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 | 125 | #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ |
9c4fdda1 RB |
126 | if (!CAT2(PL_utf8_,class)) { \ |
127 | bool ok; \ | |
128 | ENTER; save_re_context(); \ | |
129 | ok=CAT2(is_utf8_,class)((const U8*)str); \ | |
130 | assert(ok); LEAVE; } } STMT_END | |
37e2e78e KW |
131 | |
132 | /* Doesn't do an assert to verify that is correct */ | |
133 | #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \ | |
9c4fdda1 RB |
134 | if (!CAT2(PL_utf8_,class)) { \ |
135 | bool throw_away __attribute__unused__; \ | |
136 | ENTER; save_re_context(); \ | |
137 | throw_away = CAT2(is_utf8_,class)((const U8*)" "); \ | |
138 | LEAVE; } } STMT_END | |
37e2e78e | 139 | |
1a4fad37 AL |
140 | #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a") |
141 | #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0") | |
142 | #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ") | |
51371543 | 143 | |
37e2e78e | 144 | #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \ |
d275fa5e CB |
145 | LOAD_UTF8_CHARCLASS(X_begin, " "); \ |
146 | LOAD_UTF8_CHARCLASS(X_non_hangul, "A"); \ | |
37e2e78e KW |
147 | /* These are utf8 constants, and not utf-ebcdic constants, so the \ |
148 | * assert should likely and hopefully fail on an EBCDIC machine */ \ | |
d275fa5e | 149 | LOAD_UTF8_CHARCLASS(X_extend, "\xcc\x80"); /* U+0300 */ \ |
37e2e78e KW |
150 | \ |
151 | /* No asserts are done for these, in case called on an early \ | |
152 | * Unicode version in which they map to nothing */ \ | |
d275fa5e CB |
153 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_prepend);/* U+0E40 "\xe0\xb9\x80" */ \ |
154 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_L); /* U+1100 "\xe1\x84\x80" */ \ | |
155 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV); /* U+AC00 "\xea\xb0\x80" */ \ | |
156 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LVT); /* U+AC01 "\xea\xb0\x81" */ \ | |
157 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV_LVT_V);/* U+AC01 "\xea\xb0\x81" */\ | |
158 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_T); /* U+11A8 "\xe1\x86\xa8" */ \ | |
37e2e78e | 159 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_V) /* U+1160 "\xe1\x85\xa0" */ |
20d0b1e9 | 160 | |
1dcf4a1b | 161 | #define PLACEHOLDER /* Something for the preprocessor to grab onto */ |
d1eb3177 | 162 | |
ee9a90b8 KW |
163 | /* The actual code for CCC_TRY, which uses several variables from the routine |
164 | * it's callable from. It is designed to be the bulk of a case statement. | |
165 | * FUNC is the macro or function to call on non-utf8 targets that indicate if | |
166 | * nextchr matches the class. | |
167 | * UTF8_TEST is the whole test string to use for utf8 targets | |
168 | * LOAD is what to use to test, and if not present to load in the swash for the | |
169 | * class | |
170 | * POS_OR_NEG is either empty or ! to complement the results of FUNC or | |
171 | * UTF8_TEST test. | |
172 | * The logic is: Fail if we're at the end-of-string; otherwise if the target is | |
173 | * utf8 and a variant, load the swash if necessary and test using the utf8 | |
174 | * test. Advance to the next character if test is ok, otherwise fail; If not | |
175 | * utf8 or an invariant under utf8, use the non-utf8 test, and fail if it | |
176 | * fails, or advance to the next character */ | |
177 | ||
178 | #define _CCC_TRY_CODE(POS_OR_NEG, FUNC, UTF8_TEST, CLASS, STR) \ | |
179 | if (locinput >= PL_regeol) { \ | |
180 | sayNO; \ | |
181 | } \ | |
182 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ | |
183 | LOAD_UTF8_CHARCLASS(CLASS, STR); \ | |
184 | if (POS_OR_NEG (UTF8_TEST)) { \ | |
185 | sayNO; \ | |
186 | } \ | |
187 | locinput += PL_utf8skip[nextchr]; \ | |
188 | nextchr = UCHARAT(locinput); \ | |
189 | break; \ | |
190 | } \ | |
191 | if (POS_OR_NEG (FUNC(nextchr))) { \ | |
192 | sayNO; \ | |
193 | } \ | |
194 | nextchr = UCHARAT(++locinput); \ | |
980866de KW |
195 | break; |
196 | ||
ee9a90b8 KW |
197 | /* Handle the non-locale cases for a character class and its complement. It |
198 | * calls _CCC_TRY_CODE with a ! to complement the test for the character class. | |
199 | * This is because that code fails when the test succeeds, so we want to have | |
200 | * the test fail so that the code succeeds. The swash is stored in a | |
201 | * predictable PL_ place */ | |
cfaf538b KW |
202 | #define _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, \ |
203 | CLASS, STR) \ | |
ee9a90b8 KW |
204 | case NAME: \ |
205 | _CCC_TRY_CODE( !, FUNC, \ | |
206 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ | |
207 | (U8*)locinput, TRUE)), \ | |
208 | CLASS, STR) \ | |
209 | case NNAME: \ | |
1dcf4a1b | 210 | _CCC_TRY_CODE( PLACEHOLDER , FUNC, \ |
ee9a90b8 KW |
211 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ |
212 | (U8*)locinput, TRUE)), \ | |
213 | CLASS, STR) \ | |
214 | ||
215 | /* Generate the case statements for both locale and non-locale character | |
216 | * classes in regmatch for classes that don't have special unicode semantics. | |
217 | * Locales don't use an immediate swash, but an intermediary special locale | |
218 | * function that is called on the pointer to the current place in the input | |
219 | * string. That function will resolve to needing the same swash. One might | |
220 | * think that because we don't know what the locale will match, we shouldn't | |
221 | * check with the swash loading function that it loaded properly; ie, that we | |
222 | * should use LOAD_UTF8_CHARCLASS_NO_CHECK for those, but what is passed to the | |
223 | * regular LOAD_UTF8_CHARCLASS is in non-locale terms, and so locale is | |
224 | * irrelevant here */ | |
225 | #define CCC_TRY(NAME, NNAME, FUNC, \ | |
226 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
cfaf538b | 227 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 KW |
228 | CLASS, STR) \ |
229 | case NAMEL: \ | |
230 | PL_reg_flags |= RF_tainted; \ | |
231 | _CCC_TRY_CODE( !, LCFUNC, LCFUNC_utf8((U8*)locinput), CLASS, STR) \ | |
232 | case NNAMEL: \ | |
233 | PL_reg_flags |= RF_tainted; \ | |
1dcf4a1b KW |
234 | _CCC_TRY_CODE( PLACEHOLDER, LCFUNC, LCFUNC_utf8((U8*)locinput), \ |
235 | CLASS, STR) \ | |
cfaf538b KW |
236 | case NAMEA: \ |
237 | if (locinput >= PL_regeol || ! FUNCA(nextchr)) { \ | |
238 | sayNO; \ | |
239 | } \ | |
240 | /* Matched a utf8-invariant, so don't have to worry about utf8 */ \ | |
241 | nextchr = UCHARAT(++locinput); \ | |
242 | break; \ | |
243 | case NNAMEA: \ | |
244 | if (locinput >= PL_regeol || FUNCA(nextchr)) { \ | |
245 | sayNO; \ | |
246 | } \ | |
247 | if (utf8_target) { \ | |
248 | locinput += PL_utf8skip[nextchr]; \ | |
249 | nextchr = UCHARAT(locinput); \ | |
250 | } \ | |
251 | else { \ | |
252 | nextchr = UCHARAT(++locinput); \ | |
253 | } \ | |
254 | break; \ | |
ee9a90b8 KW |
255 | /* Generate the non-locale cases */ \ |
256 | _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, CLASS, STR) | |
257 | ||
258 | /* This is like CCC_TRY, but has an extra set of parameters for generating case | |
259 | * statements to handle separate Unicode semantics nodes */ | |
260 | #define CCC_TRY_U(NAME, NNAME, FUNC, \ | |
261 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
262 | NAMEU, NNAMEU, FUNCU, \ | |
cfaf538b | 263 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 | 264 | CLASS, STR) \ |
cfaf538b KW |
265 | CCC_TRY(NAME, NNAME, FUNC, \ |
266 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
267 | NAMEA, NNAMEA, FUNCA, \ | |
268 | CLASS, STR) \ | |
ee9a90b8 | 269 | _CCC_TRY_NONLOCALE(NAMEU, NNAMEU, FUNCU, CLASS, STR) |
d1eb3177 | 270 | |
3dab1dad YO |
271 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
272 | ||
5f80c4cf | 273 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
c35dcbe2 YO |
274 | /* it would be nice to rework regcomp.sym to generate this stuff. sigh |
275 | * | |
276 | * NOTE that *nothing* that affects backtracking should be in here, specifically | |
277 | * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a | |
278 | * node that is in between two EXACT like nodes when ascertaining what the required | |
279 | * "follow" character is. This should probably be moved to regex compile time | |
280 | * although it may be done at run time beause of the REF possibility - more | |
281 | * investigation required. -- demerphq | |
282 | */ | |
3e901dc0 YO |
283 | #define JUMPABLE(rn) ( \ |
284 | OP(rn) == OPEN || \ | |
285 | (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \ | |
286 | OP(rn) == EVAL || \ | |
cca55fe3 JP |
287 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
288 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
d1c771f5 | 289 | OP(rn) == KEEPS || \ |
3dab1dad | 290 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 | 291 | ) |
ee9b8eae | 292 | #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT) |
e2d8ce26 | 293 | |
ee9b8eae YO |
294 | #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF ) |
295 | ||
296 | #if 0 | |
297 | /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so | |
298 | we don't need this definition. */ | |
299 | #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) | |
2f7f8cb1 | 300 | #define IS_TEXTF(rn) ( (OP(rn)==EXACTFU || OP(rn)==EXACTFA || OP(rn)==EXACTF) || OP(rn)==REFF || OP(rn)==NREFF ) |
ee9b8eae YO |
301 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) |
302 | ||
303 | #else | |
304 | /* ... so we use this as its faster. */ | |
305 | #define IS_TEXT(rn) ( OP(rn)==EXACT ) | |
2f7f8cb1 | 306 | #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn) == EXACTFA) |
ee9b8eae YO |
307 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) |
308 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) | |
309 | ||
310 | #endif | |
e2d8ce26 | 311 | |
a84d97b6 HS |
312 | /* |
313 | Search for mandatory following text node; for lookahead, the text must | |
314 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
315 | */ | |
cca55fe3 | 316 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
317 | while (JUMPABLE(rn)) { \ |
318 | const OPCODE type = OP(rn); \ | |
319 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 320 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 321 | else if (type == PLUS) \ |
cca55fe3 | 322 | rn = NEXTOPER(rn); \ |
3dab1dad | 323 | else if (type == IFMATCH) \ |
a84d97b6 | 324 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 325 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 326 | } \ |
5f80c4cf | 327 | } STMT_END |
74750237 | 328 | |
c476f425 | 329 | |
acfe0abc | 330 | static void restore_pos(pTHX_ void *arg); |
51371543 | 331 | |
620d5b66 NC |
332 | #define REGCP_PAREN_ELEMS 4 |
333 | #define REGCP_OTHER_ELEMS 5 | |
e0fa7e2b | 334 | #define REGCP_FRAME_ELEMS 1 |
620d5b66 NC |
335 | /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and |
336 | * are needed for the regexp context stack bookkeeping. */ | |
337 | ||
76e3520e | 338 | STATIC CHECKPOINT |
cea2e8a9 | 339 | S_regcppush(pTHX_ I32 parenfloor) |
a0d0e21e | 340 | { |
97aff369 | 341 | dVAR; |
a3b680e6 | 342 | const int retval = PL_savestack_ix; |
a3b680e6 | 343 | const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; |
e0fa7e2b NC |
344 | const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS; |
345 | const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT; | |
a0d0e21e | 346 | int p; |
40a82448 | 347 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 348 | |
e49a9654 IH |
349 | if (paren_elems_to_push < 0) |
350 | Perl_croak(aTHX_ "panic: paren_elems_to_push < 0"); | |
351 | ||
e0fa7e2b NC |
352 | if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems) |
353 | Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf | |
5df417d0 JH |
354 | " out of range (%lu-%ld)", |
355 | total_elems, (unsigned long)PL_regsize, (long)parenfloor); | |
e0fa7e2b | 356 | |
620d5b66 | 357 | SSGROW(total_elems + REGCP_FRAME_ELEMS); |
7f69552c | 358 | |
3280af22 | 359 | for (p = PL_regsize; p > parenfloor; p--) { |
b1ce53c5 | 360 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
f0ab9afb NC |
361 | SSPUSHINT(PL_regoffs[p].end); |
362 | SSPUSHINT(PL_regoffs[p].start); | |
3280af22 | 363 | SSPUSHPTR(PL_reg_start_tmp[p]); |
a0d0e21e | 364 | SSPUSHINT(p); |
e7707071 | 365 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
40a82448 | 366 | " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n", |
f0ab9afb | 367 | (UV)p, (IV)PL_regoffs[p].start, |
40a82448 | 368 | (IV)(PL_reg_start_tmp[p] - PL_bostr), |
f0ab9afb | 369 | (IV)PL_regoffs[p].end |
40a82448 | 370 | )); |
a0d0e21e | 371 | } |
b1ce53c5 | 372 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
f0ab9afb | 373 | SSPUSHPTR(PL_regoffs); |
3280af22 NIS |
374 | SSPUSHINT(PL_regsize); |
375 | SSPUSHINT(*PL_reglastparen); | |
a01268b5 | 376 | SSPUSHINT(*PL_reglastcloseparen); |
3280af22 | 377 | SSPUSHPTR(PL_reginput); |
e0fa7e2b | 378 | SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */ |
41123dfd | 379 | |
a0d0e21e LW |
380 | return retval; |
381 | } | |
382 | ||
c277df42 | 383 | /* These are needed since we do not localize EVAL nodes: */ |
ab3bbdeb YO |
384 | #define REGCP_SET(cp) \ |
385 | DEBUG_STATE_r( \ | |
ab3bbdeb | 386 | PerlIO_printf(Perl_debug_log, \ |
e4f74956 | 387 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
ab3bbdeb YO |
388 | (IV)PL_savestack_ix)); \ |
389 | cp = PL_savestack_ix | |
c3464db5 | 390 | |
ab3bbdeb | 391 | #define REGCP_UNWIND(cp) \ |
e4f74956 | 392 | DEBUG_STATE_r( \ |
ab3bbdeb | 393 | if (cp != PL_savestack_ix) \ |
e4f74956 YO |
394 | PerlIO_printf(Perl_debug_log, \ |
395 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ | |
ab3bbdeb YO |
396 | (IV)(cp), (IV)PL_savestack_ix)); \ |
397 | regcpblow(cp) | |
c277df42 | 398 | |
76e3520e | 399 | STATIC char * |
097eb12c | 400 | S_regcppop(pTHX_ const regexp *rex) |
a0d0e21e | 401 | { |
97aff369 | 402 | dVAR; |
e0fa7e2b | 403 | UV i; |
a0d0e21e | 404 | char *input; |
a3621e74 YO |
405 | GET_RE_DEBUG_FLAGS_DECL; |
406 | ||
7918f24d NC |
407 | PERL_ARGS_ASSERT_REGCPPOP; |
408 | ||
b1ce53c5 | 409 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
c6bf6a65 | 410 | i = SSPOPUV; |
e0fa7e2b NC |
411 | assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
412 | i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */ | |
a0d0e21e | 413 | input = (char *) SSPOPPTR; |
a01268b5 | 414 | *PL_reglastcloseparen = SSPOPINT; |
3280af22 NIS |
415 | *PL_reglastparen = SSPOPINT; |
416 | PL_regsize = SSPOPINT; | |
f0ab9afb | 417 | PL_regoffs=(regexp_paren_pair *) SSPOPPTR; |
b1ce53c5 | 418 | |
620d5b66 | 419 | i -= REGCP_OTHER_ELEMS; |
b1ce53c5 | 420 | /* Now restore the parentheses context. */ |
620d5b66 | 421 | for ( ; i > 0; i -= REGCP_PAREN_ELEMS) { |
1df70142 | 422 | I32 tmps; |
097eb12c | 423 | U32 paren = (U32)SSPOPINT; |
3280af22 | 424 | PL_reg_start_tmp[paren] = (char *) SSPOPPTR; |
f0ab9afb | 425 | PL_regoffs[paren].start = SSPOPINT; |
cf93c79d | 426 | tmps = SSPOPINT; |
3280af22 | 427 | if (paren <= *PL_reglastparen) |
f0ab9afb | 428 | PL_regoffs[paren].end = tmps; |
e7707071 | 429 | DEBUG_BUFFERS_r( |
c3464db5 | 430 | PerlIO_printf(Perl_debug_log, |
b900a521 | 431 | " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n", |
f0ab9afb | 432 | (UV)paren, (IV)PL_regoffs[paren].start, |
b900a521 | 433 | (IV)(PL_reg_start_tmp[paren] - PL_bostr), |
f0ab9afb | 434 | (IV)PL_regoffs[paren].end, |
3280af22 | 435 | (paren > *PL_reglastparen ? "(no)" : "")); |
c277df42 | 436 | ); |
a0d0e21e | 437 | } |
e7707071 | 438 | DEBUG_BUFFERS_r( |
bb7a0f54 | 439 | if (*PL_reglastparen + 1 <= rex->nparens) { |
c3464db5 | 440 | PerlIO_printf(Perl_debug_log, |
faccc32b | 441 | " restoring \\%"IVdf"..\\%"IVdf" to undef\n", |
4f639d21 | 442 | (IV)(*PL_reglastparen + 1), (IV)rex->nparens); |
c277df42 IZ |
443 | } |
444 | ); | |
daf18116 | 445 | #if 1 |
dafc8851 JH |
446 | /* It would seem that the similar code in regtry() |
447 | * already takes care of this, and in fact it is in | |
448 | * a better location to since this code can #if 0-ed out | |
449 | * but the code in regtry() is needed or otherwise tests | |
450 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
451 | * (as of patchlevel 7877) will fail. Then again, |
452 | * this code seems to be necessary or otherwise | |
225593e1 DM |
453 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ |
454 | * --jhi updated by dapm */ | |
3b6647e0 | 455 | for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) { |
097eb12c | 456 | if (i > PL_regsize) |
f0ab9afb NC |
457 | PL_regoffs[i].start = -1; |
458 | PL_regoffs[i].end = -1; | |
a0d0e21e | 459 | } |
dafc8851 | 460 | #endif |
a0d0e21e LW |
461 | return input; |
462 | } | |
463 | ||
02db2b7b | 464 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 465 | |
a687059c | 466 | /* |
e50aee73 | 467 | * pregexec and friends |
a687059c LW |
468 | */ |
469 | ||
76234dfb | 470 | #ifndef PERL_IN_XSUB_RE |
a687059c | 471 | /* |
c277df42 | 472 | - pregexec - match a regexp against a string |
a687059c | 473 | */ |
c277df42 | 474 | I32 |
49d7dfbc | 475 | Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend, |
c3464db5 | 476 | char *strbeg, I32 minend, SV *screamer, U32 nosave) |
c277df42 IZ |
477 | /* strend: pointer to null at end of string */ |
478 | /* strbeg: real beginning of string */ | |
479 | /* minend: end of match must be >=minend after stringarg. */ | |
480 | /* nosave: For optimizations. */ | |
481 | { | |
7918f24d NC |
482 | PERL_ARGS_ASSERT_PREGEXEC; |
483 | ||
c277df42 | 484 | return |
9041c2e3 | 485 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
486 | nosave ? 0 : REXEC_COPY_STR); |
487 | } | |
76234dfb | 488 | #endif |
22e551b9 | 489 | |
9041c2e3 | 490 | /* |
cad2e5aa JH |
491 | * Need to implement the following flags for reg_anch: |
492 | * | |
493 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
494 | * USE_INTUIT_ML | |
495 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
496 | * INTUIT_AUTORITATIVE_ML | |
497 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
498 | * INTUIT_ONCE_ML | |
499 | * | |
500 | * Another flag for this function: SECOND_TIME (so that float substrs | |
501 | * with giant delta may be not rechecked). | |
502 | */ | |
503 | ||
504 | /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ | |
505 | ||
3f7c398e | 506 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
507 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
508 | ||
509 | /* XXXX We assume that strpos is strbeg unless sv. */ | |
510 | ||
6eb5f6b9 JH |
511 | /* XXXX Some places assume that there is a fixed substring. |
512 | An update may be needed if optimizer marks as "INTUITable" | |
513 | RExen without fixed substrings. Similarly, it is assumed that | |
514 | lengths of all the strings are no more than minlen, thus they | |
515 | cannot come from lookahead. | |
40d049e4 YO |
516 | (Or minlen should take into account lookahead.) |
517 | NOTE: Some of this comment is not correct. minlen does now take account | |
518 | of lookahead/behind. Further research is required. -- demerphq | |
519 | ||
520 | */ | |
6eb5f6b9 | 521 | |
2c2d71f5 JH |
522 | /* A failure to find a constant substring means that there is no need to make |
523 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
524 | finding a substring too deep into the string means that less calls to | |
30944b6d IZ |
525 | regtry() should be needed. |
526 | ||
527 | REx compiler's optimizer found 4 possible hints: | |
528 | a) Anchored substring; | |
529 | b) Fixed substring; | |
530 | c) Whether we are anchored (beginning-of-line or \G); | |
486ec47a | 531 | d) First node (of those at offset 0) which may distinguish positions; |
6eb5f6b9 | 532 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
533 | string which does not contradict any of them. |
534 | */ | |
2c2d71f5 | 535 | |
6eb5f6b9 JH |
536 | /* Most of decisions we do here should have been done at compile time. |
537 | The nodes of the REx which we used for the search should have been | |
538 | deleted from the finite automaton. */ | |
539 | ||
cad2e5aa | 540 | char * |
288b8c02 | 541 | Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, |
9f61653a | 542 | char *strend, const U32 flags, re_scream_pos_data *data) |
cad2e5aa | 543 | { |
97aff369 | 544 | dVAR; |
288b8c02 | 545 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
b7953727 | 546 | register I32 start_shift = 0; |
cad2e5aa | 547 | /* Should be nonnegative! */ |
b7953727 | 548 | register I32 end_shift = 0; |
2c2d71f5 JH |
549 | register char *s; |
550 | register SV *check; | |
a1933d95 | 551 | char *strbeg; |
cad2e5aa | 552 | char *t; |
f2ed9b32 | 553 | const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 554 | I32 ml_anch; |
bd61b366 SS |
555 | register char *other_last = NULL; /* other substr checked before this */ |
556 | char *check_at = NULL; /* check substr found at this pos */ | |
bbe252da | 557 | const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; |
f8fc2ecf | 558 | RXi_GET_DECL(prog,progi); |
30944b6d | 559 | #ifdef DEBUGGING |
890ce7af | 560 | const char * const i_strpos = strpos; |
30944b6d | 561 | #endif |
a3621e74 YO |
562 | GET_RE_DEBUG_FLAGS_DECL; |
563 | ||
7918f24d NC |
564 | PERL_ARGS_ASSERT_RE_INTUIT_START; |
565 | ||
f2ed9b32 | 566 | RX_MATCH_UTF8_set(rx,utf8_target); |
cad2e5aa | 567 | |
3c8556c3 | 568 | if (RX_UTF8(rx)) { |
b8d68ded JH |
569 | PL_reg_flags |= RF_utf8; |
570 | } | |
ab3bbdeb | 571 | DEBUG_EXECUTE_r( |
f2ed9b32 | 572 | debug_start_match(rx, utf8_target, strpos, strend, |
1de06328 YO |
573 | sv ? "Guessing start of match in sv for" |
574 | : "Guessing start of match in string for"); | |
2a782b5b | 575 | ); |
cad2e5aa | 576 | |
c344f387 JH |
577 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
578 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 579 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 580 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 581 | goto fail; |
2c2d71f5 | 582 | } |
1de06328 | 583 | |
a1933d95 | 584 | strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; |
1aa99e6b | 585 | PL_regeol = strend; |
f2ed9b32 | 586 | if (utf8_target) { |
33b8afdf JH |
587 | if (!prog->check_utf8 && prog->check_substr) |
588 | to_utf8_substr(prog); | |
589 | check = prog->check_utf8; | |
590 | } else { | |
591 | if (!prog->check_substr && prog->check_utf8) | |
592 | to_byte_substr(prog); | |
593 | check = prog->check_substr; | |
594 | } | |
1de06328 | 595 | if (check == &PL_sv_undef) { |
a3621e74 | 596 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
1de06328 | 597 | "Non-utf8 string cannot match utf8 check string\n")); |
33b8afdf JH |
598 | goto fail; |
599 | } | |
bbe252da YO |
600 | if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */ |
601 | ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) | |
602 | || ( (prog->extflags & RXf_ANCH_BOL) | |
7fba1cd6 | 603 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 604 | |
7e25d62c | 605 | if (!ml_anch) { |
bbe252da YO |
606 | if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */ |
607 | && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ | |
3f7c398e | 608 | /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ |
7e25d62c JH |
609 | && sv && !SvROK(sv) |
610 | && (strpos != strbeg)) { | |
a3621e74 | 611 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
612 | goto fail; |
613 | } | |
614 | if (prog->check_offset_min == prog->check_offset_max && | |
bbe252da | 615 | !(prog->extflags & RXf_CANY_SEEN)) { |
2c2d71f5 | 616 | /* Substring at constant offset from beg-of-str... */ |
cad2e5aa JH |
617 | I32 slen; |
618 | ||
1aa99e6b | 619 | s = HOP3c(strpos, prog->check_offset_min, strend); |
1de06328 | 620 | |
653099ff GS |
621 | if (SvTAIL(check)) { |
622 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 623 | |
9041c2e3 | 624 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 625 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 626 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 627 | goto fail_finish; |
cad2e5aa JH |
628 | } |
629 | /* Now should match s[0..slen-2] */ | |
630 | slen--; | |
3f7c398e | 631 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 632 | || (slen > 1 |
3f7c398e | 633 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 634 | report_neq: |
a3621e74 | 635 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
636 | goto fail_finish; |
637 | } | |
cad2e5aa | 638 | } |
3f7c398e | 639 | else if (*SvPVX_const(check) != *s |
653099ff | 640 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 641 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 642 | goto report_neq; |
c315bfe8 | 643 | check_at = s; |
2c2d71f5 | 644 | goto success_at_start; |
7e25d62c | 645 | } |
cad2e5aa | 646 | } |
2c2d71f5 | 647 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 648 | s = strpos; |
2c2d71f5 | 649 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
1de06328 YO |
650 | end_shift = prog->check_end_shift; |
651 | ||
2c2d71f5 | 652 | if (!ml_anch) { |
a3b680e6 | 653 | const I32 end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 654 | - (SvTAIL(check) != 0); |
a3b680e6 | 655 | const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
656 | |
657 | if (end_shift < eshift) | |
658 | end_shift = eshift; | |
659 | } | |
cad2e5aa | 660 | } |
2c2d71f5 | 661 | else { /* Can match at random position */ |
cad2e5aa JH |
662 | ml_anch = 0; |
663 | s = strpos; | |
1de06328 YO |
664 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
665 | end_shift = prog->check_end_shift; | |
666 | ||
667 | /* end shift should be non negative here */ | |
cad2e5aa JH |
668 | } |
669 | ||
bcdf7404 | 670 | #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 671 | if (end_shift < 0) |
1de06328 | 672 | Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", |
220fc49f | 673 | (IV)end_shift, RX_PRECOMP(prog)); |
2c2d71f5 JH |
674 | #endif |
675 | ||
2c2d71f5 JH |
676 | restart: |
677 | /* Find a possible match in the region s..strend by looking for | |
678 | the "check" substring in the region corrected by start/end_shift. */ | |
1de06328 YO |
679 | |
680 | { | |
681 | I32 srch_start_shift = start_shift; | |
682 | I32 srch_end_shift = end_shift; | |
683 | if (srch_start_shift < 0 && strbeg - s > srch_start_shift) { | |
684 | srch_end_shift -= ((strbeg - s) - srch_start_shift); | |
685 | srch_start_shift = strbeg - s; | |
686 | } | |
6bda09f9 | 687 | DEBUG_OPTIMISE_MORE_r({ |
1de06328 YO |
688 | PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n", |
689 | (IV)prog->check_offset_min, | |
690 | (IV)srch_start_shift, | |
691 | (IV)srch_end_shift, | |
692 | (IV)prog->check_end_shift); | |
693 | }); | |
694 | ||
cad2e5aa | 695 | if (flags & REXEC_SCREAM) { |
cad2e5aa | 696 | I32 p = -1; /* Internal iterator of scream. */ |
a3b680e6 | 697 | I32 * const pp = data ? data->scream_pos : &p; |
cad2e5aa | 698 | |
2c2d71f5 JH |
699 | if (PL_screamfirst[BmRARE(check)] >= 0 |
700 | || ( BmRARE(check) == '\n' | |
85c508c3 | 701 | && (BmPREVIOUS(check) == SvCUR(check) - 1) |
2c2d71f5 | 702 | && SvTAIL(check) )) |
9041c2e3 | 703 | s = screaminstr(sv, check, |
1de06328 | 704 | srch_start_shift + (s - strbeg), srch_end_shift, pp, 0); |
cad2e5aa | 705 | else |
2c2d71f5 | 706 | goto fail_finish; |
4addbd3b | 707 | /* we may be pointing at the wrong string */ |
07bc277f | 708 | if (s && RXp_MATCH_COPIED(prog)) |
3f7c398e | 709 | s = strbeg + (s - SvPVX_const(sv)); |
cad2e5aa JH |
710 | if (data) |
711 | *data->scream_olds = s; | |
712 | } | |
1de06328 YO |
713 | else { |
714 | U8* start_point; | |
715 | U8* end_point; | |
bbe252da | 716 | if (prog->extflags & RXf_CANY_SEEN) { |
1de06328 YO |
717 | start_point= (U8*)(s + srch_start_shift); |
718 | end_point= (U8*)(strend - srch_end_shift); | |
719 | } else { | |
720 | start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); | |
721 | end_point= HOP3(strend, -srch_end_shift, strbeg); | |
722 | } | |
6bda09f9 | 723 | DEBUG_OPTIMISE_MORE_r({ |
56570a2c | 724 | PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", |
1de06328 | 725 | (int)(end_point - start_point), |
fc8cd66c | 726 | (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), |
1de06328 YO |
727 | start_point); |
728 | }); | |
729 | ||
730 | s = fbm_instr( start_point, end_point, | |
7fba1cd6 | 731 | check, multiline ? FBMrf_MULTILINE : 0); |
1de06328 YO |
732 | } |
733 | } | |
cad2e5aa JH |
734 | /* Update the count-of-usability, remove useless subpatterns, |
735 | unshift s. */ | |
2c2d71f5 | 736 | |
ab3bbdeb | 737 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 738 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
739 | SvPVX_const(check), RE_SV_DUMPLEN(check), 30); |
740 | PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", | |
2c2d71f5 | 741 | (s ? "Found" : "Did not find"), |
f2ed9b32 | 742 | (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) |
ab3bbdeb YO |
743 | ? "anchored" : "floating"), |
744 | quoted, | |
745 | RE_SV_TAIL(check), | |
746 | (s ? " at offset " : "...\n") ); | |
747 | }); | |
2c2d71f5 JH |
748 | |
749 | if (!s) | |
750 | goto fail_finish; | |
2c2d71f5 | 751 | /* Finish the diagnostic message */ |
a3621e74 | 752 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 | 753 | |
1de06328 YO |
754 | /* XXX dmq: first branch is for positive lookbehind... |
755 | Our check string is offset from the beginning of the pattern. | |
756 | So we need to do any stclass tests offset forward from that | |
757 | point. I think. :-( | |
758 | */ | |
759 | ||
760 | ||
761 | ||
762 | check_at=s; | |
763 | ||
764 | ||
2c2d71f5 JH |
765 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. |
766 | Start with the other substr. | |
767 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 768 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
769 | *always* match. Probably should be marked during compile... |
770 | Probably it is right to do no SCREAM here... | |
771 | */ | |
772 | ||
f2ed9b32 | 773 | if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8) |
1de06328 YO |
774 | : (prog->float_substr && prog->anchored_substr)) |
775 | { | |
30944b6d | 776 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
777 | /* XXXX May be hopelessly wrong for UTF... */ |
778 | if (!other_last) | |
6eb5f6b9 | 779 | other_last = strpos; |
f2ed9b32 | 780 | if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
781 | do_other_anchored: |
782 | { | |
890ce7af AL |
783 | char * const last = HOP3c(s, -start_shift, strbeg); |
784 | char *last1, *last2; | |
be8e71aa | 785 | char * const saved_s = s; |
33b8afdf | 786 | SV* must; |
2c2d71f5 | 787 | |
2c2d71f5 JH |
788 | t = s - prog->check_offset_max; |
789 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
f2ed9b32 | 790 | && (!utf8_target |
0ce71af7 | 791 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 792 | && t > strpos))) |
6f207bd3 | 793 | NOOP; |
2c2d71f5 JH |
794 | else |
795 | t = strpos; | |
1aa99e6b | 796 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
797 | if (t < other_last) /* These positions already checked */ |
798 | t = other_last; | |
1aa99e6b | 799 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
800 | if (last < last1) |
801 | last1 = last; | |
1de06328 YO |
802 | /* XXXX It is not documented what units *_offsets are in. |
803 | We assume bytes, but this is clearly wrong. | |
804 | Meaning this code needs to be carefully reviewed for errors. | |
805 | dmq. | |
806 | */ | |
807 | ||
2c2d71f5 | 808 | /* On end-of-str: see comment below. */ |
f2ed9b32 | 809 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; |
33b8afdf JH |
810 | if (must == &PL_sv_undef) { |
811 | s = (char*)NULL; | |
1de06328 | 812 | DEBUG_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
813 | } |
814 | else | |
815 | s = fbm_instr( | |
816 | (unsigned char*)t, | |
817 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
818 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
819 | must, | |
7fba1cd6 | 820 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 821 | ); |
ab3bbdeb | 822 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 823 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
824 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
825 | PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", | |
2c2d71f5 | 826 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
827 | quoted, RE_SV_TAIL(must)); |
828 | }); | |
829 | ||
830 | ||
2c2d71f5 JH |
831 | if (!s) { |
832 | if (last1 >= last2) { | |
a3621e74 | 833 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
834 | ", giving up...\n")); |
835 | goto fail_finish; | |
836 | } | |
a3621e74 | 837 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 838 | ", trying floating at offset %ld...\n", |
be8e71aa | 839 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
840 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
841 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
842 | goto restart; |
843 | } | |
844 | else { | |
a3621e74 | 845 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 846 | (long)(s - i_strpos))); |
1aa99e6b IH |
847 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
848 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 849 | s = saved_s; |
2c2d71f5 JH |
850 | if (t == strpos) |
851 | goto try_at_start; | |
2c2d71f5 JH |
852 | goto try_at_offset; |
853 | } | |
30944b6d | 854 | } |
2c2d71f5 JH |
855 | } |
856 | else { /* Take into account the floating substring. */ | |
33b8afdf | 857 | char *last, *last1; |
be8e71aa | 858 | char * const saved_s = s; |
33b8afdf JH |
859 | SV* must; |
860 | ||
861 | t = HOP3c(s, -start_shift, strbeg); | |
862 | last1 = last = | |
863 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
864 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
865 | last = HOP3c(t, prog->float_max_offset, strend); | |
866 | s = HOP3c(t, prog->float_min_offset, strend); | |
867 | if (s < other_last) | |
868 | s = other_last; | |
2c2d71f5 | 869 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
f2ed9b32 | 870 | must = utf8_target ? prog->float_utf8 : prog->float_substr; |
33b8afdf JH |
871 | /* fbm_instr() takes into account exact value of end-of-str |
872 | if the check is SvTAIL(ed). Since false positives are OK, | |
873 | and end-of-str is not later than strend we are OK. */ | |
874 | if (must == &PL_sv_undef) { | |
875 | s = (char*)NULL; | |
1de06328 | 876 | DEBUG_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
877 | } |
878 | else | |
2c2d71f5 | 879 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
880 | (unsigned char*)last + SvCUR(must) |
