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 | |
ef07e810 | 83 | #define RF_tainted 1 /* tainted information used? e.g. locale */ |
c277df42 | 84 | #define RF_warned 2 /* warned about big count? */ |
faec1544 | 85 | |
ab3bbdeb | 86 | #define RF_utf8 8 /* Pattern contains multibyte chars? */ |
a0ed51b3 | 87 | |
f2ed9b32 | 88 | #define UTF_PATTERN ((PL_reg_flags & RF_utf8) != 0) |
ce862d02 | 89 | |
a687059c LW |
90 | #ifndef STATIC |
91 | #define STATIC static | |
92 | #endif | |
93 | ||
7e2509c1 KW |
94 | /* Valid for non-utf8 strings, non-ANYOFV nodes only: avoids the reginclass |
95 | * call if there are no complications: i.e., if everything matchable is | |
96 | * straight forward in the bitmap */ | |
af364d03 KW |
97 | #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) \ |
98 | : ANYOF_BITMAP_TEST(p,*(c))) | |
7d3e948e | 99 | |
c277df42 IZ |
100 | /* |
101 | * Forwards. | |
102 | */ | |
103 | ||
f2ed9b32 | 104 | #define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv)) |
53c4c00c | 105 | #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b) |
a0ed51b3 | 106 | |
3dab1dad YO |
107 | #define HOPc(pos,off) \ |
108 | (char *)(PL_reg_match_utf8 \ | |
52657f30 | 109 | ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \ |
3dab1dad YO |
110 | : (U8*)(pos + off)) |
111 | #define HOPBACKc(pos, off) \ | |
07be1b83 YO |
112 | (char*)(PL_reg_match_utf8\ |
113 | ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \ | |
114 | : (pos - off >= PL_bostr) \ | |
8e11feef | 115 | ? (U8*)pos - off \ |
3dab1dad | 116 | : NULL) |
efb30f32 | 117 | |
e7409c1b | 118 | #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off)) |
1aa99e6b | 119 | #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim)) |
1aa99e6b | 120 | |
20d0b1e9 | 121 | /* these are unrolled below in the CCC_TRY_XXX defined */ |
a9a337f8 KW |
122 | #ifdef EBCDIC |
123 | /* Often 'str' is a hard-coded utf8 string instead of utfebcdic. so just | |
124 | * skip the check on EBCDIC platforms */ | |
125 | # define LOAD_UTF8_CHARCLASS(class,str) LOAD_UTF8_CHARCLASS_NO_CHECK(class) | |
126 | #else | |
127 | # define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ | |
9c4fdda1 | 128 | if (!CAT2(PL_utf8_,class)) { \ |
cf54f63a | 129 | bool ok; \ |
9c4fdda1 | 130 | ENTER; save_re_context(); \ |
cf54f63a JL |
131 | ok=CAT2(is_utf8_,class)((const U8*)str); \ |
132 | PERL_UNUSED_VAR(ok); \ | |
133 | assert(ok); assert(CAT2(PL_utf8_,class)); LEAVE; } } STMT_END | |
a9a337f8 | 134 | #endif |
37e2e78e KW |
135 | |
136 | /* Doesn't do an assert to verify that is correct */ | |
137 | #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \ | |
9c4fdda1 | 138 | if (!CAT2(PL_utf8_,class)) { \ |
9d63fa07 | 139 | bool throw_away PERL_UNUSED_DECL; \ |
9c4fdda1 RB |
140 | ENTER; save_re_context(); \ |
141 | throw_away = CAT2(is_utf8_,class)((const U8*)" "); \ | |
142 | LEAVE; } } STMT_END | |
37e2e78e | 143 | |
1a4fad37 AL |
144 | #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a") |
145 | #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0") | |
146 | #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ") | |
51371543 | 147 | |
37e2e78e | 148 | #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \ |
d275fa5e CB |
149 | LOAD_UTF8_CHARCLASS(X_begin, " "); \ |
150 | LOAD_UTF8_CHARCLASS(X_non_hangul, "A"); \ | |
37e2e78e KW |
151 | /* These are utf8 constants, and not utf-ebcdic constants, so the \ |
152 | * assert should likely and hopefully fail on an EBCDIC machine */ \ | |
d275fa5e | 153 | LOAD_UTF8_CHARCLASS(X_extend, "\xcc\x80"); /* U+0300 */ \ |
37e2e78e KW |
154 | \ |
155 | /* No asserts are done for these, in case called on an early \ | |
156 | * Unicode version in which they map to nothing */ \ | |
d275fa5e CB |
157 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_prepend);/* U+0E40 "\xe0\xb9\x80" */ \ |
158 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_L); /* U+1100 "\xe1\x84\x80" */ \ | |
159 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV); /* U+AC00 "\xea\xb0\x80" */ \ | |
160 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LVT); /* U+AC01 "\xea\xb0\x81" */ \ | |
161 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV_LVT_V);/* U+AC01 "\xea\xb0\x81" */\ | |
162 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_T); /* U+11A8 "\xe1\x86\xa8" */ \ | |
37e2e78e | 163 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_V) /* U+1160 "\xe1\x85\xa0" */ |
20d0b1e9 | 164 | |
1dcf4a1b | 165 | #define PLACEHOLDER /* Something for the preprocessor to grab onto */ |
d1eb3177 | 166 | |
ee9a90b8 KW |
167 | /* The actual code for CCC_TRY, which uses several variables from the routine |
168 | * it's callable from. It is designed to be the bulk of a case statement. | |
169 | * FUNC is the macro or function to call on non-utf8 targets that indicate if | |
170 | * nextchr matches the class. | |
171 | * UTF8_TEST is the whole test string to use for utf8 targets | |
172 | * LOAD is what to use to test, and if not present to load in the swash for the | |
173 | * class | |
174 | * POS_OR_NEG is either empty or ! to complement the results of FUNC or | |
175 | * UTF8_TEST test. | |
176 | * The logic is: Fail if we're at the end-of-string; otherwise if the target is | |
177 | * utf8 and a variant, load the swash if necessary and test using the utf8 | |
178 | * test. Advance to the next character if test is ok, otherwise fail; If not | |
179 | * utf8 or an invariant under utf8, use the non-utf8 test, and fail if it | |
180 | * fails, or advance to the next character */ | |
181 | ||
182 | #define _CCC_TRY_CODE(POS_OR_NEG, FUNC, UTF8_TEST, CLASS, STR) \ | |
183 | if (locinput >= PL_regeol) { \ | |
184 | sayNO; \ | |
185 | } \ | |
186 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ | |
187 | LOAD_UTF8_CHARCLASS(CLASS, STR); \ | |
188 | if (POS_OR_NEG (UTF8_TEST)) { \ | |
189 | sayNO; \ | |
190 | } \ | |
191 | locinput += PL_utf8skip[nextchr]; \ | |
192 | nextchr = UCHARAT(locinput); \ | |
193 | break; \ | |
194 | } \ | |
195 | if (POS_OR_NEG (FUNC(nextchr))) { \ | |
196 | sayNO; \ | |
197 | } \ | |
198 | nextchr = UCHARAT(++locinput); \ | |
980866de KW |
199 | break; |
200 | ||
ee9a90b8 KW |
201 | /* Handle the non-locale cases for a character class and its complement. It |
202 | * calls _CCC_TRY_CODE with a ! to complement the test for the character class. | |
203 | * This is because that code fails when the test succeeds, so we want to have | |
204 | * the test fail so that the code succeeds. The swash is stored in a | |
205 | * predictable PL_ place */ | |
cfaf538b KW |
206 | #define _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, \ |
207 | CLASS, STR) \ | |
ee9a90b8 KW |
208 | case NAME: \ |
209 | _CCC_TRY_CODE( !, FUNC, \ | |
210 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ | |
211 | (U8*)locinput, TRUE)), \ | |
212 | CLASS, STR) \ | |
213 | case NNAME: \ | |
1dcf4a1b | 214 | _CCC_TRY_CODE( PLACEHOLDER , FUNC, \ |
ee9a90b8 KW |
215 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ |
216 | (U8*)locinput, TRUE)), \ | |
217 | CLASS, STR) \ | |
218 | ||
219 | /* Generate the case statements for both locale and non-locale character | |
220 | * classes in regmatch for classes that don't have special unicode semantics. | |
221 | * Locales don't use an immediate swash, but an intermediary special locale | |
222 | * function that is called on the pointer to the current place in the input | |
223 | * string. That function will resolve to needing the same swash. One might | |
224 | * think that because we don't know what the locale will match, we shouldn't | |
225 | * check with the swash loading function that it loaded properly; ie, that we | |
226 | * should use LOAD_UTF8_CHARCLASS_NO_CHECK for those, but what is passed to the | |
227 | * regular LOAD_UTF8_CHARCLASS is in non-locale terms, and so locale is | |
228 | * irrelevant here */ | |
229 | #define CCC_TRY(NAME, NNAME, FUNC, \ | |
230 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
cfaf538b | 231 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 KW |
232 | CLASS, STR) \ |
233 | case NAMEL: \ | |
234 | PL_reg_flags |= RF_tainted; \ | |
235 | _CCC_TRY_CODE( !, LCFUNC, LCFUNC_utf8((U8*)locinput), CLASS, STR) \ | |
236 | case NNAMEL: \ | |
237 | PL_reg_flags |= RF_tainted; \ | |
1dcf4a1b KW |
238 | _CCC_TRY_CODE( PLACEHOLDER, LCFUNC, LCFUNC_utf8((U8*)locinput), \ |
239 | CLASS, STR) \ | |
cfaf538b KW |
240 | case NAMEA: \ |
241 | if (locinput >= PL_regeol || ! FUNCA(nextchr)) { \ | |
242 | sayNO; \ | |
243 | } \ | |
244 | /* Matched a utf8-invariant, so don't have to worry about utf8 */ \ | |
245 | nextchr = UCHARAT(++locinput); \ | |
246 | break; \ | |
247 | case NNAMEA: \ | |
248 | if (locinput >= PL_regeol || FUNCA(nextchr)) { \ | |
249 | sayNO; \ | |
250 | } \ | |
251 | if (utf8_target) { \ | |
252 | locinput += PL_utf8skip[nextchr]; \ | |
253 | nextchr = UCHARAT(locinput); \ | |
254 | } \ | |
255 | else { \ | |
256 | nextchr = UCHARAT(++locinput); \ | |
257 | } \ | |
258 | break; \ | |
ee9a90b8 KW |
259 | /* Generate the non-locale cases */ \ |
260 | _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, CLASS, STR) | |
261 | ||
262 | /* This is like CCC_TRY, but has an extra set of parameters for generating case | |
263 | * statements to handle separate Unicode semantics nodes */ | |
264 | #define CCC_TRY_U(NAME, NNAME, FUNC, \ | |
265 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
266 | NAMEU, NNAMEU, FUNCU, \ | |
cfaf538b | 267 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 | 268 | CLASS, STR) \ |
cfaf538b KW |
269 | CCC_TRY(NAME, NNAME, FUNC, \ |
270 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
271 | NAMEA, NNAMEA, FUNCA, \ | |
272 | CLASS, STR) \ | |
ee9a90b8 | 273 | _CCC_TRY_NONLOCALE(NAMEU, NNAMEU, FUNCU, CLASS, STR) |
d1eb3177 | 274 | |
3dab1dad YO |
275 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
276 | ||
5f80c4cf | 277 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
c35dcbe2 YO |
278 | /* it would be nice to rework regcomp.sym to generate this stuff. sigh |
279 | * | |
280 | * NOTE that *nothing* that affects backtracking should be in here, specifically | |
281 | * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a | |
282 | * node that is in between two EXACT like nodes when ascertaining what the required | |
283 | * "follow" character is. This should probably be moved to regex compile time | |
284 | * although it may be done at run time beause of the REF possibility - more | |
285 | * investigation required. -- demerphq | |
286 | */ | |
3e901dc0 YO |
287 | #define JUMPABLE(rn) ( \ |
288 | OP(rn) == OPEN || \ | |
289 | (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \ | |
290 | OP(rn) == EVAL || \ | |
cca55fe3 JP |
291 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
292 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
d1c771f5 | 293 | OP(rn) == KEEPS || \ |
3dab1dad | 294 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 | 295 | ) |
ee9b8eae | 296 | #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT) |
e2d8ce26 | 297 | |
ee9b8eae YO |
298 | #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF ) |
299 | ||
300 | #if 0 | |
301 | /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so | |
302 | we don't need this definition. */ | |
303 | #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) | |
fab2782b | 304 | #define IS_TEXTF(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFU_TRICKYFOLD || OP(rn)==EXACTFA || OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF ) |
ee9b8eae YO |
305 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) |
306 | ||
307 | #else | |
308 | /* ... so we use this as its faster. */ | |
309 | #define IS_TEXT(rn) ( OP(rn)==EXACT ) | |
fab2782b | 310 | #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFU_TRICKYFOLD || OP(rn) == EXACTFA) |
ee9b8eae YO |
311 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) |
312 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) | |
313 | ||
314 | #endif | |
e2d8ce26 | 315 | |
a84d97b6 HS |
316 | /* |
317 | Search for mandatory following text node; for lookahead, the text must | |
318 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
319 | */ | |
cca55fe3 | 320 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
321 | while (JUMPABLE(rn)) { \ |
322 | const OPCODE type = OP(rn); \ | |
323 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 324 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 325 | else if (type == PLUS) \ |
cca55fe3 | 326 | rn = NEXTOPER(rn); \ |
3dab1dad | 327 | else if (type == IFMATCH) \ |
a84d97b6 | 328 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 329 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 330 | } \ |
5f80c4cf | 331 | } STMT_END |
74750237 | 332 | |
c476f425 | 333 | |
acfe0abc | 334 | static void restore_pos(pTHX_ void *arg); |
51371543 | 335 | |
87c0511b | 336 | #define REGCP_PAREN_ELEMS 3 |
f067efbf | 337 | #define REGCP_OTHER_ELEMS 3 |
e0fa7e2b | 338 | #define REGCP_FRAME_ELEMS 1 |
620d5b66 NC |
339 | /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and |
340 | * are needed for the regexp context stack bookkeeping. */ | |
341 | ||
76e3520e | 342 | STATIC CHECKPOINT |
b93070ed | 343 | S_regcppush(pTHX_ const regexp *rex, I32 parenfloor) |
a0d0e21e | 344 | { |
97aff369 | 345 | dVAR; |
a3b680e6 | 346 | const int retval = PL_savestack_ix; |
a3b680e6 | 347 | const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; |
e0fa7e2b NC |
348 | const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS; |
349 | const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT; | |
87c0511b | 350 | I32 p; |
40a82448 | 351 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 352 | |
b93070ed DM |
353 | PERL_ARGS_ASSERT_REGCPPUSH; |
354 | ||
e49a9654 | 355 | if (paren_elems_to_push < 0) |
5637ef5b NC |
356 | Perl_croak(aTHX_ "panic: paren_elems_to_push, %i < 0", |
357 | paren_elems_to_push); | |
e49a9654 | 358 | |
e0fa7e2b NC |
359 | if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems) |
360 | Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf | |
5df417d0 JH |
361 | " out of range (%lu-%ld)", |
362 | total_elems, (unsigned long)PL_regsize, (long)parenfloor); | |
e0fa7e2b | 363 | |
620d5b66 | 364 | SSGROW(total_elems + REGCP_FRAME_ELEMS); |
7f69552c | 365 | |
495f47a5 DM |
366 | DEBUG_BUFFERS_r( |
367 | if ((int)PL_regsize > (int)parenfloor) | |
368 | PerlIO_printf(Perl_debug_log, | |
369 | "rex=0x%"UVxf" offs=0x%"UVxf": saving capture indices:\n", | |
370 | PTR2UV(rex), | |
371 | PTR2UV(rex->offs) | |
372 | ); | |
373 | ); | |
87c0511b | 374 | for (p = parenfloor+1; p <= (I32)PL_regsize; p++) { |
b1ce53c5 | 375 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
b93070ed DM |
376 | SSPUSHINT(rex->offs[p].end); |
377 | SSPUSHINT(rex->offs[p].start); | |
1ca2007e | 378 | SSPUSHINT(rex->offs[p].start_tmp); |
e7707071 | 379 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
495f47a5 DM |
380 | " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"\n", |
381 | (UV)p, | |
382 | (IV)rex->offs[p].start, | |
383 | (IV)rex->offs[p].start_tmp, | |
384 | (IV)rex->offs[p].end | |
40a82448 | 385 | )); |
a0d0e21e | 386 | } |
b1ce53c5 | 387 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
3280af22 | 388 | SSPUSHINT(PL_regsize); |
b93070ed DM |
389 | SSPUSHINT(rex->lastparen); |
390 | SSPUSHINT(rex->lastcloseparen); | |
e0fa7e2b | 391 | SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */ |
41123dfd | 392 | |
a0d0e21e LW |
393 | return retval; |
394 | } | |
395 | ||
c277df42 | 396 | /* These are needed since we do not localize EVAL nodes: */ |
ab3bbdeb YO |
397 | #define REGCP_SET(cp) \ |
398 | DEBUG_STATE_r( \ | |
ab3bbdeb | 399 | PerlIO_printf(Perl_debug_log, \ |
e4f74956 | 400 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
ab3bbdeb YO |
401 | (IV)PL_savestack_ix)); \ |
402 | cp = PL_savestack_ix | |
c3464db5 | 403 | |
ab3bbdeb | 404 | #define REGCP_UNWIND(cp) \ |
e4f74956 | 405 | DEBUG_STATE_r( \ |
ab3bbdeb | 406 | if (cp != PL_savestack_ix) \ |
e4f74956 YO |
407 | PerlIO_printf(Perl_debug_log, \ |
408 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ | |
ab3bbdeb YO |
409 | (IV)(cp), (IV)PL_savestack_ix)); \ |
410 | regcpblow(cp) | |
c277df42 | 411 | |
a8d1f4b4 DM |
412 | #define UNWIND_PAREN(lp, lcp) \ |
413 | for (n = rex->lastparen; n > lp; n--) \ | |
414 | rex->offs[n].end = -1; \ | |
415 | rex->lastparen = n; \ | |
416 | rex->lastcloseparen = lcp; | |
417 | ||
418 | ||
f067efbf | 419 | STATIC void |
b93070ed | 420 | S_regcppop(pTHX_ regexp *rex) |
a0d0e21e | 421 | { |
97aff369 | 422 | dVAR; |
e0fa7e2b | 423 | UV i; |
87c0511b | 424 | U32 paren; |
a3621e74 YO |
425 | GET_RE_DEBUG_FLAGS_DECL; |
426 | ||
7918f24d NC |
427 | PERL_ARGS_ASSERT_REGCPPOP; |
428 | ||
b1ce53c5 | 429 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
c6bf6a65 | 430 | i = SSPOPUV; |
e0fa7e2b NC |
431 | assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
432 | i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */ | |
b93070ed DM |
433 | rex->lastcloseparen = SSPOPINT; |
434 | rex->lastparen = SSPOPINT; | |
3280af22 | 435 | PL_regsize = SSPOPINT; |
b1ce53c5 | 436 | |
620d5b66 | 437 | i -= REGCP_OTHER_ELEMS; |
b1ce53c5 | 438 | /* Now restore the parentheses context. */ |
495f47a5 DM |
439 | DEBUG_BUFFERS_r( |
440 | if (i || rex->lastparen + 1 <= rex->nparens) | |
441 | PerlIO_printf(Perl_debug_log, | |
442 | "rex=0x%"UVxf" offs=0x%"UVxf": restoring capture indices to:\n", | |
443 | PTR2UV(rex), | |
444 | PTR2UV(rex->offs) | |
445 | ); | |
446 | ); | |
87c0511b | 447 | paren = PL_regsize; |
620d5b66 | 448 | for ( ; i > 0; i -= REGCP_PAREN_ELEMS) { |
1df70142 | 449 | I32 tmps; |
1ca2007e | 450 | rex->offs[paren].start_tmp = SSPOPINT; |
b93070ed | 451 | rex->offs[paren].start = SSPOPINT; |
cf93c79d | 452 | tmps = SSPOPINT; |
b93070ed DM |
453 | if (paren <= rex->lastparen) |
454 | rex->offs[paren].end = tmps; | |
495f47a5 DM |
455 | DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log, |
456 | " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"%s\n", | |
457 | (UV)paren, | |
458 | (IV)rex->offs[paren].start, | |
459 | (IV)rex->offs[paren].start_tmp, | |
460 | (IV)rex->offs[paren].end, | |
461 | (paren > rex->lastparen ? "(skipped)" : "")); | |
c277df42 | 462 | ); |
87c0511b | 463 | paren--; |
a0d0e21e | 464 | } |
daf18116 | 465 | #if 1 |
dafc8851 JH |
466 | /* It would seem that the similar code in regtry() |
467 | * already takes care of this, and in fact it is in | |
468 | * a better location to since this code can #if 0-ed out | |
469 | * but the code in regtry() is needed or otherwise tests | |
470 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
471 | * (as of patchlevel 7877) will fail. Then again, |
472 | * this code seems to be necessary or otherwise | |
225593e1 DM |
473 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ |
474 | * --jhi updated by dapm */ | |
b93070ed | 475 | for (i = rex->lastparen + 1; i <= rex->nparens; i++) { |
097eb12c | 476 | if (i > PL_regsize) |
b93070ed DM |
477 | rex->offs[i].start = -1; |
478 | rex->offs[i].end = -1; | |
495f47a5 DM |
479 | DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log, |
480 | " \\%"UVuf": %s ..-1 undeffing\n", | |
481 | (UV)i, | |
482 | (i > PL_regsize) ? "-1" : " " | |
483 | )); | |
a0d0e21e | 484 | } |
dafc8851 | 485 | #endif |
a0d0e21e LW |
486 | } |
487 | ||
74088413 DM |
488 | /* restore the parens and associated vars at savestack position ix, |
489 | * but without popping the stack */ | |
490 | ||
491 | STATIC void | |
492 | S_regcp_restore(pTHX_ regexp *rex, I32 ix) | |
493 | { | |
494 | I32 tmpix = PL_savestack_ix; | |
495 | PL_savestack_ix = ix; | |
496 | regcppop(rex); | |
497 | PL_savestack_ix = tmpix; | |
498 | } | |
499 | ||
02db2b7b | 500 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 501 | |
a687059c | 502 | /* |
e50aee73 | 503 | * pregexec and friends |
a687059c LW |
504 | */ |
505 | ||
76234dfb | 506 | #ifndef PERL_IN_XSUB_RE |
a687059c | 507 | /* |
c277df42 | 508 | - pregexec - match a regexp against a string |
a687059c | 509 | */ |
c277df42 | 510 | I32 |
49d7dfbc | 511 | Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend, |
c3464db5 | 512 | char *strbeg, I32 minend, SV *screamer, U32 nosave) |
c277df42 IZ |
513 | /* strend: pointer to null at end of string */ |
514 | /* strbeg: real beginning of string */ | |
515 | /* minend: end of match must be >=minend after stringarg. */ | |
516 | /* nosave: For optimizations. */ | |
517 | { | |
7918f24d NC |
518 | PERL_ARGS_ASSERT_PREGEXEC; |
519 | ||
c277df42 | 520 | return |
9041c2e3 | 521 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
522 | nosave ? 0 : REXEC_COPY_STR); |
523 | } | |
76234dfb | 524 | #endif |
22e551b9 | 525 | |
9041c2e3 | 526 | /* |
cad2e5aa JH |
527 | * Need to implement the following flags for reg_anch: |
528 | * | |
529 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
530 | * USE_INTUIT_ML | |
531 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
532 | * INTUIT_AUTORITATIVE_ML | |
533 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
534 | * INTUIT_ONCE_ML | |
535 | * | |
536 | * Another flag for this function: SECOND_TIME (so that float substrs | |
537 | * with giant delta may be not rechecked). | |
538 | */ | |
539 | ||
540 | /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ | |
541 | ||
3f7c398e | 542 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
543 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
544 | ||
545 | /* XXXX We assume that strpos is strbeg unless sv. */ | |
546 | ||
6eb5f6b9 JH |
547 | /* XXXX Some places assume that there is a fixed substring. |
548 | An update may be needed if optimizer marks as "INTUITable" | |
549 | RExen without fixed substrings. Similarly, it is assumed that | |
550 | lengths of all the strings are no more than minlen, thus they | |
551 | cannot come from lookahead. | |
40d049e4 YO |
552 | (Or minlen should take into account lookahead.) |
553 | NOTE: Some of this comment is not correct. minlen does now take account | |
554 | of lookahead/behind. Further research is required. -- demerphq | |
555 | ||
556 | */ | |
6eb5f6b9 | 557 | |
2c2d71f5 JH |
558 | /* A failure to find a constant substring means that there is no need to make |
559 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
560 | finding a substring too deep into the string means that less calls to | |
30944b6d IZ |
561 | regtry() should be needed. |
562 | ||
563 | REx compiler's optimizer found 4 possible hints: | |
564 | a) Anchored substring; | |
565 | b) Fixed substring; | |
566 | c) Whether we are anchored (beginning-of-line or \G); | |
486ec47a | 567 | d) First node (of those at offset 0) which may distinguish positions; |
6eb5f6b9 | 568 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
569 | string which does not contradict any of them. |
570 | */ | |
2c2d71f5 | 571 | |
6eb5f6b9 JH |
572 | /* Most of decisions we do here should have been done at compile time. |
573 | The nodes of the REx which we used for the search should have been | |
574 | deleted from the finite automaton. */ | |
575 | ||
cad2e5aa | 576 | char * |
288b8c02 | 577 | Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, |
9f61653a | 578 | char *strend, const U32 flags, re_scream_pos_data *data) |
cad2e5aa | 579 | { |
97aff369 | 580 | dVAR; |
288b8c02 | 581 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
b7953727 | 582 | register I32 start_shift = 0; |
cad2e5aa | 583 | /* Should be nonnegative! */ |
b7953727 | 584 | register I32 end_shift = 0; |
2c2d71f5 JH |
585 | register char *s; |
586 | register SV *check; | |
a1933d95 | 587 | char *strbeg; |
cad2e5aa | 588 | char *t; |
f2ed9b32 | 589 | const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 590 | I32 ml_anch; |
bd61b366 SS |
591 | register char *other_last = NULL; /* other substr checked before this */ |
592 | char *check_at = NULL; /* check substr found at this pos */ | |
d8080198 | 593 | char *checked_upto = NULL; /* how far into the string we have already checked using find_byclass*/ |
bbe252da | 594 | const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; |
f8fc2ecf | 595 | RXi_GET_DECL(prog,progi); |
30944b6d | 596 | #ifdef DEBUGGING |
890ce7af | 597 | const char * const i_strpos = strpos; |
30944b6d | 598 | #endif |
a3621e74 YO |
599 | GET_RE_DEBUG_FLAGS_DECL; |
600 | ||
7918f24d | 601 | PERL_ARGS_ASSERT_RE_INTUIT_START; |
c33e64f0 FC |
602 | PERL_UNUSED_ARG(flags); |
603 | PERL_UNUSED_ARG(data); | |
7918f24d | 604 | |
f2ed9b32 | 605 | RX_MATCH_UTF8_set(rx,utf8_target); |
cad2e5aa | 606 | |
3c8556c3 | 607 | if (RX_UTF8(rx)) { |
b8d68ded JH |
608 | PL_reg_flags |= RF_utf8; |
609 | } | |
ab3bbdeb | 610 | DEBUG_EXECUTE_r( |
f2ed9b32 | 611 | debug_start_match(rx, utf8_target, strpos, strend, |
1de06328 YO |
612 | sv ? "Guessing start of match in sv for" |
613 | : "Guessing start of match in string for"); | |
2a782b5b | 614 | ); |
cad2e5aa | 615 | |
c344f387 JH |
616 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
617 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 618 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 619 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 620 | goto fail; |
2c2d71f5 | 621 | } |
1de06328 | 622 | |
a1933d95 | 623 | strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; |
1aa99e6b | 624 | PL_regeol = strend; |
f2ed9b32 | 625 | if (utf8_target) { |
33b8afdf JH |
626 | if (!prog->check_utf8 && prog->check_substr) |
627 | to_utf8_substr(prog); | |
628 | check = prog->check_utf8; | |
629 | } else { | |
630 | if (!prog->check_substr && prog->check_utf8) | |
631 | to_byte_substr(prog); | |
632 | check = prog->check_substr; | |
633 | } | |
1de06328 | 634 | if (check == &PL_sv_undef) { |
a3621e74 | 635 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
1de06328 | 636 | "Non-utf8 string cannot match utf8 check string\n")); |
33b8afdf JH |
637 | goto fail; |
638 | } | |
bbe252da YO |
639 | if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */ |
640 | ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) | |
641 | || ( (prog->extflags & RXf_ANCH_BOL) | |
7fba1cd6 | 642 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 643 | |
7e25d62c | 644 | if (!ml_anch) { |
bbe252da YO |
645 | if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */ |
646 | && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ | |
3f7c398e | 647 | /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ |
7e25d62c JH |
648 | && sv && !SvROK(sv) |
649 | && (strpos != strbeg)) { | |
a3621e74 | 650 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
651 | goto fail; |
652 | } | |
653 | if (prog->check_offset_min == prog->check_offset_max && | |
bbe252da | 654 | !(prog->extflags & RXf_CANY_SEEN)) { |
2c2d71f5 | 655 | /* Substring at constant offset from beg-of-str... */ |
cad2e5aa JH |
656 | I32 slen; |
657 | ||
1aa99e6b | 658 | s = HOP3c(strpos, prog->check_offset_min, strend); |
1de06328 | 659 | |
653099ff GS |
660 | if (SvTAIL(check)) { |
661 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 662 | |
9041c2e3 | 663 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 664 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 665 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 666 | goto fail_finish; |
cad2e5aa JH |
667 | } |
668 | /* Now should match s[0..slen-2] */ | |
669 | slen--; | |
3f7c398e | 670 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 671 | || (slen > 1 |
3f7c398e | 672 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 673 | report_neq: |
a3621e74 | 674 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
675 | goto fail_finish; |
676 | } | |
cad2e5aa | 677 | } |
3f7c398e | 678 | else if (*SvPVX_const(check) != *s |
653099ff | 679 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 680 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 681 | goto report_neq; |
c315bfe8 | 682 | check_at = s; |
2c2d71f5 | 683 | goto success_at_start; |
7e25d62c | 684 | } |
cad2e5aa | 685 | } |
2c2d71f5 | 686 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 687 | s = strpos; |
2c2d71f5 | 688 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
1de06328 YO |
689 | end_shift = prog->check_end_shift; |
690 | ||
2c2d71f5 | 691 | if (!ml_anch) { |
a3b680e6 | 692 | const I32 end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 693 | - (SvTAIL(check) != 0); |
a3b680e6 | 694 | const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
695 | |
696 | if (end_shift < eshift) | |
697 | end_shift = eshift; | |
698 | } | |
cad2e5aa | 699 | } |
2c2d71f5 | 700 | else { /* Can match at random position */ |
cad2e5aa JH |
701 | ml_anch = 0; |
702 | s = strpos; | |
1de06328 YO |
703 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
704 | end_shift = prog->check_end_shift; | |
705 | ||
706 | /* end shift should be non negative here */ | |
cad2e5aa JH |
707 | } |
708 | ||
bcdf7404 | 709 | #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 710 | if (end_shift < 0) |
1de06328 | 711 | Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", |
220fc49f | 712 | (IV)end_shift, RX_PRECOMP(prog)); |
2c2d71f5 JH |
713 | #endif |
714 | ||
2c2d71f5 JH |
715 | restart: |
716 | /* Find a possible match in the region s..strend by looking for | |
717 | the "check" substring in the region corrected by start/end_shift. */ | |
1de06328 YO |
718 | |
719 | { | |
720 | I32 srch_start_shift = start_shift; | |
721 | I32 srch_end_shift = end_shift; | |
c33e64f0 FC |
722 | U8* start_point; |
723 | U8* end_point; | |
1de06328 YO |
724 | if (srch_start_shift < 0 && strbeg - s > srch_start_shift) { |
725 | srch_end_shift -= ((strbeg - s) - srch_start_shift); | |
726 | srch_start_shift = strbeg - s; | |
727 | } | |
6bda09f9 | 728 | DEBUG_OPTIMISE_MORE_r({ |
1de06328 YO |
729 | PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n", |
730 | (IV)prog->check_offset_min, | |
731 | (IV)srch_start_shift, | |
732 | (IV)srch_end_shift, | |
733 | (IV)prog->check_end_shift); | |
734 | }); | |
735 | ||
bbe252da | 736 | if (prog->extflags & RXf_CANY_SEEN) { |
1de06328 YO |
737 | start_point= (U8*)(s + srch_start_shift); |
738 | end_point= (U8*)(strend - srch_end_shift); | |
739 | } else { | |
740 | start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); | |
741 | end_point= HOP3(strend, -srch_end_shift, strbeg); | |
742 | } | |
6bda09f9 | 743 | DEBUG_OPTIMISE_MORE_r({ |
56570a2c | 744 | PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", |
1de06328 | 745 | (int)(end_point - start_point), |
fc8cd66c | 746 | (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), |
1de06328 YO |
747 | start_point); |
748 | }); | |
749 | ||
750 | s = fbm_instr( start_point, end_point, | |
7fba1cd6 | 751 | check, multiline ? FBMrf_MULTILINE : 0); |
1de06328 | 752 | } |
cad2e5aa JH |
753 | /* Update the count-of-usability, remove useless subpatterns, |
754 | unshift s. */ | |
2c2d71f5 | 755 | |
ab3bbdeb | 756 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 757 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
758 | SvPVX_const(check), RE_SV_DUMPLEN(check), 30); |
759 | PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", | |
2c2d71f5 | 760 | (s ? "Found" : "Did not find"), |
f2ed9b32 | 761 | (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) |
ab3bbdeb YO |
762 | ? "anchored" : "floating"), |
763 | quoted, | |
764 | RE_SV_TAIL(check), | |
765 | (s ? " at offset " : "...\n") ); | |
766 | }); | |
2c2d71f5 JH |
767 | |
768 | if (!s) | |
769 | goto fail_finish; | |
2c2d71f5 | 770 | /* Finish the diagnostic message */ |
a3621e74 | 771 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 | 772 | |
1de06328 YO |
773 | /* XXX dmq: first branch is for positive lookbehind... |
774 | Our check string is offset from the beginning of the pattern. | |
775 | So we need to do any stclass tests offset forward from that | |
776 | point. I think. :-( | |
777 | */ | |
778 | ||
779 | ||
780 | ||
781 | check_at=s; | |
782 | ||
783 | ||
2c2d71f5 JH |
784 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. |
785 | Start with the other substr. | |
786 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 787 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
788 | *always* match. Probably should be marked during compile... |
789 | Probably it is right to do no SCREAM here... | |
790 | */ | |
791 | ||
f2ed9b32 | 792 | if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8) |
1de06328 YO |
793 | : (prog->float_substr && prog->anchored_substr)) |
794 | { | |
30944b6d | 795 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
796 | /* XXXX May be hopelessly wrong for UTF... */ |
797 | if (!other_last) | |
6eb5f6b9 | 798 | other_last = strpos; |
f2ed9b32 | 799 | if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
800 | do_other_anchored: |
801 | { | |
890ce7af AL |
802 | char * const last = HOP3c(s, -start_shift, strbeg); |
803 | char *last1, *last2; | |
be8e71aa | 804 | char * const saved_s = s; |
33b8afdf | 805 | SV* must; |
2c2d71f5 | 806 | |
2c2d71f5 JH |
807 | t = s - prog->check_offset_max; |
808 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
f2ed9b32 | 809 | && (!utf8_target |
0ce71af7 | 810 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 811 | && t > strpos))) |
6f207bd3 | 812 | NOOP; |
2c2d71f5 JH |
813 | else |
814 | t = strpos; | |
1aa99e6b | 815 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
816 | if (t < other_last) /* These positions already checked */ |
817 | t = other_last; | |
1aa99e6b | 818 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
819 | if (last < last1) |
820 | last1 = last; | |
1de06328 YO |
821 | /* XXXX It is not documented what units *_offsets are in. |
822 | We assume bytes, but this is clearly wrong. | |
823 | Meaning this code needs to be carefully reviewed for errors. | |
824 | dmq. | |
825 | */ | |
826 | ||
2c2d71f5 | 827 | /* On end-of-str: see comment below. */ |
f2ed9b32 | 828 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; |
33b8afdf JH |
829 | if (must == &PL_sv_undef) { |
830 | s = (char*)NULL; | |
1de06328 | 831 | DEBUG_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
832 | } |
833 | else | |
834 | s = fbm_instr( | |
835 | (unsigned char*)t, | |
836 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
837 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
838 | must, | |
7fba1cd6 | 839 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 840 | ); |
ab3bbdeb | 841 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 842 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
843 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
844 | PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", | |
2c2d71f5 | 845 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
846 | quoted, RE_SV_TAIL(must)); |
847 | }); | |
848 | ||
849 | ||
2c2d71f5 JH |
850 | if (!s) { |
851 | if (last1 >= last2) { | |
a3621e74 | 852 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
853 | ", giving up...\n")); |
854 | goto fail_finish; | |
855 | } | |
a3621e74 | 856 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 857 | ", trying floating at offset %ld...\n", |
be8e71aa | 858 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
859 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
860 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
861 | goto restart; |
862 | } | |
863 | else { | |
a3621e74 | 864 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 865 | (long)(s - i_strpos))); |
1aa99e6b IH |
866 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
867 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 868 | s = saved_s; |
2c2d71f5 JH |
869 | if (t == strpos) |
870 | goto try_at_start; | |
2c2d71f5 JH |
871 | goto try_at_offset; |
872 | } | |
30944b6d | 873 | } |
2c2d71f5 JH |
874 | } |
875 | else { /* Take into account the floating substring. */ | |
33b8afdf | 876 | char *last, *last1; |
be8e71aa | 877 | char * const saved_s = s; |
33b8afdf JH |
878 | SV* must; |
879 | ||
880 | t = HOP3c(s, -start_shift, strbeg); | |
881 | last1 = last = | |
882 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
883 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
884 | last = HOP3c(t, prog->float_max_offset, strend); | |
885 | s = HOP3c(t, prog->float_min_offset, strend); | |
886 | if (s < other_last) | |
887 | s = other_last; | |
2c2d71f5 | 888 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
