2a0e8 KW |
3459 | |
3460 | /* Force loading this by folding an above-Latin1 char */ | |
3461 | to_utf8_fold((U8*) HYPHEN_UTF8, dummy, NULL); | |
3462 | assert(PL_utf8_tofold); /* Verify that worked */ | |
3463 | } | |
3464 | PL_utf8_foldclosures = _swash_inversion_hash(PL_utf8_tofold); | |
3465 | } | |
3466 | ||
3467 | /* The fold closures data structure is a hash with the keys being | |
3468 | * the UTF-8 of every character that is folded to, like 'k', and | |
3469 | * the values each an array of all code points that fold to its | |
3470 | * key. e.g. [ 'k', 'K', KELVIN_SIGN ]. Multi-character folds are | |
3471 | * not included */ | |
3472 | if ((! (listp = hv_fetch(PL_utf8_foldclosures, | |
3473 | (char *) pat, | |
3474 | UTF8SKIP(pat), | |
3475 | FALSE)))) | |
3476 | { | |
3477 | /* Not found in the hash, therefore there are no folds | |
3478 | * containing it, so there is only a single character that | |
3479 | * could match */ | |
3480 | c2 = c1; | |
3481 | } | |
3482 | else { /* Does participate in folds */ | |
3483 | AV* list = (AV*) *listp; | |
3484 | if (av_len(list) != 1) { | |
3485 | ||
3486 | /* If there aren't exactly two folds to this, it is outside | |
3487 | * the scope of this function */ | |
3488 | use_chrtest_void = TRUE; | |
3489 | } | |
3490 | else { /* There are two. Get them */ | |
3491 | SV** c_p = av_fetch(list, 0, FALSE); | |
3492 | if (c_p == NULL) { | |
3493 | Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure"); | |
3494 | } | |
3495 | c1 = SvUV(*c_p); | |
3496 | ||
3497 | c_p = av_fetch(list, 1, FALSE); | |
3498 | if (c_p == NULL) { | |
3499 | Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure"); | |
3500 | } | |
3501 | c2 = SvUV(*c_p); | |
3502 | ||
3503 | /* Folds that cross the 255/256 boundary are forbidden if | |
3504 | * EXACTFL, or EXACTFA and one is ASCIII. Since the | |
3505 | * pattern character is above 256, and its only other match | |
3506 | * is below 256, the only legal match will be to itself. | |
3507 | * We have thrown away the original, so have to compute | |
3508 | * which is the one above 255 */ | |
3509 | if ((c1 < 256) != (c2 < 256)) { | |
3510 | if (OP(text_node) == EXACTFL | |
098b07d5 KW |
3511 | || ((OP(text_node) == EXACTFA |
3512 | || OP(text_node) == EXACTFA_NO_TRIE) | |
79a2a0e8 KW |
3513 | && (isASCII(c1) || isASCII(c2)))) |
3514 | { | |
3515 | if (c1 < 256) { | |
3516 | c1 = c2; | |
3517 | } | |
3518 | else { | |
3519 | c2 = c1; | |
3520 | } | |
3521 | } | |
3522 | } | |
3523 | } | |
3524 | } | |
3525 | } | |
3526 | else /* Here, c1 is < 255 */ | |
3527 | if (utf8_target | |
3528 | && HAS_NONLATIN1_FOLD_CLOSURE(c1) | |
3529 | && OP(text_node) != EXACTFL | |
098b07d5 KW |
3530 | && ((OP(text_node) != EXACTFA |
3531 | && OP(text_node) != EXACTFA_NO_TRIE) | |
3532 | || ! isASCII(c1))) | |
c74f6de9 KW |
3533 | { |
3534 | /* Here, there could be something above Latin1 in the target which | |
79a2a0e8 KW |
3535 | * folds to this character in the pattern. All such cases except |
3536 | * LATIN SMALL LETTER Y WITH DIAERESIS have more than two characters | |
3537 | * involved in their folds, so are outside the scope of this | |
3538 | * function */ | |
3539 | if (UNLIKELY(c1 == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)) { | |
3540 | c2 = LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS; | |
3541 | } | |
3542 | else { | |
3543 | use_chrtest_void = TRUE; | |
3544 | } | |
c74f6de9 KW |
3545 | } |
3546 | else { /* Here nothing above Latin1 can fold to the pattern character */ | |
3547 | switch (OP(text_node)) { | |
3548 | ||
3549 | case EXACTFL: /* /l rules */ | |
79a2a0e8 | 3550 | c2 = PL_fold_locale[c1]; |
c74f6de9 KW |
3551 | break; |
3552 | ||
2fdb7295 KW |
3553 | case EXACTF: /* This node only generated for non-utf8 |
3554 | patterns */ | |
3555 | assert(! is_utf8_pat); | |
c74f6de9 | 3556 | if (! utf8_target) { /* /d rules */ |
79a2a0e8 | 3557 | c2 = PL_fold[c1]; |
c74f6de9 KW |
3558 | break; |
3559 | } | |
3560 | /* FALLTHROUGH */ | |
3561 | /* /u rules for all these. This happens to work for | |
79a2a0e8 | 3562 | * EXACTFA as nothing in Latin1 folds to ASCII */ |
098b07d5 KW |
3563 | case EXACTFA_NO_TRIE: /* This node only generated for |
3564 | non-utf8 patterns */ | |
3565 | assert(! is_utf8_pat); | |
3566 | /* FALL THROUGH */ | |
c74f6de9 | 3567 | case EXACTFA: |
79a2a0e8 | 3568 | case EXACTFU_SS: |
c74f6de9 | 3569 | case EXACTFU: |
79a2a0e8 | 3570 | c2 = PL_fold_latin1[c1]; |
c74f6de9 KW |
3571 | break; |
3572 | ||
878531d3 KW |
3573 | default: |
3574 | Perl_croak(aTHX_ "panic: Unexpected op %u", OP(text_node)); | |
3575 | assert(0); /* NOTREACHED */ | |
c74f6de9 KW |
3576 | } |
3577 | } | |
3578 | } | |
79a2a0e8 KW |
3579 | |
3580 | /* Here have figured things out. Set up the returns */ | |
3581 | if (use_chrtest_void) { | |
3582 | *c2p = *c1p = CHRTEST_VOID; | |
3583 | } | |
3584 | else if (utf8_target) { | |
3585 | if (! utf8_has_been_setup) { /* Don't have the utf8; must get it */ | |
3586 | uvchr_to_utf8(c1_utf8, c1); | |
3587 | uvchr_to_utf8(c2_utf8, c2); | |
c74f6de9 | 3588 | } |
c74f6de9 | 3589 | |
79a2a0e8 KW |
3590 | /* Invariants are stored in both the utf8 and byte outputs; Use |
3591 | * negative numbers otherwise for the byte ones. Make sure that the | |
3592 | * byte ones are the same iff the utf8 ones are the same */ | |
3593 | *c1p = (UTF8_IS_INVARIANT(*c1_utf8)) ? *c1_utf8 : CHRTEST_NOT_A_CP_1; | |
3594 | *c2p = (UTF8_IS_INVARIANT(*c2_utf8)) | |
3595 | ? *c2_utf8 | |
3596 | : (c1 == c2) | |
3597 | ? CHRTEST_NOT_A_CP_1 | |
3598 | : CHRTEST_NOT_A_CP_2; | |
3599 | } | |
3600 | else if (c1 > 255) { | |
3601 | if (c2 > 255) { /* both possibilities are above what a non-utf8 string | |
3602 | can represent */ | |
3603 | return FALSE; | |
3604 | } | |
c74f6de9 | 3605 | |
79a2a0e8 KW |
3606 | *c1p = *c2p = c2; /* c2 is the only representable value */ |
3607 | } | |
3608 | else { /* c1 is representable; see about c2 */ | |
3609 | *c1p = c1; | |
3610 | *c2p = (c2 < 256) ? c2 : c1; | |
c74f6de9 | 3611 | } |
2f554ef7 | 3612 | |
c74f6de9 KW |
3613 | return TRUE; |
3614 | } | |
2f554ef7 | 3615 | |
f73aaa43 | 3616 | /* returns -1 on failure, $+[0] on success */ |
99a90e59 | 3617 | STATIC SSize_t |
f73aaa43 | 3618 | S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) |
d6a28714 | 3619 | { |
a35a87e7 | 3620 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
3621 | dMY_CXT; |
3622 | #endif | |
27da23d5 | 3623 | dVAR; |
ba44c216 | 3624 | const bool utf8_target = reginfo->is_utf8_target; |
4ad0818d | 3625 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
288b8c02 | 3626 | REGEXP *rex_sv = reginfo->prog; |
8d919b0a | 3627 | regexp *rex = ReANY(rex_sv); |
f8fc2ecf | 3628 | RXi_GET_DECL(rex,rexi); |
5d9a96ca | 3629 | /* the current state. This is a cached copy of PL_regmatch_state */ |
eb578fdb | 3630 | regmatch_state *st; |
5d9a96ca | 3631 | /* cache heavy used fields of st in registers */ |
eb578fdb KW |
3632 | regnode *scan; |
3633 | regnode *next; | |
3634 | U32 n = 0; /* general value; init to avoid compiler warning */ | |
ea3daa5d | 3635 | SSize_t ln = 0; /* len or last; init to avoid compiler warning */ |
d60de1d1 | 3636 | char *locinput = startpos; |
4d5016e5 | 3637 | char *pushinput; /* where to continue after a PUSH */ |
eb578fdb | 3638 | I32 nextchr; /* is always set to UCHARAT(locinput) */ |
24d3c4a9 | 3639 | |
b69b0499 | 3640 | bool result = 0; /* return value of S_regmatch */ |
24d3c4a9 | 3641 | int depth = 0; /* depth of backtrack stack */ |
4b196cd4 YO |
3642 | U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */ |
3643 | const U32 max_nochange_depth = | |
3644 | (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? | |
3645 | 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; | |
77cb431f DM |
3646 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
3647 | subpattern */ | |
e2e6a0f1 YO |
3648 | /* mark_state piggy backs on the yes_state logic so that when we unwind |
3649 | the stack on success we can update the mark_state as we go */ | |
3650 | regmatch_state *mark_state = NULL; /* last mark state we have seen */ | |
faec1544 | 3651 | regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ |
b8591aee | 3652 | struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ |
40a82448 | 3653 | U32 state_num; |
5d458dd8 YO |
3654 | bool no_final = 0; /* prevent failure from backtracking? */ |
3655 | bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ | |
d60de1d1 | 3656 | char *startpoint = locinput; |
5d458dd8 YO |
3657 | SV *popmark = NULL; /* are we looking for a mark? */ |
3658 | SV *sv_commit = NULL; /* last mark name seen in failure */ | |
3659 | SV *sv_yes_mark = NULL; /* last mark name we have seen | |
486ec47a | 3660 | during a successful match */ |
5d458dd8 YO |
3661 | U32 lastopen = 0; /* last open we saw */ |
3662 | bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; | |
4a2b275c | 3663 | SV* const oreplsv = GvSVn(PL_replgv); |
24d3c4a9 DM |
3664 | /* these three flags are set by various ops to signal information to |
3665 | * the very next op. They have a useful lifetime of exactly one loop | |
3666 | * iteration, and are not preserved or restored by state pushes/pops | |
3667 | */ | |
3668 | bool sw = 0; /* the condition value in (?(cond)a|b) */ | |
3669 | bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ | |
3670 | int logical = 0; /* the following EVAL is: | |
3671 | 0: (?{...}) | |
3672 | 1: (?(?{...})X|Y) | |
3673 | 2: (??{...}) | |
3674 | or the following IFMATCH/UNLESSM is: | |
3675 | false: plain (?=foo) | |
3676 | true: used as a condition: (?(?=foo)) | |
3677 | */ | |
81ed78b2 DM |
3678 | PAD* last_pad = NULL; |
3679 | dMULTICALL; | |
3680 | I32 gimme = G_SCALAR; | |
3681 | CV *caller_cv = NULL; /* who called us */ | |
3682 | CV *last_pushed_cv = NULL; /* most recently called (?{}) CV */ | |
74088413 | 3683 | CHECKPOINT runops_cp; /* savestack position before executing EVAL */ |
92da3157 | 3684 | U32 maxopenparen = 0; /* max '(' index seen so far */ |
3018b823 KW |
3685 | int to_complement; /* Invert the result? */ |
3686 | _char_class_number classnum; | |
984e6dd1 | 3687 | bool is_utf8_pat = reginfo->is_utf8_pat; |
81ed78b2 | 3688 | |
95b24440 | 3689 | #ifdef DEBUGGING |
e68ec53f | 3690 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
3691 | #endif |
3692 | ||
81ed78b2 DM |
3693 | /* shut up 'may be used uninitialized' compiler warnings for dMULTICALL */ |
3694 | multicall_oldcatch = 0; | |
3695 | multicall_cv = NULL; | |
3696 | cx = NULL; | |
4f8dbb2d JL |
3697 | PERL_UNUSED_VAR(multicall_cop); |
3698 | PERL_UNUSED_VAR(newsp); | |
81ed78b2 DM |
3699 | |
3700 | ||
7918f24d NC |
3701 | PERL_ARGS_ASSERT_REGMATCH; |
3702 | ||
3b57cd43 | 3703 | DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ |
24b23f37 | 3704 | PerlIO_printf(Perl_debug_log,"regmatch start\n"); |
3b57cd43 | 3705 | })); |
5d9a96ca | 3706 | |
331b2dcc | 3707 | st = PL_regmatch_state; |
5d9a96ca | 3708 | |
d6a28714 | 3709 | /* Note that nextchr is a byte even in UTF */ |
7016d6eb | 3710 | SET_nextchr; |
d6a28714 JH |
3711 | scan = prog; |
3712 | while (scan != NULL) { | |
8ba1375e | 3713 | |
a3621e74 | 3714 | DEBUG_EXECUTE_r( { |
6136c704 | 3715 | SV * const prop = sv_newmortal(); |
1de06328 | 3716 | regnode *rnext=regnext(scan); |
f2ed9b32 | 3717 | DUMP_EXEC_POS( locinput, scan, utf8_target ); |
32fc9b6a | 3718 | regprop(rex, prop, scan); |
07be1b83 YO |
3719 | |
3720 | PerlIO_printf(Perl_debug_log, | |
3721 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
f8fc2ecf | 3722 | (IV)(scan - rexi->program), depth*2, "", |
07be1b83 | 3723 | SvPVX_const(prop), |
1de06328 | 3724 | (PL_regkind[OP(scan)] == END || !rnext) ? |
f8fc2ecf | 3725 | 0 : (IV)(rnext - rexi->program)); |
2a782b5b | 3726 | }); |
d6a28714 JH |
3727 | |
3728 | next = scan + NEXT_OFF(scan); | |
3729 | if (next == scan) | |
3730 | next = NULL; | |
40a82448 | 3731 | state_num = OP(scan); |
d6a28714 | 3732 | |
40a82448 | 3733 | reenter_switch: |
3018b823 | 3734 | to_complement = 0; |
34a81e2b | 3735 | |
7016d6eb | 3736 | SET_nextchr; |
e6ca698c | 3737 | assert(nextchr < 256 && (nextchr >= 0 || nextchr == NEXTCHR_EOS)); |
bf798dc4 | 3738 | |
40a82448 | 3739 | switch (state_num) { |
3c0563b9 | 3740 | case BOL: /* /^../ */ |
9d9163fb | 3741 | if (locinput == reginfo->strbeg) |
b8c5462f | 3742 | break; |
d6a28714 | 3743 | sayNO; |
3c0563b9 DM |
3744 | |
3745 | case MBOL: /* /^../m */ | |
9d9163fb | 3746 | if (locinput == reginfo->strbeg || |
7016d6eb | 3747 | (!NEXTCHR_IS_EOS && locinput[-1] == '\n')) |
d6a28714 | 3748 | { |
b8c5462f JH |
3749 | break; |
3750 | } | |
d6a28714 | 3751 | sayNO; |
3c0563b9 DM |
3752 | |
3753 | case SBOL: /* /^../s */ | |
9d9163fb | 3754 | if (locinput == reginfo->strbeg) |
b8c5462f | 3755 | break; |
d6a28714 | 3756 | sayNO; |
3c0563b9 DM |
3757 | |
3758 | case GPOS: /* \G */ | |
3b0527fe | 3759 | if (locinput == reginfo->ganch) |
d6a28714 JH |
3760 | break; |
3761 | sayNO; | |
ee9b8eae | 3762 | |
3c0563b9 | 3763 | case KEEPS: /* \K */ |
ee9b8eae | 3764 | /* update the startpoint */ |
b93070ed | 3765 | st->u.keeper.val = rex->offs[0].start; |
9d9163fb | 3766 | rex->offs[0].start = locinput - reginfo->strbeg; |
4d5016e5 | 3767 | PUSH_STATE_GOTO(KEEPS_next, next, locinput); |
878531d3 | 3768 | assert(0); /*NOTREACHED*/ |
ee9b8eae YO |
3769 | case KEEPS_next_fail: |
3770 | /* rollback the start point change */ | |
b93070ed | 3771 | rex->offs[0].start = st->u.keeper.val; |
ee9b8eae | 3772 | sayNO_SILENT; |
878531d3 | 3773 | assert(0); /*NOTREACHED*/ |
3c0563b9 | 3774 | |
3c0563b9 | 3775 | case MEOL: /* /..$/m */ |
7016d6eb | 3776 | if (!NEXTCHR_IS_EOS && nextchr != '\n') |
b8c5462f | 3777 | sayNO; |
b8c5462f | 3778 | break; |
3c0563b9 | 3779 | |
62e16618 KW |
3780 | case EOL: /* /..$/ */ |
3781 | /* FALL THROUGH */ | |
3c0563b9 | 3782 | case SEOL: /* /..$/s */ |
7016d6eb | 3783 | if (!NEXTCHR_IS_EOS && nextchr != '\n') |
b8c5462f | 3784 | sayNO; |
220db18a | 3785 | if (reginfo->strend - locinput > 1) |
b8c5462f | 3786 | sayNO; |
b8c5462f | 3787 | break; |
3c0563b9 DM |
3788 | |
3789 | case EOS: /* \z */ | |
7016d6eb | 3790 | if (!NEXTCHR_IS_EOS) |
b8c5462f | 3791 | sayNO; |
d6a28714 | 3792 | break; |
3c0563b9 DM |
3793 | |
3794 | case SANY: /* /./s */ | |
7016d6eb | 3795 | if (NEXTCHR_IS_EOS) |
4633a7c4 | 3796 | sayNO; |
28b98f76 | 3797 | goto increment_locinput; |
3c0563b9 DM |
3798 | |
3799 | case CANY: /* \C */ | |
7016d6eb | 3800 | if (NEXTCHR_IS_EOS) |
f33976b4 | 3801 | sayNO; |
3640db6b | 3802 | locinput++; |
a0d0e21e | 3803 | break; |
3c0563b9 DM |
3804 | |
3805 | case REG_ANY: /* /./ */ | |
7016d6eb | 3806 | if ((NEXTCHR_IS_EOS) || nextchr == '\n') |
1aa99e6b | 3807 | sayNO; |
28b98f76 DM |
3808 | goto increment_locinput; |
3809 | ||
166ba7cd DM |
3810 | |
3811 | #undef ST | |
3812 | #define ST st->u.trie | |
3c0563b9 | 3813 | case TRIEC: /* (ab|cd) with known charclass */ |
786e8c11 YO |
3814 | /* In this case the charclass data is available inline so |
3815 | we can fail fast without a lot of extra overhead. | |
3816 | */ | |
7016d6eb | 3817 | if(!NEXTCHR_IS_EOS && !ANYOF_BITMAP_TEST(scan, nextchr)) { |
fab2782b YO |
3818 | DEBUG_EXECUTE_r( |
3819 | PerlIO_printf(Perl_debug_log, | |
3820 | "%*s %sfailed to match trie start class...%s\n", | |
3821 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) | |
3822 | ); | |
3823 | sayNO_SILENT; | |
118e2215 | 3824 | assert(0); /* NOTREACHED */ |
786e8c11 YO |
3825 | } |
3826 | /* FALL THROUGH */ | |
3c0563b9 | 3827 | case TRIE: /* (ab|cd) */ |
2e64971a DM |
3828 | /* the basic plan of execution of the trie is: |
3829 | * At the beginning, run though all the states, and | |
3830 | * find the longest-matching word. Also remember the position | |
3831 | * of the shortest matching word. For example, this pattern: | |
3832 | * 1 2 3 4 5 | |
3833 | * ab|a|x|abcd|abc | |
3834 | * when matched against the string "abcde", will generate | |
3835 | * accept states for all words except 3, with the longest | |
895cc420 | 3836 | * matching word being 4, and the shortest being 2 (with |
2e64971a DM |
3837 | * the position being after char 1 of the string). |
3838 | * | |
3839 | * Then for each matching word, in word order (i.e. 1,2,4,5), | |
3840 | * we run the remainder of the pattern; on each try setting | |
3841 | * the current position to the character following the word, | |
3842 | * returning to try the next word on failure. | |
3843 | * | |
3844 | * We avoid having to build a list of words at runtime by | |
3845 | * using a compile-time structure, wordinfo[].prev, which | |
3846 | * gives, for each word, the previous accepting word (if any). | |
3847 | * In the case above it would contain the mappings 1->2, 2->0, | |
3848 | * 3->0, 4->5, 5->1. We can use this table to generate, from | |
3849 | * the longest word (4 above), a list of all words, by | |
3850 | * following the list of prev pointers; this gives us the | |
3851 | * unordered list 4,5,1,2. Then given the current word we have | |
3852 | * just tried, we can go through the list and find the | |
3853 | * next-biggest word to try (so if we just failed on word 2, | |
3854 | * the next in the list is 4). | |
3855 | * | |
3856 | * Since at runtime we don't record the matching position in | |
3857 | * the string for each word, we have to work that out for | |
3858 | * each word we're about to process. The wordinfo table holds | |
3859 | * the character length of each word; given that we recorded | |
3860 | * at the start: the position of the shortest word and its | |
3861 | * length in chars, we just need to move the pointer the | |
3862 | * difference between the two char lengths. Depending on | |
3863 | * Unicode status and folding, that's cheap or expensive. | |
3864 | * | |
3865 | * This algorithm is optimised for the case where are only a | |
3866 | * small number of accept states, i.e. 0,1, or maybe 2. | |
3867 | * With lots of accepts states, and having to try all of them, | |
3868 | * it becomes quadratic on number of accept states to find all | |
3869 | * the next words. | |
3870 | */ | |
3871 | ||
3dab1dad | 3872 | { |
07be1b83 | 3873 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
a0a388a1 | 3874 | DECL_TRIE_TYPE(scan); |
3dab1dad YO |
3875 | |
3876 | /* what trie are we using right now */ | |
be8e71aa | 3877 | reg_trie_data * const trie |
f8fc2ecf | 3878 | = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; |
85fbaab2 | 3879 | HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); |
3dab1dad | 3880 | U32 state = trie->startstate; |
166ba7cd | 3881 | |
7016d6eb DM |
3882 | if ( trie->bitmap |
3883 | && (NEXTCHR_IS_EOS || !TRIE_BITMAP_TEST(trie, nextchr))) | |
3884 | { | |
3dab1dad YO |
3885 | if (trie->states[ state ].wordnum) { |
3886 | DEBUG_EXECUTE_r( | |
3887 | PerlIO_printf(Perl_debug_log, | |
3888 | "%*s %smatched empty string...%s\n", | |
5bc10b2c | 3889 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad | 3890 | ); |
20dbff7c YO |
3891 | if (!trie->jump) |
3892 | break; | |
3dab1dad YO |
3893 | } else { |
3894 | DEBUG_EXECUTE_r( | |
3895 | PerlIO_printf(Perl_debug_log, | |
786e8c11 | 3896 | "%*s %sfailed to match trie start class...%s\n", |
5bc10b2c | 3897 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad YO |
3898 | ); |
3899 | sayNO_SILENT; | |
3900 | } | |
3901 | } | |
166ba7cd | 3902 | |
786e8c11 YO |
3903 | { |
3904 | U8 *uc = ( U8* )locinput; | |
3905 | ||
3906 | STRLEN len = 0; | |
3907 | STRLEN foldlen = 0; | |
3908 | U8 *uscan = (U8*)NULL; | |
786e8c11 | 3909 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
2e64971a DM |
3910 | U32 charcount = 0; /* how many input chars we have matched */ |
3911 | U32 accepted = 0; /* have we seen any accepting states? */ | |
786e8c11 | 3912 | |
786e8c11 | 3913 | ST.jump = trie->jump; |
786e8c11 | 3914 | ST.me = scan; |
2e64971a DM |
3915 | ST.firstpos = NULL; |
3916 | ST.longfold = FALSE; /* char longer if folded => it's harder */ | |
3917 | ST.nextword = 0; | |
3918 | ||
3919 | /* fully traverse the TRIE; note the position of the | |
3920 | shortest accept state and the wordnum of the longest | |
3921 | accept state */ | |
07be1b83 | 3922 | |
220db18a | 3923 | while ( state && uc <= (U8*)(reginfo->strend) ) { |
786e8c11 | 3924 | U32 base = trie->states[ state ].trans.base; |
f9f4320a | 3925 | UV uvc = 0; |
acb909b4 | 3926 | U16 charid = 0; |
2e64971a DM |
3927 | U16 wordnum; |
3928 | wordnum = trie->states[ state ].wordnum; | |
3929 | ||
3930 | if (wordnum) { /* it's an accept state */ | |
3931 | if (!accepted) { | |
3932 | accepted = 1; | |
3933 | /* record first match position */ | |
3934 | if (ST.longfold) { | |
3935 | ST.firstpos = (U8*)locinput; | |
3936 | ST.firstchars = 0; | |
5b47454d | 3937 | } |
2e64971a DM |
3938 | else { |
3939 | ST.firstpos = uc; | |
3940 | ST.firstchars = charcount; | |
3941 | } | |
3942 | } | |
3943 | if (!ST.nextword || wordnum < ST.nextword) | |
3944 | ST.nextword = wordnum; | |
3945 | ST.topword = wordnum; | |
786e8c11 | 3946 | } |
a3621e74 | 3947 | |
07be1b83 | 3948 | DEBUG_TRIE_EXECUTE_r({ |
f2ed9b32 | 3949 | DUMP_EXEC_POS( (char *)uc, scan, utf8_target ); |
a3621e74 | 3950 | PerlIO_printf( Perl_debug_log, |
2e64971a | 3951 | "%*s %sState: %4"UVxf" Accepted: %c ", |
5bc10b2c | 3952 | 2+depth * 2, "", PL_colors[4], |
2e64971a | 3953 | (UV)state, (accepted ? 'Y' : 'N')); |
07be1b83 | 3954 | }); |
a3621e74 | 3955 | |
2e64971a | 3956 | /* read a char and goto next state */ |
220db18a | 3957 | if ( base && (foldlen || uc < (U8*)(reginfo->strend))) { |
6dd2be57 | 3958 | I32 offset; |
55eed653 NC |
3959 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
3960 | uscan, len, uvc, charid, foldlen, | |
3961 | foldbuf, uniflags); | |
2e64971a DM |
3962 | charcount++; |
3963 | if (foldlen>0) | |
3964 | ST.longfold = TRUE; | |
5b47454d | 3965 | if (charid && |
6dd2be57 DM |
3966 | ( ((offset = |
3967 | base + charid - 1 - trie->uniquecharcount)) >= 0) | |
3968 | ||
3969 | && ((U32)offset < trie->lasttrans) | |
3970 | && trie->trans[offset].check == state) | |
5b47454d | 3971 | { |
6dd2be57 | 3972 | state = trie->trans[offset].next; |
5b47454d DM |
3973 | } |
3974 | else { | |
3975 | state = 0; | |
3976 | } | |
3977 | uc += len; | |
3978 | ||
3979 | } | |
3980 | else { | |
a3621e74 YO |
3981 | state = 0; |