881 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 882 | must, multiline ? FBMrf_MULTILINE : 0); |
ab3bbdeb | 883 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 884 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
885 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
886 | PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", | |
33b8afdf | 887 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
888 | quoted, RE_SV_TAIL(must)); |
889 | }); | |
33b8afdf JH |
890 | if (!s) { |
891 | if (last1 == last) { | |
a3621e74 | 892 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
893 | ", giving up...\n")); |
894 | goto fail_finish; | |
2c2d71f5 | 895 | } |
a3621e74 | 896 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 897 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 898 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
899 | other_last = last; |
900 | s = HOP3c(t, 1, strend); | |
901 | goto restart; | |
902 | } | |
903 | else { | |
a3621e74 | 904 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
905 | (long)(s - i_strpos))); |
906 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 907 | s = saved_s; |
33b8afdf JH |
908 | if (t == strpos) |
909 | goto try_at_start; | |
910 | goto try_at_offset; | |
911 | } | |
2c2d71f5 | 912 | } |
cad2e5aa | 913 | } |
2c2d71f5 | 914 | |
1de06328 | 915 | |
9ef43ace | 916 | t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); |
1de06328 | 917 | |
6bda09f9 | 918 | DEBUG_OPTIMISE_MORE_r( |
1de06328 YO |
919 | PerlIO_printf(Perl_debug_log, |
920 | "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", | |
921 | (IV)prog->check_offset_min, | |
922 | (IV)prog->check_offset_max, | |
923 | (IV)(s-strpos), | |
924 | (IV)(t-strpos), | |
925 | (IV)(t-s), | |
926 | (IV)(strend-strpos) | |
927 | ) | |
928 | ); | |
929 | ||
2c2d71f5 | 930 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
f2ed9b32 | 931 | && (!utf8_target |
9ef43ace | 932 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) |
1de06328 YO |
933 | && t > strpos))) |
934 | { | |
2c2d71f5 JH |
935 | /* Fixed substring is found far enough so that the match |
936 | cannot start at strpos. */ | |
937 | try_at_offset: | |
cad2e5aa | 938 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
939 | /* Eventually fbm_*() should handle this, but often |
940 | anchored_offset is not 0, so this check will not be wasted. */ | |
941 | /* XXXX In the code below we prefer to look for "^" even in | |
942 | presence of anchored substrings. And we search even | |
943 | beyond the found float position. These pessimizations | |
944 | are historical artefacts only. */ | |
945 | find_anchor: | |
2c2d71f5 | 946 | while (t < strend - prog->minlen) { |
cad2e5aa | 947 | if (*t == '\n') { |
4ee3650e | 948 | if (t < check_at - prog->check_offset_min) { |
f2ed9b32 | 949 | if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
950 | /* Since we moved from the found position, |
951 | we definitely contradict the found anchored | |
30944b6d IZ |
952 | substr. Due to the above check we do not |
953 | contradict "check" substr. | |
954 | Thus we can arrive here only if check substr | |
955 | is float. Redo checking for "other"=="fixed". | |
956 | */ | |
9041c2e3 | 957 | strpos = t + 1; |
a3621e74 | 958 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 959 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
960 | goto do_other_anchored; |
961 | } | |
4ee3650e GS |
962 | /* We don't contradict the found floating substring. */ |
963 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 964 | s = t + 1; |
a3621e74 | 965 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 966 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
967 | goto set_useful; |
968 | } | |
4ee3650e GS |
969 | /* Position contradicts check-string */ |
970 | /* XXXX probably better to look for check-string | |
971 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 972 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 973 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 974 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
975 | goto restart; |
976 | } | |
977 | t++; | |
978 | } | |
a3621e74 | 979 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 980 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 981 | goto fail_finish; |
cad2e5aa | 982 | } |
f5952150 | 983 | else { |
a3621e74 | 984 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 985 | PL_colors[0], PL_colors[1])); |
f5952150 | 986 | } |
cad2e5aa JH |
987 | s = t; |
988 | set_useful: | |
f2ed9b32 | 989 | ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
990 | } |
991 | else { | |
f5952150 | 992 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 993 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
994 | unless it was an MBOL and we are not after MBOL, |
995 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
996 | try_at_start: |
997 | /* Even in this situation we may use MBOL flag if strpos is offset | |
998 | wrt the start of the string. */ | |
05b4157f | 999 | if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ |
a1933d95 | 1000 | && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d | 1001 | /* May be due to an implicit anchor of m{.*foo} */ |
bbe252da | 1002 | && !(prog->intflags & PREGf_IMPLICIT)) |
d506a20d | 1003 | { |
cad2e5aa JH |
1004 | t = strpos; |
1005 | goto find_anchor; | |
1006 | } | |
a3621e74 | 1007 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 1008 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
70685ca0 | 1009 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 1010 | ); |
2c2d71f5 | 1011 | success_at_start: |
bbe252da | 1012 | if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ |
f2ed9b32 | 1013 | && (utf8_target ? ( |
33b8afdf JH |
1014 | prog->check_utf8 /* Could be deleted already */ |
1015 | && --BmUSEFUL(prog->check_utf8) < 0 | |
1016 | && (prog->check_utf8 == prog->float_utf8) | |
1017 | ) : ( | |
1018 | prog->check_substr /* Could be deleted already */ | |
1019 | && --BmUSEFUL(prog->check_substr) < 0 | |
1020 | && (prog->check_substr == prog->float_substr) | |
1021 | ))) | |
66e933ab | 1022 | { |
cad2e5aa | 1023 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 1024 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
f2ed9b32 KW |
1025 | /* XXX Does the destruction order has to change with utf8_target? */ |
1026 | SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr); | |
1027 | SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
1028 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
1029 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
1030 | check = NULL; /* abort */ | |
cad2e5aa | 1031 | s = strpos; |
486ec47a | 1032 | /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag |
c9415951 YO |
1033 | see http://bugs.activestate.com/show_bug.cgi?id=87173 */ |
1034 | if (prog->intflags & PREGf_IMPLICIT) | |
1035 | prog->extflags &= ~RXf_ANCH_MBOL; | |
3cf5c195 IZ |
1036 | /* XXXX This is a remnant of the old implementation. It |
1037 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 1038 | other heuristics. */ |
bbe252da | 1039 | prog->extflags &= ~RXf_USE_INTUIT; |
c9415951 | 1040 | /* XXXX What other flags might need to be cleared in this branch? */ |
cad2e5aa JH |
1041 | } |
1042 | else | |
1043 | s = strpos; | |
1044 | } | |
1045 | ||
6eb5f6b9 JH |
1046 | /* Last resort... */ |
1047 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
1de06328 YO |
1048 | /* trie stclasses are too expensive to use here, we are better off to |
1049 | leave it to regmatch itself */ | |
f8fc2ecf | 1050 | if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { |
6eb5f6b9 JH |
1051 | /* minlen == 0 is possible if regstclass is \b or \B, |
1052 | and the fixed substr is ''$. | |
1053 | Since minlen is already taken into account, s+1 is before strend; | |
1054 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
1055 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
1056 | regstclass does not come from lookahead... */ | |
1057 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
af944926 | 1058 | This leaves EXACTF-ish only, which are dealt with in find_byclass(). */ |
f8fc2ecf YO |
1059 | const U8* const str = (U8*)STRING(progi->regstclass); |
1060 | const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT | |
1061 | ? CHR_DIST(str+STR_LEN(progi->regstclass), str) | |
66e933ab | 1062 | : 1); |
1de06328 YO |
1063 | char * endpos; |
1064 | if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) | |
1065 | endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); | |
1066 | else if (prog->float_substr || prog->float_utf8) | |
1067 | endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); | |
1068 | else | |
1069 | endpos= strend; | |
1070 | ||
70685ca0 JH |
1071 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n", |
1072 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg))); | |
1de06328 | 1073 | |
6eb5f6b9 | 1074 | t = s; |
f8fc2ecf | 1075 | s = find_byclass(prog, progi->regstclass, s, endpos, NULL); |
6eb5f6b9 JH |
1076 | if (!s) { |
1077 | #ifdef DEBUGGING | |
cbbf8932 | 1078 | const char *what = NULL; |
6eb5f6b9 JH |
1079 | #endif |
1080 | if (endpos == strend) { | |
a3621e74 | 1081 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
1082 | "Could not match STCLASS...\n") ); |
1083 | goto fail; | |
1084 | } | |
a3621e74 | 1085 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 1086 | "This position contradicts STCLASS...\n") ); |
bbe252da | 1087 | if ((prog->extflags & RXf_ANCH) && !ml_anch) |
653099ff | 1088 | goto fail; |
6eb5f6b9 | 1089 | /* Contradict one of substrings */ |
33b8afdf | 1090 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 | 1091 | if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) { |
a3621e74 | 1092 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 1093 | hop_and_restart: |
1aa99e6b | 1094 | s = HOP3c(t, 1, strend); |
66e933ab GS |
1095 | if (s + start_shift + end_shift > strend) { |
1096 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 1097 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
1098 | "Could not match STCLASS...\n") ); |
1099 | goto fail; | |
1100 | } | |
5e39e1e5 HS |
1101 | if (!check) |
1102 | goto giveup; | |
a3621e74 | 1103 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1104 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1105 | what, (long)(s + start_shift - i_strpos)) ); |
1106 | goto restart; | |
1107 | } | |
66e933ab | 1108 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
1109 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
1110 | goto retry_floating_check; | |
1111 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 1112 | s = check_at; |
5e39e1e5 HS |
1113 | if (!check) |
1114 | goto giveup; | |
a3621e74 | 1115 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1116 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1117 | (long)(other_last - i_strpos)) ); |
1118 | goto do_other_anchored; | |
1119 | } | |
60e71179 GS |
1120 | /* Another way we could have checked stclass at the |
1121 | current position only: */ | |
1122 | if (ml_anch) { | |
1123 | s = t = t + 1; | |
5e39e1e5 HS |
1124 | if (!check) |
1125 | goto giveup; | |
a3621e74 | 1126 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1127 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 1128 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 1129 | goto try_at_offset; |
66e933ab | 1130 | } |
f2ed9b32 | 1131 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 1132 | goto fail; |
486ec47a | 1133 | /* Check is floating substring. */ |
6eb5f6b9 JH |
1134 | retry_floating_check: |
1135 | t = check_at - start_shift; | |
a3621e74 | 1136 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
1137 | goto hop_and_restart; |
1138 | } | |
b7953727 | 1139 | if (t != s) { |
a3621e74 | 1140 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 1141 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
1142 | (long)(t - i_strpos), (long)(s - i_strpos)) |
1143 | ); | |
1144 | } | |
1145 | else { | |
a3621e74 | 1146 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1147 | "Does not contradict STCLASS...\n"); |
1148 | ); | |
1149 | } | |
6eb5f6b9 | 1150 | } |
5e39e1e5 | 1151 | giveup: |
a3621e74 | 1152 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
1153 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
1154 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 1155 | return s; |
2c2d71f5 JH |
1156 | |
1157 | fail_finish: /* Substring not found */ | |
33b8afdf | 1158 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
f2ed9b32 | 1159 | BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ |
cad2e5aa | 1160 | fail: |
a3621e74 | 1161 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 1162 | PL_colors[4], PL_colors[5])); |
bd61b366 | 1163 | return NULL; |
cad2e5aa | 1164 | } |
9661b544 | 1165 | |
a0a388a1 YO |
1166 | #define DECL_TRIE_TYPE(scan) \ |
1167 | const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \ | |
1168 | trie_type = (scan->flags != EXACT) \ | |
f2ed9b32 KW |
1169 | ? (utf8_target ? trie_utf8_fold : (UTF_PATTERN ? trie_latin_utf8_fold : trie_plain)) \ |
1170 | : (utf8_target ? trie_utf8 : trie_plain) | |
3b0527fe | 1171 | |
55eed653 NC |
1172 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ |
1173 | uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ | |
4cadc6a9 YO |
1174 | switch (trie_type) { \ |
1175 | case trie_utf8_fold: \ | |
1176 | if ( foldlen>0 ) { \ | |
0abd0d78 | 1177 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ |
4cadc6a9 YO |
1178 | foldlen -= len; \ |
1179 | uscan += len; \ | |
1180 | len=0; \ | |
1181 | } else { \ | |
0abd0d78 | 1182 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ |
4cadc6a9 YO |
1183 | uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \ |
1184 | foldlen -= UNISKIP( uvc ); \ | |
1185 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1186 | } \ | |
1187 | break; \ | |
a0a388a1 YO |
1188 | case trie_latin_utf8_fold: \ |
1189 | if ( foldlen>0 ) { \ | |
1190 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1191 | foldlen -= len; \ | |
1192 | uscan += len; \ | |
1193 | len=0; \ | |
1194 | } else { \ | |
1195 | len = 1; \ | |
1196 | uvc = to_uni_fold( *(U8*)uc, foldbuf, &foldlen ); \ | |
1197 | foldlen -= UNISKIP( uvc ); \ | |
1198 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1199 | } \ | |
1200 | break; \ | |
4cadc6a9 YO |
1201 | case trie_utf8: \ |
1202 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ | |
1203 | break; \ | |
1204 | case trie_plain: \ | |
1205 | uvc = (UV)*uc; \ | |
1206 | len = 1; \ | |
1207 | } \ | |
4cadc6a9 YO |
1208 | if (uvc < 256) { \ |
1209 | charid = trie->charmap[ uvc ]; \ | |
1210 | } \ | |
1211 | else { \ | |
1212 | charid = 0; \ | |
55eed653 NC |
1213 | if (widecharmap) { \ |
1214 | SV** const svpp = hv_fetch(widecharmap, \ | |
4cadc6a9 YO |
1215 | (char*)&uvc, sizeof(UV), 0); \ |
1216 | if (svpp) \ | |
1217 | charid = (U16)SvIV(*svpp); \ | |
1218 | } \ | |
1219 | } \ | |
1220 | } STMT_END | |
1221 | ||
4cadc6a9 YO |
1222 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ |
1223 | STMT_START { \ | |
1224 | while (s <= e) { \ | |
1225 | if ( (CoNd) \ | |
fac1af77 | 1226 | && (ln == 1 || folder(s, pat_string, ln)) \ |
9a5a5549 | 1227 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1228 | goto got_it; \ |
1229 | s++; \ | |
1230 | } \ | |
1231 | } STMT_END | |
1232 | ||
1233 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
1234 | STMT_START { \ | |
1235 | while (s + (uskip = UTF8SKIP(s)) <= strend) { \ | |
1236 | CoDe \ | |
1237 | s += uskip; \ | |
1238 | } \ | |
1239 | } STMT_END | |
1240 | ||
1241 | #define REXEC_FBC_SCAN(CoDe) \ | |
1242 | STMT_START { \ | |
1243 | while (s < strend) { \ | |
1244 | CoDe \ | |
1245 | s++; \ | |
1246 | } \ | |
1247 | } STMT_END | |
1248 | ||
1249 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1250 | REXEC_FBC_UTF8_SCAN( \ | |
1251 | if (CoNd) { \ | |
24b23f37 | 1252 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1253 | goto got_it; \ |
1254 | else \ | |
1255 | tmp = doevery; \ | |
1256 | } \ | |
1257 | else \ | |
1258 | tmp = 1; \ | |
1259 | ) | |
1260 | ||
1261 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1262 | REXEC_FBC_SCAN( \ | |
1263 | if (CoNd) { \ | |
24b23f37 | 1264 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1265 | goto got_it; \ |
1266 | else \ | |
1267 | tmp = doevery; \ | |
1268 | } \ | |
1269 | else \ | |
1270 | tmp = 1; \ | |
1271 | ) | |
1272 | ||
1273 | #define REXEC_FBC_TRYIT \ | |
24b23f37 | 1274 | if ((!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1275 | goto got_it |
1276 | ||
e1d1eefb | 1277 | #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \ |
f2ed9b32 | 1278 | if (utf8_target) { \ |
e1d1eefb YO |
1279 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1280 | } \ | |
1281 | else { \ | |
1282 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1283 | } |
e1d1eefb | 1284 | |
4cadc6a9 | 1285 | #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ |
f2ed9b32 | 1286 | if (utf8_target) { \ |
4cadc6a9 YO |
1287 | UtFpReLoAd; \ |
1288 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1289 | } \ | |
1290 | else { \ | |
1291 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1292 | } |
4cadc6a9 YO |
1293 | |
1294 | #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ | |
1295 | PL_reg_flags |= RF_tainted; \ | |
f2ed9b32 | 1296 | if (utf8_target) { \ |
4cadc6a9 YO |
1297 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1298 | } \ | |
1299 | else { \ | |
1300 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1301 | } |
4cadc6a9 | 1302 | |
786e8c11 YO |
1303 | #define DUMP_EXEC_POS(li,s,doutf8) \ |
1304 | dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8) | |
1305 | ||
cfaf538b KW |
1306 | |
1307 | #define UTF8_NOLOAD(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ | |
1308 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1309 | tmp = TEST_NON_UTF8(tmp); \ | |
1310 | REXEC_FBC_UTF8_SCAN( \ | |
1311 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1312 | tmp = !tmp; \ | |
1313 | IF_SUCCESS; \ | |
1314 | } \ | |
1315 | else { \ | |
1316 | IF_FAIL; \ | |
1317 | } \ | |
1318 | ); \ | |
1319 | ||
1320 | #define UTF8_LOAD(TeSt1_UtF8, TeSt2_UtF8, IF_SUCCESS, IF_FAIL) \ | |
1321 | if (s == PL_bostr) { \ | |
1322 | tmp = '\n'; \ | |
1323 | } \ | |
1324 | else { \ | |
1325 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); \ | |
1326 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); \ | |
1327 | } \ | |
1328 | tmp = TeSt1_UtF8; \ | |
1329 | LOAD_UTF8_CHARCLASS_ALNUM(); \ | |
1330 | REXEC_FBC_UTF8_SCAN( \ | |
1331 | if (tmp == ! (TeSt2_UtF8)) { \ | |
1332 | tmp = !tmp; \ | |
1333 | IF_SUCCESS; \ | |
1334 | } \ | |
1335 | else { \ | |
1336 | IF_FAIL; \ | |
1337 | } \ | |
1338 | ); \ | |
1339 | ||
63ac0dad KW |
1340 | /* The only difference between the BOUND and NBOUND cases is that |
1341 | * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in | |
1342 | * NBOUND. This is accomplished by passing it in either the if or else clause, | |
1343 | * with the other one being empty */ | |
1344 | #define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1345 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
cfaf538b KW |
1346 | |
1347 | #define FBC_BOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1348 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
63ac0dad KW |
1349 | |
1350 | #define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1351 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b KW |
1352 | |
1353 | #define FBC_NBOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1354 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b | 1355 | |
63ac0dad KW |
1356 | |
1357 | /* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to | |
1358 | * be passed in completely with the variable name being tested, which isn't | |
1359 | * such a clean interface, but this is easier to read than it was before. We | |
1360 | * are looking for the boundary (or non-boundary between a word and non-word | |
1361 | * character. The utf8 and non-utf8 cases have the same logic, but the details | |
1362 | * must be different. Find the "wordness" of the character just prior to this | |
1363 | * one, and compare it with the wordness of this one. If they differ, we have | |
1364 | * a boundary. At the beginning of the string, pretend that the previous | |
1365 | * character was a new-line */ | |
cfaf538b | 1366 | #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ |
63ac0dad | 1367 | if (utf8_target) { \ |
cfaf538b | 1368 | UTF8_CODE \ |
63ac0dad KW |
1369 | } \ |
1370 | else { /* Not utf8 */ \ | |
1371 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1372 | tmp = TEST_NON_UTF8(tmp); \ | |
1373 | REXEC_FBC_SCAN( \ | |
1374 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1375 | tmp = !tmp; \ | |
1376 | IF_SUCCESS; \ | |
1377 | } \ | |
1378 | else { \ | |
1379 | IF_FAIL; \ | |
1380 | } \ | |
1381 | ); \ | |
1382 | } \ | |
1383 | if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) \ | |
1384 | goto got_it; | |
1385 | ||
786e8c11 YO |
1386 | /* We know what class REx starts with. Try to find this position... */ |
1387 | /* if reginfo is NULL, its a dryrun */ | |
1388 | /* annoyingly all the vars in this routine have different names from their counterparts | |
1389 | in regmatch. /grrr */ | |
1390 | ||
3c3eec57 | 1391 | STATIC char * |
07be1b83 | 1392 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
24b23f37 | 1393 | const char *strend, regmatch_info *reginfo) |
a687059c | 1394 | { |
27da23d5 | 1395 | dVAR; |
bbe252da | 1396 | const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; |
fac1af77 KW |
1397 | char *pat_string; /* The pattern's exactish string */ |
1398 | char *pat_end; /* ptr to end char of pat_string */ | |
1399 | re_fold_t folder; /* Function for computing non-utf8 folds */ | |
1400 | const U8 *fold_array; /* array for folding ords < 256 */ | |
d8093b23 | 1401 | STRLEN ln; |
5dab1207 | 1402 | STRLEN lnc; |
078c425b | 1403 | register STRLEN uskip; |
fac1af77 KW |
1404 | U8 c1; |
1405 | U8 c2; | |
6eb5f6b9 JH |
1406 | char *e; |
1407 | register I32 tmp = 1; /* Scratch variable? */ | |
f2ed9b32 | 1408 | register const bool utf8_target = PL_reg_match_utf8; |
453bfd44 | 1409 | UV utf8_fold_flags = 0; |
f8fc2ecf | 1410 | RXi_GET_DECL(prog,progi); |
7918f24d NC |
1411 | |
1412 | PERL_ARGS_ASSERT_FIND_BYCLASS; | |
f8fc2ecf | 1413 | |
6eb5f6b9 JH |
1414 | /* We know what class it must start with. */ |
1415 | switch (OP(c)) { | |
f56b6394 | 1416 | case ANYOFV: |
6eb5f6b9 | 1417 | case ANYOF: |
f56b6394 | 1418 | if (utf8_target || OP(c) == ANYOFV) { |
b1e3e569 KW |
1419 | STRLEN inclasslen = strend - s; |
1420 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1421 | reginclass(prog, c, (U8*)s, &inclasslen, utf8_target)); | |
388cc4de HS |
1422 | } |
1423 | else { | |
6ef69d56 | 1424 | REXEC_FBC_CLASS_SCAN(REGINCLASS(prog, c, (U8*)s)); |
a0d0e21e | 1425 | } |
6eb5f6b9 | 1426 | break; |
f33976b4 | 1427 | case CANY: |
4cadc6a9 | 1428 | REXEC_FBC_SCAN( |
24b23f37 | 1429 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
f33976b4 DB |
1430 | goto got_it; |
1431 | else | |
1432 | tmp = doevery; | |
4cadc6a9 | 1433 | ); |
f33976b4 | 1434 | break; |
2f7f8cb1 KW |
1435 | |
1436 | case EXACTFA: | |
1437 | if (UTF_PATTERN || utf8_target) { | |
1438 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
1439 | goto do_exactf_utf8; | |
1440 | } | |
1441 | fold_array = PL_fold_latin1; /* Latin1 folds are not affected by */ | |
1442 | folder = foldEQ_latin1; /* /a, except the sharp s one which */ | |
1443 | goto do_exactf_non_utf8; /* isn't dealt with by these */ | |
1444 | ||
9a5a5549 | 1445 | case EXACTFU: |
fac1af77 KW |
1446 | if (UTF_PATTERN || utf8_target) { |
1447 | utf8_fold_flags = 0; | |
1448 | goto do_exactf_utf8; | |
1449 | } | |
0a185c45 KW |
1450 | |
1451 | /* Any 'ss' in the pattern should have been replaced by regcomp, | |
1452 | * so we don't have to worry here about this single special case | |
1453 | * in the Latin1 range */ | |
fac1af77 KW |
1454 | fold_array = PL_fold_latin1; |
1455 | folder = foldEQ_latin1; | |
0a185c45 | 1456 | goto do_exactf_non_utf8; |
fac1af77 | 1457 | |
6eb5f6b9 | 1458 | case EXACTF: |
fac1af77 KW |
1459 | if (UTF_PATTERN || utf8_target) { |
1460 | utf8_fold_flags = 0; | |
1461 | goto do_exactf_utf8; | |
1462 | } | |
1463 | fold_array = PL_fold; | |
1464 | folder = foldEQ; | |
1465 | goto do_exactf_non_utf8; | |
1466 | ||
1467 | case EXACTFL: | |
1468 | if (UTF_PATTERN || utf8_target) { | |
17580e7a | 1469 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
fac1af77 KW |
1470 | goto do_exactf_utf8; |
1471 | } | |
1472 | fold_array = PL_fold_locale; | |
1473 | folder = foldEQ_locale; | |
1474 | ||
1475 | /* FALL THROUGH */ | |
1476 | ||
1477 | do_exactf_non_utf8: /* Neither pattern nor string are UTF8 */ | |
1478 | ||
1479 | /* The idea in the non-utf8 EXACTF* cases is to first find the | |
1480 | * first character of the EXACTF* node and then, if necessary, | |
1481 | * case-insensitively compare the full text of the node. c1 is the | |
1482 | * first character. c2 is its fold. This logic will not work for | |
1483 | * Unicode semantics and the german sharp ss, which hence should | |
1484 | * not be compiled into a node that gets here. */ | |
1485 | pat_string = STRING(c); | |
1486 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1487 | ||
1488 | e = HOP3c(strend, -((I32)ln), s); | |
1489 | ||
1490 | if (!reginfo && e < s) { | |
1491 | e = s; /* Due to minlen logic of intuit() */ | |
1492 | } | |
1493 | ||
1494 | c1 = *pat_string; | |
1495 | c2 = fold_array[c1]; | |
1496 | if (c1 == c2) { /* If char and fold are the same */ | |
1497 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); | |
1498 | } | |
1499 | else { | |
1500 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); | |
1501 | } | |
1502 | break; | |
1503 | ||
1504 | do_exactf_utf8: | |
1505 | ||
1506 | /* If one of the operands is in utf8, we can't use the simpler | |
1507 | * folding above, due to the fact that many different characters | |
1508 | * can have the same fold, or portion of a fold, or different- | |
1509 | * length fold */ | |
1510 | pat_string = STRING(c); | |
1511 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1512 | pat_end = pat_string + ln; | |
1513 | lnc = (UTF_PATTERN) /* length to match in characters */ | |
1514 | ? utf8_length((U8 *) pat_string, (U8 *) pat_end) | |
1515 | : ln; | |
1516 | ||
1517 | e = HOP3c(strend, -((I32)lnc), s); | |
1518 | ||
1519 | if (!reginfo && e < s) { | |
1520 | e = s; /* Due to minlen logic of intuit() */ | |
1521 | } | |
1522 | ||
1523 | while (s <= e) { | |
1524 | char *my_strend= (char *)strend; | |
1525 | if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target, | |
1526 | pat_string, NULL, ln, cBOOL(UTF_PATTERN), utf8_fold_flags) | |
1527 | && (!reginfo || regtry(reginfo, &s)) ) | |
1528 | { | |
1529 | goto got_it; | |
1530 | } | |
1531 | s += UTF8SKIP(s); | |
1532 | } | |
1533 | break; | |
bbce6d69 | 1534 | case BOUNDL: |
3280af22 | 1535 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1536 | FBC_BOUND(isALNUM_LC, |
1537 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1538 | isALNUM_LC_utf8((U8*)s)); | |
a0ed51b3 | 1539 | break; |
bbce6d69 | 1540 | case NBOUNDL: |
3280af22 | 1541 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1542 | FBC_NBOUND(isALNUM_LC, |
1543 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1544 | isALNUM_LC_utf8((U8*)s)); | |
1545 | break; | |
1546 | case BOUND: | |
1547 | FBC_BOUND(isWORDCHAR, | |
1548 | isALNUM_uni(tmp), | |
1549 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1550 | break; | |
cfaf538b KW |
1551 | case BOUNDA: |
1552 | FBC_BOUND_NOLOAD(isWORDCHAR_A, | |
1553 | isWORDCHAR_A(tmp), | |
1554 | isWORDCHAR_A((U8*)s)); | |
1555 | break; | |
a0d0e21e | 1556 | case NBOUND: |
63ac0dad KW |
1557 | FBC_NBOUND(isWORDCHAR, |
1558 | isALNUM_uni(tmp), | |
1559 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1560 | break; | |
cfaf538b KW |
1561 | case NBOUNDA: |
1562 | FBC_NBOUND_NOLOAD(isWORDCHAR_A, | |
1563 | isWORDCHAR_A(tmp), | |
1564 | isWORDCHAR_A((U8*)s)); | |
1565 | break; | |
63ac0dad KW |
1566 | case BOUNDU: |
1567 | FBC_BOUND(isWORDCHAR_L1, | |
1568 | isALNUM_uni(tmp), | |
1569 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1570 | break; | |
1571 | case NBOUNDU: | |
1572 | FBC_NBOUND(isWORDCHAR_L1, | |
1573 | isALNUM_uni(tmp), | |
1574 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
a0ed51b3 | 1575 | break; |
bbce6d69 | 1576 | case ALNUML: |
4cadc6a9 YO |
1577 | REXEC_FBC_CSCAN_TAINT( |
1578 | isALNUM_LC_utf8((U8*)s), | |
1579 | isALNUM_LC(*s) | |
1580 | ); | |
6895a8aa | 1581 | break; |
980866de KW |
1582 | case ALNUMU: |
1583 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1584 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1585 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1586 | isWORDCHAR_L1((U8) *s) |
1587 | ); | |
6895a8aa | 1588 | break; |
980866de KW |
1589 | case ALNUM: |
1590 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1591 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1592 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1593 | isWORDCHAR((U8) *s) |
1594 | ); | |
6895a8aa | 1595 | break; |
cfaf538b | 1596 | case ALNUMA: |
8e9da4d4 KW |
1597 | /* Don't need to worry about utf8, as it can match only a single |
1598 | * byte invariant character */ | |
cfaf538b | 1599 | REXEC_FBC_CLASS_SCAN( isWORDCHAR_A(*s)); |
6895a8aa | 1600 | break; |
980866de KW |
1601 | case NALNUMU: |
1602 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 | 1603 | LOAD_UTF8_CHARCLASS_ALNUM(), |
359960d4 | 1604 | !swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), |
980866de KW |
1605 | ! isWORDCHAR_L1((U8) *s) |
1606 | ); | |
6895a8aa | 1607 | break; |
a0d0e21e | 1608 | case NALNUM: |
4cadc6a9 | 1609 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1610 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1611 | !swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target), | |
980866de | 1612 | ! isALNUM(*s) |
4cadc6a9 | 1613 | ); |
6895a8aa | 1614 | break; |
cfaf538b | 1615 | case NALNUMA: |
8e9da4d4 KW |
1616 | REXEC_FBC_CSCAN( |
1617 | !isWORDCHAR_A(*s), | |
1618 | !isWORDCHAR_A(*s) | |
1619 | ); | |
1620 | break; | |
bbce6d69 | 1621 | case NALNUML: |
4cadc6a9 YO |
1622 | REXEC_FBC_CSCAN_TAINT( |
1623 | !isALNUM_LC_utf8((U8*)s), | |
1624 | !isALNUM_LC(*s) | |
1625 | ); | |
6895a8aa | 1626 | break; |
980866de KW |
1627 | case SPACEU: |
1628 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1629 | LOAD_UTF8_CHARCLASS_SPACE(), |
1630 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de KW |
1631 | isSPACE_L1((U8) *s) |
1632 | ); | |
6895a8aa | 1633 | break; |
a0d0e21e | 1634 | case SPACE: |
4cadc6a9 | 1635 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1636 | LOAD_UTF8_CHARCLASS_SPACE(), |
1637 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de | 1638 | isSPACE((U8) *s) |
4cadc6a9 | 1639 | ); |
6895a8aa | 1640 | break; |
cfaf538b | 1641 | case SPACEA: |
8e9da4d4 KW |
1642 | /* Don't need to worry about utf8, as it can match only a single |
1643 | * byte invariant character */ | |
cfaf538b | 1644 | REXEC_FBC_CLASS_SCAN( isSPACE_A(*s)); |
6895a8aa | 1645 | break; |
bbce6d69 | 1646 | case SPACEL: |
4cadc6a9 | 1647 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1648 | isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1649 | isSPACE_LC(*s) |
1650 | ); | |
6895a8aa | 1651 | break; |
980866de KW |
1652 | case NSPACEU: |
1653 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1654 | LOAD_UTF8_CHARCLASS_SPACE(), |
1655 | !( *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de KW |
1656 | ! isSPACE_L1((U8) *s) |
1657 | ); | |
6895a8aa | 1658 | break; |
a0d0e21e | 1659 | case NSPACE: |
4cadc6a9 | 1660 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1661 | LOAD_UTF8_CHARCLASS_SPACE(), |
1662 | !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de | 1663 | ! isSPACE((U8) *s) |
4cadc6a9 | 1664 | ); |
6895a8aa | 1665 | break; |
cfaf538b | 1666 | case NSPACEA: |
8e9da4d4 KW |
1667 | REXEC_FBC_CSCAN( |
1668 | !isSPACE_A(*s), | |
1669 | !isSPACE_A(*s) | |
1670 | ); | |
1671 | break; | |
bbce6d69 | 1672 | case NSPACEL: |
4cadc6a9 | 1673 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1674 | !isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1675 | !isSPACE_LC(*s) |
1676 | ); | |
6895a8aa | 1677 | break; |
a0d0e21e | 1678 | case DIGIT: |
4cadc6a9 | 1679 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1680 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1681 | swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1682 | isDIGIT(*s) |
1683 | ); | |
6895a8aa | 1684 | break; |
cfaf538b | 1685 | case DIGITA: |
8e9da4d4 KW |
1686 | /* Don't need to worry about utf8, as it can match only a single |
1687 | * byte invariant character */ | |
cfaf538b | 1688 | REXEC_FBC_CLASS_SCAN( isDIGIT_A(*s)); |
6895a8aa | 1689 | break; |
b8c5462f | 1690 | case DIGITL: |
4cadc6a9 YO |
1691 | REXEC_FBC_CSCAN_TAINT( |
1692 | isDIGIT_LC_utf8((U8*)s), | |
1693 | isDIGIT_LC(*s) | |
1694 | ); | |
6895a8aa | 1695 | break; |
a0d0e21e | 1696 | case NDIGIT: |
4cadc6a9 | 1697 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1698 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1699 | !swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1700 | !isDIGIT(*s) |
1701 | ); | |
6895a8aa | 1702 | break; |
cfaf538b | 1703 | case NDIGITA: |
8e9da4d4 KW |
1704 | REXEC_FBC_CSCAN( |
1705 | !isDIGIT_A(*s), | |
1706 | !isDIGIT_A(*s) | |
1707 | ); | |
1708 | break; | |
b8c5462f | 1709 | case NDIGITL: |
4cadc6a9 YO |
1710 | REXEC_FBC_CSCAN_TAINT( |
1711 | !isDIGIT_LC_utf8((U8*)s), | |
1712 | !isDIGIT_LC(*s) | |
1713 | ); | |
6895a8aa | 1714 | break; |
e1d1eefb YO |
1715 | case LNBREAK: |
1716 | REXEC_FBC_CSCAN( | |
1717 | is_LNBREAK_utf8(s), | |
1718 | is_LNBREAK_latin1(s) | |
1719 | ); | |
6895a8aa | 1720 | break; |
e1d1eefb YO |
1721 | case VERTWS: |
1722 | REXEC_FBC_CSCAN( | |
1723 | is_VERTWS_utf8(s), | |
1724 | is_VERTWS_latin1(s) | |
1725 | ); | |
6895a8aa | 1726 | break; |
e1d1eefb YO |
1727 | case NVERTWS: |
1728 | REXEC_FBC_CSCAN( | |
1729 | !is_VERTWS_utf8(s), | |
1730 | !is_VERTWS_latin1(s) | |
1731 | ); | |
6895a8aa | 1732 | break; |
e1d1eefb YO |
1733 | case HORIZWS: |
1734 | REXEC_FBC_CSCAN( | |
1735 | is_HORIZWS_utf8(s), | |
1736 | is_HORIZWS_latin1(s) | |
1737 | ); | |
6895a8aa | 1738 | break; |
