Commit | Line | Data |
---|---|---|
a0d0e21e LW |
1 | /* regexec.c |
2 | */ | |
3 | ||
4 | /* | |
4ac71550 TC |
5 | * One Ring to rule them all, One Ring to find them |
6 | & | |
7 | * [p.v of _The Lord of the Rings_, opening poem] | |
8 | * [p.50 of _The Lord of the Rings_, I/iii: "The Shadow of the Past"] | |
9 | * [p.254 of _The Lord of the Rings_, II/ii: "The Council of Elrond"] | |
a0d0e21e LW |
10 | */ |
11 | ||
61296642 DM |
12 | /* This file contains functions for executing a regular expression. See |
13 | * also regcomp.c which funnily enough, contains functions for compiling | |
166f8a29 | 14 | * a regular expression. |
e4a054ea DM |
15 | * |
16 | * This file is also copied at build time to ext/re/re_exec.c, where | |
17 | * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT. | |
18 | * This causes the main functions to be compiled under new names and with | |
19 | * debugging support added, which makes "use re 'debug'" work. | |
166f8a29 DM |
20 | */ |
21 | ||
a687059c LW |
22 | /* NOTE: this is derived from Henry Spencer's regexp code, and should not |
23 | * confused with the original package (see point 3 below). Thanks, Henry! | |
24 | */ | |
25 | ||
26 | /* Additional note: this code is very heavily munged from Henry's version | |
27 | * in places. In some spots I've traded clarity for efficiency, so don't | |
28 | * blame Henry for some of the lack of readability. | |
29 | */ | |
30 | ||
e50aee73 AD |
31 | /* The names of the functions have been changed from regcomp and |
32 | * regexec to pregcomp and pregexec in order to avoid conflicts | |
33 | * with the POSIX routines of the same names. | |
34 | */ | |
35 | ||
b9d5759e | 36 | #ifdef PERL_EXT_RE_BUILD |
54df2634 | 37 | #include "re_top.h" |
9041c2e3 | 38 | #endif |
56953603 | 39 | |
a687059c | 40 | /* |
e50aee73 | 41 | * pregcomp and pregexec -- regsub and regerror are not used in perl |
a687059c LW |
42 | * |
43 | * Copyright (c) 1986 by University of Toronto. | |
44 | * Written by Henry Spencer. Not derived from licensed software. | |
45 | * | |
46 | * Permission is granted to anyone to use this software for any | |
47 | * purpose on any computer system, and to redistribute it freely, | |
48 | * subject to the following restrictions: | |
49 | * | |
50 | * 1. The author is not responsible for the consequences of use of | |
51 | * this software, no matter how awful, even if they arise | |
52 | * from defects in it. | |
53 | * | |
54 | * 2. The origin of this software must not be misrepresented, either | |
55 | * by explicit claim or by omission. | |
56 | * | |
57 | * 3. Altered versions must be plainly marked as such, and must not | |
58 | * be misrepresented as being the original software. | |
59 | * | |
60 | **** Alterations to Henry's code are... | |
61 | **** | |
4bb101f2 | 62 | **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, |
1129b882 NC |
63 | **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 |
64 | **** by Larry Wall and others | |
a687059c | 65 | **** |
9ef589d8 LW |
66 | **** You may distribute under the terms of either the GNU General Public |
67 | **** License or the Artistic License, as specified in the README file. | |
a687059c LW |
68 | * |
69 | * Beware that some of this code is subtly aware of the way operator | |
70 | * precedence is structured in regular expressions. Serious changes in | |
71 | * regular-expression syntax might require a total rethink. | |
72 | */ | |
73 | #include "EXTERN.h" | |
864dbfa3 | 74 | #define PERL_IN_REGEXEC_C |
a687059c | 75 | #include "perl.h" |
0f5d15d6 | 76 | |
54df2634 NC |
77 | #ifdef PERL_IN_XSUB_RE |
78 | # include "re_comp.h" | |
79 | #else | |
80 | # include "regcomp.h" | |
81 | #endif | |
a687059c | 82 | |
ef07e810 | 83 | #define RF_tainted 1 /* tainted information used? e.g. locale */ |
c277df42 | 84 | #define RF_warned 2 /* warned about big count? */ |
faec1544 | 85 | |
ab3bbdeb | 86 | #define RF_utf8 8 /* Pattern contains multibyte chars? */ |
a0ed51b3 | 87 | |
f2ed9b32 | 88 | #define UTF_PATTERN ((PL_reg_flags & RF_utf8) != 0) |
ce862d02 IZ |
89 | |
90 | #define RS_init 1 /* eval environment created */ | |
91 | #define RS_set 2 /* replsv value is set */ | |
c277df42 | 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 */ |
1a4fad37 AL |
125 | #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ |
126 | if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END | |
37e2e78e KW |
127 | |
128 | /* Doesn't do an assert to verify that is correct */ | |
129 | #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \ | |
c269f403 | 130 | if (!CAT2(PL_utf8_,class)) { bool throw_away; ENTER; save_re_context(); throw_away = CAT2(is_utf8_,class)((const U8*)" "); LEAVE; } } STMT_END |
37e2e78e | 131 | |
1a4fad37 AL |
132 | #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a") |
133 | #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0") | |
134 | #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ") | |
51371543 | 135 | |
37e2e78e | 136 | #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \ |
d275fa5e CB |
137 | LOAD_UTF8_CHARCLASS(X_begin, " "); \ |
138 | LOAD_UTF8_CHARCLASS(X_non_hangul, "A"); \ | |
37e2e78e KW |
139 | /* These are utf8 constants, and not utf-ebcdic constants, so the \ |
140 | * assert should likely and hopefully fail on an EBCDIC machine */ \ | |
d275fa5e | 141 | LOAD_UTF8_CHARCLASS(X_extend, "\xcc\x80"); /* U+0300 */ \ |
37e2e78e KW |
142 | \ |
143 | /* No asserts are done for these, in case called on an early \ | |
144 | * Unicode version in which they map to nothing */ \ | |
d275fa5e CB |
145 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_prepend);/* U+0E40 "\xe0\xb9\x80" */ \ |
146 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_L); /* U+1100 "\xe1\x84\x80" */ \ | |
147 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV); /* U+AC00 "\xea\xb0\x80" */ \ | |
148 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LVT); /* U+AC01 "\xea\xb0\x81" */ \ | |
149 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV_LVT_V);/* U+AC01 "\xea\xb0\x81" */\ | |
150 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_T); /* U+11A8 "\xe1\x86\xa8" */ \ | |
37e2e78e | 151 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_V) /* U+1160 "\xe1\x85\xa0" */ |
20d0b1e9 | 152 | |
1dcf4a1b | 153 | #define PLACEHOLDER /* Something for the preprocessor to grab onto */ |
d1eb3177 | 154 | |
ee9a90b8 KW |
155 | /* The actual code for CCC_TRY, which uses several variables from the routine |
156 | * it's callable from. It is designed to be the bulk of a case statement. | |
157 | * FUNC is the macro or function to call on non-utf8 targets that indicate if | |
158 | * nextchr matches the class. | |
159 | * UTF8_TEST is the whole test string to use for utf8 targets | |
160 | * LOAD is what to use to test, and if not present to load in the swash for the | |
161 | * class | |
162 | * POS_OR_NEG is either empty or ! to complement the results of FUNC or | |
163 | * UTF8_TEST test. | |
164 | * The logic is: Fail if we're at the end-of-string; otherwise if the target is | |
165 | * utf8 and a variant, load the swash if necessary and test using the utf8 | |
166 | * test. Advance to the next character if test is ok, otherwise fail; If not | |
167 | * utf8 or an invariant under utf8, use the non-utf8 test, and fail if it | |
168 | * fails, or advance to the next character */ | |
169 | ||
170 | #define _CCC_TRY_CODE(POS_OR_NEG, FUNC, UTF8_TEST, CLASS, STR) \ | |
171 | if (locinput >= PL_regeol) { \ | |
172 | sayNO; \ | |
173 | } \ | |
174 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ | |
175 | LOAD_UTF8_CHARCLASS(CLASS, STR); \ | |
176 | if (POS_OR_NEG (UTF8_TEST)) { \ | |
177 | sayNO; \ | |
178 | } \ | |
179 | locinput += PL_utf8skip[nextchr]; \ | |
180 | nextchr = UCHARAT(locinput); \ | |
181 | break; \ | |
182 | } \ | |
183 | if (POS_OR_NEG (FUNC(nextchr))) { \ | |
184 | sayNO; \ | |
185 | } \ | |
186 | nextchr = UCHARAT(++locinput); \ | |
980866de KW |
187 | break; |
188 | ||
ee9a90b8 KW |
189 | /* Handle the non-locale cases for a character class and its complement. It |
190 | * calls _CCC_TRY_CODE with a ! to complement the test for the character class. | |
191 | * This is because that code fails when the test succeeds, so we want to have | |
192 | * the test fail so that the code succeeds. The swash is stored in a | |
193 | * predictable PL_ place */ | |
cfaf538b KW |
194 | #define _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, \ |
195 | CLASS, STR) \ | |
ee9a90b8 KW |
196 | case NAME: \ |
197 | _CCC_TRY_CODE( !, FUNC, \ | |
198 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ | |
199 | (U8*)locinput, TRUE)), \ | |
200 | CLASS, STR) \ | |
201 | case NNAME: \ | |
1dcf4a1b | 202 | _CCC_TRY_CODE( PLACEHOLDER , FUNC, \ |
ee9a90b8 KW |
203 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ |
204 | (U8*)locinput, TRUE)), \ | |
205 | CLASS, STR) \ | |
206 | ||
207 | /* Generate the case statements for both locale and non-locale character | |
208 | * classes in regmatch for classes that don't have special unicode semantics. | |
209 | * Locales don't use an immediate swash, but an intermediary special locale | |
210 | * function that is called on the pointer to the current place in the input | |
211 | * string. That function will resolve to needing the same swash. One might | |
212 | * think that because we don't know what the locale will match, we shouldn't | |
213 | * check with the swash loading function that it loaded properly; ie, that we | |
214 | * should use LOAD_UTF8_CHARCLASS_NO_CHECK for those, but what is passed to the | |
215 | * regular LOAD_UTF8_CHARCLASS is in non-locale terms, and so locale is | |
216 | * irrelevant here */ | |
217 | #define CCC_TRY(NAME, NNAME, FUNC, \ | |
218 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
cfaf538b | 219 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 KW |
220 | CLASS, STR) \ |
221 | case NAMEL: \ | |
222 | PL_reg_flags |= RF_tainted; \ | |
223 | _CCC_TRY_CODE( !, LCFUNC, LCFUNC_utf8((U8*)locinput), CLASS, STR) \ | |
224 | case NNAMEL: \ | |
225 | PL_reg_flags |= RF_tainted; \ | |
1dcf4a1b KW |
226 | _CCC_TRY_CODE( PLACEHOLDER, LCFUNC, LCFUNC_utf8((U8*)locinput), \ |
227 | CLASS, STR) \ | |
cfaf538b KW |
228 | case NAMEA: \ |
229 | if (locinput >= PL_regeol || ! FUNCA(nextchr)) { \ | |
230 | sayNO; \ | |
231 | } \ | |
232 | /* Matched a utf8-invariant, so don't have to worry about utf8 */ \ | |
233 | nextchr = UCHARAT(++locinput); \ | |
234 | break; \ | |
235 | case NNAMEA: \ | |
236 | if (locinput >= PL_regeol || FUNCA(nextchr)) { \ | |
237 | sayNO; \ | |
238 | } \ | |
239 | if (utf8_target) { \ | |
240 | locinput += PL_utf8skip[nextchr]; \ | |
241 | nextchr = UCHARAT(locinput); \ | |
242 | } \ | |
243 | else { \ | |
244 | nextchr = UCHARAT(++locinput); \ | |
245 | } \ | |
246 | break; \ | |
ee9a90b8 KW |
247 | /* Generate the non-locale cases */ \ |
248 | _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, CLASS, STR) | |
249 | ||
250 | /* This is like CCC_TRY, but has an extra set of parameters for generating case | |
251 | * statements to handle separate Unicode semantics nodes */ | |
252 | #define CCC_TRY_U(NAME, NNAME, FUNC, \ | |
253 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
254 | NAMEU, NNAMEU, FUNCU, \ | |
cfaf538b | 255 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 | 256 | CLASS, STR) \ |
cfaf538b KW |
257 | CCC_TRY(NAME, NNAME, FUNC, \ |
258 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
259 | NAMEA, NNAMEA, FUNCA, \ | |
260 | CLASS, STR) \ | |
ee9a90b8 | 261 | _CCC_TRY_NONLOCALE(NAMEU, NNAMEU, FUNCU, CLASS, STR) |
d1eb3177 | 262 | |
3dab1dad YO |
263 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
264 | ||
5f80c4cf | 265 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
c35dcbe2 YO |
266 | /* it would be nice to rework regcomp.sym to generate this stuff. sigh |
267 | * | |
268 | * NOTE that *nothing* that affects backtracking should be in here, specifically | |
269 | * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a | |
270 | * node that is in between two EXACT like nodes when ascertaining what the required | |
271 | * "follow" character is. This should probably be moved to regex compile time | |
272 | * although it may be done at run time beause of the REF possibility - more | |
273 | * investigation required. -- demerphq | |
274 | */ | |
3e901dc0 YO |
275 | #define JUMPABLE(rn) ( \ |
276 | OP(rn) == OPEN || \ | |
277 | (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \ | |
278 | OP(rn) == EVAL || \ | |
cca55fe3 JP |
279 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
280 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
d1c771f5 | 281 | OP(rn) == KEEPS || \ |
3dab1dad | 282 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 | 283 | ) |
ee9b8eae | 284 | #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT) |
e2d8ce26 | 285 | |
ee9b8eae YO |
286 | #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF ) |
287 | ||
288 | #if 0 | |
289 | /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so | |
290 | we don't need this definition. */ | |
291 | #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) | |
2f7f8cb1 | 292 | #define IS_TEXTF(rn) ( (OP(rn)==EXACTFU || OP(rn)==EXACTFA || OP(rn)==EXACTF) || OP(rn)==REFF || OP(rn)==NREFF ) |
ee9b8eae YO |
293 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) |
294 | ||
295 | #else | |
296 | /* ... so we use this as its faster. */ | |
297 | #define IS_TEXT(rn) ( OP(rn)==EXACT ) | |
2f7f8cb1 | 298 | #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn) == EXACTFA) |
ee9b8eae YO |
299 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) |
300 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) | |
301 | ||
302 | #endif | |
e2d8ce26 | 303 | |
a84d97b6 HS |
304 | /* |
305 | Search for mandatory following text node; for lookahead, the text must | |
306 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
307 | */ | |
cca55fe3 | 308 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
309 | while (JUMPABLE(rn)) { \ |
310 | const OPCODE type = OP(rn); \ | |
311 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 312 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 313 | else if (type == PLUS) \ |
cca55fe3 | 314 | rn = NEXTOPER(rn); \ |
3dab1dad | 315 | else if (type == IFMATCH) \ |
a84d97b6 | 316 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 317 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 318 | } \ |
5f80c4cf | 319 | } STMT_END |
74750237 | 320 | |
c476f425 | 321 | |
acfe0abc | 322 | static void restore_pos(pTHX_ void *arg); |
51371543 | 323 | |
620d5b66 NC |
324 | #define REGCP_PAREN_ELEMS 4 |
325 | #define REGCP_OTHER_ELEMS 5 | |
e0fa7e2b | 326 | #define REGCP_FRAME_ELEMS 1 |
620d5b66 NC |
327 | /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and |
328 | * are needed for the regexp context stack bookkeeping. */ | |
329 | ||
76e3520e | 330 | STATIC CHECKPOINT |
cea2e8a9 | 331 | S_regcppush(pTHX_ I32 parenfloor) |
a0d0e21e | 332 | { |
97aff369 | 333 | dVAR; |
a3b680e6 | 334 | const int retval = PL_savestack_ix; |
a3b680e6 | 335 | const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; |
e0fa7e2b NC |
336 | const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS; |
337 | const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT; | |
a0d0e21e | 338 | int p; |
40a82448 | 339 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 340 | |
e49a9654 IH |
341 | if (paren_elems_to_push < 0) |
342 | Perl_croak(aTHX_ "panic: paren_elems_to_push < 0"); | |
343 | ||
e0fa7e2b NC |
344 | if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems) |
345 | Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf | |
5df417d0 JH |
346 | " out of range (%lu-%ld)", |
347 | total_elems, (unsigned long)PL_regsize, (long)parenfloor); | |
e0fa7e2b | 348 | |
620d5b66 | 349 | SSGROW(total_elems + REGCP_FRAME_ELEMS); |
7f69552c | 350 | |
3280af22 | 351 | for (p = PL_regsize; p > parenfloor; p--) { |
b1ce53c5 | 352 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
f0ab9afb NC |
353 | SSPUSHINT(PL_regoffs[p].end); |
354 | SSPUSHINT(PL_regoffs[p].start); | |
3280af22 | 355 | SSPUSHPTR(PL_reg_start_tmp[p]); |
a0d0e21e | 356 | SSPUSHINT(p); |
e7707071 | 357 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
40a82448 | 358 | " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n", |
f0ab9afb | 359 | (UV)p, (IV)PL_regoffs[p].start, |
40a82448 | 360 | (IV)(PL_reg_start_tmp[p] - PL_bostr), |
f0ab9afb | 361 | (IV)PL_regoffs[p].end |
40a82448 | 362 | )); |
a0d0e21e | 363 | } |
b1ce53c5 | 364 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
f0ab9afb | 365 | SSPUSHPTR(PL_regoffs); |
3280af22 NIS |
366 | SSPUSHINT(PL_regsize); |
367 | SSPUSHINT(*PL_reglastparen); | |
a01268b5 | 368 | SSPUSHINT(*PL_reglastcloseparen); |
3280af22 | 369 | SSPUSHPTR(PL_reginput); |
e0fa7e2b | 370 | SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */ |
41123dfd | 371 | |
a0d0e21e LW |
372 | return retval; |
373 | } | |
374 | ||
c277df42 | 375 | /* These are needed since we do not localize EVAL nodes: */ |
ab3bbdeb YO |
376 | #define REGCP_SET(cp) \ |
377 | DEBUG_STATE_r( \ | |
ab3bbdeb | 378 | PerlIO_printf(Perl_debug_log, \ |
e4f74956 | 379 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
ab3bbdeb YO |
380 | (IV)PL_savestack_ix)); \ |
381 | cp = PL_savestack_ix | |
c3464db5 | 382 | |
ab3bbdeb | 383 | #define REGCP_UNWIND(cp) \ |
e4f74956 | 384 | DEBUG_STATE_r( \ |
ab3bbdeb | 385 | if (cp != PL_savestack_ix) \ |
e4f74956 YO |
386 | PerlIO_printf(Perl_debug_log, \ |
387 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ | |
ab3bbdeb YO |
388 | (IV)(cp), (IV)PL_savestack_ix)); \ |
389 | regcpblow(cp) | |
c277df42 | 390 | |
76e3520e | 391 | STATIC char * |
097eb12c | 392 | S_regcppop(pTHX_ const regexp *rex) |
a0d0e21e | 393 | { |
97aff369 | 394 | dVAR; |
e0fa7e2b | 395 | UV i; |
a0d0e21e | 396 | char *input; |
a3621e74 YO |
397 | GET_RE_DEBUG_FLAGS_DECL; |
398 | ||
7918f24d NC |
399 | PERL_ARGS_ASSERT_REGCPPOP; |
400 | ||
b1ce53c5 | 401 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
c6bf6a65 | 402 | i = SSPOPUV; |
e0fa7e2b NC |
403 | assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
404 | i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */ | |
a0d0e21e | 405 | input = (char *) SSPOPPTR; |
a01268b5 | 406 | *PL_reglastcloseparen = SSPOPINT; |
3280af22 NIS |
407 | *PL_reglastparen = SSPOPINT; |
408 | PL_regsize = SSPOPINT; | |
f0ab9afb | 409 | PL_regoffs=(regexp_paren_pair *) SSPOPPTR; |
b1ce53c5 | 410 | |
620d5b66 | 411 | i -= REGCP_OTHER_ELEMS; |
b1ce53c5 | 412 | /* Now restore the parentheses context. */ |
620d5b66 | 413 | for ( ; i > 0; i -= REGCP_PAREN_ELEMS) { |
1df70142 | 414 | I32 tmps; |
097eb12c | 415 | U32 paren = (U32)SSPOPINT; |
3280af22 | 416 | PL_reg_start_tmp[paren] = (char *) SSPOPPTR; |
f0ab9afb | 417 | PL_regoffs[paren].start = SSPOPINT; |
cf93c79d | 418 | tmps = SSPOPINT; |
3280af22 | 419 | if (paren <= *PL_reglastparen) |
f0ab9afb | 420 | PL_regoffs[paren].end = tmps; |
e7707071 | 421 | DEBUG_BUFFERS_r( |
c3464db5 | 422 | PerlIO_printf(Perl_debug_log, |
b900a521 | 423 | " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n", |
f0ab9afb | 424 | (UV)paren, (IV)PL_regoffs[paren].start, |
b900a521 | 425 | (IV)(PL_reg_start_tmp[paren] - PL_bostr), |
f0ab9afb | 426 | (IV)PL_regoffs[paren].end, |
3280af22 | 427 | (paren > *PL_reglastparen ? "(no)" : "")); |
c277df42 | 428 | ); |
a0d0e21e | 429 | } |
e7707071 | 430 | DEBUG_BUFFERS_r( |
bb7a0f54 | 431 | if (*PL_reglastparen + 1 <= rex->nparens) { |
c3464db5 | 432 | PerlIO_printf(Perl_debug_log, |
faccc32b | 433 | " restoring \\%"IVdf"..\\%"IVdf" to undef\n", |
4f639d21 | 434 | (IV)(*PL_reglastparen + 1), (IV)rex->nparens); |
c277df42 IZ |
435 | } |
436 | ); | |
daf18116 | 437 | #if 1 |
dafc8851 JH |
438 | /* It would seem that the similar code in regtry() |
439 | * already takes care of this, and in fact it is in | |
440 | * a better location to since this code can #if 0-ed out | |
441 | * but the code in regtry() is needed or otherwise tests | |
442 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
443 | * (as of patchlevel 7877) will fail. Then again, |
444 | * this code seems to be necessary or otherwise | |
225593e1 DM |
445 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ |
446 | * --jhi updated by dapm */ | |
3b6647e0 | 447 | for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) { |
097eb12c | 448 | if (i > PL_regsize) |
f0ab9afb NC |
449 | PL_regoffs[i].start = -1; |
450 | PL_regoffs[i].end = -1; | |
a0d0e21e | 451 | } |
dafc8851 | 452 | #endif |
a0d0e21e LW |
453 | return input; |
454 | } | |
455 | ||
02db2b7b | 456 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 457 | |
a687059c | 458 | /* |
e50aee73 | 459 | * pregexec and friends |
a687059c LW |
460 | */ |
461 | ||
76234dfb | 462 | #ifndef PERL_IN_XSUB_RE |
a687059c | 463 | /* |
c277df42 | 464 | - pregexec - match a regexp against a string |
a687059c | 465 | */ |
c277df42 | 466 | I32 |
49d7dfbc | 467 | Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend, |
c3464db5 | 468 | char *strbeg, I32 minend, SV *screamer, U32 nosave) |
c277df42 IZ |
469 | /* strend: pointer to null at end of string */ |
470 | /* strbeg: real beginning of string */ | |
471 | /* minend: end of match must be >=minend after stringarg. */ | |
472 | /* nosave: For optimizations. */ | |
473 | { | |
7918f24d NC |
474 | PERL_ARGS_ASSERT_PREGEXEC; |
475 | ||
c277df42 | 476 | return |
9041c2e3 | 477 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
478 | nosave ? 0 : REXEC_COPY_STR); |
479 | } | |
76234dfb | 480 | #endif |
22e551b9 | 481 | |
9041c2e3 | 482 | /* |
cad2e5aa JH |
483 | * Need to implement the following flags for reg_anch: |
484 | * | |
485 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
486 | * USE_INTUIT_ML | |
487 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
488 | * INTUIT_AUTORITATIVE_ML | |
489 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
490 | * INTUIT_ONCE_ML | |
491 | * | |
492 | * Another flag for this function: SECOND_TIME (so that float substrs | |
493 | * with giant delta may be not rechecked). | |
494 | */ | |
495 | ||
496 | /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ | |
497 | ||
3f7c398e | 498 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
499 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
500 | ||
501 | /* XXXX We assume that strpos is strbeg unless sv. */ | |
502 | ||
6eb5f6b9 JH |
503 | /* XXXX Some places assume that there is a fixed substring. |
504 | An update may be needed if optimizer marks as "INTUITable" | |
505 | RExen without fixed substrings. Similarly, it is assumed that | |
506 | lengths of all the strings are no more than minlen, thus they | |
507 | cannot come from lookahead. | |
40d049e4 YO |
508 | (Or minlen should take into account lookahead.) |
509 | NOTE: Some of this comment is not correct. minlen does now take account | |
510 | of lookahead/behind. Further research is required. -- demerphq | |
511 | ||
512 | */ | |
6eb5f6b9 | 513 | |
2c2d71f5 JH |
514 | /* A failure to find a constant substring means that there is no need to make |
515 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
516 | finding a substring too deep into the string means that less calls to | |
30944b6d IZ |
517 | regtry() should be needed. |
518 | ||
519 | REx compiler's optimizer found 4 possible hints: | |
520 | a) Anchored substring; | |
521 | b) Fixed substring; | |
522 | c) Whether we are anchored (beginning-of-line or \G); | |
486ec47a | 523 | d) First node (of those at offset 0) which may distinguish positions; |
6eb5f6b9 | 524 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
525 | string which does not contradict any of them. |
526 | */ | |
2c2d71f5 | 527 | |
6eb5f6b9 JH |
528 | /* Most of decisions we do here should have been done at compile time. |
529 | The nodes of the REx which we used for the search should have been | |
530 | deleted from the finite automaton. */ | |
531 | ||
cad2e5aa | 532 | char * |
288b8c02 | 533 | Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, |
9f61653a | 534 | char *strend, const U32 flags, re_scream_pos_data *data) |
cad2e5aa | 535 | { |
97aff369 | 536 | dVAR; |
288b8c02 | 537 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
b7953727 | 538 | register I32 start_shift = 0; |
cad2e5aa | 539 | /* Should be nonnegative! */ |
b7953727 | 540 | register I32 end_shift = 0; |
2c2d71f5 JH |
541 | register char *s; |
542 | register SV *check; | |
a1933d95 | 543 | char *strbeg; |
cad2e5aa | 544 | char *t; |
f2ed9b32 | 545 | const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 546 | I32 ml_anch; |
bd61b366 SS |
547 | register char *other_last = NULL; /* other substr checked before this */ |
548 | char *check_at = NULL; /* check substr found at this pos */ | |
bbe252da | 549 | const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; |
f8fc2ecf | 550 | RXi_GET_DECL(prog,progi); |
30944b6d | 551 | #ifdef DEBUGGING |
890ce7af | 552 | const char * const i_strpos = strpos; |
30944b6d | 553 | #endif |
a3621e74 YO |
554 | GET_RE_DEBUG_FLAGS_DECL; |
555 | ||
7918f24d NC |
556 | PERL_ARGS_ASSERT_RE_INTUIT_START; |
557 | ||
f2ed9b32 | 558 | RX_MATCH_UTF8_set(rx,utf8_target); |
cad2e5aa | 559 | |
3c8556c3 | 560 | if (RX_UTF8(rx)) { |
b8d68ded JH |
561 | PL_reg_flags |= RF_utf8; |
562 | } | |
ab3bbdeb | 563 | DEBUG_EXECUTE_r( |
f2ed9b32 | 564 | debug_start_match(rx, utf8_target, strpos, strend, |
1de06328 YO |
565 | sv ? "Guessing start of match in sv for" |
566 | : "Guessing start of match in string for"); | |
2a782b5b | 567 | ); |
cad2e5aa | 568 | |
c344f387 JH |
569 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
570 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 571 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 572 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 573 | goto fail; |
2c2d71f5 | 574 | } |
1de06328 | 575 | |
a1933d95 | 576 | strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; |
1aa99e6b | 577 | PL_regeol = strend; |
f2ed9b32 | 578 | if (utf8_target) { |
33b8afdf JH |
579 | if (!prog->check_utf8 && prog->check_substr) |
580 | to_utf8_substr(prog); | |
581 | check = prog->check_utf8; | |
582 | } else { | |
583 | if (!prog->check_substr && prog->check_utf8) | |
584 | to_byte_substr(prog); | |
585 | check = prog->check_substr; | |
586 | } | |
1de06328 | 587 | if (check == &PL_sv_undef) { |
a3621e74 | 588 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
1de06328 | 589 | "Non-utf8 string cannot match utf8 check string\n")); |
33b8afdf JH |
590 | goto fail; |
591 | } | |
bbe252da YO |
592 | if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */ |
593 | ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) | |
594 | || ( (prog->extflags & RXf_ANCH_BOL) | |
7fba1cd6 | 595 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 596 | |
7e25d62c | 597 | if (!ml_anch) { |
bbe252da YO |
598 | if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */ |
599 | && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ | |
3f7c398e | 600 | /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ |
7e25d62c JH |
601 | && sv && !SvROK(sv) |
602 | && (strpos != strbeg)) { | |
a3621e74 | 603 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
604 | goto fail; |
605 | } | |
606 | if (prog->check_offset_min == prog->check_offset_max && | |
bbe252da | 607 | !(prog->extflags & RXf_CANY_SEEN)) { |
2c2d71f5 | 608 | /* Substring at constant offset from beg-of-str... */ |
cad2e5aa JH |
609 | I32 slen; |
610 | ||
1aa99e6b | 611 | s = HOP3c(strpos, prog->check_offset_min, strend); |
1de06328 | 612 | |
653099ff GS |
613 | if (SvTAIL(check)) { |
614 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 615 | |
9041c2e3 | 616 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 617 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 618 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 619 | goto fail_finish; |
cad2e5aa JH |
620 | } |
621 | /* Now should match s[0..slen-2] */ | |
622 | slen--; | |
3f7c398e | 623 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 624 | || (slen > 1 |
3f7c398e | 625 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 626 | report_neq: |
a3621e74 | 627 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
628 | goto fail_finish; |
629 | } | |
cad2e5aa | 630 | } |
3f7c398e | 631 | else if (*SvPVX_const(check) != *s |
653099ff | 632 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 633 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 634 | goto report_neq; |
c315bfe8 | 635 | check_at = s; |
2c2d71f5 | 636 | goto success_at_start; |
7e25d62c | 637 | } |
cad2e5aa | 638 | } |
2c2d71f5 | 639 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 640 | s = strpos; |
2c2d71f5 | 641 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
1de06328 YO |
642 | end_shift = prog->check_end_shift; |
643 | ||
2c2d71f5 | 644 | if (!ml_anch) { |
a3b680e6 | 645 | const I32 end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 646 | - (SvTAIL(check) != 0); |
a3b680e6 | 647 | const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
648 | |
649 | if (end_shift < eshift) | |
650 | end_shift = eshift; | |
651 | } | |
cad2e5aa | 652 | } |
2c2d71f5 | 653 | else { /* Can match at random position */ |
cad2e5aa JH |
654 | ml_anch = 0; |
655 | s = strpos; | |
1de06328 YO |
656 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
657 | end_shift = prog->check_end_shift; | |
658 | ||
659 | /* end shift should be non negative here */ | |
cad2e5aa JH |
660 | } |
661 | ||
bcdf7404 | 662 | #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 663 | if (end_shift < 0) |
1de06328 | 664 | Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", |
220fc49f | 665 | (IV)end_shift, RX_PRECOMP(prog)); |
2c2d71f5 JH |
666 | #endif |
667 | ||
2c2d71f5 JH |
668 | restart: |
669 | /* Find a possible match in the region s..strend by looking for | |
670 | the "check" substring in the region corrected by start/end_shift. */ | |
1de06328 YO |
671 | |
672 | { | |
673 | I32 srch_start_shift = start_shift; | |
674 | I32 srch_end_shift = end_shift; | |
675 | if (srch_start_shift < 0 && strbeg - s > srch_start_shift) { | |
676 | srch_end_shift -= ((strbeg - s) - srch_start_shift); | |
677 | srch_start_shift = strbeg - s; | |
678 | } | |
6bda09f9 | 679 | DEBUG_OPTIMISE_MORE_r({ |
1de06328 YO |
680 | PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n", |
681 | (IV)prog->check_offset_min, | |
682 | (IV)srch_start_shift, | |
683 | (IV)srch_end_shift, | |
684 | (IV)prog->check_end_shift); | |
685 | }); | |
686 | ||
cad2e5aa | 687 | if (flags & REXEC_SCREAM) { |
cad2e5aa | 688 | I32 p = -1; /* Internal iterator of scream. */ |
a3b680e6 | 689 | I32 * const pp = data ? data->scream_pos : &p; |
cad2e5aa | 690 | |
2c2d71f5 JH |
691 | if (PL_screamfirst[BmRARE(check)] >= 0 |
692 | || ( BmRARE(check) == '\n' | |
85c508c3 | 693 | && (BmPREVIOUS(check) == SvCUR(check) - 1) |
2c2d71f5 | 694 | && SvTAIL(check) )) |
9041c2e3 | 695 | s = screaminstr(sv, check, |
1de06328 | 696 | srch_start_shift + (s - strbeg), srch_end_shift, pp, 0); |
cad2e5aa | 697 | else |
2c2d71f5 | 698 | goto fail_finish; |
4addbd3b | 699 | /* we may be pointing at the wrong string */ |
07bc277f | 700 | if (s && RXp_MATCH_COPIED(prog)) |
3f7c398e | 701 | s = strbeg + (s - SvPVX_const(sv)); |
cad2e5aa JH |
702 | if (data) |
703 | *data->scream_olds = s; | |
704 | } | |
1de06328 YO |
705 | else { |
706 | U8* start_point; | |
707 | U8* end_point; | |
bbe252da | 708 | if (prog->extflags & RXf_CANY_SEEN) { |
1de06328 YO |
709 | start_point= (U8*)(s + srch_start_shift); |
710 | end_point= (U8*)(strend - srch_end_shift); | |
711 | } else { | |
712 | start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); | |
713 | end_point= HOP3(strend, -srch_end_shift, strbeg); | |
714 | } | |
6bda09f9 | 715 | DEBUG_OPTIMISE_MORE_r({ |
56570a2c | 716 | PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", |
1de06328 | 717 | (int)(end_point - start_point), |
fc8cd66c | 718 | (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), |
1de06328 YO |
719 | start_point); |
720 | }); | |
721 | ||
722 | s = fbm_instr( start_point, end_point, | |
7fba1cd6 | 723 | check, multiline ? FBMrf_MULTILINE : 0); |
1de06328 YO |
724 | } |
725 | } | |
cad2e5aa JH |
726 | /* Update the count-of-usability, remove useless subpatterns, |
727 | unshift s. */ | |
2c2d71f5 | 728 | |
ab3bbdeb | 729 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 730 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
731 | SvPVX_const(check), RE_SV_DUMPLEN(check), 30); |
732 | PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", | |
2c2d71f5 | 733 | (s ? "Found" : "Did not find"), |
f2ed9b32 | 734 | (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) |
ab3bbdeb YO |
735 | ? "anchored" : "floating"), |
736 | quoted, | |
737 | RE_SV_TAIL(check), | |
738 | (s ? " at offset " : "...\n") ); | |
739 | }); | |
2c2d71f5 JH |
740 | |
741 | if (!s) | |
742 | goto fail_finish; | |
2c2d71f5 | 743 | /* Finish the diagnostic message */ |
a3621e74 | 744 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 | 745 | |
1de06328 YO |
746 | /* XXX dmq: first branch is for positive lookbehind... |
747 | Our check string is offset from the beginning of the pattern. | |
748 | So we need to do any stclass tests offset forward from that | |
749 | point. I think. :-( | |
750 | */ | |
751 | ||
752 | ||
753 | ||
754 | check_at=s; | |
755 | ||
756 | ||
2c2d71f5 JH |
757 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. |
758 | Start with the other substr. | |
759 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 760 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
761 | *always* match. Probably should be marked during compile... |
762 | Probably it is right to do no SCREAM here... | |
763 | */ | |
764 | ||
f2ed9b32 | 765 | if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8) |
1de06328 YO |
766 | : (prog->float_substr && prog->anchored_substr)) |
767 | { | |
30944b6d | 768 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
769 | /* XXXX May be hopelessly wrong for UTF... */ |
770 | if (!other_last) | |
6eb5f6b9 | 771 | other_last = strpos; |
f2ed9b32 | 772 | if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
773 | do_other_anchored: |
774 | { | |
890ce7af AL |
775 | char * const last = HOP3c(s, -start_shift, strbeg); |
776 | char *last1, *last2; | |
be8e71aa | 777 | char * const saved_s = s; |
33b8afdf | 778 | SV* must; |
2c2d71f5 | 779 | |
2c2d71f5 JH |
780 | t = s - prog->check_offset_max; |
781 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
f2ed9b32 | 782 | && (!utf8_target |
0ce71af7 | 783 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 784 | && t > strpos))) |
6f207bd3 | 785 | NOOP; |
2c2d71f5 JH |
786 | else |
787 | t = strpos; | |
1aa99e6b | 788 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
789 | if (t < other_last) /* These positions already checked */ |
790 | t = other_last; | |
1aa99e6b | 791 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
