Commit | Line | Data |
---|---|---|
a0d0e21e LW |
1 | /* regexec.c |
2 | */ | |
3 | ||
4 | /* | |
4ac71550 TC |
5 | * One Ring to rule them all, One Ring to find them |
6 | & | |
7 | * [p.v of _The Lord of the Rings_, opening poem] | |
8 | * [p.50 of _The Lord of the Rings_, I/iii: "The Shadow of the Past"] | |
9 | * [p.254 of _The Lord of the Rings_, II/ii: "The Council of Elrond"] | |
a0d0e21e LW |
10 | */ |
11 | ||
61296642 DM |
12 | /* This file contains functions for executing a regular expression. See |
13 | * also regcomp.c which funnily enough, contains functions for compiling | |
166f8a29 | 14 | * a regular expression. |
e4a054ea DM |
15 | * |
16 | * This file is also copied at build time to ext/re/re_exec.c, where | |
17 | * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT. | |
18 | * This causes the main functions to be compiled under new names and with | |
19 | * debugging support added, which makes "use re 'debug'" work. | |
166f8a29 DM |
20 | */ |
21 | ||
a687059c LW |
22 | /* NOTE: this is derived from Henry Spencer's regexp code, and should not |
23 | * confused with the original package (see point 3 below). Thanks, Henry! | |
24 | */ | |
25 | ||
26 | /* Additional note: this code is very heavily munged from Henry's version | |
27 | * in places. In some spots I've traded clarity for efficiency, so don't | |
28 | * blame Henry for some of the lack of readability. | |
29 | */ | |
30 | ||
e50aee73 AD |
31 | /* The names of the functions have been changed from regcomp and |
32 | * regexec to pregcomp and pregexec in order to avoid conflicts | |
33 | * with the POSIX routines of the same names. | |
34 | */ | |
35 | ||
b9d5759e | 36 | #ifdef PERL_EXT_RE_BUILD |
54df2634 | 37 | #include "re_top.h" |
9041c2e3 | 38 | #endif |
56953603 | 39 | |
a687059c | 40 | /* |
e50aee73 | 41 | * pregcomp and pregexec -- regsub and regerror are not used in perl |
a687059c LW |
42 | * |
43 | * Copyright (c) 1986 by University of Toronto. | |
44 | * Written by Henry Spencer. Not derived from licensed software. | |
45 | * | |
46 | * Permission is granted to anyone to use this software for any | |
47 | * purpose on any computer system, and to redistribute it freely, | |
48 | * subject to the following restrictions: | |
49 | * | |
50 | * 1. The author is not responsible for the consequences of use of | |
51 | * this software, no matter how awful, even if they arise | |
52 | * from defects in it. | |
53 | * | |
54 | * 2. The origin of this software must not be misrepresented, either | |
55 | * by explicit claim or by omission. | |
56 | * | |
57 | * 3. Altered versions must be plainly marked as such, and must not | |
58 | * be misrepresented as being the original software. | |
59 | * | |
60 | **** Alterations to Henry's code are... | |
61 | **** | |
4bb101f2 | 62 | **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, |
1129b882 NC |
63 | **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 |
64 | **** by Larry Wall and others | |
a687059c | 65 | **** |
9ef589d8 LW |
66 | **** You may distribute under the terms of either the GNU General Public |
67 | **** License or the Artistic License, as specified in the README file. | |
a687059c LW |
68 | * |
69 | * Beware that some of this code is subtly aware of the way operator | |
70 | * precedence is structured in regular expressions. Serious changes in | |
71 | * regular-expression syntax might require a total rethink. | |
72 | */ | |
73 | #include "EXTERN.h" | |
864dbfa3 | 74 | #define PERL_IN_REGEXEC_C |
a687059c | 75 | #include "perl.h" |
0f5d15d6 | 76 | |
54df2634 NC |
77 | #ifdef PERL_IN_XSUB_RE |
78 | # include "re_comp.h" | |
79 | #else | |
80 | # include "regcomp.h" | |
81 | #endif | |
a687059c | 82 | |
81e983c1 | 83 | #include "inline_invlist.c" |
1b0f46bf | 84 | #include "unicode_constants.h" |
81e983c1 | 85 | |
ef07e810 | 86 | #define RF_tainted 1 /* tainted information used? e.g. locale */ |
c277df42 | 87 | #define RF_warned 2 /* warned about big count? */ |
faec1544 | 88 | |
ab3bbdeb | 89 | #define RF_utf8 8 /* Pattern contains multibyte chars? */ |
a0ed51b3 | 90 | |
f2ed9b32 | 91 | #define UTF_PATTERN ((PL_reg_flags & RF_utf8) != 0) |
ce862d02 | 92 | |
a687059c LW |
93 | #ifndef STATIC |
94 | #define STATIC static | |
95 | #endif | |
96 | ||
7e2509c1 KW |
97 | /* Valid for non-utf8 strings, non-ANYOFV nodes only: avoids the reginclass |
98 | * call if there are no complications: i.e., if everything matchable is | |
99 | * straight forward in the bitmap */ | |
af364d03 KW |
100 | #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) \ |
101 | : ANYOF_BITMAP_TEST(p,*(c))) | |
7d3e948e | 102 | |
c277df42 IZ |
103 | /* |
104 | * Forwards. | |
105 | */ | |
106 | ||
f2ed9b32 | 107 | #define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv)) |
53c4c00c | 108 | #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b) |
a0ed51b3 | 109 | |
3dab1dad YO |
110 | #define HOPc(pos,off) \ |
111 | (char *)(PL_reg_match_utf8 \ | |
52657f30 | 112 | ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \ |
3dab1dad YO |
113 | : (U8*)(pos + off)) |
114 | #define HOPBACKc(pos, off) \ | |
07be1b83 YO |
115 | (char*)(PL_reg_match_utf8\ |
116 | ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \ | |
117 | : (pos - off >= PL_bostr) \ | |
8e11feef | 118 | ? (U8*)pos - off \ |
3dab1dad | 119 | : NULL) |
efb30f32 | 120 | |
e7409c1b | 121 | #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off)) |
1aa99e6b | 122 | #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim)) |
1aa99e6b | 123 | |
20d0b1e9 | 124 | /* these are unrolled below in the CCC_TRY_XXX defined */ |
61dad979 | 125 | #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ |
9c4fdda1 | 126 | if (!CAT2(PL_utf8_,class)) { \ |
cf54f63a | 127 | bool ok; \ |
9c4fdda1 | 128 | ENTER; save_re_context(); \ |
cf54f63a JL |
129 | ok=CAT2(is_utf8_,class)((const U8*)str); \ |
130 | PERL_UNUSED_VAR(ok); \ | |
131 | assert(ok); assert(CAT2(PL_utf8_,class)); LEAVE; } } STMT_END | |
37e2e78e KW |
132 | /* Doesn't do an assert to verify that is correct */ |
133 | #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \ | |
9c4fdda1 | 134 | if (!CAT2(PL_utf8_,class)) { \ |
9d63fa07 | 135 | bool throw_away PERL_UNUSED_DECL; \ |
9c4fdda1 RB |
136 | ENTER; save_re_context(); \ |
137 | throw_away = CAT2(is_utf8_,class)((const U8*)" "); \ | |
138 | LEAVE; } } STMT_END | |
37e2e78e | 139 | |
1a4fad37 AL |
140 | #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a") |
141 | #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0") | |
142 | #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ") | |
51371543 | 143 | |
37e2e78e | 144 | #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \ |
61dad979 KW |
145 | /* No asserts are done for some of these, in case called on a */ \ |
146 | /* Unicode version in which they map to nothing */ \ | |
27d4fc33 | 147 | LOAD_UTF8_CHARCLASS(X_regular_begin, HYPHEN_UTF8); \ |
61dad979 | 148 | LOAD_UTF8_CHARCLASS(X_extend, COMBINING_GRAVE_ACCENT_UTF8); \ |
20d0b1e9 | 149 | |
1dcf4a1b | 150 | #define PLACEHOLDER /* Something for the preprocessor to grab onto */ |
d1eb3177 | 151 | |
ee9a90b8 KW |
152 | /* The actual code for CCC_TRY, which uses several variables from the routine |
153 | * it's callable from. It is designed to be the bulk of a case statement. | |
154 | * FUNC is the macro or function to call on non-utf8 targets that indicate if | |
155 | * nextchr matches the class. | |
156 | * UTF8_TEST is the whole test string to use for utf8 targets | |
157 | * LOAD is what to use to test, and if not present to load in the swash for the | |
158 | * class | |
159 | * POS_OR_NEG is either empty or ! to complement the results of FUNC or | |
160 | * UTF8_TEST test. | |
161 | * The logic is: Fail if we're at the end-of-string; otherwise if the target is | |
162 | * utf8 and a variant, load the swash if necessary and test using the utf8 | |
163 | * test. Advance to the next character if test is ok, otherwise fail; If not | |
164 | * utf8 or an invariant under utf8, use the non-utf8 test, and fail if it | |
165 | * fails, or advance to the next character */ | |
166 | ||
167 | #define _CCC_TRY_CODE(POS_OR_NEG, FUNC, UTF8_TEST, CLASS, STR) \ | |
168 | if (locinput >= PL_regeol) { \ | |
169 | sayNO; \ | |
170 | } \ | |
171 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ | |
172 | LOAD_UTF8_CHARCLASS(CLASS, STR); \ | |
173 | if (POS_OR_NEG (UTF8_TEST)) { \ | |
174 | sayNO; \ | |
175 | } \ | |
ee9a90b8 | 176 | } \ |
28b98f76 DM |
177 | else if (POS_OR_NEG (FUNC(nextchr))) { \ |
178 | sayNO; \ | |
ee9a90b8 | 179 | } \ |
28b98f76 | 180 | goto increment_locinput; |
980866de | 181 | |
ee9a90b8 KW |
182 | /* Handle the non-locale cases for a character class and its complement. It |
183 | * calls _CCC_TRY_CODE with a ! to complement the test for the character class. | |
184 | * This is because that code fails when the test succeeds, so we want to have | |
185 | * the test fail so that the code succeeds. The swash is stored in a | |
186 | * predictable PL_ place */ | |
cfaf538b KW |
187 | #define _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, \ |
188 | CLASS, STR) \ | |
ee9a90b8 KW |
189 | case NAME: \ |
190 | _CCC_TRY_CODE( !, FUNC, \ | |
191 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ | |
192 | (U8*)locinput, TRUE)), \ | |
193 | CLASS, STR) \ | |
194 | case NNAME: \ | |
1dcf4a1b | 195 | _CCC_TRY_CODE( PLACEHOLDER , FUNC, \ |
ee9a90b8 KW |
196 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ |
197 | (U8*)locinput, TRUE)), \ | |
198 | CLASS, STR) \ | |
199 | ||
200 | /* Generate the case statements for both locale and non-locale character | |
201 | * classes in regmatch for classes that don't have special unicode semantics. | |
202 | * Locales don't use an immediate swash, but an intermediary special locale | |
203 | * function that is called on the pointer to the current place in the input | |
204 | * string. That function will resolve to needing the same swash. One might | |
205 | * think that because we don't know what the locale will match, we shouldn't | |
206 | * check with the swash loading function that it loaded properly; ie, that we | |
207 | * should use LOAD_UTF8_CHARCLASS_NO_CHECK for those, but what is passed to the | |
208 | * regular LOAD_UTF8_CHARCLASS is in non-locale terms, and so locale is | |
209 | * irrelevant here */ | |
210 | #define CCC_TRY(NAME, NNAME, FUNC, \ | |
211 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
cfaf538b | 212 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 KW |
213 | CLASS, STR) \ |
214 | case NAMEL: \ | |
215 | PL_reg_flags |= RF_tainted; \ | |
216 | _CCC_TRY_CODE( !, LCFUNC, LCFUNC_utf8((U8*)locinput), CLASS, STR) \ | |
217 | case NNAMEL: \ | |
218 | PL_reg_flags |= RF_tainted; \ | |
1dcf4a1b KW |
219 | _CCC_TRY_CODE( PLACEHOLDER, LCFUNC, LCFUNC_utf8((U8*)locinput), \ |
220 | CLASS, STR) \ | |
cfaf538b KW |
221 | case NAMEA: \ |
222 | if (locinput >= PL_regeol || ! FUNCA(nextchr)) { \ | |
223 | sayNO; \ | |
224 | } \ | |
225 | /* Matched a utf8-invariant, so don't have to worry about utf8 */ \ | |
3640db6b | 226 | locinput++; \ |
cfaf538b KW |
227 | break; \ |
228 | case NNAMEA: \ | |
229 | if (locinput >= PL_regeol || FUNCA(nextchr)) { \ | |
230 | sayNO; \ | |
231 | } \ | |
28b98f76 | 232 | goto increment_locinput; \ |
ee9a90b8 KW |
233 | /* Generate the non-locale cases */ \ |
234 | _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, CLASS, STR) | |
235 | ||
236 | /* This is like CCC_TRY, but has an extra set of parameters for generating case | |
237 | * statements to handle separate Unicode semantics nodes */ | |
238 | #define CCC_TRY_U(NAME, NNAME, FUNC, \ | |
239 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
240 | NAMEU, NNAMEU, FUNCU, \ | |
cfaf538b | 241 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 | 242 | CLASS, STR) \ |
cfaf538b KW |
243 | CCC_TRY(NAME, NNAME, FUNC, \ |
244 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
245 | NAMEA, NNAMEA, FUNCA, \ | |
246 | CLASS, STR) \ | |
ee9a90b8 | 247 | _CCC_TRY_NONLOCALE(NAMEU, NNAMEU, FUNCU, CLASS, STR) |
d1eb3177 | 248 | |
3dab1dad YO |
249 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
250 | ||
5f80c4cf | 251 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
c35dcbe2 YO |
252 | /* it would be nice to rework regcomp.sym to generate this stuff. sigh |
253 | * | |
254 | * NOTE that *nothing* that affects backtracking should be in here, specifically | |
255 | * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a | |
256 | * node that is in between two EXACT like nodes when ascertaining what the required | |
257 | * "follow" character is. This should probably be moved to regex compile time | |
258 | * although it may be done at run time beause of the REF possibility - more | |
259 | * investigation required. -- demerphq | |
260 | */ | |
3e901dc0 YO |
261 | #define JUMPABLE(rn) ( \ |
262 | OP(rn) == OPEN || \ | |
263 | (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \ | |
264 | OP(rn) == EVAL || \ | |
cca55fe3 JP |
265 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
266 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
d1c771f5 | 267 | OP(rn) == KEEPS || \ |
3dab1dad | 268 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 | 269 | ) |
ee9b8eae | 270 | #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT) |
e2d8ce26 | 271 | |
ee9b8eae YO |
272 | #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF ) |
273 | ||
274 | #if 0 | |
275 | /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so | |
276 | we don't need this definition. */ | |
277 | #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) | |
fab2782b | 278 | #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 |
279 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) |
280 | ||
281 | #else | |
282 | /* ... so we use this as its faster. */ | |
283 | #define IS_TEXT(rn) ( OP(rn)==EXACT ) | |
fab2782b | 284 | #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFU_TRICKYFOLD || OP(rn) == EXACTFA) |
ee9b8eae YO |
285 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) |
286 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) | |
287 | ||
288 | #endif | |
e2d8ce26 | 289 | |
a84d97b6 HS |
290 | /* |
291 | Search for mandatory following text node; for lookahead, the text must | |
292 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
293 | */ | |
cca55fe3 | 294 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
295 | while (JUMPABLE(rn)) { \ |
296 | const OPCODE type = OP(rn); \ | |
297 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 298 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 299 | else if (type == PLUS) \ |
cca55fe3 | 300 | rn = NEXTOPER(rn); \ |
3dab1dad | 301 | else if (type == IFMATCH) \ |
a84d97b6 | 302 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 303 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 304 | } \ |
5f80c4cf | 305 | } STMT_END |
74750237 | 306 | |
c476f425 | 307 | |
acfe0abc | 308 | static void restore_pos(pTHX_ void *arg); |
51371543 | 309 | |
87c0511b | 310 | #define REGCP_PAREN_ELEMS 3 |
f067efbf | 311 | #define REGCP_OTHER_ELEMS 3 |
e0fa7e2b | 312 | #define REGCP_FRAME_ELEMS 1 |
620d5b66 NC |
313 | /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and |
314 | * are needed for the regexp context stack bookkeeping. */ | |
315 | ||
76e3520e | 316 | STATIC CHECKPOINT |
b93070ed | 317 | S_regcppush(pTHX_ const regexp *rex, I32 parenfloor) |
a0d0e21e | 318 | { |
97aff369 | 319 | dVAR; |
a3b680e6 | 320 | const int retval = PL_savestack_ix; |
a3b680e6 | 321 | const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; |
e0fa7e2b NC |
322 | const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS; |
323 | const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT; | |
87c0511b | 324 | I32 p; |
40a82448 | 325 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 326 | |
b93070ed DM |
327 | PERL_ARGS_ASSERT_REGCPPUSH; |
328 | ||
e49a9654 | 329 | if (paren_elems_to_push < 0) |
5637ef5b NC |
330 | Perl_croak(aTHX_ "panic: paren_elems_to_push, %i < 0", |
331 | paren_elems_to_push); | |
e49a9654 | 332 | |
e0fa7e2b NC |
333 | if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems) |
334 | Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf | |
5df417d0 JH |
335 | " out of range (%lu-%ld)", |
336 | total_elems, (unsigned long)PL_regsize, (long)parenfloor); | |
e0fa7e2b | 337 | |
620d5b66 | 338 | SSGROW(total_elems + REGCP_FRAME_ELEMS); |
7f69552c | 339 | |
495f47a5 DM |
340 | DEBUG_BUFFERS_r( |
341 | if ((int)PL_regsize > (int)parenfloor) | |
342 | PerlIO_printf(Perl_debug_log, | |
343 | "rex=0x%"UVxf" offs=0x%"UVxf": saving capture indices:\n", | |
344 | PTR2UV(rex), | |
345 | PTR2UV(rex->offs) | |
346 | ); | |
347 | ); | |
87c0511b | 348 | for (p = parenfloor+1; p <= (I32)PL_regsize; p++) { |
b1ce53c5 | 349 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
b93070ed DM |
350 | SSPUSHINT(rex->offs[p].end); |
351 | SSPUSHINT(rex->offs[p].start); | |
1ca2007e | 352 | SSPUSHINT(rex->offs[p].start_tmp); |
e7707071 | 353 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
495f47a5 DM |
354 | " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"\n", |
355 | (UV)p, | |
356 | (IV)rex->offs[p].start, | |
357 | (IV)rex->offs[p].start_tmp, | |
358 | (IV)rex->offs[p].end | |
40a82448 | 359 | )); |
a0d0e21e | 360 | } |
b1ce53c5 | 361 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
3280af22 | 362 | SSPUSHINT(PL_regsize); |
b93070ed DM |
363 | SSPUSHINT(rex->lastparen); |
364 | SSPUSHINT(rex->lastcloseparen); | |
e0fa7e2b | 365 | SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */ |
41123dfd | 366 | |
a0d0e21e LW |
367 | return retval; |
368 | } | |
369 | ||
c277df42 | 370 | /* These are needed since we do not localize EVAL nodes: */ |
ab3bbdeb YO |
371 | #define REGCP_SET(cp) \ |
372 | DEBUG_STATE_r( \ | |
ab3bbdeb | 373 | PerlIO_printf(Perl_debug_log, \ |
e4f74956 | 374 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
ab3bbdeb YO |
375 | (IV)PL_savestack_ix)); \ |
376 | cp = PL_savestack_ix | |
c3464db5 | 377 | |
ab3bbdeb | 378 | #define REGCP_UNWIND(cp) \ |
e4f74956 | 379 | DEBUG_STATE_r( \ |
ab3bbdeb | 380 | if (cp != PL_savestack_ix) \ |
e4f74956 YO |
381 | PerlIO_printf(Perl_debug_log, \ |
382 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ | |
ab3bbdeb YO |
383 | (IV)(cp), (IV)PL_savestack_ix)); \ |
384 | regcpblow(cp) | |
c277df42 | 385 | |
a8d1f4b4 DM |
386 | #define UNWIND_PAREN(lp, lcp) \ |
387 | for (n = rex->lastparen; n > lp; n--) \ | |
388 | rex->offs[n].end = -1; \ | |
389 | rex->lastparen = n; \ | |
390 | rex->lastcloseparen = lcp; | |
391 | ||
392 | ||
f067efbf | 393 | STATIC void |
b93070ed | 394 | S_regcppop(pTHX_ regexp *rex) |
a0d0e21e | 395 | { |
97aff369 | 396 | dVAR; |
e0fa7e2b | 397 | UV i; |
87c0511b | 398 | U32 paren; |
a3621e74 YO |
399 | GET_RE_DEBUG_FLAGS_DECL; |
400 | ||
7918f24d NC |
401 | PERL_ARGS_ASSERT_REGCPPOP; |
402 | ||
b1ce53c5 | 403 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
c6bf6a65 | 404 | i = SSPOPUV; |
e0fa7e2b NC |
405 | assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
406 | i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */ | |
b93070ed DM |
407 | rex->lastcloseparen = SSPOPINT; |
408 | rex->lastparen = SSPOPINT; | |
3280af22 | 409 | PL_regsize = SSPOPINT; |
b1ce53c5 | 410 | |
620d5b66 | 411 | i -= REGCP_OTHER_ELEMS; |
b1ce53c5 | 412 | /* Now restore the parentheses context. */ |
495f47a5 DM |
413 | DEBUG_BUFFERS_r( |
414 | if (i || rex->lastparen + 1 <= rex->nparens) | |
415 | PerlIO_printf(Perl_debug_log, | |
416 | "rex=0x%"UVxf" offs=0x%"UVxf": restoring capture indices to:\n", | |
417 | PTR2UV(rex), | |
418 | PTR2UV(rex->offs) | |
419 | ); | |
420 | ); | |
87c0511b | 421 | paren = PL_regsize; |
620d5b66 | 422 | for ( ; i > 0; i -= REGCP_PAREN_ELEMS) { |
1df70142 | 423 | I32 tmps; |
1ca2007e | 424 | rex->offs[paren].start_tmp = SSPOPINT; |
b93070ed | 425 | rex->offs[paren].start = SSPOPINT; |
cf93c79d | 426 | tmps = SSPOPINT; |
b93070ed DM |
427 | if (paren <= rex->lastparen) |
428 | rex->offs[paren].end = tmps; | |
495f47a5 DM |
429 | DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log, |
430 | " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"%s\n", | |
431 | (UV)paren, | |
432 | (IV)rex->offs[paren].start, | |
433 | (IV)rex->offs[paren].start_tmp, | |
434 | (IV)rex->offs[paren].end, | |
435 | (paren > rex->lastparen ? "(skipped)" : "")); | |
c277df42 | 436 | ); |
87c0511b | 437 | paren--; |
a0d0e21e | 438 | } |
daf18116 | 439 | #if 1 |
dafc8851 JH |
440 | /* It would seem that the similar code in regtry() |
441 | * already takes care of this, and in fact it is in | |
442 | * a better location to since this code can #if 0-ed out | |
443 | * but the code in regtry() is needed or otherwise tests | |
444 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
445 | * (as of patchlevel 7877) will fail. Then again, |
446 | * this code seems to be necessary or otherwise | |
225593e1 DM |
447 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ |
448 | * --jhi updated by dapm */ | |
b93070ed | 449 | for (i = rex->lastparen + 1; i <= rex->nparens; i++) { |
097eb12c | 450 | if (i > PL_regsize) |
b93070ed DM |
451 | rex->offs[i].start = -1; |
452 | rex->offs[i].end = -1; | |
495f47a5 DM |
453 | DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log, |
454 | " \\%"UVuf": %s ..-1 undeffing\n", | |
455 | (UV)i, | |
456 | (i > PL_regsize) ? "-1" : " " | |
457 | )); | |
a0d0e21e | 458 | } |
dafc8851 | 459 | #endif |
a0d0e21e LW |
460 | } |
461 | ||
74088413 DM |
462 | /* restore the parens and associated vars at savestack position ix, |
463 | * but without popping the stack */ | |
464 | ||
465 | STATIC void | |
466 | S_regcp_restore(pTHX_ regexp *rex, I32 ix) | |
467 | { | |
468 | I32 tmpix = PL_savestack_ix; | |
469 | PL_savestack_ix = ix; | |
470 | regcppop(rex); | |
471 | PL_savestack_ix = tmpix; | |
472 | } | |
473 | ||
02db2b7b | 474 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 475 | |
a687059c | 476 | /* |
e50aee73 | 477 | * pregexec and friends |
a687059c LW |
478 | */ |
479 | ||
76234dfb | 480 | #ifndef PERL_IN_XSUB_RE |
a687059c | 481 | /* |
c277df42 | 482 | - pregexec - match a regexp against a string |
a687059c | 483 | */ |
c277df42 | 484 | I32 |
49d7dfbc | 485 | Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend, |
c3464db5 | 486 | char *strbeg, I32 minend, SV *screamer, U32 nosave) |
8fd1a950 DM |
487 | /* stringarg: the point in the string at which to begin matching */ |
488 | /* strend: pointer to null at end of string */ | |
489 | /* strbeg: real beginning of string */ | |
490 | /* minend: end of match must be >= minend bytes after stringarg. */ | |
491 | /* screamer: SV being matched: only used for utf8 flag, pos() etc; string | |
492 | * itself is accessed via the pointers above */ | |
493 | /* nosave: For optimizations. */ | |
c277df42 | 494 | { |
7918f24d NC |
495 | PERL_ARGS_ASSERT_PREGEXEC; |
496 | ||
c277df42 | 497 | return |
9041c2e3 | 498 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
499 | nosave ? 0 : REXEC_COPY_STR); |
500 | } | |
76234dfb | 501 | #endif |
22e551b9 | 502 | |
9041c2e3 | 503 | /* |
cad2e5aa JH |
504 | * Need to implement the following flags for reg_anch: |
505 | * | |
506 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
507 | * USE_INTUIT_ML | |
508 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
509 | * INTUIT_AUTORITATIVE_ML | |
510 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
511 | * INTUIT_ONCE_ML | |
512 | * | |
513 | * Another flag for this function: SECOND_TIME (so that float substrs | |
514 | * with giant delta may be not rechecked). | |
515 | */ | |
516 | ||
517 | /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ | |
518 | ||
3f7c398e | 519 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
520 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
521 | ||
522 | /* XXXX We assume that strpos is strbeg unless sv. */ | |
523 | ||
6eb5f6b9 JH |
524 | /* XXXX Some places assume that there is a fixed substring. |
525 | An update may be needed if optimizer marks as "INTUITable" | |
526 | RExen without fixed substrings. Similarly, it is assumed that | |
527 | lengths of all the strings are no more than minlen, thus they | |
528 | cannot come from lookahead. | |
40d049e4 YO |
529 | (Or minlen should take into account lookahead.) |
530 | NOTE: Some of this comment is not correct. minlen does now take account | |
531 | of lookahead/behind. Further research is required. -- demerphq | |
532 | ||
533 | */ | |
6eb5f6b9 | 534 | |
2c2d71f5 JH |
535 | /* A failure to find a constant substring means that there is no need to make |
536 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
537 | finding a substring too deep into the string means that less calls to | |
30944b6d IZ |
538 | regtry() should be needed. |
539 | ||
540 | REx compiler's optimizer found 4 possible hints: | |
541 | a) Anchored substring; | |
542 | b) Fixed substring; | |
543 | c) Whether we are anchored (beginning-of-line or \G); | |
486ec47a | 544 | d) First node (of those at offset 0) which may distinguish positions; |
6eb5f6b9 | 545 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
546 | string which does not contradict any of them. |
547 | */ | |
2c2d71f5 | 548 | |
6eb5f6b9 JH |
549 | /* Most of decisions we do here should have been done at compile time. |
550 | The nodes of the REx which we used for the search should have been | |
551 | deleted from the finite automaton. */ | |
552 | ||
cad2e5aa | 553 | char * |
288b8c02 | 554 | Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, |
9f61653a | 555 | char *strend, const U32 flags, re_scream_pos_data *data) |
cad2e5aa | 556 | { |
97aff369 | 557 | dVAR; |
288b8c02 | 558 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
eb578fdb | 559 | I32 start_shift = 0; |
cad2e5aa | 560 | /* Should be nonnegative! */ |
eb578fdb KW |
561 | I32 end_shift = 0; |
562 | char *s; | |
563 | SV *check; | |
a1933d95 | 564 | char *strbeg; |
cad2e5aa | 565 | char *t; |
f2ed9b32 | 566 | const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 567 | I32 ml_anch; |
eb578fdb | 568 | char *other_last = NULL; /* other substr checked before this */ |
bd61b366 | 569 | char *check_at = NULL; /* check substr found at this pos */ |
d8080198 | 570 | char *checked_upto = NULL; /* how far into the string we have already checked using find_byclass*/ |
bbe252da | 571 | const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; |
f8fc2ecf | 572 | RXi_GET_DECL(prog,progi); |
30944b6d | 573 | #ifdef DEBUGGING |
890ce7af | 574 | const char * const i_strpos = strpos; |
30944b6d | 575 | #endif |
a3621e74 YO |
576 | GET_RE_DEBUG_FLAGS_DECL; |
577 | ||
7918f24d | 578 | PERL_ARGS_ASSERT_RE_INTUIT_START; |
c33e64f0 FC |
579 | PERL_UNUSED_ARG(flags); |
580 | PERL_UNUSED_ARG(data); | |
7918f24d | 581 | |
f2ed9b32 | 582 | RX_MATCH_UTF8_set(rx,utf8_target); |
cad2e5aa | 583 | |
3c8556c3 | 584 | if (RX_UTF8(rx)) { |
b8d68ded JH |
585 | PL_reg_flags |= RF_utf8; |
586 | } | |
ab3bbdeb | 587 | DEBUG_EXECUTE_r( |
f2ed9b32 | 588 | debug_start_match(rx, utf8_target, strpos, strend, |
1de06328 YO |
589 | sv ? "Guessing start of match in sv for" |
590 | : "Guessing start of match in string for"); | |
2a782b5b | 591 | ); |
cad2e5aa | 592 | |
c344f387 JH |
593 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
594 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 595 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 596 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 597 | goto fail; |
2c2d71f5 | 598 | } |
1de06328 | 599 | |
a1933d95 | 600 | strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; |
1aa99e6b | 601 | PL_regeol = strend; |
f2ed9b32 | 602 | if (utf8_target) { |
33b8afdf JH |
603 | if (!prog->check_utf8 && prog->check_substr) |
604 | to_utf8_substr(prog); | |
605 | check = prog->check_utf8; | |
606 | } else { | |
607 | if (!prog->check_substr && prog->check_utf8) | |
608 | to_byte_substr(prog); | |
609 | check = prog->check_substr; | |
610 | } | |
1de06328 | 611 | if (check == &PL_sv_undef) { |
a3621e74 | 612 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
1de06328 | 613 | "Non-utf8 string cannot match utf8 check string\n")); |
33b8afdf JH |
614 | goto fail; |
615 | } | |
bbe252da YO |
616 | if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */ |
617 | ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) | |
618 | || ( (prog->extflags & RXf_ANCH_BOL) | |
7fba1cd6 | 619 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 620 | |
7e25d62c | 621 | if (!ml_anch) { |
bbe252da YO |
622 | if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */ |
623 | && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ | |
3f7c398e | 624 | /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ |
7e25d62c JH |
625 | && sv && !SvROK(sv) |
626 | && (strpos != strbeg)) { | |
a3621e74 | 627 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
628 | goto fail; |
629 | } | |
630 | if (prog->check_offset_min == prog->check_offset_max && | |
bbe252da | 631 | !(prog->extflags & RXf_CANY_SEEN)) { |
2c2d71f5 | 632 | /* Substring at constant offset from beg-of-str... */ |
cad2e5aa JH |
633 | I32 slen; |
634 | ||
1aa99e6b | 635 | s = HOP3c(strpos, prog->check_offset_min, strend); |
1de06328 | 636 | |
653099ff GS |
637 | if (SvTAIL(check)) { |
638 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 639 | |
9041c2e3 | 640 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 641 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 642 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 643 | goto fail_finish; |
cad2e5aa JH |
644 | } |
645 | /* Now should match s[0..slen-2] */ | |
646 | slen--; | |
3f7c398e | 647 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 648 | || (slen > 1 |
3f7c398e | 649 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 650 | report_neq: |
a3621e74 | 651 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
652 | goto fail_finish; |
653 | } | |
cad2e5aa | 654 | } |
3f7c398e | 655 | else if (*SvPVX_const(check) != *s |
653099ff | 656 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 657 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 658 | goto report_neq; |
c315bfe8 | 659 | check_at = s; |
2c2d71f5 | 660 | goto success_at_start; |
7e25d62c | 661 | } |
cad2e5aa | 662 | } |
2c2d71f5 | 663 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 664 | s = strpos; |
2c2d71f5 | 665 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
1de06328 YO |
666 | end_shift = prog->check_end_shift; |
667 | ||
2c2d71f5 | 668 | if (!ml_anch) { |
a3b680e6 | 669 | const I32 end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 670 | - (SvTAIL(check) != 0); |
a3b680e6 | 671 | const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
672 | |
673 | if (end_shift < eshift) | |
674 | end_shift = eshift; | |
675 | } | |
cad2e5aa | 676 | } |
2c2d71f5 | 677 | else { /* Can match at random position */ |
cad2e5aa JH |
678 | ml_anch = 0; |
679 | s = strpos; | |
1de06328 YO |
680 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
681 | end_shift = prog->check_end_shift; | |
682 | ||
683 | /* end shift should be non negative here */ | |
cad2e5aa JH |
684 | } |
685 | ||
bcdf7404 | 686 | #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 687 | if (end_shift < 0) |
1de06328 | 688 | Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", |
220fc49f | 689 | (IV)end_shift, RX_PRECOMP(prog)); |
2c2d71f5 JH |
690 | #endif |
691 | ||
2c2d71f5 JH |
692 | restart: |
693 | /* Find a possible match in the region s..strend by looking for | |
694 | the "check" substring in the region corrected by start/end_shift. */ | |
1de06328 YO |
695 | |
696 | { | |
697 | I32 srch_start_shift = start_shift; | |
698 | I32 srch_end_shift = end_shift; | |
c33e64f0 FC |
699 | U8* start_point; |
700 | U8* end_point; | |
1de06328 YO |
701 | if (srch_start_shift < 0 && strbeg - s > srch_start_shift) { |
702 | srch_end_shift -= ((strbeg - s) - srch_start_shift); | |
703 | srch_start_shift = strbeg - s; | |
704 | } | |
6bda09f9 | 705 | DEBUG_OPTIMISE_MORE_r({ |
1de06328 YO |
706 | PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n", |
707 | (IV)prog->check_offset_min, | |
708 | (IV)srch_start_shift, | |
709 | (IV)srch_end_shift, | |
710 | (IV)prog->check_end_shift); | |
711 | }); | |
712 | ||
bbe252da | 713 | if (prog->extflags & RXf_CANY_SEEN) { |
1de06328 YO |
714 | start_point= (U8*)(s + srch_start_shift); |
715 | end_point= (U8*)(strend - srch_end_shift); | |
716 | } else { | |
717 | start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); | |
718 | end_point= HOP3(strend, -srch_end_shift, strbeg); | |
719 | } | |
6bda09f9 | 720 | DEBUG_OPTIMISE_MORE_r({ |
56570a2c | 721 | PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", |
1de06328 | 722 | (int)(end_point - start_point), |
fc8cd66c | 723 | (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), |
1de06328 YO |
724 | start_point); |
725 | }); | |
726 | ||
727 | s = fbm_instr( start_point, end_point, | |
7fba1cd6 | 728 | check, multiline ? FBMrf_MULTILINE : 0); |
1de06328 | 729 | } |
cad2e5aa JH |
730 | /* Update the count-of-usability, remove useless subpatterns, |
731 | unshift s. */ | |
2c2d71f5 | 732 | |
ab3bbdeb | 733 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 734 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
735 | SvPVX_const(check), RE_SV_DUMPLEN(check), 30); |
736 | PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", | |
2c2d71f5 | 737 | (s ? "Found" : "Did not find"), |
f2ed9b32 | 738 | (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) |
ab3bbdeb YO |
739 | ? "anchored" : "floating"), |
740 | quoted, | |
741 | RE_SV_TAIL(check), | |
742 | (s ? " at offset " : "...\n") ); | |
743 | }); | |
2c2d71f5 JH |
744 | |
745 | if (!s) | |
746 | goto fail_finish; | |
2c2d71f5 | 747 | /* Finish the diagnostic message */ |
a3621e74 | 748 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 | 749 | |
1de06328 YO |
750 | /* XXX dmq: first branch is for positive lookbehind... |
751 | Our check string is offset from the beginning of the pattern. | |
752 | So we need to do any stclass tests offset forward from that | |
753 | point. I think. :-( | |
754 | */ | |
755 | ||
756 | ||
757 | ||
758 | check_at=s; | |
759 | ||
760 | ||
2c2d71f5 JH |
761 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. |
762 | Start with the other substr. | |
763 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 764 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
765 | *always* match. Probably should be marked during compile... |
766 | Probably it is right to do no SCREAM here... | |
767 | */ | |
768 | ||
f2ed9b32 | 769 | if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8) |
1de06328 YO |
770 | : (prog->float_substr && prog->anchored_substr)) |
771 | { | |
30944b6d | 772 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
773 | /* XXXX May be hopelessly wrong for UTF... */ |
774 | if (!other_last) | |
6eb5f6b9 | 775 | other_last = strpos; |
f2ed9b32 | 776 | if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
777 | do_other_anchored: |
778 | { | |
890ce7af AL |
779 | char * const last = HOP3c(s, -start_shift, strbeg); |
780 | char *last1, *last2; | |
be8e71aa | 781 | char * const saved_s = s; |
33b8afdf | 782 | SV* must; |
2c2d71f5 | 783 | |
2c2d71f5 JH |
784 | t = s - prog->check_offset_max; |
785 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
f2ed9b32 | 786 | && (!utf8_target |
0ce71af7 | 787 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 788 | && t > strpos))) |
6f207bd3 | 789 | NOOP; |
2c2d71f5 JH |
790 | else |
791 | t = strpos; | |
1aa99e6b | 792 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
793 | if (t < other_last) /* These positions already checked */ |
794 | t = other_last; | |
1aa99e6b | 795 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
796 | if (last < last1) |
797 | last1 = last; | |
1de06328 YO |
798 | /* XXXX It is not documented what units *_offsets are in. |
799 | We assume bytes, but this is clearly wrong. | |
800 | Meaning this code needs to be carefully reviewed for errors. | |
801 | dmq. | |
802 | */ | |
803 | ||
2c2d71f5 | 804 | /* On end-of-str: see comment below. */ |
f2ed9b32 | 805 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; |
33b8afdf JH |
806 | if (must == &PL_sv_undef) { |
807 | s = (char*)NULL; | |
1de06328 | 808 | DEBUG_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
809 | } |
810 | else | |
811 | s = fbm_instr( | |
812 | (unsigned char*)t, | |
813 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
814 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
815 | must, | |
7fba1cd6 | 816 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 817 | ); |
ab3bbdeb | 818 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 819 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
820 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
821 | PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", | |
2c2d71f5 | 822 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
823 | quoted, RE_SV_TAIL(must)); |
824 | }); | |
825 | ||
826 | ||
2c2d71f5 JH |
827 | if (!s) { |
828 | if (last1 >= last2) { | |
a3621e74 | 829 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
830 | ", giving up...\n")); |
831 | goto fail_finish; | |
832 | } | |
a3621e74 | 833 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 834 | ", trying floating at offset %ld...\n", |
be8e71aa | 835 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
836 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
837 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
838 | goto restart; |
839 | } | |
840 | else { | |
a3621e74 | 841 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 842 | (long)(s - i_strpos))); |
1aa99e6b IH |
843 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
844 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 845 | s = saved_s; |
2c2d71f5 JH |
846 | if (t == strpos) |
847 | goto try_at_start; | |