e1d1eefb YO |
1739 | case NHORIZWS: |
1740 | REXEC_FBC_CSCAN( | |
1741 | !is_HORIZWS_utf8(s), | |
1742 | !is_HORIZWS_latin1(s) | |
1743 | ); | |
6895a8aa | 1744 | break; |
1de06328 YO |
1745 | case AHOCORASICKC: |
1746 | case AHOCORASICK: | |
07be1b83 | 1747 | { |
a0a388a1 | 1748 | DECL_TRIE_TYPE(c); |
07be1b83 YO |
1749 | /* what trie are we using right now */ |
1750 | reg_ac_data *aho | |
f8fc2ecf | 1751 | = (reg_ac_data*)progi->data->data[ ARG( c ) ]; |
3251b653 NC |
1752 | reg_trie_data *trie |
1753 | = (reg_trie_data*)progi->data->data[ aho->trie ]; | |
85fbaab2 | 1754 | HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); |
07be1b83 YO |
1755 | |
1756 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1757 | #ifdef DEBUGGING |
07be1b83 | 1758 | const char *real_start = s; |
6148ee25 | 1759 | #endif |
07be1b83 | 1760 | STRLEN maxlen = trie->maxlen; |
be8e71aa YO |
1761 | SV *sv_points; |
1762 | U8 **points; /* map of where we were in the input string | |
786e8c11 | 1763 | when reading a given char. For ASCII this |
be8e71aa | 1764 | is unnecessary overhead as the relationship |
38a44b82 NC |
1765 | is always 1:1, but for Unicode, especially |
1766 | case folded Unicode this is not true. */ | |
f9e705e8 | 1767 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
786e8c11 YO |
1768 | U8 *bitmap=NULL; |
1769 | ||
07be1b83 YO |
1770 | |
1771 | GET_RE_DEBUG_FLAGS_DECL; | |
1772 | ||
be8e71aa YO |
1773 | /* We can't just allocate points here. We need to wrap it in |
1774 | * an SV so it gets freed properly if there is a croak while | |
1775 | * running the match */ | |
1776 | ENTER; | |
1777 | SAVETMPS; | |
1778 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1779 | SvCUR_set(sv_points, | |
1780 | maxlen * sizeof(U8 *)); | |
1781 | SvPOK_on(sv_points); | |
1782 | sv_2mortal(sv_points); | |
1783 | points=(U8**)SvPV_nolen(sv_points ); | |
1de06328 YO |
1784 | if ( trie_type != trie_utf8_fold |
1785 | && (trie->bitmap || OP(c)==AHOCORASICKC) ) | |
1786 | { | |
786e8c11 YO |
1787 | if (trie->bitmap) |
1788 | bitmap=(U8*)trie->bitmap; | |
1789 | else | |
1790 | bitmap=(U8*)ANYOF_BITMAP(c); | |
07be1b83 | 1791 | } |
786e8c11 YO |
1792 | /* this is the Aho-Corasick algorithm modified a touch |
1793 | to include special handling for long "unknown char" | |
1794 | sequences. The basic idea being that we use AC as long | |
1795 | as we are dealing with a possible matching char, when | |
1796 | we encounter an unknown char (and we have not encountered | |
1797 | an accepting state) we scan forward until we find a legal | |
1798 | starting char. | |
1799 | AC matching is basically that of trie matching, except | |
1800 | that when we encounter a failing transition, we fall back | |
1801 | to the current states "fail state", and try the current char | |
1802 | again, a process we repeat until we reach the root state, | |
1803 | state 1, or a legal transition. If we fail on the root state | |
1804 | then we can either terminate if we have reached an accepting | |
1805 | state previously, or restart the entire process from the beginning | |
1806 | if we have not. | |
1807 | ||
1808 | */ | |
07be1b83 YO |
1809 | while (s <= last_start) { |
1810 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1811 | U8 *uc = (U8*)s; | |
1812 | U16 charid = 0; | |
1813 | U32 base = 1; | |
1814 | U32 state = 1; | |
1815 | UV uvc = 0; | |
1816 | STRLEN len = 0; | |
1817 | STRLEN foldlen = 0; | |
1818 | U8 *uscan = (U8*)NULL; | |
1819 | U8 *leftmost = NULL; | |
786e8c11 YO |
1820 | #ifdef DEBUGGING |
1821 | U32 accepted_word= 0; | |
1822 | #endif | |
07be1b83 YO |
1823 | U32 pointpos = 0; |
1824 | ||
1825 | while ( state && uc <= (U8*)strend ) { | |
1826 | int failed=0; | |
786e8c11 YO |
1827 | U32 word = aho->states[ state ].wordnum; |
1828 | ||
1de06328 YO |
1829 | if( state==1 ) { |
1830 | if ( bitmap ) { | |
1831 | DEBUG_TRIE_EXECUTE_r( | |
1832 | if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1833 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1834 | (char *)uc, utf8_target ); |
1de06328 YO |
1835 | PerlIO_printf( Perl_debug_log, |
1836 | " Scanning for legal start char...\n"); | |
1837 | } | |
d085b490 YO |
1838 | ); |
1839 | if (utf8_target) { | |
1840 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1841 | uc += UTF8SKIP(uc); | |
1842 | } | |
1843 | } else { | |
1844 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1845 | uc++; | |
1846 | } | |
1847 | } | |
1de06328 | 1848 | s= (char *)uc; |
786e8c11 | 1849 | } |
786e8c11 YO |
1850 | if (uc >(U8*)last_start) break; |
1851 | } | |
1852 | ||
1853 | if ( word ) { | |
2e64971a | 1854 | U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ]; |
786e8c11 YO |
1855 | if (!leftmost || lpos < leftmost) { |
1856 | DEBUG_r(accepted_word=word); | |
07be1b83 | 1857 | leftmost= lpos; |
786e8c11 | 1858 | } |
07be1b83 | 1859 | if (base==0) break; |
786e8c11 | 1860 | |
07be1b83 YO |
1861 | } |
1862 | points[pointpos++ % maxlen]= uc; | |
55eed653 NC |
1863 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
1864 | uscan, len, uvc, charid, foldlen, | |
1865 | foldbuf, uniflags); | |
786e8c11 YO |
1866 | DEBUG_TRIE_EXECUTE_r({ |
1867 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1868 | s, utf8_target ); |
07be1b83 | 1869 | PerlIO_printf(Perl_debug_log, |
786e8c11 YO |
1870 | " Charid:%3u CP:%4"UVxf" ", |
1871 | charid, uvc); | |
1872 | }); | |
07be1b83 YO |
1873 | |
1874 | do { | |
6148ee25 | 1875 | #ifdef DEBUGGING |
786e8c11 | 1876 | word = aho->states[ state ].wordnum; |
6148ee25 | 1877 | #endif |
07be1b83 YO |
1878 | base = aho->states[ state ].trans.base; |
1879 | ||
786e8c11 YO |
1880 | DEBUG_TRIE_EXECUTE_r({ |
1881 | if (failed) | |
1882 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1883 | s, utf8_target ); |
07be1b83 | 1884 | PerlIO_printf( Perl_debug_log, |
786e8c11 YO |
1885 | "%sState: %4"UVxf", word=%"UVxf, |
1886 | failed ? " Fail transition to " : "", | |
1887 | (UV)state, (UV)word); | |
1888 | }); | |
07be1b83 YO |
1889 | if ( base ) { |
1890 | U32 tmp; | |
6dd2be57 | 1891 | I32 offset; |
07be1b83 | 1892 | if (charid && |
6dd2be57 DM |
1893 | ( ((offset = base + charid |
1894 | - 1 - trie->uniquecharcount)) >= 0) | |
1895 | && ((U32)offset < trie->lasttrans) | |
1896 | && trie->trans[offset].check == state | |
1897 | && (tmp=trie->trans[offset].next)) | |
07be1b83 | 1898 | { |
786e8c11 YO |
1899 | DEBUG_TRIE_EXECUTE_r( |
1900 | PerlIO_printf( Perl_debug_log," - legal\n")); | |
07be1b83 YO |
1901 | state = tmp; |
1902 | break; | |
1903 | } | |
1904 | else { | |
786e8c11 YO |
1905 | DEBUG_TRIE_EXECUTE_r( |
1906 | PerlIO_printf( Perl_debug_log," - fail\n")); | |
1907 | failed = 1; | |
1908 | state = aho->fail[state]; | |
07be1b83 YO |
1909 | } |
1910 | } | |
1911 | else { | |
1912 | /* we must be accepting here */ | |
786e8c11 YO |
1913 | DEBUG_TRIE_EXECUTE_r( |
1914 | PerlIO_printf( Perl_debug_log," - accepting\n")); | |
1915 | failed = 1; | |
07be1b83 YO |
1916 | break; |
1917 | } | |
1918 | } while(state); | |
786e8c11 | 1919 | uc += len; |
07be1b83 YO |
1920 | if (failed) { |
1921 | if (leftmost) | |
1922 | break; | |
786e8c11 | 1923 | if (!state) state = 1; |
07be1b83 YO |
1924 | } |
1925 | } | |
1926 | if ( aho->states[ state ].wordnum ) { | |
2e64971a | 1927 | U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ]; |
786e8c11 YO |
1928 | if (!leftmost || lpos < leftmost) { |
1929 | DEBUG_r(accepted_word=aho->states[ state ].wordnum); | |
07be1b83 | 1930 | leftmost = lpos; |
786e8c11 | 1931 | } |
07be1b83 | 1932 | } |
07be1b83 YO |
1933 | if (leftmost) { |
1934 | s = (char*)leftmost; | |
786e8c11 YO |
1935 | DEBUG_TRIE_EXECUTE_r({ |
1936 | PerlIO_printf( | |
70685ca0 JH |
1937 | Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", |
1938 | (UV)accepted_word, (IV)(s - real_start) | |
786e8c11 YO |
1939 | ); |
1940 | }); | |
24b23f37 | 1941 | if (!reginfo || regtry(reginfo, &s)) { |
be8e71aa YO |
1942 | FREETMPS; |
1943 | LEAVE; | |
07be1b83 | 1944 | goto got_it; |
be8e71aa | 1945 | } |
07be1b83 | 1946 | s = HOPc(s,1); |
786e8c11 YO |
1947 | DEBUG_TRIE_EXECUTE_r({ |
1948 | PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); | |
1949 | }); | |
07be1b83 | 1950 | } else { |
786e8c11 YO |
1951 | DEBUG_TRIE_EXECUTE_r( |
1952 | PerlIO_printf( Perl_debug_log,"No match.\n")); | |
07be1b83 YO |
1953 | break; |
1954 | } | |
1955 | } | |
be8e71aa YO |
1956 | FREETMPS; |
1957 | LEAVE; | |
07be1b83 YO |
1958 | } |
1959 | break; | |
b3c9acc1 | 1960 | default: |
3c3eec57 GS |
1961 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); |
1962 | break; | |
d6a28714 | 1963 | } |
6eb5f6b9 JH |
1964 | return 0; |
1965 | got_it: | |
1966 | return s; | |
1967 | } | |
1968 | ||
fae667d5 | 1969 | |
6eb5f6b9 JH |
1970 | /* |
1971 | - regexec_flags - match a regexp against a string | |
1972 | */ | |
1973 | I32 | |
288b8c02 | 1974 | Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend, |
6eb5f6b9 JH |
1975 | char *strbeg, I32 minend, SV *sv, void *data, U32 flags) |
1976 | /* strend: pointer to null at end of string */ | |
1977 | /* strbeg: real beginning of string */ | |
1978 | /* minend: end of match must be >=minend after stringarg. */ | |
58e23c8d YO |
1979 | /* data: May be used for some additional optimizations. |
1980 | Currently its only used, with a U32 cast, for transmitting | |
1981 | the ganch offset when doing a /g match. This will change */ | |
6eb5f6b9 JH |
1982 | /* nosave: For optimizations. */ |
1983 | { | |
97aff369 | 1984 | dVAR; |
288b8c02 | 1985 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
24b23f37 | 1986 | /*register*/ char *s; |
6eb5f6b9 | 1987 | register regnode *c; |
24b23f37 | 1988 | /*register*/ char *startpos = stringarg; |
6eb5f6b9 JH |
1989 | I32 minlen; /* must match at least this many chars */ |
1990 | I32 dontbother = 0; /* how many characters not to try at end */ | |
6eb5f6b9 JH |
1991 | I32 end_shift = 0; /* Same for the end. */ /* CC */ |
1992 | I32 scream_pos = -1; /* Internal iterator of scream. */ | |
ccac19ea | 1993 | char *scream_olds = NULL; |
f2ed9b32 | 1994 | const bool utf8_target = cBOOL(DO_UTF8(sv)); |
2757e526 | 1995 | I32 multiline; |
f8fc2ecf | 1996 | RXi_GET_DECL(prog,progi); |
3b0527fe | 1997 | regmatch_info reginfo; /* create some info to pass to regtry etc */ |
e9105d30 | 1998 | regexp_paren_pair *swap = NULL; |
a3621e74 YO |
1999 | GET_RE_DEBUG_FLAGS_DECL; |
2000 | ||
7918f24d | 2001 | PERL_ARGS_ASSERT_REGEXEC_FLAGS; |
9d4ba2ae | 2002 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
2003 | |
2004 | /* Be paranoid... */ | |
2005 | if (prog == NULL || startpos == NULL) { | |
2006 | Perl_croak(aTHX_ "NULL regexp parameter"); | |
2007 | return 0; | |
2008 | } | |
2009 | ||
bbe252da | 2010 | multiline = prog->extflags & RXf_PMf_MULTILINE; |
288b8c02 | 2011 | reginfo.prog = rx; /* Yes, sorry that this is confusing. */ |
2757e526 | 2012 | |
f2ed9b32 | 2013 | RX_MATCH_UTF8_set(rx, utf8_target); |
1de06328 | 2014 | DEBUG_EXECUTE_r( |
f2ed9b32 | 2015 | debug_start_match(rx, utf8_target, startpos, strend, |
1de06328 YO |
2016 | "Matching"); |
2017 | ); | |
bac06658 | 2018 | |
6eb5f6b9 | 2019 | minlen = prog->minlen; |
1de06328 YO |
2020 | |
2021 | if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { | |
a3621e74 | 2022 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
2023 | "String too short [regexec_flags]...\n")); |
2024 | goto phooey; | |
1aa99e6b | 2025 | } |
6eb5f6b9 | 2026 | |
1de06328 | 2027 | |
6eb5f6b9 | 2028 | /* Check validity of program. */ |
f8fc2ecf | 2029 | if (UCHARAT(progi->program) != REG_MAGIC) { |
6eb5f6b9 JH |
2030 | Perl_croak(aTHX_ "corrupted regexp program"); |
2031 | } | |
2032 | ||
2033 | PL_reg_flags = 0; | |
2034 | PL_reg_eval_set = 0; | |
2035 | PL_reg_maxiter = 0; | |
2036 | ||
3c8556c3 | 2037 | if (RX_UTF8(rx)) |
6eb5f6b9 JH |
2038 | PL_reg_flags |= RF_utf8; |
2039 | ||
2040 | /* Mark beginning of line for ^ and lookbehind. */ | |
3b0527fe | 2041 | reginfo.bol = startpos; /* XXX not used ??? */ |
6eb5f6b9 | 2042 | PL_bostr = strbeg; |
3b0527fe | 2043 | reginfo.sv = sv; |
6eb5f6b9 JH |
2044 | |
2045 | /* Mark end of line for $ (and such) */ | |
2046 | PL_regeol = strend; | |
2047 | ||
2048 | /* see how far we have to get to not match where we matched before */ | |
3b0527fe | 2049 | reginfo.till = startpos+minend; |
6eb5f6b9 | 2050 | |
6eb5f6b9 JH |
2051 | /* If there is a "must appear" string, look for it. */ |
2052 | s = startpos; | |
2053 | ||
bbe252da | 2054 | if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ |
6eb5f6b9 | 2055 | MAGIC *mg; |
2c296965 | 2056 | if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */ |
58e23c8d | 2057 | reginfo.ganch = startpos + prog->gofs; |
2c296965 | 2058 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2059 | "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs)); |
2c296965 | 2060 | } else if (sv && SvTYPE(sv) >= SVt_PVMG |
6eb5f6b9 | 2061 | && SvMAGIC(sv) |
14befaf4 DM |
2062 | && (mg = mg_find(sv, PERL_MAGIC_regex_global)) |
2063 | && mg->mg_len >= 0) { | |
3b0527fe | 2064 | reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ |
2c296965 | 2065 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2066 | "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len)); |
2c296965 | 2067 | |
bbe252da | 2068 | if (prog->extflags & RXf_ANCH_GPOS) { |
3b0527fe | 2069 | if (s > reginfo.ganch) |
6eb5f6b9 | 2070 | goto phooey; |
58e23c8d | 2071 | s = reginfo.ganch - prog->gofs; |
2c296965 | 2072 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2073 | "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs)); |
c584a96e YO |
2074 | if (s < strbeg) |
2075 | goto phooey; | |
6eb5f6b9 JH |
2076 | } |
2077 | } | |
58e23c8d | 2078 | else if (data) { |
70685ca0 | 2079 | reginfo.ganch = strbeg + PTR2UV(data); |
2c296965 YO |
2080 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2081 | "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data))); | |
2082 | ||
2083 | } else { /* pos() not defined */ | |
3b0527fe | 2084 | reginfo.ganch = strbeg; |
2c296965 YO |
2085 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2086 | "GPOS: reginfo.ganch = strbeg\n")); | |
2087 | } | |
6eb5f6b9 | 2088 | } |
288b8c02 | 2089 | if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { |
e9105d30 GG |
2090 | /* We have to be careful. If the previous successful match |
2091 | was from this regex we don't want a subsequent partially | |
2092 | successful match to clobber the old results. | |
2093 | So when we detect this possibility we add a swap buffer | |
2094 | to the re, and switch the buffer each match. If we fail | |
2095 | we switch it back, otherwise we leave it swapped. | |
2096 | */ | |
2097 | swap = prog->offs; | |
2098 | /* do we need a save destructor here for eval dies? */ | |
2099 | Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); | |
c74340f9 | 2100 | } |
a0714e2c | 2101 | if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { |
6eb5f6b9 JH |
2102 | re_scream_pos_data d; |
2103 | ||
2104 | d.scream_olds = &scream_olds; | |
2105 | d.scream_pos = &scream_pos; | |
288b8c02 | 2106 | s = re_intuit_start(rx, sv, s, strend, flags, &d); |
3fa9c3d7 | 2107 | if (!s) { |
a3621e74 | 2108 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n")); |
6eb5f6b9 | 2109 | goto phooey; /* not present */ |
3fa9c3d7 | 2110 | } |
6eb5f6b9 JH |
2111 | } |
2112 | ||
1de06328 | 2113 | |
6eb5f6b9 JH |
2114 | |
2115 | /* Simplest case: anchored match need be tried only once. */ | |
2116 | /* [unless only anchor is BOL and multiline is set] */ | |
bbe252da | 2117 | if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { |
24b23f37 | 2118 | if (s == startpos && regtry(®info, &startpos)) |
6eb5f6b9 | 2119 | goto got_it; |
bbe252da YO |
2120 | else if (multiline || (prog->intflags & PREGf_IMPLICIT) |
2121 | || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ | |
6eb5f6b9 JH |
2122 | { |
2123 | char *end; | |
2124 | ||
2125 | if (minlen) | |
2126 | dontbother = minlen - 1; | |
1aa99e6b | 2127 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 2128 | /* for multiline we only have to try after newlines */ |
33b8afdf | 2129 | if (prog->check_substr || prog->check_utf8) { |
92f3d482 YO |
2130 | /* because of the goto we can not easily reuse the macros for bifurcating the |
2131 | unicode/non-unicode match modes here like we do elsewhere - demerphq */ | |
2132 | if (utf8_target) { | |
2133 | if (s == startpos) | |
2134 | goto after_try_utf8; | |
2135 | while (1) { | |
2136 | if (regtry(®info, &s)) { | |
2137 | goto got_it; | |
2138 | } | |
2139 | after_try_utf8: | |
2140 | if (s > end) { | |
2141 | goto phooey; | |
2142 | } | |
2143 | if (prog->extflags & RXf_USE_INTUIT) { | |
2144 | s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL); | |
2145 | if (!s) { | |
2146 | goto phooey; | |
2147 | } | |
2148 | } | |
2149 | else { | |
2150 | s += UTF8SKIP(s); | |
2151 | } | |
2152 | } | |
2153 | } /* end search for check string in unicode */ | |
2154 | else { | |
2155 | if (s == startpos) { | |
2156 | goto after_try_latin; | |
2157 | } | |
2158 | while (1) { | |
2159 | if (regtry(®info, &s)) { | |
2160 | goto got_it; | |
2161 | } | |
2162 | after_try_latin: | |
2163 | if (s > end) { | |
2164 | goto phooey; | |
2165 | } | |
2166 | if (prog->extflags & RXf_USE_INTUIT) { | |
2167 | s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL); | |
2168 | if (!s) { | |
2169 | goto phooey; | |
2170 | } | |
2171 | } | |
2172 | else { | |
2173 | s++; | |
2174 | } | |
2175 | } | |
2176 | } /* end search for check string in latin*/ | |
2177 | } /* end search for check string */ | |
2178 | else { /* search for newline */ | |
2179 | if (s > startpos) { | |
2180 | /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/ | |
6eb5f6b9 | 2181 | s--; |
92f3d482 YO |
2182 | } |
2183 | /* We can use a more efficient search as newlines are the same in unicode as they are in latin */ | |
6eb5f6b9 JH |
2184 | while (s < end) { |
2185 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ | |
24b23f37 | 2186 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2187 | goto got_it; |
2188 | } | |
92f3d482 YO |
2189 | } |
2190 | } /* end search for newline */ | |
2191 | } /* end anchored/multiline check string search */ | |
6eb5f6b9 | 2192 | goto phooey; |
bbe252da | 2193 | } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) |
f9f4320a | 2194 | { |
486ec47a | 2195 | /* the warning about reginfo.ganch being used without initialization |
bbe252da | 2196 | is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN |
f9f4320a | 2197 | and we only enter this block when the same bit is set. */ |
58e23c8d | 2198 | char *tmp_s = reginfo.ganch - prog->gofs; |
c584a96e YO |
2199 | |
2200 | if (tmp_s >= strbeg && regtry(®info, &tmp_s)) | |
6eb5f6b9 JH |
2201 | goto got_it; |
2202 | goto phooey; | |
2203 | } | |
2204 | ||
2205 | /* Messy cases: unanchored match. */ | |
bbe252da | 2206 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { |
6eb5f6b9 | 2207 | /* we have /x+whatever/ */ |
f2ed9b32 | 2208 | /* it must be a one character string (XXXX Except UTF_PATTERN?) */ |
33b8afdf | 2209 | char ch; |
bf93d4cc GS |
2210 | #ifdef DEBUGGING |
2211 | int did_match = 0; | |
2212 | #endif | |
f2ed9b32 KW |
2213 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2214 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2215 | ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0]; | |
bf93d4cc | 2216 | |
f2ed9b32 | 2217 | if (utf8_target) { |
4cadc6a9 | 2218 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2219 | if (*s == ch) { |
a3621e74 | 2220 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2221 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2222 | s += UTF8SKIP(s); |
2223 | while (s < strend && *s == ch) | |
2224 | s += UTF8SKIP(s); | |
2225 | } | |
4cadc6a9 | 2226 | ); |
6eb5f6b9 JH |
2227 | } |
2228 | else { | |
4cadc6a9 | 2229 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2230 | if (*s == ch) { |
a3621e74 | 2231 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2232 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2233 | s++; |
2234 | while (s < strend && *s == ch) | |
2235 | s++; | |
2236 | } | |
4cadc6a9 | 2237 | ); |
6eb5f6b9 | 2238 | } |
a3621e74 | 2239 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 2240 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
2241 | "Did not find anchored character...\n") |
2242 | ); | |
6eb5f6b9 | 2243 | } |
a0714e2c SS |
2244 | else if (prog->anchored_substr != NULL |
2245 | || prog->anchored_utf8 != NULL | |
2246 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
2247 | && prog->float_max_offset < strend - s)) { |
2248 | SV *must; | |
2249 | I32 back_max; | |
2250 | I32 back_min; | |
2251 | char *last; | |
6eb5f6b9 | 2252 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
2253 | #ifdef DEBUGGING |
2254 | int did_match = 0; | |
2255 | #endif | |
33b8afdf | 2256 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 KW |
2257 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2258 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2259 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; | |
33b8afdf JH |
2260 | back_max = back_min = prog->anchored_offset; |
2261 | } else { | |
f2ed9b32 KW |
2262 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2263 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2264 | must = utf8_target ? prog->float_utf8 : prog->float_substr; | |
33b8afdf JH |
2265 | back_max = prog->float_max_offset; |
2266 | back_min = prog->float_min_offset; | |
2267 | } | |
1de06328 YO |
2268 | |
2269 | ||
33b8afdf JH |
2270 | if (must == &PL_sv_undef) |
2271 | /* could not downgrade utf8 check substring, so must fail */ | |
2272 | goto phooey; | |
2273 | ||
1de06328 YO |
2274 | if (back_min<0) { |
2275 | last = strend; | |
2276 | } else { | |
2277 | last = HOP3c(strend, /* Cannot start after this */ | |
2278 | -(I32)(CHR_SVLEN(must) | |
2279 | - (SvTAIL(must) != 0) + back_min), strbeg); | |
2280 | } | |
6eb5f6b9 JH |
2281 | if (s > PL_bostr) |
2282 | last1 = HOPc(s, -1); | |
2283 | else | |
2284 | last1 = s - 1; /* bogus */ | |
2285 | ||
a0288114 | 2286 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 JH |
2287 | check_substr==must. */ |
2288 | scream_pos = -1; | |
2289 | dontbother = end_shift; | |
2290 | strend = HOPc(strend, -dontbother); | |
2291 | while ( (s <= last) && | |
9041c2e3 | 2292 | ((flags & REXEC_SCREAM) |
1de06328 | 2293 | ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg, |
6eb5f6b9 | 2294 | end_shift, &scream_pos, 0)) |
1de06328 | 2295 | : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), |
9041c2e3 | 2296 | (unsigned char*)strend, must, |
7fba1cd6 | 2297 | multiline ? FBMrf_MULTILINE : 0))) ) { |
4addbd3b | 2298 | /* we may be pointing at the wrong string */ |
07bc277f | 2299 | if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog)) |
3f7c398e | 2300 | s = strbeg + (s - SvPVX_const(sv)); |
a3621e74 | 2301 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
2302 | if (HOPc(s, -back_max) > last1) { |
2303 | last1 = HOPc(s, -back_min); | |
2304 | s = HOPc(s, -back_max); | |
2305 | } | |
2306 | else { | |
52657f30 | 2307 | char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; |
6eb5f6b9 JH |
2308 | |
2309 | last1 = HOPc(s, -back_min); | |
52657f30 | 2310 | s = t; |
6eb5f6b9 | 2311 | } |
f2ed9b32 | 2312 | if (utf8_target) { |
6eb5f6b9 | 2313 | while (s <= last1) { |
24b23f37 | 2314 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2315 | goto got_it; |
2316 | s += UTF8SKIP(s); | |
2317 | } | |
2318 | } | |
2319 | else { | |
2320 | while (s <= last1) { | |
24b23f37 | 2321 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2322 | goto got_it; |
2323 | s++; | |
2324 | } | |
2325 | } | |
2326 | } | |
ab3bbdeb | 2327 | DEBUG_EXECUTE_r(if (!did_match) { |
f2ed9b32 | 2328 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
2329 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
2330 | PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", | |
33b8afdf | 2331 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc | 2332 | ? "anchored" : "floating"), |
ab3bbdeb YO |
2333 | quoted, RE_SV_TAIL(must)); |
2334 | }); | |
6eb5f6b9 JH |
2335 | goto phooey; |
2336 | } | |
f8fc2ecf | 2337 | else if ( (c = progi->regstclass) ) { |
f14c76ed | 2338 | if (minlen) { |
f8fc2ecf | 2339 | const OPCODE op = OP(progi->regstclass); |
66e933ab | 2340 | /* don't bother with what can't match */ |
786e8c11 | 2341 | if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) |
f14c76ed RGS |
2342 | strend = HOPc(strend, -(minlen - 1)); |
2343 | } | |
a3621e74 | 2344 | DEBUG_EXECUTE_r({ |
be8e71aa | 2345 | SV * const prop = sv_newmortal(); |
32fc9b6a | 2346 | regprop(prog, prop, c); |
0df25f3d | 2347 | { |
f2ed9b32 | 2348 | RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb | 2349 | s,strend-s,60); |
0df25f3d | 2350 | PerlIO_printf(Perl_debug_log, |
1c8f8eb1 | 2351 | "Matching stclass %.*s against %s (%d bytes)\n", |
e4f74956 | 2352 | (int)SvCUR(prop), SvPVX_const(prop), |
ab3bbdeb | 2353 | quoted, (int)(strend - s)); |
0df25f3d | 2354 | } |
ffc61ed2 | 2355 | }); |
3b0527fe | 2356 | if (find_byclass(prog, c, s, strend, ®info)) |
6eb5f6b9 | 2357 | goto got_it; |
07be1b83 | 2358 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
2359 | } |
2360 | else { | |
2361 | dontbother = 0; | |
a0714e2c | 2362 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 2363 | /* Trim the end. */ |
d6a28714 | 2364 | char *last; |
33b8afdf JH |
2365 | SV* float_real; |
2366 | ||
f2ed9b32 KW |
2367 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2368 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2369 | float_real = utf8_target ? prog->float_utf8 : prog->float_substr; | |
d6a28714 JH |
2370 | |
2371 | if (flags & REXEC_SCREAM) { | |
33b8afdf | 2372 | last = screaminstr(sv, float_real, s - strbeg, |
d6a28714 JH |
2373 | end_shift, &scream_pos, 1); /* last one */ |
2374 | if (!last) | |
ffc61ed2 | 2375 | last = scream_olds; /* Only one occurrence. */ |
4addbd3b | 2376 | /* we may be pointing at the wrong string */ |
07bc277f | 2377 | else if (RXp_MATCH_COPIED(prog)) |
3f7c398e | 2378 | s = strbeg + (s - SvPVX_const(sv)); |
b8c5462f | 2379 | } |
d6a28714 JH |
2380 | else { |
2381 | STRLEN len; | |
cfd0369c | 2382 | const char * const little = SvPV_const(float_real, len); |
d6a28714 | 2383 | |
33b8afdf | 2384 | if (SvTAIL(float_real)) { |
d6a28714 JH |
2385 | if (memEQ(strend - len + 1, little, len - 1)) |
2386 | last = strend - len + 1; | |
7fba1cd6 | 2387 | else if (!multiline) |
9041c2e3 | 2388 | last = memEQ(strend - len, little, len) |
bd61b366 | 2389 | ? strend - len : NULL; |
b8c5462f | 2390 | else |
d6a28714 JH |
2391 | goto find_last; |
2392 | } else { | |
2393 | find_last: | |
9041c2e3 | 2394 | if (len) |
d6a28714 | 2395 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 2396 | else |
a0288114 | 2397 | last = strend; /* matching "$" */ |
b8c5462f | 2398 | } |
b8c5462f | 2399 | } |
bf93d4cc | 2400 | if (last == NULL) { |
6bda09f9 YO |
2401 | DEBUG_EXECUTE_r( |
2402 | PerlIO_printf(Perl_debug_log, | |
2403 | "%sCan't trim the tail, match fails (should not happen)%s\n", | |
2404 | PL_colors[4], PL_colors[5])); | |
bf93d4cc GS |
2405 | goto phooey; /* Should not happen! */ |
2406 | } | |
d6a28714 JH |
2407 | dontbother = strend - last + prog->float_min_offset; |
2408 | } | |
2409 | if (minlen && (dontbother < minlen)) | |
2410 | dontbother = minlen - 1; | |
2411 | strend -= dontbother; /* this one's always in bytes! */ | |
2412 | /* We don't know much -- general case. */ | |
f2ed9b32 | 2413 | if (utf8_target) { |
d6a28714 | 2414 | for (;;) { |
24b23f37 | 2415 | if (regtry(®info, &s)) |
d6a28714 JH |
2416 | goto got_it; |
2417 | if (s >= strend) | |
2418 | break; | |
b8c5462f | 2419 | s += UTF8SKIP(s); |
d6a28714 JH |
2420 | }; |
2421 | } | |
2422 | else { | |
2423 | do { | |
24b23f37 | 2424 | if (regtry(®info, &s)) |
d6a28714 JH |
2425 | goto got_it; |
2426 | } while (s++ < strend); | |
2427 | } | |
2428 | } | |
2429 | ||
2430 | /* Failure. */ | |
2431 | goto phooey; | |
2432 | ||
2433 | got_it: | |
e9105d30 | 2434 | Safefree(swap); |
288b8c02 | 2435 | RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted); |
d6a28714 | 2436 | |
19b95bf0 | 2437 | if (PL_reg_eval_set) |
4f639d21 | 2438 | restore_pos(aTHX_ prog); |
5daac39c NC |
2439 | if (RXp_PAREN_NAMES(prog)) |
2440 | (void)hv_iterinit(RXp_PAREN_NAMES(prog)); | |
d6a28714 JH |
2441 | |
2442 | /* make sure $`, $&, $', and $digit will work later */ | |
2443 | if ( !(flags & REXEC_NOT_FIRST) ) { | |
288b8c02 | 2444 | RX_MATCH_COPY_FREE(rx); |
d6a28714 | 2445 | if (flags & REXEC_COPY_STR) { |
be8e71aa | 2446 | const I32 i = PL_regeol - startpos + (stringarg - strbeg); |
f8c7b90f | 2447 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2448 | if ((SvIsCOW(sv) |
2449 | || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { | |
2450 | if (DEBUG_C_TEST) { | |
2451 | PerlIO_printf(Perl_debug_log, | |
2452 | "Copy on write: regexp capture, type %d\n", | |
2453 | (int) SvTYPE(sv)); | |
2454 | } | |
2455 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); | |
d5263905 | 2456 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
ed252734 NC |
2457 | assert (SvPOKp(prog->saved_copy)); |
2458 | } else | |
2459 | #endif | |
2460 | { | |
288b8c02 | 2461 | RX_MATCH_COPIED_on(rx); |
ed252734 NC |
2462 | s = savepvn(strbeg, i); |
2463 | prog->subbeg = s; | |
2464 | } | |
d6a28714 | 2465 | prog->sublen = i; |
d6a28714 JH |
2466 | } |
2467 | else { | |
2468 | prog->subbeg = strbeg; | |
2469 | prog->sublen = PL_regeol - strbeg; /* strend may have been modified */ | |
2470 | } | |
2471 | } | |
9041c2e3 | 2472 | |
d6a28714 JH |
2473 | return 1; |
2474 | ||
2475 | phooey: | |
a3621e74 | 2476 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 2477 | PL_colors[4], PL_colors[5])); |
d6a28714 | 2478 | if (PL_reg_eval_set) |
4f639d21 | 2479 | restore_pos(aTHX_ prog); |
e9105d30 | 2480 | if (swap) { |
c74340f9 | 2481 | /* we failed :-( roll it back */ |
e9105d30 GG |
2482 | Safefree(prog->offs); |
2483 | prog->offs = swap; | |
2484 | } | |
2485 | ||
d6a28714 JH |
2486 | return 0; |
2487 | } | |
2488 | ||
6bda09f9 | 2489 | |
d6a28714 JH |
2490 | /* |
2491 | - regtry - try match at specific point | |
2492 | */ | |
2493 | STATIC I32 /* 0 failure, 1 success */ | |
24b23f37 | 2494 | S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) |
d6a28714 | 2495 | { |
97aff369 | 2496 | dVAR; |
d6a28714 | 2497 | CHECKPOINT lastcp; |
288b8c02 NC |
2498 | REGEXP *const rx = reginfo->prog; |
2499 | regexp *const prog = (struct regexp *)SvANY(rx); | |
f8fc2ecf | 2500 | RXi_GET_DECL(prog,progi); |
a3621e74 | 2501 | GET_RE_DEBUG_FLAGS_DECL; |
7918f24d NC |
2502 | |
2503 | PERL_ARGS_ASSERT_REGTRY; | |
2504 | ||
24b23f37 | 2505 | reginfo->cutpoint=NULL; |
d6a28714 | 2506 | |
bbe252da | 2507 | if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) { |
d6a28714 JH |
2508 | MAGIC *mg; |
2509 | ||
2510 | PL_reg_eval_set = RS_init; | |
a3621e74 | 2511 | DEBUG_EXECUTE_r(DEBUG_s( |
b900a521 JH |
2512 | PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n", |
2513 | (IV)(PL_stack_sp - PL_stack_base)); | |
d6a28714 | 2514 | )); |
ea8d6ae1 | 2515 | SAVESTACK_CXPOS(); |
d6a28714 JH |
2516 | cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base; |
2517 | /* Otherwise OP_NEXTSTATE will free whatever on stack now. */ | |
2518 | SAVETMPS; | |
2519 | /* Apparently this is not needed, judging by wantarray. */ | |
e8347627 | 2520 | /* SAVEI8(cxstack[cxstack_ix].blk_gimme); |
d6a28714 JH |
2521 | cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ |
2522 | ||
3b0527fe | 2523 | if (reginfo->sv) { |
d6a28714 | 2524 | /* Make $_ available to executed code. */ |
3b0527fe | 2525 | if (reginfo->sv != DEFSV) { |
59f00321 | 2526 | SAVE_DEFSV; |
414bf5ae | 2527 | DEFSV_set(reginfo->sv); |
b8c5462f | 2528 | } |
d6a28714 | 2529 | |
3b0527fe DM |
2530 | if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) |
2531 | && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { | |
d6a28714 | 2532 | /* prepare for quick setting of pos */ |
d300d9fa | 2533 | #ifdef PERL_OLD_COPY_ON_WRITE |
51a9ea20 NC |
2534 | if (SvIsCOW(reginfo->sv)) |
2535 | sv_force_normal_flags(reginfo->sv, 0); | |
d300d9fa | 2536 | #endif |
3dab1dad | 2537 | mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, |
d300d9fa | 2538 | &PL_vtbl_mglob, NULL, 0); |
d6a28714 | 2539 | mg->mg_len = -1; |
b8c5462f | 2540 | } |
d6a28714 JH |
2541 | PL_reg_magic = mg; |
2542 | PL_reg_oldpos = mg->mg_len; | |
4f639d21 | 2543 | SAVEDESTRUCTOR_X(restore_pos, prog); |
d6a28714 | 2544 | } |
09687e5a | 2545 | if (!PL_reg_curpm) { |
a02a5408 | 2546 | Newxz(PL_reg_curpm, 1, PMOP); |
09687e5a AB |
2547 | #ifdef USE_ITHREADS |
2548 | { | |
14a49a24 | 2549 | SV* const repointer = &PL_sv_undef; |
92313705 NC |
2550 | /* this regexp is also owned by the new PL_reg_curpm, which |
2551 | will try to free it. */ | |
d2ece331 | 2552 | av_push(PL_regex_padav, repointer); |
09687e5a AB |
2553 | PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); |
2554 | PL_regex_pad = AvARRAY(PL_regex_padav); | |
2555 | } | |
2556 | #endif | |
2557 | } | |
86c29d75 NC |
2558 | #ifdef USE_ITHREADS |
2559 | /* It seems that non-ithreads works both with and without this code. | |
2560 | So for efficiency reasons it seems best not to have the code | |
2561 | compiled when it is not needed. */ | |
92313705 NC |
2562 | /* This is safe against NULLs: */ |
2563 | ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); | |
2564 | /* PM_reg_curpm owns a reference to this regexp. */ | |
cd733337 | 2565 | (void)ReREFCNT_inc(rx); |
86c29d75 | 2566 | #endif |
288b8c02 | 2567 | PM_SETRE(PL_reg_curpm, rx); |
d6a28714 JH |
2568 | PL_reg_oldcurpm = PL_curpm; |
2569 | PL_curpm = PL_reg_curpm; | |
07bc277f | 2570 | if (RXp_MATCH_COPIED(prog)) { |
d6a28714 JH |
2571 | /* Here is a serious problem: we cannot rewrite subbeg, |
2572 | since it may be needed if this match fails. Thus | |
2573 | $` inside (?{}) could fail... */ | |
2574 | PL_reg_oldsaved = prog->subbeg; | |
2575 | PL_reg_oldsavedlen = prog->sublen; | |
f8c7b90f | 2576 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2577 | PL_nrs = prog->saved_copy; |
2578 | #endif | |
07bc277f | 2579 | RXp_MATCH_COPIED_off(prog); |
d6a28714 JH |
2580 | } |
2581 | else | |
bd61b366 | 2582 | PL_reg_oldsaved = NULL; |
d6a28714 JH |
2583 | prog->subbeg = PL_bostr; |
2584 | prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ | |
2585 | } | |