3982 | } | |
3983 | DEBUG_TRIE_EXECUTE_r( | |
e4584336 | 3984 | PerlIO_printf( Perl_debug_log, |
786e8c11 | 3985 | "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n", |
e4584336 | 3986 | charid, uvc, (UV)state, PL_colors[5] ); |
a3621e74 YO |
3987 | ); |
3988 | } | |
2e64971a | 3989 | if (!accepted) |
a3621e74 | 3990 | sayNO; |
a3621e74 | 3991 | |
2e64971a DM |
3992 | /* calculate total number of accept states */ |
3993 | { | |
3994 | U16 w = ST.topword; | |
3995 | accepted = 0; | |
3996 | while (w) { | |
3997 | w = trie->wordinfo[w].prev; | |
3998 | accepted++; | |
3999 | } | |
4000 | ST.accepted = accepted; | |
4001 | } | |
4002 | ||
166ba7cd DM |
4003 | DEBUG_EXECUTE_r( |
4004 | PerlIO_printf( Perl_debug_log, | |
4005 | "%*s %sgot %"IVdf" possible matches%s\n", | |
5bc10b2c | 4006 | REPORT_CODE_OFF + depth * 2, "", |
166ba7cd DM |
4007 | PL_colors[4], (IV)ST.accepted, PL_colors[5] ); |
4008 | ); | |
2e64971a | 4009 | goto trie_first_try; /* jump into the fail handler */ |
786e8c11 | 4010 | }} |
118e2215 | 4011 | assert(0); /* NOTREACHED */ |
2e64971a DM |
4012 | |
4013 | case TRIE_next_fail: /* we failed - try next alternative */ | |
a059a757 DM |
4014 | { |
4015 | U8 *uc; | |
fae667d5 YO |
4016 | if ( ST.jump) { |
4017 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 | 4018 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
fae667d5 | 4019 | } |
2e64971a DM |
4020 | if (!--ST.accepted) { |
4021 | DEBUG_EXECUTE_r({ | |
4022 | PerlIO_printf( Perl_debug_log, | |
4023 | "%*s %sTRIE failed...%s\n", | |
4024 | REPORT_CODE_OFF+depth*2, "", | |
4025 | PL_colors[4], | |
4026 | PL_colors[5] ); | |
4027 | }); | |
4028 | sayNO_SILENT; | |
4029 | } | |
4030 | { | |
4031 | /* Find next-highest word to process. Note that this code | |
4032 | * is O(N^2) per trie run (O(N) per branch), so keep tight */ | |
eb578fdb KW |
4033 | U16 min = 0; |
4034 | U16 word; | |
4035 | U16 const nextword = ST.nextword; | |
4036 | reg_trie_wordinfo * const wordinfo | |
2e64971a DM |
4037 | = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo; |
4038 | for (word=ST.topword; word; word=wordinfo[word].prev) { | |
4039 | if (word > nextword && (!min || word < min)) | |
4040 | min = word; | |
4041 | } | |
4042 | ST.nextword = min; | |
4043 | } | |
4044 | ||
fae667d5 | 4045 | trie_first_try: |
5d458dd8 YO |
4046 | if (do_cutgroup) { |
4047 | do_cutgroup = 0; | |
4048 | no_final = 0; | |
4049 | } | |
fae667d5 YO |
4050 | |
4051 | if ( ST.jump) { | |
b93070ed | 4052 | ST.lastparen = rex->lastparen; |
f6033a9d | 4053 | ST.lastcloseparen = rex->lastcloseparen; |
fae667d5 | 4054 | REGCP_SET(ST.cp); |
2e64971a | 4055 | } |
a3621e74 | 4056 | |
2e64971a | 4057 | /* find start char of end of current word */ |
166ba7cd | 4058 | { |
2e64971a | 4059 | U32 chars; /* how many chars to skip */ |
2e64971a DM |
4060 | reg_trie_data * const trie |
4061 | = (reg_trie_data*)rexi->data->data[ARG(ST.me)]; | |
4062 | ||
4063 | assert((trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
4064 | >= ST.firstchars); | |
4065 | chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
4066 | - ST.firstchars; | |
a059a757 | 4067 | uc = ST.firstpos; |
2e64971a DM |
4068 | |
4069 | if (ST.longfold) { | |
4070 | /* the hard option - fold each char in turn and find | |
4071 | * its folded length (which may be different */ | |
4072 | U8 foldbuf[UTF8_MAXBYTES_CASE + 1]; | |
4073 | STRLEN foldlen; | |
4074 | STRLEN len; | |
d9a396a3 | 4075 | UV uvc; |
2e64971a DM |
4076 | U8 *uscan; |
4077 | ||
4078 | while (chars) { | |
f2ed9b32 | 4079 | if (utf8_target) { |
c80e42f3 | 4080 | uvc = utf8n_to_uvchr((U8*)uc, UTF8_MAXLEN, &len, |
2e64971a DM |
4081 | uniflags); |
4082 | uc += len; | |
4083 | } | |
4084 | else { | |
4085 | uvc = *uc; | |
4086 | uc++; | |
4087 | } | |
4088 | uvc = to_uni_fold(uvc, foldbuf, &foldlen); | |
4089 | uscan = foldbuf; | |
4090 | while (foldlen) { | |
4091 | if (!--chars) | |
4092 | break; | |
c80e42f3 | 4093 | uvc = utf8n_to_uvchr(uscan, UTF8_MAXLEN, &len, |
2e64971a DM |
4094 | uniflags); |
4095 | uscan += len; | |
4096 | foldlen -= len; | |
4097 | } | |
4098 | } | |
a3621e74 | 4099 | } |
2e64971a | 4100 | else { |
f2ed9b32 | 4101 | if (utf8_target) |
2e64971a DM |
4102 | while (chars--) |
4103 | uc += UTF8SKIP(uc); | |
4104 | else | |
4105 | uc += chars; | |
4106 | } | |
2e64971a | 4107 | } |
166ba7cd | 4108 | |
6603fe3e DM |
4109 | scan = ST.me + ((ST.jump && ST.jump[ST.nextword]) |
4110 | ? ST.jump[ST.nextword] | |
4111 | : NEXT_OFF(ST.me)); | |
166ba7cd | 4112 | |
2e64971a DM |
4113 | DEBUG_EXECUTE_r({ |
4114 | PerlIO_printf( Perl_debug_log, | |
4115 | "%*s %sTRIE matched word #%d, continuing%s\n", | |
4116 | REPORT_CODE_OFF+depth*2, "", | |
4117 | PL_colors[4], | |
4118 | ST.nextword, | |
4119 | PL_colors[5] | |
4120 | ); | |
4121 | }); | |
4122 | ||
4123 | if (ST.accepted > 1 || has_cutgroup) { | |
a059a757 | 4124 | PUSH_STATE_GOTO(TRIE_next, scan, (char*)uc); |
118e2215 | 4125 | assert(0); /* NOTREACHED */ |
166ba7cd | 4126 | } |
2e64971a DM |
4127 | /* only one choice left - just continue */ |
4128 | DEBUG_EXECUTE_r({ | |
4129 | AV *const trie_words | |
4130 | = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]); | |
4131 | SV ** const tmp = av_fetch( trie_words, | |
4132 | ST.nextword-1, 0 ); | |
4133 | SV *sv= tmp ? sv_newmortal() : NULL; | |
4134 | ||
4135 | PerlIO_printf( Perl_debug_log, | |
4136 | "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n", | |
4137 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], | |
4138 | ST.nextword, | |
4139 | tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0, | |
4140 | PL_colors[0], PL_colors[1], | |
c89df6cf | 4141 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII |
2e64971a DM |
4142 | ) |
4143 | : "not compiled under -Dr", | |
4144 | PL_colors[5] ); | |
4145 | }); | |
4146 | ||
a059a757 | 4147 | locinput = (char*)uc; |
2e64971a | 4148 | continue; /* execute rest of RE */ |
118e2215 | 4149 | assert(0); /* NOTREACHED */ |
a059a757 | 4150 | } |
166ba7cd DM |
4151 | #undef ST |
4152 | ||
3c0563b9 | 4153 | case EXACT: { /* /abc/ */ |
95b24440 | 4154 | char *s = STRING(scan); |
24d3c4a9 | 4155 | ln = STR_LEN(scan); |
984e6dd1 | 4156 | if (utf8_target != is_utf8_pat) { |
bc517b45 | 4157 | /* The target and the pattern have differing utf8ness. */ |
1aa99e6b | 4158 | char *l = locinput; |
24d3c4a9 | 4159 | const char * const e = s + ln; |
a72c7584 | 4160 | |
f2ed9b32 | 4161 | if (utf8_target) { |
e6a3850e KW |
4162 | /* The target is utf8, the pattern is not utf8. |
4163 | * Above-Latin1 code points can't match the pattern; | |
4164 | * invariants match exactly, and the other Latin1 ones need | |
4165 | * to be downgraded to a single byte in order to do the | |
4166 | * comparison. (If we could be confident that the target | |
4167 | * is not malformed, this could be refactored to have fewer | |
4168 | * tests by just assuming that if the first bytes match, it | |
4169 | * is an invariant, but there are tests in the test suite | |
4170 | * dealing with (??{...}) which violate this) */ | |
1aa99e6b | 4171 | while (s < e) { |
220db18a DM |
4172 | if (l >= reginfo->strend |
4173 | || UTF8_IS_ABOVE_LATIN1(* (U8*) l)) | |
4174 | { | |
e6a3850e KW |
4175 | sayNO; |
4176 | } | |
4177 | if (UTF8_IS_INVARIANT(*(U8*)l)) { | |
4178 | if (*l != *s) { | |
4179 | sayNO; | |
4180 | } | |
4181 | l++; | |
4182 | } | |
4183 | else { | |
94bb8c36 KW |
4184 | if (TWO_BYTE_UTF8_TO_NATIVE(*l, *(l+1)) != * (U8*) s) |
4185 | { | |
e6a3850e KW |
4186 | sayNO; |
4187 | } | |
4188 | l += 2; | |
4189 | } | |
4190 | s++; | |
1aa99e6b | 4191 | } |
5ff6fc6d JH |
4192 | } |
4193 | else { | |
4194 | /* The target is not utf8, the pattern is utf8. */ | |
1aa99e6b | 4195 | while (s < e) { |
220db18a DM |
4196 | if (l >= reginfo->strend |
4197 | || UTF8_IS_ABOVE_LATIN1(* (U8*) s)) | |
e6a3850e KW |
4198 | { |
4199 | sayNO; | |
4200 | } | |
4201 | if (UTF8_IS_INVARIANT(*(U8*)s)) { | |
4202 | if (*s != *l) { | |
4203 | sayNO; | |
4204 | } | |
4205 | s++; | |
4206 | } | |
4207 | else { | |
94bb8c36 KW |
4208 | if (TWO_BYTE_UTF8_TO_NATIVE(*s, *(s+1)) != * (U8*) l) |
4209 | { | |
e6a3850e KW |
4210 | sayNO; |
4211 | } | |
4212 | s += 2; | |
4213 | } | |
4214 | l++; | |
1aa99e6b | 4215 | } |
5ff6fc6d | 4216 | } |
1aa99e6b | 4217 | locinput = l; |
1aa99e6b | 4218 | } |
5ac65bff KW |
4219 | else { |
4220 | /* The target and the pattern have the same utf8ness. */ | |
4221 | /* Inline the first character, for speed. */ | |
220db18a | 4222 | if (reginfo->strend - locinput < ln |
5ac65bff KW |
4223 | || UCHARAT(s) != nextchr |
4224 | || (ln > 1 && memNE(s, locinput, ln))) | |
4225 | { | |
4226 | sayNO; | |
4227 | } | |
4228 | locinput += ln; | |
4229 | } | |
d6a28714 | 4230 | break; |
95b24440 | 4231 | } |
7016d6eb | 4232 | |
3c0563b9 | 4233 | case EXACTFL: { /* /abc/il */ |
a932d541 | 4234 | re_fold_t folder; |
9a5a5549 KW |
4235 | const U8 * fold_array; |
4236 | const char * s; | |
d513472c | 4237 | U32 fold_utf8_flags; |
9a5a5549 | 4238 | |
272d35c9 | 4239 | RX_MATCH_TAINTED_on(reginfo->prog); |
f67f9e53 KW |
4240 | folder = foldEQ_locale; |
4241 | fold_array = PL_fold_locale; | |
17580e7a | 4242 | fold_utf8_flags = FOLDEQ_UTF8_LOCALE; |
9a5a5549 KW |
4243 | goto do_exactf; |
4244 | ||
3c0563b9 | 4245 | case EXACTFU_SS: /* /\x{df}/iu */ |
3c0563b9 | 4246 | case EXACTFU: /* /abc/iu */ |
9a5a5549 KW |
4247 | folder = foldEQ_latin1; |
4248 | fold_array = PL_fold_latin1; | |
984e6dd1 | 4249 | fold_utf8_flags = is_utf8_pat ? FOLDEQ_S1_ALREADY_FOLDED : 0; |
9a5a5549 KW |
4250 | goto do_exactf; |
4251 | ||
098b07d5 KW |
4252 | case EXACTFA_NO_TRIE: /* This node only generated for non-utf8 |
4253 | patterns */ | |
4254 | assert(! is_utf8_pat); | |
4255 | /* FALL THROUGH */ | |
3c0563b9 | 4256 | case EXACTFA: /* /abc/iaa */ |
2f7f8cb1 KW |
4257 | folder = foldEQ_latin1; |
4258 | fold_array = PL_fold_latin1; | |
57014d77 | 4259 | fold_utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; |
2f7f8cb1 KW |
4260 | goto do_exactf; |
4261 | ||
2fdb7295 KW |
4262 | case EXACTF: /* /abc/i This node only generated for |
4263 | non-utf8 patterns */ | |
4264 | assert(! is_utf8_pat); | |
9a5a5549 KW |
4265 | folder = foldEQ; |
4266 | fold_array = PL_fold; | |
62bf7766 | 4267 | fold_utf8_flags = 0; |
9a5a5549 KW |
4268 | |
4269 | do_exactf: | |
4270 | s = STRING(scan); | |
24d3c4a9 | 4271 | ln = STR_LEN(scan); |
d6a28714 | 4272 | |
984e6dd1 | 4273 | if (utf8_target || is_utf8_pat || state_num == EXACTFU_SS) { |
3c760661 KW |
4274 | /* Either target or the pattern are utf8, or has the issue where |
4275 | * the fold lengths may differ. */ | |
be8e71aa | 4276 | const char * const l = locinput; |
220db18a | 4277 | char *e = reginfo->strend; |
bc517b45 | 4278 | |
984e6dd1 | 4279 | if (! foldEQ_utf8_flags(s, 0, ln, is_utf8_pat, |
fa5b1667 | 4280 | l, &e, 0, utf8_target, fold_utf8_flags)) |
c3e1d013 KW |
4281 | { |
4282 | sayNO; | |
5486206c | 4283 | } |
d07ddd77 | 4284 | locinput = e; |
d07ddd77 | 4285 | break; |
a0ed51b3 | 4286 | } |
d6a28714 | 4287 | |
0a138b74 | 4288 | /* Neither the target nor the pattern are utf8 */ |
1443c94c DM |
4289 | if (UCHARAT(s) != nextchr |
4290 | && !NEXTCHR_IS_EOS | |
4291 | && UCHARAT(s) != fold_array[nextchr]) | |
9a5a5549 | 4292 | { |
a0ed51b3 | 4293 | sayNO; |
9a5a5549 | 4294 | } |
220db18a | 4295 | if (reginfo->strend - locinput < ln) |
b8c5462f | 4296 | sayNO; |
9a5a5549 | 4297 | if (ln > 1 && ! folder(s, locinput, ln)) |
4633a7c4 | 4298 | sayNO; |
24d3c4a9 | 4299 | locinput += ln; |
a0d0e21e | 4300 | break; |
9a5a5549 | 4301 | } |
63ac0dad KW |
4302 | |
4303 | /* XXX Could improve efficiency by separating these all out using a | |
4304 | * macro or in-line function. At that point regcomp.c would no longer | |
4305 | * have to set the FLAGS fields of these */ | |
3c0563b9 DM |
4306 | case BOUNDL: /* /\b/l */ |
4307 | case NBOUNDL: /* /\B/l */ | |
272d35c9 | 4308 | RX_MATCH_TAINTED_on(reginfo->prog); |
b2680017 | 4309 | /* FALL THROUGH */ |
3c0563b9 DM |
4310 | case BOUND: /* /\b/ */ |
4311 | case BOUNDU: /* /\b/u */ | |
4312 | case BOUNDA: /* /\b/a */ | |
4313 | case NBOUND: /* /\B/ */ | |
4314 | case NBOUNDU: /* /\B/u */ | |
4315 | case NBOUNDA: /* /\B/a */ | |
b2680017 | 4316 | /* was last char in word? */ |
f2e96b5d KW |
4317 | if (utf8_target |
4318 | && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET | |
4319 | && FLAGS(scan) != REGEX_ASCII_MORE_RESTRICTED_CHARSET) | |
4320 | { | |
9d9163fb | 4321 | if (locinput == reginfo->strbeg) |
b2680017 YO |
4322 | ln = '\n'; |
4323 | else { | |
9d9163fb DM |
4324 | const U8 * const r = |
4325 | reghop3((U8*)locinput, -1, (U8*)(reginfo->strbeg)); | |
b2680017 YO |
4326 | |
4327 | ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); | |
4328 | } | |
63ac0dad | 4329 | if (FLAGS(scan) != REGEX_LOCALE_CHARSET) { |
0eb30aeb | 4330 | ln = isWORDCHAR_uni(ln); |
7016d6eb DM |
4331 | if (NEXTCHR_IS_EOS) |
4332 | n = 0; | |
4333 | else { | |
4334 | LOAD_UTF8_CHARCLASS_ALNUM(); | |
03940dc2 | 4335 | n = swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)locinput, |
7016d6eb DM |
4336 | utf8_target); |
4337 | } | |
b2680017 YO |
4338 | } |
4339 | else { | |
f4cd282c | 4340 | ln = isWORDCHAR_LC_uvchr(ln); |
0eb30aeb | 4341 | n = NEXTCHR_IS_EOS ? 0 : isWORDCHAR_LC_utf8((U8*)locinput); |
b2680017 YO |
4342 | } |
4343 | } | |
4344 | else { | |
cfaf538b KW |
4345 | |
4346 | /* Here the string isn't utf8, or is utf8 and only ascii | |
4347 | * characters are to match \w. In the latter case looking at | |
4348 | * the byte just prior to the current one may be just the final | |
4349 | * byte of a multi-byte character. This is ok. There are two | |
4350 | * cases: | |
4351 | * 1) it is a single byte character, and then the test is doing | |
4352 | * just what it's supposed to. | |
4353 | * 2) it is a multi-byte character, in which case the final | |
4354 | * byte is never mistakable for ASCII, and so the test | |
4355 | * will say it is not a word character, which is the | |
4356 | * correct answer. */ | |
9d9163fb | 4357 | ln = (locinput != reginfo->strbeg) ? |
b2680017 | 4358 | UCHARAT(locinput - 1) : '\n'; |
63ac0dad KW |
4359 | switch (FLAGS(scan)) { |
4360 | case REGEX_UNICODE_CHARSET: | |
4361 | ln = isWORDCHAR_L1(ln); | |
7016d6eb | 4362 | n = NEXTCHR_IS_EOS ? 0 : isWORDCHAR_L1(nextchr); |
63ac0dad KW |
4363 | break; |
4364 | case REGEX_LOCALE_CHARSET: | |
0eb30aeb KW |
4365 | ln = isWORDCHAR_LC(ln); |
4366 | n = NEXTCHR_IS_EOS ? 0 : isWORDCHAR_LC(nextchr); | |
63ac0dad KW |
4367 | break; |
4368 | case REGEX_DEPENDS_CHARSET: | |
0eb30aeb KW |
4369 | ln = isWORDCHAR(ln); |
4370 | n = NEXTCHR_IS_EOS ? 0 : isWORDCHAR(nextchr); | |
63ac0dad | 4371 | break; |
cfaf538b | 4372 | case REGEX_ASCII_RESTRICTED_CHARSET: |
c973bd4f | 4373 | case REGEX_ASCII_MORE_RESTRICTED_CHARSET: |
cfaf538b | 4374 | ln = isWORDCHAR_A(ln); |
7016d6eb | 4375 | n = NEXTCHR_IS_EOS ? 0 : isWORDCHAR_A(nextchr); |
cfaf538b | 4376 | break; |
63ac0dad KW |
4377 | default: |
4378 | Perl_croak(aTHX_ "panic: Unexpected FLAGS %u in op %u", FLAGS(scan), OP(scan)); | |
4379 | break; | |
b2680017 YO |
4380 | } |
4381 | } | |
63ac0dad KW |
4382 | /* Note requires that all BOUNDs be lower than all NBOUNDs in |
4383 | * regcomp.sym */ | |
4384 | if (((!ln) == (!n)) == (OP(scan) < NBOUND)) | |
b2680017 YO |
4385 | sayNO; |
4386 | break; | |
3c0563b9 | 4387 | |
3c0563b9 | 4388 | case ANYOF: /* /[abc]/ */ |
954a2af6 | 4389 | case ANYOF_WARN_SUPER: |
7016d6eb DM |
4390 | if (NEXTCHR_IS_EOS) |
4391 | sayNO; | |
e0193e47 | 4392 | if (utf8_target) { |
635cd5d4 | 4393 | if (!reginclass(rex, scan, (U8*)locinput, utf8_target)) |
09b08e9b | 4394 | sayNO; |
635cd5d4 | 4395 | locinput += UTF8SKIP(locinput); |
ffc61ed2 JH |
4396 | } |
4397 | else { | |
20ed0b26 | 4398 | if (!REGINCLASS(rex, scan, (U8*)locinput)) |
09b08e9b | 4399 | sayNO; |
3640db6b | 4400 | locinput++; |
e0f9d4a8 | 4401 | } |
b8c5462f | 4402 | break; |
3c0563b9 | 4403 | |
3018b823 KW |
4404 | /* The argument (FLAGS) to all the POSIX node types is the class number |
4405 | * */ | |
ee9a90b8 | 4406 | |
3018b823 KW |
4407 | case NPOSIXL: /* \W or [:^punct:] etc. under /l */ |
4408 | to_complement = 1; | |
4409 | /* FALLTHROUGH */ | |
4410 | ||
4411 | case POSIXL: /* \w or [:punct:] etc. under /l */ | |
4412 | if (NEXTCHR_IS_EOS) | |
bedac28b | 4413 | sayNO; |
bedac28b | 4414 | |
3018b823 KW |
4415 | /* The locale hasn't influenced the outcome before this, so defer |
4416 | * tainting until now */ | |
272d35c9 | 4417 | RX_MATCH_TAINTED_on(reginfo->prog); |
3018b823 KW |
4418 | |
4419 | /* Use isFOO_lc() for characters within Latin1. (Note that | |
4420 | * UTF8_IS_INVARIANT works even on non-UTF-8 strings, or else | |
4421 | * wouldn't be invariant) */ | |
4422 | if (UTF8_IS_INVARIANT(nextchr) || ! utf8_target) { | |
eb4e9c04 | 4423 | if (! (to_complement ^ cBOOL(isFOO_lc(FLAGS(scan), (U8) nextchr)))) { |
bedac28b KW |
4424 | sayNO; |
4425 | } | |
4426 | } | |
3018b823 KW |
4427 | else if (UTF8_IS_DOWNGRADEABLE_START(nextchr)) { |
4428 | if (! (to_complement ^ cBOOL(isFOO_lc(FLAGS(scan), | |
94bb8c36 | 4429 | (U8) TWO_BYTE_UTF8_TO_NATIVE(nextchr, |
3018b823 KW |
4430 | *(locinput + 1)))))) |
4431 | { | |
bedac28b | 4432 | sayNO; |
3018b823 | 4433 | } |
bedac28b | 4434 | } |
3018b823 KW |
4435 | else { /* Here, must be an above Latin-1 code point */ |
4436 | goto utf8_posix_not_eos; | |
bedac28b | 4437 | } |
3018b823 KW |
4438 | |
4439 | /* Here, must be utf8 */ | |
4440 | locinput += UTF8SKIP(locinput); | |
bedac28b KW |
4441 | break; |
4442 | ||
3018b823 KW |
4443 | case NPOSIXD: /* \W or [:^punct:] etc. under /d */ |
4444 | to_complement = 1; | |
4445 | /* FALLTHROUGH */ | |
4446 | ||
4447 | case POSIXD: /* \w or [:punct:] etc. under /d */ | |
bedac28b | 4448 | if (utf8_target) { |
3018b823 | 4449 | goto utf8_posix; |
bedac28b | 4450 | } |
3018b823 KW |
4451 | goto posixa; |
4452 | ||
4453 | case NPOSIXA: /* \W or [:^punct:] etc. under /a */ | |
bedac28b | 4454 | |
3018b823 | 4455 | if (NEXTCHR_IS_EOS) { |
bedac28b KW |
4456 | sayNO; |
4457 | } | |
bedac28b | 4458 | |
3018b823 KW |
4459 | /* All UTF-8 variants match */ |
4460 | if (! UTF8_IS_INVARIANT(nextchr)) { | |
4461 | goto increment_locinput; | |
bedac28b | 4462 | } |
ee9a90b8 | 4463 | |
3018b823 KW |
4464 | to_complement = 1; |
4465 | /* FALLTHROUGH */ | |
4466 | ||
4467 | case POSIXA: /* \w or [:punct:] etc. under /a */ | |
4468 | ||
4469 | posixa: | |
4470 | /* We get here through POSIXD, NPOSIXD, and NPOSIXA when not in | |
4471 | * UTF-8, and also from NPOSIXA even in UTF-8 when the current | |
4472 | * character is a single byte */ | |
20d0b1e9 | 4473 | |
3018b823 KW |
4474 | if (NEXTCHR_IS_EOS |
4475 | || ! (to_complement ^ cBOOL(_generic_isCC_A(nextchr, | |
4476 | FLAGS(scan))))) | |
4477 | { | |
0658cdde KW |
4478 | sayNO; |
4479 | } | |
3018b823 KW |
4480 | |
4481 | /* Here we are either not in utf8, or we matched a utf8-invariant, | |
4482 | * so the next char is the next byte */ | |
3640db6b | 4483 | locinput++; |
0658cdde | 4484 | break; |
3c0563b9 | 4485 | |
3018b823 KW |
4486 | case NPOSIXU: /* \W or [:^punct:] etc. under /u */ |
4487 | to_complement = 1; | |
4488 | /* FALLTHROUGH */ | |
4489 | ||
4490 | case POSIXU: /* \w or [:punct:] etc. under /u */ | |
4491 | utf8_posix: | |
4492 | if (NEXTCHR_IS_EOS) { | |
0658cdde KW |
4493 | sayNO; |
4494 | } | |
3018b823 KW |
4495 | utf8_posix_not_eos: |
4496 | ||
4497 | /* Use _generic_isCC() for characters within Latin1. (Note that | |
4498 | * UTF8_IS_INVARIANT works even on non-UTF-8 strings, or else | |
4499 | * wouldn't be invariant) */ | |
4500 | if (UTF8_IS_INVARIANT(nextchr) || ! utf8_target) { | |
4501 | if (! (to_complement ^ cBOOL(_generic_isCC(nextchr, | |
4502 | FLAGS(scan))))) | |
4503 | { | |
4504 | sayNO; | |
4505 | } | |
4506 | locinput++; | |
4507 | } | |
4508 | else if (UTF8_IS_DOWNGRADEABLE_START(nextchr)) { | |
4509 | if (! (to_complement | |
94bb8c36 | 4510 | ^ cBOOL(_generic_isCC(TWO_BYTE_UTF8_TO_NATIVE(nextchr, |
3018b823 | 4511 | *(locinput + 1)), |
94bb8c36 | 4512 | FLAGS(scan))))) |
3018b823 KW |
4513 | { |
4514 | sayNO; | |
4515 | } | |
4516 | locinput += 2; | |
4517 | } | |
4518 | else { /* Handle above Latin-1 code points */ | |
4519 | classnum = (_char_class_number) FLAGS(scan); | |
4520 | if (classnum < _FIRST_NON_SWASH_CC) { | |
4521 | ||
4522 | /* Here, uses a swash to find such code points. Load if if | |
4523 | * not done already */ | |
4524 | if (! PL_utf8_swash_ptrs[classnum]) { | |
4525 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; | |
4526 | PL_utf8_swash_ptrs[classnum] | |
4527 | = _core_swash_init("utf8", | |
4528 | swash_property_names[classnum], | |
4529 | &PL_sv_undef, 1, 0, NULL, &flags); | |
4530 | } | |
4531 | if (! (to_complement | |
4532 | ^ cBOOL(swash_fetch(PL_utf8_swash_ptrs[classnum], | |
4533 | (U8 *) locinput, TRUE)))) | |
4534 | { | |
4535 | sayNO; | |
4536 | } | |
4537 | } | |
4538 | else { /* Here, uses macros to find above Latin-1 code points */ | |
4539 | switch (classnum) { | |
4540 | case _CC_ENUM_SPACE: /* XXX would require separate | |
4541 | code if we revert the change | |
4542 | of \v matching this */ | |
4543 | case _CC_ENUM_PSXSPC: | |
4544 | if (! (to_complement | |
4545 | ^ cBOOL(is_XPERLSPACE_high(locinput)))) | |
4546 | { | |
4547 | sayNO; | |
4548 | } | |
4549 | break; | |
4550 | case _CC_ENUM_BLANK: | |
4551 | if (! (to_complement | |
4552 | ^ cBOOL(is_HORIZWS_high(locinput)))) | |
4553 | { | |
4554 | sayNO; | |
4555 | } | |
4556 | break; | |
4557 | case _CC_ENUM_XDIGIT: | |
4558 | if (! (to_complement | |
4559 | ^ cBOOL(is_XDIGIT_high(locinput)))) | |
4560 | { | |
4561 | sayNO; | |
4562 | } | |
4563 | break; | |
4564 | case _CC_ENUM_VERTSPACE: | |
4565 | if (! (to_complement | |
4566 | ^ cBOOL(is_VERTWS_high(locinput)))) | |
4567 | { | |
4568 | sayNO; | |
4569 | } | |
4570 | break; | |
4571 | default: /* The rest, e.g. [:cntrl:], can't match | |
4572 | above Latin1 */ | |
4573 | if (! to_complement) { | |
4574 | sayNO; | |
4575 | } | |
4576 | break; | |
4577 | } | |
4578 | } | |
4579 | locinput += UTF8SKIP(locinput); | |
4580 | } | |
4581 | break; | |
0658cdde | 4582 | |
37e2e78e KW |
4583 | case CLUMP: /* Match \X: logical Unicode character. This is defined as |