f2ed9b32 | 889 | must = utf8_target ? prog->float_utf8 : prog->float_substr; |
33b8afdf JH |
890 | /* fbm_instr() takes into account exact value of end-of-str |
891 | if the check is SvTAIL(ed). Since false positives are OK, | |
892 | and end-of-str is not later than strend we are OK. */ | |
893 | if (must == &PL_sv_undef) { | |
894 | s = (char*)NULL; | |
1de06328 | 895 | DEBUG_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
896 | } |
897 | else | |
2c2d71f5 | 898 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
899 | (unsigned char*)last + SvCUR(must) |
900 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 901 | must, multiline ? FBMrf_MULTILINE : 0); |
ab3bbdeb | 902 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 903 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
904 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
905 | PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", | |
33b8afdf | 906 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
907 | quoted, RE_SV_TAIL(must)); |
908 | }); | |
33b8afdf JH |
909 | if (!s) { |
910 | if (last1 == last) { | |
a3621e74 | 911 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
912 | ", giving up...\n")); |
913 | goto fail_finish; | |
2c2d71f5 | 914 | } |
a3621e74 | 915 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 916 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 917 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
918 | other_last = last; |
919 | s = HOP3c(t, 1, strend); | |
920 | goto restart; | |
921 | } | |
922 | else { | |
a3621e74 | 923 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
924 | (long)(s - i_strpos))); |
925 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 926 | s = saved_s; |
33b8afdf JH |
927 | if (t == strpos) |
928 | goto try_at_start; | |
929 | goto try_at_offset; | |
930 | } | |
2c2d71f5 | 931 | } |
cad2e5aa | 932 | } |
2c2d71f5 | 933 | |
1de06328 | 934 | |
9ef43ace | 935 | t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); |
1de06328 | 936 | |
6bda09f9 | 937 | DEBUG_OPTIMISE_MORE_r( |
1de06328 YO |
938 | PerlIO_printf(Perl_debug_log, |
939 | "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", | |
940 | (IV)prog->check_offset_min, | |
941 | (IV)prog->check_offset_max, | |
942 | (IV)(s-strpos), | |
943 | (IV)(t-strpos), | |
944 | (IV)(t-s), | |
945 | (IV)(strend-strpos) | |
946 | ) | |
947 | ); | |
948 | ||
2c2d71f5 | 949 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
f2ed9b32 | 950 | && (!utf8_target |
9ef43ace | 951 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) |
1de06328 YO |
952 | && t > strpos))) |
953 | { | |
2c2d71f5 JH |
954 | /* Fixed substring is found far enough so that the match |
955 | cannot start at strpos. */ | |
956 | try_at_offset: | |
cad2e5aa | 957 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
958 | /* Eventually fbm_*() should handle this, but often |
959 | anchored_offset is not 0, so this check will not be wasted. */ | |
960 | /* XXXX In the code below we prefer to look for "^" even in | |
961 | presence of anchored substrings. And we search even | |
962 | beyond the found float position. These pessimizations | |
963 | are historical artefacts only. */ | |
964 | find_anchor: | |
2c2d71f5 | 965 | while (t < strend - prog->minlen) { |
cad2e5aa | 966 | if (*t == '\n') { |
4ee3650e | 967 | if (t < check_at - prog->check_offset_min) { |
f2ed9b32 | 968 | if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
969 | /* Since we moved from the found position, |
970 | we definitely contradict the found anchored | |
30944b6d IZ |
971 | substr. Due to the above check we do not |
972 | contradict "check" substr. | |
973 | Thus we can arrive here only if check substr | |
974 | is float. Redo checking for "other"=="fixed". | |
975 | */ | |
9041c2e3 | 976 | strpos = t + 1; |
a3621e74 | 977 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 978 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
979 | goto do_other_anchored; |
980 | } | |
4ee3650e GS |
981 | /* We don't contradict the found floating substring. */ |
982 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 983 | s = t + 1; |
a3621e74 | 984 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 985 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
986 | goto set_useful; |
987 | } | |
4ee3650e GS |
988 | /* Position contradicts check-string */ |
989 | /* XXXX probably better to look for check-string | |
990 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 991 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 992 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 993 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
994 | goto restart; |
995 | } | |
996 | t++; | |
997 | } | |
a3621e74 | 998 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 999 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 1000 | goto fail_finish; |
cad2e5aa | 1001 | } |
f5952150 | 1002 | else { |
a3621e74 | 1003 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 1004 | PL_colors[0], PL_colors[1])); |
f5952150 | 1005 | } |
cad2e5aa JH |
1006 | s = t; |
1007 | set_useful: | |
f2ed9b32 | 1008 | ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
1009 | } |
1010 | else { | |
f5952150 | 1011 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 1012 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
1013 | unless it was an MBOL and we are not after MBOL, |
1014 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
1015 | try_at_start: |
1016 | /* Even in this situation we may use MBOL flag if strpos is offset | |
1017 | wrt the start of the string. */ | |
05b4157f | 1018 | if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ |
a1933d95 | 1019 | && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d | 1020 | /* May be due to an implicit anchor of m{.*foo} */ |
bbe252da | 1021 | && !(prog->intflags & PREGf_IMPLICIT)) |
d506a20d | 1022 | { |
cad2e5aa JH |
1023 | t = strpos; |
1024 | goto find_anchor; | |
1025 | } | |
a3621e74 | 1026 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 1027 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
70685ca0 | 1028 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 1029 | ); |
2c2d71f5 | 1030 | success_at_start: |
bbe252da | 1031 | if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ |
f2ed9b32 | 1032 | && (utf8_target ? ( |
33b8afdf JH |
1033 | prog->check_utf8 /* Could be deleted already */ |
1034 | && --BmUSEFUL(prog->check_utf8) < 0 | |
1035 | && (prog->check_utf8 == prog->float_utf8) | |
1036 | ) : ( | |
1037 | prog->check_substr /* Could be deleted already */ | |
1038 | && --BmUSEFUL(prog->check_substr) < 0 | |
1039 | && (prog->check_substr == prog->float_substr) | |
1040 | ))) | |
66e933ab | 1041 | { |
cad2e5aa | 1042 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 1043 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
f2ed9b32 KW |
1044 | /* XXX Does the destruction order has to change with utf8_target? */ |
1045 | SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr); | |
1046 | SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
1047 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
1048 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
1049 | check = NULL; /* abort */ | |
cad2e5aa | 1050 | s = strpos; |
486ec47a | 1051 | /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag |
c9415951 YO |
1052 | see http://bugs.activestate.com/show_bug.cgi?id=87173 */ |
1053 | if (prog->intflags & PREGf_IMPLICIT) | |
1054 | prog->extflags &= ~RXf_ANCH_MBOL; | |
3cf5c195 IZ |
1055 | /* XXXX This is a remnant of the old implementation. It |
1056 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 1057 | other heuristics. */ |
bbe252da | 1058 | prog->extflags &= ~RXf_USE_INTUIT; |
c9415951 | 1059 | /* XXXX What other flags might need to be cleared in this branch? */ |
cad2e5aa JH |
1060 | } |
1061 | else | |
1062 | s = strpos; | |
1063 | } | |
1064 | ||
6eb5f6b9 JH |
1065 | /* Last resort... */ |
1066 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
1de06328 YO |
1067 | /* trie stclasses are too expensive to use here, we are better off to |
1068 | leave it to regmatch itself */ | |
f8fc2ecf | 1069 | if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { |
6eb5f6b9 JH |
1070 | /* minlen == 0 is possible if regstclass is \b or \B, |
1071 | and the fixed substr is ''$. | |
1072 | Since minlen is already taken into account, s+1 is before strend; | |
1073 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
1074 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
1075 | regstclass does not come from lookahead... */ | |
1076 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
af944926 | 1077 | This leaves EXACTF-ish only, which are dealt with in find_byclass(). */ |
f8fc2ecf YO |
1078 | const U8* const str = (U8*)STRING(progi->regstclass); |
1079 | const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT | |
1080 | ? CHR_DIST(str+STR_LEN(progi->regstclass), str) | |
66e933ab | 1081 | : 1); |
1de06328 YO |
1082 | char * endpos; |
1083 | if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) | |
1084 | endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); | |
1085 | else if (prog->float_substr || prog->float_utf8) | |
1086 | endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); | |
1087 | else | |
1088 | endpos= strend; | |
1089 | ||
d8080198 YO |
1090 | if (checked_upto < s) |
1091 | checked_upto = s; | |
1092 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n", | |
1093 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg))); | |
1094 | ||
6eb5f6b9 | 1095 | t = s; |
d8080198 YO |
1096 | s = find_byclass(prog, progi->regstclass, checked_upto, endpos, NULL); |
1097 | if (s) { | |
1098 | checked_upto = s; | |
1099 | } else { | |
6eb5f6b9 | 1100 | #ifdef DEBUGGING |
cbbf8932 | 1101 | const char *what = NULL; |
6eb5f6b9 JH |
1102 | #endif |
1103 | if (endpos == strend) { | |
a3621e74 | 1104 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
1105 | "Could not match STCLASS...\n") ); |
1106 | goto fail; | |
1107 | } | |
a3621e74 | 1108 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 1109 | "This position contradicts STCLASS...\n") ); |
bbe252da | 1110 | if ((prog->extflags & RXf_ANCH) && !ml_anch) |
653099ff | 1111 | goto fail; |
d8080198 YO |
1112 | checked_upto = HOPBACKc(endpos, start_shift); |
1113 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n", | |
1114 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg))); | |
6eb5f6b9 | 1115 | /* Contradict one of substrings */ |
33b8afdf | 1116 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 | 1117 | if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) { |
a3621e74 | 1118 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 1119 | hop_and_restart: |
1aa99e6b | 1120 | s = HOP3c(t, 1, strend); |
66e933ab GS |
1121 | if (s + start_shift + end_shift > strend) { |
1122 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 1123 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
1124 | "Could not match STCLASS...\n") ); |
1125 | goto fail; | |
1126 | } | |
5e39e1e5 HS |
1127 | if (!check) |
1128 | goto giveup; | |
a3621e74 | 1129 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1130 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1131 | what, (long)(s + start_shift - i_strpos)) ); |
1132 | goto restart; | |
1133 | } | |
66e933ab | 1134 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
1135 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
1136 | goto retry_floating_check; | |
1137 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 1138 | s = check_at; |
5e39e1e5 HS |
1139 | if (!check) |
1140 | goto giveup; | |
a3621e74 | 1141 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1142 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1143 | (long)(other_last - i_strpos)) ); |
1144 | goto do_other_anchored; | |
1145 | } | |
60e71179 GS |
1146 | /* Another way we could have checked stclass at the |
1147 | current position only: */ | |
1148 | if (ml_anch) { | |
1149 | s = t = t + 1; | |
5e39e1e5 HS |
1150 | if (!check) |
1151 | goto giveup; | |
a3621e74 | 1152 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1153 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 1154 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 1155 | goto try_at_offset; |
66e933ab | 1156 | } |
f2ed9b32 | 1157 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 1158 | goto fail; |
486ec47a | 1159 | /* Check is floating substring. */ |
6eb5f6b9 JH |
1160 | retry_floating_check: |
1161 | t = check_at - start_shift; | |
a3621e74 | 1162 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
1163 | goto hop_and_restart; |
1164 | } | |
b7953727 | 1165 | if (t != s) { |
a3621e74 | 1166 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 1167 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
1168 | (long)(t - i_strpos), (long)(s - i_strpos)) |
1169 | ); | |
1170 | } | |
1171 | else { | |
a3621e74 | 1172 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1173 | "Does not contradict STCLASS...\n"); |
1174 | ); | |
1175 | } | |
6eb5f6b9 | 1176 | } |
5e39e1e5 | 1177 | giveup: |
a3621e74 | 1178 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
1179 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
1180 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 1181 | return s; |
2c2d71f5 JH |
1182 | |
1183 | fail_finish: /* Substring not found */ | |
33b8afdf | 1184 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
f2ed9b32 | 1185 | BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ |
cad2e5aa | 1186 | fail: |
a3621e74 | 1187 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 1188 | PL_colors[4], PL_colors[5])); |
bd61b366 | 1189 | return NULL; |
cad2e5aa | 1190 | } |
9661b544 | 1191 | |
a0a388a1 YO |
1192 | #define DECL_TRIE_TYPE(scan) \ |
1193 | const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \ | |
fab2782b YO |
1194 | trie_type = ((scan->flags == EXACT) \ |
1195 | ? (utf8_target ? trie_utf8 : trie_plain) \ | |
1196 | : (utf8_target ? trie_utf8_fold : trie_latin_utf8_fold)) | |
1197 | ||
1198 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ | |
1199 | uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ | |
1200 | STRLEN skiplen; \ | |
1201 | switch (trie_type) { \ | |
1202 | case trie_utf8_fold: \ | |
1203 | if ( foldlen>0 ) { \ | |
1204 | uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1205 | foldlen -= len; \ | |
1206 | uscan += len; \ | |
1207 | len=0; \ | |
1208 | } else { \ | |
1209 | uvc = to_utf8_fold( (const U8*) uc, foldbuf, &foldlen ); \ | |
1210 | len = UTF8SKIP(uc); \ | |
1211 | skiplen = UNISKIP( uvc ); \ | |
1212 | foldlen -= skiplen; \ | |
1213 | uscan = foldbuf + skiplen; \ | |
1214 | } \ | |
1215 | break; \ | |
1216 | case trie_latin_utf8_fold: \ | |
1217 | if ( foldlen>0 ) { \ | |
1218 | uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1219 | foldlen -= len; \ | |
1220 | uscan += len; \ | |
1221 | len=0; \ | |
1222 | } else { \ | |
1223 | len = 1; \ | |
1224 | uvc = _to_fold_latin1( (U8) *uc, foldbuf, &foldlen, 1); \ | |
1225 | skiplen = UNISKIP( uvc ); \ | |
1226 | foldlen -= skiplen; \ | |
1227 | uscan = foldbuf + skiplen; \ | |
1228 | } \ | |
1229 | break; \ | |
1230 | case trie_utf8: \ | |
1231 | uvc = utf8n_to_uvuni( (const U8*) uc, UTF8_MAXLEN, &len, uniflags ); \ | |
1232 | break; \ | |
1233 | case trie_plain: \ | |
1234 | uvc = (UV)*uc; \ | |
1235 | len = 1; \ | |
1236 | } \ | |
1237 | if (uvc < 256) { \ | |
1238 | charid = trie->charmap[ uvc ]; \ | |
1239 | } \ | |
1240 | else { \ | |
1241 | charid = 0; \ | |
1242 | if (widecharmap) { \ | |
1243 | SV** const svpp = hv_fetch(widecharmap, \ | |
1244 | (char*)&uvc, sizeof(UV), 0); \ | |
1245 | if (svpp) \ | |
1246 | charid = (U16)SvIV(*svpp); \ | |
1247 | } \ | |
1248 | } \ | |
4cadc6a9 YO |
1249 | } STMT_END |
1250 | ||
4cadc6a9 YO |
1251 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ |
1252 | STMT_START { \ | |
1253 | while (s <= e) { \ | |
1254 | if ( (CoNd) \ | |
fac1af77 | 1255 | && (ln == 1 || folder(s, pat_string, ln)) \ |
9a5a5549 | 1256 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1257 | goto got_it; \ |
1258 | s++; \ | |
1259 | } \ | |
1260 | } STMT_END | |
1261 | ||
1262 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
1263 | STMT_START { \ | |
1264 | while (s + (uskip = UTF8SKIP(s)) <= strend) { \ | |
1265 | CoDe \ | |
1266 | s += uskip; \ | |
1267 | } \ | |
1268 | } STMT_END | |
1269 | ||
1270 | #define REXEC_FBC_SCAN(CoDe) \ | |
1271 | STMT_START { \ | |
1272 | while (s < strend) { \ | |
1273 | CoDe \ | |
1274 | s++; \ | |
1275 | } \ | |
1276 | } STMT_END | |
1277 | ||
1278 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1279 | REXEC_FBC_UTF8_SCAN( \ | |
1280 | if (CoNd) { \ | |
24b23f37 | 1281 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1282 | goto got_it; \ |
1283 | else \ | |
1284 | tmp = doevery; \ | |
1285 | } \ | |
1286 | else \ | |
1287 | tmp = 1; \ | |
1288 | ) | |
1289 | ||
1290 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1291 | REXEC_FBC_SCAN( \ | |
1292 | if (CoNd) { \ | |
24b23f37 | 1293 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1294 | goto got_it; \ |
1295 | else \ | |
1296 | tmp = doevery; \ | |
1297 | } \ | |
1298 | else \ | |
1299 | tmp = 1; \ | |
1300 | ) | |
1301 | ||
1302 | #define REXEC_FBC_TRYIT \ | |
24b23f37 | 1303 | if ((!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1304 | goto got_it |
1305 | ||
e1d1eefb | 1306 | #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \ |
f2ed9b32 | 1307 | if (utf8_target) { \ |
e1d1eefb YO |
1308 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1309 | } \ | |
1310 | else { \ | |
1311 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1312 | } |
e1d1eefb | 1313 | |
4cadc6a9 | 1314 | #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ |
f2ed9b32 | 1315 | if (utf8_target) { \ |
4cadc6a9 YO |
1316 | UtFpReLoAd; \ |
1317 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1318 | } \ | |
1319 | else { \ | |
1320 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1321 | } |
4cadc6a9 YO |
1322 | |
1323 | #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ | |
1324 | PL_reg_flags |= RF_tainted; \ | |
f2ed9b32 | 1325 | if (utf8_target) { \ |
4cadc6a9 YO |
1326 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1327 | } \ | |
1328 | else { \ | |
1329 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1330 | } |
4cadc6a9 | 1331 | |
786e8c11 YO |
1332 | #define DUMP_EXEC_POS(li,s,doutf8) \ |
1333 | dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8) | |
1334 | ||
cfaf538b KW |
1335 | |
1336 | #define UTF8_NOLOAD(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ | |
1337 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1338 | tmp = TEST_NON_UTF8(tmp); \ | |
1339 | REXEC_FBC_UTF8_SCAN( \ | |
1340 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1341 | tmp = !tmp; \ | |
1342 | IF_SUCCESS; \ | |
1343 | } \ | |
1344 | else { \ | |
1345 | IF_FAIL; \ | |
1346 | } \ | |
1347 | ); \ | |
1348 | ||
1349 | #define UTF8_LOAD(TeSt1_UtF8, TeSt2_UtF8, IF_SUCCESS, IF_FAIL) \ | |
1350 | if (s == PL_bostr) { \ | |
1351 | tmp = '\n'; \ | |
1352 | } \ | |
1353 | else { \ | |
1354 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); \ | |
1355 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); \ | |
1356 | } \ | |
1357 | tmp = TeSt1_UtF8; \ | |
1358 | LOAD_UTF8_CHARCLASS_ALNUM(); \ | |
1359 | REXEC_FBC_UTF8_SCAN( \ | |
1360 | if (tmp == ! (TeSt2_UtF8)) { \ | |
1361 | tmp = !tmp; \ | |
1362 | IF_SUCCESS; \ | |
1363 | } \ | |
1364 | else { \ | |
1365 | IF_FAIL; \ | |
1366 | } \ | |
1367 | ); \ | |
1368 | ||
63ac0dad KW |
1369 | /* The only difference between the BOUND and NBOUND cases is that |
1370 | * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in | |
1371 | * NBOUND. This is accomplished by passing it in either the if or else clause, | |
1372 | * with the other one being empty */ | |
1373 | #define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1374 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
cfaf538b KW |
1375 | |
1376 | #define FBC_BOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1377 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
63ac0dad KW |
1378 | |
1379 | #define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1380 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b KW |
1381 | |
1382 | #define FBC_NBOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1383 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b | 1384 | |
63ac0dad KW |
1385 | |
1386 | /* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to | |
1387 | * be passed in completely with the variable name being tested, which isn't | |
1388 | * such a clean interface, but this is easier to read than it was before. We | |
1389 | * are looking for the boundary (or non-boundary between a word and non-word | |
1390 | * character. The utf8 and non-utf8 cases have the same logic, but the details | |
1391 | * must be different. Find the "wordness" of the character just prior to this | |
1392 | * one, and compare it with the wordness of this one. If they differ, we have | |
1393 | * a boundary. At the beginning of the string, pretend that the previous | |
1394 | * character was a new-line */ | |
cfaf538b | 1395 | #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ |
63ac0dad | 1396 | if (utf8_target) { \ |
cfaf538b | 1397 | UTF8_CODE \ |
63ac0dad KW |
1398 | } \ |
1399 | else { /* Not utf8 */ \ | |
1400 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1401 | tmp = TEST_NON_UTF8(tmp); \ | |
1402 | REXEC_FBC_SCAN( \ | |
1403 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1404 | tmp = !tmp; \ | |
1405 | IF_SUCCESS; \ | |
1406 | } \ | |
1407 | else { \ | |
1408 | IF_FAIL; \ | |
1409 | } \ | |
1410 | ); \ | |
1411 | } \ | |
1412 | if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) \ | |
1413 | goto got_it; | |
1414 | ||
786e8c11 YO |
1415 | /* We know what class REx starts with. Try to find this position... */ |
1416 | /* if reginfo is NULL, its a dryrun */ | |
1417 | /* annoyingly all the vars in this routine have different names from their counterparts | |
1418 | in regmatch. /grrr */ | |
1419 | ||
3c3eec57 | 1420 | STATIC char * |
07be1b83 | 1421 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
24b23f37 | 1422 | const char *strend, regmatch_info *reginfo) |
a687059c | 1423 | { |
27da23d5 | 1424 | dVAR; |
bbe252da | 1425 | const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; |
fac1af77 KW |
1426 | char *pat_string; /* The pattern's exactish string */ |
1427 | char *pat_end; /* ptr to end char of pat_string */ | |
1428 | re_fold_t folder; /* Function for computing non-utf8 folds */ | |
1429 | const U8 *fold_array; /* array for folding ords < 256 */ | |
d8093b23 | 1430 | STRLEN ln; |
5dab1207 | 1431 | STRLEN lnc; |
078c425b | 1432 | register STRLEN uskip; |
fac1af77 KW |
1433 | U8 c1; |
1434 | U8 c2; | |
6eb5f6b9 JH |
1435 | char *e; |
1436 | register I32 tmp = 1; /* Scratch variable? */ | |
f2ed9b32 | 1437 | register const bool utf8_target = PL_reg_match_utf8; |
453bfd44 | 1438 | UV utf8_fold_flags = 0; |
f8fc2ecf | 1439 | RXi_GET_DECL(prog,progi); |
7918f24d NC |
1440 | |
1441 | PERL_ARGS_ASSERT_FIND_BYCLASS; | |
f8fc2ecf | 1442 | |
6eb5f6b9 JH |
1443 | /* We know what class it must start with. */ |
1444 | switch (OP(c)) { | |
f56b6394 | 1445 | case ANYOFV: |
6eb5f6b9 | 1446 | case ANYOF: |
f56b6394 | 1447 | if (utf8_target || OP(c) == ANYOFV) { |
b1e3e569 KW |
1448 | STRLEN inclasslen = strend - s; |
1449 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1450 | reginclass(prog, c, (U8*)s, &inclasslen, utf8_target)); | |
388cc4de HS |
1451 | } |
1452 | else { | |
6ef69d56 | 1453 | REXEC_FBC_CLASS_SCAN(REGINCLASS(prog, c, (U8*)s)); |
a0d0e21e | 1454 | } |
6eb5f6b9 | 1455 | break; |
f33976b4 | 1456 | case CANY: |
4cadc6a9 | 1457 | REXEC_FBC_SCAN( |
24b23f37 | 1458 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
f33976b4 DB |
1459 | goto got_it; |
1460 | else | |
1461 | tmp = doevery; | |
4cadc6a9 | 1462 | ); |
f33976b4 | 1463 | break; |
2f7f8cb1 KW |
1464 | |
1465 | case EXACTFA: | |
1466 | if (UTF_PATTERN || utf8_target) { | |
1467 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
1468 | goto do_exactf_utf8; | |
1469 | } | |
1470 | fold_array = PL_fold_latin1; /* Latin1 folds are not affected by */ | |
1471 | folder = foldEQ_latin1; /* /a, except the sharp s one which */ | |
1472 | goto do_exactf_non_utf8; /* isn't dealt with by these */ | |
1473 | ||
6eb5f6b9 | 1474 | case EXACTF: |
62bf7766 | 1475 | if (utf8_target) { |
77a6d856 KW |
1476 | |
1477 | /* regcomp.c already folded this if pattern is in UTF-8 */ | |
62bf7766 | 1478 | utf8_fold_flags = 0; |
fac1af77 KW |
1479 | goto do_exactf_utf8; |
1480 | } | |
1481 | fold_array = PL_fold; | |
1482 | folder = foldEQ; | |
1483 | goto do_exactf_non_utf8; | |
1484 | ||
1485 | case EXACTFL: | |
1486 | if (UTF_PATTERN || utf8_target) { | |
17580e7a | 1487 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
fac1af77 KW |
1488 | goto do_exactf_utf8; |
1489 | } | |
1490 | fold_array = PL_fold_locale; | |
1491 | folder = foldEQ_locale; | |
16d951b7 KW |
1492 | goto do_exactf_non_utf8; |
1493 | ||
3c760661 KW |
1494 | case EXACTFU_SS: |
1495 | if (UTF_PATTERN) { | |
1496 | utf8_fold_flags = FOLDEQ_S2_ALREADY_FOLDED; | |
1497 | } | |
1498 | goto do_exactf_utf8; | |
1499 | ||
fab2782b | 1500 | case EXACTFU_TRICKYFOLD: |
16d951b7 KW |
1501 | case EXACTFU: |
1502 | if (UTF_PATTERN || utf8_target) { | |
77a6d856 | 1503 | utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0; |
16d951b7 KW |
1504 | goto do_exactf_utf8; |
1505 | } | |
1506 | ||
1507 | /* Any 'ss' in the pattern should have been replaced by regcomp, | |
1508 | * so we don't have to worry here about this single special case | |
1509 | * in the Latin1 range */ | |
1510 | fold_array = PL_fold_latin1; | |
1511 | folder = foldEQ_latin1; | |
fac1af77 KW |
1512 | |
1513 | /* FALL THROUGH */ | |
1514 | ||
62bf7766 KW |
1515 | do_exactf_non_utf8: /* Neither pattern nor string are UTF8, and there |
1516 | are no glitches with fold-length differences | |
1517 | between the target string and pattern */ | |
fac1af77 KW |
1518 | |
1519 | /* The idea in the non-utf8 EXACTF* cases is to first find the | |
1520 | * first character of the EXACTF* node and then, if necessary, | |
1521 | * case-insensitively compare the full text of the node. c1 is the | |
1522 | * first character. c2 is its fold. This logic will not work for | |
1523 | * Unicode semantics and the german sharp ss, which hence should | |
1524 | * not be compiled into a node that gets here. */ | |
1525 | pat_string = STRING(c); | |
1526 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1527 | ||
8a90a8fe KW |
1528 | /* We know that we have to match at least 'ln' bytes (which is the |
1529 | * same as characters, since not utf8). If we have to match 3 | |
1530 | * characters, and there are only 2 availabe, we know without | |
1531 | * trying that it will fail; so don't start a match past the | |
1532 | * required minimum number from the far end */ | |
fac1af77 KW |
1533 | e = HOP3c(strend, -((I32)ln), s); |
1534 | ||
1535 | if (!reginfo && e < s) { | |
1536 | e = s; /* Due to minlen logic of intuit() */ | |
1537 | } | |
1538 | ||
1539 | c1 = *pat_string; | |
1540 | c2 = fold_array[c1]; | |
1541 | if (c1 == c2) { /* If char and fold are the same */ | |
1542 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); | |
1543 | } | |
1544 | else { | |
1545 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); | |
1546 | } | |
1547 | break; | |
1548 | ||
1549 | do_exactf_utf8: | |
e067297c KW |
1550 | { |
1551 | unsigned expansion; | |
1552 | ||
fac1af77 KW |
1553 | |
1554 | /* If one of the operands is in utf8, we can't use the simpler | |
1555 | * folding above, due to the fact that many different characters | |
1556 | * can have the same fold, or portion of a fold, or different- | |
1557 | * length fold */ | |
1558 | pat_string = STRING(c); | |
1559 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1560 | pat_end = pat_string + ln; | |
1561 | lnc = (UTF_PATTERN) /* length to match in characters */ | |
1562 | ? utf8_length((U8 *) pat_string, (U8 *) pat_end) | |
1563 | : ln; | |
1564 | ||
e067297c KW |
1565 | /* We have 'lnc' characters to match in the pattern, but because of |
1566 | * multi-character folding, each character in the target can match | |
1567 | * up to 3 characters (Unicode guarantees it will never exceed | |
1568 | * this) if it is utf8-encoded; and up to 2 if not (based on the | |
1569 | * fact that the Latin 1 folds are already determined, and the | |
1570 | * only multi-char fold in that range is the sharp-s folding to | |
1571 | * 'ss'. Thus, a pattern character can match as little as 1/3 of a | |
89378d8a KW |
1572 | * string character. Adjust lnc accordingly, rounding up, so that |
1573 | * if we need to match at least 4+1/3 chars, that really is 5. */ | |
e067297c | 1574 | expansion = (utf8_target) ? UTF8_MAX_FOLD_CHAR_EXPAND : 2; |
89378d8a | 1575 | lnc = (lnc + expansion - 1) / expansion; |
e067297c KW |
1576 | |
1577 | /* As in the non-UTF8 case, if we have to match 3 characters, and | |
1578 | * only 2 are left, it's guaranteed to fail, so don't start a | |
1579 | * match that would require us to go beyond the end of the string | |
1580 | */ | |
1581 | e = HOP3c(strend, -((I32)lnc), s); | |
fac1af77 KW |
1582 | |
1583 | if (!reginfo && e < s) { | |
1584 | e = s; /* Due to minlen logic of intuit() */ | |
1585 | } | |
1586 | ||
b33105db KW |
1587 | /* XXX Note that we could recalculate e to stop the loop earlier, |
1588 | * as the worst case expansion above will rarely be met, and as we | |
1589 | * go along we would usually find that e moves further to the left. | |
1590 | * This would happen only after we reached the point in the loop | |
1591 | * where if there were no expansion we should fail. Unclear if | |
1592 | * worth the expense */ | |
e067297c | 1593 | |
fac1af77 KW |
1594 | while (s <= e) { |
1595 | char *my_strend= (char *)strend; | |
1596 | if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target, | |
1597 | pat_string, NULL, ln, cBOOL(UTF_PATTERN), utf8_fold_flags) | |
1598 | && (!reginfo || regtry(reginfo, &s)) ) | |
1599 | { | |
1600 | goto got_it; | |
1601 | } | |
bbdd8bad | 1602 | s += (utf8_target) ? UTF8SKIP(s) : 1; |
fac1af77 KW |
1603 | } |
1604 | break; | |
e067297c | 1605 | } |
bbce6d69 | 1606 | case BOUNDL: |
3280af22 | 1607 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1608 | FBC_BOUND(isALNUM_LC, |
1609 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1610 | isALNUM_LC_utf8((U8*)s)); | |
a0ed51b3 | 1611 | break; |
bbce6d69 | 1612 | case NBOUNDL: |
3280af22 | 1613 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1614 | FBC_NBOUND(isALNUM_LC, |
1615 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1616 | isALNUM_LC_utf8((U8*)s)); | |
1617 | break; | |
1618 | case BOUND: | |
1619 | FBC_BOUND(isWORDCHAR, | |
1620 | isALNUM_uni(tmp), | |
1621 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1622 | break; | |
cfaf538b KW |
1623 | case BOUNDA: |
1624 | FBC_BOUND_NOLOAD(isWORDCHAR_A, | |
1625 | isWORDCHAR_A(tmp), | |
1626 | isWORDCHAR_A((U8*)s)); | |
1627 | break; | |
a0d0e21e | 1628 | case NBOUND: |
63ac0dad KW |
1629 | FBC_NBOUND(isWORDCHAR, |
1630 | isALNUM_uni(tmp), | |
1631 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1632 | break; | |
cfaf538b KW |
1633 | case NBOUNDA: |
1634 | FBC_NBOUND_NOLOAD(isWORDCHAR_A, | |
1635 | isWORDCHAR_A(tmp), | |
1636 | isWORDCHAR_A((U8*)s)); | |
1637 | break; | |
63ac0dad KW |
1638 | case BOUNDU: |
1639 | FBC_BOUND(isWORDCHAR_L1, | |
1640 | isALNUM_uni(tmp), | |
1641 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1642 | break; | |
1643 | case NBOUNDU: | |
1644 | FBC_NBOUND(isWORDCHAR_L1, | |
1645 | isALNUM_uni(tmp), | |
1646 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
a0ed51b3 | 1647 | break; |
bbce6d69 | 1648 | case ALNUML: |
4cadc6a9 YO |
1649 | REXEC_FBC_CSCAN_TAINT( |
1650 | isALNUM_LC_utf8((U8*)s), | |
1651 | isALNUM_LC(*s) | |
1652 | ); | |
6895a8aa | 1653 | break; |
980866de KW |
1654 | case ALNUMU: |
1655 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1656 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1657 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1658 | isWORDCHAR_L1((U8) *s) |
1659 | ); | |
6895a8aa | 1660 | break; |
980866de KW |
1661 | case ALNUM: |
1662 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1663 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1664 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1665 | isWORDCHAR((U8) *s) |
1666 | ); | |
6895a8aa | 1667 | break; |
cfaf538b | 1668 | case ALNUMA: |
8e9da4d4 KW |
1669 | /* Don't need to worry about utf8, as it can match only a single |
1670 | * byte invariant character */ | |
cfaf538b | 1671 | REXEC_FBC_CLASS_SCAN( isWORDCHAR_A(*s)); |
6895a8aa | 1672 | break; |
980866de KW |
1673 | case NALNUMU: |
1674 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 | 1675 | LOAD_UTF8_CHARCLASS_ALNUM(), |
359960d4 | 1676 | !swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), |
980866de KW |
1677 | ! isWORDCHAR_L1((U8) *s) |
1678 | ); | |
6895a8aa | 1679 | break; |
a0d0e21e | 1680 | case NALNUM: |
4cadc6a9 | 1681 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1682 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1683 | !swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target), | |
980866de | 1684 | ! isALNUM(*s) |
4cadc6a9 | 1685 | ); |
6895a8aa | 1686 | break; |
cfaf538b | 1687 | case NALNUMA: |
8e9da4d4 KW |
1688 | REXEC_FBC_CSCAN( |
1689 | !isWORDCHAR_A(*s), | |
1690 | !isWORDCHAR_A(*s) | |
1691 | ); | |
1692 | break; | |
bbce6d69 | 1693 | case NALNUML: |
4cadc6a9 YO |
1694 | REXEC_FBC_CSCAN_TAINT( |
1695 | !isALNUM_LC_utf8((U8*)s), | |
1696 | !isALNUM_LC(*s) | |
1697 | ); | |
6895a8aa | 1698 | break; |
980866de KW |
1699 | case SPACEU: |
1700 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1701 | LOAD_UTF8_CHARCLASS_SPACE(), |
1702 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de KW |
1703 | isSPACE_L1((U8) *s) |
1704 | ); | |
6895a8aa | 1705 | break; |
a0d0e21e | 1706 | case SPACE: |
4cadc6a9 | 1707 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1708 | LOAD_UTF8_CHARCLASS_SPACE(), |
1709 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de | 1710 | isSPACE((U8) *s) |
4cadc6a9 | 1711 | ); |
6895a8aa | 1712 | break; |
cfaf538b | 1713 | case SPACEA: |
8e9da4d4 KW |
1714 | /* Don't need to worry about utf8, as it can match only a single |
1715 | * byte invariant character */ | |
cfaf538b | 1716 | REXEC_FBC_CLASS_SCAN( isSPACE_A(*s)); |
6895a8aa | 1717 | break; |
bbce6d69 | 1718 | case SPACEL: |
4cadc6a9 | 1719 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1720 | isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1721 | isSPACE_LC(*s) |
1722 | ); | |
6895a8aa | 1723 | break; |
980866de KW |
1724 | case NSPACEU: |
1725 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1726 | LOAD_UTF8_CHARCLASS_SPACE(), |
1727 | !( *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de KW |
1728 | ! isSPACE_L1((U8) *s) |
1729 | ); | |
6895a8aa | 1730 | break; |
a0d0e21e | 1731 | case NSPACE: |
4cadc6a9 | 1732 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1733 | LOAD_UTF8_CHARCLASS_SPACE(), |
1734 | !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de | 1735 | ! isSPACE((U8) *s) |
4cadc6a9 | 1736 | ); |
6895a8aa | 1737 | break; |
cfaf538b | 1738 | case NSPACEA: |
8e9da4d4 KW |
1739 | REXEC_FBC_CSCAN( |
1740 | !isSPACE_A(*s), | |
1741 | !isSPACE_A(*s) | |
1742 | ); | |
1743 | break; | |
bbce6d69 | 1744 | case NSPACEL: |
4cadc6a9 | 1745 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1746 | !isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1747 | !isSPACE_LC(*s) |
1748 | ); | |
6895a8aa | 1749 | break; |
a0d0e21e | 1750 | case DIGIT: |
4cadc6a9 | 1751 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1752 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1753 | swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1754 | isDIGIT(*s) |
1755 | ); | |
6895a8aa | 1756 | break; |
cfaf538b | 1757 | case DIGITA: |
8e9da4d4 KW |
1758 | /* Don't need to worry about utf8, as it can match only a single |