2c2d71f5 JH |
848 | goto try_at_offset; |
849 | } | |
30944b6d | 850 | } |
2c2d71f5 JH |
851 | } |
852 | else { /* Take into account the floating substring. */ | |
33b8afdf | 853 | char *last, *last1; |
be8e71aa | 854 | char * const saved_s = s; |
33b8afdf JH |
855 | SV* must; |
856 | ||
857 | t = HOP3c(s, -start_shift, strbeg); | |
858 | last1 = last = | |
859 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
860 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
861 | last = HOP3c(t, prog->float_max_offset, strend); | |
862 | s = HOP3c(t, prog->float_min_offset, strend); | |
863 | if (s < other_last) | |
864 | s = other_last; | |
2c2d71f5 | 865 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
f2ed9b32 | 866 | must = utf8_target ? prog->float_utf8 : prog->float_substr; |
33b8afdf JH |
867 | /* fbm_instr() takes into account exact value of end-of-str |
868 | if the check is SvTAIL(ed). Since false positives are OK, | |
869 | and end-of-str is not later than strend we are OK. */ | |
870 | if (must == &PL_sv_undef) { | |
871 | s = (char*)NULL; | |
1de06328 | 872 | DEBUG_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
873 | } |
874 | else | |
2c2d71f5 | 875 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
876 | (unsigned char*)last + SvCUR(must) |
877 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 878 | must, multiline ? FBMrf_MULTILINE : 0); |
ab3bbdeb | 879 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 880 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
881 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
882 | PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", | |
33b8afdf | 883 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
884 | quoted, RE_SV_TAIL(must)); |
885 | }); | |
33b8afdf JH |
886 | if (!s) { |
887 | if (last1 == last) { | |
a3621e74 | 888 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
889 | ", giving up...\n")); |
890 | goto fail_finish; | |
2c2d71f5 | 891 | } |
a3621e74 | 892 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 893 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 894 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
895 | other_last = last; |
896 | s = HOP3c(t, 1, strend); | |
897 | goto restart; | |
898 | } | |
899 | else { | |
a3621e74 | 900 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
901 | (long)(s - i_strpos))); |
902 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 903 | s = saved_s; |
33b8afdf JH |
904 | if (t == strpos) |
905 | goto try_at_start; | |
906 | goto try_at_offset; | |
907 | } | |
2c2d71f5 | 908 | } |
cad2e5aa | 909 | } |
2c2d71f5 | 910 | |
1de06328 | 911 | |
9ef43ace | 912 | t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); |
1de06328 | 913 | |
6bda09f9 | 914 | DEBUG_OPTIMISE_MORE_r( |
1de06328 YO |
915 | PerlIO_printf(Perl_debug_log, |
916 | "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", | |
917 | (IV)prog->check_offset_min, | |
918 | (IV)prog->check_offset_max, | |
919 | (IV)(s-strpos), | |
920 | (IV)(t-strpos), | |
921 | (IV)(t-s), | |
922 | (IV)(strend-strpos) | |
923 | ) | |
924 | ); | |
925 | ||
2c2d71f5 | 926 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
f2ed9b32 | 927 | && (!utf8_target |
9ef43ace | 928 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) |
1de06328 YO |
929 | && t > strpos))) |
930 | { | |
2c2d71f5 JH |
931 | /* Fixed substring is found far enough so that the match |
932 | cannot start at strpos. */ | |
933 | try_at_offset: | |
cad2e5aa | 934 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
935 | /* Eventually fbm_*() should handle this, but often |
936 | anchored_offset is not 0, so this check will not be wasted. */ | |
937 | /* XXXX In the code below we prefer to look for "^" even in | |
938 | presence of anchored substrings. And we search even | |
939 | beyond the found float position. These pessimizations | |
940 | are historical artefacts only. */ | |
941 | find_anchor: | |
2c2d71f5 | 942 | while (t < strend - prog->minlen) { |
cad2e5aa | 943 | if (*t == '\n') { |
4ee3650e | 944 | if (t < check_at - prog->check_offset_min) { |
f2ed9b32 | 945 | if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
946 | /* Since we moved from the found position, |
947 | we definitely contradict the found anchored | |
30944b6d IZ |
948 | substr. Due to the above check we do not |
949 | contradict "check" substr. | |
950 | Thus we can arrive here only if check substr | |
951 | is float. Redo checking for "other"=="fixed". | |
952 | */ | |
9041c2e3 | 953 | strpos = t + 1; |
a3621e74 | 954 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 955 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
956 | goto do_other_anchored; |
957 | } | |
4ee3650e GS |
958 | /* We don't contradict the found floating substring. */ |
959 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 960 | s = t + 1; |
a3621e74 | 961 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 962 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
963 | goto set_useful; |
964 | } | |
4ee3650e GS |
965 | /* Position contradicts check-string */ |
966 | /* XXXX probably better to look for check-string | |
967 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 968 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 969 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 970 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
971 | goto restart; |
972 | } | |
973 | t++; | |
974 | } | |
a3621e74 | 975 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 976 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 977 | goto fail_finish; |
cad2e5aa | 978 | } |
f5952150 | 979 | else { |
a3621e74 | 980 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 981 | PL_colors[0], PL_colors[1])); |
f5952150 | 982 | } |
cad2e5aa JH |
983 | s = t; |
984 | set_useful: | |
f2ed9b32 | 985 | ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
986 | } |
987 | else { | |
f5952150 | 988 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 989 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
990 | unless it was an MBOL and we are not after MBOL, |
991 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
992 | try_at_start: |
993 | /* Even in this situation we may use MBOL flag if strpos is offset | |
994 | wrt the start of the string. */ | |
05b4157f | 995 | if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ |
a1933d95 | 996 | && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d | 997 | /* May be due to an implicit anchor of m{.*foo} */ |
bbe252da | 998 | && !(prog->intflags & PREGf_IMPLICIT)) |
d506a20d | 999 | { |
cad2e5aa JH |
1000 | t = strpos; |
1001 | goto find_anchor; | |
1002 | } | |
a3621e74 | 1003 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 1004 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
70685ca0 | 1005 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 1006 | ); |
2c2d71f5 | 1007 | success_at_start: |
bbe252da | 1008 | if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ |
f2ed9b32 | 1009 | && (utf8_target ? ( |
33b8afdf JH |
1010 | prog->check_utf8 /* Could be deleted already */ |
1011 | && --BmUSEFUL(prog->check_utf8) < 0 | |
1012 | && (prog->check_utf8 == prog->float_utf8) | |
1013 | ) : ( | |
1014 | prog->check_substr /* Could be deleted already */ | |
1015 | && --BmUSEFUL(prog->check_substr) < 0 | |
1016 | && (prog->check_substr == prog->float_substr) | |
1017 | ))) | |
66e933ab | 1018 | { |
cad2e5aa | 1019 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 1020 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
f2ed9b32 KW |
1021 | /* XXX Does the destruction order has to change with utf8_target? */ |
1022 | SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr); | |
1023 | SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
1024 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
1025 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
1026 | check = NULL; /* abort */ | |
cad2e5aa | 1027 | s = strpos; |
486ec47a | 1028 | /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag |
c9415951 YO |
1029 | see http://bugs.activestate.com/show_bug.cgi?id=87173 */ |
1030 | if (prog->intflags & PREGf_IMPLICIT) | |
1031 | prog->extflags &= ~RXf_ANCH_MBOL; | |
3cf5c195 IZ |
1032 | /* XXXX This is a remnant of the old implementation. It |
1033 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 1034 | other heuristics. */ |
bbe252da | 1035 | prog->extflags &= ~RXf_USE_INTUIT; |
c9415951 | 1036 | /* XXXX What other flags might need to be cleared in this branch? */ |
cad2e5aa JH |
1037 | } |
1038 | else | |
1039 | s = strpos; | |
1040 | } | |
1041 | ||
6eb5f6b9 JH |
1042 | /* Last resort... */ |
1043 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
1de06328 YO |
1044 | /* trie stclasses are too expensive to use here, we are better off to |
1045 | leave it to regmatch itself */ | |
f8fc2ecf | 1046 | if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { |
6eb5f6b9 JH |
1047 | /* minlen == 0 is possible if regstclass is \b or \B, |
1048 | and the fixed substr is ''$. | |
1049 | Since minlen is already taken into account, s+1 is before strend; | |
1050 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
1051 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
1052 | regstclass does not come from lookahead... */ | |
1053 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
af944926 | 1054 | This leaves EXACTF-ish only, which are dealt with in find_byclass(). */ |
f8fc2ecf YO |
1055 | const U8* const str = (U8*)STRING(progi->regstclass); |
1056 | const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT | |
1057 | ? CHR_DIST(str+STR_LEN(progi->regstclass), str) | |
66e933ab | 1058 | : 1); |
1de06328 YO |
1059 | char * endpos; |
1060 | if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) | |
1061 | endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); | |
1062 | else if (prog->float_substr || prog->float_utf8) | |
1063 | endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); | |
1064 | else | |
1065 | endpos= strend; | |
1066 | ||
d8080198 YO |
1067 | if (checked_upto < s) |
1068 | checked_upto = s; | |
1069 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n", | |
1070 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg))); | |
1071 | ||
6eb5f6b9 | 1072 | t = s; |
d8080198 YO |
1073 | s = find_byclass(prog, progi->regstclass, checked_upto, endpos, NULL); |
1074 | if (s) { | |
1075 | checked_upto = s; | |
1076 | } else { | |
6eb5f6b9 | 1077 | #ifdef DEBUGGING |
cbbf8932 | 1078 | const char *what = NULL; |
6eb5f6b9 JH |
1079 | #endif |
1080 | if (endpos == strend) { | |
a3621e74 | 1081 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
1082 | "Could not match STCLASS...\n") ); |
1083 | goto fail; | |
1084 | } | |
a3621e74 | 1085 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 1086 | "This position contradicts STCLASS...\n") ); |
bbe252da | 1087 | if ((prog->extflags & RXf_ANCH) && !ml_anch) |
653099ff | 1088 | goto fail; |
d8080198 YO |
1089 | checked_upto = HOPBACKc(endpos, start_shift); |
1090 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n", | |
1091 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg))); | |
6eb5f6b9 | 1092 | /* Contradict one of substrings */ |
33b8afdf | 1093 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 | 1094 | if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) { |
a3621e74 | 1095 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 1096 | hop_and_restart: |
1aa99e6b | 1097 | s = HOP3c(t, 1, strend); |
66e933ab GS |
1098 | if (s + start_shift + end_shift > strend) { |
1099 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 1100 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
1101 | "Could not match STCLASS...\n") ); |
1102 | goto fail; | |
1103 | } | |
5e39e1e5 HS |
1104 | if (!check) |
1105 | goto giveup; | |
a3621e74 | 1106 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1107 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1108 | what, (long)(s + start_shift - i_strpos)) ); |
1109 | goto restart; | |
1110 | } | |
66e933ab | 1111 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
1112 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
1113 | goto retry_floating_check; | |
1114 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 1115 | s = check_at; |
5e39e1e5 HS |
1116 | if (!check) |
1117 | goto giveup; | |
a3621e74 | 1118 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1119 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1120 | (long)(other_last - i_strpos)) ); |
1121 | goto do_other_anchored; | |
1122 | } | |
60e71179 GS |
1123 | /* Another way we could have checked stclass at the |
1124 | current position only: */ | |
1125 | if (ml_anch) { | |
1126 | s = t = t + 1; | |
5e39e1e5 HS |
1127 | if (!check) |
1128 | goto giveup; | |
a3621e74 | 1129 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1130 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 1131 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 1132 | goto try_at_offset; |
66e933ab | 1133 | } |
f2ed9b32 | 1134 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 1135 | goto fail; |
486ec47a | 1136 | /* Check is floating substring. */ |
6eb5f6b9 JH |
1137 | retry_floating_check: |
1138 | t = check_at - start_shift; | |
a3621e74 | 1139 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
1140 | goto hop_and_restart; |
1141 | } | |
b7953727 | 1142 | if (t != s) { |
a3621e74 | 1143 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 1144 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
1145 | (long)(t - i_strpos), (long)(s - i_strpos)) |
1146 | ); | |
1147 | } | |
1148 | else { | |
a3621e74 | 1149 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1150 | "Does not contradict STCLASS...\n"); |
1151 | ); | |
1152 | } | |
6eb5f6b9 | 1153 | } |
5e39e1e5 | 1154 | giveup: |
a3621e74 | 1155 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
1156 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
1157 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 1158 | return s; |
2c2d71f5 JH |
1159 | |
1160 | fail_finish: /* Substring not found */ | |
33b8afdf | 1161 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
f2ed9b32 | 1162 | BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ |
cad2e5aa | 1163 | fail: |
a3621e74 | 1164 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 1165 | PL_colors[4], PL_colors[5])); |
bd61b366 | 1166 | return NULL; |
cad2e5aa | 1167 | } |
9661b544 | 1168 | |
a0a388a1 YO |
1169 | #define DECL_TRIE_TYPE(scan) \ |
1170 | const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \ | |
fab2782b YO |
1171 | trie_type = ((scan->flags == EXACT) \ |
1172 | ? (utf8_target ? trie_utf8 : trie_plain) \ | |
1173 | : (utf8_target ? trie_utf8_fold : trie_latin_utf8_fold)) | |
1174 | ||
1175 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ | |
1176 | uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ | |
1177 | STRLEN skiplen; \ | |
1178 | switch (trie_type) { \ | |
1179 | case trie_utf8_fold: \ | |
1180 | if ( foldlen>0 ) { \ | |
1181 | uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1182 | foldlen -= len; \ | |
1183 | uscan += len; \ | |
1184 | len=0; \ | |
1185 | } else { \ | |
1186 | uvc = to_utf8_fold( (const U8*) uc, foldbuf, &foldlen ); \ | |
1187 | len = UTF8SKIP(uc); \ | |
1188 | skiplen = UNISKIP( uvc ); \ | |
1189 | foldlen -= skiplen; \ | |
1190 | uscan = foldbuf + skiplen; \ | |
1191 | } \ | |
1192 | break; \ | |
1193 | case trie_latin_utf8_fold: \ | |
1194 | if ( foldlen>0 ) { \ | |
1195 | uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1196 | foldlen -= len; \ | |
1197 | uscan += len; \ | |
1198 | len=0; \ | |
1199 | } else { \ | |
1200 | len = 1; \ | |
1201 | uvc = _to_fold_latin1( (U8) *uc, foldbuf, &foldlen, 1); \ | |
1202 | skiplen = UNISKIP( uvc ); \ | |
1203 | foldlen -= skiplen; \ | |
1204 | uscan = foldbuf + skiplen; \ | |
1205 | } \ | |
1206 | break; \ | |
1207 | case trie_utf8: \ | |
1208 | uvc = utf8n_to_uvuni( (const U8*) uc, UTF8_MAXLEN, &len, uniflags ); \ | |
1209 | break; \ | |
1210 | case trie_plain: \ | |
1211 | uvc = (UV)*uc; \ | |
1212 | len = 1; \ | |
1213 | } \ | |
1214 | if (uvc < 256) { \ | |
1215 | charid = trie->charmap[ uvc ]; \ | |
1216 | } \ | |
1217 | else { \ | |
1218 | charid = 0; \ | |
1219 | if (widecharmap) { \ | |
1220 | SV** const svpp = hv_fetch(widecharmap, \ | |
1221 | (char*)&uvc, sizeof(UV), 0); \ | |
1222 | if (svpp) \ | |
1223 | charid = (U16)SvIV(*svpp); \ | |
1224 | } \ | |
1225 | } \ | |
4cadc6a9 YO |
1226 | } STMT_END |
1227 | ||
4cadc6a9 YO |
1228 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ |
1229 | STMT_START { \ | |
1230 | while (s <= e) { \ | |
1231 | if ( (CoNd) \ | |
fac1af77 | 1232 | && (ln == 1 || folder(s, pat_string, ln)) \ |
9a5a5549 | 1233 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1234 | goto got_it; \ |
1235 | s++; \ | |
1236 | } \ | |
1237 | } STMT_END | |
1238 | ||
1239 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
1240 | STMT_START { \ | |
1241 | while (s + (uskip = UTF8SKIP(s)) <= strend) { \ | |
1242 | CoDe \ | |
1243 | s += uskip; \ | |
1244 | } \ | |
1245 | } STMT_END | |
1246 | ||
1247 | #define REXEC_FBC_SCAN(CoDe) \ | |
1248 | STMT_START { \ | |
1249 | while (s < strend) { \ | |
1250 | CoDe \ | |
1251 | s++; \ | |
1252 | } \ | |
1253 | } STMT_END | |
1254 | ||
1255 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1256 | REXEC_FBC_UTF8_SCAN( \ | |
1257 | if (CoNd) { \ | |
24b23f37 | 1258 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1259 | goto got_it; \ |
1260 | else \ | |
1261 | tmp = doevery; \ | |
1262 | } \ | |
1263 | else \ | |
1264 | tmp = 1; \ | |
1265 | ) | |
1266 | ||
1267 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1268 | REXEC_FBC_SCAN( \ | |
1269 | if (CoNd) { \ | |
24b23f37 | 1270 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1271 | goto got_it; \ |
1272 | else \ | |
1273 | tmp = doevery; \ | |
1274 | } \ | |
1275 | else \ | |
1276 | tmp = 1; \ | |
1277 | ) | |
1278 | ||
1279 | #define REXEC_FBC_TRYIT \ | |
24b23f37 | 1280 | if ((!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1281 | goto got_it |
1282 | ||
e1d1eefb | 1283 | #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \ |
f2ed9b32 | 1284 | if (utf8_target) { \ |
e1d1eefb YO |
1285 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1286 | } \ | |
1287 | else { \ | |
1288 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1289 | } |
e1d1eefb | 1290 | |
4cadc6a9 | 1291 | #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ |
f2ed9b32 | 1292 | if (utf8_target) { \ |
4cadc6a9 YO |
1293 | UtFpReLoAd; \ |
1294 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1295 | } \ | |
1296 | else { \ | |
1297 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1298 | } |
4cadc6a9 YO |
1299 | |
1300 | #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ | |
1301 | PL_reg_flags |= RF_tainted; \ | |
f2ed9b32 | 1302 | if (utf8_target) { \ |
4cadc6a9 YO |
1303 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1304 | } \ | |
1305 | else { \ | |
1306 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1307 | } |
4cadc6a9 | 1308 | |
786e8c11 YO |
1309 | #define DUMP_EXEC_POS(li,s,doutf8) \ |
1310 | dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8) | |
1311 | ||
cfaf538b KW |
1312 | |
1313 | #define UTF8_NOLOAD(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ | |
1314 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1315 | tmp = TEST_NON_UTF8(tmp); \ | |
1316 | REXEC_FBC_UTF8_SCAN( \ | |
1317 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1318 | tmp = !tmp; \ | |
1319 | IF_SUCCESS; \ | |
1320 | } \ | |
1321 | else { \ | |
1322 | IF_FAIL; \ | |
1323 | } \ | |
1324 | ); \ | |
1325 | ||
1326 | #define UTF8_LOAD(TeSt1_UtF8, TeSt2_UtF8, IF_SUCCESS, IF_FAIL) \ | |
1327 | if (s == PL_bostr) { \ | |
1328 | tmp = '\n'; \ | |
1329 | } \ | |
1330 | else { \ | |
1331 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); \ | |
1332 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); \ | |
1333 | } \ | |
1334 | tmp = TeSt1_UtF8; \ | |
1335 | LOAD_UTF8_CHARCLASS_ALNUM(); \ | |
1336 | REXEC_FBC_UTF8_SCAN( \ | |
1337 | if (tmp == ! (TeSt2_UtF8)) { \ | |
1338 | tmp = !tmp; \ | |
1339 | IF_SUCCESS; \ | |
1340 | } \ | |
1341 | else { \ | |
1342 | IF_FAIL; \ | |
1343 | } \ | |
1344 | ); \ | |
1345 | ||
63ac0dad KW |
1346 | /* The only difference between the BOUND and NBOUND cases is that |
1347 | * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in | |
1348 | * NBOUND. This is accomplished by passing it in either the if or else clause, | |
1349 | * with the other one being empty */ | |
1350 | #define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1351 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
cfaf538b KW |
1352 | |
1353 | #define FBC_BOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1354 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
63ac0dad KW |
1355 | |
1356 | #define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1357 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b KW |
1358 | |
1359 | #define FBC_NBOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1360 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b | 1361 | |
63ac0dad KW |
1362 | |
1363 | /* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to | |
1364 | * be passed in completely with the variable name being tested, which isn't | |
1365 | * such a clean interface, but this is easier to read than it was before. We | |
1366 | * are looking for the boundary (or non-boundary between a word and non-word | |
1367 | * character. The utf8 and non-utf8 cases have the same logic, but the details | |
1368 | * must be different. Find the "wordness" of the character just prior to this | |
1369 | * one, and compare it with the wordness of this one. If they differ, we have | |
1370 | * a boundary. At the beginning of the string, pretend that the previous | |
1371 | * character was a new-line */ | |
cfaf538b | 1372 | #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ |
63ac0dad | 1373 | if (utf8_target) { \ |
cfaf538b | 1374 | UTF8_CODE \ |
63ac0dad KW |
1375 | } \ |
1376 | else { /* Not utf8 */ \ | |
1377 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1378 | tmp = TEST_NON_UTF8(tmp); \ | |
1379 | REXEC_FBC_SCAN( \ | |
1380 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1381 | tmp = !tmp; \ | |
1382 | IF_SUCCESS; \ | |
1383 | } \ | |
1384 | else { \ | |
1385 | IF_FAIL; \ | |
1386 | } \ | |
1387 | ); \ | |
1388 | } \ | |
1389 | if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) \ | |
1390 | goto got_it; | |
1391 | ||
786e8c11 YO |
1392 | /* We know what class REx starts with. Try to find this position... */ |
1393 | /* if reginfo is NULL, its a dryrun */ | |
1394 | /* annoyingly all the vars in this routine have different names from their counterparts | |
1395 | in regmatch. /grrr */ | |
1396 | ||
3c3eec57 | 1397 | STATIC char * |
07be1b83 | 1398 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
24b23f37 | 1399 | const char *strend, regmatch_info *reginfo) |
a687059c | 1400 | { |
27da23d5 | 1401 | dVAR; |
bbe252da | 1402 | const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; |
fac1af77 KW |
1403 | char *pat_string; /* The pattern's exactish string */ |
1404 | char *pat_end; /* ptr to end char of pat_string */ | |
1405 | re_fold_t folder; /* Function for computing non-utf8 folds */ | |
1406 | const U8 *fold_array; /* array for folding ords < 256 */ | |
d8093b23 | 1407 | STRLEN ln; |
5dab1207 | 1408 | STRLEN lnc; |
eb578fdb | 1409 | STRLEN uskip; |
fac1af77 KW |
1410 | U8 c1; |
1411 | U8 c2; | |
6eb5f6b9 | 1412 | char *e; |
eb578fdb KW |
1413 | I32 tmp = 1; /* Scratch variable? */ |
1414 | const bool utf8_target = PL_reg_match_utf8; | |
453bfd44 | 1415 | UV utf8_fold_flags = 0; |
f8fc2ecf | 1416 | RXi_GET_DECL(prog,progi); |
7918f24d NC |
1417 | |
1418 | PERL_ARGS_ASSERT_FIND_BYCLASS; | |
f8fc2ecf | 1419 | |
6eb5f6b9 JH |
1420 | /* We know what class it must start with. */ |
1421 | switch (OP(c)) { | |
f56b6394 | 1422 | case ANYOFV: |
6eb5f6b9 | 1423 | case ANYOF: |
f56b6394 | 1424 | if (utf8_target || OP(c) == ANYOFV) { |
b1e3e569 KW |
1425 | STRLEN inclasslen = strend - s; |
1426 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1427 | reginclass(prog, c, (U8*)s, &inclasslen, utf8_target)); | |
388cc4de HS |
1428 | } |
1429 | else { | |
6ef69d56 | 1430 | REXEC_FBC_CLASS_SCAN(REGINCLASS(prog, c, (U8*)s)); |
a0d0e21e | 1431 | } |
6eb5f6b9 | 1432 | break; |
f33976b4 | 1433 | case CANY: |
4cadc6a9 | 1434 | REXEC_FBC_SCAN( |
24b23f37 | 1435 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
f33976b4 DB |
1436 | goto got_it; |
1437 | else | |
1438 | tmp = doevery; | |
4cadc6a9 | 1439 | ); |
f33976b4 | 1440 | break; |
2f7f8cb1 KW |
1441 | |
1442 | case EXACTFA: | |
1443 | if (UTF_PATTERN || utf8_target) { | |
1444 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
1445 | goto do_exactf_utf8; | |
1446 | } | |
1447 | fold_array = PL_fold_latin1; /* Latin1 folds are not affected by */ | |
1448 | folder = foldEQ_latin1; /* /a, except the sharp s one which */ | |
1449 | goto do_exactf_non_utf8; /* isn't dealt with by these */ | |
1450 | ||
6eb5f6b9 | 1451 | case EXACTF: |
62bf7766 | 1452 | if (utf8_target) { |
77a6d856 KW |
1453 | |
1454 | /* regcomp.c already folded this if pattern is in UTF-8 */ | |
62bf7766 | 1455 | utf8_fold_flags = 0; |
fac1af77 KW |
1456 | goto do_exactf_utf8; |
1457 | } | |
1458 | fold_array = PL_fold; | |
1459 | folder = foldEQ; | |
1460 | goto do_exactf_non_utf8; | |
1461 | ||
1462 | case EXACTFL: | |
1463 | if (UTF_PATTERN || utf8_target) { | |
17580e7a | 1464 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
fac1af77 KW |
1465 | goto do_exactf_utf8; |
1466 | } | |
1467 | fold_array = PL_fold_locale; | |
1468 | folder = foldEQ_locale; | |
16d951b7 KW |
1469 | goto do_exactf_non_utf8; |
1470 | ||
3c760661 KW |
1471 | case EXACTFU_SS: |
1472 | if (UTF_PATTERN) { | |
1473 | utf8_fold_flags = FOLDEQ_S2_ALREADY_FOLDED; | |
1474 | } | |
1475 | goto do_exactf_utf8; | |
1476 | ||
fab2782b | 1477 | case EXACTFU_TRICKYFOLD: |
16d951b7 KW |
1478 | case EXACTFU: |
1479 | if (UTF_PATTERN || utf8_target) { | |
77a6d856 | 1480 | utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0; |
16d951b7 KW |
1481 | goto do_exactf_utf8; |
1482 | } | |
1483 | ||
1484 | /* Any 'ss' in the pattern should have been replaced by regcomp, | |
1485 | * so we don't have to worry here about this single special case | |
1486 | * in the Latin1 range */ | |
1487 | fold_array = PL_fold_latin1; | |
1488 | folder = foldEQ_latin1; | |
fac1af77 KW |
1489 | |
1490 | /* FALL THROUGH */ | |
1491 | ||
62bf7766 KW |
1492 | do_exactf_non_utf8: /* Neither pattern nor string are UTF8, and there |
1493 | are no glitches with fold-length differences | |
1494 | between the target string and pattern */ | |
fac1af77 KW |
1495 | |
1496 | /* The idea in the non-utf8 EXACTF* cases is to first find the | |
1497 | * first character of the EXACTF* node and then, if necessary, | |
1498 | * case-insensitively compare the full text of the node. c1 is the | |
1499 | * first character. c2 is its fold. This logic will not work for | |
1500 | * Unicode semantics and the german sharp ss, which hence should | |
1501 | * not be compiled into a node that gets here. */ | |
1502 | pat_string = STRING(c); | |
1503 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1504 | ||
8a90a8fe KW |
1505 | /* We know that we have to match at least 'ln' bytes (which is the |
1506 | * same as characters, since not utf8). If we have to match 3 | |
1507 | * characters, and there are only 2 availabe, we know without | |
1508 | * trying that it will fail; so don't start a match past the | |
1509 | * required minimum number from the far end */ | |
fac1af77 KW |
1510 | e = HOP3c(strend, -((I32)ln), s); |
1511 | ||
1512 | if (!reginfo && e < s) { | |
1513 | e = s; /* Due to minlen logic of intuit() */ | |
1514 | } | |
1515 | ||
1516 | c1 = *pat_string; | |
1517 | c2 = fold_array[c1]; | |
1518 | if (c1 == c2) { /* If char and fold are the same */ | |
1519 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); | |
1520 | } | |
1521 | else { | |
1522 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); | |
1523 | } | |
1524 | break; | |
1525 | ||
1526 | do_exactf_utf8: | |
e067297c KW |
1527 | { |
1528 | unsigned expansion; | |
1529 | ||
fac1af77 KW |
1530 | |
1531 | /* If one of the operands is in utf8, we can't use the simpler | |
1532 | * folding above, due to the fact that many different characters | |
1533 | * can have the same fold, or portion of a fold, or different- | |
1534 | * length fold */ | |
1535 | pat_string = STRING(c); | |
1536 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1537 | pat_end = pat_string + ln; | |
1538 | lnc = (UTF_PATTERN) /* length to match in characters */ | |
1539 | ? utf8_length((U8 *) pat_string, (U8 *) pat_end) | |
1540 | : ln; | |
1541 | ||
e067297c KW |
1542 | /* We have 'lnc' characters to match in the pattern, but because of |
1543 | * multi-character folding, each character in the target can match | |
1544 | * up to 3 characters (Unicode guarantees it will never exceed | |
1545 | * this) if it is utf8-encoded; and up to 2 if not (based on the | |
1546 | * fact that the Latin 1 folds are already determined, and the | |
1547 | * only multi-char fold in that range is the sharp-s folding to | |
1548 | * 'ss'. Thus, a pattern character can match as little as 1/3 of a | |
89378d8a KW |
1549 | * string character. Adjust lnc accordingly, rounding up, so that |
1550 | * if we need to match at least 4+1/3 chars, that really is 5. */ | |
e067297c | 1551 | expansion = (utf8_target) ? UTF8_MAX_FOLD_CHAR_EXPAND : 2; |
89378d8a | 1552 | lnc = (lnc + expansion - 1) / expansion; |
e067297c KW |
1553 | |
1554 | /* As in the non-UTF8 case, if we have to match 3 characters, and | |
1555 | * only 2 are left, it's guaranteed to fail, so don't start a | |
1556 | * match that would require us to go beyond the end of the string | |
1557 | */ | |
1558 | e = HOP3c(strend, -((I32)lnc), s); | |
fac1af77 KW |
1559 | |
1560 | if (!reginfo && e < s) { | |
1561 | e = s; /* Due to minlen logic of intuit() */ | |
1562 | } | |
1563 | ||
b33105db KW |
1564 | /* XXX Note that we could recalculate e to stop the loop earlier, |
1565 | * as the worst case expansion above will rarely be met, and as we | |
1566 | * go along we would usually find that e moves further to the left. | |
1567 | * This would happen only after we reached the point in the loop | |
1568 | * where if there were no expansion we should fail. Unclear if | |
1569 | * worth the expense */ | |
e067297c | 1570 | |
fac1af77 KW |
1571 | while (s <= e) { |
1572 | char *my_strend= (char *)strend; | |
1573 | if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target, | |
1574 | pat_string, NULL, ln, cBOOL(UTF_PATTERN), utf8_fold_flags) | |
1575 | && (!reginfo || regtry(reginfo, &s)) ) | |
1576 | { | |
1577 | goto got_it; | |
1578 | } | |
bbdd8bad | 1579 | s += (utf8_target) ? UTF8SKIP(s) : 1; |
fac1af77 KW |
1580 | } |
1581 | break; | |
e067297c | 1582 | } |
bbce6d69 | 1583 | case BOUNDL: |
3280af22 | 1584 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1585 | FBC_BOUND(isALNUM_LC, |
1586 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1587 | isALNUM_LC_utf8((U8*)s)); | |
a0ed51b3 | 1588 | break; |
bbce6d69 | 1589 | case NBOUNDL: |
3280af22 | 1590 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1591 | FBC_NBOUND(isALNUM_LC, |
1592 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1593 | isALNUM_LC_utf8((U8*)s)); | |
1594 | break; | |
1595 | case BOUND: | |
1596 | FBC_BOUND(isWORDCHAR, | |
1597 | isALNUM_uni(tmp), | |
1598 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1599 | break; | |
cfaf538b KW |
1600 | case BOUNDA: |
1601 | FBC_BOUND_NOLOAD(isWORDCHAR_A, | |
1602 | isWORDCHAR_A(tmp), | |
1603 | isWORDCHAR_A((U8*)s)); | |
1604 | break; | |
a0d0e21e | 1605 | case NBOUND: |
63ac0dad KW |
1606 | FBC_NBOUND(isWORDCHAR, |
1607 | isALNUM_uni(tmp), | |
1608 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1609 | break; | |
cfaf538b KW |
1610 | case NBOUNDA: |
1611 | FBC_NBOUND_NOLOAD(isWORDCHAR_A, | |
1612 | isWORDCHAR_A(tmp), | |
1613 | isWORDCHAR_A((U8*)s)); | |
1614 | break; | |
63ac0dad KW |
1615 | case BOUNDU: |
1616 | FBC_BOUND(isWORDCHAR_L1, | |
1617 | isALNUM_uni(tmp), | |
1618 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1619 | break; | |
1620 | case NBOUNDU: | |
1621 | FBC_NBOUND(isWORDCHAR_L1, | |
1622 | isALNUM_uni(tmp), | |
1623 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
a0ed51b3 | 1624 | break; |
bbce6d69 | 1625 | case ALNUML: |
4cadc6a9 YO |
1626 | REXEC_FBC_CSCAN_TAINT( |
1627 | isALNUM_LC_utf8((U8*)s), | |
1628 | isALNUM_LC(*s) | |
1629 | ); | |
6895a8aa | 1630 | break; |
980866de KW |
1631 | case ALNUMU: |
1632 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1633 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1634 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1635 | isWORDCHAR_L1((U8) *s) |
1636 | ); | |
6895a8aa | 1637 | break; |
980866de KW |
1638 | case ALNUM: |
1639 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1640 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1641 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1642 | isWORDCHAR((U8) *s) |
1643 | ); | |
6895a8aa | 1644 | break; |
cfaf538b | 1645 | case ALNUMA: |
8e9da4d4 KW |
1646 | /* Don't need to worry about utf8, as it can match only a single |
1647 | * byte invariant character */ | |
cfaf538b | 1648 | REXEC_FBC_CLASS_SCAN( isWORDCHAR_A(*s)); |
6895a8aa | 1649 | break; |
980866de KW |
1650 | case NALNUMU: |
1651 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 | 1652 | LOAD_UTF8_CHARCLASS_ALNUM(), |
359960d4 | 1653 | !swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), |
980866de KW |
1654 | ! isWORDCHAR_L1((U8) *s) |
1655 | ); | |
6895a8aa | 1656 | break; |
a0d0e21e | 1657 | case NALNUM: |
4cadc6a9 | 1658 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1659 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1660 | !swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target), | |
980866de | 1661 | ! isALNUM(*s) |
4cadc6a9 | 1662 | ); |
6895a8aa | 1663 | break; |
cfaf538b | 1664 | case NALNUMA: |
8e9da4d4 KW |
1665 | REXEC_FBC_CSCAN( |
1666 | !isWORDCHAR_A(*s), | |
1667 | !isWORDCHAR_A(*s) | |
1668 | ); | |
1669 | break; | |
bbce6d69 | 1670 | case NALNUML: |
4cadc6a9 YO |
1671 | REXEC_FBC_CSCAN_TAINT( |
1672 | !isALNUM_LC_utf8((U8*)s), | |
1673 | !isALNUM_LC(*s) | |
1674 | ); | |
6895a8aa | 1675 | break; |
980866de KW |
1676 | case SPACEU: |
1677 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1678 | LOAD_UTF8_CHARCLASS_SPACE(), |
1679 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de KW |
1680 | isSPACE_L1((U8) *s) |
1681 | ); | |
6895a8aa | 1682 | break; |
a0d0e21e | 1683 | case SPACE: |
4cadc6a9 | 1684 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1685 | LOAD_UTF8_CHARCLASS_SPACE(), |
1686 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de | 1687 | isSPACE((U8) *s) |
4cadc6a9 | 1688 | ); |
6895a8aa | 1689 | break; |
cfaf538b | 1690 | case SPACEA: |
8e9da4d4 KW |
1691 | /* Don't need to worry about utf8, as it can match only a single |
1692 | * byte invariant character */ | |
cfaf538b | 1693 | REXEC_FBC_CLASS_SCAN( isSPACE_A(*s)); |
6895a8aa | 1694 | break; |
bbce6d69 | 1695 | case SPACEL: |
4cadc6a9 | 1696 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1697 | isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1698 | isSPACE_LC(*s) |
1699 | ); | |
6895a8aa | 1700 | break; |
980866de KW |
1701 | case NSPACEU: |
1702 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1703 | LOAD_UTF8_CHARCLASS_SPACE(), |
1704 | !( *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de KW |
1705 | ! isSPACE_L1((U8) *s) |
1706 | ); | |
6895a8aa | 1707 | break; |
a0d0e21e | 1708 | case NSPACE: |
4cadc6a9 | 1709 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1710 | LOAD_UTF8_CHARCLASS_SPACE(), |
1711 | !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de | 1712 | ! isSPACE((U8) *s) |
4cadc6a9 | 1713 | ); |
6895a8aa | 1714 | break; |
cfaf538b | 1715 | case NSPACEA: |
8e9da4d4 KW |
1716 | REXEC_FBC_CSCAN( |
1717 | !isSPACE_A(*s), | |
1718 | !isSPACE_A(*s) | |
1719 | ); | |
1720 | break; | |
bbce6d69 | 1721 | case NSPACEL: |
4cadc6a9 | 1722 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1723 | !isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1724 | !isSPACE_LC(*s) |
1725 | ); | |
6895a8aa | 1726 | break; |
a0d0e21e | 1727 | case DIGIT: |
4cadc6a9 | 1728 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1729 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1730 | swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1731 | isDIGIT(*s) |
1732 | ); | |
6895a8aa | 1733 | break; |
cfaf538b | 1734 | case DIGITA: |
8e9da4d4 KW |
1735 | /* Don't need to worry about utf8, as it can match only a single |
1736 | * byte invariant character */ | |
cfaf538b | 1737 | REXEC_FBC_CLASS_SCAN( isDIGIT_A(*s)); |
6895a8aa | 1738 | break; |
b8c5462f | 1739 | case DIGITL: |
4cadc6a9 YO |
1740 | REXEC_FBC_CSCAN_TAINT( |
1741 | isDIGIT_LC_utf8((U8*)s), | |
1742 | isDIGIT_LC(*s) | |
1743 | ); | |
6895a8aa | 1744 | break; |
a0d0e21e | 1745 | case NDIGIT: |
4cadc6a9 | 1746 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1747 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1748 | !swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1749 | !isDIGIT(*s) |
1750 | ); | |
6895a8aa | 1751 | break; |
cfaf538b | 1752 | case NDIGITA: |
8e9da4d4 KW |
1753 | REXEC_FBC_CSCAN( |
1754 | !isDIGIT_A(*s), | |
1755 | !isDIGIT_A(*s) | |
1756 | ); | |
1757 | break; | |
b8c5462f | 1758 | case NDIGITL: |
4cadc6a9 YO |
1759 | REXEC_FBC_CSCAN_TAINT( |
1760 | !isDIGIT_LC_utf8((U8*)s), | |
1761 | !isDIGIT_LC(*s) | |
1762 | ); | |
6895a8aa | 1763 | break; |
e1d1eefb YO |
1764 | case LNBREAK: |
1765 | REXEC_FBC_CSCAN( | |
1766 | is_LNBREAK_utf8(s), | |