4584 | a Unicode extended Grapheme Cluster */ | |
4585 | /* From http://www.unicode.org/reports/tr29 (5.2 version). An | |
4586 | extended Grapheme Cluster is: | |
4587 | ||
7aee35ff KW |
4588 | CR LF |
4589 | | Prepend* Begin Extend* | |
4590 | | . | |
37e2e78e | 4591 | |
7aee35ff KW |
4592 | Begin is: ( Special_Begin | ! Control ) |
4593 | Special_Begin is: ( Regional-Indicator+ | Hangul-syllable ) | |
4594 | Extend is: ( Grapheme_Extend | Spacing_Mark ) | |
4595 | Control is: [ GCB_Control | CR | LF ] | |
4596 | Hangul-syllable is: ( T+ | ( L* ( L | ( LVT | ( V | LV ) V* ) T* ) )) | |
37e2e78e | 4597 | |
27d4fc33 KW |
4598 | If we create a 'Regular_Begin' = Begin - Special_Begin, then |
4599 | we can rewrite | |
4600 | ||
4601 | Begin is ( Regular_Begin + Special Begin ) | |
4602 | ||
4603 | It turns out that 98.4% of all Unicode code points match | |
4604 | Regular_Begin. Doing it this way eliminates a table match in | |
c101f46d | 4605 | the previous implementation for almost all Unicode code points. |
27d4fc33 | 4606 | |
37e2e78e KW |
4607 | There is a subtlety with Prepend* which showed up in testing. |
4608 | Note that the Begin, and only the Begin is required in: | |
4609 | | Prepend* Begin Extend* | |
cc3b396d KW |
4610 | Also, Begin contains '! Control'. A Prepend must be a |
4611 | '! Control', which means it must also be a Begin. What it | |
4612 | comes down to is that if we match Prepend* and then find no | |
4613 | suitable Begin afterwards, that if we backtrack the last | |
4614 | Prepend, that one will be a suitable Begin. | |
37e2e78e KW |
4615 | */ |
4616 | ||
7016d6eb | 4617 | if (NEXTCHR_IS_EOS) |
a0ed51b3 | 4618 | sayNO; |
f2ed9b32 | 4619 | if (! utf8_target) { |
37e2e78e KW |
4620 | |
4621 | /* Match either CR LF or '.', as all the other possibilities | |
4622 | * require utf8 */ | |
4623 | locinput++; /* Match the . or CR */ | |
cc3b396d KW |
4624 | if (nextchr == '\r' /* And if it was CR, and the next is LF, |
4625 | match the LF */ | |
220db18a | 4626 | && locinput < reginfo->strend |
e699a1d5 KW |
4627 | && UCHARAT(locinput) == '\n') |
4628 | { | |
4629 | locinput++; | |
4630 | } | |
37e2e78e KW |
4631 | } |
4632 | else { | |
4633 | ||
4634 | /* Utf8: See if is ( CR LF ); already know that locinput < | |
220db18a DM |
4635 | * reginfo->strend, so locinput+1 is in bounds */ |
4636 | if ( nextchr == '\r' && locinput+1 < reginfo->strend | |
f67f9e53 | 4637 | && UCHARAT(locinput + 1) == '\n') |
7016d6eb | 4638 | { |
37e2e78e KW |
4639 | locinput += 2; |
4640 | } | |
4641 | else { | |
45fdf108 KW |
4642 | STRLEN len; |
4643 | ||
37e2e78e KW |
4644 | /* In case have to backtrack to beginning, then match '.' */ |
4645 | char *starting = locinput; | |
4646 | ||
4647 | /* In case have to backtrack the last prepend */ | |
e699a1d5 | 4648 | char *previous_prepend = NULL; |
37e2e78e KW |
4649 | |
4650 | LOAD_UTF8_CHARCLASS_GCB(); | |
4651 | ||
45fdf108 | 4652 | /* Match (prepend)* */ |
220db18a | 4653 | while (locinput < reginfo->strend |
45fdf108 KW |
4654 | && (len = is_GCB_Prepend_utf8(locinput))) |
4655 | { | |
4656 | previous_prepend = locinput; | |
4657 | locinput += len; | |
a1853d78 | 4658 | } |
37e2e78e KW |
4659 | |
4660 | /* As noted above, if we matched a prepend character, but | |
4661 | * the next thing won't match, back off the last prepend we | |
4662 | * matched, as it is guaranteed to match the begin */ | |
4663 | if (previous_prepend | |
220db18a | 4664 | && (locinput >= reginfo->strend |
c101f46d KW |
4665 | || (! swash_fetch(PL_utf8_X_regular_begin, |
4666 | (U8*)locinput, utf8_target) | |
bff53399 | 4667 | && ! is_GCB_SPECIAL_BEGIN_START_utf8(locinput))) |
c101f46d | 4668 | ) |
37e2e78e KW |
4669 | { |
4670 | locinput = previous_prepend; | |
4671 | } | |
4672 | ||
220db18a | 4673 | /* Note that here we know reginfo->strend > locinput, as we |
37e2e78e KW |
4674 | * tested that upon input to this switch case, and if we |
4675 | * moved locinput forward, we tested the result just above | |
4676 | * and it either passed, or we backed off so that it will | |
4677 | * now pass */ | |
11dfcd49 KW |
4678 | if (swash_fetch(PL_utf8_X_regular_begin, |
4679 | (U8*)locinput, utf8_target)) { | |
27d4fc33 KW |
4680 | locinput += UTF8SKIP(locinput); |
4681 | } | |
bff53399 | 4682 | else if (! is_GCB_SPECIAL_BEGIN_START_utf8(locinput)) { |
37e2e78e KW |
4683 | |
4684 | /* Here did not match the required 'Begin' in the | |
4685 | * second term. So just match the very first | |
4686 | * character, the '.' of the final term of the regex */ | |
4687 | locinput = starting + UTF8SKIP(starting); | |
27d4fc33 | 4688 | goto exit_utf8; |
37e2e78e KW |
4689 | } else { |
4690 | ||
11dfcd49 KW |
4691 | /* Here is a special begin. It can be composed of |
4692 | * several individual characters. One possibility is | |
4693 | * RI+ */ | |
45fdf108 KW |
4694 | if ((len = is_GCB_RI_utf8(locinput))) { |
4695 | locinput += len; | |
220db18a | 4696 | while (locinput < reginfo->strend |
45fdf108 | 4697 | && (len = is_GCB_RI_utf8(locinput))) |
11dfcd49 | 4698 | { |
45fdf108 | 4699 | locinput += len; |
11dfcd49 | 4700 | } |
45fdf108 KW |
4701 | } else if ((len = is_GCB_T_utf8(locinput))) { |
4702 | /* Another possibility is T+ */ | |
4703 | locinput += len; | |
220db18a | 4704 | while (locinput < reginfo->strend |
45fdf108 | 4705 | && (len = is_GCB_T_utf8(locinput))) |
11dfcd49 | 4706 | { |
45fdf108 | 4707 | locinput += len; |
11dfcd49 KW |
4708 | } |
4709 | } else { | |
4710 | ||
4711 | /* Here, neither RI+ nor T+; must be some other | |
4712 | * Hangul. That means it is one of the others: L, | |
4713 | * LV, LVT or V, and matches: | |
4714 | * L* (L | LVT T* | V * V* T* | LV V* T*) */ | |
4715 | ||
4716 | /* Match L* */ | |
220db18a | 4717 | while (locinput < reginfo->strend |
45fdf108 | 4718 | && (len = is_GCB_L_utf8(locinput))) |
11dfcd49 | 4719 | { |
45fdf108 | 4720 | locinput += len; |
11dfcd49 | 4721 | } |
37e2e78e | 4722 | |
11dfcd49 KW |
4723 | /* Here, have exhausted L*. If the next character |
4724 | * is not an LV, LVT nor V, it means we had to have | |
4725 | * at least one L, so matches L+ in the original | |
4726 | * equation, we have a complete hangul syllable. | |
4727 | * Are done. */ | |
4728 | ||
220db18a | 4729 | if (locinput < reginfo->strend |
45fdf108 | 4730 | && is_GCB_LV_LVT_V_utf8(locinput)) |
11dfcd49 | 4731 | { |
11dfcd49 | 4732 | /* Otherwise keep going. Must be LV, LVT or V. |
7d43c479 KW |
4733 | * See if LVT, by first ruling out V, then LV */ |
4734 | if (! is_GCB_V_utf8(locinput) | |
4735 | /* All but every TCount one is LV */ | |
4736 | && (valid_utf8_to_uvchr((U8 *) locinput, | |
4737 | NULL) | |
4738 | - SBASE) | |
4739 | % TCount != 0) | |
4740 | { | |
11dfcd49 KW |
4741 | locinput += UTF8SKIP(locinput); |
4742 | } else { | |
4743 | ||
4744 | /* Must be V or LV. Take it, then match | |
4745 | * V* */ | |
4746 | locinput += UTF8SKIP(locinput); | |
220db18a | 4747 | while (locinput < reginfo->strend |
45fdf108 | 4748 | && (len = is_GCB_V_utf8(locinput))) |
11dfcd49 | 4749 | { |
45fdf108 | 4750 | locinput += len; |
11dfcd49 KW |
4751 | } |
4752 | } | |
37e2e78e | 4753 | |
11dfcd49 | 4754 | /* And any of LV, LVT, or V can be followed |
45fdf108 | 4755 | * by T* */ |
220db18a | 4756 | while (locinput < reginfo->strend |
45fdf108 | 4757 | && (len = is_GCB_T_utf8(locinput))) |
11dfcd49 | 4758 | { |
45fdf108 | 4759 | locinput += len; |
11dfcd49 KW |
4760 | } |
4761 | } | |
cd94d768 | 4762 | } |
11dfcd49 | 4763 | } |
37e2e78e | 4764 | |
11dfcd49 | 4765 | /* Match any extender */ |
220db18a | 4766 | while (locinput < reginfo->strend |
11dfcd49 KW |
4767 | && swash_fetch(PL_utf8_X_extend, |
4768 | (U8*)locinput, utf8_target)) | |
4769 | { | |
4770 | locinput += UTF8SKIP(locinput); | |
4771 | } | |
37e2e78e | 4772 | } |
27d4fc33 | 4773 | exit_utf8: |
220db18a | 4774 | if (locinput > reginfo->strend) sayNO; |
37e2e78e | 4775 | } |
a0ed51b3 | 4776 | break; |
81714fb9 | 4777 | |
3c0563b9 | 4778 | case NREFFL: /* /\g{name}/il */ |
d7ef4b73 KW |
4779 | { /* The capture buffer cases. The ones beginning with N for the |
4780 | named buffers just convert to the equivalent numbered and | |
4781 | pretend they were called as the corresponding numbered buffer | |
4782 | op. */ | |
26ecd678 TC |
4783 | /* don't initialize these in the declaration, it makes C++ |
4784 | unhappy */ | |
9d9163fb | 4785 | const char *s; |
ff1157ca | 4786 | char type; |
8368298a TC |
4787 | re_fold_t folder; |
4788 | const U8 *fold_array; | |
26ecd678 | 4789 | UV utf8_fold_flags; |
8368298a | 4790 | |
272d35c9 | 4791 | RX_MATCH_TAINTED_on(reginfo->prog); |
d7ef4b73 KW |
4792 | folder = foldEQ_locale; |
4793 | fold_array = PL_fold_locale; | |
4794 | type = REFFL; | |
17580e7a | 4795 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4796 | goto do_nref; |
4797 | ||
3c0563b9 | 4798 | case NREFFA: /* /\g{name}/iaa */ |
2f7f8cb1 KW |
4799 | folder = foldEQ_latin1; |
4800 | fold_array = PL_fold_latin1; | |
4801 | type = REFFA; | |
4802 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4803 | goto do_nref; | |
4804 | ||
3c0563b9 | 4805 | case NREFFU: /* /\g{name}/iu */ |
d7ef4b73 KW |
4806 | folder = foldEQ_latin1; |
4807 | fold_array = PL_fold_latin1; | |
4808 | type = REFFU; | |
d513472c | 4809 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4810 | goto do_nref; |
4811 | ||
3c0563b9 | 4812 | case NREFF: /* /\g{name}/i */ |
d7ef4b73 KW |
4813 | folder = foldEQ; |
4814 | fold_array = PL_fold; | |
4815 | type = REFF; | |
d513472c | 4816 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4817 | goto do_nref; |
4818 | ||
3c0563b9 | 4819 | case NREF: /* /\g{name}/ */ |
d7ef4b73 | 4820 | type = REF; |
83d7b90b KW |
4821 | folder = NULL; |
4822 | fold_array = NULL; | |
d513472c | 4823 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4824 | do_nref: |
4825 | ||
4826 | /* For the named back references, find the corresponding buffer | |
4827 | * number */ | |
0a4db386 YO |
4828 | n = reg_check_named_buff_matched(rex,scan); |
4829 | ||
d7ef4b73 | 4830 | if ( ! n ) { |
81714fb9 | 4831 | sayNO; |
d7ef4b73 KW |
4832 | } |
4833 | goto do_nref_ref_common; | |
4834 | ||
3c0563b9 | 4835 | case REFFL: /* /\1/il */ |
272d35c9 | 4836 | RX_MATCH_TAINTED_on(reginfo->prog); |
d7ef4b73 KW |
4837 | folder = foldEQ_locale; |
4838 | fold_array = PL_fold_locale; | |
17580e7a | 4839 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4840 | goto do_ref; |
4841 | ||
3c0563b9 | 4842 | case REFFA: /* /\1/iaa */ |
2f7f8cb1 KW |
4843 | folder = foldEQ_latin1; |
4844 | fold_array = PL_fold_latin1; | |
4845 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4846 | goto do_ref; | |
4847 | ||
3c0563b9 | 4848 | case REFFU: /* /\1/iu */ |
d7ef4b73 KW |
4849 | folder = foldEQ_latin1; |
4850 | fold_array = PL_fold_latin1; | |
d513472c | 4851 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4852 | goto do_ref; |
4853 | ||
3c0563b9 | 4854 | case REFF: /* /\1/i */ |
d7ef4b73 KW |
4855 | folder = foldEQ; |
4856 | fold_array = PL_fold; | |
d513472c | 4857 | utf8_fold_flags = 0; |
83d7b90b | 4858 | goto do_ref; |
d7ef4b73 | 4859 | |
3c0563b9 | 4860 | case REF: /* /\1/ */ |
83d7b90b KW |
4861 | folder = NULL; |
4862 | fold_array = NULL; | |
d513472c | 4863 | utf8_fold_flags = 0; |
83d7b90b | 4864 | |
d7ef4b73 | 4865 | do_ref: |
81714fb9 | 4866 | type = OP(scan); |
d7ef4b73 KW |
4867 | n = ARG(scan); /* which paren pair */ |
4868 | ||
4869 | do_nref_ref_common: | |
b93070ed | 4870 | ln = rex->offs[n].start; |
1cb48e53 | 4871 | reginfo->poscache_iter = reginfo->poscache_maxiter; /* Void cache */ |
b93070ed | 4872 | if (rex->lastparen < n || ln == -1) |
af3f8c16 | 4873 | sayNO; /* Do not match unless seen CLOSEn. */ |
b93070ed | 4874 | if (ln == rex->offs[n].end) |
a0d0e21e | 4875 | break; |
a0ed51b3 | 4876 | |
9d9163fb | 4877 | s = reginfo->strbeg + ln; |
d7ef4b73 | 4878 | if (type != REF /* REF can do byte comparison */ |
2f65c56d | 4879 | && (utf8_target || type == REFFU)) |
d7ef4b73 | 4880 | { /* XXX handle REFFL better */ |
220db18a | 4881 | char * limit = reginfo->strend; |
d7ef4b73 KW |
4882 | |
4883 | /* This call case insensitively compares the entire buffer | |
4884 | * at s, with the current input starting at locinput, but | |
220db18a DM |
4885 | * not going off the end given by reginfo->strend, and |
4886 | * returns in <limit> upon success, how much of the | |
4887 | * current input was matched */ | |
b93070ed | 4888 | if (! foldEQ_utf8_flags(s, NULL, rex->offs[n].end - ln, utf8_target, |
d513472c | 4889 | locinput, &limit, 0, utf8_target, utf8_fold_flags)) |
d7ef4b73 KW |
4890 | { |
4891 | sayNO; | |
a0ed51b3 | 4892 | } |
d7ef4b73 | 4893 | locinput = limit; |
a0ed51b3 LW |
4894 | break; |
4895 | } | |
4896 | ||
d7ef4b73 | 4897 | /* Not utf8: Inline the first character, for speed. */ |
7016d6eb DM |
4898 | if (!NEXTCHR_IS_EOS && |
4899 | UCHARAT(s) != nextchr && | |
81714fb9 | 4900 | (type == REF || |
d7ef4b73 | 4901 | UCHARAT(s) != fold_array[nextchr])) |
4633a7c4 | 4902 | sayNO; |
b93070ed | 4903 | ln = rex->offs[n].end - ln; |
220db18a | 4904 | if (locinput + ln > reginfo->strend) |
4633a7c4 | 4905 | sayNO; |
81714fb9 | 4906 | if (ln > 1 && (type == REF |
24d3c4a9 | 4907 | ? memNE(s, locinput, ln) |
d7ef4b73 | 4908 | : ! folder(s, locinput, ln))) |
4633a7c4 | 4909 | sayNO; |
24d3c4a9 | 4910 | locinput += ln; |
a0d0e21e | 4911 | break; |
81714fb9 | 4912 | } |
3c0563b9 DM |
4913 | |
4914 | case NOTHING: /* null op; e.g. the 'nothing' following | |
4915 | * the '*' in m{(a+|b)*}' */ | |
4916 | break; | |
4917 | case TAIL: /* placeholder while compiling (A|B|C) */ | |
a0d0e21e | 4918 | break; |
3c0563b9 DM |
4919 | |
4920 | case BACK: /* ??? doesn't appear to be used ??? */ | |
a0d0e21e | 4921 | break; |
40a82448 DM |
4922 | |
4923 | #undef ST | |
4924 | #define ST st->u.eval | |
c277df42 | 4925 | { |
c277df42 | 4926 | SV *ret; |
d2f13c59 | 4927 | REGEXP *re_sv; |
6bda09f9 | 4928 | regexp *re; |
f8fc2ecf | 4929 | regexp_internal *rei; |
1a147d38 YO |
4930 | regnode *startpoint; |
4931 | ||
3c0563b9 | 4932 | case GOSTART: /* (?R) */ |
e7707071 YO |
4933 | case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */ |
4934 | if (cur_eval && cur_eval->locinput==locinput) { | |
24b23f37 | 4935 | if (cur_eval->u.eval.close_paren == (U32)ARG(scan)) |
1a147d38 | 4936 | Perl_croak(aTHX_ "Infinite recursion in regex"); |
4b196cd4 | 4937 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 YO |
4938 | Perl_croak(aTHX_ |
4939 | "Pattern subroutine nesting without pos change" | |
4940 | " exceeded limit in regex"); | |
6bda09f9 YO |
4941 | } else { |
4942 | nochange_depth = 0; | |
1a147d38 | 4943 | } |
288b8c02 | 4944 | re_sv = rex_sv; |
6bda09f9 | 4945 | re = rex; |
f8fc2ecf | 4946 | rei = rexi; |
1a147d38 | 4947 | if (OP(scan)==GOSUB) { |
6bda09f9 YO |
4948 | startpoint = scan + ARG2L(scan); |
4949 | ST.close_paren = ARG(scan); | |
4950 | } else { | |
f8fc2ecf | 4951 | startpoint = rei->program+1; |
6bda09f9 YO |
4952 | ST.close_paren = 0; |
4953 | } | |
4954 | goto eval_recurse_doit; | |
118e2215 | 4955 | assert(0); /* NOTREACHED */ |
3c0563b9 | 4956 | |
6bda09f9 YO |
4957 | case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ |
4958 | if (cur_eval && cur_eval->locinput==locinput) { | |
4b196cd4 | 4959 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 | 4960 | Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex"); |
6bda09f9 YO |
4961 | } else { |
4962 | nochange_depth = 0; | |
4963 | } | |
8e5e9ebe | 4964 | { |
4aabdb9b | 4965 | /* execute the code in the {...} */ |
81ed78b2 | 4966 | |
4aabdb9b | 4967 | dSP; |
a6dc34f1 | 4968 | IV before; |
1f4d1a1e | 4969 | OP * const oop = PL_op; |
4aabdb9b | 4970 | COP * const ocurcop = PL_curcop; |
81ed78b2 | 4971 | OP *nop; |
81ed78b2 | 4972 | CV *newcv; |
91332126 | 4973 | |
74088413 | 4974 | /* save *all* paren positions */ |
92da3157 | 4975 | regcppush(rex, 0, maxopenparen); |
74088413 DM |
4976 | REGCP_SET(runops_cp); |
4977 | ||
81ed78b2 DM |
4978 | if (!caller_cv) |
4979 | caller_cv = find_runcv(NULL); | |
4980 | ||
4aabdb9b | 4981 | n = ARG(scan); |
81ed78b2 | 4982 | |
b30fcab9 | 4983 | if (rexi->data->what[n] == 'r') { /* code from an external qr */ |
8d919b0a | 4984 | newcv = (ReANY( |
b30fcab9 DM |
4985 | (REGEXP*)(rexi->data->data[n]) |
4986 | ))->qr_anoncv | |
81ed78b2 DM |
4987 | ; |
4988 | nop = (OP*)rexi->data->data[n+1]; | |
b30fcab9 DM |
4989 | } |
4990 | else if (rexi->data->what[n] == 'l') { /* literal code */ | |
81ed78b2 DM |
4991 | newcv = caller_cv; |
4992 | nop = (OP*)rexi->data->data[n]; | |
4993 | assert(CvDEPTH(newcv)); | |
68e2671b DM |
4994 | } |
4995 | else { | |
d24ca0c5 DM |
4996 | /* literal with own CV */ |
4997 | assert(rexi->data->what[n] == 'L'); | |
81ed78b2 DM |
4998 | newcv = rex->qr_anoncv; |
4999 | nop = (OP*)rexi->data->data[n]; | |
68e2671b | 5000 | } |
81ed78b2 | 5001 | |
0e458318 DM |
5002 | /* normally if we're about to execute code from the same |
5003 | * CV that we used previously, we just use the existing | |
5004 | * CX stack entry. However, its possible that in the | |
5005 | * meantime we may have backtracked, popped from the save | |
5006 | * stack, and undone the SAVECOMPPAD(s) associated with | |
5007 | * PUSH_MULTICALL; in which case PL_comppad no longer | |
5008 | * points to newcv's pad. */ | |
5009 | if (newcv != last_pushed_cv || PL_comppad != last_pad) | |
5010 | { | |
b0065247 DM |
5011 | U8 flags = (CXp_SUB_RE | |
5012 | ((newcv == caller_cv) ? CXp_SUB_RE_FAKE : 0)); | |
0e458318 | 5013 | if (last_pushed_cv) { |
b0065247 | 5014 | CHANGE_MULTICALL_FLAGS(newcv, flags); |
0e458318 DM |
5015 | } |
5016 | else { | |
b0065247 | 5017 | PUSH_MULTICALL_FLAGS(newcv, flags); |
0e458318 DM |
5018 | } |
5019 | last_pushed_cv = newcv; | |
5020 | } | |
c31ee3bb DM |
5021 | else { |
5022 | /* these assignments are just to silence compiler | |
5023 | * warnings */ | |
5024 | multicall_cop = NULL; | |
5025 | newsp = NULL; | |
5026 | } | |
0e458318 DM |
5027 | last_pad = PL_comppad; |
5028 | ||
2e2e3f36 DM |
5029 | /* the initial nextstate you would normally execute |
5030 | * at the start of an eval (which would cause error | |
5031 | * messages to come from the eval), may be optimised | |
5032 | * away from the execution path in the regex code blocks; | |
5033 | * so manually set PL_curcop to it initially */ | |
5034 | { | |
81ed78b2 | 5035 | OP *o = cUNOPx(nop)->op_first; |
2e2e3f36 DM |
5036 | assert(o->op_type == OP_NULL); |
5037 | if (o->op_targ == OP_SCOPE) { | |
5038 | o = cUNOPo->op_first; | |
5039 | } | |
5040 | else { | |
5041 | assert(o->op_targ == OP_LEAVE); | |
5042 | o = cUNOPo->op_first; | |
5043 | assert(o->op_type == OP_ENTER); | |
5044 | o = o->op_sibling; | |
5045 | } | |
5046 | ||
5047 | if (o->op_type != OP_STUB) { | |
5048 | assert( o->op_type == OP_NEXTSTATE | |
5049 | || o->op_type == OP_DBSTATE | |
5050 | || (o->op_type == OP_NULL | |
5051 | && ( o->op_targ == OP_NEXTSTATE | |
5052 | || o->op_targ == OP_DBSTATE | |
5053 | ) | |
5054 | ) | |
5055 | ); | |
5056 | PL_curcop = (COP*)o; | |
5057 | } | |
5058 | } | |
81ed78b2 | 5059 | nop = nop->op_next; |
2e2e3f36 | 5060 | |
24b23f37 | 5061 | DEBUG_STATE_r( PerlIO_printf(Perl_debug_log, |
81ed78b2 DM |
5062 | " re EVAL PL_op=0x%"UVxf"\n", PTR2UV(nop)) ); |
5063 | ||
8adc0f72 | 5064 | rex->offs[0].end = locinput - reginfo->strbeg; |
bf2039a9 | 5065 | if (reginfo->info_aux_eval->pos_magic) |
25fdce4a FC |
5066 | MgBYTEPOS_set(reginfo->info_aux_eval->pos_magic, |
5067 | reginfo->sv, reginfo->strbeg, | |
5068 | locinput - reginfo->strbeg); | |
4aabdb9b | 5069 | |
2bf803e2 YO |
5070 | if (sv_yes_mark) { |
5071 | SV *sv_mrk = get_sv("REGMARK", 1); | |
5072 | sv_setsv(sv_mrk, sv_yes_mark); | |
5073 | } | |
5074 | ||
81ed78b2 DM |
5075 | /* we don't use MULTICALL here as we want to call the |
5076 | * first op of the block of interest, rather than the | |
5077 | * first op of the sub */ | |
a6dc34f1 | 5078 | before = (IV)(SP-PL_stack_base); |
81ed78b2 | 5079 | PL_op = nop; |
8e5e9ebe RGS |
5080 | CALLRUNOPS(aTHX); /* Scalar context. */ |
5081 | SPAGAIN; | |
a6dc34f1 | 5082 | if ((IV)(SP-PL_stack_base) == before) |
075aa684 | 5083 | ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */ |
8e5e9ebe RGS |
5084 | else { |
5085 | ret = POPs; | |
5086 | PUTBACK; | |
5087 | } | |
4aabdb9b | 5088 | |
e4bfbed3 DM |
5089 | /* before restoring everything, evaluate the returned |
5090 | * value, so that 'uninit' warnings don't use the wrong | |
497d0a96 DM |
5091 | * PL_op or pad. Also need to process any magic vars |
5092 | * (e.g. $1) *before* parentheses are restored */ | |
e4bfbed3 DM |
5093 | |
5094 | PL_op = NULL; | |
5095 | ||
5e98dac2 | 5096 | re_sv = NULL; |
e4bfbed3 DM |
5097 | if (logical == 0) /* (?{})/ */ |
5098 | sv_setsv(save_scalar(PL_replgv), ret); /* $^R */ | |
5099 | else if (logical == 1) { /* /(?(?{...})X|Y)/ */ | |
5100 | sw = cBOOL(SvTRUE(ret)); | |
5101 | logical = 0; | |
5102 | } | |
5103 | else { /* /(??{}) */ | |
497d0a96 DM |
5104 | /* if its overloaded, let the regex compiler handle |
5105 | * it; otherwise extract regex, or stringify */ | |
237da807 | 5106 | if (SvGMAGICAL(ret)) |
2685dc2d | 5107 | ret = sv_mortalcopy(ret); |
497d0a96 DM |
5108 | if (!SvAMAGIC(ret)) { |
5109 | SV *sv = ret; | |
5110 | if (SvROK(sv)) | |
5111 | sv = SvRV(sv); | |
5112 | if (SvTYPE(sv) == SVt_REGEXP) | |
5113 | re_sv = (REGEXP*) sv; | |
63620942 FC |
5114 | else if (SvSMAGICAL(ret)) { |
5115 | MAGIC *mg = mg_find(ret, PERL_MAGIC_qr); | |
497d0a96 DM |
5116 | if (mg) |
5117 | re_sv = (REGEXP *) mg->mg_obj; | |
5118 | } | |
e4bfbed3 | 5119 | |
2685dc2d | 5120 | /* force any undef warnings here */ |
237da807 FC |
5121 | if (!re_sv && !SvPOK(ret) && !SvNIOK(ret)) { |
5122 | ret = sv_mortalcopy(ret); | |
497d0a96 DM |
5123 | (void) SvPV_force_nolen(ret); |
5124 | } | |
e4bfbed3 DM |
5125 | } |
5126 | ||
5127 | } | |
5128 | ||
81ed78b2 DM |
5129 | /* *** Note that at this point we don't restore |
5130 | * PL_comppad, (or pop the CxSUB) on the assumption it may | |
5131 | * be used again soon. This is safe as long as nothing | |
5132 | * in the regexp code uses the pad ! */ | |
4aabdb9b | 5133 | PL_op = oop; |
4aabdb9b | 5134 | PL_curcop = ocurcop; |
92da3157 | 5135 | S_regcp_restore(aTHX_ rex, runops_cp, &maxopenparen); |