24b23f37 | 2586 | DEBUG_EXECUTE_r(PL_reg_starttry = *startpos); |
f0ab9afb | 2587 | prog->offs[0].start = *startpos - PL_bostr; |
24b23f37 | 2588 | PL_reginput = *startpos; |
d6a28714 | 2589 | PL_reglastparen = &prog->lastparen; |
a01268b5 | 2590 | PL_reglastcloseparen = &prog->lastcloseparen; |
d6a28714 | 2591 | prog->lastparen = 0; |
03994de8 | 2592 | prog->lastcloseparen = 0; |
d6a28714 | 2593 | PL_regsize = 0; |
f0ab9afb | 2594 | PL_regoffs = prog->offs; |
d6a28714 JH |
2595 | if (PL_reg_start_tmpl <= prog->nparens) { |
2596 | PL_reg_start_tmpl = prog->nparens*3/2 + 3; | |
2597 | if(PL_reg_start_tmp) | |
2598 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
2599 | else | |
a02a5408 | 2600 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); |
d6a28714 JH |
2601 | } |
2602 | ||
2603 | /* XXXX What this code is doing here?!!! There should be no need | |
2604 | to do this again and again, PL_reglastparen should take care of | |
3dd2943c | 2605 | this! --ilya*/ |
dafc8851 JH |
2606 | |
2607 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2608 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
daf18116 | 2609 | * PL_reglastparen), is not needed at all by the test suite |
225593e1 DM |
2610 | * (op/regexp, op/pat, op/split), but that code is needed otherwise |
2611 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ | |
2612 | * Meanwhile, this code *is* needed for the | |
daf18116 JH |
2613 | * above-mentioned test suite tests to succeed. The common theme |
2614 | * on those tests seems to be returning null fields from matches. | |
225593e1 | 2615 | * --jhi updated by dapm */ |
dafc8851 | 2616 | #if 1 |
d6a28714 | 2617 | if (prog->nparens) { |
f0ab9afb | 2618 | regexp_paren_pair *pp = PL_regoffs; |
097eb12c | 2619 | register I32 i; |
eb160463 | 2620 | for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) { |
f0ab9afb NC |
2621 | ++pp; |
2622 | pp->start = -1; | |
2623 | pp->end = -1; | |
d6a28714 JH |
2624 | } |
2625 | } | |
dafc8851 | 2626 | #endif |
02db2b7b | 2627 | REGCP_SET(lastcp); |
f8fc2ecf | 2628 | if (regmatch(reginfo, progi->program + 1)) { |
f0ab9afb | 2629 | PL_regoffs[0].end = PL_reginput - PL_bostr; |
d6a28714 JH |
2630 | return 1; |
2631 | } | |
24b23f37 YO |
2632 | if (reginfo->cutpoint) |
2633 | *startpos= reginfo->cutpoint; | |
02db2b7b | 2634 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
2635 | return 0; |
2636 | } | |
2637 | ||
02db2b7b | 2638 | |
8ba1375e MJD |
2639 | #define sayYES goto yes |
2640 | #define sayNO goto no | |
262b90c4 | 2641 | #define sayNO_SILENT goto no_silent |
8ba1375e | 2642 | |
f9f4320a YO |
2643 | /* we dont use STMT_START/END here because it leads to |
2644 | "unreachable code" warnings, which are bogus, but distracting. */ | |
2645 | #define CACHEsayNO \ | |
c476f425 DM |
2646 | if (ST.cache_mask) \ |
2647 | PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \ | |
f9f4320a | 2648 | sayNO |
3298f257 | 2649 | |
a3621e74 | 2650 | /* this is used to determine how far from the left messages like |
265c4333 YO |
2651 | 'failed...' are printed. It should be set such that messages |
2652 | are inline with the regop output that created them. | |
a3621e74 | 2653 | */ |
265c4333 | 2654 | #define REPORT_CODE_OFF 32 |
a3621e74 YO |
2655 | |
2656 | ||
40a82448 DM |
2657 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
2658 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
9e137952 | 2659 | |
86545054 DM |
2660 | #define SLAB_FIRST(s) (&(s)->states[0]) |
2661 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
2662 | ||
5d9a96ca DM |
2663 | /* grab a new slab and return the first slot in it */ |
2664 | ||
2665 | STATIC regmatch_state * | |
2666 | S_push_slab(pTHX) | |
2667 | { | |
a35a87e7 | 2668 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2669 | dMY_CXT; |
2670 | #endif | |
5d9a96ca DM |
2671 | regmatch_slab *s = PL_regmatch_slab->next; |
2672 | if (!s) { | |
2673 | Newx(s, 1, regmatch_slab); | |
2674 | s->prev = PL_regmatch_slab; | |
2675 | s->next = NULL; | |
2676 | PL_regmatch_slab->next = s; | |
2677 | } | |
2678 | PL_regmatch_slab = s; | |
86545054 | 2679 | return SLAB_FIRST(s); |
5d9a96ca | 2680 | } |
5b47454d | 2681 | |
95b24440 | 2682 | |
40a82448 DM |
2683 | /* push a new state then goto it */ |
2684 | ||
2685 | #define PUSH_STATE_GOTO(state, node) \ | |
2686 | scan = node; \ | |
2687 | st->resume_state = state; \ | |
2688 | goto push_state; | |
2689 | ||
2690 | /* push a new state with success backtracking, then goto it */ | |
2691 | ||
2692 | #define PUSH_YES_STATE_GOTO(state, node) \ | |
2693 | scan = node; \ | |
2694 | st->resume_state = state; \ | |
2695 | goto push_yes_state; | |
2696 | ||
aa283a38 | 2697 | |
aa283a38 | 2698 | |
d6a28714 | 2699 | /* |
95b24440 | 2700 | |
bf1f174e DM |
2701 | regmatch() - main matching routine |
2702 | ||
2703 | This is basically one big switch statement in a loop. We execute an op, | |
2704 | set 'next' to point the next op, and continue. If we come to a point which | |
2705 | we may need to backtrack to on failure such as (A|B|C), we push a | |
2706 | backtrack state onto the backtrack stack. On failure, we pop the top | |
2707 | state, and re-enter the loop at the state indicated. If there are no more | |
2708 | states to pop, we return failure. | |
2709 | ||
2710 | Sometimes we also need to backtrack on success; for example /A+/, where | |
2711 | after successfully matching one A, we need to go back and try to | |
2712 | match another one; similarly for lookahead assertions: if the assertion | |
2713 | completes successfully, we backtrack to the state just before the assertion | |
2714 | and then carry on. In these cases, the pushed state is marked as | |
2715 | 'backtrack on success too'. This marking is in fact done by a chain of | |
2716 | pointers, each pointing to the previous 'yes' state. On success, we pop to | |
2717 | the nearest yes state, discarding any intermediate failure-only states. | |
2718 | Sometimes a yes state is pushed just to force some cleanup code to be | |
2719 | called at the end of a successful match or submatch; e.g. (??{$re}) uses | |
2720 | it to free the inner regex. | |
2721 | ||
2722 | Note that failure backtracking rewinds the cursor position, while | |
2723 | success backtracking leaves it alone. | |
2724 | ||
2725 | A pattern is complete when the END op is executed, while a subpattern | |
2726 | such as (?=foo) is complete when the SUCCESS op is executed. Both of these | |
2727 | ops trigger the "pop to last yes state if any, otherwise return true" | |
2728 | behaviour. | |
2729 | ||
2730 | A common convention in this function is to use A and B to refer to the two | |
2731 | subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is | |
2732 | the subpattern to be matched possibly multiple times, while B is the entire | |
2733 | rest of the pattern. Variable and state names reflect this convention. | |
2734 | ||
2735 | The states in the main switch are the union of ops and failure/success of | |
2736 | substates associated with with that op. For example, IFMATCH is the op | |
2737 | that does lookahead assertions /(?=A)B/ and so the IFMATCH state means | |
2738 | 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just | |
2739 | successfully matched A and IFMATCH_A_fail is a state saying that we have | |
2740 | just failed to match A. Resume states always come in pairs. The backtrack | |
2741 | state we push is marked as 'IFMATCH_A', but when that is popped, we resume | |
2742 | at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking | |
2743 | on success or failure. | |
2744 | ||
2745 | The struct that holds a backtracking state is actually a big union, with | |
2746 | one variant for each major type of op. The variable st points to the | |
2747 | top-most backtrack struct. To make the code clearer, within each | |
2748 | block of code we #define ST to alias the relevant union. | |
2749 | ||
2750 | Here's a concrete example of a (vastly oversimplified) IFMATCH | |
2751 | implementation: | |
2752 | ||
2753 | switch (state) { | |
2754 | .... | |
2755 | ||
2756 | #define ST st->u.ifmatch | |
2757 | ||
2758 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B | |
2759 | ST.foo = ...; // some state we wish to save | |
95b24440 | 2760 | ... |
bf1f174e DM |
2761 | // push a yes backtrack state with a resume value of |
2762 | // IFMATCH_A/IFMATCH_A_fail, then continue execution at the | |
2763 | // first node of A: | |
2764 | PUSH_YES_STATE_GOTO(IFMATCH_A, A); | |
2765 | // NOTREACHED | |
2766 | ||
2767 | case IFMATCH_A: // we have successfully executed A; now continue with B | |
2768 | next = B; | |
2769 | bar = ST.foo; // do something with the preserved value | |
2770 | break; | |
2771 | ||
2772 | case IFMATCH_A_fail: // A failed, so the assertion failed | |
2773 | ...; // do some housekeeping, then ... | |
2774 | sayNO; // propagate the failure | |
2775 | ||
2776 | #undef ST | |
95b24440 | 2777 | |
bf1f174e DM |
2778 | ... |
2779 | } | |
95b24440 | 2780 | |
bf1f174e DM |
2781 | For any old-timers reading this who are familiar with the old recursive |
2782 | approach, the code above is equivalent to: | |
95b24440 | 2783 | |
bf1f174e DM |
2784 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B |
2785 | { | |
2786 | int foo = ... | |
95b24440 | 2787 | ... |
bf1f174e DM |
2788 | if (regmatch(A)) { |
2789 | next = B; | |
2790 | bar = foo; | |
2791 | break; | |
95b24440 | 2792 | } |
bf1f174e DM |
2793 | ...; // do some housekeeping, then ... |
2794 | sayNO; // propagate the failure | |
95b24440 | 2795 | } |
bf1f174e DM |
2796 | |
2797 | The topmost backtrack state, pointed to by st, is usually free. If you | |
2798 | want to claim it, populate any ST.foo fields in it with values you wish to | |
2799 | save, then do one of | |
2800 | ||
2801 | PUSH_STATE_GOTO(resume_state, node); | |
2802 | PUSH_YES_STATE_GOTO(resume_state, node); | |
2803 | ||
2804 | which sets that backtrack state's resume value to 'resume_state', pushes a | |
2805 | new free entry to the top of the backtrack stack, then goes to 'node'. | |
2806 | On backtracking, the free slot is popped, and the saved state becomes the | |
2807 | new free state. An ST.foo field in this new top state can be temporarily | |
2808 | accessed to retrieve values, but once the main loop is re-entered, it | |
2809 | becomes available for reuse. | |
2810 | ||
2811 | Note that the depth of the backtrack stack constantly increases during the | |
2812 | left-to-right execution of the pattern, rather than going up and down with | |
2813 | the pattern nesting. For example the stack is at its maximum at Z at the | |
2814 | end of the pattern, rather than at X in the following: | |
2815 | ||
2816 | /(((X)+)+)+....(Y)+....Z/ | |
2817 | ||
2818 | The only exceptions to this are lookahead/behind assertions and the cut, | |
2819 | (?>A), which pop all the backtrack states associated with A before | |
2820 | continuing. | |
2821 | ||
486ec47a | 2822 | Backtrack state structs are allocated in slabs of about 4K in size. |
bf1f174e DM |
2823 | PL_regmatch_state and st always point to the currently active state, |
2824 | and PL_regmatch_slab points to the slab currently containing | |
2825 | PL_regmatch_state. The first time regmatch() is called, the first slab is | |
2826 | allocated, and is never freed until interpreter destruction. When the slab | |
2827 | is full, a new one is allocated and chained to the end. At exit from | |
2828 | regmatch(), slabs allocated since entry are freed. | |
2829 | ||
2830 | */ | |
95b24440 | 2831 | |
40a82448 | 2832 | |
5bc10b2c | 2833 | #define DEBUG_STATE_pp(pp) \ |
265c4333 | 2834 | DEBUG_STATE_r({ \ |
f2ed9b32 | 2835 | DUMP_EXEC_POS(locinput, scan, utf8_target); \ |
5bc10b2c | 2836 | PerlIO_printf(Perl_debug_log, \ |
5d458dd8 | 2837 | " %*s"pp" %s%s%s%s%s\n", \ |
5bc10b2c | 2838 | depth*2, "", \ |
13d6edb4 | 2839 | PL_reg_name[st->resume_state], \ |
5d458dd8 YO |
2840 | ((st==yes_state||st==mark_state) ? "[" : ""), \ |
2841 | ((st==yes_state) ? "Y" : ""), \ | |
2842 | ((st==mark_state) ? "M" : ""), \ | |
2843 | ((st==yes_state||st==mark_state) ? "]" : "") \ | |
2844 | ); \ | |
265c4333 | 2845 | }); |
5bc10b2c | 2846 | |
40a82448 | 2847 | |
3dab1dad | 2848 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 2849 | |
3df15adc | 2850 | #ifdef DEBUGGING |
5bc10b2c | 2851 | |
ab3bbdeb | 2852 | STATIC void |
f2ed9b32 | 2853 | S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target, |
ab3bbdeb YO |
2854 | const char *start, const char *end, const char *blurb) |
2855 | { | |
efd26800 | 2856 | const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; |
7918f24d NC |
2857 | |
2858 | PERL_ARGS_ASSERT_DEBUG_START_MATCH; | |
2859 | ||
ab3bbdeb YO |
2860 | if (!PL_colorset) |
2861 | reginitcolors(); | |
2862 | { | |
2863 | RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), | |
d2c6dc5e | 2864 | RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); |
ab3bbdeb | 2865 | |
f2ed9b32 | 2866 | RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb YO |
2867 | start, end - start, 60); |
2868 | ||
2869 | PerlIO_printf(Perl_debug_log, | |
2870 | "%s%s REx%s %s against %s\n", | |
2871 | PL_colors[4], blurb, PL_colors[5], s0, s1); | |
2872 | ||
f2ed9b32 | 2873 | if (utf8_target||utf8_pat) |
1de06328 YO |
2874 | PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", |
2875 | utf8_pat ? "pattern" : "", | |
f2ed9b32 KW |
2876 | utf8_pat && utf8_target ? " and " : "", |
2877 | utf8_target ? "string" : "" | |
ab3bbdeb YO |
2878 | ); |
2879 | } | |
2880 | } | |
3df15adc YO |
2881 | |
2882 | STATIC void | |
786e8c11 YO |
2883 | S_dump_exec_pos(pTHX_ const char *locinput, |
2884 | const regnode *scan, | |
2885 | const char *loc_regeol, | |
2886 | const char *loc_bostr, | |
2887 | const char *loc_reg_starttry, | |
f2ed9b32 | 2888 | const bool utf8_target) |
07be1b83 | 2889 | { |
786e8c11 | 2890 | const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; |
07be1b83 | 2891 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ |
786e8c11 | 2892 | int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); |
07be1b83 YO |
2893 | /* The part of the string before starttry has one color |
2894 | (pref0_len chars), between starttry and current | |
2895 | position another one (pref_len - pref0_len chars), | |
2896 | after the current position the third one. | |
2897 | We assume that pref0_len <= pref_len, otherwise we | |
2898 | decrease pref0_len. */ | |
786e8c11 YO |
2899 | int pref_len = (locinput - loc_bostr) > (5 + taill) - l |
2900 | ? (5 + taill) - l : locinput - loc_bostr; | |
07be1b83 YO |
2901 | int pref0_len; |
2902 | ||
7918f24d NC |
2903 | PERL_ARGS_ASSERT_DUMP_EXEC_POS; |
2904 | ||
f2ed9b32 | 2905 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) |
07be1b83 | 2906 | pref_len++; |
786e8c11 YO |
2907 | pref0_len = pref_len - (locinput - loc_reg_starttry); |
2908 | if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) | |
2909 | l = ( loc_regeol - locinput > (5 + taill) - pref_len | |
2910 | ? (5 + taill) - pref_len : loc_regeol - locinput); | |
f2ed9b32 | 2911 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) |
07be1b83 YO |
2912 | l--; |
2913 | if (pref0_len < 0) | |
2914 | pref0_len = 0; | |
2915 | if (pref0_len > pref_len) | |
2916 | pref0_len = pref_len; | |
2917 | { | |
f2ed9b32 | 2918 | const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0; |
0df25f3d | 2919 | |
ab3bbdeb | 2920 | RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), |
1de06328 | 2921 | (locinput - pref_len),pref0_len, 60, 4, 5); |
0df25f3d | 2922 | |
ab3bbdeb | 2923 | RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), |
3df15adc | 2924 | (locinput - pref_len + pref0_len), |
1de06328 | 2925 | pref_len - pref0_len, 60, 2, 3); |
0df25f3d | 2926 | |
ab3bbdeb | 2927 | RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), |
1de06328 | 2928 | locinput, loc_regeol - locinput, 10, 0, 1); |
0df25f3d | 2929 | |
1de06328 | 2930 | const STRLEN tlen=len0+len1+len2; |
3df15adc | 2931 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb | 2932 | "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", |
786e8c11 | 2933 | (IV)(locinput - loc_bostr), |
07be1b83 | 2934 | len0, s0, |
07be1b83 | 2935 | len1, s1, |
07be1b83 | 2936 | (docolor ? "" : "> <"), |
07be1b83 | 2937 | len2, s2, |
f9f4320a | 2938 | (int)(tlen > 19 ? 0 : 19 - tlen), |
07be1b83 YO |
2939 | ""); |
2940 | } | |
2941 | } | |
3df15adc | 2942 | |
07be1b83 YO |
2943 | #endif |
2944 | ||
0a4db386 YO |
2945 | /* reg_check_named_buff_matched() |
2946 | * Checks to see if a named buffer has matched. The data array of | |
2947 | * buffer numbers corresponding to the buffer is expected to reside | |
2948 | * in the regexp->data->data array in the slot stored in the ARG() of | |
2949 | * node involved. Note that this routine doesn't actually care about the | |
2950 | * name, that information is not preserved from compilation to execution. | |
2951 | * Returns the index of the leftmost defined buffer with the given name | |
2952 | * or 0 if non of the buffers matched. | |
2953 | */ | |
2954 | STATIC I32 | |
7918f24d NC |
2955 | S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) |
2956 | { | |
0a4db386 | 2957 | I32 n; |
f8fc2ecf | 2958 | RXi_GET_DECL(rex,rexi); |
ad64d0ec | 2959 | SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
0a4db386 | 2960 | I32 *nums=(I32*)SvPVX(sv_dat); |
7918f24d NC |
2961 | |
2962 | PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; | |
2963 | ||
0a4db386 YO |
2964 | for ( n=0; n<SvIVX(sv_dat); n++ ) { |
2965 | if ((I32)*PL_reglastparen >= nums[n] && | |
f0ab9afb | 2966 | PL_regoffs[nums[n]].end != -1) |
0a4db386 YO |
2967 | { |
2968 | return nums[n]; | |
2969 | } | |
2970 | } | |
2971 | return 0; | |
2972 | } | |
2973 | ||
2f554ef7 DM |
2974 | |
2975 | /* free all slabs above current one - called during LEAVE_SCOPE */ | |
2976 | ||
2977 | STATIC void | |
2978 | S_clear_backtrack_stack(pTHX_ void *p) | |
2979 | { | |
2980 | regmatch_slab *s = PL_regmatch_slab->next; | |
2981 | PERL_UNUSED_ARG(p); | |
2982 | ||
2983 | if (!s) | |
2984 | return; | |
2985 | PL_regmatch_slab->next = NULL; | |
2986 | while (s) { | |
2987 | regmatch_slab * const osl = s; | |
2988 | s = s->next; | |
2989 | Safefree(osl); | |
2990 | } | |
2991 | } | |
2992 | ||
2993 | ||
28d8d7f4 YO |
2994 | #define SETREX(Re1,Re2) \ |
2995 | if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \ | |
2996 | Re1 = (Re2) | |
2997 | ||
d6a28714 | 2998 | STATIC I32 /* 0 failure, 1 success */ |
24b23f37 | 2999 | S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) |
d6a28714 | 3000 | { |
a35a87e7 | 3001 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
3002 | dMY_CXT; |
3003 | #endif | |
27da23d5 | 3004 | dVAR; |
f2ed9b32 | 3005 | register const bool utf8_target = PL_reg_match_utf8; |
4ad0818d | 3006 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
288b8c02 NC |
3007 | REGEXP *rex_sv = reginfo->prog; |
3008 | regexp *rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 3009 | RXi_GET_DECL(rex,rexi); |
2f554ef7 | 3010 | I32 oldsave; |
5d9a96ca DM |
3011 | /* the current state. This is a cached copy of PL_regmatch_state */ |
3012 | register regmatch_state *st; | |
5d9a96ca DM |
3013 | /* cache heavy used fields of st in registers */ |
3014 | register regnode *scan; | |
3015 | register regnode *next; | |
438e9bae | 3016 | register U32 n = 0; /* general value; init to avoid compiler warning */ |
24d3c4a9 | 3017 | register I32 ln = 0; /* len or last; init to avoid compiler warning */ |
5d9a96ca | 3018 | register char *locinput = PL_reginput; |
5d9a96ca | 3019 | register I32 nextchr; /* is always set to UCHARAT(locinput) */ |
24d3c4a9 | 3020 | |
b69b0499 | 3021 | bool result = 0; /* return value of S_regmatch */ |
24d3c4a9 | 3022 | int depth = 0; /* depth of backtrack stack */ |
4b196cd4 YO |
3023 | U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */ |
3024 | const U32 max_nochange_depth = | |
3025 | (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? | |
3026 | 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; | |
77cb431f DM |
3027 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
3028 | subpattern */ | |
e2e6a0f1 YO |
3029 | /* mark_state piggy backs on the yes_state logic so that when we unwind |
3030 | the stack on success we can update the mark_state as we go */ | |
3031 | regmatch_state *mark_state = NULL; /* last mark state we have seen */ | |
faec1544 | 3032 | regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ |
b8591aee | 3033 | struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ |
40a82448 | 3034 | U32 state_num; |
5d458dd8 YO |
3035 | bool no_final = 0; /* prevent failure from backtracking? */ |
3036 | bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ | |
e2e6a0f1 | 3037 | char *startpoint = PL_reginput; |
5d458dd8 YO |
3038 | SV *popmark = NULL; /* are we looking for a mark? */ |
3039 | SV *sv_commit = NULL; /* last mark name seen in failure */ | |
3040 | SV *sv_yes_mark = NULL; /* last mark name we have seen | |
486ec47a | 3041 | during a successful match */ |
5d458dd8 YO |
3042 | U32 lastopen = 0; /* last open we saw */ |
3043 | bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; | |
19b95bf0 | 3044 | SV* const oreplsv = GvSV(PL_replgv); |
24d3c4a9 DM |
3045 | /* these three flags are set by various ops to signal information to |
3046 | * the very next op. They have a useful lifetime of exactly one loop | |
3047 | * iteration, and are not preserved or restored by state pushes/pops | |
3048 | */ | |
3049 | bool sw = 0; /* the condition value in (?(cond)a|b) */ | |
3050 | bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ | |
3051 | int logical = 0; /* the following EVAL is: | |
3052 | 0: (?{...}) | |
3053 | 1: (?(?{...})X|Y) | |
3054 | 2: (??{...}) | |
3055 | or the following IFMATCH/UNLESSM is: | |
3056 | false: plain (?=foo) | |
3057 | true: used as a condition: (?(?=foo)) | |
3058 | */ | |
95b24440 | 3059 | #ifdef DEBUGGING |
e68ec53f | 3060 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
3061 | #endif |
3062 | ||
7918f24d NC |
3063 | PERL_ARGS_ASSERT_REGMATCH; |
3064 | ||
3b57cd43 | 3065 | DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ |
24b23f37 | 3066 | PerlIO_printf(Perl_debug_log,"regmatch start\n"); |
3b57cd43 | 3067 | })); |
5d9a96ca DM |
3068 | /* on first ever call to regmatch, allocate first slab */ |
3069 | if (!PL_regmatch_slab) { | |
3070 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
3071 | PL_regmatch_slab->prev = NULL; | |
3072 | PL_regmatch_slab->next = NULL; | |
86545054 | 3073 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); |
5d9a96ca DM |
3074 | } |
3075 | ||
2f554ef7 DM |
3076 | oldsave = PL_savestack_ix; |
3077 | SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL); | |
3078 | SAVEVPTR(PL_regmatch_slab); | |
3079 | SAVEVPTR(PL_regmatch_state); | |
5d9a96ca DM |
3080 | |
3081 | /* grab next free state slot */ | |
3082 | st = ++PL_regmatch_state; | |
86545054 | 3083 | if (st > SLAB_LAST(PL_regmatch_slab)) |
5d9a96ca DM |
3084 | st = PL_regmatch_state = S_push_slab(aTHX); |
3085 | ||
d6a28714 JH |
3086 | /* Note that nextchr is a byte even in UTF */ |
3087 | nextchr = UCHARAT(locinput); | |
3088 | scan = prog; | |
3089 | while (scan != NULL) { | |
8ba1375e | 3090 | |
a3621e74 | 3091 | DEBUG_EXECUTE_r( { |
6136c704 | 3092 | SV * const prop = sv_newmortal(); |
1de06328 | 3093 | regnode *rnext=regnext(scan); |
f2ed9b32 | 3094 | DUMP_EXEC_POS( locinput, scan, utf8_target ); |
32fc9b6a | 3095 | regprop(rex, prop, scan); |
07be1b83 YO |
3096 | |
3097 | PerlIO_printf(Perl_debug_log, | |
3098 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
f8fc2ecf | 3099 | (IV)(scan - rexi->program), depth*2, "", |
07be1b83 | 3100 | SvPVX_const(prop), |
1de06328 | 3101 | (PL_regkind[OP(scan)] == END || !rnext) ? |
f8fc2ecf | 3102 | 0 : (IV)(rnext - rexi->program)); |
2a782b5b | 3103 | }); |
d6a28714 JH |
3104 | |
3105 | next = scan + NEXT_OFF(scan); | |
3106 | if (next == scan) | |
3107 | next = NULL; | |
40a82448 | 3108 | state_num = OP(scan); |
d6a28714 | 3109 | |
40a82448 | 3110 | reenter_switch: |
34a81e2b B |
3111 | |
3112 | assert(PL_reglastparen == &rex->lastparen); | |
3113 | assert(PL_reglastcloseparen == &rex->lastcloseparen); | |
3114 | assert(PL_regoffs == rex->offs); | |
3115 | ||
40a82448 | 3116 | switch (state_num) { |
d6a28714 | 3117 | case BOL: |
7fba1cd6 | 3118 | if (locinput == PL_bostr) |
d6a28714 | 3119 | { |
3b0527fe | 3120 | /* reginfo->till = reginfo->bol; */ |
b8c5462f JH |
3121 | break; |
3122 | } | |
d6a28714 JH |
3123 | sayNO; |
3124 | case MBOL: | |
12d33761 HS |
3125 | if (locinput == PL_bostr || |
3126 | ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n')) | |
d6a28714 | 3127 | { |
b8c5462f JH |
3128 | break; |
3129 | } | |
d6a28714 JH |
3130 | sayNO; |
3131 | case SBOL: | |
c2a73568 | 3132 | if (locinput == PL_bostr) |
b8c5462f | 3133 | break; |
d6a28714 JH |
3134 | sayNO; |
3135 | case GPOS: | |
3b0527fe | 3136 | if (locinput == reginfo->ganch) |
d6a28714 JH |
3137 | break; |
3138 | sayNO; | |
ee9b8eae YO |
3139 | |
3140 | case KEEPS: | |
3141 | /* update the startpoint */ | |
f0ab9afb | 3142 | st->u.keeper.val = PL_regoffs[0].start; |
ee9b8eae | 3143 | PL_reginput = locinput; |
f0ab9afb | 3144 | PL_regoffs[0].start = locinput - PL_bostr; |
ee9b8eae YO |
3145 | PUSH_STATE_GOTO(KEEPS_next, next); |
3146 | /*NOT-REACHED*/ | |
3147 | case KEEPS_next_fail: | |
3148 | /* rollback the start point change */ | |
f0ab9afb | 3149 | PL_regoffs[0].start = st->u.keeper.val; |
ee9b8eae YO |
3150 | sayNO_SILENT; |
3151 | /*NOT-REACHED*/ | |
d6a28714 | 3152 | case EOL: |
d6a28714 JH |
3153 | goto seol; |
3154 | case MEOL: | |
d6a28714 | 3155 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') |
b8c5462f | 3156 | sayNO; |
b8c5462f | 3157 | break; |
d6a28714 JH |
3158 | case SEOL: |
3159 | seol: | |
3160 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') | |
b8c5462f | 3161 | sayNO; |
d6a28714 | 3162 | if (PL_regeol - locinput > 1) |
b8c5462f | 3163 | sayNO; |
b8c5462f | 3164 | break; |
d6a28714 JH |
3165 | case EOS: |
3166 | if (PL_regeol != locinput) | |
b8c5462f | 3167 | sayNO; |
d6a28714 | 3168 | break; |
ffc61ed2 | 3169 | case SANY: |
d6a28714 | 3170 | if (!nextchr && locinput >= PL_regeol) |
4633a7c4 | 3171 | sayNO; |
f2ed9b32 | 3172 | if (utf8_target) { |
f33976b4 DB |
3173 | locinput += PL_utf8skip[nextchr]; |
3174 | if (locinput > PL_regeol) | |
3175 | sayNO; | |
3176 | nextchr = UCHARAT(locinput); | |
3177 | } | |
3178 | else | |
3179 | nextchr = UCHARAT(++locinput); | |
3180 | break; | |
3181 | case CANY: | |
3182 | if (!nextchr && locinput >= PL_regeol) | |
3183 | sayNO; | |
b8c5462f | 3184 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 3185 | break; |
ffc61ed2 | 3186 | case REG_ANY: |
1aa99e6b IH |
3187 | if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n') |
3188 | sayNO; | |
f2ed9b32 | 3189 | if (utf8_target) { |
b8c5462f | 3190 | locinput += PL_utf8skip[nextchr]; |
d6a28714 JH |
3191 | if (locinput > PL_regeol) |
3192 | sayNO; | |
a0ed51b3 | 3193 | nextchr = UCHARAT(locinput); |
a0ed51b3 | 3194 | } |
1aa99e6b IH |
3195 | else |
3196 | nextchr = UCHARAT(++locinput); | |
a0ed51b3 | 3197 | break; |
166ba7cd DM |
3198 | |
3199 | #undef ST | |
3200 | #define ST st->u.trie | |
786e8c11 YO |
3201 | case TRIEC: |
3202 | /* In this case the charclass data is available inline so | |
3203 | we can fail fast without a lot of extra overhead. | |
3204 | */ | |
f2ed9b32 | 3205 | if (scan->flags == EXACT || !utf8_target) { |
786e8c11 YO |
3206 | if(!ANYOF_BITMAP_TEST(scan, *locinput)) { |
3207 | DEBUG_EXECUTE_r( | |
3208 | PerlIO_printf(Perl_debug_log, | |
3209 | "%*s %sfailed to match trie start class...%s\n", | |
5bc10b2c | 3210 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
786e8c11 YO |
3211 | ); |
3212 | sayNO_SILENT; | |
3213 | /* NOTREACHED */ | |
3214 | } | |
3215 | } | |
3216 | /* FALL THROUGH */ | |
5b47454d | 3217 | case TRIE: |
2e64971a DM |
3218 | /* the basic plan of execution of the trie is: |
3219 | * At the beginning, run though all the states, and | |
3220 | * find the longest-matching word. Also remember the position | |
3221 | * of the shortest matching word. For example, this pattern: | |
3222 | * 1 2 3 4 5 | |
3223 | * ab|a|x|abcd|abc | |
3224 | * when matched against the string "abcde", will generate | |
3225 | * accept states for all words except 3, with the longest | |
3226 | * matching word being 4, and the shortest being 1 (with | |
3227 | * the position being after char 1 of the string). | |
3228 | * | |
3229 | * Then for each matching word, in word order (i.e. 1,2,4,5), | |
3230 | * we run the remainder of the pattern; on each try setting | |
3231 | * the current position to the character following the word, | |
3232 | * returning to try the next word on failure. | |
3233 | * | |
3234 | * We avoid having to build a list of words at runtime by | |
3235 | * using a compile-time structure, wordinfo[].prev, which | |
3236 | * gives, for each word, the previous accepting word (if any). | |
3237 | * In the case above it would contain the mappings 1->2, 2->0, | |
3238 | * 3->0, 4->5, 5->1. We can use this table to generate, from | |
3239 | * the longest word (4 above), a list of all words, by | |
3240 | * following the list of prev pointers; this gives us the | |
3241 | * unordered list 4,5,1,2. Then given the current word we have | |
3242 | * just tried, we can go through the list and find the | |
3243 | * next-biggest word to try (so if we just failed on word 2, | |
3244 | * the next in the list is 4). | |
3245 | * | |
3246 | * Since at runtime we don't record the matching position in | |
3247 | * the string for each word, we have to work that out for | |
3248 | * each word we're about to process. The wordinfo table holds | |
3249 | * the character length of each word; given that we recorded | |
3250 | * at the start: the position of the shortest word and its | |
3251 | * length in chars, we just need to move the pointer the | |
3252 | * difference between the two char lengths. Depending on | |
3253 | * Unicode status and folding, that's cheap or expensive. | |
3254 | * | |
3255 | * This algorithm is optimised for the case where are only a | |
3256 | * small number of accept states, i.e. 0,1, or maybe 2. | |
3257 | * With lots of accepts states, and having to try all of them, | |
3258 | * it becomes quadratic on number of accept states to find all | |
3259 | * the next words. | |
3260 | */ | |
3261 | ||
3dab1dad | 3262 | { |
07be1b83 | 3263 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
a0a388a1 | 3264 | DECL_TRIE_TYPE(scan); |
3dab1dad YO |
3265 | |
3266 | /* what trie are we using right now */ | |
be8e71aa | 3267 | reg_trie_data * const trie |
f8fc2ecf | 3268 | = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; |
85fbaab2 | 3269 | HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); |
3dab1dad | 3270 | U32 state = trie->startstate; |
166ba7cd | 3271 | |
3dab1dad YO |
3272 | if (trie->bitmap && trie_type != trie_utf8_fold && |
3273 | !TRIE_BITMAP_TEST(trie,*locinput) | |
3274 | ) { | |
3275 | if (trie->states[ state ].wordnum) { | |
3276 | DEBUG_EXECUTE_r( | |
3277 | PerlIO_printf(Perl_debug_log, | |
3278 | "%*s %smatched empty string...%s\n", | |
5bc10b2c | 3279 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad | 3280 | ); |
20dbff7c YO |
3281 | if (!trie->jump) |
3282 | break; | |
3dab1dad YO |
3283 | } else { |
3284 | DEBUG_EXECUTE_r( | |
3285 | PerlIO_printf(Perl_debug_log, | |
786e8c11 | 3286 | "%*s %sfailed to match trie start class...%s\n", |
5bc10b2c | 3287 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad YO |
3288 | ); |
3289 | sayNO_SILENT; | |
3290 | } | |
3291 | } | |
166ba7cd | 3292 | |
786e8c11 YO |
3293 | { |
3294 | U8 *uc = ( U8* )locinput; | |
3295 | ||
3296 | STRLEN len = 0; | |
3297 | STRLEN foldlen = 0; | |
3298 | U8 *uscan = (U8*)NULL; | |
786e8c11 | 3299 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
2e64971a DM |
3300 | U32 charcount = 0; /* how many input chars we have matched */ |
3301 | U32 accepted = 0; /* have we seen any accepting states? */ | |
786e8c11 | 3302 | |
786e8c11 YO |
3303 | ST.B = next; |
3304 | ST.jump = trie->jump; | |
786e8c11 | 3305 | ST.me = scan; |
2e64971a DM |
3306 | ST.firstpos = NULL; |
3307 | ST.longfold = FALSE; /* char longer if folded => it's harder */ | |
3308 | ST.nextword = 0; | |
3309 | ||
3310 | /* fully traverse the TRIE; note the position of the | |
3311 | shortest accept state and the wordnum of the longest | |
3312 | accept state */ | |
07be1b83 | 3313 | |
a3621e74 | 3314 | while ( state && uc <= (U8*)PL_regeol ) { |
786e8c11 | 3315 | U32 base = trie->states[ state ].trans.base; |
f9f4320a | 3316 | UV uvc = 0; |
acb909b4 | 3317 | U16 charid = 0; |
2e64971a DM |
3318 | U16 wordnum; |
3319 | wordnum = trie->states[ state ].wordnum; | |
3320 | ||
3321 | if (wordnum) { /* it's an accept state */ | |
3322 | if (!accepted) { | |
3323 | accepted = 1; | |
3324 | /* record first match position */ | |
3325 | if (ST.longfold) { | |
3326 | ST.firstpos = (U8*)locinput; | |
3327 | ST.firstchars = 0; | |
5b47454d | 3328 | } |
2e64971a DM |
3329 | else { |
3330 | ST.firstpos = uc; | |
3331 | ST.firstchars = charcount; | |
3332 | } | |
3333 | } | |
3334 | if (!ST.nextword || wordnum < ST.nextword) | |
3335 | ST.nextword = wordnum; | |
3336 | ST.topword = wordnum; | |
786e8c11 | 3337 | } |
a3621e74 | 3338 | |
07be1b83 | 3339 | DEBUG_TRIE_EXECUTE_r({ |
f2ed9b32 | 3340 | DUMP_EXEC_POS( (char *)uc, scan, utf8_target ); |
a3621e74 | 3341 | PerlIO_printf( Perl_debug_log, |
2e64971a | 3342 | "%*s %sState: %4"UVxf" Accepted: %c ", |
5bc10b2c | 3343 | 2+depth * 2, "", PL_colors[4], |
2e64971a | 3344 | (UV)state, (accepted ? 'Y' : 'N')); |
07be1b83 | 3345 | }); |
a3621e74 | 3346 | |
2e64971a | 3347 | /* read a char and goto next state */ |
a3621e74 | 3348 | if ( base ) { |
6dd2be57 | 3349 | I32 offset; |
55eed653 NC |
3350 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
3351 | uscan, len, uvc, charid, foldlen, | |
3352 | foldbuf, uniflags); | |
2e64971a DM |
3353 | charcount++; |
3354 | if (foldlen>0) | |
3355 | ST.longfold = TRUE; | |
5b47454d | 3356 | if (charid && |
6dd2be57 DM |
3357 | ( ((offset = |
3358 | base + charid - 1 - trie->uniquecharcount)) >= 0) | |
3359 | ||
3360 | && ((U32)offset < trie->lasttrans) | |
3361 | && trie->trans[offset].check == state) | |
5b47454d | 3362 | { |
6dd2be57 | 3363 | state = trie->trans[offset].next; |
5b47454d DM |
3364 | } |
3365 | else { | |