1767 | is_LNBREAK_latin1(s) | |
1768 | ); | |
6895a8aa | 1769 | break; |
e1d1eefb YO |
1770 | case VERTWS: |
1771 | REXEC_FBC_CSCAN( | |
1772 | is_VERTWS_utf8(s), | |
1773 | is_VERTWS_latin1(s) | |
1774 | ); | |
6895a8aa | 1775 | break; |
e1d1eefb YO |
1776 | case NVERTWS: |
1777 | REXEC_FBC_CSCAN( | |
1778 | !is_VERTWS_utf8(s), | |
1779 | !is_VERTWS_latin1(s) | |
1780 | ); | |
6895a8aa | 1781 | break; |
e1d1eefb YO |
1782 | case HORIZWS: |
1783 | REXEC_FBC_CSCAN( | |
1784 | is_HORIZWS_utf8(s), | |
1785 | is_HORIZWS_latin1(s) | |
1786 | ); | |
6895a8aa | 1787 | break; |
e1d1eefb YO |
1788 | case NHORIZWS: |
1789 | REXEC_FBC_CSCAN( | |
1790 | !is_HORIZWS_utf8(s), | |
1791 | !is_HORIZWS_latin1(s) | |
1792 | ); | |
6895a8aa | 1793 | break; |
0658cdde KW |
1794 | case POSIXA: |
1795 | /* Don't need to worry about utf8, as it can match only a single | |
1796 | * byte invariant character. The flag in this node type is the | |
1797 | * class number to pass to _generic_isCC() to build a mask for | |
1798 | * searching in PL_charclass[] */ | |
1799 | REXEC_FBC_CLASS_SCAN( _generic_isCC_A(*s, FLAGS(c))); | |
1800 | break; | |
1801 | case NPOSIXA: | |
1802 | REXEC_FBC_CSCAN( | |
1803 | !_generic_isCC_A(*s, FLAGS(c)), | |
1804 | !_generic_isCC_A(*s, FLAGS(c)) | |
1805 | ); | |
1806 | break; | |
1807 | ||
1de06328 YO |
1808 | case AHOCORASICKC: |
1809 | case AHOCORASICK: | |
07be1b83 | 1810 | { |
a0a388a1 | 1811 | DECL_TRIE_TYPE(c); |
07be1b83 YO |
1812 | /* what trie are we using right now */ |
1813 | reg_ac_data *aho | |
f8fc2ecf | 1814 | = (reg_ac_data*)progi->data->data[ ARG( c ) ]; |
3251b653 NC |
1815 | reg_trie_data *trie |
1816 | = (reg_trie_data*)progi->data->data[ aho->trie ]; | |
85fbaab2 | 1817 | HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); |
07be1b83 YO |
1818 | |
1819 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1820 | #ifdef DEBUGGING |
07be1b83 | 1821 | const char *real_start = s; |
6148ee25 | 1822 | #endif |
07be1b83 | 1823 | STRLEN maxlen = trie->maxlen; |
be8e71aa YO |
1824 | SV *sv_points; |
1825 | U8 **points; /* map of where we were in the input string | |
786e8c11 | 1826 | when reading a given char. For ASCII this |
be8e71aa | 1827 | is unnecessary overhead as the relationship |
38a44b82 NC |
1828 | is always 1:1, but for Unicode, especially |
1829 | case folded Unicode this is not true. */ | |
f9e705e8 | 1830 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
786e8c11 YO |
1831 | U8 *bitmap=NULL; |
1832 | ||
07be1b83 YO |
1833 | |
1834 | GET_RE_DEBUG_FLAGS_DECL; | |
1835 | ||
be8e71aa YO |
1836 | /* We can't just allocate points here. We need to wrap it in |
1837 | * an SV so it gets freed properly if there is a croak while | |
1838 | * running the match */ | |
1839 | ENTER; | |
1840 | SAVETMPS; | |
1841 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1842 | SvCUR_set(sv_points, | |
1843 | maxlen * sizeof(U8 *)); | |
1844 | SvPOK_on(sv_points); | |
1845 | sv_2mortal(sv_points); | |
1846 | points=(U8**)SvPV_nolen(sv_points ); | |
1de06328 YO |
1847 | if ( trie_type != trie_utf8_fold |
1848 | && (trie->bitmap || OP(c)==AHOCORASICKC) ) | |
1849 | { | |
786e8c11 YO |
1850 | if (trie->bitmap) |
1851 | bitmap=(U8*)trie->bitmap; | |
1852 | else | |
1853 | bitmap=(U8*)ANYOF_BITMAP(c); | |
07be1b83 | 1854 | } |
786e8c11 YO |
1855 | /* this is the Aho-Corasick algorithm modified a touch |
1856 | to include special handling for long "unknown char" | |
1857 | sequences. The basic idea being that we use AC as long | |
1858 | as we are dealing with a possible matching char, when | |
1859 | we encounter an unknown char (and we have not encountered | |
1860 | an accepting state) we scan forward until we find a legal | |
1861 | starting char. | |
1862 | AC matching is basically that of trie matching, except | |
1863 | that when we encounter a failing transition, we fall back | |
1864 | to the current states "fail state", and try the current char | |
1865 | again, a process we repeat until we reach the root state, | |
1866 | state 1, or a legal transition. If we fail on the root state | |
1867 | then we can either terminate if we have reached an accepting | |
1868 | state previously, or restart the entire process from the beginning | |
1869 | if we have not. | |
1870 | ||
1871 | */ | |
07be1b83 YO |
1872 | while (s <= last_start) { |
1873 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1874 | U8 *uc = (U8*)s; | |
1875 | U16 charid = 0; | |
1876 | U32 base = 1; | |
1877 | U32 state = 1; | |
1878 | UV uvc = 0; | |
1879 | STRLEN len = 0; | |
1880 | STRLEN foldlen = 0; | |
1881 | U8 *uscan = (U8*)NULL; | |
1882 | U8 *leftmost = NULL; | |
786e8c11 YO |
1883 | #ifdef DEBUGGING |
1884 | U32 accepted_word= 0; | |
1885 | #endif | |
07be1b83 YO |
1886 | U32 pointpos = 0; |
1887 | ||
1888 | while ( state && uc <= (U8*)strend ) { | |
1889 | int failed=0; | |
786e8c11 YO |
1890 | U32 word = aho->states[ state ].wordnum; |
1891 | ||
1de06328 YO |
1892 | if( state==1 ) { |
1893 | if ( bitmap ) { | |
1894 | DEBUG_TRIE_EXECUTE_r( | |
1895 | if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1896 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1897 | (char *)uc, utf8_target ); |
1de06328 YO |
1898 | PerlIO_printf( Perl_debug_log, |
1899 | " Scanning for legal start char...\n"); | |
1900 | } | |
d085b490 YO |
1901 | ); |
1902 | if (utf8_target) { | |
1903 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1904 | uc += UTF8SKIP(uc); | |
1905 | } | |
1906 | } else { | |
1907 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1908 | uc++; | |
1909 | } | |
1910 | } | |
1de06328 | 1911 | s= (char *)uc; |
786e8c11 | 1912 | } |
786e8c11 YO |
1913 | if (uc >(U8*)last_start) break; |
1914 | } | |
1915 | ||
1916 | if ( word ) { | |
2e64971a | 1917 | U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ]; |
786e8c11 YO |
1918 | if (!leftmost || lpos < leftmost) { |
1919 | DEBUG_r(accepted_word=word); | |
07be1b83 | 1920 | leftmost= lpos; |
786e8c11 | 1921 | } |
07be1b83 | 1922 | if (base==0) break; |
786e8c11 | 1923 | |
07be1b83 YO |
1924 | } |
1925 | points[pointpos++ % maxlen]= uc; | |
55eed653 NC |
1926 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
1927 | uscan, len, uvc, charid, foldlen, | |
1928 | foldbuf, uniflags); | |
786e8c11 YO |
1929 | DEBUG_TRIE_EXECUTE_r({ |
1930 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1931 | s, utf8_target ); |
07be1b83 | 1932 | PerlIO_printf(Perl_debug_log, |
786e8c11 YO |
1933 | " Charid:%3u CP:%4"UVxf" ", |
1934 | charid, uvc); | |
1935 | }); | |
07be1b83 YO |
1936 | |
1937 | do { | |
6148ee25 | 1938 | #ifdef DEBUGGING |
786e8c11 | 1939 | word = aho->states[ state ].wordnum; |
6148ee25 | 1940 | #endif |
07be1b83 YO |
1941 | base = aho->states[ state ].trans.base; |
1942 | ||
786e8c11 YO |
1943 | DEBUG_TRIE_EXECUTE_r({ |
1944 | if (failed) | |
1945 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1946 | s, utf8_target ); |
07be1b83 | 1947 | PerlIO_printf( Perl_debug_log, |
786e8c11 YO |
1948 | "%sState: %4"UVxf", word=%"UVxf, |
1949 | failed ? " Fail transition to " : "", | |
1950 | (UV)state, (UV)word); | |
1951 | }); | |
07be1b83 YO |
1952 | if ( base ) { |
1953 | U32 tmp; | |
6dd2be57 | 1954 | I32 offset; |
07be1b83 | 1955 | if (charid && |
6dd2be57 DM |
1956 | ( ((offset = base + charid |
1957 | - 1 - trie->uniquecharcount)) >= 0) | |
1958 | && ((U32)offset < trie->lasttrans) | |
1959 | && trie->trans[offset].check == state | |
1960 | && (tmp=trie->trans[offset].next)) | |
07be1b83 | 1961 | { |
786e8c11 YO |
1962 | DEBUG_TRIE_EXECUTE_r( |
1963 | PerlIO_printf( Perl_debug_log," - legal\n")); | |
07be1b83 YO |
1964 | state = tmp; |
1965 | break; | |
1966 | } | |
1967 | else { | |
786e8c11 YO |
1968 | DEBUG_TRIE_EXECUTE_r( |
1969 | PerlIO_printf( Perl_debug_log," - fail\n")); | |
1970 | failed = 1; | |
1971 | state = aho->fail[state]; | |
07be1b83 YO |
1972 | } |
1973 | } | |
1974 | else { | |
1975 | /* we must be accepting here */ | |
786e8c11 YO |
1976 | DEBUG_TRIE_EXECUTE_r( |
1977 | PerlIO_printf( Perl_debug_log," - accepting\n")); | |
1978 | failed = 1; | |
07be1b83 YO |
1979 | break; |
1980 | } | |
1981 | } while(state); | |
786e8c11 | 1982 | uc += len; |
07be1b83 YO |
1983 | if (failed) { |
1984 | if (leftmost) | |
1985 | break; | |
786e8c11 | 1986 | if (!state) state = 1; |
07be1b83 YO |
1987 | } |
1988 | } | |
1989 | if ( aho->states[ state ].wordnum ) { | |
2e64971a | 1990 | U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ]; |
786e8c11 YO |
1991 | if (!leftmost || lpos < leftmost) { |
1992 | DEBUG_r(accepted_word=aho->states[ state ].wordnum); | |
07be1b83 | 1993 | leftmost = lpos; |
786e8c11 | 1994 | } |
07be1b83 | 1995 | } |
07be1b83 YO |
1996 | if (leftmost) { |
1997 | s = (char*)leftmost; | |
786e8c11 YO |
1998 | DEBUG_TRIE_EXECUTE_r({ |
1999 | PerlIO_printf( | |
70685ca0 JH |
2000 | Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", |
2001 | (UV)accepted_word, (IV)(s - real_start) | |
786e8c11 YO |
2002 | ); |
2003 | }); | |
24b23f37 | 2004 | if (!reginfo || regtry(reginfo, &s)) { |
be8e71aa YO |
2005 | FREETMPS; |
2006 | LEAVE; | |
07be1b83 | 2007 | goto got_it; |
be8e71aa | 2008 | } |
07be1b83 | 2009 | s = HOPc(s,1); |
786e8c11 YO |
2010 | DEBUG_TRIE_EXECUTE_r({ |
2011 | PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); | |
2012 | }); | |
07be1b83 | 2013 | } else { |
786e8c11 YO |
2014 | DEBUG_TRIE_EXECUTE_r( |
2015 | PerlIO_printf( Perl_debug_log,"No match.\n")); | |
07be1b83 YO |
2016 | break; |
2017 | } | |
2018 | } | |
be8e71aa YO |
2019 | FREETMPS; |
2020 | LEAVE; | |
07be1b83 YO |
2021 | } |
2022 | break; | |
b3c9acc1 | 2023 | default: |
3c3eec57 GS |
2024 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); |
2025 | break; | |
d6a28714 | 2026 | } |
6eb5f6b9 JH |
2027 | return 0; |
2028 | got_it: | |
2029 | return s; | |
2030 | } | |
2031 | ||
fae667d5 | 2032 | |
6eb5f6b9 JH |
2033 | /* |
2034 | - regexec_flags - match a regexp against a string | |
2035 | */ | |
2036 | I32 | |
288b8c02 | 2037 | Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend, |
6eb5f6b9 | 2038 | char *strbeg, I32 minend, SV *sv, void *data, U32 flags) |
8fd1a950 DM |
2039 | /* stringarg: the point in the string at which to begin matching */ |
2040 | /* strend: pointer to null at end of string */ | |
2041 | /* strbeg: real beginning of string */ | |
2042 | /* minend: end of match must be >= minend bytes after stringarg. */ | |
2043 | /* sv: SV being matched: only used for utf8 flag, pos() etc; string | |
2044 | * itself is accessed via the pointers above */ | |
2045 | /* data: May be used for some additional optimizations. | |
2046 | Currently its only used, with a U32 cast, for transmitting | |
2047 | the ganch offset when doing a /g match. This will change */ | |
2048 | /* nosave: For optimizations. */ | |
2049 | ||
6eb5f6b9 | 2050 | { |
97aff369 | 2051 | dVAR; |
288b8c02 | 2052 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
24b23f37 | 2053 | /*register*/ char *s; |
eb578fdb | 2054 | regnode *c; |
24b23f37 | 2055 | /*register*/ char *startpos = stringarg; |
6eb5f6b9 JH |
2056 | I32 minlen; /* must match at least this many chars */ |
2057 | I32 dontbother = 0; /* how many characters not to try at end */ | |
6eb5f6b9 JH |
2058 | I32 end_shift = 0; /* Same for the end. */ /* CC */ |
2059 | I32 scream_pos = -1; /* Internal iterator of scream. */ | |
ccac19ea | 2060 | char *scream_olds = NULL; |
f2ed9b32 | 2061 | const bool utf8_target = cBOOL(DO_UTF8(sv)); |
2757e526 | 2062 | I32 multiline; |
f8fc2ecf | 2063 | RXi_GET_DECL(prog,progi); |
3b0527fe | 2064 | regmatch_info reginfo; /* create some info to pass to regtry etc */ |
e9105d30 | 2065 | regexp_paren_pair *swap = NULL; |
a3621e74 YO |
2066 | GET_RE_DEBUG_FLAGS_DECL; |
2067 | ||
7918f24d | 2068 | PERL_ARGS_ASSERT_REGEXEC_FLAGS; |
9d4ba2ae | 2069 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
2070 | |
2071 | /* Be paranoid... */ | |
2072 | if (prog == NULL || startpos == NULL) { | |
2073 | Perl_croak(aTHX_ "NULL regexp parameter"); | |
2074 | return 0; | |
2075 | } | |
2076 | ||
bbe252da | 2077 | multiline = prog->extflags & RXf_PMf_MULTILINE; |
288b8c02 | 2078 | reginfo.prog = rx; /* Yes, sorry that this is confusing. */ |
2757e526 | 2079 | |
f2ed9b32 | 2080 | RX_MATCH_UTF8_set(rx, utf8_target); |
1de06328 | 2081 | DEBUG_EXECUTE_r( |
f2ed9b32 | 2082 | debug_start_match(rx, utf8_target, startpos, strend, |
1de06328 YO |
2083 | "Matching"); |
2084 | ); | |
bac06658 | 2085 | |
6eb5f6b9 | 2086 | minlen = prog->minlen; |
1de06328 YO |
2087 | |
2088 | if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { | |
a3621e74 | 2089 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
2090 | "String too short [regexec_flags]...\n")); |
2091 | goto phooey; | |
1aa99e6b | 2092 | } |
6eb5f6b9 | 2093 | |
1de06328 | 2094 | |
6eb5f6b9 | 2095 | /* Check validity of program. */ |
f8fc2ecf | 2096 | if (UCHARAT(progi->program) != REG_MAGIC) { |
6eb5f6b9 JH |
2097 | Perl_croak(aTHX_ "corrupted regexp program"); |
2098 | } | |
2099 | ||
2100 | PL_reg_flags = 0; | |
ed301438 | 2101 | PL_reg_state.re_state_eval_setup_done = FALSE; |
6eb5f6b9 JH |
2102 | PL_reg_maxiter = 0; |
2103 | ||
3c8556c3 | 2104 | if (RX_UTF8(rx)) |
6eb5f6b9 JH |
2105 | PL_reg_flags |= RF_utf8; |
2106 | ||
2107 | /* Mark beginning of line for ^ and lookbehind. */ | |
3b0527fe | 2108 | reginfo.bol = startpos; /* XXX not used ??? */ |
6eb5f6b9 | 2109 | PL_bostr = strbeg; |
3b0527fe | 2110 | reginfo.sv = sv; |
6eb5f6b9 JH |
2111 | |
2112 | /* Mark end of line for $ (and such) */ | |
2113 | PL_regeol = strend; | |
2114 | ||
2115 | /* see how far we have to get to not match where we matched before */ | |
3b0527fe | 2116 | reginfo.till = startpos+minend; |
6eb5f6b9 | 2117 | |
6eb5f6b9 JH |
2118 | /* If there is a "must appear" string, look for it. */ |
2119 | s = startpos; | |
2120 | ||
bbe252da | 2121 | if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ |
6eb5f6b9 | 2122 | MAGIC *mg; |
2c296965 | 2123 | if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */ |
58e23c8d | 2124 | reginfo.ganch = startpos + prog->gofs; |
2c296965 | 2125 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2126 | "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs)); |
2c296965 | 2127 | } else if (sv && SvTYPE(sv) >= SVt_PVMG |
6eb5f6b9 | 2128 | && SvMAGIC(sv) |
14befaf4 DM |
2129 | && (mg = mg_find(sv, PERL_MAGIC_regex_global)) |
2130 | && mg->mg_len >= 0) { | |
3b0527fe | 2131 | reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ |
2c296965 | 2132 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2133 | "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len)); |
2c296965 | 2134 | |
bbe252da | 2135 | if (prog->extflags & RXf_ANCH_GPOS) { |
3b0527fe | 2136 | if (s > reginfo.ganch) |
6eb5f6b9 | 2137 | goto phooey; |
58e23c8d | 2138 | s = reginfo.ganch - prog->gofs; |
2c296965 | 2139 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2140 | "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs)); |
c584a96e YO |
2141 | if (s < strbeg) |
2142 | goto phooey; | |
6eb5f6b9 JH |
2143 | } |
2144 | } | |
58e23c8d | 2145 | else if (data) { |
70685ca0 | 2146 | reginfo.ganch = strbeg + PTR2UV(data); |
2c296965 YO |
2147 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2148 | "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data))); | |
2149 | ||
2150 | } else { /* pos() not defined */ | |
3b0527fe | 2151 | reginfo.ganch = strbeg; |
2c296965 YO |
2152 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2153 | "GPOS: reginfo.ganch = strbeg\n")); | |
2154 | } | |
6eb5f6b9 | 2155 | } |
288b8c02 | 2156 | if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { |
e9105d30 GG |
2157 | /* We have to be careful. If the previous successful match |
2158 | was from this regex we don't want a subsequent partially | |
2159 | successful match to clobber the old results. | |
2160 | So when we detect this possibility we add a swap buffer | |
2161 | to the re, and switch the buffer each match. If we fail | |
2162 | we switch it back, otherwise we leave it swapped. | |
2163 | */ | |
2164 | swap = prog->offs; | |
2165 | /* do we need a save destructor here for eval dies? */ | |
2166 | Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); | |
495f47a5 DM |
2167 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
2168 | "rex=0x%"UVxf" saving offs: orig=0x%"UVxf" new=0x%"UVxf"\n", | |
2169 | PTR2UV(prog), | |
2170 | PTR2UV(swap), | |
2171 | PTR2UV(prog->offs) | |
2172 | )); | |
c74340f9 | 2173 | } |
a0714e2c | 2174 | if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { |
6eb5f6b9 JH |
2175 | re_scream_pos_data d; |
2176 | ||
2177 | d.scream_olds = &scream_olds; | |
2178 | d.scream_pos = &scream_pos; | |
288b8c02 | 2179 | s = re_intuit_start(rx, sv, s, strend, flags, &d); |
3fa9c3d7 | 2180 | if (!s) { |
a3621e74 | 2181 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n")); |
6eb5f6b9 | 2182 | goto phooey; /* not present */ |
3fa9c3d7 | 2183 | } |
6eb5f6b9 JH |
2184 | } |
2185 | ||
1de06328 | 2186 | |
6eb5f6b9 JH |
2187 | |
2188 | /* Simplest case: anchored match need be tried only once. */ | |
2189 | /* [unless only anchor is BOL and multiline is set] */ | |
bbe252da | 2190 | if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { |
24b23f37 | 2191 | if (s == startpos && regtry(®info, &startpos)) |
6eb5f6b9 | 2192 | goto got_it; |
bbe252da YO |
2193 | else if (multiline || (prog->intflags & PREGf_IMPLICIT) |
2194 | || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ | |
6eb5f6b9 JH |
2195 | { |
2196 | char *end; | |
2197 | ||
2198 | if (minlen) | |
2199 | dontbother = minlen - 1; | |
1aa99e6b | 2200 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 2201 | /* for multiline we only have to try after newlines */ |
33b8afdf | 2202 | if (prog->check_substr || prog->check_utf8) { |
92f3d482 YO |
2203 | /* because of the goto we can not easily reuse the macros for bifurcating the |
2204 | unicode/non-unicode match modes here like we do elsewhere - demerphq */ | |
2205 | if (utf8_target) { | |
2206 | if (s == startpos) | |
2207 | goto after_try_utf8; | |
2208 | while (1) { | |
2209 | if (regtry(®info, &s)) { | |
2210 | goto got_it; | |
2211 | } | |
2212 | after_try_utf8: | |
2213 | if (s > end) { | |
2214 | goto phooey; | |
2215 | } | |
2216 | if (prog->extflags & RXf_USE_INTUIT) { | |
2217 | s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL); | |
2218 | if (!s) { | |
2219 | goto phooey; | |
2220 | } | |
2221 | } | |
2222 | else { | |
2223 | s += UTF8SKIP(s); | |
2224 | } | |
2225 | } | |
2226 | } /* end search for check string in unicode */ | |
2227 | else { | |
2228 | if (s == startpos) { | |
2229 | goto after_try_latin; | |
2230 | } | |
2231 | while (1) { | |
2232 | if (regtry(®info, &s)) { | |
2233 | goto got_it; | |
2234 | } | |
2235 | after_try_latin: | |
2236 | if (s > end) { | |
2237 | goto phooey; | |
2238 | } | |
2239 | if (prog->extflags & RXf_USE_INTUIT) { | |
2240 | s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL); | |
2241 | if (!s) { | |
2242 | goto phooey; | |
2243 | } | |
2244 | } | |
2245 | else { | |
2246 | s++; | |
2247 | } | |
2248 | } | |
2249 | } /* end search for check string in latin*/ | |
2250 | } /* end search for check string */ | |
2251 | else { /* search for newline */ | |
2252 | if (s > startpos) { | |
2253 | /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/ | |
6eb5f6b9 | 2254 | s--; |
92f3d482 | 2255 | } |
21eede78 YO |
2256 | /* We can use a more efficient search as newlines are the same in unicode as they are in latin */ |
2257 | while (s <= end) { /* note it could be possible to match at the end of the string */ | |
6eb5f6b9 | 2258 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ |
24b23f37 | 2259 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2260 | goto got_it; |
2261 | } | |
92f3d482 YO |
2262 | } |
2263 | } /* end search for newline */ | |
2264 | } /* end anchored/multiline check string search */ | |
6eb5f6b9 | 2265 | goto phooey; |
bbe252da | 2266 | } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) |
f9f4320a | 2267 | { |
486ec47a | 2268 | /* the warning about reginfo.ganch being used without initialization |
bbe252da | 2269 | is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN |
f9f4320a | 2270 | and we only enter this block when the same bit is set. */ |
58e23c8d | 2271 | char *tmp_s = reginfo.ganch - prog->gofs; |
c584a96e YO |
2272 | |
2273 | if (tmp_s >= strbeg && regtry(®info, &tmp_s)) | |
6eb5f6b9 JH |
2274 | goto got_it; |
2275 | goto phooey; | |
2276 | } | |
2277 | ||
2278 | /* Messy cases: unanchored match. */ | |
bbe252da | 2279 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { |
6eb5f6b9 | 2280 | /* we have /x+whatever/ */ |
f2ed9b32 | 2281 | /* it must be a one character string (XXXX Except UTF_PATTERN?) */ |
33b8afdf | 2282 | char ch; |
bf93d4cc GS |
2283 | #ifdef DEBUGGING |
2284 | int did_match = 0; | |
2285 | #endif | |
f2ed9b32 KW |
2286 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2287 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2288 | ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0]; | |
bf93d4cc | 2289 | |
f2ed9b32 | 2290 | if (utf8_target) { |
4cadc6a9 | 2291 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2292 | if (*s == ch) { |
a3621e74 | 2293 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2294 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2295 | s += UTF8SKIP(s); |
2296 | while (s < strend && *s == ch) | |
2297 | s += UTF8SKIP(s); | |
2298 | } | |
4cadc6a9 | 2299 | ); |
6eb5f6b9 JH |
2300 | } |
2301 | else { | |
4cadc6a9 | 2302 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2303 | if (*s == ch) { |
a3621e74 | 2304 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2305 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2306 | s++; |
2307 | while (s < strend && *s == ch) | |
2308 | s++; | |
2309 | } | |
4cadc6a9 | 2310 | ); |
6eb5f6b9 | 2311 | } |
a3621e74 | 2312 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 2313 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
2314 | "Did not find anchored character...\n") |
2315 | ); | |
6eb5f6b9 | 2316 | } |
a0714e2c SS |
2317 | else if (prog->anchored_substr != NULL |
2318 | || prog->anchored_utf8 != NULL | |
2319 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
2320 | && prog->float_max_offset < strend - s)) { |
2321 | SV *must; | |
2322 | I32 back_max; | |
2323 | I32 back_min; | |
2324 | char *last; | |
6eb5f6b9 | 2325 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
2326 | #ifdef DEBUGGING |
2327 | int did_match = 0; | |
2328 | #endif | |
33b8afdf | 2329 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 KW |
2330 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2331 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2332 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; | |
33b8afdf JH |
2333 | back_max = back_min = prog->anchored_offset; |
2334 | } else { | |
f2ed9b32 KW |
2335 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2336 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2337 | must = utf8_target ? prog->float_utf8 : prog->float_substr; | |
33b8afdf JH |
2338 | back_max = prog->float_max_offset; |
2339 | back_min = prog->float_min_offset; | |
2340 | } | |
1de06328 YO |
2341 | |
2342 | ||
33b8afdf JH |
2343 | if (must == &PL_sv_undef) |
2344 | /* could not downgrade utf8 check substring, so must fail */ | |
2345 | goto phooey; | |
2346 | ||
1de06328 YO |
2347 | if (back_min<0) { |
2348 | last = strend; | |
2349 | } else { | |
2350 | last = HOP3c(strend, /* Cannot start after this */ | |
2351 | -(I32)(CHR_SVLEN(must) | |
2352 | - (SvTAIL(must) != 0) + back_min), strbeg); | |
2353 | } | |
6eb5f6b9 JH |
2354 | if (s > PL_bostr) |
2355 | last1 = HOPc(s, -1); | |
2356 | else | |
2357 | last1 = s - 1; /* bogus */ | |
2358 | ||
a0288114 | 2359 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 JH |
2360 | check_substr==must. */ |
2361 | scream_pos = -1; | |
2362 | dontbother = end_shift; | |
2363 | strend = HOPc(strend, -dontbother); | |
2364 | while ( (s <= last) && | |
c33e64f0 | 2365 | (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), |
9041c2e3 | 2366 | (unsigned char*)strend, must, |
c33e64f0 | 2367 | multiline ? FBMrf_MULTILINE : 0)) ) { |
a3621e74 | 2368 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
2369 | if (HOPc(s, -back_max) > last1) { |
2370 | last1 = HOPc(s, -back_min); | |
2371 | s = HOPc(s, -back_max); | |
2372 | } | |
2373 | else { | |
52657f30 | 2374 | char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; |
6eb5f6b9 JH |
2375 | |
2376 | last1 = HOPc(s, -back_min); | |
52657f30 | 2377 | s = t; |
6eb5f6b9 | 2378 | } |
f2ed9b32 | 2379 | if (utf8_target) { |
6eb5f6b9 | 2380 | while (s <= last1) { |
24b23f37 | 2381 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2382 | goto got_it; |
2383 | s += UTF8SKIP(s); | |
2384 | } | |
2385 | } | |
2386 | else { | |
2387 | while (s <= last1) { | |
24b23f37 | 2388 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2389 | goto got_it; |
2390 | s++; | |
2391 | } | |
2392 | } | |
2393 | } | |
ab3bbdeb | 2394 | DEBUG_EXECUTE_r(if (!did_match) { |
f2ed9b32 | 2395 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
2396 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
2397 | PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", | |
33b8afdf | 2398 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc | 2399 | ? "anchored" : "floating"), |
ab3bbdeb YO |
2400 | quoted, RE_SV_TAIL(must)); |
2401 | }); | |
6eb5f6b9 JH |
2402 | goto phooey; |
2403 | } | |
f8fc2ecf | 2404 | else if ( (c = progi->regstclass) ) { |
f14c76ed | 2405 | if (minlen) { |
f8fc2ecf | 2406 | const OPCODE op = OP(progi->regstclass); |
66e933ab | 2407 | /* don't bother with what can't match */ |
786e8c11 | 2408 | if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) |
f14c76ed RGS |
2409 | strend = HOPc(strend, -(minlen - 1)); |
2410 | } | |
a3621e74 | 2411 | DEBUG_EXECUTE_r({ |
be8e71aa | 2412 | SV * const prop = sv_newmortal(); |
32fc9b6a | 2413 | regprop(prog, prop, c); |
0df25f3d | 2414 | { |
f2ed9b32 | 2415 | RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb | 2416 | s,strend-s,60); |
0df25f3d | 2417 | PerlIO_printf(Perl_debug_log, |
1c8f8eb1 | 2418 | "Matching stclass %.*s against %s (%d bytes)\n", |
e4f74956 | 2419 | (int)SvCUR(prop), SvPVX_const(prop), |
ab3bbdeb | 2420 | quoted, (int)(strend - s)); |
0df25f3d | 2421 | } |
ffc61ed2 | 2422 | }); |
3b0527fe | 2423 | if (find_byclass(prog, c, s, strend, ®info)) |
6eb5f6b9 | 2424 | goto got_it; |
07be1b83 | 2425 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
2426 | } |
2427 | else { | |
2428 | dontbother = 0; | |
a0714e2c | 2429 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 2430 | /* Trim the end. */ |
6af40bd7 | 2431 | char *last= NULL; |
33b8afdf | 2432 | SV* float_real; |
c33e64f0 FC |
2433 | STRLEN len; |
2434 | const char *little; | |
33b8afdf | 2435 | |
f2ed9b32 KW |
2436 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2437 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2438 | float_real = utf8_target ? prog->float_utf8 : prog->float_substr; | |
d6a28714 | 2439 | |
c33e64f0 FC |
2440 | little = SvPV_const(float_real, len); |
2441 | if (SvTAIL(float_real)) { | |
1a13b075 YO |
2442 | /* This means that float_real contains an artificial \n on the end |
2443 | * due to the presence of something like this: /foo$/ | |
2444 | * where we can match both "foo" and "foo\n" at the end of the string. | |
2445 | * So we have to compare the end of the string first against the float_real | |
2446 | * without the \n and then against the full float_real with the string. | |
2447 | * We have to watch out for cases where the string might be smaller | |
2448 | * than the float_real or the float_real without the \n. | |
2449 | */ | |
2450 | char *checkpos= strend - len; | |
2451 | DEBUG_OPTIMISE_r( | |
2452 | PerlIO_printf(Perl_debug_log, | |
2453 | "%sChecking for float_real.%s\n", | |
2454 | PL_colors[4], PL_colors[5])); | |
2455 | if (checkpos + 1 < strbeg) { | |
2456 | /* can't match, even if we remove the trailing \n string is too short to match */ | |
2457 | DEBUG_EXECUTE_r( | |
2458 | PerlIO_printf(Perl_debug_log, | |
2459 | "%sString shorter than required trailing substring, cannot match.%s\n", | |
2460 | PL_colors[4], PL_colors[5])); | |
2461 | goto phooey; | |
2462 | } else if (memEQ(checkpos + 1, little, len - 1)) { | |
2463 | /* can match, the end of the string matches without the "\n" */ | |
2464 | last = checkpos + 1; | |
2465 | } else if (checkpos < strbeg) { | |
2466 | /* cant match, string is too short when the "\n" is included */ | |
2467 | DEBUG_EXECUTE_r( | |
2468 | PerlIO_printf(Perl_debug_log, | |
2469 | "%sString does not contain required trailing substring, cannot match.%s\n", | |
2470 | PL_colors[4], PL_colors[5])); | |
2471 | goto phooey; | |
2472 | } else if (!multiline) { | |
2473 | /* non multiline match, so compare with the "\n" at the end of the string */ | |
2474 | if (memEQ(checkpos, little, len)) { | |
2475 | last= checkpos; | |
2476 | } else { | |
2477 | DEBUG_EXECUTE_r( | |
2478 | PerlIO_printf(Perl_debug_log, | |
2479 | "%sString does not contain required trailing substring, cannot match.%s\n", | |
2480 | PL_colors[4], PL_colors[5])); | |
2481 | goto phooey; | |
2482 | } | |
2483 | } else { | |
2484 | /* multiline match, so we have to search for a place where the full string is located */ | |
d6a28714 | 2485 | goto find_last; |
1a13b075 | 2486 | } |
c33e64f0 | 2487 | } else { |
d6a28714 | 2488 | find_last: |
9041c2e3 | 2489 | if (len) |
d6a28714 | 2490 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 2491 | else |
a0288114 | 2492 | last = strend; /* matching "$" */ |
b8c5462f | 2493 | } |
6af40bd7 YO |
2494 | if (!last) { |
2495 | /* at one point this block contained a comment which was probably | |
2496 | * incorrect, which said that this was a "should not happen" case. | |
2497 | * Even if it was true when it was written I am pretty sure it is | |
2498 | * not anymore, so I have removed the comment and replaced it with | |
2499 | * this one. Yves */ | |
6bda09f9 YO |
2500 | DEBUG_EXECUTE_r( |
2501 | PerlIO_printf(Perl_debug_log, | |
6af40bd7 YO |
2502 | "String does not contain required substring, cannot match.\n" |
2503 | )); | |
2504 | goto phooey; | |
bf93d4cc | 2505 | } |
d6a28714 JH |
2506 | dontbother = strend - last + prog->float_min_offset; |
2507 | } | |
2508 | if (minlen && (dontbother < minlen)) | |
2509 | dontbother = minlen - 1; | |
2510 | strend -= dontbother; /* this one's always in bytes! */ | |
2511 | /* We don't know much -- general case. */ | |
f2ed9b32 | 2512 | if (utf8_target) { |
d6a28714 | 2513 | for (;;) { |
24b23f37 | 2514 | if (regtry(®info, &s)) |
d6a28714 JH |
2515 | goto got_it; |
2516 | if (s >= strend) | |
2517 | break; | |
b8c5462f | 2518 | s += UTF8SKIP(s); |
d6a28714 JH |
2519 | }; |
2520 | } | |
2521 | else { | |
2522 | do { | |
24b23f37 | 2523 | if (regtry(®info, &s)) |
d6a28714 JH |
2524 | goto got_it; |
2525 | } while (s++ < strend); | |
2526 | } | |
2527 | } | |
2528 | ||
2529 | /* Failure. */ | |
2530 | goto phooey; | |
2531 | ||
2532 | got_it: | |
495f47a5 DM |
2533 | DEBUG_BUFFERS_r( |
2534 | if (swap) | |
2535 | PerlIO_printf(Perl_debug_log, | |
2536 | "rex=0x%"UVxf" freeing offs: 0x%"UVxf"\n", | |
2537 | PTR2UV(prog), | |
2538 | PTR2UV(swap) | |
2539 | ); | |
2540 | ); | |
e9105d30 | 2541 | Safefree(swap); |
288b8c02 | 2542 | RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted); |
d6a28714 | 2543 | |
ed301438 | 2544 | if (PL_reg_state.re_state_eval_setup_done) |
4f639d21 | 2545 | restore_pos(aTHX_ prog); |
5daac39c NC |
2546 | if (RXp_PAREN_NAMES(prog)) |
2547 | (void)hv_iterinit(RXp_PAREN_NAMES(prog)); | |
d6a28714 JH |
2548 | |
2549 | /* make sure $`, $&, $', and $digit will work later */ | |
2550 | if ( !(flags & REXEC_NOT_FIRST) ) { | |
d6a28714 | 2551 | if (flags & REXEC_COPY_STR) { |
f8c7b90f | 2552 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2553 | if ((SvIsCOW(sv) |
2554 | || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { | |
2555 | if (DEBUG_C_TEST) { | |
2556 | PerlIO_printf(Perl_debug_log, | |
2557 | "Copy on write: regexp capture, type %d\n", | |
2558 | (int) SvTYPE(sv)); | |
2559 | } | |
77f8f7c1 | 2560 | RX_MATCH_COPY_FREE(rx); |
ed252734 | 2561 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); |
d5263905 | 2562 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
ed252734 | 2563 | assert (SvPOKp(prog->saved_copy)); |
6502e081 DM |
2564 | prog->sublen = PL_regeol - strbeg; |
2565 | prog->suboffset = 0; | |
2566 | prog->subcoffset = 0; | |
ed252734 NC |
2567 | } else |
2568 | #endif | |
2569 | { | |
6502e081 DM |
2570 | I32 min = 0; |
2571 | I32 max = PL_regeol - strbeg; | |
2572 | I32 sublen; | |
2573 | ||
2574 | if ( (flags & REXEC_COPY_SKIP_POST) | |
2575 | && !(RX_EXTFLAGS(rx) & RXf_PMf_KEEPCOPY) /* //p */ | |
2576 | && !(PL_sawampersand & SAWAMPERSAND_RIGHT) | |
2577 | ) { /* don't copy $' part of string */ | |
3de645a8 | 2578 | U32 n = 0; |
6502e081 DM |
2579 | max = -1; |
2580 | /* calculate the right-most part of the string covered | |
2581 | * by a capture. Due to look-ahead, this may be to | |
2582 | * the right of $&, so we have to scan all captures */ | |
2583 | while (n <= prog->lastparen) { | |
2584 | if (prog->offs[n].end > max) | |
2585 | max = prog->offs[n].end; | |
2586 | n++; | |
2587 | } | |
2588 | if (max == -1) | |
2589 | max = (PL_sawampersand & SAWAMPERSAND_LEFT) | |
2590 | ? prog->offs[0].start | |
2591 | : 0; | |
2592 | assert(max >= 0 && max <= PL_regeol - strbeg); | |
2593 | } | |
2594 | ||
2595 | if ( (flags & REXEC_COPY_SKIP_PRE) | |
2596 | && !(RX_EXTFLAGS(rx) & RXf_PMf_KEEPCOPY) /* //p */ | |
2597 | && !(PL_sawampersand & SAWAMPERSAND_LEFT) | |
2598 | ) { /* don't copy $` part of string */ | |
3de645a8 | 2599 | U32 n = 0; |
6502e081 DM |
2600 | min = max; |
2601 | /* calculate the left-most part of the string covered | |
2602 | * by a capture. Due to look-behind, this may be to | |
2603 | * the left of $&, so we have to scan all captures */ | |
2604 | while (min && n <= prog->lastparen) { | |
2605 | if ( prog->offs[n].start != -1 | |
2606 | && prog->offs[n].start < min) | |
2607 | { | |
2608 | min = prog->offs[n].start; | |
2609 | } | |
2610 | n++; | |
2611 | } | |
2612 | if ((PL_sawampersand & SAWAMPERSAND_RIGHT) | |
2613 | && min > prog->offs[0].end | |
2614 | ) | |
2615 | min = prog->offs[0].end; | |
2616 | ||
2617 | } | |
2618 | ||
2619 | assert(min >= 0 && min <= max && min <= PL_regeol - strbeg); | |
2620 | sublen = max - min; | |
2621 | ||
2622 | if (RX_MATCH_COPIED(rx)) { | |
2623 | if (sublen > prog->sublen) | |
2624 | prog->subbeg = | |
2625 | (char*)saferealloc(prog->subbeg, sublen+1); | |
2626 | } | |
2627 | else | |
2628 | prog->subbeg = (char*)safemalloc(sublen+1); | |
2629 | Copy(strbeg + min, prog->subbeg, sublen, char); | |
2630 | prog->subbeg[sublen] = '\0'; | |
2631 | prog->suboffset = min; | |
2632 | prog->sublen = sublen; | |
77f8f7c1 | 2633 | RX_MATCH_COPIED_on(rx); |
6502e081 | 2634 | } |
6502e081 DM |
2635 | prog->subcoffset = prog->suboffset; |
2636 | if (prog->suboffset && utf8_target) { | |
2637 | /* Convert byte offset to chars. | |
2638 | * XXX ideally should only compute this if @-/@+ | |
2639 | * has been seen, a la PL_sawampersand ??? */ | |
2640 | ||
2641 | /* If there's a direct correspondence between the | |
2642 | * string which we're matching and the original SV, | |
2643 | * then we can use the utf8 len cache associated with | |
2644 | * the SV. In particular, it means that under //g, | |
2645 | * sv_pos_b2u() will use the previously cached | |
2646 | * position to speed up working out the new length of | |
2647 | * subcoffset, rather than counting from the start of | |
2648 | * the string each time. This stops | |
2649 | * $x = "\x{100}" x 1E6; 1 while $x =~ /(.)/g; | |
2650 | * from going quadratic */ | |
2651 | if (SvPOKp(sv) && SvPVX(sv) == strbeg) | |
2652 | sv_pos_b2u(sv, &(prog->subcoffset)); | |
2653 | else | |
2654 | prog->subcoffset = utf8_length((U8*)strbeg, | |
2655 | (U8*)(strbeg+prog->suboffset)); | |
2656 | } | |
d6a28714 JH |
2657 | } |
2658 | else { | |
6502e081 | 2659 | RX_MATCH_COPY_FREE(rx); |
d6a28714 | 2660 | prog->subbeg = strbeg; |
6502e081 DM |
2661 | prog->suboffset = 0; |
2662 | prog->subcoffset = 0; | |
d6a28714 JH |
2663 | prog->sublen = PL_regeol - strbeg; /* strend may have been modified */ |
2664 | } | |
2665 | } | |
9041c2e3 | 2666 | |
d6a28714 JH |
2667 | return 1; |
2668 | ||
2669 | phooey: | |
a3621e74 | 2670 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 2671 | PL_colors[4], PL_colors[5])); |
ed301438 | 2672 | if (PL_reg_state.re_state_eval_setup_done) |
4f639d21 | 2673 | restore_pos(aTHX_ prog); |
e9105d30 | 2674 | if (swap) { |
c74340f9 | 2675 | /* we failed :-( roll it back */ |
495f47a5 DM |
2676 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
2677 | "rex=0x%"UVxf" rolling back offs: freeing=0x%"UVxf" restoring=0x%"UVxf"\n", | |
2678 | PTR2UV(prog), | |
2679 | PTR2UV(prog->offs), | |
2680 | PTR2UV(swap) | |
2681 | )); | |
e9105d30 GG |
2682 | Safefree(prog->offs); |
2683 | prog->offs = swap; | |
2684 | } | |
2685 | ||
d6a28714 JH |
2686 | return 0; |
2687 | } | |
2688 | ||
6bda09f9 | 2689 | |
ec43f78b DM |
2690 | /* Set which rex is pointed to by PL_reg_state, handling ref counting. |
2691 | * Do inc before dec, in case old and new rex are the same */ | |
2692 | #define SET_reg_curpm(Re2) \ | |
2693 | if (PL_reg_state.re_state_eval_setup_done) { \ | |
2694 | (void)ReREFCNT_inc(Re2); \ | |
2695 | ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); \ | |
2696 | PM_SETRE((PL_reg_curpm), (Re2)); \ | |
2697 | } | |
2698 | ||
2699 | ||
d6a28714 JH |
2700 | /* |
2701 | - regtry - try match at specific point | |
2702 | */ | |
2703 | STATIC I32 /* 0 failure, 1 success */ | |
f73aaa43 | 2704 | S_regtry(pTHX_ regmatch_info *reginfo, char **startposp) |
d6a28714 | 2705 | { |
97aff369 | 2706 | dVAR; |