f5df269c | 5136 | PL_curpm = PL_reg_curpm; |
e4bfbed3 DM |
5137 | |
5138 | if (logical != 2) | |
4aabdb9b | 5139 | break; |
8e5e9ebe | 5140 | } |
e4bfbed3 DM |
5141 | |
5142 | /* only /(??{})/ from now on */ | |
24d3c4a9 | 5143 | logical = 0; |
4aabdb9b | 5144 | { |
4f639d21 DM |
5145 | /* extract RE object from returned value; compiling if |
5146 | * necessary */ | |
5c35adbb | 5147 | |
575c37f6 DM |
5148 | if (re_sv) { |
5149 | re_sv = reg_temp_copy(NULL, re_sv); | |
288b8c02 | 5150 | } |
0f5d15d6 | 5151 | else { |
c737faaf | 5152 | U32 pm_flags = 0; |
0f5d15d6 | 5153 | |
9753d940 DM |
5154 | if (SvUTF8(ret) && IN_BYTES) { |
5155 | /* In use 'bytes': make a copy of the octet | |
5156 | * sequence, but without the flag on */ | |
b9ad30b4 NC |
5157 | STRLEN len; |
5158 | const char *const p = SvPV(ret, len); | |
5159 | ret = newSVpvn_flags(p, len, SVs_TEMP); | |
5160 | } | |
732caac7 DM |
5161 | if (rex->intflags & PREGf_USE_RE_EVAL) |
5162 | pm_flags |= PMf_USE_RE_EVAL; | |
5163 | ||
5164 | /* if we got here, it should be an engine which | |
5165 | * supports compiling code blocks and stuff */ | |
5166 | assert(rex->engine && rex->engine->op_comp); | |
ec841a27 | 5167 | assert(!(scan->flags & ~RXf_PMf_COMPILETIME)); |
575c37f6 | 5168 | re_sv = rex->engine->op_comp(aTHX_ &ret, 1, NULL, |
ec841a27 DM |
5169 | rex->engine, NULL, NULL, |
5170 | /* copy /msix etc to inner pattern */ | |
5171 | scan->flags, | |
5172 | pm_flags); | |
732caac7 | 5173 | |
9041c2e3 | 5174 | if (!(SvFLAGS(ret) |
237da807 FC |
5175 | & (SVs_TEMP | SVs_GMG | SVf_ROK)) |
5176 | && (!SvPADTMP(ret) || SvREADONLY(ret))) { | |
a2794585 NC |
5177 | /* This isn't a first class regexp. Instead, it's |
5178 | caching a regexp onto an existing, Perl visible | |
5179 | scalar. */ | |
575c37f6 | 5180 | sv_magic(ret, MUTABLE_SV(re_sv), PERL_MAGIC_qr, 0, 0); |
3ce3ed55 | 5181 | } |
0f5d15d6 | 5182 | } |
e1ff3a88 | 5183 | SAVEFREESV(re_sv); |
8d919b0a | 5184 | re = ReANY(re_sv); |
4aabdb9b | 5185 | } |
07bc277f | 5186 | RXp_MATCH_COPIED_off(re); |
28d8d7f4 YO |
5187 | re->subbeg = rex->subbeg; |
5188 | re->sublen = rex->sublen; | |
6502e081 DM |
5189 | re->suboffset = rex->suboffset; |
5190 | re->subcoffset = rex->subcoffset; | |
f8fc2ecf | 5191 | rei = RXi_GET(re); |
6bda09f9 | 5192 | DEBUG_EXECUTE_r( |
220db18a DM |
5193 | debug_start_match(re_sv, utf8_target, locinput, |
5194 | reginfo->strend, "Matching embedded"); | |
6bda09f9 | 5195 | ); |
f8fc2ecf | 5196 | startpoint = rei->program + 1; |
1a147d38 | 5197 | ST.close_paren = 0; /* only used for GOSUB */ |
aa283a38 | 5198 | |
1a147d38 | 5199 | eval_recurse_doit: /* Share code with GOSUB below this line */ |
aa283a38 | 5200 | /* run the pattern returned from (??{...}) */ |
92da3157 DM |
5201 | |
5202 | /* Save *all* the positions. */ | |
5203 | ST.cp = regcppush(rex, 0, maxopenparen); | |
40a82448 | 5204 | REGCP_SET(ST.lastcp); |
6bda09f9 | 5205 | |
0357f1fd ML |
5206 | re->lastparen = 0; |
5207 | re->lastcloseparen = 0; | |
5208 | ||
92da3157 | 5209 | maxopenparen = 0; |
4aabdb9b | 5210 | |
1cb95af7 DM |
5211 | /* invalidate the S-L poscache. We're now executing a |
5212 | * different set of WHILEM ops (and their associated | |
5213 | * indexes) against the same string, so the bits in the | |
5214 | * cache are meaningless. Setting maxiter to zero forces | |
5215 | * the cache to be invalidated and zeroed before reuse. | |
5216 | * XXX This is too dramatic a measure. Ideally we should | |
5217 | * save the old cache and restore when running the outer | |
5218 | * pattern again */ | |
1cb48e53 | 5219 | reginfo->poscache_maxiter = 0; |
4aabdb9b | 5220 | |
aed7b151 | 5221 | is_utf8_pat = reginfo->is_utf8_pat = cBOOL(RX_UTF8(re_sv)); |
faec1544 | 5222 | |
288b8c02 | 5223 | ST.prev_rex = rex_sv; |
faec1544 | 5224 | ST.prev_curlyx = cur_curlyx; |
ec43f78b DM |
5225 | rex_sv = re_sv; |
5226 | SET_reg_curpm(rex_sv); | |
288b8c02 | 5227 | rex = re; |
f8fc2ecf | 5228 | rexi = rei; |
faec1544 | 5229 | cur_curlyx = NULL; |
40a82448 | 5230 | ST.B = next; |
faec1544 DM |
5231 | ST.prev_eval = cur_eval; |
5232 | cur_eval = st; | |
faec1544 | 5233 | /* now continue from first node in postoned RE */ |
4d5016e5 | 5234 | PUSH_YES_STATE_GOTO(EVAL_AB, startpoint, locinput); |
118e2215 | 5235 | assert(0); /* NOTREACHED */ |
c277df42 | 5236 | } |
40a82448 | 5237 | |
faec1544 DM |
5238 | case EVAL_AB: /* cleanup after a successful (??{A})B */ |
5239 | /* note: this is called twice; first after popping B, then A */ | |
ec43f78b | 5240 | rex_sv = ST.prev_rex; |
aed7b151 | 5241 | is_utf8_pat = reginfo->is_utf8_pat = cBOOL(RX_UTF8(rex_sv)); |
ec43f78b | 5242 | SET_reg_curpm(rex_sv); |
8d919b0a | 5243 | rex = ReANY(rex_sv); |
f8fc2ecf | 5244 | rexi = RXi_GET(rex); |
faec1544 DM |
5245 | regcpblow(ST.cp); |
5246 | cur_eval = ST.prev_eval; | |
5247 | cur_curlyx = ST.prev_curlyx; | |
34a81e2b | 5248 | |
1cb95af7 | 5249 | /* Invalidate cache. See "invalidate" comment above. */ |
1cb48e53 | 5250 | reginfo->poscache_maxiter = 0; |
e7707071 | 5251 | if ( nochange_depth ) |
4b196cd4 | 5252 | nochange_depth--; |
262b90c4 | 5253 | sayYES; |
40a82448 | 5254 | |
40a82448 | 5255 | |
faec1544 DM |
5256 | case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */ |
5257 | /* note: this is called twice; first after popping B, then A */ | |
ec43f78b | 5258 | rex_sv = ST.prev_rex; |
aed7b151 | 5259 | is_utf8_pat = reginfo->is_utf8_pat = cBOOL(RX_UTF8(rex_sv)); |
ec43f78b | 5260 | SET_reg_curpm(rex_sv); |
8d919b0a | 5261 | rex = ReANY(rex_sv); |
f8fc2ecf | 5262 | rexi = RXi_GET(rex); |
0357f1fd | 5263 | |
40a82448 | 5264 | REGCP_UNWIND(ST.lastcp); |
92da3157 | 5265 | regcppop(rex, &maxopenparen); |
faec1544 DM |
5266 | cur_eval = ST.prev_eval; |
5267 | cur_curlyx = ST.prev_curlyx; | |
1cb95af7 | 5268 | /* Invalidate cache. See "invalidate" comment above. */ |
1cb48e53 | 5269 | reginfo->poscache_maxiter = 0; |
e7707071 | 5270 | if ( nochange_depth ) |
4b196cd4 | 5271 | nochange_depth--; |
40a82448 | 5272 | sayNO_SILENT; |
40a82448 DM |
5273 | #undef ST |
5274 | ||
3c0563b9 | 5275 | case OPEN: /* ( */ |
c277df42 | 5276 | n = ARG(scan); /* which paren pair */ |
9d9163fb | 5277 | rex->offs[n].start_tmp = locinput - reginfo->strbeg; |
92da3157 DM |
5278 | if (n > maxopenparen) |
5279 | maxopenparen = n; | |
495f47a5 | 5280 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
92da3157 | 5281 | "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf" tmp; maxopenparen=%"UVuf"\n", |
495f47a5 DM |
5282 | PTR2UV(rex), |
5283 | PTR2UV(rex->offs), | |
5284 | (UV)n, | |
5285 | (IV)rex->offs[n].start_tmp, | |
92da3157 | 5286 | (UV)maxopenparen |
495f47a5 | 5287 | )); |
e2e6a0f1 | 5288 | lastopen = n; |
a0d0e21e | 5289 | break; |
495f47a5 DM |
5290 | |
5291 | /* XXX really need to log other places start/end are set too */ | |
5292 | #define CLOSE_CAPTURE \ | |
5293 | rex->offs[n].start = rex->offs[n].start_tmp; \ | |
9d9163fb | 5294 | rex->offs[n].end = locinput - reginfo->strbeg; \ |
495f47a5 DM |
5295 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, \ |
5296 | "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf"..%"IVdf"\n", \ | |
5297 | PTR2UV(rex), \ | |
5298 | PTR2UV(rex->offs), \ | |
5299 | (UV)n, \ | |
5300 | (IV)rex->offs[n].start, \ | |
5301 | (IV)rex->offs[n].end \ | |
5302 | )) | |
5303 | ||
3c0563b9 | 5304 | case CLOSE: /* ) */ |
c277df42 | 5305 | n = ARG(scan); /* which paren pair */ |
495f47a5 | 5306 | CLOSE_CAPTURE; |
b93070ed DM |
5307 | if (n > rex->lastparen) |
5308 | rex->lastparen = n; | |
5309 | rex->lastcloseparen = n; | |
3b6647e0 | 5310 | if (cur_eval && cur_eval->u.eval.close_paren == n) { |
6bda09f9 YO |
5311 | goto fake_end; |
5312 | } | |
a0d0e21e | 5313 | break; |
3c0563b9 DM |
5314 | |
5315 | case ACCEPT: /* (*ACCEPT) */ | |
e2e6a0f1 YO |
5316 | if (ARG(scan)){ |
5317 | regnode *cursor; | |
5318 | for (cursor=scan; | |
5319 | cursor && OP(cursor)!=END; | |
5320 | cursor=regnext(cursor)) | |
5321 | { | |
5322 | if ( OP(cursor)==CLOSE ){ | |
5323 | n = ARG(cursor); | |
5324 | if ( n <= lastopen ) { | |
495f47a5 | 5325 | CLOSE_CAPTURE; |
b93070ed DM |
5326 | if (n > rex->lastparen) |
5327 | rex->lastparen = n; | |
5328 | rex->lastcloseparen = n; | |
3b6647e0 RB |
5329 | if ( n == ARG(scan) || (cur_eval && |
5330 | cur_eval->u.eval.close_paren == n)) | |
e2e6a0f1 YO |
5331 | break; |
5332 | } | |
5333 | } | |
5334 | } | |
5335 | } | |
5336 | goto fake_end; | |
5337 | /*NOTREACHED*/ | |
3c0563b9 DM |
5338 | |
5339 | case GROUPP: /* (?(1)) */ | |
c277df42 | 5340 | n = ARG(scan); /* which paren pair */ |
b93070ed | 5341 | sw = cBOOL(rex->lastparen >= n && rex->offs[n].end != -1); |
c277df42 | 5342 | break; |
3c0563b9 DM |
5343 | |
5344 | case NGROUPP: /* (?(<name>)) */ | |
0a4db386 | 5345 | /* reg_check_named_buff_matched returns 0 for no match */ |
f2338a2e | 5346 | sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan)); |
0a4db386 | 5347 | break; |
3c0563b9 DM |
5348 | |
5349 | case INSUBP: /* (?(R)) */ | |
0a4db386 | 5350 | n = ARG(scan); |
3b6647e0 | 5351 | sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n)); |
0a4db386 | 5352 | break; |
3c0563b9 DM |
5353 | |
5354 | case DEFINEP: /* (?(DEFINE)) */ | |
0a4db386 YO |
5355 | sw = 0; |
5356 | break; | |
3c0563b9 DM |
5357 | |
5358 | case IFTHEN: /* (?(cond)A|B) */ | |
1cb48e53 | 5359 | reginfo->poscache_iter = reginfo->poscache_maxiter; /* Void cache */ |
24d3c4a9 | 5360 | if (sw) |
c277df42 IZ |
5361 | next = NEXTOPER(NEXTOPER(scan)); |
5362 | else { | |
5363 | next = scan + ARG(scan); | |
5364 | if (OP(next) == IFTHEN) /* Fake one. */ | |
5365 | next = NEXTOPER(NEXTOPER(next)); | |
5366 | } | |
5367 | break; | |
3c0563b9 DM |
5368 | |
5369 | case LOGICAL: /* modifier for EVAL and IFMATCH */ | |
24d3c4a9 | 5370 | logical = scan->flags; |
c277df42 | 5371 | break; |
c476f425 | 5372 | |
2ab05381 | 5373 | /******************************************************************* |
2ab05381 | 5374 | |
c476f425 DM |
5375 | The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/ |
5376 | pattern, where A and B are subpatterns. (For simple A, CURLYM or | |
5377 | STAR/PLUS/CURLY/CURLYN are used instead.) | |
2ab05381 | 5378 | |
c476f425 | 5379 | A*B is compiled as <CURLYX><A><WHILEM><B> |
2ab05381 | 5380 | |
c476f425 DM |
5381 | On entry to the subpattern, CURLYX is called. This pushes a CURLYX |
5382 | state, which contains the current count, initialised to -1. It also sets | |
5383 | cur_curlyx to point to this state, with any previous value saved in the | |
5384 | state block. | |
2ab05381 | 5385 | |
c476f425 DM |
5386 | CURLYX then jumps straight to the WHILEM op, rather than executing A, |
5387 | since the pattern may possibly match zero times (i.e. it's a while {} loop | |
5388 | rather than a do {} while loop). | |
2ab05381 | 5389 | |
c476f425 DM |
5390 | Each entry to WHILEM represents a successful match of A. The count in the |
5391 | CURLYX block is incremented, another WHILEM state is pushed, and execution | |
5392 | passes to A or B depending on greediness and the current count. | |
2ab05381 | 5393 | |
c476f425 DM |
5394 | For example, if matching against the string a1a2a3b (where the aN are |
5395 | substrings that match /A/), then the match progresses as follows: (the | |
5396 | pushed states are interspersed with the bits of strings matched so far): | |
2ab05381 | 5397 | |
c476f425 DM |
5398 | <CURLYX cnt=-1> |
5399 | <CURLYX cnt=0><WHILEM> | |
5400 | <CURLYX cnt=1><WHILEM> a1 <WHILEM> | |
5401 | <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM> | |
5402 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> | |
5403 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b | |
2ab05381 | 5404 | |
c476f425 DM |
5405 | (Contrast this with something like CURLYM, which maintains only a single |
5406 | backtrack state: | |
2ab05381 | 5407 | |
c476f425 DM |
5408 | <CURLYM cnt=0> a1 |
5409 | a1 <CURLYM cnt=1> a2 | |
5410 | a1 a2 <CURLYM cnt=2> a3 | |
5411 | a1 a2 a3 <CURLYM cnt=3> b | |
5412 | ) | |
2ab05381 | 5413 | |
c476f425 DM |
5414 | Each WHILEM state block marks a point to backtrack to upon partial failure |
5415 | of A or B, and also contains some minor state data related to that | |
5416 | iteration. The CURLYX block, pointed to by cur_curlyx, contains the | |
5417 | overall state, such as the count, and pointers to the A and B ops. | |
2ab05381 | 5418 | |
c476f425 DM |
5419 | This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx |
5420 | must always point to the *current* CURLYX block, the rules are: | |
2ab05381 | 5421 | |
c476f425 DM |
5422 | When executing CURLYX, save the old cur_curlyx in the CURLYX state block, |
5423 | and set cur_curlyx to point the new block. | |
2ab05381 | 5424 | |
c476f425 DM |
5425 | When popping the CURLYX block after a successful or unsuccessful match, |
5426 | restore the previous cur_curlyx. | |
2ab05381 | 5427 | |
c476f425 DM |
5428 | When WHILEM is about to execute B, save the current cur_curlyx, and set it |
5429 | to the outer one saved in the CURLYX block. | |
2ab05381 | 5430 | |
c476f425 DM |
5431 | When popping the WHILEM block after a successful or unsuccessful B match, |
5432 | restore the previous cur_curlyx. | |
2ab05381 | 5433 | |
c476f425 DM |
5434 | Here's an example for the pattern (AI* BI)*BO |
5435 | I and O refer to inner and outer, C and W refer to CURLYX and WHILEM: | |
2ab05381 | 5436 | |
c476f425 DM |
5437 | cur_ |
5438 | curlyx backtrack stack | |
5439 | ------ --------------- | |
5440 | NULL | |
5441 | CO <CO prev=NULL> <WO> | |
5442 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
5443 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi | |
5444 | NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo | |
2ab05381 | 5445 | |
c476f425 DM |
5446 | At this point the pattern succeeds, and we work back down the stack to |
5447 | clean up, restoring as we go: | |
95b24440 | 5448 | |
c476f425 DM |
5449 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi |
5450 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
5451 | CO <CO prev=NULL> <WO> | |
5452 | NULL | |
a0374537 | 5453 | |
c476f425 DM |
5454 | *******************************************************************/ |
5455 | ||
5456 | #define ST st->u.curlyx | |
5457 | ||
5458 | case CURLYX: /* start of /A*B/ (for complex A) */ | |
5459 | { | |
5460 | /* No need to save/restore up to this paren */ | |
5461 | I32 parenfloor = scan->flags; | |
5462 | ||
5463 | assert(next); /* keep Coverity happy */ | |
5464 | if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ | |
5465 | next += ARG(next); | |
5466 | ||
5467 | /* XXXX Probably it is better to teach regpush to support | |
92da3157 | 5468 | parenfloor > maxopenparen ... */ |
b93070ed DM |
5469 | if (parenfloor > (I32)rex->lastparen) |
5470 | parenfloor = rex->lastparen; /* Pessimization... */ | |
c476f425 DM |
5471 | |
5472 | ST.prev_curlyx= cur_curlyx; | |
5473 | cur_curlyx = st; | |
5474 | ST.cp = PL_savestack_ix; | |
5475 | ||
5476 | /* these fields contain the state of the current curly. | |
5477 | * they are accessed by subsequent WHILEMs */ | |
5478 | ST.parenfloor = parenfloor; | |
d02d6d97 | 5479 | ST.me = scan; |
c476f425 | 5480 | ST.B = next; |
24d3c4a9 DM |
5481 | ST.minmod = minmod; |
5482 | minmod = 0; | |
c476f425 DM |
5483 | ST.count = -1; /* this will be updated by WHILEM */ |
5484 | ST.lastloc = NULL; /* this will be updated by WHILEM */ | |
5485 | ||
4d5016e5 | 5486 | PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next), locinput); |
118e2215 | 5487 | assert(0); /* NOTREACHED */ |
c476f425 | 5488 | } |
a0d0e21e | 5489 | |
c476f425 | 5490 | case CURLYX_end: /* just finished matching all of A*B */ |
c476f425 DM |
5491 | cur_curlyx = ST.prev_curlyx; |
5492 | sayYES; | |
118e2215 | 5493 | assert(0); /* NOTREACHED */ |
a0d0e21e | 5494 | |
c476f425 DM |
5495 | case CURLYX_end_fail: /* just failed to match all of A*B */ |
5496 | regcpblow(ST.cp); | |
5497 | cur_curlyx = ST.prev_curlyx; | |
5498 | sayNO; | |
118e2215 | 5499 | assert(0); /* NOTREACHED */ |
4633a7c4 | 5500 | |
a0d0e21e | 5501 | |
c476f425 DM |
5502 | #undef ST |
5503 | #define ST st->u.whilem | |
5504 | ||
5505 | case WHILEM: /* just matched an A in /A*B/ (for complex A) */ | |
5506 | { | |
5507 | /* see the discussion above about CURLYX/WHILEM */ | |
c476f425 | 5508 | I32 n; |
d02d6d97 DM |
5509 | int min = ARG1(cur_curlyx->u.curlyx.me); |
5510 | int max = ARG2(cur_curlyx->u.curlyx.me); | |
5511 | regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS; | |
5512 | ||
c476f425 DM |
5513 | assert(cur_curlyx); /* keep Coverity happy */ |
5514 | n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */ | |
5515 | ST.save_lastloc = cur_curlyx->u.curlyx.lastloc; | |
5516 | ST.cache_offset = 0; | |
5517 | ST.cache_mask = 0; | |
5518 | ||
c476f425 DM |
5519 | |
5520 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
d02d6d97 DM |
5521 | "%*s whilem: matched %ld out of %d..%d\n", |
5522 | REPORT_CODE_OFF+depth*2, "", (long)n, min, max) | |
c476f425 | 5523 | ); |
a0d0e21e | 5524 | |
c476f425 | 5525 | /* First just match a string of min A's. */ |
a0d0e21e | 5526 | |
d02d6d97 | 5527 | if (n < min) { |
92da3157 DM |
5528 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor, |
5529 | maxopenparen); | |
c476f425 | 5530 | cur_curlyx->u.curlyx.lastloc = locinput; |
92e82afa YO |
5531 | REGCP_SET(ST.lastcp); |
5532 | ||
4d5016e5 | 5533 | PUSH_STATE_GOTO(WHILEM_A_pre, A, locinput); |
118e2215 | 5534 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5535 | } |
5536 | ||
5537 | /* If degenerate A matches "", assume A done. */ | |
5538 | ||
5539 | if (locinput == cur_curlyx->u.curlyx.lastloc) { | |
5540 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
5541 | "%*s whilem: empty match detected, trying continuation...\n", | |
5542 | REPORT_CODE_OFF+depth*2, "") | |
5543 | ); | |
5544 | goto do_whilem_B_max; | |
5545 | } | |
5546 | ||
1cb95af7 DM |
5547 | /* super-linear cache processing. |
5548 | * | |
5549 | * The idea here is that for certain types of CURLYX/WHILEM - | |
5550 | * principally those whose upper bound is infinity (and | |
5551 | * excluding regexes that have things like \1 and other very | |
5552 | * non-regular expresssiony things), then if a pattern like | |
5553 | * /....A*.../ fails and we backtrack to the WHILEM, then we | |
5554 | * make a note that this particular WHILEM op was at string | |
5555 | * position 47 (say) when the rest of pattern failed. Then, if | |
5556 | * we ever find ourselves back at that WHILEM, and at string | |
5557 | * position 47 again, we can just fail immediately rather than | |
5558 | * running the rest of the pattern again. | |
5559 | * | |
5560 | * This is very handy when patterns start to go | |
5561 | * 'super-linear', like in (a+)*(a+)*(a+)*, where you end up | |
5562 | * with a combinatorial explosion of backtracking. | |
5563 | * | |
5564 | * The cache is implemented as a bit array, with one bit per | |
5565 | * string byte position per WHILEM op (up to 16) - so its | |
5566 | * between 0.25 and 2x the string size. | |
5567 | * | |
5568 | * To avoid allocating a poscache buffer every time, we do an | |
5569 | * initially countdown; only after we have executed a WHILEM | |
5570 | * op (string-length x #WHILEMs) times do we allocate the | |
5571 | * cache. | |
5572 | * | |
5573 | * The top 4 bits of scan->flags byte say how many different | |
5574 | * relevant CURLLYX/WHILEM op pairs there are, while the | |
5575 | * bottom 4-bits is the identifying index number of this | |
5576 | * WHILEM. | |
5577 | */ | |
c476f425 DM |
5578 | |
5579 | if (scan->flags) { | |
a0d0e21e | 5580 | |
1cb48e53 | 5581 | if (!reginfo->poscache_maxiter) { |
c476f425 DM |
5582 | /* start the countdown: Postpone detection until we |
5583 | * know the match is not *that* much linear. */ | |
1cb48e53 | 5584 | reginfo->poscache_maxiter |
9d9163fb DM |
5585 | = (reginfo->strend - reginfo->strbeg + 1) |
5586 | * (scan->flags>>4); | |
66bf836d | 5587 | /* possible overflow for long strings and many CURLYX's */ |
1cb48e53 DM |
5588 | if (reginfo->poscache_maxiter < 0) |
5589 | reginfo->poscache_maxiter = I32_MAX; | |
5590 | reginfo->poscache_iter = reginfo->poscache_maxiter; | |
2c2d71f5 | 5591 | } |
c476f425 | 5592 | |
1cb48e53 | 5593 | if (reginfo->poscache_iter-- == 0) { |
c476f425 | 5594 | /* initialise cache */ |
ea3daa5d | 5595 | const SSize_t size = (reginfo->poscache_maxiter + 7)/8; |
2ac8ff4b DM |
5596 | regmatch_info_aux *const aux = reginfo->info_aux; |
5597 | if (aux->poscache) { | |
ea3daa5d | 5598 | if ((SSize_t)reginfo->poscache_size < size) { |
2ac8ff4b DM |
5599 | Renew(aux->poscache, size, char); |
5600 | reginfo->poscache_size = size; | |
2c2d71f5 | 5601 | } |
2ac8ff4b | 5602 | Zero(aux->poscache, size, char); |
2c2d71f5 JH |
5603 | } |
5604 | else { | |
2ac8ff4b DM |
5605 | reginfo->poscache_size = size; |
5606 | Newxz(aux->poscache, size, char); | |
2c2d71f5 | 5607 | } |
c476f425 DM |
5608 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
5609 | "%swhilem: Detected a super-linear match, switching on caching%s...\n", | |
5610 | PL_colors[4], PL_colors[5]) | |
5611 | ); | |
2c2d71f5 | 5612 | } |
c476f425 | 5613 | |
1cb48e53 | 5614 | if (reginfo->poscache_iter < 0) { |
c476f425 | 5615 | /* have we already failed at this position? */ |
ea3daa5d | 5616 | SSize_t offset, mask; |
338e600a DM |
5617 | |
5618 | reginfo->poscache_iter = -1; /* stop eventual underflow */ | |
c476f425 | 5619 | offset = (scan->flags & 0xf) - 1 |
9d9163fb DM |
5620 | + (locinput - reginfo->strbeg) |
5621 | * (scan->flags>>4); | |
c476f425 DM |
5622 | mask = 1 << (offset % 8); |
5623 | offset /= 8; | |
2ac8ff4b | 5624 | if (reginfo->info_aux->poscache[offset] & mask) { |
c476f425 DM |
5625 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
5626 | "%*s whilem: (cache) already tried at this position...\n", | |
5627 | REPORT_CODE_OFF+depth*2, "") | |
2c2d71f5 | 5628 | ); |
3298f257 | 5629 | sayNO; /* cache records failure */ |
2c2d71f5 | 5630 | } |
c476f425 DM |
5631 | ST.cache_offset = offset; |
5632 | ST.cache_mask = mask; | |
2c2d71f5 | 5633 | } |
c476f425 | 5634 | } |
2c2d71f5 | 5635 | |
c476f425 | 5636 | /* Prefer B over A for minimal matching. */ |
a687059c | 5637 | |
c476f425 DM |
5638 | if (cur_curlyx->u.curlyx.minmod) { |