792 | if (last < last1) |
793 | last1 = last; | |
1de06328 YO |
794 | /* XXXX It is not documented what units *_offsets are in. |
795 | We assume bytes, but this is clearly wrong. | |
796 | Meaning this code needs to be carefully reviewed for errors. | |
797 | dmq. | |
798 | */ | |
799 | ||
2c2d71f5 | 800 | /* On end-of-str: see comment below. */ |
f2ed9b32 | 801 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; |
33b8afdf JH |
802 | if (must == &PL_sv_undef) { |
803 | s = (char*)NULL; | |
1de06328 | 804 | DEBUG_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
805 | } |
806 | else | |
807 | s = fbm_instr( | |
808 | (unsigned char*)t, | |
809 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
810 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
811 | must, | |
7fba1cd6 | 812 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 813 | ); |
ab3bbdeb | 814 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 815 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
816 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
817 | PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", | |
2c2d71f5 | 818 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
819 | quoted, RE_SV_TAIL(must)); |
820 | }); | |
821 | ||
822 | ||
2c2d71f5 JH |
823 | if (!s) { |
824 | if (last1 >= last2) { | |
a3621e74 | 825 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
826 | ", giving up...\n")); |
827 | goto fail_finish; | |
828 | } | |
a3621e74 | 829 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 830 | ", trying floating at offset %ld...\n", |
be8e71aa | 831 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
832 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
833 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
834 | goto restart; |
835 | } | |
836 | else { | |
a3621e74 | 837 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 838 | (long)(s - i_strpos))); |
1aa99e6b IH |
839 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
840 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 841 | s = saved_s; |
2c2d71f5 JH |
842 | if (t == strpos) |
843 | goto try_at_start; | |
2c2d71f5 JH |
844 | goto try_at_offset; |
845 | } | |
30944b6d | 846 | } |
2c2d71f5 JH |
847 | } |
848 | else { /* Take into account the floating substring. */ | |
33b8afdf | 849 | char *last, *last1; |
be8e71aa | 850 | char * const saved_s = s; |
33b8afdf JH |
851 | SV* must; |
852 | ||
853 | t = HOP3c(s, -start_shift, strbeg); | |
854 | last1 = last = | |
855 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
856 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
857 | last = HOP3c(t, prog->float_max_offset, strend); | |
858 | s = HOP3c(t, prog->float_min_offset, strend); | |
859 | if (s < other_last) | |
860 | s = other_last; | |
2c2d71f5 | 861 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
f2ed9b32 | 862 | must = utf8_target ? prog->float_utf8 : prog->float_substr; |
33b8afdf JH |
863 | /* fbm_instr() takes into account exact value of end-of-str |
864 | if the check is SvTAIL(ed). Since false positives are OK, | |
865 | and end-of-str is not later than strend we are OK. */ | |
866 | if (must == &PL_sv_undef) { | |
867 | s = (char*)NULL; | |
1de06328 | 868 | DEBUG_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
869 | } |
870 | else | |
2c2d71f5 | 871 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
872 | (unsigned char*)last + SvCUR(must) |
873 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 874 | must, multiline ? FBMrf_MULTILINE : 0); |
ab3bbdeb | 875 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 876 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
877 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
878 | PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", | |
33b8afdf | 879 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
880 | quoted, RE_SV_TAIL(must)); |
881 | }); | |
33b8afdf JH |
882 | if (!s) { |
883 | if (last1 == last) { | |
a3621e74 | 884 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
885 | ", giving up...\n")); |
886 | goto fail_finish; | |
2c2d71f5 | 887 | } |
a3621e74 | 888 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 889 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 890 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
891 | other_last = last; |
892 | s = HOP3c(t, 1, strend); | |
893 | goto restart; | |
894 | } | |
895 | else { | |
a3621e74 | 896 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
897 | (long)(s - i_strpos))); |
898 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 899 | s = saved_s; |
33b8afdf JH |
900 | if (t == strpos) |
901 | goto try_at_start; | |
902 | goto try_at_offset; | |
903 | } | |
2c2d71f5 | 904 | } |
cad2e5aa | 905 | } |
2c2d71f5 | 906 | |
1de06328 | 907 | |
9ef43ace | 908 | t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); |
1de06328 | 909 | |
6bda09f9 | 910 | DEBUG_OPTIMISE_MORE_r( |
1de06328 YO |
911 | PerlIO_printf(Perl_debug_log, |
912 | "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", | |
913 | (IV)prog->check_offset_min, | |
914 | (IV)prog->check_offset_max, | |
915 | (IV)(s-strpos), | |
916 | (IV)(t-strpos), | |
917 | (IV)(t-s), | |
918 | (IV)(strend-strpos) | |
919 | ) | |
920 | ); | |
921 | ||
2c2d71f5 | 922 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
f2ed9b32 | 923 | && (!utf8_target |
9ef43ace | 924 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) |
1de06328 YO |
925 | && t > strpos))) |
926 | { | |
2c2d71f5 JH |
927 | /* Fixed substring is found far enough so that the match |
928 | cannot start at strpos. */ | |
929 | try_at_offset: | |
cad2e5aa | 930 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
931 | /* Eventually fbm_*() should handle this, but often |
932 | anchored_offset is not 0, so this check will not be wasted. */ | |
933 | /* XXXX In the code below we prefer to look for "^" even in | |
934 | presence of anchored substrings. And we search even | |
935 | beyond the found float position. These pessimizations | |
936 | are historical artefacts only. */ | |
937 | find_anchor: | |
2c2d71f5 | 938 | while (t < strend - prog->minlen) { |
cad2e5aa | 939 | if (*t == '\n') { |
4ee3650e | 940 | if (t < check_at - prog->check_offset_min) { |
f2ed9b32 | 941 | if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
942 | /* Since we moved from the found position, |
943 | we definitely contradict the found anchored | |
30944b6d IZ |
944 | substr. Due to the above check we do not |
945 | contradict "check" substr. | |
946 | Thus we can arrive here only if check substr | |
947 | is float. Redo checking for "other"=="fixed". | |
948 | */ | |
9041c2e3 | 949 | strpos = t + 1; |
a3621e74 | 950 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 951 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
952 | goto do_other_anchored; |
953 | } | |
4ee3650e GS |
954 | /* We don't contradict the found floating substring. */ |
955 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 956 | s = t + 1; |
a3621e74 | 957 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 958 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
959 | goto set_useful; |
960 | } | |
4ee3650e GS |
961 | /* Position contradicts check-string */ |
962 | /* XXXX probably better to look for check-string | |
963 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 964 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 965 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 966 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
967 | goto restart; |
968 | } | |
969 | t++; | |
970 | } | |
a3621e74 | 971 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 972 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 973 | goto fail_finish; |
cad2e5aa | 974 | } |
f5952150 | 975 | else { |
a3621e74 | 976 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 977 | PL_colors[0], PL_colors[1])); |
f5952150 | 978 | } |
cad2e5aa JH |
979 | s = t; |
980 | set_useful: | |
f2ed9b32 | 981 | ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
982 | } |
983 | else { | |
f5952150 | 984 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 985 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
986 | unless it was an MBOL and we are not after MBOL, |
987 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
988 | try_at_start: |
989 | /* Even in this situation we may use MBOL flag if strpos is offset | |
990 | wrt the start of the string. */ | |
05b4157f | 991 | if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ |
a1933d95 | 992 | && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d | 993 | /* May be due to an implicit anchor of m{.*foo} */ |
bbe252da | 994 | && !(prog->intflags & PREGf_IMPLICIT)) |
d506a20d | 995 | { |
cad2e5aa JH |
996 | t = strpos; |
997 | goto find_anchor; | |
998 | } | |
a3621e74 | 999 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 1000 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
70685ca0 | 1001 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 1002 | ); |
2c2d71f5 | 1003 | success_at_start: |
bbe252da | 1004 | if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ |
f2ed9b32 | 1005 | && (utf8_target ? ( |
33b8afdf JH |
1006 | prog->check_utf8 /* Could be deleted already */ |
1007 | && --BmUSEFUL(prog->check_utf8) < 0 | |
1008 | && (prog->check_utf8 == prog->float_utf8) | |
1009 | ) : ( | |
1010 | prog->check_substr /* Could be deleted already */ | |
1011 | && --BmUSEFUL(prog->check_substr) < 0 | |
1012 | && (prog->check_substr == prog->float_substr) | |
1013 | ))) | |
66e933ab | 1014 | { |
cad2e5aa | 1015 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 1016 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
f2ed9b32 KW |
1017 | /* XXX Does the destruction order has to change with utf8_target? */ |
1018 | SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr); | |
1019 | SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
1020 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
1021 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
1022 | check = NULL; /* abort */ | |
cad2e5aa | 1023 | s = strpos; |
486ec47a | 1024 | /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag |
c9415951 YO |
1025 | see http://bugs.activestate.com/show_bug.cgi?id=87173 */ |
1026 | if (prog->intflags & PREGf_IMPLICIT) | |
1027 | prog->extflags &= ~RXf_ANCH_MBOL; | |
3cf5c195 IZ |
1028 | /* XXXX This is a remnant of the old implementation. It |
1029 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 1030 | other heuristics. */ |
bbe252da | 1031 | prog->extflags &= ~RXf_USE_INTUIT; |
c9415951 | 1032 | /* XXXX What other flags might need to be cleared in this branch? */ |
cad2e5aa JH |
1033 | } |
1034 | else | |
1035 | s = strpos; | |
1036 | } | |
1037 | ||
6eb5f6b9 JH |
1038 | /* Last resort... */ |
1039 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
1de06328 YO |
1040 | /* trie stclasses are too expensive to use here, we are better off to |
1041 | leave it to regmatch itself */ | |
f8fc2ecf | 1042 | if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { |
6eb5f6b9 JH |
1043 | /* minlen == 0 is possible if regstclass is \b or \B, |
1044 | and the fixed substr is ''$. | |
1045 | Since minlen is already taken into account, s+1 is before strend; | |
1046 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
1047 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
1048 | regstclass does not come from lookahead... */ | |
1049 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
af944926 | 1050 | This leaves EXACTF-ish only, which are dealt with in find_byclass(). */ |
f8fc2ecf YO |
1051 | const U8* const str = (U8*)STRING(progi->regstclass); |
1052 | const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT | |
1053 | ? CHR_DIST(str+STR_LEN(progi->regstclass), str) | |
66e933ab | 1054 | : 1); |
1de06328 YO |
1055 | char * endpos; |
1056 | if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) | |
1057 | endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); | |
1058 | else if (prog->float_substr || prog->float_utf8) | |
1059 | endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); | |
1060 | else | |
1061 | endpos= strend; | |
1062 | ||
70685ca0 JH |
1063 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n", |
1064 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg))); | |
1de06328 | 1065 | |
6eb5f6b9 | 1066 | t = s; |
f8fc2ecf | 1067 | s = find_byclass(prog, progi->regstclass, s, endpos, NULL); |
6eb5f6b9 JH |
1068 | if (!s) { |
1069 | #ifdef DEBUGGING | |
cbbf8932 | 1070 | const char *what = NULL; |
6eb5f6b9 JH |
1071 | #endif |
1072 | if (endpos == strend) { | |
a3621e74 | 1073 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
1074 | "Could not match STCLASS...\n") ); |
1075 | goto fail; | |
1076 | } | |
a3621e74 | 1077 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 1078 | "This position contradicts STCLASS...\n") ); |
bbe252da | 1079 | if ((prog->extflags & RXf_ANCH) && !ml_anch) |
653099ff | 1080 | goto fail; |
6eb5f6b9 | 1081 | /* Contradict one of substrings */ |
33b8afdf | 1082 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 | 1083 | if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) { |
a3621e74 | 1084 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 1085 | hop_and_restart: |
1aa99e6b | 1086 | s = HOP3c(t, 1, strend); |
66e933ab GS |
1087 | if (s + start_shift + end_shift > strend) { |
1088 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 1089 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
1090 | "Could not match STCLASS...\n") ); |
1091 | goto fail; | |
1092 | } | |
5e39e1e5 HS |
1093 | if (!check) |
1094 | goto giveup; | |
a3621e74 | 1095 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1096 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1097 | what, (long)(s + start_shift - i_strpos)) ); |
1098 | goto restart; | |
1099 | } | |
66e933ab | 1100 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
1101 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
1102 | goto retry_floating_check; | |
1103 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 1104 | s = check_at; |
5e39e1e5 HS |
1105 | if (!check) |
1106 | goto giveup; | |
a3621e74 | 1107 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1108 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1109 | (long)(other_last - i_strpos)) ); |
1110 | goto do_other_anchored; | |
1111 | } | |
60e71179 GS |
1112 | /* Another way we could have checked stclass at the |
1113 | current position only: */ | |
1114 | if (ml_anch) { | |
1115 | s = t = t + 1; | |
5e39e1e5 HS |
1116 | if (!check) |
1117 | goto giveup; | |
a3621e74 | 1118 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1119 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 1120 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 1121 | goto try_at_offset; |
66e933ab | 1122 | } |
f2ed9b32 | 1123 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 1124 | goto fail; |
486ec47a | 1125 | /* Check is floating substring. */ |
6eb5f6b9 JH |
1126 | retry_floating_check: |
1127 | t = check_at - start_shift; | |
a3621e74 | 1128 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
1129 | goto hop_and_restart; |
1130 | } | |
b7953727 | 1131 | if (t != s) { |
a3621e74 | 1132 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 1133 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
1134 | (long)(t - i_strpos), (long)(s - i_strpos)) |
1135 | ); | |
1136 | } | |
1137 | else { | |
a3621e74 | 1138 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1139 | "Does not contradict STCLASS...\n"); |
1140 | ); | |
1141 | } | |
6eb5f6b9 | 1142 | } |
5e39e1e5 | 1143 | giveup: |
a3621e74 | 1144 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
1145 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
1146 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 1147 | return s; |
2c2d71f5 JH |
1148 | |
1149 | fail_finish: /* Substring not found */ | |
33b8afdf | 1150 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
f2ed9b32 | 1151 | BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ |
cad2e5aa | 1152 | fail: |
a3621e74 | 1153 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 1154 | PL_colors[4], PL_colors[5])); |
bd61b366 | 1155 | return NULL; |
cad2e5aa | 1156 | } |
9661b544 | 1157 | |
a0a388a1 YO |
1158 | #define DECL_TRIE_TYPE(scan) \ |
1159 | const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \ | |
1160 | trie_type = (scan->flags != EXACT) \ | |
f2ed9b32 KW |
1161 | ? (utf8_target ? trie_utf8_fold : (UTF_PATTERN ? trie_latin_utf8_fold : trie_plain)) \ |
1162 | : (utf8_target ? trie_utf8 : trie_plain) | |
3b0527fe | 1163 | |
55eed653 NC |
1164 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ |
1165 | uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ | |
4cadc6a9 YO |
1166 | switch (trie_type) { \ |
1167 | case trie_utf8_fold: \ | |
1168 | if ( foldlen>0 ) { \ | |
0abd0d78 | 1169 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ |
4cadc6a9 YO |
1170 | foldlen -= len; \ |
1171 | uscan += len; \ | |
1172 | len=0; \ | |
1173 | } else { \ | |
0abd0d78 | 1174 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ |
4cadc6a9 YO |
1175 | uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \ |
1176 | foldlen -= UNISKIP( uvc ); \ | |
1177 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1178 | } \ | |
1179 | break; \ | |
a0a388a1 YO |
1180 | case trie_latin_utf8_fold: \ |
1181 | if ( foldlen>0 ) { \ | |
1182 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1183 | foldlen -= len; \ | |
1184 | uscan += len; \ | |
1185 | len=0; \ | |
1186 | } else { \ | |
1187 | len = 1; \ | |
1188 | uvc = to_uni_fold( *(U8*)uc, foldbuf, &foldlen ); \ | |
1189 | foldlen -= UNISKIP( uvc ); \ | |
1190 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1191 | } \ | |
1192 | break; \ | |
4cadc6a9 YO |
1193 | case trie_utf8: \ |
1194 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ | |
1195 | break; \ | |
1196 | case trie_plain: \ | |
1197 | uvc = (UV)*uc; \ | |
1198 | len = 1; \ | |
1199 | } \ | |
4cadc6a9 YO |
1200 | if (uvc < 256) { \ |
1201 | charid = trie->charmap[ uvc ]; \ | |
1202 | } \ | |
1203 | else { \ | |
1204 | charid = 0; \ | |
55eed653 NC |
1205 | if (widecharmap) { \ |
1206 | SV** const svpp = hv_fetch(widecharmap, \ | |
4cadc6a9 YO |
1207 | (char*)&uvc, sizeof(UV), 0); \ |
1208 | if (svpp) \ | |
1209 | charid = (U16)SvIV(*svpp); \ | |
1210 | } \ | |
1211 | } \ | |
1212 | } STMT_END | |
1213 | ||
4cadc6a9 YO |
1214 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ |
1215 | STMT_START { \ | |
1216 | while (s <= e) { \ | |
1217 | if ( (CoNd) \ | |
fac1af77 | 1218 | && (ln == 1 || folder(s, pat_string, ln)) \ |
9a5a5549 | 1219 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1220 | goto got_it; \ |
1221 | s++; \ | |
1222 | } \ | |
1223 | } STMT_END | |
1224 | ||
1225 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
1226 | STMT_START { \ | |
1227 | while (s + (uskip = UTF8SKIP(s)) <= strend) { \ | |
1228 | CoDe \ | |
1229 | s += uskip; \ | |
1230 | } \ | |
1231 | } STMT_END | |
1232 | ||
1233 | #define REXEC_FBC_SCAN(CoDe) \ | |
1234 | STMT_START { \ | |
1235 | while (s < strend) { \ | |
1236 | CoDe \ | |
1237 | s++; \ | |
1238 | } \ | |
1239 | } STMT_END | |
1240 | ||
1241 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1242 | REXEC_FBC_UTF8_SCAN( \ | |
1243 | if (CoNd) { \ | |
24b23f37 | 1244 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1245 | goto got_it; \ |
1246 | else \ | |
1247 | tmp = doevery; \ | |
1248 | } \ | |
1249 | else \ | |
1250 | tmp = 1; \ | |
1251 | ) | |
1252 | ||
1253 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1254 | REXEC_FBC_SCAN( \ | |
1255 | if (CoNd) { \ | |
24b23f37 | 1256 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1257 | goto got_it; \ |
1258 | else \ | |
1259 | tmp = doevery; \ | |
1260 | } \ | |
1261 | else \ | |
1262 | tmp = 1; \ | |
1263 | ) | |
1264 | ||
1265 | #define REXEC_FBC_TRYIT \ | |
24b23f37 | 1266 | if ((!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1267 | goto got_it |
1268 | ||
e1d1eefb | 1269 | #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \ |
f2ed9b32 | 1270 | if (utf8_target) { \ |
e1d1eefb YO |
1271 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1272 | } \ | |
1273 | else { \ | |
1274 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1275 | } |
e1d1eefb | 1276 | |
4cadc6a9 | 1277 | #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ |
f2ed9b32 | 1278 | if (utf8_target) { \ |
4cadc6a9 YO |
1279 | UtFpReLoAd; \ |
1280 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1281 | } \ | |
1282 | else { \ | |
1283 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1284 | } |
4cadc6a9 YO |
1285 | |
1286 | #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ | |
1287 | PL_reg_flags |= RF_tainted; \ | |
f2ed9b32 | 1288 | if (utf8_target) { \ |
4cadc6a9 YO |
1289 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1290 | } \ | |
1291 | else { \ | |
1292 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1293 | } |
4cadc6a9 | 1294 | |
786e8c11 YO |
1295 | #define DUMP_EXEC_POS(li,s,doutf8) \ |
1296 | dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8) | |
1297 | ||
cfaf538b KW |
1298 | |
1299 | #define UTF8_NOLOAD(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ | |
1300 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1301 | tmp = TEST_NON_UTF8(tmp); \ | |
1302 | REXEC_FBC_UTF8_SCAN( \ | |
1303 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1304 | tmp = !tmp; \ | |
1305 | IF_SUCCESS; \ | |
1306 | } \ | |
1307 | else { \ | |
1308 | IF_FAIL; \ | |
1309 | } \ | |
1310 | ); \ | |
1311 | ||
1312 | #define UTF8_LOAD(TeSt1_UtF8, TeSt2_UtF8, IF_SUCCESS, IF_FAIL) \ | |
1313 | if (s == PL_bostr) { \ | |
1314 | tmp = '\n'; \ | |
1315 | } \ | |
1316 | else { \ | |
1317 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); \ | |
1318 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); \ | |
1319 | } \ | |
1320 | tmp = TeSt1_UtF8; \ | |
1321 | LOAD_UTF8_CHARCLASS_ALNUM(); \ | |
1322 | REXEC_FBC_UTF8_SCAN( \ | |
1323 | if (tmp == ! (TeSt2_UtF8)) { \ | |
1324 | tmp = !tmp; \ | |
1325 | IF_SUCCESS; \ | |
1326 | } \ | |
1327 | else { \ | |
1328 | IF_FAIL; \ | |
1329 | } \ | |
1330 | ); \ | |
1331 | ||
63ac0dad KW |
1332 | /* The only difference between the BOUND and NBOUND cases is that |
1333 | * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in | |
1334 | * NBOUND. This is accomplished by passing it in either the if or else clause, | |
1335 | * with the other one being empty */ | |
1336 | #define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1337 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
cfaf538b KW |
1338 | |
1339 | #define FBC_BOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1340 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
63ac0dad KW |
1341 | |
1342 | #define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1343 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b KW |
1344 | |
1345 | #define FBC_NBOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1346 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b | 1347 | |
63ac0dad KW |
1348 | |
1349 | /* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to | |
1350 | * be passed in completely with the variable name being tested, which isn't | |
1351 | * such a clean interface, but this is easier to read than it was before. We | |
1352 | * are looking for the boundary (or non-boundary between a word and non-word | |
1353 | * character. The utf8 and non-utf8 cases have the same logic, but the details | |
1354 | * must be different. Find the "wordness" of the character just prior to this | |
1355 | * one, and compare it with the wordness of this one. If they differ, we have | |
1356 | * a boundary. At the beginning of the string, pretend that the previous | |
1357 | * character was a new-line */ | |
cfaf538b | 1358 | #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ |
63ac0dad | 1359 | if (utf8_target) { \ |
cfaf538b | 1360 | UTF8_CODE \ |
63ac0dad KW |
1361 | } \ |
1362 | else { /* Not utf8 */ \ | |
1363 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1364 | tmp = TEST_NON_UTF8(tmp); \ | |
1365 | REXEC_FBC_SCAN( \ | |
1366 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1367 | tmp = !tmp; \ | |
1368 | IF_SUCCESS; \ | |
1369 | } \ | |
1370 | else { \ | |
1371 | IF_FAIL; \ | |
1372 | } \ | |
1373 | ); \ | |
1374 | } \ | |
1375 | if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) \ | |
1376 | goto got_it; | |
1377 | ||
786e8c11 YO |
1378 | /* We know what class REx starts with. Try to find this position... */ |
1379 | /* if reginfo is NULL, its a dryrun */ | |
1380 | /* annoyingly all the vars in this routine have different names from their counterparts | |
1381 | in regmatch. /grrr */ | |
1382 | ||
3c3eec57 | 1383 | STATIC char * |
07be1b83 | 1384 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
24b23f37 | 1385 | const char *strend, regmatch_info *reginfo) |
a687059c | 1386 | { |
27da23d5 | 1387 | dVAR; |
bbe252da | 1388 | const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; |
fac1af77 KW |
1389 | char *pat_string; /* The pattern's exactish string */ |
1390 | char *pat_end; /* ptr to end char of pat_string */ | |
1391 | re_fold_t folder; /* Function for computing non-utf8 folds */ | |
1392 | const U8 *fold_array; /* array for folding ords < 256 */ | |
d8093b23 | 1393 | STRLEN ln; |
5dab1207 | 1394 | STRLEN lnc; |
078c425b | 1395 | register STRLEN uskip; |
fac1af77 KW |
1396 | U8 c1; |
1397 | U8 c2; | |
6eb5f6b9 JH |
1398 | char *e; |
1399 | register I32 tmp = 1; /* Scratch variable? */ | |
f2ed9b32 | 1400 | register const bool utf8_target = PL_reg_match_utf8; |
453bfd44 | 1401 | UV utf8_fold_flags = 0; |
f8fc2ecf | 1402 | RXi_GET_DECL(prog,progi); |
7918f24d NC |
1403 | |
1404 | PERL_ARGS_ASSERT_FIND_BYCLASS; | |
f8fc2ecf | 1405 | |
6eb5f6b9 JH |
1406 | /* We know what class it must start with. */ |
1407 | switch (OP(c)) { | |
f56b6394 | 1408 | case ANYOFV: |
6eb5f6b9 | 1409 | case ANYOF: |
f56b6394 | 1410 | if (utf8_target || OP(c) == ANYOFV) { |
b1e3e569 KW |
1411 | STRLEN inclasslen = strend - s; |
1412 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1413 | reginclass(prog, c, (U8*)s, &inclasslen, utf8_target)); | |
388cc4de HS |
1414 | } |
1415 | else { | |
1416 | while (s < strend) { | |
1417 | STRLEN skip = 1; | |
1418 | ||
9febde93 | 1419 | if (REGINCLASS(prog, c, (U8*)s)) { |
24b23f37 | 1420 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
388cc4de HS |
1421 | goto got_it; |
1422 | else | |
1423 | tmp = doevery; | |
1424 | } | |
1425 | else | |
1426 | tmp = 1; | |
1427 | s += skip; | |
1428 | } | |
a0d0e21e | 1429 | } |
6eb5f6b9 | 1430 | break; |
f33976b4 | 1431 | case CANY: |
4cadc6a9 | 1432 | REXEC_FBC_SCAN( |
24b23f37 | 1433 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
f33976b4 DB |
1434 | goto got_it; |
1435 | else | |
1436 | tmp = doevery; | |
4cadc6a9 | 1437 | ); |
f33976b4 | 1438 | break; |
2f7f8cb1 KW |
1439 | |
1440 | case EXACTFA: | |
1441 | if (UTF_PATTERN || utf8_target) { | |
1442 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
1443 | goto do_exactf_utf8; | |
1444 | } | |
1445 | fold_array = PL_fold_latin1; /* Latin1 folds are not affected by */ | |
1446 | folder = foldEQ_latin1; /* /a, except the sharp s one which */ | |
1447 | goto do_exactf_non_utf8; /* isn't dealt with by these */ | |
1448 | ||
9a5a5549 | 1449 | case EXACTFU: |
fac1af77 KW |
1450 | if (UTF_PATTERN || utf8_target) { |
1451 | utf8_fold_flags = 0; | |
1452 | goto do_exactf_utf8; | |
1453 | } | |
1454 | fold_array = PL_fold_latin1; | |
1455 | folder = foldEQ_latin1; | |
1456 | /* XXX This uses the full utf8 fold because if the pattern contains | |
1457 | * 'ss' it could match LATIN_SMALL_LETTER SHARP_S in the string. | |
1458 | * There could be a new node type, say EXACTFU_SS, which is | |
1459 | * generated by regcomp only if there is an 'ss', and then every | |
1460 | * other case could goto do_exactf_non_utf8;*/ | |
1461 | goto do_exactf_utf8; | |
1462 | ||
6eb5f6b9 | 1463 | case EXACTF: |
fac1af77 KW |
1464 | if (UTF_PATTERN || utf8_target) { |
1465 | utf8_fold_flags = 0; | |
1466 | goto do_exactf_utf8; | |
1467 | } | |
1468 | fold_array = PL_fold; | |
1469 | folder = foldEQ; | |
1470 | goto do_exactf_non_utf8; | |
1471 | ||
1472 | case EXACTFL: | |
1473 | if (UTF_PATTERN || utf8_target) { | |
17580e7a | 1474 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
fac1af77 KW |
1475 | goto do_exactf_utf8; |
1476 | } | |
1477 | fold_array = PL_fold_locale; | |
1478 | folder = foldEQ_locale; | |
1479 | ||
1480 | /* FALL THROUGH */ | |
1481 | ||
1482 | do_exactf_non_utf8: /* Neither pattern nor string are UTF8 */ | |
1483 | ||
1484 | /* The idea in the non-utf8 EXACTF* cases is to first find the | |
1485 | * first character of the EXACTF* node and then, if necessary, | |
1486 | * case-insensitively compare the full text of the node. c1 is the | |
1487 | * first character. c2 is its fold. This logic will not work for | |
1488 | * Unicode semantics and the german sharp ss, which hence should | |
1489 | * not be compiled into a node that gets here. */ | |
1490 | pat_string = STRING(c); | |
1491 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1492 | ||
1493 | e = HOP3c(strend, -((I32)ln), s); | |
1494 | ||
1495 | if (!reginfo && e < s) { | |
1496 | e = s; /* Due to minlen logic of intuit() */ | |
1497 | } | |
1498 | ||
1499 | c1 = *pat_string; | |
1500 | c2 = fold_array[c1]; | |
1501 | if (c1 == c2) { /* If char and fold are the same */ | |
1502 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); | |
1503 | } | |
1504 | else { | |
1505 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); | |
1506 | } | |
1507 | break; | |
1508 | ||
1509 | do_exactf_utf8: | |
1510 | ||
1511 | /* If one of the operands is in utf8, we can't use the simpler | |
1512 | * folding above, due to the fact that many different characters | |
1513 | * can have the same fold, or portion of a fold, or different- | |
1514 | * length fold */ | |
1515 | pat_string = STRING(c); | |
1516 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1517 | pat_end = pat_string + ln; | |
1518 | lnc = (UTF_PATTERN) /* length to match in characters */ | |
1519 | ? utf8_length((U8 *) pat_string, (U8 *) pat_end) | |
1520 | : ln; | |
1521 | ||
1522 | e = HOP3c(strend, -((I32)lnc), s); | |
1523 | ||
1524 | if (!reginfo && e < s) { | |
1525 | e = s; /* Due to minlen logic of intuit() */ | |
1526 | } | |
1527 | ||
1528 | while (s <= e) { | |
1529 | char *my_strend= (char *)strend; | |
1530 | if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target, | |
1531 | pat_string, NULL, ln, cBOOL(UTF_PATTERN), utf8_fold_flags) | |
1532 | && (!reginfo || regtry(reginfo, &s)) ) | |
1533 | { | |
1534 | goto got_it; | |
1535 | } | |
1536 | s += UTF8SKIP(s); | |
1537 | } | |
1538 | break; | |
bbce6d69 | 1539 | case BOUNDL: |
3280af22 | 1540 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1541 | FBC_BOUND(isALNUM_LC, |
1542 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1543 | isALNUM_LC_utf8((U8*)s)); | |
a0ed51b3 | 1544 | break; |
bbce6d69 | 1545 | case NBOUNDL: |
3280af22 | 1546 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1547 | FBC_NBOUND(isALNUM_LC, |
1548 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1549 | isALNUM_LC_utf8((U8*)s)); | |
1550 | break; | |
1551 | case BOUND: | |
1552 | FBC_BOUND(isWORDCHAR, | |
1553 | isALNUM_uni(tmp), | |
1554 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1555 | break; | |
cfaf538b KW |
1556 | case BOUNDA: |
1557 | FBC_BOUND_NOLOAD(isWORDCHAR_A, | |
1558 | isWORDCHAR_A(tmp), | |
1559 | isWORDCHAR_A((U8*)s)); | |
1560 | break; | |
a0d0e21e | 1561 | case NBOUND: |
63ac0dad KW |
1562 | FBC_NBOUND(isWORDCHAR, |
1563 | isALNUM_uni(tmp), | |
1564 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1565 | break; | |
cfaf538b KW |
1566 | case NBOUNDA: |
1567 | FBC_NBOUND_NOLOAD(isWORDCHAR_A, | |
1568 | isWORDCHAR_A(tmp), | |
1569 | isWORDCHAR_A((U8*)s)); | |
1570 | break; | |
63ac0dad KW |
1571 | case BOUNDU: |
1572 | FBC_BOUND(isWORDCHAR_L1, | |
1573 | isALNUM_uni(tmp), | |
1574 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1575 | break; | |
1576 | case NBOUNDU: | |
1577 | FBC_NBOUND(isWORDCHAR_L1, | |
1578 | isALNUM_uni(tmp), | |
1579 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
a0ed51b3 | 1580 | break; |
bbce6d69 | 1581 | case ALNUML: |
4cadc6a9 YO |
1582 | REXEC_FBC_CSCAN_TAINT( |
1583 | isALNUM_LC_utf8((U8*)s), | |
1584 | isALNUM_LC(*s) | |
1585 | ); | |
6895a8aa | 1586 | break; |
980866de KW |
1587 | case ALNUMU: |
1588 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1589 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1590 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1591 | isWORDCHAR_L1((U8) *s) |
1592 | ); | |
6895a8aa | 1593 | break; |
980866de KW |
1594 | case ALNUM: |
1595 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1596 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1597 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1598 | isWORDCHAR((U8) *s) |
1599 | ); | |
6895a8aa | 1600 | break; |
cfaf538b | 1601 | case ALNUMA: |
8e9da4d4 KW |
1602 | /* Don't need to worry about utf8, as it can match only a single |
1603 | * byte invariant character */ | |
cfaf538b | 1604 | REXEC_FBC_CLASS_SCAN( isWORDCHAR_A(*s)); |
6895a8aa | 1605 | break; |
980866de KW |
1606 | case NALNUMU: |
1607 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1608 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1609 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1610 | ! isWORDCHAR_L1((U8) *s) |
1611 | ); | |
6895a8aa | 1612 | break; |
a0d0e21e | 1613 | case NALNUM: |
4cadc6a9 | 1614 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1615 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1616 | !swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target), | |
980866de | 1617 | ! isALNUM(*s) |
4cadc6a9 | 1618 | ); |
6895a8aa | 1619 | break; |
cfaf538b | 1620 | case NALNUMA: |
8e9da4d4 KW |
1621 | REXEC_FBC_CSCAN( |
1622 | !isWORDCHAR_A(*s), | |
1623 | !isWORDCHAR_A(*s) | |
1624 | ); | |
1625 | break; | |
bbce6d69 | 1626 | case NALNUML: |
4cadc6a9 YO |
1627 | REXEC_FBC_CSCAN_TAINT( |
1628 | !isALNUM_LC_utf8((U8*)s), | |
1629 | !isALNUM_LC(*s) | |
1630 | ); | |
6895a8aa | 1631 | break; |
980866de KW |
1632 | case SPACEU: |
1633 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1634 | LOAD_UTF8_CHARCLASS_SPACE(), |
1635 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de KW |
1636 | isSPACE_L1((U8) *s) |
1637 | ); | |
6895a8aa | 1638 | break; |
a0d0e21e | 1639 | case SPACE: |
4cadc6a9 | 1640 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1641 | LOAD_UTF8_CHARCLASS_SPACE(), |
1642 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de | 1643 | isSPACE((U8) *s) |
4cadc6a9 | 1644 | ); |
6895a8aa | 1645 | break; |
cfaf538b | 1646 | case SPACEA: |
8e9da4d4 KW |
1647 | /* Don't need to worry about utf8, as it can match only a single |