d6a28714 | 2707 | CHECKPOINT lastcp; |
288b8c02 NC |
2708 | REGEXP *const rx = reginfo->prog; |
2709 | regexp *const prog = (struct regexp *)SvANY(rx); | |
f73aaa43 | 2710 | I32 result; |
f8fc2ecf | 2711 | RXi_GET_DECL(prog,progi); |
a3621e74 | 2712 | GET_RE_DEBUG_FLAGS_DECL; |
7918f24d NC |
2713 | |
2714 | PERL_ARGS_ASSERT_REGTRY; | |
2715 | ||
24b23f37 | 2716 | reginfo->cutpoint=NULL; |
d6a28714 | 2717 | |
ed301438 DM |
2718 | if ((prog->extflags & RXf_EVAL_SEEN) |
2719 | && !PL_reg_state.re_state_eval_setup_done) | |
2720 | { | |
d6a28714 JH |
2721 | MAGIC *mg; |
2722 | ||
ed301438 | 2723 | PL_reg_state.re_state_eval_setup_done = TRUE; |
3b0527fe | 2724 | if (reginfo->sv) { |
d6a28714 | 2725 | /* Make $_ available to executed code. */ |
3b0527fe | 2726 | if (reginfo->sv != DEFSV) { |
59f00321 | 2727 | SAVE_DEFSV; |
414bf5ae | 2728 | DEFSV_set(reginfo->sv); |
b8c5462f | 2729 | } |
d6a28714 | 2730 | |
3b0527fe DM |
2731 | if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) |
2732 | && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { | |
d6a28714 | 2733 | /* prepare for quick setting of pos */ |
d300d9fa | 2734 | #ifdef PERL_OLD_COPY_ON_WRITE |
51a9ea20 NC |
2735 | if (SvIsCOW(reginfo->sv)) |
2736 | sv_force_normal_flags(reginfo->sv, 0); | |
d300d9fa | 2737 | #endif |
3dab1dad | 2738 | mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, |
d300d9fa | 2739 | &PL_vtbl_mglob, NULL, 0); |
d6a28714 | 2740 | mg->mg_len = -1; |
b8c5462f | 2741 | } |
d6a28714 JH |
2742 | PL_reg_magic = mg; |
2743 | PL_reg_oldpos = mg->mg_len; | |
4f639d21 | 2744 | SAVEDESTRUCTOR_X(restore_pos, prog); |
d6a28714 | 2745 | } |
09687e5a | 2746 | if (!PL_reg_curpm) { |
a02a5408 | 2747 | Newxz(PL_reg_curpm, 1, PMOP); |
09687e5a AB |
2748 | #ifdef USE_ITHREADS |
2749 | { | |
14a49a24 | 2750 | SV* const repointer = &PL_sv_undef; |
92313705 NC |
2751 | /* this regexp is also owned by the new PL_reg_curpm, which |
2752 | will try to free it. */ | |
d2ece331 | 2753 | av_push(PL_regex_padav, repointer); |
09687e5a AB |
2754 | PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); |
2755 | PL_regex_pad = AvARRAY(PL_regex_padav); | |
2756 | } | |
2757 | #endif | |
2758 | } | |
ec43f78b | 2759 | SET_reg_curpm(rx); |
d6a28714 JH |
2760 | PL_reg_oldcurpm = PL_curpm; |
2761 | PL_curpm = PL_reg_curpm; | |
07bc277f | 2762 | if (RXp_MATCH_COPIED(prog)) { |
d6a28714 JH |
2763 | /* Here is a serious problem: we cannot rewrite subbeg, |
2764 | since it may be needed if this match fails. Thus | |
2765 | $` inside (?{}) could fail... */ | |
2766 | PL_reg_oldsaved = prog->subbeg; | |
2767 | PL_reg_oldsavedlen = prog->sublen; | |
6502e081 DM |
2768 | PL_reg_oldsavedoffset = prog->suboffset; |
2769 | PL_reg_oldsavedcoffset = prog->suboffset; | |
f8c7b90f | 2770 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2771 | PL_nrs = prog->saved_copy; |
2772 | #endif | |
07bc277f | 2773 | RXp_MATCH_COPIED_off(prog); |
d6a28714 JH |
2774 | } |
2775 | else | |
bd61b366 | 2776 | PL_reg_oldsaved = NULL; |
d6a28714 | 2777 | prog->subbeg = PL_bostr; |
6502e081 DM |
2778 | prog->suboffset = 0; |
2779 | prog->subcoffset = 0; | |
d6a28714 JH |
2780 | prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ |
2781 | } | |
97ca13b7 | 2782 | #ifdef DEBUGGING |
f73aaa43 | 2783 | PL_reg_starttry = *startposp; |
97ca13b7 | 2784 | #endif |
f73aaa43 | 2785 | prog->offs[0].start = *startposp - PL_bostr; |
d6a28714 | 2786 | prog->lastparen = 0; |
03994de8 | 2787 | prog->lastcloseparen = 0; |
d6a28714 | 2788 | PL_regsize = 0; |
d6a28714 JH |
2789 | |
2790 | /* XXXX What this code is doing here?!!! There should be no need | |
b93070ed | 2791 | to do this again and again, prog->lastparen should take care of |
3dd2943c | 2792 | this! --ilya*/ |
dafc8851 JH |
2793 | |
2794 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2795 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
b93070ed | 2796 | * prog->lastparen), is not needed at all by the test suite |
225593e1 DM |
2797 | * (op/regexp, op/pat, op/split), but that code is needed otherwise |
2798 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ | |
2799 | * Meanwhile, this code *is* needed for the | |
daf18116 JH |
2800 | * above-mentioned test suite tests to succeed. The common theme |
2801 | * on those tests seems to be returning null fields from matches. | |
225593e1 | 2802 | * --jhi updated by dapm */ |
dafc8851 | 2803 | #if 1 |
d6a28714 | 2804 | if (prog->nparens) { |
b93070ed | 2805 | regexp_paren_pair *pp = prog->offs; |
eb578fdb | 2806 | I32 i; |
b93070ed | 2807 | for (i = prog->nparens; i > (I32)prog->lastparen; i--) { |
f0ab9afb NC |
2808 | ++pp; |
2809 | pp->start = -1; | |
2810 | pp->end = -1; | |
d6a28714 JH |
2811 | } |
2812 | } | |
dafc8851 | 2813 | #endif |
02db2b7b | 2814 | REGCP_SET(lastcp); |
f73aaa43 DM |
2815 | result = regmatch(reginfo, *startposp, progi->program + 1); |
2816 | if (result != -1) { | |
2817 | prog->offs[0].end = result; | |
d6a28714 JH |
2818 | return 1; |
2819 | } | |
24b23f37 | 2820 | if (reginfo->cutpoint) |
f73aaa43 | 2821 | *startposp= reginfo->cutpoint; |
02db2b7b | 2822 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
2823 | return 0; |
2824 | } | |
2825 | ||
02db2b7b | 2826 | |
8ba1375e MJD |
2827 | #define sayYES goto yes |
2828 | #define sayNO goto no | |
262b90c4 | 2829 | #define sayNO_SILENT goto no_silent |
8ba1375e | 2830 | |
f9f4320a YO |
2831 | /* we dont use STMT_START/END here because it leads to |
2832 | "unreachable code" warnings, which are bogus, but distracting. */ | |
2833 | #define CACHEsayNO \ | |
c476f425 DM |
2834 | if (ST.cache_mask) \ |
2835 | PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \ | |
f9f4320a | 2836 | sayNO |
3298f257 | 2837 | |
a3621e74 | 2838 | /* this is used to determine how far from the left messages like |
265c4333 YO |
2839 | 'failed...' are printed. It should be set such that messages |
2840 | are inline with the regop output that created them. | |
a3621e74 | 2841 | */ |
265c4333 | 2842 | #define REPORT_CODE_OFF 32 |
a3621e74 YO |
2843 | |
2844 | ||
40a82448 DM |
2845 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
2846 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
9e137952 | 2847 | |
86545054 DM |
2848 | #define SLAB_FIRST(s) (&(s)->states[0]) |
2849 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
2850 | ||
5d9a96ca DM |
2851 | /* grab a new slab and return the first slot in it */ |
2852 | ||
2853 | STATIC regmatch_state * | |
2854 | S_push_slab(pTHX) | |
2855 | { | |
a35a87e7 | 2856 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2857 | dMY_CXT; |
2858 | #endif | |
5d9a96ca DM |
2859 | regmatch_slab *s = PL_regmatch_slab->next; |
2860 | if (!s) { | |
2861 | Newx(s, 1, regmatch_slab); | |
2862 | s->prev = PL_regmatch_slab; | |
2863 | s->next = NULL; | |
2864 | PL_regmatch_slab->next = s; | |
2865 | } | |
2866 | PL_regmatch_slab = s; | |
86545054 | 2867 | return SLAB_FIRST(s); |
5d9a96ca | 2868 | } |
5b47454d | 2869 | |
95b24440 | 2870 | |
40a82448 DM |
2871 | /* push a new state then goto it */ |
2872 | ||
4d5016e5 DM |
2873 | #define PUSH_STATE_GOTO(state, node, input) \ |
2874 | pushinput = input; \ | |
40a82448 DM |
2875 | scan = node; \ |
2876 | st->resume_state = state; \ | |
2877 | goto push_state; | |
2878 | ||
2879 | /* push a new state with success backtracking, then goto it */ | |
2880 | ||
4d5016e5 DM |
2881 | #define PUSH_YES_STATE_GOTO(state, node, input) \ |
2882 | pushinput = input; \ | |
40a82448 DM |
2883 | scan = node; \ |
2884 | st->resume_state = state; \ | |
2885 | goto push_yes_state; | |
2886 | ||
aa283a38 | 2887 | |
aa283a38 | 2888 | |
4d5016e5 | 2889 | |
d6a28714 | 2890 | /* |
95b24440 | 2891 | |
bf1f174e DM |
2892 | regmatch() - main matching routine |
2893 | ||
2894 | This is basically one big switch statement in a loop. We execute an op, | |
2895 | set 'next' to point the next op, and continue. If we come to a point which | |
2896 | we may need to backtrack to on failure such as (A|B|C), we push a | |
2897 | backtrack state onto the backtrack stack. On failure, we pop the top | |
2898 | state, and re-enter the loop at the state indicated. If there are no more | |
2899 | states to pop, we return failure. | |
2900 | ||
2901 | Sometimes we also need to backtrack on success; for example /A+/, where | |
2902 | after successfully matching one A, we need to go back and try to | |
2903 | match another one; similarly for lookahead assertions: if the assertion | |
2904 | completes successfully, we backtrack to the state just before the assertion | |
2905 | and then carry on. In these cases, the pushed state is marked as | |
2906 | 'backtrack on success too'. This marking is in fact done by a chain of | |
2907 | pointers, each pointing to the previous 'yes' state. On success, we pop to | |
2908 | the nearest yes state, discarding any intermediate failure-only states. | |
2909 | Sometimes a yes state is pushed just to force some cleanup code to be | |
2910 | called at the end of a successful match or submatch; e.g. (??{$re}) uses | |
2911 | it to free the inner regex. | |
2912 | ||
2913 | Note that failure backtracking rewinds the cursor position, while | |
2914 | success backtracking leaves it alone. | |
2915 | ||
2916 | A pattern is complete when the END op is executed, while a subpattern | |
2917 | such as (?=foo) is complete when the SUCCESS op is executed. Both of these | |
2918 | ops trigger the "pop to last yes state if any, otherwise return true" | |
2919 | behaviour. | |
2920 | ||
2921 | A common convention in this function is to use A and B to refer to the two | |
2922 | subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is | |
2923 | the subpattern to be matched possibly multiple times, while B is the entire | |
2924 | rest of the pattern. Variable and state names reflect this convention. | |
2925 | ||
2926 | The states in the main switch are the union of ops and failure/success of | |
2927 | substates associated with with that op. For example, IFMATCH is the op | |
2928 | that does lookahead assertions /(?=A)B/ and so the IFMATCH state means | |
2929 | 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just | |
2930 | successfully matched A and IFMATCH_A_fail is a state saying that we have | |
2931 | just failed to match A. Resume states always come in pairs. The backtrack | |
2932 | state we push is marked as 'IFMATCH_A', but when that is popped, we resume | |
2933 | at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking | |
2934 | on success or failure. | |
2935 | ||
2936 | The struct that holds a backtracking state is actually a big union, with | |
2937 | one variant for each major type of op. The variable st points to the | |
2938 | top-most backtrack struct. To make the code clearer, within each | |
2939 | block of code we #define ST to alias the relevant union. | |
2940 | ||
2941 | Here's a concrete example of a (vastly oversimplified) IFMATCH | |
2942 | implementation: | |
2943 | ||
2944 | switch (state) { | |
2945 | .... | |
2946 | ||
2947 | #define ST st->u.ifmatch | |
2948 | ||
2949 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B | |
2950 | ST.foo = ...; // some state we wish to save | |
95b24440 | 2951 | ... |
bf1f174e DM |
2952 | // push a yes backtrack state with a resume value of |
2953 | // IFMATCH_A/IFMATCH_A_fail, then continue execution at the | |
2954 | // first node of A: | |
4d5016e5 | 2955 | PUSH_YES_STATE_GOTO(IFMATCH_A, A, newinput); |
bf1f174e DM |
2956 | // NOTREACHED |
2957 | ||
2958 | case IFMATCH_A: // we have successfully executed A; now continue with B | |
2959 | next = B; | |
2960 | bar = ST.foo; // do something with the preserved value | |
2961 | break; | |
2962 | ||
2963 | case IFMATCH_A_fail: // A failed, so the assertion failed | |
2964 | ...; // do some housekeeping, then ... | |
2965 | sayNO; // propagate the failure | |
2966 | ||
2967 | #undef ST | |
95b24440 | 2968 | |
bf1f174e DM |
2969 | ... |
2970 | } | |
95b24440 | 2971 | |
bf1f174e DM |
2972 | For any old-timers reading this who are familiar with the old recursive |
2973 | approach, the code above is equivalent to: | |
95b24440 | 2974 | |
bf1f174e DM |
2975 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B |
2976 | { | |
2977 | int foo = ... | |
95b24440 | 2978 | ... |
bf1f174e DM |
2979 | if (regmatch(A)) { |
2980 | next = B; | |
2981 | bar = foo; | |
2982 | break; | |
95b24440 | 2983 | } |
bf1f174e DM |
2984 | ...; // do some housekeeping, then ... |
2985 | sayNO; // propagate the failure | |
95b24440 | 2986 | } |
bf1f174e DM |
2987 | |
2988 | The topmost backtrack state, pointed to by st, is usually free. If you | |
2989 | want to claim it, populate any ST.foo fields in it with values you wish to | |
2990 | save, then do one of | |
2991 | ||
4d5016e5 DM |
2992 | PUSH_STATE_GOTO(resume_state, node, newinput); |
2993 | PUSH_YES_STATE_GOTO(resume_state, node, newinput); | |
bf1f174e DM |
2994 | |
2995 | which sets that backtrack state's resume value to 'resume_state', pushes a | |
2996 | new free entry to the top of the backtrack stack, then goes to 'node'. | |
2997 | On backtracking, the free slot is popped, and the saved state becomes the | |
2998 | new free state. An ST.foo field in this new top state can be temporarily | |
2999 | accessed to retrieve values, but once the main loop is re-entered, it | |
3000 | becomes available for reuse. | |
3001 | ||
3002 | Note that the depth of the backtrack stack constantly increases during the | |
3003 | left-to-right execution of the pattern, rather than going up and down with | |
3004 | the pattern nesting. For example the stack is at its maximum at Z at the | |
3005 | end of the pattern, rather than at X in the following: | |
3006 | ||
3007 | /(((X)+)+)+....(Y)+....Z/ | |
3008 | ||
3009 | The only exceptions to this are lookahead/behind assertions and the cut, | |
3010 | (?>A), which pop all the backtrack states associated with A before | |
3011 | continuing. | |
3012 | ||
486ec47a | 3013 | Backtrack state structs are allocated in slabs of about 4K in size. |
bf1f174e DM |
3014 | PL_regmatch_state and st always point to the currently active state, |
3015 | and PL_regmatch_slab points to the slab currently containing | |
3016 | PL_regmatch_state. The first time regmatch() is called, the first slab is | |
3017 | allocated, and is never freed until interpreter destruction. When the slab | |
3018 | is full, a new one is allocated and chained to the end. At exit from | |
3019 | regmatch(), slabs allocated since entry are freed. | |
3020 | ||
3021 | */ | |
95b24440 | 3022 | |
40a82448 | 3023 | |
5bc10b2c | 3024 | #define DEBUG_STATE_pp(pp) \ |
265c4333 | 3025 | DEBUG_STATE_r({ \ |
f2ed9b32 | 3026 | DUMP_EXEC_POS(locinput, scan, utf8_target); \ |
5bc10b2c | 3027 | PerlIO_printf(Perl_debug_log, \ |
5d458dd8 | 3028 | " %*s"pp" %s%s%s%s%s\n", \ |
5bc10b2c | 3029 | depth*2, "", \ |
13d6edb4 | 3030 | PL_reg_name[st->resume_state], \ |
5d458dd8 YO |
3031 | ((st==yes_state||st==mark_state) ? "[" : ""), \ |
3032 | ((st==yes_state) ? "Y" : ""), \ | |
3033 | ((st==mark_state) ? "M" : ""), \ | |
3034 | ((st==yes_state||st==mark_state) ? "]" : "") \ | |
3035 | ); \ | |
265c4333 | 3036 | }); |
5bc10b2c | 3037 | |
40a82448 | 3038 | |
3dab1dad | 3039 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 3040 | |
3df15adc | 3041 | #ifdef DEBUGGING |
5bc10b2c | 3042 | |
ab3bbdeb | 3043 | STATIC void |
f2ed9b32 | 3044 | S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target, |
ab3bbdeb YO |
3045 | const char *start, const char *end, const char *blurb) |
3046 | { | |
efd26800 | 3047 | const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; |
7918f24d NC |
3048 | |
3049 | PERL_ARGS_ASSERT_DEBUG_START_MATCH; | |
3050 | ||
ab3bbdeb YO |
3051 | if (!PL_colorset) |
3052 | reginitcolors(); | |
3053 | { | |
3054 | RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), | |
d2c6dc5e | 3055 | RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); |
ab3bbdeb | 3056 | |
f2ed9b32 | 3057 | RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb YO |
3058 | start, end - start, 60); |
3059 | ||
3060 | PerlIO_printf(Perl_debug_log, | |
3061 | "%s%s REx%s %s against %s\n", | |
3062 | PL_colors[4], blurb, PL_colors[5], s0, s1); | |
3063 | ||
f2ed9b32 | 3064 | if (utf8_target||utf8_pat) |
1de06328 YO |
3065 | PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", |
3066 | utf8_pat ? "pattern" : "", | |
f2ed9b32 KW |
3067 | utf8_pat && utf8_target ? " and " : "", |
3068 | utf8_target ? "string" : "" | |
ab3bbdeb YO |
3069 | ); |
3070 | } | |
3071 | } | |
3df15adc YO |
3072 | |
3073 | STATIC void | |
786e8c11 YO |
3074 | S_dump_exec_pos(pTHX_ const char *locinput, |
3075 | const regnode *scan, | |
3076 | const char *loc_regeol, | |
3077 | const char *loc_bostr, | |
3078 | const char *loc_reg_starttry, | |
f2ed9b32 | 3079 | const bool utf8_target) |
07be1b83 | 3080 | { |
786e8c11 | 3081 | const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; |
07be1b83 | 3082 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ |
786e8c11 | 3083 | int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); |
07be1b83 YO |
3084 | /* The part of the string before starttry has one color |
3085 | (pref0_len chars), between starttry and current | |
3086 | position another one (pref_len - pref0_len chars), | |
3087 | after the current position the third one. | |
3088 | We assume that pref0_len <= pref_len, otherwise we | |
3089 | decrease pref0_len. */ | |
786e8c11 YO |
3090 | int pref_len = (locinput - loc_bostr) > (5 + taill) - l |
3091 | ? (5 + taill) - l : locinput - loc_bostr; | |
07be1b83 YO |
3092 | int pref0_len; |
3093 | ||
7918f24d NC |
3094 | PERL_ARGS_ASSERT_DUMP_EXEC_POS; |
3095 | ||
f2ed9b32 | 3096 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) |
07be1b83 | 3097 | pref_len++; |
786e8c11 YO |
3098 | pref0_len = pref_len - (locinput - loc_reg_starttry); |
3099 | if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) | |
3100 | l = ( loc_regeol - locinput > (5 + taill) - pref_len | |
3101 | ? (5 + taill) - pref_len : loc_regeol - locinput); | |
f2ed9b32 | 3102 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) |
07be1b83 YO |
3103 | l--; |
3104 | if (pref0_len < 0) | |
3105 | pref0_len = 0; | |
3106 | if (pref0_len > pref_len) | |
3107 | pref0_len = pref_len; | |
3108 | { | |
f2ed9b32 | 3109 | const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0; |
0df25f3d | 3110 | |
ab3bbdeb | 3111 | RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), |
1de06328 | 3112 | (locinput - pref_len),pref0_len, 60, 4, 5); |
0df25f3d | 3113 | |
ab3bbdeb | 3114 | RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), |
3df15adc | 3115 | (locinput - pref_len + pref0_len), |
1de06328 | 3116 | pref_len - pref0_len, 60, 2, 3); |
0df25f3d | 3117 | |
ab3bbdeb | 3118 | RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), |
1de06328 | 3119 | locinput, loc_regeol - locinput, 10, 0, 1); |
0df25f3d | 3120 | |
1de06328 | 3121 | const STRLEN tlen=len0+len1+len2; |
3df15adc | 3122 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb | 3123 | "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", |
786e8c11 | 3124 | (IV)(locinput - loc_bostr), |
07be1b83 | 3125 | len0, s0, |
07be1b83 | 3126 | len1, s1, |
07be1b83 | 3127 | (docolor ? "" : "> <"), |
07be1b83 | 3128 | len2, s2, |
f9f4320a | 3129 | (int)(tlen > 19 ? 0 : 19 - tlen), |
07be1b83 YO |
3130 | ""); |
3131 | } | |
3132 | } | |
3df15adc | 3133 | |
07be1b83 YO |
3134 | #endif |
3135 | ||
0a4db386 YO |
3136 | /* reg_check_named_buff_matched() |
3137 | * Checks to see if a named buffer has matched. The data array of | |
3138 | * buffer numbers corresponding to the buffer is expected to reside | |
3139 | * in the regexp->data->data array in the slot stored in the ARG() of | |
3140 | * node involved. Note that this routine doesn't actually care about the | |
3141 | * name, that information is not preserved from compilation to execution. | |
3142 | * Returns the index of the leftmost defined buffer with the given name | |
3143 | * or 0 if non of the buffers matched. | |
3144 | */ | |
3145 | STATIC I32 | |
7918f24d NC |
3146 | S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) |
3147 | { | |
0a4db386 | 3148 | I32 n; |
f8fc2ecf | 3149 | RXi_GET_DECL(rex,rexi); |
ad64d0ec | 3150 | SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
0a4db386 | 3151 | I32 *nums=(I32*)SvPVX(sv_dat); |
7918f24d NC |
3152 | |
3153 | PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; | |
3154 | ||
0a4db386 | 3155 | for ( n=0; n<SvIVX(sv_dat); n++ ) { |
b93070ed DM |
3156 | if ((I32)rex->lastparen >= nums[n] && |
3157 | rex->offs[nums[n]].end != -1) | |
0a4db386 YO |
3158 | { |
3159 | return nums[n]; | |
3160 | } | |
3161 | } | |
3162 | return 0; | |
3163 | } | |
3164 | ||
2f554ef7 DM |
3165 | |
3166 | /* free all slabs above current one - called during LEAVE_SCOPE */ | |
3167 | ||
3168 | STATIC void | |
3169 | S_clear_backtrack_stack(pTHX_ void *p) | |
3170 | { | |
3171 | regmatch_slab *s = PL_regmatch_slab->next; | |
3172 | PERL_UNUSED_ARG(p); | |
3173 | ||
3174 | if (!s) | |
3175 | return; | |
3176 | PL_regmatch_slab->next = NULL; | |
3177 | while (s) { | |
3178 | regmatch_slab * const osl = s; | |
3179 | s = s->next; | |
3180 | Safefree(osl); | |
3181 | } | |
3182 | } | |
3183 | ||
3184 | ||
f73aaa43 DM |
3185 | /* returns -1 on failure, $+[0] on success */ |
3186 | STATIC I32 | |
3187 | S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) | |
d6a28714 | 3188 | { |
a35a87e7 | 3189 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
3190 | dMY_CXT; |
3191 | #endif | |
27da23d5 | 3192 | dVAR; |
eb578fdb | 3193 | const bool utf8_target = PL_reg_match_utf8; |
4ad0818d | 3194 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
288b8c02 NC |
3195 | REGEXP *rex_sv = reginfo->prog; |
3196 | regexp *rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 3197 | RXi_GET_DECL(rex,rexi); |
2f554ef7 | 3198 | I32 oldsave; |
5d9a96ca | 3199 | /* the current state. This is a cached copy of PL_regmatch_state */ |
eb578fdb | 3200 | regmatch_state *st; |
5d9a96ca | 3201 | /* cache heavy used fields of st in registers */ |
eb578fdb KW |
3202 | regnode *scan; |
3203 | regnode *next; | |
3204 | U32 n = 0; /* general value; init to avoid compiler warning */ | |
3205 | I32 ln = 0; /* len or last; init to avoid compiler warning */ | |
d60de1d1 | 3206 | char *locinput = startpos; |
4d5016e5 | 3207 | char *pushinput; /* where to continue after a PUSH */ |
eb578fdb | 3208 | I32 nextchr; /* is always set to UCHARAT(locinput) */ |
24d3c4a9 | 3209 | |
b69b0499 | 3210 | bool result = 0; /* return value of S_regmatch */ |
24d3c4a9 | 3211 | int depth = 0; /* depth of backtrack stack */ |
4b196cd4 YO |
3212 | U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */ |
3213 | const U32 max_nochange_depth = | |
3214 | (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? | |
3215 | 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; | |
77cb431f DM |
3216 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
3217 | subpattern */ | |
e2e6a0f1 YO |
3218 | /* mark_state piggy backs on the yes_state logic so that when we unwind |
3219 | the stack on success we can update the mark_state as we go */ | |
3220 | regmatch_state *mark_state = NULL; /* last mark state we have seen */ | |
faec1544 | 3221 | regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ |
b8591aee | 3222 | struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ |
40a82448 | 3223 | U32 state_num; |
5d458dd8 YO |
3224 | bool no_final = 0; /* prevent failure from backtracking? */ |
3225 | bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ | |
d60de1d1 | 3226 | char *startpoint = locinput; |
5d458dd8 YO |
3227 | SV *popmark = NULL; /* are we looking for a mark? */ |
3228 | SV *sv_commit = NULL; /* last mark name seen in failure */ | |
3229 | SV *sv_yes_mark = NULL; /* last mark name we have seen | |
486ec47a | 3230 | during a successful match */ |
5d458dd8 YO |
3231 | U32 lastopen = 0; /* last open we saw */ |
3232 | bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; | |
19b95bf0 | 3233 | SV* const oreplsv = GvSV(PL_replgv); |
24d3c4a9 DM |
3234 | /* these three flags are set by various ops to signal information to |
3235 | * the very next op. They have a useful lifetime of exactly one loop | |
3236 | * iteration, and are not preserved or restored by state pushes/pops | |
3237 | */ | |
3238 | bool sw = 0; /* the condition value in (?(cond)a|b) */ | |
3239 | bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ | |
3240 | int logical = 0; /* the following EVAL is: | |
3241 | 0: (?{...}) | |
3242 | 1: (?(?{...})X|Y) | |
3243 | 2: (??{...}) | |
3244 | or the following IFMATCH/UNLESSM is: | |
3245 | false: plain (?=foo) | |
3246 | true: used as a condition: (?(?=foo)) | |
3247 | */ | |
81ed78b2 DM |
3248 | PAD* last_pad = NULL; |
3249 | dMULTICALL; | |
3250 | I32 gimme = G_SCALAR; | |
3251 | CV *caller_cv = NULL; /* who called us */ | |
3252 | CV *last_pushed_cv = NULL; /* most recently called (?{}) CV */ | |
74088413 | 3253 | CHECKPOINT runops_cp; /* savestack position before executing EVAL */ |
81ed78b2 | 3254 | |
95b24440 | 3255 | #ifdef DEBUGGING |
e68ec53f | 3256 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
3257 | #endif |
3258 | ||
81ed78b2 DM |
3259 | /* shut up 'may be used uninitialized' compiler warnings for dMULTICALL */ |
3260 | multicall_oldcatch = 0; | |
3261 | multicall_cv = NULL; | |
3262 | cx = NULL; | |
4f8dbb2d JL |
3263 | PERL_UNUSED_VAR(multicall_cop); |
3264 | PERL_UNUSED_VAR(newsp); | |
81ed78b2 DM |
3265 | |
3266 | ||
7918f24d NC |
3267 | PERL_ARGS_ASSERT_REGMATCH; |
3268 | ||
3b57cd43 | 3269 | DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ |
24b23f37 | 3270 | PerlIO_printf(Perl_debug_log,"regmatch start\n"); |
3b57cd43 | 3271 | })); |
5d9a96ca DM |
3272 | /* on first ever call to regmatch, allocate first slab */ |
3273 | if (!PL_regmatch_slab) { | |
3274 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
3275 | PL_regmatch_slab->prev = NULL; | |
3276 | PL_regmatch_slab->next = NULL; | |
86545054 | 3277 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); |
5d9a96ca DM |
3278 | } |
3279 | ||
2f554ef7 DM |
3280 | oldsave = PL_savestack_ix; |
3281 | SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL); | |
3282 | SAVEVPTR(PL_regmatch_slab); | |
3283 | SAVEVPTR(PL_regmatch_state); | |
5d9a96ca DM |
3284 | |
3285 | /* grab next free state slot */ | |
3286 | st = ++PL_regmatch_state; | |
86545054 | 3287 | if (st > SLAB_LAST(PL_regmatch_slab)) |
5d9a96ca DM |
3288 | st = PL_regmatch_state = S_push_slab(aTHX); |
3289 | ||
d6a28714 JH |
3290 | /* Note that nextchr is a byte even in UTF */ |
3291 | nextchr = UCHARAT(locinput); | |
3292 | scan = prog; | |
3293 | while (scan != NULL) { | |
8ba1375e | 3294 | |
a3621e74 | 3295 | DEBUG_EXECUTE_r( { |
6136c704 | 3296 | SV * const prop = sv_newmortal(); |
1de06328 | 3297 | regnode *rnext=regnext(scan); |
f2ed9b32 | 3298 | DUMP_EXEC_POS( locinput, scan, utf8_target ); |
32fc9b6a | 3299 | regprop(rex, prop, scan); |
07be1b83 YO |
3300 | |
3301 | PerlIO_printf(Perl_debug_log, | |
3302 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
f8fc2ecf | 3303 | (IV)(scan - rexi->program), depth*2, "", |
07be1b83 | 3304 | SvPVX_const(prop), |
1de06328 | 3305 | (PL_regkind[OP(scan)] == END || !rnext) ? |
f8fc2ecf | 3306 | 0 : (IV)(rnext - rexi->program)); |
2a782b5b | 3307 | }); |
d6a28714 JH |
3308 | |
3309 | next = scan + NEXT_OFF(scan); | |
3310 | if (next == scan) | |
3311 | next = NULL; | |
40a82448 | 3312 | state_num = OP(scan); |
d6a28714 | 3313 | |
40a82448 | 3314 | reenter_switch: |
34a81e2b | 3315 | |
3640db6b | 3316 | nextchr = UCHARAT(locinput); |
bf798dc4 DM |
3317 | assert(nextchr >= 0); |
3318 | ||
40a82448 | 3319 | switch (state_num) { |
3c0563b9 | 3320 | case BOL: /* /^../ */ |
7fba1cd6 | 3321 | if (locinput == PL_bostr) |
d6a28714 | 3322 | { |
3b0527fe | 3323 | /* reginfo->till = reginfo->bol; */ |
b8c5462f JH |
3324 | break; |
3325 | } | |
d6a28714 | 3326 | sayNO; |
3c0563b9 DM |
3327 | |
3328 | case MBOL: /* /^../m */ | |
12d33761 HS |
3329 | if (locinput == PL_bostr || |
3330 | ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n')) | |
d6a28714 | 3331 | { |
b8c5462f JH |
3332 | break; |
3333 | } | |
d6a28714 | 3334 | sayNO; |
3c0563b9 DM |
3335 | |
3336 | case SBOL: /* /^../s */ | |
c2a73568 | 3337 | if (locinput == PL_bostr) |
b8c5462f | 3338 | break; |
d6a28714 | 3339 | sayNO; |
3c0563b9 DM |
3340 | |
3341 | case GPOS: /* \G */ | |
3b0527fe | 3342 | if (locinput == reginfo->ganch) |
d6a28714 JH |
3343 | break; |
3344 | sayNO; | |
ee9b8eae | 3345 | |
3c0563b9 | 3346 | case KEEPS: /* \K */ |
ee9b8eae | 3347 | /* update the startpoint */ |
b93070ed | 3348 | st->u.keeper.val = rex->offs[0].start; |
b93070ed | 3349 | rex->offs[0].start = locinput - PL_bostr; |
4d5016e5 | 3350 | PUSH_STATE_GOTO(KEEPS_next, next, locinput); |
ee9b8eae YO |
3351 | /*NOT-REACHED*/ |
3352 | case KEEPS_next_fail: | |
3353 | /* rollback the start point change */ | |
b93070ed | 3354 | rex->offs[0].start = st->u.keeper.val; |
ee9b8eae YO |
3355 | sayNO_SILENT; |
3356 | /*NOT-REACHED*/ | |
3c0563b9 DM |
3357 | |
3358 | case EOL: /* /..$/ */ | |
d6a28714 | 3359 | goto seol; |
3c0563b9 DM |
3360 | |
3361 | case MEOL: /* /..$/m */ | |
d6a28714 | 3362 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') |
b8c5462f | 3363 | sayNO; |
b8c5462f | 3364 | break; |
3c0563b9 DM |
3365 | |
3366 | case SEOL: /* /..$/s */ | |
d6a28714 JH |
3367 | seol: |
3368 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') | |
b8c5462f | 3369 | sayNO; |
d6a28714 | 3370 | if (PL_regeol - locinput > 1) |
b8c5462f | 3371 | sayNO; |
b8c5462f | 3372 | break; |
3c0563b9 DM |
3373 | |
3374 | case EOS: /* \z */ | |
d6a28714 | 3375 | if (PL_regeol != locinput) |
b8c5462f | 3376 | sayNO; |
d6a28714 | 3377 | break; |
3c0563b9 DM |
3378 | |
3379 | case SANY: /* /./s */ | |
d6a28714 | 3380 | if (!nextchr && locinput >= PL_regeol) |
4633a7c4 | 3381 | sayNO; |
28b98f76 | 3382 | goto increment_locinput; |
3c0563b9 DM |
3383 | |
3384 | case CANY: /* \C */ | |
f33976b4 DB |
3385 | if (!nextchr && locinput >= PL_regeol) |
3386 | sayNO; | |
3640db6b | 3387 | locinput++; |
a0d0e21e | 3388 | break; |
3c0563b9 DM |
3389 | |
3390 | case REG_ANY: /* /./ */ | |
1aa99e6b IH |
3391 | if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n') |
3392 | sayNO; | |
28b98f76 DM |
3393 | goto increment_locinput; |
3394 | ||
166ba7cd DM |
3395 | |
3396 | #undef ST | |
3397 | #define ST st->u.trie | |
3c0563b9 | 3398 | case TRIEC: /* (ab|cd) with known charclass */ |
786e8c11 YO |
3399 | /* In this case the charclass data is available inline so |
3400 | we can fail fast without a lot of extra overhead. | |
3401 | */ | |
b9b31e9d | 3402 | if(!ANYOF_BITMAP_TEST(scan, nextchr)) { |
fab2782b YO |
3403 | DEBUG_EXECUTE_r( |
3404 | PerlIO_printf(Perl_debug_log, | |
3405 | "%*s %sfailed to match trie start class...%s\n", | |
3406 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) | |
3407 | ); | |
3408 | sayNO_SILENT; | |
118e2215 | 3409 | assert(0); /* NOTREACHED */ |
786e8c11 YO |
3410 | } |
3411 | /* FALL THROUGH */ | |
3c0563b9 | 3412 | case TRIE: /* (ab|cd) */ |
2e64971a DM |
3413 | /* the basic plan of execution of the trie is: |
3414 | * At the beginning, run though all the states, and | |
3415 | * find the longest-matching word. Also remember the position | |
3416 | * of the shortest matching word. For example, this pattern: | |
3417 | * 1 2 3 4 5 | |
3418 | * ab|a|x|abcd|abc | |
3419 | * when matched against the string "abcde", will generate | |
3420 | * accept states for all words except 3, with the longest | |
895cc420 | 3421 | * matching word being 4, and the shortest being 2 (with |
2e64971a DM |
3422 | * the position being after char 1 of the string). |
3423 | * | |
3424 | * Then for each matching word, in word order (i.e. 1,2,4,5), | |
3425 | * we run the remainder of the pattern; on each try setting | |
3426 | * the current position to the character following the word, | |
3427 | * returning to try the next word on failure. | |
3428 | * | |
3429 | * We avoid having to build a list of words at runtime by | |
3430 | * using a compile-time structure, wordinfo[].prev, which | |
3431 | * gives, for each word, the previous accepting word (if any). | |
3432 | * In the case above it would contain the mappings 1->2, 2->0, | |
3433 | * 3->0, 4->5, 5->1. We can use this table to generate, from | |
3434 | * the longest word (4 above), a list of all words, by | |
3435 | * following the list of prev pointers; this gives us the | |
3436 | * unordered list 4,5,1,2. Then given the current word we have | |
3437 | * just tried, we can go through the list and find the | |
3438 | * next-biggest word to try (so if we just failed on word 2, | |
3439 | * the next in the list is 4). | |
3440 | * | |
3441 | * Since at runtime we don't record the matching position in | |
3442 | * the string for each word, we have to work that out for | |
3443 | * each word we're about to process. The wordinfo table holds | |
3444 | * the character length of each word; given that we recorded | |
3445 | * at the start: the position of the shortest word and its | |
3446 | * length in chars, we just need to move the pointer the | |
3447 | * difference between the two char lengths. Depending on | |
3448 | * Unicode status and folding, that's cheap or expensive. | |
3449 | * | |
3450 | * This algorithm is optimised for the case where are only a | |
3451 | * small number of accept states, i.e. 0,1, or maybe 2. | |
3452 | * With lots of accepts states, and having to try all of them, | |
3453 | * it becomes quadratic on number of accept states to find all | |
3454 | * the next words. | |
3455 | */ | |
3456 | ||
3dab1dad | 3457 | { |
07be1b83 | 3458 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
a0a388a1 | 3459 | DECL_TRIE_TYPE(scan); |
3dab1dad YO |
3460 | |
3461 | /* what trie are we using right now */ | |
be8e71aa | 3462 | reg_trie_data * const trie |
f8fc2ecf | 3463 | = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; |
85fbaab2 | 3464 | HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); |
3dab1dad | 3465 | U32 state = trie->startstate; |
166ba7cd | 3466 | |
b9b31e9d | 3467 | if (trie->bitmap && !TRIE_BITMAP_TEST(trie, nextchr) ) { |
3dab1dad YO |
3468 | if (trie->states[ state ].wordnum) { |
3469 | DEBUG_EXECUTE_r( | |
3470 | PerlIO_printf(Perl_debug_log, | |
3471 | "%*s %smatched empty string...%s\n", | |
5bc10b2c | 3472 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad | 3473 | ); |
20dbff7c YO |
3474 | if (!trie->jump) |
3475 | break; | |
3dab1dad YO |
3476 | } else { |
3477 | DEBUG_EXECUTE_r( | |
3478 | PerlIO_printf(Perl_debug_log, | |
786e8c11 | 3479 | "%*s %sfailed to match trie start class...%s\n", |
5bc10b2c | 3480 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad YO |
3481 | ); |
3482 | sayNO_SILENT; | |
3483 | } | |
3484 | } | |
166ba7cd | 3485 | |
786e8c11 YO |
3486 | { |
3487 | U8 *uc = ( U8* )locinput; | |
3488 | ||
3489 | STRLEN len = 0; | |
3490 | STRLEN foldlen = 0; | |
3491 | U8 *uscan = (U8*)NULL; | |
786e8c11 | 3492 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
2e64971a DM |
3493 | U32 charcount = 0; /* how many input chars we have matched */ |
3494 | U32 accepted = 0; /* have we seen any accepting states? */ | |
786e8c11 | 3495 | |
786e8c11 | 3496 | ST.jump = trie->jump; |
786e8c11 | 3497 | ST.me = scan; |
2e64971a DM |
3498 | ST.firstpos = NULL; |
3499 | ST.longfold = FALSE; /* char longer if folded => it's harder */ | |
3500 | ST.nextword = 0; | |
3501 | ||
3502 | /* fully traverse the TRIE; note the position of the | |
3503 | shortest accept state and the wordnum of the longest | |
3504 | accept state */ | |
07be1b83 | 3505 | |
a3621e74 | 3506 | while ( state && uc <= (U8*)PL_regeol ) { |
786e8c11 | 3507 | U32 base = trie->states[ state ].trans.base; |
f9f4320a | 3508 | UV uvc = 0; |
acb909b4 | 3509 | U16 charid = 0; |
2e64971a DM |
3510 | U16 wordnum; |
3511 | wordnum = trie->states[ state ].wordnum; | |
3512 | ||
3513 | if (wordnum) { /* it's an accept state */ | |
3514 | if (!accepted) { | |
3515 | accepted = 1; | |
3516 | /* record first match position */ | |
3517 | if (ST.longfold) { | |
3518 | ST.firstpos = (U8*)locinput; | |
3519 | ST.firstchars = 0; | |
5b47454d | 3520 | } |
2e64971a DM |
3521 | else { |
3522 | ST.firstpos = uc; | |
3523 | ST.firstchars = charcount; | |
3524 | } | |
3525 | } | |
3526 | if (!ST.nextword || wordnum < ST.nextword) | |
3527 | ST.nextword = wordnum; | |
3528 | ST.topword = wordnum; | |
786e8c11 | 3529 | } |
a3621e74 | 3530 | |
07be1b83 | 3531 | DEBUG_TRIE_EXECUTE_r({ |
f2ed9b32 | 3532 | DUMP_EXEC_POS( (char *)uc, scan, utf8_target ); |
a3621e74 | 3533 | PerlIO_printf( Perl_debug_log, |
2e64971a | 3534 | "%*s %sState: %4"UVxf" Accepted: %c ", |
5bc10b2c | 3535 | 2+depth * 2, "", PL_colors[4], |
2e64971a | 3536 | (UV)state, (accepted ? 'Y' : 'N')); |
07be1b83 | 3537 | }); |
a3621e74 | 3538 | |
2e64971a | 3539 | /* read a char and goto next state */ |
a3621e74 | 3540 | if ( base ) { |
6dd2be57 | 3541 | I32 offset; |
55eed653 NC |
3542 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
3543 | uscan, len, uvc, charid, foldlen, | |
3544 | foldbuf, uniflags); | |
2e64971a DM |
3545 | charcount++; |
3546 | if (foldlen>0) | |
3547 | ST.longfold = TRUE; | |
5b47454d | 3548 | if (charid && |
6dd2be57 DM |
3549 | ( ((offset = |
3550 | base + charid - 1 - trie->uniquecharcount)) >= 0) | |
3551 | ||
3552 | && ((U32)offset < trie->lasttrans) | |
3553 | && trie->trans[offset].check == state) | |
5b47454d | 3554 | { |
6dd2be57 | 3555 | state = trie->trans[offset].next; |
5b47454d DM |
3556 | } |
3557 | else { | |
3558 | state = 0; | |
3559 | } | |
3560 | uc += len; | |
3561 | ||
3562 | } | |
3563 | else { | |
a3621e74 YO |
3564 | state = 0; |
3565 | } | |
3566 | DEBUG_TRIE_EXECUTE_r( | |
e4584336 | 3567 | PerlIO_printf( Perl_debug_log, |
786e8c11 | 3568 | "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n", |