5639 | ST.save_curlyx = cur_curlyx; | |
5640 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
92da3157 DM |
5641 | ST.cp = regcppush(rex, ST.save_curlyx->u.curlyx.parenfloor, |
5642 | maxopenparen); | |
c476f425 | 5643 | REGCP_SET(ST.lastcp); |
4d5016e5 DM |
5644 | PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B, |
5645 | locinput); | |
118e2215 | 5646 | assert(0); /* NOTREACHED */ |
c476f425 | 5647 | } |
a0d0e21e | 5648 | |
c476f425 DM |
5649 | /* Prefer A over B for maximal matching. */ |
5650 | ||
d02d6d97 | 5651 | if (n < max) { /* More greed allowed? */ |
92da3157 DM |
5652 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor, |
5653 | maxopenparen); | |
c476f425 DM |
5654 | cur_curlyx->u.curlyx.lastloc = locinput; |
5655 | REGCP_SET(ST.lastcp); | |
4d5016e5 | 5656 | PUSH_STATE_GOTO(WHILEM_A_max, A, locinput); |
118e2215 | 5657 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5658 | } |
5659 | goto do_whilem_B_max; | |
5660 | } | |
118e2215 | 5661 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5662 | |
5663 | case WHILEM_B_min: /* just matched B in a minimal match */ | |
5664 | case WHILEM_B_max: /* just matched B in a maximal match */ | |
5665 | cur_curlyx = ST.save_curlyx; | |
5666 | sayYES; | |
118e2215 | 5667 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5668 | |
5669 | case WHILEM_B_max_fail: /* just failed to match B in a maximal match */ | |
5670 | cur_curlyx = ST.save_curlyx; | |
5671 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
5672 | cur_curlyx->u.curlyx.count--; | |
5673 | CACHEsayNO; | |
118e2215 | 5674 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5675 | |
5676 | case WHILEM_A_min_fail: /* just failed to match A in a minimal match */ | |
c476f425 DM |
5677 | /* FALL THROUGH */ |
5678 | case WHILEM_A_pre_fail: /* just failed to match even minimal A */ | |
92e82afa | 5679 | REGCP_UNWIND(ST.lastcp); |
92da3157 | 5680 | regcppop(rex, &maxopenparen); |
c476f425 DM |
5681 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; |
5682 | cur_curlyx->u.curlyx.count--; | |
5683 | CACHEsayNO; | |
118e2215 | 5684 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5685 | |
5686 | case WHILEM_A_max_fail: /* just failed to match A in a maximal match */ | |
5687 | REGCP_UNWIND(ST.lastcp); | |
92da3157 | 5688 | regcppop(rex, &maxopenparen); /* Restore some previous $<digit>s? */ |
c476f425 DM |
5689 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
5690 | "%*s whilem: failed, trying continuation...\n", | |
5691 | REPORT_CODE_OFF+depth*2, "") | |
5692 | ); | |
5693 | do_whilem_B_max: | |
5694 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
5695 | && ckWARN(WARN_REGEXP) | |
39819bd9 | 5696 | && !reginfo->warned) |
c476f425 | 5697 | { |
39819bd9 | 5698 | reginfo->warned = TRUE; |
dcbac5bb FC |
5699 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
5700 | "Complex regular subexpression recursion limit (%d) " | |
5701 | "exceeded", | |
c476f425 DM |
5702 | REG_INFTY - 1); |
5703 | } | |
5704 | ||
5705 | /* now try B */ | |
5706 | ST.save_curlyx = cur_curlyx; | |
5707 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4d5016e5 DM |
5708 | PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B, |
5709 | locinput); | |
118e2215 | 5710 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5711 | |
5712 | case WHILEM_B_min_fail: /* just failed to match B in a minimal match */ | |
5713 | cur_curlyx = ST.save_curlyx; | |
5714 | REGCP_UNWIND(ST.lastcp); | |
92da3157 | 5715 | regcppop(rex, &maxopenparen); |
c476f425 | 5716 | |
d02d6d97 | 5717 | if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) { |
c476f425 DM |
5718 | /* Maximum greed exceeded */ |
5719 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
5720 | && ckWARN(WARN_REGEXP) | |
39819bd9 | 5721 | && !reginfo->warned) |
c476f425 | 5722 | { |
39819bd9 | 5723 | reginfo->warned = TRUE; |
c476f425 | 5724 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
dcbac5bb FC |
5725 | "Complex regular subexpression recursion " |
5726 | "limit (%d) exceeded", | |
c476f425 | 5727 | REG_INFTY - 1); |
a0d0e21e | 5728 | } |
c476f425 | 5729 | cur_curlyx->u.curlyx.count--; |
3ab3c9b4 | 5730 | CACHEsayNO; |
a0d0e21e | 5731 | } |
c476f425 DM |
5732 | |
5733 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
5734 | "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "") | |
5735 | ); | |
5736 | /* Try grabbing another A and see if it helps. */ | |
c476f425 | 5737 | cur_curlyx->u.curlyx.lastloc = locinput; |
92da3157 DM |
5738 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor, |
5739 | maxopenparen); | |
c476f425 | 5740 | REGCP_SET(ST.lastcp); |
d02d6d97 | 5741 | PUSH_STATE_GOTO(WHILEM_A_min, |
4d5016e5 DM |
5742 | /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS, |
5743 | locinput); | |
118e2215 | 5744 | assert(0); /* NOTREACHED */ |
40a82448 DM |
5745 | |
5746 | #undef ST | |
5747 | #define ST st->u.branch | |
5748 | ||
5749 | case BRANCHJ: /* /(...|A|...)/ with long next pointer */ | |
c277df42 IZ |
5750 | next = scan + ARG(scan); |
5751 | if (next == scan) | |
5752 | next = NULL; | |
40a82448 DM |
5753 | scan = NEXTOPER(scan); |
5754 | /* FALL THROUGH */ | |
c277df42 | 5755 | |
40a82448 DM |
5756 | case BRANCH: /* /(...|A|...)/ */ |
5757 | scan = NEXTOPER(scan); /* scan now points to inner node */ | |
b93070ed | 5758 | ST.lastparen = rex->lastparen; |
f6033a9d | 5759 | ST.lastcloseparen = rex->lastcloseparen; |
40a82448 DM |
5760 | ST.next_branch = next; |
5761 | REGCP_SET(ST.cp); | |
02db2b7b | 5762 | |
40a82448 | 5763 | /* Now go into the branch */ |
5d458dd8 | 5764 | if (has_cutgroup) { |
4d5016e5 | 5765 | PUSH_YES_STATE_GOTO(BRANCH_next, scan, locinput); |
5d458dd8 | 5766 | } else { |
4d5016e5 | 5767 | PUSH_STATE_GOTO(BRANCH_next, scan, locinput); |
5d458dd8 | 5768 | } |
118e2215 | 5769 | assert(0); /* NOTREACHED */ |
3c0563b9 DM |
5770 | |
5771 | case CUTGROUP: /* /(*THEN)/ */ | |
5d458dd8 | 5772 | sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : |
ad64d0ec | 5773 | MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
4d5016e5 | 5774 | PUSH_STATE_GOTO(CUTGROUP_next, next, locinput); |
118e2215 | 5775 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5776 | |
5d458dd8 YO |
5777 | case CUTGROUP_next_fail: |
5778 | do_cutgroup = 1; | |
5779 | no_final = 1; | |
5780 | if (st->u.mark.mark_name) | |
5781 | sv_commit = st->u.mark.mark_name; | |
5782 | sayNO; | |
118e2215 | 5783 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5784 | |
5d458dd8 YO |
5785 | case BRANCH_next: |
5786 | sayYES; | |
118e2215 | 5787 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5788 | |
40a82448 | 5789 | case BRANCH_next_fail: /* that branch failed; try the next, if any */ |
5d458dd8 YO |
5790 | if (do_cutgroup) { |
5791 | do_cutgroup = 0; | |
5792 | no_final = 0; | |
5793 | } | |
40a82448 | 5794 | REGCP_UNWIND(ST.cp); |
a8d1f4b4 | 5795 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
40a82448 DM |
5796 | scan = ST.next_branch; |
5797 | /* no more branches? */ | |
5d458dd8 YO |
5798 | if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) { |
5799 | DEBUG_EXECUTE_r({ | |
5800 | PerlIO_printf( Perl_debug_log, | |
5801 | "%*s %sBRANCH failed...%s\n", | |
5802 | REPORT_CODE_OFF+depth*2, "", | |
5803 | PL_colors[4], | |
5804 | PL_colors[5] ); | |
5805 | }); | |
5806 | sayNO_SILENT; | |
5807 | } | |
40a82448 | 5808 | continue; /* execute next BRANCH[J] op */ |
118e2215 | 5809 | assert(0); /* NOTREACHED */ |
40a82448 | 5810 | |
3c0563b9 | 5811 | case MINMOD: /* next op will be non-greedy, e.g. A*? */ |
24d3c4a9 | 5812 | minmod = 1; |
a0d0e21e | 5813 | break; |
40a82448 DM |
5814 | |
5815 | #undef ST | |
5816 | #define ST st->u.curlym | |
5817 | ||
5818 | case CURLYM: /* /A{m,n}B/ where A is fixed-length */ | |
5819 | ||
5820 | /* This is an optimisation of CURLYX that enables us to push | |
84d2fa14 | 5821 | * only a single backtracking state, no matter how many matches |
40a82448 DM |
5822 | * there are in {m,n}. It relies on the pattern being constant |
5823 | * length, with no parens to influence future backrefs | |
5824 | */ | |
5825 | ||
5826 | ST.me = scan; | |
dc45a647 | 5827 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; |
40a82448 | 5828 | |
f6033a9d DM |
5829 | ST.lastparen = rex->lastparen; |
5830 | ST.lastcloseparen = rex->lastcloseparen; | |
5831 | ||
40a82448 DM |
5832 | /* if paren positive, emulate an OPEN/CLOSE around A */ |
5833 | if (ST.me->flags) { | |
3b6647e0 | 5834 | U32 paren = ST.me->flags; |
92da3157 DM |
5835 | if (paren > maxopenparen) |
5836 | maxopenparen = paren; | |
c277df42 | 5837 | scan += NEXT_OFF(scan); /* Skip former OPEN. */ |
6407bf3b | 5838 | } |
40a82448 DM |
5839 | ST.A = scan; |
5840 | ST.B = next; | |
5841 | ST.alen = 0; | |
5842 | ST.count = 0; | |
24d3c4a9 DM |
5843 | ST.minmod = minmod; |
5844 | minmod = 0; | |
40a82448 DM |
5845 | ST.c1 = CHRTEST_UNINIT; |
5846 | REGCP_SET(ST.cp); | |
6407bf3b | 5847 | |
40a82448 DM |
5848 | if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */ |
5849 | goto curlym_do_B; | |
5850 | ||
5851 | curlym_do_A: /* execute the A in /A{m,n}B/ */ | |
4d5016e5 | 5852 | PUSH_YES_STATE_GOTO(CURLYM_A, ST.A, locinput); /* match A */ |
118e2215 | 5853 | assert(0); /* NOTREACHED */ |
5f80c4cf | 5854 | |
40a82448 | 5855 | case CURLYM_A: /* we've just matched an A */ |
40a82448 DM |
5856 | ST.count++; |
5857 | /* after first match, determine A's length: u.curlym.alen */ | |
5858 | if (ST.count == 1) { | |
ba44c216 | 5859 | if (reginfo->is_utf8_target) { |
c07e9d7b DM |
5860 | char *s = st->locinput; |
5861 | while (s < locinput) { | |
40a82448 DM |
5862 | ST.alen++; |
5863 | s += UTF8SKIP(s); | |
5864 | } | |
5865 | } | |
5866 | else { | |
c07e9d7b | 5867 | ST.alen = locinput - st->locinput; |
40a82448 DM |
5868 | } |
5869 | if (ST.alen == 0) | |
5870 | ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me); | |
5871 | } | |
0cadcf80 DM |
5872 | DEBUG_EXECUTE_r( |
5873 | PerlIO_printf(Perl_debug_log, | |
40a82448 | 5874 | "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n", |
5bc10b2c | 5875 | (int)(REPORT_CODE_OFF+(depth*2)), "", |
40a82448 | 5876 | (IV) ST.count, (IV)ST.alen) |
0cadcf80 DM |
5877 | ); |
5878 | ||
0a4db386 | 5879 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 5880 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5881 | goto fake_end; |
5882 | ||
c966426a DM |
5883 | { |
5884 | I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)); | |
5885 | if ( max == REG_INFTY || ST.count < max ) | |
5886 | goto curlym_do_A; /* try to match another A */ | |
5887 | } | |
40a82448 | 5888 | goto curlym_do_B; /* try to match B */ |
5f80c4cf | 5889 | |
40a82448 DM |
5890 | case CURLYM_A_fail: /* just failed to match an A */ |
5891 | REGCP_UNWIND(ST.cp); | |
0a4db386 YO |
5892 | |
5893 | if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ | |
5894 | || (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 5895 | cur_eval->u.eval.close_paren == (U32)ST.me->flags)) |
40a82448 | 5896 | sayNO; |
0cadcf80 | 5897 | |
40a82448 | 5898 | curlym_do_B: /* execute the B in /A{m,n}B/ */ |
40a82448 DM |
5899 | if (ST.c1 == CHRTEST_UNINIT) { |
5900 | /* calculate c1 and c2 for possible match of 1st char | |
5901 | * following curly */ | |
5902 | ST.c1 = ST.c2 = CHRTEST_VOID; | |
5903 | if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) { | |
5904 | regnode *text_node = ST.B; | |
5905 | if (! HAS_TEXT(text_node)) | |
5906 | FIND_NEXT_IMPT(text_node); | |
ee9b8eae YO |
5907 | /* this used to be |
5908 | ||
5909 | (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT) | |
5910 | ||
5911 | But the former is redundant in light of the latter. | |
5912 | ||
5913 | if this changes back then the macro for | |
5914 | IS_TEXT and friends need to change. | |
5915 | */ | |
c74f6de9 | 5916 | if (PL_regkind[OP(text_node)] == EXACT) { |
79a2a0e8 | 5917 | if (! S_setup_EXACTISH_ST_c1_c2(aTHX_ |
984e6dd1 | 5918 | text_node, &ST.c1, ST.c1_utf8, &ST.c2, ST.c2_utf8, |
aed7b151 | 5919 | reginfo)) |
c74f6de9 KW |
5920 | { |
5921 | sayNO; | |
5922 | } | |
c277df42 | 5923 | } |
c277df42 | 5924 | } |
40a82448 DM |
5925 | } |
5926 | ||
5927 | DEBUG_EXECUTE_r( | |
5928 | PerlIO_printf(Perl_debug_log, | |
5929 | "%*s CURLYM trying tail with matches=%"IVdf"...\n", | |
5bc10b2c | 5930 | (int)(REPORT_CODE_OFF+(depth*2)), |
40a82448 DM |
5931 | "", (IV)ST.count) |
5932 | ); | |
c74f6de9 | 5933 | if (! NEXTCHR_IS_EOS && ST.c1 != CHRTEST_VOID) { |
79a2a0e8 KW |
5934 | if (! UTF8_IS_INVARIANT(nextchr) && utf8_target) { |
5935 | if (memNE(locinput, ST.c1_utf8, UTF8SKIP(locinput)) | |
5936 | && memNE(locinput, ST.c2_utf8, UTF8SKIP(locinput))) | |
5937 | { | |
5938 | /* simulate B failing */ | |
5939 | DEBUG_OPTIMISE_r( | |
5940 | PerlIO_printf(Perl_debug_log, | |
33daa3a5 | 5941 | "%*s CURLYM Fast bail next target=0x%"UVXf" c1=0x%"UVXf" c2=0x%"UVXf"\n", |
79a2a0e8 KW |
5942 | (int)(REPORT_CODE_OFF+(depth*2)),"", |
5943 | valid_utf8_to_uvchr((U8 *) locinput, NULL), | |
5944 | valid_utf8_to_uvchr(ST.c1_utf8, NULL), | |
5945 | valid_utf8_to_uvchr(ST.c2_utf8, NULL)) | |
5946 | ); | |
5947 | state_num = CURLYM_B_fail; | |
5948 | goto reenter_switch; | |
5949 | } | |
5950 | } | |
5951 | else if (nextchr != ST.c1 && nextchr != ST.c2) { | |
5400f398 KW |
5952 | /* simulate B failing */ |
5953 | DEBUG_OPTIMISE_r( | |
5954 | PerlIO_printf(Perl_debug_log, | |
33daa3a5 | 5955 | "%*s CURLYM Fast bail next target=0x%X c1=0x%X c2=0x%X\n", |
5400f398 | 5956 | (int)(REPORT_CODE_OFF+(depth*2)),"", |
79a2a0e8 KW |
5957 | (int) nextchr, ST.c1, ST.c2) |
5958 | ); | |
5400f398 KW |
5959 | state_num = CURLYM_B_fail; |
5960 | goto reenter_switch; | |
5961 | } | |
c74f6de9 | 5962 | } |
40a82448 DM |
5963 | |
5964 | if (ST.me->flags) { | |
f6033a9d | 5965 | /* emulate CLOSE: mark current A as captured */ |
40a82448 DM |
5966 | I32 paren = ST.me->flags; |
5967 | if (ST.count) { | |
b93070ed | 5968 | rex->offs[paren].start |
9d9163fb DM |
5969 | = HOPc(locinput, -ST.alen) - reginfo->strbeg; |
5970 | rex->offs[paren].end = locinput - reginfo->strbeg; | |
f6033a9d DM |
5971 | if ((U32)paren > rex->lastparen) |
5972 | rex->lastparen = paren; | |
5973 | rex->lastcloseparen = paren; | |
c277df42 | 5974 | } |
40a82448 | 5975 | else |
b93070ed | 5976 | rex->offs[paren].end = -1; |
0a4db386 | 5977 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 5978 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5979 | { |
5980 | if (ST.count) | |
5981 | goto fake_end; | |
5982 | else | |
5983 | sayNO; | |
5984 | } | |
c277df42 | 5985 | } |
0a4db386 | 5986 | |
4d5016e5 | 5987 | PUSH_STATE_GOTO(CURLYM_B, ST.B, locinput); /* match B */ |
118e2215 | 5988 | assert(0); /* NOTREACHED */ |
40a82448 DM |
5989 | |
5990 | case CURLYM_B_fail: /* just failed to match a B */ | |
5991 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 | 5992 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
40a82448 | 5993 | if (ST.minmod) { |
84d2fa14 HS |
5994 | I32 max = ARG2(ST.me); |
5995 | if (max != REG_INFTY && ST.count == max) | |
40a82448 DM |
5996 | sayNO; |
5997 | goto curlym_do_A; /* try to match a further A */ | |
5998 | } | |
5999 | /* backtrack one A */ | |
6000 | if (ST.count == ARG1(ST.me) /* min */) | |
6001 | sayNO; | |
6002 | ST.count--; | |
7016d6eb | 6003 | SET_locinput(HOPc(locinput, -ST.alen)); |
40a82448 DM |
6004 | goto curlym_do_B; /* try to match B */ |
6005 | ||
c255a977 DM |
6006 | #undef ST |
6007 | #define ST st->u.curly | |
40a82448 | 6008 | |
c255a977 DM |
6009 | #define CURLY_SETPAREN(paren, success) \ |
6010 | if (paren) { \ | |
6011 | if (success) { \ | |
9d9163fb DM |
6012 | rex->offs[paren].start = HOPc(locinput, -1) - reginfo->strbeg; \ |
6013 | rex->offs[paren].end = locinput - reginfo->strbeg; \ | |
f6033a9d DM |
6014 | if (paren > rex->lastparen) \ |
6015 | rex->lastparen = paren; \ | |
b93070ed | 6016 | rex->lastcloseparen = paren; \ |
c255a977 | 6017 | } \ |
f6033a9d | 6018 | else { \ |
b93070ed | 6019 | rex->offs[paren].end = -1; \ |
f6033a9d DM |
6020 | rex->lastparen = ST.lastparen; \ |
6021 | rex->lastcloseparen = ST.lastcloseparen; \ | |
6022 | } \ | |
c255a977 DM |
6023 | } |
6024 | ||
b40a2c17 | 6025 | case STAR: /* /A*B/ where A is width 1 char */ |
c255a977 DM |
6026 | ST.paren = 0; |
6027 | ST.min = 0; | |
6028 | ST.max = REG_INFTY; | |
a0d0e21e LW |
6029 | scan = NEXTOPER(scan); |
6030 | goto repeat; | |
3c0563b9 | 6031 | |
b40a2c17 | 6032 | case PLUS: /* /A+B/ where A is width 1 char */ |
c255a977 DM |
6033 | ST.paren = 0; |
6034 | ST.min = 1; | |
6035 | ST.max = REG_INFTY; | |
c277df42 | 6036 | scan = NEXTOPER(scan); |
c255a977 | 6037 | goto repeat; |
3c0563b9 | 6038 | |
b40a2c17 | 6039 | case CURLYN: /* /(A){m,n}B/ where A is width 1 char */ |
5400f398 KW |
6040 | ST.paren = scan->flags; /* Which paren to set */ |
6041 | ST.lastparen = rex->lastparen; | |
f6033a9d | 6042 | ST.lastcloseparen = rex->lastcloseparen; |
92da3157 DM |
6043 | if (ST.paren > maxopenparen) |
6044 | maxopenparen = ST.paren; | |
c255a977 DM |
6045 | ST.min = ARG1(scan); /* min to match */ |
6046 | ST.max = ARG2(scan); /* max to match */ | |
0a4db386 | 6047 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 6048 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
6049 | ST.min=1; |
6050 | ST.max=1; | |
6051 | } | |
c255a977 DM |
6052 | scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); |
6053 | goto repeat; | |
3c0563b9 | 6054 | |
b40a2c17 | 6055 | case CURLY: /* /A{m,n}B/ where A is width 1 char */ |
c255a977 DM |
6056 | ST.paren = 0; |
6057 | ST.min = ARG1(scan); /* min to match */ | |
6058 | ST.max = ARG2(scan); /* max to match */ | |
6059 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; | |
c277df42 | 6060 | repeat: |
a0d0e21e LW |
6061 | /* |
6062 | * Lookahead to avoid useless match attempts | |
6063 | * when we know what character comes next. | |
c255a977 | 6064 | * |
5f80c4cf JP |
6065 | * Used to only do .*x and .*?x, but now it allows |
6066 | * for )'s, ('s and (?{ ... })'s to be in the way | |
6067 | * of the quantifier and the EXACT-like node. -- japhy | |
6068 | */ | |
6069 | ||
eb5c1be8 | 6070 | assert(ST.min <= ST.max); |
3337dfe3 KW |
6071 | if (! HAS_TEXT(next) && ! JUMPABLE(next)) { |
6072 | ST.c1 = ST.c2 = CHRTEST_VOID; | |
6073 | } | |
6074 | else { | |
5f80c4cf JP |
6075 | regnode *text_node = next; |
6076 | ||
3dab1dad YO |
6077 | if (! HAS_TEXT(text_node)) |
6078 | FIND_NEXT_IMPT(text_node); | |
5f80c4cf | 6079 | |
9e137952 | 6080 | if (! HAS_TEXT(text_node)) |
c255a977 | 6081 | ST.c1 = ST.c2 = CHRTEST_VOID; |
5f80c4cf | 6082 | else { |
ee9b8eae | 6083 | if ( PL_regkind[OP(text_node)] != EXACT ) { |
c255a977 | 6084 | ST.c1 = ST.c2 = CHRTEST_VOID; |
cca55fe3 | 6085 | } |
c74f6de9 | 6086 | else { |
ee9b8eae YO |
6087 | |
6088 | /* Currently we only get here when | |
6089 | ||
6090 | PL_rekind[OP(text_node)] == EXACT | |
6091 | ||
6092 | if this changes back then the macro for IS_TEXT and | |
6093 | friends need to change. */ | |
79a2a0e8 | 6094 | if (! S_setup_EXACTISH_ST_c1_c2(aTHX_ |
984e6dd1 | 6095 | text_node, &ST.c1, ST.c1_utf8, &ST.c2, ST.c2_utf8, |
aed7b151 | 6096 | reginfo)) |
c74f6de9 KW |
6097 | { |
6098 | sayNO; | |
6099 | } | |
6100 | } | |
1aa99e6b | 6101 | } |
bbce6d69 | 6102 | } |
c255a977 DM |
6103 | |
6104 | ST.A = scan; | |
6105 | ST.B = next; | |
24d3c4a9 | 6106 | if (minmod) { |
eb72505d | 6107 | char *li = locinput; |
24d3c4a9 | 6108 | minmod = 0; |
984e6dd1 | 6109 | if (ST.min && |
f9176b44 | 6110 | regrepeat(rex, &li, ST.A, reginfo, ST.min, depth) |
984e6dd1 | 6111 | < ST.min) |
4633a7c4 | 6112 | sayNO; |
7016d6eb | 6113 | SET_locinput(li); |
c255a977 | 6114 | ST.count = ST.min; |
c255a977 DM |
6115 | REGCP_SET(ST.cp); |
6116 | if (ST.c1 == CHRTEST_VOID) | |
6117 | goto curly_try_B_min; | |
6118 | ||
6119 | ST.oldloc = locinput; | |
6120 | ||
6121 | /* set ST.maxpos to the furthest point along the | |
6122 | * string that could possibly match */ | |
6123 | if (ST.max == REG_INFTY) { | |
220db18a | 6124 | ST.maxpos = reginfo->strend - 1; |
f2ed9b32 | 6125 | if (utf8_target) |
c255a977 DM |
6126 | while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos)) |
6127 | ST.maxpos--; | |
6128 | } | |
f2ed9b32 | 6129 | else if (utf8_target) { |
c255a977 DM |
6130 | int m = ST.max - ST.min; |
6131 | for (ST.maxpos = locinput; | |
220db18a | 6132 | m >0 && ST.maxpos < reginfo->strend; m--) |
c255a977 DM |
6133 | ST.maxpos += UTF8SKIP(ST.maxpos); |
6134 | } | |
6135 | else { | |
6136 | ST.maxpos = locinput + ST.max - ST.min; | |
220db18a DM |
6137 | if (ST.maxpos >= reginfo->strend) |
6138 | ST.maxpos = reginfo->strend - 1; | |
c255a977 DM |
6139 | } |
6140 | goto curly_try_B_min_known; | |
6141 | ||
6142 | } | |
6143 | else { | |
eb72505d DM |
6144 | /* avoid taking address of locinput, so it can remain |
6145 | * a register var */ | |
6146 | char *li = locinput; | |
f9176b44 | 6147 | ST.count = regrepeat(rex, &li, ST.A, reginfo, ST.max, depth); |
c255a977 DM |
6148 | if (ST.count < ST.min) |
6149 | sayNO; | |
7016d6eb | 6150 | SET_locinput(li); |
c255a977 DM |
6151 | if ((ST.count > ST.min) |
6152 | && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL)) | |
6153 | { | |
6154 | /* A{m,n} must come at the end of the string, there's | |
6155 | * no point in backing off ... */ | |
6156 | ST.min = ST.count; | |
6157 | /* ...except that $ and \Z can match before *and* after | |
6158 | newline at the end. Consider "\n\n" =~ /\n+\Z\n/. | |
6159 | We may back off by one in this case. */ | |
eb72505d | 6160 | if (UCHARAT(locinput - 1) == '\n' && OP(ST.B) != EOS) |
c255a977 DM |
6161 | ST.min--; |
6162 | } | |
6163 | REGCP_SET(ST.cp); | |
6164 | goto curly_try_B_max; | |
6165 | } | |
118e2215 | 6166 | assert(0); /* NOTREACHED */ |
c255a977 DM |
6167 | |
6168 | ||
6169 | case CURLY_B_min_known_fail: | |
6170 | /* failed to find B in a non-greedy match where c1,c2 valid */ | |
c255a977 | 6171 | |
c255a977 | 6172 | REGCP_UNWIND(ST.cp); |
a8d1f4b4 DM |
6173 | if (ST.paren) { |
6174 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
6175 | } | |
c255a977 DM |
6176 | /* Couldn't or didn't -- move forward. */ |
6177 | ST.oldloc = locinput; | |
f2ed9b32 | 6178 | if (utf8_target) |
c255a977 DM |
6179 | locinput += UTF8SKIP(locinput); |