3366 | state = 0; | |
3367 | } | |
3368 | uc += len; | |
3369 | ||
3370 | } | |
3371 | else { | |
a3621e74 YO |
3372 | state = 0; |
3373 | } | |
3374 | DEBUG_TRIE_EXECUTE_r( | |
e4584336 | 3375 | PerlIO_printf( Perl_debug_log, |
786e8c11 | 3376 | "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n", |
e4584336 | 3377 | charid, uvc, (UV)state, PL_colors[5] ); |
a3621e74 YO |
3378 | ); |
3379 | } | |
2e64971a | 3380 | if (!accepted) |
a3621e74 | 3381 | sayNO; |
a3621e74 | 3382 | |
2e64971a DM |
3383 | /* calculate total number of accept states */ |
3384 | { | |
3385 | U16 w = ST.topword; | |
3386 | accepted = 0; | |
3387 | while (w) { | |
3388 | w = trie->wordinfo[w].prev; | |
3389 | accepted++; | |
3390 | } | |
3391 | ST.accepted = accepted; | |
3392 | } | |
3393 | ||
166ba7cd DM |
3394 | DEBUG_EXECUTE_r( |
3395 | PerlIO_printf( Perl_debug_log, | |
3396 | "%*s %sgot %"IVdf" possible matches%s\n", | |
5bc10b2c | 3397 | REPORT_CODE_OFF + depth * 2, "", |
166ba7cd DM |
3398 | PL_colors[4], (IV)ST.accepted, PL_colors[5] ); |
3399 | ); | |
2e64971a | 3400 | goto trie_first_try; /* jump into the fail handler */ |
786e8c11 | 3401 | }} |
fae667d5 | 3402 | /* NOTREACHED */ |
2e64971a DM |
3403 | |
3404 | case TRIE_next_fail: /* we failed - try next alternative */ | |
fae667d5 YO |
3405 | if ( ST.jump) { |
3406 | REGCP_UNWIND(ST.cp); | |
3407 | for (n = *PL_reglastparen; n > ST.lastparen; n--) | |
f0ab9afb | 3408 | PL_regoffs[n].end = -1; |
fae667d5 YO |
3409 | *PL_reglastparen = n; |
3410 | } | |
2e64971a DM |
3411 | if (!--ST.accepted) { |
3412 | DEBUG_EXECUTE_r({ | |
3413 | PerlIO_printf( Perl_debug_log, | |
3414 | "%*s %sTRIE failed...%s\n", | |
3415 | REPORT_CODE_OFF+depth*2, "", | |
3416 | PL_colors[4], | |
3417 | PL_colors[5] ); | |
3418 | }); | |
3419 | sayNO_SILENT; | |
3420 | } | |
3421 | { | |
3422 | /* Find next-highest word to process. Note that this code | |
3423 | * is O(N^2) per trie run (O(N) per branch), so keep tight */ | |
d9a396a3 DM |
3424 | register U16 min = 0; |
3425 | register U16 word; | |
2e64971a DM |
3426 | register U16 const nextword = ST.nextword; |
3427 | register reg_trie_wordinfo * const wordinfo | |
3428 | = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo; | |
3429 | for (word=ST.topword; word; word=wordinfo[word].prev) { | |
3430 | if (word > nextword && (!min || word < min)) | |
3431 | min = word; | |
3432 | } | |
3433 | ST.nextword = min; | |
3434 | } | |
3435 | ||
fae667d5 | 3436 | trie_first_try: |
5d458dd8 YO |
3437 | if (do_cutgroup) { |
3438 | do_cutgroup = 0; | |
3439 | no_final = 0; | |
3440 | } | |
fae667d5 YO |
3441 | |
3442 | if ( ST.jump) { | |
3443 | ST.lastparen = *PL_reglastparen; | |
3444 | REGCP_SET(ST.cp); | |
2e64971a | 3445 | } |
a3621e74 | 3446 | |
2e64971a | 3447 | /* find start char of end of current word */ |
166ba7cd | 3448 | { |
2e64971a DM |
3449 | U32 chars; /* how many chars to skip */ |
3450 | U8 *uc = ST.firstpos; | |
3451 | reg_trie_data * const trie | |
3452 | = (reg_trie_data*)rexi->data->data[ARG(ST.me)]; | |
3453 | ||
3454 | assert((trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3455 | >= ST.firstchars); | |
3456 | chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3457 | - ST.firstchars; | |
3458 | ||
3459 | if (ST.longfold) { | |
3460 | /* the hard option - fold each char in turn and find | |
3461 | * its folded length (which may be different */ | |
3462 | U8 foldbuf[UTF8_MAXBYTES_CASE + 1]; | |
3463 | STRLEN foldlen; | |
3464 | STRLEN len; | |
d9a396a3 | 3465 | UV uvc; |
2e64971a DM |
3466 | U8 *uscan; |
3467 | ||
3468 | while (chars) { | |
f2ed9b32 | 3469 | if (utf8_target) { |
2e64971a DM |
3470 | uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len, |
3471 | uniflags); | |
3472 | uc += len; | |
3473 | } | |
3474 | else { | |
3475 | uvc = *uc; | |
3476 | uc++; | |
3477 | } | |
3478 | uvc = to_uni_fold(uvc, foldbuf, &foldlen); | |
3479 | uscan = foldbuf; | |
3480 | while (foldlen) { | |
3481 | if (!--chars) | |
3482 | break; | |
3483 | uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len, | |
3484 | uniflags); | |
3485 | uscan += len; | |
3486 | foldlen -= len; | |
3487 | } | |
3488 | } | |
a3621e74 | 3489 | } |
2e64971a | 3490 | else { |
f2ed9b32 | 3491 | if (utf8_target) |
2e64971a DM |
3492 | while (chars--) |
3493 | uc += UTF8SKIP(uc); | |
3494 | else | |
3495 | uc += chars; | |
3496 | } | |
3497 | PL_reginput = (char *)uc; | |
3498 | } | |
166ba7cd | 3499 | |
2e64971a DM |
3500 | scan = (ST.jump && ST.jump[ST.nextword]) |
3501 | ? ST.me + ST.jump[ST.nextword] | |
3502 | : ST.B; | |
166ba7cd | 3503 | |
2e64971a DM |
3504 | DEBUG_EXECUTE_r({ |
3505 | PerlIO_printf( Perl_debug_log, | |
3506 | "%*s %sTRIE matched word #%d, continuing%s\n", | |
3507 | REPORT_CODE_OFF+depth*2, "", | |
3508 | PL_colors[4], | |
3509 | ST.nextword, | |
3510 | PL_colors[5] | |
3511 | ); | |
3512 | }); | |
3513 | ||
3514 | if (ST.accepted > 1 || has_cutgroup) { | |
3515 | PUSH_STATE_GOTO(TRIE_next, scan); | |
3516 | /* NOTREACHED */ | |
166ba7cd | 3517 | } |
2e64971a DM |
3518 | /* only one choice left - just continue */ |
3519 | DEBUG_EXECUTE_r({ | |
3520 | AV *const trie_words | |
3521 | = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]); | |
3522 | SV ** const tmp = av_fetch( trie_words, | |
3523 | ST.nextword-1, 0 ); | |
3524 | SV *sv= tmp ? sv_newmortal() : NULL; | |
3525 | ||
3526 | PerlIO_printf( Perl_debug_log, | |
3527 | "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n", | |
3528 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], | |
3529 | ST.nextword, | |
3530 | tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0, | |
3531 | PL_colors[0], PL_colors[1], | |
c89df6cf | 3532 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII |
2e64971a DM |
3533 | ) |
3534 | : "not compiled under -Dr", | |
3535 | PL_colors[5] ); | |
3536 | }); | |
3537 | ||
3538 | locinput = PL_reginput; | |
3539 | nextchr = UCHARAT(locinput); | |
3540 | continue; /* execute rest of RE */ | |
166ba7cd | 3541 | /* NOTREACHED */ |
166ba7cd DM |
3542 | #undef ST |
3543 | ||
95b24440 DM |
3544 | case EXACT: { |
3545 | char *s = STRING(scan); | |
24d3c4a9 | 3546 | ln = STR_LEN(scan); |
f2ed9b32 | 3547 | if (utf8_target != UTF_PATTERN) { |
bc517b45 | 3548 | /* The target and the pattern have differing utf8ness. */ |
1aa99e6b | 3549 | char *l = locinput; |
24d3c4a9 | 3550 | const char * const e = s + ln; |
a72c7584 | 3551 | |
f2ed9b32 | 3552 | if (utf8_target) { |
5ff6fc6d | 3553 | /* The target is utf8, the pattern is not utf8. */ |
1aa99e6b | 3554 | while (s < e) { |
a3b680e6 | 3555 | STRLEN ulen; |
1aa99e6b | 3556 | if (l >= PL_regeol) |
5ff6fc6d JH |
3557 | sayNO; |
3558 | if (NATIVE_TO_UNI(*(U8*)s) != | |
89ebb4a3 | 3559 | utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen, |
041457d9 | 3560 | uniflags)) |
5ff6fc6d | 3561 | sayNO; |
bc517b45 | 3562 | l += ulen; |
5ff6fc6d | 3563 | s ++; |
1aa99e6b | 3564 | } |
5ff6fc6d JH |
3565 | } |
3566 | else { | |
3567 | /* The target is not utf8, the pattern is utf8. */ | |
1aa99e6b | 3568 | while (s < e) { |
a3b680e6 | 3569 | STRLEN ulen; |
1aa99e6b IH |
3570 | if (l >= PL_regeol) |
3571 | sayNO; | |
5ff6fc6d | 3572 | if (NATIVE_TO_UNI(*((U8*)l)) != |
89ebb4a3 | 3573 | utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen, |
041457d9 | 3574 | uniflags)) |
1aa99e6b | 3575 | sayNO; |
bc517b45 | 3576 | s += ulen; |
a72c7584 | 3577 | l ++; |
1aa99e6b | 3578 | } |
5ff6fc6d | 3579 | } |
1aa99e6b IH |
3580 | locinput = l; |
3581 | nextchr = UCHARAT(locinput); | |
3582 | break; | |
3583 | } | |
bc517b45 | 3584 | /* The target and the pattern have the same utf8ness. */ |
d6a28714 JH |
3585 | /* Inline the first character, for speed. */ |
3586 | if (UCHARAT(s) != nextchr) | |
3587 | sayNO; | |
24d3c4a9 | 3588 | if (PL_regeol - locinput < ln) |
d6a28714 | 3589 | sayNO; |
24d3c4a9 | 3590 | if (ln > 1 && memNE(s, locinput, ln)) |
d6a28714 | 3591 | sayNO; |
24d3c4a9 | 3592 | locinput += ln; |
d6a28714 JH |
3593 | nextchr = UCHARAT(locinput); |
3594 | break; | |
95b24440 | 3595 | } |
9a5a5549 | 3596 | case EXACTFL: { |
a932d541 | 3597 | re_fold_t folder; |
9a5a5549 KW |
3598 | const U8 * fold_array; |
3599 | const char * s; | |
d513472c | 3600 | U32 fold_utf8_flags; |
9a5a5549 | 3601 | |
b8c5462f | 3602 | PL_reg_flags |= RF_tainted; |
9a5a5549 KW |
3603 | folder = foldEQ_locale; |
3604 | fold_array = PL_fold_locale; | |
17580e7a | 3605 | fold_utf8_flags = FOLDEQ_UTF8_LOCALE; |
9a5a5549 KW |
3606 | goto do_exactf; |
3607 | ||
3608 | case EXACTFU: | |
3609 | folder = foldEQ_latin1; | |
3610 | fold_array = PL_fold_latin1; | |
d513472c | 3611 | fold_utf8_flags = 0; |
9a5a5549 KW |
3612 | goto do_exactf; |
3613 | ||
2f7f8cb1 KW |
3614 | case EXACTFA: |
3615 | folder = foldEQ_latin1; | |
3616 | fold_array = PL_fold_latin1; | |
57014d77 | 3617 | fold_utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; |
2f7f8cb1 KW |
3618 | goto do_exactf; |
3619 | ||
9a5a5549 KW |
3620 | case EXACTF: |
3621 | folder = foldEQ; | |
3622 | fold_array = PL_fold; | |
d513472c | 3623 | fold_utf8_flags = 0; |
9a5a5549 KW |
3624 | |
3625 | do_exactf: | |
3626 | s = STRING(scan); | |
24d3c4a9 | 3627 | ln = STR_LEN(scan); |
d6a28714 | 3628 | |
f2ed9b32 | 3629 | if (utf8_target || UTF_PATTERN) { |
d07ddd77 | 3630 | /* Either target or the pattern are utf8. */ |
be8e71aa | 3631 | const char * const l = locinput; |
d07ddd77 | 3632 | char *e = PL_regeol; |
bc517b45 | 3633 | |
d513472c | 3634 | if (! foldEQ_utf8_flags(s, 0, ln, cBOOL(UTF_PATTERN), |
c3e1d013 KW |
3635 | l, &e, 0, utf8_target, fold_utf8_flags)) |
3636 | { | |
3637 | sayNO; | |
5486206c | 3638 | } |
d07ddd77 JH |
3639 | locinput = e; |
3640 | nextchr = UCHARAT(locinput); | |
3641 | break; | |
a0ed51b3 | 3642 | } |
d6a28714 | 3643 | |
0a138b74 | 3644 | /* Neither the target nor the pattern are utf8 */ |
d6a28714 | 3645 | if (UCHARAT(s) != nextchr && |
9a5a5549 KW |
3646 | UCHARAT(s) != fold_array[nextchr]) |
3647 | { | |
a0ed51b3 | 3648 | sayNO; |
9a5a5549 | 3649 | } |
24d3c4a9 | 3650 | if (PL_regeol - locinput < ln) |
b8c5462f | 3651 | sayNO; |
9a5a5549 | 3652 | if (ln > 1 && ! folder(s, locinput, ln)) |
4633a7c4 | 3653 | sayNO; |
24d3c4a9 | 3654 | locinput += ln; |
d6a28714 | 3655 | nextchr = UCHARAT(locinput); |
a0d0e21e | 3656 | break; |
9a5a5549 | 3657 | } |
63ac0dad KW |
3658 | |
3659 | /* XXX Could improve efficiency by separating these all out using a | |
3660 | * macro or in-line function. At that point regcomp.c would no longer | |
3661 | * have to set the FLAGS fields of these */ | |
b2680017 YO |
3662 | case BOUNDL: |
3663 | case NBOUNDL: | |
3664 | PL_reg_flags |= RF_tainted; | |
3665 | /* FALL THROUGH */ | |
3666 | case BOUND: | |
63ac0dad | 3667 | case BOUNDU: |
cfaf538b | 3668 | case BOUNDA: |
b2680017 | 3669 | case NBOUND: |
63ac0dad | 3670 | case NBOUNDU: |
cfaf538b | 3671 | case NBOUNDA: |
b2680017 | 3672 | /* was last char in word? */ |
cfaf538b | 3673 | if (utf8_target && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET) { |
b2680017 YO |
3674 | if (locinput == PL_bostr) |
3675 | ln = '\n'; | |
3676 | else { | |
3677 | const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); | |
3678 | ||
3679 | ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); | |
3680 | } | |
63ac0dad | 3681 | if (FLAGS(scan) != REGEX_LOCALE_CHARSET) { |
b2680017 YO |
3682 | ln = isALNUM_uni(ln); |
3683 | LOAD_UTF8_CHARCLASS_ALNUM(); | |
f2ed9b32 | 3684 | n = swash_fetch(PL_utf8_alnum, (U8*)locinput, utf8_target); |
b2680017 YO |
3685 | } |
3686 | else { | |
3687 | ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); | |
3688 | n = isALNUM_LC_utf8((U8*)locinput); | |
3689 | } | |
3690 | } | |
3691 | else { | |
cfaf538b KW |
3692 | |
3693 | /* Here the string isn't utf8, or is utf8 and only ascii | |
3694 | * characters are to match \w. In the latter case looking at | |
3695 | * the byte just prior to the current one may be just the final | |
3696 | * byte of a multi-byte character. This is ok. There are two | |
3697 | * cases: | |
3698 | * 1) it is a single byte character, and then the test is doing | |
3699 | * just what it's supposed to. | |
3700 | * 2) it is a multi-byte character, in which case the final | |
3701 | * byte is never mistakable for ASCII, and so the test | |
3702 | * will say it is not a word character, which is the | |
3703 | * correct answer. */ | |
b2680017 YO |
3704 | ln = (locinput != PL_bostr) ? |
3705 | UCHARAT(locinput - 1) : '\n'; | |
63ac0dad KW |
3706 | switch (FLAGS(scan)) { |
3707 | case REGEX_UNICODE_CHARSET: | |
3708 | ln = isWORDCHAR_L1(ln); | |
3709 | n = isWORDCHAR_L1(nextchr); | |
3710 | break; | |
3711 | case REGEX_LOCALE_CHARSET: | |
3712 | ln = isALNUM_LC(ln); | |
3713 | n = isALNUM_LC(nextchr); | |
3714 | break; | |
3715 | case REGEX_DEPENDS_CHARSET: | |
3716 | ln = isALNUM(ln); | |
3717 | n = isALNUM(nextchr); | |
3718 | break; | |
cfaf538b KW |
3719 | case REGEX_ASCII_RESTRICTED_CHARSET: |
3720 | ln = isWORDCHAR_A(ln); | |
3721 | n = isWORDCHAR_A(nextchr); | |
3722 | break; | |
63ac0dad KW |
3723 | default: |
3724 | Perl_croak(aTHX_ "panic: Unexpected FLAGS %u in op %u", FLAGS(scan), OP(scan)); | |
3725 | break; | |
b2680017 YO |
3726 | } |
3727 | } | |
63ac0dad KW |
3728 | /* Note requires that all BOUNDs be lower than all NBOUNDs in |
3729 | * regcomp.sym */ | |
3730 | if (((!ln) == (!n)) == (OP(scan) < NBOUND)) | |
b2680017 YO |
3731 | sayNO; |
3732 | break; | |
f56b6394 | 3733 | case ANYOFV: |
d6a28714 | 3734 | case ANYOF: |
f56b6394 | 3735 | if (utf8_target || state_num == ANYOFV) { |
9e55ce06 | 3736 | STRLEN inclasslen = PL_regeol - locinput; |
20ed0b26 KW |
3737 | if (locinput >= PL_regeol) |
3738 | sayNO; | |
9e55ce06 | 3739 | |
f2ed9b32 | 3740 | if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target)) |
09b08e9b | 3741 | sayNO; |
b32d7d3e | 3742 | locinput += inclasslen; |
b8c5462f | 3743 | nextchr = UCHARAT(locinput); |
e0f9d4a8 | 3744 | break; |
ffc61ed2 JH |
3745 | } |
3746 | else { | |
3747 | if (nextchr < 0) | |
3748 | nextchr = UCHARAT(locinput); | |
ffc61ed2 JH |
3749 | if (!nextchr && locinput >= PL_regeol) |
3750 | sayNO; | |
20ed0b26 | 3751 | if (!REGINCLASS(rex, scan, (U8*)locinput)) |
09b08e9b | 3752 | sayNO; |
ffc61ed2 | 3753 | nextchr = UCHARAT(++locinput); |
e0f9d4a8 JH |
3754 | break; |
3755 | } | |
b8c5462f | 3756 | break; |
20d0b1e9 | 3757 | /* Special char classes - The defines start on line 129 or so */ |
ee9a90b8 KW |
3758 | CCC_TRY_U(ALNUM, NALNUM, isWORDCHAR, |
3759 | ALNUML, NALNUML, isALNUM_LC, isALNUM_LC_utf8, | |
3760 | ALNUMU, NALNUMU, isWORDCHAR_L1, | |
cfaf538b | 3761 | ALNUMA, NALNUMA, isWORDCHAR_A, |
779d7b58 | 3762 | alnum, "a"); |
ee9a90b8 KW |
3763 | |
3764 | CCC_TRY_U(SPACE, NSPACE, isSPACE, | |
3765 | SPACEL, NSPACEL, isSPACE_LC, isSPACE_LC_utf8, | |
3766 | SPACEU, NSPACEU, isSPACE_L1, | |
cfaf538b | 3767 | SPACEA, NSPACEA, isSPACE_A, |
779d7b58 | 3768 | space, " "); |
ee9a90b8 KW |
3769 | |
3770 | CCC_TRY(DIGIT, NDIGIT, isDIGIT, | |
3771 | DIGITL, NDIGITL, isDIGIT_LC, isDIGIT_LC_utf8, | |
cfaf538b | 3772 | DIGITA, NDIGITA, isDIGIT_A, |
779d7b58 | 3773 | digit, "0"); |
20d0b1e9 | 3774 | |
37e2e78e KW |
3775 | case CLUMP: /* Match \X: logical Unicode character. This is defined as |
3776 | a Unicode extended Grapheme Cluster */ | |
3777 | /* From http://www.unicode.org/reports/tr29 (5.2 version). An | |
3778 | extended Grapheme Cluster is: | |
3779 | ||
3780 | CR LF | |
3781 | | Prepend* Begin Extend* | |
3782 | | . | |
3783 | ||
3784 | Begin is (Hangul-syllable | ! Control) | |
3785 | Extend is (Grapheme_Extend | Spacing_Mark) | |
3786 | Control is [ GCB_Control CR LF ] | |
3787 | ||
3788 | The discussion below shows how the code for CLUMP is derived | |
3789 | from this regex. Note that most of these concepts are from | |
3790 | property values of the Grapheme Cluster Boundary (GCB) property. | |
3791 | No code point can have multiple property values for a given | |
3792 | property. Thus a code point in Prepend can't be in Control, but | |
3793 | it must be in !Control. This is why Control above includes | |
3794 | GCB_Control plus CR plus LF. The latter two are used in the GCB | |
3795 | property separately, and so can't be in GCB_Control, even though | |
3796 | they logically are controls. Control is not the same as gc=cc, | |
3797 | but includes format and other characters as well. | |
3798 | ||
3799 | The Unicode definition of Hangul-syllable is: | |
3800 | L+ | |
3801 | | (L* ( ( V | LV ) V* | LVT ) T*) | |
3802 | | T+ | |
3803 | ) | |
3804 | Each of these is a value for the GCB property, and hence must be | |
3805 | disjoint, so the order they are tested is immaterial, so the | |
3806 | above can safely be changed to | |
3807 | T+ | |
3808 | | L+ | |
3809 | | (L* ( LVT | ( V | LV ) V*) T*) | |
3810 | ||
3811 | The last two terms can be combined like this: | |
3812 | L* ( L | |
3813 | | (( LVT | ( V | LV ) V*) T*)) | |
3814 | ||
3815 | And refactored into this: | |
3816 | L* (L | LVT T* | V V* T* | LV V* T*) | |
3817 | ||
3818 | That means that if we have seen any L's at all we can quit | |
3819 | there, but if the next character is a LVT, a V or and LV we | |
3820 | should keep going. | |
3821 | ||
3822 | There is a subtlety with Prepend* which showed up in testing. | |
3823 | Note that the Begin, and only the Begin is required in: | |
3824 | | Prepend* Begin Extend* | |
3825 | Also, Begin contains '! Control'. A Prepend must be a '! | |
3826 | Control', which means it must be a Begin. What it comes down to | |
3827 | is that if we match Prepend* and then find no suitable Begin | |
3828 | afterwards, that if we backtrack the last Prepend, that one will | |
3829 | be a suitable Begin. | |
3830 | */ | |
3831 | ||
b7c83a7e | 3832 | if (locinput >= PL_regeol) |
a0ed51b3 | 3833 | sayNO; |
f2ed9b32 | 3834 | if (! utf8_target) { |
37e2e78e KW |
3835 | |
3836 | /* Match either CR LF or '.', as all the other possibilities | |
3837 | * require utf8 */ | |
3838 | locinput++; /* Match the . or CR */ | |
3839 | if (nextchr == '\r' | |
3840 | && locinput < PL_regeol | |
3841 | && UCHARAT(locinput) == '\n') locinput++; | |
3842 | } | |
3843 | else { | |
3844 | ||
3845 | /* Utf8: See if is ( CR LF ); already know that locinput < | |
3846 | * PL_regeol, so locinput+1 is in bounds */ | |
3847 | if (nextchr == '\r' && UCHARAT(locinput + 1) == '\n') { | |
3848 | locinput += 2; | |
3849 | } | |
3850 | else { | |
3851 | /* In case have to backtrack to beginning, then match '.' */ | |
3852 | char *starting = locinput; | |
3853 | ||
3854 | /* In case have to backtrack the last prepend */ | |
3855 | char *previous_prepend = 0; | |
3856 | ||
3857 | LOAD_UTF8_CHARCLASS_GCB(); | |
3858 | ||
3859 | /* Match (prepend)* */ | |
3860 | while (locinput < PL_regeol | |
3861 | && swash_fetch(PL_utf8_X_prepend, | |
f2ed9b32 | 3862 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3863 | { |
3864 | previous_prepend = locinput; | |
3865 | locinput += UTF8SKIP(locinput); | |
3866 | } | |
3867 | ||
3868 | /* As noted above, if we matched a prepend character, but | |
3869 | * the next thing won't match, back off the last prepend we | |
3870 | * matched, as it is guaranteed to match the begin */ | |
3871 | if (previous_prepend | |
3872 | && (locinput >= PL_regeol | |
3873 | || ! swash_fetch(PL_utf8_X_begin, | |
f2ed9b32 | 3874 | (U8*)locinput, utf8_target))) |
37e2e78e KW |
3875 | { |
3876 | locinput = previous_prepend; | |
3877 | } | |
3878 | ||
3879 | /* Note that here we know PL_regeol > locinput, as we | |
3880 | * tested that upon input to this switch case, and if we | |
3881 | * moved locinput forward, we tested the result just above | |
3882 | * and it either passed, or we backed off so that it will | |
3883 | * now pass */ | |
f2ed9b32 | 3884 | if (! swash_fetch(PL_utf8_X_begin, (U8*)locinput, utf8_target)) { |
37e2e78e KW |
3885 | |
3886 | /* Here did not match the required 'Begin' in the | |
3887 | * second term. So just match the very first | |
3888 | * character, the '.' of the final term of the regex */ | |
3889 | locinput = starting + UTF8SKIP(starting); | |
3890 | } else { | |
3891 | ||
3892 | /* Here is the beginning of a character that can have | |
3893 | * an extender. It is either a hangul syllable, or a | |
3894 | * non-control */ | |
3895 | if (swash_fetch(PL_utf8_X_non_hangul, | |
f2ed9b32 | 3896 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3897 | { |
3898 | ||
3899 | /* Here not a Hangul syllable, must be a | |
3900 | * ('! * Control') */ | |
3901 | locinput += UTF8SKIP(locinput); | |
3902 | } else { | |
3903 | ||
3904 | /* Here is a Hangul syllable. It can be composed | |
3905 | * of several individual characters. One | |
3906 | * possibility is T+ */ | |
3907 | if (swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 3908 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3909 | { |
3910 | while (locinput < PL_regeol | |
3911 | && swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 3912 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3913 | { |
3914 | locinput += UTF8SKIP(locinput); | |
3915 | } | |
3916 | } else { | |
3917 | ||
3918 | /* Here, not T+, but is a Hangul. That means | |
3919 | * it is one of the others: L, LV, LVT or V, | |
3920 | * and matches: | |
3921 | * L* (L | LVT T* | V V* T* | LV V* T*) */ | |
3922 | ||
3923 | /* Match L* */ | |
3924 | while (locinput < PL_regeol | |
3925 | && swash_fetch(PL_utf8_X_L, | |
f2ed9b32 | 3926 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3927 | { |
3928 | locinput += UTF8SKIP(locinput); | |
3929 | } | |
3930 | ||
3931 | /* Here, have exhausted L*. If the next | |
3932 | * character is not an LV, LVT nor V, it means | |
3933 | * we had to have at least one L, so matches L+ | |
3934 | * in the original equation, we have a complete | |
3935 | * hangul syllable. Are done. */ | |
3936 | ||
3937 | if (locinput < PL_regeol | |
3938 | && swash_fetch(PL_utf8_X_LV_LVT_V, | |
f2ed9b32 | 3939 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3940 | { |
3941 | ||
3942 | /* Otherwise keep going. Must be LV, LVT | |
3943 | * or V. See if LVT */ | |
3944 | if (swash_fetch(PL_utf8_X_LVT, | |
f2ed9b32 | 3945 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3946 | { |
3947 | locinput += UTF8SKIP(locinput); | |
3948 | } else { | |
3949 | ||
3950 | /* Must be V or LV. Take it, then | |
3951 | * match V* */ | |
3952 | locinput += UTF8SKIP(locinput); | |
3953 | while (locinput < PL_regeol | |
3954 | && swash_fetch(PL_utf8_X_V, | |
f2ed9b32 | 3955 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3956 | { |
3957 | locinput += UTF8SKIP(locinput); | |
3958 | } | |
3959 | } | |
3960 | ||
3961 | /* And any of LV, LVT, or V can be followed | |
3962 | * by T* */ | |
3963 | while (locinput < PL_regeol | |
3964 | && swash_fetch(PL_utf8_X_T, | |
3965 | (U8*)locinput, | |
f2ed9b32 | 3966 | utf8_target)) |
37e2e78e KW |
3967 | { |
3968 | locinput += UTF8SKIP(locinput); | |
3969 | } | |
3970 | } | |
3971 | } | |
3972 | } | |
3973 | ||
3974 | /* Match any extender */ | |
3975 | while (locinput < PL_regeol | |
3976 | && swash_fetch(PL_utf8_X_extend, | |
f2ed9b32 | 3977 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3978 | { |
3979 | locinput += UTF8SKIP(locinput); | |
3980 | } | |
3981 | } | |
3982 | } | |
3983 | if (locinput > PL_regeol) sayNO; | |
3984 | } | |
a0ed51b3 LW |
3985 | nextchr = UCHARAT(locinput); |
3986 | break; | |
81714fb9 YO |
3987 | |
3988 | case NREFFL: | |
d7ef4b73 KW |
3989 | { /* The capture buffer cases. The ones beginning with N for the |
3990 | named buffers just convert to the equivalent numbered and | |
3991 | pretend they were called as the corresponding numbered buffer | |
3992 | op. */ | |
26ecd678 TC |
3993 | /* don't initialize these in the declaration, it makes C++ |
3994 | unhappy */ | |
81714fb9 | 3995 | char *s; |
ff1157ca | 3996 | char type; |
8368298a TC |
3997 | re_fold_t folder; |
3998 | const U8 *fold_array; | |
26ecd678 | 3999 | UV utf8_fold_flags; |
8368298a | 4000 | |
81714fb9 | 4001 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4002 | folder = foldEQ_locale; |
4003 | fold_array = PL_fold_locale; | |
4004 | type = REFFL; | |
17580e7a | 4005 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4006 | goto do_nref; |
4007 | ||
2f7f8cb1 KW |
4008 | case NREFFA: |
4009 | folder = foldEQ_latin1; | |
4010 | fold_array = PL_fold_latin1; | |
4011 | type = REFFA; | |
4012 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4013 | goto do_nref; | |
4014 | ||
d7ef4b73 KW |
4015 | case NREFFU: |
4016 | folder = foldEQ_latin1; | |
4017 | fold_array = PL_fold_latin1; | |
4018 | type = REFFU; | |
d513472c | 4019 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4020 | goto do_nref; |
4021 | ||
81714fb9 | 4022 | case NREFF: |
d7ef4b73 KW |
4023 | folder = foldEQ; |
4024 | fold_array = PL_fold; | |
4025 | type = REFF; | |
d513472c | 4026 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4027 | goto do_nref; |
4028 | ||
4029 | case NREF: | |
4030 | type = REF; | |
83d7b90b KW |
4031 | folder = NULL; |
4032 | fold_array = NULL; | |
d513472c | 4033 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4034 | do_nref: |
4035 | ||
4036 | /* For the named back references, find the corresponding buffer | |
4037 | * number */ | |
0a4db386 YO |
4038 | n = reg_check_named_buff_matched(rex,scan); |
4039 | ||
d7ef4b73 | 4040 | if ( ! n ) { |
81714fb9 | 4041 | sayNO; |
d7ef4b73 KW |
4042 | } |
4043 | goto do_nref_ref_common; | |
4044 | ||
c8756f30 | 4045 | case REFFL: |
3280af22 | 4046 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4047 | folder = foldEQ_locale; |
4048 | fold_array = PL_fold_locale; | |
17580e7a | 4049 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4050 | goto do_ref; |
4051 | ||
2f7f8cb1 KW |
4052 | case REFFA: |
4053 | folder = foldEQ_latin1; | |
4054 | fold_array = PL_fold_latin1; | |
4055 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4056 | goto do_ref; | |
4057 | ||
d7ef4b73 KW |
4058 | case REFFU: |
4059 | folder = foldEQ_latin1; | |
4060 | fold_array = PL_fold_latin1; | |
d513472c | 4061 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4062 | goto do_ref; |
4063 | ||
4064 | case REFF: | |
4065 | folder = foldEQ; | |
4066 | fold_array = PL_fold; | |
d513472c | 4067 | utf8_fold_flags = 0; |
83d7b90b | 4068 | goto do_ref; |
d7ef4b73 | 4069 | |
c277df42 | 4070 | case REF: |
83d7b90b KW |
4071 | folder = NULL; |
4072 | fold_array = NULL; | |
d513472c | 4073 | utf8_fold_flags = 0; |
83d7b90b | 4074 | |
d7ef4b73 | 4075 | do_ref: |
81714fb9 | 4076 | type = OP(scan); |
d7ef4b73 KW |
4077 | n = ARG(scan); /* which paren pair */ |
4078 | ||
4079 | do_nref_ref_common: | |
f0ab9afb | 4080 | ln = PL_regoffs[n].start; |
2c2d71f5 | 4081 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
3b6647e0 | 4082 | if (*PL_reglastparen < n || ln == -1) |
af3f8c16 | 4083 | sayNO; /* Do not match unless seen CLOSEn. */ |
f0ab9afb | 4084 | if (ln == PL_regoffs[n].end) |
a0d0e21e | 4085 | break; |
a0ed51b3 | 4086 | |
24d3c4a9 | 4087 | s = PL_bostr + ln; |
d7ef4b73 | 4088 | if (type != REF /* REF can do byte comparison */ |
2f65c56d | 4089 | && (utf8_target || type == REFFU)) |
d7ef4b73 KW |
4090 | { /* XXX handle REFFL better */ |
4091 | char * limit = PL_regeol; | |
4092 | ||
4093 | /* This call case insensitively compares the entire buffer | |
4094 | * at s, with the current input starting at locinput, but | |
4095 | * not going off the end given by PL_regeol, and returns in | |
4096 | * limit upon success, how much of the current input was | |
4097 | * matched */ | |
d513472c KW |
4098 | if (! foldEQ_utf8_flags(s, NULL, PL_regoffs[n].end - ln, utf8_target, |
4099 | locinput, &limit, 0, utf8_target, utf8_fold_flags)) | |
d7ef4b73 KW |
4100 | { |
4101 | sayNO; | |
a0ed51b3 | 4102 | } |
d7ef4b73 | 4103 | locinput = limit; |
a0ed51b3 LW |
4104 | nextchr = UCHARAT(locinput); |
4105 | break; | |
4106 | } | |
4107 | ||
d7ef4b73 | 4108 | /* Not utf8: Inline the first character, for speed. */ |
76e3520e | 4109 | if (UCHARAT(s) != nextchr && |
81714fb9 | 4110 | (type == REF || |
d7ef4b73 | 4111 | UCHARAT(s) != fold_array[nextchr])) |
4633a7c4 | 4112 | sayNO; |
f0ab9afb | 4113 | ln = PL_regoffs[n].end - ln; |
24d3c4a9 | 4114 | if (locinput + ln > PL_regeol) |
4633a7c4 | 4115 | sayNO; |
81714fb9 | 4116 | if (ln > 1 && (type == REF |
24d3c4a9 | 4117 | ? memNE(s, locinput, ln) |
d7ef4b73 | 4118 | : ! folder(s, locinput, ln))) |
4633a7c4 | 4119 | sayNO; |
24d3c4a9 | 4120 | locinput += ln; |
76e3520e | 4121 | nextchr = UCHARAT(locinput); |
a0d0e21e | 4122 | break; |
81714fb9 | 4123 | } |
a0d0e21e | 4124 | case NOTHING: |
c277df42 | 4125 | case TAIL: |
a0d0e21e LW |
4126 | break; |
4127 | case BACK: | |
4128 | break; | |
40a82448 DM |
4129 | |
4130 | #undef ST | |
4131 | #define ST st->u.eval | |
c277df42 | 4132 | { |
c277df42 | 4133 | SV *ret; |
d2f13c59 | 4134 | REGEXP *re_sv; |
6bda09f9 | 4135 | regexp *re; |
f8fc2ecf | 4136 | regexp_internal *rei; |
1a147d38 YO |
4137 | regnode *startpoint; |
4138 | ||
4139 | case GOSTART: | |
e7707071 YO |
4140 | case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */ |
4141 | if (cur_eval && cur_eval->locinput==locinput) { | |
24b23f37 | 4142 | if (cur_eval->u.eval.close_paren == (U32)ARG(scan)) |
1a147d38 | 4143 | Perl_croak(aTHX_ "Infinite recursion in regex"); |
4b196cd4 | 4144 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 YO |
4145 | Perl_croak(aTHX_ |
4146 | "Pattern subroutine nesting without pos change" | |
4147 | " exceeded limit in regex"); | |
6bda09f9 YO |
4148 | } else { |
4149 | nochange_depth = 0; | |
1a147d38 | 4150 | } |
288b8c02 | 4151 | re_sv = rex_sv; |
6bda09f9 | 4152 | re = rex; |
f8fc2ecf | 4153 | rei = rexi; |
288b8c02 | 4154 | (void)ReREFCNT_inc(rex_sv); |
1a147d38 | 4155 | if (OP(scan)==GOSUB) { |
6bda09f9 YO |
4156 | startpoint = scan + ARG2L(scan); |
4157 | ST.close_paren = ARG(scan); | |
4158 | } else { | |
f8fc2ecf | 4159 | startpoint = rei->program+1; |
6bda09f9 YO |
4160 | ST.close_paren = 0; |
4161 | } | |
4162 | goto eval_recurse_doit; | |
4163 | /* NOTREACHED */ | |
4164 | case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ | |
4165 | if (cur_eval && cur_eval->locinput==locinput) { | |
4b196cd4 | 4166 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 | 4167 | Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex"); |
6bda09f9 YO |
4168 | } else { |
4169 | nochange_depth = 0; | |
4170 | } | |
8e5e9ebe | 4171 | { |
4aabdb9b DM |
4172 | /* execute the code in the {...} */ |
4173 | dSP; | |
6136c704 | 4174 | SV ** const before = SP; |
4aabdb9b DM |
4175 | OP_4tree * const oop = PL_op; |
4176 | COP * const ocurcop = PL_curcop; | |
4177 | PAD *old_comppad; | |
d80618d2 | 4178 | char *saved_regeol = PL_regeol; |
91332126 FR |
4179 | struct re_save_state saved_state; |
4180 | ||
6562f1c4 | 4181 | /* To not corrupt the existing regex state while executing the |
b7f4cd04 FR |
4182 | * eval we would normally put it on the save stack, like with |
4183 | * save_re_context. However, re-evals have a weird scoping so we | |
4184 | * can't just add ENTER/LEAVE here. With that, things like | |
4185 | * | |
4186 | * (?{$a=2})(a(?{local$a=$a+1}))*aak*c(?{$b=$a}) | |
4187 | * | |
4188 | * would break, as they expect the localisation to be unwound | |
4189 | * only when the re-engine backtracks through the bit that | |
4190 | * localised it. | |
4191 | * | |
4192 | * What we do instead is just saving the state in a local c | |
4193 | * variable. | |
4194 | */ | |
91332126 FR |
4195 | Copy(&PL_reg_state, &saved_state, 1, struct re_save_state); |
4196 | ||
4aabdb9b | 4197 | n = ARG(scan); |
f8fc2ecf | 4198 | PL_op = (OP_4tree*)rexi->data->data[n]; |
24b23f37 YO |
4199 | DEBUG_STATE_r( PerlIO_printf(Perl_debug_log, |
4200 | " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); | |
f8fc2ecf | 4201 | PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]); |
f0ab9afb | 4202 | PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr; |
4aabdb9b | 4203 | |
2bf803e2 YO |
4204 | if (sv_yes_mark) { |
4205 | SV *sv_mrk = get_sv("REGMARK", 1); | |
4206 | sv_setsv(sv_mrk, sv_yes_mark); | |
4207 | } | |
4208 | ||
8e5e9ebe RGS |
4209 | CALLRUNOPS(aTHX); /* Scalar context. */ |
4210 | SPAGAIN; | |
4211 | if (SP == before) | |
075aa684 | 4212 | ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */ |
8e5e9ebe RGS |
4213 | else { |
4214 | ret = POPs; | |
4215 | PUTBACK; | |
4216 | } | |
4aabdb9b | 4217 | |
91332126 FR |
4218 | Copy(&saved_state, &PL_reg_state, 1, struct re_save_state); |
4219 | ||
4aabdb9b DM |
4220 | PL_op = oop; |
4221 | PAD_RESTORE_LOCAL(old_comppad); | |
4222 | PL_curcop = ocurcop; | |
d80618d2 | 4223 | PL_regeol = saved_regeol; |
24d3c4a9 | 4224 | if (!logical) { |
4aabdb9b DM |
4225 | /* /(?{...})/ */ |
4226 | sv_setsv(save_scalar(PL_replgv), ret); | |
4aabdb9b DM |
4227 | break; |
4228 | } | |
8e5e9ebe | 4229 | } |
24d3c4a9 DM |
4230 | if (logical == 2) { /* Postponed subexpression: /(??{...})/ */ |
4231 | logical = 0; | |
4aabdb9b | 4232 | { |
4f639d21 DM |
4233 | /* extract RE object from returned value; compiling if |
4234 | * necessary */ | |
6136c704 | 4235 | MAGIC *mg = NULL; |
288b8c02 | 4236 | REGEXP *rx = NULL; |
5c35adbb NC |
4237 | |
4238 | if (SvROK(ret)) { | |
288b8c02 | 4239 | SV *const sv = SvRV(ret); |
5c35adbb NC |
4240 | |
4241 | if (SvTYPE(sv) == SVt_REGEXP) { | |
d2f13c59 | 4242 | rx = (REGEXP*) sv; |
5c35adbb NC |
4243 | } else if (SvSMAGICAL(sv)) { |
4244 | mg = mg_find(sv, PERL_MAGIC_qr); | |
4245 | assert(mg); | |
4246 | } | |
4247 | } else if (SvTYPE(ret) == SVt_REGEXP) { | |
d2f13c59 | 4248 | rx = (REGEXP*) ret; |
5c35adbb | 4249 | } else if (SvSMAGICAL(ret)) { |
124ee91a NC |
4250 | if (SvGMAGICAL(ret)) { |
4251 | /* I don't believe that there is ever qr magic | |
4252 | here. */ | |
4253 | assert(!mg_find(ret, PERL_MAGIC_qr)); | |
faf82a0b | 4254 | sv_unmagic(ret, PERL_MAGIC_qr); |
124ee91a NC |
4255 | } |
4256 | else { | |
faf82a0b | 4257 | mg = mg_find(ret, PERL_MAGIC_qr); |
124ee91a NC |
4258 | /* testing suggests mg only ends up non-NULL for |
4259 | scalars who were upgraded and compiled in the | |
4260 | else block below. In turn, this is only | |
4261 | triggered in the "postponed utf8 string" tests | |
4262 | in t/op/pat.t */ | |
4263 | } | |
0f5d15d6 | 4264 | } |
faf82a0b | 4265 | |
0f5d15d6 | 4266 | if (mg) { |
d2f13c59 | 4267 | rx = (REGEXP *) mg->mg_obj; /*XXX:dmq*/ |
e2560c33 | 4268 | assert(rx); |
0f5d15d6 | 4269 | } |
288b8c02 | 4270 | if (rx) { |
f0826785 | 4271 | rx = reg_temp_copy(NULL, rx); |
288b8c02 | 4272 | } |