1759 | * byte invariant character */ | |
cfaf538b | 1760 | REXEC_FBC_CLASS_SCAN( isDIGIT_A(*s)); |
6895a8aa | 1761 | break; |
b8c5462f | 1762 | case DIGITL: |
4cadc6a9 YO |
1763 | REXEC_FBC_CSCAN_TAINT( |
1764 | isDIGIT_LC_utf8((U8*)s), | |
1765 | isDIGIT_LC(*s) | |
1766 | ); | |
6895a8aa | 1767 | break; |
a0d0e21e | 1768 | case NDIGIT: |
4cadc6a9 | 1769 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1770 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1771 | !swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1772 | !isDIGIT(*s) |
1773 | ); | |
6895a8aa | 1774 | break; |
cfaf538b | 1775 | case NDIGITA: |
8e9da4d4 KW |
1776 | REXEC_FBC_CSCAN( |
1777 | !isDIGIT_A(*s), | |
1778 | !isDIGIT_A(*s) | |
1779 | ); | |
1780 | break; | |
b8c5462f | 1781 | case NDIGITL: |
4cadc6a9 YO |
1782 | REXEC_FBC_CSCAN_TAINT( |
1783 | !isDIGIT_LC_utf8((U8*)s), | |
1784 | !isDIGIT_LC(*s) | |
1785 | ); | |
6895a8aa | 1786 | break; |
e1d1eefb YO |
1787 | case LNBREAK: |
1788 | REXEC_FBC_CSCAN( | |
1789 | is_LNBREAK_utf8(s), | |
1790 | is_LNBREAK_latin1(s) | |
1791 | ); | |
6895a8aa | 1792 | break; |
e1d1eefb YO |
1793 | case VERTWS: |
1794 | REXEC_FBC_CSCAN( | |
1795 | is_VERTWS_utf8(s), | |
1796 | is_VERTWS_latin1(s) | |
1797 | ); | |
6895a8aa | 1798 | break; |
e1d1eefb YO |
1799 | case NVERTWS: |
1800 | REXEC_FBC_CSCAN( | |
1801 | !is_VERTWS_utf8(s), | |
1802 | !is_VERTWS_latin1(s) | |
1803 | ); | |
6895a8aa | 1804 | break; |
e1d1eefb YO |
1805 | case HORIZWS: |
1806 | REXEC_FBC_CSCAN( | |
1807 | is_HORIZWS_utf8(s), | |
1808 | is_HORIZWS_latin1(s) | |
1809 | ); | |
6895a8aa | 1810 | break; |
e1d1eefb YO |
1811 | case NHORIZWS: |
1812 | REXEC_FBC_CSCAN( | |
1813 | !is_HORIZWS_utf8(s), | |
1814 | !is_HORIZWS_latin1(s) | |
1815 | ); | |
6895a8aa | 1816 | break; |
1de06328 YO |
1817 | case AHOCORASICKC: |
1818 | case AHOCORASICK: | |
07be1b83 | 1819 | { |
a0a388a1 | 1820 | DECL_TRIE_TYPE(c); |
07be1b83 YO |
1821 | /* what trie are we using right now */ |
1822 | reg_ac_data *aho | |
f8fc2ecf | 1823 | = (reg_ac_data*)progi->data->data[ ARG( c ) ]; |
3251b653 NC |
1824 | reg_trie_data *trie |
1825 | = (reg_trie_data*)progi->data->data[ aho->trie ]; | |
85fbaab2 | 1826 | HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); |
07be1b83 YO |
1827 | |
1828 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1829 | #ifdef DEBUGGING |
07be1b83 | 1830 | const char *real_start = s; |
6148ee25 | 1831 | #endif |
07be1b83 | 1832 | STRLEN maxlen = trie->maxlen; |
be8e71aa YO |
1833 | SV *sv_points; |
1834 | U8 **points; /* map of where we were in the input string | |
786e8c11 | 1835 | when reading a given char. For ASCII this |
be8e71aa | 1836 | is unnecessary overhead as the relationship |
38a44b82 NC |
1837 | is always 1:1, but for Unicode, especially |
1838 | case folded Unicode this is not true. */ | |
f9e705e8 | 1839 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
786e8c11 YO |
1840 | U8 *bitmap=NULL; |
1841 | ||
07be1b83 YO |
1842 | |
1843 | GET_RE_DEBUG_FLAGS_DECL; | |
1844 | ||
be8e71aa YO |
1845 | /* We can't just allocate points here. We need to wrap it in |
1846 | * an SV so it gets freed properly if there is a croak while | |
1847 | * running the match */ | |
1848 | ENTER; | |
1849 | SAVETMPS; | |
1850 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1851 | SvCUR_set(sv_points, | |
1852 | maxlen * sizeof(U8 *)); | |
1853 | SvPOK_on(sv_points); | |
1854 | sv_2mortal(sv_points); | |
1855 | points=(U8**)SvPV_nolen(sv_points ); | |
1de06328 YO |
1856 | if ( trie_type != trie_utf8_fold |
1857 | && (trie->bitmap || OP(c)==AHOCORASICKC) ) | |
1858 | { | |
786e8c11 YO |
1859 | if (trie->bitmap) |
1860 | bitmap=(U8*)trie->bitmap; | |
1861 | else | |
1862 | bitmap=(U8*)ANYOF_BITMAP(c); | |
07be1b83 | 1863 | } |
786e8c11 YO |
1864 | /* this is the Aho-Corasick algorithm modified a touch |
1865 | to include special handling for long "unknown char" | |
1866 | sequences. The basic idea being that we use AC as long | |
1867 | as we are dealing with a possible matching char, when | |
1868 | we encounter an unknown char (and we have not encountered | |
1869 | an accepting state) we scan forward until we find a legal | |
1870 | starting char. | |
1871 | AC matching is basically that of trie matching, except | |
1872 | that when we encounter a failing transition, we fall back | |
1873 | to the current states "fail state", and try the current char | |
1874 | again, a process we repeat until we reach the root state, | |
1875 | state 1, or a legal transition. If we fail on the root state | |
1876 | then we can either terminate if we have reached an accepting | |
1877 | state previously, or restart the entire process from the beginning | |
1878 | if we have not. | |
1879 | ||
1880 | */ | |
07be1b83 YO |
1881 | while (s <= last_start) { |
1882 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1883 | U8 *uc = (U8*)s; | |
1884 | U16 charid = 0; | |
1885 | U32 base = 1; | |
1886 | U32 state = 1; | |
1887 | UV uvc = 0; | |
1888 | STRLEN len = 0; | |
1889 | STRLEN foldlen = 0; | |
1890 | U8 *uscan = (U8*)NULL; | |
1891 | U8 *leftmost = NULL; | |
786e8c11 YO |
1892 | #ifdef DEBUGGING |
1893 | U32 accepted_word= 0; | |
1894 | #endif | |
07be1b83 YO |
1895 | U32 pointpos = 0; |
1896 | ||
1897 | while ( state && uc <= (U8*)strend ) { | |
1898 | int failed=0; | |
786e8c11 YO |
1899 | U32 word = aho->states[ state ].wordnum; |
1900 | ||
1de06328 YO |
1901 | if( state==1 ) { |
1902 | if ( bitmap ) { | |
1903 | DEBUG_TRIE_EXECUTE_r( | |
1904 | if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1905 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1906 | (char *)uc, utf8_target ); |
1de06328 YO |
1907 | PerlIO_printf( Perl_debug_log, |
1908 | " Scanning for legal start char...\n"); | |
1909 | } | |
d085b490 YO |
1910 | ); |
1911 | if (utf8_target) { | |
1912 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1913 | uc += UTF8SKIP(uc); | |
1914 | } | |
1915 | } else { | |
1916 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1917 | uc++; | |
1918 | } | |
1919 | } | |
1de06328 | 1920 | s= (char *)uc; |
786e8c11 | 1921 | } |
786e8c11 YO |
1922 | if (uc >(U8*)last_start) break; |
1923 | } | |
1924 | ||
1925 | if ( word ) { | |
2e64971a | 1926 | U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ]; |
786e8c11 YO |
1927 | if (!leftmost || lpos < leftmost) { |
1928 | DEBUG_r(accepted_word=word); | |
07be1b83 | 1929 | leftmost= lpos; |
786e8c11 | 1930 | } |
07be1b83 | 1931 | if (base==0) break; |
786e8c11 | 1932 | |
07be1b83 YO |
1933 | } |
1934 | points[pointpos++ % maxlen]= uc; | |
55eed653 NC |
1935 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
1936 | uscan, len, uvc, charid, foldlen, | |
1937 | foldbuf, uniflags); | |
786e8c11 YO |
1938 | DEBUG_TRIE_EXECUTE_r({ |
1939 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1940 | s, utf8_target ); |
07be1b83 | 1941 | PerlIO_printf(Perl_debug_log, |
786e8c11 YO |
1942 | " Charid:%3u CP:%4"UVxf" ", |
1943 | charid, uvc); | |
1944 | }); | |
07be1b83 YO |
1945 | |
1946 | do { | |
6148ee25 | 1947 | #ifdef DEBUGGING |
786e8c11 | 1948 | word = aho->states[ state ].wordnum; |
6148ee25 | 1949 | #endif |
07be1b83 YO |
1950 | base = aho->states[ state ].trans.base; |
1951 | ||
786e8c11 YO |
1952 | DEBUG_TRIE_EXECUTE_r({ |
1953 | if (failed) | |
1954 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1955 | s, utf8_target ); |
07be1b83 | 1956 | PerlIO_printf( Perl_debug_log, |
786e8c11 YO |
1957 | "%sState: %4"UVxf", word=%"UVxf, |
1958 | failed ? " Fail transition to " : "", | |
1959 | (UV)state, (UV)word); | |
1960 | }); | |
07be1b83 YO |
1961 | if ( base ) { |
1962 | U32 tmp; | |
6dd2be57 | 1963 | I32 offset; |
07be1b83 | 1964 | if (charid && |
6dd2be57 DM |
1965 | ( ((offset = base + charid |
1966 | - 1 - trie->uniquecharcount)) >= 0) | |
1967 | && ((U32)offset < trie->lasttrans) | |
1968 | && trie->trans[offset].check == state | |
1969 | && (tmp=trie->trans[offset].next)) | |
07be1b83 | 1970 | { |
786e8c11 YO |
1971 | DEBUG_TRIE_EXECUTE_r( |
1972 | PerlIO_printf( Perl_debug_log," - legal\n")); | |
07be1b83 YO |
1973 | state = tmp; |
1974 | break; | |
1975 | } | |
1976 | else { | |
786e8c11 YO |
1977 | DEBUG_TRIE_EXECUTE_r( |
1978 | PerlIO_printf( Perl_debug_log," - fail\n")); | |
1979 | failed = 1; | |
1980 | state = aho->fail[state]; | |
07be1b83 YO |
1981 | } |
1982 | } | |
1983 | else { | |
1984 | /* we must be accepting here */ | |
786e8c11 YO |
1985 | DEBUG_TRIE_EXECUTE_r( |
1986 | PerlIO_printf( Perl_debug_log," - accepting\n")); | |
1987 | failed = 1; | |
07be1b83 YO |
1988 | break; |
1989 | } | |
1990 | } while(state); | |
786e8c11 | 1991 | uc += len; |
07be1b83 YO |
1992 | if (failed) { |
1993 | if (leftmost) | |
1994 | break; | |
786e8c11 | 1995 | if (!state) state = 1; |
07be1b83 YO |
1996 | } |
1997 | } | |
1998 | if ( aho->states[ state ].wordnum ) { | |
2e64971a | 1999 | U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ]; |
786e8c11 YO |
2000 | if (!leftmost || lpos < leftmost) { |
2001 | DEBUG_r(accepted_word=aho->states[ state ].wordnum); | |
07be1b83 | 2002 | leftmost = lpos; |
786e8c11 | 2003 | } |
07be1b83 | 2004 | } |
07be1b83 YO |
2005 | if (leftmost) { |
2006 | s = (char*)leftmost; | |
786e8c11 YO |
2007 | DEBUG_TRIE_EXECUTE_r({ |
2008 | PerlIO_printf( | |
70685ca0 JH |
2009 | Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", |
2010 | (UV)accepted_word, (IV)(s - real_start) | |
786e8c11 YO |
2011 | ); |
2012 | }); | |
24b23f37 | 2013 | if (!reginfo || regtry(reginfo, &s)) { |
be8e71aa YO |
2014 | FREETMPS; |
2015 | LEAVE; | |
07be1b83 | 2016 | goto got_it; |
be8e71aa | 2017 | } |
07be1b83 | 2018 | s = HOPc(s,1); |
786e8c11 YO |
2019 | DEBUG_TRIE_EXECUTE_r({ |
2020 | PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); | |
2021 | }); | |
07be1b83 | 2022 | } else { |
786e8c11 YO |
2023 | DEBUG_TRIE_EXECUTE_r( |
2024 | PerlIO_printf( Perl_debug_log,"No match.\n")); | |
07be1b83 YO |
2025 | break; |
2026 | } | |
2027 | } | |
be8e71aa YO |
2028 | FREETMPS; |
2029 | LEAVE; | |
07be1b83 YO |
2030 | } |
2031 | break; | |
b3c9acc1 | 2032 | default: |
3c3eec57 GS |
2033 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); |
2034 | break; | |
d6a28714 | 2035 | } |
6eb5f6b9 JH |
2036 | return 0; |
2037 | got_it: | |
2038 | return s; | |
2039 | } | |
2040 | ||
fae667d5 | 2041 | |
6eb5f6b9 JH |
2042 | /* |
2043 | - regexec_flags - match a regexp against a string | |
2044 | */ | |
2045 | I32 | |
288b8c02 | 2046 | Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend, |
6eb5f6b9 JH |
2047 | char *strbeg, I32 minend, SV *sv, void *data, U32 flags) |
2048 | /* strend: pointer to null at end of string */ | |
2049 | /* strbeg: real beginning of string */ | |
2050 | /* minend: end of match must be >=minend after stringarg. */ | |
58e23c8d YO |
2051 | /* data: May be used for some additional optimizations. |
2052 | Currently its only used, with a U32 cast, for transmitting | |
2053 | the ganch offset when doing a /g match. This will change */ | |
6eb5f6b9 JH |
2054 | /* nosave: For optimizations. */ |
2055 | { | |
97aff369 | 2056 | dVAR; |
288b8c02 | 2057 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
24b23f37 | 2058 | /*register*/ char *s; |
6eb5f6b9 | 2059 | register regnode *c; |
24b23f37 | 2060 | /*register*/ char *startpos = stringarg; |
6eb5f6b9 JH |
2061 | I32 minlen; /* must match at least this many chars */ |
2062 | I32 dontbother = 0; /* how many characters not to try at end */ | |
6eb5f6b9 JH |
2063 | I32 end_shift = 0; /* Same for the end. */ /* CC */ |
2064 | I32 scream_pos = -1; /* Internal iterator of scream. */ | |
ccac19ea | 2065 | char *scream_olds = NULL; |
f2ed9b32 | 2066 | const bool utf8_target = cBOOL(DO_UTF8(sv)); |
2757e526 | 2067 | I32 multiline; |
f8fc2ecf | 2068 | RXi_GET_DECL(prog,progi); |
3b0527fe | 2069 | regmatch_info reginfo; /* create some info to pass to regtry etc */ |
e9105d30 | 2070 | regexp_paren_pair *swap = NULL; |
a3621e74 YO |
2071 | GET_RE_DEBUG_FLAGS_DECL; |
2072 | ||
7918f24d | 2073 | PERL_ARGS_ASSERT_REGEXEC_FLAGS; |
9d4ba2ae | 2074 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
2075 | |
2076 | /* Be paranoid... */ | |
2077 | if (prog == NULL || startpos == NULL) { | |
2078 | Perl_croak(aTHX_ "NULL regexp parameter"); | |
2079 | return 0; | |
2080 | } | |
2081 | ||
bbe252da | 2082 | multiline = prog->extflags & RXf_PMf_MULTILINE; |
288b8c02 | 2083 | reginfo.prog = rx; /* Yes, sorry that this is confusing. */ |
2757e526 | 2084 | |
f2ed9b32 | 2085 | RX_MATCH_UTF8_set(rx, utf8_target); |
1de06328 | 2086 | DEBUG_EXECUTE_r( |
f2ed9b32 | 2087 | debug_start_match(rx, utf8_target, startpos, strend, |
1de06328 YO |
2088 | "Matching"); |
2089 | ); | |
bac06658 | 2090 | |
6eb5f6b9 | 2091 | minlen = prog->minlen; |
1de06328 YO |
2092 | |
2093 | if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { | |
a3621e74 | 2094 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
2095 | "String too short [regexec_flags]...\n")); |
2096 | goto phooey; | |
1aa99e6b | 2097 | } |
6eb5f6b9 | 2098 | |
1de06328 | 2099 | |
6eb5f6b9 | 2100 | /* Check validity of program. */ |
f8fc2ecf | 2101 | if (UCHARAT(progi->program) != REG_MAGIC) { |
6eb5f6b9 JH |
2102 | Perl_croak(aTHX_ "corrupted regexp program"); |
2103 | } | |
2104 | ||
2105 | PL_reg_flags = 0; | |
ed301438 | 2106 | PL_reg_state.re_state_eval_setup_done = FALSE; |
6eb5f6b9 JH |
2107 | PL_reg_maxiter = 0; |
2108 | ||
3c8556c3 | 2109 | if (RX_UTF8(rx)) |
6eb5f6b9 JH |
2110 | PL_reg_flags |= RF_utf8; |
2111 | ||
2112 | /* Mark beginning of line for ^ and lookbehind. */ | |
3b0527fe | 2113 | reginfo.bol = startpos; /* XXX not used ??? */ |
6eb5f6b9 | 2114 | PL_bostr = strbeg; |
3b0527fe | 2115 | reginfo.sv = sv; |
6eb5f6b9 JH |
2116 | |
2117 | /* Mark end of line for $ (and such) */ | |
2118 | PL_regeol = strend; | |
2119 | ||
2120 | /* see how far we have to get to not match where we matched before */ | |
3b0527fe | 2121 | reginfo.till = startpos+minend; |
6eb5f6b9 | 2122 | |
6eb5f6b9 JH |
2123 | /* If there is a "must appear" string, look for it. */ |
2124 | s = startpos; | |
2125 | ||
bbe252da | 2126 | if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ |
6eb5f6b9 | 2127 | MAGIC *mg; |
2c296965 | 2128 | if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */ |
58e23c8d | 2129 | reginfo.ganch = startpos + prog->gofs; |
2c296965 | 2130 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2131 | "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs)); |
2c296965 | 2132 | } else if (sv && SvTYPE(sv) >= SVt_PVMG |
6eb5f6b9 | 2133 | && SvMAGIC(sv) |
14befaf4 DM |
2134 | && (mg = mg_find(sv, PERL_MAGIC_regex_global)) |
2135 | && mg->mg_len >= 0) { | |
3b0527fe | 2136 | reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ |
2c296965 | 2137 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2138 | "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len)); |
2c296965 | 2139 | |
bbe252da | 2140 | if (prog->extflags & RXf_ANCH_GPOS) { |
3b0527fe | 2141 | if (s > reginfo.ganch) |
6eb5f6b9 | 2142 | goto phooey; |
58e23c8d | 2143 | s = reginfo.ganch - prog->gofs; |
2c296965 | 2144 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2145 | "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs)); |
c584a96e YO |
2146 | if (s < strbeg) |
2147 | goto phooey; | |
6eb5f6b9 JH |
2148 | } |
2149 | } | |
58e23c8d | 2150 | else if (data) { |
70685ca0 | 2151 | reginfo.ganch = strbeg + PTR2UV(data); |
2c296965 YO |
2152 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2153 | "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data))); | |
2154 | ||
2155 | } else { /* pos() not defined */ | |
3b0527fe | 2156 | reginfo.ganch = strbeg; |
2c296965 YO |
2157 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2158 | "GPOS: reginfo.ganch = strbeg\n")); | |
2159 | } | |
6eb5f6b9 | 2160 | } |
288b8c02 | 2161 | if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { |
e9105d30 GG |
2162 | /* We have to be careful. If the previous successful match |
2163 | was from this regex we don't want a subsequent partially | |
2164 | successful match to clobber the old results. | |
2165 | So when we detect this possibility we add a swap buffer | |
2166 | to the re, and switch the buffer each match. If we fail | |
2167 | we switch it back, otherwise we leave it swapped. | |
2168 | */ | |
2169 | swap = prog->offs; | |
2170 | /* do we need a save destructor here for eval dies? */ | |
2171 | Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); | |
495f47a5 DM |
2172 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
2173 | "rex=0x%"UVxf" saving offs: orig=0x%"UVxf" new=0x%"UVxf"\n", | |
2174 | PTR2UV(prog), | |
2175 | PTR2UV(swap), | |
2176 | PTR2UV(prog->offs) | |
2177 | )); | |
c74340f9 | 2178 | } |
a0714e2c | 2179 | if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { |
6eb5f6b9 JH |
2180 | re_scream_pos_data d; |
2181 | ||
2182 | d.scream_olds = &scream_olds; | |
2183 | d.scream_pos = &scream_pos; | |
288b8c02 | 2184 | s = re_intuit_start(rx, sv, s, strend, flags, &d); |
3fa9c3d7 | 2185 | if (!s) { |
a3621e74 | 2186 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n")); |
6eb5f6b9 | 2187 | goto phooey; /* not present */ |
3fa9c3d7 | 2188 | } |
6eb5f6b9 JH |
2189 | } |
2190 | ||
1de06328 | 2191 | |
6eb5f6b9 JH |
2192 | |
2193 | /* Simplest case: anchored match need be tried only once. */ | |
2194 | /* [unless only anchor is BOL and multiline is set] */ | |
bbe252da | 2195 | if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { |
24b23f37 | 2196 | if (s == startpos && regtry(®info, &startpos)) |
6eb5f6b9 | 2197 | goto got_it; |
bbe252da YO |
2198 | else if (multiline || (prog->intflags & PREGf_IMPLICIT) |
2199 | || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ | |
6eb5f6b9 JH |
2200 | { |
2201 | char *end; | |
2202 | ||
2203 | if (minlen) | |
2204 | dontbother = minlen - 1; | |
1aa99e6b | 2205 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 2206 | /* for multiline we only have to try after newlines */ |
33b8afdf | 2207 | if (prog->check_substr || prog->check_utf8) { |
92f3d482 YO |
2208 | /* because of the goto we can not easily reuse the macros for bifurcating the |
2209 | unicode/non-unicode match modes here like we do elsewhere - demerphq */ | |
2210 | if (utf8_target) { | |
2211 | if (s == startpos) | |
2212 | goto after_try_utf8; | |
2213 | while (1) { | |
2214 | if (regtry(®info, &s)) { | |
2215 | goto got_it; | |
2216 | } | |
2217 | after_try_utf8: | |
2218 | if (s > end) { | |
2219 | goto phooey; | |
2220 | } | |
2221 | if (prog->extflags & RXf_USE_INTUIT) { | |
2222 | s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL); | |
2223 | if (!s) { | |
2224 | goto phooey; | |
2225 | } | |
2226 | } | |
2227 | else { | |
2228 | s += UTF8SKIP(s); | |
2229 | } | |
2230 | } | |
2231 | } /* end search for check string in unicode */ | |
2232 | else { | |
2233 | if (s == startpos) { | |
2234 | goto after_try_latin; | |
2235 | } | |
2236 | while (1) { | |
2237 | if (regtry(®info, &s)) { | |
2238 | goto got_it; | |
2239 | } | |
2240 | after_try_latin: | |
2241 | if (s > end) { | |
2242 | goto phooey; | |
2243 | } | |
2244 | if (prog->extflags & RXf_USE_INTUIT) { | |
2245 | s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL); | |
2246 | if (!s) { | |
2247 | goto phooey; | |
2248 | } | |
2249 | } | |
2250 | else { | |
2251 | s++; | |
2252 | } | |
2253 | } | |
2254 | } /* end search for check string in latin*/ | |
2255 | } /* end search for check string */ | |
2256 | else { /* search for newline */ | |
2257 | if (s > startpos) { | |
2258 | /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/ | |
6eb5f6b9 | 2259 | s--; |
92f3d482 | 2260 | } |
21eede78 YO |
2261 | /* We can use a more efficient search as newlines are the same in unicode as they are in latin */ |
2262 | while (s <= end) { /* note it could be possible to match at the end of the string */ | |
6eb5f6b9 | 2263 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ |
24b23f37 | 2264 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2265 | goto got_it; |
2266 | } | |
92f3d482 YO |
2267 | } |
2268 | } /* end search for newline */ | |
2269 | } /* end anchored/multiline check string search */ | |
6eb5f6b9 | 2270 | goto phooey; |
bbe252da | 2271 | } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) |
f9f4320a | 2272 | { |
486ec47a | 2273 | /* the warning about reginfo.ganch being used without initialization |
bbe252da | 2274 | is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN |
f9f4320a | 2275 | and we only enter this block when the same bit is set. */ |
58e23c8d | 2276 | char *tmp_s = reginfo.ganch - prog->gofs; |
c584a96e YO |
2277 | |
2278 | if (tmp_s >= strbeg && regtry(®info, &tmp_s)) | |
6eb5f6b9 JH |
2279 | goto got_it; |
2280 | goto phooey; | |
2281 | } | |
2282 | ||
2283 | /* Messy cases: unanchored match. */ | |
bbe252da | 2284 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { |
6eb5f6b9 | 2285 | /* we have /x+whatever/ */ |
f2ed9b32 | 2286 | /* it must be a one character string (XXXX Except UTF_PATTERN?) */ |
33b8afdf | 2287 | char ch; |
bf93d4cc GS |
2288 | #ifdef DEBUGGING |
2289 | int did_match = 0; | |
2290 | #endif | |
f2ed9b32 KW |
2291 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2292 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2293 | ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0]; | |
bf93d4cc | 2294 | |
f2ed9b32 | 2295 | if (utf8_target) { |
4cadc6a9 | 2296 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2297 | if (*s == ch) { |
a3621e74 | 2298 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2299 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2300 | s += UTF8SKIP(s); |
2301 | while (s < strend && *s == ch) | |
2302 | s += UTF8SKIP(s); | |
2303 | } | |
4cadc6a9 | 2304 | ); |
6eb5f6b9 JH |
2305 | } |
2306 | else { | |
4cadc6a9 | 2307 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2308 | if (*s == ch) { |
a3621e74 | 2309 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2310 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2311 | s++; |
2312 | while (s < strend && *s == ch) | |
2313 | s++; | |
2314 | } | |
4cadc6a9 | 2315 | ); |
6eb5f6b9 | 2316 | } |
a3621e74 | 2317 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 2318 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
2319 | "Did not find anchored character...\n") |
2320 | ); | |
6eb5f6b9 | 2321 | } |
a0714e2c SS |
2322 | else if (prog->anchored_substr != NULL |
2323 | || prog->anchored_utf8 != NULL | |
2324 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
2325 | && prog->float_max_offset < strend - s)) { |
2326 | SV *must; | |
2327 | I32 back_max; | |
2328 | I32 back_min; | |
2329 | char *last; | |
6eb5f6b9 | 2330 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
2331 | #ifdef DEBUGGING |
2332 | int did_match = 0; | |
2333 | #endif | |
33b8afdf | 2334 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 KW |
2335 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2336 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2337 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; | |
33b8afdf JH |
2338 | back_max = back_min = prog->anchored_offset; |
2339 | } else { | |
f2ed9b32 KW |
2340 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2341 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2342 | must = utf8_target ? prog->float_utf8 : prog->float_substr; | |
33b8afdf JH |
2343 | back_max = prog->float_max_offset; |
2344 | back_min = prog->float_min_offset; | |
2345 | } | |
1de06328 YO |
2346 | |
2347 | ||
33b8afdf JH |
2348 | if (must == &PL_sv_undef) |
2349 | /* could not downgrade utf8 check substring, so must fail */ | |
2350 | goto phooey; | |
2351 | ||
1de06328 YO |
2352 | if (back_min<0) { |
2353 | last = strend; | |
2354 | } else { | |
2355 | last = HOP3c(strend, /* Cannot start after this */ | |
2356 | -(I32)(CHR_SVLEN(must) | |
2357 | - (SvTAIL(must) != 0) + back_min), strbeg); | |
2358 | } | |
6eb5f6b9 JH |
2359 | if (s > PL_bostr) |
2360 | last1 = HOPc(s, -1); | |
2361 | else | |
2362 | last1 = s - 1; /* bogus */ | |
2363 | ||
a0288114 | 2364 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 JH |
2365 | check_substr==must. */ |
2366 | scream_pos = -1; | |
2367 | dontbother = end_shift; | |
2368 | strend = HOPc(strend, -dontbother); | |
2369 | while ( (s <= last) && | |
c33e64f0 | 2370 | (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), |
9041c2e3 | 2371 | (unsigned char*)strend, must, |
c33e64f0 | 2372 | multiline ? FBMrf_MULTILINE : 0)) ) { |
a3621e74 | 2373 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
2374 | if (HOPc(s, -back_max) > last1) { |
2375 | last1 = HOPc(s, -back_min); | |
2376 | s = HOPc(s, -back_max); | |
2377 | } | |
2378 | else { | |
52657f30 | 2379 | char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; |
6eb5f6b9 JH |
2380 | |
2381 | last1 = HOPc(s, -back_min); | |
52657f30 | 2382 | s = t; |
6eb5f6b9 | 2383 | } |
f2ed9b32 | 2384 | if (utf8_target) { |
6eb5f6b9 | 2385 | while (s <= last1) { |
24b23f37 | 2386 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2387 | goto got_it; |
2388 | s += UTF8SKIP(s); | |
2389 | } | |
2390 | } | |
2391 | else { | |
2392 | while (s <= last1) { | |
24b23f37 | 2393 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2394 | goto got_it; |
2395 | s++; | |
2396 | } | |
2397 | } | |
2398 | } | |
ab3bbdeb | 2399 | DEBUG_EXECUTE_r(if (!did_match) { |
f2ed9b32 | 2400 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
2401 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
2402 | PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", | |
33b8afdf | 2403 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc | 2404 | ? "anchored" : "floating"), |
ab3bbdeb YO |
2405 | quoted, RE_SV_TAIL(must)); |
2406 | }); | |
6eb5f6b9 JH |
2407 | goto phooey; |
2408 | } | |
f8fc2ecf | 2409 | else if ( (c = progi->regstclass) ) { |
f14c76ed | 2410 | if (minlen) { |
f8fc2ecf | 2411 | const OPCODE op = OP(progi->regstclass); |
66e933ab | 2412 | /* don't bother with what can't match */ |
786e8c11 | 2413 | if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) |
f14c76ed RGS |
2414 | strend = HOPc(strend, -(minlen - 1)); |
2415 | } | |
a3621e74 | 2416 | DEBUG_EXECUTE_r({ |
be8e71aa | 2417 | SV * const prop = sv_newmortal(); |
32fc9b6a | 2418 | regprop(prog, prop, c); |
0df25f3d | 2419 | { |
f2ed9b32 | 2420 | RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb | 2421 | s,strend-s,60); |
0df25f3d | 2422 | PerlIO_printf(Perl_debug_log, |
1c8f8eb1 | 2423 | "Matching stclass %.*s against %s (%d bytes)\n", |
e4f74956 | 2424 | (int)SvCUR(prop), SvPVX_const(prop), |
ab3bbdeb | 2425 | quoted, (int)(strend - s)); |
0df25f3d | 2426 | } |
ffc61ed2 | 2427 | }); |
3b0527fe | 2428 | if (find_byclass(prog, c, s, strend, ®info)) |
6eb5f6b9 | 2429 | goto got_it; |
07be1b83 | 2430 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
2431 | } |
2432 | else { | |
2433 | dontbother = 0; | |
a0714e2c | 2434 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 2435 | /* Trim the end. */ |
6af40bd7 | 2436 | char *last= NULL; |
33b8afdf | 2437 | SV* float_real; |
c33e64f0 FC |
2438 | STRLEN len; |
2439 | const char *little; | |
33b8afdf | 2440 | |
f2ed9b32 KW |
2441 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2442 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2443 | float_real = utf8_target ? prog->float_utf8 : prog->float_substr; | |
d6a28714 | 2444 | |
c33e64f0 FC |
2445 | little = SvPV_const(float_real, len); |
2446 | if (SvTAIL(float_real)) { | |
1a13b075 YO |
2447 | /* This means that float_real contains an artificial \n on the end |
2448 | * due to the presence of something like this: /foo$/ | |
2449 | * where we can match both "foo" and "foo\n" at the end of the string. | |
2450 | * So we have to compare the end of the string first against the float_real | |
2451 | * without the \n and then against the full float_real with the string. | |
2452 | * We have to watch out for cases where the string might be smaller | |
2453 | * than the float_real or the float_real without the \n. | |
2454 | */ | |
2455 | char *checkpos= strend - len; | |
2456 | DEBUG_OPTIMISE_r( | |
2457 | PerlIO_printf(Perl_debug_log, | |
2458 | "%sChecking for float_real.%s\n", | |
2459 | PL_colors[4], PL_colors[5])); | |
2460 | if (checkpos + 1 < strbeg) { | |
2461 | /* can't match, even if we remove the trailing \n string is too short to match */ | |
2462 | DEBUG_EXECUTE_r( | |
2463 | PerlIO_printf(Perl_debug_log, | |
2464 | "%sString shorter than required trailing substring, cannot match.%s\n", | |
2465 | PL_colors[4], PL_colors[5])); | |
2466 | goto phooey; | |
2467 | } else if (memEQ(checkpos + 1, little, len - 1)) { | |
2468 | /* can match, the end of the string matches without the "\n" */ | |
2469 | last = checkpos + 1; | |
2470 | } else if (checkpos < strbeg) { | |
2471 | /* cant match, string is too short when the "\n" is included */ | |
2472 | DEBUG_EXECUTE_r( | |
2473 | PerlIO_printf(Perl_debug_log, | |
2474 | "%sString does not contain required trailing substring, cannot match.%s\n", | |
2475 | PL_colors[4], PL_colors[5])); | |
2476 | goto phooey; | |
2477 | } else if (!multiline) { | |
2478 | /* non multiline match, so compare with the "\n" at the end of the string */ | |
2479 | if (memEQ(checkpos, little, len)) { | |
2480 | last= checkpos; | |
2481 | } else { | |
2482 | DEBUG_EXECUTE_r( | |
2483 | PerlIO_printf(Perl_debug_log, | |
2484 | "%sString does not contain required trailing substring, cannot match.%s\n", | |
2485 | PL_colors[4], PL_colors[5])); | |
2486 | goto phooey; | |
2487 | } | |
2488 | } else { | |
2489 | /* multiline match, so we have to search for a place where the full string is located */ | |
d6a28714 | 2490 | goto find_last; |
1a13b075 | 2491 | } |
c33e64f0 | 2492 | } else { |
d6a28714 | 2493 | find_last: |
9041c2e3 | 2494 | if (len) |
d6a28714 | 2495 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 2496 | else |
a0288114 | 2497 | last = strend; /* matching "$" */ |
b8c5462f | 2498 | } |
6af40bd7 YO |
2499 | if (!last) { |
2500 | /* at one point this block contained a comment which was probably | |
2501 | * incorrect, which said that this was a "should not happen" case. | |
2502 | * Even if it was true when it was written I am pretty sure it is | |
2503 | * not anymore, so I have removed the comment and replaced it with | |
2504 | * this one. Yves */ | |
6bda09f9 YO |
2505 | DEBUG_EXECUTE_r( |
2506 | PerlIO_printf(Perl_debug_log, | |
6af40bd7 YO |
2507 | "String does not contain required substring, cannot match.\n" |
2508 | )); | |
2509 | goto phooey; | |
bf93d4cc | 2510 | } |
d6a28714 JH |
2511 | dontbother = strend - last + prog->float_min_offset; |
2512 | } | |
2513 | if (minlen && (dontbother < minlen)) | |
2514 | dontbother = minlen - 1; | |
2515 | strend -= dontbother; /* this one's always in bytes! */ | |
2516 | /* We don't know much -- general case. */ | |
f2ed9b32 | 2517 | if (utf8_target) { |
d6a28714 | 2518 | for (;;) { |
24b23f37 | 2519 | if (regtry(®info, &s)) |
d6a28714 JH |
2520 | goto got_it; |
2521 | if (s >= strend) | |
2522 | break; | |
b8c5462f | 2523 | s += UTF8SKIP(s); |
d6a28714 JH |
2524 | }; |
2525 | } | |
2526 | else { | |
2527 | do { | |
24b23f37 | 2528 | if (regtry(®info, &s)) |
d6a28714 JH |
2529 | goto got_it; |
2530 | } while (s++ < strend); | |
2531 | } | |
2532 | } | |
2533 | ||
2534 | /* Failure. */ | |
2535 | goto phooey; | |
2536 | ||
2537 | got_it: | |
495f47a5 DM |
2538 | DEBUG_BUFFERS_r( |
2539 | if (swap) | |
2540 | PerlIO_printf(Perl_debug_log, | |
2541 | "rex=0x%"UVxf" freeing offs: 0x%"UVxf"\n", | |
2542 | PTR2UV(prog), | |
2543 | PTR2UV(swap) | |
2544 | ); | |
2545 | ); | |
e9105d30 | 2546 | Safefree(swap); |
288b8c02 | 2547 | RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted); |
d6a28714 | 2548 | |
ed301438 | 2549 | if (PL_reg_state.re_state_eval_setup_done) |
4f639d21 | 2550 | restore_pos(aTHX_ prog); |
5daac39c NC |
2551 | if (RXp_PAREN_NAMES(prog)) |
2552 | (void)hv_iterinit(RXp_PAREN_NAMES(prog)); | |
d6a28714 JH |
2553 | |
2554 | /* make sure $`, $&, $', and $digit will work later */ | |
2555 | if ( !(flags & REXEC_NOT_FIRST) ) { | |
288b8c02 | 2556 | RX_MATCH_COPY_FREE(rx); |
d6a28714 | 2557 | if (flags & REXEC_COPY_STR) { |
be8e71aa | 2558 | const I32 i = PL_regeol - startpos + (stringarg - strbeg); |
f8c7b90f | 2559 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2560 | if ((SvIsCOW(sv) |
2561 | || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { | |
2562 | if (DEBUG_C_TEST) { | |
2563 | PerlIO_printf(Perl_debug_log, | |
2564 | "Copy on write: regexp capture, type %d\n", | |
2565 | (int) SvTYPE(sv)); | |
2566 | } | |
2567 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); | |
d5263905 | 2568 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
ed252734 NC |
2569 | assert (SvPOKp(prog->saved_copy)); |
2570 | } else | |
2571 | #endif | |
2572 | { | |
288b8c02 | 2573 | RX_MATCH_COPIED_on(rx); |
ed252734 NC |
2574 | s = savepvn(strbeg, i); |
2575 | prog->subbeg = s; | |
2576 | } | |
d6a28714 | 2577 | prog->sublen = i; |
d6a28714 JH |
2578 | } |
2579 | else { | |
2580 | prog->subbeg = strbeg; | |
2581 | prog->sublen = PL_regeol - strbeg; /* strend may have been modified */ | |
2582 | } | |
2583 | } | |
9041c2e3 | 2584 | |
d6a28714 JH |
2585 | return 1; |
2586 | ||
2587 | phooey: | |
a3621e74 | 2588 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 2589 | PL_colors[4], PL_colors[5])); |
ed301438 | 2590 | if (PL_reg_state.re_state_eval_setup_done) |
4f639d21 | 2591 | restore_pos(aTHX_ prog); |
e9105d30 | 2592 | if (swap) { |
c74340f9 | 2593 | /* we failed :-( roll it back */ |
495f47a5 DM |
2594 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
2595 | "rex=0x%"UVxf" rolling back offs: freeing=0x%"UVxf" restoring=0x%"UVxf"\n", | |
2596 | PTR2UV(prog), | |
2597 | PTR2UV(prog->offs), | |
2598 | PTR2UV(swap) | |
2599 | )); | |
e9105d30 GG |
2600 | Safefree(prog->offs); |
2601 | prog->offs = swap; | |
2602 | } | |
2603 | ||
d6a28714 JH |
2604 | return 0; |
2605 | } | |
2606 | ||
6bda09f9 | 2607 | |
ec43f78b DM |
2608 | /* Set which rex is pointed to by PL_reg_state, handling ref counting. |
2609 | * Do inc before dec, in case old and new rex are the same */ | |
2610 | #define SET_reg_curpm(Re2) \ | |
2611 | if (PL_reg_state.re_state_eval_setup_done) { \ | |
2612 | (void)ReREFCNT_inc(Re2); \ | |
2613 | ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); \ | |
2614 | PM_SETRE((PL_reg_curpm), (Re2)); \ | |
2615 | } | |
2616 | ||
2617 | ||
d6a28714 JH |
2618 | /* |
2619 | - regtry - try match at specific point | |
2620 | */ | |
2621 | STATIC I32 /* 0 failure, 1 success */ | |
24b23f37 | 2622 | S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) |
d6a28714 | 2623 | { |
97aff369 | 2624 | dVAR; |
d6a28714 | 2625 | CHECKPOINT lastcp; |
288b8c02 NC |
2626 | REGEXP *const rx = reginfo->prog; |
2627 | regexp *const prog = (struct regexp *)SvANY(rx); | |
f8fc2ecf | 2628 | RXi_GET_DECL(prog,progi); |
a3621e74 | 2629 | GET_RE_DEBUG_FLAGS_DECL; |
7918f24d NC |
2630 | |
2631 | PERL_ARGS_ASSERT_REGTRY; | |
2632 | ||
24b23f37 | 2633 | reginfo->cutpoint=NULL; |
d6a28714 | 2634 | |
ed301438 DM |
2635 | if ((prog->extflags & RXf_EVAL_SEEN) |
2636 | && !PL_reg_state.re_state_eval_setup_done) | |
2637 | { | |
d6a28714 JH |
2638 | MAGIC *mg; |
2639 | ||
ed301438 | 2640 | PL_reg_state.re_state_eval_setup_done = TRUE; |
3b0527fe | 2641 | if (reginfo->sv) { |
d6a28714 | 2642 | /* Make $_ available to executed code. */ |
3b0527fe | 2643 | if (reginfo->sv != DEFSV) { |
59f00321 | 2644 | SAVE_DEFSV; |
414bf5ae | 2645 | DEFSV_set(reginfo->sv); |
b8c5462f | 2646 | } |
d6a28714 | 2647 | |
3b0527fe DM |
2648 | if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) |
2649 | && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { | |
d6a28714 | 2650 | /* prepare for quick setting of pos */ |
d300d9fa | 2651 | #ifdef PERL_OLD_COPY_ON_WRITE |
51a9ea20 NC |
2652 | if (SvIsCOW(reginfo->sv)) |
2653 | sv_force_normal_flags(reginfo->sv, 0); | |
d300d9fa | 2654 | #endif |
3dab1dad | 2655 | mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, |
d300d9fa | 2656 | &PL_vtbl_mglob, NULL, 0); |
d6a28714 | 2657 | mg->mg_len = -1; |
b8c5462f | 2658 | } |
d6a28714 JH |
2659 | PL_reg_magic = mg; |
2660 | PL_reg_oldpos = mg->mg_len; | |
4f639d21 | 2661 | SAVEDESTRUCTOR_X(restore_pos, prog); |
d6a28714 | 2662 | } |
