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 */ |
a9a337f8 KW |
125 | #ifdef EBCDIC |
126 | /* Often 'str' is a hard-coded utf8 string instead of utfebcdic. so just | |
127 | * skip the check on EBCDIC platforms */ | |
128 | # define LOAD_UTF8_CHARCLASS(class,str) LOAD_UTF8_CHARCLASS_NO_CHECK(class) | |
129 | #else | |
130 | # define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ | |
9c4fdda1 RB |
131 | if (!CAT2(PL_utf8_,class)) { \ |
132 | bool ok; \ | |
133 | ENTER; save_re_context(); \ | |
134 | ok=CAT2(is_utf8_,class)((const U8*)str); \ | |
9ac792f4 | 135 | assert(ok); assert(CAT2(PL_utf8_,class)); LEAVE; } } STMT_END |
a9a337f8 | 136 | #endif |
37e2e78e KW |
137 | |
138 | /* Doesn't do an assert to verify that is correct */ | |
139 | #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \ | |
9c4fdda1 | 140 | if (!CAT2(PL_utf8_,class)) { \ |
9d63fa07 | 141 | bool throw_away PERL_UNUSED_DECL; \ |
9c4fdda1 RB |
142 | ENTER; save_re_context(); \ |
143 | throw_away = CAT2(is_utf8_,class)((const U8*)" "); \ | |
144 | LEAVE; } } STMT_END | |
37e2e78e | 145 | |
1a4fad37 AL |
146 | #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a") |
147 | #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0") | |
148 | #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ") | |
51371543 | 149 | |
37e2e78e | 150 | #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \ |
d275fa5e CB |
151 | LOAD_UTF8_CHARCLASS(X_begin, " "); \ |
152 | LOAD_UTF8_CHARCLASS(X_non_hangul, "A"); \ | |
37e2e78e KW |
153 | /* These are utf8 constants, and not utf-ebcdic constants, so the \ |
154 | * assert should likely and hopefully fail on an EBCDIC machine */ \ | |
d275fa5e | 155 | LOAD_UTF8_CHARCLASS(X_extend, "\xcc\x80"); /* U+0300 */ \ |
37e2e78e KW |
156 | \ |
157 | /* No asserts are done for these, in case called on an early \ | |
158 | * Unicode version in which they map to nothing */ \ | |
d275fa5e CB |
159 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_prepend);/* U+0E40 "\xe0\xb9\x80" */ \ |
160 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_L); /* U+1100 "\xe1\x84\x80" */ \ | |
161 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV); /* U+AC00 "\xea\xb0\x80" */ \ | |
162 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LVT); /* U+AC01 "\xea\xb0\x81" */ \ | |
163 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV_LVT_V);/* U+AC01 "\xea\xb0\x81" */\ | |
164 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_T); /* U+11A8 "\xe1\x86\xa8" */ \ | |
37e2e78e | 165 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_V) /* U+1160 "\xe1\x85\xa0" */ |
20d0b1e9 | 166 | |
1dcf4a1b | 167 | #define PLACEHOLDER /* Something for the preprocessor to grab onto */ |
d1eb3177 | 168 | |
ee9a90b8 KW |
169 | /* The actual code for CCC_TRY, which uses several variables from the routine |
170 | * it's callable from. It is designed to be the bulk of a case statement. | |
171 | * FUNC is the macro or function to call on non-utf8 targets that indicate if | |
172 | * nextchr matches the class. | |
173 | * UTF8_TEST is the whole test string to use for utf8 targets | |
174 | * LOAD is what to use to test, and if not present to load in the swash for the | |
175 | * class | |
176 | * POS_OR_NEG is either empty or ! to complement the results of FUNC or | |
177 | * UTF8_TEST test. | |
178 | * The logic is: Fail if we're at the end-of-string; otherwise if the target is | |
179 | * utf8 and a variant, load the swash if necessary and test using the utf8 | |
180 | * test. Advance to the next character if test is ok, otherwise fail; If not | |