0f5d15d6 | 4273 | else { |
c737faaf | 4274 | U32 pm_flags = 0; |
a3b680e6 | 4275 | const I32 osize = PL_regsize; |
0f5d15d6 | 4276 | |
b9ad30b4 NC |
4277 | if (DO_UTF8(ret)) { |
4278 | assert (SvUTF8(ret)); | |
4279 | } else if (SvUTF8(ret)) { | |
4280 | /* Not doing UTF-8, despite what the SV says. Is | |
4281 | this only if we're trapped in use 'bytes'? */ | |
4282 | /* Make a copy of the octet sequence, but without | |
4283 | the flag on, as the compiler now honours the | |
4284 | SvUTF8 flag on ret. */ | |
4285 | STRLEN len; | |
4286 | const char *const p = SvPV(ret, len); | |
4287 | ret = newSVpvn_flags(p, len, SVs_TEMP); | |
4288 | } | |
288b8c02 | 4289 | rx = CALLREGCOMP(ret, pm_flags); |
9041c2e3 | 4290 | if (!(SvFLAGS(ret) |
faf82a0b | 4291 | & (SVs_TEMP | SVs_PADTMP | SVf_READONLY |
3ce3ed55 | 4292 | | SVs_GMG))) { |
a2794585 NC |
4293 | /* This isn't a first class regexp. Instead, it's |
4294 | caching a regexp onto an existing, Perl visible | |
4295 | scalar. */ | |
ad64d0ec | 4296 | sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0); |
3ce3ed55 | 4297 | } |
0f5d15d6 | 4298 | PL_regsize = osize; |
0f5d15d6 | 4299 | } |
288b8c02 NC |
4300 | re_sv = rx; |
4301 | re = (struct regexp *)SvANY(rx); | |
4aabdb9b | 4302 | } |
07bc277f | 4303 | RXp_MATCH_COPIED_off(re); |
28d8d7f4 YO |
4304 | re->subbeg = rex->subbeg; |
4305 | re->sublen = rex->sublen; | |
f8fc2ecf | 4306 | rei = RXi_GET(re); |
6bda09f9 | 4307 | DEBUG_EXECUTE_r( |
f2ed9b32 | 4308 | debug_start_match(re_sv, utf8_target, locinput, PL_regeol, |
6bda09f9 YO |
4309 | "Matching embedded"); |
4310 | ); | |
f8fc2ecf | 4311 | startpoint = rei->program + 1; |
1a147d38 | 4312 | ST.close_paren = 0; /* only used for GOSUB */ |
6bda09f9 YO |
4313 | /* borrowed from regtry */ |
4314 | if (PL_reg_start_tmpl <= re->nparens) { | |
4315 | PL_reg_start_tmpl = re->nparens*3/2 + 3; | |
4316 | if(PL_reg_start_tmp) | |
4317 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
4318 | else | |
4319 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
dd5def09 | 4320 | } |
aa283a38 | 4321 | |
1a147d38 | 4322 | eval_recurse_doit: /* Share code with GOSUB below this line */ |
aa283a38 | 4323 | /* run the pattern returned from (??{...}) */ |
40a82448 DM |
4324 | ST.cp = regcppush(0); /* Save *all* the positions. */ |
4325 | REGCP_SET(ST.lastcp); | |
6bda09f9 | 4326 | |
f0ab9afb | 4327 | PL_regoffs = re->offs; /* essentially NOOP on GOSUB */ |
6bda09f9 | 4328 | |
0357f1fd ML |
4329 | /* see regtry, specifically PL_reglast(?:close)?paren is a pointer! (i dont know why) :dmq */ |
4330 | PL_reglastparen = &re->lastparen; | |
4331 | PL_reglastcloseparen = &re->lastcloseparen; | |
4332 | re->lastparen = 0; | |
4333 | re->lastcloseparen = 0; | |
4334 | ||
4aabdb9b | 4335 | PL_reginput = locinput; |
ae0beba1 | 4336 | PL_regsize = 0; |
4aabdb9b DM |
4337 | |
4338 | /* XXXX This is too dramatic a measure... */ | |
4339 | PL_reg_maxiter = 0; | |
4340 | ||
faec1544 | 4341 | ST.toggle_reg_flags = PL_reg_flags; |
3c8556c3 | 4342 | if (RX_UTF8(re_sv)) |
faec1544 DM |
4343 | PL_reg_flags |= RF_utf8; |
4344 | else | |
4345 | PL_reg_flags &= ~RF_utf8; | |
4346 | ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */ | |
4347 | ||
288b8c02 | 4348 | ST.prev_rex = rex_sv; |
faec1544 | 4349 | ST.prev_curlyx = cur_curlyx; |
288b8c02 NC |
4350 | SETREX(rex_sv,re_sv); |
4351 | rex = re; | |
f8fc2ecf | 4352 | rexi = rei; |
faec1544 | 4353 | cur_curlyx = NULL; |
40a82448 | 4354 | ST.B = next; |
faec1544 DM |
4355 | ST.prev_eval = cur_eval; |
4356 | cur_eval = st; | |
faec1544 | 4357 | /* now continue from first node in postoned RE */ |
6bda09f9 | 4358 | PUSH_YES_STATE_GOTO(EVAL_AB, startpoint); |
4aabdb9b | 4359 | /* NOTREACHED */ |
a0ed51b3 | 4360 | } |
24d3c4a9 | 4361 | /* logical is 1, /(?(?{...})X|Y)/ */ |
f2338a2e | 4362 | sw = cBOOL(SvTRUE(ret)); |
24d3c4a9 | 4363 | logical = 0; |
c277df42 IZ |
4364 | break; |
4365 | } | |
40a82448 | 4366 | |
faec1544 DM |
4367 | case EVAL_AB: /* cleanup after a successful (??{A})B */ |
4368 | /* note: this is called twice; first after popping B, then A */ | |
4369 | PL_reg_flags ^= ST.toggle_reg_flags; | |
288b8c02 NC |
4370 | ReREFCNT_dec(rex_sv); |
4371 | SETREX(rex_sv,ST.prev_rex); | |
4372 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 4373 | rexi = RXi_GET(rex); |
faec1544 DM |
4374 | regcpblow(ST.cp); |
4375 | cur_eval = ST.prev_eval; | |
4376 | cur_curlyx = ST.prev_curlyx; | |
34a81e2b B |
4377 | |
4378 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ | |
0357f1fd ML |
4379 | PL_reglastparen = &rex->lastparen; |
4380 | PL_reglastcloseparen = &rex->lastcloseparen; | |
34a81e2b B |
4381 | /* also update PL_regoffs */ |
4382 | PL_regoffs = rex->offs; | |
0357f1fd | 4383 | |
40a82448 DM |
4384 | /* XXXX This is too dramatic a measure... */ |
4385 | PL_reg_maxiter = 0; | |
e7707071 | 4386 | if ( nochange_depth ) |
4b196cd4 | 4387 | nochange_depth--; |
262b90c4 | 4388 | sayYES; |
40a82448 | 4389 | |
40a82448 | 4390 | |
faec1544 DM |
4391 | case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */ |
4392 | /* note: this is called twice; first after popping B, then A */ | |
4393 | PL_reg_flags ^= ST.toggle_reg_flags; | |
288b8c02 NC |
4394 | ReREFCNT_dec(rex_sv); |
4395 | SETREX(rex_sv,ST.prev_rex); | |
4396 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 4397 | rexi = RXi_GET(rex); |
34a81e2b | 4398 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ |
0357f1fd ML |
4399 | PL_reglastparen = &rex->lastparen; |
4400 | PL_reglastcloseparen = &rex->lastcloseparen; | |
4401 | ||
40a82448 DM |
4402 | PL_reginput = locinput; |
4403 | REGCP_UNWIND(ST.lastcp); | |
4404 | regcppop(rex); | |
faec1544 DM |
4405 | cur_eval = ST.prev_eval; |
4406 | cur_curlyx = ST.prev_curlyx; | |
4407 | /* XXXX This is too dramatic a measure... */ | |
4408 | PL_reg_maxiter = 0; | |
e7707071 | 4409 | if ( nochange_depth ) |
4b196cd4 | 4410 | nochange_depth--; |
40a82448 | 4411 | sayNO_SILENT; |
40a82448 DM |
4412 | #undef ST |
4413 | ||
a0d0e21e | 4414 | case OPEN: |
c277df42 | 4415 | n = ARG(scan); /* which paren pair */ |
3280af22 NIS |
4416 | PL_reg_start_tmp[n] = locinput; |
4417 | if (n > PL_regsize) | |
4418 | PL_regsize = n; | |
e2e6a0f1 | 4419 | lastopen = n; |
a0d0e21e LW |
4420 | break; |
4421 | case CLOSE: | |
c277df42 | 4422 | n = ARG(scan); /* which paren pair */ |
f0ab9afb NC |
4423 | PL_regoffs[n].start = PL_reg_start_tmp[n] - PL_bostr; |
4424 | PL_regoffs[n].end = locinput - PL_bostr; | |
7f69552c YO |
4425 | /*if (n > PL_regsize) |
4426 | PL_regsize = n;*/ | |
3b6647e0 | 4427 | if (n > *PL_reglastparen) |
3280af22 | 4428 | *PL_reglastparen = n; |
a01268b5 | 4429 | *PL_reglastcloseparen = n; |
3b6647e0 | 4430 | if (cur_eval && cur_eval->u.eval.close_paren == n) { |
6bda09f9 YO |
4431 | goto fake_end; |
4432 | } | |
a0d0e21e | 4433 | break; |
e2e6a0f1 YO |
4434 | case ACCEPT: |
4435 | if (ARG(scan)){ | |
4436 | regnode *cursor; | |
4437 | for (cursor=scan; | |
4438 | cursor && OP(cursor)!=END; | |
4439 | cursor=regnext(cursor)) | |
4440 | { | |
4441 | if ( OP(cursor)==CLOSE ){ | |
4442 | n = ARG(cursor); | |
4443 | if ( n <= lastopen ) { | |
f0ab9afb NC |
4444 | PL_regoffs[n].start |
4445 | = PL_reg_start_tmp[n] - PL_bostr; | |
4446 | PL_regoffs[n].end = locinput - PL_bostr; | |
e2e6a0f1 YO |
4447 | /*if (n > PL_regsize) |
4448 | PL_regsize = n;*/ | |
3b6647e0 | 4449 | if (n > *PL_reglastparen) |
e2e6a0f1 YO |
4450 | *PL_reglastparen = n; |
4451 | *PL_reglastcloseparen = n; | |
3b6647e0 RB |
4452 | if ( n == ARG(scan) || (cur_eval && |
4453 | cur_eval->u.eval.close_paren == n)) | |
e2e6a0f1 YO |
4454 | break; |
4455 | } | |
4456 | } | |
4457 | } | |
4458 | } | |
4459 | goto fake_end; | |
4460 | /*NOTREACHED*/ | |
c277df42 IZ |
4461 | case GROUPP: |
4462 | n = ARG(scan); /* which paren pair */ | |
f2338a2e | 4463 | sw = cBOOL(*PL_reglastparen >= n && PL_regoffs[n].end != -1); |
c277df42 | 4464 | break; |
0a4db386 YO |
4465 | case NGROUPP: |
4466 | /* reg_check_named_buff_matched returns 0 for no match */ | |
f2338a2e | 4467 | sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan)); |
0a4db386 | 4468 | break; |
1a147d38 | 4469 | case INSUBP: |
0a4db386 | 4470 | n = ARG(scan); |
3b6647e0 | 4471 | sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n)); |
0a4db386 YO |
4472 | break; |
4473 | case DEFINEP: | |
4474 | sw = 0; | |
4475 | break; | |
c277df42 | 4476 | case IFTHEN: |
2c2d71f5 | 4477 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
24d3c4a9 | 4478 | if (sw) |
c277df42 IZ |
4479 | next = NEXTOPER(NEXTOPER(scan)); |
4480 | else { | |
4481 | next = scan + ARG(scan); | |
4482 | if (OP(next) == IFTHEN) /* Fake one. */ | |
4483 | next = NEXTOPER(NEXTOPER(next)); | |
4484 | } | |
4485 | break; | |
4486 | case LOGICAL: | |
24d3c4a9 | 4487 | logical = scan->flags; |
c277df42 | 4488 | break; |
c476f425 | 4489 | |
2ab05381 | 4490 | /******************************************************************* |
2ab05381 | 4491 | |
c476f425 DM |
4492 | The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/ |
4493 | pattern, where A and B are subpatterns. (For simple A, CURLYM or | |
4494 | STAR/PLUS/CURLY/CURLYN are used instead.) | |
2ab05381 | 4495 | |
c476f425 | 4496 | A*B is compiled as <CURLYX><A><WHILEM><B> |
2ab05381 | 4497 | |
c476f425 DM |
4498 | On entry to the subpattern, CURLYX is called. This pushes a CURLYX |
4499 | state, which contains the current count, initialised to -1. It also sets | |
4500 | cur_curlyx to point to this state, with any previous value saved in the | |
4501 | state block. | |
2ab05381 | 4502 | |
c476f425 DM |
4503 | CURLYX then jumps straight to the WHILEM op, rather than executing A, |
4504 | since the pattern may possibly match zero times (i.e. it's a while {} loop | |
4505 | rather than a do {} while loop). | |
2ab05381 | 4506 | |
c476f425 DM |
4507 | Each entry to WHILEM represents a successful match of A. The count in the |
4508 | CURLYX block is incremented, another WHILEM state is pushed, and execution | |
4509 | passes to A or B depending on greediness and the current count. | |
2ab05381 | 4510 | |
c476f425 DM |
4511 | For example, if matching against the string a1a2a3b (where the aN are |
4512 | substrings that match /A/), then the match progresses as follows: (the | |
4513 | pushed states are interspersed with the bits of strings matched so far): | |
2ab05381 | 4514 | |
c476f425 DM |
4515 | <CURLYX cnt=-1> |
4516 | <CURLYX cnt=0><WHILEM> | |
4517 | <CURLYX cnt=1><WHILEM> a1 <WHILEM> | |
4518 | <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM> | |
4519 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> | |
4520 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b | |
2ab05381 | 4521 | |
c476f425 DM |
4522 | (Contrast this with something like CURLYM, which maintains only a single |
4523 | backtrack state: | |
2ab05381 | 4524 | |
c476f425 DM |
4525 | <CURLYM cnt=0> a1 |
4526 | a1 <CURLYM cnt=1> a2 | |
4527 | a1 a2 <CURLYM cnt=2> a3 | |
4528 | a1 a2 a3 <CURLYM cnt=3> b | |
4529 | ) | |
2ab05381 | 4530 | |
c476f425 DM |
4531 | Each WHILEM state block marks a point to backtrack to upon partial failure |
4532 | of A or B, and also contains some minor state data related to that | |
4533 | iteration. The CURLYX block, pointed to by cur_curlyx, contains the | |
4534 | overall state, such as the count, and pointers to the A and B ops. | |
2ab05381 | 4535 | |
c476f425 DM |
4536 | This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx |
4537 | must always point to the *current* CURLYX block, the rules are: | |
2ab05381 | 4538 | |
c476f425 DM |
4539 | When executing CURLYX, save the old cur_curlyx in the CURLYX state block, |
4540 | and set cur_curlyx to point the new block. | |
2ab05381 | 4541 | |
c476f425 DM |
4542 | When popping the CURLYX block after a successful or unsuccessful match, |
4543 | restore the previous cur_curlyx. | |
2ab05381 | 4544 | |
c476f425 DM |
4545 | When WHILEM is about to execute B, save the current cur_curlyx, and set it |
4546 | to the outer one saved in the CURLYX block. | |
2ab05381 | 4547 | |
c476f425 DM |
4548 | When popping the WHILEM block after a successful or unsuccessful B match, |
4549 | restore the previous cur_curlyx. | |
2ab05381 | 4550 | |
c476f425 DM |
4551 | Here's an example for the pattern (AI* BI)*BO |
4552 | I and O refer to inner and outer, C and W refer to CURLYX and WHILEM: | |
2ab05381 | 4553 | |
c476f425 DM |
4554 | cur_ |
4555 | curlyx backtrack stack | |
4556 | ------ --------------- | |
4557 | NULL | |
4558 | CO <CO prev=NULL> <WO> | |
4559 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4560 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi | |
4561 | NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo | |
2ab05381 | 4562 | |
c476f425 DM |
4563 | At this point the pattern succeeds, and we work back down the stack to |
4564 | clean up, restoring as we go: | |
95b24440 | 4565 | |
c476f425 DM |
4566 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi |
4567 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4568 | CO <CO prev=NULL> <WO> | |
4569 | NULL | |
a0374537 | 4570 | |
c476f425 DM |
4571 | *******************************************************************/ |
4572 | ||
4573 | #define ST st->u.curlyx | |
4574 | ||
4575 | case CURLYX: /* start of /A*B/ (for complex A) */ | |
4576 | { | |
4577 | /* No need to save/restore up to this paren */ | |
4578 | I32 parenfloor = scan->flags; | |
4579 | ||
4580 | assert(next); /* keep Coverity happy */ | |
4581 | if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ | |
4582 | next += ARG(next); | |
4583 | ||
4584 | /* XXXX Probably it is better to teach regpush to support | |
4585 | parenfloor > PL_regsize... */ | |
4586 | if (parenfloor > (I32)*PL_reglastparen) | |
4587 | parenfloor = *PL_reglastparen; /* Pessimization... */ | |
4588 | ||
4589 | ST.prev_curlyx= cur_curlyx; | |
4590 | cur_curlyx = st; | |
4591 | ST.cp = PL_savestack_ix; | |
4592 | ||
4593 | /* these fields contain the state of the current curly. | |
4594 | * they are accessed by subsequent WHILEMs */ | |
4595 | ST.parenfloor = parenfloor; | |
d02d6d97 | 4596 | ST.me = scan; |
c476f425 | 4597 | ST.B = next; |
24d3c4a9 DM |
4598 | ST.minmod = minmod; |
4599 | minmod = 0; | |
c476f425 DM |
4600 | ST.count = -1; /* this will be updated by WHILEM */ |
4601 | ST.lastloc = NULL; /* this will be updated by WHILEM */ | |
4602 | ||
4603 | PL_reginput = locinput; | |
4604 | PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next)); | |
5f66b61c | 4605 | /* NOTREACHED */ |
c476f425 | 4606 | } |
a0d0e21e | 4607 | |
c476f425 | 4608 | case CURLYX_end: /* just finished matching all of A*B */ |
c476f425 DM |
4609 | cur_curlyx = ST.prev_curlyx; |
4610 | sayYES; | |
4611 | /* NOTREACHED */ | |
a0d0e21e | 4612 | |
c476f425 DM |
4613 | case CURLYX_end_fail: /* just failed to match all of A*B */ |
4614 | regcpblow(ST.cp); | |
4615 | cur_curlyx = ST.prev_curlyx; | |
4616 | sayNO; | |
4617 | /* NOTREACHED */ | |
4633a7c4 | 4618 | |
a0d0e21e | 4619 | |
c476f425 DM |
4620 | #undef ST |
4621 | #define ST st->u.whilem | |
4622 | ||
4623 | case WHILEM: /* just matched an A in /A*B/ (for complex A) */ | |
4624 | { | |
4625 | /* see the discussion above about CURLYX/WHILEM */ | |
c476f425 | 4626 | I32 n; |
d02d6d97 DM |
4627 | int min = ARG1(cur_curlyx->u.curlyx.me); |
4628 | int max = ARG2(cur_curlyx->u.curlyx.me); | |
4629 | regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS; | |
4630 | ||
c476f425 DM |
4631 | assert(cur_curlyx); /* keep Coverity happy */ |
4632 | n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */ | |
4633 | ST.save_lastloc = cur_curlyx->u.curlyx.lastloc; | |
4634 | ST.cache_offset = 0; | |
4635 | ST.cache_mask = 0; | |
4636 | ||
4637 | PL_reginput = locinput; | |
4638 | ||
4639 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
d02d6d97 DM |
4640 | "%*s whilem: matched %ld out of %d..%d\n", |
4641 | REPORT_CODE_OFF+depth*2, "", (long)n, min, max) | |
c476f425 | 4642 | ); |
a0d0e21e | 4643 | |
c476f425 | 4644 | /* First just match a string of min A's. */ |
a0d0e21e | 4645 | |
d02d6d97 | 4646 | if (n < min) { |
92e82afa | 4647 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); |
c476f425 | 4648 | cur_curlyx->u.curlyx.lastloc = locinput; |
92e82afa YO |
4649 | REGCP_SET(ST.lastcp); |
4650 | ||
d02d6d97 | 4651 | PUSH_STATE_GOTO(WHILEM_A_pre, A); |
c476f425 DM |
4652 | /* NOTREACHED */ |
4653 | } | |
4654 | ||
4655 | /* If degenerate A matches "", assume A done. */ | |
4656 | ||
4657 | if (locinput == cur_curlyx->u.curlyx.lastloc) { | |
4658 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4659 | "%*s whilem: empty match detected, trying continuation...\n", | |
4660 | REPORT_CODE_OFF+depth*2, "") | |
4661 | ); | |
4662 | goto do_whilem_B_max; | |
4663 | } | |
4664 | ||
4665 | /* super-linear cache processing */ | |
4666 | ||
4667 | if (scan->flags) { | |
a0d0e21e | 4668 | |
2c2d71f5 | 4669 | if (!PL_reg_maxiter) { |
c476f425 DM |
4670 | /* start the countdown: Postpone detection until we |
4671 | * know the match is not *that* much linear. */ | |
2c2d71f5 | 4672 | PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4); |
66bf836d DM |
4673 | /* possible overflow for long strings and many CURLYX's */ |
4674 | if (PL_reg_maxiter < 0) | |
4675 | PL_reg_maxiter = I32_MAX; | |
2c2d71f5 JH |
4676 | PL_reg_leftiter = PL_reg_maxiter; |
4677 | } | |
c476f425 | 4678 | |
2c2d71f5 | 4679 | if (PL_reg_leftiter-- == 0) { |
c476f425 | 4680 | /* initialise cache */ |
3298f257 | 4681 | const I32 size = (PL_reg_maxiter + 7)/8; |
2c2d71f5 | 4682 | if (PL_reg_poscache) { |
eb160463 | 4683 | if ((I32)PL_reg_poscache_size < size) { |
2c2d71f5 JH |
4684 | Renew(PL_reg_poscache, size, char); |
4685 | PL_reg_poscache_size = size; | |
4686 | } | |
4687 | Zero(PL_reg_poscache, size, char); | |
4688 | } | |
4689 | else { | |
4690 | PL_reg_poscache_size = size; | |
a02a5408 | 4691 | Newxz(PL_reg_poscache, size, char); |
2c2d71f5 | 4692 | } |
c476f425 DM |
4693 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
4694 | "%swhilem: Detected a super-linear match, switching on caching%s...\n", | |
4695 | PL_colors[4], PL_colors[5]) | |
4696 | ); | |
2c2d71f5 | 4697 | } |
c476f425 | 4698 | |
2c2d71f5 | 4699 | if (PL_reg_leftiter < 0) { |
c476f425 DM |
4700 | /* have we already failed at this position? */ |
4701 | I32 offset, mask; | |
4702 | offset = (scan->flags & 0xf) - 1 | |
4703 | + (locinput - PL_bostr) * (scan->flags>>4); | |
4704 | mask = 1 << (offset % 8); | |
4705 | offset /= 8; | |
4706 | if (PL_reg_poscache[offset] & mask) { | |
4707 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4708 | "%*s whilem: (cache) already tried at this position...\n", | |
4709 | REPORT_CODE_OFF+depth*2, "") | |
2c2d71f5 | 4710 | ); |
3298f257 | 4711 | sayNO; /* cache records failure */ |
2c2d71f5 | 4712 | } |
c476f425 DM |
4713 | ST.cache_offset = offset; |
4714 | ST.cache_mask = mask; | |
2c2d71f5 | 4715 | } |
c476f425 | 4716 | } |
2c2d71f5 | 4717 | |
c476f425 | 4718 | /* Prefer B over A for minimal matching. */ |
a687059c | 4719 | |
c476f425 DM |
4720 | if (cur_curlyx->u.curlyx.minmod) { |
4721 | ST.save_curlyx = cur_curlyx; | |
4722 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4723 | ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor); | |
4724 | REGCP_SET(ST.lastcp); | |
4725 | PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B); | |
4726 | /* NOTREACHED */ | |
4727 | } | |
a0d0e21e | 4728 | |
c476f425 DM |
4729 | /* Prefer A over B for maximal matching. */ |
4730 | ||
d02d6d97 | 4731 | if (n < max) { /* More greed allowed? */ |
c476f425 DM |
4732 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); |
4733 | cur_curlyx->u.curlyx.lastloc = locinput; | |
4734 | REGCP_SET(ST.lastcp); | |
d02d6d97 | 4735 | PUSH_STATE_GOTO(WHILEM_A_max, A); |
c476f425 DM |
4736 | /* NOTREACHED */ |
4737 | } | |
4738 | goto do_whilem_B_max; | |
4739 | } | |
4740 | /* NOTREACHED */ | |
4741 | ||
4742 | case WHILEM_B_min: /* just matched B in a minimal match */ | |
4743 | case WHILEM_B_max: /* just matched B in a maximal match */ | |
4744 | cur_curlyx = ST.save_curlyx; | |
4745 | sayYES; | |
4746 | /* NOTREACHED */ | |
4747 | ||
4748 | case WHILEM_B_max_fail: /* just failed to match B in a maximal match */ | |
4749 | cur_curlyx = ST.save_curlyx; | |
4750 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
4751 | cur_curlyx->u.curlyx.count--; | |
4752 | CACHEsayNO; | |
4753 | /* NOTREACHED */ | |
4754 | ||
4755 | case WHILEM_A_min_fail: /* just failed to match A in a minimal match */ | |
c476f425 DM |
4756 | /* FALL THROUGH */ |
4757 | case WHILEM_A_pre_fail: /* just failed to match even minimal A */ | |
92e82afa YO |
4758 | REGCP_UNWIND(ST.lastcp); |
4759 | regcppop(rex); | |
c476f425 DM |
4760 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; |
4761 | cur_curlyx->u.curlyx.count--; | |
4762 | CACHEsayNO; | |
4763 | /* NOTREACHED */ | |
4764 | ||
4765 | case WHILEM_A_max_fail: /* just failed to match A in a maximal match */ | |
4766 | REGCP_UNWIND(ST.lastcp); | |
4767 | regcppop(rex); /* Restore some previous $<digit>s? */ | |
4768 | PL_reginput = locinput; | |
4769 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
4770 | "%*s whilem: failed, trying continuation...\n", | |
4771 | REPORT_CODE_OFF+depth*2, "") | |
4772 | ); | |
4773 | do_whilem_B_max: | |
4774 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
4775 | && ckWARN(WARN_REGEXP) | |
4776 | && !(PL_reg_flags & RF_warned)) | |
4777 | { | |
4778 | PL_reg_flags |= RF_warned; | |
4779 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded", | |
4780 | "Complex regular subexpression recursion", | |
4781 | REG_INFTY - 1); | |
4782 | } | |
4783 | ||
4784 | /* now try B */ | |
4785 | ST.save_curlyx = cur_curlyx; | |
4786 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4787 | PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B); | |
4788 | /* NOTREACHED */ | |
4789 | ||
4790 | case WHILEM_B_min_fail: /* just failed to match B in a minimal match */ | |
4791 | cur_curlyx = ST.save_curlyx; | |
4792 | REGCP_UNWIND(ST.lastcp); | |
4793 | regcppop(rex); | |
4794 | ||
d02d6d97 | 4795 | if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) { |
c476f425 DM |
4796 | /* Maximum greed exceeded */ |
4797 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
4798 | && ckWARN(WARN_REGEXP) | |
4799 | && !(PL_reg_flags & RF_warned)) | |
4800 | { | |
3280af22 | 4801 | PL_reg_flags |= RF_warned; |
c476f425 DM |
4802 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
4803 | "%s limit (%d) exceeded", | |
4804 | "Complex regular subexpression recursion", | |
4805 | REG_INFTY - 1); | |
a0d0e21e | 4806 | } |
c476f425 | 4807 | cur_curlyx->u.curlyx.count--; |
3ab3c9b4 | 4808 | CACHEsayNO; |
a0d0e21e | 4809 | } |
c476f425 DM |
4810 | |
4811 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
4812 | "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "") | |
4813 | ); | |
4814 | /* Try grabbing another A and see if it helps. */ | |
4815 | PL_reginput = locinput; | |
4816 | cur_curlyx->u.curlyx.lastloc = locinput; | |
4817 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); | |
4818 | REGCP_SET(ST.lastcp); | |
d02d6d97 DM |
4819 | PUSH_STATE_GOTO(WHILEM_A_min, |
4820 | /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS); | |
5f66b61c | 4821 | /* NOTREACHED */ |
40a82448 DM |
4822 | |
4823 | #undef ST | |
4824 | #define ST st->u.branch | |
4825 | ||
4826 | case BRANCHJ: /* /(...|A|...)/ with long next pointer */ | |
c277df42 IZ |
4827 | next = scan + ARG(scan); |
4828 | if (next == scan) | |
4829 | next = NULL; | |
40a82448 DM |
4830 | scan = NEXTOPER(scan); |
4831 | /* FALL THROUGH */ | |
c277df42 | 4832 | |
40a82448 DM |
4833 | case BRANCH: /* /(...|A|...)/ */ |
4834 | scan = NEXTOPER(scan); /* scan now points to inner node */ | |
40a82448 DM |
4835 | ST.lastparen = *PL_reglastparen; |
4836 | ST.next_branch = next; | |
4837 | REGCP_SET(ST.cp); | |
4838 | PL_reginput = locinput; | |
02db2b7b | 4839 | |
40a82448 | 4840 | /* Now go into the branch */ |
5d458dd8 YO |
4841 | if (has_cutgroup) { |
4842 | PUSH_YES_STATE_GOTO(BRANCH_next, scan); | |
4843 | } else { | |
4844 | PUSH_STATE_GOTO(BRANCH_next, scan); | |
4845 | } | |
40a82448 | 4846 | /* NOTREACHED */ |
5d458dd8 YO |
4847 | case CUTGROUP: |
4848 | PL_reginput = locinput; | |
4849 | sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : | |
ad64d0ec | 4850 | MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
4851 | PUSH_STATE_GOTO(CUTGROUP_next,next); |
4852 | /* NOTREACHED */ | |
4853 | case CUTGROUP_next_fail: | |
4854 | do_cutgroup = 1; | |
4855 | no_final = 1; | |
4856 | if (st->u.mark.mark_name) | |
4857 | sv_commit = st->u.mark.mark_name; | |
4858 | sayNO; | |
4859 | /* NOTREACHED */ | |
4860 | case BRANCH_next: | |
4861 | sayYES; | |
4862 | /* NOTREACHED */ | |
40a82448 | 4863 | case BRANCH_next_fail: /* that branch failed; try the next, if any */ |
5d458dd8 YO |
4864 | if (do_cutgroup) { |
4865 | do_cutgroup = 0; | |
4866 | no_final = 0; | |
4867 | } | |
40a82448 DM |
4868 | REGCP_UNWIND(ST.cp); |
4869 | for (n = *PL_reglastparen; n > ST.lastparen; n--) | |
f0ab9afb | 4870 | PL_regoffs[n].end = -1; |
40a82448 | 4871 | *PL_reglastparen = n; |
0a4db386 | 4872 | /*dmq: *PL_reglastcloseparen = n; */ |
40a82448 DM |
4873 | scan = ST.next_branch; |
4874 | /* no more branches? */ | |
5d458dd8 YO |
4875 | if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) { |
4876 | DEBUG_EXECUTE_r({ | |
4877 | PerlIO_printf( Perl_debug_log, | |
4878 | "%*s %sBRANCH failed...%s\n", | |
4879 | REPORT_CODE_OFF+depth*2, "", | |
4880 | PL_colors[4], | |
4881 | PL_colors[5] ); | |
4882 | }); | |
4883 | sayNO_SILENT; | |
4884 | } | |
40a82448 DM |
4885 | continue; /* execute next BRANCH[J] op */ |
4886 | /* NOTREACHED */ | |
4887 | ||
a0d0e21e | 4888 | case MINMOD: |
24d3c4a9 | 4889 | minmod = 1; |
a0d0e21e | 4890 | break; |
40a82448 DM |
4891 | |
4892 | #undef ST | |
4893 | #define ST st->u.curlym | |
4894 | ||
4895 | case CURLYM: /* /A{m,n}B/ where A is fixed-length */ | |
4896 | ||
4897 | /* This is an optimisation of CURLYX that enables us to push | |
84d2fa14 | 4898 | * only a single backtracking state, no matter how many matches |
40a82448 DM |
4899 | * there are in {m,n}. It relies on the pattern being constant |
4900 | * length, with no parens to influence future backrefs | |
4901 | */ | |
4902 | ||
4903 | ST.me = scan; | |
dc45a647 | 4904 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; |
40a82448 DM |
4905 | |
4906 | /* if paren positive, emulate an OPEN/CLOSE around A */ | |
4907 | if (ST.me->flags) { | |
3b6647e0 | 4908 | U32 paren = ST.me->flags; |
40a82448 DM |
4909 | if (paren > PL_regsize) |
4910 | PL_regsize = paren; | |
3b6647e0 | 4911 | if (paren > *PL_reglastparen) |
40a82448 | 4912 | *PL_reglastparen = paren; |
c277df42 | 4913 | scan += NEXT_OFF(scan); /* Skip former OPEN. */ |
6407bf3b | 4914 | } |
40a82448 DM |
4915 | ST.A = scan; |
4916 | ST.B = next; | |
4917 | ST.alen = 0; | |
4918 | ST.count = 0; | |
24d3c4a9 DM |
4919 | ST.minmod = minmod; |
4920 | minmod = 0; | |
40a82448 DM |
4921 | ST.c1 = CHRTEST_UNINIT; |
4922 | REGCP_SET(ST.cp); | |
6407bf3b | 4923 | |
40a82448 DM |
4924 | if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */ |
4925 | goto curlym_do_B; | |
4926 | ||
4927 | curlym_do_A: /* execute the A in /A{m,n}B/ */ | |
6407bf3b | 4928 | PL_reginput = locinput; |
40a82448 DM |
4929 | PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */ |
4930 | /* NOTREACHED */ | |
5f80c4cf | 4931 | |
40a82448 DM |
4932 | case CURLYM_A: /* we've just matched an A */ |
4933 | locinput = st->locinput; | |
4934 | nextchr = UCHARAT(locinput); | |
4935 | ||
4936 | ST.count++; | |
4937 | /* after first match, determine A's length: u.curlym.alen */ | |
4938 | if (ST.count == 1) { | |
4939 | if (PL_reg_match_utf8) { | |
4940 | char *s = locinput; | |
4941 | while (s < PL_reginput) { | |
4942 | ST.alen++; | |
4943 | s += UTF8SKIP(s); | |
4944 | } | |
4945 | } | |
4946 | else { | |
4947 | ST.alen = PL_reginput - locinput; | |
4948 | } | |
4949 | if (ST.alen == 0) | |
4950 | ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me); | |
4951 | } | |
0cadcf80 DM |
4952 | DEBUG_EXECUTE_r( |
4953 | PerlIO_printf(Perl_debug_log, | |
40a82448 | 4954 | "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n", |
5bc10b2c | 4955 | (int)(REPORT_CODE_OFF+(depth*2)), "", |
40a82448 | 4956 | (IV) ST.count, (IV)ST.alen) |
0cadcf80 DM |
4957 | ); |
4958 | ||
40a82448 | 4959 | locinput = PL_reginput; |
0a4db386 YO |
4960 | |
4961 | if (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 4962 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
4963 | goto fake_end; |
4964 | ||
c966426a DM |
4965 | { |
4966 | I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)); | |
4967 | if ( max == REG_INFTY || ST.count < max ) | |
4968 | goto curlym_do_A; /* try to match another A */ | |
4969 | } | |
40a82448 | 4970 | goto curlym_do_B; /* try to match B */ |
5f80c4cf | 4971 | |
40a82448 DM |
4972 | case CURLYM_A_fail: /* just failed to match an A */ |
4973 | REGCP_UNWIND(ST.cp); | |
0a4db386 YO |
4974 | |
4975 | if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ | |
4976 | || (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 4977 | cur_eval->u.eval.close_paren == (U32)ST.me->flags)) |
40a82448 | 4978 | sayNO; |
0cadcf80 | 4979 | |
40a82448 DM |
4980 | curlym_do_B: /* execute the B in /A{m,n}B/ */ |
4981 | PL_reginput = locinput; | |
4982 | if (ST.c1 == CHRTEST_UNINIT) { | |
4983 | /* calculate c1 and c2 for possible match of 1st char | |
4984 | * following curly */ | |
4985 | ST.c1 = ST.c2 = CHRTEST_VOID; | |
4986 | if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) { | |
4987 | regnode *text_node = ST.B; | |
4988 | if (! HAS_TEXT(text_node)) | |
4989 | FIND_NEXT_IMPT(text_node); | |
ee9b8eae YO |
4990 | /* this used to be |
4991 | ||
4992 | (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT) | |
4993 | ||
4994 | But the former is redundant in light of the latter. | |
4995 | ||
4996 | if this changes back then the macro for | |
4997 | IS_TEXT and friends need to change. | |
4998 | */ | |
4999 | if (PL_regkind[OP(text_node)] == EXACT) | |
40a82448 | 5000 | { |
ee9b8eae | 5001 | |
40a82448 | 5002 | ST.c1 = (U8)*STRING(text_node); |
9a5a5549 KW |
5003 | switch (OP(text_node)) { |
5004 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
2f7f8cb1 | 5005 | case EXACTFA: |
9a5a5549 KW |
5006 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; |
5007 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
5008 | default: ST.c2 = ST.c1; | |
5009 | } | |
c277df42 | 5010 | } |
c277df42 | 5011 | } |
40a82448 DM |
5012 | } |
5013 | ||
5014 | DEBUG_EXECUTE_r( | |
5015 | PerlIO_printf(Perl_debug_log, | |
5016 | "%*s CURLYM trying tail with matches=%"IVdf"...\n", | |
5bc10b2c | 5017 | (int)(REPORT_CODE_OFF+(depth*2)), |
40a82448 DM |
5018 | "", (IV)ST.count) |
5019 | ); | |
5020 | if (ST.c1 != CHRTEST_VOID | |
5021 | && UCHARAT(PL_reginput) != ST.c1 | |
5022 | && UCHARAT(PL_reginput) != ST.c2) | |
5023 | { | |
5024 | /* simulate B failing */ | |
3e901dc0 YO |
5025 | DEBUG_OPTIMISE_r( |
5026 | PerlIO_printf(Perl_debug_log, | |
5027 | "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n", | |
5028 | (int)(REPORT_CODE_OFF+(depth*2)),"", | |
5029 | (IV)ST.c1,(IV)ST.c2 | |
5030 | )); | |
40a82448 DM |
5031 | state_num = CURLYM_B_fail; |
5032 | goto reenter_switch; | |
5033 | } | |
5034 | ||
5035 | if (ST.me->flags) { | |
5036 | /* mark current A as captured */ | |
5037 | I32 paren = ST.me->flags; | |
5038 | if (ST.count) { | |
f0ab9afb | 5039 | PL_regoffs[paren].start |
40a82448 | 5040 | = HOPc(PL_reginput, -ST.alen) - PL_bostr; |
f0ab9afb | 5041 | PL_regoffs[paren].end = PL_reginput - PL_bostr; |
0a4db386 | 5042 | /*dmq: *PL_reglastcloseparen = paren; */ |
c277df42 | 5043 | } |
40a82448 | 5044 | else |
f0ab9afb | 5045 | PL_regoffs[paren].end = -1; |
0a4db386 | 5046 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 5047 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5048 | { |
5049 | if (ST.count) | |
5050 | goto fake_end; | |
5051 | else | |
5052 | sayNO; | |
5053 | } | |
c277df42 | 5054 | } |
0a4db386 | 5055 | |
40a82448 | 5056 | PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */ |
5f66b61c | 5057 | /* NOTREACHED */ |
40a82448 DM |
5058 | |
5059 | case CURLYM_B_fail: /* just failed to match a B */ | |
5060 | REGCP_UNWIND(ST.cp); | |
5061 | if (ST.minmod) { | |
84d2fa14 HS |
5062 | I32 max = ARG2(ST.me); |
5063 | if (max != REG_INFTY && ST.count == max) | |
40a82448 DM |
5064 | sayNO; |
5065 | goto curlym_do_A; /* try to match a further A */ | |
5066 | } | |
5067 | /* backtrack one A */ | |
5068 | if (ST.count == ARG1(ST.me) /* min */) | |
5069 | sayNO; | |
5070 | ST.count--; | |
5071 | locinput = HOPc(locinput, -ST.alen); | |
5072 | goto curlym_do_B; /* try to match B */ | |
5073 | ||
c255a977 DM |
5074 | #undef ST |
5075 | #define ST st->u.curly | |
40a82448 | 5076 | |
c255a977 DM |
5077 | #define CURLY_SETPAREN(paren, success) \ |
5078 | if (paren) { \ | |
5079 | if (success) { \ | |
f0ab9afb NC |
5080 | PL_regoffs[paren].start = HOPc(locinput, -1) - PL_bostr; \ |
5081 | PL_regoffs[paren].end = locinput - PL_bostr; \ | |
0a4db386 | 5082 | *PL_reglastcloseparen = paren; \ |
c255a977 DM |
5083 | } \ |
5084 | else \ | |
f0ab9afb | 5085 | PL_regoffs[paren].end = -1; \ |
c255a977 DM |
5086 | } |
5087 | ||
5088 | case STAR: /* /A*B/ where A is width 1 */ | |
5089 | ST.paren = 0; | |
5090 | ST.min = 0; | |
5091 | ST.max = REG_INFTY; | |
a0d0e21e LW |
5092 | scan = NEXTOPER(scan); |
5093 | goto repeat; | |
c255a977 DM |
5094 | case PLUS: /* /A+B/ where A is width 1 */ |
5095 | ST.paren = 0; | |
5096 | ST.min = 1; | |
5097 | ST.max = REG_INFTY; | |
c277df42 | 5098 | scan = NEXTOPER(scan); |
c255a977 DM |
5099 | goto repeat; |
5100 | case CURLYN: /* /(A){m,n}B/ where A is width 1 */ | |
5101 | ST.paren = scan->flags; /* Which paren to set */ | |
5102 | if (ST.paren > PL_regsize) | |
5103 | PL_regsize = ST.paren; | |
3b6647e0 | 5104 | if (ST.paren > *PL_reglastparen) |
c255a977 DM |
5105 | *PL_reglastparen = ST.paren; |
5106 | ST.min = ARG1(scan); /* min to match */ | |
5107 | ST.max = ARG2(scan); /* max to match */ | |
0a4db386 | 5108 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5109 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5110 | ST.min=1; |
5111 | ST.max=1; | |
5112 | } | |
c255a977 DM |
5113 | scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); |