09687e5a | 2663 | if (!PL_reg_curpm) { |
a02a5408 | 2664 | Newxz(PL_reg_curpm, 1, PMOP); |
09687e5a AB |
2665 | #ifdef USE_ITHREADS |
2666 | { | |
14a49a24 | 2667 | SV* const repointer = &PL_sv_undef; |
92313705 NC |
2668 | /* this regexp is also owned by the new PL_reg_curpm, which |
2669 | will try to free it. */ | |
d2ece331 | 2670 | av_push(PL_regex_padav, repointer); |
09687e5a AB |
2671 | PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); |
2672 | PL_regex_pad = AvARRAY(PL_regex_padav); | |
2673 | } | |
2674 | #endif | |
2675 | } | |
ec43f78b | 2676 | SET_reg_curpm(rx); |
d6a28714 JH |
2677 | PL_reg_oldcurpm = PL_curpm; |
2678 | PL_curpm = PL_reg_curpm; | |
07bc277f | 2679 | if (RXp_MATCH_COPIED(prog)) { |
d6a28714 JH |
2680 | /* Here is a serious problem: we cannot rewrite subbeg, |
2681 | since it may be needed if this match fails. Thus | |
2682 | $` inside (?{}) could fail... */ | |
2683 | PL_reg_oldsaved = prog->subbeg; | |
2684 | PL_reg_oldsavedlen = prog->sublen; | |
f8c7b90f | 2685 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2686 | PL_nrs = prog->saved_copy; |
2687 | #endif | |
07bc277f | 2688 | RXp_MATCH_COPIED_off(prog); |
d6a28714 JH |
2689 | } |
2690 | else | |
bd61b366 | 2691 | PL_reg_oldsaved = NULL; |
d6a28714 JH |
2692 | prog->subbeg = PL_bostr; |
2693 | prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ | |
2694 | } | |
97ca13b7 DM |
2695 | #ifdef DEBUGGING |
2696 | PL_reg_starttry = *startpos; | |
2697 | #endif | |
f0ab9afb | 2698 | prog->offs[0].start = *startpos - PL_bostr; |
24b23f37 | 2699 | PL_reginput = *startpos; |
d6a28714 | 2700 | prog->lastparen = 0; |
03994de8 | 2701 | prog->lastcloseparen = 0; |
d6a28714 | 2702 | PL_regsize = 0; |
d6a28714 JH |
2703 | |
2704 | /* XXXX What this code is doing here?!!! There should be no need | |
b93070ed | 2705 | to do this again and again, prog->lastparen should take care of |
3dd2943c | 2706 | this! --ilya*/ |
dafc8851 JH |
2707 | |
2708 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2709 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
b93070ed | 2710 | * prog->lastparen), is not needed at all by the test suite |
225593e1 DM |
2711 | * (op/regexp, op/pat, op/split), but that code is needed otherwise |
2712 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ | |
2713 | * Meanwhile, this code *is* needed for the | |
daf18116 JH |
2714 | * above-mentioned test suite tests to succeed. The common theme |
2715 | * on those tests seems to be returning null fields from matches. | |
225593e1 | 2716 | * --jhi updated by dapm */ |
dafc8851 | 2717 | #if 1 |
d6a28714 | 2718 | if (prog->nparens) { |
b93070ed | 2719 | regexp_paren_pair *pp = prog->offs; |
097eb12c | 2720 | register I32 i; |
b93070ed | 2721 | for (i = prog->nparens; i > (I32)prog->lastparen; i--) { |
f0ab9afb NC |
2722 | ++pp; |
2723 | pp->start = -1; | |
2724 | pp->end = -1; | |
d6a28714 JH |
2725 | } |
2726 | } | |
dafc8851 | 2727 | #endif |
02db2b7b | 2728 | REGCP_SET(lastcp); |
f8fc2ecf | 2729 | if (regmatch(reginfo, progi->program + 1)) { |
b93070ed | 2730 | prog->offs[0].end = PL_reginput - PL_bostr; |
d6a28714 JH |
2731 | return 1; |
2732 | } | |
24b23f37 YO |
2733 | if (reginfo->cutpoint) |
2734 | *startpos= reginfo->cutpoint; | |
02db2b7b | 2735 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
2736 | return 0; |
2737 | } | |
2738 | ||
02db2b7b | 2739 | |
8ba1375e MJD |
2740 | #define sayYES goto yes |
2741 | #define sayNO goto no | |
262b90c4 | 2742 | #define sayNO_SILENT goto no_silent |
8ba1375e | 2743 | |
f9f4320a YO |
2744 | /* we dont use STMT_START/END here because it leads to |
2745 | "unreachable code" warnings, which are bogus, but distracting. */ | |
2746 | #define CACHEsayNO \ | |
c476f425 DM |
2747 | if (ST.cache_mask) \ |
2748 | PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \ | |
f9f4320a | 2749 | sayNO |
3298f257 | 2750 | |
a3621e74 | 2751 | /* this is used to determine how far from the left messages like |
265c4333 YO |
2752 | 'failed...' are printed. It should be set such that messages |
2753 | are inline with the regop output that created them. | |
a3621e74 | 2754 | */ |
265c4333 | 2755 | #define REPORT_CODE_OFF 32 |
a3621e74 YO |
2756 | |
2757 | ||
40a82448 DM |
2758 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
2759 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
9e137952 | 2760 | |
86545054 DM |
2761 | #define SLAB_FIRST(s) (&(s)->states[0]) |
2762 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
2763 | ||
5d9a96ca DM |
2764 | /* grab a new slab and return the first slot in it */ |
2765 | ||
2766 | STATIC regmatch_state * | |
2767 | S_push_slab(pTHX) | |
2768 | { | |
a35a87e7 | 2769 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2770 | dMY_CXT; |
2771 | #endif | |
5d9a96ca DM |
2772 | regmatch_slab *s = PL_regmatch_slab->next; |
2773 | if (!s) { | |
2774 | Newx(s, 1, regmatch_slab); | |
2775 | s->prev = PL_regmatch_slab; | |
2776 | s->next = NULL; | |
2777 | PL_regmatch_slab->next = s; | |
2778 | } | |
2779 | PL_regmatch_slab = s; | |
86545054 | 2780 | return SLAB_FIRST(s); |
5d9a96ca | 2781 | } |
5b47454d | 2782 | |
95b24440 | 2783 | |
40a82448 DM |
2784 | /* push a new state then goto it */ |
2785 | ||
2786 | #define PUSH_STATE_GOTO(state, node) \ | |
2787 | scan = node; \ | |
2788 | st->resume_state = state; \ | |
2789 | goto push_state; | |
2790 | ||
2791 | /* push a new state with success backtracking, then goto it */ | |
2792 | ||
2793 | #define PUSH_YES_STATE_GOTO(state, node) \ | |
2794 | scan = node; \ | |
2795 | st->resume_state = state; \ | |
2796 | goto push_yes_state; | |
2797 | ||
aa283a38 | 2798 | |
aa283a38 | 2799 | |
d6a28714 | 2800 | /* |
95b24440 | 2801 | |
bf1f174e DM |
2802 | regmatch() - main matching routine |
2803 | ||
2804 | This is basically one big switch statement in a loop. We execute an op, | |
2805 | set 'next' to point the next op, and continue. If we come to a point which | |
2806 | we may need to backtrack to on failure such as (A|B|C), we push a | |
2807 | backtrack state onto the backtrack stack. On failure, we pop the top | |
2808 | state, and re-enter the loop at the state indicated. If there are no more | |
2809 | states to pop, we return failure. | |
2810 | ||
2811 | Sometimes we also need to backtrack on success; for example /A+/, where | |
2812 | after successfully matching one A, we need to go back and try to | |
2813 | match another one; similarly for lookahead assertions: if the assertion | |
2814 | completes successfully, we backtrack to the state just before the assertion | |
2815 | and then carry on. In these cases, the pushed state is marked as | |
2816 | 'backtrack on success too'. This marking is in fact done by a chain of | |
2817 | pointers, each pointing to the previous 'yes' state. On success, we pop to | |
2818 | the nearest yes state, discarding any intermediate failure-only states. | |
2819 | Sometimes a yes state is pushed just to force some cleanup code to be | |
2820 | called at the end of a successful match or submatch; e.g. (??{$re}) uses | |
2821 | it to free the inner regex. | |
2822 | ||
2823 | Note that failure backtracking rewinds the cursor position, while | |
2824 | success backtracking leaves it alone. | |
2825 | ||
2826 | A pattern is complete when the END op is executed, while a subpattern | |
2827 | such as (?=foo) is complete when the SUCCESS op is executed. Both of these | |
2828 | ops trigger the "pop to last yes state if any, otherwise return true" | |
2829 | behaviour. | |
2830 | ||
2831 | A common convention in this function is to use A and B to refer to the two | |
2832 | subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is | |
2833 | the subpattern to be matched possibly multiple times, while B is the entire | |
2834 | rest of the pattern. Variable and state names reflect this convention. | |
2835 | ||
2836 | The states in the main switch are the union of ops and failure/success of | |
2837 | substates associated with with that op. For example, IFMATCH is the op | |
2838 | that does lookahead assertions /(?=A)B/ and so the IFMATCH state means | |
2839 | 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just | |
2840 | successfully matched A and IFMATCH_A_fail is a state saying that we have | |
2841 | just failed to match A. Resume states always come in pairs. The backtrack | |
2842 | state we push is marked as 'IFMATCH_A', but when that is popped, we resume | |
2843 | at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking | |
2844 | on success or failure. | |
2845 | ||
2846 | The struct that holds a backtracking state is actually a big union, with | |
2847 | one variant for each major type of op. The variable st points to the | |
2848 | top-most backtrack struct. To make the code clearer, within each | |
2849 | block of code we #define ST to alias the relevant union. | |
2850 | ||
2851 | Here's a concrete example of a (vastly oversimplified) IFMATCH | |
2852 | implementation: | |
2853 | ||
2854 | switch (state) { | |
2855 | .... | |
2856 | ||
2857 | #define ST st->u.ifmatch | |
2858 | ||
2859 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B | |
2860 | ST.foo = ...; // some state we wish to save | |
95b24440 | 2861 | ... |
bf1f174e DM |
2862 | // push a yes backtrack state with a resume value of |
2863 | // IFMATCH_A/IFMATCH_A_fail, then continue execution at the | |
2864 | // first node of A: | |
2865 | PUSH_YES_STATE_GOTO(IFMATCH_A, A); | |
2866 | // NOTREACHED | |
2867 | ||
2868 | case IFMATCH_A: // we have successfully executed A; now continue with B | |
2869 | next = B; | |
2870 | bar = ST.foo; // do something with the preserved value | |
2871 | break; | |
2872 | ||
2873 | case IFMATCH_A_fail: // A failed, so the assertion failed | |
2874 | ...; // do some housekeeping, then ... | |
2875 | sayNO; // propagate the failure | |
2876 | ||
2877 | #undef ST | |
95b24440 | 2878 | |
bf1f174e DM |
2879 | ... |
2880 | } | |
95b24440 | 2881 | |
bf1f174e DM |
2882 | For any old-timers reading this who are familiar with the old recursive |
2883 | approach, the code above is equivalent to: | |
95b24440 | 2884 | |
bf1f174e DM |
2885 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B |
2886 | { | |
2887 | int foo = ... | |
95b24440 | 2888 | ... |
bf1f174e DM |
2889 | if (regmatch(A)) { |
2890 | next = B; | |
2891 | bar = foo; | |
2892 | break; | |
95b24440 | 2893 | } |
bf1f174e DM |
2894 | ...; // do some housekeeping, then ... |
2895 | sayNO; // propagate the failure | |
95b24440 | 2896 | } |
bf1f174e DM |
2897 | |
2898 | The topmost backtrack state, pointed to by st, is usually free. If you | |
2899 | want to claim it, populate any ST.foo fields in it with values you wish to | |
2900 | save, then do one of | |
2901 | ||
2902 | PUSH_STATE_GOTO(resume_state, node); | |
2903 | PUSH_YES_STATE_GOTO(resume_state, node); | |
2904 | ||
2905 | which sets that backtrack state's resume value to 'resume_state', pushes a | |
2906 | new free entry to the top of the backtrack stack, then goes to 'node'. | |
2907 | On backtracking, the free slot is popped, and the saved state becomes the | |
2908 | new free state. An ST.foo field in this new top state can be temporarily | |
2909 | accessed to retrieve values, but once the main loop is re-entered, it | |
2910 | becomes available for reuse. | |
2911 | ||
2912 | Note that the depth of the backtrack stack constantly increases during the | |
2913 | left-to-right execution of the pattern, rather than going up and down with | |
2914 | the pattern nesting. For example the stack is at its maximum at Z at the | |
2915 | end of the pattern, rather than at X in the following: | |
2916 | ||
2917 | /(((X)+)+)+....(Y)+....Z/ | |
2918 | ||
2919 | The only exceptions to this are lookahead/behind assertions and the cut, | |
2920 | (?>A), which pop all the backtrack states associated with A before | |
2921 | continuing. | |
2922 | ||
486ec47a | 2923 | Backtrack state structs are allocated in slabs of about 4K in size. |
bf1f174e DM |
2924 | PL_regmatch_state and st always point to the currently active state, |
2925 | and PL_regmatch_slab points to the slab currently containing | |
2926 | PL_regmatch_state. The first time regmatch() is called, the first slab is | |
2927 | allocated, and is never freed until interpreter destruction. When the slab | |
2928 | is full, a new one is allocated and chained to the end. At exit from | |
2929 | regmatch(), slabs allocated since entry are freed. | |
2930 | ||
2931 | */ | |
95b24440 | 2932 | |
40a82448 | 2933 | |
5bc10b2c | 2934 | #define DEBUG_STATE_pp(pp) \ |
265c4333 | 2935 | DEBUG_STATE_r({ \ |
f2ed9b32 | 2936 | DUMP_EXEC_POS(locinput, scan, utf8_target); \ |
5bc10b2c | 2937 | PerlIO_printf(Perl_debug_log, \ |
5d458dd8 | 2938 | " %*s"pp" %s%s%s%s%s\n", \ |
5bc10b2c | 2939 | depth*2, "", \ |
13d6edb4 | 2940 | PL_reg_name[st->resume_state], \ |
5d458dd8 YO |
2941 | ((st==yes_state||st==mark_state) ? "[" : ""), \ |
2942 | ((st==yes_state) ? "Y" : ""), \ | |
2943 | ((st==mark_state) ? "M" : ""), \ | |
2944 | ((st==yes_state||st==mark_state) ? "]" : "") \ | |
2945 | ); \ | |
265c4333 | 2946 | }); |
5bc10b2c | 2947 | |
40a82448 | 2948 | |
3dab1dad | 2949 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 2950 | |
3df15adc | 2951 | #ifdef DEBUGGING |
5bc10b2c | 2952 | |
ab3bbdeb | 2953 | STATIC void |
f2ed9b32 | 2954 | S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target, |
ab3bbdeb YO |
2955 | const char *start, const char *end, const char *blurb) |
2956 | { | |
efd26800 | 2957 | const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; |
7918f24d NC |
2958 | |
2959 | PERL_ARGS_ASSERT_DEBUG_START_MATCH; | |
2960 | ||
ab3bbdeb YO |
2961 | if (!PL_colorset) |
2962 | reginitcolors(); | |
2963 | { | |
2964 | RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), | |
d2c6dc5e | 2965 | RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); |
ab3bbdeb | 2966 | |
f2ed9b32 | 2967 | RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb YO |
2968 | start, end - start, 60); |
2969 | ||
2970 | PerlIO_printf(Perl_debug_log, | |
2971 | "%s%s REx%s %s against %s\n", | |
2972 | PL_colors[4], blurb, PL_colors[5], s0, s1); | |
2973 | ||
f2ed9b32 | 2974 | if (utf8_target||utf8_pat) |
1de06328 YO |
2975 | PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", |
2976 | utf8_pat ? "pattern" : "", | |
f2ed9b32 KW |
2977 | utf8_pat && utf8_target ? " and " : "", |
2978 | utf8_target ? "string" : "" | |
ab3bbdeb YO |
2979 | ); |
2980 | } | |
2981 | } | |
3df15adc YO |
2982 | |
2983 | STATIC void | |
786e8c11 YO |
2984 | S_dump_exec_pos(pTHX_ const char *locinput, |
2985 | const regnode *scan, | |
2986 | const char *loc_regeol, | |
2987 | const char *loc_bostr, | |
2988 | const char *loc_reg_starttry, | |
f2ed9b32 | 2989 | const bool utf8_target) |
07be1b83 | 2990 | { |
786e8c11 | 2991 | const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; |
07be1b83 | 2992 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ |
786e8c11 | 2993 | int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); |
07be1b83 YO |
2994 | /* The part of the string before starttry has one color |
2995 | (pref0_len chars), between starttry and current | |
2996 | position another one (pref_len - pref0_len chars), | |
2997 | after the current position the third one. | |
2998 | We assume that pref0_len <= pref_len, otherwise we | |
2999 | decrease pref0_len. */ | |
786e8c11 YO |
3000 | int pref_len = (locinput - loc_bostr) > (5 + taill) - l |
3001 | ? (5 + taill) - l : locinput - loc_bostr; | |
07be1b83 YO |
3002 | int pref0_len; |
3003 | ||
7918f24d NC |
3004 | PERL_ARGS_ASSERT_DUMP_EXEC_POS; |
3005 | ||
f2ed9b32 | 3006 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) |
07be1b83 | 3007 | pref_len++; |
786e8c11 YO |
3008 | pref0_len = pref_len - (locinput - loc_reg_starttry); |
3009 | if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) | |
3010 | l = ( loc_regeol - locinput > (5 + taill) - pref_len | |
3011 | ? (5 + taill) - pref_len : loc_regeol - locinput); | |
f2ed9b32 | 3012 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) |
07be1b83 YO |
3013 | l--; |
3014 | if (pref0_len < 0) | |
3015 | pref0_len = 0; | |
3016 | if (pref0_len > pref_len) | |
3017 | pref0_len = pref_len; | |
3018 | { | |
f2ed9b32 | 3019 | const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0; |
0df25f3d | 3020 | |
ab3bbdeb | 3021 | RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), |
1de06328 | 3022 | (locinput - pref_len),pref0_len, 60, 4, 5); |
0df25f3d | 3023 | |
ab3bbdeb | 3024 | RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), |
3df15adc | 3025 | (locinput - pref_len + pref0_len), |
1de06328 | 3026 | pref_len - pref0_len, 60, 2, 3); |
0df25f3d | 3027 | |
ab3bbdeb | 3028 | RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), |
1de06328 | 3029 | locinput, loc_regeol - locinput, 10, 0, 1); |
0df25f3d | 3030 | |
1de06328 | 3031 | const STRLEN tlen=len0+len1+len2; |
3df15adc | 3032 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb | 3033 | "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", |
786e8c11 | 3034 | (IV)(locinput - loc_bostr), |
07be1b83 | 3035 | len0, s0, |
07be1b83 | 3036 | len1, s1, |
07be1b83 | 3037 | (docolor ? "" : "> <"), |
07be1b83 | 3038 | len2, s2, |
f9f4320a | 3039 | (int)(tlen > 19 ? 0 : 19 - tlen), |
07be1b83 YO |
3040 | ""); |
3041 | } | |
3042 | } | |
3df15adc | 3043 | |
07be1b83 YO |
3044 | #endif |
3045 | ||
0a4db386 YO |
3046 | /* reg_check_named_buff_matched() |
3047 | * Checks to see if a named buffer has matched. The data array of | |
3048 | * buffer numbers corresponding to the buffer is expected to reside | |
3049 | * in the regexp->data->data array in the slot stored in the ARG() of | |
3050 | * node involved. Note that this routine doesn't actually care about the | |
3051 | * name, that information is not preserved from compilation to execution. | |
3052 | * Returns the index of the leftmost defined buffer with the given name | |
3053 | * or 0 if non of the buffers matched. | |
3054 | */ | |
3055 | STATIC I32 | |
7918f24d NC |
3056 | S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) |
3057 | { | |
0a4db386 | 3058 | I32 n; |
f8fc2ecf | 3059 | RXi_GET_DECL(rex,rexi); |
ad64d0ec | 3060 | SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
0a4db386 | 3061 | I32 *nums=(I32*)SvPVX(sv_dat); |
7918f24d NC |
3062 | |
3063 | PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; | |
3064 | ||
0a4db386 | 3065 | for ( n=0; n<SvIVX(sv_dat); n++ ) { |
b93070ed DM |
3066 | if ((I32)rex->lastparen >= nums[n] && |
3067 | rex->offs[nums[n]].end != -1) | |
0a4db386 YO |
3068 | { |
3069 | return nums[n]; | |
3070 | } | |
3071 | } | |
3072 | return 0; | |
3073 | } | |
3074 | ||
2f554ef7 DM |
3075 | |
3076 | /* free all slabs above current one - called during LEAVE_SCOPE */ | |
3077 | ||
3078 | STATIC void | |
3079 | S_clear_backtrack_stack(pTHX_ void *p) | |
3080 | { | |
3081 | regmatch_slab *s = PL_regmatch_slab->next; | |
3082 | PERL_UNUSED_ARG(p); | |
3083 | ||
3084 | if (!s) | |
3085 | return; | |
3086 | PL_regmatch_slab->next = NULL; | |
3087 | while (s) { | |
3088 | regmatch_slab * const osl = s; | |
3089 | s = s->next; | |
3090 | Safefree(osl); | |
3091 | } | |
3092 | } | |
3093 | ||
3094 | ||
d6a28714 | 3095 | STATIC I32 /* 0 failure, 1 success */ |
24b23f37 | 3096 | S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) |
d6a28714 | 3097 | { |
a35a87e7 | 3098 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
3099 | dMY_CXT; |
3100 | #endif | |
27da23d5 | 3101 | dVAR; |
f2ed9b32 | 3102 | register const bool utf8_target = PL_reg_match_utf8; |
4ad0818d | 3103 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
288b8c02 NC |
3104 | REGEXP *rex_sv = reginfo->prog; |
3105 | regexp *rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 3106 | RXi_GET_DECL(rex,rexi); |
2f554ef7 | 3107 | I32 oldsave; |
5d9a96ca DM |
3108 | /* the current state. This is a cached copy of PL_regmatch_state */ |
3109 | register regmatch_state *st; | |
5d9a96ca DM |
3110 | /* cache heavy used fields of st in registers */ |
3111 | register regnode *scan; | |
3112 | register regnode *next; | |
438e9bae | 3113 | register U32 n = 0; /* general value; init to avoid compiler warning */ |
24d3c4a9 | 3114 | register I32 ln = 0; /* len or last; init to avoid compiler warning */ |
5d9a96ca | 3115 | register char *locinput = PL_reginput; |
5d9a96ca | 3116 | register I32 nextchr; /* is always set to UCHARAT(locinput) */ |
24d3c4a9 | 3117 | |
b69b0499 | 3118 | bool result = 0; /* return value of S_regmatch */ |
24d3c4a9 | 3119 | int depth = 0; /* depth of backtrack stack */ |
4b196cd4 YO |
3120 | U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */ |
3121 | const U32 max_nochange_depth = | |
3122 | (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? | |
3123 | 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; | |
77cb431f DM |
3124 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
3125 | subpattern */ | |
e2e6a0f1 YO |
3126 | /* mark_state piggy backs on the yes_state logic so that when we unwind |
3127 | the stack on success we can update the mark_state as we go */ | |
3128 | regmatch_state *mark_state = NULL; /* last mark state we have seen */ | |
faec1544 | 3129 | regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ |
b8591aee | 3130 | struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ |
40a82448 | 3131 | U32 state_num; |
5d458dd8 YO |
3132 | bool no_final = 0; /* prevent failure from backtracking? */ |
3133 | bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ | |
e2e6a0f1 | 3134 | char *startpoint = PL_reginput; |
5d458dd8 YO |
3135 | SV *popmark = NULL; /* are we looking for a mark? */ |
3136 | SV *sv_commit = NULL; /* last mark name seen in failure */ | |
3137 | SV *sv_yes_mark = NULL; /* last mark name we have seen | |
486ec47a | 3138 | during a successful match */ |
5d458dd8 YO |
3139 | U32 lastopen = 0; /* last open we saw */ |
3140 | bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; | |
19b95bf0 | 3141 | SV* const oreplsv = GvSV(PL_replgv); |
24d3c4a9 DM |
3142 | /* these three flags are set by various ops to signal information to |
3143 | * the very next op. They have a useful lifetime of exactly one loop | |
3144 | * iteration, and are not preserved or restored by state pushes/pops | |
3145 | */ | |
3146 | bool sw = 0; /* the condition value in (?(cond)a|b) */ | |
3147 | bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ | |
3148 | int logical = 0; /* the following EVAL is: | |
3149 | 0: (?{...}) | |
3150 | 1: (?(?{...})X|Y) | |
3151 | 2: (??{...}) | |
3152 | or the following IFMATCH/UNLESSM is: | |
3153 | false: plain (?=foo) | |
3154 | true: used as a condition: (?(?=foo)) | |
3155 | */ | |
81ed78b2 DM |
3156 | PAD* last_pad = NULL; |
3157 | dMULTICALL; | |
3158 | I32 gimme = G_SCALAR; | |
3159 | CV *caller_cv = NULL; /* who called us */ | |
3160 | CV *last_pushed_cv = NULL; /* most recently called (?{}) CV */ | |
74088413 | 3161 | CHECKPOINT runops_cp; /* savestack position before executing EVAL */ |
81ed78b2 | 3162 | |
95b24440 | 3163 | #ifdef DEBUGGING |
e68ec53f | 3164 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
3165 | #endif |
3166 | ||
81ed78b2 DM |
3167 | /* shut up 'may be used uninitialized' compiler warnings for dMULTICALL */ |
3168 | multicall_oldcatch = 0; | |
3169 | multicall_cv = NULL; | |
3170 | cx = NULL; | |
4f8dbb2d JL |
3171 | PERL_UNUSED_VAR(multicall_cop); |
3172 | PERL_UNUSED_VAR(newsp); | |
81ed78b2 DM |
3173 | |
3174 | ||
7918f24d NC |
3175 | PERL_ARGS_ASSERT_REGMATCH; |
3176 | ||
3b57cd43 | 3177 | DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ |
24b23f37 | 3178 | PerlIO_printf(Perl_debug_log,"regmatch start\n"); |
3b57cd43 | 3179 | })); |
5d9a96ca DM |
3180 | /* on first ever call to regmatch, allocate first slab */ |
3181 | if (!PL_regmatch_slab) { | |
3182 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
3183 | PL_regmatch_slab->prev = NULL; | |
3184 | PL_regmatch_slab->next = NULL; | |
86545054 | 3185 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); |
5d9a96ca DM |
3186 | } |
3187 | ||
2f554ef7 DM |
3188 | oldsave = PL_savestack_ix; |
3189 | SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL); | |
3190 | SAVEVPTR(PL_regmatch_slab); | |
3191 | SAVEVPTR(PL_regmatch_state); | |
5d9a96ca DM |
3192 | |
3193 | /* grab next free state slot */ | |
3194 | st = ++PL_regmatch_state; | |
86545054 | 3195 | if (st > SLAB_LAST(PL_regmatch_slab)) |
5d9a96ca DM |
3196 | st = PL_regmatch_state = S_push_slab(aTHX); |
3197 | ||
d6a28714 JH |
3198 | /* Note that nextchr is a byte even in UTF */ |
3199 | nextchr = UCHARAT(locinput); | |
3200 | scan = prog; | |
3201 | while (scan != NULL) { | |
8ba1375e | 3202 | |
a3621e74 | 3203 | DEBUG_EXECUTE_r( { |
6136c704 | 3204 | SV * const prop = sv_newmortal(); |
1de06328 | 3205 | regnode *rnext=regnext(scan); |
f2ed9b32 | 3206 | DUMP_EXEC_POS( locinput, scan, utf8_target ); |
32fc9b6a | 3207 | regprop(rex, prop, scan); |
07be1b83 YO |
3208 | |
3209 | PerlIO_printf(Perl_debug_log, | |
3210 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
f8fc2ecf | 3211 | (IV)(scan - rexi->program), depth*2, "", |
07be1b83 | 3212 | SvPVX_const(prop), |
1de06328 | 3213 | (PL_regkind[OP(scan)] == END || !rnext) ? |
f8fc2ecf | 3214 | 0 : (IV)(rnext - rexi->program)); |
2a782b5b | 3215 | }); |
d6a28714 JH |
3216 | |
3217 | next = scan + NEXT_OFF(scan); | |
3218 | if (next == scan) | |
3219 | next = NULL; | |
40a82448 | 3220 | state_num = OP(scan); |
d6a28714 | 3221 | |
40a82448 | 3222 | reenter_switch: |
34a81e2b | 3223 | |
40a82448 | 3224 | switch (state_num) { |
d6a28714 | 3225 | case BOL: |
7fba1cd6 | 3226 | if (locinput == PL_bostr) |
d6a28714 | 3227 | { |
3b0527fe | 3228 | /* reginfo->till = reginfo->bol; */ |
b8c5462f JH |
3229 | break; |
3230 | } | |
d6a28714 JH |
3231 | sayNO; |
3232 | case MBOL: | |
12d33761 HS |
3233 | if (locinput == PL_bostr || |
3234 | ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n')) | |
d6a28714 | 3235 | { |
b8c5462f JH |
3236 | break; |
3237 | } | |
d6a28714 JH |
3238 | sayNO; |
3239 | case SBOL: | |
c2a73568 | 3240 | if (locinput == PL_bostr) |
b8c5462f | 3241 | break; |
d6a28714 JH |
3242 | sayNO; |
3243 | case GPOS: | |
3b0527fe | 3244 | if (locinput == reginfo->ganch) |
d6a28714 JH |
3245 | break; |
3246 | sayNO; | |
ee9b8eae YO |
3247 | |
3248 | case KEEPS: | |
3249 | /* update the startpoint */ | |
b93070ed | 3250 | st->u.keeper.val = rex->offs[0].start; |
ee9b8eae | 3251 | PL_reginput = locinput; |
b93070ed | 3252 | rex->offs[0].start = locinput - PL_bostr; |
ee9b8eae YO |
3253 | PUSH_STATE_GOTO(KEEPS_next, next); |
3254 | /*NOT-REACHED*/ | |
3255 | case KEEPS_next_fail: | |
3256 | /* rollback the start point change */ | |
b93070ed | 3257 | rex->offs[0].start = st->u.keeper.val; |
ee9b8eae YO |
3258 | sayNO_SILENT; |
3259 | /*NOT-REACHED*/ | |
d6a28714 | 3260 | case EOL: |
d6a28714 JH |
3261 | goto seol; |
3262 | case MEOL: | |
d6a28714 | 3263 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') |
b8c5462f | 3264 | sayNO; |
b8c5462f | 3265 | break; |
d6a28714 JH |
3266 | case SEOL: |
3267 | seol: | |
3268 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') | |
b8c5462f | 3269 | sayNO; |
d6a28714 | 3270 | if (PL_regeol - locinput > 1) |
b8c5462f | 3271 | sayNO; |
b8c5462f | 3272 | break; |
d6a28714 JH |
3273 | case EOS: |
3274 | if (PL_regeol != locinput) | |
b8c5462f | 3275 | sayNO; |
d6a28714 | 3276 | break; |
ffc61ed2 | 3277 | case SANY: |
d6a28714 | 3278 | if (!nextchr && locinput >= PL_regeol) |
4633a7c4 | 3279 | sayNO; |
f2ed9b32 | 3280 | if (utf8_target) { |
f33976b4 DB |
3281 | locinput += PL_utf8skip[nextchr]; |
3282 | if (locinput > PL_regeol) | |
3283 | sayNO; | |
3284 | nextchr = UCHARAT(locinput); | |
3285 | } | |
3286 | else | |
3287 | nextchr = UCHARAT(++locinput); | |
3288 | break; | |
3289 | case CANY: | |
3290 | if (!nextchr && locinput >= PL_regeol) | |
3291 | sayNO; | |
b8c5462f | 3292 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 3293 | break; |
ffc61ed2 | 3294 | case REG_ANY: |
1aa99e6b IH |
3295 | if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n') |
3296 | sayNO; | |
f2ed9b32 | 3297 | if (utf8_target) { |
b8c5462f | 3298 | locinput += PL_utf8skip[nextchr]; |
d6a28714 JH |
3299 | if (locinput > PL_regeol) |
3300 | sayNO; | |
a0ed51b3 | 3301 | nextchr = UCHARAT(locinput); |
a0ed51b3 | 3302 | } |
1aa99e6b IH |
3303 | else |
3304 | nextchr = UCHARAT(++locinput); | |
a0ed51b3 | 3305 | break; |
166ba7cd DM |
3306 | |
3307 | #undef ST | |
3308 | #define ST st->u.trie | |
786e8c11 YO |
3309 | case TRIEC: |
3310 | /* In this case the charclass data is available inline so | |
3311 | we can fail fast without a lot of extra overhead. | |
3312 | */ | |
fab2782b YO |
3313 | if(!ANYOF_BITMAP_TEST(scan, *locinput)) { |
3314 | DEBUG_EXECUTE_r( | |
3315 | PerlIO_printf(Perl_debug_log, | |
3316 | "%*s %sfailed to match trie start class...%s\n", | |
3317 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) | |
3318 | ); | |
3319 | sayNO_SILENT; | |
118e2215 | 3320 | assert(0); /* NOTREACHED */ |
786e8c11 YO |
3321 | } |
3322 | /* FALL THROUGH */ | |
5b47454d | 3323 | case TRIE: |
2e64971a DM |
3324 | /* the basic plan of execution of the trie is: |
3325 | * At the beginning, run though all the states, and | |
3326 | * find the longest-matching word. Also remember the position | |
3327 | * of the shortest matching word. For example, this pattern: | |
3328 | * 1 2 3 4 5 | |
3329 | * ab|a|x|abcd|abc | |
3330 | * when matched against the string "abcde", will generate | |
3331 | * accept states for all words except 3, with the longest | |
3332 | * matching word being 4, and the shortest being 1 (with | |
3333 | * the position being after char 1 of the string). | |
3334 | * | |
3335 | * Then for each matching word, in word order (i.e. 1,2,4,5), | |
3336 | * we run the remainder of the pattern; on each try setting | |
3337 | * the current position to the character following the word, | |
3338 | * returning to try the next word on failure. | |
3339 | * | |
3340 | * We avoid having to build a list of words at runtime by | |
3341 | * using a compile-time structure, wordinfo[].prev, which | |
3342 | * gives, for each word, the previous accepting word (if any). | |
3343 | * In the case above it would contain the mappings 1->2, 2->0, | |
3344 | * 3->0, 4->5, 5->1. We can use this table to generate, from | |
3345 | * the longest word (4 above), a list of all words, by | |
3346 | * following the list of prev pointers; this gives us the | |
3347 | * unordered list 4,5,1,2. Then given the current word we have | |
3348 | * just tried, we can go through the list and find the | |
3349 | * next-biggest word to try (so if we just failed on word 2, | |
3350 | * the next in the list is 4). | |
3351 | * | |
3352 | * Since at runtime we don't record the matching position in | |
3353 | * the string for each word, we have to work that out for | |
3354 | * each word we're about to process. The wordinfo table holds | |
3355 | * the character length of each word; given that we recorded | |
3356 | * at the start: the position of the shortest word and its | |
3357 | * length in chars, we just need to move the pointer the | |
3358 | * difference between the two char lengths. Depending on | |
3359 | * Unicode status and folding, that's cheap or expensive. | |
3360 | * | |
3361 | * This algorithm is optimised for the case where are only a | |
3362 | * small number of accept states, i.e. 0,1, or maybe 2. | |
3363 | * With lots of accepts states, and having to try all of them, | |
3364 | * it becomes quadratic on number of accept states to find all | |
3365 | * the next words. | |
3366 | */ | |
3367 | ||
3dab1dad | 3368 | { |
07be1b83 | 3369 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
a0a388a1 | 3370 | DECL_TRIE_TYPE(scan); |
3dab1dad YO |
3371 | |
3372 | /* what trie are we using right now */ | |
be8e71aa | 3373 | reg_trie_data * const trie |
f8fc2ecf | 3374 | = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; |
85fbaab2 | 3375 | HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); |
3dab1dad | 3376 | U32 state = trie->startstate; |
166ba7cd | 3377 | |
fab2782b | 3378 | if (trie->bitmap && !TRIE_BITMAP_TEST(trie,*locinput) ) { |
3dab1dad YO |
3379 | if (trie->states[ state ].wordnum) { |
3380 | DEBUG_EXECUTE_r( | |
3381 | PerlIO_printf(Perl_debug_log, | |
3382 | "%*s %smatched empty string...%s\n", | |
5bc10b2c | 3383 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad | 3384 | ); |
20dbff7c YO |
3385 | if (!trie->jump) |
3386 | break; | |
3dab1dad YO |
3387 | } else { |
3388 | DEBUG_EXECUTE_r( | |
3389 | PerlIO_printf(Perl_debug_log, | |
786e8c11 | 3390 | "%*s %sfailed to match trie start class...%s\n", |
5bc10b2c | 3391 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad YO |
3392 | ); |
3393 | sayNO_SILENT; | |
3394 | } | |
3395 | } | |
166ba7cd | 3396 | |
786e8c11 YO |
3397 | { |
3398 | U8 *uc = ( U8* )locinput; | |
3399 | ||
3400 | STRLEN len = 0; | |
3401 | STRLEN foldlen = 0; | |
3402 | U8 *uscan = (U8*)NULL; | |
786e8c11 | 3403 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
2e64971a DM |
3404 | U32 charcount = 0; /* how many input chars we have matched */ |
3405 | U32 accepted = 0; /* have we seen any accepting states? */ | |
786e8c11 | 3406 | |
786e8c11 | 3407 | ST.jump = trie->jump; |
786e8c11 | 3408 | ST.me = scan; |
2e64971a DM |
3409 | ST.firstpos = NULL; |
3410 | ST.longfold = FALSE; /* char longer if folded => it's harder */ | |
3411 | ST.nextword = 0; | |
3412 | ||
3413 | /* fully traverse the TRIE; note the position of the | |
3414 | shortest accept state and the wordnum of the longest | |
3415 | accept state */ | |
07be1b83 | 3416 | |
a3621e74 | 3417 | while ( state && uc <= (U8*)PL_regeol ) { |
786e8c11 | 3418 | U32 base = trie->states[ state ].trans.base; |
f9f4320a | 3419 | UV uvc = 0; |
acb909b4 | 3420 | U16 charid = 0; |
2e64971a DM |
3421 | U16 wordnum; |
3422 | wordnum = trie->states[ state ].wordnum; | |
3423 | ||
3424 | if (wordnum) { /* it's an accept state */ | |
3425 | if (!accepted) { | |
3426 | accepted = 1; | |
3427 | /* record first match position */ | |
3428 | if (ST.longfold) { | |
3429 | ST.firstpos = (U8*)locinput; | |
3430 | ST.firstchars = 0; | |
5b47454d | 3431 | } |
2e64971a DM |
3432 | else { |
3433 | ST.firstpos = uc; | |
3434 | ST.firstchars = charcount; | |
3435 | } | |
3436 | } | |
3437 | if (!ST.nextword || wordnum < ST.nextword) | |
3438 | ST.nextword = wordnum; | |
3439 | ST.topword = wordnum; | |
786e8c11 | 3440 | } |
a3621e74 | 3441 | |
07be1b83 | 3442 | DEBUG_TRIE_EXECUTE_r({ |
f2ed9b32 | 3443 | DUMP_EXEC_POS( (char *)uc, scan, utf8_target ); |
a3621e74 | 3444 | PerlIO_printf( Perl_debug_log, |
2e64971a | 3445 | "%*s %sState: %4"UVxf" Accepted: %c ", |
5bc10b2c | 3446 | 2+depth * 2, "", PL_colors[4], |
2e64971a | 3447 | (UV)state, (accepted ? 'Y' : 'N')); |
07be1b83 | 3448 | }); |
a3621e74 | 3449 | |
2e64971a | 3450 | /* read a char and goto next state */ |
a3621e74 | 3451 | if ( base ) { |
6dd2be57 | 3452 | I32 offset; |
55eed653 NC |
3453 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
3454 | uscan, len, uvc, charid, foldlen, | |
3455 | foldbuf, uniflags); | |
2e64971a DM |
3456 | charcount++; |
3457 | if (foldlen>0) | |
3458 | ST.longfold = TRUE; | |
5b47454d | 3459 | if (charid && |
6dd2be57 DM |
3460 | ( ((offset = |
3461 | base + charid - 1 - trie->uniquecharcount)) >= 0) | |
3462 | ||
3463 | && ((U32)offset < trie->lasttrans) | |
3464 | && trie->trans[offset].check == state) | |
5b47454d | 3465 | { |
6dd2be57 | 3466 | state = trie->trans[offset].next; |
5b47454d DM |
3467 | } |
3468 | else { | |
3469 | state = 0; | |
3470 | } | |
3471 | uc += len; | |
3472 | ||
3473 | } | |
3474 | else { | |
a3621e74 YO |
3475 | state = 0; |
3476 | } | |
3477 | DEBUG_TRIE_EXECUTE_r( | |
e4584336 | 3478 | PerlIO_printf( Perl_debug_log, |
786e8c11 | 3479 | "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n", |
e4584336 | 3480 | charid, uvc, (UV)state, PL_colors[5] ); |
a3621e74 YO |
3481 | ); |
3482 | } | |
2e64971a | 3483 | if (!accepted) |
a3621e74 | 3484 | sayNO; |
a3621e74 | 3485 | |
2e64971a DM |
3486 | /* calculate total number of accept states */ |
3487 | { | |
3488 | U16 w = ST.topword; | |
3489 | accepted = 0; | |
3490 | while (w) { | |
3491 | w = trie->wordinfo[w].prev; | |