181 | * utf8 or an invariant under utf8, use the non-utf8 test, and fail if it | |
182 | * fails, or advance to the next character */ | |
183 | ||
184 | #define _CCC_TRY_CODE(POS_OR_NEG, FUNC, UTF8_TEST, CLASS, STR) \ | |
185 | if (locinput >= PL_regeol) { \ | |
186 | sayNO; \ | |
187 | } \ | |
188 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ | |
189 | LOAD_UTF8_CHARCLASS(CLASS, STR); \ | |
190 | if (POS_OR_NEG (UTF8_TEST)) { \ | |
191 | sayNO; \ | |
192 | } \ | |
193 | locinput += PL_utf8skip[nextchr]; \ | |
194 | nextchr = UCHARAT(locinput); \ | |
195 | break; \ | |
196 | } \ | |
197 | if (POS_OR_NEG (FUNC(nextchr))) { \ | |
198 | sayNO; \ | |
199 | } \ | |
200 | nextchr = UCHARAT(++locinput); \ | |
980866de KW |
201 | break; |
202 | ||
ee9a90b8 KW |
203 | /* Handle the non-locale cases for a character class and its complement. It |
204 | * calls _CCC_TRY_CODE with a ! to complement the test for the character class. | |
205 | * This is because that code fails when the test succeeds, so we want to have | |
206 | * the test fail so that the code succeeds. The swash is stored in a | |
207 | * predictable PL_ place */ | |
cfaf538b KW |
208 | #define _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, \ |
209 | CLASS, STR) \ | |
ee9a90b8 KW |
210 | case NAME: \ |
211 | _CCC_TRY_CODE( !, FUNC, \ | |
212 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ | |
213 | (U8*)locinput, TRUE)), \ | |
214 | CLASS, STR) \ | |
215 | case NNAME: \ | |
1dcf4a1b | 216 | _CCC_TRY_CODE( PLACEHOLDER , FUNC, \ |
ee9a90b8 KW |
217 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ |
218 | (U8*)locinput, TRUE)), \ | |
219 | CLASS, STR) \ | |
220 | ||
221 | /* Generate the case statements for both locale and non-locale character | |
222 | * classes in regmatch for classes that don't have special unicode semantics. | |
223 | * Locales don't use an immediate swash, but an intermediary special locale | |
224 | * function that is called on the pointer to the current place in the input | |
225 | * string. That function will resolve to needing the same swash. One might | |
226 | * think that because we don't know what the locale will match, we shouldn't | |
227 | * check with the swash loading function that it loaded properly; ie, that we | |
228 | * should use LOAD_UTF8_CHARCLASS_NO_CHECK for those, but what is passed to the | |
229 | * regular LOAD_UTF8_CHARCLASS is in non-locale terms, and so locale is | |
230 | * irrelevant here */ | |
231 | #define CCC_TRY(NAME, NNAME, FUNC, \ | |
232 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
cfaf538b | 233 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 KW |
234 | CLASS, STR) \ |
235 | case NAMEL: \ | |
236 | PL_reg_flags |= RF_tainted; \ | |
237 | _CCC_TRY_CODE( !, LCFUNC, LCFUNC_utf8((U8*)locinput), CLASS, STR) \ | |
238 | case NNAMEL: \ | |
239 | PL_reg_flags |= RF_tainted; \ | |
1dcf4a1b KW |
240 | _CCC_TRY_CODE( PLACEHOLDER, LCFUNC, LCFUNC_utf8((U8*)locinput), \ |
241 | CLASS, STR) \ | |
cfaf538b KW |
242 | case NAMEA: \ |
243 | if (locinput >= PL_regeol || ! FUNCA(nextchr)) { \ | |
244 | sayNO; \ | |
245 | } \ | |
246 | /* Matched a utf8-invariant, so don't have to worry about utf8 */ \ | |
247 | nextchr = UCHARAT(++locinput); \ | |
248 | break; \ | |
249 | case NNAMEA: \ | |
250 | if (locinput >= PL_regeol || FUNCA(nextchr)) { \ | |
251 | sayNO; \ | |
252 | } \ | |
253 | if (utf8_target) { \ | |
254 | locinput += PL_utf8skip[nextchr]; \ | |
255 | nextchr = UCHARAT(locinput); \ | |
256 | } \ | |
257 | else { \ | |
258 | nextchr = UCHARAT(++locinput); \ | |
259 | } \ | |