e4584336 | 3569 | charid, uvc, (UV)state, PL_colors[5] ); |
a3621e74 YO |
3570 | ); |
3571 | } | |
2e64971a | 3572 | if (!accepted) |
a3621e74 | 3573 | sayNO; |
a3621e74 | 3574 | |
2e64971a DM |
3575 | /* calculate total number of accept states */ |
3576 | { | |
3577 | U16 w = ST.topword; | |
3578 | accepted = 0; | |
3579 | while (w) { | |
3580 | w = trie->wordinfo[w].prev; | |
3581 | accepted++; | |
3582 | } | |
3583 | ST.accepted = accepted; | |
3584 | } | |
3585 | ||
166ba7cd DM |
3586 | DEBUG_EXECUTE_r( |
3587 | PerlIO_printf( Perl_debug_log, | |
3588 | "%*s %sgot %"IVdf" possible matches%s\n", | |
5bc10b2c | 3589 | REPORT_CODE_OFF + depth * 2, "", |
166ba7cd DM |
3590 | PL_colors[4], (IV)ST.accepted, PL_colors[5] ); |
3591 | ); | |
2e64971a | 3592 | goto trie_first_try; /* jump into the fail handler */ |
786e8c11 | 3593 | }} |
118e2215 | 3594 | assert(0); /* NOTREACHED */ |
2e64971a DM |
3595 | |
3596 | case TRIE_next_fail: /* we failed - try next alternative */ | |
a059a757 DM |
3597 | { |
3598 | U8 *uc; | |
fae667d5 YO |
3599 | if ( ST.jump) { |
3600 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 | 3601 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
fae667d5 | 3602 | } |
2e64971a DM |
3603 | if (!--ST.accepted) { |
3604 | DEBUG_EXECUTE_r({ | |
3605 | PerlIO_printf( Perl_debug_log, | |
3606 | "%*s %sTRIE failed...%s\n", | |
3607 | REPORT_CODE_OFF+depth*2, "", | |
3608 | PL_colors[4], | |
3609 | PL_colors[5] ); | |
3610 | }); | |
3611 | sayNO_SILENT; | |
3612 | } | |
3613 | { | |
3614 | /* Find next-highest word to process. Note that this code | |
3615 | * is O(N^2) per trie run (O(N) per branch), so keep tight */ | |
eb578fdb KW |
3616 | U16 min = 0; |
3617 | U16 word; | |
3618 | U16 const nextword = ST.nextword; | |
3619 | reg_trie_wordinfo * const wordinfo | |
2e64971a DM |
3620 | = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo; |
3621 | for (word=ST.topword; word; word=wordinfo[word].prev) { | |
3622 | if (word > nextword && (!min || word < min)) | |
3623 | min = word; | |
3624 | } | |
3625 | ST.nextword = min; | |
3626 | } | |
3627 | ||
fae667d5 | 3628 | trie_first_try: |
5d458dd8 YO |
3629 | if (do_cutgroup) { |
3630 | do_cutgroup = 0; | |
3631 | no_final = 0; | |
3632 | } | |
fae667d5 YO |
3633 | |
3634 | if ( ST.jump) { | |
b93070ed | 3635 | ST.lastparen = rex->lastparen; |
f6033a9d | 3636 | ST.lastcloseparen = rex->lastcloseparen; |
fae667d5 | 3637 | REGCP_SET(ST.cp); |
2e64971a | 3638 | } |
a3621e74 | 3639 | |
2e64971a | 3640 | /* find start char of end of current word */ |
166ba7cd | 3641 | { |
2e64971a | 3642 | U32 chars; /* how many chars to skip */ |
2e64971a DM |
3643 | reg_trie_data * const trie |
3644 | = (reg_trie_data*)rexi->data->data[ARG(ST.me)]; | |
3645 | ||
3646 | assert((trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3647 | >= ST.firstchars); | |
3648 | chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3649 | - ST.firstchars; | |
a059a757 | 3650 | uc = ST.firstpos; |
2e64971a DM |
3651 | |
3652 | if (ST.longfold) { | |
3653 | /* the hard option - fold each char in turn and find | |
3654 | * its folded length (which may be different */ | |
3655 | U8 foldbuf[UTF8_MAXBYTES_CASE + 1]; | |
3656 | STRLEN foldlen; | |
3657 | STRLEN len; | |
d9a396a3 | 3658 | UV uvc; |
2e64971a DM |
3659 | U8 *uscan; |
3660 | ||
3661 | while (chars) { | |
f2ed9b32 | 3662 | if (utf8_target) { |
2e64971a DM |
3663 | uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len, |
3664 | uniflags); | |
3665 | uc += len; | |
3666 | } | |
3667 | else { | |
3668 | uvc = *uc; | |
3669 | uc++; | |
3670 | } | |
3671 | uvc = to_uni_fold(uvc, foldbuf, &foldlen); | |
3672 | uscan = foldbuf; | |
3673 | while (foldlen) { | |
3674 | if (!--chars) | |
3675 | break; | |
3676 | uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len, | |
3677 | uniflags); | |
3678 | uscan += len; | |
3679 | foldlen -= len; | |
3680 | } | |
3681 | } | |
a3621e74 | 3682 | } |
2e64971a | 3683 | else { |
f2ed9b32 | 3684 | if (utf8_target) |
2e64971a DM |
3685 | while (chars--) |
3686 | uc += UTF8SKIP(uc); | |
3687 | else | |
3688 | uc += chars; | |
3689 | } | |
2e64971a | 3690 | } |
166ba7cd | 3691 | |
6603fe3e DM |
3692 | scan = ST.me + ((ST.jump && ST.jump[ST.nextword]) |
3693 | ? ST.jump[ST.nextword] | |
3694 | : NEXT_OFF(ST.me)); | |
166ba7cd | 3695 | |
2e64971a DM |
3696 | DEBUG_EXECUTE_r({ |
3697 | PerlIO_printf( Perl_debug_log, | |
3698 | "%*s %sTRIE matched word #%d, continuing%s\n", | |
3699 | REPORT_CODE_OFF+depth*2, "", | |
3700 | PL_colors[4], | |
3701 | ST.nextword, | |
3702 | PL_colors[5] | |
3703 | ); | |
3704 | }); | |
3705 | ||
3706 | if (ST.accepted > 1 || has_cutgroup) { | |
a059a757 | 3707 | PUSH_STATE_GOTO(TRIE_next, scan, (char*)uc); |
118e2215 | 3708 | assert(0); /* NOTREACHED */ |
166ba7cd | 3709 | } |
2e64971a DM |
3710 | /* only one choice left - just continue */ |
3711 | DEBUG_EXECUTE_r({ | |
3712 | AV *const trie_words | |
3713 | = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]); | |
3714 | SV ** const tmp = av_fetch( trie_words, | |
3715 | ST.nextword-1, 0 ); | |
3716 | SV *sv= tmp ? sv_newmortal() : NULL; | |
3717 | ||
3718 | PerlIO_printf( Perl_debug_log, | |
3719 | "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n", | |
3720 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], | |
3721 | ST.nextword, | |
3722 | tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0, | |
3723 | PL_colors[0], PL_colors[1], | |
c89df6cf | 3724 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII |
2e64971a DM |
3725 | ) |
3726 | : "not compiled under -Dr", | |
3727 | PL_colors[5] ); | |
3728 | }); | |
3729 | ||
a059a757 | 3730 | locinput = (char*)uc; |
2e64971a | 3731 | continue; /* execute rest of RE */ |
118e2215 | 3732 | assert(0); /* NOTREACHED */ |
a059a757 | 3733 | } |
166ba7cd DM |
3734 | #undef ST |
3735 | ||
3c0563b9 | 3736 | case EXACT: { /* /abc/ */ |
95b24440 | 3737 | char *s = STRING(scan); |
24d3c4a9 | 3738 | ln = STR_LEN(scan); |
f2ed9b32 | 3739 | if (utf8_target != UTF_PATTERN) { |
bc517b45 | 3740 | /* The target and the pattern have differing utf8ness. */ |
1aa99e6b | 3741 | char *l = locinput; |
24d3c4a9 | 3742 | const char * const e = s + ln; |
a72c7584 | 3743 | |
f2ed9b32 | 3744 | if (utf8_target) { |
e6a3850e KW |
3745 | /* The target is utf8, the pattern is not utf8. |
3746 | * Above-Latin1 code points can't match the pattern; | |
3747 | * invariants match exactly, and the other Latin1 ones need | |
3748 | * to be downgraded to a single byte in order to do the | |
3749 | * comparison. (If we could be confident that the target | |
3750 | * is not malformed, this could be refactored to have fewer | |
3751 | * tests by just assuming that if the first bytes match, it | |
3752 | * is an invariant, but there are tests in the test suite | |
3753 | * dealing with (??{...}) which violate this) */ | |
1aa99e6b IH |
3754 | while (s < e) { |
3755 | if (l >= PL_regeol) | |
5ff6fc6d | 3756 | sayNO; |
e6a3850e KW |
3757 | if (UTF8_IS_ABOVE_LATIN1(* (U8*) l)) { |
3758 | sayNO; | |
3759 | } | |
3760 | if (UTF8_IS_INVARIANT(*(U8*)l)) { | |
3761 | if (*l != *s) { | |
3762 | sayNO; | |
3763 | } | |
3764 | l++; | |
3765 | } | |
3766 | else { | |
3767 | if (TWO_BYTE_UTF8_TO_UNI(*l, *(l+1)) != * (U8*) s) { | |
3768 | sayNO; | |
3769 | } | |
3770 | l += 2; | |
3771 | } | |
3772 | s++; | |
1aa99e6b | 3773 | } |
5ff6fc6d JH |
3774 | } |
3775 | else { | |
3776 | /* The target is not utf8, the pattern is utf8. */ | |
1aa99e6b | 3777 | while (s < e) { |
e6a3850e KW |
3778 | if (l >= PL_regeol || UTF8_IS_ABOVE_LATIN1(* (U8*) s)) |
3779 | { | |
3780 | sayNO; | |
3781 | } | |
3782 | if (UTF8_IS_INVARIANT(*(U8*)s)) { | |
3783 | if (*s != *l) { | |
3784 | sayNO; | |
3785 | } | |
3786 | s++; | |
3787 | } | |
3788 | else { | |
3789 | if (TWO_BYTE_UTF8_TO_UNI(*s, *(s+1)) != * (U8*) l) { | |
3790 | sayNO; | |
3791 | } | |
3792 | s += 2; | |
3793 | } | |
3794 | l++; | |
1aa99e6b | 3795 | } |
5ff6fc6d | 3796 | } |
1aa99e6b | 3797 | locinput = l; |
1aa99e6b IH |
3798 | break; |
3799 | } | |
bc517b45 | 3800 | /* The target and the pattern have the same utf8ness. */ |
d6a28714 JH |
3801 | /* Inline the first character, for speed. */ |
3802 | if (UCHARAT(s) != nextchr) | |
3803 | sayNO; | |
24d3c4a9 | 3804 | if (PL_regeol - locinput < ln) |
d6a28714 | 3805 | sayNO; |
24d3c4a9 | 3806 | if (ln > 1 && memNE(s, locinput, ln)) |
d6a28714 | 3807 | sayNO; |
24d3c4a9 | 3808 | locinput += ln; |
d6a28714 | 3809 | break; |
95b24440 | 3810 | } |
3c0563b9 | 3811 | case EXACTFL: { /* /abc/il */ |
a932d541 | 3812 | re_fold_t folder; |
9a5a5549 KW |
3813 | const U8 * fold_array; |
3814 | const char * s; | |
d513472c | 3815 | U32 fold_utf8_flags; |
9a5a5549 | 3816 | |
b8c5462f | 3817 | PL_reg_flags |= RF_tainted; |
9a5a5549 KW |
3818 | folder = foldEQ_locale; |
3819 | fold_array = PL_fold_locale; | |
17580e7a | 3820 | fold_utf8_flags = FOLDEQ_UTF8_LOCALE; |
9a5a5549 KW |
3821 | goto do_exactf; |
3822 | ||
3c0563b9 DM |
3823 | case EXACTFU_SS: /* /\x{df}/iu */ |
3824 | case EXACTFU_TRICKYFOLD: /* /\x{390}/iu */ | |
3825 | case EXACTFU: /* /abc/iu */ | |
9a5a5549 KW |
3826 | folder = foldEQ_latin1; |
3827 | fold_array = PL_fold_latin1; | |
2daa8fee | 3828 | fold_utf8_flags = (UTF_PATTERN) ? FOLDEQ_S1_ALREADY_FOLDED : 0; |
9a5a5549 KW |
3829 | goto do_exactf; |
3830 | ||
3c0563b9 | 3831 | case EXACTFA: /* /abc/iaa */ |
2f7f8cb1 KW |
3832 | folder = foldEQ_latin1; |
3833 | fold_array = PL_fold_latin1; | |
57014d77 | 3834 | fold_utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; |
2f7f8cb1 KW |
3835 | goto do_exactf; |
3836 | ||
3c0563b9 | 3837 | case EXACTF: /* /abc/i */ |
9a5a5549 KW |
3838 | folder = foldEQ; |
3839 | fold_array = PL_fold; | |
62bf7766 | 3840 | fold_utf8_flags = 0; |
9a5a5549 KW |
3841 | |
3842 | do_exactf: | |
3843 | s = STRING(scan); | |
24d3c4a9 | 3844 | ln = STR_LEN(scan); |
d6a28714 | 3845 | |
3c760661 KW |
3846 | if (utf8_target || UTF_PATTERN || state_num == EXACTFU_SS) { |
3847 | /* Either target or the pattern are utf8, or has the issue where | |
3848 | * the fold lengths may differ. */ | |
be8e71aa | 3849 | const char * const l = locinput; |
d07ddd77 | 3850 | char *e = PL_regeol; |
bc517b45 | 3851 | |
d513472c | 3852 | if (! foldEQ_utf8_flags(s, 0, ln, cBOOL(UTF_PATTERN), |
fa5b1667 | 3853 | l, &e, 0, utf8_target, fold_utf8_flags)) |
c3e1d013 KW |
3854 | { |
3855 | sayNO; | |
5486206c | 3856 | } |
d07ddd77 | 3857 | locinput = e; |
d07ddd77 | 3858 | break; |
a0ed51b3 | 3859 | } |
d6a28714 | 3860 | |
0a138b74 | 3861 | /* Neither the target nor the pattern are utf8 */ |
d6a28714 | 3862 | if (UCHARAT(s) != nextchr && |
9a5a5549 KW |
3863 | UCHARAT(s) != fold_array[nextchr]) |
3864 | { | |
a0ed51b3 | 3865 | sayNO; |
9a5a5549 | 3866 | } |
24d3c4a9 | 3867 | if (PL_regeol - locinput < ln) |
b8c5462f | 3868 | sayNO; |
9a5a5549 | 3869 | if (ln > 1 && ! folder(s, locinput, ln)) |
4633a7c4 | 3870 | sayNO; |
24d3c4a9 | 3871 | locinput += ln; |
a0d0e21e | 3872 | break; |
9a5a5549 | 3873 | } |
63ac0dad KW |
3874 | |
3875 | /* XXX Could improve efficiency by separating these all out using a | |
3876 | * macro or in-line function. At that point regcomp.c would no longer | |
3877 | * have to set the FLAGS fields of these */ | |
3c0563b9 DM |
3878 | case BOUNDL: /* /\b/l */ |
3879 | case NBOUNDL: /* /\B/l */ | |
b2680017 YO |
3880 | PL_reg_flags |= RF_tainted; |
3881 | /* FALL THROUGH */ | |
3c0563b9 DM |
3882 | case BOUND: /* /\b/ */ |
3883 | case BOUNDU: /* /\b/u */ | |
3884 | case BOUNDA: /* /\b/a */ | |
3885 | case NBOUND: /* /\B/ */ | |
3886 | case NBOUNDU: /* /\B/u */ | |
3887 | case NBOUNDA: /* /\B/a */ | |
b2680017 | 3888 | /* was last char in word? */ |
f2e96b5d KW |
3889 | if (utf8_target |
3890 | && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET | |
3891 | && FLAGS(scan) != REGEX_ASCII_MORE_RESTRICTED_CHARSET) | |
3892 | { | |
b2680017 YO |
3893 | if (locinput == PL_bostr) |
3894 | ln = '\n'; | |
3895 | else { | |
3896 | const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); | |
3897 | ||
3898 | ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); | |
3899 | } | |
63ac0dad | 3900 | if (FLAGS(scan) != REGEX_LOCALE_CHARSET) { |
b2680017 YO |
3901 | ln = isALNUM_uni(ln); |
3902 | LOAD_UTF8_CHARCLASS_ALNUM(); | |
f2ed9b32 | 3903 | n = swash_fetch(PL_utf8_alnum, (U8*)locinput, utf8_target); |
b2680017 YO |
3904 | } |
3905 | else { | |
3906 | ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); | |
3907 | n = isALNUM_LC_utf8((U8*)locinput); | |
3908 | } | |
3909 | } | |
3910 | else { | |
cfaf538b KW |
3911 | |
3912 | /* Here the string isn't utf8, or is utf8 and only ascii | |
3913 | * characters are to match \w. In the latter case looking at | |
3914 | * the byte just prior to the current one may be just the final | |
3915 | * byte of a multi-byte character. This is ok. There are two | |
3916 | * cases: | |
3917 | * 1) it is a single byte character, and then the test is doing | |
3918 | * just what it's supposed to. | |
3919 | * 2) it is a multi-byte character, in which case the final | |
3920 | * byte is never mistakable for ASCII, and so the test | |
3921 | * will say it is not a word character, which is the | |
3922 | * correct answer. */ | |
b2680017 YO |
3923 | ln = (locinput != PL_bostr) ? |
3924 | UCHARAT(locinput - 1) : '\n'; | |
63ac0dad KW |
3925 | switch (FLAGS(scan)) { |
3926 | case REGEX_UNICODE_CHARSET: | |
3927 | ln = isWORDCHAR_L1(ln); | |
3928 | n = isWORDCHAR_L1(nextchr); | |
3929 | break; | |
3930 | case REGEX_LOCALE_CHARSET: | |
3931 | ln = isALNUM_LC(ln); | |
3932 | n = isALNUM_LC(nextchr); | |
3933 | break; | |
3934 | case REGEX_DEPENDS_CHARSET: | |
3935 | ln = isALNUM(ln); | |
3936 | n = isALNUM(nextchr); | |
3937 | break; | |
cfaf538b | 3938 | case REGEX_ASCII_RESTRICTED_CHARSET: |
c973bd4f | 3939 | case REGEX_ASCII_MORE_RESTRICTED_CHARSET: |
cfaf538b KW |
3940 | ln = isWORDCHAR_A(ln); |
3941 | n = isWORDCHAR_A(nextchr); | |
3942 | break; | |
63ac0dad KW |
3943 | default: |
3944 | Perl_croak(aTHX_ "panic: Unexpected FLAGS %u in op %u", FLAGS(scan), OP(scan)); | |
3945 | break; | |
b2680017 YO |
3946 | } |
3947 | } | |
63ac0dad KW |
3948 | /* Note requires that all BOUNDs be lower than all NBOUNDs in |
3949 | * regcomp.sym */ | |
3950 | if (((!ln) == (!n)) == (OP(scan) < NBOUND)) | |
b2680017 YO |
3951 | sayNO; |
3952 | break; | |
3c0563b9 DM |
3953 | |
3954 | case ANYOFV: /* /[abx{df}]/i */ | |
3955 | case ANYOF: /* /[abc]/ */ | |
f56b6394 | 3956 | if (utf8_target || state_num == ANYOFV) { |
9e55ce06 | 3957 | STRLEN inclasslen = PL_regeol - locinput; |
20ed0b26 KW |
3958 | if (locinput >= PL_regeol) |
3959 | sayNO; | |
9e55ce06 | 3960 | |
f2ed9b32 | 3961 | if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target)) |
09b08e9b | 3962 | sayNO; |
b32d7d3e | 3963 | locinput += inclasslen; |
e0f9d4a8 | 3964 | break; |
ffc61ed2 JH |
3965 | } |
3966 | else { | |
ffc61ed2 JH |
3967 | if (!nextchr && locinput >= PL_regeol) |
3968 | sayNO; | |
20ed0b26 | 3969 | if (!REGINCLASS(rex, scan, (U8*)locinput)) |
09b08e9b | 3970 | sayNO; |
3640db6b | 3971 | locinput++; |
e0f9d4a8 JH |
3972 | break; |
3973 | } | |
b8c5462f | 3974 | break; |
3c0563b9 DM |
3975 | |
3976 | /* Special char classes: \d, \w etc. | |
3977 | * The defines start on line 166 or so */ | |
ee9a90b8 KW |
3978 | CCC_TRY_U(ALNUM, NALNUM, isWORDCHAR, |
3979 | ALNUML, NALNUML, isALNUM_LC, isALNUM_LC_utf8, | |
3980 | ALNUMU, NALNUMU, isWORDCHAR_L1, | |
cfaf538b | 3981 | ALNUMA, NALNUMA, isWORDCHAR_A, |
779d7b58 | 3982 | alnum, "a"); |
ee9a90b8 KW |
3983 | |
3984 | CCC_TRY_U(SPACE, NSPACE, isSPACE, | |
3985 | SPACEL, NSPACEL, isSPACE_LC, isSPACE_LC_utf8, | |
3986 | SPACEU, NSPACEU, isSPACE_L1, | |
cfaf538b | 3987 | SPACEA, NSPACEA, isSPACE_A, |
779d7b58 | 3988 | space, " "); |
ee9a90b8 KW |
3989 | |
3990 | CCC_TRY(DIGIT, NDIGIT, isDIGIT, | |
3991 | DIGITL, NDIGITL, isDIGIT_LC, isDIGIT_LC_utf8, | |
cfaf538b | 3992 | DIGITA, NDIGITA, isDIGIT_A, |
779d7b58 | 3993 | digit, "0"); |
20d0b1e9 | 3994 | |
3c0563b9 | 3995 | case POSIXA: /* /[[:ascii:]]/ etc */ |
0658cdde KW |
3996 | if (locinput >= PL_regeol || ! _generic_isCC_A(nextchr, FLAGS(scan))) { |
3997 | sayNO; | |
3998 | } | |
3999 | /* Matched a utf8-invariant, so don't have to worry about utf8 */ | |
3640db6b | 4000 | locinput++; |
0658cdde | 4001 | break; |
3c0563b9 DM |
4002 | |
4003 | case NPOSIXA: /* /[^[:ascii:]]/ etc */ | |
0658cdde KW |
4004 | if (locinput >= PL_regeol || _generic_isCC_A(nextchr, FLAGS(scan))) { |
4005 | sayNO; | |
4006 | } | |
28b98f76 | 4007 | goto increment_locinput; |
0658cdde | 4008 | |
37e2e78e KW |
4009 | case CLUMP: /* Match \X: logical Unicode character. This is defined as |
4010 | a Unicode extended Grapheme Cluster */ | |
4011 | /* From http://www.unicode.org/reports/tr29 (5.2 version). An | |
4012 | extended Grapheme Cluster is: | |
4013 | ||
4014 | CR LF | |
4015 | | Prepend* Begin Extend* | |
4016 | | . | |
4017 | ||
1e958ea9 KW |
4018 | Begin is: ( Special_Begin | ! Control ) |
4019 | Special_Begin is: ( Regional-Indicator+ | Hangul-syllable ) | |
4020 | Extend is: ( Grapheme_Extend | Spacing_Mark ) | |
4021 | Control is: [ GCB_Control CR LF ] | |
4022 | Hangul-syllable is: ( T+ | ( L* ( L | ( LVT | ( V | LV ) V* ) T* ) )) | |
37e2e78e | 4023 | |
27d4fc33 KW |
4024 | If we create a 'Regular_Begin' = Begin - Special_Begin, then |
4025 | we can rewrite | |
4026 | ||
4027 | Begin is ( Regular_Begin + Special Begin ) | |
4028 | ||
4029 | It turns out that 98.4% of all Unicode code points match | |
4030 | Regular_Begin. Doing it this way eliminates a table match in | |
c101f46d | 4031 | the previous implementation for almost all Unicode code points. |
27d4fc33 | 4032 | |
37e2e78e KW |
4033 | There is a subtlety with Prepend* which showed up in testing. |
4034 | Note that the Begin, and only the Begin is required in: | |
4035 | | Prepend* Begin Extend* | |
cc3b396d KW |
4036 | Also, Begin contains '! Control'. A Prepend must be a |
4037 | '! Control', which means it must also be a Begin. What it | |
4038 | comes down to is that if we match Prepend* and then find no | |
4039 | suitable Begin afterwards, that if we backtrack the last | |
4040 | Prepend, that one will be a suitable Begin. | |
37e2e78e KW |
4041 | */ |
4042 | ||
b7c83a7e | 4043 | if (locinput >= PL_regeol) |
a0ed51b3 | 4044 | sayNO; |
f2ed9b32 | 4045 | if (! utf8_target) { |
37e2e78e KW |
4046 | |
4047 | /* Match either CR LF or '.', as all the other possibilities | |
4048 | * require utf8 */ | |
4049 | locinput++; /* Match the . or CR */ | |
cc3b396d KW |
4050 | if (nextchr == '\r' /* And if it was CR, and the next is LF, |
4051 | match the LF */ | |
37e2e78e KW |
4052 | && locinput < PL_regeol |
4053 | && UCHARAT(locinput) == '\n') locinput++; | |
4054 | } | |
4055 | else { | |
4056 | ||
4057 | /* Utf8: See if is ( CR LF ); already know that locinput < | |
4058 | * PL_regeol, so locinput+1 is in bounds */ | |
4059 | if (nextchr == '\r' && UCHARAT(locinput + 1) == '\n') { | |
4060 | locinput += 2; | |
4061 | } | |
4062 | else { | |
45fdf108 KW |
4063 | STRLEN len; |
4064 | ||
37e2e78e KW |
4065 | /* In case have to backtrack to beginning, then match '.' */ |
4066 | char *starting = locinput; | |
4067 | ||
4068 | /* In case have to backtrack the last prepend */ | |
4069 | char *previous_prepend = 0; | |
4070 | ||
4071 | LOAD_UTF8_CHARCLASS_GCB(); | |
4072 | ||
45fdf108 KW |
4073 | /* Match (prepend)* */ |
4074 | while (locinput < PL_regeol | |
4075 | && (len = is_GCB_Prepend_utf8(locinput))) | |
4076 | { | |
4077 | previous_prepend = locinput; | |
4078 | locinput += len; | |
a1853d78 | 4079 | } |
37e2e78e KW |
4080 | |
4081 | /* As noted above, if we matched a prepend character, but | |
4082 | * the next thing won't match, back off the last prepend we | |
4083 | * matched, as it is guaranteed to match the begin */ | |
4084 | if (previous_prepend | |
4085 | && (locinput >= PL_regeol | |
c101f46d KW |
4086 | || (! swash_fetch(PL_utf8_X_regular_begin, |
4087 | (U8*)locinput, utf8_target) | |
45fdf108 | 4088 | && ! is_GCB_SPECIAL_BEGIN_utf8(locinput))) |
c101f46d | 4089 | ) |
37e2e78e KW |
4090 | { |
4091 | locinput = previous_prepend; | |
4092 | } | |
4093 | ||
4094 | /* Note that here we know PL_regeol > locinput, as we | |
4095 | * tested that upon input to this switch case, and if we | |
4096 | * moved locinput forward, we tested the result just above | |
4097 | * and it either passed, or we backed off so that it will | |
4098 | * now pass */ | |
11dfcd49 KW |
4099 | if (swash_fetch(PL_utf8_X_regular_begin, |
4100 | (U8*)locinput, utf8_target)) { | |
27d4fc33 KW |
4101 | locinput += UTF8SKIP(locinput); |
4102 | } | |
45fdf108 | 4103 | else if (! is_GCB_SPECIAL_BEGIN_utf8(locinput)) { |
37e2e78e KW |
4104 | |
4105 | /* Here did not match the required 'Begin' in the | |
4106 | * second term. So just match the very first | |
4107 | * character, the '.' of the final term of the regex */ | |
4108 | locinput = starting + UTF8SKIP(starting); | |
27d4fc33 | 4109 | goto exit_utf8; |
37e2e78e KW |
4110 | } else { |
4111 | ||
11dfcd49 KW |
4112 | /* Here is a special begin. It can be composed of |
4113 | * several individual characters. One possibility is | |
4114 | * RI+ */ | |
45fdf108 KW |
4115 | if ((len = is_GCB_RI_utf8(locinput))) { |
4116 | locinput += len; | |
11dfcd49 | 4117 | while (locinput < PL_regeol |
45fdf108 | 4118 | && (len = is_GCB_RI_utf8(locinput))) |
11dfcd49 | 4119 | { |
45fdf108 | 4120 | locinput += len; |
11dfcd49 | 4121 | } |
45fdf108 KW |
4122 | } else if ((len = is_GCB_T_utf8(locinput))) { |
4123 | /* Another possibility is T+ */ | |
4124 | locinput += len; | |
11dfcd49 | 4125 | while (locinput < PL_regeol |
45fdf108 | 4126 | && (len = is_GCB_T_utf8(locinput))) |
11dfcd49 | 4127 | { |
45fdf108 | 4128 | locinput += len; |
11dfcd49 KW |
4129 | } |
4130 | } else { | |
4131 | ||
4132 | /* Here, neither RI+ nor T+; must be some other | |
4133 | * Hangul. That means it is one of the others: L, | |
4134 | * LV, LVT or V, and matches: | |
4135 | * L* (L | LVT T* | V * V* T* | LV V* T*) */ | |
4136 | ||
4137 | /* Match L* */ | |
4138 | while (locinput < PL_regeol | |
45fdf108 | 4139 | && (len = is_GCB_L_utf8(locinput))) |
11dfcd49 | 4140 | { |
45fdf108 | 4141 | locinput += len; |
11dfcd49 | 4142 | } |
37e2e78e | 4143 | |
11dfcd49 KW |
4144 | /* Here, have exhausted L*. If the next character |
4145 | * is not an LV, LVT nor V, it means we had to have | |
4146 | * at least one L, so matches L+ in the original | |
4147 | * equation, we have a complete hangul syllable. | |
4148 | * Are done. */ | |
4149 | ||
4150 | if (locinput < PL_regeol | |
45fdf108 | 4151 | && is_GCB_LV_LVT_V_utf8(locinput)) |
11dfcd49 KW |
4152 | { |
4153 | ||
4154 | /* Otherwise keep going. Must be LV, LVT or V. | |
4155 | * See if LVT */ | |
4156 | if (is_utf8_X_LVT((U8*)locinput)) { | |
4157 | locinput += UTF8SKIP(locinput); | |
4158 | } else { | |
4159 | ||
4160 | /* Must be V or LV. Take it, then match | |
4161 | * V* */ | |
4162 | locinput += UTF8SKIP(locinput); | |
4163 | while (locinput < PL_regeol | |
45fdf108 | 4164 | && (len = is_GCB_V_utf8(locinput))) |
11dfcd49 | 4165 | { |
45fdf108 | 4166 | locinput += len; |
11dfcd49 KW |
4167 | } |
4168 | } | |
37e2e78e | 4169 | |
11dfcd49 | 4170 | /* And any of LV, LVT, or V can be followed |
45fdf108 | 4171 | * by T* */ |
11dfcd49 | 4172 | while (locinput < PL_regeol |
45fdf108 | 4173 | && (len = is_GCB_T_utf8(locinput))) |
11dfcd49 | 4174 | { |
45fdf108 | 4175 | locinput += len; |
11dfcd49 KW |
4176 | } |
4177 | } | |
cd94d768 | 4178 | } |
11dfcd49 | 4179 | } |
37e2e78e | 4180 | |
11dfcd49 KW |
4181 | /* Match any extender */ |
4182 | while (locinput < PL_regeol | |
4183 | && swash_fetch(PL_utf8_X_extend, | |
4184 | (U8*)locinput, utf8_target)) | |
4185 | { | |
4186 | locinput += UTF8SKIP(locinput); | |
4187 | } | |
37e2e78e | 4188 | } |
27d4fc33 | 4189 | exit_utf8: |
37e2e78e KW |
4190 | if (locinput > PL_regeol) sayNO; |
4191 | } | |
a0ed51b3 | 4192 | break; |
81714fb9 | 4193 | |
3c0563b9 | 4194 | case NREFFL: /* /\g{name}/il */ |
d7ef4b73 KW |
4195 | { /* The capture buffer cases. The ones beginning with N for the |
4196 | named buffers just convert to the equivalent numbered and | |
4197 | pretend they were called as the corresponding numbered buffer | |
4198 | op. */ | |
26ecd678 TC |
4199 | /* don't initialize these in the declaration, it makes C++ |
4200 | unhappy */ | |
81714fb9 | 4201 | char *s; |
ff1157ca | 4202 | char type; |
8368298a TC |
4203 | re_fold_t folder; |
4204 | const U8 *fold_array; | |
26ecd678 | 4205 | UV utf8_fold_flags; |
8368298a | 4206 | |
81714fb9 | 4207 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4208 | folder = foldEQ_locale; |
4209 | fold_array = PL_fold_locale; | |
4210 | type = REFFL; | |
17580e7a | 4211 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4212 | goto do_nref; |
4213 | ||
3c0563b9 | 4214 | case NREFFA: /* /\g{name}/iaa */ |
2f7f8cb1 KW |
4215 | folder = foldEQ_latin1; |
4216 | fold_array = PL_fold_latin1; | |
4217 | type = REFFA; | |
4218 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4219 | goto do_nref; | |
4220 | ||
3c0563b9 | 4221 | case NREFFU: /* /\g{name}/iu */ |
d7ef4b73 KW |
4222 | folder = foldEQ_latin1; |
4223 | fold_array = PL_fold_latin1; | |
4224 | type = REFFU; | |
d513472c | 4225 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4226 | goto do_nref; |
4227 | ||
3c0563b9 | 4228 | case NREFF: /* /\g{name}/i */ |
d7ef4b73 KW |
4229 | folder = foldEQ; |
4230 | fold_array = PL_fold; | |
4231 | type = REFF; | |
d513472c | 4232 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4233 | goto do_nref; |
4234 | ||
3c0563b9 | 4235 | case NREF: /* /\g{name}/ */ |
d7ef4b73 | 4236 | type = REF; |
83d7b90b KW |
4237 | folder = NULL; |
4238 | fold_array = NULL; | |
d513472c | 4239 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4240 | do_nref: |
4241 | ||
4242 | /* For the named back references, find the corresponding buffer | |
4243 | * number */ | |
0a4db386 YO |
4244 | n = reg_check_named_buff_matched(rex,scan); |
4245 | ||
d7ef4b73 | 4246 | if ( ! n ) { |
81714fb9 | 4247 | sayNO; |
d7ef4b73 KW |
4248 | } |
4249 | goto do_nref_ref_common; | |
4250 | ||
3c0563b9 | 4251 | case REFFL: /* /\1/il */ |
3280af22 | 4252 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4253 | folder = foldEQ_locale; |
4254 | fold_array = PL_fold_locale; | |
17580e7a | 4255 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4256 | goto do_ref; |
4257 | ||
3c0563b9 | 4258 | case REFFA: /* /\1/iaa */ |
2f7f8cb1 KW |
4259 | folder = foldEQ_latin1; |
4260 | fold_array = PL_fold_latin1; | |
4261 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4262 | goto do_ref; | |
4263 | ||
3c0563b9 | 4264 | case REFFU: /* /\1/iu */ |
d7ef4b73 KW |
4265 | folder = foldEQ_latin1; |
4266 | fold_array = PL_fold_latin1; | |
d513472c | 4267 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4268 | goto do_ref; |
4269 | ||
3c0563b9 | 4270 | case REFF: /* /\1/i */ |
d7ef4b73 KW |
4271 | folder = foldEQ; |
4272 | fold_array = PL_fold; | |
d513472c | 4273 | utf8_fold_flags = 0; |
83d7b90b | 4274 | goto do_ref; |
d7ef4b73 | 4275 | |
3c0563b9 | 4276 | case REF: /* /\1/ */ |
83d7b90b KW |
4277 | folder = NULL; |
4278 | fold_array = NULL; | |
d513472c | 4279 | utf8_fold_flags = 0; |
83d7b90b | 4280 | |
d7ef4b73 | 4281 | do_ref: |
81714fb9 | 4282 | type = OP(scan); |
d7ef4b73 KW |
4283 | n = ARG(scan); /* which paren pair */ |
4284 | ||
4285 | do_nref_ref_common: | |
b93070ed | 4286 | ln = rex->offs[n].start; |
2c2d71f5 | 4287 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
b93070ed | 4288 | if (rex->lastparen < n || ln == -1) |
af3f8c16 | 4289 | sayNO; /* Do not match unless seen CLOSEn. */ |
b93070ed | 4290 | if (ln == rex->offs[n].end) |
a0d0e21e | 4291 | break; |
a0ed51b3 | 4292 | |
24d3c4a9 | 4293 | s = PL_bostr + ln; |
d7ef4b73 | 4294 | if (type != REF /* REF can do byte comparison */ |
2f65c56d | 4295 | && (utf8_target || type == REFFU)) |
d7ef4b73 KW |
4296 | { /* XXX handle REFFL better */ |
4297 | char * limit = PL_regeol; | |
4298 | ||
4299 | /* This call case insensitively compares the entire buffer | |
4300 | * at s, with the current input starting at locinput, but | |
4301 | * not going off the end given by PL_regeol, and returns in | |
4302 | * limit upon success, how much of the current input was | |
4303 | * matched */ | |
b93070ed | 4304 | if (! foldEQ_utf8_flags(s, NULL, rex->offs[n].end - ln, utf8_target, |
d513472c | 4305 | locinput, &limit, 0, utf8_target, utf8_fold_flags)) |
d7ef4b73 KW |
4306 | { |
4307 | sayNO; | |
a0ed51b3 | 4308 | } |
d7ef4b73 | 4309 | locinput = limit; |
a0ed51b3 LW |
4310 | break; |
4311 | } | |
4312 | ||
d7ef4b73 | 4313 | /* Not utf8: Inline the first character, for speed. */ |
76e3520e | 4314 | if (UCHARAT(s) != nextchr && |
81714fb9 | 4315 | (type == REF || |
d7ef4b73 | 4316 | UCHARAT(s) != fold_array[nextchr])) |
4633a7c4 | 4317 | sayNO; |
b93070ed | 4318 | ln = rex->offs[n].end - ln; |
24d3c4a9 | 4319 | if (locinput + ln > PL_regeol) |
4633a7c4 | 4320 | sayNO; |
81714fb9 | 4321 | if (ln > 1 && (type == REF |
24d3c4a9 | 4322 | ? memNE(s, locinput, ln) |
d7ef4b73 | 4323 | : ! folder(s, locinput, ln))) |
4633a7c4 | 4324 | sayNO; |
24d3c4a9 | 4325 | locinput += ln; |
a0d0e21e | 4326 | break; |
81714fb9 | 4327 | } |
3c0563b9 DM |
4328 | |
4329 | case NOTHING: /* null op; e.g. the 'nothing' following | |
4330 | * the '*' in m{(a+|b)*}' */ | |
4331 | break; | |
4332 | case TAIL: /* placeholder while compiling (A|B|C) */ | |
a0d0e21e | 4333 | break; |
3c0563b9 DM |
4334 | |
4335 | case BACK: /* ??? doesn't appear to be used ??? */ | |
a0d0e21e | 4336 | break; |
40a82448 DM |
4337 | |
4338 | #undef ST | |
4339 | #define ST st->u.eval | |
c277df42 | 4340 | { |
c277df42 | 4341 | SV *ret; |
d2f13c59 | 4342 | REGEXP *re_sv; |
6bda09f9 | 4343 | regexp *re; |
f8fc2ecf | 4344 | regexp_internal *rei; |
1a147d38 YO |
4345 | regnode *startpoint; |
4346 | ||
3c0563b9 | 4347 | case GOSTART: /* (?R) */ |
e7707071 YO |
4348 | case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */ |
4349 | if (cur_eval && cur_eval->locinput==locinput) { | |
24b23f37 | 4350 | if (cur_eval->u.eval.close_paren == (U32)ARG(scan)) |
1a147d38 | 4351 | Perl_croak(aTHX_ "Infinite recursion in regex"); |
4b196cd4 | 4352 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 YO |
4353 | Perl_croak(aTHX_ |
4354 | "Pattern subroutine nesting without pos change" | |
4355 | " exceeded limit in regex"); | |
6bda09f9 YO |
4356 | } else { |
4357 | nochange_depth = 0; | |
1a147d38 | 4358 | } |
288b8c02 | 4359 | re_sv = rex_sv; |
6bda09f9 | 4360 | re = rex; |
f8fc2ecf | 4361 | rei = rexi; |
1a147d38 | 4362 | if (OP(scan)==GOSUB) { |
6bda09f9 YO |
4363 | startpoint = scan + ARG2L(scan); |
4364 | ST.close_paren = ARG(scan); | |
4365 | } else { | |
f8fc2ecf | 4366 | startpoint = rei->program+1; |
6bda09f9 YO |
4367 | ST.close_paren = 0; |
4368 | } | |
4369 | goto eval_recurse_doit; | |
118e2215 | 4370 | assert(0); /* NOTREACHED */ |
3c0563b9 | 4371 | |
6bda09f9 YO |
4372 | case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ |
4373 | if (cur_eval && cur_eval->locinput==locinput) { | |
4b196cd4 | 4374 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 | 4375 | Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex"); |
6bda09f9 YO |
4376 | } else { |
4377 | nochange_depth = 0; | |
4378 | } | |
8e5e9ebe | 4379 | { |
4aabdb9b | 4380 | /* execute the code in the {...} */ |
81ed78b2 | 4381 | |
4aabdb9b | 4382 | dSP; |
81ed78b2 | 4383 | SV ** before; |
1f4d1a1e | 4384 | OP * const oop = PL_op; |
4aabdb9b | 4385 | COP * const ocurcop = PL_curcop; |
81ed78b2 | 4386 | OP *nop; |
d80618d2 | 4387 | char *saved_regeol = PL_regeol; |
91332126 | 4388 | struct re_save_state saved_state; |
81ed78b2 | 4389 | CV *newcv; |
91332126 | 4390 | |
74088413 DM |
4391 | /* save *all* paren positions */ |
4392 | regcppush(rex, 0); | |
4393 | REGCP_SET(runops_cp); | |
4394 | ||
6562f1c4 | 4395 | /* To not corrupt the existing regex state while executing the |
b7f4cd04 FR |
4396 | * eval we would normally put it on the save stack, like with |
4397 | * save_re_context. However, re-evals have a weird scoping so we | |
4398 | * can't just add ENTER/LEAVE here. With that, things like | |
4399 | * | |
4400 | * (?{$a=2})(a(?{local$a=$a+1}))*aak*c(?{$b=$a}) | |
4401 | * | |
4402 | * would break, as they expect the localisation to be unwound | |
4403 | * only when the re-engine backtracks through the bit that | |
4404 | * localised it. | |
4405 | * | |
4406 | * What we do instead is just saving the state in a local c | |
4407 | * variable. | |
4408 | */ | |
91332126 | 4409 | Copy(&PL_reg_state, &saved_state, 1, struct re_save_state); |
81ed78b2 | 4410 | |
d24ca0c5 | 4411 | PL_reg_state.re_reparsing = FALSE; |
91332126 | 4412 | |
81ed78b2 DM |
4413 | if (!caller_cv) |
4414 | caller_cv = find_runcv(NULL); | |
4415 | ||
4aabdb9b | 4416 | n = ARG(scan); |
81ed78b2 | 4417 | |
b30fcab9 | 4418 | if (rexi->data->what[n] == 'r') { /* code from an external qr */ |
81ed78b2 | 4419 | newcv = ((struct regexp *)SvANY( |
b30fcab9 DM |
4420 | (REGEXP*)(rexi->data->data[n]) |
4421 | ))->qr_anoncv | |
81ed78b2 DM |
4422 | ; |
4423 | nop = (OP*)rexi->data->data[n+1]; | |
b30fcab9 DM |
4424 | } |
4425 | else if (rexi->data->what[n] == 'l') { /* literal code */ | |
81ed78b2 DM |
4426 | newcv = caller_cv; |
4427 | nop = (OP*)rexi->data->data[n]; | |
4428 | assert(CvDEPTH(newcv)); | |
68e2671b DM |
4429 | } |
4430 | else { | |
d24ca0c5 DM |
4431 | /* literal with own CV */ |
4432 | assert(rexi->data->what[n] == 'L'); | |
81ed78b2 DM |
4433 | newcv = rex->qr_anoncv; |
4434 | nop = (OP*)rexi->data->data[n]; | |
68e2671b | 4435 | } |
81ed78b2 | 4436 | |
0e458318 DM |
4437 | /* normally if we're about to execute code from the same |
4438 | * CV that we used previously, we just use the existing | |
4439 | * CX stack entry. However, its possible that in the | |
4440 | * meantime we may have backtracked, popped from the save | |
4441 | * stack, and undone the SAVECOMPPAD(s) associated with | |
4442 | * PUSH_MULTICALL; in which case PL_comppad no longer | |
4443 | * points to newcv's pad. */ | |
4444 | if (newcv != last_pushed_cv || PL_comppad != last_pad) | |
4445 | { | |
4446 | I32 depth = (newcv == caller_cv) ? 0 : 1; | |
4447 | if (last_pushed_cv) { | |
4448 | CHANGE_MULTICALL_WITHDEPTH(newcv, depth); | |
4449 | } | |
4450 | else { | |
4451 | PUSH_MULTICALL_WITHDEPTH(newcv, depth); | |
4452 | } | |
4453 | last_pushed_cv = newcv; | |
4454 | } | |
4455 | last_pad = PL_comppad; | |
4456 | ||
2e2e3f36 DM |
4457 | /* the initial nextstate you would normally execute |
4458 | * at the start of an eval (which would cause error | |
4459 | * messages to come from the eval), may be optimised | |
4460 | * away from the execution path in the regex code blocks; | |
4461 | * so manually set PL_curcop to it initially */ | |
4462 | { | |
81ed78b2 | 4463 | OP *o = cUNOPx(nop)->op_first; |
2e2e3f36 DM |
4464 | assert(o->op_type == OP_NULL); |
4465 | if (o->op_targ == OP_SCOPE) { | |
4466 | o = cUNOPo->op_first; | |
4467 | } | |
4468 | else { | |
4469 | assert(o->op_targ == OP_LEAVE); | |
4470 | o = cUNOPo->op_first; | |
4471 | assert(o->op_type == OP_ENTER); | |
4472 | o = o->op_sibling; | |
4473 | } | |
4474 | ||
4475 | if (o->op_type != OP_STUB) { | |
4476 | assert( o->op_type == OP_NEXTSTATE | |
4477 | || o->op_type == OP_DBSTATE | |
4478 | || (o->op_type == OP_NULL | |
4479 | && ( o->op_targ == OP_NEXTSTATE | |
4480 | || o->op_targ == OP_DBSTATE | |
4481 | ) | |
4482 | ) | |
4483 | ); | |
4484 | PL_curcop = (COP*)o; | |
4485 | } | |
4486 | } | |
81ed78b2 | 4487 | nop = nop->op_next; |
2e2e3f36 | 4488 | |
24b23f37 | 4489 | DEBUG_STATE_r( PerlIO_printf(Perl_debug_log, |
81ed78b2 DM |
4490 | " re EVAL PL_op=0x%"UVxf"\n", PTR2UV(nop)) ); |
4491 | ||
b93070ed | 4492 | rex->offs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr; |
4aabdb9b | 4493 | |
2bf803e2 YO |
4494 | if (sv_yes_mark) { |
4495 | SV *sv_mrk = get_sv("REGMARK", 1); | |
4496 | sv_setsv(sv_mrk, sv_yes_mark); | |
4497 | } | |
4498 | ||
81ed78b2 DM |
4499 | /* we don't use MULTICALL here as we want to call the |
4500 | * first op of the block of interest, rather than the | |
4501 | * first op of the sub */ | |
4502 | before = SP; | |
4503 | PL_op = nop; | |
8e5e9ebe RGS |
4504 | CALLRUNOPS(aTHX); /* Scalar context. */ |
4505 | SPAGAIN; | |
4506 | if (SP == before) | |
075aa684 | 4507 | ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */ |
8e5e9ebe RGS |
4508 | else { |
4509 | ret = POPs; | |
4510 | PUTBACK; | |
4511 | } | |
4aabdb9b | 4512 | |
e4bfbed3 DM |
4513 | /* before restoring everything, evaluate the returned |
4514 | * value, so that 'uninit' warnings don't use the wrong | |
497d0a96 DM |
4515 | * PL_op or pad. Also need to process any magic vars |
4516 | * (e.g. $1) *before* parentheses are restored */ | |
e4bfbed3 DM |
4517 | |
4518 | PL_op = NULL; | |
4519 | ||
5e98dac2 | 4520 | re_sv = NULL; |
e4bfbed3 DM |
4521 | if (logical == 0) /* (?{})/ */ |
4522 | sv_setsv(save_scalar(PL_replgv), ret); /* $^R */ | |
4523 | else if (logical == 1) { /* /(?(?{...})X|Y)/ */ | |
4524 | sw = cBOOL(SvTRUE(ret)); | |
4525 | logical = 0; | |
4526 | } | |
4527 | else { /* /(??{}) */ | |
497d0a96 DM |
4528 | /* if its overloaded, let the regex compiler handle |
4529 | * it; otherwise extract regex, or stringify */ | |
4530 | if (!SvAMAGIC(ret)) { | |
4531 | SV *sv = ret; | |
4532 | if (SvROK(sv)) | |
4533 | sv = SvRV(sv); | |
4534 | if (SvTYPE(sv) == SVt_REGEXP) | |
4535 | re_sv = (REGEXP*) sv; | |
4536 | else if (SvSMAGICAL(sv)) { | |
4537 | MAGIC *mg = mg_find(sv, PERL_MAGIC_qr); | |
4538 | if (mg) | |
4539 | re_sv = (REGEXP *) mg->mg_obj; | |
4540 | } | |
e4bfbed3 | 4541 | |
497d0a96 DM |
4542 | /* force any magic, undef warnings here */ |
4543 | if (!re_sv) { | |
4544 | ret = sv_mortalcopy(ret); | |
4545 | (void) SvPV_force_nolen(ret); | |
4546 | } | |
e4bfbed3 DM |
4547 | } |
4548 | ||
4549 | } | |
4550 | ||