3492 | accepted++; | |
3493 | } | |
3494 | ST.accepted = accepted; | |
3495 | } | |
3496 | ||
166ba7cd DM |
3497 | DEBUG_EXECUTE_r( |
3498 | PerlIO_printf( Perl_debug_log, | |
3499 | "%*s %sgot %"IVdf" possible matches%s\n", | |
5bc10b2c | 3500 | REPORT_CODE_OFF + depth * 2, "", |
166ba7cd DM |
3501 | PL_colors[4], (IV)ST.accepted, PL_colors[5] ); |
3502 | ); | |
2e64971a | 3503 | goto trie_first_try; /* jump into the fail handler */ |
786e8c11 | 3504 | }} |
118e2215 | 3505 | assert(0); /* NOTREACHED */ |
2e64971a DM |
3506 | |
3507 | case TRIE_next_fail: /* we failed - try next alternative */ | |
fae667d5 YO |
3508 | if ( ST.jump) { |
3509 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 | 3510 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
fae667d5 | 3511 | } |
2e64971a DM |
3512 | if (!--ST.accepted) { |
3513 | DEBUG_EXECUTE_r({ | |
3514 | PerlIO_printf( Perl_debug_log, | |
3515 | "%*s %sTRIE failed...%s\n", | |
3516 | REPORT_CODE_OFF+depth*2, "", | |
3517 | PL_colors[4], | |
3518 | PL_colors[5] ); | |
3519 | }); | |
3520 | sayNO_SILENT; | |
3521 | } | |
3522 | { | |
3523 | /* Find next-highest word to process. Note that this code | |
3524 | * is O(N^2) per trie run (O(N) per branch), so keep tight */ | |
d9a396a3 DM |
3525 | register U16 min = 0; |
3526 | register U16 word; | |
2e64971a DM |
3527 | register U16 const nextword = ST.nextword; |
3528 | register reg_trie_wordinfo * const wordinfo | |
3529 | = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo; | |
3530 | for (word=ST.topword; word; word=wordinfo[word].prev) { | |
3531 | if (word > nextword && (!min || word < min)) | |
3532 | min = word; | |
3533 | } | |
3534 | ST.nextword = min; | |
3535 | } | |
3536 | ||
fae667d5 | 3537 | trie_first_try: |
5d458dd8 YO |
3538 | if (do_cutgroup) { |
3539 | do_cutgroup = 0; | |
3540 | no_final = 0; | |
3541 | } | |
fae667d5 YO |
3542 | |
3543 | if ( ST.jump) { | |
b93070ed | 3544 | ST.lastparen = rex->lastparen; |
f6033a9d | 3545 | ST.lastcloseparen = rex->lastcloseparen; |
fae667d5 | 3546 | REGCP_SET(ST.cp); |
2e64971a | 3547 | } |
a3621e74 | 3548 | |
2e64971a | 3549 | /* find start char of end of current word */ |
166ba7cd | 3550 | { |
2e64971a DM |
3551 | U32 chars; /* how many chars to skip */ |
3552 | U8 *uc = ST.firstpos; | |
3553 | reg_trie_data * const trie | |
3554 | = (reg_trie_data*)rexi->data->data[ARG(ST.me)]; | |
3555 | ||
3556 | assert((trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3557 | >= ST.firstchars); | |
3558 | chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3559 | - ST.firstchars; | |
3560 | ||
3561 | if (ST.longfold) { | |
3562 | /* the hard option - fold each char in turn and find | |
3563 | * its folded length (which may be different */ | |
3564 | U8 foldbuf[UTF8_MAXBYTES_CASE + 1]; | |
3565 | STRLEN foldlen; | |
3566 | STRLEN len; | |
d9a396a3 | 3567 | UV uvc; |
2e64971a DM |
3568 | U8 *uscan; |
3569 | ||
3570 | while (chars) { | |
f2ed9b32 | 3571 | if (utf8_target) { |
2e64971a DM |
3572 | uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len, |
3573 | uniflags); | |
3574 | uc += len; | |
3575 | } | |
3576 | else { | |
3577 | uvc = *uc; | |
3578 | uc++; | |
3579 | } | |
3580 | uvc = to_uni_fold(uvc, foldbuf, &foldlen); | |
3581 | uscan = foldbuf; | |
3582 | while (foldlen) { | |
3583 | if (!--chars) | |
3584 | break; | |
3585 | uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len, | |
3586 | uniflags); | |
3587 | uscan += len; | |
3588 | foldlen -= len; | |
3589 | } | |
3590 | } | |
a3621e74 | 3591 | } |
2e64971a | 3592 | else { |
f2ed9b32 | 3593 | if (utf8_target) |
2e64971a DM |
3594 | while (chars--) |
3595 | uc += UTF8SKIP(uc); | |
3596 | else | |
3597 | uc += chars; | |
3598 | } | |
3599 | PL_reginput = (char *)uc; | |
3600 | } | |
166ba7cd | 3601 | |
6603fe3e DM |
3602 | scan = ST.me + ((ST.jump && ST.jump[ST.nextword]) |
3603 | ? ST.jump[ST.nextword] | |
3604 | : NEXT_OFF(ST.me)); | |
166ba7cd | 3605 | |
2e64971a DM |
3606 | DEBUG_EXECUTE_r({ |
3607 | PerlIO_printf( Perl_debug_log, | |
3608 | "%*s %sTRIE matched word #%d, continuing%s\n", | |
3609 | REPORT_CODE_OFF+depth*2, "", | |
3610 | PL_colors[4], | |
3611 | ST.nextword, | |
3612 | PL_colors[5] | |
3613 | ); | |
3614 | }); | |
3615 | ||
3616 | if (ST.accepted > 1 || has_cutgroup) { | |
3617 | PUSH_STATE_GOTO(TRIE_next, scan); | |
118e2215 | 3618 | assert(0); /* NOTREACHED */ |
166ba7cd | 3619 | } |
2e64971a DM |
3620 | /* only one choice left - just continue */ |
3621 | DEBUG_EXECUTE_r({ | |
3622 | AV *const trie_words | |
3623 | = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]); | |
3624 | SV ** const tmp = av_fetch( trie_words, | |
3625 | ST.nextword-1, 0 ); | |
3626 | SV *sv= tmp ? sv_newmortal() : NULL; | |
3627 | ||
3628 | PerlIO_printf( Perl_debug_log, | |
3629 | "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n", | |
3630 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], | |
3631 | ST.nextword, | |
3632 | tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0, | |
3633 | PL_colors[0], PL_colors[1], | |
c89df6cf | 3634 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII |
2e64971a DM |
3635 | ) |
3636 | : "not compiled under -Dr", | |
3637 | PL_colors[5] ); | |
3638 | }); | |
3639 | ||
3640 | locinput = PL_reginput; | |
3641 | nextchr = UCHARAT(locinput); | |
3642 | continue; /* execute rest of RE */ | |
118e2215 | 3643 | assert(0); /* NOTREACHED */ |
166ba7cd DM |
3644 | #undef ST |
3645 | ||
95b24440 DM |
3646 | case EXACT: { |
3647 | char *s = STRING(scan); | |
24d3c4a9 | 3648 | ln = STR_LEN(scan); |
f2ed9b32 | 3649 | if (utf8_target != UTF_PATTERN) { |
bc517b45 | 3650 | /* The target and the pattern have differing utf8ness. */ |
1aa99e6b | 3651 | char *l = locinput; |
24d3c4a9 | 3652 | const char * const e = s + ln; |
a72c7584 | 3653 | |
f2ed9b32 | 3654 | if (utf8_target) { |
5ff6fc6d | 3655 | /* The target is utf8, the pattern is not utf8. */ |
1aa99e6b | 3656 | while (s < e) { |
a3b680e6 | 3657 | STRLEN ulen; |
1aa99e6b | 3658 | if (l >= PL_regeol) |
5ff6fc6d JH |
3659 | sayNO; |
3660 | if (NATIVE_TO_UNI(*(U8*)s) != | |
89ebb4a3 | 3661 | utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen, |
041457d9 | 3662 | uniflags)) |
5ff6fc6d | 3663 | sayNO; |
bc517b45 | 3664 | l += ulen; |
5ff6fc6d | 3665 | s ++; |
1aa99e6b | 3666 | } |
5ff6fc6d JH |
3667 | } |
3668 | else { | |
3669 | /* The target is not utf8, the pattern is utf8. */ | |
1aa99e6b | 3670 | while (s < e) { |
a3b680e6 | 3671 | STRLEN ulen; |
1aa99e6b IH |
3672 | if (l >= PL_regeol) |
3673 | sayNO; | |
5ff6fc6d | 3674 | if (NATIVE_TO_UNI(*((U8*)l)) != |
89ebb4a3 | 3675 | utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen, |
041457d9 | 3676 | uniflags)) |
1aa99e6b | 3677 | sayNO; |
bc517b45 | 3678 | s += ulen; |
a72c7584 | 3679 | l ++; |
1aa99e6b | 3680 | } |
5ff6fc6d | 3681 | } |
1aa99e6b IH |
3682 | locinput = l; |
3683 | nextchr = UCHARAT(locinput); | |
3684 | break; | |
3685 | } | |
bc517b45 | 3686 | /* The target and the pattern have the same utf8ness. */ |
d6a28714 JH |
3687 | /* Inline the first character, for speed. */ |
3688 | if (UCHARAT(s) != nextchr) | |
3689 | sayNO; | |
24d3c4a9 | 3690 | if (PL_regeol - locinput < ln) |
d6a28714 | 3691 | sayNO; |
24d3c4a9 | 3692 | if (ln > 1 && memNE(s, locinput, ln)) |
d6a28714 | 3693 | sayNO; |
24d3c4a9 | 3694 | locinput += ln; |
d6a28714 JH |
3695 | nextchr = UCHARAT(locinput); |
3696 | break; | |
95b24440 | 3697 | } |
9a5a5549 | 3698 | case EXACTFL: { |
a932d541 | 3699 | re_fold_t folder; |
9a5a5549 KW |
3700 | const U8 * fold_array; |
3701 | const char * s; | |
d513472c | 3702 | U32 fold_utf8_flags; |
9a5a5549 | 3703 | |
b8c5462f | 3704 | PL_reg_flags |= RF_tainted; |
9a5a5549 KW |
3705 | folder = foldEQ_locale; |
3706 | fold_array = PL_fold_locale; | |
17580e7a | 3707 | fold_utf8_flags = FOLDEQ_UTF8_LOCALE; |
9a5a5549 KW |
3708 | goto do_exactf; |
3709 | ||
3c760661 | 3710 | case EXACTFU_SS: |
fab2782b | 3711 | case EXACTFU_TRICKYFOLD: |
9a5a5549 KW |
3712 | case EXACTFU: |
3713 | folder = foldEQ_latin1; | |
3714 | fold_array = PL_fold_latin1; | |
2daa8fee | 3715 | fold_utf8_flags = (UTF_PATTERN) ? FOLDEQ_S1_ALREADY_FOLDED : 0; |
9a5a5549 KW |
3716 | goto do_exactf; |
3717 | ||
2f7f8cb1 KW |
3718 | case EXACTFA: |
3719 | folder = foldEQ_latin1; | |
3720 | fold_array = PL_fold_latin1; | |
57014d77 | 3721 | fold_utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; |
2f7f8cb1 KW |
3722 | goto do_exactf; |
3723 | ||
9a5a5549 KW |
3724 | case EXACTF: |
3725 | folder = foldEQ; | |
3726 | fold_array = PL_fold; | |
62bf7766 | 3727 | fold_utf8_flags = 0; |
9a5a5549 KW |
3728 | |
3729 | do_exactf: | |
3730 | s = STRING(scan); | |
24d3c4a9 | 3731 | ln = STR_LEN(scan); |
d6a28714 | 3732 | |
3c760661 KW |
3733 | if (utf8_target || UTF_PATTERN || state_num == EXACTFU_SS) { |
3734 | /* Either target or the pattern are utf8, or has the issue where | |
3735 | * the fold lengths may differ. */ | |
be8e71aa | 3736 | const char * const l = locinput; |
d07ddd77 | 3737 | char *e = PL_regeol; |
bc517b45 | 3738 | |
d513472c | 3739 | if (! foldEQ_utf8_flags(s, 0, ln, cBOOL(UTF_PATTERN), |
fa5b1667 | 3740 | l, &e, 0, utf8_target, fold_utf8_flags)) |
c3e1d013 KW |
3741 | { |
3742 | sayNO; | |
5486206c | 3743 | } |
d07ddd77 JH |
3744 | locinput = e; |
3745 | nextchr = UCHARAT(locinput); | |
3746 | break; | |
a0ed51b3 | 3747 | } |
d6a28714 | 3748 | |
0a138b74 | 3749 | /* Neither the target nor the pattern are utf8 */ |
d6a28714 | 3750 | if (UCHARAT(s) != nextchr && |
9a5a5549 KW |
3751 | UCHARAT(s) != fold_array[nextchr]) |
3752 | { | |
a0ed51b3 | 3753 | sayNO; |
9a5a5549 | 3754 | } |
24d3c4a9 | 3755 | if (PL_regeol - locinput < ln) |
b8c5462f | 3756 | sayNO; |
9a5a5549 | 3757 | if (ln > 1 && ! folder(s, locinput, ln)) |
4633a7c4 | 3758 | sayNO; |
24d3c4a9 | 3759 | locinput += ln; |
d6a28714 | 3760 | nextchr = UCHARAT(locinput); |
a0d0e21e | 3761 | break; |
9a5a5549 | 3762 | } |
63ac0dad KW |
3763 | |
3764 | /* XXX Could improve efficiency by separating these all out using a | |
3765 | * macro or in-line function. At that point regcomp.c would no longer | |
3766 | * have to set the FLAGS fields of these */ | |
b2680017 YO |
3767 | case BOUNDL: |
3768 | case NBOUNDL: | |
3769 | PL_reg_flags |= RF_tainted; | |
3770 | /* FALL THROUGH */ | |
3771 | case BOUND: | |
63ac0dad | 3772 | case BOUNDU: |
cfaf538b | 3773 | case BOUNDA: |
b2680017 | 3774 | case NBOUND: |
63ac0dad | 3775 | case NBOUNDU: |
cfaf538b | 3776 | case NBOUNDA: |
b2680017 | 3777 | /* was last char in word? */ |
f2e96b5d KW |
3778 | if (utf8_target |
3779 | && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET | |
3780 | && FLAGS(scan) != REGEX_ASCII_MORE_RESTRICTED_CHARSET) | |
3781 | { | |
b2680017 YO |
3782 | if (locinput == PL_bostr) |
3783 | ln = '\n'; | |
3784 | else { | |
3785 | const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); | |
3786 | ||
3787 | ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); | |
3788 | } | |
63ac0dad | 3789 | if (FLAGS(scan) != REGEX_LOCALE_CHARSET) { |
b2680017 YO |
3790 | ln = isALNUM_uni(ln); |
3791 | LOAD_UTF8_CHARCLASS_ALNUM(); | |
f2ed9b32 | 3792 | n = swash_fetch(PL_utf8_alnum, (U8*)locinput, utf8_target); |
b2680017 YO |
3793 | } |
3794 | else { | |
3795 | ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); | |
3796 | n = isALNUM_LC_utf8((U8*)locinput); | |
3797 | } | |
3798 | } | |
3799 | else { | |
cfaf538b KW |
3800 | |
3801 | /* Here the string isn't utf8, or is utf8 and only ascii | |
3802 | * characters are to match \w. In the latter case looking at | |
3803 | * the byte just prior to the current one may be just the final | |
3804 | * byte of a multi-byte character. This is ok. There are two | |
3805 | * cases: | |
3806 | * 1) it is a single byte character, and then the test is doing | |
3807 | * just what it's supposed to. | |
3808 | * 2) it is a multi-byte character, in which case the final | |
3809 | * byte is never mistakable for ASCII, and so the test | |
3810 | * will say it is not a word character, which is the | |
3811 | * correct answer. */ | |
b2680017 YO |
3812 | ln = (locinput != PL_bostr) ? |
3813 | UCHARAT(locinput - 1) : '\n'; | |
63ac0dad KW |
3814 | switch (FLAGS(scan)) { |
3815 | case REGEX_UNICODE_CHARSET: | |
3816 | ln = isWORDCHAR_L1(ln); | |
3817 | n = isWORDCHAR_L1(nextchr); | |
3818 | break; | |
3819 | case REGEX_LOCALE_CHARSET: | |
3820 | ln = isALNUM_LC(ln); | |
3821 | n = isALNUM_LC(nextchr); | |
3822 | break; | |
3823 | case REGEX_DEPENDS_CHARSET: | |
3824 | ln = isALNUM(ln); | |
3825 | n = isALNUM(nextchr); | |
3826 | break; | |
cfaf538b | 3827 | case REGEX_ASCII_RESTRICTED_CHARSET: |
c973bd4f | 3828 | case REGEX_ASCII_MORE_RESTRICTED_CHARSET: |
cfaf538b KW |
3829 | ln = isWORDCHAR_A(ln); |
3830 | n = isWORDCHAR_A(nextchr); | |
3831 | break; | |
63ac0dad KW |
3832 | default: |
3833 | Perl_croak(aTHX_ "panic: Unexpected FLAGS %u in op %u", FLAGS(scan), OP(scan)); | |
3834 | break; | |
b2680017 YO |
3835 | } |
3836 | } | |
63ac0dad KW |
3837 | /* Note requires that all BOUNDs be lower than all NBOUNDs in |
3838 | * regcomp.sym */ | |
3839 | if (((!ln) == (!n)) == (OP(scan) < NBOUND)) | |
b2680017 YO |
3840 | sayNO; |
3841 | break; | |
f56b6394 | 3842 | case ANYOFV: |
d6a28714 | 3843 | case ANYOF: |
f56b6394 | 3844 | if (utf8_target || state_num == ANYOFV) { |
9e55ce06 | 3845 | STRLEN inclasslen = PL_regeol - locinput; |
20ed0b26 KW |
3846 | if (locinput >= PL_regeol) |
3847 | sayNO; | |
9e55ce06 | 3848 | |
f2ed9b32 | 3849 | if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target)) |
09b08e9b | 3850 | sayNO; |
b32d7d3e | 3851 | locinput += inclasslen; |
b8c5462f | 3852 | nextchr = UCHARAT(locinput); |
e0f9d4a8 | 3853 | break; |
ffc61ed2 JH |
3854 | } |
3855 | else { | |
3856 | if (nextchr < 0) | |
3857 | nextchr = UCHARAT(locinput); | |
ffc61ed2 JH |
3858 | if (!nextchr && locinput >= PL_regeol) |
3859 | sayNO; | |
20ed0b26 | 3860 | if (!REGINCLASS(rex, scan, (U8*)locinput)) |
09b08e9b | 3861 | sayNO; |
ffc61ed2 | 3862 | nextchr = UCHARAT(++locinput); |
e0f9d4a8 JH |
3863 | break; |
3864 | } | |
b8c5462f | 3865 | break; |
20d0b1e9 | 3866 | /* Special char classes - The defines start on line 129 or so */ |
ee9a90b8 KW |
3867 | CCC_TRY_U(ALNUM, NALNUM, isWORDCHAR, |
3868 | ALNUML, NALNUML, isALNUM_LC, isALNUM_LC_utf8, | |
3869 | ALNUMU, NALNUMU, isWORDCHAR_L1, | |
cfaf538b | 3870 | ALNUMA, NALNUMA, isWORDCHAR_A, |
779d7b58 | 3871 | alnum, "a"); |
ee9a90b8 KW |
3872 | |
3873 | CCC_TRY_U(SPACE, NSPACE, isSPACE, | |
3874 | SPACEL, NSPACEL, isSPACE_LC, isSPACE_LC_utf8, | |
3875 | SPACEU, NSPACEU, isSPACE_L1, | |
cfaf538b | 3876 | SPACEA, NSPACEA, isSPACE_A, |
779d7b58 | 3877 | space, " "); |
ee9a90b8 KW |
3878 | |
3879 | CCC_TRY(DIGIT, NDIGIT, isDIGIT, | |
3880 | DIGITL, NDIGITL, isDIGIT_LC, isDIGIT_LC_utf8, | |
cfaf538b | 3881 | DIGITA, NDIGITA, isDIGIT_A, |
779d7b58 | 3882 | digit, "0"); |
20d0b1e9 | 3883 | |
37e2e78e KW |
3884 | case CLUMP: /* Match \X: logical Unicode character. This is defined as |
3885 | a Unicode extended Grapheme Cluster */ | |
3886 | /* From http://www.unicode.org/reports/tr29 (5.2 version). An | |
3887 | extended Grapheme Cluster is: | |
3888 | ||
3889 | CR LF | |
3890 | | Prepend* Begin Extend* | |
3891 | | . | |
3892 | ||
3893 | Begin is (Hangul-syllable | ! Control) | |
3894 | Extend is (Grapheme_Extend | Spacing_Mark) | |
3895 | Control is [ GCB_Control CR LF ] | |
3896 | ||
3897 | The discussion below shows how the code for CLUMP is derived | |
3898 | from this regex. Note that most of these concepts are from | |
3899 | property values of the Grapheme Cluster Boundary (GCB) property. | |
3900 | No code point can have multiple property values for a given | |
3901 | property. Thus a code point in Prepend can't be in Control, but | |
3902 | it must be in !Control. This is why Control above includes | |
3903 | GCB_Control plus CR plus LF. The latter two are used in the GCB | |
3904 | property separately, and so can't be in GCB_Control, even though | |
3905 | they logically are controls. Control is not the same as gc=cc, | |
3906 | but includes format and other characters as well. | |
3907 | ||
3908 | The Unicode definition of Hangul-syllable is: | |
3909 | L+ | |
3910 | | (L* ( ( V | LV ) V* | LVT ) T*) | |
3911 | | T+ | |
3912 | ) | |
3913 | Each of these is a value for the GCB property, and hence must be | |
3914 | disjoint, so the order they are tested is immaterial, so the | |
3915 | above can safely be changed to | |
3916 | T+ | |
3917 | | L+ | |
3918 | | (L* ( LVT | ( V | LV ) V*) T*) | |
3919 | ||
3920 | The last two terms can be combined like this: | |
3921 | L* ( L | |
3922 | | (( LVT | ( V | LV ) V*) T*)) | |
3923 | ||
3924 | And refactored into this: | |
3925 | L* (L | LVT T* | V V* T* | LV V* T*) | |
3926 | ||
3927 | That means that if we have seen any L's at all we can quit | |
cc3b396d | 3928 | there, but if the next character is an LVT, a V, or an LV we |
37e2e78e KW |
3929 | should keep going. |
3930 | ||
3931 | There is a subtlety with Prepend* which showed up in testing. | |
3932 | Note that the Begin, and only the Begin is required in: | |
3933 | | Prepend* Begin Extend* | |
cc3b396d KW |
3934 | Also, Begin contains '! Control'. A Prepend must be a |
3935 | '! Control', which means it must also be a Begin. What it | |
3936 | comes down to is that if we match Prepend* and then find no | |
3937 | suitable Begin afterwards, that if we backtrack the last | |
3938 | Prepend, that one will be a suitable Begin. | |
37e2e78e KW |
3939 | */ |
3940 | ||
b7c83a7e | 3941 | if (locinput >= PL_regeol) |
a0ed51b3 | 3942 | sayNO; |
f2ed9b32 | 3943 | if (! utf8_target) { |
37e2e78e KW |
3944 | |
3945 | /* Match either CR LF or '.', as all the other possibilities | |
3946 | * require utf8 */ | |
3947 | locinput++; /* Match the . or CR */ | |
cc3b396d KW |
3948 | if (nextchr == '\r' /* And if it was CR, and the next is LF, |
3949 | match the LF */ | |
37e2e78e KW |
3950 | && locinput < PL_regeol |
3951 | && UCHARAT(locinput) == '\n') locinput++; | |
3952 | } | |
3953 | else { | |
3954 | ||
3955 | /* Utf8: See if is ( CR LF ); already know that locinput < | |
3956 | * PL_regeol, so locinput+1 is in bounds */ | |
3957 | if (nextchr == '\r' && UCHARAT(locinput + 1) == '\n') { | |
3958 | locinput += 2; | |
3959 | } | |
3960 | else { | |
3961 | /* In case have to backtrack to beginning, then match '.' */ | |
3962 | char *starting = locinput; | |
3963 | ||
3964 | /* In case have to backtrack the last prepend */ | |
3965 | char *previous_prepend = 0; | |
3966 | ||
3967 | LOAD_UTF8_CHARCLASS_GCB(); | |
3968 | ||
3969 | /* Match (prepend)* */ | |
3970 | while (locinput < PL_regeol | |
3971 | && swash_fetch(PL_utf8_X_prepend, | |
f2ed9b32 | 3972 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3973 | { |
3974 | previous_prepend = locinput; | |
3975 | locinput += UTF8SKIP(locinput); | |
3976 | } | |
3977 | ||
3978 | /* As noted above, if we matched a prepend character, but | |
3979 | * the next thing won't match, back off the last prepend we | |
3980 | * matched, as it is guaranteed to match the begin */ | |
3981 | if (previous_prepend | |
3982 | && (locinput >= PL_regeol | |
3983 | || ! swash_fetch(PL_utf8_X_begin, | |
f2ed9b32 | 3984 | (U8*)locinput, utf8_target))) |
37e2e78e KW |
3985 | { |
3986 | locinput = previous_prepend; | |
3987 | } | |
3988 | ||
3989 | /* Note that here we know PL_regeol > locinput, as we | |
3990 | * tested that upon input to this switch case, and if we | |
3991 | * moved locinput forward, we tested the result just above | |
3992 | * and it either passed, or we backed off so that it will | |
3993 | * now pass */ | |
f2ed9b32 | 3994 | if (! swash_fetch(PL_utf8_X_begin, (U8*)locinput, utf8_target)) { |
37e2e78e KW |
3995 | |
3996 | /* Here did not match the required 'Begin' in the | |
3997 | * second term. So just match the very first | |
3998 | * character, the '.' of the final term of the regex */ | |
3999 | locinput = starting + UTF8SKIP(starting); | |
4000 | } else { | |
4001 | ||
4002 | /* Here is the beginning of a character that can have | |
4003 | * an extender. It is either a hangul syllable, or a | |
4004 | * non-control */ | |
4005 | if (swash_fetch(PL_utf8_X_non_hangul, | |
f2ed9b32 | 4006 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4007 | { |
4008 | ||
4009 | /* Here not a Hangul syllable, must be a | |
4010 | * ('! * Control') */ | |
4011 | locinput += UTF8SKIP(locinput); | |
4012 | } else { | |
4013 | ||
4014 | /* Here is a Hangul syllable. It can be composed | |
4015 | * of several individual characters. One | |
4016 | * possibility is T+ */ | |
4017 | if (swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 4018 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4019 | { |
4020 | while (locinput < PL_regeol | |
4021 | && swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 4022 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4023 | { |
4024 | locinput += UTF8SKIP(locinput); | |
4025 | } | |
4026 | } else { | |
4027 | ||
4028 | /* Here, not T+, but is a Hangul. That means | |
4029 | * it is one of the others: L, LV, LVT or V, | |
4030 | * and matches: | |
4031 | * L* (L | LVT T* | V V* T* | LV V* T*) */ | |
4032 | ||
4033 | /* Match L* */ | |
4034 | while (locinput < PL_regeol | |
4035 | && swash_fetch(PL_utf8_X_L, | |
f2ed9b32 | 4036 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4037 | { |
4038 | locinput += UTF8SKIP(locinput); | |
4039 | } | |
4040 | ||
4041 | /* Here, have exhausted L*. If the next | |
4042 | * character is not an LV, LVT nor V, it means | |
4043 | * we had to have at least one L, so matches L+ | |
4044 | * in the original equation, we have a complete | |
4045 | * hangul syllable. Are done. */ | |
4046 | ||
4047 | if (locinput < PL_regeol | |
4048 | && swash_fetch(PL_utf8_X_LV_LVT_V, | |
f2ed9b32 | 4049 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4050 | { |
4051 | ||
4052 | /* Otherwise keep going. Must be LV, LVT | |
4053 | * or V. See if LVT */ | |
4054 | if (swash_fetch(PL_utf8_X_LVT, | |
f2ed9b32 | 4055 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4056 | { |
4057 | locinput += UTF8SKIP(locinput); | |
4058 | } else { | |
4059 | ||
4060 | /* Must be V or LV. Take it, then | |
4061 | * match V* */ | |
4062 | locinput += UTF8SKIP(locinput); | |
4063 | while (locinput < PL_regeol | |
4064 | && swash_fetch(PL_utf8_X_V, | |
f2ed9b32 | 4065 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4066 | { |
4067 | locinput += UTF8SKIP(locinput); | |
4068 | } | |
4069 | } | |
4070 | ||
4071 | /* And any of LV, LVT, or V can be followed | |
4072 | * by T* */ | |
4073 | while (locinput < PL_regeol | |
4074 | && swash_fetch(PL_utf8_X_T, | |
4075 | (U8*)locinput, | |
f2ed9b32 | 4076 | utf8_target)) |
37e2e78e KW |
4077 | { |
4078 | locinput += UTF8SKIP(locinput); | |
4079 | } | |
4080 | } | |
4081 | } | |
4082 | } | |
4083 | ||
4084 | /* Match any extender */ | |
4085 | while (locinput < PL_regeol | |
4086 | && swash_fetch(PL_utf8_X_extend, | |
f2ed9b32 | 4087 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4088 | { |
4089 | locinput += UTF8SKIP(locinput); | |
4090 | } | |
4091 | } | |
4092 | } | |
4093 | if (locinput > PL_regeol) sayNO; | |
4094 | } | |
a0ed51b3 LW |
4095 | nextchr = UCHARAT(locinput); |
4096 | break; | |
81714fb9 YO |
4097 | |
4098 | case NREFFL: | |
d7ef4b73 KW |
4099 | { /* The capture buffer cases. The ones beginning with N for the |
4100 | named buffers just convert to the equivalent numbered and | |
4101 | pretend they were called as the corresponding numbered buffer | |
4102 | op. */ | |
26ecd678 TC |
4103 | /* don't initialize these in the declaration, it makes C++ |
4104 | unhappy */ | |
81714fb9 | 4105 | char *s; |
ff1157ca | 4106 | char type; |
8368298a TC |
4107 | re_fold_t folder; |
4108 | const U8 *fold_array; | |
26ecd678 | 4109 | UV utf8_fold_flags; |
8368298a | 4110 | |
81714fb9 | 4111 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4112 | folder = foldEQ_locale; |
4113 | fold_array = PL_fold_locale; | |
4114 | type = REFFL; | |
17580e7a | 4115 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4116 | goto do_nref; |
4117 | ||
2f7f8cb1 KW |
4118 | case NREFFA: |
4119 | folder = foldEQ_latin1; | |
4120 | fold_array = PL_fold_latin1; | |
4121 | type = REFFA; | |
4122 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4123 | goto do_nref; | |
4124 | ||
d7ef4b73 KW |
4125 | case NREFFU: |
4126 | folder = foldEQ_latin1; | |
4127 | fold_array = PL_fold_latin1; | |
4128 | type = REFFU; | |
d513472c | 4129 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4130 | goto do_nref; |
4131 | ||
81714fb9 | 4132 | case NREFF: |
d7ef4b73 KW |
4133 | folder = foldEQ; |
4134 | fold_array = PL_fold; | |
4135 | type = REFF; | |
d513472c | 4136 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4137 | goto do_nref; |
4138 | ||
4139 | case NREF: | |
4140 | type = REF; | |
83d7b90b KW |
4141 | folder = NULL; |
4142 | fold_array = NULL; | |
d513472c | 4143 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4144 | do_nref: |
4145 | ||
4146 | /* For the named back references, find the corresponding buffer | |
4147 | * number */ | |
0a4db386 YO |
4148 | n = reg_check_named_buff_matched(rex,scan); |
4149 | ||
d7ef4b73 | 4150 | if ( ! n ) { |
81714fb9 | 4151 | sayNO; |
d7ef4b73 KW |
4152 | } |
4153 | goto do_nref_ref_common; | |
4154 | ||
c8756f30 | 4155 | case REFFL: |
3280af22 | 4156 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4157 | folder = foldEQ_locale; |
4158 | fold_array = PL_fold_locale; | |
17580e7a | 4159 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4160 | goto do_ref; |
4161 | ||
2f7f8cb1 KW |
4162 | case REFFA: |
4163 | folder = foldEQ_latin1; | |
4164 | fold_array = PL_fold_latin1; | |
4165 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4166 | goto do_ref; | |
4167 | ||
d7ef4b73 KW |
4168 | case REFFU: |
4169 | folder = foldEQ_latin1; | |
4170 | fold_array = PL_fold_latin1; | |
d513472c | 4171 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4172 | goto do_ref; |
4173 | ||
4174 | case REFF: | |
4175 | folder = foldEQ; | |
4176 | fold_array = PL_fold; | |
d513472c | 4177 | utf8_fold_flags = 0; |
83d7b90b | 4178 | goto do_ref; |
d7ef4b73 | 4179 | |
c277df42 | 4180 | case REF: |
83d7b90b KW |
4181 | folder = NULL; |
4182 | fold_array = NULL; | |
d513472c | 4183 | utf8_fold_flags = 0; |
83d7b90b | 4184 | |
d7ef4b73 | 4185 | do_ref: |
81714fb9 | 4186 | type = OP(scan); |
d7ef4b73 KW |
4187 | n = ARG(scan); /* which paren pair */ |
4188 | ||
4189 | do_nref_ref_common: | |
b93070ed | 4190 | ln = rex->offs[n].start; |
2c2d71f5 | 4191 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
b93070ed | 4192 | if (rex->lastparen < n || ln == -1) |
af3f8c16 | 4193 | sayNO; /* Do not match unless seen CLOSEn. */ |
b93070ed | 4194 | if (ln == rex->offs[n].end) |
a0d0e21e | 4195 | break; |
a0ed51b3 | 4196 | |
24d3c4a9 | 4197 | s = PL_bostr + ln; |
d7ef4b73 | 4198 | if (type != REF /* REF can do byte comparison */ |
2f65c56d | 4199 | && (utf8_target || type == REFFU)) |
d7ef4b73 KW |
4200 | { /* XXX handle REFFL better */ |
4201 | char * limit = PL_regeol; | |
4202 | ||
4203 | /* This call case insensitively compares the entire buffer | |
4204 | * at s, with the current input starting at locinput, but | |
4205 | * not going off the end given by PL_regeol, and returns in | |
4206 | * limit upon success, how much of the current input was | |
4207 | * matched */ | |
b93070ed | 4208 | if (! foldEQ_utf8_flags(s, NULL, rex->offs[n].end - ln, utf8_target, |
d513472c | 4209 | locinput, &limit, 0, utf8_target, utf8_fold_flags)) |
d7ef4b73 KW |
4210 | { |
4211 | sayNO; | |
a0ed51b3 | 4212 | } |
d7ef4b73 | 4213 | locinput = limit; |
a0ed51b3 LW |
4214 | nextchr = UCHARAT(locinput); |
4215 | break; | |
4216 | } | |
4217 | ||
d7ef4b73 | 4218 | /* Not utf8: Inline the first character, for speed. */ |
76e3520e | 4219 | if (UCHARAT(s) != nextchr && |
81714fb9 | 4220 | (type == REF || |
d7ef4b73 | 4221 | UCHARAT(s) != fold_array[nextchr])) |
4633a7c4 | 4222 | sayNO; |
b93070ed | 4223 | ln = rex->offs[n].end - ln; |
24d3c4a9 | 4224 | if (locinput + ln > PL_regeol) |
4633a7c4 | 4225 | sayNO; |
81714fb9 | 4226 | if (ln > 1 && (type == REF |
24d3c4a9 | 4227 | ? memNE(s, locinput, ln) |
d7ef4b73 | 4228 | : ! folder(s, locinput, ln))) |
4633a7c4 | 4229 | sayNO; |
24d3c4a9 | 4230 | locinput += ln; |
76e3520e | 4231 | nextchr = UCHARAT(locinput); |
a0d0e21e | 4232 | break; |
81714fb9 | 4233 | } |
a0d0e21e | 4234 | case NOTHING: |
c277df42 | 4235 | case TAIL: |
a0d0e21e LW |
4236 | break; |
4237 | case BACK: | |
4238 | break; | |
40a82448 DM |
4239 | |
4240 | #undef ST | |
4241 | #define ST st->u.eval | |
c277df42 | 4242 | { |
c277df42 | 4243 | SV *ret; |
d2f13c59 | 4244 | REGEXP *re_sv; |
6bda09f9 | 4245 | regexp *re; |
f8fc2ecf | 4246 | regexp_internal *rei; |
1a147d38 YO |
4247 | regnode *startpoint; |
4248 | ||
4249 | case GOSTART: | |
e7707071 YO |
4250 | case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */ |
4251 | if (cur_eval && cur_eval->locinput==locinput) { | |
24b23f37 | 4252 | if (cur_eval->u.eval.close_paren == (U32)ARG(scan)) |
1a147d38 | 4253 | Perl_croak(aTHX_ "Infinite recursion in regex"); |
4b196cd4 | 4254 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 YO |
4255 | Perl_croak(aTHX_ |
4256 | "Pattern subroutine nesting without pos change" | |
4257 | " exceeded limit in regex"); | |
6bda09f9 YO |
4258 | } else { |
4259 | nochange_depth = 0; | |
1a147d38 | 4260 | } |
288b8c02 | 4261 | re_sv = rex_sv; |
6bda09f9 | 4262 | re = rex; |
f8fc2ecf | 4263 | rei = rexi; |
1a147d38 | 4264 | if (OP(scan)==GOSUB) { |
6bda09f9 YO |
4265 | startpoint = scan + ARG2L(scan); |
4266 | ST.close_paren = ARG(scan); | |
4267 | } else { | |
f8fc2ecf | 4268 | startpoint = rei->program+1; |
6bda09f9 YO |
4269 | ST.close_paren = 0; |
4270 | } | |
4271 | goto eval_recurse_doit; | |
118e2215 | 4272 | assert(0); /* NOTREACHED */ |
6bda09f9 YO |
4273 | case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ |
4274 | if (cur_eval && cur_eval->locinput==locinput) { | |
4b196cd4 | 4275 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 | 4276 | Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex"); |
6bda09f9 YO |
4277 | } else { |
4278 | nochange_depth = 0; | |
4279 | } | |
8e5e9ebe | 4280 | { |
4aabdb9b | 4281 | /* execute the code in the {...} */ |
81ed78b2 | 4282 | |
4aabdb9b | 4283 | dSP; |
81ed78b2 | 4284 | SV ** before; |
1f4d1a1e | 4285 | OP * const oop = PL_op; |
4aabdb9b | 4286 | COP * const ocurcop = PL_curcop; |
81ed78b2 | 4287 | OP *nop; |
d80618d2 | 4288 | char *saved_regeol = PL_regeol; |
91332126 | 4289 | struct re_save_state saved_state; |
81ed78b2 | 4290 | CV *newcv; |
91332126 | 4291 | |
74088413 DM |
4292 | /* save *all* paren positions */ |
4293 | regcppush(rex, 0); | |
4294 | REGCP_SET(runops_cp); | |
4295 | ||
6562f1c4 | 4296 | /* To not corrupt the existing regex state while executing the |
b7f4cd04 FR |
4297 | * eval we would normally put it on the save stack, like with |
4298 | * save_re_context. However, re-evals have a weird scoping so we | |
4299 | * can't just add ENTER/LEAVE here. With that, things like | |
4300 | * | |
4301 | * (?{$a=2})(a(?{local$a=$a+1}))*aak*c(?{$b=$a}) | |
4302 | * | |
4303 | * would break, as they expect the localisation to be unwound | |
4304 | * only when the re-engine backtracks through the bit that | |
4305 | * localised it. | |
4306 | * | |
4307 | * What we do instead is just saving the state in a local c | |
4308 | * variable. | |
4309 | */ | |
91332126 | 4310 | Copy(&PL_reg_state, &saved_state, 1, struct re_save_state); |
81ed78b2 | 4311 | |
d24ca0c5 | 4312 | PL_reg_state.re_reparsing = FALSE; |
91332126 | 4313 | |
81ed78b2 DM |
4314 | if (!caller_cv) |
4315 | caller_cv = find_runcv(NULL); | |
4316 | ||
4aabdb9b | 4317 | n = ARG(scan); |
81ed78b2 | 4318 | |
b30fcab9 | 4319 | if (rexi->data->what[n] == 'r') { /* code from an external qr */ |
81ed78b2 | 4320 | newcv = ((struct regexp *)SvANY( |
b30fcab9 DM |
4321 | (REGEXP*)(rexi->data->data[n]) |
4322 | ))->qr_anoncv | |
81ed78b2 DM |
4323 | ; |
4324 | nop = (OP*)rexi->data->data[n+1]; | |
b30fcab9 DM |
4325 | } |
4326 | else if (rexi->data->what[n] == 'l') { /* literal code */ | |
81ed78b2 DM |
4327 | newcv = caller_cv; |
4328 | nop = (OP*)rexi->data->data[n]; | |
4329 | assert(CvDEPTH(newcv)); | |
68e2671b DM |
4330 | } |
4331 | else { | |
d24ca0c5 DM |
4332 | /* literal with own CV */ |
4333 | assert(rexi->data->what[n] == 'L'); | |
81ed78b2 DM |
4334 | newcv = rex->qr_anoncv; |
4335 | nop = (OP*)rexi->data->data[n]; | |
68e2671b | 4336 | } |
81ed78b2 | 4337 | |
0e458318 DM |
4338 | /* normally if we're about to execute code from the same |
4339 | * CV that we used previously, we just use the existing | |
4340 | * CX stack entry. However, its possible that in the | |
4341 | * meantime we may have backtracked, popped from the save | |
4342 | * stack, and undone the SAVECOMPPAD(s) associated with | |
4343 | * PUSH_MULTICALL; in which case PL_comppad no longer | |
4344 | * points to newcv's pad. */ | |
4345 | if (newcv != last_pushed_cv || PL_comppad != last_pad) | |
4346 | { | |
4347 | I32 depth = (newcv == caller_cv) ? 0 : 1; | |
4348 | if (last_pushed_cv) { | |
4349 | CHANGE_MULTICALL_WITHDEPTH(newcv, depth); | |
4350 | } | |
4351 | else { | |
4352 | PUSH_MULTICALL_WITHDEPTH(newcv, depth); | |
4353 | } | |
4354 | last_pushed_cv = newcv; | |
4355 | } | |
4356 | last_pad = PL_comppad; | |
4357 | ||
2e2e3f36 DM |
4358 | /* the initial nextstate you would normally execute |
4359 | * at the start of an eval (which would cause error | |
4360 | * messages to come from the eval), may be optimised | |
4361 | * away from the execution path in the regex code blocks; | |
4362 | * so manually set PL_curcop to it initially */ | |
4363 | { | |
81ed78b2 | 4364 | OP *o = cUNOPx(nop)->op_first; |
2e2e3f36 DM |
4365 | assert(o->op_type == OP_NULL); |
4366 | if (o->op_targ == OP_SCOPE) { | |
4367 | o = cUNOPo->op_first; | |
4368 | } | |
4369 | else { | |
4370 | assert(o->op_targ == OP_LEAVE); | |
4371 | o = cUNOPo->op_first; | |
4372 | assert(o->op_type == OP_ENTER); | |
4373 | o = o->op_sibling; | |
4374 | } | |
4375 | ||
4376 | if (o->op_type != OP_STUB) { | |
4377 | assert( o->op_type == OP_NEXTSTATE | |
4378 | || o->op_type == OP_DBSTATE | |
4379 | || (o->op_type == OP_NULL | |
4380 | && ( o->op_targ == OP_NEXTSTATE | |
4381 | || o->op_targ == OP_DBSTATE | |
4382 | ) | |
4383 | ) | |
4384 | ); | |
4385 | PL_curcop = (COP*)o; | |
4386 | } | |
4387 | } | |
81ed78b2 | 4388 | nop = nop->op_next; |
2e2e3f36 | 4389 | |
24b23f37 | 4390 | DEBUG_STATE_r( PerlIO_printf(Perl_debug_log, |
81ed78b2 DM |
4391 | " re EVAL PL_op=0x%"UVxf"\n", PTR2UV(nop)) ); |
4392 | ||
b93070ed | 4393 | rex->offs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr; |
4aabdb9b | 4394 | |
2bf803e2 YO |
4395 | if (sv_yes_mark) { |
4396 | SV *sv_mrk = get_sv("REGMARK", 1); | |
4397 | sv_setsv(sv_mrk, sv_yes_mark); | |
4398 | } | |
4399 | ||
81ed78b2 DM |
4400 | /* we don't use MULTICALL here as we want to call the |
4401 | * first op of the block of interest, rather than the | |
4402 | * first op of the sub */ | |
4403 | before = SP; | |
4404 | PL_op = nop; | |
8e5e9ebe RGS |
4405 | CALLRUNOPS(aTHX); /* Scalar context. */ |
4406 | SPAGAIN; | |
4407 | if (SP == before) | |
075aa684 | 4408 | ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */ |
8e5e9ebe RGS |
4409 | else { |
4410 | ret = POPs; | |
4411 | PUTBACK; | |
4412 | } | |
4aabdb9b | 4413 | |
e4bfbed3 DM |
4414 | /* before restoring everything, evaluate the returned |
4415 | * value, so that 'uninit' warnings don't use the wrong | |
497d0a96 DM |
4416 | * PL_op or pad. Also need to process any magic vars |
4417 | * (e.g. $1) *before* parentheses are restored */ | |
e4bfbed3 DM |
4418 | |
4419 | PL_op = NULL; | |
4420 | ||
5e98dac2 | 4421 | re_sv = NULL; |
e4bfbed3 DM |
4422 | if (logical == 0) /* (?{})/ */ |
4423 | sv_setsv(save_scalar(PL_replgv), ret); /* $^R */ | |
4424 | else if (logical == 1) { /* /(?(?{...})X|Y)/ */ | |
4425 | sw = cBOOL(SvTRUE(ret)); | |
4426 | logical = 0; | |
4427 | } | |
4428 | else { /* /(??{}) */ | |
497d0a96 DM |
4429 | /* if its overloaded, let the regex compiler handle |
4430 | * it; otherwise extract regex, or stringify */ | |
4431 | if (!SvAMAGIC(ret)) { | |