260 | break; \ | |
ee9a90b8 KW |
261 | /* Generate the non-locale cases */ \ |
262 | _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, CLASS, STR) | |
263 | ||
264 | /* This is like CCC_TRY, but has an extra set of parameters for generating case | |
265 | * statements to handle separate Unicode semantics nodes */ | |
266 | #define CCC_TRY_U(NAME, NNAME, FUNC, \ | |
267 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
268 | NAMEU, NNAMEU, FUNCU, \ | |
cfaf538b | 269 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 | 270 | CLASS, STR) \ |
cfaf538b KW |
271 | CCC_TRY(NAME, NNAME, FUNC, \ |
272 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
273 | NAMEA, NNAMEA, FUNCA, \ | |
274 | CLASS, STR) \ | |
ee9a90b8 | 275 | _CCC_TRY_NONLOCALE(NAMEU, NNAMEU, FUNCU, CLASS, STR) |
d1eb3177 | 276 | |
3dab1dad YO |
277 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
278 | ||
5f80c4cf | 279 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
c35dcbe2 YO |
280 | /* it would be nice to rework regcomp.sym to generate this stuff. sigh |
281 | * | |
282 | * NOTE that *nothing* that affects backtracking should be in here, specifically | |
283 | * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a | |
284 | * node that is in between two EXACT like nodes when ascertaining what the required | |
285 | * "follow" character is. This should probably be moved to regex compile time | |
286 | * although it may be done at run time beause of the REF possibility - more | |
287 | * investigation required. -- demerphq | |
288 | */ | |
3e901dc0 YO |
289 | #define JUMPABLE(rn) ( \ |
290 | OP(rn) == OPEN || \ | |
291 | (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \ | |
292 | OP(rn) == EVAL || \ | |
cca55fe3 JP |
293 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
294 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
d1c771f5 | 295 | OP(rn) == KEEPS || \ |
3dab1dad | 296 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 | 297 | ) |
ee9b8eae | 298 | #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT) |
e2d8ce26 | 299 | |
ee9b8eae YO |
300 | #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF ) |
301 | ||
302 | #if 0 | |
303 | /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so | |
304 | we don't need this definition. */ | |
305 | #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) | |
2f7f8cb1 | 306 | #define IS_TEXTF(rn) ( (OP(rn)==EXACTFU || OP(rn)==EXACTFA || OP(rn)==EXACTF) || OP(rn)==REFF || OP(rn)==NREFF ) |
ee9b8eae YO |
307 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) |
308 | ||
309 | #else | |
310 | /* ... so we use this as its faster. */ | |
311 | #define IS_TEXT(rn) ( OP(rn)==EXACT ) | |
2f7f8cb1 | 312 | #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn) == EXACTFA) |
ee9b8eae YO |
313 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) |
314 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) | |
315 | ||
316 | #endif | |
e2d8ce26 | 317 | |
a84d97b6 HS |
318 | /* |
319 | Search for mandatory following text node; for lookahead, the text must | |
320 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
321 | */ | |
cca55fe3 | 322 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
323 | while (JUMPABLE(rn)) { \ |
324 | const OPCODE type = OP(rn); \ | |
325 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 326 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 327 | else if (type == PLUS) \ |
cca55fe3 | 328 | rn = NEXTOPER(rn); \ |
3dab1dad | 329 | else if (type == IFMATCH) \ |
a84d97b6 | 330 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 331 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 332 | } \ |