6180 | else | |
6181 | locinput++; | |
6182 | ST.count++; | |
6183 | curly_try_B_min_known: | |
6184 | /* find the next place where 'B' could work, then call B */ | |
6185 | { | |
6186 | int n; | |
f2ed9b32 | 6187 | if (utf8_target) { |
c255a977 DM |
6188 | n = (ST.oldloc == locinput) ? 0 : 1; |
6189 | if (ST.c1 == ST.c2) { | |
c255a977 | 6190 | /* set n to utf8_distance(oldloc, locinput) */ |
79a2a0e8 KW |
6191 | while (locinput <= ST.maxpos |
6192 | && memNE(locinput, ST.c1_utf8, UTF8SKIP(locinput))) | |
6193 | { | |
6194 | locinput += UTF8SKIP(locinput); | |
c255a977 DM |
6195 | n++; |
6196 | } | |
1aa99e6b IH |
6197 | } |
6198 | else { | |
c255a977 | 6199 | /* set n to utf8_distance(oldloc, locinput) */ |
79a2a0e8 KW |
6200 | while (locinput <= ST.maxpos |
6201 | && memNE(locinput, ST.c1_utf8, UTF8SKIP(locinput)) | |
6202 | && memNE(locinput, ST.c2_utf8, UTF8SKIP(locinput))) | |
6203 | { | |
6204 | locinput += UTF8SKIP(locinput); | |
c255a977 | 6205 | n++; |
1aa99e6b | 6206 | } |
0fe9bf95 IZ |
6207 | } |
6208 | } | |
5400f398 | 6209 | else { /* Not utf8_target */ |
c255a977 DM |
6210 | if (ST.c1 == ST.c2) { |
6211 | while (locinput <= ST.maxpos && | |
6212 | UCHARAT(locinput) != ST.c1) | |
6213 | locinput++; | |
bbce6d69 | 6214 | } |
c255a977 DM |
6215 | else { |
6216 | while (locinput <= ST.maxpos | |
6217 | && UCHARAT(locinput) != ST.c1 | |
6218 | && UCHARAT(locinput) != ST.c2) | |
6219 | locinput++; | |
a0ed51b3 | 6220 | } |
c255a977 DM |
6221 | n = locinput - ST.oldloc; |
6222 | } | |
6223 | if (locinput > ST.maxpos) | |
6224 | sayNO; | |
c255a977 | 6225 | if (n) { |
eb72505d DM |
6226 | /* In /a{m,n}b/, ST.oldloc is at "a" x m, locinput is |
6227 | * at b; check that everything between oldloc and | |
6228 | * locinput matches */ | |
6229 | char *li = ST.oldloc; | |
c255a977 | 6230 | ST.count += n; |
f9176b44 | 6231 | if (regrepeat(rex, &li, ST.A, reginfo, n, depth) < n) |
4633a7c4 | 6232 | sayNO; |
eb72505d | 6233 | assert(n == REG_INFTY || locinput == li); |
a0d0e21e | 6234 | } |
c255a977 | 6235 | CURLY_SETPAREN(ST.paren, ST.count); |
0a4db386 | 6236 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 6237 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
6238 | goto fake_end; |
6239 | } | |
4d5016e5 | 6240 | PUSH_STATE_GOTO(CURLY_B_min_known, ST.B, locinput); |
a0d0e21e | 6241 | } |
118e2215 | 6242 | assert(0); /* NOTREACHED */ |
c255a977 DM |
6243 | |
6244 | ||
6245 | case CURLY_B_min_fail: | |
6246 | /* failed to find B in a non-greedy match where c1,c2 invalid */ | |
c255a977 DM |
6247 | |
6248 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 DM |
6249 | if (ST.paren) { |
6250 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
6251 | } | |
c255a977 | 6252 | /* failed -- move forward one */ |
f73aaa43 | 6253 | { |
eb72505d | 6254 | char *li = locinput; |
f9176b44 | 6255 | if (!regrepeat(rex, &li, ST.A, reginfo, 1, depth)) { |
f73aaa43 DM |
6256 | sayNO; |
6257 | } | |
eb72505d | 6258 | locinput = li; |
f73aaa43 DM |
6259 | } |
6260 | { | |
c255a977 | 6261 | ST.count++; |
c255a977 DM |
6262 | if (ST.count <= ST.max || (ST.max == REG_INFTY && |
6263 | ST.count > 0)) /* count overflow ? */ | |
15272685 | 6264 | { |
c255a977 DM |
6265 | curly_try_B_min: |
6266 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 6267 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 6268 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
6269 | goto fake_end; |
6270 | } | |
4d5016e5 | 6271 | PUSH_STATE_GOTO(CURLY_B_min, ST.B, locinput); |
a0d0e21e LW |
6272 | } |
6273 | } | |
c74f6de9 | 6274 | sayNO; |
118e2215 | 6275 | assert(0); /* NOTREACHED */ |
c255a977 DM |
6276 | |
6277 | ||
6278 | curly_try_B_max: | |
6279 | /* a successful greedy match: now try to match B */ | |
40d049e4 | 6280 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 6281 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
40d049e4 YO |
6282 | goto fake_end; |
6283 | } | |
c255a977 | 6284 | { |
220db18a | 6285 | bool could_match = locinput < reginfo->strend; |
79a2a0e8 | 6286 | |
c255a977 | 6287 | /* If it could work, try it. */ |
79a2a0e8 KW |
6288 | if (ST.c1 != CHRTEST_VOID && could_match) { |
6289 | if (! UTF8_IS_INVARIANT(UCHARAT(locinput)) && utf8_target) | |
6290 | { | |
6291 | could_match = memEQ(locinput, | |
6292 | ST.c1_utf8, | |
6293 | UTF8SKIP(locinput)) | |
6294 | || memEQ(locinput, | |
6295 | ST.c2_utf8, | |
6296 | UTF8SKIP(locinput)); | |
6297 | } | |
6298 | else { | |
6299 | could_match = UCHARAT(locinput) == ST.c1 | |
6300 | || UCHARAT(locinput) == ST.c2; | |
6301 | } | |
6302 | } | |
6303 | if (ST.c1 == CHRTEST_VOID || could_match) { | |
c255a977 | 6304 | CURLY_SETPAREN(ST.paren, ST.count); |
4d5016e5 | 6305 | PUSH_STATE_GOTO(CURLY_B_max, ST.B, locinput); |
118e2215 | 6306 | assert(0); /* NOTREACHED */ |
c255a977 DM |
6307 | } |
6308 | } | |
6309 | /* FALL THROUGH */ | |
3c0563b9 | 6310 | |
c255a977 DM |
6311 | case CURLY_B_max_fail: |
6312 | /* failed to find B in a greedy match */ | |
c255a977 DM |
6313 | |
6314 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 DM |
6315 | if (ST.paren) { |
6316 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
6317 | } | |
c255a977 DM |
6318 | /* back up. */ |
6319 | if (--ST.count < ST.min) | |
6320 | sayNO; | |
eb72505d | 6321 | locinput = HOPc(locinput, -1); |
c255a977 DM |
6322 | goto curly_try_B_max; |
6323 | ||
6324 | #undef ST | |
6325 | ||
3c0563b9 | 6326 | case END: /* last op of main pattern */ |
6bda09f9 | 6327 | fake_end: |
faec1544 DM |
6328 | if (cur_eval) { |
6329 | /* we've just finished A in /(??{A})B/; now continue with B */ | |
faec1544 | 6330 | |
288b8c02 | 6331 | st->u.eval.prev_rex = rex_sv; /* inner */ |
92da3157 DM |
6332 | |
6333 | /* Save *all* the positions. */ | |
6334 | st->u.eval.cp = regcppush(rex, 0, maxopenparen); | |
ec43f78b | 6335 | rex_sv = cur_eval->u.eval.prev_rex; |
aed7b151 | 6336 | is_utf8_pat = reginfo->is_utf8_pat = cBOOL(RX_UTF8(rex_sv)); |
ec43f78b | 6337 | SET_reg_curpm(rex_sv); |
8d919b0a | 6338 | rex = ReANY(rex_sv); |
f8fc2ecf | 6339 | rexi = RXi_GET(rex); |
faec1544 | 6340 | cur_curlyx = cur_eval->u.eval.prev_curlyx; |
34a81e2b | 6341 | |
faec1544 | 6342 | REGCP_SET(st->u.eval.lastcp); |
faec1544 DM |
6343 | |
6344 | /* Restore parens of the outer rex without popping the | |
6345 | * savestack */ | |
92da3157 DM |
6346 | S_regcp_restore(aTHX_ rex, cur_eval->u.eval.lastcp, |
6347 | &maxopenparen); | |
faec1544 DM |
6348 | |
6349 | st->u.eval.prev_eval = cur_eval; | |
6350 | cur_eval = cur_eval->u.eval.prev_eval; | |
6351 | DEBUG_EXECUTE_r( | |
2a49f0f5 JH |
6352 | PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n", |
6353 | REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval));); | |
e7707071 YO |
6354 | if ( nochange_depth ) |
6355 | nochange_depth--; | |
6356 | ||
4d5016e5 DM |
6357 | PUSH_YES_STATE_GOTO(EVAL_AB, st->u.eval.prev_eval->u.eval.B, |
6358 | locinput); /* match B */ | |
faec1544 DM |
6359 | } |
6360 | ||
3b0527fe | 6361 | if (locinput < reginfo->till) { |
a3621e74 | 6362 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
7821416a IZ |
6363 | "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n", |
6364 | PL_colors[4], | |
6d59b646 DM |
6365 | (long)(locinput - startpos), |
6366 | (long)(reginfo->till - startpos), | |
7821416a | 6367 | PL_colors[5])); |
58e23c8d | 6368 | |
262b90c4 | 6369 | sayNO_SILENT; /* Cannot match: too short. */ |
7821416a | 6370 | } |
262b90c4 | 6371 | sayYES; /* Success! */ |
dad79028 DM |
6372 | |
6373 | case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */ | |
6374 | DEBUG_EXECUTE_r( | |
6375 | PerlIO_printf(Perl_debug_log, | |
6376 | "%*s %ssubpattern success...%s\n", | |
5bc10b2c | 6377 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])); |
262b90c4 | 6378 | sayYES; /* Success! */ |
dad79028 | 6379 | |
40a82448 DM |
6380 | #undef ST |
6381 | #define ST st->u.ifmatch | |
6382 | ||
37f53970 DM |
6383 | { |
6384 | char *newstart; | |
6385 | ||
40a82448 DM |
6386 | case SUSPEND: /* (?>A) */ |
6387 | ST.wanted = 1; | |
37f53970 | 6388 | newstart = locinput; |
9041c2e3 | 6389 | goto do_ifmatch; |
dad79028 | 6390 | |
40a82448 DM |
6391 | case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */ |
6392 | ST.wanted = 0; | |
dad79028 DM |
6393 | goto ifmatch_trivial_fail_test; |
6394 | ||
40a82448 DM |
6395 | case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */ |
6396 | ST.wanted = 1; | |
dad79028 | 6397 | ifmatch_trivial_fail_test: |
a0ed51b3 | 6398 | if (scan->flags) { |
52657f30 | 6399 | char * const s = HOPBACKc(locinput, scan->flags); |
dad79028 DM |
6400 | if (!s) { |
6401 | /* trivial fail */ | |
24d3c4a9 DM |
6402 | if (logical) { |
6403 | logical = 0; | |
f2338a2e | 6404 | sw = 1 - cBOOL(ST.wanted); |
dad79028 | 6405 | } |
40a82448 | 6406 | else if (ST.wanted) |
dad79028 DM |
6407 | sayNO; |
6408 | next = scan + ARG(scan); | |
6409 | if (next == scan) | |
6410 | next = NULL; | |
6411 | break; | |
6412 | } | |
37f53970 | 6413 | newstart = s; |
a0ed51b3 LW |
6414 | } |
6415 | else | |
37f53970 | 6416 | newstart = locinput; |
a0ed51b3 | 6417 | |
c277df42 | 6418 | do_ifmatch: |
40a82448 | 6419 | ST.me = scan; |
24d3c4a9 | 6420 | ST.logical = logical; |
24d786f4 YO |
6421 | logical = 0; /* XXX: reset state of logical once it has been saved into ST */ |
6422 | ||
40a82448 | 6423 | /* execute body of (?...A) */ |
37f53970 | 6424 | PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)), newstart); |
118e2215 | 6425 | assert(0); /* NOTREACHED */ |
37f53970 | 6426 | } |
40a82448 DM |
6427 | |
6428 | case IFMATCH_A_fail: /* body of (?...A) failed */ | |
6429 | ST.wanted = !ST.wanted; | |
6430 | /* FALL THROUGH */ | |
6431 | ||
6432 | case IFMATCH_A: /* body of (?...A) succeeded */ | |
24d3c4a9 | 6433 | if (ST.logical) { |
f2338a2e | 6434 | sw = cBOOL(ST.wanted); |
40a82448 DM |
6435 | } |
6436 | else if (!ST.wanted) | |
6437 | sayNO; | |
6438 | ||
37f53970 DM |
6439 | if (OP(ST.me) != SUSPEND) { |
6440 | /* restore old position except for (?>...) */ | |
6441 | locinput = st->locinput; | |
40a82448 DM |
6442 | } |
6443 | scan = ST.me + ARG(ST.me); | |
6444 | if (scan == ST.me) | |
6445 | scan = NULL; | |
6446 | continue; /* execute B */ | |
6447 | ||
6448 | #undef ST | |
dad79028 | 6449 | |
3c0563b9 DM |
6450 | case LONGJMP: /* alternative with many branches compiles to |
6451 | * (BRANCHJ; EXACT ...; LONGJMP ) x N */ | |
c277df42 IZ |
6452 | next = scan + ARG(scan); |
6453 | if (next == scan) | |
6454 | next = NULL; | |
a0d0e21e | 6455 | break; |
3c0563b9 DM |
6456 | |
6457 | case COMMIT: /* (*COMMIT) */ | |
220db18a | 6458 | reginfo->cutpoint = reginfo->strend; |
e2e6a0f1 | 6459 | /* FALLTHROUGH */ |
3c0563b9 DM |
6460 | |
6461 | case PRUNE: /* (*PRUNE) */ | |
e2e6a0f1 | 6462 | if (!scan->flags) |
ad64d0ec | 6463 | sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
4d5016e5 | 6464 | PUSH_STATE_GOTO(COMMIT_next, next, locinput); |
118e2215 | 6465 | assert(0); /* NOTREACHED */ |
3c0563b9 | 6466 | |
54612592 YO |
6467 | case COMMIT_next_fail: |
6468 | no_final = 1; | |
6469 | /* FALLTHROUGH */ | |
3c0563b9 DM |
6470 | |
6471 | case OPFAIL: /* (*FAIL) */ | |
7f69552c | 6472 | sayNO; |
118e2215 | 6473 | assert(0); /* NOTREACHED */ |
e2e6a0f1 YO |
6474 | |
6475 | #define ST st->u.mark | |
3c0563b9 | 6476 | case MARKPOINT: /* (*MARK:foo) */ |
e2e6a0f1 | 6477 | ST.prev_mark = mark_state; |
5d458dd8 | 6478 | ST.mark_name = sv_commit = sv_yes_mark |
ad64d0ec | 6479 | = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
e2e6a0f1 | 6480 | mark_state = st; |
4d5016e5 DM |
6481 | ST.mark_loc = locinput; |
6482 | PUSH_YES_STATE_GOTO(MARKPOINT_next, next, locinput); | |
118e2215 | 6483 | assert(0); /* NOTREACHED */ |
3c0563b9 | 6484 | |
e2e6a0f1 YO |
6485 | case MARKPOINT_next: |
6486 | mark_state = ST.prev_mark; | |
6487 | sayYES; | |
118e2215 | 6488 | assert(0); /* NOTREACHED */ |
3c0563b9 | 6489 | |
e2e6a0f1 | 6490 | case MARKPOINT_next_fail: |
5d458dd8 | 6491 | if (popmark && sv_eq(ST.mark_name,popmark)) |
e2e6a0f1 YO |
6492 | { |
6493 | if (ST.mark_loc > startpoint) | |
6494 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
6495 | popmark = NULL; /* we found our mark */ | |
6496 | sv_commit = ST.mark_name; | |
6497 | ||
6498 | DEBUG_EXECUTE_r({ | |
5d458dd8 | 6499 | PerlIO_printf(Perl_debug_log, |
e2e6a0f1 YO |
6500 | "%*s %ssetting cutpoint to mark:%"SVf"...%s\n", |
6501 | REPORT_CODE_OFF+depth*2, "", | |
be2597df | 6502 | PL_colors[4], SVfARG(sv_commit), PL_colors[5]); |
e2e6a0f1 YO |
6503 | }); |
6504 | } | |
6505 | mark_state = ST.prev_mark; | |
5d458dd8 YO |
6506 | sv_yes_mark = mark_state ? |
6507 | mark_state->u.mark.mark_name : NULL; | |
e2e6a0f1 | 6508 | sayNO; |
118e2215 | 6509 | assert(0); /* NOTREACHED */ |
3c0563b9 DM |
6510 | |
6511 | case SKIP: /* (*SKIP) */ | |
5d458dd8 | 6512 | if (scan->flags) { |
2bf803e2 | 6513 | /* (*SKIP) : if we fail we cut here*/ |
5d458dd8 | 6514 | ST.mark_name = NULL; |
e2e6a0f1 | 6515 | ST.mark_loc = locinput; |
4d5016e5 | 6516 | PUSH_STATE_GOTO(SKIP_next,next, locinput); |
5d458dd8 | 6517 | } else { |
2bf803e2 | 6518 | /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was, |
5d458dd8 YO |
6519 | otherwise do nothing. Meaning we need to scan |
6520 | */ | |
6521 | regmatch_state *cur = mark_state; | |
ad64d0ec | 6522 | SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
6523 | |
6524 | while (cur) { | |
6525 | if ( sv_eq( cur->u.mark.mark_name, | |
6526 | find ) ) | |
6527 | { | |
6528 | ST.mark_name = find; | |
4d5016e5 | 6529 | PUSH_STATE_GOTO( SKIP_next, next, locinput); |
5d458dd8 YO |
6530 | } |
6531 | cur = cur->u.mark.prev_mark; | |
6532 | } | |
e2e6a0f1 | 6533 | } |
2bf803e2 | 6534 | /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */ |
5d458dd8 | 6535 | break; |
3c0563b9 | 6536 | |
5d458dd8 YO |
6537 | case SKIP_next_fail: |
6538 | if (ST.mark_name) { | |
6539 | /* (*CUT:NAME) - Set up to search for the name as we | |
6540 | collapse the stack*/ | |
6541 | popmark = ST.mark_name; | |
6542 | } else { | |
6543 | /* (*CUT) - No name, we cut here.*/ | |
e2e6a0f1 YO |
6544 | if (ST.mark_loc > startpoint) |
6545 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5d458dd8 YO |
6546 | /* but we set sv_commit to latest mark_name if there |
6547 | is one so they can test to see how things lead to this | |
6548 | cut */ | |
6549 | if (mark_state) | |
6550 | sv_commit=mark_state->u.mark.mark_name; | |
6551 | } | |
e2e6a0f1 YO |
6552 | no_final = 1; |
6553 | sayNO; | |
118e2215 | 6554 | assert(0); /* NOTREACHED */ |
e2e6a0f1 | 6555 | #undef ST |
3c0563b9 DM |
6556 | |
6557 | case LNBREAK: /* \R */ | |
220db18a | 6558 | if ((n=is_LNBREAK_safe(locinput, reginfo->strend, utf8_target))) { |
e1d1eefb | 6559 | locinput += n; |
e1d1eefb YO |
6560 | } else |
6561 | sayNO; | |
6562 | break; | |
6563 | ||
a0d0e21e | 6564 | default: |
b900a521 | 6565 | PerlIO_printf(Perl_error_log, "%"UVxf" %d\n", |
d7d93a81 | 6566 | PTR2UV(scan), OP(scan)); |
cea2e8a9 | 6567 | Perl_croak(aTHX_ "regexp memory corruption"); |
28b98f76 DM |
6568 | |
6569 | /* this is a point to jump to in order to increment | |
6570 | * locinput by one character */ | |
6571 | increment_locinput: | |
e6ca698c | 6572 | assert(!NEXTCHR_IS_EOS); |
28b98f76 DM |
6573 | if (utf8_target) { |
6574 | locinput += PL_utf8skip[nextchr]; | |
7016d6eb | 6575 | /* locinput is allowed to go 1 char off the end, but not 2+ */ |
220db18a | 6576 | if (locinput > reginfo->strend) |
28b98f76 | 6577 | sayNO; |
28b98f76 DM |
6578 | } |
6579 | else | |
3640db6b | 6580 | locinput++; |
28b98f76 | 6581 | break; |
5d458dd8 YO |
6582 | |
6583 | } /* end switch */ | |
95b24440 | 6584 | |
5d458dd8 YO |
6585 | /* switch break jumps here */ |
6586 | scan = next; /* prepare to execute the next op and ... */ | |
6587 | continue; /* ... jump back to the top, reusing st */ | |
118e2215 | 6588 | assert(0); /* NOTREACHED */ |
95b24440 | 6589 | |
40a82448 DM |
6590 | push_yes_state: |
6591 | /* push a state that backtracks on success */ | |
6592 | st->u.yes.prev_yes_state = yes_state; | |
6593 | yes_state = st; | |
6594 | /* FALL THROUGH */ | |
6595 | push_state: | |
6596 | /* push a new regex state, then continue at scan */ | |
6597 | { | |
6598 | regmatch_state *newst; | |
6599 | ||
24b23f37 YO |
6600 | DEBUG_STACK_r({ |
6601 | regmatch_state *cur = st; | |
6602 | regmatch_state *curyes = yes_state; | |
6603 | int curd = depth; | |
6604 | regmatch_slab *slab = PL_regmatch_slab; | |
6605 | for (;curd > -1;cur--,curd--) { | |
6606 | if (cur < SLAB_FIRST(slab)) { | |
6607 | slab = slab->prev; | |
6608 | cur = SLAB_LAST(slab); | |
6609 | } | |
6610 | PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n", | |
6611 | REPORT_CODE_OFF + 2 + depth * 2,"", | |
13d6edb4 | 6612 | curd, PL_reg_name[cur->resume_state], |
24b23f37 YO |
6613 | (curyes == cur) ? "yes" : "" |
6614 | ); | |
6615 | if (curyes == cur) | |
6616 | curyes = cur->u.yes.prev_yes_state; | |
6617 | } | |
6618 | } else | |
6619 | DEBUG_STATE_pp("push") | |
6620 | ); | |
40a82448 | 6621 | depth++; |
40a82448 DM |
6622 | st->locinput = locinput; |
6623 | newst = st+1; | |
6624 | if (newst > SLAB_LAST(PL_regmatch_slab)) | |
6625 | newst = S_push_slab(aTHX); | |
6626 | PL_regmatch_state = newst; | |
786e8c11 | 6627 | |
4d5016e5 | 6628 | locinput = pushinput; |
40a82448 DM |
6629 | st = newst; |
6630 | continue; | |
118e2215 | 6631 | assert(0); /* NOTREACHED */ |
40a82448 | 6632 | } |
a0d0e21e | 6633 | } |
a687059c | 6634 | |
a0d0e21e LW |
6635 | /* |
6636 | * We get here only if there's trouble -- normally "case END" is | |
6637 | * the terminating point. | |
6638 | */ | |
cea2e8a9 | 6639 | Perl_croak(aTHX_ "corrupted regexp pointers"); |
a0d0e21e | 6640 | /*NOTREACHED*/ |
4633a7c4 LW |
6641 | sayNO; |
6642 | ||
262b90c4 | 6643 | yes: |
77cb431f DM |
6644 | if (yes_state) { |
6645 | /* we have successfully completed a subexpression, but we must now | |
6646 | * pop to the state marked by yes_state and continue from there */ | |
77cb431f | 6647 | assert(st != yes_state); |
5bc10b2c DM |
6648 | #ifdef DEBUGGING |
6649 | while (st != yes_state) { | |
6650 | st--; | |
6651 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
6652 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
6653 | st = SLAB_LAST(PL_regmatch_slab); | |
6654 | } | |
e2e6a0f1 | 6655 | DEBUG_STATE_r({ |
54612592 YO |
6656 | if (no_final) { |
6657 | DEBUG_STATE_pp("pop (no final)"); | |
6658 | } else { | |
6659 | DEBUG_STATE_pp("pop (yes)"); | |
6660 | } | |
e2e6a0f1 | 6661 | }); |
5bc10b2c DM |
6662 | depth--; |
6663 | } | |
6664 | #else | |
77cb431f DM |
6665 | while (yes_state < SLAB_FIRST(PL_regmatch_slab) |
6666 | || yes_state > SLAB_LAST(PL_regmatch_slab)) | |
6667 | { | |
6668 | /* not in this slab, pop slab */ | |
6669 | depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1); | |
6670 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
6671 | st = SLAB_LAST(PL_regmatch_slab); | |
6672 | } | |
6673 | depth -= (st - yes_state); | |
5bc10b2c | 6674 | #endif |
77cb431f DM |
6675 | st = yes_state; |
6676 | yes_state = st->u.yes.prev_yes_state; | |
6677 | PL_regmatch_state = st; | |
24b23f37 | 6678 | |
3640db6b | 6679 | if (no_final) |
5d458dd8 | 6680 | locinput= st->locinput; |
54612592 | 6681 | state_num = st->resume_state + no_final; |
24d3c4a9 | 6682 | goto reenter_switch; |
77cb431f DM |
6683 | } |
6684 | ||
a3621e74 | 6685 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", |
e4584336 | 6686 | PL_colors[4], PL_colors[5])); |
02db2b7b | 6687 | |
bf2039a9 | 6688 | if (reginfo->info_aux_eval) { |
19b95bf0 DM |
6689 | /* each successfully executed (?{...}) block does the equivalent of |
6690 | * local $^R = do {...} | |
6691 | * When popping the save stack, all these locals would be undone; | |
6692 | * bypass this by setting the outermost saved $^R to the latest | |
6693 | * value */ | |
6694 | if (oreplsv != GvSV(PL_replgv)) | |
6695 | sv_setsv(oreplsv, GvSV(PL_replgv)); | |
6696 | } | |
95b24440 | 6697 | result = 1; |
aa283a38 | 6698 | goto final_exit; |
4633a7c4 LW |
6699 | |
6700 | no: | |
a3621e74 | 6701 | DEBUG_EXECUTE_r( |
7821416a | 6702 | PerlIO_printf(Perl_debug_log, |
786e8c11 | 6703 | "%*s %sfailed...%s\n", |
5bc10b2c | 6704 | REPORT_CODE_OFF+depth*2, "", |
786e8c11 | 6705 | PL_colors[4], PL_colors[5]) |
7821416a | 6706 | ); |
aa283a38 | 6707 | |
262b90c4 | 6708 | no_silent: |
54612592 YO |
6709 | if (no_final) { |
6710 | if (yes_state) { | |
6711 | goto yes; | |
6712 | } else { | |
6713 | goto final_exit; | |
6714 | } | |
6715 | } | |
aa283a38 DM |
6716 | if (depth) { |
6717 | /* there's a previous state to backtrack to */ | |
40a82448 DM |
6718 | st--; |
6719 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
6720 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
6721 | st = SLAB_LAST(PL_regmatch_slab); | |
6722 | } | |
6723 | PL_regmatch_state = st; | |
40a82448 | 6724 | locinput= st->locinput; |
40a82448 | 6725 | |
5bc10b2c DM |
6726 | DEBUG_STATE_pp("pop"); |
6727 | depth--; | |
262b90c4 DM |
6728 | if (yes_state == st) |
6729 | yes_state = st->u.yes.prev_yes_state; | |
5bc10b2c | 6730 | |
24d3c4a9 DM |
6731 | state_num = st->resume_state + 1; /* failure = success + 1 */ |
6732 | goto reenter_switch; | |
95b24440 | 6733 | } |
24d3c4a9 | 6734 | result = 0; |
aa283a38 | 6735 | |
262b90c4 | 6736 | final_exit: |
bbe252da | 6737 | if (rex->intflags & PREGf_VERBARG_SEEN) { |
5d458dd8 YO |
6738 | SV *sv_err = get_sv("REGERROR", 1); |
6739 | SV *sv_mrk = get_sv("REGMARK", 1); | |
6740 | if (result) { | |
e2e6a0f1 | 6741 | sv_commit = &PL_sv_no; |
5d458dd8 YO |
6742 | if (!sv_yes_mark) |
6743 | sv_yes_mark = &PL_sv_yes; | |
6744 | } else { | |
6745 | if (!sv_commit) | |
6746 | sv_commit = &PL_sv_yes; | |
6747 | sv_yes_mark = &PL_sv_no; | |
6748 | } | |
6749 | sv_setsv(sv_err, sv_commit); | |
6750 | sv_setsv(sv_mrk, sv_yes_mark); | |
e2e6a0f1 | 6751 | } |
19b95bf0 | 6752 | |
81ed78b2 DM |
6753 | |
6754 | if (last_pushed_cv) { | |
6755 | dSP; | |
6756 | POP_MULTICALL; | |
4f8dbb2d | 6757 | PERL_UNUSED_VAR(SP); |
81ed78b2 DM |
6758 | } |
6759 | ||
9d9163fb DM |
6760 | assert(!result || locinput - reginfo->strbeg >= 0); |
6761 | return result ? locinput - reginfo->strbeg : -1; | |
a687059c LW |
6762 | } |
6763 | ||
6764 | /* | |