4432 | SV *sv = ret; | |
4433 | if (SvROK(sv)) | |
4434 | sv = SvRV(sv); | |
4435 | if (SvTYPE(sv) == SVt_REGEXP) | |
4436 | re_sv = (REGEXP*) sv; | |
4437 | else if (SvSMAGICAL(sv)) { | |
4438 | MAGIC *mg = mg_find(sv, PERL_MAGIC_qr); | |
4439 | if (mg) | |
4440 | re_sv = (REGEXP *) mg->mg_obj; | |
4441 | } | |
e4bfbed3 | 4442 | |
497d0a96 DM |
4443 | /* force any magic, undef warnings here */ |
4444 | if (!re_sv) { | |
4445 | ret = sv_mortalcopy(ret); | |
4446 | (void) SvPV_force_nolen(ret); | |
4447 | } | |
e4bfbed3 DM |
4448 | } |
4449 | ||
4450 | } | |
4451 | ||
91332126 FR |
4452 | Copy(&saved_state, &PL_reg_state, 1, struct re_save_state); |
4453 | ||
81ed78b2 DM |
4454 | /* *** Note that at this point we don't restore |
4455 | * PL_comppad, (or pop the CxSUB) on the assumption it may | |
4456 | * be used again soon. This is safe as long as nothing | |
4457 | * in the regexp code uses the pad ! */ | |
4aabdb9b | 4458 | PL_op = oop; |
4aabdb9b | 4459 | PL_curcop = ocurcop; |
d80618d2 | 4460 | PL_regeol = saved_regeol; |
e4bfbed3 DM |
4461 | S_regcp_restore(aTHX_ rex, runops_cp); |
4462 | ||
4463 | if (logical != 2) | |
4aabdb9b | 4464 | break; |
8e5e9ebe | 4465 | } |
e4bfbed3 DM |
4466 | |
4467 | /* only /(??{})/ from now on */ | |
24d3c4a9 | 4468 | logical = 0; |
4aabdb9b | 4469 | { |
4f639d21 DM |
4470 | /* extract RE object from returned value; compiling if |
4471 | * necessary */ | |
5c35adbb | 4472 | |
575c37f6 DM |
4473 | if (re_sv) { |
4474 | re_sv = reg_temp_copy(NULL, re_sv); | |
288b8c02 | 4475 | } |
0f5d15d6 | 4476 | else { |
c737faaf | 4477 | U32 pm_flags = 0; |
a3b680e6 | 4478 | const I32 osize = PL_regsize; |
0f5d15d6 | 4479 | |
9753d940 DM |
4480 | if (SvUTF8(ret) && IN_BYTES) { |
4481 | /* In use 'bytes': make a copy of the octet | |
4482 | * sequence, but without the flag on */ | |
b9ad30b4 NC |
4483 | STRLEN len; |
4484 | const char *const p = SvPV(ret, len); | |
4485 | ret = newSVpvn_flags(p, len, SVs_TEMP); | |
4486 | } | |
732caac7 DM |
4487 | if (rex->intflags & PREGf_USE_RE_EVAL) |
4488 | pm_flags |= PMf_USE_RE_EVAL; | |
4489 | ||
4490 | /* if we got here, it should be an engine which | |
4491 | * supports compiling code blocks and stuff */ | |
4492 | assert(rex->engine && rex->engine->op_comp); | |
ec841a27 | 4493 | assert(!(scan->flags & ~RXf_PMf_COMPILETIME)); |
575c37f6 | 4494 | re_sv = rex->engine->op_comp(aTHX_ &ret, 1, NULL, |
ec841a27 DM |
4495 | rex->engine, NULL, NULL, |
4496 | /* copy /msix etc to inner pattern */ | |
4497 | scan->flags, | |
4498 | pm_flags); | |
732caac7 | 4499 | |
9041c2e3 | 4500 | if (!(SvFLAGS(ret) |
faf82a0b | 4501 | & (SVs_TEMP | SVs_PADTMP | SVf_READONLY |
3ce3ed55 | 4502 | | SVs_GMG))) { |
a2794585 NC |
4503 | /* This isn't a first class regexp. Instead, it's |
4504 | caching a regexp onto an existing, Perl visible | |
4505 | scalar. */ | |
575c37f6 | 4506 | sv_magic(ret, MUTABLE_SV(re_sv), PERL_MAGIC_qr, 0, 0); |
3ce3ed55 | 4507 | } |
0f5d15d6 | 4508 | PL_regsize = osize; |
74088413 DM |
4509 | /* safe to do now that any $1 etc has been |
4510 | * interpolated into the new pattern string and | |
4511 | * compiled */ | |
4512 | S_regcp_restore(aTHX_ rex, runops_cp); | |
0f5d15d6 | 4513 | } |
575c37f6 | 4514 | re = (struct regexp *)SvANY(re_sv); |
4aabdb9b | 4515 | } |
07bc277f | 4516 | RXp_MATCH_COPIED_off(re); |
28d8d7f4 YO |
4517 | re->subbeg = rex->subbeg; |
4518 | re->sublen = rex->sublen; | |
f8fc2ecf | 4519 | rei = RXi_GET(re); |
6bda09f9 | 4520 | DEBUG_EXECUTE_r( |
f2ed9b32 | 4521 | debug_start_match(re_sv, utf8_target, locinput, PL_regeol, |
6bda09f9 YO |
4522 | "Matching embedded"); |
4523 | ); | |
f8fc2ecf | 4524 | startpoint = rei->program + 1; |
1a147d38 | 4525 | ST.close_paren = 0; /* only used for GOSUB */ |
aa283a38 | 4526 | |
1a147d38 | 4527 | eval_recurse_doit: /* Share code with GOSUB below this line */ |
aa283a38 | 4528 | /* run the pattern returned from (??{...}) */ |
b93070ed | 4529 | ST.cp = regcppush(rex, 0); /* Save *all* the positions. */ |
40a82448 | 4530 | REGCP_SET(ST.lastcp); |
6bda09f9 | 4531 | |
0357f1fd ML |
4532 | re->lastparen = 0; |
4533 | re->lastcloseparen = 0; | |
4534 | ||
4aabdb9b | 4535 | PL_reginput = locinput; |
ae0beba1 | 4536 | PL_regsize = 0; |
4aabdb9b DM |
4537 | |
4538 | /* XXXX This is too dramatic a measure... */ | |
4539 | PL_reg_maxiter = 0; | |
4540 | ||
faec1544 | 4541 | ST.toggle_reg_flags = PL_reg_flags; |
3c8556c3 | 4542 | if (RX_UTF8(re_sv)) |
faec1544 DM |
4543 | PL_reg_flags |= RF_utf8; |
4544 | else | |
4545 | PL_reg_flags &= ~RF_utf8; | |
4546 | ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */ | |
4547 | ||
288b8c02 | 4548 | ST.prev_rex = rex_sv; |
faec1544 | 4549 | ST.prev_curlyx = cur_curlyx; |
ec43f78b DM |
4550 | rex_sv = re_sv; |
4551 | SET_reg_curpm(rex_sv); | |
288b8c02 | 4552 | rex = re; |
f8fc2ecf | 4553 | rexi = rei; |
faec1544 | 4554 | cur_curlyx = NULL; |
40a82448 | 4555 | ST.B = next; |
faec1544 DM |
4556 | ST.prev_eval = cur_eval; |
4557 | cur_eval = st; | |
faec1544 | 4558 | /* now continue from first node in postoned RE */ |
6bda09f9 | 4559 | PUSH_YES_STATE_GOTO(EVAL_AB, startpoint); |
118e2215 | 4560 | assert(0); /* NOTREACHED */ |
c277df42 | 4561 | } |
40a82448 | 4562 | |
faec1544 DM |
4563 | case EVAL_AB: /* cleanup after a successful (??{A})B */ |
4564 | /* note: this is called twice; first after popping B, then A */ | |
4565 | PL_reg_flags ^= ST.toggle_reg_flags; | |
ec43f78b DM |
4566 | rex_sv = ST.prev_rex; |
4567 | SET_reg_curpm(rex_sv); | |
288b8c02 | 4568 | rex = (struct regexp *)SvANY(rex_sv); |
f8fc2ecf | 4569 | rexi = RXi_GET(rex); |
faec1544 DM |
4570 | regcpblow(ST.cp); |
4571 | cur_eval = ST.prev_eval; | |
4572 | cur_curlyx = ST.prev_curlyx; | |
34a81e2b | 4573 | |
40a82448 DM |
4574 | /* XXXX This is too dramatic a measure... */ |
4575 | PL_reg_maxiter = 0; | |
e7707071 | 4576 | if ( nochange_depth ) |
4b196cd4 | 4577 | nochange_depth--; |
262b90c4 | 4578 | sayYES; |
40a82448 | 4579 | |
40a82448 | 4580 | |
faec1544 DM |
4581 | case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */ |
4582 | /* note: this is called twice; first after popping B, then A */ | |
4583 | PL_reg_flags ^= ST.toggle_reg_flags; | |
ec43f78b DM |
4584 | rex_sv = ST.prev_rex; |
4585 | SET_reg_curpm(rex_sv); | |
288b8c02 | 4586 | rex = (struct regexp *)SvANY(rex_sv); |
f8fc2ecf | 4587 | rexi = RXi_GET(rex); |
0357f1fd | 4588 | |
40a82448 DM |
4589 | PL_reginput = locinput; |
4590 | REGCP_UNWIND(ST.lastcp); | |
4591 | regcppop(rex); | |
faec1544 DM |
4592 | cur_eval = ST.prev_eval; |
4593 | cur_curlyx = ST.prev_curlyx; | |
4594 | /* XXXX This is too dramatic a measure... */ | |
4595 | PL_reg_maxiter = 0; | |
e7707071 | 4596 | if ( nochange_depth ) |
4b196cd4 | 4597 | nochange_depth--; |
40a82448 | 4598 | sayNO_SILENT; |
40a82448 DM |
4599 | #undef ST |
4600 | ||
a0d0e21e | 4601 | case OPEN: |
c277df42 | 4602 | n = ARG(scan); /* which paren pair */ |
1ca2007e | 4603 | rex->offs[n].start_tmp = locinput - PL_bostr; |
3280af22 NIS |
4604 | if (n > PL_regsize) |
4605 | PL_regsize = n; | |
495f47a5 DM |
4606 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
4607 | "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf" tmp; regsize=%"UVuf"\n", | |
4608 | PTR2UV(rex), | |
4609 | PTR2UV(rex->offs), | |
4610 | (UV)n, | |
4611 | (IV)rex->offs[n].start_tmp, | |
4612 | (UV)PL_regsize | |
4613 | )); | |
e2e6a0f1 | 4614 | lastopen = n; |
a0d0e21e | 4615 | break; |
495f47a5 DM |
4616 | |
4617 | /* XXX really need to log other places start/end are set too */ | |
4618 | #define CLOSE_CAPTURE \ | |
4619 | rex->offs[n].start = rex->offs[n].start_tmp; \ | |
4620 | rex->offs[n].end = locinput - PL_bostr; \ | |
4621 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, \ | |
4622 | "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf"..%"IVdf"\n", \ | |
4623 | PTR2UV(rex), \ | |
4624 | PTR2UV(rex->offs), \ | |
4625 | (UV)n, \ | |
4626 | (IV)rex->offs[n].start, \ | |
4627 | (IV)rex->offs[n].end \ | |
4628 | )) | |
4629 | ||
a0d0e21e | 4630 | case CLOSE: |
c277df42 | 4631 | n = ARG(scan); /* which paren pair */ |
495f47a5 | 4632 | CLOSE_CAPTURE; |
7f69552c YO |
4633 | /*if (n > PL_regsize) |
4634 | PL_regsize = n;*/ | |
b93070ed DM |
4635 | if (n > rex->lastparen) |
4636 | rex->lastparen = n; | |
4637 | rex->lastcloseparen = n; | |
3b6647e0 | 4638 | if (cur_eval && cur_eval->u.eval.close_paren == n) { |
6bda09f9 YO |
4639 | goto fake_end; |
4640 | } | |
a0d0e21e | 4641 | break; |
e2e6a0f1 YO |
4642 | case ACCEPT: |
4643 | if (ARG(scan)){ | |
4644 | regnode *cursor; | |
4645 | for (cursor=scan; | |
4646 | cursor && OP(cursor)!=END; | |
4647 | cursor=regnext(cursor)) | |
4648 | { | |
4649 | if ( OP(cursor)==CLOSE ){ | |
4650 | n = ARG(cursor); | |
4651 | if ( n <= lastopen ) { | |
495f47a5 | 4652 | CLOSE_CAPTURE; |
e2e6a0f1 YO |
4653 | /*if (n > PL_regsize) |
4654 | PL_regsize = n;*/ | |
b93070ed DM |
4655 | if (n > rex->lastparen) |
4656 | rex->lastparen = n; | |
4657 | rex->lastcloseparen = n; | |
3b6647e0 RB |
4658 | if ( n == ARG(scan) || (cur_eval && |
4659 | cur_eval->u.eval.close_paren == n)) | |
e2e6a0f1 YO |
4660 | break; |
4661 | } | |
4662 | } | |
4663 | } | |
4664 | } | |
4665 | goto fake_end; | |
4666 | /*NOTREACHED*/ | |
c277df42 IZ |
4667 | case GROUPP: |
4668 | n = ARG(scan); /* which paren pair */ | |
b93070ed | 4669 | sw = cBOOL(rex->lastparen >= n && rex->offs[n].end != -1); |
c277df42 | 4670 | break; |
0a4db386 YO |
4671 | case NGROUPP: |
4672 | /* reg_check_named_buff_matched returns 0 for no match */ | |
f2338a2e | 4673 | sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan)); |
0a4db386 | 4674 | break; |
1a147d38 | 4675 | case INSUBP: |
0a4db386 | 4676 | n = ARG(scan); |
3b6647e0 | 4677 | sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n)); |
0a4db386 YO |
4678 | break; |
4679 | case DEFINEP: | |
4680 | sw = 0; | |
4681 | break; | |
c277df42 | 4682 | case IFTHEN: |
2c2d71f5 | 4683 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
24d3c4a9 | 4684 | if (sw) |
c277df42 IZ |
4685 | next = NEXTOPER(NEXTOPER(scan)); |
4686 | else { | |
4687 | next = scan + ARG(scan); | |
4688 | if (OP(next) == IFTHEN) /* Fake one. */ | |
4689 | next = NEXTOPER(NEXTOPER(next)); | |
4690 | } | |
4691 | break; | |
4692 | case LOGICAL: | |
24d3c4a9 | 4693 | logical = scan->flags; |
c277df42 | 4694 | break; |
c476f425 | 4695 | |
2ab05381 | 4696 | /******************************************************************* |
2ab05381 | 4697 | |
c476f425 DM |
4698 | The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/ |
4699 | pattern, where A and B are subpatterns. (For simple A, CURLYM or | |
4700 | STAR/PLUS/CURLY/CURLYN are used instead.) | |
2ab05381 | 4701 | |
c476f425 | 4702 | A*B is compiled as <CURLYX><A><WHILEM><B> |
2ab05381 | 4703 | |
c476f425 DM |
4704 | On entry to the subpattern, CURLYX is called. This pushes a CURLYX |
4705 | state, which contains the current count, initialised to -1. It also sets | |
4706 | cur_curlyx to point to this state, with any previous value saved in the | |
4707 | state block. | |
2ab05381 | 4708 | |
c476f425 DM |
4709 | CURLYX then jumps straight to the WHILEM op, rather than executing A, |
4710 | since the pattern may possibly match zero times (i.e. it's a while {} loop | |
4711 | rather than a do {} while loop). | |
2ab05381 | 4712 | |
c476f425 DM |
4713 | Each entry to WHILEM represents a successful match of A. The count in the |
4714 | CURLYX block is incremented, another WHILEM state is pushed, and execution | |
4715 | passes to A or B depending on greediness and the current count. | |
2ab05381 | 4716 | |
c476f425 DM |
4717 | For example, if matching against the string a1a2a3b (where the aN are |
4718 | substrings that match /A/), then the match progresses as follows: (the | |
4719 | pushed states are interspersed with the bits of strings matched so far): | |
2ab05381 | 4720 | |
c476f425 DM |
4721 | <CURLYX cnt=-1> |
4722 | <CURLYX cnt=0><WHILEM> | |
4723 | <CURLYX cnt=1><WHILEM> a1 <WHILEM> | |
4724 | <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM> | |
4725 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> | |
4726 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b | |
2ab05381 | 4727 | |
c476f425 DM |
4728 | (Contrast this with something like CURLYM, which maintains only a single |
4729 | backtrack state: | |
2ab05381 | 4730 | |
c476f425 DM |
4731 | <CURLYM cnt=0> a1 |
4732 | a1 <CURLYM cnt=1> a2 | |
4733 | a1 a2 <CURLYM cnt=2> a3 | |
4734 | a1 a2 a3 <CURLYM cnt=3> b | |
4735 | ) | |
2ab05381 | 4736 | |
c476f425 DM |
4737 | Each WHILEM state block marks a point to backtrack to upon partial failure |
4738 | of A or B, and also contains some minor state data related to that | |
4739 | iteration. The CURLYX block, pointed to by cur_curlyx, contains the | |
4740 | overall state, such as the count, and pointers to the A and B ops. | |
2ab05381 | 4741 | |
c476f425 DM |
4742 | This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx |
4743 | must always point to the *current* CURLYX block, the rules are: | |
2ab05381 | 4744 | |
c476f425 DM |
4745 | When executing CURLYX, save the old cur_curlyx in the CURLYX state block, |
4746 | and set cur_curlyx to point the new block. | |
2ab05381 | 4747 | |
c476f425 DM |
4748 | When popping the CURLYX block after a successful or unsuccessful match, |
4749 | restore the previous cur_curlyx. | |
2ab05381 | 4750 | |
c476f425 DM |
4751 | When WHILEM is about to execute B, save the current cur_curlyx, and set it |
4752 | to the outer one saved in the CURLYX block. | |
2ab05381 | 4753 | |
c476f425 DM |
4754 | When popping the WHILEM block after a successful or unsuccessful B match, |
4755 | restore the previous cur_curlyx. | |
2ab05381 | 4756 | |
c476f425 DM |
4757 | Here's an example for the pattern (AI* BI)*BO |
4758 | I and O refer to inner and outer, C and W refer to CURLYX and WHILEM: | |
2ab05381 | 4759 | |
c476f425 DM |
4760 | cur_ |
4761 | curlyx backtrack stack | |
4762 | ------ --------------- | |
4763 | NULL | |
4764 | CO <CO prev=NULL> <WO> | |
4765 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4766 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi | |
4767 | NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo | |
2ab05381 | 4768 | |
c476f425 DM |
4769 | At this point the pattern succeeds, and we work back down the stack to |
4770 | clean up, restoring as we go: | |
95b24440 | 4771 | |
c476f425 DM |
4772 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi |
4773 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4774 | CO <CO prev=NULL> <WO> | |
4775 | NULL | |
a0374537 | 4776 | |
c476f425 DM |
4777 | *******************************************************************/ |
4778 | ||
4779 | #define ST st->u.curlyx | |
4780 | ||
4781 | case CURLYX: /* start of /A*B/ (for complex A) */ | |
4782 | { | |
4783 | /* No need to save/restore up to this paren */ | |
4784 | I32 parenfloor = scan->flags; | |
4785 | ||
4786 | assert(next); /* keep Coverity happy */ | |
4787 | if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ | |
4788 | next += ARG(next); | |
4789 | ||
4790 | /* XXXX Probably it is better to teach regpush to support | |
4791 | parenfloor > PL_regsize... */ | |
b93070ed DM |
4792 | if (parenfloor > (I32)rex->lastparen) |
4793 | parenfloor = rex->lastparen; /* Pessimization... */ | |
c476f425 DM |
4794 | |
4795 | ST.prev_curlyx= cur_curlyx; | |
4796 | cur_curlyx = st; | |
4797 | ST.cp = PL_savestack_ix; | |
4798 | ||
4799 | /* these fields contain the state of the current curly. | |
4800 | * they are accessed by subsequent WHILEMs */ | |
4801 | ST.parenfloor = parenfloor; | |
d02d6d97 | 4802 | ST.me = scan; |
c476f425 | 4803 | ST.B = next; |
24d3c4a9 DM |
4804 | ST.minmod = minmod; |
4805 | minmod = 0; | |
c476f425 DM |
4806 | ST.count = -1; /* this will be updated by WHILEM */ |
4807 | ST.lastloc = NULL; /* this will be updated by WHILEM */ | |
4808 | ||
4809 | PL_reginput = locinput; | |
4810 | PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next)); | |
118e2215 | 4811 | assert(0); /* NOTREACHED */ |
c476f425 | 4812 | } |
a0d0e21e | 4813 | |
c476f425 | 4814 | case CURLYX_end: /* just finished matching all of A*B */ |
c476f425 DM |
4815 | cur_curlyx = ST.prev_curlyx; |
4816 | sayYES; | |
118e2215 | 4817 | assert(0); /* NOTREACHED */ |
a0d0e21e | 4818 | |
c476f425 DM |
4819 | case CURLYX_end_fail: /* just failed to match all of A*B */ |
4820 | regcpblow(ST.cp); | |
4821 | cur_curlyx = ST.prev_curlyx; | |
4822 | sayNO; | |
118e2215 | 4823 | assert(0); /* NOTREACHED */ |
4633a7c4 | 4824 | |
a0d0e21e | 4825 | |
c476f425 DM |
4826 | #undef ST |
4827 | #define ST st->u.whilem | |
4828 | ||
4829 | case WHILEM: /* just matched an A in /A*B/ (for complex A) */ | |
4830 | { | |
4831 | /* see the discussion above about CURLYX/WHILEM */ | |
c476f425 | 4832 | I32 n; |
d02d6d97 DM |
4833 | int min = ARG1(cur_curlyx->u.curlyx.me); |
4834 | int max = ARG2(cur_curlyx->u.curlyx.me); | |
4835 | regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS; | |
4836 | ||
c476f425 DM |
4837 | assert(cur_curlyx); /* keep Coverity happy */ |
4838 | n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */ | |
4839 | ST.save_lastloc = cur_curlyx->u.curlyx.lastloc; | |
4840 | ST.cache_offset = 0; | |
4841 | ST.cache_mask = 0; | |
4842 | ||
4843 | PL_reginput = locinput; | |
4844 | ||
4845 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
d02d6d97 DM |
4846 | "%*s whilem: matched %ld out of %d..%d\n", |
4847 | REPORT_CODE_OFF+depth*2, "", (long)n, min, max) | |
c476f425 | 4848 | ); |
a0d0e21e | 4849 | |
c476f425 | 4850 | /* First just match a string of min A's. */ |
a0d0e21e | 4851 | |
d02d6d97 | 4852 | if (n < min) { |
b93070ed | 4853 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor); |
c476f425 | 4854 | cur_curlyx->u.curlyx.lastloc = locinput; |
92e82afa YO |
4855 | REGCP_SET(ST.lastcp); |
4856 | ||
d02d6d97 | 4857 | PUSH_STATE_GOTO(WHILEM_A_pre, A); |
118e2215 | 4858 | assert(0); /* NOTREACHED */ |
c476f425 DM |
4859 | } |
4860 | ||
4861 | /* If degenerate A matches "", assume A done. */ | |
4862 | ||
4863 | if (locinput == cur_curlyx->u.curlyx.lastloc) { | |
4864 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4865 | "%*s whilem: empty match detected, trying continuation...\n", | |
4866 | REPORT_CODE_OFF+depth*2, "") | |
4867 | ); | |
4868 | goto do_whilem_B_max; | |
4869 | } | |
4870 | ||
4871 | /* super-linear cache processing */ | |
4872 | ||
4873 | if (scan->flags) { | |
a0d0e21e | 4874 | |
2c2d71f5 | 4875 | if (!PL_reg_maxiter) { |
c476f425 DM |
4876 | /* start the countdown: Postpone detection until we |
4877 | * know the match is not *that* much linear. */ | |
2c2d71f5 | 4878 | PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4); |
66bf836d DM |
4879 | /* possible overflow for long strings and many CURLYX's */ |
4880 | if (PL_reg_maxiter < 0) | |
4881 | PL_reg_maxiter = I32_MAX; | |
2c2d71f5 JH |
4882 | PL_reg_leftiter = PL_reg_maxiter; |
4883 | } | |
c476f425 | 4884 | |
2c2d71f5 | 4885 | if (PL_reg_leftiter-- == 0) { |
c476f425 | 4886 | /* initialise cache */ |
3298f257 | 4887 | const I32 size = (PL_reg_maxiter + 7)/8; |
2c2d71f5 | 4888 | if (PL_reg_poscache) { |
eb160463 | 4889 | if ((I32)PL_reg_poscache_size < size) { |
2c2d71f5 JH |
4890 | Renew(PL_reg_poscache, size, char); |
4891 | PL_reg_poscache_size = size; | |
4892 | } | |
4893 | Zero(PL_reg_poscache, size, char); | |
4894 | } | |
4895 | else { | |
4896 | PL_reg_poscache_size = size; | |
a02a5408 | 4897 | Newxz(PL_reg_poscache, size, char); |
2c2d71f5 | 4898 | } |
c476f425 DM |
4899 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
4900 | "%swhilem: Detected a super-linear match, switching on caching%s...\n", | |
4901 | PL_colors[4], PL_colors[5]) | |
4902 | ); | |
2c2d71f5 | 4903 | } |
c476f425 | 4904 | |
2c2d71f5 | 4905 | if (PL_reg_leftiter < 0) { |
c476f425 DM |
4906 | /* have we already failed at this position? */ |
4907 | I32 offset, mask; | |
4908 | offset = (scan->flags & 0xf) - 1 | |
4909 | + (locinput - PL_bostr) * (scan->flags>>4); | |
4910 | mask = 1 << (offset % 8); | |
4911 | offset /= 8; | |
4912 | if (PL_reg_poscache[offset] & mask) { | |
4913 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4914 | "%*s whilem: (cache) already tried at this position...\n", | |
4915 | REPORT_CODE_OFF+depth*2, "") | |
2c2d71f5 | 4916 | ); |
3298f257 | 4917 | sayNO; /* cache records failure */ |
2c2d71f5 | 4918 | } |
c476f425 DM |
4919 | ST.cache_offset = offset; |
4920 | ST.cache_mask = mask; | |
2c2d71f5 | 4921 | } |
c476f425 | 4922 | } |
2c2d71f5 | 4923 | |
c476f425 | 4924 | /* Prefer B over A for minimal matching. */ |
a687059c | 4925 | |
c476f425 DM |
4926 | if (cur_curlyx->u.curlyx.minmod) { |
4927 | ST.save_curlyx = cur_curlyx; | |
4928 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
b93070ed | 4929 | ST.cp = regcppush(rex, ST.save_curlyx->u.curlyx.parenfloor); |
c476f425 DM |
4930 | REGCP_SET(ST.lastcp); |
4931 | PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B); | |
118e2215 | 4932 | assert(0); /* NOTREACHED */ |
c476f425 | 4933 | } |
a0d0e21e | 4934 | |
c476f425 DM |
4935 | /* Prefer A over B for maximal matching. */ |
4936 | ||
d02d6d97 | 4937 | if (n < max) { /* More greed allowed? */ |
b93070ed | 4938 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor); |
c476f425 DM |
4939 | cur_curlyx->u.curlyx.lastloc = locinput; |
4940 | REGCP_SET(ST.lastcp); | |
d02d6d97 | 4941 | PUSH_STATE_GOTO(WHILEM_A_max, A); |
118e2215 | 4942 | assert(0); /* NOTREACHED */ |
c476f425 DM |
4943 | } |
4944 | goto do_whilem_B_max; | |
4945 | } | |
118e2215 | 4946 | assert(0); /* NOTREACHED */ |
c476f425 DM |
4947 | |
4948 | case WHILEM_B_min: /* just matched B in a minimal match */ | |
4949 | case WHILEM_B_max: /* just matched B in a maximal match */ | |
4950 | cur_curlyx = ST.save_curlyx; | |
4951 | sayYES; | |
118e2215 | 4952 | assert(0); /* NOTREACHED */ |
c476f425 DM |
4953 | |
4954 | case WHILEM_B_max_fail: /* just failed to match B in a maximal match */ | |
4955 | cur_curlyx = ST.save_curlyx; | |
4956 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
4957 | cur_curlyx->u.curlyx.count--; | |
4958 | CACHEsayNO; | |
118e2215 | 4959 | assert(0); /* NOTREACHED */ |
c476f425 DM |
4960 | |
4961 | case WHILEM_A_min_fail: /* just failed to match A in a minimal match */ | |
c476f425 DM |
4962 | /* FALL THROUGH */ |
4963 | case WHILEM_A_pre_fail: /* just failed to match even minimal A */ | |
92e82afa YO |
4964 | REGCP_UNWIND(ST.lastcp); |
4965 | regcppop(rex); | |
c476f425 DM |
4966 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; |
4967 | cur_curlyx->u.curlyx.count--; | |
4968 | CACHEsayNO; | |
118e2215 | 4969 | assert(0); /* NOTREACHED */ |
c476f425 DM |
4970 | |
4971 | case WHILEM_A_max_fail: /* just failed to match A in a maximal match */ | |
4972 | REGCP_UNWIND(ST.lastcp); | |
4973 | regcppop(rex); /* Restore some previous $<digit>s? */ | |
4974 | PL_reginput = locinput; | |
4975 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
4976 | "%*s whilem: failed, trying continuation...\n", | |
4977 | REPORT_CODE_OFF+depth*2, "") | |
4978 | ); | |
4979 | do_whilem_B_max: | |
4980 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
4981 | && ckWARN(WARN_REGEXP) | |
4982 | && !(PL_reg_flags & RF_warned)) | |
4983 | { | |
4984 | PL_reg_flags |= RF_warned; | |
dcbac5bb FC |
4985 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
4986 | "Complex regular subexpression recursion limit (%d) " | |
4987 | "exceeded", | |
c476f425 DM |
4988 | REG_INFTY - 1); |
4989 | } | |
4990 | ||
4991 | /* now try B */ | |
4992 | ST.save_curlyx = cur_curlyx; | |
4993 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4994 | PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B); | |
118e2215 | 4995 | assert(0); /* NOTREACHED */ |
c476f425 DM |
4996 | |
4997 | case WHILEM_B_min_fail: /* just failed to match B in a minimal match */ | |
4998 | cur_curlyx = ST.save_curlyx; | |
4999 | REGCP_UNWIND(ST.lastcp); | |
5000 | regcppop(rex); | |
5001 | ||
d02d6d97 | 5002 | if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) { |
c476f425 DM |
5003 | /* Maximum greed exceeded */ |
5004 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
5005 | && ckWARN(WARN_REGEXP) | |
5006 | && !(PL_reg_flags & RF_warned)) | |
5007 | { | |
3280af22 | 5008 | PL_reg_flags |= RF_warned; |
c476f425 | 5009 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
dcbac5bb FC |
5010 | "Complex regular subexpression recursion " |
5011 | "limit (%d) exceeded", | |
c476f425 | 5012 | REG_INFTY - 1); |
a0d0e21e | 5013 | } |
c476f425 | 5014 | cur_curlyx->u.curlyx.count--; |
3ab3c9b4 | 5015 | CACHEsayNO; |
a0d0e21e | 5016 | } |
c476f425 DM |
5017 | |
5018 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
5019 | "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "") | |
5020 | ); | |
5021 | /* Try grabbing another A and see if it helps. */ | |
5022 | PL_reginput = locinput; | |
5023 | cur_curlyx->u.curlyx.lastloc = locinput; | |
b93070ed | 5024 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor); |
c476f425 | 5025 | REGCP_SET(ST.lastcp); |
d02d6d97 DM |
5026 | PUSH_STATE_GOTO(WHILEM_A_min, |
5027 | /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS); | |
118e2215 | 5028 | assert(0); /* NOTREACHED */ |
40a82448 DM |
5029 | |
5030 | #undef ST | |
5031 | #define ST st->u.branch | |
5032 | ||
5033 | case BRANCHJ: /* /(...|A|...)/ with long next pointer */ | |
c277df42 IZ |
5034 | next = scan + ARG(scan); |
5035 | if (next == scan) | |
5036 | next = NULL; | |
40a82448 DM |
5037 | scan = NEXTOPER(scan); |
5038 | /* FALL THROUGH */ | |
c277df42 | 5039 | |
40a82448 DM |
5040 | case BRANCH: /* /(...|A|...)/ */ |
5041 | scan = NEXTOPER(scan); /* scan now points to inner node */ | |
b93070ed | 5042 | ST.lastparen = rex->lastparen; |
f6033a9d | 5043 | ST.lastcloseparen = rex->lastcloseparen; |
40a82448 DM |
5044 | ST.next_branch = next; |
5045 | REGCP_SET(ST.cp); | |
5046 | PL_reginput = locinput; | |
02db2b7b | 5047 | |
40a82448 | 5048 | /* Now go into the branch */ |
5d458dd8 YO |
5049 | if (has_cutgroup) { |
5050 | PUSH_YES_STATE_GOTO(BRANCH_next, scan); | |
5051 | } else { | |
5052 | PUSH_STATE_GOTO(BRANCH_next, scan); | |
5053 | } | |
118e2215 | 5054 | assert(0); /* NOTREACHED */ |
5d458dd8 YO |
5055 | case CUTGROUP: |
5056 | PL_reginput = locinput; | |
5057 | sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : | |
ad64d0ec | 5058 | MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 | 5059 | PUSH_STATE_GOTO(CUTGROUP_next,next); |
118e2215 | 5060 | assert(0); /* NOTREACHED */ |
5d458dd8 YO |
5061 | case CUTGROUP_next_fail: |
5062 | do_cutgroup = 1; | |
5063 | no_final = 1; | |
5064 | if (st->u.mark.mark_name) | |
5065 | sv_commit = st->u.mark.mark_name; | |
5066 | sayNO; | |
118e2215 | 5067 | assert(0); /* NOTREACHED */ |
5d458dd8 YO |
5068 | case BRANCH_next: |
5069 | sayYES; | |
118e2215 | 5070 | assert(0); /* NOTREACHED */ |
40a82448 | 5071 | case BRANCH_next_fail: /* that branch failed; try the next, if any */ |
5d458dd8 YO |
5072 | if (do_cutgroup) { |
5073 | do_cutgroup = 0; | |
5074 | no_final = 0; | |
5075 | } | |
40a82448 | 5076 | REGCP_UNWIND(ST.cp); |
a8d1f4b4 | 5077 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
40a82448 DM |
5078 | scan = ST.next_branch; |
5079 | /* no more branches? */ | |
5d458dd8 YO |
5080 | if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) { |
5081 | DEBUG_EXECUTE_r({ | |
5082 | PerlIO_printf( Perl_debug_log, | |
5083 | "%*s %sBRANCH failed...%s\n", | |
5084 | REPORT_CODE_OFF+depth*2, "", | |
5085 | PL_colors[4], | |
5086 | PL_colors[5] ); | |
5087 | }); | |
5088 | sayNO_SILENT; | |
5089 | } | |
40a82448 | 5090 | continue; /* execute next BRANCH[J] op */ |
118e2215 | 5091 | assert(0); /* NOTREACHED */ |
40a82448 | 5092 | |
a0d0e21e | 5093 | case MINMOD: |
24d3c4a9 | 5094 | minmod = 1; |
a0d0e21e | 5095 | break; |
40a82448 DM |
5096 | |
5097 | #undef ST | |
5098 | #define ST st->u.curlym | |
5099 | ||
5100 | case CURLYM: /* /A{m,n}B/ where A is fixed-length */ | |
5101 | ||
5102 | /* This is an optimisation of CURLYX that enables us to push | |
84d2fa14 | 5103 | * only a single backtracking state, no matter how many matches |
40a82448 DM |
5104 | * there are in {m,n}. It relies on the pattern being constant |
5105 | * length, with no parens to influence future backrefs | |
5106 | */ | |
5107 | ||
5108 | ST.me = scan; | |
dc45a647 | 5109 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; |
40a82448 | 5110 | |
f6033a9d DM |
5111 | ST.lastparen = rex->lastparen; |
5112 | ST.lastcloseparen = rex->lastcloseparen; | |
5113 | ||
40a82448 DM |
5114 | /* if paren positive, emulate an OPEN/CLOSE around A */ |
5115 | if (ST.me->flags) { | |
3b6647e0 | 5116 | U32 paren = ST.me->flags; |
40a82448 DM |
5117 | if (paren > PL_regsize) |
5118 | PL_regsize = paren; | |
c277df42 | 5119 | scan += NEXT_OFF(scan); /* Skip former OPEN. */ |
6407bf3b | 5120 | } |
40a82448 DM |
5121 | ST.A = scan; |
5122 | ST.B = next; | |
5123 | ST.alen = 0; | |
5124 | ST.count = 0; | |
24d3c4a9 DM |
5125 | ST.minmod = minmod; |
5126 | minmod = 0; | |
40a82448 DM |
5127 | ST.c1 = CHRTEST_UNINIT; |
5128 | REGCP_SET(ST.cp); | |
6407bf3b | 5129 | |
40a82448 DM |
5130 | if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */ |
5131 | goto curlym_do_B; | |
5132 | ||
5133 | curlym_do_A: /* execute the A in /A{m,n}B/ */ | |
6407bf3b | 5134 | PL_reginput = locinput; |
40a82448 | 5135 | PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */ |
118e2215 | 5136 | assert(0); /* NOTREACHED */ |
5f80c4cf | 5137 | |
40a82448 DM |
5138 | case CURLYM_A: /* we've just matched an A */ |
5139 | locinput = st->locinput; | |
5140 | nextchr = UCHARAT(locinput); | |
5141 | ||
5142 | ST.count++; | |
5143 | /* after first match, determine A's length: u.curlym.alen */ | |
5144 | if (ST.count == 1) { | |
5145 | if (PL_reg_match_utf8) { | |
5146 | char *s = locinput; | |
5147 | while (s < PL_reginput) { | |
5148 | ST.alen++; | |
5149 | s += UTF8SKIP(s); | |
5150 | } | |
5151 | } | |
5152 | else { | |
5153 | ST.alen = PL_reginput - locinput; | |
5154 | } | |
5155 | if (ST.alen == 0) | |
5156 | ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me); | |
5157 | } | |
0cadcf80 DM |
5158 | DEBUG_EXECUTE_r( |
5159 | PerlIO_printf(Perl_debug_log, | |
40a82448 | 5160 | "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n", |
5bc10b2c | 5161 | (int)(REPORT_CODE_OFF+(depth*2)), "", |
40a82448 | 5162 | (IV) ST.count, (IV)ST.alen) |
0cadcf80 DM |
5163 | ); |
5164 | ||
40a82448 | 5165 | locinput = PL_reginput; |
0a4db386 YO |
5166 | |
5167 | if (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 5168 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5169 | goto fake_end; |
5170 | ||
c966426a DM |
5171 | { |
5172 | I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)); | |
5173 | if ( max == REG_INFTY || ST.count < max ) | |
5174 | goto curlym_do_A; /* try to match another A */ | |
5175 | } | |
40a82448 | 5176 | goto curlym_do_B; /* try to match B */ |
5f80c4cf | 5177 | |
40a82448 DM |
5178 | case CURLYM_A_fail: /* just failed to match an A */ |
5179 | REGCP_UNWIND(ST.cp); | |
0a4db386 YO |
5180 | |
5181 | if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ | |
5182 | || (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 5183 | cur_eval->u.eval.close_paren == (U32)ST.me->flags)) |
40a82448 | 5184 | sayNO; |
0cadcf80 | 5185 | |
40a82448 DM |
5186 | curlym_do_B: /* execute the B in /A{m,n}B/ */ |
5187 | PL_reginput = locinput; | |
5188 | if (ST.c1 == CHRTEST_UNINIT) { | |
5189 | /* calculate c1 and c2 for possible match of 1st char | |
5190 | * following curly */ | |
5191 | ST.c1 = ST.c2 = CHRTEST_VOID; | |
5192 | if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) { | |
5193 | regnode *text_node = ST.B; | |
5194 | if (! HAS_TEXT(text_node)) | |
5195 | FIND_NEXT_IMPT(text_node); | |
ee9b8eae YO |
5196 | /* this used to be |
5197 | ||
5198 | (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT) | |
5199 | ||
5200 | But the former is redundant in light of the latter. | |
5201 | ||
5202 | if this changes back then the macro for | |
5203 | IS_TEXT and friends need to change. | |
5204 | */ | |
5205 | if (PL_regkind[OP(text_node)] == EXACT) | |
40a82448 | 5206 | { |
ee9b8eae | 5207 | |
40a82448 | 5208 | ST.c1 = (U8)*STRING(text_node); |
9a5a5549 KW |
5209 | switch (OP(text_node)) { |
5210 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
2f7f8cb1 | 5211 | case EXACTFA: |
3c760661 | 5212 | case EXACTFU_SS: |
fab2782b | 5213 | case EXACTFU_TRICKYFOLD: |
9a5a5549 KW |
5214 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; |
5215 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
5216 | default: ST.c2 = ST.c1; | |
5217 | } | |
c277df42 | 5218 | } |
c277df42 | 5219 | } |
40a82448 DM |
5220 | } |
5221 | ||
5222 | DEBUG_EXECUTE_r( | |
5223 | PerlIO_printf(Perl_debug_log, | |
5224 | "%*s CURLYM trying tail with matches=%"IVdf"...\n", | |
5bc10b2c | 5225 | (int)(REPORT_CODE_OFF+(depth*2)), |
40a82448 DM |
5226 | "", (IV)ST.count) |
5227 | ); | |
5228 | if (ST.c1 != CHRTEST_VOID | |
5229 | && UCHARAT(PL_reginput) != ST.c1 | |
5230 | && UCHARAT(PL_reginput) != ST.c2) | |
5231 | { | |
5232 | /* simulate B failing */ | |
3e901dc0 YO |
5233 | DEBUG_OPTIMISE_r( |
5234 | PerlIO_printf(Perl_debug_log, | |
5235 | "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n", | |
5236 | (int)(REPORT_CODE_OFF+(depth*2)),"", | |
5237 | (IV)ST.c1,(IV)ST.c2 | |
5238 | )); | |
40a82448 DM |
5239 | state_num = CURLYM_B_fail; |
5240 | goto reenter_switch; | |
5241 | } | |
5242 | ||
5243 | if (ST.me->flags) { | |
f6033a9d | 5244 | /* emulate CLOSE: mark current A as captured */ |
40a82448 DM |
5245 | I32 paren = ST.me->flags; |
5246 | if (ST.count) { | |
b93070ed | 5247 | rex->offs[paren].start |
40a82448 | 5248 | = HOPc(PL_reginput, -ST.alen) - PL_bostr; |
b93070ed | 5249 | rex->offs[paren].end = PL_reginput - PL_bostr; |
f6033a9d DM |
5250 | if ((U32)paren > rex->lastparen) |
5251 | rex->lastparen = paren; | |
5252 | rex->lastcloseparen = paren; | |
c277df42 | 5253 | } |
40a82448 | 5254 | else |
b93070ed | 5255 | rex->offs[paren].end = -1; |
0a4db386 | 5256 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 5257 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5258 | { |
5259 | if (ST.count) | |
5260 | goto fake_end; | |
5261 | else | |
5262 | sayNO; | |
5263 | } | |
c277df42 | 5264 | } |
0a4db386 | 5265 | |
40a82448 | 5266 | PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */ |
118e2215 | 5267 | assert(0); /* NOTREACHED */ |
40a82448 DM |
5268 | |
5269 | case CURLYM_B_fail: /* just failed to match a B */ | |
5270 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 | 5271 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
40a82448 | 5272 | if (ST.minmod) { |
84d2fa14 HS |
5273 | I32 max = ARG2(ST.me); |
5274 | if (max != REG_INFTY && ST.count == max) | |
40a82448 DM |
5275 | sayNO; |
5276 | goto curlym_do_A; /* try to match a further A */ | |
5277 | } | |
5278 | /* backtrack one A */ | |
5279 | if (ST.count == ARG1(ST.me) /* min */) | |
5280 | sayNO; | |
5281 | ST.count--; | |
5282 | locinput = HOPc(locinput, -ST.alen); | |
5283 | goto curlym_do_B; /* try to match B */ | |
5284 | ||
c255a977 DM |
5285 | #undef ST |
5286 | #define ST st->u.curly | |
40a82448 | 5287 | |
c255a977 DM |
5288 | #define CURLY_SETPAREN(paren, success) \ |
5289 | if (paren) { \ | |
5290 | if (success) { \ | |
b93070ed DM |
5291 | rex->offs[paren].start = HOPc(locinput, -1) - PL_bostr; \ |
5292 | rex->offs[paren].end = locinput - PL_bostr; \ | |
f6033a9d DM |
5293 | if (paren > rex->lastparen) \ |
5294 | rex->lastparen = paren; \ | |
b93070ed | 5295 | rex->lastcloseparen = paren; \ |
c255a977 | 5296 | } \ |
f6033a9d | 5297 | else { \ |
b93070ed | 5298 | rex->offs[paren].end = -1; \ |
f6033a9d DM |
5299 | rex->lastparen = ST.lastparen; \ |
5300 | rex->lastcloseparen = ST.lastcloseparen; \ | |
5301 | } \ | |
c255a977 DM |
5302 | } |
5303 | ||
5304 | case STAR: /* /A*B/ where A is width 1 */ | |
5305 | ST.paren = 0; | |
5306 | ST.min = 0; | |
5307 | ST.max = REG_INFTY; | |
a0d0e21e LW |
5308 | scan = NEXTOPER(scan); |
5309 | goto repeat; | |
c255a977 DM |
5310 | case PLUS: /* /A+B/ where A is width 1 */ |
5311 | ST.paren = 0; | |
5312 | ST.min = 1; | |
5313 | ST.max = REG_INFTY; | |
c277df42 | 5314 | scan = NEXTOPER(scan); |