5f80c4cf | 333 | } STMT_END |
74750237 | 334 | |
c476f425 | 335 | |
acfe0abc | 336 | static void restore_pos(pTHX_ void *arg); |
51371543 | 337 | |
620d5b66 NC |
338 | #define REGCP_PAREN_ELEMS 4 |
339 | #define REGCP_OTHER_ELEMS 5 | |
e0fa7e2b | 340 | #define REGCP_FRAME_ELEMS 1 |
620d5b66 NC |
341 | /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and |
342 | * are needed for the regexp context stack bookkeeping. */ | |
343 | ||
76e3520e | 344 | STATIC CHECKPOINT |
cea2e8a9 | 345 | S_regcppush(pTHX_ I32 parenfloor) |
a0d0e21e | 346 | { |
97aff369 | 347 | dVAR; |
a3b680e6 | 348 | const int retval = PL_savestack_ix; |
a3b680e6 | 349 | const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; |
e0fa7e2b NC |
350 | const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS; |
351 | const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT; | |
a0d0e21e | 352 | int p; |
40a82448 | 353 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 354 | |
e49a9654 IH |
355 | if (paren_elems_to_push < 0) |
356 | Perl_croak(aTHX_ "panic: paren_elems_to_push < 0"); | |
357 | ||
e0fa7e2b NC |
358 | if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems) |
359 | Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf | |
5df417d0 JH |
360 | " out of range (%lu-%ld)", |
361 | total_elems, (unsigned long)PL_regsize, (long)parenfloor); | |
e0fa7e2b | 362 | |
620d5b66 | 363 | SSGROW(total_elems + REGCP_FRAME_ELEMS); |
7f69552c | 364 | |
3280af22 | 365 | for (p = PL_regsize; p > parenfloor; p--) { |
b1ce53c5 | 366 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
f0ab9afb NC |
367 | SSPUSHINT(PL_regoffs[p].end); |
368 | SSPUSHINT(PL_regoffs[p].start); | |
3280af22 | 369 | SSPUSHPTR(PL_reg_start_tmp[p]); |
a0d0e21e | 370 | SSPUSHINT(p); |
e7707071 | 371 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
40a82448 | 372 | " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n", |
f0ab9afb | 373 | (UV)p, (IV)PL_regoffs[p].start, |
40a82448 | 374 | (IV)(PL_reg_start_tmp[p] - PL_bostr), |
f0ab9afb | 375 | (IV)PL_regoffs[p].end |
40a82448 | 376 | )); |
a0d0e21e | 377 | } |
b1ce53c5 | 378 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
f0ab9afb | 379 | SSPUSHPTR(PL_regoffs); |
3280af22 NIS |
380 | SSPUSHINT(PL_regsize); |
381 | SSPUSHINT(*PL_reglastparen); | |
a01268b5 | 382 | SSPUSHINT(*PL_reglastcloseparen); |
3280af22 | 383 | SSPUSHPTR(PL_reginput); |
e0fa7e2b | 384 | SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */ |
41123dfd | 385 | |
a0d0e21e LW |
386 | return retval; |
387 | } | |
388 | ||
c277df42 | 389 | /* These are needed since we do not localize EVAL nodes: */ |
ab3bbdeb YO |
390 | #define REGCP_SET(cp) \ |
391 | DEBUG_STATE_r( \ | |
ab3bbdeb | 392 | PerlIO_printf(Perl_debug_log, \ |
e4f74956 | 393 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
ab3bbdeb YO |
394 | (IV)PL_savestack_ix)); \ |
395 | cp = PL_savestack_ix | |
c3464db5 | 396 | |
ab3bbdeb | 397 | #define REGCP_UNWIND(cp) \ |
e4f74956 | 398 | DEBUG_STATE_r( \ |
ab3bbdeb | 399 | if (cp != PL_savestack_ix) \ |
e4f74956 YO |
400 | PerlIO_printf(Perl_debug_log, \ |
401 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ | |
ab3bbdeb YO |
402 | (IV)(cp), (IV)PL_savestack_ix)); \ |
403 | regcpblow(cp) | |
c277df42 | 404 | |
76e3520e | 405 | STATIC char * |
097eb12c | 406 | S_regcppop(pTHX_ const regexp *rex) |
a0d0e21e | 407 | { |
97aff369 | 408 | dVAR; |
e0fa7e2b | 409 | UV i; |
a0d0e21e | 410 | char *input; |
a3621e74 YO |
411 | GET_RE_DEBUG_FLAGS_DECL; |
412 | ||
7918f24d NC |
413 | PERL_ARGS_ASSERT_REGCPPOP; |