1648 | * byte invariant character */ | |
cfaf538b | 1649 | REXEC_FBC_CLASS_SCAN( isSPACE_A(*s)); |
6895a8aa | 1650 | break; |
bbce6d69 | 1651 | case SPACEL: |
4cadc6a9 | 1652 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1653 | isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1654 | isSPACE_LC(*s) |
1655 | ); | |
6895a8aa | 1656 | break; |
980866de KW |
1657 | case NSPACEU: |
1658 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1659 | LOAD_UTF8_CHARCLASS_SPACE(), |
1660 | !( *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de KW |
1661 | ! isSPACE_L1((U8) *s) |
1662 | ); | |
6895a8aa | 1663 | break; |
a0d0e21e | 1664 | case NSPACE: |
4cadc6a9 | 1665 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1666 | LOAD_UTF8_CHARCLASS_SPACE(), |
1667 | !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de | 1668 | ! isSPACE((U8) *s) |
4cadc6a9 | 1669 | ); |
6895a8aa | 1670 | break; |
cfaf538b | 1671 | case NSPACEA: |
8e9da4d4 KW |
1672 | REXEC_FBC_CSCAN( |
1673 | !isSPACE_A(*s), | |
1674 | !isSPACE_A(*s) | |
1675 | ); | |
1676 | break; | |
bbce6d69 | 1677 | case NSPACEL: |
4cadc6a9 | 1678 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1679 | !isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1680 | !isSPACE_LC(*s) |
1681 | ); | |
6895a8aa | 1682 | break; |
a0d0e21e | 1683 | case DIGIT: |
4cadc6a9 | 1684 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1685 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1686 | swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1687 | isDIGIT(*s) |
1688 | ); | |
6895a8aa | 1689 | break; |
cfaf538b | 1690 | case DIGITA: |
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( isDIGIT_A(*s)); |
6895a8aa | 1694 | break; |
b8c5462f | 1695 | case DIGITL: |
4cadc6a9 YO |
1696 | REXEC_FBC_CSCAN_TAINT( |
1697 | isDIGIT_LC_utf8((U8*)s), | |
1698 | isDIGIT_LC(*s) | |
1699 | ); | |
6895a8aa | 1700 | break; |
a0d0e21e | 1701 | case NDIGIT: |
4cadc6a9 | 1702 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1703 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1704 | !swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1705 | !isDIGIT(*s) |
1706 | ); | |
6895a8aa | 1707 | break; |
cfaf538b | 1708 | case NDIGITA: |
8e9da4d4 KW |
1709 | REXEC_FBC_CSCAN( |
1710 | !isDIGIT_A(*s), | |
1711 | !isDIGIT_A(*s) | |
1712 | ); | |
1713 | break; | |
b8c5462f | 1714 | case NDIGITL: |
4cadc6a9 YO |
1715 | REXEC_FBC_CSCAN_TAINT( |
1716 | !isDIGIT_LC_utf8((U8*)s), | |
1717 | !isDIGIT_LC(*s) | |
1718 | ); | |
6895a8aa | 1719 | break; |
e1d1eefb YO |
1720 | case LNBREAK: |
1721 | REXEC_FBC_CSCAN( | |
1722 | is_LNBREAK_utf8(s), | |
1723 | is_LNBREAK_latin1(s) | |
1724 | ); | |
6895a8aa | 1725 | break; |
e1d1eefb YO |
1726 | case VERTWS: |
1727 | REXEC_FBC_CSCAN( | |
1728 | is_VERTWS_utf8(s), | |
1729 | is_VERTWS_latin1(s) | |
1730 | ); | |
6895a8aa | 1731 | break; |
e1d1eefb YO |
1732 | case NVERTWS: |
1733 | REXEC_FBC_CSCAN( | |
1734 | !is_VERTWS_utf8(s), | |
1735 | !is_VERTWS_latin1(s) | |
1736 | ); | |
6895a8aa | 1737 | break; |
e1d1eefb YO |
1738 | case HORIZWS: |
1739 | REXEC_FBC_CSCAN( | |
1740 | is_HORIZWS_utf8(s), | |
1741 | is_HORIZWS_latin1(s) | |
1742 | ); | |
6895a8aa | 1743 | break; |
e1d1eefb YO |
1744 | case NHORIZWS: |
1745 | REXEC_FBC_CSCAN( | |
1746 | !is_HORIZWS_utf8(s), | |
1747 | !is_HORIZWS_latin1(s) | |
1748 | ); | |
6895a8aa | 1749 | break; |
1de06328 YO |
1750 | case AHOCORASICKC: |
1751 | case AHOCORASICK: | |
07be1b83 | 1752 | { |
a0a388a1 | 1753 | DECL_TRIE_TYPE(c); |
07be1b83 YO |
1754 | /* what trie are we using right now */ |
1755 | reg_ac_data *aho | |
f8fc2ecf | 1756 | = (reg_ac_data*)progi->data->data[ ARG( c ) ]; |
3251b653 NC |
1757 | reg_trie_data *trie |
1758 | = (reg_trie_data*)progi->data->data[ aho->trie ]; | |
85fbaab2 | 1759 | HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); |
07be1b83 YO |
1760 | |
1761 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1762 | #ifdef DEBUGGING |
07be1b83 | 1763 | const char *real_start = s; |
6148ee25 | 1764 | #endif |
07be1b83 | 1765 | STRLEN maxlen = trie->maxlen; |
be8e71aa YO |
1766 | SV *sv_points; |
1767 | U8 **points; /* map of where we were in the input string | |
786e8c11 | 1768 | when reading a given char. For ASCII this |
be8e71aa | 1769 | is unnecessary overhead as the relationship |
38a44b82 NC |
1770 | is always 1:1, but for Unicode, especially |
1771 | case folded Unicode this is not true. */ | |
f9e705e8 | 1772 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
786e8c11 YO |
1773 | U8 *bitmap=NULL; |
1774 | ||
07be1b83 YO |
1775 | |
1776 | GET_RE_DEBUG_FLAGS_DECL; | |
1777 | ||
be8e71aa YO |
1778 | /* We can't just allocate points here. We need to wrap it in |
1779 | * an SV so it gets freed properly if there is a croak while | |
1780 | * running the match */ | |
1781 | ENTER; | |
1782 | SAVETMPS; | |
1783 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1784 | SvCUR_set(sv_points, | |
1785 | maxlen * sizeof(U8 *)); | |
1786 | SvPOK_on(sv_points); | |
1787 | sv_2mortal(sv_points); | |
1788 | points=(U8**)SvPV_nolen(sv_points ); | |
1de06328 YO |
1789 | if ( trie_type != trie_utf8_fold |
1790 | && (trie->bitmap || OP(c)==AHOCORASICKC) ) | |
1791 | { | |
786e8c11 YO |
1792 | if (trie->bitmap) |
1793 | bitmap=(U8*)trie->bitmap; | |
1794 | else | |
1795 | bitmap=(U8*)ANYOF_BITMAP(c); | |
07be1b83 | 1796 | } |
786e8c11 YO |
1797 | /* this is the Aho-Corasick algorithm modified a touch |
1798 | to include special handling for long "unknown char" | |
1799 | sequences. The basic idea being that we use AC as long | |
1800 | as we are dealing with a possible matching char, when | |
1801 | we encounter an unknown char (and we have not encountered | |
1802 | an accepting state) we scan forward until we find a legal | |
1803 | starting char. | |
1804 | AC matching is basically that of trie matching, except | |
1805 | that when we encounter a failing transition, we fall back | |
1806 | to the current states "fail state", and try the current char | |
1807 | again, a process we repeat until we reach the root state, | |
1808 | state 1, or a legal transition. If we fail on the root state | |
1809 | then we can either terminate if we have reached an accepting | |
1810 | state previously, or restart the entire process from the beginning | |
1811 | if we have not. | |
1812 | ||
1813 | */ | |
07be1b83 YO |
1814 | while (s <= last_start) { |
1815 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1816 | U8 *uc = (U8*)s; | |
1817 | U16 charid = 0; | |
1818 | U32 base = 1; | |
1819 | U32 state = 1; | |
1820 | UV uvc = 0; | |
1821 | STRLEN len = 0; | |
1822 | STRLEN foldlen = 0; | |
1823 | U8 *uscan = (U8*)NULL; | |
1824 | U8 *leftmost = NULL; | |
786e8c11 YO |
1825 | #ifdef DEBUGGING |
1826 | U32 accepted_word= 0; | |
1827 | #endif | |
07be1b83 YO |
1828 | U32 pointpos = 0; |
1829 | ||
1830 | while ( state && uc <= (U8*)strend ) { | |
1831 | int failed=0; | |
786e8c11 YO |
1832 | U32 word = aho->states[ state ].wordnum; |
1833 | ||
1de06328 YO |
1834 | if( state==1 ) { |
1835 | if ( bitmap ) { | |
1836 | DEBUG_TRIE_EXECUTE_r( | |
1837 | if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1838 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1839 | (char *)uc, utf8_target ); |
1de06328 YO |
1840 | PerlIO_printf( Perl_debug_log, |
1841 | " Scanning for legal start char...\n"); | |
1842 | } | |
d085b490 YO |
1843 | ); |
1844 | if (utf8_target) { | |
1845 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1846 | uc += UTF8SKIP(uc); | |
1847 | } | |
1848 | } else { | |
1849 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1850 | uc++; | |
1851 | } | |
1852 | } | |
1de06328 | 1853 | s= (char *)uc; |
786e8c11 | 1854 | } |
786e8c11 YO |
1855 | if (uc >(U8*)last_start) break; |
1856 | } | |
1857 | ||
1858 | if ( word ) { | |
2e64971a | 1859 | U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ]; |
786e8c11 YO |
1860 | if (!leftmost || lpos < leftmost) { |
1861 | DEBUG_r(accepted_word=word); | |
07be1b83 | 1862 | leftmost= lpos; |
786e8c11 | 1863 | } |
07be1b83 | 1864 | if (base==0) break; |
786e8c11 | 1865 | |
07be1b83 YO |
1866 | } |
1867 | points[pointpos++ % maxlen]= uc; | |
55eed653 NC |
1868 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
1869 | uscan, len, uvc, charid, foldlen, | |
1870 | foldbuf, uniflags); | |
786e8c11 YO |
1871 | DEBUG_TRIE_EXECUTE_r({ |
1872 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1873 | s, utf8_target ); |
07be1b83 | 1874 | PerlIO_printf(Perl_debug_log, |
786e8c11 YO |
1875 | " Charid:%3u CP:%4"UVxf" ", |
1876 | charid, uvc); | |
1877 | }); | |
07be1b83 YO |
1878 | |
1879 | do { | |
6148ee25 | 1880 | #ifdef DEBUGGING |
786e8c11 | 1881 | word = aho->states[ state ].wordnum; |
6148ee25 | 1882 | #endif |
07be1b83 YO |
1883 | base = aho->states[ state ].trans.base; |
1884 | ||
786e8c11 YO |
1885 | DEBUG_TRIE_EXECUTE_r({ |
1886 | if (failed) | |
1887 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1888 | s, utf8_target ); |
07be1b83 | 1889 | PerlIO_printf( Perl_debug_log, |
786e8c11 YO |
1890 | "%sState: %4"UVxf", word=%"UVxf, |
1891 | failed ? " Fail transition to " : "", | |
1892 | (UV)state, (UV)word); | |
1893 | }); | |
07be1b83 YO |
1894 | if ( base ) { |
1895 | U32 tmp; | |
6dd2be57 | 1896 | I32 offset; |
07be1b83 | 1897 | if (charid && |
6dd2be57 DM |
1898 | ( ((offset = base + charid |
1899 | - 1 - trie->uniquecharcount)) >= 0) | |
1900 | && ((U32)offset < trie->lasttrans) | |
1901 | && trie->trans[offset].check == state | |
1902 | && (tmp=trie->trans[offset].next)) | |
07be1b83 | 1903 | { |
786e8c11 YO |
1904 | DEBUG_TRIE_EXECUTE_r( |
1905 | PerlIO_printf( Perl_debug_log," - legal\n")); | |
07be1b83 YO |
1906 | state = tmp; |
1907 | break; | |
1908 | } | |
1909 | else { | |
786e8c11 YO |
1910 | DEBUG_TRIE_EXECUTE_r( |
1911 | PerlIO_printf( Perl_debug_log," - fail\n")); | |
1912 | failed = 1; | |
1913 | state = aho->fail[state]; | |
07be1b83 YO |
1914 | } |
1915 | } | |
1916 | else { | |
1917 | /* we must be accepting here */ | |
786e8c11 YO |
1918 | DEBUG_TRIE_EXECUTE_r( |
1919 | PerlIO_printf( Perl_debug_log," - accepting\n")); | |
1920 | failed = 1; | |
07be1b83 YO |
1921 | break; |
1922 | } | |
1923 | } while(state); | |
786e8c11 | 1924 | uc += len; |
07be1b83 YO |
1925 | if (failed) { |
1926 | if (leftmost) | |
1927 | break; | |
786e8c11 | 1928 | if (!state) state = 1; |
07be1b83 YO |
1929 | } |
1930 | } | |
1931 | if ( aho->states[ state ].wordnum ) { | |
2e64971a | 1932 | U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ]; |
786e8c11 YO |
1933 | if (!leftmost || lpos < leftmost) { |
1934 | DEBUG_r(accepted_word=aho->states[ state ].wordnum); | |
07be1b83 | 1935 | leftmost = lpos; |
786e8c11 | 1936 | } |
07be1b83 | 1937 | } |
07be1b83 YO |
1938 | if (leftmost) { |
1939 | s = (char*)leftmost; | |
786e8c11 YO |
1940 | DEBUG_TRIE_EXECUTE_r({ |
1941 | PerlIO_printf( | |
70685ca0 JH |
1942 | Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", |
1943 | (UV)accepted_word, (IV)(s - real_start) | |
786e8c11 YO |
1944 | ); |
1945 | }); | |
24b23f37 | 1946 | if (!reginfo || regtry(reginfo, &s)) { |
be8e71aa YO |
1947 | FREETMPS; |
1948 | LEAVE; | |
07be1b83 | 1949 | goto got_it; |
be8e71aa | 1950 | } |
07be1b83 | 1951 | s = HOPc(s,1); |
786e8c11 YO |
1952 | DEBUG_TRIE_EXECUTE_r({ |
1953 | PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); | |
1954 | }); | |
07be1b83 | 1955 | } else { |
786e8c11 YO |
1956 | DEBUG_TRIE_EXECUTE_r( |
1957 | PerlIO_printf( Perl_debug_log,"No match.\n")); | |
07be1b83 YO |
1958 | break; |
1959 | } | |
1960 | } | |
be8e71aa YO |
1961 | FREETMPS; |
1962 | LEAVE; | |
07be1b83 YO |
1963 | } |
1964 | break; | |
b3c9acc1 | 1965 | default: |
3c3eec57 GS |
1966 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); |
1967 | break; | |
d6a28714 | 1968 | } |
6eb5f6b9 JH |
1969 | return 0; |
1970 | got_it: | |
1971 | return s; | |
1972 | } | |
1973 | ||
fae667d5 | 1974 | |
6eb5f6b9 JH |
1975 | /* |
1976 | - regexec_flags - match a regexp against a string | |
1977 | */ | |
1978 | I32 | |
288b8c02 | 1979 | Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend, |
6eb5f6b9 JH |
1980 | char *strbeg, I32 minend, SV *sv, void *data, U32 flags) |
1981 | /* strend: pointer to null at end of string */ | |
1982 | /* strbeg: real beginning of string */ | |
1983 | /* minend: end of match must be >=minend after stringarg. */ | |
58e23c8d YO |
1984 | /* data: May be used for some additional optimizations. |
1985 | Currently its only used, with a U32 cast, for transmitting | |
1986 | the ganch offset when doing a /g match. This will change */ | |
6eb5f6b9 JH |
1987 | /* nosave: For optimizations. */ |
1988 | { | |
97aff369 | 1989 | dVAR; |
288b8c02 | 1990 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
24b23f37 | 1991 | /*register*/ char *s; |
6eb5f6b9 | 1992 | register regnode *c; |
24b23f37 | 1993 | /*register*/ char *startpos = stringarg; |
6eb5f6b9 JH |
1994 | I32 minlen; /* must match at least this many chars */ |
1995 | I32 dontbother = 0; /* how many characters not to try at end */ | |
6eb5f6b9 JH |
1996 | I32 end_shift = 0; /* Same for the end. */ /* CC */ |
1997 | I32 scream_pos = -1; /* Internal iterator of scream. */ | |
ccac19ea | 1998 | char *scream_olds = NULL; |
f2ed9b32 | 1999 | const bool utf8_target = cBOOL(DO_UTF8(sv)); |
2757e526 | 2000 | I32 multiline; |
f8fc2ecf | 2001 | RXi_GET_DECL(prog,progi); |
3b0527fe | 2002 | regmatch_info reginfo; /* create some info to pass to regtry etc */ |
e9105d30 | 2003 | regexp_paren_pair *swap = NULL; |
a3621e74 YO |
2004 | GET_RE_DEBUG_FLAGS_DECL; |
2005 | ||
7918f24d | 2006 | PERL_ARGS_ASSERT_REGEXEC_FLAGS; |
9d4ba2ae | 2007 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
2008 | |
2009 | /* Be paranoid... */ | |
2010 | if (prog == NULL || startpos == NULL) { | |
2011 | Perl_croak(aTHX_ "NULL regexp parameter"); | |
2012 | return 0; | |
2013 | } | |
2014 | ||
bbe252da | 2015 | multiline = prog->extflags & RXf_PMf_MULTILINE; |
288b8c02 | 2016 | reginfo.prog = rx; /* Yes, sorry that this is confusing. */ |
2757e526 | 2017 | |
f2ed9b32 | 2018 | RX_MATCH_UTF8_set(rx, utf8_target); |
1de06328 | 2019 | DEBUG_EXECUTE_r( |
f2ed9b32 | 2020 | debug_start_match(rx, utf8_target, startpos, strend, |
1de06328 YO |
2021 | "Matching"); |
2022 | ); | |
bac06658 | 2023 | |
6eb5f6b9 | 2024 | minlen = prog->minlen; |
1de06328 YO |
2025 | |
2026 | if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { | |
a3621e74 | 2027 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
2028 | "String too short [regexec_flags]...\n")); |
2029 | goto phooey; | |
1aa99e6b | 2030 | } |
6eb5f6b9 | 2031 | |
1de06328 | 2032 | |
6eb5f6b9 | 2033 | /* Check validity of program. */ |
f8fc2ecf | 2034 | if (UCHARAT(progi->program) != REG_MAGIC) { |
6eb5f6b9 JH |
2035 | Perl_croak(aTHX_ "corrupted regexp program"); |
2036 | } | |
2037 | ||
2038 | PL_reg_flags = 0; | |
2039 | PL_reg_eval_set = 0; | |
2040 | PL_reg_maxiter = 0; | |
2041 | ||
3c8556c3 | 2042 | if (RX_UTF8(rx)) |
6eb5f6b9 JH |
2043 | PL_reg_flags |= RF_utf8; |
2044 | ||
2045 | /* Mark beginning of line for ^ and lookbehind. */ | |
3b0527fe | 2046 | reginfo.bol = startpos; /* XXX not used ??? */ |
6eb5f6b9 | 2047 | PL_bostr = strbeg; |
3b0527fe | 2048 | reginfo.sv = sv; |
6eb5f6b9 JH |
2049 | |
2050 | /* Mark end of line for $ (and such) */ | |
2051 | PL_regeol = strend; | |
2052 | ||
2053 | /* see how far we have to get to not match where we matched before */ | |
3b0527fe | 2054 | reginfo.till = startpos+minend; |
6eb5f6b9 | 2055 | |
6eb5f6b9 JH |
2056 | /* If there is a "must appear" string, look for it. */ |
2057 | s = startpos; | |
2058 | ||
bbe252da | 2059 | if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ |
6eb5f6b9 | 2060 | MAGIC *mg; |
2c296965 | 2061 | if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */ |
58e23c8d | 2062 | reginfo.ganch = startpos + prog->gofs; |
2c296965 | 2063 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2064 | "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs)); |
2c296965 | 2065 | } else if (sv && SvTYPE(sv) >= SVt_PVMG |
6eb5f6b9 | 2066 | && SvMAGIC(sv) |
14befaf4 DM |
2067 | && (mg = mg_find(sv, PERL_MAGIC_regex_global)) |
2068 | && mg->mg_len >= 0) { | |
3b0527fe | 2069 | reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ |
2c296965 | 2070 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2071 | "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len)); |
2c296965 | 2072 | |
bbe252da | 2073 | if (prog->extflags & RXf_ANCH_GPOS) { |
3b0527fe | 2074 | if (s > reginfo.ganch) |
6eb5f6b9 | 2075 | goto phooey; |
58e23c8d | 2076 | s = reginfo.ganch - prog->gofs; |
2c296965 | 2077 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2078 | "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs)); |
c584a96e YO |
2079 | if (s < strbeg) |
2080 | goto phooey; | |
6eb5f6b9 JH |
2081 | } |
2082 | } | |
58e23c8d | 2083 | else if (data) { |
70685ca0 | 2084 | reginfo.ganch = strbeg + PTR2UV(data); |
2c296965 YO |
2085 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2086 | "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data))); | |
2087 | ||
2088 | } else { /* pos() not defined */ | |
3b0527fe | 2089 | reginfo.ganch = strbeg; |
2c296965 YO |
2090 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2091 | "GPOS: reginfo.ganch = strbeg\n")); | |
2092 | } | |
6eb5f6b9 | 2093 | } |
288b8c02 | 2094 | if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { |
e9105d30 GG |
2095 | /* We have to be careful. If the previous successful match |
2096 | was from this regex we don't want a subsequent partially | |
2097 | successful match to clobber the old results. | |
2098 | So when we detect this possibility we add a swap buffer | |
2099 | to the re, and switch the buffer each match. If we fail | |
2100 | we switch it back, otherwise we leave it swapped. | |
2101 | */ | |
2102 | swap = prog->offs; | |
2103 | /* do we need a save destructor here for eval dies? */ | |
2104 | Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); | |
c74340f9 | 2105 | } |
a0714e2c | 2106 | if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { |
6eb5f6b9 JH |
2107 | re_scream_pos_data d; |
2108 | ||
2109 | d.scream_olds = &scream_olds; | |
2110 | d.scream_pos = &scream_pos; | |
288b8c02 | 2111 | s = re_intuit_start(rx, sv, s, strend, flags, &d); |
3fa9c3d7 | 2112 | if (!s) { |
a3621e74 | 2113 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n")); |
6eb5f6b9 | 2114 | goto phooey; /* not present */ |
3fa9c3d7 | 2115 | } |
6eb5f6b9 JH |
2116 | } |
2117 | ||
1de06328 | 2118 | |
6eb5f6b9 JH |
2119 | |
2120 | /* Simplest case: anchored match need be tried only once. */ | |
2121 | /* [unless only anchor is BOL and multiline is set] */ | |
bbe252da | 2122 | if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { |
24b23f37 | 2123 | if (s == startpos && regtry(®info, &startpos)) |
6eb5f6b9 | 2124 | goto got_it; |
bbe252da YO |
2125 | else if (multiline || (prog->intflags & PREGf_IMPLICIT) |
2126 | || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ | |
6eb5f6b9 JH |
2127 | { |
2128 | char *end; | |
2129 | ||
2130 | if (minlen) | |
2131 | dontbother = minlen - 1; | |
1aa99e6b | 2132 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 2133 | /* for multiline we only have to try after newlines */ |
33b8afdf | 2134 | if (prog->check_substr || prog->check_utf8) { |
92f3d482 YO |
2135 | /* because of the goto we can not easily reuse the macros for bifurcating the |
2136 | unicode/non-unicode match modes here like we do elsewhere - demerphq */ | |
2137 | if (utf8_target) { | |
2138 | if (s == startpos) | |
2139 | goto after_try_utf8; | |
2140 | while (1) { | |
2141 | if (regtry(®info, &s)) { | |
2142 | goto got_it; | |
2143 | } | |
2144 | after_try_utf8: | |
2145 | if (s > end) { | |
2146 | goto phooey; | |
2147 | } | |
2148 | if (prog->extflags & RXf_USE_INTUIT) { | |
2149 | s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL); | |
2150 | if (!s) { | |
2151 | goto phooey; | |
2152 | } | |
2153 | } | |
2154 | else { | |
2155 | s += UTF8SKIP(s); | |
2156 | } | |
2157 | } | |
2158 | } /* end search for check string in unicode */ | |
2159 | else { | |
2160 | if (s == startpos) { | |
2161 | goto after_try_latin; | |
2162 | } | |
2163 | while (1) { | |
2164 | if (regtry(®info, &s)) { | |
2165 | goto got_it; | |
2166 | } | |
2167 | after_try_latin: | |
2168 | if (s > end) { | |
2169 | goto phooey; | |
2170 | } | |
2171 | if (prog->extflags & RXf_USE_INTUIT) { | |
2172 | s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL); | |
2173 | if (!s) { | |
2174 | goto phooey; | |
2175 | } | |
2176 | } | |
2177 | else { | |
2178 | s++; | |
2179 | } | |
2180 | } | |
2181 | } /* end search for check string in latin*/ | |
2182 | } /* end search for check string */ | |
2183 | else { /* search for newline */ | |
2184 | if (s > startpos) { | |
2185 | /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/ | |
6eb5f6b9 | 2186 | s--; |
92f3d482 YO |
2187 | } |
2188 | /* We can use a more efficient search as newlines are the same in unicode as they are in latin */ | |
6eb5f6b9 JH |
2189 | while (s < end) { |
2190 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ | |
24b23f37 | 2191 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2192 | goto got_it; |
2193 | } | |
92f3d482 YO |
2194 | } |
2195 | } /* end search for newline */ | |
2196 | } /* end anchored/multiline check string search */ | |
6eb5f6b9 | 2197 | goto phooey; |
bbe252da | 2198 | } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) |
f9f4320a | 2199 | { |
486ec47a | 2200 | /* the warning about reginfo.ganch being used without initialization |
bbe252da | 2201 | is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN |
f9f4320a | 2202 | and we only enter this block when the same bit is set. */ |
58e23c8d | 2203 | char *tmp_s = reginfo.ganch - prog->gofs; |
c584a96e YO |
2204 | |
2205 | if (tmp_s >= strbeg && regtry(®info, &tmp_s)) | |
6eb5f6b9 JH |
2206 | goto got_it; |
2207 | goto phooey; | |
2208 | } | |
2209 | ||
2210 | /* Messy cases: unanchored match. */ | |
bbe252da | 2211 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { |
6eb5f6b9 | 2212 | /* we have /x+whatever/ */ |
f2ed9b32 | 2213 | /* it must be a one character string (XXXX Except UTF_PATTERN?) */ |
33b8afdf | 2214 | char ch; |
bf93d4cc GS |
2215 | #ifdef DEBUGGING |
2216 | int did_match = 0; | |
2217 | #endif | |
f2ed9b32 KW |
2218 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2219 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2220 | ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0]; | |
bf93d4cc | 2221 | |
f2ed9b32 | 2222 | if (utf8_target) { |
4cadc6a9 | 2223 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2224 | if (*s == ch) { |
a3621e74 | 2225 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2226 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2227 | s += UTF8SKIP(s); |
2228 | while (s < strend && *s == ch) | |
2229 | s += UTF8SKIP(s); | |
2230 | } | |
4cadc6a9 | 2231 | ); |
6eb5f6b9 JH |
2232 | } |
2233 | else { | |
4cadc6a9 | 2234 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2235 | if (*s == ch) { |
a3621e74 | 2236 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2237 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2238 | s++; |
2239 | while (s < strend && *s == ch) | |
2240 | s++; | |
2241 | } | |
4cadc6a9 | 2242 | ); |
6eb5f6b9 | 2243 | } |
a3621e74 | 2244 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 2245 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
2246 | "Did not find anchored character...\n") |
2247 | ); | |
6eb5f6b9 | 2248 | } |
a0714e2c SS |
2249 | else if (prog->anchored_substr != NULL |
2250 | || prog->anchored_utf8 != NULL | |
2251 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
2252 | && prog->float_max_offset < strend - s)) { |
2253 | SV *must; | |
2254 | I32 back_max; | |
2255 | I32 back_min; | |
2256 | char *last; | |
6eb5f6b9 | 2257 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
2258 | #ifdef DEBUGGING |
2259 | int did_match = 0; | |
2260 | #endif | |
33b8afdf | 2261 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 KW |
2262 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2263 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2264 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; | |
33b8afdf JH |
2265 | back_max = back_min = prog->anchored_offset; |
2266 | } else { | |
f2ed9b32 KW |
2267 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2268 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2269 | must = utf8_target ? prog->float_utf8 : prog->float_substr; | |
33b8afdf JH |
2270 | back_max = prog->float_max_offset; |
2271 | back_min = prog->float_min_offset; | |
2272 | } | |
1de06328 YO |
2273 | |
2274 | ||
33b8afdf JH |
2275 | if (must == &PL_sv_undef) |
2276 | /* could not downgrade utf8 check substring, so must fail */ | |
2277 | goto phooey; | |
2278 | ||
1de06328 YO |
2279 | if (back_min<0) { |
2280 | last = strend; | |
2281 | } else { | |
2282 | last = HOP3c(strend, /* Cannot start after this */ | |
2283 | -(I32)(CHR_SVLEN(must) | |
2284 | - (SvTAIL(must) != 0) + back_min), strbeg); | |
2285 | } | |
6eb5f6b9 JH |
2286 | if (s > PL_bostr) |
2287 | last1 = HOPc(s, -1); | |
2288 | else | |
2289 | last1 = s - 1; /* bogus */ | |
2290 | ||
a0288114 | 2291 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 JH |
2292 | check_substr==must. */ |
2293 | scream_pos = -1; | |
2294 | dontbother = end_shift; | |
2295 | strend = HOPc(strend, -dontbother); | |
2296 | while ( (s <= last) && | |
9041c2e3 | 2297 | ((flags & REXEC_SCREAM) |
1de06328 | 2298 | ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg, |
6eb5f6b9 | 2299 | end_shift, &scream_pos, 0)) |
1de06328 | 2300 | : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), |
9041c2e3 | 2301 | (unsigned char*)strend, must, |
7fba1cd6 | 2302 | multiline ? FBMrf_MULTILINE : 0))) ) { |
4addbd3b | 2303 | /* we may be pointing at the wrong string */ |
07bc277f | 2304 | if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog)) |
3f7c398e | 2305 | s = strbeg + (s - SvPVX_const(sv)); |
a3621e74 | 2306 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
2307 | if (HOPc(s, -back_max) > last1) { |
2308 | last1 = HOPc(s, -back_min); | |
2309 | s = HOPc(s, -back_max); | |
2310 | } | |
2311 | else { | |
52657f30 | 2312 | char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; |
6eb5f6b9 JH |
2313 | |
2314 | last1 = HOPc(s, -back_min); | |
52657f30 | 2315 | s = t; |
6eb5f6b9 | 2316 | } |
f2ed9b32 | 2317 | if (utf8_target) { |
6eb5f6b9 | 2318 | while (s <= last1) { |
24b23f37 | 2319 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2320 | goto got_it; |
2321 | s += UTF8SKIP(s); | |
2322 | } | |
2323 | } | |
2324 | else { | |
2325 | while (s <= last1) { | |
24b23f37 | 2326 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2327 | goto got_it; |
2328 | s++; | |
2329 | } | |
2330 | } | |
2331 | } | |
ab3bbdeb | 2332 | DEBUG_EXECUTE_r(if (!did_match) { |
f2ed9b32 | 2333 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
2334 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
2335 | PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", | |
33b8afdf | 2336 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc | 2337 | ? "anchored" : "floating"), |
ab3bbdeb YO |
2338 | quoted, RE_SV_TAIL(must)); |
2339 | }); | |
6eb5f6b9 JH |
2340 | goto phooey; |
2341 | } | |
f8fc2ecf | 2342 | else if ( (c = progi->regstclass) ) { |
f14c76ed | 2343 | if (minlen) { |
f8fc2ecf | 2344 | const OPCODE op = OP(progi->regstclass); |
66e933ab | 2345 | /* don't bother with what can't match */ |
786e8c11 | 2346 | if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) |
f14c76ed RGS |
2347 | strend = HOPc(strend, -(minlen - 1)); |
2348 | } | |
a3621e74 | 2349 | DEBUG_EXECUTE_r({ |
be8e71aa | 2350 | SV * const prop = sv_newmortal(); |
32fc9b6a | 2351 | regprop(prog, prop, c); |
0df25f3d | 2352 | { |
f2ed9b32 | 2353 | RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb | 2354 | s,strend-s,60); |
0df25f3d | 2355 | PerlIO_printf(Perl_debug_log, |
1c8f8eb1 | 2356 | "Matching stclass %.*s against %s (%d bytes)\n", |
e4f74956 | 2357 | (int)SvCUR(prop), SvPVX_const(prop), |
ab3bbdeb | 2358 | quoted, (int)(strend - s)); |
0df25f3d | 2359 | } |
ffc61ed2 | 2360 | }); |
3b0527fe | 2361 | if (find_byclass(prog, c, s, strend, ®info)) |
6eb5f6b9 | 2362 | goto got_it; |
07be1b83 | 2363 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
2364 | } |
2365 | else { | |
2366 | dontbother = 0; | |
a0714e2c | 2367 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 2368 | /* Trim the end. */ |
d6a28714 | 2369 | char *last; |
33b8afdf JH |
2370 | SV* float_real; |
2371 | ||
f2ed9b32 KW |
2372 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2373 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2374 | float_real = utf8_target ? prog->float_utf8 : prog->float_substr; | |
d6a28714 JH |
2375 | |
2376 | if (flags & REXEC_SCREAM) { | |
33b8afdf | 2377 | last = screaminstr(sv, float_real, s - strbeg, |
d6a28714 JH |
2378 | end_shift, &scream_pos, 1); /* last one */ |
2379 | if (!last) | |
ffc61ed2 | 2380 | last = scream_olds; /* Only one occurrence. */ |
4addbd3b | 2381 | /* we may be pointing at the wrong string */ |
07bc277f | 2382 | else if (RXp_MATCH_COPIED(prog)) |
3f7c398e | 2383 | s = strbeg + (s - SvPVX_const(sv)); |
b8c5462f | 2384 | } |
d6a28714 JH |
2385 | else { |
2386 | STRLEN len; | |
cfd0369c | 2387 | const char * const little = SvPV_const(float_real, len); |
d6a28714 | 2388 | |
33b8afdf | 2389 | if (SvTAIL(float_real)) { |
d6a28714 JH |
2390 | if (memEQ(strend - len + 1, little, len - 1)) |
2391 | last = strend - len + 1; | |
7fba1cd6 | 2392 | else if (!multiline) |
9041c2e3 | 2393 | last = memEQ(strend - len, little, len) |
bd61b366 | 2394 | ? strend - len : NULL; |
b8c5462f | 2395 | else |
d6a28714 JH |
2396 | goto find_last; |
2397 | } else { | |
2398 | find_last: | |
9041c2e3 | 2399 | if (len) |
d6a28714 | 2400 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 2401 | else |
a0288114 | 2402 | last = strend; /* matching "$" */ |
b8c5462f | 2403 | } |
b8c5462f | 2404 | } |
bf93d4cc | 2405 | if (last == NULL) { |
6bda09f9 YO |
2406 | DEBUG_EXECUTE_r( |
2407 | PerlIO_printf(Perl_debug_log, | |
2408 | "%sCan't trim the tail, match fails (should not happen)%s\n", | |
2409 | PL_colors[4], PL_colors[5])); | |
bf93d4cc GS |
2410 | goto phooey; /* Should not happen! */ |
2411 | } | |
d6a28714 JH |
2412 | dontbother = strend - last + prog->float_min_offset; |
2413 | } | |
2414 | if (minlen && (dontbother < minlen)) | |
2415 | dontbother = minlen - 1; | |
2416 | strend -= dontbother; /* this one's always in bytes! */ | |
2417 | /* We don't know much -- general case. */ | |
f2ed9b32 | 2418 | if (utf8_target) { |
d6a28714 | 2419 | for (;;) { |
24b23f37 | 2420 | if (regtry(®info, &s)) |
d6a28714 JH |
2421 | goto got_it; |
2422 | if (s >= strend) | |
2423 | break; | |
b8c5462f | 2424 | s += UTF8SKIP(s); |
d6a28714 JH |
2425 | }; |
2426 | } | |
2427 | else { | |
2428 | do { | |
24b23f37 | 2429 | if (regtry(®info, &s)) |
d6a28714 JH |
2430 | goto got_it; |
2431 | } while (s++ < strend); | |
2432 | } | |
2433 | } | |
2434 | ||
2435 | /* Failure. */ | |
2436 | goto phooey; | |
2437 | ||
2438 | got_it: | |
e9105d30 | 2439 | Safefree(swap); |
288b8c02 | 2440 | RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted); |
d6a28714 | 2441 | |
19b95bf0 | 2442 | if (PL_reg_eval_set) |
4f639d21 | 2443 | restore_pos(aTHX_ prog); |
5daac39c NC |
2444 | if (RXp_PAREN_NAMES(prog)) |
2445 | (void)hv_iterinit(RXp_PAREN_NAMES(prog)); | |
d6a28714 JH |
2446 | |
2447 | /* make sure $`, $&, $', and $digit will work later */ | |
2448 | if ( !(flags & REXEC_NOT_FIRST) ) { | |
288b8c02 | 2449 | RX_MATCH_COPY_FREE(rx); |
d6a28714 | 2450 | if (flags & REXEC_COPY_STR) { |
be8e71aa | 2451 | const I32 i = PL_regeol - startpos + (stringarg - strbeg); |
f8c7b90f | 2452 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2453 | if ((SvIsCOW(sv) |
2454 | || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { | |
2455 | if (DEBUG_C_TEST) { | |
2456 | PerlIO_printf(Perl_debug_log, | |
2457 | "Copy on write: regexp capture, type %d\n", | |
2458 | (int) SvTYPE(sv)); | |
2459 | } | |
2460 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); | |
d5263905 | 2461 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
ed252734 NC |
2462 | assert (SvPOKp(prog->saved_copy)); |
2463 | } else | |
2464 | #endif | |
2465 | { | |
288b8c02 | 2466 | RX_MATCH_COPIED_on(rx); |
ed252734 NC |
2467 | s = savepvn(strbeg, i); |
2468 | prog->subbeg = s; | |
2469 | } | |
d6a28714 | 2470 | prog->sublen = i; |
d6a28714 JH |
2471 | } |
2472 | else { | |
2473 | prog->subbeg = strbeg; | |
2474 | prog->sublen = PL_regeol - strbeg; /* strend may have been modified */ | |
2475 | } | |
2476 | } | |
9041c2e3 | 2477 | |
d6a28714 JH |
2478 | return 1; |
2479 | ||
2480 | phooey: | |
a3621e74 | 2481 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 2482 | PL_colors[4], PL_colors[5])); |
d6a28714 | 2483 | if (PL_reg_eval_set) |
4f639d21 | 2484 | restore_pos(aTHX_ prog); |
e9105d30 | 2485 | if (swap) { |
c74340f9 | 2486 | /* we failed :-( roll it back */ |
e9105d30 GG |
2487 | Safefree(prog->offs); |
2488 | prog->offs = swap; | |
2489 | } | |
2490 | ||
d6a28714 JH |
2491 | return 0; |
2492 | } | |
2493 | ||
6bda09f9 | 2494 | |
d6a28714 JH |
2495 | /* |
2496 | - regtry - try match at specific point | |
2497 | */ | |
2498 | STATIC I32 /* 0 failure, 1 success */ | |
24b23f37 | 2499 | S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) |
d6a28714 | 2500 | { |
97aff369 | 2501 | dVAR; |
d6a28714 | 2502 | CHECKPOINT lastcp; |
288b8c02 NC |
2503 | REGEXP *const rx = reginfo->prog; |
2504 | regexp *const prog = (struct regexp *)SvANY(rx); | |
f8fc2ecf | 2505 | RXi_GET_DECL(prog,progi); |
a3621e74 | 2506 | GET_RE_DEBUG_FLAGS_DECL; |
7918f24d NC |
2507 | |
2508 | PERL_ARGS_ASSERT_REGTRY; | |
2509 | ||
24b23f37 | 2510 | reginfo->cutpoint=NULL; |
d6a28714 | 2511 | |
bbe252da | 2512 | if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) { |
d6a28714 JH |
2513 | MAGIC *mg; |
2514 | ||
2515 | PL_reg_eval_set = RS_init; | |
a3621e74 | 2516 | DEBUG_EXECUTE_r(DEBUG_s( |
b900a521 JH |
2517 | PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n", |
2518 | (IV)(PL_stack_sp - PL_stack_base)); | |
d6a28714 | 2519 | )); |
ea8d6ae1 | 2520 | SAVESTACK_CXPOS(); |
d6a28714 JH |
2521 | cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base; |
2522 | /* Otherwise OP_NEXTSTATE will free whatever on stack now. */ | |
2523 | SAVETMPS; | |
2524 | /* Apparently this is not needed, judging by wantarray. */ | |
e8347627 | 2525 | /* SAVEI8(cxstack[cxstack_ix].blk_gimme); |
d6a28714 JH |
2526 | cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ |
2527 | ||
3b0527fe | 2528 | if (reginfo->sv) { |
d6a28714 | 2529 | /* Make $_ available to executed code. */ |
3b0527fe | 2530 | if (reginfo->sv != DEFSV) { |
59f00321 | 2531 | SAVE_DEFSV; |
414bf5ae | 2532 | DEFSV_set(reginfo->sv); |
b8c5462f | 2533 | } |