c255a977 DM |
5315 | goto repeat; |
5316 | case CURLYN: /* /(A){m,n}B/ where A is width 1 */ | |
5317 | ST.paren = scan->flags; /* Which paren to set */ | |
f6033a9d DM |
5318 | ST.lastparen = rex->lastparen; |
5319 | ST.lastcloseparen = rex->lastcloseparen; | |
c255a977 DM |
5320 | if (ST.paren > PL_regsize) |
5321 | PL_regsize = ST.paren; | |
c255a977 DM |
5322 | ST.min = ARG1(scan); /* min to match */ |
5323 | ST.max = ARG2(scan); /* max to match */ | |
0a4db386 | 5324 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5325 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5326 | ST.min=1; |
5327 | ST.max=1; | |
5328 | } | |
c255a977 DM |
5329 | scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); |
5330 | goto repeat; | |
5331 | case CURLY: /* /A{m,n}B/ where A is width 1 */ | |
5332 | ST.paren = 0; | |
5333 | ST.min = ARG1(scan); /* min to match */ | |
5334 | ST.max = ARG2(scan); /* max to match */ | |
5335 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; | |
c277df42 | 5336 | repeat: |
a0d0e21e LW |
5337 | /* |
5338 | * Lookahead to avoid useless match attempts | |
5339 | * when we know what character comes next. | |
c255a977 | 5340 | * |
5f80c4cf JP |
5341 | * Used to only do .*x and .*?x, but now it allows |
5342 | * for )'s, ('s and (?{ ... })'s to be in the way | |
5343 | * of the quantifier and the EXACT-like node. -- japhy | |
5344 | */ | |
5345 | ||
c255a977 DM |
5346 | if (ST.min > ST.max) /* XXX make this a compile-time check? */ |
5347 | sayNO; | |
cca55fe3 | 5348 | if (HAS_TEXT(next) || JUMPABLE(next)) { |
5f80c4cf JP |
5349 | U8 *s; |
5350 | regnode *text_node = next; | |
5351 | ||
3dab1dad YO |
5352 | if (! HAS_TEXT(text_node)) |
5353 | FIND_NEXT_IMPT(text_node); | |
5f80c4cf | 5354 | |
9e137952 | 5355 | if (! HAS_TEXT(text_node)) |
c255a977 | 5356 | ST.c1 = ST.c2 = CHRTEST_VOID; |
5f80c4cf | 5357 | else { |
ee9b8eae | 5358 | if ( PL_regkind[OP(text_node)] != EXACT ) { |
c255a977 | 5359 | ST.c1 = ST.c2 = CHRTEST_VOID; |
44a68960 | 5360 | goto assume_ok_easy; |
cca55fe3 | 5361 | } |
be8e71aa YO |
5362 | else |
5363 | s = (U8*)STRING(text_node); | |
ee9b8eae YO |
5364 | |
5365 | /* Currently we only get here when | |
5366 | ||
5367 | PL_rekind[OP(text_node)] == EXACT | |
5368 | ||
5369 | if this changes back then the macro for IS_TEXT and | |
5370 | friends need to change. */ | |
f2ed9b32 | 5371 | if (!UTF_PATTERN) { |
9a5a5549 KW |
5372 | ST.c1 = *s; |
5373 | switch (OP(text_node)) { | |
5374 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
2f7f8cb1 | 5375 | case EXACTFA: |
3c760661 | 5376 | case EXACTFU_SS: |
fab2782b | 5377 | case EXACTFU_TRICKYFOLD: |
9a5a5549 KW |
5378 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; |
5379 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
5380 | default: ST.c2 = ST.c1; break; | |
5381 | } | |
1aa99e6b | 5382 | } |
f2ed9b32 | 5383 | else { /* UTF_PATTERN */ |
9a5a5549 | 5384 | if (IS_TEXTFU(text_node) || IS_TEXTF(text_node)) { |
b0573d8b KW |
5385 | STRLEN ulen; |
5386 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; | |
5387 | ||
5388 | to_utf8_fold((U8*)s, tmpbuf, &ulen); | |
5389 | ST.c1 = ST.c2 = utf8n_to_uvchr(tmpbuf, UTF8_MAXLEN, 0, | |
e294cc5d | 5390 | uniflags); |
5f80c4cf JP |
5391 | } |
5392 | else { | |
c255a977 | 5393 | ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, |
041457d9 | 5394 | uniflags); |
5f80c4cf | 5395 | } |
1aa99e6b IH |
5396 | } |
5397 | } | |
bbce6d69 | 5398 | } |
a0d0e21e | 5399 | else |
c255a977 | 5400 | ST.c1 = ST.c2 = CHRTEST_VOID; |
cca55fe3 | 5401 | assume_ok_easy: |
c255a977 DM |
5402 | |
5403 | ST.A = scan; | |
5404 | ST.B = next; | |
3280af22 | 5405 | PL_reginput = locinput; |
24d3c4a9 DM |
5406 | if (minmod) { |
5407 | minmod = 0; | |
e2e6a0f1 | 5408 | if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min) |
4633a7c4 | 5409 | sayNO; |
c255a977 | 5410 | ST.count = ST.min; |
a0ed51b3 | 5411 | locinput = PL_reginput; |
c255a977 DM |
5412 | REGCP_SET(ST.cp); |
5413 | if (ST.c1 == CHRTEST_VOID) | |
5414 | goto curly_try_B_min; | |
5415 | ||
5416 | ST.oldloc = locinput; | |
5417 | ||
5418 | /* set ST.maxpos to the furthest point along the | |
5419 | * string that could possibly match */ | |
5420 | if (ST.max == REG_INFTY) { | |
5421 | ST.maxpos = PL_regeol - 1; | |
f2ed9b32 | 5422 | if (utf8_target) |
c255a977 DM |
5423 | while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos)) |
5424 | ST.maxpos--; | |
5425 | } | |
f2ed9b32 | 5426 | else if (utf8_target) { |
c255a977 DM |
5427 | int m = ST.max - ST.min; |
5428 | for (ST.maxpos = locinput; | |
5429 | m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--) | |
5430 | ST.maxpos += UTF8SKIP(ST.maxpos); | |
5431 | } | |
5432 | else { | |
5433 | ST.maxpos = locinput + ST.max - ST.min; | |
5434 | if (ST.maxpos >= PL_regeol) | |
5435 | ST.maxpos = PL_regeol - 1; | |
5436 | } | |
5437 | goto curly_try_B_min_known; | |
5438 | ||
5439 | } | |
5440 | else { | |
e2e6a0f1 | 5441 | ST.count = regrepeat(rex, ST.A, ST.max, depth); |
c255a977 DM |
5442 | locinput = PL_reginput; |
5443 | if (ST.count < ST.min) | |
5444 | sayNO; | |
5445 | if ((ST.count > ST.min) | |
5446 | && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL)) | |
5447 | { | |
5448 | /* A{m,n} must come at the end of the string, there's | |
5449 | * no point in backing off ... */ | |
5450 | ST.min = ST.count; | |
5451 | /* ...except that $ and \Z can match before *and* after | |
5452 | newline at the end. Consider "\n\n" =~ /\n+\Z\n/. | |
5453 | We may back off by one in this case. */ | |
5454 | if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS) | |
5455 | ST.min--; | |
5456 | } | |
5457 | REGCP_SET(ST.cp); | |
5458 | goto curly_try_B_max; | |
5459 | } | |
118e2215 | 5460 | assert(0); /* NOTREACHED */ |
c255a977 DM |
5461 | |
5462 | ||
5463 | case CURLY_B_min_known_fail: | |
5464 | /* failed to find B in a non-greedy match where c1,c2 valid */ | |
c255a977 DM |
5465 | |
5466 | PL_reginput = locinput; /* Could be reset... */ | |
5467 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 DM |
5468 | if (ST.paren) { |
5469 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
5470 | } | |
c255a977 DM |
5471 | /* Couldn't or didn't -- move forward. */ |
5472 | ST.oldloc = locinput; | |
f2ed9b32 | 5473 | if (utf8_target) |
c255a977 DM |
5474 | locinput += UTF8SKIP(locinput); |
5475 | else | |
5476 | locinput++; | |
5477 | ST.count++; | |
5478 | curly_try_B_min_known: | |
5479 | /* find the next place where 'B' could work, then call B */ | |
5480 | { | |
5481 | int n; | |
f2ed9b32 | 5482 | if (utf8_target) { |
c255a977 DM |
5483 | n = (ST.oldloc == locinput) ? 0 : 1; |
5484 | if (ST.c1 == ST.c2) { | |
5485 | STRLEN len; | |
5486 | /* set n to utf8_distance(oldloc, locinput) */ | |
5487 | while (locinput <= ST.maxpos && | |
5488 | utf8n_to_uvchr((U8*)locinput, | |
5489 | UTF8_MAXBYTES, &len, | |
5490 | uniflags) != (UV)ST.c1) { | |
5491 | locinput += len; | |
5492 | n++; | |
5493 | } | |
1aa99e6b IH |
5494 | } |
5495 | else { | |
c255a977 DM |
5496 | /* set n to utf8_distance(oldloc, locinput) */ |
5497 | while (locinput <= ST.maxpos) { | |
5498 | STRLEN len; | |
5499 | const UV c = utf8n_to_uvchr((U8*)locinput, | |
5500 | UTF8_MAXBYTES, &len, | |
5501 | uniflags); | |
5502 | if (c == (UV)ST.c1 || c == (UV)ST.c2) | |
5503 | break; | |
5504 | locinput += len; | |
5505 | n++; | |
1aa99e6b | 5506 | } |
0fe9bf95 IZ |
5507 | } |
5508 | } | |
c255a977 DM |
5509 | else { |
5510 | if (ST.c1 == ST.c2) { | |
5511 | while (locinput <= ST.maxpos && | |
5512 | UCHARAT(locinput) != ST.c1) | |
5513 | locinput++; | |
bbce6d69 | 5514 | } |
c255a977 DM |
5515 | else { |
5516 | while (locinput <= ST.maxpos | |
5517 | && UCHARAT(locinput) != ST.c1 | |
5518 | && UCHARAT(locinput) != ST.c2) | |
5519 | locinput++; | |
a0ed51b3 | 5520 | } |
c255a977 DM |
5521 | n = locinput - ST.oldloc; |
5522 | } | |
5523 | if (locinput > ST.maxpos) | |
5524 | sayNO; | |
5525 | /* PL_reginput == oldloc now */ | |
5526 | if (n) { | |
5527 | ST.count += n; | |
e2e6a0f1 | 5528 | if (regrepeat(rex, ST.A, n, depth) < n) |
4633a7c4 | 5529 | sayNO; |
a0d0e21e | 5530 | } |
c255a977 DM |
5531 | PL_reginput = locinput; |
5532 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5533 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5534 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5535 | goto fake_end; |
5536 | } | |
c255a977 | 5537 | PUSH_STATE_GOTO(CURLY_B_min_known, ST.B); |
a0d0e21e | 5538 | } |
118e2215 | 5539 | assert(0); /* NOTREACHED */ |
c255a977 DM |
5540 | |
5541 | ||
5542 | case CURLY_B_min_fail: | |
5543 | /* failed to find B in a non-greedy match where c1,c2 invalid */ | |
c255a977 DM |
5544 | |
5545 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 DM |
5546 | if (ST.paren) { |
5547 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
5548 | } | |
c255a977 DM |
5549 | /* failed -- move forward one */ |
5550 | PL_reginput = locinput; | |
e2e6a0f1 | 5551 | if (regrepeat(rex, ST.A, 1, depth)) { |
c255a977 | 5552 | ST.count++; |
a0ed51b3 | 5553 | locinput = PL_reginput; |
c255a977 DM |
5554 | if (ST.count <= ST.max || (ST.max == REG_INFTY && |
5555 | ST.count > 0)) /* count overflow ? */ | |
15272685 | 5556 | { |
c255a977 DM |
5557 | curly_try_B_min: |
5558 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5559 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5560 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5561 | goto fake_end; |
5562 | } | |
c255a977 | 5563 | PUSH_STATE_GOTO(CURLY_B_min, ST.B); |
a0d0e21e LW |
5564 | } |
5565 | } | |
4633a7c4 | 5566 | sayNO; |
118e2215 | 5567 | assert(0); /* NOTREACHED */ |
c255a977 DM |
5568 | |
5569 | ||
5570 | curly_try_B_max: | |
5571 | /* a successful greedy match: now try to match B */ | |
40d049e4 | 5572 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5573 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
40d049e4 YO |
5574 | goto fake_end; |
5575 | } | |
c255a977 DM |
5576 | { |
5577 | UV c = 0; | |
5578 | if (ST.c1 != CHRTEST_VOID) | |
f2ed9b32 | 5579 | c = utf8_target ? utf8n_to_uvchr((U8*)PL_reginput, |
c255a977 | 5580 | UTF8_MAXBYTES, 0, uniflags) |
466787eb | 5581 | : (UV) UCHARAT(PL_reginput); |
c255a977 DM |
5582 | /* If it could work, try it. */ |
5583 | if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) { | |
5584 | CURLY_SETPAREN(ST.paren, ST.count); | |
5585 | PUSH_STATE_GOTO(CURLY_B_max, ST.B); | |
118e2215 | 5586 | assert(0); /* NOTREACHED */ |
c255a977 DM |
5587 | } |
5588 | } | |
5589 | /* FALL THROUGH */ | |
5590 | case CURLY_B_max_fail: | |
5591 | /* failed to find B in a greedy match */ | |
c255a977 DM |
5592 | |
5593 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 DM |
5594 | if (ST.paren) { |
5595 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
5596 | } | |
c255a977 DM |
5597 | /* back up. */ |
5598 | if (--ST.count < ST.min) | |
5599 | sayNO; | |
5600 | PL_reginput = locinput = HOPc(locinput, -1); | |
5601 | goto curly_try_B_max; | |
5602 | ||
5603 | #undef ST | |
5604 | ||
a0d0e21e | 5605 | case END: |
6bda09f9 | 5606 | fake_end: |
faec1544 DM |
5607 | if (cur_eval) { |
5608 | /* we've just finished A in /(??{A})B/; now continue with B */ | |
faec1544 DM |
5609 | st->u.eval.toggle_reg_flags |
5610 | = cur_eval->u.eval.toggle_reg_flags; | |
5611 | PL_reg_flags ^= st->u.eval.toggle_reg_flags; | |
5612 | ||
288b8c02 | 5613 | st->u.eval.prev_rex = rex_sv; /* inner */ |
b93070ed | 5614 | st->u.eval.cp = regcppush(rex, 0); /* Save *all* the positions. */ |
ec43f78b DM |
5615 | rex_sv = cur_eval->u.eval.prev_rex; |
5616 | SET_reg_curpm(rex_sv); | |
288b8c02 | 5617 | rex = (struct regexp *)SvANY(rex_sv); |
f8fc2ecf | 5618 | rexi = RXi_GET(rex); |
faec1544 | 5619 | cur_curlyx = cur_eval->u.eval.prev_curlyx; |
34a81e2b | 5620 | |
faec1544 DM |
5621 | REGCP_SET(st->u.eval.lastcp); |
5622 | PL_reginput = locinput; | |
5623 | ||
5624 | /* Restore parens of the outer rex without popping the | |
5625 | * savestack */ | |
74088413 | 5626 | S_regcp_restore(aTHX_ rex, cur_eval->u.eval.lastcp); |
faec1544 DM |
5627 | |
5628 | st->u.eval.prev_eval = cur_eval; | |
5629 | cur_eval = cur_eval->u.eval.prev_eval; | |
5630 | DEBUG_EXECUTE_r( | |
2a49f0f5 JH |
5631 | PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n", |
5632 | REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval));); | |
e7707071 YO |
5633 | if ( nochange_depth ) |
5634 | nochange_depth--; | |
5635 | ||
5636 | PUSH_YES_STATE_GOTO(EVAL_AB, | |
faec1544 DM |
5637 | st->u.eval.prev_eval->u.eval.B); /* match B */ |
5638 | } | |
5639 | ||
3b0527fe | 5640 | if (locinput < reginfo->till) { |
a3621e74 | 5641 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
7821416a IZ |
5642 | "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n", |
5643 | PL_colors[4], | |
5644 | (long)(locinput - PL_reg_starttry), | |
3b0527fe | 5645 | (long)(reginfo->till - PL_reg_starttry), |
7821416a | 5646 | PL_colors[5])); |
58e23c8d | 5647 | |
262b90c4 | 5648 | sayNO_SILENT; /* Cannot match: too short. */ |
7821416a IZ |
5649 | } |
5650 | PL_reginput = locinput; /* put where regtry can find it */ | |
262b90c4 | 5651 | sayYES; /* Success! */ |
dad79028 DM |
5652 | |
5653 | case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */ | |
5654 | DEBUG_EXECUTE_r( | |
5655 | PerlIO_printf(Perl_debug_log, | |
5656 | "%*s %ssubpattern success...%s\n", | |
5bc10b2c | 5657 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])); |
3280af22 | 5658 | PL_reginput = locinput; /* put where regtry can find it */ |
262b90c4 | 5659 | sayYES; /* Success! */ |
dad79028 | 5660 | |
40a82448 DM |
5661 | #undef ST |
5662 | #define ST st->u.ifmatch | |
5663 | ||
5664 | case SUSPEND: /* (?>A) */ | |
5665 | ST.wanted = 1; | |
9fe1d20c | 5666 | PL_reginput = locinput; |
9041c2e3 | 5667 | goto do_ifmatch; |
dad79028 | 5668 | |
40a82448 DM |
5669 | case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */ |
5670 | ST.wanted = 0; | |
dad79028 DM |
5671 | goto ifmatch_trivial_fail_test; |
5672 | ||
40a82448 DM |
5673 | case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */ |
5674 | ST.wanted = 1; | |
dad79028 | 5675 | ifmatch_trivial_fail_test: |
a0ed51b3 | 5676 | if (scan->flags) { |
52657f30 | 5677 | char * const s = HOPBACKc(locinput, scan->flags); |
dad79028 DM |
5678 | if (!s) { |
5679 | /* trivial fail */ | |
24d3c4a9 DM |
5680 | if (logical) { |
5681 | logical = 0; | |
f2338a2e | 5682 | sw = 1 - cBOOL(ST.wanted); |
dad79028 | 5683 | } |
40a82448 | 5684 | else if (ST.wanted) |
dad79028 DM |
5685 | sayNO; |
5686 | next = scan + ARG(scan); | |
5687 | if (next == scan) | |
5688 | next = NULL; | |
5689 | break; | |
5690 | } | |
efb30f32 | 5691 | PL_reginput = s; |
a0ed51b3 LW |
5692 | } |
5693 | else | |
5694 | PL_reginput = locinput; | |
5695 | ||
c277df42 | 5696 | do_ifmatch: |
40a82448 | 5697 | ST.me = scan; |
24d3c4a9 | 5698 | ST.logical = logical; |
24d786f4 YO |
5699 | logical = 0; /* XXX: reset state of logical once it has been saved into ST */ |
5700 | ||
40a82448 DM |
5701 | /* execute body of (?...A) */ |
5702 | PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan))); | |
118e2215 | 5703 | assert(0); /* NOTREACHED */ |
40a82448 DM |
5704 | |
5705 | case IFMATCH_A_fail: /* body of (?...A) failed */ | |
5706 | ST.wanted = !ST.wanted; | |
5707 | /* FALL THROUGH */ | |
5708 | ||
5709 | case IFMATCH_A: /* body of (?...A) succeeded */ | |
24d3c4a9 | 5710 | if (ST.logical) { |
f2338a2e | 5711 | sw = cBOOL(ST.wanted); |
40a82448 DM |
5712 | } |
5713 | else if (!ST.wanted) | |
5714 | sayNO; | |
5715 | ||
5716 | if (OP(ST.me) == SUSPEND) | |
5717 | locinput = PL_reginput; | |
5718 | else { | |
5719 | locinput = PL_reginput = st->locinput; | |
5720 | nextchr = UCHARAT(locinput); | |
5721 | } | |
5722 | scan = ST.me + ARG(ST.me); | |
5723 | if (scan == ST.me) | |
5724 | scan = NULL; | |
5725 | continue; /* execute B */ | |
5726 | ||
5727 | #undef ST | |
dad79028 | 5728 | |
c277df42 | 5729 | case LONGJMP: |
c277df42 IZ |
5730 | next = scan + ARG(scan); |
5731 | if (next == scan) | |
5732 | next = NULL; | |
a0d0e21e | 5733 | break; |
54612592 | 5734 | case COMMIT: |
e2e6a0f1 YO |
5735 | reginfo->cutpoint = PL_regeol; |
5736 | /* FALLTHROUGH */ | |
5d458dd8 | 5737 | case PRUNE: |
24b23f37 | 5738 | PL_reginput = locinput; |
e2e6a0f1 | 5739 | if (!scan->flags) |
ad64d0ec | 5740 | sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
54612592 | 5741 | PUSH_STATE_GOTO(COMMIT_next,next); |
118e2215 | 5742 | assert(0); /* NOTREACHED */ |
54612592 YO |
5743 | case COMMIT_next_fail: |
5744 | no_final = 1; | |
5745 | /* FALLTHROUGH */ | |
7f69552c YO |
5746 | case OPFAIL: |
5747 | sayNO; | |
118e2215 | 5748 | assert(0); /* NOTREACHED */ |
e2e6a0f1 YO |
5749 | |
5750 | #define ST st->u.mark | |
5751 | case MARKPOINT: | |
5752 | ST.prev_mark = mark_state; | |
5d458dd8 | 5753 | ST.mark_name = sv_commit = sv_yes_mark |
ad64d0ec | 5754 | = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
e2e6a0f1 YO |
5755 | mark_state = st; |
5756 | ST.mark_loc = PL_reginput = locinput; | |
5757 | PUSH_YES_STATE_GOTO(MARKPOINT_next,next); | |
118e2215 | 5758 | assert(0); /* NOTREACHED */ |
e2e6a0f1 YO |
5759 | case MARKPOINT_next: |
5760 | mark_state = ST.prev_mark; | |
5761 | sayYES; | |
118e2215 | 5762 | assert(0); /* NOTREACHED */ |
e2e6a0f1 | 5763 | case MARKPOINT_next_fail: |
5d458dd8 | 5764 | if (popmark && sv_eq(ST.mark_name,popmark)) |
e2e6a0f1 YO |
5765 | { |
5766 | if (ST.mark_loc > startpoint) | |
5767 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5768 | popmark = NULL; /* we found our mark */ | |
5769 | sv_commit = ST.mark_name; | |
5770 | ||
5771 | DEBUG_EXECUTE_r({ | |
5d458dd8 | 5772 | PerlIO_printf(Perl_debug_log, |
e2e6a0f1 YO |
5773 | "%*s %ssetting cutpoint to mark:%"SVf"...%s\n", |
5774 | REPORT_CODE_OFF+depth*2, "", | |
be2597df | 5775 | PL_colors[4], SVfARG(sv_commit), PL_colors[5]); |
e2e6a0f1 YO |
5776 | }); |
5777 | } | |
5778 | mark_state = ST.prev_mark; | |
5d458dd8 YO |
5779 | sv_yes_mark = mark_state ? |
5780 | mark_state->u.mark.mark_name : NULL; | |
e2e6a0f1 | 5781 | sayNO; |
118e2215 | 5782 | assert(0); /* NOTREACHED */ |
5d458dd8 YO |
5783 | case SKIP: |
5784 | PL_reginput = locinput; | |
5785 | if (scan->flags) { | |
2bf803e2 | 5786 | /* (*SKIP) : if we fail we cut here*/ |
5d458dd8 | 5787 | ST.mark_name = NULL; |
e2e6a0f1 | 5788 | ST.mark_loc = locinput; |
5d458dd8 YO |
5789 | PUSH_STATE_GOTO(SKIP_next,next); |
5790 | } else { | |
2bf803e2 | 5791 | /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was, |
5d458dd8 YO |
5792 | otherwise do nothing. Meaning we need to scan |
5793 | */ | |
5794 | regmatch_state *cur = mark_state; | |
ad64d0ec | 5795 | SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
5796 | |
5797 | while (cur) { | |
5798 | if ( sv_eq( cur->u.mark.mark_name, | |
5799 | find ) ) | |
5800 | { | |
5801 | ST.mark_name = find; | |
5802 | PUSH_STATE_GOTO( SKIP_next, next ); | |
5803 | } | |
5804 | cur = cur->u.mark.prev_mark; | |
5805 | } | |
e2e6a0f1 | 5806 | } |
2bf803e2 | 5807 | /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */ |
5d458dd8 YO |
5808 | break; |
5809 | case SKIP_next_fail: | |
5810 | if (ST.mark_name) { | |
5811 | /* (*CUT:NAME) - Set up to search for the name as we | |
5812 | collapse the stack*/ | |
5813 | popmark = ST.mark_name; | |
5814 | } else { | |
5815 | /* (*CUT) - No name, we cut here.*/ | |
e2e6a0f1 YO |
5816 | if (ST.mark_loc > startpoint) |
5817 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5d458dd8 YO |
5818 | /* but we set sv_commit to latest mark_name if there |
5819 | is one so they can test to see how things lead to this | |
5820 | cut */ | |
5821 | if (mark_state) | |
5822 | sv_commit=mark_state->u.mark.mark_name; | |
5823 | } | |
e2e6a0f1 YO |
5824 | no_final = 1; |
5825 | sayNO; | |
118e2215 | 5826 | assert(0); /* NOTREACHED */ |
e2e6a0f1 | 5827 | #undef ST |
e1d1eefb | 5828 | case LNBREAK: |
f2ed9b32 | 5829 | if ((n=is_LNBREAK(locinput,utf8_target))) { |
e1d1eefb YO |
5830 | locinput += n; |
5831 | nextchr = UCHARAT(locinput); | |
5832 | } else | |
5833 | sayNO; | |
5834 | break; | |
5835 | ||
5836 | #define CASE_CLASS(nAmE) \ | |
5837 | case nAmE: \ | |
1380b51d DM |
5838 | if (locinput >= PL_regeol) \ |
5839 | sayNO; \ | |
f2ed9b32 | 5840 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5841 | locinput += n; \ |
5842 | nextchr = UCHARAT(locinput); \ | |
5843 | } else \ | |
5844 | sayNO; \ | |
5845 | break; \ | |
5846 | case N##nAmE: \ | |
1380b51d DM |
5847 | if (locinput >= PL_regeol) \ |
5848 | sayNO; \ | |
f2ed9b32 | 5849 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5850 | sayNO; \ |
5851 | } else { \ | |
5852 | locinput += UTF8SKIP(locinput); \ | |
5853 | nextchr = UCHARAT(locinput); \ | |
5854 | } \ | |
5855 | break | |
5856 | ||
5857 | CASE_CLASS(VERTWS); | |
5858 | CASE_CLASS(HORIZWS); | |
5859 | #undef CASE_CLASS | |
5860 | ||
a0d0e21e | 5861 | default: |
b900a521 | 5862 | PerlIO_printf(Perl_error_log, "%"UVxf" %d\n", |
d7d93a81 | 5863 | PTR2UV(scan), OP(scan)); |
cea2e8a9 | 5864 | Perl_croak(aTHX_ "regexp memory corruption"); |
5d458dd8 YO |
5865 | |
5866 | } /* end switch */ | |
95b24440 | 5867 | |
5d458dd8 YO |
5868 | /* switch break jumps here */ |
5869 | scan = next; /* prepare to execute the next op and ... */ | |
5870 | continue; /* ... jump back to the top, reusing st */ | |
118e2215 | 5871 | assert(0); /* NOTREACHED */ |
95b24440 | 5872 | |
40a82448 DM |
5873 | push_yes_state: |
5874 | /* push a state that backtracks on success */ | |
5875 | st->u.yes.prev_yes_state = yes_state; | |
5876 | yes_state = st; | |
5877 | /* FALL THROUGH */ | |
5878 | push_state: | |
5879 | /* push a new regex state, then continue at scan */ | |
5880 | { | |
5881 | regmatch_state *newst; | |
5882 | ||
24b23f37 YO |
5883 | DEBUG_STACK_r({ |
5884 | regmatch_state *cur = st; | |
5885 | regmatch_state *curyes = yes_state; | |
5886 | int curd = depth; | |
5887 | regmatch_slab *slab = PL_regmatch_slab; | |
5888 | for (;curd > -1;cur--,curd--) { | |
5889 | if (cur < SLAB_FIRST(slab)) { | |
5890 | slab = slab->prev; | |
5891 | cur = SLAB_LAST(slab); | |
5892 | } | |
5893 | PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n", | |
5894 | REPORT_CODE_OFF + 2 + depth * 2,"", | |
13d6edb4 | 5895 | curd, PL_reg_name[cur->resume_state], |
24b23f37 YO |
5896 | (curyes == cur) ? "yes" : "" |
5897 | ); | |
5898 | if (curyes == cur) | |
5899 | curyes = cur->u.yes.prev_yes_state; | |
5900 | } | |
5901 | } else | |
5902 | DEBUG_STATE_pp("push") | |
5903 | ); | |
40a82448 | 5904 | depth++; |
40a82448 DM |
5905 | st->locinput = locinput; |
5906 | newst = st+1; | |
5907 | if (newst > SLAB_LAST(PL_regmatch_slab)) | |
5908 | newst = S_push_slab(aTHX); | |
5909 | PL_regmatch_state = newst; | |
786e8c11 | 5910 | |
40a82448 DM |
5911 | locinput = PL_reginput; |
5912 | nextchr = UCHARAT(locinput); | |
5913 | st = newst; | |
5914 | continue; | |
118e2215 | 5915 | assert(0); /* NOTREACHED */ |
40a82448 | 5916 | } |
a0d0e21e | 5917 | } |
a687059c | 5918 | |
a0d0e21e LW |
5919 | /* |
5920 | * We get here only if there's trouble -- normally "case END" is | |
5921 | * the terminating point. | |
5922 | */ | |
cea2e8a9 | 5923 | Perl_croak(aTHX_ "corrupted regexp pointers"); |
a0d0e21e | 5924 | /*NOTREACHED*/ |
4633a7c4 LW |
5925 | sayNO; |
5926 | ||
262b90c4 | 5927 | yes: |
77cb431f DM |
5928 | if (yes_state) { |
5929 | /* we have successfully completed a subexpression, but we must now | |
5930 | * pop to the state marked by yes_state and continue from there */ | |
77cb431f | 5931 | assert(st != yes_state); |
5bc10b2c DM |
5932 | #ifdef DEBUGGING |
5933 | while (st != yes_state) { | |
5934 | st--; | |
5935 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
5936 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5937 | st = SLAB_LAST(PL_regmatch_slab); | |
5938 | } | |
e2e6a0f1 | 5939 | DEBUG_STATE_r({ |
54612592 YO |
5940 | if (no_final) { |
5941 | DEBUG_STATE_pp("pop (no final)"); | |
5942 | } else { | |
5943 | DEBUG_STATE_pp("pop (yes)"); | |
5944 | } | |
e2e6a0f1 | 5945 | }); |
5bc10b2c DM |
5946 | depth--; |
5947 | } | |
5948 | #else | |
77cb431f DM |
5949 | while (yes_state < SLAB_FIRST(PL_regmatch_slab) |
5950 | || yes_state > SLAB_LAST(PL_regmatch_slab)) | |
5951 | { | |
5952 | /* not in this slab, pop slab */ | |
5953 | depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1); | |
5954 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5955 | st = SLAB_LAST(PL_regmatch_slab); | |
5956 | } | |
5957 | depth -= (st - yes_state); | |
5bc10b2c | 5958 | #endif |
77cb431f DM |
5959 | st = yes_state; |
5960 | yes_state = st->u.yes.prev_yes_state; | |
5961 | PL_regmatch_state = st; | |
24b23f37 | 5962 | |
5d458dd8 YO |
5963 | if (no_final) { |
5964 | locinput= st->locinput; | |
5965 | nextchr = UCHARAT(locinput); | |
5966 | } | |
54612592 | 5967 | state_num = st->resume_state + no_final; |
24d3c4a9 | 5968 | goto reenter_switch; |
77cb431f DM |
5969 | } |
5970 | ||
a3621e74 | 5971 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", |
e4584336 | 5972 | PL_colors[4], PL_colors[5])); |
02db2b7b | 5973 | |
ed301438 | 5974 | if (PL_reg_state.re_state_eval_setup_done) { |
19b95bf0 DM |
5975 | /* each successfully executed (?{...}) block does the equivalent of |
5976 | * local $^R = do {...} | |
5977 | * When popping the save stack, all these locals would be undone; | |
5978 | * bypass this by setting the outermost saved $^R to the latest | |
5979 | * value */ | |
5980 | if (oreplsv != GvSV(PL_replgv)) | |
5981 | sv_setsv(oreplsv, GvSV(PL_replgv)); | |
5982 | } | |
95b24440 | 5983 | result = 1; |
aa283a38 | 5984 | goto final_exit; |
4633a7c4 LW |
5985 | |
5986 | no: | |
a3621e74 | 5987 | DEBUG_EXECUTE_r( |
7821416a | 5988 | PerlIO_printf(Perl_debug_log, |
786e8c11 | 5989 | "%*s %sfailed...%s\n", |
5bc10b2c | 5990 | REPORT_CODE_OFF+depth*2, "", |
786e8c11 | 5991 | PL_colors[4], PL_colors[5]) |
7821416a | 5992 | ); |
aa283a38 | 5993 | |
262b90c4 | 5994 | no_silent: |
54612592 YO |
5995 | if (no_final) { |
5996 | if (yes_state) { | |
5997 | goto yes; | |
5998 | } else { | |
5999 | goto final_exit; | |
6000 | } | |
6001 | } | |
aa283a38 DM |
6002 | if (depth) { |
6003 | /* there's a previous state to backtrack to */ | |
40a82448 DM |
6004 | st--; |
6005 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
6006 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
6007 | st = SLAB_LAST(PL_regmatch_slab); | |
6008 | } | |
6009 | PL_regmatch_state = st; | |
40a82448 DM |
6010 | locinput= st->locinput; |
6011 | nextchr = UCHARAT(locinput); | |
6012 | ||
5bc10b2c DM |
6013 | DEBUG_STATE_pp("pop"); |
6014 | depth--; | |
262b90c4 DM |
6015 | if (yes_state == st) |
6016 | yes_state = st->u.yes.prev_yes_state; | |
5bc10b2c | 6017 | |
24d3c4a9 DM |
6018 | state_num = st->resume_state + 1; /* failure = success + 1 */ |
6019 | goto reenter_switch; | |
95b24440 | 6020 | } |
24d3c4a9 | 6021 | result = 0; |
aa283a38 | 6022 | |
262b90c4 | 6023 | final_exit: |
bbe252da | 6024 | if (rex->intflags & PREGf_VERBARG_SEEN) { |
5d458dd8 YO |
6025 | SV *sv_err = get_sv("REGERROR", 1); |
6026 | SV *sv_mrk = get_sv("REGMARK", 1); | |
6027 | if (result) { | |
e2e6a0f1 | 6028 | sv_commit = &PL_sv_no; |
5d458dd8 YO |
6029 | if (!sv_yes_mark) |
6030 | sv_yes_mark = &PL_sv_yes; | |
6031 | } else { | |
6032 | if (!sv_commit) | |
6033 | sv_commit = &PL_sv_yes; | |
6034 | sv_yes_mark = &PL_sv_no; | |
6035 | } | |
6036 | sv_setsv(sv_err, sv_commit); | |
6037 | sv_setsv(sv_mrk, sv_yes_mark); | |
e2e6a0f1 | 6038 | } |
19b95bf0 | 6039 | |
81ed78b2 DM |
6040 | |
6041 | if (last_pushed_cv) { | |
6042 | dSP; | |
6043 | POP_MULTICALL; | |
4f8dbb2d | 6044 | PERL_UNUSED_VAR(SP); |
81ed78b2 DM |
6045 | } |
6046 | ||
2f554ef7 DM |
6047 | /* clean up; in particular, free all slabs above current one */ |
6048 | LEAVE_SCOPE(oldsave); | |
5d9a96ca | 6049 | |
95b24440 | 6050 | return result; |
a687059c LW |
6051 | } |
6052 | ||
6053 | /* | |
6054 | - regrepeat - repeatedly match something simple, report how many | |
6055 | */ | |
6056 | /* | |
6057 | * [This routine now assumes that it will only match on things of length 1. | |
6058 | * That was true before, but now we assume scan - reginput is the count, | |
a0ed51b3 | 6059 | * rather than incrementing count on every character. [Er, except utf8.]] |
a687059c | 6060 | */ |
76e3520e | 6061 | STATIC I32 |
e2e6a0f1 | 6062 | S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth) |
a687059c | 6063 | { |
27da23d5 | 6064 | dVAR; |
a0d0e21e | 6065 | register char *scan; |
a0d0e21e | 6066 | register I32 c; |
3280af22 | 6067 | register char *loceol = PL_regeol; |
a0ed51b3 | 6068 | register I32 hardcount = 0; |
f2ed9b32 | 6069 | register bool utf8_target = PL_reg_match_utf8; |
d513472c | 6070 | UV utf8_flags; |
4f55667c SP |
6071 | #ifndef DEBUGGING |
6072 | PERL_UNUSED_ARG(depth); | |
6073 | #endif | |
a0d0e21e | 6074 | |
7918f24d NC |
6075 | PERL_ARGS_ASSERT_REGREPEAT; |
6076 | ||
3280af22 | 6077 | scan = PL_reginput; |
faf11cac HS |
6078 | if (max == REG_INFTY) |
6079 | max = I32_MAX; | |
6080 | else if (max < loceol - scan) | |
7f596f4c | 6081 | loceol = scan + max; |
a0d0e21e | 6082 | switch (OP(p)) { |
22c35a8c | 6083 | case REG_ANY: |
f2ed9b32 | 6084 | if (utf8_target) { |
ffc61ed2 | 6085 | loceol = PL_regeol; |
1aa99e6b | 6086 | while (scan < loceol && hardcount < max && *scan != '\n') { |
ffc61ed2 JH |
6087 | scan += UTF8SKIP(scan); |
6088 | hardcount++; | |
6089 | } | |
6090 | } else { | |
6091 | while (scan < loceol && *scan != '\n') | |
6092 | scan++; | |
a0ed51b3 LW |
6093 | } |
6094 | break; | |
ffc61ed2 | 6095 | case SANY: |
f2ed9b32 | 6096 | if (utf8_target) { |
def8e4ea | 6097 | loceol = PL_regeol; |
a0804c9e | 6098 | while (scan < loceol && hardcount < max) { |
def8e4ea JH |
6099 | scan += UTF8SKIP(scan); |
6100 | hardcount++; | |
6101 | } | |
6102 | } | |
6103 | else | |
6104 | scan = loceol; | |
a0ed51b3 | 6105 | break; |
f33976b4 DB |
6106 | case CANY: |
6107 | scan = loceol; | |
6108 | break; | |
59d32103 KW |
6109 | case EXACT: |
6110 | /* To get here, EXACTish nodes must have *byte* length == 1. That | |
6111 | * means they match only characters in the string that can be expressed | |
6112 | * as a single byte. For non-utf8 strings, that means a simple match. | |
6113 | * For utf8 strings, the character matched must be an invariant, or | |
6114 | * downgradable to a single byte. The pattern's utf8ness is | |
6115 | * irrelevant, as since it's a single byte, it either isn't utf8, or if | |
6116 | * it is, it's an invariant */ | |
6117 | ||
6118 | c = (U8)*STRING(p); | |
6119 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); | |
6120 | ||
6121 | if (! utf8_target || UNI_IS_INVARIANT(c)) { | |
6122 | while (scan < loceol && UCHARAT(scan) == c) { | |
6123 | scan++; | |
6124 | } | |
6125 | } | |
6126 | else { | |
6127 | ||
6128 | /* Here, the string is utf8, and the pattern char is different | |
6129 | * in utf8 than not, so can't compare them directly. Outside the | |
5908807f | 6130 | * loop, find the two utf8 bytes that represent c, and then |
59d32103 KW |
6131 | * look for those in sequence in the utf8 string */ |
6132 | U8 high = UTF8_TWO_BYTE_HI(c); | |
6133 | U8 low = UTF8_TWO_BYTE_LO(c); | |
6134 | loceol = PL_regeol; | |
6135 | ||
6136 | while (hardcount < max | |
6137 | && scan + 1 < loceol | |
6138 | && UCHARAT(scan) == high | |
6139 | && UCHARAT(scan + 1) == low) | |
6140 | { | |
6141 | scan += 2; | |
6142 | hardcount++; | |
6143 | } | |
6144 | } | |
6145 | break; | |
2f7f8cb1 KW |
6146 | case EXACTFA: |
6147 | utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
6148 | goto do_exactf; | |
6149 | ||
d4e0b827 KW |
6150 | case EXACTFL: |
6151 | PL_reg_flags |= RF_tainted; | |
17580e7a KW |
6152 | utf8_flags = FOLDEQ_UTF8_LOCALE; |
6153 | goto do_exactf; | |
6154 | ||
d4e0b827 | 6155 | case EXACTF: |
62bf7766 KW |
6156 | utf8_flags = 0; |
6157 | goto do_exactf; | |
6158 | ||
3c760661 | 6159 | case EXACTFU_SS: |
fab2782b | 6160 | case EXACTFU_TRICKYFOLD: |
9a5a5549 | 6161 | case EXACTFU: |
05f861a2 | 6162 | utf8_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0; |
59d32103 | 6163 | |
2be3e190 KW |
6164 | /* The comments for the EXACT case above apply as well to these fold |
6165 | * ones */ | |
634c83a2 | 6166 | |
2f7f8cb1 | 6167 | do_exactf: |
090f7165 | 6168 | c = (U8)*STRING(p); |
634c83a2 | 6169 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); |
d4e0b827 | 6170 | |
3c760661 | 6171 | if (utf8_target || OP(p) == EXACTFU_SS) { /* Use full Unicode fold matching */ |
59d32103 KW |
6172 | char *tmpeol = loceol; |
6173 | while (hardcount < max | |
d513472c KW |
6174 | && foldEQ_utf8_flags(scan, &tmpeol, 0, utf8_target, |
6175 | STRING(p), NULL, 1, cBOOL(UTF_PATTERN), utf8_flags)) | |
59d32103 KW |
6176 | { |
6177 | scan = tmpeol; | |
6178 | tmpeol = loceol; | |
6179 | hardcount++; | |
6180 | } | |
634c83a2 | 6181 | |
59d32103 KW |
6182 | /* XXX Note that the above handles properly the German sharp s in |
6183 | * the pattern matching ss in the string. But it doesn't handle | |
6184 | * properly cases where the string contains say 'LIGATURE ff' and | |
6185 | * the pattern is 'f+'. This would require, say, a new function or | |
6186 | * revised interface to foldEQ_utf8(), in which the maximum number | |
6187 | * of characters to match could be passed and it would return how | |
6188 | * many actually did. This is just one of many cases where | |
6189 | * multi-char folds don't work properly, and so the fix is being | |
6190 | * deferred */ | |
6191 | } | |
6192 | else { | |
87381386 | 6193 | U8 folded; |
59d32103 | 6194 | |
2be3e190 KW |
6195 | /* Here, the string isn't utf8 and c is a single byte; and either |
6196 | * the pattern isn't utf8 or c is an invariant, so its utf8ness | |
6197 | * doesn't affect c. Can just do simple comparisons for exact or | |
6198 | * fold matching. */ | |
d4e0b827 | 6199 | switch (OP(p)) { |
87381386 | 6200 | case EXACTF: folded = PL_fold[c]; break; |
2f7f8cb1 | 6201 | case EXACTFA: |
fab2782b | 6202 | case EXACTFU_TRICKYFOLD: |
9a5a5549 | 6203 | case EXACTFU: folded = PL_fold_latin1[c]; break; |
87381386 KW |
6204 | case EXACTFL: folded = PL_fold_locale[c]; break; |
6205 | default: Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p)); | |
6206 | } | |
6207 | while (scan < loceol && | |
6208 | (UCHARAT(scan) == c || UCHARAT(scan) == folded)) | |
6209 | { | |
6210 | scan++; | |
634c83a2 KW |
6211 | } |
6212 | } | |
bbce6d69 | 6213 | break; |
f56b6394 | 6214 | case ANYOFV: |
a0d0e21e | 6215 | case ANYOF: |
4e8910e0 KW |
6216 | if (utf8_target || OP(p) == ANYOFV) { |
6217 | STRLEN inclasslen; | |
ffc61ed2 | 6218 | loceol = PL_regeol; |
4e8910e0 KW |
6219 | inclasslen = loceol - scan; |
6220 | while (hardcount < max | |
6221 | && ((inclasslen = loceol - scan) > 0) | |
6222 | && reginclass(prog, p, (U8*)scan, &inclasslen, utf8_target)) | |
6223 | { | |
6224 | scan += inclasslen; | |
ffc61ed2 JH |
6225 | hardcount++; |
6226 | } | |
6227 | } else { | |
32fc9b6a | 6228 | while (scan < loceol && REGINCLASS(prog, p, (U8*)scan)) |
ffc61ed2 JH |
6229 | scan++; |
6230 | } | |
a0d0e21e | 6231 | break; |
980866de | 6232 | case ALNUMU: |
f2ed9b32 | 6233 | if (utf8_target) { |
980866de | 6234 | utf8_wordchar: |
ffc61ed2 | 6235 | loceol = PL_regeol; |
1a4fad37 | 6236 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6237 | while (hardcount < max && scan < loceol && |
a12cf05f KW |
6238 | swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
6239 | { | |
ffc61ed2 JH |
6240 | scan += UTF8SKIP(scan); |
6241 | hardcount++; | |
6242 | } | |
980866de | 6243 | } else { |
a12cf05f KW |
6244 | while (scan < loceol && isWORDCHAR_L1((U8) *scan)) { |
6245 | scan++; | |
6246 | } | |
980866de KW |
6247 | } |
6248 | break; | |
6249 | case ALNUM: | |
6250 | if (utf8_target) | |
6251 | goto utf8_wordchar; | |
6252 | while (scan < loceol && isALNUM((U8) *scan)) { | |
6253 | scan++; | |
a0ed51b3 LW |
6254 | } |
6255 | break; | |
cfaf538b KW |
6256 | case ALNUMA: |
6257 | while (scan < loceol && isWORDCHAR_A((U8) *scan)) { | |
6258 | scan++; | |
6259 | } | |
6260 | break; | |
bbce6d69 | 6261 | case ALNUML: |
3280af22 | 6262 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6263 | if (utf8_target) { |
ffc61ed2 | 6264 | loceol = PL_regeol; |
1aa99e6b IH |
6265 | while (hardcount < max && scan < loceol && |