414 | ||
b1ce53c5 | 415 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
c6bf6a65 | 416 | i = SSPOPUV; |
e0fa7e2b NC |
417 | assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
418 | i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */ | |
a0d0e21e | 419 | input = (char *) SSPOPPTR; |
a01268b5 | 420 | *PL_reglastcloseparen = SSPOPINT; |
3280af22 NIS |
421 | *PL_reglastparen = SSPOPINT; |
422 | PL_regsize = SSPOPINT; | |
f0ab9afb | 423 | PL_regoffs=(regexp_paren_pair *) SSPOPPTR; |
b1ce53c5 | 424 | |
620d5b66 | 425 | i -= REGCP_OTHER_ELEMS; |
b1ce53c5 | 426 | /* Now restore the parentheses context. */ |
620d5b66 | 427 | for ( ; i > 0; i -= REGCP_PAREN_ELEMS) { |
1df70142 | 428 | I32 tmps; |
097eb12c | 429 | U32 paren = (U32)SSPOPINT; |
3280af22 | 430 | PL_reg_start_tmp[paren] = (char *) SSPOPPTR; |
f0ab9afb | 431 | PL_regoffs[paren].start = SSPOPINT; |
cf93c79d | 432 | tmps = SSPOPINT; |
3280af22 | 433 | if (paren <= *PL_reglastparen) |
f0ab9afb | 434 | PL_regoffs[paren].end = tmps; |
e7707071 | 435 | DEBUG_BUFFERS_r( |
c3464db5 | 436 | PerlIO_printf(Perl_debug_log, |
b900a521 | 437 | " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n", |
f0ab9afb | 438 | (UV)paren, (IV)PL_regoffs[paren].start, |
b900a521 | 439 | (IV)(PL_reg_start_tmp[paren] - PL_bostr), |
f0ab9afb | 440 | (IV)PL_regoffs[paren].end, |
3280af22 | 441 | (paren > *PL_reglastparen ? "(no)" : "")); |
c277df42 | 442 | ); |
a0d0e21e | 443 | } |
e7707071 | 444 | DEBUG_BUFFERS_r( |
bb7a0f54 | 445 | if (*PL_reglastparen + 1 <= rex->nparens) { |
c3464db5 | 446 | PerlIO_printf(Perl_debug_log, |
faccc32b | 447 | " restoring \\%"IVdf"..\\%"IVdf" to undef\n", |
4f639d21 | 448 | (IV)(*PL_reglastparen + 1), (IV)rex->nparens); |
c277df42 IZ |
449 | } |
450 | ); | |
daf18116 | 451 | #if 1 |
dafc8851 JH |
452 | /* It would seem that the similar code in regtry() |
453 | * already takes care of this, and in fact it is in | |
454 | * a better location to since this code can #if 0-ed out | |
455 | * but the code in regtry() is needed or otherwise tests | |
456 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
457 | * (as of patchlevel 7877) will fail. Then again, |
458 | * this code seems to be necessary or otherwise | |
225593e1 DM |
459 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ |
460 | * --jhi updated by dapm */ | |
3b6647e0 | 461 | for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) { |
097eb12c | 462 | if (i > PL_regsize) |
f0ab9afb NC |
463 | PL_regoffs[i].start = -1; |
464 | PL_regoffs[i].end = -1; | |
a0d0e21e | 465 | } |
dafc8851 | 466 | #endif |
a0d0e21e LW |
467 | return input; |
468 | } | |
469 | ||
02db2b7b | 470 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 471 | |
a687059c | 472 | /* |
e50aee73 | 473 | * pregexec and friends |
a687059c LW |
474 | */ |
475 | ||
76234dfb | 476 | #ifndef PERL_IN_XSUB_RE |
a687059c | 477 | /* |
c277df42 | 478 | - pregexec - match a regexp against a string |
a687059c | 479 | */ |
c277df42 | 480 | I32 |
49d7dfbc | 481 | Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend, |
c3464db5 | 482 | char *strbeg, I32 minend, SV *screamer, U32 nosave) |
c277df42 IZ |
483 | /* strend: pointer to null at end of string */ |
484 | /* strbeg: real beginning of string */ | |
485 | /* minend: end of match must be >=minend after stringarg. */ | |
486 | /* nosave: For optimizations. */ | |
487 | { | |
7918f24d NC |
488 | PERL_ARGS_ASSERT_PREGEXEC; |
489 | ||
c277df42 | 490 | return |
9041c2e3 | 491 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
492 | nosave ? 0 : REXEC_COPY_STR); |
493 | } | |