d6a28714 | 2534 | |
3b0527fe DM |
2535 | if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) |
2536 | && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { | |
d6a28714 | 2537 | /* prepare for quick setting of pos */ |
d300d9fa | 2538 | #ifdef PERL_OLD_COPY_ON_WRITE |
51a9ea20 NC |
2539 | if (SvIsCOW(reginfo->sv)) |
2540 | sv_force_normal_flags(reginfo->sv, 0); | |
d300d9fa | 2541 | #endif |
3dab1dad | 2542 | mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, |
d300d9fa | 2543 | &PL_vtbl_mglob, NULL, 0); |
d6a28714 | 2544 | mg->mg_len = -1; |
b8c5462f | 2545 | } |
d6a28714 JH |
2546 | PL_reg_magic = mg; |
2547 | PL_reg_oldpos = mg->mg_len; | |
4f639d21 | 2548 | SAVEDESTRUCTOR_X(restore_pos, prog); |
d6a28714 | 2549 | } |
09687e5a | 2550 | if (!PL_reg_curpm) { |
a02a5408 | 2551 | Newxz(PL_reg_curpm, 1, PMOP); |
09687e5a AB |
2552 | #ifdef USE_ITHREADS |
2553 | { | |
14a49a24 | 2554 | SV* const repointer = &PL_sv_undef; |
92313705 NC |
2555 | /* this regexp is also owned by the new PL_reg_curpm, which |
2556 | will try to free it. */ | |
d2ece331 | 2557 | av_push(PL_regex_padav, repointer); |
09687e5a AB |
2558 | PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); |
2559 | PL_regex_pad = AvARRAY(PL_regex_padav); | |
2560 | } | |
2561 | #endif | |
2562 | } | |
86c29d75 NC |
2563 | #ifdef USE_ITHREADS |
2564 | /* It seems that non-ithreads works both with and without this code. | |
2565 | So for efficiency reasons it seems best not to have the code | |
2566 | compiled when it is not needed. */ | |
92313705 NC |
2567 | /* This is safe against NULLs: */ |
2568 | ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); | |
2569 | /* PM_reg_curpm owns a reference to this regexp. */ | |
2570 | ReREFCNT_inc(rx); | |
86c29d75 | 2571 | #endif |
288b8c02 | 2572 | PM_SETRE(PL_reg_curpm, rx); |
d6a28714 JH |
2573 | PL_reg_oldcurpm = PL_curpm; |
2574 | PL_curpm = PL_reg_curpm; | |
07bc277f | 2575 | if (RXp_MATCH_COPIED(prog)) { |
d6a28714 JH |
2576 | /* Here is a serious problem: we cannot rewrite subbeg, |
2577 | since it may be needed if this match fails. Thus | |
2578 | $` inside (?{}) could fail... */ | |
2579 | PL_reg_oldsaved = prog->subbeg; | |
2580 | PL_reg_oldsavedlen = prog->sublen; | |
f8c7b90f | 2581 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2582 | PL_nrs = prog->saved_copy; |
2583 | #endif | |
07bc277f | 2584 | RXp_MATCH_COPIED_off(prog); |
d6a28714 JH |
2585 | } |
2586 | else | |
bd61b366 | 2587 | PL_reg_oldsaved = NULL; |
d6a28714 JH |
2588 | prog->subbeg = PL_bostr; |
2589 | prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ | |
2590 | } | |
24b23f37 | 2591 | DEBUG_EXECUTE_r(PL_reg_starttry = *startpos); |
f0ab9afb | 2592 | prog->offs[0].start = *startpos - PL_bostr; |
24b23f37 | 2593 | PL_reginput = *startpos; |
d6a28714 | 2594 | PL_reglastparen = &prog->lastparen; |
a01268b5 | 2595 | PL_reglastcloseparen = &prog->lastcloseparen; |
d6a28714 | 2596 | prog->lastparen = 0; |
03994de8 | 2597 | prog->lastcloseparen = 0; |
d6a28714 | 2598 | PL_regsize = 0; |
f0ab9afb | 2599 | PL_regoffs = prog->offs; |
d6a28714 JH |
2600 | if (PL_reg_start_tmpl <= prog->nparens) { |
2601 | PL_reg_start_tmpl = prog->nparens*3/2 + 3; | |
2602 | if(PL_reg_start_tmp) | |
2603 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
2604 | else | |
a02a5408 | 2605 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); |
d6a28714 JH |
2606 | } |
2607 | ||
2608 | /* XXXX What this code is doing here?!!! There should be no need | |
2609 | to do this again and again, PL_reglastparen should take care of | |
3dd2943c | 2610 | this! --ilya*/ |
dafc8851 JH |
2611 | |
2612 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2613 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
daf18116 | 2614 | * PL_reglastparen), is not needed at all by the test suite |
225593e1 DM |
2615 | * (op/regexp, op/pat, op/split), but that code is needed otherwise |
2616 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ | |
2617 | * Meanwhile, this code *is* needed for the | |
daf18116 JH |
2618 | * above-mentioned test suite tests to succeed. The common theme |
2619 | * on those tests seems to be returning null fields from matches. | |
225593e1 | 2620 | * --jhi updated by dapm */ |
dafc8851 | 2621 | #if 1 |
d6a28714 | 2622 | if (prog->nparens) { |
f0ab9afb | 2623 | regexp_paren_pair *pp = PL_regoffs; |
097eb12c | 2624 | register I32 i; |
eb160463 | 2625 | for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) { |
f0ab9afb NC |
2626 | ++pp; |
2627 | pp->start = -1; | |
2628 | pp->end = -1; | |
d6a28714 JH |
2629 | } |
2630 | } | |
dafc8851 | 2631 | #endif |
02db2b7b | 2632 | REGCP_SET(lastcp); |
f8fc2ecf | 2633 | if (regmatch(reginfo, progi->program + 1)) { |
f0ab9afb | 2634 | PL_regoffs[0].end = PL_reginput - PL_bostr; |
d6a28714 JH |
2635 | return 1; |
2636 | } | |
24b23f37 YO |
2637 | if (reginfo->cutpoint) |
2638 | *startpos= reginfo->cutpoint; | |
02db2b7b | 2639 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
2640 | return 0; |
2641 | } | |
2642 | ||
02db2b7b | 2643 | |
8ba1375e MJD |
2644 | #define sayYES goto yes |
2645 | #define sayNO goto no | |
262b90c4 | 2646 | #define sayNO_SILENT goto no_silent |
8ba1375e | 2647 | |
f9f4320a YO |
2648 | /* we dont use STMT_START/END here because it leads to |
2649 | "unreachable code" warnings, which are bogus, but distracting. */ | |
2650 | #define CACHEsayNO \ | |
c476f425 DM |
2651 | if (ST.cache_mask) \ |
2652 | PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \ | |
f9f4320a | 2653 | sayNO |
3298f257 | 2654 | |
a3621e74 | 2655 | /* this is used to determine how far from the left messages like |
265c4333 YO |
2656 | 'failed...' are printed. It should be set such that messages |
2657 | are inline with the regop output that created them. | |
a3621e74 | 2658 | */ |
265c4333 | 2659 | #define REPORT_CODE_OFF 32 |
a3621e74 YO |
2660 | |
2661 | ||
40a82448 DM |
2662 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
2663 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
9e137952 | 2664 | |
86545054 DM |
2665 | #define SLAB_FIRST(s) (&(s)->states[0]) |
2666 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
2667 | ||
5d9a96ca DM |
2668 | /* grab a new slab and return the first slot in it */ |
2669 | ||
2670 | STATIC regmatch_state * | |
2671 | S_push_slab(pTHX) | |
2672 | { | |
a35a87e7 | 2673 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2674 | dMY_CXT; |
2675 | #endif | |
5d9a96ca DM |
2676 | regmatch_slab *s = PL_regmatch_slab->next; |
2677 | if (!s) { | |
2678 | Newx(s, 1, regmatch_slab); | |
2679 | s->prev = PL_regmatch_slab; | |
2680 | s->next = NULL; | |
2681 | PL_regmatch_slab->next = s; | |
2682 | } | |
2683 | PL_regmatch_slab = s; | |
86545054 | 2684 | return SLAB_FIRST(s); |
5d9a96ca | 2685 | } |
5b47454d | 2686 | |
95b24440 | 2687 | |
40a82448 DM |
2688 | /* push a new state then goto it */ |
2689 | ||
2690 | #define PUSH_STATE_GOTO(state, node) \ | |
2691 | scan = node; \ | |
2692 | st->resume_state = state; \ | |
2693 | goto push_state; | |
2694 | ||
2695 | /* push a new state with success backtracking, then goto it */ | |
2696 | ||
2697 | #define PUSH_YES_STATE_GOTO(state, node) \ | |
2698 | scan = node; \ | |
2699 | st->resume_state = state; \ | |
2700 | goto push_yes_state; | |
2701 | ||
aa283a38 | 2702 | |
aa283a38 | 2703 | |
d6a28714 | 2704 | /* |
95b24440 | 2705 | |
bf1f174e DM |
2706 | regmatch() - main matching routine |
2707 | ||
2708 | This is basically one big switch statement in a loop. We execute an op, | |
2709 | set 'next' to point the next op, and continue. If we come to a point which | |
2710 | we may need to backtrack to on failure such as (A|B|C), we push a | |
2711 | backtrack state onto the backtrack stack. On failure, we pop the top | |
2712 | state, and re-enter the loop at the state indicated. If there are no more | |
2713 | states to pop, we return failure. | |
2714 | ||
2715 | Sometimes we also need to backtrack on success; for example /A+/, where | |
2716 | after successfully matching one A, we need to go back and try to | |
2717 | match another one; similarly for lookahead assertions: if the assertion | |
2718 | completes successfully, we backtrack to the state just before the assertion | |
2719 | and then carry on. In these cases, the pushed state is marked as | |
2720 | 'backtrack on success too'. This marking is in fact done by a chain of | |
2721 | pointers, each pointing to the previous 'yes' state. On success, we pop to | |
2722 | the nearest yes state, discarding any intermediate failure-only states. | |
2723 | Sometimes a yes state is pushed just to force some cleanup code to be | |
2724 | called at the end of a successful match or submatch; e.g. (??{$re}) uses | |
2725 | it to free the inner regex. | |
2726 | ||
2727 | Note that failure backtracking rewinds the cursor position, while | |
2728 | success backtracking leaves it alone. | |
2729 | ||
2730 | A pattern is complete when the END op is executed, while a subpattern | |
2731 | such as (?=foo) is complete when the SUCCESS op is executed. Both of these | |
2732 | ops trigger the "pop to last yes state if any, otherwise return true" | |
2733 | behaviour. | |
2734 | ||
2735 | A common convention in this function is to use A and B to refer to the two | |
2736 | subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is | |
2737 | the subpattern to be matched possibly multiple times, while B is the entire | |
2738 | rest of the pattern. Variable and state names reflect this convention. | |
2739 | ||
2740 | The states in the main switch are the union of ops and failure/success of | |
2741 | substates associated with with that op. For example, IFMATCH is the op | |
2742 | that does lookahead assertions /(?=A)B/ and so the IFMATCH state means | |
2743 | 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just | |
2744 | successfully matched A and IFMATCH_A_fail is a state saying that we have | |
2745 | just failed to match A. Resume states always come in pairs. The backtrack | |
2746 | state we push is marked as 'IFMATCH_A', but when that is popped, we resume | |
2747 | at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking | |
2748 | on success or failure. | |
2749 | ||
2750 | The struct that holds a backtracking state is actually a big union, with | |
2751 | one variant for each major type of op. The variable st points to the | |
2752 | top-most backtrack struct. To make the code clearer, within each | |
2753 | block of code we #define ST to alias the relevant union. | |
2754 | ||
2755 | Here's a concrete example of a (vastly oversimplified) IFMATCH | |
2756 | implementation: | |
2757 | ||
2758 | switch (state) { | |
2759 | .... | |
2760 | ||
2761 | #define ST st->u.ifmatch | |
2762 | ||
2763 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B | |
2764 | ST.foo = ...; // some state we wish to save | |
95b24440 | 2765 | ... |
bf1f174e DM |
2766 | // push a yes backtrack state with a resume value of |
2767 | // IFMATCH_A/IFMATCH_A_fail, then continue execution at the | |
2768 | // first node of A: | |
2769 | PUSH_YES_STATE_GOTO(IFMATCH_A, A); | |
2770 | // NOTREACHED | |
2771 | ||
2772 | case IFMATCH_A: // we have successfully executed A; now continue with B | |
2773 | next = B; | |
2774 | bar = ST.foo; // do something with the preserved value | |
2775 | break; | |
2776 | ||
2777 | case IFMATCH_A_fail: // A failed, so the assertion failed | |
2778 | ...; // do some housekeeping, then ... | |
2779 | sayNO; // propagate the failure | |
2780 | ||
2781 | #undef ST | |
95b24440 | 2782 | |
bf1f174e DM |
2783 | ... |
2784 | } | |
95b24440 | 2785 | |
bf1f174e DM |
2786 | For any old-timers reading this who are familiar with the old recursive |
2787 | approach, the code above is equivalent to: | |
95b24440 | 2788 | |
bf1f174e DM |
2789 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B |
2790 | { | |
2791 | int foo = ... | |
95b24440 | 2792 | ... |
bf1f174e DM |
2793 | if (regmatch(A)) { |
2794 | next = B; | |
2795 | bar = foo; | |
2796 | break; | |
95b24440 | 2797 | } |
bf1f174e DM |
2798 | ...; // do some housekeeping, then ... |
2799 | sayNO; // propagate the failure | |
95b24440 | 2800 | } |
bf1f174e DM |
2801 | |
2802 | The topmost backtrack state, pointed to by st, is usually free. If you | |
2803 | want to claim it, populate any ST.foo fields in it with values you wish to | |
2804 | save, then do one of | |
2805 | ||
2806 | PUSH_STATE_GOTO(resume_state, node); | |
2807 | PUSH_YES_STATE_GOTO(resume_state, node); | |
2808 | ||
2809 | which sets that backtrack state's resume value to 'resume_state', pushes a | |
2810 | new free entry to the top of the backtrack stack, then goes to 'node'. | |
2811 | On backtracking, the free slot is popped, and the saved state becomes the | |
2812 | new free state. An ST.foo field in this new top state can be temporarily | |
2813 | accessed to retrieve values, but once the main loop is re-entered, it | |
2814 | becomes available for reuse. | |
2815 | ||
2816 | Note that the depth of the backtrack stack constantly increases during the | |
2817 | left-to-right execution of the pattern, rather than going up and down with | |
2818 | the pattern nesting. For example the stack is at its maximum at Z at the | |
2819 | end of the pattern, rather than at X in the following: | |
2820 | ||
2821 | /(((X)+)+)+....(Y)+....Z/ | |
2822 | ||
2823 | The only exceptions to this are lookahead/behind assertions and the cut, | |
2824 | (?>A), which pop all the backtrack states associated with A before | |
2825 | continuing. | |
2826 | ||
486ec47a | 2827 | Backtrack state structs are allocated in slabs of about 4K in size. |
bf1f174e DM |
2828 | PL_regmatch_state and st always point to the currently active state, |
2829 | and PL_regmatch_slab points to the slab currently containing | |
2830 | PL_regmatch_state. The first time regmatch() is called, the first slab is | |
2831 | allocated, and is never freed until interpreter destruction. When the slab | |
2832 | is full, a new one is allocated and chained to the end. At exit from | |
2833 | regmatch(), slabs allocated since entry are freed. | |
2834 | ||
2835 | */ | |
95b24440 | 2836 | |
40a82448 | 2837 | |
5bc10b2c | 2838 | #define DEBUG_STATE_pp(pp) \ |
265c4333 | 2839 | DEBUG_STATE_r({ \ |
f2ed9b32 | 2840 | DUMP_EXEC_POS(locinput, scan, utf8_target); \ |
5bc10b2c | 2841 | PerlIO_printf(Perl_debug_log, \ |
5d458dd8 | 2842 | " %*s"pp" %s%s%s%s%s\n", \ |
5bc10b2c | 2843 | depth*2, "", \ |
13d6edb4 | 2844 | PL_reg_name[st->resume_state], \ |
5d458dd8 YO |
2845 | ((st==yes_state||st==mark_state) ? "[" : ""), \ |
2846 | ((st==yes_state) ? "Y" : ""), \ | |
2847 | ((st==mark_state) ? "M" : ""), \ | |
2848 | ((st==yes_state||st==mark_state) ? "]" : "") \ | |
2849 | ); \ | |
265c4333 | 2850 | }); |
5bc10b2c | 2851 | |
40a82448 | 2852 | |
3dab1dad | 2853 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 2854 | |
3df15adc | 2855 | #ifdef DEBUGGING |
5bc10b2c | 2856 | |
ab3bbdeb | 2857 | STATIC void |
f2ed9b32 | 2858 | S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target, |
ab3bbdeb YO |
2859 | const char *start, const char *end, const char *blurb) |
2860 | { | |
efd26800 | 2861 | const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; |
7918f24d NC |
2862 | |
2863 | PERL_ARGS_ASSERT_DEBUG_START_MATCH; | |
2864 | ||
ab3bbdeb YO |
2865 | if (!PL_colorset) |
2866 | reginitcolors(); | |
2867 | { | |
2868 | RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), | |
d2c6dc5e | 2869 | RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); |
ab3bbdeb | 2870 | |
f2ed9b32 | 2871 | RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb YO |
2872 | start, end - start, 60); |
2873 | ||
2874 | PerlIO_printf(Perl_debug_log, | |
2875 | "%s%s REx%s %s against %s\n", | |
2876 | PL_colors[4], blurb, PL_colors[5], s0, s1); | |
2877 | ||
f2ed9b32 | 2878 | if (utf8_target||utf8_pat) |
1de06328 YO |
2879 | PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", |
2880 | utf8_pat ? "pattern" : "", | |
f2ed9b32 KW |
2881 | utf8_pat && utf8_target ? " and " : "", |
2882 | utf8_target ? "string" : "" | |
ab3bbdeb YO |
2883 | ); |
2884 | } | |
2885 | } | |
3df15adc YO |
2886 | |
2887 | STATIC void | |
786e8c11 YO |
2888 | S_dump_exec_pos(pTHX_ const char *locinput, |
2889 | const regnode *scan, | |
2890 | const char *loc_regeol, | |
2891 | const char *loc_bostr, | |
2892 | const char *loc_reg_starttry, | |
f2ed9b32 | 2893 | const bool utf8_target) |
07be1b83 | 2894 | { |
786e8c11 | 2895 | const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; |
07be1b83 | 2896 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ |
786e8c11 | 2897 | int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); |
07be1b83 YO |
2898 | /* The part of the string before starttry has one color |
2899 | (pref0_len chars), between starttry and current | |
2900 | position another one (pref_len - pref0_len chars), | |
2901 | after the current position the third one. | |
2902 | We assume that pref0_len <= pref_len, otherwise we | |
2903 | decrease pref0_len. */ | |
786e8c11 YO |
2904 | int pref_len = (locinput - loc_bostr) > (5 + taill) - l |
2905 | ? (5 + taill) - l : locinput - loc_bostr; | |
07be1b83 YO |
2906 | int pref0_len; |
2907 | ||
7918f24d NC |
2908 | PERL_ARGS_ASSERT_DUMP_EXEC_POS; |
2909 | ||
f2ed9b32 | 2910 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) |
07be1b83 | 2911 | pref_len++; |
786e8c11 YO |
2912 | pref0_len = pref_len - (locinput - loc_reg_starttry); |
2913 | if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) | |
2914 | l = ( loc_regeol - locinput > (5 + taill) - pref_len | |
2915 | ? (5 + taill) - pref_len : loc_regeol - locinput); | |
f2ed9b32 | 2916 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) |
07be1b83 YO |
2917 | l--; |
2918 | if (pref0_len < 0) | |
2919 | pref0_len = 0; | |
2920 | if (pref0_len > pref_len) | |
2921 | pref0_len = pref_len; | |
2922 | { | |
f2ed9b32 | 2923 | const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0; |
0df25f3d | 2924 | |
ab3bbdeb | 2925 | RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), |
1de06328 | 2926 | (locinput - pref_len),pref0_len, 60, 4, 5); |
0df25f3d | 2927 | |
ab3bbdeb | 2928 | RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), |
3df15adc | 2929 | (locinput - pref_len + pref0_len), |
1de06328 | 2930 | pref_len - pref0_len, 60, 2, 3); |
0df25f3d | 2931 | |
ab3bbdeb | 2932 | RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), |
1de06328 | 2933 | locinput, loc_regeol - locinput, 10, 0, 1); |
0df25f3d | 2934 | |
1de06328 | 2935 | const STRLEN tlen=len0+len1+len2; |
3df15adc | 2936 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb | 2937 | "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", |
786e8c11 | 2938 | (IV)(locinput - loc_bostr), |
07be1b83 | 2939 | len0, s0, |
07be1b83 | 2940 | len1, s1, |
07be1b83 | 2941 | (docolor ? "" : "> <"), |
07be1b83 | 2942 | len2, s2, |
f9f4320a | 2943 | (int)(tlen > 19 ? 0 : 19 - tlen), |
07be1b83 YO |
2944 | ""); |
2945 | } | |
2946 | } | |
3df15adc | 2947 | |
07be1b83 YO |
2948 | #endif |
2949 | ||
0a4db386 YO |
2950 | /* reg_check_named_buff_matched() |
2951 | * Checks to see if a named buffer has matched. The data array of | |
2952 | * buffer numbers corresponding to the buffer is expected to reside | |
2953 | * in the regexp->data->data array in the slot stored in the ARG() of | |
2954 | * node involved. Note that this routine doesn't actually care about the | |
2955 | * name, that information is not preserved from compilation to execution. | |
2956 | * Returns the index of the leftmost defined buffer with the given name | |
2957 | * or 0 if non of the buffers matched. | |
2958 | */ | |
2959 | STATIC I32 | |
7918f24d NC |
2960 | S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) |
2961 | { | |
0a4db386 | 2962 | I32 n; |
f8fc2ecf | 2963 | RXi_GET_DECL(rex,rexi); |
ad64d0ec | 2964 | SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
0a4db386 | 2965 | I32 *nums=(I32*)SvPVX(sv_dat); |
7918f24d NC |
2966 | |
2967 | PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; | |
2968 | ||
0a4db386 YO |
2969 | for ( n=0; n<SvIVX(sv_dat); n++ ) { |
2970 | if ((I32)*PL_reglastparen >= nums[n] && | |
f0ab9afb | 2971 | PL_regoffs[nums[n]].end != -1) |
0a4db386 YO |
2972 | { |
2973 | return nums[n]; | |
2974 | } | |
2975 | } | |
2976 | return 0; | |
2977 | } | |
2978 | ||
2f554ef7 DM |
2979 | |
2980 | /* free all slabs above current one - called during LEAVE_SCOPE */ | |
2981 | ||
2982 | STATIC void | |
2983 | S_clear_backtrack_stack(pTHX_ void *p) | |
2984 | { | |
2985 | regmatch_slab *s = PL_regmatch_slab->next; | |
2986 | PERL_UNUSED_ARG(p); | |
2987 | ||
2988 | if (!s) | |
2989 | return; | |
2990 | PL_regmatch_slab->next = NULL; | |
2991 | while (s) { | |
2992 | regmatch_slab * const osl = s; | |
2993 | s = s->next; | |
2994 | Safefree(osl); | |
2995 | } | |
2996 | } | |
2997 | ||
2998 | ||
28d8d7f4 YO |
2999 | #define SETREX(Re1,Re2) \ |
3000 | if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \ | |
3001 | Re1 = (Re2) | |
3002 | ||
d6a28714 | 3003 | STATIC I32 /* 0 failure, 1 success */ |
24b23f37 | 3004 | S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) |
d6a28714 | 3005 | { |
a35a87e7 | 3006 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
3007 | dMY_CXT; |
3008 | #endif | |
27da23d5 | 3009 | dVAR; |
f2ed9b32 | 3010 | register const bool utf8_target = PL_reg_match_utf8; |
4ad0818d | 3011 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
288b8c02 NC |
3012 | REGEXP *rex_sv = reginfo->prog; |
3013 | regexp *rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 3014 | RXi_GET_DECL(rex,rexi); |
2f554ef7 | 3015 | I32 oldsave; |
5d9a96ca DM |
3016 | /* the current state. This is a cached copy of PL_regmatch_state */ |
3017 | register regmatch_state *st; | |
5d9a96ca DM |
3018 | /* cache heavy used fields of st in registers */ |
3019 | register regnode *scan; | |
3020 | register regnode *next; | |
438e9bae | 3021 | register U32 n = 0; /* general value; init to avoid compiler warning */ |
24d3c4a9 | 3022 | register I32 ln = 0; /* len or last; init to avoid compiler warning */ |
5d9a96ca | 3023 | register char *locinput = PL_reginput; |
5d9a96ca | 3024 | register I32 nextchr; /* is always set to UCHARAT(locinput) */ |
24d3c4a9 | 3025 | |
b69b0499 | 3026 | bool result = 0; /* return value of S_regmatch */ |
24d3c4a9 | 3027 | int depth = 0; /* depth of backtrack stack */ |
4b196cd4 YO |
3028 | U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */ |
3029 | const U32 max_nochange_depth = | |
3030 | (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? | |
3031 | 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; | |
77cb431f DM |
3032 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
3033 | subpattern */ | |
e2e6a0f1 YO |
3034 | /* mark_state piggy backs on the yes_state logic so that when we unwind |
3035 | the stack on success we can update the mark_state as we go */ | |
3036 | regmatch_state *mark_state = NULL; /* last mark state we have seen */ | |
faec1544 | 3037 | regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ |
b8591aee | 3038 | struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ |
40a82448 | 3039 | U32 state_num; |
5d458dd8 YO |
3040 | bool no_final = 0; /* prevent failure from backtracking? */ |
3041 | bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ | |
e2e6a0f1 | 3042 | char *startpoint = PL_reginput; |
5d458dd8 YO |
3043 | SV *popmark = NULL; /* are we looking for a mark? */ |
3044 | SV *sv_commit = NULL; /* last mark name seen in failure */ | |
3045 | SV *sv_yes_mark = NULL; /* last mark name we have seen | |
486ec47a | 3046 | during a successful match */ |
5d458dd8 YO |
3047 | U32 lastopen = 0; /* last open we saw */ |
3048 | bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; | |
19b95bf0 | 3049 | SV* const oreplsv = GvSV(PL_replgv); |
24d3c4a9 DM |
3050 | /* these three flags are set by various ops to signal information to |
3051 | * the very next op. They have a useful lifetime of exactly one loop | |
3052 | * iteration, and are not preserved or restored by state pushes/pops | |
3053 | */ | |
3054 | bool sw = 0; /* the condition value in (?(cond)a|b) */ | |
3055 | bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ | |
3056 | int logical = 0; /* the following EVAL is: | |
3057 | 0: (?{...}) | |
3058 | 1: (?(?{...})X|Y) | |
3059 | 2: (??{...}) | |
3060 | or the following IFMATCH/UNLESSM is: | |
3061 | false: plain (?=foo) | |
3062 | true: used as a condition: (?(?=foo)) | |
3063 | */ | |
95b24440 | 3064 | #ifdef DEBUGGING |
e68ec53f | 3065 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
3066 | #endif |
3067 | ||
7918f24d NC |
3068 | PERL_ARGS_ASSERT_REGMATCH; |
3069 | ||
3b57cd43 | 3070 | DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ |
24b23f37 | 3071 | PerlIO_printf(Perl_debug_log,"regmatch start\n"); |
3b57cd43 | 3072 | })); |
5d9a96ca DM |
3073 | /* on first ever call to regmatch, allocate first slab */ |
3074 | if (!PL_regmatch_slab) { | |
3075 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
3076 | PL_regmatch_slab->prev = NULL; | |
3077 | PL_regmatch_slab->next = NULL; | |
86545054 | 3078 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); |
5d9a96ca DM |
3079 | } |
3080 | ||
2f554ef7 DM |
3081 | oldsave = PL_savestack_ix; |
3082 | SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL); | |
3083 | SAVEVPTR(PL_regmatch_slab); | |
3084 | SAVEVPTR(PL_regmatch_state); | |
5d9a96ca DM |
3085 | |
3086 | /* grab next free state slot */ | |
3087 | st = ++PL_regmatch_state; | |
86545054 | 3088 | if (st > SLAB_LAST(PL_regmatch_slab)) |
5d9a96ca DM |
3089 | st = PL_regmatch_state = S_push_slab(aTHX); |
3090 | ||
d6a28714 JH |
3091 | /* Note that nextchr is a byte even in UTF */ |
3092 | nextchr = UCHARAT(locinput); | |
3093 | scan = prog; | |
3094 | while (scan != NULL) { | |
8ba1375e | 3095 | |
a3621e74 | 3096 | DEBUG_EXECUTE_r( { |
6136c704 | 3097 | SV * const prop = sv_newmortal(); |
1de06328 | 3098 | regnode *rnext=regnext(scan); |
f2ed9b32 | 3099 | DUMP_EXEC_POS( locinput, scan, utf8_target ); |
32fc9b6a | 3100 | regprop(rex, prop, scan); |
07be1b83 YO |
3101 | |
3102 | PerlIO_printf(Perl_debug_log, | |
3103 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
f8fc2ecf | 3104 | (IV)(scan - rexi->program), depth*2, "", |
07be1b83 | 3105 | SvPVX_const(prop), |
1de06328 | 3106 | (PL_regkind[OP(scan)] == END || !rnext) ? |
f8fc2ecf | 3107 | 0 : (IV)(rnext - rexi->program)); |
2a782b5b | 3108 | }); |
d6a28714 JH |
3109 | |
3110 | next = scan + NEXT_OFF(scan); | |
3111 | if (next == scan) | |
3112 | next = NULL; | |
40a82448 | 3113 | state_num = OP(scan); |
d6a28714 | 3114 | |
40a82448 | 3115 | reenter_switch: |
34a81e2b B |
3116 | |
3117 | assert(PL_reglastparen == &rex->lastparen); | |
3118 | assert(PL_reglastcloseparen == &rex->lastcloseparen); | |
3119 | assert(PL_regoffs == rex->offs); | |
3120 | ||
40a82448 | 3121 | switch (state_num) { |
d6a28714 | 3122 | case BOL: |
7fba1cd6 | 3123 | if (locinput == PL_bostr) |
d6a28714 | 3124 | { |
3b0527fe | 3125 | /* reginfo->till = reginfo->bol; */ |
b8c5462f JH |
3126 | break; |
3127 | } | |
d6a28714 JH |
3128 | sayNO; |
3129 | case MBOL: | |
12d33761 HS |
3130 | if (locinput == PL_bostr || |
3131 | ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n')) | |
d6a28714 | 3132 | { |
b8c5462f JH |
3133 | break; |
3134 | } | |
d6a28714 JH |
3135 | sayNO; |
3136 | case SBOL: | |
c2a73568 | 3137 | if (locinput == PL_bostr) |
b8c5462f | 3138 | break; |
d6a28714 JH |
3139 | sayNO; |
3140 | case GPOS: | |
3b0527fe | 3141 | if (locinput == reginfo->ganch) |
d6a28714 JH |
3142 | break; |
3143 | sayNO; | |
ee9b8eae YO |
3144 | |
3145 | case KEEPS: | |
3146 | /* update the startpoint */ | |
f0ab9afb | 3147 | st->u.keeper.val = PL_regoffs[0].start; |
ee9b8eae | 3148 | PL_reginput = locinput; |
f0ab9afb | 3149 | PL_regoffs[0].start = locinput - PL_bostr; |
ee9b8eae YO |
3150 | PUSH_STATE_GOTO(KEEPS_next, next); |
3151 | /*NOT-REACHED*/ | |
3152 | case KEEPS_next_fail: | |
3153 | /* rollback the start point change */ | |
f0ab9afb | 3154 | PL_regoffs[0].start = st->u.keeper.val; |
ee9b8eae YO |
3155 | sayNO_SILENT; |
3156 | /*NOT-REACHED*/ | |
d6a28714 | 3157 | case EOL: |
d6a28714 JH |
3158 | goto seol; |
3159 | case MEOL: | |
d6a28714 | 3160 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') |
b8c5462f | 3161 | sayNO; |
b8c5462f | 3162 | break; |
d6a28714 JH |
3163 | case SEOL: |
3164 | seol: | |
3165 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') | |
b8c5462f | 3166 | sayNO; |
d6a28714 | 3167 | if (PL_regeol - locinput > 1) |
b8c5462f | 3168 | sayNO; |
b8c5462f | 3169 | break; |
d6a28714 JH |
3170 | case EOS: |
3171 | if (PL_regeol != locinput) | |
b8c5462f | 3172 | sayNO; |
d6a28714 | 3173 | break; |
ffc61ed2 | 3174 | case SANY: |
d6a28714 | 3175 | if (!nextchr && locinput >= PL_regeol) |
4633a7c4 | 3176 | sayNO; |
f2ed9b32 | 3177 | if (utf8_target) { |
f33976b4 DB |
3178 | locinput += PL_utf8skip[nextchr]; |
3179 | if (locinput > PL_regeol) | |
3180 | sayNO; | |
3181 | nextchr = UCHARAT(locinput); | |
3182 | } | |
3183 | else | |
3184 | nextchr = UCHARAT(++locinput); | |
3185 | break; | |
3186 | case CANY: | |
3187 | if (!nextchr && locinput >= PL_regeol) | |
3188 | sayNO; | |
b8c5462f | 3189 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 3190 | break; |
ffc61ed2 | 3191 | case REG_ANY: |
1aa99e6b IH |
3192 | if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n') |
3193 | sayNO; | |
f2ed9b32 | 3194 | if (utf8_target) { |
b8c5462f | 3195 | locinput += PL_utf8skip[nextchr]; |
d6a28714 JH |
3196 | if (locinput > PL_regeol) |
3197 | sayNO; | |
a0ed51b3 | 3198 | nextchr = UCHARAT(locinput); |
a0ed51b3 | 3199 | } |
1aa99e6b IH |
3200 | else |
3201 | nextchr = UCHARAT(++locinput); | |
a0ed51b3 | 3202 | break; |
166ba7cd DM |
3203 | |
3204 | #undef ST | |
3205 | #define ST st->u.trie | |
786e8c11 YO |
3206 | case TRIEC: |
3207 | /* In this case the charclass data is available inline so | |
3208 | we can fail fast without a lot of extra overhead. | |
3209 | */ | |
f2ed9b32 | 3210 | if (scan->flags == EXACT || !utf8_target) { |
786e8c11 YO |
3211 | if(!ANYOF_BITMAP_TEST(scan, *locinput)) { |
3212 | DEBUG_EXECUTE_r( | |
3213 | PerlIO_printf(Perl_debug_log, | |
3214 | "%*s %sfailed to match trie start class...%s\n", | |
5bc10b2c | 3215 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
786e8c11 YO |
3216 | ); |
3217 | sayNO_SILENT; | |
3218 | /* NOTREACHED */ | |
3219 | } | |
3220 | } | |
3221 | /* FALL THROUGH */ | |
5b47454d | 3222 | case TRIE: |
2e64971a DM |
3223 | /* the basic plan of execution of the trie is: |
3224 | * At the beginning, run though all the states, and | |
3225 | * find the longest-matching word. Also remember the position | |
3226 | * of the shortest matching word. For example, this pattern: | |
3227 | * 1 2 3 4 5 | |
3228 | * ab|a|x|abcd|abc | |
3229 | * when matched against the string "abcde", will generate | |
3230 | * accept states for all words except 3, with the longest | |
3231 | * matching word being 4, and the shortest being 1 (with | |
3232 | * the position being after char 1 of the string). | |
3233 | * | |
3234 | * Then for each matching word, in word order (i.e. 1,2,4,5), | |
3235 | * we run the remainder of the pattern; on each try setting | |
3236 | * the current position to the character following the word, | |
3237 | * returning to try the next word on failure. | |
3238 | * | |
3239 | * We avoid having to build a list of words at runtime by | |
3240 | * using a compile-time structure, wordinfo[].prev, which | |
3241 | * gives, for each word, the previous accepting word (if any). | |
3242 | * In the case above it would contain the mappings 1->2, 2->0, | |
3243 | * 3->0, 4->5, 5->1. We can use this table to generate, from | |
3244 | * the longest word (4 above), a list of all words, by | |
3245 | * following the list of prev pointers; this gives us the | |
3246 | * unordered list 4,5,1,2. Then given the current word we have | |
3247 | * just tried, we can go through the list and find the | |
3248 | * next-biggest word to try (so if we just failed on word 2, | |
3249 | * the next in the list is 4). | |
3250 | * | |
3251 | * Since at runtime we don't record the matching position in | |
3252 | * the string for each word, we have to work that out for | |
3253 | * each word we're about to process. The wordinfo table holds | |
3254 | * the character length of each word; given that we recorded | |
3255 | * at the start: the position of the shortest word and its | |
3256 | * length in chars, we just need to move the pointer the | |
3257 | * difference between the two char lengths. Depending on | |
3258 | * Unicode status and folding, that's cheap or expensive. | |
3259 | * | |
3260 | * This algorithm is optimised for the case where are only a | |
3261 | * small number of accept states, i.e. 0,1, or maybe 2. | |
3262 | * With lots of accepts states, and having to try all of them, | |
3263 | * it becomes quadratic on number of accept states to find all | |
3264 | * the next words. | |
3265 | */ | |
3266 | ||
3dab1dad | 3267 | { |
07be1b83 | 3268 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
a0a388a1 | 3269 | DECL_TRIE_TYPE(scan); |
3dab1dad YO |
3270 | |
3271 | /* what trie are we using right now */ | |
be8e71aa | 3272 | reg_trie_data * const trie |
f8fc2ecf | 3273 | = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; |
85fbaab2 | 3274 | HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); |
3dab1dad | 3275 | U32 state = trie->startstate; |
166ba7cd | 3276 | |
3dab1dad YO |
3277 | if (trie->bitmap && trie_type != trie_utf8_fold && |
3278 | !TRIE_BITMAP_TEST(trie,*locinput) | |
3279 | ) { | |
3280 | if (trie->states[ state ].wordnum) { | |
3281 | DEBUG_EXECUTE_r( | |
3282 | PerlIO_printf(Perl_debug_log, | |
3283 | "%*s %smatched empty string...%s\n", | |
5bc10b2c | 3284 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad | 3285 | ); |
20dbff7c YO |
3286 | if (!trie->jump) |
3287 | break; | |
3dab1dad YO |
3288 | } else { |
3289 | DEBUG_EXECUTE_r( | |
3290 | PerlIO_printf(Perl_debug_log, | |
786e8c11 | 3291 | "%*s %sfailed to match trie start class...%s\n", |
5bc10b2c | 3292 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad YO |
3293 | ); |
3294 | sayNO_SILENT; | |
3295 | } | |
3296 | } | |
166ba7cd | 3297 | |
786e8c11 YO |
3298 | { |
3299 | U8 *uc = ( U8* )locinput; | |
3300 | ||
3301 | STRLEN len = 0; | |
3302 | STRLEN foldlen = 0; | |
3303 | U8 *uscan = (U8*)NULL; | |
786e8c11 | 3304 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
2e64971a DM |
3305 | U32 charcount = 0; /* how many input chars we have matched */ |
3306 | U32 accepted = 0; /* have we seen any accepting states? */ | |
786e8c11 | 3307 | |
786e8c11 YO |
3308 | ST.B = next; |
3309 | ST.jump = trie->jump; | |
786e8c11 | 3310 | ST.me = scan; |
2e64971a DM |
3311 | ST.firstpos = NULL; |
3312 | ST.longfold = FALSE; /* char longer if folded => it's harder */ | |
3313 | ST.nextword = 0; | |