91332126 FR |
4551 | Copy(&saved_state, &PL_reg_state, 1, struct re_save_state); |
4552 | ||
81ed78b2 DM |
4553 | /* *** Note that at this point we don't restore |
4554 | * PL_comppad, (or pop the CxSUB) on the assumption it may | |
4555 | * be used again soon. This is safe as long as nothing | |
4556 | * in the regexp code uses the pad ! */ | |
4aabdb9b | 4557 | PL_op = oop; |
4aabdb9b | 4558 | PL_curcop = ocurcop; |
d80618d2 | 4559 | PL_regeol = saved_regeol; |
e4bfbed3 DM |
4560 | S_regcp_restore(aTHX_ rex, runops_cp); |
4561 | ||
4562 | if (logical != 2) | |
4aabdb9b | 4563 | break; |
8e5e9ebe | 4564 | } |
e4bfbed3 DM |
4565 | |
4566 | /* only /(??{})/ from now on */ | |
24d3c4a9 | 4567 | logical = 0; |
4aabdb9b | 4568 | { |
4f639d21 DM |
4569 | /* extract RE object from returned value; compiling if |
4570 | * necessary */ | |
5c35adbb | 4571 | |
575c37f6 DM |
4572 | if (re_sv) { |
4573 | re_sv = reg_temp_copy(NULL, re_sv); | |
288b8c02 | 4574 | } |
0f5d15d6 | 4575 | else { |
c737faaf | 4576 | U32 pm_flags = 0; |
a3b680e6 | 4577 | const I32 osize = PL_regsize; |
0f5d15d6 | 4578 | |
9753d940 DM |
4579 | if (SvUTF8(ret) && IN_BYTES) { |
4580 | /* In use 'bytes': make a copy of the octet | |
4581 | * sequence, but without the flag on */ | |
b9ad30b4 NC |
4582 | STRLEN len; |
4583 | const char *const p = SvPV(ret, len); | |
4584 | ret = newSVpvn_flags(p, len, SVs_TEMP); | |
4585 | } | |
732caac7 DM |
4586 | if (rex->intflags & PREGf_USE_RE_EVAL) |
4587 | pm_flags |= PMf_USE_RE_EVAL; | |
4588 | ||
4589 | /* if we got here, it should be an engine which | |
4590 | * supports compiling code blocks and stuff */ | |
4591 | assert(rex->engine && rex->engine->op_comp); | |
ec841a27 | 4592 | assert(!(scan->flags & ~RXf_PMf_COMPILETIME)); |
575c37f6 | 4593 | re_sv = rex->engine->op_comp(aTHX_ &ret, 1, NULL, |
ec841a27 DM |
4594 | rex->engine, NULL, NULL, |
4595 | /* copy /msix etc to inner pattern */ | |
4596 | scan->flags, | |
4597 | pm_flags); | |
732caac7 | 4598 | |
9041c2e3 | 4599 | if (!(SvFLAGS(ret) |
faf82a0b | 4600 | & (SVs_TEMP | SVs_PADTMP | SVf_READONLY |
3ce3ed55 | 4601 | | SVs_GMG))) { |
a2794585 NC |
4602 | /* This isn't a first class regexp. Instead, it's |
4603 | caching a regexp onto an existing, Perl visible | |
4604 | scalar. */ | |
575c37f6 | 4605 | sv_magic(ret, MUTABLE_SV(re_sv), PERL_MAGIC_qr, 0, 0); |
3ce3ed55 | 4606 | } |
0f5d15d6 | 4607 | PL_regsize = osize; |
74088413 DM |
4608 | /* safe to do now that any $1 etc has been |
4609 | * interpolated into the new pattern string and | |
4610 | * compiled */ | |
4611 | S_regcp_restore(aTHX_ rex, runops_cp); | |
0f5d15d6 | 4612 | } |
575c37f6 | 4613 | re = (struct regexp *)SvANY(re_sv); |
4aabdb9b | 4614 | } |
07bc277f | 4615 | RXp_MATCH_COPIED_off(re); |
28d8d7f4 YO |
4616 | re->subbeg = rex->subbeg; |
4617 | re->sublen = rex->sublen; | |
6502e081 DM |
4618 | re->suboffset = rex->suboffset; |
4619 | re->subcoffset = rex->subcoffset; | |
f8fc2ecf | 4620 | rei = RXi_GET(re); |
6bda09f9 | 4621 | DEBUG_EXECUTE_r( |
f2ed9b32 | 4622 | debug_start_match(re_sv, utf8_target, locinput, PL_regeol, |
6bda09f9 YO |
4623 | "Matching embedded"); |
4624 | ); | |
f8fc2ecf | 4625 | startpoint = rei->program + 1; |
1a147d38 | 4626 | ST.close_paren = 0; /* only used for GOSUB */ |
aa283a38 | 4627 | |
1a147d38 | 4628 | eval_recurse_doit: /* Share code with GOSUB below this line */ |
aa283a38 | 4629 | /* run the pattern returned from (??{...}) */ |
b93070ed | 4630 | ST.cp = regcppush(rex, 0); /* Save *all* the positions. */ |
40a82448 | 4631 | REGCP_SET(ST.lastcp); |
6bda09f9 | 4632 | |
0357f1fd ML |
4633 | re->lastparen = 0; |
4634 | re->lastcloseparen = 0; | |
4635 | ||
ae0beba1 | 4636 | PL_regsize = 0; |
4aabdb9b DM |
4637 | |
4638 | /* XXXX This is too dramatic a measure... */ | |
4639 | PL_reg_maxiter = 0; | |
4640 | ||
faec1544 | 4641 | ST.toggle_reg_flags = PL_reg_flags; |
3c8556c3 | 4642 | if (RX_UTF8(re_sv)) |
faec1544 DM |
4643 | PL_reg_flags |= RF_utf8; |
4644 | else | |
4645 | PL_reg_flags &= ~RF_utf8; | |
4646 | ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */ | |
4647 | ||
288b8c02 | 4648 | ST.prev_rex = rex_sv; |
faec1544 | 4649 | ST.prev_curlyx = cur_curlyx; |
ec43f78b DM |
4650 | rex_sv = re_sv; |
4651 | SET_reg_curpm(rex_sv); | |
288b8c02 | 4652 | rex = re; |
f8fc2ecf | 4653 | rexi = rei; |
faec1544 | 4654 | cur_curlyx = NULL; |
40a82448 | 4655 | ST.B = next; |
faec1544 DM |
4656 | ST.prev_eval = cur_eval; |
4657 | cur_eval = st; | |
faec1544 | 4658 | /* now continue from first node in postoned RE */ |
4d5016e5 | 4659 | PUSH_YES_STATE_GOTO(EVAL_AB, startpoint, locinput); |
118e2215 | 4660 | assert(0); /* NOTREACHED */ |
c277df42 | 4661 | } |
40a82448 | 4662 | |
faec1544 DM |
4663 | case EVAL_AB: /* cleanup after a successful (??{A})B */ |
4664 | /* note: this is called twice; first after popping B, then A */ | |
4665 | PL_reg_flags ^= ST.toggle_reg_flags; | |
ec43f78b DM |
4666 | rex_sv = ST.prev_rex; |
4667 | SET_reg_curpm(rex_sv); | |
288b8c02 | 4668 | rex = (struct regexp *)SvANY(rex_sv); |
f8fc2ecf | 4669 | rexi = RXi_GET(rex); |
faec1544 DM |
4670 | regcpblow(ST.cp); |
4671 | cur_eval = ST.prev_eval; | |
4672 | cur_curlyx = ST.prev_curlyx; | |
34a81e2b | 4673 | |
40a82448 DM |
4674 | /* XXXX This is too dramatic a measure... */ |
4675 | PL_reg_maxiter = 0; | |
e7707071 | 4676 | if ( nochange_depth ) |
4b196cd4 | 4677 | nochange_depth--; |
262b90c4 | 4678 | sayYES; |
40a82448 | 4679 | |
40a82448 | 4680 | |
faec1544 DM |
4681 | case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */ |
4682 | /* note: this is called twice; first after popping B, then A */ | |
4683 | PL_reg_flags ^= ST.toggle_reg_flags; | |
ec43f78b DM |
4684 | rex_sv = ST.prev_rex; |
4685 | SET_reg_curpm(rex_sv); | |
288b8c02 | 4686 | rex = (struct regexp *)SvANY(rex_sv); |
f8fc2ecf | 4687 | rexi = RXi_GET(rex); |
0357f1fd | 4688 | |
40a82448 DM |
4689 | REGCP_UNWIND(ST.lastcp); |
4690 | regcppop(rex); | |
faec1544 DM |
4691 | cur_eval = ST.prev_eval; |
4692 | cur_curlyx = ST.prev_curlyx; | |
4693 | /* XXXX This is too dramatic a measure... */ | |
4694 | PL_reg_maxiter = 0; | |
e7707071 | 4695 | if ( nochange_depth ) |
4b196cd4 | 4696 | nochange_depth--; |
40a82448 | 4697 | sayNO_SILENT; |
40a82448 DM |
4698 | #undef ST |
4699 | ||
3c0563b9 | 4700 | case OPEN: /* ( */ |
c277df42 | 4701 | n = ARG(scan); /* which paren pair */ |
1ca2007e | 4702 | rex->offs[n].start_tmp = locinput - PL_bostr; |
3280af22 NIS |
4703 | if (n > PL_regsize) |
4704 | PL_regsize = n; | |
495f47a5 DM |
4705 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
4706 | "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf" tmp; regsize=%"UVuf"\n", | |
4707 | PTR2UV(rex), | |
4708 | PTR2UV(rex->offs), | |
4709 | (UV)n, | |
4710 | (IV)rex->offs[n].start_tmp, | |
4711 | (UV)PL_regsize | |
4712 | )); | |
e2e6a0f1 | 4713 | lastopen = n; |
a0d0e21e | 4714 | break; |
495f47a5 DM |
4715 | |
4716 | /* XXX really need to log other places start/end are set too */ | |
4717 | #define CLOSE_CAPTURE \ | |
4718 | rex->offs[n].start = rex->offs[n].start_tmp; \ | |
4719 | rex->offs[n].end = locinput - PL_bostr; \ | |
4720 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, \ | |
4721 | "rex=0x%"UVxf" offs=0x%"UVxf": \\%"UVuf": set %"IVdf"..%"IVdf"\n", \ | |
4722 | PTR2UV(rex), \ | |
4723 | PTR2UV(rex->offs), \ | |
4724 | (UV)n, \ | |
4725 | (IV)rex->offs[n].start, \ | |
4726 | (IV)rex->offs[n].end \ | |
4727 | )) | |
4728 | ||
3c0563b9 | 4729 | case CLOSE: /* ) */ |
c277df42 | 4730 | n = ARG(scan); /* which paren pair */ |
495f47a5 | 4731 | CLOSE_CAPTURE; |
7f69552c YO |
4732 | /*if (n > PL_regsize) |
4733 | PL_regsize = n;*/ | |
b93070ed DM |
4734 | if (n > rex->lastparen) |
4735 | rex->lastparen = n; | |
4736 | rex->lastcloseparen = n; | |
3b6647e0 | 4737 | if (cur_eval && cur_eval->u.eval.close_paren == n) { |
6bda09f9 YO |
4738 | goto fake_end; |
4739 | } | |
a0d0e21e | 4740 | break; |
3c0563b9 DM |
4741 | |
4742 | case ACCEPT: /* (*ACCEPT) */ | |
e2e6a0f1 YO |
4743 | if (ARG(scan)){ |
4744 | regnode *cursor; | |
4745 | for (cursor=scan; | |
4746 | cursor && OP(cursor)!=END; | |
4747 | cursor=regnext(cursor)) | |
4748 | { | |
4749 | if ( OP(cursor)==CLOSE ){ | |
4750 | n = ARG(cursor); | |
4751 | if ( n <= lastopen ) { | |
495f47a5 | 4752 | CLOSE_CAPTURE; |
e2e6a0f1 YO |
4753 | /*if (n > PL_regsize) |
4754 | PL_regsize = n;*/ | |
b93070ed DM |
4755 | if (n > rex->lastparen) |
4756 | rex->lastparen = n; | |
4757 | rex->lastcloseparen = n; | |
3b6647e0 RB |
4758 | if ( n == ARG(scan) || (cur_eval && |
4759 | cur_eval->u.eval.close_paren == n)) | |
e2e6a0f1 YO |
4760 | break; |
4761 | } | |
4762 | } | |
4763 | } | |
4764 | } | |
4765 | goto fake_end; | |
4766 | /*NOTREACHED*/ | |
3c0563b9 DM |
4767 | |
4768 | case GROUPP: /* (?(1)) */ | |
c277df42 | 4769 | n = ARG(scan); /* which paren pair */ |
b93070ed | 4770 | sw = cBOOL(rex->lastparen >= n && rex->offs[n].end != -1); |
c277df42 | 4771 | break; |
3c0563b9 DM |
4772 | |
4773 | case NGROUPP: /* (?(<name>)) */ | |
0a4db386 | 4774 | /* reg_check_named_buff_matched returns 0 for no match */ |
f2338a2e | 4775 | sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan)); |
0a4db386 | 4776 | break; |
3c0563b9 DM |
4777 | |
4778 | case INSUBP: /* (?(R)) */ | |
0a4db386 | 4779 | n = ARG(scan); |
3b6647e0 | 4780 | sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n)); |
0a4db386 | 4781 | break; |
3c0563b9 DM |
4782 | |
4783 | case DEFINEP: /* (?(DEFINE)) */ | |
0a4db386 YO |
4784 | sw = 0; |
4785 | break; | |
3c0563b9 DM |
4786 | |
4787 | case IFTHEN: /* (?(cond)A|B) */ | |
2c2d71f5 | 4788 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
24d3c4a9 | 4789 | if (sw) |
c277df42 IZ |
4790 | next = NEXTOPER(NEXTOPER(scan)); |
4791 | else { | |
4792 | next = scan + ARG(scan); | |
4793 | if (OP(next) == IFTHEN) /* Fake one. */ | |
4794 | next = NEXTOPER(NEXTOPER(next)); | |
4795 | } | |
4796 | break; | |
3c0563b9 DM |
4797 | |
4798 | case LOGICAL: /* modifier for EVAL and IFMATCH */ | |
24d3c4a9 | 4799 | logical = scan->flags; |
c277df42 | 4800 | break; |
c476f425 | 4801 | |
2ab05381 | 4802 | /******************************************************************* |
2ab05381 | 4803 | |
c476f425 DM |
4804 | The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/ |
4805 | pattern, where A and B are subpatterns. (For simple A, CURLYM or | |
4806 | STAR/PLUS/CURLY/CURLYN are used instead.) | |
2ab05381 | 4807 | |
c476f425 | 4808 | A*B is compiled as <CURLYX><A><WHILEM><B> |
2ab05381 | 4809 | |
c476f425 DM |
4810 | On entry to the subpattern, CURLYX is called. This pushes a CURLYX |
4811 | state, which contains the current count, initialised to -1. It also sets | |
4812 | cur_curlyx to point to this state, with any previous value saved in the | |
4813 | state block. | |
2ab05381 | 4814 | |
c476f425 DM |
4815 | CURLYX then jumps straight to the WHILEM op, rather than executing A, |
4816 | since the pattern may possibly match zero times (i.e. it's a while {} loop | |
4817 | rather than a do {} while loop). | |
2ab05381 | 4818 | |
c476f425 DM |
4819 | Each entry to WHILEM represents a successful match of A. The count in the |
4820 | CURLYX block is incremented, another WHILEM state is pushed, and execution | |
4821 | passes to A or B depending on greediness and the current count. | |
2ab05381 | 4822 | |
c476f425 DM |
4823 | For example, if matching against the string a1a2a3b (where the aN are |
4824 | substrings that match /A/), then the match progresses as follows: (the | |
4825 | pushed states are interspersed with the bits of strings matched so far): | |
2ab05381 | 4826 | |
c476f425 DM |
4827 | <CURLYX cnt=-1> |
4828 | <CURLYX cnt=0><WHILEM> | |
4829 | <CURLYX cnt=1><WHILEM> a1 <WHILEM> | |
4830 | <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM> | |
4831 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> | |
4832 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b | |
2ab05381 | 4833 | |
c476f425 DM |
4834 | (Contrast this with something like CURLYM, which maintains only a single |
4835 | backtrack state: | |
2ab05381 | 4836 | |
c476f425 DM |
4837 | <CURLYM cnt=0> a1 |
4838 | a1 <CURLYM cnt=1> a2 | |
4839 | a1 a2 <CURLYM cnt=2> a3 | |
4840 | a1 a2 a3 <CURLYM cnt=3> b | |
4841 | ) | |
2ab05381 | 4842 | |
c476f425 DM |
4843 | Each WHILEM state block marks a point to backtrack to upon partial failure |
4844 | of A or B, and also contains some minor state data related to that | |
4845 | iteration. The CURLYX block, pointed to by cur_curlyx, contains the | |
4846 | overall state, such as the count, and pointers to the A and B ops. | |
2ab05381 | 4847 | |
c476f425 DM |
4848 | This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx |
4849 | must always point to the *current* CURLYX block, the rules are: | |
2ab05381 | 4850 | |
c476f425 DM |
4851 | When executing CURLYX, save the old cur_curlyx in the CURLYX state block, |
4852 | and set cur_curlyx to point the new block. | |
2ab05381 | 4853 | |
c476f425 DM |
4854 | When popping the CURLYX block after a successful or unsuccessful match, |
4855 | restore the previous cur_curlyx. | |
2ab05381 | 4856 | |
c476f425 DM |
4857 | When WHILEM is about to execute B, save the current cur_curlyx, and set it |
4858 | to the outer one saved in the CURLYX block. | |
2ab05381 | 4859 | |
c476f425 DM |
4860 | When popping the WHILEM block after a successful or unsuccessful B match, |
4861 | restore the previous cur_curlyx. | |
2ab05381 | 4862 | |
c476f425 DM |
4863 | Here's an example for the pattern (AI* BI)*BO |
4864 | I and O refer to inner and outer, C and W refer to CURLYX and WHILEM: | |
2ab05381 | 4865 | |
c476f425 DM |
4866 | cur_ |
4867 | curlyx backtrack stack | |
4868 | ------ --------------- | |
4869 | NULL | |
4870 | CO <CO prev=NULL> <WO> | |
4871 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4872 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi | |
4873 | NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo | |
2ab05381 | 4874 | |
c476f425 DM |
4875 | At this point the pattern succeeds, and we work back down the stack to |
4876 | clean up, restoring as we go: | |
95b24440 | 4877 | |
c476f425 DM |
4878 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi |
4879 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4880 | CO <CO prev=NULL> <WO> | |
4881 | NULL | |
a0374537 | 4882 | |
c476f425 DM |
4883 | *******************************************************************/ |
4884 | ||
4885 | #define ST st->u.curlyx | |
4886 | ||
4887 | case CURLYX: /* start of /A*B/ (for complex A) */ | |
4888 | { | |
4889 | /* No need to save/restore up to this paren */ | |
4890 | I32 parenfloor = scan->flags; | |
4891 | ||
4892 | assert(next); /* keep Coverity happy */ | |
4893 | if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ | |
4894 | next += ARG(next); | |
4895 | ||
4896 | /* XXXX Probably it is better to teach regpush to support | |
4897 | parenfloor > PL_regsize... */ | |
b93070ed DM |
4898 | if (parenfloor > (I32)rex->lastparen) |
4899 | parenfloor = rex->lastparen; /* Pessimization... */ | |
c476f425 DM |
4900 | |
4901 | ST.prev_curlyx= cur_curlyx; | |
4902 | cur_curlyx = st; | |
4903 | ST.cp = PL_savestack_ix; | |
4904 | ||
4905 | /* these fields contain the state of the current curly. | |
4906 | * they are accessed by subsequent WHILEMs */ | |
4907 | ST.parenfloor = parenfloor; | |
d02d6d97 | 4908 | ST.me = scan; |
c476f425 | 4909 | ST.B = next; |
24d3c4a9 DM |
4910 | ST.minmod = minmod; |
4911 | minmod = 0; | |
c476f425 DM |
4912 | ST.count = -1; /* this will be updated by WHILEM */ |
4913 | ST.lastloc = NULL; /* this will be updated by WHILEM */ | |
4914 | ||
4d5016e5 | 4915 | PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next), locinput); |
118e2215 | 4916 | assert(0); /* NOTREACHED */ |
c476f425 | 4917 | } |
a0d0e21e | 4918 | |
c476f425 | 4919 | case CURLYX_end: /* just finished matching all of A*B */ |
c476f425 DM |
4920 | cur_curlyx = ST.prev_curlyx; |
4921 | sayYES; | |
118e2215 | 4922 | assert(0); /* NOTREACHED */ |
a0d0e21e | 4923 | |
c476f425 DM |
4924 | case CURLYX_end_fail: /* just failed to match all of A*B */ |
4925 | regcpblow(ST.cp); | |
4926 | cur_curlyx = ST.prev_curlyx; | |
4927 | sayNO; | |
118e2215 | 4928 | assert(0); /* NOTREACHED */ |
4633a7c4 | 4929 | |
a0d0e21e | 4930 | |
c476f425 DM |
4931 | #undef ST |
4932 | #define ST st->u.whilem | |
4933 | ||
4934 | case WHILEM: /* just matched an A in /A*B/ (for complex A) */ | |
4935 | { | |
4936 | /* see the discussion above about CURLYX/WHILEM */ | |
c476f425 | 4937 | I32 n; |
d02d6d97 DM |
4938 | int min = ARG1(cur_curlyx->u.curlyx.me); |
4939 | int max = ARG2(cur_curlyx->u.curlyx.me); | |
4940 | regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS; | |
4941 | ||
c476f425 DM |
4942 | assert(cur_curlyx); /* keep Coverity happy */ |
4943 | n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */ | |
4944 | ST.save_lastloc = cur_curlyx->u.curlyx.lastloc; | |
4945 | ST.cache_offset = 0; | |
4946 | ST.cache_mask = 0; | |
4947 | ||
c476f425 DM |
4948 | |
4949 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
d02d6d97 DM |
4950 | "%*s whilem: matched %ld out of %d..%d\n", |
4951 | REPORT_CODE_OFF+depth*2, "", (long)n, min, max) | |
c476f425 | 4952 | ); |
a0d0e21e | 4953 | |
c476f425 | 4954 | /* First just match a string of min A's. */ |
a0d0e21e | 4955 | |
d02d6d97 | 4956 | if (n < min) { |
b93070ed | 4957 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor); |
c476f425 | 4958 | cur_curlyx->u.curlyx.lastloc = locinput; |
92e82afa YO |
4959 | REGCP_SET(ST.lastcp); |
4960 | ||
4d5016e5 | 4961 | PUSH_STATE_GOTO(WHILEM_A_pre, A, locinput); |
118e2215 | 4962 | assert(0); /* NOTREACHED */ |
c476f425 DM |
4963 | } |
4964 | ||
4965 | /* If degenerate A matches "", assume A done. */ | |
4966 | ||
4967 | if (locinput == cur_curlyx->u.curlyx.lastloc) { | |
4968 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4969 | "%*s whilem: empty match detected, trying continuation...\n", | |
4970 | REPORT_CODE_OFF+depth*2, "") | |
4971 | ); | |
4972 | goto do_whilem_B_max; | |
4973 | } | |
4974 | ||
4975 | /* super-linear cache processing */ | |
4976 | ||
4977 | if (scan->flags) { | |
a0d0e21e | 4978 | |
2c2d71f5 | 4979 | if (!PL_reg_maxiter) { |
c476f425 DM |
4980 | /* start the countdown: Postpone detection until we |
4981 | * know the match is not *that* much linear. */ | |
2c2d71f5 | 4982 | PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4); |
66bf836d DM |
4983 | /* possible overflow for long strings and many CURLYX's */ |
4984 | if (PL_reg_maxiter < 0) | |
4985 | PL_reg_maxiter = I32_MAX; | |
2c2d71f5 JH |
4986 | PL_reg_leftiter = PL_reg_maxiter; |
4987 | } | |
c476f425 | 4988 | |
2c2d71f5 | 4989 | if (PL_reg_leftiter-- == 0) { |
c476f425 | 4990 | /* initialise cache */ |
3298f257 | 4991 | const I32 size = (PL_reg_maxiter + 7)/8; |
2c2d71f5 | 4992 | if (PL_reg_poscache) { |
eb160463 | 4993 | if ((I32)PL_reg_poscache_size < size) { |
2c2d71f5 JH |
4994 | Renew(PL_reg_poscache, size, char); |
4995 | PL_reg_poscache_size = size; | |
4996 | } | |
4997 | Zero(PL_reg_poscache, size, char); | |
4998 | } | |
4999 | else { | |
5000 | PL_reg_poscache_size = size; | |
a02a5408 | 5001 | Newxz(PL_reg_poscache, size, char); |
2c2d71f5 | 5002 | } |
c476f425 DM |
5003 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
5004 | "%swhilem: Detected a super-linear match, switching on caching%s...\n", | |
5005 | PL_colors[4], PL_colors[5]) | |
5006 | ); | |
2c2d71f5 | 5007 | } |
c476f425 | 5008 | |
2c2d71f5 | 5009 | if (PL_reg_leftiter < 0) { |
c476f425 DM |
5010 | /* have we already failed at this position? */ |
5011 | I32 offset, mask; | |
5012 | offset = (scan->flags & 0xf) - 1 | |
5013 | + (locinput - PL_bostr) * (scan->flags>>4); | |
5014 | mask = 1 << (offset % 8); | |
5015 | offset /= 8; | |
5016 | if (PL_reg_poscache[offset] & mask) { | |
5017 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
5018 | "%*s whilem: (cache) already tried at this position...\n", | |
5019 | REPORT_CODE_OFF+depth*2, "") | |
2c2d71f5 | 5020 | ); |
3298f257 | 5021 | sayNO; /* cache records failure */ |
2c2d71f5 | 5022 | } |
c476f425 DM |
5023 | ST.cache_offset = offset; |
5024 | ST.cache_mask = mask; | |
2c2d71f5 | 5025 | } |
c476f425 | 5026 | } |
2c2d71f5 | 5027 | |
c476f425 | 5028 | /* Prefer B over A for minimal matching. */ |
a687059c | 5029 | |
c476f425 DM |
5030 | if (cur_curlyx->u.curlyx.minmod) { |
5031 | ST.save_curlyx = cur_curlyx; | |
5032 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
b93070ed | 5033 | ST.cp = regcppush(rex, ST.save_curlyx->u.curlyx.parenfloor); |
c476f425 | 5034 | REGCP_SET(ST.lastcp); |
4d5016e5 DM |
5035 | PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B, |
5036 | locinput); | |
118e2215 | 5037 | assert(0); /* NOTREACHED */ |
c476f425 | 5038 | } |
a0d0e21e | 5039 | |
c476f425 DM |
5040 | /* Prefer A over B for maximal matching. */ |
5041 | ||
d02d6d97 | 5042 | if (n < max) { /* More greed allowed? */ |
b93070ed | 5043 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor); |
c476f425 DM |
5044 | cur_curlyx->u.curlyx.lastloc = locinput; |
5045 | REGCP_SET(ST.lastcp); | |
4d5016e5 | 5046 | PUSH_STATE_GOTO(WHILEM_A_max, A, locinput); |
118e2215 | 5047 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5048 | } |
5049 | goto do_whilem_B_max; | |
5050 | } | |
118e2215 | 5051 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5052 | |
5053 | case WHILEM_B_min: /* just matched B in a minimal match */ | |
5054 | case WHILEM_B_max: /* just matched B in a maximal match */ | |
5055 | cur_curlyx = ST.save_curlyx; | |
5056 | sayYES; | |
118e2215 | 5057 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5058 | |
5059 | case WHILEM_B_max_fail: /* just failed to match B in a maximal match */ | |
5060 | cur_curlyx = ST.save_curlyx; | |
5061 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
5062 | cur_curlyx->u.curlyx.count--; | |
5063 | CACHEsayNO; | |
118e2215 | 5064 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5065 | |
5066 | case WHILEM_A_min_fail: /* just failed to match A in a minimal match */ | |
c476f425 DM |
5067 | /* FALL THROUGH */ |
5068 | case WHILEM_A_pre_fail: /* just failed to match even minimal A */ | |
92e82afa YO |
5069 | REGCP_UNWIND(ST.lastcp); |
5070 | regcppop(rex); | |
c476f425 DM |
5071 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; |
5072 | cur_curlyx->u.curlyx.count--; | |
5073 | CACHEsayNO; | |
118e2215 | 5074 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5075 | |
5076 | case WHILEM_A_max_fail: /* just failed to match A in a maximal match */ | |
5077 | REGCP_UNWIND(ST.lastcp); | |
5078 | regcppop(rex); /* Restore some previous $<digit>s? */ | |
c476f425 DM |
5079 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
5080 | "%*s whilem: failed, trying continuation...\n", | |
5081 | REPORT_CODE_OFF+depth*2, "") | |
5082 | ); | |
5083 | do_whilem_B_max: | |
5084 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
5085 | && ckWARN(WARN_REGEXP) | |
5086 | && !(PL_reg_flags & RF_warned)) | |
5087 | { | |
5088 | PL_reg_flags |= RF_warned; | |
dcbac5bb FC |
5089 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
5090 | "Complex regular subexpression recursion limit (%d) " | |
5091 | "exceeded", | |
c476f425 DM |
5092 | REG_INFTY - 1); |
5093 | } | |
5094 | ||
5095 | /* now try B */ | |
5096 | ST.save_curlyx = cur_curlyx; | |
5097 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4d5016e5 DM |
5098 | PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B, |
5099 | locinput); | |
118e2215 | 5100 | assert(0); /* NOTREACHED */ |
c476f425 DM |
5101 | |
5102 | case WHILEM_B_min_fail: /* just failed to match B in a minimal match */ | |
5103 | cur_curlyx = ST.save_curlyx; | |
5104 | REGCP_UNWIND(ST.lastcp); | |
5105 | regcppop(rex); | |
5106 | ||
d02d6d97 | 5107 | if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) { |
c476f425 DM |
5108 | /* Maximum greed exceeded */ |
5109 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
5110 | && ckWARN(WARN_REGEXP) | |
5111 | && !(PL_reg_flags & RF_warned)) | |
5112 | { | |
3280af22 | 5113 | PL_reg_flags |= RF_warned; |
c476f425 | 5114 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
dcbac5bb FC |
5115 | "Complex regular subexpression recursion " |
5116 | "limit (%d) exceeded", | |
c476f425 | 5117 | REG_INFTY - 1); |
a0d0e21e | 5118 | } |
c476f425 | 5119 | cur_curlyx->u.curlyx.count--; |
3ab3c9b4 | 5120 | CACHEsayNO; |
a0d0e21e | 5121 | } |
c476f425 DM |
5122 | |
5123 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
5124 | "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "") | |
5125 | ); | |
5126 | /* Try grabbing another A and see if it helps. */ | |
c476f425 | 5127 | cur_curlyx->u.curlyx.lastloc = locinput; |
b93070ed | 5128 | ST.cp = regcppush(rex, cur_curlyx->u.curlyx.parenfloor); |
c476f425 | 5129 | REGCP_SET(ST.lastcp); |
d02d6d97 | 5130 | PUSH_STATE_GOTO(WHILEM_A_min, |
4d5016e5 DM |
5131 | /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS, |
5132 | locinput); | |
118e2215 | 5133 | assert(0); /* NOTREACHED */ |
40a82448 DM |
5134 | |
5135 | #undef ST | |
5136 | #define ST st->u.branch | |
5137 | ||
5138 | case BRANCHJ: /* /(...|A|...)/ with long next pointer */ | |
c277df42 IZ |
5139 | next = scan + ARG(scan); |
5140 | if (next == scan) | |
5141 | next = NULL; | |
40a82448 DM |
5142 | scan = NEXTOPER(scan); |
5143 | /* FALL THROUGH */ | |
c277df42 | 5144 | |
40a82448 DM |
5145 | case BRANCH: /* /(...|A|...)/ */ |
5146 | scan = NEXTOPER(scan); /* scan now points to inner node */ | |
b93070ed | 5147 | ST.lastparen = rex->lastparen; |
f6033a9d | 5148 | ST.lastcloseparen = rex->lastcloseparen; |
40a82448 DM |
5149 | ST.next_branch = next; |
5150 | REGCP_SET(ST.cp); | |
02db2b7b | 5151 | |
40a82448 | 5152 | /* Now go into the branch */ |
5d458dd8 | 5153 | if (has_cutgroup) { |
4d5016e5 | 5154 | PUSH_YES_STATE_GOTO(BRANCH_next, scan, locinput); |
5d458dd8 | 5155 | } else { |
4d5016e5 | 5156 | PUSH_STATE_GOTO(BRANCH_next, scan, locinput); |
5d458dd8 | 5157 | } |
118e2215 | 5158 | assert(0); /* NOTREACHED */ |
3c0563b9 DM |
5159 | |
5160 | case CUTGROUP: /* /(*THEN)/ */ | |
5d458dd8 | 5161 | sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : |
ad64d0ec | 5162 | MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
4d5016e5 | 5163 | PUSH_STATE_GOTO(CUTGROUP_next, next, locinput); |
118e2215 | 5164 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5165 | |
5d458dd8 YO |
5166 | case CUTGROUP_next_fail: |
5167 | do_cutgroup = 1; | |
5168 | no_final = 1; | |
5169 | if (st->u.mark.mark_name) | |
5170 | sv_commit = st->u.mark.mark_name; | |
5171 | sayNO; | |
118e2215 | 5172 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5173 | |
5d458dd8 YO |
5174 | case BRANCH_next: |
5175 | sayYES; | |
118e2215 | 5176 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5177 | |
40a82448 | 5178 | case BRANCH_next_fail: /* that branch failed; try the next, if any */ |
5d458dd8 YO |
5179 | if (do_cutgroup) { |
5180 | do_cutgroup = 0; | |
5181 | no_final = 0; | |
5182 | } | |
40a82448 | 5183 | REGCP_UNWIND(ST.cp); |
a8d1f4b4 | 5184 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
40a82448 DM |
5185 | scan = ST.next_branch; |
5186 | /* no more branches? */ | |
5d458dd8 YO |
5187 | if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) { |
5188 | DEBUG_EXECUTE_r({ | |
5189 | PerlIO_printf( Perl_debug_log, | |
5190 | "%*s %sBRANCH failed...%s\n", | |
5191 | REPORT_CODE_OFF+depth*2, "", | |
5192 | PL_colors[4], | |
5193 | PL_colors[5] ); | |
5194 | }); | |
5195 | sayNO_SILENT; | |
5196 | } | |
40a82448 | 5197 | continue; /* execute next BRANCH[J] op */ |
118e2215 | 5198 | assert(0); /* NOTREACHED */ |
40a82448 | 5199 | |
3c0563b9 | 5200 | case MINMOD: /* next op will be non-greedy, e.g. A*? */ |
24d3c4a9 | 5201 | minmod = 1; |
a0d0e21e | 5202 | break; |
40a82448 DM |
5203 | |
5204 | #undef ST | |
5205 | #define ST st->u.curlym | |
5206 | ||
5207 | case CURLYM: /* /A{m,n}B/ where A is fixed-length */ | |
5208 | ||
5209 | /* This is an optimisation of CURLYX that enables us to push | |
84d2fa14 | 5210 | * only a single backtracking state, no matter how many matches |
40a82448 DM |
5211 | * there are in {m,n}. It relies on the pattern being constant |
5212 | * length, with no parens to influence future backrefs | |
5213 | */ | |
5214 | ||
5215 | ST.me = scan; | |
dc45a647 | 5216 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; |
40a82448 | 5217 | |
f6033a9d DM |
5218 | ST.lastparen = rex->lastparen; |
5219 | ST.lastcloseparen = rex->lastcloseparen; | |
5220 | ||
40a82448 DM |
5221 | /* if paren positive, emulate an OPEN/CLOSE around A */ |
5222 | if (ST.me->flags) { | |
3b6647e0 | 5223 | U32 paren = ST.me->flags; |
40a82448 DM |
5224 | if (paren > PL_regsize) |
5225 | PL_regsize = paren; | |
c277df42 | 5226 | scan += NEXT_OFF(scan); /* Skip former OPEN. */ |
6407bf3b | 5227 | } |
40a82448 DM |
5228 | ST.A = scan; |
5229 | ST.B = next; | |
5230 | ST.alen = 0; | |
5231 | ST.count = 0; | |
24d3c4a9 DM |
5232 | ST.minmod = minmod; |
5233 | minmod = 0; | |
40a82448 DM |
5234 | ST.c1 = CHRTEST_UNINIT; |
5235 | REGCP_SET(ST.cp); | |
6407bf3b | 5236 | |
40a82448 DM |
5237 | if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */ |
5238 | goto curlym_do_B; | |
5239 | ||
5240 | curlym_do_A: /* execute the A in /A{m,n}B/ */ | |
4d5016e5 | 5241 | PUSH_YES_STATE_GOTO(CURLYM_A, ST.A, locinput); /* match A */ |
118e2215 | 5242 | assert(0); /* NOTREACHED */ |
5f80c4cf | 5243 | |
40a82448 | 5244 | case CURLYM_A: /* we've just matched an A */ |
40a82448 DM |
5245 | ST.count++; |
5246 | /* after first match, determine A's length: u.curlym.alen */ | |
5247 | if (ST.count == 1) { | |
5248 | if (PL_reg_match_utf8) { | |
c07e9d7b DM |
5249 | char *s = st->locinput; |
5250 | while (s < locinput) { | |
40a82448 DM |
5251 | ST.alen++; |
5252 | s += UTF8SKIP(s); | |
5253 | } | |
5254 | } | |
5255 | else { | |
c07e9d7b | 5256 | ST.alen = locinput - st->locinput; |
40a82448 DM |
5257 | } |
5258 | if (ST.alen == 0) | |
5259 | ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me); | |
5260 | } | |
0cadcf80 DM |
5261 | DEBUG_EXECUTE_r( |
5262 | PerlIO_printf(Perl_debug_log, | |
40a82448 | 5263 | "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n", |
5bc10b2c | 5264 | (int)(REPORT_CODE_OFF+(depth*2)), "", |
40a82448 | 5265 | (IV) ST.count, (IV)ST.alen) |
0cadcf80 DM |
5266 | ); |
5267 | ||
0a4db386 | 5268 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 5269 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5270 | goto fake_end; |
5271 | ||
c966426a DM |
5272 | { |
5273 | I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)); | |
5274 | if ( max == REG_INFTY || ST.count < max ) | |
5275 | goto curlym_do_A; /* try to match another A */ | |
5276 | } | |
40a82448 | 5277 | goto curlym_do_B; /* try to match B */ |
5f80c4cf | 5278 | |
40a82448 DM |
5279 | case CURLYM_A_fail: /* just failed to match an A */ |
5280 | REGCP_UNWIND(ST.cp); | |
0a4db386 YO |
5281 | |
5282 | if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ | |
5283 | || (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 5284 | cur_eval->u.eval.close_paren == (U32)ST.me->flags)) |
40a82448 | 5285 | sayNO; |
0cadcf80 | 5286 | |
40a82448 | 5287 | curlym_do_B: /* execute the B in /A{m,n}B/ */ |
40a82448 DM |
5288 | if (ST.c1 == CHRTEST_UNINIT) { |
5289 | /* calculate c1 and c2 for possible match of 1st char | |
5290 | * following curly */ | |
5291 | ST.c1 = ST.c2 = CHRTEST_VOID; | |
5292 | if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) { | |
5293 | regnode *text_node = ST.B; | |
5294 | if (! HAS_TEXT(text_node)) | |
5295 | FIND_NEXT_IMPT(text_node); | |
ee9b8eae YO |
5296 | /* this used to be |
5297 | ||
5298 | (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT) | |
5299 | ||
5300 | But the former is redundant in light of the latter. | |
5301 | ||
5302 | if this changes back then the macro for | |
5303 | IS_TEXT and friends need to change. | |
5304 | */ | |
5305 | if (PL_regkind[OP(text_node)] == EXACT) | |
40a82448 | 5306 | { |
ee9b8eae | 5307 | |
40a82448 | 5308 | ST.c1 = (U8)*STRING(text_node); |
9a5a5549 KW |
5309 | switch (OP(text_node)) { |
5310 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
2f7f8cb1 | 5311 | case EXACTFA: |
3c760661 | 5312 | case EXACTFU_SS: |
fab2782b | 5313 | case EXACTFU_TRICKYFOLD: |
9a5a5549 KW |
5314 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; |
5315 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
5316 | default: ST.c2 = ST.c1; | |
5317 | } | |
c277df42 | 5318 | } |
c277df42 | 5319 | } |
40a82448 DM |
5320 | } |
5321 | ||
5322 | DEBUG_EXECUTE_r( | |
5323 | PerlIO_printf(Perl_debug_log, | |
5324 | "%*s CURLYM trying tail with matches=%"IVdf"...\n", | |
5bc10b2c | 5325 | (int)(REPORT_CODE_OFF+(depth*2)), |
40a82448 DM |
5326 | "", (IV)ST.count) |
5327 | ); | |
5328 | if (ST.c1 != CHRTEST_VOID | |
c07e9d7b DM |
5329 | && nextchr != ST.c1 |
5330 | && nextchr != ST.c2) | |
40a82448 DM |
5331 | { |
5332 | /* simulate B failing */ | |
3e901dc0 YO |
5333 | DEBUG_OPTIMISE_r( |
5334 | PerlIO_printf(Perl_debug_log, | |
5335 | "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n", | |
5336 | (int)(REPORT_CODE_OFF+(depth*2)),"", | |
5337 | (IV)ST.c1,(IV)ST.c2 | |
5338 | )); | |
40a82448 DM |
5339 | state_num = CURLYM_B_fail; |
5340 | goto reenter_switch; | |
5341 | } | |
5342 | ||
5343 | if (ST.me->flags) { | |
f6033a9d | 5344 | /* emulate CLOSE: mark current A as captured */ |
40a82448 DM |
5345 | I32 paren = ST.me->flags; |
5346 | if (ST.count) { | |
b93070ed | 5347 | rex->offs[paren].start |
c07e9d7b DM |
5348 | = HOPc(locinput, -ST.alen) - PL_bostr; |
5349 | rex->offs[paren].end = locinput - PL_bostr; | |
f6033a9d DM |
5350 | if ((U32)paren > rex->lastparen) |
5351 | rex->lastparen = paren; | |
5352 | rex->lastcloseparen = paren; | |
c277df42 | 5353 | } |
40a82448 | 5354 | else |
b93070ed | 5355 | rex->offs[paren].end = -1; |
0a4db386 | 5356 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 5357 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5358 | { |
5359 | if (ST.count) | |
5360 | goto fake_end; | |
5361 | else | |
5362 | sayNO; | |
5363 | } | |
c277df42 | 5364 | } |
0a4db386 | 5365 | |
4d5016e5 | 5366 | PUSH_STATE_GOTO(CURLYM_B, ST.B, locinput); /* match B */ |
118e2215 | 5367 | assert(0); /* NOTREACHED */ |
40a82448 DM |
5368 | |
5369 | case CURLYM_B_fail: /* just failed to match a B */ | |
5370 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 | 5371 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); |
40a82448 | 5372 | if (ST.minmod) { |
84d2fa14 HS |
5373 | I32 max = ARG2(ST.me); |
5374 | if (max != REG_INFTY && ST.count == max) | |
40a82448 DM |
5375 | sayNO; |
5376 | goto curlym_do_A; /* try to match a further A */ | |
5377 | } | |
5378 | /* backtrack one A */ | |
5379 | if (ST.count == ARG1(ST.me) /* min */) | |
5380 | sayNO; | |
5381 | ST.count--; | |
5382 | locinput = HOPc(locinput, -ST.alen); | |
c07e9d7b | 5383 | nextchr = UCHARAT(locinput); |
40a82448 DM |
5384 | goto curlym_do_B; /* try to match B */ |
5385 | ||
c255a977 DM |
5386 | #undef ST |
5387 | #define ST st->u.curly | |
40a82448 | 5388 | |
c255a977 DM |
5389 | #define CURLY_SETPAREN(paren, success) \ |
5390 | if (paren) { \ | |
5391 | if (success) { \ | |
b93070ed DM |
5392 | rex->offs[paren].start = HOPc(locinput, -1) - PL_bostr; \ |
5393 | rex->offs[paren].end = locinput - PL_bostr; \ | |
f6033a9d DM |
5394 | if (paren > rex->lastparen) \ |
5395 | rex->lastparen = paren; \ | |
b93070ed | 5396 | rex->lastcloseparen = paren; \ |
c255a977 | 5397 | } \ |
f6033a9d | 5398 | else { \ |
b93070ed | 5399 | rex->offs[paren].end = -1; \ |
f6033a9d DM |
5400 | rex->lastparen = ST.lastparen; \ |
5401 | rex->lastcloseparen = ST.lastcloseparen; \ | |
5402 | } \ | |
c255a977 DM |
5403 | } |
5404 | ||
5405 | case STAR: /* /A*B/ where A is width 1 */ | |
5406 | ST.paren = 0; | |
5407 | ST.min = 0; | |
5408 | ST.max = REG_INFTY; | |
a0d0e21e LW |
5409 | scan = NEXTOPER(scan); |
5410 | goto repeat; | |
3c0563b9 | 5411 | |
c255a977 DM |
5412 | case PLUS: /* /A+B/ where A is width 1 */ |
5413 | ST.paren = 0; | |
5414 | ST.min = 1; | |
5415 | ST.max = REG_INFTY; | |
c277df42 | 5416 | scan = NEXTOPER(scan); |
c255a977 | 5417 | goto repeat; |
3c0563b9 | 5418 | |
c255a977 DM |
5419 | case CURLYN: /* /(A){m,n}B/ where A is width 1 */ |
5420 | ST.paren = scan->flags; /* Which paren to set */ | |
f6033a9d DM |
5421 | ST.lastparen = rex->lastparen; |
5422 | ST.lastcloseparen = rex->lastcloseparen; | |
c255a977 DM |
5423 | if (ST.paren > PL_regsize) |
5424 | PL_regsize = ST.paren; | |
c255a977 DM |
5425 | ST.min = ARG1(scan); /* min to match */ |
5426 | ST.max = ARG2(scan); /* max to match */ | |
0a4db386 | 5427 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5428 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5429 | ST.min=1; |
5430 | ST.max=1; | |
5431 | } | |
c255a977 DM |
5432 | scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); |
5433 | goto repeat; | |
3c0563b9 | 5434 | |
c255a977 DM |
5435 | case CURLY: /* /A{m,n}B/ where A is width 1 */ |
5436 | ST.paren = 0; | |
5437 | ST.min = ARG1(scan); /* min to match */ | |
5438 | ST.max = ARG2(scan); /* max to match */ | |
5439 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; | |
c277df42 | 5440 | repeat: |
a0d0e21e LW |
5441 | /* |
5442 | * Lookahead to avoid useless match attempts | |
5443 | * when we know what character comes next. | |
c255a977 | 5444 | * |
5f80c4cf JP |
5445 | * Used to only do .*x and .*?x, but now it allows |
5446 | * for )'s, ('s and (?{ ... })'s to be in the way | |
5447 | * of the quantifier and the EXACT-like node. -- japhy | |
5448 | */ | |
5449 | ||
c255a977 DM |
5450 | if (ST.min > ST.max) /* XXX make this a compile-time check? */ |
5451 | sayNO; | |
cca55fe3 | 5452 | if (HAS_TEXT(next) || JUMPABLE(next)) { |
5f80c4cf JP |
5453 | U8 *s; |
5454 | regnode *text_node = next; | |
5455 | ||
3dab1dad YO |
5456 | if (! HAS_TEXT(text_node)) |