5114 | goto repeat; | |
5115 | case CURLY: /* /A{m,n}B/ where A is width 1 */ | |
5116 | ST.paren = 0; | |
5117 | ST.min = ARG1(scan); /* min to match */ | |
5118 | ST.max = ARG2(scan); /* max to match */ | |
5119 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; | |
c277df42 | 5120 | repeat: |
a0d0e21e LW |
5121 | /* |
5122 | * Lookahead to avoid useless match attempts | |
5123 | * when we know what character comes next. | |
c255a977 | 5124 | * |
5f80c4cf JP |
5125 | * Used to only do .*x and .*?x, but now it allows |
5126 | * for )'s, ('s and (?{ ... })'s to be in the way | |
5127 | * of the quantifier and the EXACT-like node. -- japhy | |
5128 | */ | |
5129 | ||
c255a977 DM |
5130 | if (ST.min > ST.max) /* XXX make this a compile-time check? */ |
5131 | sayNO; | |
cca55fe3 | 5132 | if (HAS_TEXT(next) || JUMPABLE(next)) { |
5f80c4cf JP |
5133 | U8 *s; |
5134 | regnode *text_node = next; | |
5135 | ||
3dab1dad YO |
5136 | if (! HAS_TEXT(text_node)) |
5137 | FIND_NEXT_IMPT(text_node); | |
5f80c4cf | 5138 | |
9e137952 | 5139 | if (! HAS_TEXT(text_node)) |
c255a977 | 5140 | ST.c1 = ST.c2 = CHRTEST_VOID; |
5f80c4cf | 5141 | else { |
ee9b8eae | 5142 | if ( PL_regkind[OP(text_node)] != EXACT ) { |
c255a977 | 5143 | ST.c1 = ST.c2 = CHRTEST_VOID; |
44a68960 | 5144 | goto assume_ok_easy; |
cca55fe3 | 5145 | } |
be8e71aa YO |
5146 | else |
5147 | s = (U8*)STRING(text_node); | |
ee9b8eae YO |
5148 | |
5149 | /* Currently we only get here when | |
5150 | ||
5151 | PL_rekind[OP(text_node)] == EXACT | |
5152 | ||
5153 | if this changes back then the macro for IS_TEXT and | |
5154 | friends need to change. */ | |
f2ed9b32 | 5155 | if (!UTF_PATTERN) { |
9a5a5549 KW |
5156 | ST.c1 = *s; |
5157 | switch (OP(text_node)) { | |
5158 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
2f7f8cb1 | 5159 | case EXACTFA: |
9a5a5549 KW |
5160 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; |
5161 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
5162 | default: ST.c2 = ST.c1; break; | |
5163 | } | |
1aa99e6b | 5164 | } |
f2ed9b32 | 5165 | else { /* UTF_PATTERN */ |
9a5a5549 | 5166 | if (IS_TEXTFU(text_node) || IS_TEXTF(text_node)) { |
a2a2844f | 5167 | STRLEN ulen1, ulen2; |
89ebb4a3 JH |
5168 | U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; |
5169 | U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; | |
a2a2844f JH |
5170 | |
5171 | to_utf8_lower((U8*)s, tmpbuf1, &ulen1); | |
5172 | to_utf8_upper((U8*)s, tmpbuf2, &ulen2); | |
e294cc5d JH |
5173 | #ifdef EBCDIC |
5174 | ST.c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN, 0, | |
5175 | ckWARN(WARN_UTF8) ? | |
5176 | 0 : UTF8_ALLOW_ANY); | |
5177 | ST.c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN, 0, | |
5178 | ckWARN(WARN_UTF8) ? | |
5179 | 0 : UTF8_ALLOW_ANY); | |
5180 | #else | |
c255a977 | 5181 | ST.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0, |
e294cc5d | 5182 | uniflags); |
c255a977 | 5183 | ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0, |
e294cc5d JH |
5184 | uniflags); |
5185 | #endif | |
5f80c4cf JP |
5186 | } |
5187 | else { | |
c255a977 | 5188 | ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, |
041457d9 | 5189 | uniflags); |
5f80c4cf | 5190 | } |
1aa99e6b IH |
5191 | } |
5192 | } | |
bbce6d69 | 5193 | } |
a0d0e21e | 5194 | else |
c255a977 | 5195 | ST.c1 = ST.c2 = CHRTEST_VOID; |
cca55fe3 | 5196 | assume_ok_easy: |
c255a977 DM |
5197 | |
5198 | ST.A = scan; | |
5199 | ST.B = next; | |
3280af22 | 5200 | PL_reginput = locinput; |
24d3c4a9 DM |
5201 | if (minmod) { |
5202 | minmod = 0; | |
e2e6a0f1 | 5203 | if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min) |
4633a7c4 | 5204 | sayNO; |
c255a977 | 5205 | ST.count = ST.min; |
a0ed51b3 | 5206 | locinput = PL_reginput; |
c255a977 DM |
5207 | REGCP_SET(ST.cp); |
5208 | if (ST.c1 == CHRTEST_VOID) | |
5209 | goto curly_try_B_min; | |
5210 | ||
5211 | ST.oldloc = locinput; | |
5212 | ||
5213 | /* set ST.maxpos to the furthest point along the | |
5214 | * string that could possibly match */ | |
5215 | if (ST.max == REG_INFTY) { | |
5216 | ST.maxpos = PL_regeol - 1; | |
f2ed9b32 | 5217 | if (utf8_target) |
c255a977 DM |
5218 | while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos)) |
5219 | ST.maxpos--; | |
5220 | } | |
f2ed9b32 | 5221 | else if (utf8_target) { |
c255a977 DM |
5222 | int m = ST.max - ST.min; |
5223 | for (ST.maxpos = locinput; | |
5224 | m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--) | |
5225 | ST.maxpos += UTF8SKIP(ST.maxpos); | |
5226 | } | |
5227 | else { | |
5228 | ST.maxpos = locinput + ST.max - ST.min; | |
5229 | if (ST.maxpos >= PL_regeol) | |
5230 | ST.maxpos = PL_regeol - 1; | |
5231 | } | |
5232 | goto curly_try_B_min_known; | |
5233 | ||
5234 | } | |
5235 | else { | |
e2e6a0f1 | 5236 | ST.count = regrepeat(rex, ST.A, ST.max, depth); |
c255a977 DM |
5237 | locinput = PL_reginput; |
5238 | if (ST.count < ST.min) | |
5239 | sayNO; | |
5240 | if ((ST.count > ST.min) | |
5241 | && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL)) | |
5242 | { | |
5243 | /* A{m,n} must come at the end of the string, there's | |
5244 | * no point in backing off ... */ | |
5245 | ST.min = ST.count; | |
5246 | /* ...except that $ and \Z can match before *and* after | |
5247 | newline at the end. Consider "\n\n" =~ /\n+\Z\n/. | |
5248 | We may back off by one in this case. */ | |
5249 | if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS) | |
5250 | ST.min--; | |
5251 | } | |
5252 | REGCP_SET(ST.cp); | |
5253 | goto curly_try_B_max; | |
5254 | } | |
5255 | /* NOTREACHED */ | |
5256 | ||
5257 | ||
5258 | case CURLY_B_min_known_fail: | |
5259 | /* failed to find B in a non-greedy match where c1,c2 valid */ | |
5260 | if (ST.paren && ST.count) | |
f0ab9afb | 5261 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5262 | |
5263 | PL_reginput = locinput; /* Could be reset... */ | |
5264 | REGCP_UNWIND(ST.cp); | |
5265 | /* Couldn't or didn't -- move forward. */ | |
5266 | ST.oldloc = locinput; | |
f2ed9b32 | 5267 | if (utf8_target) |
c255a977 DM |
5268 | locinput += UTF8SKIP(locinput); |
5269 | else | |
5270 | locinput++; | |
5271 | ST.count++; | |
5272 | curly_try_B_min_known: | |
5273 | /* find the next place where 'B' could work, then call B */ | |
5274 | { | |
5275 | int n; | |
f2ed9b32 | 5276 | if (utf8_target) { |
c255a977 DM |
5277 | n = (ST.oldloc == locinput) ? 0 : 1; |
5278 | if (ST.c1 == ST.c2) { | |
5279 | STRLEN len; | |
5280 | /* set n to utf8_distance(oldloc, locinput) */ | |
5281 | while (locinput <= ST.maxpos && | |
5282 | utf8n_to_uvchr((U8*)locinput, | |
5283 | UTF8_MAXBYTES, &len, | |
5284 | uniflags) != (UV)ST.c1) { | |
5285 | locinput += len; | |
5286 | n++; | |
5287 | } | |
1aa99e6b IH |
5288 | } |
5289 | else { | |
c255a977 DM |
5290 | /* set n to utf8_distance(oldloc, locinput) */ |
5291 | while (locinput <= ST.maxpos) { | |
5292 | STRLEN len; | |
5293 | const UV c = utf8n_to_uvchr((U8*)locinput, | |
5294 | UTF8_MAXBYTES, &len, | |
5295 | uniflags); | |
5296 | if (c == (UV)ST.c1 || c == (UV)ST.c2) | |
5297 | break; | |
5298 | locinput += len; | |
5299 | n++; | |
1aa99e6b | 5300 | } |
0fe9bf95 IZ |
5301 | } |
5302 | } | |
c255a977 DM |
5303 | else { |
5304 | if (ST.c1 == ST.c2) { | |
5305 | while (locinput <= ST.maxpos && | |
5306 | UCHARAT(locinput) != ST.c1) | |
5307 | locinput++; | |
bbce6d69 | 5308 | } |
c255a977 DM |
5309 | else { |
5310 | while (locinput <= ST.maxpos | |
5311 | && UCHARAT(locinput) != ST.c1 | |
5312 | && UCHARAT(locinput) != ST.c2) | |
5313 | locinput++; | |
a0ed51b3 | 5314 | } |
c255a977 DM |
5315 | n = locinput - ST.oldloc; |
5316 | } | |
5317 | if (locinput > ST.maxpos) | |
5318 | sayNO; | |
5319 | /* PL_reginput == oldloc now */ | |
5320 | if (n) { | |
5321 | ST.count += n; | |
e2e6a0f1 | 5322 | if (regrepeat(rex, ST.A, n, depth) < n) |
4633a7c4 | 5323 | sayNO; |
a0d0e21e | 5324 | } |
c255a977 DM |
5325 | PL_reginput = locinput; |
5326 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5327 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5328 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5329 | goto fake_end; |
5330 | } | |
c255a977 | 5331 | PUSH_STATE_GOTO(CURLY_B_min_known, ST.B); |
a0d0e21e | 5332 | } |
c255a977 DM |
5333 | /* NOTREACHED */ |
5334 | ||
5335 | ||
5336 | case CURLY_B_min_fail: | |
5337 | /* failed to find B in a non-greedy match where c1,c2 invalid */ | |
5338 | if (ST.paren && ST.count) | |
f0ab9afb | 5339 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5340 | |
5341 | REGCP_UNWIND(ST.cp); | |
5342 | /* failed -- move forward one */ | |
5343 | PL_reginput = locinput; | |
e2e6a0f1 | 5344 | if (regrepeat(rex, ST.A, 1, depth)) { |
c255a977 | 5345 | ST.count++; |
a0ed51b3 | 5346 | locinput = PL_reginput; |
c255a977 DM |
5347 | if (ST.count <= ST.max || (ST.max == REG_INFTY && |
5348 | ST.count > 0)) /* count overflow ? */ | |
15272685 | 5349 | { |
c255a977 DM |
5350 | curly_try_B_min: |
5351 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5352 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5353 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5354 | goto fake_end; |
5355 | } | |
c255a977 | 5356 | PUSH_STATE_GOTO(CURLY_B_min, ST.B); |
a0d0e21e LW |
5357 | } |
5358 | } | |
4633a7c4 | 5359 | sayNO; |
c255a977 DM |
5360 | /* NOTREACHED */ |
5361 | ||
5362 | ||
5363 | curly_try_B_max: | |
5364 | /* a successful greedy match: now try to match B */ | |
40d049e4 | 5365 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5366 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
40d049e4 YO |
5367 | goto fake_end; |
5368 | } | |
c255a977 DM |
5369 | { |
5370 | UV c = 0; | |
5371 | if (ST.c1 != CHRTEST_VOID) | |
f2ed9b32 | 5372 | c = utf8_target ? utf8n_to_uvchr((U8*)PL_reginput, |
c255a977 | 5373 | UTF8_MAXBYTES, 0, uniflags) |
466787eb | 5374 | : (UV) UCHARAT(PL_reginput); |
c255a977 DM |
5375 | /* If it could work, try it. */ |
5376 | if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) { | |
5377 | CURLY_SETPAREN(ST.paren, ST.count); | |
5378 | PUSH_STATE_GOTO(CURLY_B_max, ST.B); | |
5379 | /* NOTREACHED */ | |
5380 | } | |
5381 | } | |
5382 | /* FALL THROUGH */ | |
5383 | case CURLY_B_max_fail: | |
5384 | /* failed to find B in a greedy match */ | |
5385 | if (ST.paren && ST.count) | |
f0ab9afb | 5386 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5387 | |
5388 | REGCP_UNWIND(ST.cp); | |
5389 | /* back up. */ | |
5390 | if (--ST.count < ST.min) | |
5391 | sayNO; | |
5392 | PL_reginput = locinput = HOPc(locinput, -1); | |
5393 | goto curly_try_B_max; | |
5394 | ||
5395 | #undef ST | |
5396 | ||
a0d0e21e | 5397 | case END: |
6bda09f9 | 5398 | fake_end: |
faec1544 DM |
5399 | if (cur_eval) { |
5400 | /* we've just finished A in /(??{A})B/; now continue with B */ | |
5401 | I32 tmpix; | |
faec1544 DM |
5402 | st->u.eval.toggle_reg_flags |
5403 | = cur_eval->u.eval.toggle_reg_flags; | |
5404 | PL_reg_flags ^= st->u.eval.toggle_reg_flags; | |
5405 | ||
288b8c02 NC |
5406 | st->u.eval.prev_rex = rex_sv; /* inner */ |
5407 | SETREX(rex_sv,cur_eval->u.eval.prev_rex); | |
5408 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 5409 | rexi = RXi_GET(rex); |
faec1544 | 5410 | cur_curlyx = cur_eval->u.eval.prev_curlyx; |
cd733337 | 5411 | (void)ReREFCNT_inc(rex_sv); |
faec1544 | 5412 | st->u.eval.cp = regcppush(0); /* Save *all* the positions. */ |
34a81e2b B |
5413 | |
5414 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ | |
5415 | PL_reglastparen = &rex->lastparen; | |
5416 | PL_reglastcloseparen = &rex->lastcloseparen; | |
5417 | ||
faec1544 DM |
5418 | REGCP_SET(st->u.eval.lastcp); |
5419 | PL_reginput = locinput; | |
5420 | ||
5421 | /* Restore parens of the outer rex without popping the | |
5422 | * savestack */ | |
5423 | tmpix = PL_savestack_ix; | |
5424 | PL_savestack_ix = cur_eval->u.eval.lastcp; | |
5425 | regcppop(rex); | |
5426 | PL_savestack_ix = tmpix; | |
5427 | ||
5428 | st->u.eval.prev_eval = cur_eval; | |
5429 | cur_eval = cur_eval->u.eval.prev_eval; | |
5430 | DEBUG_EXECUTE_r( | |
2a49f0f5 JH |
5431 | PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n", |
5432 | REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval));); | |
e7707071 YO |
5433 | if ( nochange_depth ) |
5434 | nochange_depth--; | |
5435 | ||
5436 | PUSH_YES_STATE_GOTO(EVAL_AB, | |
faec1544 DM |
5437 | st->u.eval.prev_eval->u.eval.B); /* match B */ |
5438 | } | |
5439 | ||
3b0527fe | 5440 | if (locinput < reginfo->till) { |
a3621e74 | 5441 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
7821416a IZ |
5442 | "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n", |
5443 | PL_colors[4], | |
5444 | (long)(locinput - PL_reg_starttry), | |
3b0527fe | 5445 | (long)(reginfo->till - PL_reg_starttry), |
7821416a | 5446 | PL_colors[5])); |
58e23c8d | 5447 | |
262b90c4 | 5448 | sayNO_SILENT; /* Cannot match: too short. */ |
7821416a IZ |
5449 | } |
5450 | PL_reginput = locinput; /* put where regtry can find it */ | |
262b90c4 | 5451 | sayYES; /* Success! */ |
dad79028 DM |
5452 | |
5453 | case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */ | |
5454 | DEBUG_EXECUTE_r( | |
5455 | PerlIO_printf(Perl_debug_log, | |
5456 | "%*s %ssubpattern success...%s\n", | |
5bc10b2c | 5457 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])); |
3280af22 | 5458 | PL_reginput = locinput; /* put where regtry can find it */ |
262b90c4 | 5459 | sayYES; /* Success! */ |
dad79028 | 5460 | |
40a82448 DM |
5461 | #undef ST |
5462 | #define ST st->u.ifmatch | |
5463 | ||
5464 | case SUSPEND: /* (?>A) */ | |
5465 | ST.wanted = 1; | |
9fe1d20c | 5466 | PL_reginput = locinput; |
9041c2e3 | 5467 | goto do_ifmatch; |
dad79028 | 5468 | |
40a82448 DM |
5469 | case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */ |
5470 | ST.wanted = 0; | |
dad79028 DM |
5471 | goto ifmatch_trivial_fail_test; |
5472 | ||
40a82448 DM |
5473 | case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */ |
5474 | ST.wanted = 1; | |
dad79028 | 5475 | ifmatch_trivial_fail_test: |
a0ed51b3 | 5476 | if (scan->flags) { |
52657f30 | 5477 | char * const s = HOPBACKc(locinput, scan->flags); |
dad79028 DM |
5478 | if (!s) { |
5479 | /* trivial fail */ | |
24d3c4a9 DM |
5480 | if (logical) { |
5481 | logical = 0; | |
f2338a2e | 5482 | sw = 1 - cBOOL(ST.wanted); |
dad79028 | 5483 | } |
40a82448 | 5484 | else if (ST.wanted) |
dad79028 DM |
5485 | sayNO; |
5486 | next = scan + ARG(scan); | |
5487 | if (next == scan) | |
5488 | next = NULL; | |
5489 | break; | |
5490 | } | |
efb30f32 | 5491 | PL_reginput = s; |
a0ed51b3 LW |
5492 | } |
5493 | else | |
5494 | PL_reginput = locinput; | |
5495 | ||
c277df42 | 5496 | do_ifmatch: |
40a82448 | 5497 | ST.me = scan; |
24d3c4a9 | 5498 | ST.logical = logical; |
24d786f4 YO |
5499 | logical = 0; /* XXX: reset state of logical once it has been saved into ST */ |
5500 | ||
40a82448 DM |
5501 | /* execute body of (?...A) */ |
5502 | PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan))); | |
5503 | /* NOTREACHED */ | |
5504 | ||
5505 | case IFMATCH_A_fail: /* body of (?...A) failed */ | |
5506 | ST.wanted = !ST.wanted; | |
5507 | /* FALL THROUGH */ | |
5508 | ||
5509 | case IFMATCH_A: /* body of (?...A) succeeded */ | |
24d3c4a9 | 5510 | if (ST.logical) { |
f2338a2e | 5511 | sw = cBOOL(ST.wanted); |
40a82448 DM |
5512 | } |
5513 | else if (!ST.wanted) | |
5514 | sayNO; | |
5515 | ||
5516 | if (OP(ST.me) == SUSPEND) | |
5517 | locinput = PL_reginput; | |
5518 | else { | |
5519 | locinput = PL_reginput = st->locinput; | |
5520 | nextchr = UCHARAT(locinput); | |
5521 | } | |
5522 | scan = ST.me + ARG(ST.me); | |
5523 | if (scan == ST.me) | |
5524 | scan = NULL; | |
5525 | continue; /* execute B */ | |
5526 | ||
5527 | #undef ST | |
dad79028 | 5528 | |
c277df42 | 5529 | case LONGJMP: |
c277df42 IZ |
5530 | next = scan + ARG(scan); |
5531 | if (next == scan) | |
5532 | next = NULL; | |
a0d0e21e | 5533 | break; |
54612592 | 5534 | case COMMIT: |
e2e6a0f1 YO |
5535 | reginfo->cutpoint = PL_regeol; |
5536 | /* FALLTHROUGH */ | |
5d458dd8 | 5537 | case PRUNE: |
24b23f37 | 5538 | PL_reginput = locinput; |
e2e6a0f1 | 5539 | if (!scan->flags) |
ad64d0ec | 5540 | sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
54612592 YO |
5541 | PUSH_STATE_GOTO(COMMIT_next,next); |
5542 | /* NOTREACHED */ | |
5543 | case COMMIT_next_fail: | |
5544 | no_final = 1; | |
5545 | /* FALLTHROUGH */ | |
7f69552c YO |
5546 | case OPFAIL: |
5547 | sayNO; | |
e2e6a0f1 YO |
5548 | /* NOTREACHED */ |
5549 | ||
5550 | #define ST st->u.mark | |
5551 | case MARKPOINT: | |
5552 | ST.prev_mark = mark_state; | |
5d458dd8 | 5553 | ST.mark_name = sv_commit = sv_yes_mark |
ad64d0ec | 5554 | = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
e2e6a0f1 YO |
5555 | mark_state = st; |
5556 | ST.mark_loc = PL_reginput = locinput; | |
5557 | PUSH_YES_STATE_GOTO(MARKPOINT_next,next); | |
5558 | /* NOTREACHED */ | |
5559 | case MARKPOINT_next: | |
5560 | mark_state = ST.prev_mark; | |
5561 | sayYES; | |
5562 | /* NOTREACHED */ | |
5563 | case MARKPOINT_next_fail: | |
5d458dd8 | 5564 | if (popmark && sv_eq(ST.mark_name,popmark)) |
e2e6a0f1 YO |
5565 | { |
5566 | if (ST.mark_loc > startpoint) | |
5567 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5568 | popmark = NULL; /* we found our mark */ | |
5569 | sv_commit = ST.mark_name; | |
5570 | ||
5571 | DEBUG_EXECUTE_r({ | |
5d458dd8 | 5572 | PerlIO_printf(Perl_debug_log, |
e2e6a0f1 YO |
5573 | "%*s %ssetting cutpoint to mark:%"SVf"...%s\n", |
5574 | REPORT_CODE_OFF+depth*2, "", | |
be2597df | 5575 | PL_colors[4], SVfARG(sv_commit), PL_colors[5]); |
e2e6a0f1 YO |
5576 | }); |
5577 | } | |
5578 | mark_state = ST.prev_mark; | |
5d458dd8 YO |
5579 | sv_yes_mark = mark_state ? |
5580 | mark_state->u.mark.mark_name : NULL; | |
e2e6a0f1 YO |
5581 | sayNO; |
5582 | /* NOTREACHED */ | |
5d458dd8 YO |
5583 | case SKIP: |
5584 | PL_reginput = locinput; | |
5585 | if (scan->flags) { | |
2bf803e2 | 5586 | /* (*SKIP) : if we fail we cut here*/ |
5d458dd8 | 5587 | ST.mark_name = NULL; |
e2e6a0f1 | 5588 | ST.mark_loc = locinput; |
5d458dd8 YO |
5589 | PUSH_STATE_GOTO(SKIP_next,next); |
5590 | } else { | |
2bf803e2 | 5591 | /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was, |
5d458dd8 YO |
5592 | otherwise do nothing. Meaning we need to scan |
5593 | */ | |
5594 | regmatch_state *cur = mark_state; | |
ad64d0ec | 5595 | SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
5596 | |
5597 | while (cur) { | |
5598 | if ( sv_eq( cur->u.mark.mark_name, | |
5599 | find ) ) | |
5600 | { | |
5601 | ST.mark_name = find; | |
5602 | PUSH_STATE_GOTO( SKIP_next, next ); | |
5603 | } | |
5604 | cur = cur->u.mark.prev_mark; | |
5605 | } | |
e2e6a0f1 | 5606 | } |
2bf803e2 | 5607 | /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */ |
5d458dd8 YO |
5608 | break; |
5609 | case SKIP_next_fail: | |
5610 | if (ST.mark_name) { | |
5611 | /* (*CUT:NAME) - Set up to search for the name as we | |
5612 | collapse the stack*/ | |
5613 | popmark = ST.mark_name; | |
5614 | } else { | |
5615 | /* (*CUT) - No name, we cut here.*/ | |
e2e6a0f1 YO |
5616 | if (ST.mark_loc > startpoint) |
5617 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5d458dd8 YO |
5618 | /* but we set sv_commit to latest mark_name if there |
5619 | is one so they can test to see how things lead to this | |
5620 | cut */ | |
5621 | if (mark_state) | |
5622 | sv_commit=mark_state->u.mark.mark_name; | |
5623 | } | |
e2e6a0f1 YO |
5624 | no_final = 1; |
5625 | sayNO; | |
5626 | /* NOTREACHED */ | |
5627 | #undef ST | |
32e6a07c YO |
5628 | case FOLDCHAR: |
5629 | n = ARG(scan); | |
f2ed9b32 | 5630 | if ( n == (U32)what_len_TRICKYFOLD(locinput,utf8_target,ln) ) { |
e64b1bd1 | 5631 | locinput += ln; |
ced7f090 | 5632 | } else if ( LATIN_SMALL_LETTER_SHARP_S == n && !utf8_target && !UTF_PATTERN ) { |
e64b1bd1 YO |
5633 | sayNO; |
5634 | } else { | |
5635 | U8 folded[UTF8_MAXBYTES_CASE+1]; | |
5636 | STRLEN foldlen; | |
5637 | const char * const l = locinput; | |
5638 | char *e = PL_regeol; | |
5639 | to_uni_fold(n, folded, &foldlen); | |
5640 | ||
4c1b470c | 5641 | if (! foldEQ_utf8((const char*) folded, 0, foldlen, 1, |
f2ed9b32 | 5642 | l, &e, 0, utf8_target)) { |
32e6a07c | 5643 | sayNO; |
e64b1bd1 YO |
5644 | } |
5645 | locinput = e; | |
32e6a07c YO |
5646 | } |
5647 | nextchr = UCHARAT(locinput); | |
5648 | break; | |
e1d1eefb | 5649 | case LNBREAK: |
f2ed9b32 | 5650 | if ((n=is_LNBREAK(locinput,utf8_target))) { |
e1d1eefb YO |
5651 | locinput += n; |
5652 | nextchr = UCHARAT(locinput); | |
5653 | } else | |
5654 | sayNO; | |
5655 | break; | |
5656 | ||
5657 | #define CASE_CLASS(nAmE) \ | |
5658 | case nAmE: \ | |
f2ed9b32 | 5659 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5660 | locinput += n; \ |
5661 | nextchr = UCHARAT(locinput); \ | |
5662 | } else \ | |
5663 | sayNO; \ | |
5664 | break; \ | |
5665 | case N##nAmE: \ | |
f2ed9b32 | 5666 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5667 | sayNO; \ |
5668 | } else { \ | |
5669 | locinput += UTF8SKIP(locinput); \ | |
5670 | nextchr = UCHARAT(locinput); \ | |
5671 | } \ | |
5672 | break | |
5673 | ||
5674 | CASE_CLASS(VERTWS); | |
5675 | CASE_CLASS(HORIZWS); | |
5676 | #undef CASE_CLASS | |
5677 | ||
a0d0e21e | 5678 | default: |
b900a521 | 5679 | PerlIO_printf(Perl_error_log, "%"UVxf" %d\n", |
d7d93a81 | 5680 | PTR2UV(scan), OP(scan)); |
cea2e8a9 | 5681 | Perl_croak(aTHX_ "regexp memory corruption"); |
5d458dd8 YO |
5682 | |
5683 | } /* end switch */ | |
95b24440 | 5684 | |
5d458dd8 YO |
5685 | /* switch break jumps here */ |
5686 | scan = next; /* prepare to execute the next op and ... */ | |
5687 | continue; /* ... jump back to the top, reusing st */ | |
95b24440 DM |
5688 | /* NOTREACHED */ |
5689 | ||
40a82448 DM |
5690 | push_yes_state: |
5691 | /* push a state that backtracks on success */ | |
5692 | st->u.yes.prev_yes_state = yes_state; | |
5693 | yes_state = st; | |
5694 | /* FALL THROUGH */ | |
5695 | push_state: | |
5696 | /* push a new regex state, then continue at scan */ | |
5697 | { | |
5698 | regmatch_state *newst; | |
5699 | ||
24b23f37 YO |
5700 | DEBUG_STACK_r({ |
5701 | regmatch_state *cur = st; | |
5702 | regmatch_state *curyes = yes_state; | |
5703 | int curd = depth; | |
5704 | regmatch_slab *slab = PL_regmatch_slab; | |
5705 | for (;curd > -1;cur--,curd--) { | |
5706 | if (cur < SLAB_FIRST(slab)) { | |
5707 | slab = slab->prev; | |
5708 | cur = SLAB_LAST(slab); | |
5709 | } | |
5710 | PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n", | |
5711 | REPORT_CODE_OFF + 2 + depth * 2,"", | |
13d6edb4 | 5712 | curd, PL_reg_name[cur->resume_state], |
24b23f37 YO |
5713 | (curyes == cur) ? "yes" : "" |
5714 | ); | |
5715 | if (curyes == cur) | |
5716 | curyes = cur->u.yes.prev_yes_state; | |
5717 | } | |
5718 | } else | |
5719 | DEBUG_STATE_pp("push") | |
5720 | ); | |
40a82448 | 5721 | depth++; |
40a82448 DM |
5722 | st->locinput = locinput; |
5723 | newst = st+1; | |
5724 | if (newst > SLAB_LAST(PL_regmatch_slab)) | |
5725 | newst = S_push_slab(aTHX); | |
5726 | PL_regmatch_state = newst; | |
786e8c11 | 5727 | |
40a82448 DM |
5728 | locinput = PL_reginput; |
5729 | nextchr = UCHARAT(locinput); | |
5730 | st = newst; | |
5731 | continue; | |
5732 | /* NOTREACHED */ | |
5733 | } | |
a0d0e21e | 5734 | } |
a687059c | 5735 | |
a0d0e21e LW |
5736 | /* |
5737 | * We get here only if there's trouble -- normally "case END" is | |
5738 | * the terminating point. | |
5739 | */ | |
cea2e8a9 | 5740 | Perl_croak(aTHX_ "corrupted regexp pointers"); |
a0d0e21e | 5741 | /*NOTREACHED*/ |
4633a7c4 LW |
5742 | sayNO; |
5743 | ||
262b90c4 | 5744 | yes: |
77cb431f DM |
5745 | if (yes_state) { |
5746 | /* we have successfully completed a subexpression, but we must now | |
5747 | * pop to the state marked by yes_state and continue from there */ | |
77cb431f | 5748 | assert(st != yes_state); |
5bc10b2c DM |
5749 | #ifdef DEBUGGING |
5750 | while (st != yes_state) { | |
5751 | st--; | |
5752 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
5753 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5754 | st = SLAB_LAST(PL_regmatch_slab); | |
5755 | } | |
e2e6a0f1 | 5756 | DEBUG_STATE_r({ |
54612592 YO |
5757 | if (no_final) { |
5758 | DEBUG_STATE_pp("pop (no final)"); | |
5759 | } else { | |
5760 | DEBUG_STATE_pp("pop (yes)"); | |
5761 | } | |
e2e6a0f1 | 5762 | }); |
5bc10b2c DM |
5763 | depth--; |
5764 | } | |
5765 | #else | |
77cb431f DM |
5766 | while (yes_state < SLAB_FIRST(PL_regmatch_slab) |
5767 | || yes_state > SLAB_LAST(PL_regmatch_slab)) | |
5768 | { | |
5769 | /* not in this slab, pop slab */ | |
5770 | depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1); | |
5771 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5772 | st = SLAB_LAST(PL_regmatch_slab); | |
5773 | } | |
5774 | depth -= (st - yes_state); | |
5bc10b2c | 5775 | #endif |
77cb431f DM |
5776 | st = yes_state; |
5777 | yes_state = st->u.yes.prev_yes_state; | |
5778 | PL_regmatch_state = st; | |
24b23f37 | 5779 | |
5d458dd8 YO |
5780 | if (no_final) { |
5781 | locinput= st->locinput; | |
5782 | nextchr = UCHARAT(locinput); | |
5783 | } | |
54612592 | 5784 | state_num = st->resume_state + no_final; |
24d3c4a9 | 5785 | goto reenter_switch; |
77cb431f DM |
5786 | } |
5787 | ||
a3621e74 | 5788 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", |
e4584336 | 5789 | PL_colors[4], PL_colors[5])); |
02db2b7b | 5790 | |
19b95bf0 DM |
5791 | if (PL_reg_eval_set) { |
5792 | /* each successfully executed (?{...}) block does the equivalent of | |
5793 | * local $^R = do {...} | |
5794 | * When popping the save stack, all these locals would be undone; | |
5795 | * bypass this by setting the outermost saved $^R to the latest | |
5796 | * value */ | |
5797 | if (oreplsv != GvSV(PL_replgv)) | |
5798 | sv_setsv(oreplsv, GvSV(PL_replgv)); | |
5799 | } | |
95b24440 | 5800 | result = 1; |
aa283a38 | 5801 | goto final_exit; |
4633a7c4 LW |
5802 | |
5803 | no: | |
a3621e74 | 5804 | DEBUG_EXECUTE_r( |
7821416a | 5805 | PerlIO_printf(Perl_debug_log, |
786e8c11 | 5806 | "%*s %sfailed...%s\n", |
5bc10b2c | 5807 | REPORT_CODE_OFF+depth*2, "", |
786e8c11 | 5808 | PL_colors[4], PL_colors[5]) |
7821416a | 5809 | ); |
aa283a38 | 5810 | |
262b90c4 | 5811 | no_silent: |
54612592 YO |
5812 | if (no_final) { |
5813 | if (yes_state) { | |
5814 | goto yes; | |
5815 | } else { | |
5816 | goto final_exit; | |
5817 | } | |
5818 | } | |
aa283a38 DM |
5819 | if (depth) { |
5820 | /* there's a previous state to backtrack to */ | |
40a82448 DM |
5821 | st--; |
5822 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
5823 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5824 | st = SLAB_LAST(PL_regmatch_slab); | |
5825 | } | |
5826 | PL_regmatch_state = st; | |
40a82448 DM |
5827 | locinput= st->locinput; |
5828 | nextchr = UCHARAT(locinput); | |
5829 | ||
5bc10b2c DM |
5830 | DEBUG_STATE_pp("pop"); |
5831 | depth--; | |
262b90c4 DM |
5832 | if (yes_state == st) |
5833 | yes_state = st->u.yes.prev_yes_state; | |
5bc10b2c | 5834 | |
24d3c4a9 DM |
5835 | state_num = st->resume_state + 1; /* failure = success + 1 */ |
5836 | goto reenter_switch; | |
95b24440 | 5837 | } |
24d3c4a9 | 5838 | result = 0; |
aa283a38 | 5839 | |
262b90c4 | 5840 | final_exit: |
bbe252da | 5841 | if (rex->intflags & PREGf_VERBARG_SEEN) { |
5d458dd8 YO |
5842 | SV *sv_err = get_sv("REGERROR", 1); |
5843 | SV *sv_mrk = get_sv("REGMARK", 1); | |
5844 | if (result) { | |
e2e6a0f1 | 5845 | sv_commit = &PL_sv_no; |
5d458dd8 YO |
5846 | if (!sv_yes_mark) |
5847 | sv_yes_mark = &PL_sv_yes; | |
5848 | } else { | |
5849 | if (!sv_commit) | |
5850 | sv_commit = &PL_sv_yes; | |
5851 | sv_yes_mark = &PL_sv_no; | |
5852 | } | |
5853 | sv_setsv(sv_err, sv_commit); | |
5854 | sv_setsv(sv_mrk, sv_yes_mark); | |
e2e6a0f1 | 5855 | } |
19b95bf0 | 5856 | |
2f554ef7 DM |
5857 | /* clean up; in particular, free all slabs above current one */ |
5858 | LEAVE_SCOPE(oldsave); | |
5d9a96ca | 5859 | |
95b24440 | 5860 | return result; |
a687059c LW |
5861 | } |
5862 | ||
5863 | /* | |
5864 | - regrepeat - repeatedly match something simple, report how many | |
5865 | */ | |
5866 | /* | |
5867 | * [This routine now assumes that it will only match on things of length 1. | |
5868 | * That was true before, but now we assume scan - reginput is the count, | |
a0ed51b3 | 5869 | * rather than incrementing count on every character. [Er, except utf8.]] |
a687059c | 5870 | */ |
76e3520e | 5871 | STATIC I32 |
e2e6a0f1 | 5872 | S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth) |
a687059c | 5873 | { |
27da23d5 | 5874 | dVAR; |
a0d0e21e | 5875 | register char *scan; |
a0d0e21e | 5876 | register I32 c; |
3280af22 | 5877 | register char *loceol = PL_regeol; |
a0ed51b3 | 5878 | register I32 hardcount = 0; |
f2ed9b32 | 5879 | register bool utf8_target = PL_reg_match_utf8; |
d513472c | 5880 | UV utf8_flags; |
4f55667c SP |
5881 | #ifndef DEBUGGING |
5882 | PERL_UNUSED_ARG(depth); | |
5883 | #endif | |
a0d0e21e | 5884 | |
7918f24d NC |
5885 | PERL_ARGS_ASSERT_REGREPEAT; |
5886 | ||
3280af22 | 5887 | scan = PL_reginput; |
faf11cac HS |
5888 | if (max == REG_INFTY) |
5889 | max = I32_MAX; | |
5890 | else if (max < loceol - scan) | |
7f596f4c | 5891 | loceol = scan + max; |
a0d0e21e | 5892 | switch (OP(p)) { |
22c35a8c | 5893 | case REG_ANY: |
f2ed9b32 | 5894 | if (utf8_target) { |
ffc61ed2 | 5895 | loceol = PL_regeol; |
1aa99e6b | 5896 | while (scan < loceol && hardcount < max && *scan != '\n') { |
ffc61ed2 JH |
5897 | scan += UTF8SKIP(scan); |
5898 | hardcount++; | |
5899 | } | |
5900 | } else { | |
5901 | while (scan < loceol && *scan != '\n') | |
5902 | scan++; | |
a0ed51b3 LW |
5903 | } |
5904 | break; | |
ffc61ed2 | 5905 | case SANY: |
f2ed9b32 | 5906 | if (utf8_target) { |
def8e4ea | 5907 | loceol = PL_regeol; |
a0804c9e | 5908 | while (scan < loceol && hardcount < max) { |
def8e4ea JH |
5909 | scan += UTF8SKIP(scan); |
5910 | hardcount++; | |
5911 | } | |
5912 | } | |
5913 | else | |
5914 | scan = loceol; | |
a0ed51b3 | 5915 | break; |
f33976b4 DB |
5916 | case CANY: |
5917 | scan = loceol; | |
5918 | break; | |
59d32103 KW |
5919 | case EXACT: |
5920 | /* To get here, EXACTish nodes must have *byte* length == 1. That | |
5921 | * means they match only characters in the string that can be expressed | |
5922 | * as a single byte. For non-utf8 strings, that means a simple match. | |
5923 | * For utf8 strings, the character matched must be an invariant, or | |
5924 | * downgradable to a single byte. The pattern's utf8ness is | |
5925 | * irrelevant, as since it's a single byte, it either isn't utf8, or if | |
5926 | * it is, it's an invariant */ | |
5927 | ||
5928 | c = (U8)*STRING(p); | |
5929 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); | |
5930 | ||
5931 | if (! utf8_target || UNI_IS_INVARIANT(c)) { | |
5932 | while (scan < loceol && UCHARAT(scan) == c) { | |
5933 | scan++; | |
5934 | } | |
5935 | } | |
5936 | else { | |
5937 | ||
5938 | /* Here, the string is utf8, and the pattern char is different | |
5939 | * in utf8 than not, so can't compare them directly. Outside the | |
5940 | * loop, find find the two utf8 bytes that represent c, and then | |
5941 | * look for those in sequence in the utf8 string */ | |
5942 | U8 high = UTF8_TWO_BYTE_HI(c); | |
5943 | U8 low = UTF8_TWO_BYTE_LO(c); | |
5944 | loceol = PL_regeol; | |
5945 | ||
5946 | while (hardcount < max | |
5947 | && scan + 1 < loceol | |
5948 | && UCHARAT(scan) == high | |
5949 | && UCHARAT(scan + 1) == low) | |
5950 | { | |
5951 | scan += 2; | |
5952 | hardcount++; | |
5953 | } | |
5954 | } | |
5955 | break; | |
2f7f8cb1 KW |
5956 | case EXACTFA: |
5957 | utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
5958 | goto do_exactf; | |
5959 | ||
d4e0b827 KW |
5960 | case EXACTFL: |
5961 | PL_reg_flags |= RF_tainted; | |
17580e7a KW |
5962 | utf8_flags = FOLDEQ_UTF8_LOCALE; |
5963 | goto do_exactf; | |
5964 | ||
d4e0b827 | 5965 | case EXACTF: |
9a5a5549 | 5966 | case EXACTFU: |
d513472c | 5967 | utf8_flags = 0; |
59d32103 | 5968 | |
2be3e190 KW |
5969 | /* The comments for the EXACT case above apply as well to these fold |
5970 | * ones */ | |
634c83a2 | 5971 | |
2f7f8cb1 | 5972 | do_exactf: |
090f7165 | 5973 | c = (U8)*STRING(p); |
634c83a2 | 5974 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); |
d4e0b827 | 5975 | |
59d32103 | 5976 | if (utf8_target) { /* Use full Unicode fold matching */ |
59d32103 KW |
5977 | char *tmpeol = loceol; |
5978 | while (hardcount < max | |
d513472c KW |
5979 | && foldEQ_utf8_flags(scan, &tmpeol, 0, utf8_target, |
5980 | STRING(p), NULL, 1, cBOOL(UTF_PATTERN), utf8_flags)) | |
59d32103 KW |
5981 | { |
5982 | scan = tmpeol; | |
5983 | tmpeol = loceol; | |
5984 | hardcount++; | |
5985 | } | |
634c83a2 | 5986 | |
59d32103 KW |
5987 | /* XXX Note that the above handles properly the German sharp s in |
5988 | * the pattern matching ss in the string. But it doesn't handle | |
5989 | * properly cases where the string contains say 'LIGATURE ff' and | |
5990 | * the pattern is 'f+'. This would require, say, a new function or | |
5991 | * revised interface to foldEQ_utf8(), in which the maximum number | |
5992 | * of characters to match could be passed and it would return how | |
5993 | * many actually did. This is just one of many cases where | |
5994 | * multi-char folds don't work properly, and so the fix is being | |
5995 | * deferred */ | |
5996 | } | |
5997 | else { | |
87381386 | 5998 | U8 folded; |
59d32103 | 5999 | |
2be3e190 KW |
6000 | /* Here, the string isn't utf8 and c is a single byte; and either |
6001 | * the pattern isn't utf8 or c is an invariant, so its utf8ness | |
6002 | * doesn't affect c. Can just do simple comparisons for exact or | |
6003 | * fold matching. */ | |
d4e0b827 | 6004 | switch (OP(p)) { |
87381386 | 6005 | case EXACTF: folded = PL_fold[c]; break; |
2f7f8cb1 | 6006 | case EXACTFA: |
9a5a5549 | 6007 | case EXACTFU: folded = PL_fold_latin1[c]; break; |
87381386 KW |
6008 | case EXACTFL: folded = PL_fold_locale[c]; break; |
6009 | default: Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p)); | |
6010 | } | |
6011 | while (scan < loceol && | |
6012 | (UCHARAT(scan) == c || UCHARAT(scan) == folded)) | |
6013 | { | |
6014 | scan++; | |
634c83a2 KW |
6015 | } |
6016 | } | |
bbce6d69 | 6017 | break; |
f56b6394 | 6018 | case ANYOFV: |
a0d0e21e | 6019 | case ANYOF: |
4e8910e0 KW |
6020 | if (utf8_target || OP(p) == ANYOFV) { |
6021 | STRLEN inclasslen; | |
ffc61ed2 | 6022 | loceol = PL_regeol; |
4e8910e0 KW |
6023 | inclasslen = loceol - scan; |
6024 | while (hardcount < max | |