3314 | ||
3315 | /* fully traverse the TRIE; note the position of the | |
3316 | shortest accept state and the wordnum of the longest | |
3317 | accept state */ | |
07be1b83 | 3318 | |
a3621e74 | 3319 | while ( state && uc <= (U8*)PL_regeol ) { |
786e8c11 | 3320 | U32 base = trie->states[ state ].trans.base; |
f9f4320a | 3321 | UV uvc = 0; |
acb909b4 | 3322 | U16 charid = 0; |
2e64971a DM |
3323 | U16 wordnum; |
3324 | wordnum = trie->states[ state ].wordnum; | |
3325 | ||
3326 | if (wordnum) { /* it's an accept state */ | |
3327 | if (!accepted) { | |
3328 | accepted = 1; | |
3329 | /* record first match position */ | |
3330 | if (ST.longfold) { | |
3331 | ST.firstpos = (U8*)locinput; | |
3332 | ST.firstchars = 0; | |
5b47454d | 3333 | } |
2e64971a DM |
3334 | else { |
3335 | ST.firstpos = uc; | |
3336 | ST.firstchars = charcount; | |
3337 | } | |
3338 | } | |
3339 | if (!ST.nextword || wordnum < ST.nextword) | |
3340 | ST.nextword = wordnum; | |
3341 | ST.topword = wordnum; | |
786e8c11 | 3342 | } |
a3621e74 | 3343 | |
07be1b83 | 3344 | DEBUG_TRIE_EXECUTE_r({ |
f2ed9b32 | 3345 | DUMP_EXEC_POS( (char *)uc, scan, utf8_target ); |
a3621e74 | 3346 | PerlIO_printf( Perl_debug_log, |
2e64971a | 3347 | "%*s %sState: %4"UVxf" Accepted: %c ", |
5bc10b2c | 3348 | 2+depth * 2, "", PL_colors[4], |
2e64971a | 3349 | (UV)state, (accepted ? 'Y' : 'N')); |
07be1b83 | 3350 | }); |
a3621e74 | 3351 | |
2e64971a | 3352 | /* read a char and goto next state */ |
a3621e74 | 3353 | if ( base ) { |
6dd2be57 | 3354 | I32 offset; |
55eed653 NC |
3355 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
3356 | uscan, len, uvc, charid, foldlen, | |
3357 | foldbuf, uniflags); | |
2e64971a DM |
3358 | charcount++; |
3359 | if (foldlen>0) | |
3360 | ST.longfold = TRUE; | |
5b47454d | 3361 | if (charid && |
6dd2be57 DM |
3362 | ( ((offset = |
3363 | base + charid - 1 - trie->uniquecharcount)) >= 0) | |
3364 | ||
3365 | && ((U32)offset < trie->lasttrans) | |
3366 | && trie->trans[offset].check == state) | |
5b47454d | 3367 | { |
6dd2be57 | 3368 | state = trie->trans[offset].next; |
5b47454d DM |
3369 | } |
3370 | else { | |
3371 | state = 0; | |
3372 | } | |
3373 | uc += len; | |
3374 | ||
3375 | } | |
3376 | else { | |
a3621e74 YO |
3377 | state = 0; |
3378 | } | |
3379 | DEBUG_TRIE_EXECUTE_r( | |
e4584336 | 3380 | PerlIO_printf( Perl_debug_log, |
786e8c11 | 3381 | "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n", |
e4584336 | 3382 | charid, uvc, (UV)state, PL_colors[5] ); |
a3621e74 YO |
3383 | ); |
3384 | } | |
2e64971a | 3385 | if (!accepted) |
a3621e74 | 3386 | sayNO; |
a3621e74 | 3387 | |
2e64971a DM |
3388 | /* calculate total number of accept states */ |
3389 | { | |
3390 | U16 w = ST.topword; | |
3391 | accepted = 0; | |
3392 | while (w) { | |
3393 | w = trie->wordinfo[w].prev; | |
3394 | accepted++; | |
3395 | } | |
3396 | ST.accepted = accepted; | |
3397 | } | |
3398 | ||
166ba7cd DM |
3399 | DEBUG_EXECUTE_r( |
3400 | PerlIO_printf( Perl_debug_log, | |
3401 | "%*s %sgot %"IVdf" possible matches%s\n", | |
5bc10b2c | 3402 | REPORT_CODE_OFF + depth * 2, "", |
166ba7cd DM |
3403 | PL_colors[4], (IV)ST.accepted, PL_colors[5] ); |
3404 | ); | |
2e64971a | 3405 | goto trie_first_try; /* jump into the fail handler */ |
786e8c11 | 3406 | }} |
fae667d5 | 3407 | /* NOTREACHED */ |
2e64971a DM |
3408 | |
3409 | case TRIE_next_fail: /* we failed - try next alternative */ | |
fae667d5 YO |
3410 | if ( ST.jump) { |
3411 | REGCP_UNWIND(ST.cp); | |
3412 | for (n = *PL_reglastparen; n > ST.lastparen; n--) | |
f0ab9afb | 3413 | PL_regoffs[n].end = -1; |
fae667d5 YO |
3414 | *PL_reglastparen = n; |
3415 | } | |
2e64971a DM |
3416 | if (!--ST.accepted) { |
3417 | DEBUG_EXECUTE_r({ | |
3418 | PerlIO_printf( Perl_debug_log, | |
3419 | "%*s %sTRIE failed...%s\n", | |
3420 | REPORT_CODE_OFF+depth*2, "", | |
3421 | PL_colors[4], | |
3422 | PL_colors[5] ); | |
3423 | }); | |
3424 | sayNO_SILENT; | |
3425 | } | |
3426 | { | |
3427 | /* Find next-highest word to process. Note that this code | |
3428 | * is O(N^2) per trie run (O(N) per branch), so keep tight */ | |
d9a396a3 DM |
3429 | register U16 min = 0; |
3430 | register U16 word; | |
2e64971a DM |
3431 | register U16 const nextword = ST.nextword; |
3432 | register reg_trie_wordinfo * const wordinfo | |
3433 | = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo; | |
3434 | for (word=ST.topword; word; word=wordinfo[word].prev) { | |
3435 | if (word > nextword && (!min || word < min)) | |
3436 | min = word; | |
3437 | } | |
3438 | ST.nextword = min; | |
3439 | } | |
3440 | ||
fae667d5 | 3441 | trie_first_try: |
5d458dd8 YO |
3442 | if (do_cutgroup) { |
3443 | do_cutgroup = 0; | |
3444 | no_final = 0; | |
3445 | } | |
fae667d5 YO |
3446 | |
3447 | if ( ST.jump) { | |
3448 | ST.lastparen = *PL_reglastparen; | |
3449 | REGCP_SET(ST.cp); | |
2e64971a | 3450 | } |
a3621e74 | 3451 | |
2e64971a | 3452 | /* find start char of end of current word */ |
166ba7cd | 3453 | { |
2e64971a DM |
3454 | U32 chars; /* how many chars to skip */ |
3455 | U8 *uc = ST.firstpos; | |
3456 | reg_trie_data * const trie | |
3457 | = (reg_trie_data*)rexi->data->data[ARG(ST.me)]; | |
3458 | ||
3459 | assert((trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3460 | >= ST.firstchars); | |
3461 | chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3462 | - ST.firstchars; | |
3463 | ||
3464 | if (ST.longfold) { | |
3465 | /* the hard option - fold each char in turn and find | |
3466 | * its folded length (which may be different */ | |
3467 | U8 foldbuf[UTF8_MAXBYTES_CASE + 1]; | |
3468 | STRLEN foldlen; | |
3469 | STRLEN len; | |
d9a396a3 | 3470 | UV uvc; |
2e64971a DM |
3471 | U8 *uscan; |
3472 | ||
3473 | while (chars) { | |
f2ed9b32 | 3474 | if (utf8_target) { |
2e64971a DM |
3475 | uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len, |
3476 | uniflags); | |
3477 | uc += len; | |
3478 | } | |
3479 | else { | |
3480 | uvc = *uc; | |
3481 | uc++; | |
3482 | } | |
3483 | uvc = to_uni_fold(uvc, foldbuf, &foldlen); | |
3484 | uscan = foldbuf; | |
3485 | while (foldlen) { | |
3486 | if (!--chars) | |
3487 | break; | |
3488 | uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len, | |
3489 | uniflags); | |
3490 | uscan += len; | |
3491 | foldlen -= len; | |
3492 | } | |
3493 | } | |
a3621e74 | 3494 | } |
2e64971a | 3495 | else { |
f2ed9b32 | 3496 | if (utf8_target) |
2e64971a DM |
3497 | while (chars--) |
3498 | uc += UTF8SKIP(uc); | |
3499 | else | |
3500 | uc += chars; | |
3501 | } | |
3502 | PL_reginput = (char *)uc; | |
3503 | } | |
166ba7cd | 3504 | |
2e64971a DM |
3505 | scan = (ST.jump && ST.jump[ST.nextword]) |
3506 | ? ST.me + ST.jump[ST.nextword] | |
3507 | : ST.B; | |
166ba7cd | 3508 | |
2e64971a DM |
3509 | DEBUG_EXECUTE_r({ |
3510 | PerlIO_printf( Perl_debug_log, | |
3511 | "%*s %sTRIE matched word #%d, continuing%s\n", | |
3512 | REPORT_CODE_OFF+depth*2, "", | |
3513 | PL_colors[4], | |
3514 | ST.nextword, | |
3515 | PL_colors[5] | |
3516 | ); | |
3517 | }); | |
3518 | ||
3519 | if (ST.accepted > 1 || has_cutgroup) { | |
3520 | PUSH_STATE_GOTO(TRIE_next, scan); | |
3521 | /* NOTREACHED */ | |
166ba7cd | 3522 | } |
2e64971a DM |
3523 | /* only one choice left - just continue */ |
3524 | DEBUG_EXECUTE_r({ | |
3525 | AV *const trie_words | |
3526 | = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]); | |
3527 | SV ** const tmp = av_fetch( trie_words, | |
3528 | ST.nextword-1, 0 ); | |
3529 | SV *sv= tmp ? sv_newmortal() : NULL; | |
3530 | ||
3531 | PerlIO_printf( Perl_debug_log, | |
3532 | "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n", | |
3533 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], | |
3534 | ST.nextword, | |
3535 | tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0, | |
3536 | PL_colors[0], PL_colors[1], | |
c89df6cf | 3537 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII |
2e64971a DM |
3538 | ) |
3539 | : "not compiled under -Dr", | |
3540 | PL_colors[5] ); | |
3541 | }); | |
3542 | ||
3543 | locinput = PL_reginput; | |
3544 | nextchr = UCHARAT(locinput); | |
3545 | continue; /* execute rest of RE */ | |
166ba7cd | 3546 | /* NOTREACHED */ |
166ba7cd DM |
3547 | #undef ST |
3548 | ||
95b24440 DM |
3549 | case EXACT: { |
3550 | char *s = STRING(scan); | |
24d3c4a9 | 3551 | ln = STR_LEN(scan); |
f2ed9b32 | 3552 | if (utf8_target != UTF_PATTERN) { |
bc517b45 | 3553 | /* The target and the pattern have differing utf8ness. */ |
1aa99e6b | 3554 | char *l = locinput; |
24d3c4a9 | 3555 | const char * const e = s + ln; |
a72c7584 | 3556 | |
f2ed9b32 | 3557 | if (utf8_target) { |
5ff6fc6d | 3558 | /* The target is utf8, the pattern is not utf8. */ |
1aa99e6b | 3559 | while (s < e) { |
a3b680e6 | 3560 | STRLEN ulen; |
1aa99e6b | 3561 | if (l >= PL_regeol) |
5ff6fc6d JH |
3562 | sayNO; |
3563 | if (NATIVE_TO_UNI(*(U8*)s) != | |
89ebb4a3 | 3564 | utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen, |
041457d9 | 3565 | uniflags)) |
5ff6fc6d | 3566 | sayNO; |
bc517b45 | 3567 | l += ulen; |
5ff6fc6d | 3568 | s ++; |
1aa99e6b | 3569 | } |
5ff6fc6d JH |
3570 | } |
3571 | else { | |
3572 | /* The target is not utf8, the pattern is utf8. */ | |
1aa99e6b | 3573 | while (s < e) { |
a3b680e6 | 3574 | STRLEN ulen; |
1aa99e6b IH |
3575 | if (l >= PL_regeol) |
3576 | sayNO; | |
5ff6fc6d | 3577 | if (NATIVE_TO_UNI(*((U8*)l)) != |
89ebb4a3 | 3578 | utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen, |
041457d9 | 3579 | uniflags)) |
1aa99e6b | 3580 | sayNO; |
bc517b45 | 3581 | s += ulen; |
a72c7584 | 3582 | l ++; |
1aa99e6b | 3583 | } |
5ff6fc6d | 3584 | } |
1aa99e6b IH |
3585 | locinput = l; |
3586 | nextchr = UCHARAT(locinput); | |
3587 | break; | |
3588 | } | |
bc517b45 | 3589 | /* The target and the pattern have the same utf8ness. */ |
d6a28714 JH |
3590 | /* Inline the first character, for speed. */ |
3591 | if (UCHARAT(s) != nextchr) | |
3592 | sayNO; | |
24d3c4a9 | 3593 | if (PL_regeol - locinput < ln) |
d6a28714 | 3594 | sayNO; |
24d3c4a9 | 3595 | if (ln > 1 && memNE(s, locinput, ln)) |
d6a28714 | 3596 | sayNO; |
24d3c4a9 | 3597 | locinput += ln; |
d6a28714 JH |
3598 | nextchr = UCHARAT(locinput); |
3599 | break; | |
95b24440 | 3600 | } |
9a5a5549 | 3601 | case EXACTFL: { |
a932d541 | 3602 | re_fold_t folder; |
9a5a5549 KW |
3603 | const U8 * fold_array; |
3604 | const char * s; | |
d513472c | 3605 | U32 fold_utf8_flags; |
9a5a5549 | 3606 | |
b8c5462f | 3607 | PL_reg_flags |= RF_tainted; |
9a5a5549 KW |
3608 | folder = foldEQ_locale; |
3609 | fold_array = PL_fold_locale; | |
17580e7a | 3610 | fold_utf8_flags = FOLDEQ_UTF8_LOCALE; |
9a5a5549 KW |
3611 | goto do_exactf; |
3612 | ||
3613 | case EXACTFU: | |
3614 | folder = foldEQ_latin1; | |
3615 | fold_array = PL_fold_latin1; | |
d513472c | 3616 | fold_utf8_flags = 0; |
9a5a5549 KW |
3617 | goto do_exactf; |
3618 | ||
2f7f8cb1 KW |
3619 | case EXACTFA: |
3620 | folder = foldEQ_latin1; | |
3621 | fold_array = PL_fold_latin1; | |
57014d77 | 3622 | fold_utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; |
2f7f8cb1 KW |
3623 | goto do_exactf; |
3624 | ||
9a5a5549 KW |
3625 | case EXACTF: |
3626 | folder = foldEQ; | |
3627 | fold_array = PL_fold; | |
d513472c | 3628 | fold_utf8_flags = 0; |
9a5a5549 KW |
3629 | |
3630 | do_exactf: | |
3631 | s = STRING(scan); | |
24d3c4a9 | 3632 | ln = STR_LEN(scan); |
d6a28714 | 3633 | |
f2ed9b32 | 3634 | if (utf8_target || UTF_PATTERN) { |
d07ddd77 | 3635 | /* Either target or the pattern are utf8. */ |
be8e71aa | 3636 | const char * const l = locinput; |
d07ddd77 | 3637 | char *e = PL_regeol; |
bc517b45 | 3638 | |
d513472c KW |
3639 | if (! foldEQ_utf8_flags(s, 0, ln, cBOOL(UTF_PATTERN), |
3640 | l, &e, 0, utf8_target, fold_utf8_flags)) { | |
5486206c JH |
3641 | /* One more case for the sharp s: |
3642 | * pack("U0U*", 0xDF) =~ /ss/i, | |
3643 | * the 0xC3 0x9F are the UTF-8 | |
3644 | * byte sequence for the U+00DF. */ | |
e1d1eefb | 3645 | |
f2ed9b32 | 3646 | if (!(utf8_target && |
e1d1eefb | 3647 | toLOWER(s[0]) == 's' && |
24d3c4a9 | 3648 | ln >= 2 && |
5486206c JH |
3649 | toLOWER(s[1]) == 's' && |
3650 | (U8)l[0] == 0xC3 && | |
3651 | e - l >= 2 && | |
3652 | (U8)l[1] == 0x9F)) | |
3653 | sayNO; | |
3654 | } | |
d07ddd77 JH |
3655 | locinput = e; |
3656 | nextchr = UCHARAT(locinput); | |
3657 | break; | |
a0ed51b3 | 3658 | } |
d6a28714 | 3659 | |
0a138b74 | 3660 | /* Neither the target nor the pattern are utf8 */ |
d6a28714 | 3661 | if (UCHARAT(s) != nextchr && |
9a5a5549 KW |
3662 | UCHARAT(s) != fold_array[nextchr]) |
3663 | { | |
a0ed51b3 | 3664 | sayNO; |
9a5a5549 | 3665 | } |
24d3c4a9 | 3666 | if (PL_regeol - locinput < ln) |
b8c5462f | 3667 | sayNO; |
9a5a5549 | 3668 | if (ln > 1 && ! folder(s, locinput, ln)) |
4633a7c4 | 3669 | sayNO; |
24d3c4a9 | 3670 | locinput += ln; |
d6a28714 | 3671 | nextchr = UCHARAT(locinput); |
a0d0e21e | 3672 | break; |
9a5a5549 | 3673 | } |
63ac0dad KW |
3674 | |
3675 | /* XXX Could improve efficiency by separating these all out using a | |
3676 | * macro or in-line function. At that point regcomp.c would no longer | |
3677 | * have to set the FLAGS fields of these */ | |
b2680017 YO |
3678 | case BOUNDL: |
3679 | case NBOUNDL: | |
3680 | PL_reg_flags |= RF_tainted; | |
3681 | /* FALL THROUGH */ | |
3682 | case BOUND: | |
63ac0dad | 3683 | case BOUNDU: |
cfaf538b | 3684 | case BOUNDA: |
b2680017 | 3685 | case NBOUND: |
63ac0dad | 3686 | case NBOUNDU: |
cfaf538b | 3687 | case NBOUNDA: |
b2680017 | 3688 | /* was last char in word? */ |
cfaf538b | 3689 | if (utf8_target && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET) { |
b2680017 YO |
3690 | if (locinput == PL_bostr) |
3691 | ln = '\n'; | |
3692 | else { | |
3693 | const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); | |
3694 | ||
3695 | ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); | |
3696 | } | |
63ac0dad | 3697 | if (FLAGS(scan) != REGEX_LOCALE_CHARSET) { |
b2680017 YO |
3698 | ln = isALNUM_uni(ln); |
3699 | LOAD_UTF8_CHARCLASS_ALNUM(); | |
f2ed9b32 | 3700 | n = swash_fetch(PL_utf8_alnum, (U8*)locinput, utf8_target); |
b2680017 YO |
3701 | } |
3702 | else { | |
3703 | ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); | |
3704 | n = isALNUM_LC_utf8((U8*)locinput); | |
3705 | } | |
3706 | } | |
3707 | else { | |
cfaf538b KW |
3708 | |
3709 | /* Here the string isn't utf8, or is utf8 and only ascii | |
3710 | * characters are to match \w. In the latter case looking at | |
3711 | * the byte just prior to the current one may be just the final | |
3712 | * byte of a multi-byte character. This is ok. There are two | |
3713 | * cases: | |
3714 | * 1) it is a single byte character, and then the test is doing | |
3715 | * just what it's supposed to. | |
3716 | * 2) it is a multi-byte character, in which case the final | |
3717 | * byte is never mistakable for ASCII, and so the test | |
3718 | * will say it is not a word character, which is the | |
3719 | * correct answer. */ | |
b2680017 YO |
3720 | ln = (locinput != PL_bostr) ? |
3721 | UCHARAT(locinput - 1) : '\n'; | |
63ac0dad KW |
3722 | switch (FLAGS(scan)) { |
3723 | case REGEX_UNICODE_CHARSET: | |
3724 | ln = isWORDCHAR_L1(ln); | |
3725 | n = isWORDCHAR_L1(nextchr); | |
3726 | break; | |
3727 | case REGEX_LOCALE_CHARSET: | |
3728 | ln = isALNUM_LC(ln); | |
3729 | n = isALNUM_LC(nextchr); | |
3730 | break; | |
3731 | case REGEX_DEPENDS_CHARSET: | |
3732 | ln = isALNUM(ln); | |
3733 | n = isALNUM(nextchr); | |
3734 | break; | |
cfaf538b KW |
3735 | case REGEX_ASCII_RESTRICTED_CHARSET: |
3736 | ln = isWORDCHAR_A(ln); | |
3737 | n = isWORDCHAR_A(nextchr); | |
3738 | break; | |
63ac0dad KW |
3739 | default: |
3740 | Perl_croak(aTHX_ "panic: Unexpected FLAGS %u in op %u", FLAGS(scan), OP(scan)); | |
3741 | break; | |
b2680017 YO |
3742 | } |
3743 | } | |
63ac0dad KW |
3744 | /* Note requires that all BOUNDs be lower than all NBOUNDs in |
3745 | * regcomp.sym */ | |
3746 | if (((!ln) == (!n)) == (OP(scan) < NBOUND)) | |
b2680017 YO |
3747 | sayNO; |
3748 | break; | |
f56b6394 | 3749 | case ANYOFV: |
d6a28714 | 3750 | case ANYOF: |
f56b6394 | 3751 | if (utf8_target || state_num == ANYOFV) { |
9e55ce06 | 3752 | STRLEN inclasslen = PL_regeol - locinput; |
20ed0b26 KW |
3753 | if (locinput >= PL_regeol) |
3754 | sayNO; | |
9e55ce06 | 3755 | |
f2ed9b32 | 3756 | if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target)) |
09b08e9b | 3757 | sayNO; |
b32d7d3e | 3758 | locinput += inclasslen; |
b8c5462f | 3759 | nextchr = UCHARAT(locinput); |
e0f9d4a8 | 3760 | break; |
ffc61ed2 JH |
3761 | } |
3762 | else { | |
3763 | if (nextchr < 0) | |
3764 | nextchr = UCHARAT(locinput); | |
ffc61ed2 JH |
3765 | if (!nextchr && locinput >= PL_regeol) |
3766 | sayNO; | |
20ed0b26 | 3767 | if (!REGINCLASS(rex, scan, (U8*)locinput)) |
09b08e9b | 3768 | sayNO; |
ffc61ed2 | 3769 | nextchr = UCHARAT(++locinput); |
e0f9d4a8 JH |
3770 | break; |
3771 | } | |
b8c5462f | 3772 | break; |
20d0b1e9 | 3773 | /* Special char classes - The defines start on line 129 or so */ |
ee9a90b8 KW |
3774 | CCC_TRY_U(ALNUM, NALNUM, isWORDCHAR, |
3775 | ALNUML, NALNUML, isALNUM_LC, isALNUM_LC_utf8, | |
3776 | ALNUMU, NALNUMU, isWORDCHAR_L1, | |
cfaf538b | 3777 | ALNUMA, NALNUMA, isWORDCHAR_A, |
779d7b58 | 3778 | alnum, "a"); |
ee9a90b8 KW |
3779 | |
3780 | CCC_TRY_U(SPACE, NSPACE, isSPACE, | |
3781 | SPACEL, NSPACEL, isSPACE_LC, isSPACE_LC_utf8, | |
3782 | SPACEU, NSPACEU, isSPACE_L1, | |
cfaf538b | 3783 | SPACEA, NSPACEA, isSPACE_A, |
779d7b58 | 3784 | space, " "); |
ee9a90b8 KW |
3785 | |
3786 | CCC_TRY(DIGIT, NDIGIT, isDIGIT, | |
3787 | DIGITL, NDIGITL, isDIGIT_LC, isDIGIT_LC_utf8, | |
cfaf538b | 3788 | DIGITA, NDIGITA, isDIGIT_A, |
779d7b58 | 3789 | digit, "0"); |
20d0b1e9 | 3790 | |
37e2e78e KW |
3791 | case CLUMP: /* Match \X: logical Unicode character. This is defined as |
3792 | a Unicode extended Grapheme Cluster */ | |
3793 | /* From http://www.unicode.org/reports/tr29 (5.2 version). An | |
3794 | extended Grapheme Cluster is: | |
3795 | ||
3796 | CR LF | |
3797 | | Prepend* Begin Extend* | |
3798 | | . | |
3799 | ||
3800 | Begin is (Hangul-syllable | ! Control) | |
3801 | Extend is (Grapheme_Extend | Spacing_Mark) | |
3802 | Control is [ GCB_Control CR LF ] | |
3803 | ||
3804 | The discussion below shows how the code for CLUMP is derived | |
3805 | from this regex. Note that most of these concepts are from | |
3806 | property values of the Grapheme Cluster Boundary (GCB) property. | |
3807 | No code point can have multiple property values for a given | |
3808 | property. Thus a code point in Prepend can't be in Control, but | |
3809 | it must be in !Control. This is why Control above includes | |
3810 | GCB_Control plus CR plus LF. The latter two are used in the GCB | |
3811 | property separately, and so can't be in GCB_Control, even though | |
3812 | they logically are controls. Control is not the same as gc=cc, | |
3813 | but includes format and other characters as well. | |
3814 | ||
3815 | The Unicode definition of Hangul-syllable is: | |
3816 | L+ | |
3817 | | (L* ( ( V | LV ) V* | LVT ) T*) | |
3818 | | T+ | |
3819 | ) | |
3820 | Each of these is a value for the GCB property, and hence must be | |
3821 | disjoint, so the order they are tested is immaterial, so the | |
3822 | above can safely be changed to | |
3823 | T+ | |
3824 | | L+ | |
3825 | | (L* ( LVT | ( V | LV ) V*) T*) | |
3826 | ||
3827 | The last two terms can be combined like this: | |
3828 | L* ( L | |
3829 | | (( LVT | ( V | LV ) V*) T*)) | |
3830 | ||
3831 | And refactored into this: | |
3832 | L* (L | LVT T* | V V* T* | LV V* T*) | |
3833 | ||
3834 | That means that if we have seen any L's at all we can quit | |
3835 | there, but if the next character is a LVT, a V or and LV we | |
3836 | should keep going. | |
3837 | ||
3838 | There is a subtlety with Prepend* which showed up in testing. | |
3839 | Note that the Begin, and only the Begin is required in: | |
3840 | | Prepend* Begin Extend* | |
3841 | Also, Begin contains '! Control'. A Prepend must be a '! | |
3842 | Control', which means it must be a Begin. What it comes down to | |
3843 | is that if we match Prepend* and then find no suitable Begin | |
3844 | afterwards, that if we backtrack the last Prepend, that one will | |
3845 | be a suitable Begin. | |
3846 | */ | |
3847 | ||
b7c83a7e | 3848 | if (locinput >= PL_regeol) |
a0ed51b3 | 3849 | sayNO; |
f2ed9b32 | 3850 | if (! utf8_target) { |
37e2e78e KW |
3851 | |
3852 | /* Match either CR LF or '.', as all the other possibilities | |
3853 | * require utf8 */ | |
3854 | locinput++; /* Match the . or CR */ | |
3855 | if (nextchr == '\r' | |
3856 | && locinput < PL_regeol | |
3857 | && UCHARAT(locinput) == '\n') locinput++; | |
3858 | } | |
3859 | else { | |
3860 | ||
3861 | /* Utf8: See if is ( CR LF ); already know that locinput < | |
3862 | * PL_regeol, so locinput+1 is in bounds */ | |
3863 | if (nextchr == '\r' && UCHARAT(locinput + 1) == '\n') { | |
3864 | locinput += 2; | |
3865 | } | |
3866 | else { | |
3867 | /* In case have to backtrack to beginning, then match '.' */ | |
3868 | char *starting = locinput; | |
3869 | ||
3870 | /* In case have to backtrack the last prepend */ | |
3871 | char *previous_prepend = 0; | |
3872 | ||
3873 | LOAD_UTF8_CHARCLASS_GCB(); | |
3874 | ||
3875 | /* Match (prepend)* */ | |
3876 | while (locinput < PL_regeol | |
3877 | && swash_fetch(PL_utf8_X_prepend, | |
f2ed9b32 | 3878 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3879 | { |
3880 | previous_prepend = locinput; | |
3881 | locinput += UTF8SKIP(locinput); | |
3882 | } | |
3883 | ||
3884 | /* As noted above, if we matched a prepend character, but | |
3885 | * the next thing won't match, back off the last prepend we | |
3886 | * matched, as it is guaranteed to match the begin */ | |
3887 | if (previous_prepend | |
3888 | && (locinput >= PL_regeol | |
3889 | || ! swash_fetch(PL_utf8_X_begin, | |
f2ed9b32 | 3890 | (U8*)locinput, utf8_target))) |
37e2e78e KW |
3891 | { |
3892 | locinput = previous_prepend; | |
3893 | } | |
3894 | ||
3895 | /* Note that here we know PL_regeol > locinput, as we | |
3896 | * tested that upon input to this switch case, and if we | |
3897 | * moved locinput forward, we tested the result just above | |
3898 | * and it either passed, or we backed off so that it will | |
3899 | * now pass */ | |
f2ed9b32 | 3900 | if (! swash_fetch(PL_utf8_X_begin, (U8*)locinput, utf8_target)) { |
37e2e78e KW |
3901 | |
3902 | /* Here did not match the required 'Begin' in the | |
3903 | * second term. So just match the very first | |
3904 | * character, the '.' of the final term of the regex */ | |
3905 | locinput = starting + UTF8SKIP(starting); | |
3906 | } else { | |
3907 | ||
3908 | /* Here is the beginning of a character that can have | |
3909 | * an extender. It is either a hangul syllable, or a | |
3910 | * non-control */ | |
3911 | if (swash_fetch(PL_utf8_X_non_hangul, | |
f2ed9b32 | 3912 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3913 | { |
3914 | ||
3915 | /* Here not a Hangul syllable, must be a | |
3916 | * ('! * Control') */ | |
3917 | locinput += UTF8SKIP(locinput); | |
3918 | } else { | |
3919 | ||
3920 | /* Here is a Hangul syllable. It can be composed | |
3921 | * of several individual characters. One | |
3922 | * possibility is T+ */ | |
3923 | if (swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 3924 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3925 | { |
3926 | while (locinput < PL_regeol | |
3927 | && swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 3928 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3929 | { |
3930 | locinput += UTF8SKIP(locinput); | |
3931 | } | |
3932 | } else { | |
3933 | ||
3934 | /* Here, not T+, but is a Hangul. That means | |
3935 | * it is one of the others: L, LV, LVT or V, | |
3936 | * and matches: | |
3937 | * L* (L | LVT T* | V V* T* | LV V* T*) */ | |
3938 | ||
3939 | /* Match L* */ | |
3940 | while (locinput < PL_regeol | |
3941 | && swash_fetch(PL_utf8_X_L, | |
f2ed9b32 | 3942 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3943 | { |
3944 | locinput += UTF8SKIP(locinput); | |
3945 | } | |
3946 | ||
3947 | /* Here, have exhausted L*. If the next | |
3948 | * character is not an LV, LVT nor V, it means | |
3949 | * we had to have at least one L, so matches L+ | |
3950 | * in the original equation, we have a complete | |
3951 | * hangul syllable. Are done. */ | |
3952 | ||
3953 | if (locinput < PL_regeol | |
3954 | && swash_fetch(PL_utf8_X_LV_LVT_V, | |
f2ed9b32 | 3955 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3956 | { |
3957 | ||
3958 | /* Otherwise keep going. Must be LV, LVT | |
3959 | * or V. See if LVT */ | |
3960 | if (swash_fetch(PL_utf8_X_LVT, | |
f2ed9b32 | 3961 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3962 | { |
3963 | locinput += UTF8SKIP(locinput); | |
3964 | } else { | |
3965 | ||
3966 | /* Must be V or LV. Take it, then | |
3967 | * match V* */ | |
3968 | locinput += UTF8SKIP(locinput); | |
3969 | while (locinput < PL_regeol | |
3970 | && swash_fetch(PL_utf8_X_V, | |
f2ed9b32 | 3971 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3972 | { |
3973 | locinput += UTF8SKIP(locinput); | |
3974 | } | |
3975 | } | |
3976 | ||
3977 | /* And any of LV, LVT, or V can be followed | |
3978 | * by T* */ | |
3979 | while (locinput < PL_regeol | |
3980 | && swash_fetch(PL_utf8_X_T, | |
3981 | (U8*)locinput, | |
f2ed9b32 | 3982 | utf8_target)) |
37e2e78e KW |
3983 | { |
3984 | locinput += UTF8SKIP(locinput); | |
3985 | } | |
3986 | } | |
3987 | } | |
3988 | } | |
3989 | ||
3990 | /* Match any extender */ | |
3991 | while (locinput < PL_regeol | |
3992 | && swash_fetch(PL_utf8_X_extend, | |
f2ed9b32 | 3993 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3994 | { |
3995 | locinput += UTF8SKIP(locinput); | |
3996 | } | |
3997 | } | |
3998 | } | |
3999 | if (locinput > PL_regeol) sayNO; | |
4000 | } | |
a0ed51b3 LW |
4001 | nextchr = UCHARAT(locinput); |
4002 | break; | |
81714fb9 YO |
4003 | |
4004 | case NREFFL: | |
d7ef4b73 KW |
4005 | { /* The capture buffer cases. The ones beginning with N for the |
4006 | named buffers just convert to the equivalent numbered and | |
4007 | pretend they were called as the corresponding numbered buffer | |
4008 | op. */ | |
26ecd678 TC |
4009 | /* don't initialize these in the declaration, it makes C++ |
4010 | unhappy */ | |
81714fb9 | 4011 | char *s; |
ff1157ca | 4012 | char type; |
8368298a TC |
4013 | re_fold_t folder; |
4014 | const U8 *fold_array; | |
26ecd678 | 4015 | UV utf8_fold_flags; |
8368298a | 4016 | |
81714fb9 | 4017 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4018 | folder = foldEQ_locale; |
4019 | fold_array = PL_fold_locale; | |
4020 | type = REFFL; | |
17580e7a | 4021 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4022 | goto do_nref; |
4023 | ||
2f7f8cb1 KW |
4024 | case NREFFA: |
4025 | folder = foldEQ_latin1; | |
4026 | fold_array = PL_fold_latin1; | |
4027 | type = REFFA; | |
4028 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4029 | goto do_nref; | |
4030 | ||
d7ef4b73 KW |
4031 | case NREFFU: |
4032 | folder = foldEQ_latin1; | |
4033 | fold_array = PL_fold_latin1; | |
4034 | type = REFFU; | |
d513472c | 4035 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4036 | goto do_nref; |
4037 | ||
81714fb9 | 4038 | case NREFF: |
d7ef4b73 KW |
4039 | folder = foldEQ; |
4040 | fold_array = PL_fold; | |
4041 | type = REFF; | |
d513472c | 4042 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4043 | goto do_nref; |
4044 | ||
4045 | case NREF: | |
4046 | type = REF; | |
83d7b90b KW |
4047 | folder = NULL; |
4048 | fold_array = NULL; | |
d513472c | 4049 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4050 | do_nref: |
4051 | ||
4052 | /* For the named back references, find the corresponding buffer | |
4053 | * number */ | |
0a4db386 YO |
4054 | n = reg_check_named_buff_matched(rex,scan); |
4055 | ||
d7ef4b73 | 4056 | if ( ! n ) { |
81714fb9 | 4057 | sayNO; |
d7ef4b73 KW |
4058 | } |
4059 | goto do_nref_ref_common; | |
4060 | ||
c8756f30 | 4061 | case REFFL: |
3280af22 | 4062 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4063 | folder = foldEQ_locale; |
4064 | fold_array = PL_fold_locale; | |
17580e7a | 4065 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4066 | goto do_ref; |
4067 | ||
2f7f8cb1 KW |
4068 | case REFFA: |
4069 | folder = foldEQ_latin1; | |
4070 | fold_array = PL_fold_latin1; | |
4071 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4072 | goto do_ref; | |
4073 | ||
d7ef4b73 KW |
4074 | case REFFU: |
4075 | folder = foldEQ_latin1; | |
4076 | fold_array = PL_fold_latin1; | |
d513472c | 4077 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4078 | goto do_ref; |
4079 | ||
4080 | case REFF: | |
4081 | folder = foldEQ; | |
4082 | fold_array = PL_fold; | |
d513472c | 4083 | utf8_fold_flags = 0; |
83d7b90b | 4084 | goto do_ref; |
d7ef4b73 | 4085 | |
c277df42 | 4086 | case REF: |
83d7b90b KW |
4087 | folder = NULL; |
4088 | fold_array = NULL; | |
d513472c | 4089 | utf8_fold_flags = 0; |
83d7b90b | 4090 | |
d7ef4b73 | 4091 | do_ref: |
81714fb9 | 4092 | type = OP(scan); |
d7ef4b73 KW |
4093 | n = ARG(scan); /* which paren pair */ |
4094 | ||
4095 | do_nref_ref_common: | |
f0ab9afb | 4096 | ln = PL_regoffs[n].start; |
2c2d71f5 | 4097 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
3b6647e0 | 4098 | if (*PL_reglastparen < n || ln == -1) |
af3f8c16 | 4099 | sayNO; /* Do not match unless seen CLOSEn. */ |
f0ab9afb | 4100 | if (ln == PL_regoffs[n].end) |
a0d0e21e | 4101 | break; |
a0ed51b3 | 4102 | |
24d3c4a9 | 4103 | s = PL_bostr + ln; |
d7ef4b73 | 4104 | if (type != REF /* REF can do byte comparison */ |
2f65c56d | 4105 | && (utf8_target || type == REFFU)) |
d7ef4b73 KW |
4106 | { /* XXX handle REFFL better */ |
4107 | char * limit = PL_regeol; | |
4108 | ||
4109 | /* This call case insensitively compares the entire buffer | |
4110 | * at s, with the current input starting at locinput, but | |
4111 | * not going off the end given by PL_regeol, and returns in | |
4112 | * limit upon success, how much of the current input was | |
4113 | * matched */ | |
d513472c KW |
4114 | if (! foldEQ_utf8_flags(s, NULL, PL_regoffs[n].end - ln, utf8_target, |
4115 | locinput, &limit, 0, utf8_target, utf8_fold_flags)) | |
d7ef4b73 KW |
4116 | { |
4117 | sayNO; | |
a0ed51b3 | 4118 | } |
d7ef4b73 | 4119 | locinput = limit; |
a0ed51b3 LW |
4120 | nextchr = UCHARAT(locinput); |
4121 | break; | |
4122 | } | |
4123 | ||
d7ef4b73 | 4124 | /* Not utf8: Inline the first character, for speed. */ |
76e3520e | 4125 | if (UCHARAT(s) != nextchr && |
81714fb9 | 4126 | (type == REF || |
d7ef4b73 | 4127 | UCHARAT(s) != fold_array[nextchr])) |
4633a7c4 | 4128 | sayNO; |
f0ab9afb | 4129 | ln = PL_regoffs[n].end - ln; |
24d3c4a9 | 4130 | if (locinput + ln > PL_regeol) |
4633a7c4 | 4131 | sayNO; |
81714fb9 | 4132 | if (ln > 1 && (type == REF |
24d3c4a9 | 4133 | ? memNE(s, locinput, ln) |
d7ef4b73 | 4134 | : ! folder(s, locinput, ln))) |
4633a7c4 | 4135 | sayNO; |
24d3c4a9 | 4136 | locinput += ln; |
76e3520e | 4137 | nextchr = UCHARAT(locinput); |
a0d0e21e | 4138 | break; |
81714fb9 | 4139 | } |
a0d0e21e | 4140 | case NOTHING: |
c277df42 | 4141 | case TAIL: |
a0d0e21e LW |
4142 | break; |
4143 | case BACK: | |
4144 | break; | |
40a82448 DM |
4145 | |
4146 | #undef ST | |
4147 | #define ST st->u.eval | |
c277df42 | 4148 | { |
c277df42 | 4149 | SV *ret; |
d2f13c59 | 4150 | REGEXP *re_sv; |
6bda09f9 | 4151 | regexp *re; |
f8fc2ecf | 4152 | regexp_internal *rei; |
1a147d38 YO |
4153 | regnode *startpoint; |
4154 | ||
4155 | case GOSTART: | |
e7707071 YO |
4156 | case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */ |
4157 | if (cur_eval && cur_eval->locinput==locinput) { | |
24b23f37 | 4158 | if (cur_eval->u.eval.close_paren == (U32)ARG(scan)) |
1a147d38 | 4159 | Perl_croak(aTHX_ "Infinite recursion in regex"); |
4b196cd4 | 4160 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 YO |
4161 | Perl_croak(aTHX_ |
4162 | "Pattern subroutine nesting without pos change" | |
4163 | " exceeded limit in regex"); | |
6bda09f9 YO |
4164 | } else { |
4165 | nochange_depth = 0; | |
1a147d38 | 4166 | } |
288b8c02 | 4167 | re_sv = rex_sv; |
6bda09f9 | 4168 | re = rex; |
f8fc2ecf | 4169 | rei = rexi; |
288b8c02 | 4170 | (void)ReREFCNT_inc(rex_sv); |
1a147d38 | 4171 | if (OP(scan)==GOSUB) { |
6bda09f9 YO |
4172 | startpoint = scan + ARG2L(scan); |
4173 | ST.close_paren = ARG(scan); | |
4174 | } else { | |
f8fc2ecf | 4175 | startpoint = rei->program+1; |
6bda09f9 YO |
4176 | ST.close_paren = 0; |
4177 | } | |
4178 | goto eval_recurse_doit; | |
4179 | /* NOTREACHED */ | |
4180 | case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ | |
4181 | if (cur_eval && cur_eval->locinput==locinput) { | |
4b196cd4 | 4182 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 | 4183 | Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex"); |
6bda09f9 YO |
4184 | } else { |
4185 | nochange_depth = 0; | |
4186 | } | |
8e5e9ebe | 4187 | { |
4aabdb9b DM |
4188 | /* execute the code in the {...} */ |
4189 | dSP; | |
6136c704 | 4190 | SV ** const before = SP; |
4aabdb9b DM |
4191 | OP_4tree * const oop = PL_op; |
4192 | COP * const ocurcop = PL_curcop; | |
4193 | PAD *old_comppad; | |
d80618d2 | 4194 | char *saved_regeol = PL_regeol; |
91332126 FR |
4195 | struct re_save_state saved_state; |
4196 | ||
6562f1c4 | 4197 | /* To not corrupt the existing regex state while executing the |
b7f4cd04 FR |
4198 | * eval we would normally put it on the save stack, like with |
4199 | * save_re_context. However, re-evals have a weird scoping so we | |
4200 | * can't just add ENTER/LEAVE here. With that, things like | |
4201 | * | |
4202 | * (?{$a=2})(a(?{local$a=$a+1}))*aak*c(?{$b=$a}) | |
4203 | * | |
4204 | * would break, as they expect the localisation to be unwound | |
4205 | * only when the re-engine backtracks through the bit that | |
4206 | * localised it. | |
4207 | * | |
4208 | * What we do instead is just saving the state in a local c | |
4209 | * variable. | |
4210 | */ | |
91332126 FR |
4211 | Copy(&PL_reg_state, &saved_state, 1, struct re_save_state); |
4212 | ||
4aabdb9b | 4213 | n = ARG(scan); |
f8fc2ecf | 4214 | PL_op = (OP_4tree*)rexi->data->data[n]; |
24b23f37 YO |
4215 | DEBUG_STATE_r( PerlIO_printf(Perl_debug_log, |
4216 | " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); | |
f8fc2ecf | 4217 | PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]); |
f0ab9afb | 4218 | PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr; |
4aabdb9b | 4219 | |
2bf803e2 YO |
4220 | if (sv_yes_mark) { |
4221 | SV *sv_mrk = get_sv("REGMARK", 1); | |
4222 | sv_setsv(sv_mrk, sv_yes_mark); | |
4223 | } | |
4224 | ||
8e5e9ebe RGS |
4225 | CALLRUNOPS(aTHX); /* Scalar context. */ |
4226 | SPAGAIN; | |
4227 | if (SP == before) | |
075aa684 | 4228 | ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */ |
8e5e9ebe RGS |
4229 | else { |
4230 | ret = POPs; | |
4231 | PUTBACK; | |
4232 | } | |
4aabdb9b | 4233 | |
91332126 FR |
4234 | Copy(&saved_state, &PL_reg_state, 1, struct re_save_state); |
4235 | ||