6765 | - regrepeat - repeatedly match something simple, report how many | |
d60de1d1 | 6766 | * |
e64f369d KW |
6767 | * What 'simple' means is a node which can be the operand of a quantifier like |
6768 | * '+', or {1,3} | |
6769 | * | |
d60de1d1 DM |
6770 | * startposp - pointer a pointer to the start position. This is updated |
6771 | * to point to the byte following the highest successful | |
6772 | * match. | |
6773 | * p - the regnode to be repeatedly matched against. | |
220db18a | 6774 | * reginfo - struct holding match state, such as strend |
4063ade8 | 6775 | * max - maximum number of things to match. |
d60de1d1 | 6776 | * depth - (for debugging) backtracking depth. |
a687059c | 6777 | */ |
76e3520e | 6778 | STATIC I32 |
272d35c9 | 6779 | S_regrepeat(pTHX_ regexp *prog, char **startposp, const regnode *p, |
f9176b44 | 6780 | regmatch_info *const reginfo, I32 max, int depth) |
a687059c | 6781 | { |
27da23d5 | 6782 | dVAR; |
4063ade8 | 6783 | char *scan; /* Pointer to current position in target string */ |
eb578fdb | 6784 | I32 c; |
220db18a | 6785 | char *loceol = reginfo->strend; /* local version */ |
4063ade8 | 6786 | I32 hardcount = 0; /* How many matches so far */ |
ba44c216 | 6787 | bool utf8_target = reginfo->is_utf8_target; |
3018b823 | 6788 | int to_complement = 0; /* Invert the result? */ |
d513472c | 6789 | UV utf8_flags; |
3018b823 | 6790 | _char_class_number classnum; |
4f55667c SP |
6791 | #ifndef DEBUGGING |
6792 | PERL_UNUSED_ARG(depth); | |
6793 | #endif | |
a0d0e21e | 6794 | |
7918f24d NC |
6795 | PERL_ARGS_ASSERT_REGREPEAT; |
6796 | ||
f73aaa43 | 6797 | scan = *startposp; |
faf11cac HS |
6798 | if (max == REG_INFTY) |
6799 | max = I32_MAX; | |
dfb8f192 | 6800 | else if (! utf8_target && loceol - scan > max) |
7f596f4c | 6801 | loceol = scan + max; |
4063ade8 KW |
6802 | |
6803 | /* Here, for the case of a non-UTF-8 target we have adjusted <loceol> down | |
6804 | * to the maximum of how far we should go in it (leaving it set to the real | |
6805 | * end, if the maximum permissible would take us beyond that). This allows | |
6806 | * us to make the loop exit condition that we haven't gone past <loceol> to | |
6807 | * also mean that we haven't exceeded the max permissible count, saving a | |
6808 | * test each time through the loop. But it assumes that the OP matches a | |
6809 | * single byte, which is true for most of the OPs below when applied to a | |
6810 | * non-UTF-8 target. Those relatively few OPs that don't have this | |
6811 | * characteristic will have to compensate. | |
6812 | * | |
6813 | * There is no adjustment for UTF-8 targets, as the number of bytes per | |
6814 | * character varies. OPs will have to test both that the count is less | |
6815 | * than the max permissible (using <hardcount> to keep track), and that we | |
6816 | * are still within the bounds of the string (using <loceol>. A few OPs | |
6817 | * match a single byte no matter what the encoding. They can omit the max | |
6818 | * test if, for the UTF-8 case, they do the adjustment that was skipped | |
6819 | * above. | |
6820 | * | |
6821 | * Thus, the code above sets things up for the common case; and exceptional | |
6822 | * cases need extra work; the common case is to make sure <scan> doesn't | |
6823 | * go past <loceol>, and for UTF-8 to also use <hardcount> to make sure the | |
6824 | * count doesn't exceed the maximum permissible */ | |
6825 | ||
a0d0e21e | 6826 | switch (OP(p)) { |
22c35a8c | 6827 | case REG_ANY: |
f2ed9b32 | 6828 | if (utf8_target) { |
1aa99e6b | 6829 | while (scan < loceol && hardcount < max && *scan != '\n') { |
ffc61ed2 JH |
6830 | scan += UTF8SKIP(scan); |
6831 | hardcount++; | |
6832 | } | |
6833 | } else { | |
6834 | while (scan < loceol && *scan != '\n') | |
6835 | scan++; | |
a0ed51b3 LW |
6836 | } |
6837 | break; | |
ffc61ed2 | 6838 | case SANY: |
f2ed9b32 | 6839 | if (utf8_target) { |
a0804c9e | 6840 | while (scan < loceol && hardcount < max) { |
def8e4ea JH |
6841 | scan += UTF8SKIP(scan); |
6842 | hardcount++; | |
6843 | } | |
6844 | } | |
6845 | else | |
6846 | scan = loceol; | |
a0ed51b3 | 6847 | break; |
4063ade8 | 6848 | case CANY: /* Move <scan> forward <max> bytes, unless goes off end */ |
9597860a | 6849 | if (utf8_target && loceol - scan > max) { |
4063ade8 KW |
6850 | |
6851 | /* <loceol> hadn't been adjusted in the UTF-8 case */ | |
6852 | scan += max; | |
6853 | } | |
6854 | else { | |
6855 | scan = loceol; | |
6856 | } | |
f33976b4 | 6857 | break; |
59d32103 | 6858 | case EXACT: |
f9176b44 | 6859 | assert(STR_LEN(p) == reginfo->is_utf8_pat ? UTF8SKIP(STRING(p)) : 1); |
613a425d | 6860 | |
59d32103 | 6861 | c = (U8)*STRING(p); |
59d32103 | 6862 | |
5e4a1da1 KW |
6863 | /* Can use a simple loop if the pattern char to match on is invariant |
6864 | * under UTF-8, or both target and pattern aren't UTF-8. Note that we | |
6865 | * can use UTF8_IS_INVARIANT() even if the pattern isn't UTF-8, as it's | |
6866 | * true iff it doesn't matter if the argument is in UTF-8 or not */ | |
f9176b44 | 6867 | if (UTF8_IS_INVARIANT(c) || (! utf8_target && ! reginfo->is_utf8_pat)) { |
e9369824 | 6868 | if (utf8_target && loceol - scan > max) { |
4063ade8 KW |
6869 | /* We didn't adjust <loceol> because is UTF-8, but ok to do so, |
6870 | * since here, to match at all, 1 char == 1 byte */ | |
6871 | loceol = scan + max; | |
6872 | } | |
59d32103 KW |
6873 | while (scan < loceol && UCHARAT(scan) == c) { |
6874 | scan++; | |
6875 | } | |
6876 | } | |
f9176b44 | 6877 | else if (reginfo->is_utf8_pat) { |
5e4a1da1 KW |
6878 | if (utf8_target) { |
6879 | STRLEN scan_char_len; | |
5e4a1da1 | 6880 | |
4063ade8 | 6881 | /* When both target and pattern are UTF-8, we have to do |
5e4a1da1 KW |
6882 | * string EQ */ |
6883 | while (hardcount < max | |
9a902117 KW |
6884 | && scan < loceol |
6885 | && (scan_char_len = UTF8SKIP(scan)) <= STR_LEN(p) | |
5e4a1da1 KW |
6886 | && memEQ(scan, STRING(p), scan_char_len)) |
6887 | { | |
4200a00c | 6888 | scan += scan_char_len; |
5e4a1da1 KW |
6889 | hardcount++; |
6890 | } | |
6891 | } | |
6892 | else if (! UTF8_IS_ABOVE_LATIN1(c)) { | |
b40a2c17 | 6893 | |
5e4a1da1 KW |
6894 | /* Target isn't utf8; convert the character in the UTF-8 |
6895 | * pattern to non-UTF8, and do a simple loop */ | |
94bb8c36 | 6896 | c = TWO_BYTE_UTF8_TO_NATIVE(c, *(STRING(p) + 1)); |
5e4a1da1 KW |
6897 | while (scan < loceol && UCHARAT(scan) == c) { |
6898 | scan++; | |
6899 | } | |
6900 | } /* else pattern char is above Latin1, can't possibly match the | |
6901 | non-UTF-8 target */ | |
b40a2c17 | 6902 | } |
5e4a1da1 | 6903 | else { |
59d32103 | 6904 | |
5e4a1da1 KW |
6905 | /* Here, the string must be utf8; pattern isn't, and <c> is |
6906 | * different in utf8 than not, so can't compare them directly. | |
6907 | * Outside the loop, find the two utf8 bytes that represent c, and | |
6908 | * then look for those in sequence in the utf8 string */ | |
59d32103 KW |
6909 | U8 high = UTF8_TWO_BYTE_HI(c); |
6910 | U8 low = UTF8_TWO_BYTE_LO(c); | |
59d32103 KW |
6911 | |
6912 | while (hardcount < max | |
6913 | && scan + 1 < loceol | |
6914 | && UCHARAT(scan) == high | |
6915 | && UCHARAT(scan + 1) == low) | |
6916 | { | |
6917 | scan += 2; | |
6918 | hardcount++; | |
6919 | } | |
6920 | } | |
6921 | break; | |
5e4a1da1 | 6922 | |
098b07d5 KW |
6923 | case EXACTFA_NO_TRIE: /* This node only generated for non-utf8 patterns */ |
6924 | assert(! reginfo->is_utf8_pat); | |
6925 | /* FALL THROUGH */ | |
2f7f8cb1 | 6926 | case EXACTFA: |
098b07d5 | 6927 | utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; |
2f7f8cb1 KW |
6928 | goto do_exactf; |
6929 | ||
d4e0b827 | 6930 | case EXACTFL: |
272d35c9 | 6931 | RXp_MATCH_TAINTED_on(prog); |
17580e7a KW |
6932 | utf8_flags = FOLDEQ_UTF8_LOCALE; |
6933 | goto do_exactf; | |
6934 | ||
2fdb7295 KW |
6935 | case EXACTF: /* This node only generated for non-utf8 patterns */ |
6936 | assert(! reginfo->is_utf8_pat); | |
098b07d5 KW |
6937 | utf8_flags = 0; |
6938 | goto do_exactf; | |
62bf7766 | 6939 | |
3c760661 | 6940 | case EXACTFU_SS: |
9a5a5549 | 6941 | case EXACTFU: |
f9176b44 | 6942 | utf8_flags = reginfo->is_utf8_pat ? FOLDEQ_S2_ALREADY_FOLDED : 0; |
59d32103 | 6943 | |
613a425d KW |
6944 | do_exactf: { |
6945 | int c1, c2; | |
6946 | U8 c1_utf8[UTF8_MAXBYTES+1], c2_utf8[UTF8_MAXBYTES+1]; | |
d4e0b827 | 6947 | |
f9176b44 | 6948 | assert(STR_LEN(p) == reginfo->is_utf8_pat ? UTF8SKIP(STRING(p)) : 1); |
613a425d | 6949 | |
984e6dd1 | 6950 | if (S_setup_EXACTISH_ST_c1_c2(aTHX_ p, &c1, c1_utf8, &c2, c2_utf8, |
aed7b151 | 6951 | reginfo)) |
984e6dd1 | 6952 | { |
613a425d | 6953 | if (c1 == CHRTEST_VOID) { |
49b95fad | 6954 | /* Use full Unicode fold matching */ |
220db18a | 6955 | char *tmpeol = reginfo->strend; |
f9176b44 | 6956 | STRLEN pat_len = reginfo->is_utf8_pat ? UTF8SKIP(STRING(p)) : 1; |
49b95fad KW |
6957 | while (hardcount < max |
6958 | && foldEQ_utf8_flags(scan, &tmpeol, 0, utf8_target, | |
6959 | STRING(p), NULL, pat_len, | |
f9176b44 | 6960 | reginfo->is_utf8_pat, utf8_flags)) |
49b95fad KW |
6961 | { |
6962 | scan = tmpeol; | |
220db18a | 6963 | tmpeol = reginfo->strend; |
49b95fad KW |
6964 | hardcount++; |
6965 | } | |
613a425d KW |
6966 | } |
6967 | else if (utf8_target) { | |
6968 | if (c1 == c2) { | |
4063ade8 KW |
6969 | while (scan < loceol |
6970 | && hardcount < max | |
613a425d KW |
6971 | && memEQ(scan, c1_utf8, UTF8SKIP(scan))) |
6972 | { | |
6973 | scan += UTF8SKIP(scan); | |
6974 | hardcount++; | |
6975 | } | |
6976 | } | |
6977 | else { | |
4063ade8 KW |
6978 | while (scan < loceol |
6979 | && hardcount < max | |
613a425d KW |
6980 | && (memEQ(scan, c1_utf8, UTF8SKIP(scan)) |
6981 | || memEQ(scan, c2_utf8, UTF8SKIP(scan)))) | |
6982 | { | |
6983 | scan += UTF8SKIP(scan); | |
6984 | hardcount++; | |
6985 | } | |
6986 | } | |
6987 | } | |
6988 | else if (c1 == c2) { | |
6989 | while (scan < loceol && UCHARAT(scan) == c1) { | |
6990 | scan++; | |
6991 | } | |
6992 | } | |
6993 | else { | |
6994 | while (scan < loceol && | |
6995 | (UCHARAT(scan) == c1 || UCHARAT(scan) == c2)) | |
6996 | { | |
6997 | scan++; | |
6998 | } | |
6999 | } | |
634c83a2 | 7000 | } |
bbce6d69 | 7001 | break; |
613a425d | 7002 | } |
a0d0e21e | 7003 | case ANYOF: |
954a2af6 | 7004 | case ANYOF_WARN_SUPER: |
e0193e47 | 7005 | if (utf8_target) { |
4e8910e0 | 7006 | while (hardcount < max |
9a902117 | 7007 | && scan < loceol |
635cd5d4 | 7008 | && reginclass(prog, p, (U8*)scan, utf8_target)) |
4e8910e0 | 7009 | { |
9a902117 | 7010 | scan += UTF8SKIP(scan); |
ffc61ed2 JH |
7011 | hardcount++; |
7012 | } | |
7013 | } else { | |
32fc9b6a | 7014 | while (scan < loceol && REGINCLASS(prog, p, (U8*)scan)) |
ffc61ed2 JH |
7015 | scan++; |
7016 | } | |
a0d0e21e | 7017 | break; |
4063ade8 | 7018 | |
3018b823 | 7019 | /* The argument (FLAGS) to all the POSIX node types is the class number */ |
980866de | 7020 | |
3018b823 KW |
7021 | case NPOSIXL: |
7022 | to_complement = 1; | |
7023 | /* FALLTHROUGH */ | |
980866de | 7024 | |
3018b823 | 7025 | case POSIXL: |
272d35c9 | 7026 | RXp_MATCH_TAINTED_on(prog); |
3018b823 KW |
7027 | if (! utf8_target) { |
7028 | while (scan < loceol && to_complement ^ cBOOL(isFOO_lc(FLAGS(p), | |
7029 | *scan))) | |
a12cf05f | 7030 | { |
3018b823 KW |
7031 | scan++; |
7032 | } | |
7033 | } else { | |
7034 | while (hardcount < max && scan < loceol | |
7035 | && to_complement ^ cBOOL(isFOO_utf8_lc(FLAGS(p), | |
7036 | (U8 *) scan))) | |
7037 | { | |
7038 | scan += UTF8SKIP(scan); | |
ffc61ed2 JH |
7039 | hardcount++; |
7040 | } | |
a0ed51b3 LW |
7041 | } |
7042 | break; | |
0658cdde | 7043 | |
3018b823 KW |
7044 | case POSIXD: |
7045 | if (utf8_target) { | |
7046 | goto utf8_posix; | |
7047 | } | |
7048 | /* FALLTHROUGH */ | |
7049 | ||
0658cdde | 7050 | case POSIXA: |
0430522f | 7051 | if (utf8_target && loceol - scan > max) { |
4063ade8 | 7052 | |
7aee35ff KW |
7053 | /* We didn't adjust <loceol> at the beginning of this routine |
7054 | * because is UTF-8, but it is actually ok to do so, since here, to | |
7055 | * match, 1 char == 1 byte. */ | |
4063ade8 KW |
7056 | loceol = scan + max; |
7057 | } | |
7058 | while (scan < loceol && _generic_isCC_A((U8) *scan, FLAGS(p))) { | |
0658cdde KW |
7059 | scan++; |
7060 | } | |
7061 | break; | |
980866de | 7062 | |
3018b823 KW |
7063 | case NPOSIXD: |
7064 | if (utf8_target) { | |
7065 | to_complement = 1; | |
7066 | goto utf8_posix; | |
7067 | } | |
7068 | /* FALL THROUGH */ | |
980866de | 7069 | |
3018b823 KW |
7070 | case NPOSIXA: |
7071 | if (! utf8_target) { | |
7072 | while (scan < loceol && ! _generic_isCC_A((U8) *scan, FLAGS(p))) { | |
a12cf05f KW |
7073 | scan++; |
7074 | } | |
4063ade8 | 7075 | } |
3018b823 | 7076 | else { |
980866de | 7077 | |
3018b823 KW |
7078 | /* The complement of something that matches only ASCII matches all |
7079 | * UTF-8 variant code points, plus everything in ASCII that isn't | |
7080 | * in the class. */ | |
bedac28b | 7081 | while (hardcount < max && scan < loceol |
3018b823 KW |
7082 | && (! UTF8_IS_INVARIANT(*scan) |
7083 | || ! _generic_isCC_A((U8) *scan, FLAGS(p)))) | |
a12cf05f | 7084 | { |
3018b823 | 7085 | scan += UTF8SKIP(scan); |
ffc61ed2 JH |
7086 | hardcount++; |
7087 | } | |
3018b823 KW |
7088 | } |
7089 | break; | |
980866de | 7090 | |
3018b823 KW |
7091 | case NPOSIXU: |
7092 | to_complement = 1; | |
7093 | /* FALLTHROUGH */ | |
7094 | ||
7095 | case POSIXU: | |
7096 | if (! utf8_target) { | |
7097 | while (scan < loceol && to_complement | |
7098 | ^ cBOOL(_generic_isCC((U8) *scan, FLAGS(p)))) | |
4063ade8 | 7099 | { |
3018b823 KW |
7100 | scan++; |
7101 | } | |
cfaf538b KW |
7102 | } |
7103 | else { | |
3018b823 KW |
7104 | utf8_posix: |
7105 | classnum = (_char_class_number) FLAGS(p); | |
7106 | if (classnum < _FIRST_NON_SWASH_CC) { | |
7107 | ||
7108 | /* Here, a swash is needed for above-Latin1 code points. | |
7109 | * Process as many Latin1 code points using the built-in rules. | |
7110 | * Go to another loop to finish processing upon encountering | |
7111 | * the first Latin1 code point. We could do that in this loop | |
7112 | * as well, but the other way saves having to test if the swash | |
7113 | * has been loaded every time through the loop: extra space to | |
7114 | * save a test. */ | |
7115 | while (hardcount < max && scan < loceol) { | |
7116 | if (UTF8_IS_INVARIANT(*scan)) { | |
7117 | if (! (to_complement ^ cBOOL(_generic_isCC((U8) *scan, | |
7118 | classnum)))) | |
7119 | { | |
7120 | break; | |
7121 | } | |
7122 | scan++; | |
7123 | } | |
7124 | else if (UTF8_IS_DOWNGRADEABLE_START(*scan)) { | |
7125 | if (! (to_complement | |
94bb8c36 KW |
7126 | ^ cBOOL(_generic_isCC(TWO_BYTE_UTF8_TO_NATIVE(*scan, |
7127 | *(scan + 1)), | |
3018b823 KW |
7128 | classnum)))) |
7129 | { | |
7130 | break; | |
7131 | } | |
7132 | scan += 2; | |
7133 | } | |
7134 | else { | |
7135 | goto found_above_latin1; | |
7136 | } | |
7137 | ||
7138 | hardcount++; | |
7139 | } | |
7140 | } | |
7141 | else { | |
7142 | /* For these character classes, the knowledge of how to handle | |
7143 | * every code point is compiled in to Perl via a macro. This | |
7144 | * code is written for making the loops as tight as possible. | |
7145 | * It could be refactored to save space instead */ | |
7146 | switch (classnum) { | |
7147 | case _CC_ENUM_SPACE: /* XXX would require separate code | |
7148 | if we revert the change of \v | |
7149 | matching this */ | |
7150 | /* FALL THROUGH */ | |
7151 | case _CC_ENUM_PSXSPC: | |
7152 | while (hardcount < max | |
7153 | && scan < loceol | |
7154 | && (to_complement ^ cBOOL(isSPACE_utf8(scan)))) | |
7155 | { | |
7156 | scan += UTF8SKIP(scan); | |
7157 | hardcount++; | |
7158 | } | |
7159 | break; | |
7160 | case _CC_ENUM_BLANK: | |
7161 | while (hardcount < max | |
7162 | && scan < loceol | |
7163 | && (to_complement ^ cBOOL(isBLANK_utf8(scan)))) | |
7164 | { | |
7165 | scan += UTF8SKIP(scan); | |
7166 | hardcount++; | |
7167 | } | |
7168 | break; | |
7169 | case _CC_ENUM_XDIGIT: | |
7170 | while (hardcount < max | |
7171 | && scan < loceol | |
7172 | && (to_complement ^ cBOOL(isXDIGIT_utf8(scan)))) | |
7173 | { | |
7174 | scan += UTF8SKIP(scan); | |
7175 | hardcount++; | |
7176 | } | |
7177 | break; | |
7178 | case _CC_ENUM_VERTSPACE: | |
7179 | while (hardcount < max | |
7180 | && scan < loceol | |
7181 | && (to_complement ^ cBOOL(isVERTWS_utf8(scan)))) | |
7182 | { | |
7183 | scan += UTF8SKIP(scan); | |
7184 | hardcount++; | |
7185 | } | |
7186 | break; | |
7187 | case _CC_ENUM_CNTRL: | |
7188 | while (hardcount < max | |
7189 | && scan < loceol | |
7190 | && (to_complement ^ cBOOL(isCNTRL_utf8(scan)))) | |
7191 | { | |
7192 | scan += UTF8SKIP(scan); | |
7193 | hardcount++; | |
7194 | } | |
7195 | break; | |
7196 | default: | |
7197 | Perl_croak(aTHX_ "panic: regrepeat() node %d='%s' has an unexpected character class '%d'", OP(p), PL_reg_name[OP(p)], classnum); | |
7198 | } | |
7199 | } | |
a0ed51b3 | 7200 | } |
3018b823 | 7201 | break; |
4063ade8 | 7202 | |
3018b823 KW |
7203 | found_above_latin1: /* Continuation of POSIXU and NPOSIXU */ |
7204 | ||
7205 | /* Load the swash if not already present */ | |
7206 | if (! PL_utf8_swash_ptrs[classnum]) { | |
7207 | U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; | |
7208 | PL_utf8_swash_ptrs[classnum] = _core_swash_init( | |
7209 | "utf8", swash_property_names[classnum], | |
7210 | &PL_sv_undef, 1, 0, NULL, &flags); | |
4063ade8 | 7211 | } |
3018b823 KW |
7212 | |
7213 | while (hardcount < max && scan < loceol | |
7214 | && to_complement ^ cBOOL(_generic_utf8( | |
7215 | classnum, | |
7216 | scan, | |
7217 | swash_fetch(PL_utf8_swash_ptrs[classnum], | |
7218 | (U8 *) scan, | |
7219 | TRUE)))) | |
7220 | { | |
7221 | scan += UTF8SKIP(scan); | |
7222 | hardcount++; | |
7223 | } | |
7224 | break; | |
7225 | ||
e1d1eefb | 7226 | case LNBREAK: |
e64f369d KW |
7227 | if (utf8_target) { |
7228 | while (hardcount < max && scan < loceol && | |
7229 | (c=is_LNBREAK_utf8_safe(scan, loceol))) { | |
7230 | scan += c; | |
7231 | hardcount++; | |
7232 | } | |
7233 | } else { | |
7234 | /* LNBREAK can match one or two latin chars, which is ok, but we | |
7235 | * have to use hardcount in this situation, and throw away the | |
7236 | * adjustment to <loceol> done before the switch statement */ | |
220db18a | 7237 | loceol = reginfo->strend; |
e64f369d KW |
7238 | while (scan < loceol && (c=is_LNBREAK_latin1_safe(scan, loceol))) { |
7239 | scan+=c; | |
7240 | hardcount++; | |
7241 | } | |
7242 | } | |
7243 | break; | |
e1d1eefb | 7244 | |
584b1f02 KW |
7245 | case BOUND: |
7246 | case BOUNDA: | |
7247 | case BOUNDL: | |
7248 | case BOUNDU: | |
7249 | case EOS: | |
7250 | case GPOS: | |
7251 | case KEEPS: | |
7252 | case NBOUND: | |
7253 | case NBOUNDA: | |
7254 | case NBOUNDL: | |
7255 | case NBOUNDU: | |
7256 | case OPFAIL: | |
7257 | case SBOL: | |
7258 | case SEOL: | |
7259 | /* These are all 0 width, so match right here or not at all. */ | |
7260 | break; | |
7261 | ||
7262 | default: | |
7263 | Perl_croak(aTHX_ "panic: regrepeat() called with unrecognized node type %d='%s'", OP(p), PL_reg_name[OP(p)]); | |
7264 | assert(0); /* NOTREACHED */ | |
7265 | ||
a0d0e21e | 7266 | } |
a687059c | 7267 | |
a0ed51b3 LW |
7268 | if (hardcount) |
7269 | c = hardcount; | |
7270 | else | |
f73aaa43 DM |
7271 | c = scan - *startposp; |
7272 | *startposp = scan; | |
a687059c | 7273 | |
a3621e74 | 7274 | DEBUG_r({ |
e68ec53f | 7275 | GET_RE_DEBUG_FLAGS_DECL; |
be8e71aa | 7276 | DEBUG_EXECUTE_r({ |
e68ec53f YO |
7277 | SV * const prop = sv_newmortal(); |
7278 | regprop(prog, prop, p); | |
7279 | PerlIO_printf(Perl_debug_log, | |
be8e71aa | 7280 | "%*s %s can match %"IVdf" times out of %"IVdf"...\n", |
e2e6a0f1 | 7281 | REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max); |
a3621e74 | 7282 | }); |
be8e71aa | 7283 | }); |
9041c2e3 | 7284 | |
a0d0e21e | 7285 | return(c); |
a687059c LW |
7286 | } |
7287 | ||
c277df42 | 7288 | |
be8e71aa | 7289 | #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) |
c277df42 | 7290 | /* |
6c6525b8 | 7291 | - regclass_swash - prepare the utf8 swash. Wraps the shared core version to |
e0193e47 KW |
7292 | create a copy so that changes the caller makes won't change the shared one. |
7293 | If <altsvp> is non-null, will return NULL in it, for back-compat. | |
6c6525b8 | 7294 | */ |
ffc61ed2 | 7295 | SV * |
5aaab254 | 7296 | Perl_regclass_swash(pTHX_ const regexp *prog, const regnode* node, bool doinit, SV** listsvp, SV **altsvp) |
ffc61ed2 | 7297 | { |
6c6525b8 | 7298 | PERL_ARGS_ASSERT_REGCLASS_SWASH; |
e0193e47 KW |
7299 | |
7300 | if (altsvp) { | |
7301 | *altsvp = NULL; | |
7302 | } | |
7303 | ||
7304 | return newSVsv(core_regclass_swash(prog, node, doinit, listsvp)); | |
6c6525b8 KW |
7305 | } |
7306 | #endif | |
7307 | ||
7308 | STATIC SV * | |
5aaab254 | 7309 | S_core_regclass_swash(pTHX_ const regexp *prog, const regnode* node, bool doinit, SV** listsvp) |
6c6525b8 | 7310 | { |
8c9eb58f | 7311 | /* Returns the swash for the input 'node' in the regex 'prog'. |
a498e2f8 | 7312 | * If <doinit> is 'true', will attempt to create the swash if not already |
8c9eb58f | 7313 | * done. |
a498e2f8 KW |
7314 | * If <listsvp> is non-null, will return the printable contents of the |
7315 | * swash. This can be used to get debugging information even before the | |
7316 | * swash exists, by calling this function with 'doinit' set to false, in | |
7317 | * which case the components that will be used to eventually create the | |
7318 | * swash are returned (in a printable form). | |
8c9eb58f KW |
7319 | * Tied intimately to how regcomp.c sets up the data structure */ |
7320 | ||
97aff369 | 7321 | dVAR; |
9e55ce06 | 7322 | SV *sw = NULL; |
a498e2f8 | 7323 | SV *si = NULL; /* Input swash initialization string */ |
7a6c6baa KW |
7324 | SV* invlist = NULL; |
7325 | ||
f8fc2ecf YO |
7326 | RXi_GET_DECL(prog,progi); |
7327 | const struct reg_data * const data = prog ? progi->data : NULL; | |
ffc61ed2 | 7328 | |
6c6525b8 | 7329 | PERL_ARGS_ASSERT_CORE_REGCLASS_SWASH; |
7918f24d | 7330 | |
ccb2541c KW |
7331 | assert(ANYOF_NONBITMAP(node)); |
7332 | ||
4f639d21 | 7333 | if (data && data->count) { |
a3b680e6 | 7334 | const U32 n = ARG(node); |
ffc61ed2 | 7335 | |
4f639d21 | 7336 | if (data->what[n] == 's') { |