5457 | FIND_NEXT_IMPT(text_node); | |
5f80c4cf | 5458 | |
9e137952 | 5459 | if (! HAS_TEXT(text_node)) |
c255a977 | 5460 | ST.c1 = ST.c2 = CHRTEST_VOID; |
5f80c4cf | 5461 | else { |
ee9b8eae | 5462 | if ( PL_regkind[OP(text_node)] != EXACT ) { |
c255a977 | 5463 | ST.c1 = ST.c2 = CHRTEST_VOID; |
44a68960 | 5464 | goto assume_ok_easy; |
cca55fe3 | 5465 | } |
be8e71aa YO |
5466 | else |
5467 | s = (U8*)STRING(text_node); | |
ee9b8eae YO |
5468 | |
5469 | /* Currently we only get here when | |
5470 | ||
5471 | PL_rekind[OP(text_node)] == EXACT | |
5472 | ||
5473 | if this changes back then the macro for IS_TEXT and | |
5474 | friends need to change. */ | |
f2ed9b32 | 5475 | if (!UTF_PATTERN) { |
9a5a5549 KW |
5476 | ST.c1 = *s; |
5477 | switch (OP(text_node)) { | |
5478 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
2f7f8cb1 | 5479 | case EXACTFA: |
3c760661 | 5480 | case EXACTFU_SS: |
fab2782b | 5481 | case EXACTFU_TRICKYFOLD: |
9a5a5549 KW |
5482 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; |
5483 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
5484 | default: ST.c2 = ST.c1; break; | |
5485 | } | |
1aa99e6b | 5486 | } |
f2ed9b32 | 5487 | else { /* UTF_PATTERN */ |
9a5a5549 | 5488 | if (IS_TEXTFU(text_node) || IS_TEXTF(text_node)) { |
b0573d8b KW |
5489 | STRLEN ulen; |
5490 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; | |
5491 | ||
5492 | to_utf8_fold((U8*)s, tmpbuf, &ulen); | |
5493 | ST.c1 = ST.c2 = utf8n_to_uvchr(tmpbuf, UTF8_MAXLEN, 0, | |
e294cc5d | 5494 | uniflags); |
5f80c4cf JP |
5495 | } |
5496 | else { | |
c255a977 | 5497 | ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, |
041457d9 | 5498 | uniflags); |
5f80c4cf | 5499 | } |
1aa99e6b IH |
5500 | } |
5501 | } | |
bbce6d69 | 5502 | } |
a0d0e21e | 5503 | else |
c255a977 | 5504 | ST.c1 = ST.c2 = CHRTEST_VOID; |
cca55fe3 | 5505 | assume_ok_easy: |
c255a977 DM |
5506 | |
5507 | ST.A = scan; | |
5508 | ST.B = next; | |
24d3c4a9 | 5509 | if (minmod) { |
eb72505d | 5510 | char *li = locinput; |
24d3c4a9 | 5511 | minmod = 0; |
eb72505d | 5512 | if (ST.min && regrepeat(rex, &li, ST.A, ST.min, depth) < ST.min) |
4633a7c4 | 5513 | sayNO; |
eb72505d DM |
5514 | locinput = li; |
5515 | nextchr = UCHARAT(locinput); | |
c255a977 | 5516 | ST.count = ST.min; |
c255a977 DM |
5517 | REGCP_SET(ST.cp); |
5518 | if (ST.c1 == CHRTEST_VOID) | |
5519 | goto curly_try_B_min; | |
5520 | ||
5521 | ST.oldloc = locinput; | |
5522 | ||
5523 | /* set ST.maxpos to the furthest point along the | |
5524 | * string that could possibly match */ | |
5525 | if (ST.max == REG_INFTY) { | |
5526 | ST.maxpos = PL_regeol - 1; | |
f2ed9b32 | 5527 | if (utf8_target) |
c255a977 DM |
5528 | while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos)) |
5529 | ST.maxpos--; | |
5530 | } | |
f2ed9b32 | 5531 | else if (utf8_target) { |
c255a977 DM |
5532 | int m = ST.max - ST.min; |
5533 | for (ST.maxpos = locinput; | |
5534 | m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--) | |
5535 | ST.maxpos += UTF8SKIP(ST.maxpos); | |
5536 | } | |
5537 | else { | |
5538 | ST.maxpos = locinput + ST.max - ST.min; | |
5539 | if (ST.maxpos >= PL_regeol) | |
5540 | ST.maxpos = PL_regeol - 1; | |
5541 | } | |
5542 | goto curly_try_B_min_known; | |
5543 | ||
5544 | } | |
5545 | else { | |
eb72505d DM |
5546 | /* avoid taking address of locinput, so it can remain |
5547 | * a register var */ | |
5548 | char *li = locinput; | |
5549 | ST.count = regrepeat(rex, &li, ST.A, ST.max, depth); | |
c255a977 DM |
5550 | if (ST.count < ST.min) |
5551 | sayNO; | |
eb72505d DM |
5552 | locinput = li; |
5553 | nextchr = UCHARAT(locinput); | |
c255a977 DM |
5554 | if ((ST.count > ST.min) |
5555 | && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL)) | |
5556 | { | |
5557 | /* A{m,n} must come at the end of the string, there's | |
5558 | * no point in backing off ... */ | |
5559 | ST.min = ST.count; | |
5560 | /* ...except that $ and \Z can match before *and* after | |
5561 | newline at the end. Consider "\n\n" =~ /\n+\Z\n/. | |
5562 | We may back off by one in this case. */ | |
eb72505d | 5563 | if (UCHARAT(locinput - 1) == '\n' && OP(ST.B) != EOS) |
c255a977 DM |
5564 | ST.min--; |
5565 | } | |
5566 | REGCP_SET(ST.cp); | |
5567 | goto curly_try_B_max; | |
5568 | } | |
118e2215 | 5569 | assert(0); /* NOTREACHED */ |
c255a977 DM |
5570 | |
5571 | ||
5572 | case CURLY_B_min_known_fail: | |
5573 | /* failed to find B in a non-greedy match where c1,c2 valid */ | |
c255a977 | 5574 | |
c255a977 | 5575 | REGCP_UNWIND(ST.cp); |
a8d1f4b4 DM |
5576 | if (ST.paren) { |
5577 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
5578 | } | |
c255a977 DM |
5579 | /* Couldn't or didn't -- move forward. */ |
5580 | ST.oldloc = locinput; | |
f2ed9b32 | 5581 | if (utf8_target) |
c255a977 DM |
5582 | locinput += UTF8SKIP(locinput); |
5583 | else | |
5584 | locinput++; | |
5585 | ST.count++; | |
5586 | curly_try_B_min_known: | |
5587 | /* find the next place where 'B' could work, then call B */ | |
5588 | { | |
5589 | int n; | |
f2ed9b32 | 5590 | if (utf8_target) { |
c255a977 DM |
5591 | n = (ST.oldloc == locinput) ? 0 : 1; |
5592 | if (ST.c1 == ST.c2) { | |
5593 | STRLEN len; | |
5594 | /* set n to utf8_distance(oldloc, locinput) */ | |
5595 | while (locinput <= ST.maxpos && | |
5596 | utf8n_to_uvchr((U8*)locinput, | |
5597 | UTF8_MAXBYTES, &len, | |
5598 | uniflags) != (UV)ST.c1) { | |
5599 | locinput += len; | |
5600 | n++; | |
5601 | } | |
1aa99e6b IH |
5602 | } |
5603 | else { | |
c255a977 DM |
5604 | /* set n to utf8_distance(oldloc, locinput) */ |
5605 | while (locinput <= ST.maxpos) { | |
5606 | STRLEN len; | |
5607 | const UV c = utf8n_to_uvchr((U8*)locinput, | |
5608 | UTF8_MAXBYTES, &len, | |
5609 | uniflags); | |
5610 | if (c == (UV)ST.c1 || c == (UV)ST.c2) | |
5611 | break; | |
5612 | locinput += len; | |
5613 | n++; | |
1aa99e6b | 5614 | } |
0fe9bf95 IZ |
5615 | } |
5616 | } | |
c255a977 DM |
5617 | else { |
5618 | if (ST.c1 == ST.c2) { | |
5619 | while (locinput <= ST.maxpos && | |
5620 | UCHARAT(locinput) != ST.c1) | |
5621 | locinput++; | |
bbce6d69 | 5622 | } |
c255a977 DM |
5623 | else { |
5624 | while (locinput <= ST.maxpos | |
5625 | && UCHARAT(locinput) != ST.c1 | |
5626 | && UCHARAT(locinput) != ST.c2) | |
5627 | locinput++; | |
a0ed51b3 | 5628 | } |
c255a977 DM |
5629 | n = locinput - ST.oldloc; |
5630 | } | |
5631 | if (locinput > ST.maxpos) | |
5632 | sayNO; | |
c255a977 | 5633 | if (n) { |
eb72505d DM |
5634 | /* In /a{m,n}b/, ST.oldloc is at "a" x m, locinput is |
5635 | * at b; check that everything between oldloc and | |
5636 | * locinput matches */ | |
5637 | char *li = ST.oldloc; | |
c255a977 | 5638 | ST.count += n; |
eb72505d | 5639 | if (regrepeat(rex, &li, ST.A, n, depth) < n) |
4633a7c4 | 5640 | sayNO; |
eb72505d | 5641 | assert(n == REG_INFTY || locinput == li); |
a0d0e21e | 5642 | } |
c255a977 | 5643 | CURLY_SETPAREN(ST.paren, ST.count); |
0a4db386 | 5644 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5645 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5646 | goto fake_end; |
5647 | } | |
4d5016e5 | 5648 | PUSH_STATE_GOTO(CURLY_B_min_known, ST.B, locinput); |
a0d0e21e | 5649 | } |
118e2215 | 5650 | assert(0); /* NOTREACHED */ |
c255a977 DM |
5651 | |
5652 | ||
5653 | case CURLY_B_min_fail: | |
5654 | /* failed to find B in a non-greedy match where c1,c2 invalid */ | |
c255a977 DM |
5655 | |
5656 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 DM |
5657 | if (ST.paren) { |
5658 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
5659 | } | |
c255a977 | 5660 | /* failed -- move forward one */ |
f73aaa43 | 5661 | { |
eb72505d DM |
5662 | char *li = locinput; |
5663 | if (!regrepeat(rex, &li, ST.A, 1, depth)) { | |
f73aaa43 DM |
5664 | sayNO; |
5665 | } | |
eb72505d | 5666 | locinput = li; |
f73aaa43 DM |
5667 | } |
5668 | { | |
c255a977 | 5669 | ST.count++; |
c255a977 DM |
5670 | if (ST.count <= ST.max || (ST.max == REG_INFTY && |
5671 | ST.count > 0)) /* count overflow ? */ | |
15272685 | 5672 | { |
c255a977 DM |
5673 | curly_try_B_min: |
5674 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5675 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5676 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5677 | goto fake_end; |
5678 | } | |
4d5016e5 | 5679 | PUSH_STATE_GOTO(CURLY_B_min, ST.B, locinput); |
a0d0e21e LW |
5680 | } |
5681 | } | |
118e2215 | 5682 | assert(0); /* NOTREACHED */ |
c255a977 DM |
5683 | |
5684 | ||
5685 | curly_try_B_max: | |
5686 | /* a successful greedy match: now try to match B */ | |
40d049e4 | 5687 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5688 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
40d049e4 YO |
5689 | goto fake_end; |
5690 | } | |
c255a977 DM |
5691 | { |
5692 | UV c = 0; | |
5693 | if (ST.c1 != CHRTEST_VOID) | |
eb72505d | 5694 | c = utf8_target ? utf8n_to_uvchr((U8*)locinput, |
c255a977 | 5695 | UTF8_MAXBYTES, 0, uniflags) |
eb72505d | 5696 | : (UV) UCHARAT(locinput); |
c255a977 DM |
5697 | /* If it could work, try it. */ |
5698 | if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) { | |
5699 | CURLY_SETPAREN(ST.paren, ST.count); | |
4d5016e5 | 5700 | PUSH_STATE_GOTO(CURLY_B_max, ST.B, locinput); |
118e2215 | 5701 | assert(0); /* NOTREACHED */ |
c255a977 DM |
5702 | } |
5703 | } | |
5704 | /* FALL THROUGH */ | |
3c0563b9 | 5705 | |
c255a977 DM |
5706 | case CURLY_B_max_fail: |
5707 | /* failed to find B in a greedy match */ | |
c255a977 DM |
5708 | |
5709 | REGCP_UNWIND(ST.cp); | |
a8d1f4b4 DM |
5710 | if (ST.paren) { |
5711 | UNWIND_PAREN(ST.lastparen, ST.lastcloseparen); | |
5712 | } | |
c255a977 DM |
5713 | /* back up. */ |
5714 | if (--ST.count < ST.min) | |
5715 | sayNO; | |
eb72505d | 5716 | locinput = HOPc(locinput, -1); |
c255a977 DM |
5717 | goto curly_try_B_max; |
5718 | ||
5719 | #undef ST | |
5720 | ||
3c0563b9 | 5721 | case END: /* last op of main pattern */ |
6bda09f9 | 5722 | fake_end: |
faec1544 DM |
5723 | if (cur_eval) { |
5724 | /* we've just finished A in /(??{A})B/; now continue with B */ | |
faec1544 DM |
5725 | st->u.eval.toggle_reg_flags |
5726 | = cur_eval->u.eval.toggle_reg_flags; | |
5727 | PL_reg_flags ^= st->u.eval.toggle_reg_flags; | |
5728 | ||
288b8c02 | 5729 | st->u.eval.prev_rex = rex_sv; /* inner */ |
b93070ed | 5730 | st->u.eval.cp = regcppush(rex, 0); /* Save *all* the positions. */ |
ec43f78b DM |
5731 | rex_sv = cur_eval->u.eval.prev_rex; |
5732 | SET_reg_curpm(rex_sv); | |
288b8c02 | 5733 | rex = (struct regexp *)SvANY(rex_sv); |
f8fc2ecf | 5734 | rexi = RXi_GET(rex); |
faec1544 | 5735 | cur_curlyx = cur_eval->u.eval.prev_curlyx; |
34a81e2b | 5736 | |
faec1544 | 5737 | REGCP_SET(st->u.eval.lastcp); |
faec1544 DM |
5738 | |
5739 | /* Restore parens of the outer rex without popping the | |
5740 | * savestack */ | |
74088413 | 5741 | S_regcp_restore(aTHX_ rex, cur_eval->u.eval.lastcp); |
faec1544 DM |
5742 | |
5743 | st->u.eval.prev_eval = cur_eval; | |
5744 | cur_eval = cur_eval->u.eval.prev_eval; | |
5745 | DEBUG_EXECUTE_r( | |
2a49f0f5 JH |
5746 | PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n", |
5747 | REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval));); | |
e7707071 YO |
5748 | if ( nochange_depth ) |
5749 | nochange_depth--; | |
5750 | ||
4d5016e5 DM |
5751 | PUSH_YES_STATE_GOTO(EVAL_AB, st->u.eval.prev_eval->u.eval.B, |
5752 | locinput); /* match B */ | |
faec1544 DM |
5753 | } |
5754 | ||
3b0527fe | 5755 | if (locinput < reginfo->till) { |
a3621e74 | 5756 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
7821416a IZ |
5757 | "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n", |
5758 | PL_colors[4], | |
5759 | (long)(locinput - PL_reg_starttry), | |
3b0527fe | 5760 | (long)(reginfo->till - PL_reg_starttry), |
7821416a | 5761 | PL_colors[5])); |
58e23c8d | 5762 | |
262b90c4 | 5763 | sayNO_SILENT; /* Cannot match: too short. */ |
7821416a | 5764 | } |
262b90c4 | 5765 | sayYES; /* Success! */ |
dad79028 DM |
5766 | |
5767 | case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */ | |
5768 | DEBUG_EXECUTE_r( | |
5769 | PerlIO_printf(Perl_debug_log, | |
5770 | "%*s %ssubpattern success...%s\n", | |
5bc10b2c | 5771 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])); |
262b90c4 | 5772 | sayYES; /* Success! */ |
dad79028 | 5773 | |
40a82448 DM |
5774 | #undef ST |
5775 | #define ST st->u.ifmatch | |
5776 | ||
37f53970 DM |
5777 | { |
5778 | char *newstart; | |
5779 | ||
40a82448 DM |
5780 | case SUSPEND: /* (?>A) */ |
5781 | ST.wanted = 1; | |
37f53970 | 5782 | newstart = locinput; |
9041c2e3 | 5783 | goto do_ifmatch; |
dad79028 | 5784 | |
40a82448 DM |
5785 | case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */ |
5786 | ST.wanted = 0; | |
dad79028 DM |
5787 | goto ifmatch_trivial_fail_test; |
5788 | ||
40a82448 DM |
5789 | case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */ |
5790 | ST.wanted = 1; | |
dad79028 | 5791 | ifmatch_trivial_fail_test: |
a0ed51b3 | 5792 | if (scan->flags) { |
52657f30 | 5793 | char * const s = HOPBACKc(locinput, scan->flags); |
dad79028 DM |
5794 | if (!s) { |
5795 | /* trivial fail */ | |
24d3c4a9 DM |
5796 | if (logical) { |
5797 | logical = 0; | |
f2338a2e | 5798 | sw = 1 - cBOOL(ST.wanted); |
dad79028 | 5799 | } |
40a82448 | 5800 | else if (ST.wanted) |
dad79028 DM |
5801 | sayNO; |
5802 | next = scan + ARG(scan); | |
5803 | if (next == scan) | |
5804 | next = NULL; | |
5805 | break; | |
5806 | } | |
37f53970 | 5807 | newstart = s; |
a0ed51b3 LW |
5808 | } |
5809 | else | |
37f53970 | 5810 | newstart = locinput; |
a0ed51b3 | 5811 | |
c277df42 | 5812 | do_ifmatch: |
40a82448 | 5813 | ST.me = scan; |
24d3c4a9 | 5814 | ST.logical = logical; |
24d786f4 YO |
5815 | logical = 0; /* XXX: reset state of logical once it has been saved into ST */ |
5816 | ||
40a82448 | 5817 | /* execute body of (?...A) */ |
37f53970 | 5818 | PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)), newstart); |
118e2215 | 5819 | assert(0); /* NOTREACHED */ |
37f53970 | 5820 | } |
40a82448 DM |
5821 | |
5822 | case IFMATCH_A_fail: /* body of (?...A) failed */ | |
5823 | ST.wanted = !ST.wanted; | |
5824 | /* FALL THROUGH */ | |
5825 | ||
5826 | case IFMATCH_A: /* body of (?...A) succeeded */ | |
24d3c4a9 | 5827 | if (ST.logical) { |
f2338a2e | 5828 | sw = cBOOL(ST.wanted); |
40a82448 DM |
5829 | } |
5830 | else if (!ST.wanted) | |
5831 | sayNO; | |
5832 | ||
37f53970 DM |
5833 | if (OP(ST.me) != SUSPEND) { |
5834 | /* restore old position except for (?>...) */ | |
5835 | locinput = st->locinput; | |
40a82448 DM |
5836 | } |
5837 | scan = ST.me + ARG(ST.me); | |
5838 | if (scan == ST.me) | |
5839 | scan = NULL; | |
5840 | continue; /* execute B */ | |
5841 | ||
5842 | #undef ST | |
dad79028 | 5843 | |
3c0563b9 DM |
5844 | case LONGJMP: /* alternative with many branches compiles to |
5845 | * (BRANCHJ; EXACT ...; LONGJMP ) x N */ | |
c277df42 IZ |
5846 | next = scan + ARG(scan); |
5847 | if (next == scan) | |
5848 | next = NULL; | |
a0d0e21e | 5849 | break; |
3c0563b9 DM |
5850 | |
5851 | case COMMIT: /* (*COMMIT) */ | |
e2e6a0f1 YO |
5852 | reginfo->cutpoint = PL_regeol; |
5853 | /* FALLTHROUGH */ | |
3c0563b9 DM |
5854 | |
5855 | case PRUNE: /* (*PRUNE) */ | |
e2e6a0f1 | 5856 | if (!scan->flags) |
ad64d0ec | 5857 | sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
4d5016e5 | 5858 | PUSH_STATE_GOTO(COMMIT_next, next, locinput); |
118e2215 | 5859 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5860 | |
54612592 YO |
5861 | case COMMIT_next_fail: |
5862 | no_final = 1; | |
5863 | /* FALLTHROUGH */ | |
3c0563b9 DM |
5864 | |
5865 | case OPFAIL: /* (*FAIL) */ | |
7f69552c | 5866 | sayNO; |
118e2215 | 5867 | assert(0); /* NOTREACHED */ |
e2e6a0f1 YO |
5868 | |
5869 | #define ST st->u.mark | |
3c0563b9 | 5870 | case MARKPOINT: /* (*MARK:foo) */ |
e2e6a0f1 | 5871 | ST.prev_mark = mark_state; |
5d458dd8 | 5872 | ST.mark_name = sv_commit = sv_yes_mark |
ad64d0ec | 5873 | = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
e2e6a0f1 | 5874 | mark_state = st; |
4d5016e5 DM |
5875 | ST.mark_loc = locinput; |
5876 | PUSH_YES_STATE_GOTO(MARKPOINT_next, next, locinput); | |
118e2215 | 5877 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5878 | |
e2e6a0f1 YO |
5879 | case MARKPOINT_next: |
5880 | mark_state = ST.prev_mark; | |
5881 | sayYES; | |
118e2215 | 5882 | assert(0); /* NOTREACHED */ |
3c0563b9 | 5883 | |
e2e6a0f1 | 5884 | case MARKPOINT_next_fail: |
5d458dd8 | 5885 | if (popmark && sv_eq(ST.mark_name,popmark)) |
e2e6a0f1 YO |
5886 | { |
5887 | if (ST.mark_loc > startpoint) | |
5888 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5889 | popmark = NULL; /* we found our mark */ | |
5890 | sv_commit = ST.mark_name; | |
5891 | ||
5892 | DEBUG_EXECUTE_r({ | |
5d458dd8 | 5893 | PerlIO_printf(Perl_debug_log, |
e2e6a0f1 YO |
5894 | "%*s %ssetting cutpoint to mark:%"SVf"...%s\n", |
5895 | REPORT_CODE_OFF+depth*2, "", | |
be2597df | 5896 | PL_colors[4], SVfARG(sv_commit), PL_colors[5]); |
e2e6a0f1 YO |
5897 | }); |
5898 | } | |
5899 | mark_state = ST.prev_mark; | |
5d458dd8 YO |
5900 | sv_yes_mark = mark_state ? |
5901 | mark_state->u.mark.mark_name : NULL; | |
e2e6a0f1 | 5902 | sayNO; |
118e2215 | 5903 | assert(0); /* NOTREACHED */ |
3c0563b9 DM |
5904 | |
5905 | case SKIP: /* (*SKIP) */ | |
5d458dd8 | 5906 | if (scan->flags) { |
2bf803e2 | 5907 | /* (*SKIP) : if we fail we cut here*/ |
5d458dd8 | 5908 | ST.mark_name = NULL; |
e2e6a0f1 | 5909 | ST.mark_loc = locinput; |
4d5016e5 | 5910 | PUSH_STATE_GOTO(SKIP_next,next, locinput); |
5d458dd8 | 5911 | } else { |
2bf803e2 | 5912 | /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was, |
5d458dd8 YO |
5913 | otherwise do nothing. Meaning we need to scan |
5914 | */ | |
5915 | regmatch_state *cur = mark_state; | |
ad64d0ec | 5916 | SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
5917 | |
5918 | while (cur) { | |
5919 | if ( sv_eq( cur->u.mark.mark_name, | |
5920 | find ) ) | |
5921 | { | |
5922 | ST.mark_name = find; | |
4d5016e5 | 5923 | PUSH_STATE_GOTO( SKIP_next, next, locinput); |
5d458dd8 YO |
5924 | } |
5925 | cur = cur->u.mark.prev_mark; | |
5926 | } | |
e2e6a0f1 | 5927 | } |
2bf803e2 | 5928 | /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */ |
5d458dd8 | 5929 | break; |
3c0563b9 | 5930 | |
5d458dd8 YO |
5931 | case SKIP_next_fail: |
5932 | if (ST.mark_name) { | |
5933 | /* (*CUT:NAME) - Set up to search for the name as we | |
5934 | collapse the stack*/ | |
5935 | popmark = ST.mark_name; | |
5936 | } else { | |
5937 | /* (*CUT) - No name, we cut here.*/ | |
e2e6a0f1 YO |
5938 | if (ST.mark_loc > startpoint) |
5939 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5d458dd8 YO |
5940 | /* but we set sv_commit to latest mark_name if there |
5941 | is one so they can test to see how things lead to this | |
5942 | cut */ | |
5943 | if (mark_state) | |
5944 | sv_commit=mark_state->u.mark.mark_name; | |
5945 | } | |
e2e6a0f1 YO |
5946 | no_final = 1; |
5947 | sayNO; | |
118e2215 | 5948 | assert(0); /* NOTREACHED */ |
e2e6a0f1 | 5949 | #undef ST |
3c0563b9 DM |
5950 | |
5951 | case LNBREAK: /* \R */ | |
f2ed9b32 | 5952 | if ((n=is_LNBREAK(locinput,utf8_target))) { |
e1d1eefb | 5953 | locinput += n; |
e1d1eefb YO |
5954 | } else |
5955 | sayNO; | |
5956 | break; | |
5957 | ||
5958 | #define CASE_CLASS(nAmE) \ | |
5959 | case nAmE: \ | |
1380b51d DM |
5960 | if (locinput >= PL_regeol) \ |
5961 | sayNO; \ | |
f2ed9b32 | 5962 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb | 5963 | locinput += n; \ |
e1d1eefb YO |
5964 | } else \ |
5965 | sayNO; \ | |
5966 | break; \ | |
5967 | case N##nAmE: \ | |
1380b51d DM |
5968 | if (locinput >= PL_regeol) \ |
5969 | sayNO; \ | |
f2ed9b32 | 5970 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5971 | sayNO; \ |
5972 | } else { \ | |
5973 | locinput += UTF8SKIP(locinput); \ | |
e1d1eefb YO |
5974 | } \ |
5975 | break | |
5976 | ||
3c0563b9 DM |
5977 | CASE_CLASS(VERTWS); /* \v \V */ |
5978 | CASE_CLASS(HORIZWS); /* \h \H */ | |
e1d1eefb YO |
5979 | #undef CASE_CLASS |
5980 | ||
a0d0e21e | 5981 | default: |
b900a521 | 5982 | PerlIO_printf(Perl_error_log, "%"UVxf" %d\n", |
d7d93a81 | 5983 | PTR2UV(scan), OP(scan)); |
cea2e8a9 | 5984 | Perl_croak(aTHX_ "regexp memory corruption"); |
28b98f76 DM |
5985 | |
5986 | /* this is a point to jump to in order to increment | |
5987 | * locinput by one character */ | |
5988 | increment_locinput: | |
5989 | if (utf8_target) { | |
5990 | locinput += PL_utf8skip[nextchr]; | |
5991 | if (locinput > PL_regeol) | |
5992 | sayNO; | |
28b98f76 DM |
5993 | } |
5994 | else | |
3640db6b | 5995 | locinput++; |
28b98f76 | 5996 | break; |
5d458dd8 YO |
5997 | |
5998 | } /* end switch */ | |
95b24440 | 5999 | |
5d458dd8 YO |
6000 | /* switch break jumps here */ |
6001 | scan = next; /* prepare to execute the next op and ... */ | |
6002 | continue; /* ... jump back to the top, reusing st */ | |
118e2215 | 6003 | assert(0); /* NOTREACHED */ |
95b24440 | 6004 | |
40a82448 DM |
6005 | push_yes_state: |
6006 | /* push a state that backtracks on success */ | |
6007 | st->u.yes.prev_yes_state = yes_state; | |
6008 | yes_state = st; | |
6009 | /* FALL THROUGH */ | |
6010 | push_state: | |
6011 | /* push a new regex state, then continue at scan */ | |
6012 | { | |
6013 | regmatch_state *newst; | |
6014 | ||
24b23f37 YO |
6015 | DEBUG_STACK_r({ |
6016 | regmatch_state *cur = st; | |
6017 | regmatch_state *curyes = yes_state; | |
6018 | int curd = depth; | |
6019 | regmatch_slab *slab = PL_regmatch_slab; | |
6020 | for (;curd > -1;cur--,curd--) { | |
6021 | if (cur < SLAB_FIRST(slab)) { | |
6022 | slab = slab->prev; | |
6023 | cur = SLAB_LAST(slab); | |
6024 | } | |
6025 | PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n", | |
6026 | REPORT_CODE_OFF + 2 + depth * 2,"", | |
13d6edb4 | 6027 | curd, PL_reg_name[cur->resume_state], |
24b23f37 YO |
6028 | (curyes == cur) ? "yes" : "" |
6029 | ); | |
6030 | if (curyes == cur) | |
6031 | curyes = cur->u.yes.prev_yes_state; | |
6032 | } | |
6033 | } else | |
6034 | DEBUG_STATE_pp("push") | |
6035 | ); | |
40a82448 | 6036 | depth++; |
40a82448 DM |
6037 | st->locinput = locinput; |
6038 | newst = st+1; | |
6039 | if (newst > SLAB_LAST(PL_regmatch_slab)) | |
6040 | newst = S_push_slab(aTHX); | |
6041 | PL_regmatch_state = newst; | |
786e8c11 | 6042 | |
4d5016e5 | 6043 | locinput = pushinput; |
40a82448 DM |
6044 | st = newst; |
6045 | continue; | |
118e2215 | 6046 | assert(0); /* NOTREACHED */ |
40a82448 | 6047 | } |
a0d0e21e | 6048 | } |
a687059c | 6049 | |
a0d0e21e LW |
6050 | /* |
6051 | * We get here only if there's trouble -- normally "case END" is | |
6052 | * the terminating point. | |
6053 | */ | |
cea2e8a9 | 6054 | Perl_croak(aTHX_ "corrupted regexp pointers"); |
a0d0e21e | 6055 | /*NOTREACHED*/ |
4633a7c4 LW |
6056 | sayNO; |
6057 | ||
262b90c4 | 6058 | yes: |
77cb431f DM |
6059 | if (yes_state) { |
6060 | /* we have successfully completed a subexpression, but we must now | |
6061 | * pop to the state marked by yes_state and continue from there */ | |
77cb431f | 6062 | assert(st != yes_state); |
5bc10b2c DM |
6063 | #ifdef DEBUGGING |
6064 | while (st != yes_state) { | |
6065 | st--; | |
6066 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
6067 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
6068 | st = SLAB_LAST(PL_regmatch_slab); | |
6069 | } | |
e2e6a0f1 | 6070 | DEBUG_STATE_r({ |
54612592 YO |
6071 | if (no_final) { |
6072 | DEBUG_STATE_pp("pop (no final)"); | |
6073 | } else { | |
6074 | DEBUG_STATE_pp("pop (yes)"); | |
6075 | } | |
e2e6a0f1 | 6076 | }); |
5bc10b2c DM |
6077 | depth--; |
6078 | } | |
6079 | #else | |
77cb431f DM |
6080 | while (yes_state < SLAB_FIRST(PL_regmatch_slab) |
6081 | || yes_state > SLAB_LAST(PL_regmatch_slab)) | |
6082 | { | |
6083 | /* not in this slab, pop slab */ | |
6084 | depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1); | |
6085 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
6086 | st = SLAB_LAST(PL_regmatch_slab); | |
6087 | } | |
6088 | depth -= (st - yes_state); | |
5bc10b2c | 6089 | #endif |
77cb431f DM |
6090 | st = yes_state; |
6091 | yes_state = st->u.yes.prev_yes_state; | |
6092 | PL_regmatch_state = st; | |
24b23f37 | 6093 | |
3640db6b | 6094 | if (no_final) |
5d458dd8 | 6095 | locinput= st->locinput; |
54612592 | 6096 | state_num = st->resume_state + no_final; |
24d3c4a9 | 6097 | goto reenter_switch; |
77cb431f DM |
6098 | } |
6099 | ||
a3621e74 | 6100 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", |
e4584336 | 6101 | PL_colors[4], PL_colors[5])); |
02db2b7b | 6102 | |
ed301438 | 6103 | if (PL_reg_state.re_state_eval_setup_done) { |
19b95bf0 DM |
6104 | /* each successfully executed (?{...}) block does the equivalent of |
6105 | * local $^R = do {...} | |
6106 | * When popping the save stack, all these locals would be undone; | |
6107 | * bypass this by setting the outermost saved $^R to the latest | |
6108 | * value */ | |
6109 | if (oreplsv != GvSV(PL_replgv)) | |
6110 | sv_setsv(oreplsv, GvSV(PL_replgv)); | |
6111 | } | |
95b24440 | 6112 | result = 1; |
aa283a38 | 6113 | goto final_exit; |
4633a7c4 LW |
6114 | |
6115 | no: | |
a3621e74 | 6116 | DEBUG_EXECUTE_r( |
7821416a | 6117 | PerlIO_printf(Perl_debug_log, |
786e8c11 | 6118 | "%*s %sfailed...%s\n", |
5bc10b2c | 6119 | REPORT_CODE_OFF+depth*2, "", |
786e8c11 | 6120 | PL_colors[4], PL_colors[5]) |
7821416a | 6121 | ); |
aa283a38 | 6122 | |
262b90c4 | 6123 | no_silent: |
54612592 YO |
6124 | if (no_final) { |
6125 | if (yes_state) { | |
6126 | goto yes; | |
6127 | } else { | |
6128 | goto final_exit; | |
6129 | } | |
6130 | } | |
aa283a38 DM |
6131 | if (depth) { |
6132 | /* there's a previous state to backtrack to */ | |
40a82448 DM |
6133 | st--; |
6134 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
6135 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
6136 | st = SLAB_LAST(PL_regmatch_slab); | |
6137 | } | |
6138 | PL_regmatch_state = st; | |
40a82448 | 6139 | locinput= st->locinput; |
40a82448 | 6140 | |
5bc10b2c DM |
6141 | DEBUG_STATE_pp("pop"); |
6142 | depth--; | |
262b90c4 DM |
6143 | if (yes_state == st) |
6144 | yes_state = st->u.yes.prev_yes_state; | |
5bc10b2c | 6145 | |
24d3c4a9 DM |
6146 | state_num = st->resume_state + 1; /* failure = success + 1 */ |
6147 | goto reenter_switch; | |
95b24440 | 6148 | } |
24d3c4a9 | 6149 | result = 0; |
aa283a38 | 6150 | |
262b90c4 | 6151 | final_exit: |
bbe252da | 6152 | if (rex->intflags & PREGf_VERBARG_SEEN) { |
5d458dd8 YO |
6153 | SV *sv_err = get_sv("REGERROR", 1); |
6154 | SV *sv_mrk = get_sv("REGMARK", 1); | |
6155 | if (result) { | |
e2e6a0f1 | 6156 | sv_commit = &PL_sv_no; |
5d458dd8 YO |
6157 | if (!sv_yes_mark) |
6158 | sv_yes_mark = &PL_sv_yes; | |
6159 | } else { | |
6160 | if (!sv_commit) | |
6161 | sv_commit = &PL_sv_yes; | |
6162 | sv_yes_mark = &PL_sv_no; | |
6163 | } | |
6164 | sv_setsv(sv_err, sv_commit); | |
6165 | sv_setsv(sv_mrk, sv_yes_mark); | |
e2e6a0f1 | 6166 | } |
19b95bf0 | 6167 | |
81ed78b2 DM |
6168 | |
6169 | if (last_pushed_cv) { | |
6170 | dSP; | |
6171 | POP_MULTICALL; | |
4f8dbb2d | 6172 | PERL_UNUSED_VAR(SP); |
81ed78b2 DM |
6173 | } |
6174 | ||
2f554ef7 DM |
6175 | /* clean up; in particular, free all slabs above current one */ |
6176 | LEAVE_SCOPE(oldsave); | |
5d9a96ca | 6177 | |
730f4c74 DM |
6178 | assert(!result || locinput - PL_bostr >= 0); |
6179 | return result ? locinput - PL_bostr : -1; | |
a687059c LW |
6180 | } |
6181 | ||
6182 | /* | |
6183 | - regrepeat - repeatedly match something simple, report how many | |
d60de1d1 DM |
6184 | * |
6185 | * startposp - pointer a pointer to the start position. This is updated | |
6186 | * to point to the byte following the highest successful | |
6187 | * match. | |
6188 | * p - the regnode to be repeatedly matched against. | |
6189 | * max - maximum number of characters to match. | |
6190 | * depth - (for debugging) backtracking depth. | |
a687059c | 6191 | */ |
76e3520e | 6192 | STATIC I32 |
f73aaa43 | 6193 | S_regrepeat(pTHX_ const regexp *prog, char **startposp, const regnode *p, I32 max, int depth) |
a687059c | 6194 | { |
27da23d5 | 6195 | dVAR; |
eb578fdb KW |
6196 | char *scan; |
6197 | I32 c; | |
6198 | char *loceol = PL_regeol; | |
6199 | I32 hardcount = 0; | |
6200 | bool utf8_target = PL_reg_match_utf8; | |
d513472c | 6201 | UV utf8_flags; |
4f55667c SP |
6202 | #ifndef DEBUGGING |
6203 | PERL_UNUSED_ARG(depth); | |
6204 | #endif | |
a0d0e21e | 6205 | |
7918f24d NC |
6206 | PERL_ARGS_ASSERT_REGREPEAT; |
6207 | ||
f73aaa43 | 6208 | scan = *startposp; |
faf11cac HS |
6209 | if (max == REG_INFTY) |
6210 | max = I32_MAX; | |
6211 | else if (max < loceol - scan) | |
7f596f4c | 6212 | loceol = scan + max; |
a0d0e21e | 6213 | switch (OP(p)) { |
22c35a8c | 6214 | case REG_ANY: |
f2ed9b32 | 6215 | if (utf8_target) { |
ffc61ed2 | 6216 | loceol = PL_regeol; |
1aa99e6b | 6217 | while (scan < loceol && hardcount < max && *scan != '\n') { |
ffc61ed2 JH |
6218 | scan += UTF8SKIP(scan); |
6219 | hardcount++; | |
6220 | } | |
6221 | } else { | |
6222 | while (scan < loceol && *scan != '\n') | |
6223 | scan++; | |
a0ed51b3 LW |
6224 | } |
6225 | break; | |
ffc61ed2 | 6226 | case SANY: |
f2ed9b32 | 6227 | if (utf8_target) { |
def8e4ea | 6228 | loceol = PL_regeol; |
a0804c9e | 6229 | while (scan < loceol && hardcount < max) { |
def8e4ea JH |
6230 | scan += UTF8SKIP(scan); |
6231 | hardcount++; | |
6232 | } | |
6233 | } | |
6234 | else | |
6235 | scan = loceol; | |
a0ed51b3 | 6236 | break; |
f33976b4 DB |
6237 | case CANY: |
6238 | scan = loceol; | |
6239 | break; | |
59d32103 KW |
6240 | case EXACT: |
6241 | /* To get here, EXACTish nodes must have *byte* length == 1. That | |
6242 | * means they match only characters in the string that can be expressed | |
6243 | * as a single byte. For non-utf8 strings, that means a simple match. | |
6244 | * For utf8 strings, the character matched must be an invariant, or | |
6245 | * downgradable to a single byte. The pattern's utf8ness is | |
6246 | * irrelevant, as since it's a single byte, it either isn't utf8, or if | |
6247 | * it is, it's an invariant */ | |
6248 | ||
6249 | c = (U8)*STRING(p); | |
6250 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); | |
6251 | ||
6252 | if (! utf8_target || UNI_IS_INVARIANT(c)) { | |
6253 | while (scan < loceol && UCHARAT(scan) == c) { | |
6254 | scan++; | |
6255 | } | |
6256 | } | |
6257 | else { | |
6258 | ||
6259 | /* Here, the string is utf8, and the pattern char is different | |
6260 | * in utf8 than not, so can't compare them directly. Outside the | |
5908807f | 6261 | * loop, find the two utf8 bytes that represent c, and then |
59d32103 KW |
6262 | * look for those in sequence in the utf8 string */ |
6263 | U8 high = UTF8_TWO_BYTE_HI(c); | |
6264 | U8 low = UTF8_TWO_BYTE_LO(c); | |
6265 | loceol = PL_regeol; | |
6266 | ||
6267 | while (hardcount < max | |
6268 | && scan + 1 < loceol | |
6269 | && UCHARAT(scan) == high | |
6270 | && UCHARAT(scan + 1) == low) | |
6271 | { | |
6272 | scan += 2; | |
6273 | hardcount++; | |
6274 | } | |
6275 | } | |
6276 | break; | |
2f7f8cb1 KW |
6277 | case EXACTFA: |
6278 | utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
6279 | goto do_exactf; | |
6280 | ||
d4e0b827 KW |
6281 | case EXACTFL: |
6282 | PL_reg_flags |= RF_tainted; | |
17580e7a KW |
6283 | utf8_flags = FOLDEQ_UTF8_LOCALE; |
6284 | goto do_exactf; | |
6285 | ||
d4e0b827 | 6286 | case EXACTF: |
62bf7766 KW |
6287 | utf8_flags = 0; |
6288 | goto do_exactf; | |
6289 | ||
3c760661 | 6290 | case EXACTFU_SS: |
fab2782b | 6291 | case EXACTFU_TRICKYFOLD: |
9a5a5549 | 6292 | case EXACTFU: |
05f861a2 | 6293 | utf8_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0; |
59d32103 | 6294 | |
2be3e190 KW |
6295 | /* The comments for the EXACT case above apply as well to these fold |
6296 | * ones */ | |
634c83a2 | 6297 | |
2f7f8cb1 | 6298 | do_exactf: |
090f7165 | 6299 | c = (U8)*STRING(p); |
634c83a2 | 6300 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); |
d4e0b827 | 6301 | |
3c760661 | 6302 | if (utf8_target || OP(p) == EXACTFU_SS) { /* Use full Unicode fold matching */ |
59d32103 KW |
6303 | char *tmpeol = loceol; |
6304 | while (hardcount < max | |
d513472c KW |
6305 | && foldEQ_utf8_flags(scan, &tmpeol, 0, utf8_target, |
6306 | STRING(p), NULL, 1, cBOOL(UTF_PATTERN), utf8_flags)) | |
59d32103 KW |
6307 | { |
6308 | scan = tmpeol; | |
6309 | tmpeol = loceol; | |
6310 | hardcount++; | |
6311 | } | |
634c83a2 | 6312 | |
59d32103 KW |
6313 | /* XXX Note that the above handles properly the German sharp s in |
6314 | * the pattern matching ss in the string. But it doesn't handle | |
6315 | * properly cases where the string contains say 'LIGATURE ff' and | |
6316 | * the pattern is 'f+'. This would require, say, a new function or | |
6317 | * revised interface to foldEQ_utf8(), in which the maximum number | |
6318 | * of characters to match could be passed and it would return how | |
6319 | * many actually did. This is just one of many cases where | |
6320 | * multi-char folds don't work properly, and so the fix is being | |
6321 | * deferred */ | |
6322 | } | |
6323 | else { | |
87381386 | 6324 | U8 folded; |
59d32103 | 6325 | |
2be3e190 KW |
6326 | /* Here, the string isn't utf8 and c is a single byte; and either |
6327 | * the pattern isn't utf8 or c is an invariant, so its utf8ness | |
6328 | * doesn't affect c. Can just do simple comparisons for exact or | |
6329 | * fold matching. */ | |
d4e0b827 | 6330 | switch (OP(p)) { |
87381386 | 6331 | case EXACTF: folded = PL_fold[c]; break; |
2f7f8cb1 | 6332 | case EXACTFA: |
fab2782b | 6333 | case EXACTFU_TRICKYFOLD: |
9a5a5549 | 6334 | case EXACTFU: folded = PL_fold_latin1[c]; break; |
87381386 KW |
6335 | case EXACTFL: folded = PL_fold_locale[c]; break; |
6336 | default: Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p)); | |
6337 | } | |
6338 | while (scan < loceol && | |
6339 | (UCHARAT(scan) == c || UCHARAT(scan) == folded)) | |
6340 | { | |
6341 | scan++; | |
634c83a2 KW |
6342 | } |
6343 | } | |
bbce6d69 | 6344 | break; |
f56b6394 | 6345 | case ANYOFV: |
a0d0e21e | 6346 | case ANYOF: |
4e8910e0 KW |
6347 | if (utf8_target || OP(p) == ANYOFV) { |
6348 | STRLEN inclasslen; | |
ffc61ed2 | 6349 | loceol = PL_regeol; |
4e8910e0 KW |
6350 | inclasslen = loceol - scan; |
6351 | while (hardcount < max | |
6352 | && ((inclasslen = loceol - scan) > 0) | |
6353 | && reginclass(prog, p, (U8*)scan, &inclasslen, utf8_target)) | |
6354 | { | |
6355 | scan += inclasslen; | |
ffc61ed2 JH |
6356 | hardcount++; |
6357 | } | |
6358 | } else { | |
32fc9b6a | 6359 | while (scan < loceol && REGINCLASS(prog, p, (U8*)scan)) |
ffc61ed2 JH |
6360 | scan++; |
6361 | } | |
a0d0e21e | 6362 | break; |
980866de | 6363 | case ALNUMU: |
f2ed9b32 | 6364 | if (utf8_target) { |
980866de | 6365 | utf8_wordchar: |
ffc61ed2 | 6366 | loceol = PL_regeol; |
1a4fad37 | 6367 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6368 | while (hardcount < max && scan < loceol && |
a12cf05f KW |
6369 | swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
6370 | { | |
ffc61ed2 JH |
6371 | scan += UTF8SKIP(scan); |
6372 | hardcount++; | |
6373 | } | |
980866de | 6374 | } else { |
a12cf05f KW |
6375 | while (scan < loceol && isWORDCHAR_L1((U8) *scan)) { |
6376 | scan++; | |
6377 | } | |
980866de KW |
6378 | } |
6379 | break; | |
6380 | case ALNUM: | |
6381 | if (utf8_target) | |
6382 | goto utf8_wordchar; | |
6383 | while (scan < loceol && isALNUM((U8) *scan)) { | |
6384 | scan++; | |
a0ed51b3 LW |
6385 | } |
6386 | break; | |
cfaf538b KW |
6387 | case ALNUMA: |
6388 | while (scan < loceol && isWORDCHAR_A((U8) *scan)) { | |
6389 | scan++; | |
6390 | } | |
6391 | break; | |
bbce6d69 | 6392 | case ALNUML: |
3280af22 | 6393 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6394 | if (utf8_target) { |
ffc61ed2 | 6395 | loceol = PL_regeol; |
1aa99e6b IH |
6396 | while (hardcount < max && scan < loceol && |
6397 | isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6398 | scan += UTF8SKIP(scan); |
6399 | hardcount++; | |
6400 | } | |
6401 | } else { | |
6402 | while (scan < loceol && isALNUM_LC(*scan)) | |
6403 | scan++; | |
a0ed51b3 LW |
6404 | } |
6405 | break; | |
980866de | 6406 | case NALNUMU: |
f2ed9b32 | 6407 | if (utf8_target) { |
980866de KW |
6408 | |
6409 | utf8_Nwordchar: | |
6410 | ||
ffc61ed2 | 6411 | loceol = PL_regeol; |
1a4fad37 | 6412 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6413 | while (hardcount < max && scan < loceol && |
980866de | 6414 | ! swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
a12cf05f | 6415 | { |
ffc61ed2 JH |
6416 | scan += UTF8SKIP(scan); |
6417 | hardcount++; | |
6418 | } | |
980866de | 6419 | } else { |
a12cf05f KW |
6420 | while (scan < loceol && ! isWORDCHAR_L1((U8) *scan)) { |
6421 | scan++; | |
6422 | } | |
980866de KW |
6423 | } |
6424 | break; | |
6425 | case NALNUM: | |
6426 | if (utf8_target) | |
6427 | goto utf8_Nwordchar; | |
6428 | while (scan < loceol && ! isALNUM((U8) *scan)) { | |
6429 | scan++; | |
a0ed51b3 LW |
6430 | } |
6431 | break; | |
0658cdde KW |
6432 | |
6433 | case POSIXA: | |
6434 | while (scan < loceol && _generic_isCC_A((U8) *scan, FLAGS(p))) { | |
6435 | scan++; | |
6436 | } | |
6437 | break; | |
6438 | case NPOSIXA: | |
6439 | if (utf8_target) { | |
6440 | while (scan < loceol && ! _generic_isCC_A((U8) *scan, FLAGS(p))) { | |
6441 | scan += UTF8SKIP(scan); | |
6442 | } | |
6443 | } | |
6444 | else { | |
6445 | while (scan < loceol && ! _generic_isCC_A((U8) *scan, FLAGS(p))) { | |
6446 | scan++; | |
6447 | } | |
6448 | } | |
6449 | break; | |
cfaf538b KW |
6450 | case NALNUMA: |
6451 | if (utf8_target) { | |
6452 | while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) { | |
6453 | scan += UTF8SKIP(scan); | |
6454 | } | |
6455 | } | |
6456 | else { | |
6457 | while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) { | |
6458 | scan++; | |
6459 | } | |
6460 | } | |
6461 | break; | |
bbce6d69 | 6462 | case NALNUML: |
3280af22 | 6463 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6464 | if (utf8_target) { |
ffc61ed2 | 6465 | loceol = PL_regeol; |