6025 | && ((inclasslen = loceol - scan) > 0) | |
6026 | && reginclass(prog, p, (U8*)scan, &inclasslen, utf8_target)) | |
6027 | { | |
6028 | scan += inclasslen; | |
ffc61ed2 JH |
6029 | hardcount++; |
6030 | } | |
6031 | } else { | |
32fc9b6a | 6032 | while (scan < loceol && REGINCLASS(prog, p, (U8*)scan)) |
ffc61ed2 JH |
6033 | scan++; |
6034 | } | |
a0d0e21e | 6035 | break; |
980866de | 6036 | case ALNUMU: |
f2ed9b32 | 6037 | if (utf8_target) { |
980866de | 6038 | utf8_wordchar: |
ffc61ed2 | 6039 | loceol = PL_regeol; |
1a4fad37 | 6040 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6041 | while (hardcount < max && scan < loceol && |
a12cf05f KW |
6042 | swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
6043 | { | |
ffc61ed2 JH |
6044 | scan += UTF8SKIP(scan); |
6045 | hardcount++; | |
6046 | } | |
980866de | 6047 | } else { |
a12cf05f KW |
6048 | while (scan < loceol && isWORDCHAR_L1((U8) *scan)) { |
6049 | scan++; | |
6050 | } | |
980866de KW |
6051 | } |
6052 | break; | |
6053 | case ALNUM: | |
6054 | if (utf8_target) | |
6055 | goto utf8_wordchar; | |
6056 | while (scan < loceol && isALNUM((U8) *scan)) { | |
6057 | scan++; | |
a0ed51b3 LW |
6058 | } |
6059 | break; | |
cfaf538b KW |
6060 | case ALNUMA: |
6061 | while (scan < loceol && isWORDCHAR_A((U8) *scan)) { | |
6062 | scan++; | |
6063 | } | |
6064 | break; | |
bbce6d69 | 6065 | case ALNUML: |
3280af22 | 6066 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6067 | if (utf8_target) { |
ffc61ed2 | 6068 | loceol = PL_regeol; |
1aa99e6b IH |
6069 | while (hardcount < max && scan < loceol && |
6070 | isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6071 | scan += UTF8SKIP(scan); |
6072 | hardcount++; | |
6073 | } | |
6074 | } else { | |
6075 | while (scan < loceol && isALNUM_LC(*scan)) | |
6076 | scan++; | |
a0ed51b3 LW |
6077 | } |
6078 | break; | |
980866de | 6079 | case NALNUMU: |
f2ed9b32 | 6080 | if (utf8_target) { |
980866de KW |
6081 | |
6082 | utf8_Nwordchar: | |
6083 | ||
ffc61ed2 | 6084 | loceol = PL_regeol; |
1a4fad37 | 6085 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6086 | while (hardcount < max && scan < loceol && |
980866de | 6087 | ! swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
a12cf05f | 6088 | { |
ffc61ed2 JH |
6089 | scan += UTF8SKIP(scan); |
6090 | hardcount++; | |
6091 | } | |
980866de | 6092 | } else { |
a12cf05f KW |
6093 | while (scan < loceol && ! isWORDCHAR_L1((U8) *scan)) { |
6094 | scan++; | |
6095 | } | |
980866de KW |
6096 | } |
6097 | break; | |
6098 | case NALNUM: | |
6099 | if (utf8_target) | |
6100 | goto utf8_Nwordchar; | |
6101 | while (scan < loceol && ! isALNUM((U8) *scan)) { | |
6102 | scan++; | |
a0ed51b3 LW |
6103 | } |
6104 | break; | |
cfaf538b KW |
6105 | case NALNUMA: |
6106 | if (utf8_target) { | |
6107 | while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) { | |
6108 | scan += UTF8SKIP(scan); | |
6109 | } | |
6110 | } | |
6111 | else { | |
6112 | while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) { | |
6113 | scan++; | |
6114 | } | |
6115 | } | |
6116 | break; | |
bbce6d69 | 6117 | case NALNUML: |
3280af22 | 6118 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6119 | if (utf8_target) { |
ffc61ed2 | 6120 | loceol = PL_regeol; |
1aa99e6b IH |
6121 | while (hardcount < max && scan < loceol && |
6122 | !isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6123 | scan += UTF8SKIP(scan); |
6124 | hardcount++; | |
6125 | } | |
6126 | } else { | |
6127 | while (scan < loceol && !isALNUM_LC(*scan)) | |
6128 | scan++; | |
a0ed51b3 LW |
6129 | } |
6130 | break; | |
980866de | 6131 | case SPACEU: |
f2ed9b32 | 6132 | if (utf8_target) { |
980866de KW |
6133 | |
6134 | utf8_space: | |
6135 | ||
ffc61ed2 | 6136 | loceol = PL_regeol; |
1a4fad37 | 6137 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6138 | while (hardcount < max && scan < loceol && |
3568d838 | 6139 | (*scan == ' ' || |
a12cf05f KW |
6140 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) |
6141 | { | |
ffc61ed2 JH |
6142 | scan += UTF8SKIP(scan); |
6143 | hardcount++; | |
6144 | } | |
980866de KW |
6145 | break; |
6146 | } | |
6147 | else { | |
a12cf05f KW |
6148 | while (scan < loceol && isSPACE_L1((U8) *scan)) { |
6149 | scan++; | |
6150 | } | |
980866de KW |
6151 | break; |
6152 | } | |
6153 | case SPACE: | |
6154 | if (utf8_target) | |
6155 | goto utf8_space; | |
6156 | ||
6157 | while (scan < loceol && isSPACE((U8) *scan)) { | |
6158 | scan++; | |
a0ed51b3 LW |
6159 | } |
6160 | break; | |
cfaf538b KW |
6161 | case SPACEA: |
6162 | while (scan < loceol && isSPACE_A((U8) *scan)) { | |
6163 | scan++; | |
6164 | } | |
6165 | break; | |
bbce6d69 | 6166 | case SPACEL: |
3280af22 | 6167 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6168 | if (utf8_target) { |
ffc61ed2 | 6169 | loceol = PL_regeol; |
1aa99e6b | 6170 | while (hardcount < max && scan < loceol && |
6bbba904 | 6171 | isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6172 | scan += UTF8SKIP(scan); |
6173 | hardcount++; | |
6174 | } | |
6175 | } else { | |
6176 | while (scan < loceol && isSPACE_LC(*scan)) | |
6177 | scan++; | |
a0ed51b3 LW |
6178 | } |
6179 | break; | |
980866de | 6180 | case NSPACEU: |
f2ed9b32 | 6181 | if (utf8_target) { |
980866de KW |
6182 | |
6183 | utf8_Nspace: | |
6184 | ||
ffc61ed2 | 6185 | loceol = PL_regeol; |
1a4fad37 | 6186 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6187 | while (hardcount < max && scan < loceol && |
980866de KW |
6188 | ! (*scan == ' ' || |
6189 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) | |
a12cf05f | 6190 | { |
ffc61ed2 JH |
6191 | scan += UTF8SKIP(scan); |
6192 | hardcount++; | |
6193 | } | |
980866de KW |
6194 | break; |
6195 | } | |
6196 | else { | |
a12cf05f KW |
6197 | while (scan < loceol && ! isSPACE_L1((U8) *scan)) { |
6198 | scan++; | |
6199 | } | |
980866de KW |
6200 | } |
6201 | break; | |
6202 | case NSPACE: | |
6203 | if (utf8_target) | |
6204 | goto utf8_Nspace; | |
6205 | ||
6206 | while (scan < loceol && ! isSPACE((U8) *scan)) { | |
6207 | scan++; | |
a0ed51b3 | 6208 | } |
0008a298 | 6209 | break; |
cfaf538b KW |
6210 | case NSPACEA: |
6211 | if (utf8_target) { | |
6212 | while (scan < loceol && ! isSPACE_A((U8) *scan)) { | |
6213 | scan += UTF8SKIP(scan); | |
6214 | } | |
6215 | } | |
6216 | else { | |
6217 | while (scan < loceol && ! isSPACE_A((U8) *scan)) { | |
6218 | scan++; | |
6219 | } | |
6220 | } | |
6221 | break; | |
bbce6d69 | 6222 | case NSPACEL: |
3280af22 | 6223 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6224 | if (utf8_target) { |
ffc61ed2 | 6225 | loceol = PL_regeol; |
1aa99e6b | 6226 | while (hardcount < max && scan < loceol && |
6bbba904 | 6227 | !isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6228 | scan += UTF8SKIP(scan); |
6229 | hardcount++; | |
6230 | } | |
6231 | } else { | |
6232 | while (scan < loceol && !isSPACE_LC(*scan)) | |
6233 | scan++; | |
a0ed51b3 LW |
6234 | } |
6235 | break; | |
a0d0e21e | 6236 | case DIGIT: |
f2ed9b32 | 6237 | if (utf8_target) { |
ffc61ed2 | 6238 | loceol = PL_regeol; |
1a4fad37 | 6239 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6240 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6241 | swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6242 | scan += UTF8SKIP(scan); |
6243 | hardcount++; | |
6244 | } | |
6245 | } else { | |
6246 | while (scan < loceol && isDIGIT(*scan)) | |
6247 | scan++; | |
a0ed51b3 LW |
6248 | } |
6249 | break; | |
cfaf538b KW |
6250 | case DIGITA: |
6251 | while (scan < loceol && isDIGIT_A((U8) *scan)) { | |
6252 | scan++; | |
6253 | } | |
6254 | break; | |
b77393f6 KW |
6255 | case DIGITL: |
6256 | PL_reg_flags |= RF_tainted; | |
6257 | if (utf8_target) { | |
6258 | loceol = PL_regeol; | |
6259 | while (hardcount < max && scan < loceol && | |
6260 | isDIGIT_LC_utf8((U8*)scan)) { | |
6261 | scan += UTF8SKIP(scan); | |
6262 | hardcount++; | |
6263 | } | |
6264 | } else { | |
6265 | while (scan < loceol && isDIGIT_LC(*scan)) | |
6266 | scan++; | |
6267 | } | |
6268 | break; | |
a0d0e21e | 6269 | case NDIGIT: |
f2ed9b32 | 6270 | if (utf8_target) { |
ffc61ed2 | 6271 | loceol = PL_regeol; |
1a4fad37 | 6272 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6273 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6274 | !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6275 | scan += UTF8SKIP(scan); |
6276 | hardcount++; | |
6277 | } | |
6278 | } else { | |
6279 | while (scan < loceol && !isDIGIT(*scan)) | |
6280 | scan++; | |
a0ed51b3 | 6281 | } |
cfaf538b KW |
6282 | break; |
6283 | case NDIGITA: | |
6284 | if (utf8_target) { | |
6285 | while (scan < loceol && ! isDIGIT_A((U8) *scan)) { | |
6286 | scan += UTF8SKIP(scan); | |
6287 | } | |
6288 | } | |
6289 | else { | |
6290 | while (scan < loceol && ! isDIGIT_A((U8) *scan)) { | |
6291 | scan++; | |
6292 | } | |
6293 | } | |
6294 | break; | |
b77393f6 KW |
6295 | case NDIGITL: |
6296 | PL_reg_flags |= RF_tainted; | |
6297 | if (utf8_target) { | |
6298 | loceol = PL_regeol; | |
6299 | while (hardcount < max && scan < loceol && | |
6300 | !isDIGIT_LC_utf8((U8*)scan)) { | |
6301 | scan += UTF8SKIP(scan); | |
6302 | hardcount++; | |
6303 | } | |
6304 | } else { | |
6305 | while (scan < loceol && !isDIGIT_LC(*scan)) | |
6306 | scan++; | |
6307 | } | |
6308 | break; | |
e1d1eefb | 6309 | case LNBREAK: |
f2ed9b32 | 6310 | if (utf8_target) { |
e1d1eefb YO |
6311 | loceol = PL_regeol; |
6312 | while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) { | |
6313 | scan += c; | |
6314 | hardcount++; | |
6315 | } | |
6316 | } else { | |
6317 | /* | |
6318 | LNBREAK can match two latin chars, which is ok, | |
6319 | because we have a null terminated string, but we | |
6320 | have to use hardcount in this situation | |
6321 | */ | |
6322 | while (scan < loceol && (c=is_LNBREAK_latin1(scan))) { | |
6323 | scan+=c; | |
6324 | hardcount++; | |
6325 | } | |
6326 | } | |
6327 | break; | |
6328 | case HORIZWS: | |
f2ed9b32 | 6329 | if (utf8_target) { |
e1d1eefb YO |
6330 | loceol = PL_regeol; |
6331 | while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) { | |
6332 | scan += c; | |
6333 | hardcount++; | |
6334 | } | |
6335 | } else { | |
6336 | while (scan < loceol && is_HORIZWS_latin1(scan)) | |
6337 | scan++; | |
6338 | } | |
a0ed51b3 | 6339 | break; |
e1d1eefb | 6340 | case NHORIZWS: |
f2ed9b32 | 6341 | if (utf8_target) { |
e1d1eefb YO |
6342 | loceol = PL_regeol; |
6343 | while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) { | |
6344 | scan += UTF8SKIP(scan); | |
6345 | hardcount++; | |
6346 | } | |
6347 | } else { | |
6348 | while (scan < loceol && !is_HORIZWS_latin1(scan)) | |
6349 | scan++; | |
6350 | ||
6351 | } | |
6352 | break; | |
6353 | case VERTWS: | |
f2ed9b32 | 6354 | if (utf8_target) { |
e1d1eefb YO |
6355 | loceol = PL_regeol; |
6356 | while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) { | |
6357 | scan += c; | |
6358 | hardcount++; | |
6359 | } | |
6360 | } else { | |
6361 | while (scan < loceol && is_VERTWS_latin1(scan)) | |
6362 | scan++; | |
6363 | ||
6364 | } | |
6365 | break; | |
6366 | case NVERTWS: | |
f2ed9b32 | 6367 | if (utf8_target) { |
e1d1eefb YO |
6368 | loceol = PL_regeol; |
6369 | while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) { | |
6370 | scan += UTF8SKIP(scan); | |
6371 | hardcount++; | |
6372 | } | |
6373 | } else { | |
6374 | while (scan < loceol && !is_VERTWS_latin1(scan)) | |
6375 | scan++; | |
6376 | ||
6377 | } | |
6378 | break; | |
6379 | ||
a0d0e21e LW |
6380 | default: /* Called on something of 0 width. */ |
6381 | break; /* So match right here or not at all. */ | |
6382 | } | |
a687059c | 6383 | |
a0ed51b3 LW |
6384 | if (hardcount) |
6385 | c = hardcount; | |
6386 | else | |
6387 | c = scan - PL_reginput; | |
3280af22 | 6388 | PL_reginput = scan; |
a687059c | 6389 | |
a3621e74 | 6390 | DEBUG_r({ |
e68ec53f | 6391 | GET_RE_DEBUG_FLAGS_DECL; |
be8e71aa | 6392 | DEBUG_EXECUTE_r({ |
e68ec53f YO |
6393 | SV * const prop = sv_newmortal(); |
6394 | regprop(prog, prop, p); | |
6395 | PerlIO_printf(Perl_debug_log, | |
be8e71aa | 6396 | "%*s %s can match %"IVdf" times out of %"IVdf"...\n", |
e2e6a0f1 | 6397 | REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max); |
a3621e74 | 6398 | }); |
be8e71aa | 6399 | }); |
9041c2e3 | 6400 | |
a0d0e21e | 6401 | return(c); |
a687059c LW |
6402 | } |
6403 | ||
c277df42 | 6404 | |
be8e71aa | 6405 | #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) |
c277df42 | 6406 | /* |
ffc61ed2 JH |
6407 | - regclass_swash - prepare the utf8 swash |
6408 | */ | |
6409 | ||
6410 | SV * | |
32fc9b6a | 6411 | Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) |
ffc61ed2 | 6412 | { |
97aff369 | 6413 | dVAR; |
9e55ce06 JH |
6414 | SV *sw = NULL; |
6415 | SV *si = NULL; | |
6416 | SV *alt = NULL; | |
f8fc2ecf YO |
6417 | RXi_GET_DECL(prog,progi); |
6418 | const struct reg_data * const data = prog ? progi->data : NULL; | |
ffc61ed2 | 6419 | |
7918f24d NC |
6420 | PERL_ARGS_ASSERT_REGCLASS_SWASH; |
6421 | ||
ccb2541c KW |
6422 | assert(ANYOF_NONBITMAP(node)); |
6423 | ||
4f639d21 | 6424 | if (data && data->count) { |
a3b680e6 | 6425 | const U32 n = ARG(node); |
ffc61ed2 | 6426 | |
4f639d21 | 6427 | if (data->what[n] == 's') { |
ad64d0ec NC |
6428 | SV * const rv = MUTABLE_SV(data->data[n]); |
6429 | AV * const av = MUTABLE_AV(SvRV(rv)); | |
2d03de9c | 6430 | SV **const ary = AvARRAY(av); |
9e55ce06 | 6431 | SV **a, **b; |
9041c2e3 | 6432 | |
711a919c | 6433 | /* See the end of regcomp.c:S_regclass() for |
9e55ce06 JH |
6434 | * documentation of these array elements. */ |
6435 | ||
b11f357e | 6436 | si = *ary; |
fe5bfecd JH |
6437 | a = SvROK(ary[1]) ? &ary[1] : NULL; |
6438 | b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : NULL; | |
b11f357e | 6439 | |
ffc61ed2 JH |
6440 | if (a) |
6441 | sw = *a; | |
6442 | else if (si && doinit) { | |
6443 | sw = swash_init("utf8", "", si, 1, 0); | |
6444 | (void)av_store(av, 1, sw); | |
6445 | } | |
9e55ce06 JH |
6446 | if (b) |
6447 | alt = *b; | |
ffc61ed2 JH |
6448 | } |
6449 | } | |
6450 | ||
9e55ce06 JH |
6451 | if (listsvp) |
6452 | *listsvp = si; | |
6453 | if (altsvp) | |
6454 | *altsvp = alt; | |
ffc61ed2 JH |
6455 | |
6456 | return sw; | |
6457 | } | |
76234dfb | 6458 | #endif |
ffc61ed2 JH |
6459 | |
6460 | /* | |
ba7b4546 | 6461 | - reginclass - determine if a character falls into a character class |
832705d4 | 6462 | |
6698fab5 KW |
6463 | n is the ANYOF regnode |
6464 | p is the target string | |
6465 | lenp is pointer to the maximum number of bytes of how far to go in p | |
6466 | (This is assumed wthout checking to always be at least the current | |
6467 | character's size) | |
6468 | utf8_target tells whether p is in UTF-8. | |
832705d4 | 6469 | |
4b3cda86 KW |
6470 | Returns true if matched; false otherwise. If lenp is not NULL, on return |
6471 | from a successful match, the value it points to will be updated to how many | |
6472 | bytes in p were matched. If there was no match, the value is undefined, | |
6473 | possibly changed from the input. | |
eba1359e | 6474 | |
d5788240 KW |
6475 | Note that this can be a synthetic start class, a combination of various |
6476 | nodes, so things you think might be mutually exclusive, such as locale, | |
6477 | aren't. It can match both locale and non-locale | |
6478 | ||
bbce6d69 | 6479 | */ |
6480 | ||
76e3520e | 6481 | STATIC bool |
f6ad78d8 | 6482 | 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 | 6483 | { |
27da23d5 | 6484 | dVAR; |
a3b680e6 | 6485 | const char flags = ANYOF_FLAGS(n); |
bbce6d69 | 6486 | bool match = FALSE; |
cc07378b | 6487 | UV c = *p; |
f7ab54c6 | 6488 | STRLEN c_len = 0; |
6698fab5 | 6489 | STRLEN maxlen; |
1aa99e6b | 6490 | |
7918f24d NC |
6491 | PERL_ARGS_ASSERT_REGINCLASS; |
6492 | ||
4b3cda86 | 6493 | /* If c is not already the code point, get it */ |
f2ed9b32 | 6494 | if (utf8_target && !UTF8_IS_INVARIANT(c)) { |
f7ab54c6 | 6495 | c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &c_len, |
6182169b KW |
6496 | (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) |
6497 | | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY); | |
6498 | /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for | |
6499 | * UTF8_ALLOW_FFFF */ | |
f7ab54c6 | 6500 | if (c_len == (STRLEN)-1) |
e8a70c6f | 6501 | Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); |
19f67299 | 6502 | } |
a5a291f5 KW |
6503 | else { |
6504 | c_len = 1; | |
6505 | } | |
bbce6d69 | 6506 | |
a5a291f5 KW |
6507 | /* Use passed in max length, or one character if none passed in or less |
6508 | * than one character. And assume will match just one character. This is | |
6509 | * overwritten later if matched more. */ | |
4b3cda86 | 6510 | if (lenp) { |
a5a291f5 KW |
6511 | maxlen = (*lenp > c_len) ? *lenp : c_len; |
6512 | *lenp = c_len; | |
4b3cda86 KW |
6513 | |
6514 | } | |
6515 | else { | |
a5a291f5 | 6516 | maxlen = c_len; |
4b3cda86 KW |
6517 | } |
6518 | ||
7cdde544 KW |
6519 | /* If this character is potentially in the bitmap, check it */ |
6520 | if (c < 256) { | |
ffc61ed2 JH |
6521 | if (ANYOF_BITMAP_TEST(n, c)) |
6522 | match = TRUE; | |
11454c59 KW |
6523 | else if (flags & ANYOF_NON_UTF8_LATIN1_ALL |
6524 | && ! utf8_target | |
6525 | && ! isASCII(c)) | |
6526 | { | |
6527 | match = TRUE; | |
6528 | } | |
a0ed51b3 | 6529 | |
78969a98 KW |
6530 | else if (flags & ANYOF_LOCALE) { |
6531 | PL_reg_flags |= RF_tainted; | |
6532 | ||
6533 | if ((flags & ANYOF_LOC_NONBITMAP_FOLD) | |
6534 | && ANYOF_BITMAP_TEST(n, PL_fold_locale[c])) | |
6535 | { | |
ffc61ed2 | 6536 | match = TRUE; |
78969a98 | 6537 | } |
040aea3a KW |
6538 | else if (ANYOF_CLASS_TEST_ANY_SET(n) && |
6539 | ((ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) || | |
6540 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) || | |
6541 | (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) || | |
6542 | (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) || | |
6543 | (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) || | |
6544 | (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) || | |
6545 | (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) || | |
6546 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) || | |
6547 | (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) || | |
6548 | (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) || | |
6549 | (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) || | |
6550 | (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) || | |
6551 | (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) || | |
6552 | (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) || | |
6553 | (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) || | |
6554 | (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) || | |
6555 | (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) || | |
6556 | (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) || | |
6557 | (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) || | |
6558 | (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) || | |
6559 | (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) || | |
6560 | (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) || | |
6561 | (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) || | |
6562 | (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) || | |
6563 | (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) || | |
6564 | (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) || | |
6565 | (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) || | |
6566 | (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) || | |
6567 | (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) || | |
6568 | (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c)) | |
6569 | ) /* How's that for a conditional? */ | |
78969a98 | 6570 | ) { |
ffc61ed2 JH |
6571 | match = TRUE; |
6572 | } | |
a0ed51b3 | 6573 | } |
a0ed51b3 LW |
6574 | } |
6575 | ||
7cdde544 | 6576 | /* If the bitmap didn't (or couldn't) match, and something outside the |
de87c4fe KW |
6577 | * bitmap could match, try that. Locale nodes specifiy completely the |
6578 | * behavior of code points in the bit map (otherwise, a utf8 target would | |
c613755a | 6579 | * cause them to be treated as Unicode and not locale), except in |
de87c4fe | 6580 | * the very unlikely event when this node is a synthetic start class, which |
c613755a KW |
6581 | * could be a combination of locale and non-locale nodes. So allow locale |
6582 | * to match for the synthetic start class, which will give a false | |
6583 | * positive that will be resolved when the match is done again as not part | |
6584 | * of the synthetic start class */ | |
ef87b810 | 6585 | if (!match) { |
10ee90d2 KW |
6586 | if (utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256) { |
6587 | match = TRUE; /* Everything above 255 matches */ | |
e051a21d | 6588 | } |
6f8d7d0d KW |
6589 | else if (ANYOF_NONBITMAP(n) |
6590 | && ((flags & ANYOF_NONBITMAP_NON_UTF8) | |
6591 | || (utf8_target | |
6592 | && (c >=256 | |
6593 | || (! (flags & ANYOF_LOCALE)) | |
6594 | || (flags & ANYOF_IS_SYNTHETIC))))) | |
ef87b810 | 6595 | { |
7cdde544 KW |
6596 | AV *av; |
6597 | SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av); | |
6598 | ||
6599 | if (sw) { | |
6600 | U8 * utf8_p; | |
6601 | if (utf8_target) { | |
6602 | utf8_p = (U8 *) p; | |
6603 | } else { | |
f56b6394 KW |
6604 | |
6605 | /* Not utf8. Convert as much of the string as available up | |
6606 | * to the limit of how far the (single) character in the | |
6607 | * pattern can possibly match (no need to go further). If | |
6608 | * the node is a straight ANYOF or not folding, it can't | |
6609 | * match more than one. Otherwise, It can match up to how | |
6610 | * far a single char can fold to. Since not utf8, each | |
6611 | * character is a single byte, so the max it can be in | |
6612 | * bytes is the same as the max it can be in characters */ | |
6613 | STRLEN len = (OP(n) == ANYOF | |
6614 | || ! (flags & ANYOF_LOC_NONBITMAP_FOLD)) | |
6615 | ? 1 | |
6616 | : (maxlen < UTF8_MAX_FOLD_CHAR_EXPAND) | |
6617 | ? maxlen | |
6618 | : UTF8_MAX_FOLD_CHAR_EXPAND; | |
7cdde544 KW |
6619 | utf8_p = bytes_to_utf8(p, &len); |
6620 | } | |
f56b6394 | 6621 | |
f7eb7452 | 6622 | if (swash_fetch(sw, utf8_p, TRUE)) |
7cdde544 | 6623 | match = TRUE; |
39065660 | 6624 | else if (flags & ANYOF_LOC_NONBITMAP_FOLD) { |
f56b6394 KW |
6625 | |
6626 | /* Here, we need to test if the fold of the target string | |
8d5d17fa KW |
6627 | * matches. The non-multi char folds have all been moved to |
6628 | * the compilation phase, and the multi-char folds have | |
6629 | * been stored by regcomp into 'av'; we linearly check to | |
6630 | * see if any match the target string (folded). We know | |
6631 | * that the originals were each one character, but we don't | |
6632 | * currently know how many characters/bytes each folded to, | |
6633 | * except we do know that there are small limits imposed by | |
6634 | * Unicode. XXX A performance enhancement would be to have | |
6635 | * regcomp.c store the max number of chars/bytes that are | |
6636 | * in an av entry, as, say the 0th element. Even better | |
6637 | * would be to have a hash of the few characters that can | |
6638 | * start a multi-char fold to the max number of chars of | |
6639 | * those folds. | |
f56b6394 KW |
6640 | * |
6641 | * If there is a match, we will need to advance (if lenp is | |
6642 | * specified) the match pointer in the target string. But | |
6643 | * what we are comparing here isn't that string directly, | |
6644 | * but its fold, whose length may differ from the original. | |
6645 | * As we go along in constructing the fold, therefore, we | |
6646 | * create a map so that we know how many bytes in the | |
6647 | * source to advance given that we have matched a certain | |
6648 | * number of bytes in the fold. This map is stored in | |
6438af90 KW |
6649 | * 'map_fold_len_back'. Let n mean the number of bytes in |
6650 | * the fold of the first character that we are folding. | |
6651 | * Then map_fold_len_back[n] is set to the number of bytes | |
6652 | * in that first character. Similarly let m be the | |
6653 | * corresponding number for the second character to be | |
6654 | * folded. Then map_fold_len_back[n+m] is set to the | |
6655 | * number of bytes occupied by the first two source | |
6656 | * characters. ... */ | |
6657 | U8 map_fold_len_back[UTF8_MAXBYTES_CASE+1] = { 0 }; | |
f56b6394 KW |
6658 | U8 folded[UTF8_MAXBYTES_CASE+1]; |
6659 | STRLEN foldlen = 0; /* num bytes in fold of 1st char */ | |
deba3d96 | 6660 | STRLEN total_foldlen = 0; /* num bytes in fold of all |
6438af90 | 6661 | chars */ |
f56b6394 KW |
6662 | |
6663 | if (OP(n) == ANYOF || maxlen == 1 || ! lenp || ! av) { | |
6664 | ||
6665 | /* Here, only need to fold the first char of the target | |
6438af90 | 6666 | * string. It the source wasn't utf8, is 1 byte long */ |
f56b6394 | 6667 | to_utf8_fold(utf8_p, folded, &foldlen); |
deba3d96 | 6668 | total_foldlen = foldlen; |
6438af90 KW |
6669 | map_fold_len_back[foldlen] = (utf8_target) |
6670 | ? UTF8SKIP(utf8_p) | |
6671 | : 1; | |
f56b6394 KW |
6672 | } |
6673 | else { | |
6674 | ||
6675 | /* Here, need to fold more than the first char. Do so | |
6676 | * up to the limits */ | |
f56b6394 KW |
6677 | U8* source_ptr = utf8_p; /* The source for the fold |
6678 | is the regex target | |
6679 | string */ | |
6680 | U8* folded_ptr = folded; | |
6681 | U8* e = utf8_p + maxlen; /* Can't go beyond last | |
6682 | available byte in the | |
6683 | target string */ | |
6438af90 KW |
6684 | U8 i; |
6685 | for (i = 0; | |
6686 | i < UTF8_MAX_FOLD_CHAR_EXPAND && source_ptr < e; | |
6687 | i++) | |
f56b6394 KW |
6688 | { |
6689 | ||
6690 | /* Fold the next character */ | |
6691 | U8 this_char_folded[UTF8_MAXBYTES_CASE+1]; | |
6692 | STRLEN this_char_foldlen; | |
6693 | to_utf8_fold(source_ptr, | |
6694 | this_char_folded, | |
6695 | &this_char_foldlen); | |
6696 | ||
6697 | /* Bail if it would exceed the byte limit for | |
6698 | * folding a single char. */ | |
6699 | if (this_char_foldlen + folded_ptr - folded > | |
6700 | UTF8_MAXBYTES_CASE) | |
6701 | { | |
6702 | break; | |
6703 | } | |
6704 | ||
6438af90 | 6705 | /* Add the fold of this character */ |
f56b6394 KW |
6706 | Copy(this_char_folded, |
6707 | folded_ptr, | |
6708 | this_char_foldlen, | |
6709 | U8); | |
f56b6394 | 6710 | source_ptr += UTF8SKIP(source_ptr); |
6438af90 | 6711 | folded_ptr += this_char_foldlen; |
deba3d96 | 6712 | total_foldlen = folded_ptr - folded; |
6438af90 KW |
6713 | |
6714 | /* Create map from the number of bytes in the fold | |
6715 | * back to the number of bytes in the source. If | |
6716 | * the source isn't utf8, the byte count is just | |
6717 | * the number of characters so far */ | |
deba3d96 | 6718 | map_fold_len_back[total_foldlen] |
6438af90 KW |
6719 | = (utf8_target) |
6720 | ? source_ptr - utf8_p | |
6721 | : i + 1; | |
f56b6394 KW |
6722 | } |
6723 | *folded_ptr = '\0'; | |
f56b6394 KW |
6724 | } |
6725 | ||
6726 | ||
6727 | /* Do the linear search to see if the fold is in the list | |
8f2655f7 KW |
6728 | * of multi-char folds. */ |
6729 | if (av) { | |
7cdde544 KW |
6730 | I32 i; |
6731 | for (i = 0; i <= av_len(av); i++) { | |
6732 | SV* const sv = *av_fetch(av, i, FALSE); | |
6733 | STRLEN len; | |
6734 | const char * const s = SvPV_const(sv, len); | |
6438af90 | 6735 | |
f9126265 KW |
6736 | if (len <= total_foldlen |
6737 | && memEQ(s, (char*)folded, len) | |
6738 | ||
6739 | /* If 0, means matched a partial char. See | |
6740 | * [perl #90536] */ | |
6741 | && map_fold_len_back[len]) | |
f56b6394 KW |
6742 | { |
6743 | ||
6744 | /* Advance the target string ptr to account for | |
6745 | * this fold, but have to translate from the | |
6746 | * folded length to the corresponding source | |
6438af90 | 6747 | * length. */ |
8f2655f7 KW |
6748 | if (lenp) { |
6749 | *lenp = map_fold_len_back[len]; | |
8f2655f7 | 6750 | } |
7cdde544 KW |
6751 | match = TRUE; |
6752 | break; | |
6753 | } | |
6754 | } | |
6755 | } | |
7cdde544 KW |
6756 | } |
6757 | ||
6758 | /* If we allocated a string above, free it */ | |
6759 | if (! utf8_target) Safefree(utf8_p); | |
6760 | } | |
6761 | } | |
6762 | } | |
6763 | ||
a0ed51b3 LW |
6764 | return (flags & ANYOF_INVERT) ? !match : match; |
6765 | } | |
161b471a | 6766 | |
dfe13c55 | 6767 | STATIC U8 * |
0ce71af7 | 6768 | S_reghop3(U8 *s, I32 off, const U8* lim) |
9041c2e3 | 6769 | { |
97aff369 | 6770 | dVAR; |
7918f24d NC |
6771 | |
6772 | PERL_ARGS_ASSERT_REGHOP3; | |
6773 | ||
a0ed51b3 | 6774 | if (off >= 0) { |
1aa99e6b | 6775 | while (off-- && s < lim) { |
ffc61ed2 | 6776 | /* XXX could check well-formedness here */ |
a0ed51b3 | 6777 | s += UTF8SKIP(s); |
ffc61ed2 | 6778 | } |
a0ed51b3 LW |
6779 | } |
6780 | else { | |
1de06328 YO |
6781 | while (off++ && s > lim) { |
6782 | s--; | |
6783 | if (UTF8_IS_CONTINUED(*s)) { | |
6784 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
6785 | s--; | |
a0ed51b3 | 6786 | } |
1de06328 | 6787 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
6788 | } |
6789 | } | |
6790 | return s; | |
6791 | } | |
161b471a | 6792 | |
f9f4320a YO |
6793 | #ifdef XXX_dmq |
6794 | /* there are a bunch of places where we use two reghop3's that should | |
6795 | be replaced with this routine. but since thats not done yet | |
6796 | we ifdef it out - dmq | |
6797 | */ | |
dfe13c55 | 6798 | STATIC U8 * |
1de06328 YO |
6799 | S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim) |
6800 | { | |
6801 | dVAR; | |
7918f24d NC |
6802 | |
6803 | PERL_ARGS_ASSERT_REGHOP4; | |
6804 | ||
1de06328 YO |
6805 | if (off >= 0) { |
6806 | while (off-- && s < rlim) { | |
6807 | /* XXX could check well-formedness here */ | |
6808 | s += UTF8SKIP(s); | |
6809 | } | |
6810 | } | |
6811 | else { | |
6812 | while (off++ && s > llim) { | |
6813 | s--; | |
6814 | if (UTF8_IS_CONTINUED(*s)) { | |
6815 | while (s > llim && UTF8_IS_CONTINUATION(*s)) | |
6816 | s--; | |
6817 | } | |
6818 | /* XXX could check well-formedness here */ | |
6819 | } | |
6820 | } | |
6821 | return s; | |
6822 | } | |
f9f4320a | 6823 | #endif |
1de06328 YO |
6824 | |
6825 | STATIC U8 * | |
0ce71af7 | 6826 | S_reghopmaybe3(U8* s, I32 off, const U8* lim) |
a0ed51b3 | 6827 | { |
97aff369 | 6828 | dVAR; |
7918f24d NC |
6829 | |
6830 | PERL_ARGS_ASSERT_REGHOPMAYBE3; | |
6831 | ||
a0ed51b3 | 6832 | if (off >= 0) { |
1aa99e6b | 6833 | while (off-- && s < lim) { |
ffc61ed2 | 6834 | /* XXX could check well-formedness here */ |
a0ed51b3 | 6835 | s += UTF8SKIP(s); |
ffc61ed2 | 6836 | } |
a0ed51b3 | 6837 | if (off >= 0) |
3dab1dad | 6838 | return NULL; |
a0ed51b3 LW |
6839 | } |
6840 | else { | |
1de06328 YO |
6841 | while (off++ && s > lim) { |
6842 | s--; | |
6843 | if (UTF8_IS_CONTINUED(*s)) { | |
6844 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
6845 | s--; | |
a0ed51b3 | 6846 | } |
1de06328 | 6847 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
6848 | } |
6849 | if (off <= 0) | |
3dab1dad | 6850 | return NULL; |
a0ed51b3 LW |
6851 | } |
6852 | return s; | |
6853 | } | |
51371543 | 6854 | |
51371543 | 6855 | static void |
acfe0abc | 6856 | restore_pos(pTHX_ void *arg) |
51371543 | 6857 | { |
97aff369 | 6858 | dVAR; |
097eb12c | 6859 | regexp * const rex = (regexp *)arg; |
51371543 GS |
6860 | if (PL_reg_eval_set) { |
6861 | if (PL_reg_oldsaved) { | |
4f639d21 DM |
6862 | rex->subbeg = PL_reg_oldsaved; |
6863 | rex->sublen = PL_reg_oldsavedlen; | |
f8c7b90f | 6864 | #ifdef PERL_OLD_COPY_ON_WRITE |
4f639d21 | 6865 | rex->saved_copy = PL_nrs; |
ed252734 | 6866 | #endif |
07bc277f | 6867 | RXp_MATCH_COPIED_on(rex); |
51371543 GS |
6868 | } |
6869 | PL_reg_magic->mg_len = PL_reg_oldpos; | |
6870 | PL_reg_eval_set = 0; | |
6871 | PL_curpm = PL_reg_oldcurpm; | |
6872 | } | |
6873 | } | |
33b8afdf JH |
6874 | |
6875 | STATIC void | |
6876 | S_to_utf8_substr(pTHX_ register regexp *prog) | |
6877 | { | |
a1cac82e | 6878 | int i = 1; |
7918f24d NC |
6879 | |
6880 | PERL_ARGS_ASSERT_TO_UTF8_SUBSTR; | |
6881 | ||
a1cac82e NC |
6882 | do { |
6883 | if (prog->substrs->data[i].substr | |
6884 | && !prog->substrs->data[i].utf8_substr) { | |
6885 | SV* const sv = newSVsv(prog->substrs->data[i].substr); | |
6886 | prog->substrs->data[i].utf8_substr = sv; | |
6887 | sv_utf8_upgrade(sv); | |
610460f9 | 6888 | if (SvVALID(prog->substrs->data[i].substr)) { |
cffe132d | 6889 | if (SvTAIL(prog->substrs->data[i].substr)) { |
610460f9 NC |
6890 | /* Trim the trailing \n that fbm_compile added last |
6891 | time. */ | |
6892 | SvCUR_set(sv, SvCUR(sv) - 1); | |
6893 | /* Whilst this makes the SV technically "invalid" (as its | |
6894 | buffer is no longer followed by "\0") when fbm_compile() | |
6895 | adds the "\n" back, a "\0" is restored. */ | |
cffe132d NC |
6896 | fbm_compile(sv, FBMcf_TAIL); |
6897 | } else | |
6898 | fbm_compile(sv, 0); | |
610460f9 | 6899 | } |
a1cac82e NC |
6900 | if (prog->substrs->data[i].substr == prog->check_substr) |
6901 | prog->check_utf8 = sv; | |
6902 | } | |
6903 | } while (i--); | |
33b8afdf JH |
6904 | } |
6905 | ||
6906 | STATIC void | |
6907 | S_to_byte_substr(pTHX_ register regexp *prog) | |
6908 | { | |
97aff369 | 6909 | dVAR; |
a1cac82e | 6910 | int i = 1; |
7918f24d NC |
6911 | |
6912 | PERL_ARGS_ASSERT_TO_BYTE_SUBSTR; | |
6913 | ||
a1cac82e NC |
6914 | do { |
6915 | if (prog->substrs->data[i].utf8_substr | |
6916 | && !prog->substrs->data[i].substr) { | |
6917 | SV* sv = newSVsv(prog->substrs->data[i].utf8_substr); | |
6918 | if (sv_utf8_downgrade(sv, TRUE)) { | |
610460f9 | 6919 | if (SvVALID(prog->substrs->data[i].utf8_substr)) { |
cffe132d | 6920 | if (SvTAIL(prog->substrs->data[i].utf8_substr)) { |
610460f9 NC |
6921 | /* Trim the trailing \n that fbm_compile added last |
6922 | time. */ | |
6923 | SvCUR_set(sv, SvCUR(sv) - 1); | |
cffe132d NC |
6924 | fbm_compile(sv, FBMcf_TAIL); |
6925 | } else | |
6926 | fbm_compile(sv, 0); | |
6927 | } | |
a1cac82e NC |
6928 | } else { |
6929 | SvREFCNT_dec(sv); | |
6930 | sv = &PL_sv_undef; | |
6931 | } | |
6932 | prog->substrs->data[i].substr = sv; | |
6933 | if (prog->substrs->data[i].utf8_substr == prog->check_utf8) | |
6934 | prog->check_substr = sv; | |
33b8afdf | 6935 | } |
a1cac82e | 6936 | } while (i--); |
33b8afdf | 6937 | } |
66610fdd RGS |
6938 | |
6939 | /* | |
6940 | * Local variables: | |
6941 | * c-indentation-style: bsd | |
6942 | * c-basic-offset: 4 | |
6943 | * indent-tabs-mode: t | |
6944 | * End: | |
6945 | * | |
37442d52 RGS |
6946 | * ex: set ts=8 sts=4 sw=4 noet: |
6947 | */ |