ad64d0ec NC |
7337 | SV * const rv = MUTABLE_SV(data->data[n]); |
7338 | AV * const av = MUTABLE_AV(SvRV(rv)); | |
2d03de9c | 7339 | SV **const ary = AvARRAY(av); |
87367d5f | 7340 | U8 swash_init_flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; |
9041c2e3 | 7341 | |
8c9eb58f | 7342 | si = *ary; /* ary[0] = the string to initialize the swash with */ |
b11f357e | 7343 | |
88675427 KW |
7344 | /* Elements 2 and 3 are either both present or both absent. [2] is |
7345 | * any inversion list generated at compile time; [3] indicates if | |
7a6c6baa | 7346 | * that inversion list has any user-defined properties in it. */ |
88675427 KW |
7347 | if (av_len(av) >= 2) { |
7348 | invlist = ary[2]; | |
7349 | if (SvUV(ary[3])) { | |
83199d38 KW |
7350 | swash_init_flags |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY; |
7351 | } | |
7a6c6baa KW |
7352 | } |
7353 | else { | |
7354 | invlist = NULL; | |
7a6c6baa KW |
7355 | } |
7356 | ||
8c9eb58f KW |
7357 | /* Element [1] is reserved for the set-up swash. If already there, |
7358 | * return it; if not, create it and store it there */ | |
ce0d59fd | 7359 | if (ary[1] && SvROK(ary[1])) { |
f192cf32 KW |
7360 | sw = ary[1]; |
7361 | } | |
ffc61ed2 | 7362 | else if (si && doinit) { |
7a6c6baa KW |
7363 | |
7364 | sw = _core_swash_init("utf8", /* the utf8 package */ | |
7365 | "", /* nameless */ | |
7366 | si, | |
7367 | 1, /* binary */ | |
7368 | 0, /* not from tr/// */ | |
7a6c6baa | 7369 | invlist, |
83199d38 | 7370 | &swash_init_flags); |
ffc61ed2 JH |
7371 | (void)av_store(av, 1, sw); |
7372 | } | |
7373 | } | |
7374 | } | |
7375 | ||
a498e2f8 | 7376 | /* If requested, return a printable version of what this swash matches */ |
7a6c6baa KW |
7377 | if (listsvp) { |
7378 | SV* matches_string = newSVpvn("", 0); | |
7a6c6baa | 7379 | |
a498e2f8 KW |
7380 | /* The swash should be used, if possible, to get the data, as it |
7381 | * contains the resolved data. But this function can be called at | |
7382 | * compile-time, before everything gets resolved, in which case we | |
7383 | * return the currently best available information, which is the string | |
7384 | * that will eventually be used to do that resolving, 'si' */ | |
872dd7e0 KW |
7385 | if ((! sw || (invlist = _get_swash_invlist(sw)) == NULL) |
7386 | && (si && si != &PL_sv_undef)) | |
7387 | { | |
7a6c6baa KW |
7388 | sv_catsv(matches_string, si); |
7389 | } | |
7390 | ||
7391 | /* Add the inversion list to whatever we have. This may have come from | |
7392 | * the swash, or from an input parameter */ | |
7393 | if (invlist) { | |
7394 | sv_catsv(matches_string, _invlist_contents(invlist)); | |
7395 | } | |
7396 | *listsvp = matches_string; | |
7397 | } | |
7398 | ||
ffc61ed2 JH |
7399 | return sw; |
7400 | } | |
7401 | ||
7402 | /* | |
ba7b4546 | 7403 | - reginclass - determine if a character falls into a character class |
832705d4 | 7404 | |
6698fab5 KW |
7405 | n is the ANYOF regnode |
7406 | p is the target string | |
6698fab5 | 7407 | utf8_target tells whether p is in UTF-8. |
832705d4 | 7408 | |
635cd5d4 | 7409 | Returns true if matched; false otherwise. |
eba1359e | 7410 | |
d5788240 KW |
7411 | Note that this can be a synthetic start class, a combination of various |
7412 | nodes, so things you think might be mutually exclusive, such as locale, | |
7413 | aren't. It can match both locale and non-locale | |
7414 | ||
bbce6d69 | 7415 | */ |
7416 | ||
76e3520e | 7417 | STATIC bool |
272d35c9 | 7418 | S_reginclass(pTHX_ regexp * const prog, const regnode * const n, const U8* const p, const bool utf8_target) |
bbce6d69 | 7419 | { |
27da23d5 | 7420 | dVAR; |
a3b680e6 | 7421 | const char flags = ANYOF_FLAGS(n); |
bbce6d69 | 7422 | bool match = FALSE; |
cc07378b | 7423 | UV c = *p; |
1aa99e6b | 7424 | |
7918f24d NC |
7425 | PERL_ARGS_ASSERT_REGINCLASS; |
7426 | ||
afd2eb18 KW |
7427 | /* If c is not already the code point, get it. Note that |
7428 | * UTF8_IS_INVARIANT() works even if not in UTF-8 */ | |
7429 | if (! UTF8_IS_INVARIANT(c) && utf8_target) { | |
635cd5d4 | 7430 | STRLEN c_len = 0; |
f7ab54c6 | 7431 | c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &c_len, |
6182169b KW |
7432 | (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) |
7433 | | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY); | |
7434 | /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for | |
7435 | * UTF8_ALLOW_FFFF */ | |
f7ab54c6 | 7436 | if (c_len == (STRLEN)-1) |
e8a70c6f | 7437 | Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); |
19f67299 | 7438 | } |
4b3cda86 | 7439 | |
7cdde544 KW |
7440 | /* If this character is potentially in the bitmap, check it */ |
7441 | if (c < 256) { | |
ffc61ed2 JH |
7442 | if (ANYOF_BITMAP_TEST(n, c)) |
7443 | match = TRUE; | |
11454c59 KW |
7444 | else if (flags & ANYOF_NON_UTF8_LATIN1_ALL |
7445 | && ! utf8_target | |
7446 | && ! isASCII(c)) | |
7447 | { | |
7448 | match = TRUE; | |
7449 | } | |
78969a98 | 7450 | else if (flags & ANYOF_LOCALE) { |
b99851e1 KW |
7451 | if (flags & ANYOF_LOC_FOLD) { |
7452 | RXp_MATCH_TAINTED_on(prog); | |
7453 | if (ANYOF_BITMAP_TEST(n, PL_fold_locale[c])) { | |
7454 | match = TRUE; | |
7455 | } | |
7456 | } | |
8efd3f97 | 7457 | else if (ANYOF_POSIXL_TEST_ANY_SET(n)) { |
31c7f561 KW |
7458 | |
7459 | /* The data structure is arranged so bits 0, 2, 4, ... are set | |
7460 | * if the class includes the Posix character class given by | |
7461 | * bit/2; and 1, 3, 5, ... are set if the class includes the | |
7462 | * complemented Posix class given by int(bit/2). So we loop | |
7463 | * through the bits, each time changing whether we complement | |
7464 | * the result or not. Suppose for the sake of illustration | |
7465 | * that bits 0-3 mean respectively, \w, \W, \s, \S. If bit 0 | |
7466 | * is set, it means there is a match for this ANYOF node if the | |
7467 | * character is in the class given by the expression (0 / 2 = 0 | |
7468 | * = \w). If it is in that class, isFOO_lc() will return 1, | |
7469 | * and since 'to_complement' is 0, the result will stay TRUE, | |
7470 | * and we exit the loop. Suppose instead that bit 0 is 0, but | |
7471 | * bit 1 is 1. That means there is a match if the character | |
7472 | * matches \W. We won't bother to call isFOO_lc() on bit 0, | |
7473 | * but will on bit 1. On the second iteration 'to_complement' | |
7474 | * will be 1, so the exclusive or will reverse things, so we | |
7475 | * are testing for \W. On the third iteration, 'to_complement' | |
7476 | * will be 0, and we would be testing for \s; the fourth | |
b0d691b2 KW |
7477 | * iteration would test for \S, etc. |
7478 | * | |
7479 | * Note that this code assumes that all the classes are closed | |
7480 | * under folding. For example, if a character matches \w, then | |
7481 | * its fold does too; and vice versa. This should be true for | |
7482 | * any well-behaved locale for all the currently defined Posix | |
7483 | * classes, except for :lower: and :upper:, which are handled | |
7484 | * by the pseudo-class :cased: which matches if either of the | |
7485 | * other two does. To get rid of this assumption, an outer | |
7486 | * loop could be used below to iterate over both the source | |
7487 | * character, and its fold (if different) */ | |
31c7f561 KW |
7488 | |
7489 | int count = 0; | |
7490 | int to_complement = 0; | |
522b3c1e KW |
7491 | |
7492 | RXp_MATCH_TAINTED_on(prog); | |
31c7f561 | 7493 | while (count < ANYOF_MAX) { |
8efd3f97 | 7494 | if (ANYOF_POSIXL_TEST(n, count) |
31c7f561 KW |
7495 | && to_complement ^ cBOOL(isFOO_lc(count/2, (U8) c))) |
7496 | { | |
7497 | match = TRUE; | |
7498 | break; | |
7499 | } | |
7500 | count++; | |
7501 | to_complement ^= 1; | |
7502 | } | |
ffc61ed2 | 7503 | } |
a0ed51b3 | 7504 | } |
a0ed51b3 LW |
7505 | } |
7506 | ||
7cdde544 | 7507 | /* If the bitmap didn't (or couldn't) match, and something outside the |
1f327b5e | 7508 | * bitmap could match, try that. Locale nodes specify completely the |
de87c4fe | 7509 | * behavior of code points in the bit map (otherwise, a utf8 target would |
c613755a | 7510 | * cause them to be treated as Unicode and not locale), except in |
de87c4fe | 7511 | * the very unlikely event when this node is a synthetic start class, which |
c613755a KW |
7512 | * could be a combination of locale and non-locale nodes. So allow locale |
7513 | * to match for the synthetic start class, which will give a false | |
7514 | * positive that will be resolved when the match is done again as not part | |
7515 | * of the synthetic start class */ | |
ef87b810 | 7516 | if (!match) { |
9b1eb067 | 7517 | if (utf8_target && (flags & ANYOF_ABOVE_LATIN1_ALL) && c >= 256) { |
10ee90d2 | 7518 | match = TRUE; /* Everything above 255 matches */ |
e051a21d | 7519 | } |
6f8d7d0d KW |
7520 | else if (ANYOF_NONBITMAP(n) |
7521 | && ((flags & ANYOF_NONBITMAP_NON_UTF8) | |
7522 | || (utf8_target | |
7523 | && (c >=256 | |
7524 | || (! (flags & ANYOF_LOCALE)) | |
9aa1e39f | 7525 | || OP(n) == ANYOF_SYNTHETIC)))) |
ef87b810 | 7526 | { |
e0193e47 | 7527 | SV * const sw = core_regclass_swash(prog, n, TRUE, 0); |
7cdde544 KW |
7528 | if (sw) { |
7529 | U8 * utf8_p; | |
7530 | if (utf8_target) { | |
7531 | utf8_p = (U8 *) p; | |
e0193e47 KW |
7532 | } else { /* Convert to utf8 */ |
7533 | STRLEN len = 1; | |
7cdde544 KW |
7534 | utf8_p = bytes_to_utf8(p, &len); |
7535 | } | |
f56b6394 | 7536 | |
e0193e47 | 7537 | if (swash_fetch(sw, utf8_p, TRUE)) { |
7cdde544 | 7538 | match = TRUE; |
e0193e47 | 7539 | } |
7cdde544 KW |
7540 | |
7541 | /* If we allocated a string above, free it */ | |
7542 | if (! utf8_target) Safefree(utf8_p); | |
7543 | } | |
7544 | } | |
5073ffbd KW |
7545 | |
7546 | if (UNICODE_IS_SUPER(c) | |
954a2af6 | 7547 | && OP(n) == ANYOF_WARN_SUPER |
5073ffbd KW |
7548 | && ckWARN_d(WARN_NON_UNICODE)) |
7549 | { | |
7550 | Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE), | |
7551 | "Code point 0x%04"UVXf" is not Unicode, all \\p{} matches fail; all \\P{} matches succeed", c); | |
7552 | } | |
7cdde544 KW |
7553 | } |
7554 | ||
f0fdc1c9 KW |
7555 | /* The xor complements the return if to invert: 1^1 = 0, 1^0 = 1 */ |
7556 | return cBOOL(flags & ANYOF_INVERT) ^ match; | |
a0ed51b3 | 7557 | } |
161b471a | 7558 | |
dfe13c55 | 7559 | STATIC U8 * |
ea3daa5d | 7560 | S_reghop3(U8 *s, SSize_t off, const U8* lim) |
9041c2e3 | 7561 | { |
6af86488 KW |
7562 | /* return the position 'off' UTF-8 characters away from 's', forward if |
7563 | * 'off' >= 0, backwards if negative. But don't go outside of position | |
7564 | * 'lim', which better be < s if off < 0 */ | |
7565 | ||
97aff369 | 7566 | dVAR; |
7918f24d NC |
7567 | |
7568 | PERL_ARGS_ASSERT_REGHOP3; | |
7569 | ||
a0ed51b3 | 7570 | if (off >= 0) { |
1aa99e6b | 7571 | while (off-- && s < lim) { |
ffc61ed2 | 7572 | /* XXX could check well-formedness here */ |
a0ed51b3 | 7573 | s += UTF8SKIP(s); |
ffc61ed2 | 7574 | } |
a0ed51b3 LW |
7575 | } |
7576 | else { | |
1de06328 YO |
7577 | while (off++ && s > lim) { |
7578 | s--; | |
7579 | if (UTF8_IS_CONTINUED(*s)) { | |
7580 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
7581 | s--; | |
a0ed51b3 | 7582 | } |
1de06328 | 7583 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
7584 | } |
7585 | } | |
7586 | return s; | |
7587 | } | |
161b471a | 7588 | |
f9f4320a YO |
7589 | #ifdef XXX_dmq |
7590 | /* there are a bunch of places where we use two reghop3's that should | |
7591 | be replaced with this routine. but since thats not done yet | |
7592 | we ifdef it out - dmq | |
7593 | */ | |
dfe13c55 | 7594 | STATIC U8 * |
ea3daa5d | 7595 | S_reghop4(U8 *s, SSize_t off, const U8* llim, const U8* rlim) |
1de06328 YO |
7596 | { |
7597 | dVAR; | |
7918f24d NC |
7598 | |
7599 | PERL_ARGS_ASSERT_REGHOP4; | |
7600 | ||
1de06328 YO |
7601 | if (off >= 0) { |
7602 | while (off-- && s < rlim) { | |
7603 | /* XXX could check well-formedness here */ | |
7604 | s += UTF8SKIP(s); | |
7605 | } | |
7606 | } | |
7607 | else { | |
7608 | while (off++ && s > llim) { | |
7609 | s--; | |
7610 | if (UTF8_IS_CONTINUED(*s)) { | |
7611 | while (s > llim && UTF8_IS_CONTINUATION(*s)) | |
7612 | s--; | |
7613 | } | |
7614 | /* XXX could check well-formedness here */ | |
7615 | } | |
7616 | } | |
7617 | return s; | |
7618 | } | |
f9f4320a | 7619 | #endif |
1de06328 YO |
7620 | |
7621 | STATIC U8 * | |
ea3daa5d | 7622 | S_reghopmaybe3(U8* s, SSize_t off, const U8* lim) |
a0ed51b3 | 7623 | { |
97aff369 | 7624 | dVAR; |
7918f24d NC |
7625 | |
7626 | PERL_ARGS_ASSERT_REGHOPMAYBE3; | |
7627 | ||
a0ed51b3 | 7628 | if (off >= 0) { |
1aa99e6b | 7629 | while (off-- && s < lim) { |
ffc61ed2 | 7630 | /* XXX could check well-formedness here */ |
a0ed51b3 | 7631 | s += UTF8SKIP(s); |
ffc61ed2 | 7632 | } |
a0ed51b3 | 7633 | if (off >= 0) |
3dab1dad | 7634 | return NULL; |
a0ed51b3 LW |
7635 | } |
7636 | else { | |
1de06328 YO |
7637 | while (off++ && s > lim) { |
7638 | s--; | |
7639 | if (UTF8_IS_CONTINUED(*s)) { | |
7640 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
7641 | s--; | |
a0ed51b3 | 7642 | } |
1de06328 | 7643 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
7644 | } |
7645 | if (off <= 0) | |
3dab1dad | 7646 | return NULL; |
a0ed51b3 LW |
7647 | } |
7648 | return s; | |
7649 | } | |
51371543 | 7650 | |
a75351a1 DM |
7651 | |
7652 | /* when executing a regex that may have (?{}), extra stuff needs setting | |
7653 | up that will be visible to the called code, even before the current | |
7654 | match has finished. In particular: | |
7655 | ||
7656 | * $_ is localised to the SV currently being matched; | |
7657 | * pos($_) is created if necessary, ready to be updated on each call-out | |
7658 | to code; | |
7659 | * a fake PMOP is created that can be set to PL_curpm (normally PL_curpm | |
7660 | isn't set until the current pattern is successfully finished), so that | |
7661 | $1 etc of the match-so-far can be seen; | |
7662 | * save the old values of subbeg etc of the current regex, and set then | |
7663 | to the current string (again, this is normally only done at the end | |
7664 | of execution) | |
a75351a1 DM |
7665 | */ |
7666 | ||
7667 | static void | |
7668 | S_setup_eval_state(pTHX_ regmatch_info *const reginfo) | |
7669 | { | |
7670 | MAGIC *mg; | |
7671 | regexp *const rex = ReANY(reginfo->prog); | |
bf2039a9 | 7672 | regmatch_info_aux_eval *eval_state = reginfo->info_aux_eval; |
8adc0f72 | 7673 | |
8adc0f72 | 7674 | eval_state->rex = rex; |
a75351a1 | 7675 | |
a75351a1 DM |
7676 | if (reginfo->sv) { |
7677 | /* Make $_ available to executed code. */ | |
7678 | if (reginfo->sv != DEFSV) { | |
7679 | SAVE_DEFSV; | |
7680 | DEFSV_set(reginfo->sv); | |
7681 | } | |
7682 | ||
96c2a8ff | 7683 | if (!(mg = mg_find_mglob(reginfo->sv))) { |
a75351a1 | 7684 | /* prepare for quick setting of pos */ |
96c2a8ff | 7685 | mg = sv_magicext_mglob(reginfo->sv); |
a75351a1 DM |
7686 | mg->mg_len = -1; |
7687 | } | |
8adc0f72 DM |
7688 | eval_state->pos_magic = mg; |
7689 | eval_state->pos = mg->mg_len; | |
25fdce4a | 7690 | eval_state->pos_flags = mg->mg_flags; |
a75351a1 | 7691 | } |
8adc0f72 DM |
7692 | else |
7693 | eval_state->pos_magic = NULL; | |
7694 | ||
a75351a1 | 7695 | if (!PL_reg_curpm) { |
f65e70f5 DM |
7696 | /* PL_reg_curpm is a fake PMOP that we can attach the current |
7697 | * regex to and point PL_curpm at, so that $1 et al are visible | |
7698 | * within a /(?{})/. It's just allocated once per interpreter the | |
7699 | * first time its needed */ | |
a75351a1 DM |
7700 | Newxz(PL_reg_curpm, 1, PMOP); |
7701 | #ifdef USE_ITHREADS | |
7702 | { | |
7703 | SV* const repointer = &PL_sv_undef; | |
7704 | /* this regexp is also owned by the new PL_reg_curpm, which | |
7705 | will try to free it. */ | |
7706 | av_push(PL_regex_padav, repointer); | |
7707 | PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); | |
7708 | PL_regex_pad = AvARRAY(PL_regex_padav); | |
7709 | } | |
7710 | #endif | |
7711 | } | |
7712 | SET_reg_curpm(reginfo->prog); | |
8adc0f72 | 7713 | eval_state->curpm = PL_curpm; |
a75351a1 DM |
7714 | PL_curpm = PL_reg_curpm; |
7715 | if (RXp_MATCH_COPIED(rex)) { | |
7716 | /* Here is a serious problem: we cannot rewrite subbeg, | |
7717 | since it may be needed if this match fails. Thus | |
7718 | $` inside (?{}) could fail... */ | |
8adc0f72 DM |
7719 | eval_state->subbeg = rex->subbeg; |
7720 | eval_state->sublen = rex->sublen; | |
7721 | eval_state->suboffset = rex->suboffset; | |
a8ee055f | 7722 | eval_state->subcoffset = rex->subcoffset; |
a75351a1 | 7723 | #ifdef PERL_ANY_COW |
8adc0f72 | 7724 | eval_state->saved_copy = rex->saved_copy; |
a75351a1 DM |
7725 | #endif |
7726 | RXp_MATCH_COPIED_off(rex); | |
7727 | } | |
7728 | else | |
8adc0f72 | 7729 | eval_state->subbeg = NULL; |
a75351a1 DM |
7730 | rex->subbeg = (char *)reginfo->strbeg; |
7731 | rex->suboffset = 0; | |
7732 | rex->subcoffset = 0; | |
7733 | rex->sublen = reginfo->strend - reginfo->strbeg; | |
7734 | } | |
7735 | ||
bf2039a9 DM |
7736 | |
7737 | /* destructor to clear up regmatch_info_aux and regmatch_info_aux_eval */ | |
a75351a1 | 7738 | |
51371543 | 7739 | static void |
bf2039a9 | 7740 | S_cleanup_regmatch_info_aux(pTHX_ void *arg) |
51371543 | 7741 | { |
97aff369 | 7742 | dVAR; |
bf2039a9 DM |
7743 | regmatch_info_aux *aux = (regmatch_info_aux *) arg; |
7744 | regmatch_info_aux_eval *eval_state = aux->info_aux_eval; | |
331b2dcc | 7745 | regmatch_slab *s; |
bf2039a9 | 7746 | |
2ac8ff4b DM |
7747 | Safefree(aux->poscache); |
7748 | ||
331b2dcc | 7749 | if (eval_state) { |
bf2039a9 | 7750 | |
331b2dcc | 7751 | /* undo the effects of S_setup_eval_state() */ |
bf2039a9 | 7752 | |
331b2dcc DM |
7753 | if (eval_state->subbeg) { |
7754 | regexp * const rex = eval_state->rex; | |
7755 | rex->subbeg = eval_state->subbeg; | |
7756 | rex->sublen = eval_state->sublen; | |
7757 | rex->suboffset = eval_state->suboffset; | |
7758 | rex->subcoffset = eval_state->subcoffset; | |
db2c6cb3 | 7759 | #ifdef PERL_ANY_COW |
331b2dcc | 7760 | rex->saved_copy = eval_state->saved_copy; |
ed252734 | 7761 | #endif |
331b2dcc DM |
7762 | RXp_MATCH_COPIED_on(rex); |
7763 | } | |
7764 | if (eval_state->pos_magic) | |
25fdce4a | 7765 | { |
331b2dcc | 7766 | eval_state->pos_magic->mg_len = eval_state->pos; |
25fdce4a FC |
7767 | eval_state->pos_magic->mg_flags = |
7768 | (eval_state->pos_magic->mg_flags & ~MGf_BYTES) | |
7769 | | (eval_state->pos_flags & MGf_BYTES); | |
7770 | } | |
331b2dcc DM |
7771 | |
7772 | PL_curpm = eval_state->curpm; | |
8adc0f72 | 7773 | } |
bf2039a9 | 7774 | |
331b2dcc DM |
7775 | PL_regmatch_state = aux->old_regmatch_state; |
7776 | PL_regmatch_slab = aux->old_regmatch_slab; | |
7777 | ||
7778 | /* free all slabs above current one - this must be the last action | |
7779 | * of this function, as aux and eval_state are allocated within | |
7780 | * slabs and may be freed here */ | |
7781 | ||
7782 | s = PL_regmatch_slab->next; | |
7783 | if (s) { | |
7784 | PL_regmatch_slab->next = NULL; | |
7785 | while (s) { | |
7786 | regmatch_slab * const osl = s; | |
7787 | s = s->next; | |
7788 | Safefree(osl); | |
7789 | } | |
7790 | } | |
51371543 | 7791 | } |
33b8afdf | 7792 | |
8adc0f72 | 7793 | |
33b8afdf | 7794 | STATIC void |
5aaab254 | 7795 | S_to_utf8_substr(pTHX_ regexp *prog) |
33b8afdf | 7796 | { |
7e0d5ad7 KW |
7797 | /* Converts substr fields in prog from bytes to UTF-8, calling fbm_compile |
7798 | * on the converted value */ | |
7799 | ||
a1cac82e | 7800 | int i = 1; |
7918f24d NC |
7801 | |
7802 | PERL_ARGS_ASSERT_TO_UTF8_SUBSTR; | |
7803 | ||
a1cac82e NC |
7804 | do { |
7805 | if (prog->substrs->data[i].substr | |
7806 | && !prog->substrs->data[i].utf8_substr) { | |
7807 | SV* const sv = newSVsv(prog->substrs->data[i].substr); | |
7808 | prog->substrs->data[i].utf8_substr = sv; | |
7809 | sv_utf8_upgrade(sv); | |
610460f9 | 7810 | if (SvVALID(prog->substrs->data[i].substr)) { |
cffe132d | 7811 | if (SvTAIL(prog->substrs->data[i].substr)) { |
610460f9 NC |
7812 | /* Trim the trailing \n that fbm_compile added last |
7813 | time. */ | |
7814 | SvCUR_set(sv, SvCUR(sv) - 1); | |
7815 | /* Whilst this makes the SV technically "invalid" (as its | |
7816 | buffer is no longer followed by "\0") when fbm_compile() | |
7817 | adds the "\n" back, a "\0" is restored. */ | |
cffe132d NC |
7818 | fbm_compile(sv, FBMcf_TAIL); |
7819 | } else | |
7820 | fbm_compile(sv, 0); | |
610460f9 | 7821 | } |
a1cac82e NC |
7822 | if (prog->substrs->data[i].substr == prog->check_substr) |
7823 | prog->check_utf8 = sv; | |
7824 | } | |
7825 | } while (i--); | |
33b8afdf JH |
7826 | } |
7827 | ||
7e0d5ad7 | 7828 | STATIC bool |
5aaab254 | 7829 | S_to_byte_substr(pTHX_ regexp *prog) |
33b8afdf | 7830 | { |
7e0d5ad7 KW |
7831 | /* Converts substr fields in prog from UTF-8 to bytes, calling fbm_compile |
7832 | * on the converted value; returns FALSE if can't be converted. */ | |
7833 | ||
97aff369 | 7834 | dVAR; |
a1cac82e | 7835 | int i = 1; |
7918f24d NC |
7836 | |
7837 | PERL_ARGS_ASSERT_TO_BYTE_SUBSTR; | |
7838 | ||
a1cac82e NC |
7839 | do { |
7840 | if (prog->substrs->data[i].utf8_substr | |
7841 | && !prog->substrs->data[i].substr) { | |
7842 | SV* sv = newSVsv(prog->substrs->data[i].utf8_substr); | |
7e0d5ad7 KW |
7843 | if (! sv_utf8_downgrade(sv, TRUE)) { |
7844 | return FALSE; | |
7845 | } | |
5400f398 KW |
7846 | if (SvVALID(prog->substrs->data[i].utf8_substr)) { |
7847 | if (SvTAIL(prog->substrs->data[i].utf8_substr)) { | |
7848 | /* Trim the trailing \n that fbm_compile added last | |
7849 | time. */ | |
7850 | SvCUR_set(sv, SvCUR(sv) - 1); | |
7851 | fbm_compile(sv, FBMcf_TAIL); | |
7852 | } else | |
7853 | fbm_compile(sv, 0); | |
7854 | } | |
a1cac82e NC |
7855 | prog->substrs->data[i].substr = sv; |
7856 | if (prog->substrs->data[i].utf8_substr == prog->check_utf8) | |
7857 | prog->check_substr = sv; | |
33b8afdf | 7858 | } |
a1cac82e | 7859 | } while (i--); |
7e0d5ad7 KW |
7860 | |
7861 | return TRUE; | |
33b8afdf | 7862 | } |
66610fdd RGS |
7863 | |
7864 | /* | |
7865 | * Local variables: | |
7866 | * c-indentation-style: bsd | |
7867 | * c-basic-offset: 4 | |
14d04a33 | 7868 | * indent-tabs-mode: nil |
66610fdd RGS |
7869 | * End: |
7870 | * | |
14d04a33 | 7871 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 7872 | */ |