4aabdb9b DM |
4236 | PL_op = oop; |
4237 | PAD_RESTORE_LOCAL(old_comppad); | |
4238 | PL_curcop = ocurcop; | |
d80618d2 | 4239 | PL_regeol = saved_regeol; |
24d3c4a9 | 4240 | if (!logical) { |
4aabdb9b DM |
4241 | /* /(?{...})/ */ |
4242 | sv_setsv(save_scalar(PL_replgv), ret); | |
4aabdb9b DM |
4243 | break; |
4244 | } | |
8e5e9ebe | 4245 | } |
24d3c4a9 DM |
4246 | if (logical == 2) { /* Postponed subexpression: /(??{...})/ */ |
4247 | logical = 0; | |
4aabdb9b | 4248 | { |
4f639d21 DM |
4249 | /* extract RE object from returned value; compiling if |
4250 | * necessary */ | |
6136c704 | 4251 | MAGIC *mg = NULL; |
288b8c02 | 4252 | REGEXP *rx = NULL; |
5c35adbb NC |
4253 | |
4254 | if (SvROK(ret)) { | |
288b8c02 | 4255 | SV *const sv = SvRV(ret); |
5c35adbb NC |
4256 | |
4257 | if (SvTYPE(sv) == SVt_REGEXP) { | |
d2f13c59 | 4258 | rx = (REGEXP*) sv; |
5c35adbb NC |
4259 | } else if (SvSMAGICAL(sv)) { |
4260 | mg = mg_find(sv, PERL_MAGIC_qr); | |
4261 | assert(mg); | |
4262 | } | |
4263 | } else if (SvTYPE(ret) == SVt_REGEXP) { | |
d2f13c59 | 4264 | rx = (REGEXP*) ret; |
5c35adbb | 4265 | } else if (SvSMAGICAL(ret)) { |
124ee91a NC |
4266 | if (SvGMAGICAL(ret)) { |
4267 | /* I don't believe that there is ever qr magic | |
4268 | here. */ | |
4269 | assert(!mg_find(ret, PERL_MAGIC_qr)); | |
faf82a0b | 4270 | sv_unmagic(ret, PERL_MAGIC_qr); |
124ee91a NC |
4271 | } |
4272 | else { | |
faf82a0b | 4273 | mg = mg_find(ret, PERL_MAGIC_qr); |
124ee91a NC |
4274 | /* testing suggests mg only ends up non-NULL for |
4275 | scalars who were upgraded and compiled in the | |
4276 | else block below. In turn, this is only | |
4277 | triggered in the "postponed utf8 string" tests | |
4278 | in t/op/pat.t */ | |
4279 | } | |
0f5d15d6 | 4280 | } |
faf82a0b | 4281 | |
0f5d15d6 | 4282 | if (mg) { |
d2f13c59 | 4283 | rx = (REGEXP *) mg->mg_obj; /*XXX:dmq*/ |
e2560c33 | 4284 | assert(rx); |
0f5d15d6 | 4285 | } |
288b8c02 | 4286 | if (rx) { |
f0826785 | 4287 | rx = reg_temp_copy(NULL, rx); |
288b8c02 | 4288 | } |
0f5d15d6 | 4289 | else { |
c737faaf | 4290 | U32 pm_flags = 0; |
a3b680e6 | 4291 | const I32 osize = PL_regsize; |
0f5d15d6 | 4292 | |
b9ad30b4 NC |
4293 | if (DO_UTF8(ret)) { |
4294 | assert (SvUTF8(ret)); | |
4295 | } else if (SvUTF8(ret)) { | |
4296 | /* Not doing UTF-8, despite what the SV says. Is | |
4297 | this only if we're trapped in use 'bytes'? */ | |
4298 | /* Make a copy of the octet sequence, but without | |
4299 | the flag on, as the compiler now honours the | |
4300 | SvUTF8 flag on ret. */ | |
4301 | STRLEN len; | |
4302 | const char *const p = SvPV(ret, len); | |
4303 | ret = newSVpvn_flags(p, len, SVs_TEMP); | |
4304 | } | |
288b8c02 | 4305 | rx = CALLREGCOMP(ret, pm_flags); |
9041c2e3 | 4306 | if (!(SvFLAGS(ret) |
faf82a0b | 4307 | & (SVs_TEMP | SVs_PADTMP | SVf_READONLY |
3ce3ed55 | 4308 | | SVs_GMG))) { |
a2794585 NC |
4309 | /* This isn't a first class regexp. Instead, it's |
4310 | caching a regexp onto an existing, Perl visible | |
4311 | scalar. */ | |
ad64d0ec | 4312 | sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0); |
3ce3ed55 | 4313 | } |
0f5d15d6 | 4314 | PL_regsize = osize; |
0f5d15d6 | 4315 | } |
288b8c02 NC |
4316 | re_sv = rx; |
4317 | re = (struct regexp *)SvANY(rx); | |
4aabdb9b | 4318 | } |
07bc277f | 4319 | RXp_MATCH_COPIED_off(re); |
28d8d7f4 YO |
4320 | re->subbeg = rex->subbeg; |
4321 | re->sublen = rex->sublen; | |
f8fc2ecf | 4322 | rei = RXi_GET(re); |
6bda09f9 | 4323 | DEBUG_EXECUTE_r( |
f2ed9b32 | 4324 | debug_start_match(re_sv, utf8_target, locinput, PL_regeol, |
6bda09f9 YO |
4325 | "Matching embedded"); |
4326 | ); | |
f8fc2ecf | 4327 | startpoint = rei->program + 1; |
1a147d38 | 4328 | ST.close_paren = 0; /* only used for GOSUB */ |
6bda09f9 YO |
4329 | /* borrowed from regtry */ |
4330 | if (PL_reg_start_tmpl <= re->nparens) { | |
4331 | PL_reg_start_tmpl = re->nparens*3/2 + 3; | |
4332 | if(PL_reg_start_tmp) | |
4333 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
4334 | else | |
4335 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
dd5def09 | 4336 | } |
aa283a38 | 4337 | |
1a147d38 | 4338 | eval_recurse_doit: /* Share code with GOSUB below this line */ |
aa283a38 | 4339 | /* run the pattern returned from (??{...}) */ |
40a82448 DM |
4340 | ST.cp = regcppush(0); /* Save *all* the positions. */ |
4341 | REGCP_SET(ST.lastcp); | |
6bda09f9 | 4342 | |
f0ab9afb | 4343 | PL_regoffs = re->offs; /* essentially NOOP on GOSUB */ |
6bda09f9 | 4344 | |
0357f1fd ML |
4345 | /* see regtry, specifically PL_reglast(?:close)?paren is a pointer! (i dont know why) :dmq */ |
4346 | PL_reglastparen = &re->lastparen; | |
4347 | PL_reglastcloseparen = &re->lastcloseparen; | |
4348 | re->lastparen = 0; | |
4349 | re->lastcloseparen = 0; | |
4350 | ||
4aabdb9b | 4351 | PL_reginput = locinput; |
ae0beba1 | 4352 | PL_regsize = 0; |
4aabdb9b DM |
4353 | |
4354 | /* XXXX This is too dramatic a measure... */ | |
4355 | PL_reg_maxiter = 0; | |
4356 | ||
faec1544 | 4357 | ST.toggle_reg_flags = PL_reg_flags; |
3c8556c3 | 4358 | if (RX_UTF8(re_sv)) |
faec1544 DM |
4359 | PL_reg_flags |= RF_utf8; |
4360 | else | |
4361 | PL_reg_flags &= ~RF_utf8; | |
4362 | ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */ | |
4363 | ||
288b8c02 | 4364 | ST.prev_rex = rex_sv; |
faec1544 | 4365 | ST.prev_curlyx = cur_curlyx; |
288b8c02 NC |
4366 | SETREX(rex_sv,re_sv); |
4367 | rex = re; | |
f8fc2ecf | 4368 | rexi = rei; |
faec1544 | 4369 | cur_curlyx = NULL; |
40a82448 | 4370 | ST.B = next; |
faec1544 DM |
4371 | ST.prev_eval = cur_eval; |
4372 | cur_eval = st; | |
faec1544 | 4373 | /* now continue from first node in postoned RE */ |
6bda09f9 | 4374 | PUSH_YES_STATE_GOTO(EVAL_AB, startpoint); |
4aabdb9b | 4375 | /* NOTREACHED */ |
a0ed51b3 | 4376 | } |
24d3c4a9 | 4377 | /* logical is 1, /(?(?{...})X|Y)/ */ |
f2338a2e | 4378 | sw = cBOOL(SvTRUE(ret)); |
24d3c4a9 | 4379 | logical = 0; |
c277df42 IZ |
4380 | break; |
4381 | } | |
40a82448 | 4382 | |
faec1544 DM |
4383 | case EVAL_AB: /* cleanup after a successful (??{A})B */ |
4384 | /* note: this is called twice; first after popping B, then A */ | |
4385 | PL_reg_flags ^= ST.toggle_reg_flags; | |
288b8c02 NC |
4386 | ReREFCNT_dec(rex_sv); |
4387 | SETREX(rex_sv,ST.prev_rex); | |
4388 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 4389 | rexi = RXi_GET(rex); |
faec1544 DM |
4390 | regcpblow(ST.cp); |
4391 | cur_eval = ST.prev_eval; | |
4392 | cur_curlyx = ST.prev_curlyx; | |
34a81e2b B |
4393 | |
4394 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ | |
0357f1fd ML |
4395 | PL_reglastparen = &rex->lastparen; |
4396 | PL_reglastcloseparen = &rex->lastcloseparen; | |
34a81e2b B |
4397 | /* also update PL_regoffs */ |
4398 | PL_regoffs = rex->offs; | |
0357f1fd | 4399 | |
40a82448 DM |
4400 | /* XXXX This is too dramatic a measure... */ |
4401 | PL_reg_maxiter = 0; | |
e7707071 | 4402 | if ( nochange_depth ) |
4b196cd4 | 4403 | nochange_depth--; |
262b90c4 | 4404 | sayYES; |
40a82448 | 4405 | |
40a82448 | 4406 | |
faec1544 DM |
4407 | case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */ |
4408 | /* note: this is called twice; first after popping B, then A */ | |
4409 | PL_reg_flags ^= ST.toggle_reg_flags; | |
288b8c02 NC |
4410 | ReREFCNT_dec(rex_sv); |
4411 | SETREX(rex_sv,ST.prev_rex); | |
4412 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 4413 | rexi = RXi_GET(rex); |
34a81e2b | 4414 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ |
0357f1fd ML |
4415 | PL_reglastparen = &rex->lastparen; |
4416 | PL_reglastcloseparen = &rex->lastcloseparen; | |
4417 | ||
40a82448 DM |
4418 | PL_reginput = locinput; |
4419 | REGCP_UNWIND(ST.lastcp); | |
4420 | regcppop(rex); | |
faec1544 DM |
4421 | cur_eval = ST.prev_eval; |
4422 | cur_curlyx = ST.prev_curlyx; | |
4423 | /* XXXX This is too dramatic a measure... */ | |
4424 | PL_reg_maxiter = 0; | |
e7707071 | 4425 | if ( nochange_depth ) |
4b196cd4 | 4426 | nochange_depth--; |
40a82448 | 4427 | sayNO_SILENT; |
40a82448 DM |
4428 | #undef ST |
4429 | ||
a0d0e21e | 4430 | case OPEN: |
c277df42 | 4431 | n = ARG(scan); /* which paren pair */ |
3280af22 NIS |
4432 | PL_reg_start_tmp[n] = locinput; |
4433 | if (n > PL_regsize) | |
4434 | PL_regsize = n; | |
e2e6a0f1 | 4435 | lastopen = n; |
a0d0e21e LW |
4436 | break; |
4437 | case CLOSE: | |
c277df42 | 4438 | n = ARG(scan); /* which paren pair */ |
f0ab9afb NC |
4439 | PL_regoffs[n].start = PL_reg_start_tmp[n] - PL_bostr; |
4440 | PL_regoffs[n].end = locinput - PL_bostr; | |
7f69552c YO |
4441 | /*if (n > PL_regsize) |
4442 | PL_regsize = n;*/ | |
3b6647e0 | 4443 | if (n > *PL_reglastparen) |
3280af22 | 4444 | *PL_reglastparen = n; |
a01268b5 | 4445 | *PL_reglastcloseparen = n; |
3b6647e0 | 4446 | if (cur_eval && cur_eval->u.eval.close_paren == n) { |
6bda09f9 YO |
4447 | goto fake_end; |
4448 | } | |
a0d0e21e | 4449 | break; |
e2e6a0f1 YO |
4450 | case ACCEPT: |
4451 | if (ARG(scan)){ | |
4452 | regnode *cursor; | |
4453 | for (cursor=scan; | |
4454 | cursor && OP(cursor)!=END; | |
4455 | cursor=regnext(cursor)) | |
4456 | { | |
4457 | if ( OP(cursor)==CLOSE ){ | |
4458 | n = ARG(cursor); | |
4459 | if ( n <= lastopen ) { | |
f0ab9afb NC |
4460 | PL_regoffs[n].start |
4461 | = PL_reg_start_tmp[n] - PL_bostr; | |
4462 | PL_regoffs[n].end = locinput - PL_bostr; | |
e2e6a0f1 YO |
4463 | /*if (n > PL_regsize) |
4464 | PL_regsize = n;*/ | |
3b6647e0 | 4465 | if (n > *PL_reglastparen) |
e2e6a0f1 YO |
4466 | *PL_reglastparen = n; |
4467 | *PL_reglastcloseparen = n; | |
3b6647e0 RB |
4468 | if ( n == ARG(scan) || (cur_eval && |
4469 | cur_eval->u.eval.close_paren == n)) | |
e2e6a0f1 YO |
4470 | break; |
4471 | } | |
4472 | } | |
4473 | } | |
4474 | } | |
4475 | goto fake_end; | |
4476 | /*NOTREACHED*/ | |
c277df42 IZ |
4477 | case GROUPP: |
4478 | n = ARG(scan); /* which paren pair */ | |
f2338a2e | 4479 | sw = cBOOL(*PL_reglastparen >= n && PL_regoffs[n].end != -1); |
c277df42 | 4480 | break; |
0a4db386 YO |
4481 | case NGROUPP: |
4482 | /* reg_check_named_buff_matched returns 0 for no match */ | |
f2338a2e | 4483 | sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan)); |
0a4db386 | 4484 | break; |
1a147d38 | 4485 | case INSUBP: |
0a4db386 | 4486 | n = ARG(scan); |
3b6647e0 | 4487 | sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n)); |
0a4db386 YO |
4488 | break; |
4489 | case DEFINEP: | |
4490 | sw = 0; | |
4491 | break; | |
c277df42 | 4492 | case IFTHEN: |
2c2d71f5 | 4493 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
24d3c4a9 | 4494 | if (sw) |
c277df42 IZ |
4495 | next = NEXTOPER(NEXTOPER(scan)); |
4496 | else { | |
4497 | next = scan + ARG(scan); | |
4498 | if (OP(next) == IFTHEN) /* Fake one. */ | |
4499 | next = NEXTOPER(NEXTOPER(next)); | |
4500 | } | |
4501 | break; | |
4502 | case LOGICAL: | |
24d3c4a9 | 4503 | logical = scan->flags; |
c277df42 | 4504 | break; |
c476f425 | 4505 | |
2ab05381 | 4506 | /******************************************************************* |
2ab05381 | 4507 | |
c476f425 DM |
4508 | The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/ |
4509 | pattern, where A and B are subpatterns. (For simple A, CURLYM or | |
4510 | STAR/PLUS/CURLY/CURLYN are used instead.) | |
2ab05381 | 4511 | |
c476f425 | 4512 | A*B is compiled as <CURLYX><A><WHILEM><B> |
2ab05381 | 4513 | |
c476f425 DM |
4514 | On entry to the subpattern, CURLYX is called. This pushes a CURLYX |
4515 | state, which contains the current count, initialised to -1. It also sets | |
4516 | cur_curlyx to point to this state, with any previous value saved in the | |
4517 | state block. | |
2ab05381 | 4518 | |
c476f425 DM |
4519 | CURLYX then jumps straight to the WHILEM op, rather than executing A, |
4520 | since the pattern may possibly match zero times (i.e. it's a while {} loop | |
4521 | rather than a do {} while loop). | |
2ab05381 | 4522 | |
c476f425 DM |
4523 | Each entry to WHILEM represents a successful match of A. The count in the |
4524 | CURLYX block is incremented, another WHILEM state is pushed, and execution | |
4525 | passes to A or B depending on greediness and the current count. | |
2ab05381 | 4526 | |
c476f425 DM |
4527 | For example, if matching against the string a1a2a3b (where the aN are |
4528 | substrings that match /A/), then the match progresses as follows: (the | |
4529 | pushed states are interspersed with the bits of strings matched so far): | |
2ab05381 | 4530 | |
c476f425 DM |
4531 | <CURLYX cnt=-1> |
4532 | <CURLYX cnt=0><WHILEM> | |
4533 | <CURLYX cnt=1><WHILEM> a1 <WHILEM> | |
4534 | <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM> | |
4535 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> | |
4536 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b | |
2ab05381 | 4537 | |
c476f425 DM |
4538 | (Contrast this with something like CURLYM, which maintains only a single |
4539 | backtrack state: | |
2ab05381 | 4540 | |
c476f425 DM |
4541 | <CURLYM cnt=0> a1 |
4542 | a1 <CURLYM cnt=1> a2 | |
4543 | a1 a2 <CURLYM cnt=2> a3 | |
4544 | a1 a2 a3 <CURLYM cnt=3> b | |
4545 | ) | |
2ab05381 | 4546 | |
c476f425 DM |
4547 | Each WHILEM state block marks a point to backtrack to upon partial failure |
4548 | of A or B, and also contains some minor state data related to that | |
4549 | iteration. The CURLYX block, pointed to by cur_curlyx, contains the | |
4550 | overall state, such as the count, and pointers to the A and B ops. | |
2ab05381 | 4551 | |
c476f425 DM |
4552 | This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx |
4553 | must always point to the *current* CURLYX block, the rules are: | |
2ab05381 | 4554 | |
c476f425 DM |
4555 | When executing CURLYX, save the old cur_curlyx in the CURLYX state block, |
4556 | and set cur_curlyx to point the new block. | |
2ab05381 | 4557 | |
c476f425 DM |
4558 | When popping the CURLYX block after a successful or unsuccessful match, |
4559 | restore the previous cur_curlyx. | |
2ab05381 | 4560 | |
c476f425 DM |
4561 | When WHILEM is about to execute B, save the current cur_curlyx, and set it |
4562 | to the outer one saved in the CURLYX block. | |
2ab05381 | 4563 | |
c476f425 DM |
4564 | When popping the WHILEM block after a successful or unsuccessful B match, |
4565 | restore the previous cur_curlyx. | |
2ab05381 | 4566 | |
c476f425 DM |
4567 | Here's an example for the pattern (AI* BI)*BO |
4568 | I and O refer to inner and outer, C and W refer to CURLYX and WHILEM: | |
2ab05381 | 4569 | |
c476f425 DM |
4570 | cur_ |
4571 | curlyx backtrack stack | |
4572 | ------ --------------- | |
4573 | NULL | |
4574 | CO <CO prev=NULL> <WO> | |
4575 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4576 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi | |
4577 | NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo | |
2ab05381 | 4578 | |
c476f425 DM |
4579 | At this point the pattern succeeds, and we work back down the stack to |
4580 | clean up, restoring as we go: | |
95b24440 | 4581 | |
c476f425 DM |
4582 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi |
4583 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4584 | CO <CO prev=NULL> <WO> | |
4585 | NULL | |
a0374537 | 4586 | |
c476f425 DM |
4587 | *******************************************************************/ |
4588 | ||
4589 | #define ST st->u.curlyx | |
4590 | ||
4591 | case CURLYX: /* start of /A*B/ (for complex A) */ | |
4592 | { | |
4593 | /* No need to save/restore up to this paren */ | |
4594 | I32 parenfloor = scan->flags; | |
4595 | ||
4596 | assert(next); /* keep Coverity happy */ | |
4597 | if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ | |
4598 | next += ARG(next); | |
4599 | ||
4600 | /* XXXX Probably it is better to teach regpush to support | |
4601 | parenfloor > PL_regsize... */ | |
4602 | if (parenfloor > (I32)*PL_reglastparen) | |
4603 | parenfloor = *PL_reglastparen; /* Pessimization... */ | |
4604 | ||
4605 | ST.prev_curlyx= cur_curlyx; | |
4606 | cur_curlyx = st; | |
4607 | ST.cp = PL_savestack_ix; | |
4608 | ||
4609 | /* these fields contain the state of the current curly. | |
4610 | * they are accessed by subsequent WHILEMs */ | |
4611 | ST.parenfloor = parenfloor; | |
d02d6d97 | 4612 | ST.me = scan; |
c476f425 | 4613 | ST.B = next; |
24d3c4a9 DM |
4614 | ST.minmod = minmod; |
4615 | minmod = 0; | |
c476f425 DM |
4616 | ST.count = -1; /* this will be updated by WHILEM */ |
4617 | ST.lastloc = NULL; /* this will be updated by WHILEM */ | |
4618 | ||
4619 | PL_reginput = locinput; | |
4620 | PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next)); | |
5f66b61c | 4621 | /* NOTREACHED */ |
c476f425 | 4622 | } |
a0d0e21e | 4623 | |
c476f425 | 4624 | case CURLYX_end: /* just finished matching all of A*B */ |
c476f425 DM |
4625 | cur_curlyx = ST.prev_curlyx; |
4626 | sayYES; | |
4627 | /* NOTREACHED */ | |
a0d0e21e | 4628 | |
c476f425 DM |
4629 | case CURLYX_end_fail: /* just failed to match all of A*B */ |
4630 | regcpblow(ST.cp); | |
4631 | cur_curlyx = ST.prev_curlyx; | |
4632 | sayNO; | |
4633 | /* NOTREACHED */ | |
4633a7c4 | 4634 | |
a0d0e21e | 4635 | |
c476f425 DM |
4636 | #undef ST |
4637 | #define ST st->u.whilem | |
4638 | ||
4639 | case WHILEM: /* just matched an A in /A*B/ (for complex A) */ | |
4640 | { | |
4641 | /* see the discussion above about CURLYX/WHILEM */ | |
c476f425 | 4642 | I32 n; |
d02d6d97 DM |
4643 | int min = ARG1(cur_curlyx->u.curlyx.me); |
4644 | int max = ARG2(cur_curlyx->u.curlyx.me); | |
4645 | regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS; | |
4646 | ||
c476f425 DM |
4647 | assert(cur_curlyx); /* keep Coverity happy */ |
4648 | n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */ | |
4649 | ST.save_lastloc = cur_curlyx->u.curlyx.lastloc; | |
4650 | ST.cache_offset = 0; | |
4651 | ST.cache_mask = 0; | |
4652 | ||
4653 | PL_reginput = locinput; | |
4654 | ||
4655 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
d02d6d97 DM |
4656 | "%*s whilem: matched %ld out of %d..%d\n", |
4657 | REPORT_CODE_OFF+depth*2, "", (long)n, min, max) | |
c476f425 | 4658 | ); |
a0d0e21e | 4659 | |
c476f425 | 4660 | /* First just match a string of min A's. */ |
a0d0e21e | 4661 | |
d02d6d97 | 4662 | if (n < min) { |
c476f425 | 4663 | cur_curlyx->u.curlyx.lastloc = locinput; |
d02d6d97 | 4664 | PUSH_STATE_GOTO(WHILEM_A_pre, A); |
c476f425 DM |
4665 | /* NOTREACHED */ |
4666 | } | |
4667 | ||
4668 | /* If degenerate A matches "", assume A done. */ | |
4669 | ||
4670 | if (locinput == cur_curlyx->u.curlyx.lastloc) { | |
4671 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4672 | "%*s whilem: empty match detected, trying continuation...\n", | |
4673 | REPORT_CODE_OFF+depth*2, "") | |
4674 | ); | |
4675 | goto do_whilem_B_max; | |
4676 | } | |
4677 | ||
4678 | /* super-linear cache processing */ | |
4679 | ||
4680 | if (scan->flags) { | |
a0d0e21e | 4681 | |
2c2d71f5 | 4682 | if (!PL_reg_maxiter) { |
c476f425 DM |
4683 | /* start the countdown: Postpone detection until we |
4684 | * know the match is not *that* much linear. */ | |
2c2d71f5 | 4685 | PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4); |
66bf836d DM |
4686 | /* possible overflow for long strings and many CURLYX's */ |
4687 | if (PL_reg_maxiter < 0) | |
4688 | PL_reg_maxiter = I32_MAX; | |
2c2d71f5 JH |
4689 | PL_reg_leftiter = PL_reg_maxiter; |
4690 | } | |
c476f425 | 4691 | |
2c2d71f5 | 4692 | if (PL_reg_leftiter-- == 0) { |
c476f425 | 4693 | /* initialise cache */ |
3298f257 | 4694 | const I32 size = (PL_reg_maxiter + 7)/8; |
2c2d71f5 | 4695 | if (PL_reg_poscache) { |
eb160463 | 4696 | if ((I32)PL_reg_poscache_size < size) { |
2c2d71f5 JH |
4697 | Renew(PL_reg_poscache, size, char); |
4698 | PL_reg_poscache_size = size; | |
4699 | } | |
4700 | Zero(PL_reg_poscache, size, char); | |
4701 | } | |
4702 | else { | |
4703 | PL_reg_poscache_size = size; | |
a02a5408 | 4704 | Newxz(PL_reg_poscache, size, char); |
2c2d71f5 | 4705 | } |
c476f425 DM |
4706 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
4707 | "%swhilem: Detected a super-linear match, switching on caching%s...\n", | |
4708 | PL_colors[4], PL_colors[5]) | |
4709 | ); | |
2c2d71f5 | 4710 | } |
c476f425 | 4711 | |
2c2d71f5 | 4712 | if (PL_reg_leftiter < 0) { |
c476f425 DM |
4713 | /* have we already failed at this position? */ |
4714 | I32 offset, mask; | |
4715 | offset = (scan->flags & 0xf) - 1 | |
4716 | + (locinput - PL_bostr) * (scan->flags>>4); | |
4717 | mask = 1 << (offset % 8); | |
4718 | offset /= 8; | |
4719 | if (PL_reg_poscache[offset] & mask) { | |
4720 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4721 | "%*s whilem: (cache) already tried at this position...\n", | |
4722 | REPORT_CODE_OFF+depth*2, "") | |
2c2d71f5 | 4723 | ); |
3298f257 | 4724 | sayNO; /* cache records failure */ |
2c2d71f5 | 4725 | } |
c476f425 DM |
4726 | ST.cache_offset = offset; |
4727 | ST.cache_mask = mask; | |
2c2d71f5 | 4728 | } |
c476f425 | 4729 | } |
2c2d71f5 | 4730 | |
c476f425 | 4731 | /* Prefer B over A for minimal matching. */ |
a687059c | 4732 | |
c476f425 DM |
4733 | if (cur_curlyx->u.curlyx.minmod) { |
4734 | ST.save_curlyx = cur_curlyx; | |
4735 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4736 | ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor); | |
4737 | REGCP_SET(ST.lastcp); | |
4738 | PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B); | |
4739 | /* NOTREACHED */ | |
4740 | } | |
a0d0e21e | 4741 | |
c476f425 DM |
4742 | /* Prefer A over B for maximal matching. */ |
4743 | ||
d02d6d97 | 4744 | if (n < max) { /* More greed allowed? */ |
c476f425 DM |
4745 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); |
4746 | cur_curlyx->u.curlyx.lastloc = locinput; | |
4747 | REGCP_SET(ST.lastcp); | |
d02d6d97 | 4748 | PUSH_STATE_GOTO(WHILEM_A_max, A); |
c476f425 DM |
4749 | /* NOTREACHED */ |
4750 | } | |
4751 | goto do_whilem_B_max; | |
4752 | } | |
4753 | /* NOTREACHED */ | |
4754 | ||
4755 | case WHILEM_B_min: /* just matched B in a minimal match */ | |
4756 | case WHILEM_B_max: /* just matched B in a maximal match */ | |
4757 | cur_curlyx = ST.save_curlyx; | |
4758 | sayYES; | |
4759 | /* NOTREACHED */ | |
4760 | ||
4761 | case WHILEM_B_max_fail: /* just failed to match B in a maximal match */ | |
4762 | cur_curlyx = ST.save_curlyx; | |
4763 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
4764 | cur_curlyx->u.curlyx.count--; | |
4765 | CACHEsayNO; | |
4766 | /* NOTREACHED */ | |
4767 | ||
4768 | case WHILEM_A_min_fail: /* just failed to match A in a minimal match */ | |
4769 | REGCP_UNWIND(ST.lastcp); | |
4770 | regcppop(rex); | |
4771 | /* FALL THROUGH */ | |
4772 | case WHILEM_A_pre_fail: /* just failed to match even minimal A */ | |
4773 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
4774 | cur_curlyx->u.curlyx.count--; | |
4775 | CACHEsayNO; | |
4776 | /* NOTREACHED */ | |
4777 | ||
4778 | case WHILEM_A_max_fail: /* just failed to match A in a maximal match */ | |
4779 | REGCP_UNWIND(ST.lastcp); | |
4780 | regcppop(rex); /* Restore some previous $<digit>s? */ | |
4781 | PL_reginput = locinput; | |
4782 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
4783 | "%*s whilem: failed, trying continuation...\n", | |
4784 | REPORT_CODE_OFF+depth*2, "") | |
4785 | ); | |
4786 | do_whilem_B_max: | |
4787 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
4788 | && ckWARN(WARN_REGEXP) | |
4789 | && !(PL_reg_flags & RF_warned)) | |
4790 | { | |
4791 | PL_reg_flags |= RF_warned; | |
4792 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded", | |
4793 | "Complex regular subexpression recursion", | |
4794 | REG_INFTY - 1); | |
4795 | } | |
4796 | ||
4797 | /* now try B */ | |
4798 | ST.save_curlyx = cur_curlyx; | |
4799 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4800 | PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B); | |
4801 | /* NOTREACHED */ | |
4802 | ||
4803 | case WHILEM_B_min_fail: /* just failed to match B in a minimal match */ | |
4804 | cur_curlyx = ST.save_curlyx; | |
4805 | REGCP_UNWIND(ST.lastcp); | |
4806 | regcppop(rex); | |
4807 | ||
d02d6d97 | 4808 | if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) { |
c476f425 DM |
4809 | /* Maximum greed exceeded */ |
4810 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
4811 | && ckWARN(WARN_REGEXP) | |
4812 | && !(PL_reg_flags & RF_warned)) | |
4813 | { | |
3280af22 | 4814 | PL_reg_flags |= RF_warned; |
c476f425 DM |
4815 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
4816 | "%s limit (%d) exceeded", | |
4817 | "Complex regular subexpression recursion", | |
4818 | REG_INFTY - 1); | |
a0d0e21e | 4819 | } |
c476f425 | 4820 | cur_curlyx->u.curlyx.count--; |
3ab3c9b4 | 4821 | CACHEsayNO; |
a0d0e21e | 4822 | } |
c476f425 DM |
4823 | |
4824 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
4825 | "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "") | |
4826 | ); | |
4827 | /* Try grabbing another A and see if it helps. */ | |
4828 | PL_reginput = locinput; | |
4829 | cur_curlyx->u.curlyx.lastloc = locinput; | |
4830 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); | |
4831 | REGCP_SET(ST.lastcp); | |
d02d6d97 DM |
4832 | PUSH_STATE_GOTO(WHILEM_A_min, |
4833 | /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS); | |
5f66b61c | 4834 | /* NOTREACHED */ |
40a82448 DM |
4835 | |
4836 | #undef ST | |
4837 | #define ST st->u.branch | |
4838 | ||
4839 | case BRANCHJ: /* /(...|A|...)/ with long next pointer */ | |
c277df42 IZ |
4840 | next = scan + ARG(scan); |
4841 | if (next == scan) | |
4842 | next = NULL; | |
40a82448 DM |
4843 | scan = NEXTOPER(scan); |
4844 | /* FALL THROUGH */ | |
c277df42 | 4845 | |
40a82448 DM |
4846 | case BRANCH: /* /(...|A|...)/ */ |
4847 | scan = NEXTOPER(scan); /* scan now points to inner node */ | |
40a82448 DM |
4848 | ST.lastparen = *PL_reglastparen; |
4849 | ST.next_branch = next; | |
4850 | REGCP_SET(ST.cp); | |
4851 | PL_reginput = locinput; | |
02db2b7b | 4852 | |
40a82448 | 4853 | /* Now go into the branch */ |
5d458dd8 YO |
4854 | if (has_cutgroup) { |
4855 | PUSH_YES_STATE_GOTO(BRANCH_next, scan); | |
4856 | } else { | |
4857 | PUSH_STATE_GOTO(BRANCH_next, scan); | |
4858 | } | |
40a82448 | 4859 | /* NOTREACHED */ |
5d458dd8 YO |
4860 | case CUTGROUP: |
4861 | PL_reginput = locinput; | |
4862 | sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : | |
ad64d0ec | 4863 | MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
4864 | PUSH_STATE_GOTO(CUTGROUP_next,next); |
4865 | /* NOTREACHED */ | |
4866 | case CUTGROUP_next_fail: | |
4867 | do_cutgroup = 1; | |
4868 | no_final = 1; | |
4869 | if (st->u.mark.mark_name) | |
4870 | sv_commit = st->u.mark.mark_name; | |
4871 | sayNO; | |
4872 | /* NOTREACHED */ | |
4873 | case BRANCH_next: | |
4874 | sayYES; | |
4875 | /* NOTREACHED */ | |
40a82448 | 4876 | case BRANCH_next_fail: /* that branch failed; try the next, if any */ |
5d458dd8 YO |
4877 | if (do_cutgroup) { |
4878 | do_cutgroup = 0; | |
4879 | no_final = 0; | |
4880 | } | |
40a82448 DM |
4881 | REGCP_UNWIND(ST.cp); |
4882 | for (n = *PL_reglastparen; n > ST.lastparen; n--) | |
f0ab9afb | 4883 | PL_regoffs[n].end = -1; |
40a82448 | 4884 | *PL_reglastparen = n; |
0a4db386 | 4885 | /*dmq: *PL_reglastcloseparen = n; */ |
40a82448 DM |
4886 | scan = ST.next_branch; |
4887 | /* no more branches? */ | |
5d458dd8 YO |
4888 | if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) { |
4889 | DEBUG_EXECUTE_r({ | |
4890 | PerlIO_printf( Perl_debug_log, | |
4891 | "%*s %sBRANCH failed...%s\n", | |
4892 | REPORT_CODE_OFF+depth*2, "", | |
4893 | PL_colors[4], | |
4894 | PL_colors[5] ); | |
4895 | }); | |
4896 | sayNO_SILENT; | |
4897 | } | |
40a82448 DM |
4898 | continue; /* execute next BRANCH[J] op */ |
4899 | /* NOTREACHED */ | |
4900 | ||
a0d0e21e | 4901 | case MINMOD: |
24d3c4a9 | 4902 | minmod = 1; |
a0d0e21e | 4903 | break; |
40a82448 DM |
4904 | |
4905 | #undef ST | |
4906 | #define ST st->u.curlym | |
4907 | ||
4908 | case CURLYM: /* /A{m,n}B/ where A is fixed-length */ | |
4909 | ||
4910 | /* This is an optimisation of CURLYX that enables us to push | |
84d2fa14 | 4911 | * only a single backtracking state, no matter how many matches |
40a82448 DM |
4912 | * there are in {m,n}. It relies on the pattern being constant |
4913 | * length, with no parens to influence future backrefs | |
4914 | */ | |
4915 | ||
4916 | ST.me = scan; | |
dc45a647 | 4917 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; |
40a82448 DM |
4918 | |
4919 | /* if paren positive, emulate an OPEN/CLOSE around A */ | |
4920 | if (ST.me->flags) { | |
3b6647e0 | 4921 | U32 paren = ST.me->flags; |
40a82448 DM |
4922 | if (paren > PL_regsize) |
4923 | PL_regsize = paren; | |
3b6647e0 | 4924 | if (paren > *PL_reglastparen) |
40a82448 | 4925 | *PL_reglastparen = paren; |
c277df42 | 4926 | scan += NEXT_OFF(scan); /* Skip former OPEN. */ |
6407bf3b | 4927 | } |
40a82448 DM |
4928 | ST.A = scan; |
4929 | ST.B = next; | |
4930 | ST.alen = 0; | |
4931 | ST.count = 0; | |
24d3c4a9 DM |
4932 | ST.minmod = minmod; |
4933 | minmod = 0; | |
40a82448 DM |
4934 | ST.c1 = CHRTEST_UNINIT; |
4935 | REGCP_SET(ST.cp); | |
6407bf3b | 4936 | |
40a82448 DM |
4937 | if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */ |
4938 | goto curlym_do_B; | |
4939 | ||
4940 | curlym_do_A: /* execute the A in /A{m,n}B/ */ | |
6407bf3b | 4941 | PL_reginput = locinput; |
40a82448 DM |
4942 | PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */ |
4943 | /* NOTREACHED */ | |
5f80c4cf | 4944 | |
40a82448 DM |
4945 | case CURLYM_A: /* we've just matched an A */ |
4946 | locinput = st->locinput; | |
4947 | nextchr = UCHARAT(locinput); | |
4948 | ||
4949 | ST.count++; | |
4950 | /* after first match, determine A's length: u.curlym.alen */ | |
4951 | if (ST.count == 1) { | |
4952 | if (PL_reg_match_utf8) { | |
4953 | char *s = locinput; | |
4954 | while (s < PL_reginput) { | |
4955 | ST.alen++; | |
4956 | s += UTF8SKIP(s); | |
4957 | } | |
4958 | } | |
4959 | else { | |
4960 | ST.alen = PL_reginput - locinput; | |
4961 | } | |
4962 | if (ST.alen == 0) | |
4963 | ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me); | |
4964 | } | |
0cadcf80 DM |
4965 | DEBUG_EXECUTE_r( |
4966 | PerlIO_printf(Perl_debug_log, | |
40a82448 | 4967 | "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n", |
5bc10b2c | 4968 | (int)(REPORT_CODE_OFF+(depth*2)), "", |
40a82448 | 4969 | (IV) ST.count, (IV)ST.alen) |
0cadcf80 DM |
4970 | ); |
4971 | ||
40a82448 | 4972 | locinput = PL_reginput; |
0a4db386 YO |
4973 | |
4974 | if (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 4975 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
4976 | goto fake_end; |
4977 | ||
c966426a DM |
4978 | { |
4979 | I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)); | |
4980 | if ( max == REG_INFTY || ST.count < max ) | |
4981 | goto curlym_do_A; /* try to match another A */ | |
4982 | } | |
40a82448 | 4983 | goto curlym_do_B; /* try to match B */ |
5f80c4cf | 4984 | |
40a82448 DM |
4985 | case CURLYM_A_fail: /* just failed to match an A */ |
4986 | REGCP_UNWIND(ST.cp); | |
0a4db386 YO |
4987 | |
4988 | if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ | |
4989 | || (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 4990 | cur_eval->u.eval.close_paren == (U32)ST.me->flags)) |
40a82448 | 4991 | sayNO; |
0cadcf80 | 4992 | |
40a82448 DM |
4993 | curlym_do_B: /* execute the B in /A{m,n}B/ */ |
4994 | PL_reginput = locinput; | |
4995 | if (ST.c1 == CHRTEST_UNINIT) { | |
4996 | /* calculate c1 and c2 for possible match of 1st char | |
4997 | * following curly */ | |
4998 | ST.c1 = ST.c2 = CHRTEST_VOID; | |
4999 | if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) { | |
5000 | regnode *text_node = ST.B; | |
5001 | if (! HAS_TEXT(text_node)) | |
5002 | FIND_NEXT_IMPT(text_node); | |
ee9b8eae YO |
5003 | /* this used to be |
5004 | ||
5005 | (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT) | |
5006 | ||
5007 | But the former is redundant in light of the latter. | |
5008 | ||
5009 | if this changes back then the macro for | |
5010 | IS_TEXT and friends need to change. | |
5011 | */ | |
5012 | if (PL_regkind[OP(text_node)] == EXACT) | |
40a82448 | 5013 | { |
ee9b8eae | 5014 | |
40a82448 | 5015 | ST.c1 = (U8)*STRING(text_node); |
9a5a5549 KW |
5016 | switch (OP(text_node)) { |
5017 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
2f7f8cb1 | 5018 | case EXACTFA: |
9a5a5549 KW |
5019 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; |
5020 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
5021 | default: ST.c2 = ST.c1; | |
5022 | } | |
c277df42 | 5023 | } |
c277df42 | 5024 | } |
40a82448 DM |
5025 | } |
5026 | ||
5027 | DEBUG_EXECUTE_r( | |
5028 | PerlIO_printf(Perl_debug_log, | |
5029 | "%*s CURLYM trying tail with matches=%"IVdf"...\n", | |
5bc10b2c | 5030 | (int)(REPORT_CODE_OFF+(depth*2)), |
40a82448 DM |
5031 | "", (IV)ST.count) |
5032 | ); | |
5033 | if (ST.c1 != CHRTEST_VOID | |
5034 | && UCHARAT(PL_reginput) != ST.c1 | |
5035 | && UCHARAT(PL_reginput) != ST.c2) | |
5036 | { | |
5037 | /* simulate B failing */ | |
3e901dc0 YO |
5038 | DEBUG_OPTIMISE_r( |
5039 | PerlIO_printf(Perl_debug_log, | |
5040 | "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n", | |
5041 | (int)(REPORT_CODE_OFF+(depth*2)),"", | |
5042 | (IV)ST.c1,(IV)ST.c2 | |
5043 | )); | |
40a82448 DM |
5044 | state_num = CURLYM_B_fail; |
5045 | goto reenter_switch; | |
5046 | } | |
5047 | ||
5048 | if (ST.me->flags) { | |
5049 | /* mark current A as captured */ | |
5050 | I32 paren = ST.me->flags; | |
5051 | if (ST.count) { | |
f0ab9afb | 5052 | PL_regoffs[paren].start |
40a82448 | 5053 | = HOPc(PL_reginput, -ST.alen) - PL_bostr; |
f0ab9afb | 5054 | PL_regoffs[paren].end = PL_reginput - PL_bostr; |
0a4db386 | 5055 | /*dmq: *PL_reglastcloseparen = paren; */ |
c277df42 | 5056 | } |
40a82448 | 5057 | else |
f0ab9afb | 5058 | PL_regoffs[paren].end = -1; |
0a4db386 | 5059 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 5060 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5061 | { |
5062 | if (ST.count) | |
5063 | goto fake_end; | |
5064 | else | |
5065 | sayNO; | |
5066 | } | |
c277df42 | 5067 | } |
0a4db386 | 5068 | |
40a82448 | 5069 | PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */ |
5f66b61c | 5070 | /* NOTREACHED */ |
40a82448 DM |
5071 | |
5072 | case CURLYM_B_fail: /* just failed to match a B */ | |
5073 | REGCP_UNWIND(ST.cp); | |
5074 | if (ST.minmod) { | |
84d2fa14 HS |
5075 | I32 max = ARG2(ST.me); |
5076 | if (max != REG_INFTY && ST.count == max) | |
40a82448 DM |
5077 | sayNO; |
5078 | goto curlym_do_A; /* try to match a further A */ | |
5079 | } | |
5080 | /* backtrack one A */ | |
5081 | if (ST.count == ARG1(ST.me) /* min */) | |
5082 | sayNO; | |
5083 | ST.count--; | |
5084 | locinput = HOPc(locinput, -ST.alen); | |
5085 | goto curlym_do_B; /* try to match B */ | |
5086 | ||
c255a977 DM |
5087 | #undef ST |
5088 | #define ST st->u.curly | |
40a82448 | 5089 | |
c255a977 DM |
5090 | #define CURLY_SETPAREN(paren, success) \ |
5091 | if (paren) { \ | |
5092 | if (success) { \ | |
f0ab9afb NC |
5093 | PL_regoffs[paren].start = HOPc(locinput, -1) - PL_bostr; \ |