1aa99e6b IH |
6466 | while (hardcount < max && scan < loceol && |
6467 | !isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6468 | scan += UTF8SKIP(scan); |
6469 | hardcount++; | |
6470 | } | |
6471 | } else { | |
6472 | while (scan < loceol && !isALNUM_LC(*scan)) | |
6473 | scan++; | |
a0ed51b3 LW |
6474 | } |
6475 | break; | |
980866de | 6476 | case SPACEU: |
f2ed9b32 | 6477 | if (utf8_target) { |
980866de KW |
6478 | |
6479 | utf8_space: | |
6480 | ||
ffc61ed2 | 6481 | loceol = PL_regeol; |
1a4fad37 | 6482 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6483 | while (hardcount < max && scan < loceol && |
3568d838 | 6484 | (*scan == ' ' || |
a12cf05f KW |
6485 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) |
6486 | { | |
ffc61ed2 JH |
6487 | scan += UTF8SKIP(scan); |
6488 | hardcount++; | |
6489 | } | |
980866de KW |
6490 | break; |
6491 | } | |
6492 | else { | |
a12cf05f KW |
6493 | while (scan < loceol && isSPACE_L1((U8) *scan)) { |
6494 | scan++; | |
6495 | } | |
980866de KW |
6496 | break; |
6497 | } | |
6498 | case SPACE: | |
6499 | if (utf8_target) | |
6500 | goto utf8_space; | |
6501 | ||
6502 | while (scan < loceol && isSPACE((U8) *scan)) { | |
6503 | scan++; | |
a0ed51b3 LW |
6504 | } |
6505 | break; | |
cfaf538b KW |
6506 | case SPACEA: |
6507 | while (scan < loceol && isSPACE_A((U8) *scan)) { | |
6508 | scan++; | |
6509 | } | |
6510 | break; | |
bbce6d69 | 6511 | case SPACEL: |
3280af22 | 6512 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6513 | if (utf8_target) { |
ffc61ed2 | 6514 | loceol = PL_regeol; |
1aa99e6b | 6515 | while (hardcount < max && scan < loceol && |
6bbba904 | 6516 | isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6517 | scan += UTF8SKIP(scan); |
6518 | hardcount++; | |
6519 | } | |
6520 | } else { | |
6521 | while (scan < loceol && isSPACE_LC(*scan)) | |
6522 | scan++; | |
a0ed51b3 LW |
6523 | } |
6524 | break; | |
980866de | 6525 | case NSPACEU: |
f2ed9b32 | 6526 | if (utf8_target) { |
980866de KW |
6527 | |
6528 | utf8_Nspace: | |
6529 | ||
ffc61ed2 | 6530 | loceol = PL_regeol; |
1a4fad37 | 6531 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6532 | while (hardcount < max && scan < loceol && |
980866de KW |
6533 | ! (*scan == ' ' || |
6534 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) | |
a12cf05f | 6535 | { |
ffc61ed2 JH |
6536 | scan += UTF8SKIP(scan); |
6537 | hardcount++; | |
6538 | } | |
980866de KW |
6539 | break; |
6540 | } | |
6541 | else { | |
a12cf05f KW |
6542 | while (scan < loceol && ! isSPACE_L1((U8) *scan)) { |
6543 | scan++; | |
6544 | } | |
980866de KW |
6545 | } |
6546 | break; | |
6547 | case NSPACE: | |
6548 | if (utf8_target) | |
6549 | goto utf8_Nspace; | |
6550 | ||
6551 | while (scan < loceol && ! isSPACE((U8) *scan)) { | |
6552 | scan++; | |
a0ed51b3 | 6553 | } |
0008a298 | 6554 | break; |
cfaf538b KW |
6555 | case NSPACEA: |
6556 | if (utf8_target) { | |
6557 | while (scan < loceol && ! isSPACE_A((U8) *scan)) { | |
6558 | scan += UTF8SKIP(scan); | |
6559 | } | |
6560 | } | |
6561 | else { | |
6562 | while (scan < loceol && ! isSPACE_A((U8) *scan)) { | |
6563 | scan++; | |
6564 | } | |
6565 | } | |
6566 | break; | |
bbce6d69 | 6567 | case NSPACEL: |
3280af22 | 6568 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6569 | if (utf8_target) { |
ffc61ed2 | 6570 | loceol = PL_regeol; |
1aa99e6b | 6571 | while (hardcount < max && scan < loceol && |
6bbba904 | 6572 | !isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6573 | scan += UTF8SKIP(scan); |
6574 | hardcount++; | |
6575 | } | |
6576 | } else { | |
6577 | while (scan < loceol && !isSPACE_LC(*scan)) | |
6578 | scan++; | |
a0ed51b3 LW |
6579 | } |
6580 | break; | |
a0d0e21e | 6581 | case DIGIT: |
f2ed9b32 | 6582 | if (utf8_target) { |
ffc61ed2 | 6583 | loceol = PL_regeol; |
1a4fad37 | 6584 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6585 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6586 | swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6587 | scan += UTF8SKIP(scan); |
6588 | hardcount++; | |
6589 | } | |
6590 | } else { | |
6591 | while (scan < loceol && isDIGIT(*scan)) | |
6592 | scan++; | |
a0ed51b3 LW |
6593 | } |
6594 | break; | |
cfaf538b KW |
6595 | case DIGITA: |
6596 | while (scan < loceol && isDIGIT_A((U8) *scan)) { | |
6597 | scan++; | |
6598 | } | |
6599 | break; | |
b77393f6 KW |
6600 | case DIGITL: |
6601 | PL_reg_flags |= RF_tainted; | |
6602 | if (utf8_target) { | |
6603 | loceol = PL_regeol; | |
6604 | while (hardcount < max && scan < loceol && | |
6605 | isDIGIT_LC_utf8((U8*)scan)) { | |
6606 | scan += UTF8SKIP(scan); | |
6607 | hardcount++; | |
6608 | } | |
6609 | } else { | |
6610 | while (scan < loceol && isDIGIT_LC(*scan)) | |
6611 | scan++; | |
6612 | } | |
6613 | break; | |
a0d0e21e | 6614 | case NDIGIT: |
f2ed9b32 | 6615 | if (utf8_target) { |
ffc61ed2 | 6616 | loceol = PL_regeol; |
1a4fad37 | 6617 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6618 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6619 | !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6620 | scan += UTF8SKIP(scan); |
6621 | hardcount++; | |
6622 | } | |
6623 | } else { | |
6624 | while (scan < loceol && !isDIGIT(*scan)) | |
6625 | scan++; | |
a0ed51b3 | 6626 | } |
cfaf538b KW |
6627 | break; |
6628 | case NDIGITA: | |
6629 | if (utf8_target) { | |
6630 | while (scan < loceol && ! isDIGIT_A((U8) *scan)) { | |
6631 | scan += UTF8SKIP(scan); | |
6632 | } | |
6633 | } | |
6634 | else { | |
6635 | while (scan < loceol && ! isDIGIT_A((U8) *scan)) { | |
6636 | scan++; | |
6637 | } | |
6638 | } | |
6639 | break; | |
b77393f6 KW |
6640 | case NDIGITL: |
6641 | PL_reg_flags |= RF_tainted; | |
6642 | if (utf8_target) { | |
6643 | loceol = PL_regeol; | |
6644 | while (hardcount < max && scan < loceol && | |
6645 | !isDIGIT_LC_utf8((U8*)scan)) { | |
6646 | scan += UTF8SKIP(scan); | |
6647 | hardcount++; | |
6648 | } | |
6649 | } else { | |
6650 | while (scan < loceol && !isDIGIT_LC(*scan)) | |
6651 | scan++; | |
6652 | } | |
6653 | break; | |
e1d1eefb | 6654 | case LNBREAK: |
f2ed9b32 | 6655 | if (utf8_target) { |
e1d1eefb YO |
6656 | loceol = PL_regeol; |
6657 | while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) { | |
6658 | scan += c; | |
6659 | hardcount++; | |
6660 | } | |
6661 | } else { | |
6662 | /* | |
6663 | LNBREAK can match two latin chars, which is ok, | |
6664 | because we have a null terminated string, but we | |
6665 | have to use hardcount in this situation | |
6666 | */ | |
6667 | while (scan < loceol && (c=is_LNBREAK_latin1(scan))) { | |
6668 | scan+=c; | |
6669 | hardcount++; | |
6670 | } | |
6671 | } | |
6672 | break; | |
6673 | case HORIZWS: | |
f2ed9b32 | 6674 | if (utf8_target) { |
e1d1eefb YO |
6675 | loceol = PL_regeol; |
6676 | while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) { | |
6677 | scan += c; | |
6678 | hardcount++; | |
6679 | } | |
6680 | } else { | |
6681 | while (scan < loceol && is_HORIZWS_latin1(scan)) | |
6682 | scan++; | |
6683 | } | |
a0ed51b3 | 6684 | break; |
e1d1eefb | 6685 | case NHORIZWS: |
f2ed9b32 | 6686 | if (utf8_target) { |
e1d1eefb YO |
6687 | loceol = PL_regeol; |
6688 | while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) { | |
6689 | scan += UTF8SKIP(scan); | |
6690 | hardcount++; | |
6691 | } | |
6692 | } else { | |
6693 | while (scan < loceol && !is_HORIZWS_latin1(scan)) | |
6694 | scan++; | |
6695 | ||
6696 | } | |
6697 | break; | |
6698 | case VERTWS: | |
f2ed9b32 | 6699 | if (utf8_target) { |
e1d1eefb YO |
6700 | loceol = PL_regeol; |
6701 | while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) { | |
6702 | scan += c; | |
6703 | hardcount++; | |
6704 | } | |
6705 | } else { | |
6706 | while (scan < loceol && is_VERTWS_latin1(scan)) | |
6707 | scan++; | |
6708 | ||
6709 | } | |
6710 | break; | |
6711 | case NVERTWS: | |
f2ed9b32 | 6712 | if (utf8_target) { |
e1d1eefb YO |
6713 | loceol = PL_regeol; |
6714 | while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) { | |
6715 | scan += UTF8SKIP(scan); | |
6716 | hardcount++; | |
6717 | } | |
6718 | } else { | |
6719 | while (scan < loceol && !is_VERTWS_latin1(scan)) | |
6720 | scan++; | |
6721 | ||
6722 | } | |
6723 | break; | |
6724 | ||
a0d0e21e LW |
6725 | default: /* Called on something of 0 width. */ |
6726 | break; /* So match right here or not at all. */ | |
6727 | } | |
a687059c | 6728 | |
a0ed51b3 LW |
6729 | if (hardcount) |
6730 | c = hardcount; | |
6731 | else | |
f73aaa43 DM |
6732 | c = scan - *startposp; |
6733 | *startposp = scan; | |
a687059c | 6734 | |
a3621e74 | 6735 | DEBUG_r({ |
e68ec53f | 6736 | GET_RE_DEBUG_FLAGS_DECL; |
be8e71aa | 6737 | DEBUG_EXECUTE_r({ |
e68ec53f YO |
6738 | SV * const prop = sv_newmortal(); |
6739 | regprop(prog, prop, p); | |
6740 | PerlIO_printf(Perl_debug_log, | |
be8e71aa | 6741 | "%*s %s can match %"IVdf" times out of %"IVdf"...\n", |
e2e6a0f1 | 6742 | REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max); |
a3621e74 | 6743 | }); |
be8e71aa | 6744 | }); |
9041c2e3 | 6745 | |
a0d0e21e | 6746 | return(c); |
a687059c LW |
6747 | } |
6748 | ||
c277df42 | 6749 | |
be8e71aa | 6750 | #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) |
c277df42 | 6751 | /* |
6c6525b8 KW |
6752 | - regclass_swash - prepare the utf8 swash. Wraps the shared core version to |
6753 | create a copy so that changes the caller makes won't change the shared one | |
6754 | */ | |
ffc61ed2 | 6755 | SV * |
32fc9b6a | 6756 | Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) |
ffc61ed2 | 6757 | { |
6c6525b8 KW |
6758 | PERL_ARGS_ASSERT_REGCLASS_SWASH; |
6759 | return newSVsv(core_regclass_swash(prog, node, doinit, listsvp, altsvp)); | |
6760 | } | |
6761 | #endif | |
6762 | ||
6763 | STATIC SV * | |
6764 | S_core_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) | |
6765 | { | |
8c9eb58f KW |
6766 | /* Returns the swash for the input 'node' in the regex 'prog'. |
6767 | * If <doinit> is true, will attempt to create the swash if not already | |
6768 | * done. | |
6769 | * If <listsvp> is non-null, will return the swash initialization string in | |
6770 | * it. | |
6771 | * If <altsvp> is non-null, will return the alternates to the regular swash | |
6772 | * in it | |
6773 | * Tied intimately to how regcomp.c sets up the data structure */ | |
6774 | ||
97aff369 | 6775 | dVAR; |
9e55ce06 JH |
6776 | SV *sw = NULL; |
6777 | SV *si = NULL; | |
6778 | SV *alt = NULL; | |
7a6c6baa KW |
6779 | SV* invlist = NULL; |
6780 | ||
f8fc2ecf YO |
6781 | RXi_GET_DECL(prog,progi); |
6782 | const struct reg_data * const data = prog ? progi->data : NULL; | |
ffc61ed2 | 6783 | |
6c6525b8 | 6784 | PERL_ARGS_ASSERT_CORE_REGCLASS_SWASH; |
7918f24d | 6785 | |
ccb2541c KW |
6786 | assert(ANYOF_NONBITMAP(node)); |
6787 | ||
4f639d21 | 6788 | if (data && data->count) { |
a3b680e6 | 6789 | const U32 n = ARG(node); |
ffc61ed2 | 6790 | |
4f639d21 | 6791 | if (data->what[n] == 's') { |
ad64d0ec NC |
6792 | SV * const rv = MUTABLE_SV(data->data[n]); |
6793 | AV * const av = MUTABLE_AV(SvRV(rv)); | |
2d03de9c | 6794 | SV **const ary = AvARRAY(av); |
87367d5f | 6795 | U8 swash_init_flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; |
9041c2e3 | 6796 | |
8c9eb58f | 6797 | si = *ary; /* ary[0] = the string to initialize the swash with */ |
b11f357e | 6798 | |
7a6c6baa KW |
6799 | /* Elements 3 and 4 are either both present or both absent. [3] is |
6800 | * any inversion list generated at compile time; [4] indicates if | |
6801 | * that inversion list has any user-defined properties in it. */ | |
6802 | if (av_len(av) >= 3) { | |
6803 | invlist = ary[3]; | |
83199d38 KW |
6804 | if (SvUV(ary[4])) { |
6805 | swash_init_flags |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY; | |
6806 | } | |
7a6c6baa KW |
6807 | } |
6808 | else { | |
6809 | invlist = NULL; | |
7a6c6baa KW |
6810 | } |
6811 | ||
8c9eb58f KW |
6812 | /* Element [1] is reserved for the set-up swash. If already there, |
6813 | * return it; if not, create it and store it there */ | |
f192cf32 KW |
6814 | if (SvROK(ary[1])) { |
6815 | sw = ary[1]; | |
6816 | } | |
ffc61ed2 | 6817 | else if (si && doinit) { |
7a6c6baa KW |
6818 | |
6819 | sw = _core_swash_init("utf8", /* the utf8 package */ | |
6820 | "", /* nameless */ | |
6821 | si, | |
6822 | 1, /* binary */ | |
6823 | 0, /* not from tr/// */ | |
7a6c6baa | 6824 | invlist, |
83199d38 | 6825 | &swash_init_flags); |
ffc61ed2 JH |
6826 | (void)av_store(av, 1, sw); |
6827 | } | |
8c9eb58f KW |
6828 | |
6829 | /* Element [2] is for any multi-char folds. Note that is a | |
6830 | * fundamentally flawed design, because can't backtrack and try | |
6831 | * again. See [perl #89774] */ | |
f192cf32 KW |
6832 | if (SvTYPE(ary[2]) == SVt_PVAV) { |
6833 | alt = ary[2]; | |
6834 | } | |
ffc61ed2 JH |
6835 | } |
6836 | } | |
6837 | ||
7a6c6baa KW |
6838 | if (listsvp) { |
6839 | SV* matches_string = newSVpvn("", 0); | |
7a6c6baa KW |
6840 | |
6841 | /* Use the swash, if any, which has to have incorporated into it all | |
6842 | * possibilities */ | |
872dd7e0 KW |
6843 | if ((! sw || (invlist = _get_swash_invlist(sw)) == NULL) |
6844 | && (si && si != &PL_sv_undef)) | |
6845 | { | |
7a6c6baa | 6846 | |
872dd7e0 | 6847 | /* If no swash, use the input initialization string, if available */ |
7a6c6baa KW |
6848 | sv_catsv(matches_string, si); |
6849 | } | |
6850 | ||
6851 | /* Add the inversion list to whatever we have. This may have come from | |
6852 | * the swash, or from an input parameter */ | |
6853 | if (invlist) { | |
6854 | sv_catsv(matches_string, _invlist_contents(invlist)); | |
6855 | } | |
6856 | *listsvp = matches_string; | |
6857 | } | |
6858 | ||
9e55ce06 JH |
6859 | if (altsvp) |
6860 | *altsvp = alt; | |
ffc61ed2 JH |
6861 | |
6862 | return sw; | |
6863 | } | |
6864 | ||
6865 | /* | |
ba7b4546 | 6866 | - reginclass - determine if a character falls into a character class |
832705d4 | 6867 | |
6698fab5 KW |
6868 | n is the ANYOF regnode |
6869 | p is the target string | |
6870 | lenp is pointer to the maximum number of bytes of how far to go in p | |
6871 | (This is assumed wthout checking to always be at least the current | |
6872 | character's size) | |
6873 | utf8_target tells whether p is in UTF-8. | |
832705d4 | 6874 | |
4b3cda86 KW |
6875 | Returns true if matched; false otherwise. If lenp is not NULL, on return |
6876 | from a successful match, the value it points to will be updated to how many | |
6877 | bytes in p were matched. If there was no match, the value is undefined, | |
6878 | possibly changed from the input. | |
eba1359e | 6879 | |
d5788240 KW |
6880 | Note that this can be a synthetic start class, a combination of various |
6881 | nodes, so things you think might be mutually exclusive, such as locale, | |
6882 | aren't. It can match both locale and non-locale | |
6883 | ||
bbce6d69 | 6884 | */ |
6885 | ||
76e3520e | 6886 | STATIC bool |
f6ad78d8 | 6887 | 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 | 6888 | { |
27da23d5 | 6889 | dVAR; |
a3b680e6 | 6890 | const char flags = ANYOF_FLAGS(n); |
bbce6d69 | 6891 | bool match = FALSE; |
cc07378b | 6892 | UV c = *p; |
f7ab54c6 | 6893 | STRLEN c_len = 0; |
6698fab5 | 6894 | STRLEN maxlen; |
1aa99e6b | 6895 | |
7918f24d NC |
6896 | PERL_ARGS_ASSERT_REGINCLASS; |
6897 | ||
4b3cda86 | 6898 | /* If c is not already the code point, get it */ |
f2ed9b32 | 6899 | if (utf8_target && !UTF8_IS_INVARIANT(c)) { |
f7ab54c6 | 6900 | c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &c_len, |
6182169b KW |
6901 | (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) |
6902 | | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY); | |
6903 | /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for | |
6904 | * UTF8_ALLOW_FFFF */ | |
f7ab54c6 | 6905 | if (c_len == (STRLEN)-1) |
e8a70c6f | 6906 | Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); |
19f67299 | 6907 | } |
a5a291f5 KW |
6908 | else { |
6909 | c_len = 1; | |
6910 | } | |
bbce6d69 | 6911 | |
a5a291f5 KW |
6912 | /* Use passed in max length, or one character if none passed in or less |
6913 | * than one character. And assume will match just one character. This is | |
6914 | * overwritten later if matched more. */ | |
4b3cda86 | 6915 | if (lenp) { |
a5a291f5 KW |
6916 | maxlen = (*lenp > c_len) ? *lenp : c_len; |
6917 | *lenp = c_len; | |
4b3cda86 KW |
6918 | |
6919 | } | |
6920 | else { | |
a5a291f5 | 6921 | maxlen = c_len; |
4b3cda86 KW |
6922 | } |
6923 | ||
7cdde544 KW |
6924 | /* If this character is potentially in the bitmap, check it */ |
6925 | if (c < 256) { | |
ffc61ed2 JH |
6926 | if (ANYOF_BITMAP_TEST(n, c)) |
6927 | match = TRUE; | |
11454c59 KW |
6928 | else if (flags & ANYOF_NON_UTF8_LATIN1_ALL |
6929 | && ! utf8_target | |
6930 | && ! isASCII(c)) | |
6931 | { | |
6932 | match = TRUE; | |
6933 | } | |
a0ed51b3 | 6934 | |
78969a98 KW |
6935 | else if (flags & ANYOF_LOCALE) { |
6936 | PL_reg_flags |= RF_tainted; | |
6937 | ||
6938 | if ((flags & ANYOF_LOC_NONBITMAP_FOLD) | |
6939 | && ANYOF_BITMAP_TEST(n, PL_fold_locale[c])) | |
6940 | { | |
ffc61ed2 | 6941 | match = TRUE; |
78969a98 | 6942 | } |
040aea3a KW |
6943 | else if (ANYOF_CLASS_TEST_ANY_SET(n) && |
6944 | ((ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) || | |
6945 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) || | |
6946 | (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) || | |
6947 | (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) || | |
6948 | (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) || | |
6949 | (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) || | |
6950 | (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) || | |
6951 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) || | |
6952 | (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) || | |
6953 | (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) || | |
07315176 KW |
6954 | (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII_LC(c)) || |
6955 | (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII_LC(c)) || | |
040aea3a KW |
6956 | (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) || |
6957 | (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) || | |
6958 | (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) || | |
6959 | (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) || | |
6960 | (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) || | |
6961 | (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) || | |
6962 | (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) || | |
6963 | (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) || | |
6964 | (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) || | |
6965 | (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) || | |
6966 | (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) || | |
6967 | (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) || | |
6968 | (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) || | |
6969 | (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) || | |
6970 | (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) || | |
6971 | (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) || | |
67addccf KW |
6972 | (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK_LC(c)) || |
6973 | (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK_LC(c)) | |
040aea3a | 6974 | ) /* How's that for a conditional? */ |
78969a98 | 6975 | ) { |
ffc61ed2 JH |
6976 | match = TRUE; |
6977 | } | |
a0ed51b3 | 6978 | } |
a0ed51b3 LW |
6979 | } |
6980 | ||
7cdde544 | 6981 | /* If the bitmap didn't (or couldn't) match, and something outside the |
de87c4fe KW |
6982 | * bitmap could match, try that. Locale nodes specifiy completely the |
6983 | * behavior of code points in the bit map (otherwise, a utf8 target would | |
c613755a | 6984 | * cause them to be treated as Unicode and not locale), except in |
de87c4fe | 6985 | * the very unlikely event when this node is a synthetic start class, which |
c613755a KW |
6986 | * could be a combination of locale and non-locale nodes. So allow locale |
6987 | * to match for the synthetic start class, which will give a false | |
6988 | * positive that will be resolved when the match is done again as not part | |
6989 | * of the synthetic start class */ | |
ef87b810 | 6990 | if (!match) { |
10ee90d2 KW |
6991 | if (utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256) { |
6992 | match = TRUE; /* Everything above 255 matches */ | |
e051a21d | 6993 | } |
6f8d7d0d KW |
6994 | else if (ANYOF_NONBITMAP(n) |
6995 | && ((flags & ANYOF_NONBITMAP_NON_UTF8) | |
6996 | || (utf8_target | |
6997 | && (c >=256 | |
6998 | || (! (flags & ANYOF_LOCALE)) | |
6999 | || (flags & ANYOF_IS_SYNTHETIC))))) | |
ef87b810 | 7000 | { |
7cdde544 | 7001 | AV *av; |
210e6c47 | 7002 | SV * const sw = core_regclass_swash(prog, n, TRUE, 0, (SV**)&av); |
7cdde544 KW |
7003 | |
7004 | if (sw) { | |
7005 | U8 * utf8_p; | |
7006 | if (utf8_target) { | |
7007 | utf8_p = (U8 *) p; | |
7008 | } else { | |
f56b6394 KW |
7009 | |
7010 | /* Not utf8. Convert as much of the string as available up | |
7011 | * to the limit of how far the (single) character in the | |
7012 | * pattern can possibly match (no need to go further). If | |
7013 | * the node is a straight ANYOF or not folding, it can't | |
7014 | * match more than one. Otherwise, It can match up to how | |
7015 | * far a single char can fold to. Since not utf8, each | |
7016 | * character is a single byte, so the max it can be in | |
7017 | * bytes is the same as the max it can be in characters */ | |
7018 | STRLEN len = (OP(n) == ANYOF | |
7019 | || ! (flags & ANYOF_LOC_NONBITMAP_FOLD)) | |
7020 | ? 1 | |
7021 | : (maxlen < UTF8_MAX_FOLD_CHAR_EXPAND) | |
7022 | ? maxlen | |
7023 | : UTF8_MAX_FOLD_CHAR_EXPAND; | |
7cdde544 KW |
7024 | utf8_p = bytes_to_utf8(p, &len); |
7025 | } | |
f56b6394 | 7026 | |
f7eb7452 | 7027 | if (swash_fetch(sw, utf8_p, TRUE)) |
7cdde544 | 7028 | match = TRUE; |
39065660 | 7029 | else if (flags & ANYOF_LOC_NONBITMAP_FOLD) { |
f56b6394 KW |
7030 | |
7031 | /* Here, we need to test if the fold of the target string | |
8d5d17fa KW |
7032 | * matches. The non-multi char folds have all been moved to |
7033 | * the compilation phase, and the multi-char folds have | |
7034 | * been stored by regcomp into 'av'; we linearly check to | |
7035 | * see if any match the target string (folded). We know | |
7036 | * that the originals were each one character, but we don't | |
7037 | * currently know how many characters/bytes each folded to, | |
7038 | * except we do know that there are small limits imposed by | |
7039 | * Unicode. XXX A performance enhancement would be to have | |
7040 | * regcomp.c store the max number of chars/bytes that are | |
7041 | * in an av entry, as, say the 0th element. Even better | |
7042 | * would be to have a hash of the few characters that can | |
7043 | * start a multi-char fold to the max number of chars of | |
7044 | * those folds. | |
f56b6394 KW |
7045 | * |
7046 | * If there is a match, we will need to advance (if lenp is | |
7047 | * specified) the match pointer in the target string. But | |
7048 | * what we are comparing here isn't that string directly, | |
7049 | * but its fold, whose length may differ from the original. | |
7050 | * As we go along in constructing the fold, therefore, we | |
7051 | * create a map so that we know how many bytes in the | |
7052 | * source to advance given that we have matched a certain | |
7053 | * number of bytes in the fold. This map is stored in | |
6438af90 KW |
7054 | * 'map_fold_len_back'. Let n mean the number of bytes in |
7055 | * the fold of the first character that we are folding. | |
7056 | * Then map_fold_len_back[n] is set to the number of bytes | |
7057 | * in that first character. Similarly let m be the | |
7058 | * corresponding number for the second character to be | |
7059 | * folded. Then map_fold_len_back[n+m] is set to the | |
7060 | * number of bytes occupied by the first two source | |
7061 | * characters. ... */ | |
7062 | U8 map_fold_len_back[UTF8_MAXBYTES_CASE+1] = { 0 }; | |
f56b6394 KW |
7063 | U8 folded[UTF8_MAXBYTES_CASE+1]; |
7064 | STRLEN foldlen = 0; /* num bytes in fold of 1st char */ | |
deba3d96 | 7065 | STRLEN total_foldlen = 0; /* num bytes in fold of all |
6438af90 | 7066 | chars */ |
f56b6394 KW |
7067 | |
7068 | if (OP(n) == ANYOF || maxlen == 1 || ! lenp || ! av) { | |
7069 | ||
7070 | /* Here, only need to fold the first char of the target | |
6438af90 | 7071 | * string. It the source wasn't utf8, is 1 byte long */ |
f56b6394 | 7072 | to_utf8_fold(utf8_p, folded, &foldlen); |
deba3d96 | 7073 | total_foldlen = foldlen; |
6438af90 KW |
7074 | map_fold_len_back[foldlen] = (utf8_target) |
7075 | ? UTF8SKIP(utf8_p) | |
7076 | : 1; | |
f56b6394 KW |
7077 | } |
7078 | else { | |
7079 | ||
7080 | /* Here, need to fold more than the first char. Do so | |
7081 | * up to the limits */ | |
f56b6394 KW |
7082 | U8* source_ptr = utf8_p; /* The source for the fold |
7083 | is the regex target | |
7084 | string */ | |
7085 | U8* folded_ptr = folded; | |
7086 | U8* e = utf8_p + maxlen; /* Can't go beyond last | |
7087 | available byte in the | |
7088 | target string */ | |
6438af90 KW |
7089 | U8 i; |
7090 | for (i = 0; | |
7091 | i < UTF8_MAX_FOLD_CHAR_EXPAND && source_ptr < e; | |
7092 | i++) | |
f56b6394 KW |
7093 | { |
7094 | ||
7095 | /* Fold the next character */ | |
7096 | U8 this_char_folded[UTF8_MAXBYTES_CASE+1]; | |
7097 | STRLEN this_char_foldlen; | |
7098 | to_utf8_fold(source_ptr, | |
7099 | this_char_folded, | |
7100 | &this_char_foldlen); | |
7101 | ||
7102 | /* Bail if it would exceed the byte limit for | |
7103 | * folding a single char. */ | |
7104 | if (this_char_foldlen + folded_ptr - folded > | |
7105 | UTF8_MAXBYTES_CASE) | |
7106 | { | |
7107 | break; | |
7108 | } | |
7109 | ||
6438af90 | 7110 | /* Add the fold of this character */ |
f56b6394 KW |
7111 | Copy(this_char_folded, |
7112 | folded_ptr, | |
7113 | this_char_foldlen, | |
7114 | U8); | |
f56b6394 | 7115 | source_ptr += UTF8SKIP(source_ptr); |
6438af90 | 7116 | folded_ptr += this_char_foldlen; |
deba3d96 | 7117 | total_foldlen = folded_ptr - folded; |
6438af90 KW |
7118 | |
7119 | /* Create map from the number of bytes in the fold | |
7120 | * back to the number of bytes in the source. If | |
7121 | * the source isn't utf8, the byte count is just | |
7122 | * the number of characters so far */ | |
deba3d96 | 7123 | map_fold_len_back[total_foldlen] |
6438af90 KW |
7124 | = (utf8_target) |
7125 | ? source_ptr - utf8_p | |
7126 | : i + 1; | |
f56b6394 KW |
7127 | } |
7128 | *folded_ptr = '\0'; | |
f56b6394 KW |
7129 | } |
7130 | ||
7131 | ||
7132 | /* Do the linear search to see if the fold is in the list | |
8f2655f7 KW |
7133 | * of multi-char folds. */ |
7134 | if (av) { | |
7cdde544 KW |
7135 | I32 i; |
7136 | for (i = 0; i <= av_len(av); i++) { | |
7137 | SV* const sv = *av_fetch(av, i, FALSE); | |
7138 | STRLEN len; | |
7139 | const char * const s = SvPV_const(sv, len); | |
6438af90 | 7140 | |
f9126265 KW |
7141 | if (len <= total_foldlen |
7142 | && memEQ(s, (char*)folded, len) | |
7143 | ||
7144 | /* If 0, means matched a partial char. See | |
7145 | * [perl #90536] */ | |
7146 | && map_fold_len_back[len]) | |
f56b6394 KW |
7147 | { |
7148 | ||
7149 | /* Advance the target string ptr to account for | |
7150 | * this fold, but have to translate from the | |
7151 | * folded length to the corresponding source | |
6438af90 | 7152 | * length. */ |
8f2655f7 KW |
7153 | if (lenp) { |
7154 | *lenp = map_fold_len_back[len]; | |
8f2655f7 | 7155 | } |
7cdde544 KW |
7156 | match = TRUE; |
7157 | break; | |
7158 | } | |
7159 | } | |
7160 | } | |
7cdde544 KW |
7161 | } |
7162 | ||
7163 | /* If we allocated a string above, free it */ | |
7164 | if (! utf8_target) Safefree(utf8_p); | |
7165 | } | |
7166 | } | |
5073ffbd KW |
7167 | |
7168 | if (UNICODE_IS_SUPER(c) | |
7169 | && (flags & ANYOF_WARN_SUPER) | |
7170 | && ckWARN_d(WARN_NON_UNICODE)) | |
7171 | { | |
7172 | Perl_warner(aTHX_ packWARN(WARN_NON_UNICODE), | |
7173 | "Code point 0x%04"UVXf" is not Unicode, all \\p{} matches fail; all \\P{} matches succeed", c); | |
7174 | } | |
7cdde544 KW |
7175 | } |
7176 | ||
f0fdc1c9 KW |
7177 | /* The xor complements the return if to invert: 1^1 = 0, 1^0 = 1 */ |
7178 | return cBOOL(flags & ANYOF_INVERT) ^ match; | |
a0ed51b3 | 7179 | } |
161b471a | 7180 | |
dfe13c55 | 7181 | STATIC U8 * |
0ce71af7 | 7182 | S_reghop3(U8 *s, I32 off, const U8* lim) |
9041c2e3 | 7183 | { |
6af86488 KW |
7184 | /* return the position 'off' UTF-8 characters away from 's', forward if |
7185 | * 'off' >= 0, backwards if negative. But don't go outside of position | |
7186 | * 'lim', which better be < s if off < 0 */ | |
7187 | ||
97aff369 | 7188 | dVAR; |
7918f24d NC |
7189 | |
7190 | PERL_ARGS_ASSERT_REGHOP3; | |
7191 | ||
a0ed51b3 | 7192 | if (off >= 0) { |
1aa99e6b | 7193 | while (off-- && s < lim) { |
ffc61ed2 | 7194 | /* XXX could check well-formedness here */ |
a0ed51b3 | 7195 | s += UTF8SKIP(s); |
ffc61ed2 | 7196 | } |
a0ed51b3 LW |
7197 | } |
7198 | else { | |
1de06328 YO |
7199 | while (off++ && s > lim) { |
7200 | s--; | |
7201 | if (UTF8_IS_CONTINUED(*s)) { | |
7202 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
7203 | s--; | |
a0ed51b3 | 7204 | } |
1de06328 | 7205 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
7206 | } |
7207 | } | |
7208 | return s; | |
7209 | } | |
161b471a | 7210 | |
f9f4320a YO |
7211 | #ifdef XXX_dmq |
7212 | /* there are a bunch of places where we use two reghop3's that should | |
7213 | be replaced with this routine. but since thats not done yet | |
7214 | we ifdef it out - dmq | |
7215 | */ | |
dfe13c55 | 7216 | STATIC U8 * |
1de06328 YO |
7217 | S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim) |
7218 | { | |
7219 | dVAR; | |
7918f24d NC |
7220 | |
7221 | PERL_ARGS_ASSERT_REGHOP4; | |
7222 | ||
1de06328 YO |
7223 | if (off >= 0) { |
7224 | while (off-- && s < rlim) { | |
7225 | /* XXX could check well-formedness here */ | |
7226 | s += UTF8SKIP(s); | |
7227 | } | |
7228 | } | |
7229 | else { | |
7230 | while (off++ && s > llim) { | |
7231 | s--; | |
7232 | if (UTF8_IS_CONTINUED(*s)) { | |
7233 | while (s > llim && UTF8_IS_CONTINUATION(*s)) | |
7234 | s--; | |
7235 | } | |
7236 | /* XXX could check well-formedness here */ | |
7237 | } | |
7238 | } | |
7239 | return s; | |
7240 | } | |
f9f4320a | 7241 | #endif |
1de06328 YO |
7242 | |
7243 | STATIC U8 * | |
0ce71af7 | 7244 | S_reghopmaybe3(U8* s, I32 off, const U8* lim) |
a0ed51b3 | 7245 | { |
97aff369 | 7246 | dVAR; |
7918f24d NC |
7247 | |
7248 | PERL_ARGS_ASSERT_REGHOPMAYBE3; | |
7249 | ||
a0ed51b3 | 7250 | if (off >= 0) { |
1aa99e6b | 7251 | while (off-- && s < lim) { |
ffc61ed2 | 7252 | /* XXX could check well-formedness here */ |
a0ed51b3 | 7253 | s += UTF8SKIP(s); |
ffc61ed2 | 7254 | } |
a0ed51b3 | 7255 | if (off >= 0) |
3dab1dad | 7256 | return NULL; |
a0ed51b3 LW |
7257 | } |
7258 | else { | |
1de06328 YO |
7259 | while (off++ && s > lim) { |
7260 | s--; | |
7261 | if (UTF8_IS_CONTINUED(*s)) { | |
7262 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
7263 | s--; | |
a0ed51b3 | 7264 | } |
1de06328 | 7265 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
7266 | } |
7267 | if (off <= 0) | |
3dab1dad | 7268 | return NULL; |
a0ed51b3 LW |
7269 | } |
7270 | return s; | |
7271 | } | |
51371543 | 7272 | |
51371543 | 7273 | static void |
acfe0abc | 7274 | restore_pos(pTHX_ void *arg) |
51371543 | 7275 | { |
97aff369 | 7276 | dVAR; |
097eb12c | 7277 | regexp * const rex = (regexp *)arg; |
ed301438 | 7278 | if (PL_reg_state.re_state_eval_setup_done) { |
51371543 | 7279 | if (PL_reg_oldsaved) { |
4f639d21 DM |
7280 | rex->subbeg = PL_reg_oldsaved; |
7281 | rex->sublen = PL_reg_oldsavedlen; | |
6502e081 DM |
7282 | rex->suboffset = PL_reg_oldsavedoffset; |
7283 | rex->subcoffset = PL_reg_oldsavedcoffset; | |
f8c7b90f | 7284 | #ifdef PERL_OLD_COPY_ON_WRITE |
4f639d21 | 7285 | rex->saved_copy = PL_nrs; |
ed252734 | 7286 | #endif |
07bc277f | 7287 | RXp_MATCH_COPIED_on(rex); |
51371543 GS |
7288 | } |
7289 | PL_reg_magic->mg_len = PL_reg_oldpos; | |
ed301438 | 7290 | PL_reg_state.re_state_eval_setup_done = FALSE; |
51371543 GS |
7291 | PL_curpm = PL_reg_oldcurpm; |
7292 | } | |
7293 | } | |
33b8afdf JH |
7294 | |
7295 | STATIC void | |
7296 | S_to_utf8_substr(pTHX_ register regexp *prog) | |
7297 | { | |
a1cac82e | 7298 | int i = 1; |
7918f24d NC |
7299 | |
7300 | PERL_ARGS_ASSERT_TO_UTF8_SUBSTR; | |
7301 | ||
a1cac82e NC |
7302 | do { |
7303 | if (prog->substrs->data[i].substr | |
7304 | && !prog->substrs->data[i].utf8_substr) { | |
7305 | SV* const sv = newSVsv(prog->substrs->data[i].substr); | |
7306 | prog->substrs->data[i].utf8_substr = sv; | |
7307 | sv_utf8_upgrade(sv); | |
610460f9 | 7308 | if (SvVALID(prog->substrs->data[i].substr)) { |
cffe132d | 7309 | if (SvTAIL(prog->substrs->data[i].substr)) { |
610460f9 NC |
7310 | /* Trim the trailing \n that fbm_compile added last |
7311 | time. */ | |
7312 | SvCUR_set(sv, SvCUR(sv) - 1); | |
7313 | /* Whilst this makes the SV technically "invalid" (as its | |
7314 | buffer is no longer followed by "\0") when fbm_compile() | |
7315 | adds the "\n" back, a "\0" is restored. */ | |
cffe132d NC |
7316 | fbm_compile(sv, FBMcf_TAIL); |
7317 | } else | |
7318 | fbm_compile(sv, 0); | |
610460f9 | 7319 | } |
a1cac82e NC |
7320 | if (prog->substrs->data[i].substr == prog->check_substr) |
7321 | prog->check_utf8 = sv; | |
7322 | } | |
7323 | } while (i--); | |
33b8afdf JH |
7324 | } |
7325 | ||
7326 | STATIC void | |
7327 | S_to_byte_substr(pTHX_ register regexp *prog) | |
7328 | { | |
97aff369 | 7329 | dVAR; |
a1cac82e | 7330 | int i = 1; |
7918f24d NC |
7331 | |
7332 | PERL_ARGS_ASSERT_TO_BYTE_SUBSTR; | |
7333 | ||
a1cac82e NC |
7334 | do { |
7335 | if (prog->substrs->data[i].utf8_substr | |
7336 | && !prog->substrs->data[i].substr) { | |
7337 | SV* sv = newSVsv(prog->substrs->data[i].utf8_substr); | |
7338 | if (sv_utf8_downgrade(sv, TRUE)) { | |
610460f9 | 7339 | if (SvVALID(prog->substrs->data[i].utf8_substr)) { |
cffe132d | 7340 | if (SvTAIL(prog->substrs->data[i].utf8_substr)) { |
610460f9 NC |
7341 | /* Trim the trailing \n that fbm_compile added last |
7342 | time. */ | |
7343 | SvCUR_set(sv, SvCUR(sv) - 1); | |
cffe132d NC |
7344 | fbm_compile(sv, FBMcf_TAIL); |
7345 | } else | |
7346 | fbm_compile(sv, 0); | |
7347 | } | |
a1cac82e NC |
7348 | } else { |
7349 | SvREFCNT_dec(sv); | |
7350 | sv = &PL_sv_undef; | |
7351 | } | |
7352 | prog->substrs->data[i].substr = sv; | |
7353 | if (prog->substrs->data[i].utf8_substr == prog->check_utf8) | |
7354 | prog->check_substr = sv; | |
33b8afdf | 7355 | } |
a1cac82e | 7356 | } while (i--); |
33b8afdf | 7357 | } |
66610fdd | 7358 | |
a0e786e5 KW |
7359 | /* These constants are for finding GCB=LV and GCB=LVT. These are for the |
7360 | * pre-composed Hangul syllables, which are all in a contiguous block and | |
7361 | * arranged there in such a way so as to facilitate alorithmic determination of | |
7362 | * their characteristics. As such, they don't need a swash, but can be | |
7363 | * determined by simple arithmetic. Almost all are GCB=LVT, but every 28th one | |
7364 | * is a GCB=LV */ | |
7365 | #define SBASE 0xAC00 /* Start of block */ | |
7366 | #define SCount 11172 /* Length of block */ | |
7367 | #define TCount 28 | |
7368 | ||
7369 | #if 0 /* This routine is not currently used */ | |
7370 | PERL_STATIC_INLINE bool | |
7371 | S_is_utf8_X_LV(pTHX_ const U8 *p) | |
7372 | { | |
7373 | /* Unlike most other similarly named routines here, this does not create a | |
7374 | * swash, so swash_fetch() cannot be used on PL_utf8_X_LV. */ | |
7375 | ||
7376 | dVAR; | |
7377 | ||
7378 | UV cp = valid_utf8_to_uvchr(p, NULL); | |
7379 | ||
7380 | PERL_ARGS_ASSERT_IS_UTF8_X_LV; | |
7381 | ||
7382 | /* The earliest Unicode releases did not have these precomposed Hangul | |
7383 | * syllables. Set to point to undef in that case, so will return false on | |
7384 | * every call */ | |
7385 | if (! PL_utf8_X_LV) { /* Set up if this is the first time called */ | |
7386 | PL_utf8_X_LV = swash_init("utf8", "_X_GCB_LV", &PL_sv_undef, 1, 0); | |
7387 | if (_invlist_len(_get_swash_invlist(PL_utf8_X_LV)) == 0) { | |
7388 | SvREFCNT_dec(PL_utf8_X_LV); | |
7389 | PL_utf8_X_LV = &PL_sv_undef; | |
7390 | } | |
7391 | } | |
7392 | ||
7393 | return (PL_utf8_X_LV != &PL_sv_undef | |
7394 | && cp >= SBASE && cp < SBASE + SCount | |
7395 | && (cp - SBASE) % TCount == 0); /* Only every TCount one is LV */ | |
7396 | } | |
7397 | #endif | |
7398 | ||
7399 | PERL_STATIC_INLINE bool | |
7400 | S_is_utf8_X_LVT(pTHX_ const U8 *p) | |
7401 | { | |
7402 | /* Unlike most other similarly named routines here, this does not create a | |
7403 | * swash, so swash_fetch() cannot be used on PL_utf8_X_LVT. */ | |
7404 | ||
7405 | dVAR; | |
7406 | ||
7407 | UV cp = valid_utf8_to_uvchr(p, NULL); | |
7408 | ||
7409 | PERL_ARGS_ASSERT_IS_UTF8_X_LVT; | |
7410 | ||
7411 | /* The earliest Unicode releases did not have these precomposed Hangul | |
7412 | * syllables. Set to point to undef in that case, so will return false on | |
7413 | * every call */ | |
7414 | if (! PL_utf8_X_LVT) { /* Set up if this is the first time called */ | |
7415 | PL_utf8_X_LVT = swash_init("utf8", "_X_GCB_LVT", &PL_sv_undef, 1, 0); | |
7416 | if (_invlist_len(_get_swash_invlist(PL_utf8_X_LVT)) == 0) { | |
7417 | SvREFCNT_dec(PL_utf8_X_LVT); | |
7418 | PL_utf8_X_LVT = &PL_sv_undef; | |
7419 | } | |
7420 | } | |
7421 | ||
7422 | return (PL_utf8_X_LVT != &PL_sv_undef | |
7423 | && cp >= SBASE && cp < SBASE + SCount | |
7424 | && (cp - SBASE) % TCount != 0); /* All but every TCount one is LV */ | |
7425 | } | |
7426 | ||
66610fdd RGS |
7427 | /* |
7428 | * Local variables: | |
7429 | * c-indentation-style: bsd | |
7430 | * c-basic-offset: 4 | |
14d04a33 | 7431 | * indent-tabs-mode: nil |
66610fdd RGS |
7432 | * End: |
7433 | * | |
14d04a33 | 7434 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 7435 | */ |