6266 | isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6267 | scan += UTF8SKIP(scan); |
6268 | hardcount++; | |
6269 | } | |
6270 | } else { | |
6271 | while (scan < loceol && isALNUM_LC(*scan)) | |
6272 | scan++; | |
a0ed51b3 LW |
6273 | } |
6274 | break; | |
980866de | 6275 | case NALNUMU: |
f2ed9b32 | 6276 | if (utf8_target) { |
980866de KW |
6277 | |
6278 | utf8_Nwordchar: | |
6279 | ||
ffc61ed2 | 6280 | loceol = PL_regeol; |
1a4fad37 | 6281 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6282 | while (hardcount < max && scan < loceol && |
980866de | 6283 | ! swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
a12cf05f | 6284 | { |
ffc61ed2 JH |
6285 | scan += UTF8SKIP(scan); |
6286 | hardcount++; | |
6287 | } | |
980866de | 6288 | } else { |
a12cf05f KW |
6289 | while (scan < loceol && ! isWORDCHAR_L1((U8) *scan)) { |
6290 | scan++; | |
6291 | } | |
980866de KW |
6292 | } |
6293 | break; | |
6294 | case NALNUM: | |
6295 | if (utf8_target) | |
6296 | goto utf8_Nwordchar; | |
6297 | while (scan < loceol && ! isALNUM((U8) *scan)) { | |
6298 | scan++; | |
a0ed51b3 LW |
6299 | } |
6300 | break; | |
cfaf538b KW |
6301 | case NALNUMA: |
6302 | if (utf8_target) { | |
6303 | while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) { | |
6304 | scan += UTF8SKIP(scan); | |
6305 | } | |
6306 | } | |
6307 | else { | |
6308 | while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) { | |
6309 | scan++; | |
6310 | } | |
6311 | } | |
6312 | break; | |
bbce6d69 | 6313 | case NALNUML: |
3280af22 | 6314 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6315 | if (utf8_target) { |
ffc61ed2 | 6316 | loceol = PL_regeol; |
1aa99e6b IH |
6317 | while (hardcount < max && scan < loceol && |
6318 | !isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6319 | scan += UTF8SKIP(scan); |
6320 | hardcount++; | |
6321 | } | |
6322 | } else { | |
6323 | while (scan < loceol && !isALNUM_LC(*scan)) | |
6324 | scan++; | |
a0ed51b3 LW |
6325 | } |
6326 | break; | |
980866de | 6327 | case SPACEU: |
f2ed9b32 | 6328 | if (utf8_target) { |
980866de KW |
6329 | |
6330 | utf8_space: | |
6331 | ||
ffc61ed2 | 6332 | loceol = PL_regeol; |
1a4fad37 | 6333 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6334 | while (hardcount < max && scan < loceol && |
3568d838 | 6335 | (*scan == ' ' || |
a12cf05f KW |
6336 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) |
6337 | { | |
ffc61ed2 JH |
6338 | scan += UTF8SKIP(scan); |
6339 | hardcount++; | |
6340 | } | |
980866de KW |
6341 | break; |
6342 | } | |
6343 | else { | |
a12cf05f KW |
6344 | while (scan < loceol && isSPACE_L1((U8) *scan)) { |
6345 | scan++; | |
6346 | } | |
980866de KW |
6347 | break; |
6348 | } | |
6349 | case SPACE: | |
6350 | if (utf8_target) | |
6351 | goto utf8_space; | |
6352 | ||
6353 | while (scan < loceol && isSPACE((U8) *scan)) { | |
6354 | scan++; | |
a0ed51b3 LW |
6355 | } |
6356 | break; | |
cfaf538b KW |
6357 | case SPACEA: |
6358 | while (scan < loceol && isSPACE_A((U8) *scan)) { | |
6359 | scan++; | |
6360 | } | |
6361 | break; | |
bbce6d69 | 6362 | case SPACEL: |
3280af22 | 6363 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6364 | if (utf8_target) { |
ffc61ed2 | 6365 | loceol = PL_regeol; |
1aa99e6b | 6366 | while (hardcount < max && scan < loceol && |
6bbba904 | 6367 | isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6368 | scan += UTF8SKIP(scan); |
6369 | hardcount++; | |
6370 | } | |
6371 | } else { | |
6372 | while (scan < loceol && isSPACE_LC(*scan)) | |
6373 | scan++; | |
a0ed51b3 LW |
6374 | } |
6375 | break; | |
980866de | 6376 | case NSPACEU: |
f2ed9b32 | 6377 | if (utf8_target) { |
980866de KW |
6378 | |
6379 | utf8_Nspace: | |
6380 | ||
ffc61ed2 | 6381 | loceol = PL_regeol; |
1a4fad37 | 6382 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6383 | while (hardcount < max && scan < loceol && |
980866de KW |
6384 | ! (*scan == ' ' || |
6385 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) | |
a12cf05f | 6386 | { |
ffc61ed2 JH |
6387 | scan += UTF8SKIP(scan); |
6388 | hardcount++; | |
6389 | } | |
980866de KW |
6390 | break; |
6391 | } | |
6392 | else { | |
a12cf05f KW |
6393 | while (scan < loceol && ! isSPACE_L1((U8) *scan)) { |
6394 | scan++; | |
6395 | } | |
980866de KW |
6396 | } |
6397 | break; | |
6398 | case NSPACE: | |
6399 | if (utf8_target) | |
6400 | goto utf8_Nspace; | |
6401 | ||
6402 | while (scan < loceol && ! isSPACE((U8) *scan)) { | |
6403 | scan++; | |
a0ed51b3 | 6404 | } |
0008a298 | 6405 | break; |
cfaf538b KW |
6406 | case NSPACEA: |
6407 | if (utf8_target) { | |
6408 | while (scan < loceol && ! isSPACE_A((U8) *scan)) { | |
6409 | scan += UTF8SKIP(scan); | |
6410 | } | |
6411 | } | |
6412 | else { | |
6413 | while (scan < loceol && ! isSPACE_A((U8) *scan)) { | |
6414 | scan++; | |
6415 | } | |
6416 | } | |
6417 | break; | |
bbce6d69 | 6418 | case NSPACEL: |
3280af22 | 6419 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6420 | if (utf8_target) { |
ffc61ed2 | 6421 | loceol = PL_regeol; |
1aa99e6b | 6422 | while (hardcount < max && scan < loceol && |
6bbba904 | 6423 | !isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6424 | scan += UTF8SKIP(scan); |
6425 | hardcount++; | |
6426 | } | |
6427 | } else { | |
6428 | while (scan < loceol && !isSPACE_LC(*scan)) | |
6429 | scan++; | |
a0ed51b3 LW |
6430 | } |
6431 | break; | |
a0d0e21e | 6432 | case DIGIT: |
f2ed9b32 | 6433 | if (utf8_target) { |
ffc61ed2 | 6434 | loceol = PL_regeol; |
1a4fad37 | 6435 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6436 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6437 | swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6438 | scan += UTF8SKIP(scan); |
6439 | hardcount++; | |
6440 | } | |
6441 | } else { | |
6442 | while (scan < loceol && isDIGIT(*scan)) | |
6443 | scan++; | |
a0ed51b3 LW |
6444 | } |
6445 | break; | |
cfaf538b KW |
6446 | case DIGITA: |
6447 | while (scan < loceol && isDIGIT_A((U8) *scan)) { | |
6448 | scan++; | |
6449 | } | |
6450 | break; | |
b77393f6 KW |
6451 | case DIGITL: |
6452 | PL_reg_flags |= RF_tainted; | |
6453 | if (utf8_target) { | |
6454 | loceol = PL_regeol; | |
6455 | while (hardcount < max && scan < loceol && | |
6456 | isDIGIT_LC_utf8((U8*)scan)) { | |
6457 | scan += UTF8SKIP(scan); | |
6458 | hardcount++; | |
6459 | } | |
6460 | } else { | |
6461 | while (scan < loceol && isDIGIT_LC(*scan)) | |
6462 | scan++; | |
6463 | } | |
6464 | break; | |
a0d0e21e | 6465 | case NDIGIT: |
f2ed9b32 | 6466 | if (utf8_target) { |
ffc61ed2 | 6467 | loceol = PL_regeol; |
1a4fad37 | 6468 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6469 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6470 | !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6471 | scan += UTF8SKIP(scan); |
6472 | hardcount++; | |
6473 | } | |
6474 | } else { | |
6475 | while (scan < loceol && !isDIGIT(*scan)) | |
6476 | scan++; | |
a0ed51b3 | 6477 | } |
cfaf538b KW |
6478 | break; |
6479 | case NDIGITA: | |
6480 | if (utf8_target) { | |
6481 | while (scan < loceol && ! isDIGIT_A((U8) *scan)) { | |
6482 | scan += UTF8SKIP(scan); | |
6483 | } | |
6484 | } | |
6485 | else { | |
6486 | while (scan < loceol && ! isDIGIT_A((U8) *scan)) { | |
6487 | scan++; | |
6488 | } | |
6489 | } | |
6490 | break; | |
b77393f6 KW |
6491 | case NDIGITL: |
6492 | PL_reg_flags |= RF_tainted; | |
6493 | if (utf8_target) { | |
6494 | loceol = PL_regeol; | |
6495 | while (hardcount < max && scan < loceol && | |
6496 | !isDIGIT_LC_utf8((U8*)scan)) { | |
6497 | scan += UTF8SKIP(scan); | |
6498 | hardcount++; | |
6499 | } | |
6500 | } else { | |
6501 | while (scan < loceol && !isDIGIT_LC(*scan)) | |
6502 | scan++; | |
6503 | } | |
6504 | break; | |
e1d1eefb | 6505 | case LNBREAK: |
f2ed9b32 | 6506 | if (utf8_target) { |
e1d1eefb YO |
6507 | loceol = PL_regeol; |
6508 | while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) { | |
6509 | scan += c; | |
6510 | hardcount++; | |
6511 | } | |
6512 | } else { | |
6513 | /* | |
6514 | LNBREAK can match two latin chars, which is ok, | |
6515 | because we have a null terminated string, but we | |
6516 | have to use hardcount in this situation | |
6517 | */ | |
6518 | while (scan < loceol && (c=is_LNBREAK_latin1(scan))) { | |
6519 | scan+=c; | |
6520 | hardcount++; | |
6521 | } | |
6522 | } | |
6523 | break; | |
6524 | case HORIZWS: | |
f2ed9b32 | 6525 | if (utf8_target) { |
e1d1eefb YO |
6526 | loceol = PL_regeol; |
6527 | while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) { | |
6528 | scan += c; | |
6529 | hardcount++; | |
6530 | } | |
6531 | } else { | |
6532 | while (scan < loceol && is_HORIZWS_latin1(scan)) | |
6533 | scan++; | |
6534 | } | |
a0ed51b3 | 6535 | break; |
e1d1eefb | 6536 | case NHORIZWS: |
f2ed9b32 | 6537 | if (utf8_target) { |
e1d1eefb YO |
6538 | loceol = PL_regeol; |
6539 | while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) { | |
6540 | scan += UTF8SKIP(scan); | |
6541 | hardcount++; | |
6542 | } | |
6543 | } else { | |
6544 | while (scan < loceol && !is_HORIZWS_latin1(scan)) | |
6545 | scan++; | |
6546 | ||
6547 | } | |
6548 | break; | |
6549 | case VERTWS: | |
f2ed9b32 | 6550 | if (utf8_target) { |
e1d1eefb YO |
6551 | loceol = PL_regeol; |
6552 | while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) { | |
6553 | scan += c; | |
6554 | hardcount++; | |
6555 | } | |
6556 | } else { | |
6557 | while (scan < loceol && is_VERTWS_latin1(scan)) | |
6558 | scan++; | |
6559 | ||
6560 | } | |
6561 | break; | |
6562 | case NVERTWS: | |
f2ed9b32 | 6563 | if (utf8_target) { |
e1d1eefb YO |
6564 | loceol = PL_regeol; |
6565 | while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) { | |
6566 | scan += UTF8SKIP(scan); | |
6567 | hardcount++; | |
6568 | } | |
6569 | } else { | |
6570 | while (scan < loceol && !is_VERTWS_latin1(scan)) | |
6571 | scan++; | |
6572 | ||
6573 | } | |
6574 | break; | |
6575 | ||
a0d0e21e LW |
6576 | default: /* Called on something of 0 width. */ |
6577 | break; /* So match right here or not at all. */ | |
6578 | } | |
a687059c | 6579 | |
a0ed51b3 LW |
6580 | if (hardcount) |
6581 | c = hardcount; | |
6582 | else | |
6583 | c = scan - PL_reginput; | |
3280af22 | 6584 | PL_reginput = scan; |
a687059c | 6585 | |
a3621e74 | 6586 | DEBUG_r({ |
e68ec53f | 6587 | GET_RE_DEBUG_FLAGS_DECL; |
be8e71aa | 6588 | DEBUG_EXECUTE_r({ |
e68ec53f YO |
6589 | SV * const prop = sv_newmortal(); |
6590 | regprop(prog, prop, p); | |
6591 | PerlIO_printf(Perl_debug_log, | |
be8e71aa | 6592 | "%*s %s can match %"IVdf" times out of %"IVdf"...\n", |
e2e6a0f1 | 6593 | REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max); |
a3621e74 | 6594 | }); |
be8e71aa | 6595 | }); |
9041c2e3 | 6596 | |
a0d0e21e | 6597 | return(c); |
a687059c LW |
6598 | } |
6599 | ||
c277df42 | 6600 | |
be8e71aa | 6601 | #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) |
c277df42 | 6602 | /* |
6c6525b8 KW |
6603 | - regclass_swash - prepare the utf8 swash. Wraps the shared core version to |
6604 | create a copy so that changes the caller makes won't change the shared one | |
6605 | */ | |
ffc61ed2 | 6606 | SV * |
32fc9b6a | 6607 | Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) |
ffc61ed2 | 6608 | { |
6c6525b8 KW |
6609 | PERL_ARGS_ASSERT_REGCLASS_SWASH; |
6610 | return newSVsv(core_regclass_swash(prog, node, doinit, listsvp, altsvp)); | |
6611 | } | |
6612 | #endif | |
6613 | ||
6614 | STATIC SV * | |
6615 | S_core_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) | |
6616 | { | |
8c9eb58f KW |
6617 | /* Returns the swash for the input 'node' in the regex 'prog'. |
6618 | * If <doinit> is true, will attempt to create the swash if not already | |
6619 | * done. | |
6620 | * If <listsvp> is non-null, will return the swash initialization string in | |
6621 | * it. | |
6622 | * If <altsvp> is non-null, will return the alternates to the regular swash | |
6623 | * in it | |
6624 | * Tied intimately to how regcomp.c sets up the data structure */ | |
6625 | ||
97aff369 | 6626 | dVAR; |
9e55ce06 JH |
6627 | SV *sw = NULL; |
6628 | SV *si = NULL; | |
6629 | SV *alt = NULL; | |
7a6c6baa KW |
6630 | SV* invlist = NULL; |
6631 | ||
f8fc2ecf YO |
6632 | RXi_GET_DECL(prog,progi); |
6633 | const struct reg_data * const data = prog ? progi->data : NULL; | |
ffc61ed2 | 6634 | |
6c6525b8 | 6635 | PERL_ARGS_ASSERT_CORE_REGCLASS_SWASH; |
7918f24d | 6636 | |
ccb2541c KW |
6637 | assert(ANYOF_NONBITMAP(node)); |
6638 | ||
4f639d21 | 6639 | if (data && data->count) { |
a3b680e6 | 6640 | const U32 n = ARG(node); |
ffc61ed2 | 6641 | |
4f639d21 | 6642 | if (data->what[n] == 's') { |
ad64d0ec NC |
6643 | SV * const rv = MUTABLE_SV(data->data[n]); |
6644 | AV * const av = MUTABLE_AV(SvRV(rv)); | |
2d03de9c | 6645 | SV **const ary = AvARRAY(av); |
7a6c6baa | 6646 | bool invlist_has_user_defined_property; |
9041c2e3 | 6647 | |
8c9eb58f | 6648 | si = *ary; /* ary[0] = the string to initialize the swash with */ |
b11f357e | 6649 | |
7a6c6baa KW |
6650 | /* Elements 3 and 4 are either both present or both absent. [3] is |
6651 | * any inversion list generated at compile time; [4] indicates if | |
6652 | * that inversion list has any user-defined properties in it. */ | |
6653 | if (av_len(av) >= 3) { | |
6654 | invlist = ary[3]; | |
6655 | invlist_has_user_defined_property = cBOOL(SvUV(ary[4])); | |
6656 | } | |
6657 | else { | |
6658 | invlist = NULL; | |
6659 | invlist_has_user_defined_property = FALSE; | |
6660 | } | |
6661 | ||
8c9eb58f KW |
6662 | /* Element [1] is reserved for the set-up swash. If already there, |
6663 | * return it; if not, create it and store it there */ | |
f192cf32 KW |
6664 | if (SvROK(ary[1])) { |
6665 | sw = ary[1]; | |
6666 | } | |
ffc61ed2 | 6667 | else if (si && doinit) { |
7a6c6baa KW |
6668 | |
6669 | sw = _core_swash_init("utf8", /* the utf8 package */ | |
6670 | "", /* nameless */ | |
6671 | si, | |
6672 | 1, /* binary */ | |
6673 | 0, /* not from tr/// */ | |
6674 | FALSE, /* is error if can't find | |
6675 | property */ | |
6676 | invlist, | |
6677 | invlist_has_user_defined_property); | |
ffc61ed2 JH |
6678 | (void)av_store(av, 1, sw); |
6679 | } | |
8c9eb58f KW |
6680 | |
6681 | /* Element [2] is for any multi-char folds. Note that is a | |
6682 | * fundamentally flawed design, because can't backtrack and try | |
6683 | * again. See [perl #89774] */ | |
f192cf32 KW |
6684 | if (SvTYPE(ary[2]) == SVt_PVAV) { |
6685 | alt = ary[2]; | |
6686 | } | |
ffc61ed2 JH |
6687 | } |
6688 | } | |
6689 | ||
7a6c6baa KW |
6690 | if (listsvp) { |
6691 | SV* matches_string = newSVpvn("", 0); | |
6692 | SV** invlistsvp; | |
6693 | ||
6694 | /* Use the swash, if any, which has to have incorporated into it all | |
6695 | * possibilities */ | |
6696 | if ( sw | |
6697 | && SvROK(sw) | |
6698 | && SvTYPE(SvRV(sw)) == SVt_PVHV | |
6699 | && (invlistsvp = hv_fetchs(MUTABLE_HV(SvRV(sw)), "INVLIST", FALSE))) | |
6700 | { | |
6701 | invlist = *invlistsvp; | |
6702 | } | |
6703 | else if (si && si != &PL_sv_undef) { | |
6704 | ||
6705 | /* If no swash, use the input nitialization string, if available */ | |
6706 | sv_catsv(matches_string, si); | |
6707 | } | |
6708 | ||
6709 | /* Add the inversion list to whatever we have. This may have come from | |
6710 | * the swash, or from an input parameter */ | |
6711 | if (invlist) { | |
6712 | sv_catsv(matches_string, _invlist_contents(invlist)); | |
6713 | } | |
6714 | *listsvp = matches_string; | |
6715 | } | |
6716 | ||
9e55ce06 JH |
6717 | if (altsvp) |
6718 | *altsvp = alt; | |
ffc61ed2 JH |
6719 | |
6720 | return sw; | |
6721 | } | |
6722 | ||
6723 | /* | |
ba7b4546 | 6724 | - reginclass - determine if a character falls into a character class |
832705d4 | 6725 | |
6698fab5 KW |
6726 | n is the ANYOF regnode |
6727 | p is the target string | |
6728 | lenp is pointer to the maximum number of bytes of how far to go in p | |
6729 | (This is assumed wthout checking to always be at least the current | |
6730 | character's size) | |
6731 | utf8_target tells whether p is in UTF-8. | |
832705d4 | 6732 | |
4b3cda86 KW |
6733 | Returns true if matched; false otherwise. If lenp is not NULL, on return |
6734 | from a successful match, the value it points to will be updated to how many | |
6735 | bytes in p were matched. If there was no match, the value is undefined, | |
6736 | possibly changed from the input. | |
eba1359e | 6737 | |
d5788240 KW |
6738 | Note that this can be a synthetic start class, a combination of various |
6739 | nodes, so things you think might be mutually exclusive, such as locale, | |
6740 | aren't. It can match both locale and non-locale | |
6741 | ||
bbce6d69 | 6742 | */ |
6743 | ||
76e3520e | 6744 | STATIC bool |
f6ad78d8 | 6745 | S_reginclass(pTHX_ const regexp * const prog, register const regnode * const n, register const U8* const p, STRLEN* lenp, register const bool utf8_target) |
bbce6d69 | 6746 | { |
27da23d5 | 6747 | dVAR; |
a3b680e6 | 6748 | const char flags = ANYOF_FLAGS(n); |
bbce6d69 | 6749 | bool match = FALSE; |
cc07378b | 6750 | UV c = *p; |
f7ab54c6 | 6751 | STRLEN c_len = 0; |
6698fab5 | 6752 | STRLEN maxlen; |
1aa99e6b | 6753 | |
7918f24d NC |
6754 | PERL_ARGS_ASSERT_REGINCLASS; |
6755 | ||
4b3cda86 | 6756 | /* If c is not already the code point, get it */ |
f2ed9b32 | 6757 | if (utf8_target && !UTF8_IS_INVARIANT(c)) { |
f7ab54c6 | 6758 | c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &c_len, |
6182169b KW |
6759 | (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) |
6760 | | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY); | |
6761 | /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for | |
6762 | * UTF8_ALLOW_FFFF */ | |
f7ab54c6 | 6763 | if (c_len == (STRLEN)-1) |
e8a70c6f | 6764 | Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); |
19f67299 | 6765 | } |
a5a291f5 KW |
6766 | else { |
6767 | c_len = 1; | |
6768 | } | |
bbce6d69 | 6769 | |
a5a291f5 KW |
6770 | /* Use passed in max length, or one character if none passed in or less |
6771 | * than one character. And assume will match just one character. This is | |
6772 | * overwritten later if matched more. */ | |
4b3cda86 | 6773 | if (lenp) { |
a5a291f5 KW |
6774 | maxlen = (*lenp > c_len) ? *lenp : c_len; |
6775 | *lenp = c_len; | |
4b3cda86 KW |
6776 | |
6777 | } | |
6778 | else { | |
a5a291f5 | 6779 | maxlen = c_len; |
4b3cda86 KW |
6780 | } |
6781 | ||
7cdde544 KW |
6782 | /* If this character is potentially in the bitmap, check it */ |
6783 | if (c < 256) { | |
ffc61ed2 JH |
6784 | if (ANYOF_BITMAP_TEST(n, c)) |
6785 | match = TRUE; | |
11454c59 KW |
6786 | else if (flags & ANYOF_NON_UTF8_LATIN1_ALL |
6787 | && ! utf8_target | |
6788 | && ! isASCII(c)) | |
6789 | { | |
6790 | match = TRUE; | |
6791 | } | |
a0ed51b3 | 6792 | |
78969a98 KW |
6793 | else if (flags & ANYOF_LOCALE) { |
6794 | PL_reg_flags |= RF_tainted; | |
6795 | ||
6796 | if ((flags & ANYOF_LOC_NONBITMAP_FOLD) | |
6797 | && ANYOF_BITMAP_TEST(n, PL_fold_locale[c])) | |
6798 | { | |
ffc61ed2 | 6799 | match = TRUE; |
78969a98 | 6800 | } |
040aea3a KW |
6801 | else if (ANYOF_CLASS_TEST_ANY_SET(n) && |
6802 | ((ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) || | |
6803 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) || | |
6804 | (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) || | |
6805 | (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) || | |
6806 | (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) || | |
6807 | (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) || | |
6808 | (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) || | |
6809 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) || | |
6810 | (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) || | |
6811 | (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) || | |
07315176 KW |
6812 | (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII_LC(c)) || |
6813 | (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII_LC(c)) || | |
040aea3a KW |
6814 | (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) || |
6815 | (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) || | |
6816 | (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) || | |
6817 | (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) || | |
6818 | (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) || | |
6819 | (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) || | |
6820 | (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) || | |
6821 | (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) || | |
6822 | (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) || | |
6823 | (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) || | |
6824 | (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) || | |
6825 | (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) || | |
6826 | (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) || | |
6827 | (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) || | |
6828 | (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) || | |
6829 | (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) || | |
67addccf KW |
6830 | (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK_LC(c)) || |
6831 | (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK_LC(c)) | |
040aea3a | 6832 | ) /* How's that for a conditional? */ |
78969a98 | 6833 | ) { |
ffc61ed2 JH |
6834 | match = TRUE; |
6835 | } | |
a0ed51b3 | 6836 | } |
a0ed51b3 LW |
6837 | } |
6838 | ||
7cdde544 | 6839 | /* If the bitmap didn't (or couldn't) match, and something outside the |
de87c4fe KW |
6840 | * bitmap could match, try that. Locale nodes specifiy completely the |
6841 | * behavior of code points in the bit map (otherwise, a utf8 target would | |
c613755a | 6842 | * cause them to be treated as Unicode and not locale), except in |
de87c4fe | 6843 | * the very unlikely event when this node is a synthetic start class, which |
c613755a KW |
6844 | * could be a combination of locale and non-locale nodes. So allow locale |
6845 | * to match for the synthetic start class, which will give a false | |
6846 | * positive that will be resolved when the match is done again as not part | |
6847 | * of the synthetic start class */ | |
ef87b810 | 6848 | if (!match) { |
10ee90d2 KW |
6849 | if (utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256) { |
6850 | match = TRUE; /* Everything above 255 matches */ | |
e051a21d | 6851 | } |
6f8d7d0d KW |
6852 | else if (ANYOF_NONBITMAP(n) |
6853 | && ((flags & ANYOF_NONBITMAP_NON_UTF8) | |
6854 | || (utf8_target | |
6855 | && (c >=256 | |
6856 | || (! (flags & ANYOF_LOCALE)) | |
6857 | || (flags & ANYOF_IS_SYNTHETIC))))) | |
ef87b810 | 6858 | { |
7cdde544 | 6859 | AV *av; |
210e6c47 | 6860 | SV * const sw = core_regclass_swash(prog, n, TRUE, 0, (SV**)&av); |
7cdde544 KW |
6861 | |
6862 | if (sw) { | |
6863 | U8 * utf8_p; | |
6864 | if (utf8_target) { | |
6865 | utf8_p = (U8 *) p; | |
6866 | } else { | |
f56b6394 KW |
6867 | |
6868 | /* Not utf8. Convert as much of the string as available up | |
6869 | * to the limit of how far the (single) character in the | |
6870 | * pattern can possibly match (no need to go further). If | |
6871 | * the node is a straight ANYOF or not folding, it can't | |
6872 | * match more than one. Otherwise, It can match up to how | |
6873 | * far a single char can fold to. Since not utf8, each | |
6874 | * character is a single byte, so the max it can be in | |
6875 | * bytes is the same as the max it can be in characters */ | |
6876 | STRLEN len = (OP(n) == ANYOF | |
6877 | || ! (flags & ANYOF_LOC_NONBITMAP_FOLD)) | |
6878 | ? 1 | |
6879 | : (maxlen < UTF8_MAX_FOLD_CHAR_EXPAND) | |
6880 | ? maxlen | |
6881 | : UTF8_MAX_FOLD_CHAR_EXPAND; | |
7cdde544 KW |
6882 | utf8_p = bytes_to_utf8(p, &len); |
6883 | } | |
f56b6394 | 6884 | |
f7eb7452 | 6885 | if (swash_fetch(sw, utf8_p, TRUE)) |
7cdde544 | 6886 | match = TRUE; |
39065660 | 6887 | else if (flags & ANYOF_LOC_NONBITMAP_FOLD) { |
f56b6394 KW |
6888 | |
6889 | /* Here, we need to test if the fold of the target string | |
8d5d17fa KW |
6890 | * matches. The non-multi char folds have all been moved to |
6891 | * the compilation phase, and the multi-char folds have | |
6892 | * been stored by regcomp into 'av'; we linearly check to | |
6893 | * see if any match the target string (folded). We know | |
6894 | * that the originals were each one character, but we don't | |
6895 | * currently know how many characters/bytes each folded to, | |
6896 | * except we do know that there are small limits imposed by | |
6897 | * Unicode. XXX A performance enhancement would be to have | |
6898 | * regcomp.c store the max number of chars/bytes that are | |
6899 | * in an av entry, as, say the 0th element. Even better | |
6900 | * would be to have a hash of the few characters that can | |
6901 | * start a multi-char fold to the max number of chars of | |
6902 | * those folds. | |
f56b6394 KW |
6903 | * |
6904 | * If there is a match, we will need to advance (if lenp is | |
6905 | * specified) the match pointer in the target string. But | |
6906 | * what we are comparing here isn't that string directly, | |
6907 | * but its fold, whose length may differ from the original. | |
6908 | * As we go along in constructing the fold, therefore, we | |
6909 | * create a map so that we know how many bytes in the | |
6910 | * source to advance given that we have matched a certain | |
6911 | * number of bytes in the fold. This map is stored in | |
6438af90 KW |
6912 | * 'map_fold_len_back'. Let n mean the number of bytes in |
6913 | * the fold of the first character that we are folding. | |
6914 | * Then map_fold_len_back[n] is set to the number of bytes | |
6915 | * in that first character. Similarly let m be the | |
6916 | * corresponding number for the second character to be | |
6917 | * folded. Then map_fold_len_back[n+m] is set to the | |
6918 | * number of bytes occupied by the first two source | |
6919 | * characters. ... */ | |
6920 | U8 map_fold_len_back[UTF8_MAXBYTES_CASE+1] = { 0 }; | |
f56b6394 KW |
6921 | U8 folded[UTF8_MAXBYTES_CASE+1]; |
6922 | STRLEN foldlen = 0; /* num bytes in fold of 1st char */ | |
deba3d96 | 6923 | STRLEN total_foldlen = 0; /* num bytes in fold of all |
6438af90 | 6924 | chars */ |
f56b6394 KW |
6925 | |
6926 | if (OP(n) == ANYOF || maxlen == 1 || ! lenp || ! av) { | |
6927 | ||
6928 | /* Here, only need to fold the first char of the target | |
6438af90 | 6929 | * string. It the source wasn't utf8, is 1 byte long */ |
f56b6394 | 6930 | to_utf8_fold(utf8_p, folded, &foldlen); |
deba3d96 | 6931 | total_foldlen = foldlen; |
6438af90 KW |
6932 | map_fold_len_back[foldlen] = (utf8_target) |
6933 | ? UTF8SKIP(utf8_p) | |
6934 | : 1; | |
f56b6394 KW |
6935 | } |
6936 | else { | |
6937 | ||
6938 | /* Here, need to fold more than the first char. Do so | |
6939 | * up to the limits */ | |
f56b6394 KW |
6940 | U8* source_ptr = utf8_p; /* The source for the fold |
6941 | is the regex target | |
6942 | string */ | |
6943 | U8* folded_ptr = folded; | |
6944 | U8* e = utf8_p + maxlen; /* Can't go beyond last | |
6945 | available byte in the | |
6946 | target string */ | |
6438af90 KW |
6947 | U8 i; |
6948 | for (i = 0; | |
6949 | i < UTF8_MAX_FOLD_CHAR_EXPAND && source_ptr < e; | |
6950 | i++) | |
f56b6394 KW |
6951 | { |
6952 | ||
6953 | /* Fold the next character */ | |
6954 | U8 this_char_folded[UTF8_MAXBYTES_CASE+1]; | |
6955 | STRLEN this_char_foldlen; | |
6956 | to_utf8_fold(source_ptr, | |
6957 | this_char_folded, | |
6958 | &this_char_foldlen); | |
6959 | ||
6960 | /* Bail if it would exceed the byte limit for | |
6961 | * folding a single char. */ | |
6962 | if (this_char_foldlen + folded_ptr - folded > | |
6963 | UTF8_MAXBYTES_CASE) | |
6964 | { | |
6965 | break; | |
6966 | } | |
6967 | ||
6438af90 | 6968 | /* Add the fold of this character */ |
f56b6394 KW |
6969 | Copy(this_char_folded, |
6970 | folded_ptr, | |
6971 | this_char_foldlen, | |
6972 | U8); | |
f56b6394 | 6973 | source_ptr += UTF8SKIP(source_ptr); |
6438af90 | 6974 | folded_ptr += this_char_foldlen; |
deba3d96 | 6975 | total_foldlen = folded_ptr - folded; |
6438af90 KW |
6976 | |
6977 | /* Create map from the number of bytes in the fold | |
6978 | * back to the number of bytes in the source. If | |
6979 | * the source isn't utf8, the byte count is just | |
6980 | * the number of characters so far */ | |
deba3d96 | 6981 | map_fold_len_back[total_foldlen] |
6438af90 KW |
6982 | = (utf8_target) |
6983 | ? source_ptr - utf8_p | |
6984 | : i + 1; | |
f56b6394 KW |
6985 | } |
6986 | *folded_ptr = '\0'; | |
f56b6394 KW |
6987 | } |
6988 | ||
6989 | ||
6990 | /* Do the linear search to see if the fold is in the list | |
8f2655f7 KW |
6991 | * of multi-char folds. */ |
6992 | if (av) { | |
7cdde544 KW |
6993 | I32 i; |
6994 | for (i = 0; i <= av_len(av); i++) { | |
6995 | SV* const sv = *av_fetch(av, i, FALSE); | |
6996 | STRLEN len; | |
6997 | const char * const s = SvPV_const(sv, len); | |
6438af90 | 6998 | |
f9126265 KW |
6999 | if (len <= total_foldlen |
7000 | && memEQ(s, (char*)folded, len) | |
7001 | ||
7002 | /* If 0, means matched a partial char. See | |
7003 | * [perl #90536] */ | |
7004 | && map_fold_len_back[len]) | |
f56b6394 KW |
7005 | { |
7006 | ||
7007 | /* Advance the target string ptr to account for | |
7008 | * this fold, but have to translate from the | |
7009 | * folded length to the corresponding source | |
6438af90 | 7010 | * length. */ |
8f2655f7 KW |
7011 | if (lenp) { |
7012 | *lenp = map_fold_len_back[len]; | |
8f2655f7 | 7013 | } |
7cdde544 KW |
7014 | match = TRUE; |
7015 | break; | |
7016 | } | |
7017 | } | |
7018 | } | |
7cdde544 KW |
7019 | } |
7020 | ||
7021 | /* If we allocated a string above, free it */ | |
7022 | if (! utf8_target) Safefree(utf8_p); | |
7023 | } | |
7024 | } | |
7025 | } | |
7026 | ||
a0ed51b3 LW |
7027 | return (flags & ANYOF_INVERT) ? !match : match; |
7028 | } | |
161b471a | 7029 | |
dfe13c55 | 7030 | STATIC U8 * |
0ce71af7 | 7031 | S_reghop3(U8 *s, I32 off, const U8* lim) |
9041c2e3 | 7032 | { |
6af86488 KW |
7033 | /* return the position 'off' UTF-8 characters away from 's', forward if |
7034 | * 'off' >= 0, backwards if negative. But don't go outside of position | |
7035 | * 'lim', which better be < s if off < 0 */ | |
7036 | ||
97aff369 | 7037 | dVAR; |
7918f24d NC |
7038 | |
7039 | PERL_ARGS_ASSERT_REGHOP3; | |
7040 | ||
a0ed51b3 | 7041 | if (off >= 0) { |
1aa99e6b | 7042 | while (off-- && s < lim) { |
ffc61ed2 | 7043 | /* XXX could check well-formedness here */ |
a0ed51b3 | 7044 | s += UTF8SKIP(s); |
ffc61ed2 | 7045 | } |
a0ed51b3 LW |
7046 | } |
7047 | else { | |
1de06328 YO |
7048 | while (off++ && s > lim) { |
7049 | s--; | |
7050 | if (UTF8_IS_CONTINUED(*s)) { | |
7051 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
7052 | s--; | |
a0ed51b3 | 7053 | } |
1de06328 | 7054 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
7055 | } |
7056 | } | |
7057 | return s; | |
7058 | } | |
161b471a | 7059 | |
f9f4320a YO |
7060 | #ifdef XXX_dmq |
7061 | /* there are a bunch of places where we use two reghop3's that should | |
7062 | be replaced with this routine. but since thats not done yet | |
7063 | we ifdef it out - dmq | |
7064 | */ | |
dfe13c55 | 7065 | STATIC U8 * |
1de06328 YO |
7066 | S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim) |
7067 | { | |
7068 | dVAR; | |
7918f24d NC |
7069 | |
7070 | PERL_ARGS_ASSERT_REGHOP4; | |
7071 | ||
1de06328 YO |
7072 | if (off >= 0) { |
7073 | while (off-- && s < rlim) { | |
7074 | /* XXX could check well-formedness here */ | |
7075 | s += UTF8SKIP(s); | |
7076 | } | |
7077 | } | |
7078 | else { | |
7079 | while (off++ && s > llim) { | |
7080 | s--; | |
7081 | if (UTF8_IS_CONTINUED(*s)) { | |
7082 | while (s > llim && UTF8_IS_CONTINUATION(*s)) | |
7083 | s--; | |
7084 | } | |
7085 | /* XXX could check well-formedness here */ | |
7086 | } | |
7087 | } | |
7088 | return s; | |
7089 | } | |
f9f4320a | 7090 | #endif |
1de06328 YO |
7091 | |
7092 | STATIC U8 * | |
0ce71af7 | 7093 | S_reghopmaybe3(U8* s, I32 off, const U8* lim) |
a0ed51b3 | 7094 | { |
97aff369 | 7095 | dVAR; |
7918f24d NC |
7096 | |
7097 | PERL_ARGS_ASSERT_REGHOPMAYBE3; | |
7098 | ||
a0ed51b3 | 7099 | if (off >= 0) { |
1aa99e6b | 7100 | while (off-- && s < lim) { |
ffc61ed2 | 7101 | /* XXX could check well-formedness here */ |
a0ed51b3 | 7102 | s += UTF8SKIP(s); |
ffc61ed2 | 7103 | } |
a0ed51b3 | 7104 | if (off >= 0) |
3dab1dad | 7105 | return NULL; |
a0ed51b3 LW |
7106 | } |
7107 | else { | |
1de06328 YO |
7108 | while (off++ && s > lim) { |
7109 | s--; | |
7110 | if (UTF8_IS_CONTINUED(*s)) { | |
7111 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
7112 | s--; | |
a0ed51b3 | 7113 | } |
1de06328 | 7114 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
7115 | } |
7116 | if (off <= 0) | |
3dab1dad | 7117 | return NULL; |
a0ed51b3 LW |
7118 | } |
7119 | return s; | |
7120 | } | |
51371543 | 7121 | |
51371543 | 7122 | static void |
acfe0abc | 7123 | restore_pos(pTHX_ void *arg) |
51371543 | 7124 | { |
97aff369 | 7125 | dVAR; |
097eb12c | 7126 | regexp * const rex = (regexp *)arg; |
ed301438 | 7127 | if (PL_reg_state.re_state_eval_setup_done) { |
51371543 | 7128 | if (PL_reg_oldsaved) { |
4f639d21 DM |
7129 | rex->subbeg = PL_reg_oldsaved; |
7130 | rex->sublen = PL_reg_oldsavedlen; | |
f8c7b90f | 7131 | #ifdef PERL_OLD_COPY_ON_WRITE |
4f639d21 | 7132 | rex->saved_copy = PL_nrs; |
ed252734 | 7133 | #endif |
07bc277f | 7134 | RXp_MATCH_COPIED_on(rex); |
51371543 GS |
7135 | } |
7136 | PL_reg_magic->mg_len = PL_reg_oldpos; | |
ed301438 | 7137 | PL_reg_state.re_state_eval_setup_done = FALSE; |
51371543 GS |
7138 | PL_curpm = PL_reg_oldcurpm; |
7139 | } | |
7140 | } | |
33b8afdf JH |
7141 | |
7142 | STATIC void | |
7143 | S_to_utf8_substr(pTHX_ register regexp *prog) | |
7144 | { | |
a1cac82e | 7145 | int i = 1; |
7918f24d NC |
7146 | |
7147 | PERL_ARGS_ASSERT_TO_UTF8_SUBSTR; | |
7148 | ||
a1cac82e NC |
7149 | do { |
7150 | if (prog->substrs->data[i].substr | |
7151 | && !prog->substrs->data[i].utf8_substr) { | |
7152 | SV* const sv = newSVsv(prog->substrs->data[i].substr); | |
7153 | prog->substrs->data[i].utf8_substr = sv; | |
7154 | sv_utf8_upgrade(sv); | |
610460f9 | 7155 | if (SvVALID(prog->substrs->data[i].substr)) { |
cffe132d | 7156 | if (SvTAIL(prog->substrs->data[i].substr)) { |
610460f9 NC |
7157 | /* Trim the trailing \n that fbm_compile added last |
7158 | time. */ | |
7159 | SvCUR_set(sv, SvCUR(sv) - 1); | |
7160 | /* Whilst this makes the SV technically "invalid" (as its | |
7161 | buffer is no longer followed by "\0") when fbm_compile() | |
7162 | adds the "\n" back, a "\0" is restored. */ | |
cffe132d NC |
7163 | fbm_compile(sv, FBMcf_TAIL); |
7164 | } else | |
7165 | fbm_compile(sv, 0); | |
610460f9 | 7166 | } |
a1cac82e NC |
7167 | if (prog->substrs->data[i].substr == prog->check_substr) |
7168 | prog->check_utf8 = sv; | |
7169 | } | |
7170 | } while (i--); | |
33b8afdf JH |
7171 | } |
7172 | ||
7173 | STATIC void | |
7174 | S_to_byte_substr(pTHX_ register regexp *prog) | |
7175 | { | |
97aff369 | 7176 | dVAR; |
a1cac82e | 7177 | int i = 1; |
7918f24d NC |
7178 | |
7179 | PERL_ARGS_ASSERT_TO_BYTE_SUBSTR; | |
7180 | ||
a1cac82e NC |
7181 | do { |
7182 | if (prog->substrs->data[i].utf8_substr | |
7183 | && !prog->substrs->data[i].substr) { | |
7184 | SV* sv = newSVsv(prog->substrs->data[i].utf8_substr); | |
7185 | if (sv_utf8_downgrade(sv, TRUE)) { | |
610460f9 | 7186 | if (SvVALID(prog->substrs->data[i].utf8_substr)) { |
cffe132d | 7187 | if (SvTAIL(prog->substrs->data[i].utf8_substr)) { |
610460f9 NC |
7188 | /* Trim the trailing \n that fbm_compile added last |
7189 | time. */ | |
7190 | SvCUR_set(sv, SvCUR(sv) - 1); | |
cffe132d NC |
7191 | fbm_compile(sv, FBMcf_TAIL); |
7192 | } else | |
7193 | fbm_compile(sv, 0); | |
7194 | } | |
a1cac82e NC |
7195 | } else { |
7196 | SvREFCNT_dec(sv); | |
7197 | sv = &PL_sv_undef; | |
7198 | } | |
7199 | prog->substrs->data[i].substr = sv; | |
7200 | if (prog->substrs->data[i].utf8_substr == prog->check_utf8) | |
7201 | prog->check_substr = sv; | |
33b8afdf | 7202 | } |
a1cac82e | 7203 | } while (i--); |
33b8afdf | 7204 | } |
66610fdd RGS |
7205 | |
7206 | /* | |
7207 | * Local variables: | |
7208 | * c-indentation-style: bsd | |
7209 | * c-basic-offset: 4 | |
14d04a33 | 7210 | * indent-tabs-mode: nil |
66610fdd RGS |
7211 | * End: |
7212 | * | |
14d04a33 | 7213 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 7214 | */ |