5094 | PL_regoffs[paren].end = locinput - PL_bostr; \ | |
0a4db386 | 5095 | *PL_reglastcloseparen = paren; \ |
c255a977 DM |
5096 | } \ |
5097 | else \ | |
f0ab9afb | 5098 | PL_regoffs[paren].end = -1; \ |
c255a977 DM |
5099 | } |
5100 | ||
5101 | case STAR: /* /A*B/ where A is width 1 */ | |
5102 | ST.paren = 0; | |
5103 | ST.min = 0; | |
5104 | ST.max = REG_INFTY; | |
a0d0e21e LW |
5105 | scan = NEXTOPER(scan); |
5106 | goto repeat; | |
c255a977 DM |
5107 | case PLUS: /* /A+B/ where A is width 1 */ |
5108 | ST.paren = 0; | |
5109 | ST.min = 1; | |
5110 | ST.max = REG_INFTY; | |
c277df42 | 5111 | scan = NEXTOPER(scan); |
c255a977 DM |
5112 | goto repeat; |
5113 | case CURLYN: /* /(A){m,n}B/ where A is width 1 */ | |
5114 | ST.paren = scan->flags; /* Which paren to set */ | |
5115 | if (ST.paren > PL_regsize) | |
5116 | PL_regsize = ST.paren; | |
3b6647e0 | 5117 | if (ST.paren > *PL_reglastparen) |
c255a977 DM |
5118 | *PL_reglastparen = ST.paren; |
5119 | ST.min = ARG1(scan); /* min to match */ | |
5120 | ST.max = ARG2(scan); /* max to match */ | |
0a4db386 | 5121 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5122 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5123 | ST.min=1; |
5124 | ST.max=1; | |
5125 | } | |
c255a977 DM |
5126 | scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); |
5127 | goto repeat; | |
5128 | case CURLY: /* /A{m,n}B/ where A is width 1 */ | |
5129 | ST.paren = 0; | |
5130 | ST.min = ARG1(scan); /* min to match */ | |
5131 | ST.max = ARG2(scan); /* max to match */ | |
5132 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; | |
c277df42 | 5133 | repeat: |
a0d0e21e LW |
5134 | /* |
5135 | * Lookahead to avoid useless match attempts | |
5136 | * when we know what character comes next. | |
c255a977 | 5137 | * |
5f80c4cf JP |
5138 | * Used to only do .*x and .*?x, but now it allows |
5139 | * for )'s, ('s and (?{ ... })'s to be in the way | |
5140 | * of the quantifier and the EXACT-like node. -- japhy | |
5141 | */ | |
5142 | ||
c255a977 DM |
5143 | if (ST.min > ST.max) /* XXX make this a compile-time check? */ |
5144 | sayNO; | |
cca55fe3 | 5145 | if (HAS_TEXT(next) || JUMPABLE(next)) { |
5f80c4cf JP |
5146 | U8 *s; |
5147 | regnode *text_node = next; | |
5148 | ||
3dab1dad YO |
5149 | if (! HAS_TEXT(text_node)) |
5150 | FIND_NEXT_IMPT(text_node); | |
5f80c4cf | 5151 | |
9e137952 | 5152 | if (! HAS_TEXT(text_node)) |
c255a977 | 5153 | ST.c1 = ST.c2 = CHRTEST_VOID; |
5f80c4cf | 5154 | else { |
ee9b8eae | 5155 | if ( PL_regkind[OP(text_node)] != EXACT ) { |
c255a977 | 5156 | ST.c1 = ST.c2 = CHRTEST_VOID; |
44a68960 | 5157 | goto assume_ok_easy; |
cca55fe3 | 5158 | } |
be8e71aa YO |
5159 | else |
5160 | s = (U8*)STRING(text_node); | |
ee9b8eae YO |
5161 | |
5162 | /* Currently we only get here when | |
5163 | ||
5164 | PL_rekind[OP(text_node)] == EXACT | |
5165 | ||
5166 | if this changes back then the macro for IS_TEXT and | |
5167 | friends need to change. */ | |
f2ed9b32 | 5168 | if (!UTF_PATTERN) { |
9a5a5549 KW |
5169 | ST.c1 = *s; |
5170 | switch (OP(text_node)) { | |
5171 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
2f7f8cb1 | 5172 | case EXACTFA: |
9a5a5549 KW |
5173 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; |
5174 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
5175 | default: ST.c2 = ST.c1; break; | |
5176 | } | |
1aa99e6b | 5177 | } |
f2ed9b32 | 5178 | else { /* UTF_PATTERN */ |
9a5a5549 | 5179 | if (IS_TEXTFU(text_node) || IS_TEXTF(text_node)) { |
a2a2844f | 5180 | STRLEN ulen1, ulen2; |
89ebb4a3 JH |
5181 | U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; |
5182 | U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; | |
a2a2844f JH |
5183 | |
5184 | to_utf8_lower((U8*)s, tmpbuf1, &ulen1); | |
5185 | to_utf8_upper((U8*)s, tmpbuf2, &ulen2); | |
e294cc5d JH |
5186 | #ifdef EBCDIC |
5187 | ST.c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN, 0, | |
5188 | ckWARN(WARN_UTF8) ? | |
5189 | 0 : UTF8_ALLOW_ANY); | |
5190 | ST.c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN, 0, | |
5191 | ckWARN(WARN_UTF8) ? | |
5192 | 0 : UTF8_ALLOW_ANY); | |
5193 | #else | |
c255a977 | 5194 | ST.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0, |
e294cc5d | 5195 | uniflags); |
c255a977 | 5196 | ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0, |
e294cc5d JH |
5197 | uniflags); |
5198 | #endif | |
5f80c4cf JP |
5199 | } |
5200 | else { | |
c255a977 | 5201 | ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, |
041457d9 | 5202 | uniflags); |
5f80c4cf | 5203 | } |
1aa99e6b IH |
5204 | } |
5205 | } | |
bbce6d69 | 5206 | } |
a0d0e21e | 5207 | else |
c255a977 | 5208 | ST.c1 = ST.c2 = CHRTEST_VOID; |
cca55fe3 | 5209 | assume_ok_easy: |
c255a977 DM |
5210 | |
5211 | ST.A = scan; | |
5212 | ST.B = next; | |
3280af22 | 5213 | PL_reginput = locinput; |
24d3c4a9 DM |
5214 | if (minmod) { |
5215 | minmod = 0; | |
e2e6a0f1 | 5216 | if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min) |
4633a7c4 | 5217 | sayNO; |
c255a977 | 5218 | ST.count = ST.min; |
a0ed51b3 | 5219 | locinput = PL_reginput; |
c255a977 DM |
5220 | REGCP_SET(ST.cp); |
5221 | if (ST.c1 == CHRTEST_VOID) | |
5222 | goto curly_try_B_min; | |
5223 | ||
5224 | ST.oldloc = locinput; | |
5225 | ||
5226 | /* set ST.maxpos to the furthest point along the | |
5227 | * string that could possibly match */ | |
5228 | if (ST.max == REG_INFTY) { | |
5229 | ST.maxpos = PL_regeol - 1; | |
f2ed9b32 | 5230 | if (utf8_target) |
c255a977 DM |
5231 | while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos)) |
5232 | ST.maxpos--; | |
5233 | } | |
f2ed9b32 | 5234 | else if (utf8_target) { |
c255a977 DM |
5235 | int m = ST.max - ST.min; |
5236 | for (ST.maxpos = locinput; | |
5237 | m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--) | |
5238 | ST.maxpos += UTF8SKIP(ST.maxpos); | |
5239 | } | |
5240 | else { | |
5241 | ST.maxpos = locinput + ST.max - ST.min; | |
5242 | if (ST.maxpos >= PL_regeol) | |
5243 | ST.maxpos = PL_regeol - 1; | |
5244 | } | |
5245 | goto curly_try_B_min_known; | |
5246 | ||
5247 | } | |
5248 | else { | |
e2e6a0f1 | 5249 | ST.count = regrepeat(rex, ST.A, ST.max, depth); |
c255a977 DM |
5250 | locinput = PL_reginput; |
5251 | if (ST.count < ST.min) | |
5252 | sayNO; | |
5253 | if ((ST.count > ST.min) | |
5254 | && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL)) | |
5255 | { | |
5256 | /* A{m,n} must come at the end of the string, there's | |
5257 | * no point in backing off ... */ | |
5258 | ST.min = ST.count; | |
5259 | /* ...except that $ and \Z can match before *and* after | |
5260 | newline at the end. Consider "\n\n" =~ /\n+\Z\n/. | |
5261 | We may back off by one in this case. */ | |
5262 | if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS) | |
5263 | ST.min--; | |
5264 | } | |
5265 | REGCP_SET(ST.cp); | |
5266 | goto curly_try_B_max; | |
5267 | } | |
5268 | /* NOTREACHED */ | |
5269 | ||
5270 | ||
5271 | case CURLY_B_min_known_fail: | |
5272 | /* failed to find B in a non-greedy match where c1,c2 valid */ | |
5273 | if (ST.paren && ST.count) | |
f0ab9afb | 5274 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5275 | |
5276 | PL_reginput = locinput; /* Could be reset... */ | |
5277 | REGCP_UNWIND(ST.cp); | |
5278 | /* Couldn't or didn't -- move forward. */ | |
5279 | ST.oldloc = locinput; | |
f2ed9b32 | 5280 | if (utf8_target) |
c255a977 DM |
5281 | locinput += UTF8SKIP(locinput); |
5282 | else | |
5283 | locinput++; | |
5284 | ST.count++; | |
5285 | curly_try_B_min_known: | |
5286 | /* find the next place where 'B' could work, then call B */ | |
5287 | { | |
5288 | int n; | |
f2ed9b32 | 5289 | if (utf8_target) { |
c255a977 DM |
5290 | n = (ST.oldloc == locinput) ? 0 : 1; |
5291 | if (ST.c1 == ST.c2) { | |
5292 | STRLEN len; | |
5293 | /* set n to utf8_distance(oldloc, locinput) */ | |
5294 | while (locinput <= ST.maxpos && | |
5295 | utf8n_to_uvchr((U8*)locinput, | |
5296 | UTF8_MAXBYTES, &len, | |
5297 | uniflags) != (UV)ST.c1) { | |
5298 | locinput += len; | |
5299 | n++; | |
5300 | } | |
1aa99e6b IH |
5301 | } |
5302 | else { | |
c255a977 DM |
5303 | /* set n to utf8_distance(oldloc, locinput) */ |
5304 | while (locinput <= ST.maxpos) { | |
5305 | STRLEN len; | |
5306 | const UV c = utf8n_to_uvchr((U8*)locinput, | |
5307 | UTF8_MAXBYTES, &len, | |
5308 | uniflags); | |
5309 | if (c == (UV)ST.c1 || c == (UV)ST.c2) | |
5310 | break; | |
5311 | locinput += len; | |
5312 | n++; | |
1aa99e6b | 5313 | } |
0fe9bf95 IZ |
5314 | } |
5315 | } | |
c255a977 DM |
5316 | else { |
5317 | if (ST.c1 == ST.c2) { | |
5318 | while (locinput <= ST.maxpos && | |
5319 | UCHARAT(locinput) != ST.c1) | |
5320 | locinput++; | |
bbce6d69 | 5321 | } |
c255a977 DM |
5322 | else { |
5323 | while (locinput <= ST.maxpos | |
5324 | && UCHARAT(locinput) != ST.c1 | |
5325 | && UCHARAT(locinput) != ST.c2) | |
5326 | locinput++; | |
a0ed51b3 | 5327 | } |
c255a977 DM |
5328 | n = locinput - ST.oldloc; |
5329 | } | |
5330 | if (locinput > ST.maxpos) | |
5331 | sayNO; | |
5332 | /* PL_reginput == oldloc now */ | |
5333 | if (n) { | |
5334 | ST.count += n; | |
e2e6a0f1 | 5335 | if (regrepeat(rex, ST.A, n, depth) < n) |
4633a7c4 | 5336 | sayNO; |
a0d0e21e | 5337 | } |
c255a977 DM |
5338 | PL_reginput = locinput; |
5339 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5340 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5341 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5342 | goto fake_end; |
5343 | } | |
c255a977 | 5344 | PUSH_STATE_GOTO(CURLY_B_min_known, ST.B); |
a0d0e21e | 5345 | } |
c255a977 DM |
5346 | /* NOTREACHED */ |
5347 | ||
5348 | ||
5349 | case CURLY_B_min_fail: | |
5350 | /* failed to find B in a non-greedy match where c1,c2 invalid */ | |
5351 | if (ST.paren && ST.count) | |
f0ab9afb | 5352 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5353 | |
5354 | REGCP_UNWIND(ST.cp); | |
5355 | /* failed -- move forward one */ | |
5356 | PL_reginput = locinput; | |
e2e6a0f1 | 5357 | if (regrepeat(rex, ST.A, 1, depth)) { |
c255a977 | 5358 | ST.count++; |
a0ed51b3 | 5359 | locinput = PL_reginput; |
c255a977 DM |
5360 | if (ST.count <= ST.max || (ST.max == REG_INFTY && |
5361 | ST.count > 0)) /* count overflow ? */ | |
15272685 | 5362 | { |
c255a977 DM |
5363 | curly_try_B_min: |
5364 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5365 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5366 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5367 | goto fake_end; |
5368 | } | |
c255a977 | 5369 | PUSH_STATE_GOTO(CURLY_B_min, ST.B); |
a0d0e21e LW |
5370 | } |
5371 | } | |
4633a7c4 | 5372 | sayNO; |
c255a977 DM |
5373 | /* NOTREACHED */ |
5374 | ||
5375 | ||
5376 | curly_try_B_max: | |
5377 | /* a successful greedy match: now try to match B */ | |
40d049e4 | 5378 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5379 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
40d049e4 YO |
5380 | goto fake_end; |
5381 | } | |
c255a977 DM |
5382 | { |
5383 | UV c = 0; | |
5384 | if (ST.c1 != CHRTEST_VOID) | |
f2ed9b32 | 5385 | c = utf8_target ? utf8n_to_uvchr((U8*)PL_reginput, |
c255a977 | 5386 | UTF8_MAXBYTES, 0, uniflags) |
466787eb | 5387 | : (UV) UCHARAT(PL_reginput); |
c255a977 DM |
5388 | /* If it could work, try it. */ |
5389 | if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) { | |
5390 | CURLY_SETPAREN(ST.paren, ST.count); | |
5391 | PUSH_STATE_GOTO(CURLY_B_max, ST.B); | |
5392 | /* NOTREACHED */ | |
5393 | } | |
5394 | } | |
5395 | /* FALL THROUGH */ | |
5396 | case CURLY_B_max_fail: | |
5397 | /* failed to find B in a greedy match */ | |
5398 | if (ST.paren && ST.count) | |
f0ab9afb | 5399 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5400 | |
5401 | REGCP_UNWIND(ST.cp); | |
5402 | /* back up. */ | |
5403 | if (--ST.count < ST.min) | |
5404 | sayNO; | |
5405 | PL_reginput = locinput = HOPc(locinput, -1); | |
5406 | goto curly_try_B_max; | |
5407 | ||
5408 | #undef ST | |
5409 | ||
a0d0e21e | 5410 | case END: |
6bda09f9 | 5411 | fake_end: |
faec1544 DM |
5412 | if (cur_eval) { |
5413 | /* we've just finished A in /(??{A})B/; now continue with B */ | |
5414 | I32 tmpix; | |
faec1544 DM |
5415 | st->u.eval.toggle_reg_flags |
5416 | = cur_eval->u.eval.toggle_reg_flags; | |
5417 | PL_reg_flags ^= st->u.eval.toggle_reg_flags; | |
5418 | ||
288b8c02 NC |
5419 | st->u.eval.prev_rex = rex_sv; /* inner */ |
5420 | SETREX(rex_sv,cur_eval->u.eval.prev_rex); | |
5421 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 5422 | rexi = RXi_GET(rex); |
faec1544 | 5423 | cur_curlyx = cur_eval->u.eval.prev_curlyx; |
288b8c02 | 5424 | ReREFCNT_inc(rex_sv); |
faec1544 | 5425 | st->u.eval.cp = regcppush(0); /* Save *all* the positions. */ |
34a81e2b B |
5426 | |
5427 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ | |
5428 | PL_reglastparen = &rex->lastparen; | |
5429 | PL_reglastcloseparen = &rex->lastcloseparen; | |
5430 | ||
faec1544 DM |
5431 | REGCP_SET(st->u.eval.lastcp); |
5432 | PL_reginput = locinput; | |
5433 | ||
5434 | /* Restore parens of the outer rex without popping the | |
5435 | * savestack */ | |
5436 | tmpix = PL_savestack_ix; | |
5437 | PL_savestack_ix = cur_eval->u.eval.lastcp; | |
5438 | regcppop(rex); | |
5439 | PL_savestack_ix = tmpix; | |
5440 | ||
5441 | st->u.eval.prev_eval = cur_eval; | |
5442 | cur_eval = cur_eval->u.eval.prev_eval; | |
5443 | DEBUG_EXECUTE_r( | |
2a49f0f5 JH |
5444 | PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n", |
5445 | REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval));); | |
e7707071 YO |
5446 | if ( nochange_depth ) |
5447 | nochange_depth--; | |
5448 | ||
5449 | PUSH_YES_STATE_GOTO(EVAL_AB, | |
faec1544 DM |
5450 | st->u.eval.prev_eval->u.eval.B); /* match B */ |
5451 | } | |
5452 | ||
3b0527fe | 5453 | if (locinput < reginfo->till) { |
a3621e74 | 5454 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
7821416a IZ |
5455 | "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n", |
5456 | PL_colors[4], | |
5457 | (long)(locinput - PL_reg_starttry), | |
3b0527fe | 5458 | (long)(reginfo->till - PL_reg_starttry), |
7821416a | 5459 | PL_colors[5])); |
58e23c8d | 5460 | |
262b90c4 | 5461 | sayNO_SILENT; /* Cannot match: too short. */ |
7821416a IZ |
5462 | } |
5463 | PL_reginput = locinput; /* put where regtry can find it */ | |
262b90c4 | 5464 | sayYES; /* Success! */ |
dad79028 DM |
5465 | |
5466 | case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */ | |
5467 | DEBUG_EXECUTE_r( | |
5468 | PerlIO_printf(Perl_debug_log, | |
5469 | "%*s %ssubpattern success...%s\n", | |
5bc10b2c | 5470 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])); |
3280af22 | 5471 | PL_reginput = locinput; /* put where regtry can find it */ |
262b90c4 | 5472 | sayYES; /* Success! */ |
dad79028 | 5473 | |
40a82448 DM |
5474 | #undef ST |
5475 | #define ST st->u.ifmatch | |
5476 | ||
5477 | case SUSPEND: /* (?>A) */ | |
5478 | ST.wanted = 1; | |
9fe1d20c | 5479 | PL_reginput = locinput; |
9041c2e3 | 5480 | goto do_ifmatch; |
dad79028 | 5481 | |
40a82448 DM |
5482 | case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */ |
5483 | ST.wanted = 0; | |
dad79028 DM |
5484 | goto ifmatch_trivial_fail_test; |
5485 | ||
40a82448 DM |
5486 | case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */ |
5487 | ST.wanted = 1; | |
dad79028 | 5488 | ifmatch_trivial_fail_test: |
a0ed51b3 | 5489 | if (scan->flags) { |
52657f30 | 5490 | char * const s = HOPBACKc(locinput, scan->flags); |
dad79028 DM |
5491 | if (!s) { |
5492 | /* trivial fail */ | |
24d3c4a9 DM |
5493 | if (logical) { |
5494 | logical = 0; | |
f2338a2e | 5495 | sw = 1 - cBOOL(ST.wanted); |
dad79028 | 5496 | } |
40a82448 | 5497 | else if (ST.wanted) |
dad79028 DM |
5498 | sayNO; |
5499 | next = scan + ARG(scan); | |
5500 | if (next == scan) | |
5501 | next = NULL; | |
5502 | break; | |
5503 | } | |
efb30f32 | 5504 | PL_reginput = s; |
a0ed51b3 LW |
5505 | } |
5506 | else | |
5507 | PL_reginput = locinput; | |
5508 | ||
c277df42 | 5509 | do_ifmatch: |
40a82448 | 5510 | ST.me = scan; |
24d3c4a9 | 5511 | ST.logical = logical; |
24d786f4 YO |
5512 | logical = 0; /* XXX: reset state of logical once it has been saved into ST */ |
5513 | ||
40a82448 DM |
5514 | /* execute body of (?...A) */ |
5515 | PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan))); | |
5516 | /* NOTREACHED */ | |
5517 | ||
5518 | case IFMATCH_A_fail: /* body of (?...A) failed */ | |
5519 | ST.wanted = !ST.wanted; | |
5520 | /* FALL THROUGH */ | |
5521 | ||
5522 | case IFMATCH_A: /* body of (?...A) succeeded */ | |
24d3c4a9 | 5523 | if (ST.logical) { |
f2338a2e | 5524 | sw = cBOOL(ST.wanted); |
40a82448 DM |
5525 | } |
5526 | else if (!ST.wanted) | |
5527 | sayNO; | |
5528 | ||
5529 | if (OP(ST.me) == SUSPEND) | |
5530 | locinput = PL_reginput; | |
5531 | else { | |
5532 | locinput = PL_reginput = st->locinput; | |
5533 | nextchr = UCHARAT(locinput); | |
5534 | } | |
5535 | scan = ST.me + ARG(ST.me); | |
5536 | if (scan == ST.me) | |
5537 | scan = NULL; | |
5538 | continue; /* execute B */ | |
5539 | ||
5540 | #undef ST | |
dad79028 | 5541 | |
c277df42 | 5542 | case LONGJMP: |
c277df42 IZ |
5543 | next = scan + ARG(scan); |
5544 | if (next == scan) | |
5545 | next = NULL; | |
a0d0e21e | 5546 | break; |
54612592 | 5547 | case COMMIT: |
e2e6a0f1 YO |
5548 | reginfo->cutpoint = PL_regeol; |
5549 | /* FALLTHROUGH */ | |
5d458dd8 | 5550 | case PRUNE: |
24b23f37 | 5551 | PL_reginput = locinput; |
e2e6a0f1 | 5552 | if (!scan->flags) |
ad64d0ec | 5553 | sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
54612592 YO |
5554 | PUSH_STATE_GOTO(COMMIT_next,next); |
5555 | /* NOTREACHED */ | |
5556 | case COMMIT_next_fail: | |
5557 | no_final = 1; | |
5558 | /* FALLTHROUGH */ | |
7f69552c YO |
5559 | case OPFAIL: |
5560 | sayNO; | |
e2e6a0f1 YO |
5561 | /* NOTREACHED */ |
5562 | ||
5563 | #define ST st->u.mark | |
5564 | case MARKPOINT: | |
5565 | ST.prev_mark = mark_state; | |
5d458dd8 | 5566 | ST.mark_name = sv_commit = sv_yes_mark |
ad64d0ec | 5567 | = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
e2e6a0f1 YO |
5568 | mark_state = st; |
5569 | ST.mark_loc = PL_reginput = locinput; | |
5570 | PUSH_YES_STATE_GOTO(MARKPOINT_next,next); | |
5571 | /* NOTREACHED */ | |
5572 | case MARKPOINT_next: | |
5573 | mark_state = ST.prev_mark; | |
5574 | sayYES; | |
5575 | /* NOTREACHED */ | |
5576 | case MARKPOINT_next_fail: | |
5d458dd8 | 5577 | if (popmark && sv_eq(ST.mark_name,popmark)) |
e2e6a0f1 YO |
5578 | { |
5579 | if (ST.mark_loc > startpoint) | |
5580 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5581 | popmark = NULL; /* we found our mark */ | |
5582 | sv_commit = ST.mark_name; | |
5583 | ||
5584 | DEBUG_EXECUTE_r({ | |
5d458dd8 | 5585 | PerlIO_printf(Perl_debug_log, |
e2e6a0f1 YO |
5586 | "%*s %ssetting cutpoint to mark:%"SVf"...%s\n", |
5587 | REPORT_CODE_OFF+depth*2, "", | |
be2597df | 5588 | PL_colors[4], SVfARG(sv_commit), PL_colors[5]); |
e2e6a0f1 YO |
5589 | }); |
5590 | } | |
5591 | mark_state = ST.prev_mark; | |
5d458dd8 YO |
5592 | sv_yes_mark = mark_state ? |
5593 | mark_state->u.mark.mark_name : NULL; | |
e2e6a0f1 YO |
5594 | sayNO; |
5595 | /* NOTREACHED */ | |
5d458dd8 YO |
5596 | case SKIP: |
5597 | PL_reginput = locinput; | |
5598 | if (scan->flags) { | |
2bf803e2 | 5599 | /* (*SKIP) : if we fail we cut here*/ |
5d458dd8 | 5600 | ST.mark_name = NULL; |
e2e6a0f1 | 5601 | ST.mark_loc = locinput; |
5d458dd8 YO |
5602 | PUSH_STATE_GOTO(SKIP_next,next); |
5603 | } else { | |
2bf803e2 | 5604 | /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was, |
5d458dd8 YO |
5605 | otherwise do nothing. Meaning we need to scan |
5606 | */ | |
5607 | regmatch_state *cur = mark_state; | |
ad64d0ec | 5608 | SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
5609 | |
5610 | while (cur) { | |
5611 | if ( sv_eq( cur->u.mark.mark_name, | |
5612 | find ) ) | |
5613 | { | |
5614 | ST.mark_name = find; | |
5615 | PUSH_STATE_GOTO( SKIP_next, next ); | |
5616 | } | |
5617 | cur = cur->u.mark.prev_mark; | |
5618 | } | |
e2e6a0f1 | 5619 | } |
2bf803e2 | 5620 | /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */ |
5d458dd8 YO |
5621 | break; |
5622 | case SKIP_next_fail: | |
5623 | if (ST.mark_name) { | |
5624 | /* (*CUT:NAME) - Set up to search for the name as we | |
5625 | collapse the stack*/ | |
5626 | popmark = ST.mark_name; | |
5627 | } else { | |
5628 | /* (*CUT) - No name, we cut here.*/ | |
e2e6a0f1 YO |
5629 | if (ST.mark_loc > startpoint) |
5630 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5d458dd8 YO |
5631 | /* but we set sv_commit to latest mark_name if there |
5632 | is one so they can test to see how things lead to this | |
5633 | cut */ | |
5634 | if (mark_state) | |
5635 | sv_commit=mark_state->u.mark.mark_name; | |
5636 | } | |
e2e6a0f1 YO |
5637 | no_final = 1; |
5638 | sayNO; | |
5639 | /* NOTREACHED */ | |
5640 | #undef ST | |
32e6a07c YO |
5641 | case FOLDCHAR: |
5642 | n = ARG(scan); | |
f2ed9b32 | 5643 | if ( n == (U32)what_len_TRICKYFOLD(locinput,utf8_target,ln) ) { |
e64b1bd1 | 5644 | locinput += ln; |
ced7f090 | 5645 | } else if ( LATIN_SMALL_LETTER_SHARP_S == n && !utf8_target && !UTF_PATTERN ) { |
e64b1bd1 YO |
5646 | sayNO; |
5647 | } else { | |
5648 | U8 folded[UTF8_MAXBYTES_CASE+1]; | |
5649 | STRLEN foldlen; | |
5650 | const char * const l = locinput; | |
5651 | char *e = PL_regeol; | |
5652 | to_uni_fold(n, folded, &foldlen); | |
5653 | ||
4c1b470c | 5654 | if (! foldEQ_utf8((const char*) folded, 0, foldlen, 1, |
f2ed9b32 | 5655 | l, &e, 0, utf8_target)) { |
32e6a07c | 5656 | sayNO; |
e64b1bd1 YO |
5657 | } |
5658 | locinput = e; | |
32e6a07c YO |
5659 | } |
5660 | nextchr = UCHARAT(locinput); | |
5661 | break; | |
e1d1eefb | 5662 | case LNBREAK: |
f2ed9b32 | 5663 | if ((n=is_LNBREAK(locinput,utf8_target))) { |
e1d1eefb YO |
5664 | locinput += n; |
5665 | nextchr = UCHARAT(locinput); | |
5666 | } else | |
5667 | sayNO; | |
5668 | break; | |
5669 | ||
5670 | #define CASE_CLASS(nAmE) \ | |
5671 | case nAmE: \ | |
f2ed9b32 | 5672 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5673 | locinput += n; \ |
5674 | nextchr = UCHARAT(locinput); \ | |
5675 | } else \ | |
5676 | sayNO; \ | |
5677 | break; \ | |
5678 | case N##nAmE: \ | |
f2ed9b32 | 5679 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5680 | sayNO; \ |
5681 | } else { \ | |
5682 | locinput += UTF8SKIP(locinput); \ | |
5683 | nextchr = UCHARAT(locinput); \ | |
5684 | } \ | |
5685 | break | |
5686 | ||
5687 | CASE_CLASS(VERTWS); | |
5688 | CASE_CLASS(HORIZWS); | |
5689 | #undef CASE_CLASS | |
5690 | ||
a0d0e21e | 5691 | default: |
b900a521 | 5692 | PerlIO_printf(Perl_error_log, "%"UVxf" %d\n", |
d7d93a81 | 5693 | PTR2UV(scan), OP(scan)); |
cea2e8a9 | 5694 | Perl_croak(aTHX_ "regexp memory corruption"); |
5d458dd8 YO |
5695 | |
5696 | } /* end switch */ | |
95b24440 | 5697 | |
5d458dd8 YO |
5698 | /* switch break jumps here */ |
5699 | scan = next; /* prepare to execute the next op and ... */ | |
5700 | continue; /* ... jump back to the top, reusing st */ | |
95b24440 DM |
5701 | /* NOTREACHED */ |
5702 | ||
40a82448 DM |
5703 | push_yes_state: |
5704 | /* push a state that backtracks on success */ | |
5705 | st->u.yes.prev_yes_state = yes_state; | |
5706 | yes_state = st; | |
5707 | /* FALL THROUGH */ | |
5708 | push_state: | |
5709 | /* push a new regex state, then continue at scan */ | |
5710 | { | |
5711 | regmatch_state *newst; | |
5712 | ||
24b23f37 YO |
5713 | DEBUG_STACK_r({ |
5714 | regmatch_state *cur = st; | |
5715 | regmatch_state *curyes = yes_state; | |
5716 | int curd = depth; | |
5717 | regmatch_slab *slab = PL_regmatch_slab; | |
5718 | for (;curd > -1;cur--,curd--) { | |
5719 | if (cur < SLAB_FIRST(slab)) { | |
5720 | slab = slab->prev; | |
5721 | cur = SLAB_LAST(slab); | |
5722 | } | |
5723 | PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n", | |
5724 | REPORT_CODE_OFF + 2 + depth * 2,"", | |
13d6edb4 | 5725 | curd, PL_reg_name[cur->resume_state], |
24b23f37 YO |
5726 | (curyes == cur) ? "yes" : "" |
5727 | ); | |
5728 | if (curyes == cur) | |
5729 | curyes = cur->u.yes.prev_yes_state; | |
5730 | } | |
5731 | } else | |
5732 | DEBUG_STATE_pp("push") | |
5733 | ); | |
40a82448 | 5734 | depth++; |
40a82448 DM |
5735 | st->locinput = locinput; |
5736 | newst = st+1; | |
5737 | if (newst > SLAB_LAST(PL_regmatch_slab)) | |
5738 | newst = S_push_slab(aTHX); | |
5739 | PL_regmatch_state = newst; | |
786e8c11 | 5740 | |
40a82448 DM |
5741 | locinput = PL_reginput; |
5742 | nextchr = UCHARAT(locinput); | |
5743 | st = newst; | |
5744 | continue; | |
5745 | /* NOTREACHED */ | |
5746 | } | |
a0d0e21e | 5747 | } |
a687059c | 5748 | |
a0d0e21e LW |
5749 | /* |
5750 | * We get here only if there's trouble -- normally "case END" is | |
5751 | * the terminating point. | |
5752 | */ | |
cea2e8a9 | 5753 | Perl_croak(aTHX_ "corrupted regexp pointers"); |
a0d0e21e | 5754 | /*NOTREACHED*/ |
4633a7c4 LW |
5755 | sayNO; |
5756 | ||
262b90c4 | 5757 | yes: |
77cb431f DM |
5758 | if (yes_state) { |
5759 | /* we have successfully completed a subexpression, but we must now | |
5760 | * pop to the state marked by yes_state and continue from there */ | |
77cb431f | 5761 | assert(st != yes_state); |
5bc10b2c DM |
5762 | #ifdef DEBUGGING |
5763 | while (st != yes_state) { | |
5764 | st--; | |
5765 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
5766 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5767 | st = SLAB_LAST(PL_regmatch_slab); | |
5768 | } | |
e2e6a0f1 | 5769 | DEBUG_STATE_r({ |
54612592 YO |
5770 | if (no_final) { |
5771 | DEBUG_STATE_pp("pop (no final)"); | |
5772 | } else { | |
5773 | DEBUG_STATE_pp("pop (yes)"); | |
5774 | } | |
e2e6a0f1 | 5775 | }); |
5bc10b2c DM |
5776 | depth--; |
5777 | } | |
5778 | #else | |
77cb431f DM |
5779 | while (yes_state < SLAB_FIRST(PL_regmatch_slab) |
5780 | || yes_state > SLAB_LAST(PL_regmatch_slab)) | |
5781 | { | |
5782 | /* not in this slab, pop slab */ | |
5783 | depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1); | |
5784 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5785 | st = SLAB_LAST(PL_regmatch_slab); | |
5786 | } | |
5787 | depth -= (st - yes_state); | |
5bc10b2c | 5788 | #endif |
77cb431f DM |
5789 | st = yes_state; |
5790 | yes_state = st->u.yes.prev_yes_state; | |
5791 | PL_regmatch_state = st; | |
24b23f37 | 5792 | |
5d458dd8 YO |
5793 | if (no_final) { |
5794 | locinput= st->locinput; | |
5795 | nextchr = UCHARAT(locinput); | |
5796 | } | |
54612592 | 5797 | state_num = st->resume_state + no_final; |
24d3c4a9 | 5798 | goto reenter_switch; |
77cb431f DM |
5799 | } |
5800 | ||
a3621e74 | 5801 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", |
e4584336 | 5802 | PL_colors[4], PL_colors[5])); |
02db2b7b | 5803 | |
19b95bf0 DM |
5804 | if (PL_reg_eval_set) { |
5805 | /* each successfully executed (?{...}) block does the equivalent of | |
5806 | * local $^R = do {...} | |
5807 | * When popping the save stack, all these locals would be undone; | |
5808 | * bypass this by setting the outermost saved $^R to the latest | |
5809 | * value */ | |
5810 | if (oreplsv != GvSV(PL_replgv)) | |
5811 | sv_setsv(oreplsv, GvSV(PL_replgv)); | |
5812 | } | |
95b24440 | 5813 | result = 1; |
aa283a38 | 5814 | goto final_exit; |
4633a7c4 LW |
5815 | |
5816 | no: | |
a3621e74 | 5817 | DEBUG_EXECUTE_r( |
7821416a | 5818 | PerlIO_printf(Perl_debug_log, |
786e8c11 | 5819 | "%*s %sfailed...%s\n", |
5bc10b2c | 5820 | REPORT_CODE_OFF+depth*2, "", |
786e8c11 | 5821 | PL_colors[4], PL_colors[5]) |
7821416a | 5822 | ); |
aa283a38 | 5823 | |
262b90c4 | 5824 | no_silent: |
54612592 YO |
5825 | if (no_final) { |
5826 | if (yes_state) { | |
5827 | goto yes; | |
5828 | } else { | |
5829 | goto final_exit; | |
5830 | } | |
5831 | } | |
aa283a38 DM |
5832 | if (depth) { |
5833 | /* there's a previous state to backtrack to */ | |
40a82448 DM |
5834 | st--; |
5835 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
5836 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5837 | st = SLAB_LAST(PL_regmatch_slab); | |
5838 | } | |
5839 | PL_regmatch_state = st; | |
40a82448 DM |
5840 | locinput= st->locinput; |
5841 | nextchr = UCHARAT(locinput); | |
5842 | ||
5bc10b2c DM |
5843 | DEBUG_STATE_pp("pop"); |
5844 | depth--; | |
262b90c4 DM |
5845 | if (yes_state == st) |
5846 | yes_state = st->u.yes.prev_yes_state; | |
5bc10b2c | 5847 | |
24d3c4a9 DM |
5848 | state_num = st->resume_state + 1; /* failure = success + 1 */ |
5849 | goto reenter_switch; | |
95b24440 | 5850 | } |
24d3c4a9 | 5851 | result = 0; |
aa283a38 | 5852 | |
262b90c4 | 5853 | final_exit: |
bbe252da | 5854 | if (rex->intflags & PREGf_VERBARG_SEEN) { |
5d458dd8 YO |
5855 | SV *sv_err = get_sv("REGERROR", 1); |
5856 | SV *sv_mrk = get_sv("REGMARK", 1); | |
5857 | if (result) { | |
e2e6a0f1 | 5858 | sv_commit = &PL_sv_no; |
5d458dd8 YO |
5859 | if (!sv_yes_mark) |
5860 | sv_yes_mark = &PL_sv_yes; | |
5861 | } else { | |
5862 | if (!sv_commit) | |
5863 | sv_commit = &PL_sv_yes; | |
5864 | sv_yes_mark = &PL_sv_no; | |
5865 | } | |
5866 | sv_setsv(sv_err, sv_commit); | |
5867 | sv_setsv(sv_mrk, sv_yes_mark); | |
e2e6a0f1 | 5868 | } |
19b95bf0 | 5869 | |
2f554ef7 DM |
5870 | /* clean up; in particular, free all slabs above current one */ |
5871 | LEAVE_SCOPE(oldsave); | |
5d9a96ca | 5872 | |
95b24440 | 5873 | return result; |
a687059c LW |
5874 | } |
5875 | ||
5876 | /* | |
5877 | - regrepeat - repeatedly match something simple, report how many | |
5878 | */ | |
5879 | /* | |
5880 | * [This routine now assumes that it will only match on things of length 1. | |
5881 | * That was true before, but now we assume scan - reginput is the count, | |
a0ed51b3 | 5882 | * rather than incrementing count on every character. [Er, except utf8.]] |
a687059c | 5883 | */ |
76e3520e | 5884 | STATIC I32 |
e2e6a0f1 | 5885 | S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth) |
a687059c | 5886 | { |
27da23d5 | 5887 | dVAR; |
a0d0e21e | 5888 | register char *scan; |
a0d0e21e | 5889 | register I32 c; |
3280af22 | 5890 | register char *loceol = PL_regeol; |
a0ed51b3 | 5891 | register I32 hardcount = 0; |
f2ed9b32 | 5892 | register bool utf8_target = PL_reg_match_utf8; |
d513472c | 5893 | UV utf8_flags; |
4f55667c SP |
5894 | #ifndef DEBUGGING |
5895 | PERL_UNUSED_ARG(depth); | |
5896 | #endif | |
a0d0e21e | 5897 | |
7918f24d NC |
5898 | PERL_ARGS_ASSERT_REGREPEAT; |
5899 | ||
3280af22 | 5900 | scan = PL_reginput; |
faf11cac HS |
5901 | if (max == REG_INFTY) |
5902 | max = I32_MAX; | |
5903 | else if (max < loceol - scan) | |
7f596f4c | 5904 | loceol = scan + max; |
a0d0e21e | 5905 | switch (OP(p)) { |
22c35a8c | 5906 | case REG_ANY: |
f2ed9b32 | 5907 | if (utf8_target) { |
ffc61ed2 | 5908 | loceol = PL_regeol; |
1aa99e6b | 5909 | while (scan < loceol && hardcount < max && *scan != '\n') { |
ffc61ed2 JH |
5910 | scan += UTF8SKIP(scan); |
5911 | hardcount++; | |
5912 | } | |
5913 | } else { | |
5914 | while (scan < loceol && *scan != '\n') | |
5915 | scan++; | |
a0ed51b3 LW |
5916 | } |
5917 | break; | |
ffc61ed2 | 5918 | case SANY: |
f2ed9b32 | 5919 | if (utf8_target) { |
def8e4ea | 5920 | loceol = PL_regeol; |
a0804c9e | 5921 | while (scan < loceol && hardcount < max) { |
def8e4ea JH |
5922 | scan += UTF8SKIP(scan); |
5923 | hardcount++; | |
5924 | } | |
5925 | } | |
5926 | else | |
5927 | scan = loceol; | |
a0ed51b3 | 5928 | break; |
f33976b4 DB |
5929 | case CANY: |
5930 | scan = loceol; | |
5931 | break; | |
59d32103 KW |
5932 | case EXACT: |
5933 | /* To get here, EXACTish nodes must have *byte* length == 1. That | |
5934 | * means they match only characters in the string that can be expressed | |
5935 | * as a single byte. For non-utf8 strings, that means a simple match. | |
5936 | * For utf8 strings, the character matched must be an invariant, or | |
5937 | * downgradable to a single byte. The pattern's utf8ness is | |
5938 | * irrelevant, as since it's a single byte, it either isn't utf8, or if | |
5939 | * it is, it's an invariant */ | |
5940 | ||
5941 | c = (U8)*STRING(p); | |
5942 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); | |
5943 | ||
5944 | if (! utf8_target || UNI_IS_INVARIANT(c)) { | |
5945 | while (scan < loceol && UCHARAT(scan) == c) { | |
5946 | scan++; | |
5947 | } | |
5948 | } | |
5949 | else { | |
5950 | ||
5951 | /* Here, the string is utf8, and the pattern char is different | |
5952 | * in utf8 than not, so can't compare them directly. Outside the | |
5953 | * loop, find find the two utf8 bytes that represent c, and then | |
5954 | * look for those in sequence in the utf8 string */ | |
5955 | U8 high = UTF8_TWO_BYTE_HI(c); | |
5956 | U8 low = UTF8_TWO_BYTE_LO(c); | |
5957 | loceol = PL_regeol; | |
5958 | ||
5959 | while (hardcount < max | |
5960 | && scan + 1 < loceol | |
5961 | && UCHARAT(scan) == high | |
5962 | && UCHARAT(scan + 1) == low) | |
5963 | { | |
5964 | scan += 2; | |
5965 | hardcount++; | |
5966 | } | |
5967 | } | |
5968 | break; | |
2f7f8cb1 KW |
5969 | case EXACTFA: |
5970 | utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
5971 | goto do_exactf; | |
5972 | ||
d4e0b827 KW |
5973 | case EXACTFL: |
5974 | PL_reg_flags |= RF_tainted; | |
17580e7a KW |
5975 | utf8_flags = FOLDEQ_UTF8_LOCALE; |
5976 | goto do_exactf; | |
5977 | ||
d4e0b827 | 5978 | case EXACTF: |
9a5a5549 | 5979 | case EXACTFU: |
d513472c | 5980 | utf8_flags = 0; |
59d32103 | 5981 | |
2be3e190 KW |
5982 | /* The comments for the EXACT case above apply as well to these fold |
5983 | * ones */ | |
634c83a2 | 5984 | |
2f7f8cb1 | 5985 | do_exactf: |
090f7165 | 5986 | c = (U8)*STRING(p); |
634c83a2 | 5987 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); |
d4e0b827 | 5988 | |
59d32103 | 5989 | if (utf8_target) { /* Use full Unicode fold matching */ |
59d32103 KW |
5990 | char *tmpeol = loceol; |
5991 | while (hardcount < max | |
d513472c KW |
5992 | && foldEQ_utf8_flags(scan, &tmpeol, 0, utf8_target, |
5993 | STRING(p), NULL, 1, cBOOL(UTF_PATTERN), utf8_flags)) | |
59d32103 KW |
5994 | { |
5995 | scan = tmpeol; | |
5996 | tmpeol = loceol; | |
5997 | hardcount++; | |
5998 | } | |
634c83a2 | 5999 | |
59d32103 KW |
6000 | /* XXX Note that the above handles properly the German sharp s in |
6001 | * the pattern matching ss in the string. But it doesn't handle | |
6002 | * properly cases where the string contains say 'LIGATURE ff' and | |
6003 | * the pattern is 'f+'. This would require, say, a new function or | |
6004 | * revised interface to foldEQ_utf8(), in which the maximum number | |
6005 | * of characters to match could be passed and it would return how | |
6006 | * many actually did. This is just one of many cases where | |
6007 | * multi-char folds don't work properly, and so the fix is being | |
6008 | * deferred */ | |
6009 | } | |
6010 | else { | |
87381386 | 6011 | U8 folded; |
59d32103 | 6012 | |
2be3e190 KW |
6013 | /* Here, the string isn't utf8 and c is a single byte; and either |
6014 | * the pattern isn't utf8 or c is an invariant, so its utf8ness | |
6015 | * doesn't affect c. Can just do simple comparisons for exact or | |
6016 | * fold matching. */ | |
d4e0b827 | 6017 | switch (OP(p)) { |
87381386 | 6018 | case EXACTF: folded = PL_fold[c]; break; |
2f7f8cb1 | 6019 | case EXACTFA: |
9a5a5549 | 6020 | case EXACTFU: folded = PL_fold_latin1[c]; break; |
87381386 KW |
6021 | case EXACTFL: folded = PL_fold_locale[c]; break; |
6022 | default: Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p)); | |
6023 | } | |
6024 | while (scan < loceol && | |
6025 | (UCHARAT(scan) == c || UCHARAT(scan) == folded)) | |
6026 | { | |
6027 | scan++; | |
634c83a2 KW |
6028 | } |
6029 | } | |
bbce6d69 | 6030 | break; |
f56b6394 | 6031 | case ANYOFV: |
a0d0e21e | 6032 | case ANYOF: |
4e8910e0 KW |
6033 | if (utf8_target || OP(p) == ANYOFV) { |
6034 | STRLEN inclasslen; | |
ffc61ed2 | 6035 | loceol = PL_regeol; |
4e8910e0 KW |
6036 | inclasslen = loceol - scan; |
6037 | while (hardcount < max | |
6038 | && ((inclasslen = loceol - scan) > 0) | |
6039 | && reginclass(prog, p, (U8*)scan, &inclasslen, utf8_target)) | |
6040 | { | |
6041 | scan += inclasslen; | |
ffc61ed2 JH |
6042 | hardcount++; |
6043 | } | |
6044 | } else { | |
32fc9b6a | 6045 | while (scan < loceol && REGINCLASS(prog, p, (U8*)scan)) |
ffc61ed2 JH |
6046 | scan++; |
6047 | } | |
a0d0e21e | 6048 | break; |
980866de | 6049 | case ALNUMU: |
f2ed9b32 | 6050 | if (utf8_target) { |
980866de | 6051 | utf8_wordchar: |
ffc61ed2 | 6052 | loceol = PL_regeol; |
1a4fad37 | 6053 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6054 | while (hardcount < max && scan < loceol && |
a12cf05f KW |
6055 | swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
6056 | { | |
ffc61ed2 JH |
6057 | scan += UTF8SKIP(scan); |
6058 | hardcount++; | |
6059 | } | |
980866de | 6060 | } else { |
a12cf05f KW |
6061 | while (scan < loceol && isWORDCHAR_L1((U8) *scan)) { |
6062 | scan++; | |
6063 | } | |
980866de KW |
6064 | } |
6065 | break; | |
6066 | case ALNUM: | |
6067 | if (utf8_target) | |
6068 | goto utf8_wordchar; | |
6069 | while (scan < loceol && isALNUM((U8) *scan)) { | |
6070 | scan++; | |
a0ed51b3 LW |
6071 | } |
6072 | break; | |
cfaf538b KW |
6073 | case ALNUMA: |
6074 | while (scan < loceol && isWORDCHAR_A((U8) *scan)) { | |
6075 | scan++; | |
6076 | } | |
6077 | break; | |
bbce6d69 | 6078 | case ALNUML: |
3280af22 | 6079 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6080 | if (utf8_target) { |
ffc61ed2 | 6081 | loceol = PL_regeol; |
1aa99e6b IH |
6082 | while (hardcount < max && scan < loceol && |
6083 | isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6084 | scan += UTF8SKIP(scan); |
6085 | hardcount++; | |
6086 | } | |
6087 | } else { | |
6088 | while (scan < loceol && isALNUM_LC(*scan)) | |
6089 | scan++; | |
a0ed51b3 LW |
6090 | } |
6091 | break; | |
980866de | 6092 | case NALNUMU: |
f2ed9b32 | 6093 | if (utf8_target) { |
980866de KW |
6094 | |
6095 | utf8_Nwordchar: | |
6096 | ||
ffc61ed2 | 6097 | loceol = PL_regeol; |
1a4fad37 | 6098 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6099 | while (hardcount < max && scan < loceol && |
980866de | 6100 | ! swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
a12cf05f | 6101 | { |
ffc61ed2 JH |
6102 | scan += UTF8SKIP(scan); |
6103 | hardcount++; | |
6104 | } | |
980866de | 6105 | } else { |
a12cf05f KW |
6106 | while (scan < loceol && ! isWORDCHAR_L1((U8) *scan)) { |
6107 | scan++; | |
6108 | } | |
980866de KW |
6109 | } |
6110 | break; | |
6111 | case NALNUM: | |
6112 | if (utf8_target) | |
6113 | goto utf8_Nwordchar; | |
6114 | while (scan < loceol && ! isALNUM((U8) *scan)) { | |
6115 | scan++; | |
a0ed51b3 LW |
6116 | } |
6117 | break; | |
cfaf538b KW |
6118 | case NALNUMA: |
6119 | if (utf8_target) { | |
6120 | while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) { | |
6121 | scan += UTF8SKIP(scan); | |
6122 | } | |
6123 | } | |
6124 | else { | |
6125 | while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) { | |
6126 | scan++; | |
6127 | } | |
6128 | } | |
6129 | break; | |
bbce6d69 | 6130 | case NALNUML: |
3280af22 | 6131 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6132 | if (utf8_target) { |
ffc61ed2 | 6133 | loceol = PL_regeol; |
1aa99e6b IH |
6134 | while (hardcount < max && scan < loceol && |
6135 | !isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6136 | scan += UTF8SKIP(scan); |
6137 | hardcount++; | |
6138 | } | |
6139 | } else { | |
6140 | while (scan < loceol && !isALNUM_LC(*scan)) | |
6141 | scan++; | |
a0ed51b3 LW |
6142 | } |
6143 | break; | |
980866de | 6144 | case SPACEU: |
f2ed9b32 | 6145 | if (utf8_target) { |
980866de KW |
6146 | |
6147 | utf8_space: | |
6148 | ||
ffc61ed2 | 6149 | loceol = PL_regeol; |
1a4fad37 | 6150 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6151 | while (hardcount < max && scan < loceol && |
3568d838 | 6152 | (*scan == ' ' || |
a12cf05f KW |
6153 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) |
6154 | { | |
ffc61ed2 JH |
6155 | scan += UTF8SKIP(scan); |
6156 | hardcount++; | |
6157 | } | |
980866de KW |
6158 | break; |
6159 | } | |
6160 | else { | |
a12cf05f KW |
6161 | while (scan < loceol && isSPACE_L1((U8) *scan)) { |
6162 | scan++; | |
6163 | } | |
980866de KW |
6164 | break; |
6165 | } | |
6166 | case SPACE: | |
6167 | if (utf8_target) | |
6168 | goto utf8_space; | |
6169 | ||
6170 | while (scan < loceol && isSPACE((U8) *scan)) { | |
6171 | scan++; | |
a0ed51b3 LW |
6172 | } |
6173 | break; | |
cfaf538b KW |
6174 | case SPACEA: |
6175 | while (scan < loceol && isSPACE_A((U8) *scan)) { | |
6176 | scan++; | |
6177 | } | |
6178 | break; | |
bbce6d69 | 6179 | case SPACEL: |
3280af22 | 6180 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6181 | if (utf8_target) { |
ffc61ed2 | 6182 | loceol = PL_regeol; |
1aa99e6b | 6183 | while (hardcount < max && scan < loceol && |
6bbba904 | 6184 | isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6185 | scan += UTF8SKIP(scan); |
6186 | hardcount++; | |
6187 | } | |
6188 | } else { | |
6189 | while (scan < loceol && isSPACE_LC(*scan)) | |
6190 | scan++; | |
a0ed51b3 LW |
6191 | } |
6192 | break; | |
980866de | 6193 | case NSPACEU: |
f2ed9b32 | 6194 | if (utf8_target) { |
980866de KW |
6195 | |
6196 | utf8_Nspace: | |
6197 | ||
ffc61ed2 | 6198 | loceol = PL_regeol; |
1a4fad37 | 6199 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6200 | while (hardcount < max && scan < loceol && |
980866de KW |
6201 | ! (*scan == ' ' || |
6202 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) | |
a12cf05f | 6203 | { |
ffc61ed2 JH |
6204 | scan += UTF8SKIP(scan); |
6205 | hardcount++; | |
6206 | } | |
980866de KW |
6207 | break; |
6208 | } | |
6209 | else { | |
a12cf05f KW |
6210 | while (scan < loceol && ! isSPACE_L1((U8) *scan)) { |
6211 | scan++; | |
6212 | } | |
980866de KW |
6213 | } |
6214 | break; | |
6215 | case NSPACE: | |
6216 | if (utf8_target) | |
6217 | goto utf8_Nspace; | |
6218 | ||
6219 | while (scan < loceol && ! isSPACE((U8) *scan)) { | |
6220 | scan++; | |
a0ed51b3 | 6221 | } |
0008a298 | 6222 | break; |
cfaf538b KW |
6223 | case NSPACEA: |
6224 | if (utf8_target) { | |
6225 | while (scan < loceol && ! isSPACE_A((U8) *scan)) { | |
6226 | scan += UTF8SKIP(scan); | |
6227 | } | |
6228 | } | |
6229 | else { | |
6230 | while (scan < loceol && ! isSPACE_A((U8) *scan)) { | |
6231 | scan++; | |
6232 | } | |
6233 | } | |
6234 | break; | |
bbce6d69 | 6235 | case NSPACEL: |
3280af22 | 6236 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6237 | if (utf8_target) { |
ffc61ed2 | 6238 | loceol = PL_regeol; |
1aa99e6b | 6239 | while (hardcount < max && scan < loceol && |
6bbba904 | 6240 | !isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6241 | scan += UTF8SKIP(scan); |
6242 | hardcount++; | |
6243 | } | |
6244 | } else { | |
6245 | while (scan < loceol && !isSPACE_LC(*scan)) | |
6246 | scan++; | |
a0ed51b3 LW |
6247 | } |
6248 | break; | |
a0d0e21e | 6249 | case DIGIT: |
f2ed9b32 | 6250 | if (utf8_target) { |
ffc61ed2 | 6251 | loceol = PL_regeol; |
1a4fad37 | 6252 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6253 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6254 | swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6255 | scan += UTF8SKIP(scan); |
6256 | hardcount++; | |
6257 | } | |
6258 | } else { | |
6259 | while (scan < loceol && isDIGIT(*scan)) | |
6260 | scan++; | |
a0ed51b3 LW |
6261 | } |
6262 | break; | |
cfaf538b KW |
6263 | case DIGITA: |
6264 | while (scan < loceol && isDIGIT_A((U8) *scan)) { | |
6265 | scan++; | |
6266 | } | |
6267 | break; | |
b77393f6 KW |
6268 | case DIGITL: |
6269 | PL_reg_flags |= RF_tainted; | |
6270 | if (utf8_target) { | |
6271 | loceol = PL_regeol; | |
6272 | while (hardcount < max && scan < loceol && | |
6273 | isDIGIT_LC_utf8((U8*)scan)) { | |
6274 | scan += UTF8SKIP(scan); | |
6275 | hardcount++; | |
6276 | } | |
6277 | } else { | |
6278 | while (scan < loceol && isDIGIT_LC(*scan)) | |
6279 | scan++; | |
6280 | } | |
6281 | break; | |
a0d0e21e | 6282 | case NDIGIT: |
f2ed9b32 | 6283 | if (utf8_target) { |
ffc61ed2 | 6284 | loceol = PL_regeol; |
1a4fad37 | 6285 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6286 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6287 | !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6288 | scan += UTF8SKIP(scan); |
6289 | hardcount++; | |
6290 | } | |
6291 | } else { | |
6292 | while (scan < loceol && !isDIGIT(*scan)) | |
6293 | scan++; | |
a0ed51b3 | 6294 | } |
cfaf538b KW |
6295 | break; |
6296 | case NDIGITA: | |
6297 | if (utf8_target) { | |
6298 | while (scan < loceol && ! isDIGIT_A((U8) *scan)) { | |
6299 | scan += UTF8SKIP(scan); | |
6300 | } | |
6301 | } | |
6302 | else { | |
6303 | while (scan < loceol && ! isDIGIT_A((U8) *scan)) { | |
6304 | scan++; | |
6305 | } | |
6306 | } | |
6307 | break; | |
b77393f6 KW |
6308 | case NDIGITL: |
6309 | PL_reg_flags |= RF_tainted; | |
6310 | if (utf8_target) { | |
6311 | loceol = PL_regeol; | |
6312 | while (hardcount < max && scan < loceol && | |
6313 | !isDIGIT_LC_utf8((U8*)scan)) { | |
6314 | scan += UTF8SKIP(scan); | |
6315 | hardcount++; | |
6316 | } | |
6317 | } else { | |
6318 | while (scan < loceol && !isDIGIT_LC(*scan)) | |
6319 | scan++; | |
6320 | } | |
6321 | break; | |
e1d1eefb | 6322 | case LNBREAK: |
f2ed9b32 | 6323 | if (utf8_target) { |
e1d1eefb YO |
6324 | loceol = PL_regeol; |
6325 | while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) { | |
6326 | scan += c; | |
6327 | hardcount++; | |
6328 | } | |
6329 | } else { | |
6330 | /* | |
6331 | LNBREAK can match two latin chars, which is ok, | |
6332 | because we have a null terminated string, but we | |
6333 | have to use hardcount in this situation | |
6334 | */ | |
6335 | while (scan < loceol && (c=is_LNBREAK_latin1(scan))) { | |
6336 | scan+=c; | |
6337 | hardcount++; | |
6338 | } | |
6339 | } | |
6340 | break; | |
6341 | case HORIZWS: | |
f2ed9b32 | 6342 | if (utf8_target) { |
e1d1eefb YO |
6343 | loceol = PL_regeol; |
6344 | while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) { | |
6345 | scan += c; | |
6346 | hardcount++; | |
6347 | } | |
6348 | } else { | |
6349 | while (scan < loceol && is_HORIZWS_latin1(scan)) | |
6350 | scan++; | |
6351 | } | |
a0ed51b3 | 6352 | break; |
e1d1eefb | 6353 | case NHORIZWS: |
f2ed9b32 | 6354 | if (utf8_target) { |
e1d1eefb YO |
6355 | loceol = PL_regeol; |
6356 | while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) { | |
6357 | scan += UTF8SKIP(scan); | |
6358 | hardcount++; | |
6359 | } | |
6360 | } else { | |
6361 | while (scan < loceol && !is_HORIZWS_latin1(scan)) | |
6362 | scan++; | |
6363 | ||
6364 | } | |
6365 | break; | |
6366 | case VERTWS: | |
f2ed9b32 | 6367 | if (utf8_target) { |
e1d1eefb YO |
6368 | loceol = PL_regeol; |
6369 | while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) { | |
6370 | scan += c; | |
6371 | hardcount++; | |
6372 | } | |
6373 | } else { | |
6374 | while (scan < loceol && is_VERTWS_latin1(scan)) | |
6375 | scan++; | |
6376 | ||
6377 | } | |
6378 | break; | |
6379 | case NVERTWS: | |
f2ed9b32 | 6380 | if (utf8_target) { |
e1d1eefb YO |
6381 | loceol = PL_regeol; |
6382 | while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) { | |
6383 | scan += UTF8SKIP(scan); | |
6384 | hardcount++; | |
6385 | } | |
6386 | } else { | |
6387 | while (scan < loceol && !is_VERTWS_latin1(scan)) | |
6388 | scan++; | |
6389 | ||
6390 | } | |
6391 | break; | |
6392 | ||
a0d0e21e LW |
6393 | default: /* Called on something of 0 width. */ |
6394 | break; /* So match right here or not at all. */ | |
6395 | } | |
a687059c | 6396 | |
a0ed51b3 LW |
6397 | if (hardcount) |
6398 | c = hardcount; | |
6399 | else | |
6400 | c = scan - PL_reginput; | |
3280af22 | 6401 | PL_reginput = scan; |
a687059c | 6402 | |
a3621e74 | 6403 | DEBUG_r({ |
e68ec53f | 6404 | GET_RE_DEBUG_FLAGS_DECL; |
be8e71aa | 6405 | DEBUG_EXECUTE_r({ |
e68ec53f YO |
6406 | SV * const prop = sv_newmortal(); |
6407 | regprop(prog, prop, p); | |
6408 | PerlIO_printf(Perl_debug_log, | |
be8e71aa | 6409 | "%*s %s can match %"IVdf" times out of %"IVdf"...\n", |
e2e6a0f1 | 6410 | REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max); |
a3621e74 | 6411 | }); |
be8e71aa | 6412 | }); |
9041c2e3 | 6413 | |
a0d0e21e | 6414 | return(c); |
a687059c LW |
6415 | } |
6416 | ||
c277df42 | 6417 | |
be8e71aa | 6418 | #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) |
c277df42 | 6419 | /* |
ffc61ed2 JH |
6420 | - regclass_swash - prepare the utf8 swash |
6421 | */ | |
6422 | ||
6423 | SV * | |
32fc9b6a | 6424 | Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) |
ffc61ed2 | 6425 | { |
97aff369 | 6426 | dVAR; |
9e55ce06 JH |
6427 | SV *sw = NULL; |
6428 | SV *si = NULL; | |
6429 | SV *alt = NULL; | |
f8fc2ecf YO |
6430 | RXi_GET_DECL(prog,progi); |
6431 | const struct reg_data * const data = prog ? progi->data : NULL; | |
ffc61ed2 | 6432 | |
7918f24d NC |
6433 | PERL_ARGS_ASSERT_REGCLASS_SWASH; |
6434 | ||
4f639d21 | 6435 | if (data && data->count) { |
a3b680e6 | 6436 | const U32 n = ARG(node); |
ffc61ed2 | 6437 | |
4f639d21 | 6438 | if (data->what[n] == 's') { |
ad64d0ec NC |
6439 | SV * const rv = MUTABLE_SV(data->data[n]); |
6440 | AV * const av = MUTABLE_AV(SvRV(rv)); | |
2d03de9c | 6441 | SV **const ary = AvARRAY(av); |
9e55ce06 | 6442 | SV **a, **b; |
9041c2e3 | 6443 | |
711a919c | 6444 | /* See the end of regcomp.c:S_regclass() for |
9e55ce06 JH |
6445 | * documentation of these array elements. */ |
6446 | ||
b11f357e | 6447 | si = *ary; |
fe5bfecd JH |
6448 | a = SvROK(ary[1]) ? &ary[1] : NULL; |
6449 | b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : NULL; | |
b11f357e | 6450 | |
ffc61ed2 JH |
6451 | if (a) |
6452 | sw = *a; | |
6453 | else if (si && doinit) { | |
6454 | sw = swash_init("utf8", "", si, 1, 0); | |
6455 | (void)av_store(av, 1, sw); | |
6456 | } | |
9e55ce06 JH |
6457 | if (b) |
6458 | alt = *b; | |
ffc61ed2 JH |
6459 | } |
6460 | } | |
6461 | ||
9e55ce06 JH |
6462 | if (listsvp) |
6463 | *listsvp = si; | |
6464 | if (altsvp) | |
6465 | *altsvp = alt; | |
ffc61ed2 JH |
6466 | |
6467 | return sw; | |
6468 | } | |
76234dfb | 6469 | #endif |
ffc61ed2 JH |
6470 | |
6471 | /* | |
ba7b4546 | 6472 | - reginclass - determine if a character falls into a character class |
832705d4 | 6473 | |
6698fab5 KW |
6474 | n is the ANYOF regnode |
6475 | p is the target string | |
6476 | lenp is pointer to the maximum number of bytes of how far to go in p | |
6477 | (This is assumed wthout checking to always be at least the current | |
6478 | character's size) | |
6479 | utf8_target tells whether p is in UTF-8. | |
832705d4 | 6480 | |
4b3cda86 KW |
6481 | Returns true if matched; false otherwise. If lenp is not NULL, on return |
6482 | from a successful match, the value it points to will be updated to how many | |
6483 | bytes in p were matched. If there was no match, the value is undefined, | |
6484 | possibly changed from the input. | |
eba1359e | 6485 | |
d5788240 KW |
6486 | Note that this can be a synthetic start class, a combination of various |
6487 | nodes, so things you think might be mutually exclusive, such as locale, | |
6488 | aren't. It can match both locale and non-locale | |
6489 | ||
bbce6d69 | 6490 | */ |
6491 | ||
76e3520e | 6492 | STATIC bool |
f6ad78d8 | 6493 | 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 | 6494 | { |
27da23d5 | 6495 | dVAR; |
a3b680e6 | 6496 | const char flags = ANYOF_FLAGS(n); |
bbce6d69 | 6497 | bool match = FALSE; |
cc07378b | 6498 | UV c = *p; |
f7ab54c6 | 6499 | STRLEN c_len = 0; |
6698fab5 | 6500 | STRLEN maxlen; |
1aa99e6b | 6501 | |
7918f24d NC |
6502 | PERL_ARGS_ASSERT_REGINCLASS; |
6503 | ||
4b3cda86 | 6504 | /* If c is not already the code point, get it */ |
f2ed9b32 | 6505 | if (utf8_target && !UTF8_IS_INVARIANT(c)) { |
f7ab54c6 | 6506 | c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &c_len, |
6182169b KW |
6507 | (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) |
6508 | | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY); | |
6509 | /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for | |
6510 | * UTF8_ALLOW_FFFF */ | |
f7ab54c6 | 6511 | if (c_len == (STRLEN)-1) |
e8a70c6f | 6512 | Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); |
19f67299 | 6513 | } |
a5a291f5 KW |
6514 | else { |
6515 | c_len = 1; | |
6516 | } | |
bbce6d69 | 6517 | |
a5a291f5 KW |
6518 | /* Use passed in max length, or one character if none passed in or less |
6519 | * than one character. And assume will match just one character. This is | |
6520 | * overwritten later if matched more. */ | |
4b3cda86 | 6521 | if (lenp) { |
a5a291f5 KW |
6522 | maxlen = (*lenp > c_len) ? *lenp : c_len; |
6523 | *lenp = c_len; | |
4b3cda86 KW |
6524 | |
6525 | } | |
6526 | else { | |
a5a291f5 | 6527 | maxlen = c_len; |
4b3cda86 KW |
6528 | } |
6529 | ||
7cdde544 KW |
6530 | /* If this character is potentially in the bitmap, check it */ |
6531 | if (c < 256) { | |
ffc61ed2 JH |
6532 | if (ANYOF_BITMAP_TEST(n, c)) |
6533 | match = TRUE; | |
11454c59 KW |
6534 | else if (flags & ANYOF_NON_UTF8_LATIN1_ALL |
6535 | && ! utf8_target | |
6536 | && ! isASCII(c)) | |
6537 | { | |
6538 | match = TRUE; | |
6539 | } | |
a0ed51b3 | 6540 | |
78969a98 KW |
6541 | else if (flags & ANYOF_LOCALE) { |
6542 | PL_reg_flags |= RF_tainted; | |
6543 | ||
6544 | if ((flags & ANYOF_LOC_NONBITMAP_FOLD) | |
6545 | && ANYOF_BITMAP_TEST(n, PL_fold_locale[c])) | |
6546 | { | |
ffc61ed2 | 6547 | match = TRUE; |
78969a98 | 6548 | } |
040aea3a KW |
6549 | else if (ANYOF_CLASS_TEST_ANY_SET(n) && |
6550 | ((ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) || | |
6551 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) || | |
6552 | (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) || | |
6553 | (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) || | |
6554 | (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) || | |
6555 | (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) || | |
6556 | (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) || | |
6557 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) || | |
6558 | (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) || | |
6559 | (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) || | |
6560 | (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) || | |
6561 | (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) || | |
6562 | (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) || | |
6563 | (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) || | |
6564 | (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) || | |
6565 | (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) || | |
6566 | (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) || | |
6567 | (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) || | |
6568 | (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) || | |
6569 | (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) || | |
6570 | (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) || | |
6571 | (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) || | |
6572 | (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) || | |
6573 | (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) || | |
6574 | (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) || | |
6575 | (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) || | |
6576 | (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) || | |
6577 | (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) || | |
6578 | (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) || | |
6579 | (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c)) | |
6580 | ) /* How's that for a conditional? */ | |
78969a98 | 6581 | ) { |
ffc61ed2 JH |
6582 | match = TRUE; |
6583 | } | |
a0ed51b3 | 6584 | } |
a0ed51b3 LW |
6585 | } |
6586 | ||
7cdde544 | 6587 | /* If the bitmap didn't (or couldn't) match, and something outside the |
de87c4fe KW |
6588 | * bitmap could match, try that. Locale nodes specifiy completely the |
6589 | * behavior of code points in the bit map (otherwise, a utf8 target would | |
6590 | * cause them to be treated as Unicode and not locale), except XXX in | |
6591 | * the very unlikely event when this node is a synthetic start class, which | |
6592 | * could be a combination of locale and non-locale nodes */ | |
ef87b810 | 6593 | if (!match) { |
10ee90d2 KW |
6594 | if (utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256) { |
6595 | match = TRUE; /* Everything above 255 matches */ | |
e051a21d | 6596 | } |
10ee90d2 | 6597 | else if ((flags & ANYOF_NONBITMAP_NON_UTF8 |
137165a6 | 6598 | || (utf8_target && ANYOF_NONBITMAP(n) |
de87c4fe | 6599 | && (c >=256 || ! (flags & ANYOF_LOCALE))))) |
ef87b810 | 6600 | { |
7cdde544 KW |
6601 | AV *av; |
6602 | SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av); | |
6603 | ||
6604 | if (sw) { | |
6605 | U8 * utf8_p; | |
6606 | if (utf8_target) { | |
6607 | utf8_p = (U8 *) p; | |
6608 | } else { | |
f56b6394 KW |
6609 | |
6610 | /* Not utf8. Convert as much of the string as available up | |
6611 | * to the limit of how far the (single) character in the | |
6612 | * pattern can possibly match (no need to go further). If | |
6613 | * the node is a straight ANYOF or not folding, it can't | |
6614 | * match more than one. Otherwise, It can match up to how | |
6615 | * far a single char can fold to. Since not utf8, each | |
6616 | * character is a single byte, so the max it can be in | |
6617 | * bytes is the same as the max it can be in characters */ | |
6618 | STRLEN len = (OP(n) == ANYOF | |
6619 | || ! (flags & ANYOF_LOC_NONBITMAP_FOLD)) | |
6620 | ? 1 | |
6621 | : (maxlen < UTF8_MAX_FOLD_CHAR_EXPAND) | |
6622 | ? maxlen | |
6623 | : UTF8_MAX_FOLD_CHAR_EXPAND; | |
7cdde544 KW |
6624 | utf8_p = bytes_to_utf8(p, &len); |
6625 | } | |
f56b6394 | 6626 | |
f7eb7452 | 6627 | if (swash_fetch(sw, utf8_p, TRUE)) |
7cdde544 | 6628 | match = TRUE; |
39065660 | 6629 | else if (flags & ANYOF_LOC_NONBITMAP_FOLD) { |
f56b6394 KW |
6630 | |
6631 | /* Here, we need to test if the fold of the target string | |
6632 | * matches. In the case of a multi-char fold that is | |
6633 | * caught by regcomp.c, it has stored all such folds into | |
6634 | * 'av'; we linearly check to see if any match the target | |
6635 | * string (folded). We know that the originals were each | |
6636 | * one character, but we don't currently know how many | |
6637 | * characters/bytes each folded to, except we do know that | |
6638 | * there are small limits imposed by Unicode. XXX A | |
6639 | * performance enhancement would be to have regcomp.c store | |
6640 | * the max number of chars/bytes that are in an av entry, | |
f6ed098a KW |
6641 | * as, say the 0th element. Even better would be to have a |
6642 | * hash of the few characters that can start a multi-char | |
6643 | * fold to the max number of chars of those folds. | |
6644 | * | |
6645 | * Further down, if there isn't a | |
f56b6394 KW |
6646 | * match in the av, we will check if there is another |
6647 | * fold-type match. For that, we also need the fold, but | |
6648 | * only the first character. No sense in folding it twice, | |
6649 | * so we do it here, even if there isn't any multi-char | |
6650 | * fold, so we always fold at least the first character. | |
6651 | * If the node is a straight ANYOF node, or there is only | |
6652 | * one character available in the string, or if there isn't | |
6653 | * any av, that's all we have to fold. In the case of a | |
6654 | * multi-char fold, we do have guarantees in Unicode that | |
6655 | * it can only expand up to so many characters and so many | |
6656 | * bytes. We keep track so don't exceed either. | |
6657 | * | |
6658 | * If there is a match, we will need to advance (if lenp is | |
6659 | * specified) the match pointer in the target string. But | |
6660 | * what we are comparing here isn't that string directly, | |
6661 | * but its fold, whose length may differ from the original. | |
6662 | * As we go along in constructing the fold, therefore, we | |
6663 | * create a map so that we know how many bytes in the | |
6664 | * source to advance given that we have matched a certain | |
6665 | * number of bytes in the fold. This map is stored in | |
6666 | * 'map_fold_len_back'. The first character in the fold | |
6667 | * has array element 1 contain the number of bytes in the | |
6668 | * source that folded to it; the 2nd is the cumulative | |
6669 | * number to match it; ... */ | |
4d919e5b | 6670 | U8 map_fold_len_back[UTF8_MAX_FOLD_CHAR_EXPAND+1] = { 0 }; |
f56b6394 KW |
6671 | U8 folded[UTF8_MAXBYTES_CASE+1]; |
6672 | STRLEN foldlen = 0; /* num bytes in fold of 1st char */ | |
6673 | STRLEN foldlen_for_av; /* num bytes in fold of all chars */ | |
6674 | ||
6675 | if (OP(n) == ANYOF || maxlen == 1 || ! lenp || ! av) { | |
6676 | ||
6677 | /* Here, only need to fold the first char of the target | |
6678 | * string */ | |
6679 | to_utf8_fold(utf8_p, folded, &foldlen); | |
6680 | foldlen_for_av = foldlen; | |
6681 | map_fold_len_back[1] = UTF8SKIP(utf8_p); | |
6682 | } | |
6683 | else { | |
6684 | ||
6685 | /* Here, need to fold more than the first char. Do so | |
6686 | * up to the limits */ | |
6687 | UV which_char = 0; | |
6688 | U8* source_ptr = utf8_p; /* The source for the fold | |
6689 | is the regex target | |
6690 | string */ | |
6691 | U8* folded_ptr = folded; | |
6692 | U8* e = utf8_p + maxlen; /* Can't go beyond last | |
6693 | available byte in the | |
6694 | target string */ | |
6695 | while (which_char < UTF8_MAX_FOLD_CHAR_EXPAND | |
6696 | && source_ptr < e) | |
6697 | { | |
6698 | ||
6699 | /* Fold the next character */ | |
6700 | U8 this_char_folded[UTF8_MAXBYTES_CASE+1]; | |
6701 | STRLEN this_char_foldlen; | |
6702 | to_utf8_fold(source_ptr, | |
6703 | this_char_folded, | |
6704 | &this_char_foldlen); | |
6705 | ||
6706 | /* Bail if it would exceed the byte limit for | |
6707 | * folding a single char. */ | |
6708 | if (this_char_foldlen + folded_ptr - folded > | |
6709 | UTF8_MAXBYTES_CASE) | |
6710 | { | |
6711 | break; | |
6712 | } | |
6713 | ||
6714 | /* Save the first character's folded length, in | |
6715 | * case we have to use it later */ | |
6716 | if (! foldlen) { | |
6717 | foldlen = this_char_foldlen; | |
6718 | } | |
6719 | ||
6720 | /* Here, add the fold of this character */ | |
6721 | Copy(this_char_folded, | |
6722 | folded_ptr, | |
6723 | this_char_foldlen, | |
6724 | U8); | |
6725 | which_char++; | |
6726 | map_fold_len_back[which_char] = | |
6727 | map_fold_len_back[which_char - 1] | |
6728 | + UTF8SKIP(source_ptr); | |
6729 | folded_ptr += this_char_foldlen; | |
6730 | source_ptr += UTF8SKIP(source_ptr); | |
6731 | } | |
6732 | *folded_ptr = '\0'; | |
6733 | foldlen_for_av = folded_ptr - folded; | |
6734 | } | |
6735 | ||
6736 | ||
6737 | /* Do the linear search to see if the fold is in the list | |
6738 | * of multi-char folds. (Useless to look if won't be able | |
6739 | * to store that it is a multi-char fold in *lenp) */ | |
6740 | if (lenp && av) { | |
7cdde544 KW |
6741 | I32 i; |
6742 | for (i = 0; i <= av_len(av); i++) { | |
6743 | SV* const sv = *av_fetch(av, i, FALSE); | |
6744 | STRLEN len; | |
6745 | const char * const s = SvPV_const(sv, len); | |
f56b6394 KW |
6746 | if (len <= foldlen_for_av && memEQ(s, |
6747 | (char*)folded, | |
6748 | len)) | |
6749 | { | |
6750 | ||
6751 | /* Advance the target string ptr to account for | |
6752 | * this fold, but have to translate from the | |
6753 | * folded length to the corresponding source | |
6754 | * length. The array is indexed by how many | |
6755 | * characters in the match */ | |
6756 | *lenp = map_fold_len_back[ | |
6757 | utf8_length(folded, folded + len)]; | |
7cdde544 KW |
6758 | match = TRUE; |
6759 | break; | |
6760 | } | |
6761 | } | |
6762 | } | |
7cdde544 KW |
6763 | } |
6764 | ||
6765 | /* If we allocated a string above, free it */ | |
6766 | if (! utf8_target) Safefree(utf8_p); | |
6767 | } | |
6768 | } | |
6769 | } | |
6770 | ||
a0ed51b3 LW |
6771 | return (flags & ANYOF_INVERT) ? !match : match; |
6772 | } | |
161b471a | 6773 | |
dfe13c55 | 6774 | STATIC U8 * |
0ce71af7 | 6775 | S_reghop3(U8 *s, I32 off, const U8* lim) |
9041c2e3 | 6776 | { |
97aff369 | 6777 | dVAR; |
7918f24d NC |
6778 | |
6779 | PERL_ARGS_ASSERT_REGHOP3; | |
6780 | ||
a0ed51b3 | 6781 | if (off >= 0) { |
1aa99e6b | 6782 | while (off-- && s < lim) { |
ffc61ed2 | 6783 | /* XXX could check well-formedness here */ |
a0ed51b3 | 6784 | s += UTF8SKIP(s); |
ffc61ed2 | 6785 | } |
a0ed51b3 LW |
6786 | } |
6787 | else { | |
1de06328 YO |
6788 | while (off++ && s > lim) { |
6789 | s--; | |
6790 | if (UTF8_IS_CONTINUED(*s)) { | |
6791 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
6792 | s--; | |
a0ed51b3 | 6793 | } |
1de06328 | 6794 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
6795 | } |
6796 | } | |
6797 | return s; | |
6798 | } | |
161b471a | 6799 | |
f9f4320a YO |
6800 | #ifdef XXX_dmq |
6801 | /* there are a bunch of places where we use two reghop3's that should | |
6802 | be replaced with this routine. but since thats not done yet | |
6803 | we ifdef it out - dmq | |
6804 | */ | |
dfe13c55 | 6805 | STATIC U8 * |
1de06328 YO |
6806 | S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim) |
6807 | { | |
6808 | dVAR; | |
7918f24d NC |
6809 | |
6810 | PERL_ARGS_ASSERT_REGHOP4; | |
6811 | ||
1de06328 YO |
6812 | if (off >= 0) { |
6813 | while (off-- && s < rlim) { | |
6814 | /* XXX could check well-formedness here */ | |
6815 | s += UTF8SKIP(s); | |
6816 | } | |
6817 | } | |
6818 | else { | |
6819 | while (off++ && s > llim) { | |
6820 | s--; | |
6821 | if (UTF8_IS_CONTINUED(*s)) { | |
6822 | while (s > llim && UTF8_IS_CONTINUATION(*s)) | |
6823 | s--; | |
6824 | } | |
6825 | /* XXX could check well-formedness here */ | |
6826 | } | |
6827 | } | |
6828 | return s; | |
6829 | } | |
f9f4320a | 6830 | #endif |
1de06328 YO |
6831 | |
6832 | STATIC U8 * | |
0ce71af7 | 6833 | S_reghopmaybe3(U8* s, I32 off, const U8* lim) |
a0ed51b3 | 6834 | { |
97aff369 | 6835 | dVAR; |
7918f24d NC |
6836 | |
6837 | PERL_ARGS_ASSERT_REGHOPMAYBE3; | |
6838 | ||
a0ed51b3 | 6839 | if (off >= 0) { |
1aa99e6b | 6840 | while (off-- && s < lim) { |
ffc61ed2 | 6841 | /* XXX could check well-formedness here */ |
a0ed51b3 | 6842 | s += UTF8SKIP(s); |
ffc61ed2 | 6843 | } |
a0ed51b3 | 6844 | if (off >= 0) |
3dab1dad | 6845 | return NULL; |
a0ed51b3 LW |
6846 | } |
6847 | else { | |
1de06328 YO |
6848 | while (off++ && s > lim) { |
6849 | s--; | |
6850 | if (UTF8_IS_CONTINUED(*s)) { | |
6851 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
6852 | s--; | |
a0ed51b3 | 6853 | } |
1de06328 | 6854 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
6855 | } |
6856 | if (off <= 0) | |
3dab1dad | 6857 | return NULL; |
a0ed51b3 LW |
6858 | } |
6859 | return s; | |
6860 | } | |
51371543 | 6861 | |
51371543 | 6862 | static void |
acfe0abc | 6863 | restore_pos(pTHX_ void *arg) |
51371543 | 6864 | { |
97aff369 | 6865 | dVAR; |
097eb12c | 6866 | regexp * const rex = (regexp *)arg; |
51371543 GS |
6867 | if (PL_reg_eval_set) { |
6868 | if (PL_reg_oldsaved) { | |
4f639d21 DM |
6869 | rex->subbeg = PL_reg_oldsaved; |
6870 | rex->sublen = PL_reg_oldsavedlen; | |
f8c7b90f | 6871 | #ifdef PERL_OLD_COPY_ON_WRITE |
4f639d21 | 6872 | rex->saved_copy = PL_nrs; |
ed252734 | 6873 | #endif |
07bc277f | 6874 | RXp_MATCH_COPIED_on(rex); |
51371543 GS |
6875 | } |
6876 | PL_reg_magic->mg_len = PL_reg_oldpos; | |
6877 | PL_reg_eval_set = 0; | |
6878 | PL_curpm = PL_reg_oldcurpm; | |
6879 | } | |
6880 | } | |
33b8afdf JH |
6881 | |
6882 | STATIC void | |
6883 | S_to_utf8_substr(pTHX_ register regexp *prog) | |
6884 | { | |
a1cac82e | 6885 | int i = 1; |
7918f24d NC |
6886 | |
6887 | PERL_ARGS_ASSERT_TO_UTF8_SUBSTR; | |
6888 | ||
a1cac82e NC |
6889 | do { |
6890 | if (prog->substrs->data[i].substr | |
6891 | && !prog->substrs->data[i].utf8_substr) { | |
6892 | SV* const sv = newSVsv(prog->substrs->data[i].substr); | |
6893 | prog->substrs->data[i].utf8_substr = sv; | |
6894 | sv_utf8_upgrade(sv); | |
610460f9 NC |
6895 | if (SvVALID(prog->substrs->data[i].substr)) { |
6896 | const U8 flags = BmFLAGS(prog->substrs->data[i].substr); | |
6897 | if (flags & FBMcf_TAIL) { | |
6898 | /* Trim the trailing \n that fbm_compile added last | |
6899 | time. */ | |
6900 | SvCUR_set(sv, SvCUR(sv) - 1); | |
6901 | /* Whilst this makes the SV technically "invalid" (as its | |
6902 | buffer is no longer followed by "\0") when fbm_compile() | |
6903 | adds the "\n" back, a "\0" is restored. */ | |
6904 | } | |
6905 | fbm_compile(sv, flags); | |
6906 | } | |
a1cac82e NC |
6907 | if (prog->substrs->data[i].substr == prog->check_substr) |
6908 | prog->check_utf8 = sv; | |
6909 | } | |
6910 | } while (i--); | |
33b8afdf JH |
6911 | } |
6912 | ||
6913 | STATIC void | |
6914 | S_to_byte_substr(pTHX_ register regexp *prog) | |
6915 | { | |
97aff369 | 6916 | dVAR; |
a1cac82e | 6917 | int i = 1; |
7918f24d NC |
6918 | |
6919 | PERL_ARGS_ASSERT_TO_BYTE_SUBSTR; | |
6920 | ||
a1cac82e NC |
6921 | do { |
6922 | if (prog->substrs->data[i].utf8_substr | |
6923 | && !prog->substrs->data[i].substr) { | |
6924 | SV* sv = newSVsv(prog->substrs->data[i].utf8_substr); | |
6925 | if (sv_utf8_downgrade(sv, TRUE)) { | |
610460f9 NC |
6926 | if (SvVALID(prog->substrs->data[i].utf8_substr)) { |
6927 | const U8 flags | |
6928 | = BmFLAGS(prog->substrs->data[i].utf8_substr); | |
6929 | if (flags & FBMcf_TAIL) { | |
6930 | /* Trim the trailing \n that fbm_compile added last | |
6931 | time. */ | |
6932 | SvCUR_set(sv, SvCUR(sv) - 1); | |
6933 | } | |
6934 | fbm_compile(sv, flags); | |
6935 | } | |
a1cac82e NC |
6936 | } else { |
6937 | SvREFCNT_dec(sv); | |
6938 | sv = &PL_sv_undef; | |
6939 | } | |
6940 | prog->substrs->data[i].substr = sv; | |
6941 | if (prog->substrs->data[i].utf8_substr == prog->check_utf8) | |
6942 | prog->check_substr = sv; | |
33b8afdf | 6943 | } |
a1cac82e | 6944 | } while (i--); |
33b8afdf | 6945 | } |
66610fdd RGS |
6946 | |
6947 | /* | |
6948 | * Local variables: | |
6949 | * c-indentation-style: bsd | |
6950 | * c-basic-offset: 4 | |
6951 | * indent-tabs-mode: t | |
6952 | * End: | |
6953 | * | |
37442d52 RGS |
6954 | * ex: set ts=8 sts=4 sw=4 noet: |
6955 | */ |