76234dfb | 494 | #endif |
22e551b9 | 495 | |
9041c2e3 | 496 | /* |
cad2e5aa JH |
497 | * Need to implement the following flags for reg_anch: |
498 | * | |
499 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
500 | * USE_INTUIT_ML | |
501 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
502 | * INTUIT_AUTORITATIVE_ML | |
503 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
504 | * INTUIT_ONCE_ML | |
505 | * | |
506 | * Another flag for this function: SECOND_TIME (so that float substrs | |
507 | * with giant delta may be not rechecked). | |
508 | */ | |
509 | ||
510 | /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ | |
511 | ||
3f7c398e | 512 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
513 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
514 | ||
515 | /* XXXX We assume that strpos is strbeg unless sv. */ | |
516 | ||
6eb5f6b9 JH |
517 | /* XXXX Some places assume that there is a fixed substring. |
518 | An update may be needed if optimizer marks as "INTUITable" | |
519 | RExen without fixed substrings. Similarly, it is assumed that | |
520 | lengths of all the strings are no more than minlen, thus they | |
521 | cannot come from lookahead. | |
40d049e4 YO |
522 | (Or minlen should take into account lookahead.) |
523 | NOTE: Some of this comment is not correct. minlen does now take account | |
524 | of lookahead/behind. Further research is required. -- demerphq | |
525 | ||
526 | */ | |
6eb5f6b9 | 527 | |
2c2d71f5 JH |
528 | /* A failure to find a constant substring means that there is no need to make |
529 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
530 | finding a substring too deep into the string means that less calls to | |
30944b6d IZ |
531 | regtry() should be needed. |
532 | ||
533 | REx compiler's optimizer found 4 possible hints: | |
534 | a) Anchored substring; | |
535 | b) Fixed substring; | |
536 | c) Whether we are anchored (beginning-of-line or \G); | |
486ec47a | 537 | d) First node (of those at offset 0) which may distinguish positions; |
6eb5f6b9 | 538 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
539 | string which does not contradict any of them. |
540 | */ | |
2c2d71f5 | 541 | |
6eb5f6b9 JH |
542 | /* Most of decisions we do here should have been done at compile time. |
543 | The nodes of the REx which we used for the search should have been | |
544 | deleted from the finite automaton. */ | |
545 | ||
cad2e5aa | 546 | char * |
288b8c02 | 547 | Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, |
9f61653a | 548 | char *strend, const U32 flags, re_scream_pos_data *data) |
cad2e5aa | 549 | { |
97aff369 | 550 | dVAR; |
288b8c02 | 551 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
b7953727 | 552 | register I32 start_shift = 0; |
cad2e5aa | 553 | /* Should be nonnegative! */ |
b7953727 | 554 | register I32 end_shift = 0; |
2c2d71f5 JH |
555 | register char *s; |
556 | register SV *check; | |
a1933d95 | 557 | char *strbeg; |
cad2e5aa | 558 | char *t; |
f2ed9b32 | 559 | const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 560 | I32 ml_anch; |
bd61b366 SS |
561 | register char *other_last = NULL; /* other substr checked before this */ |
562 | char *check_at = NULL; /* check substr found at this pos */ | |
bbe252da | 563 | const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; |
f8fc2ecf | 564 | RXi_GET_DECL(prog,progi); |
30944b6d | 565 | #ifdef DEBUGGING |
890ce7af | 566 | const char * const i_strpos = strpos; |
30944b6d | 567 | #endif |
a3621e74 YO |
568 | GET_RE_DEBUG_FLAGS_DECL; |
569 | ||
7918f24d NC |
570 | PERL_ARGS_ASSERT_RE_INTUIT_START; |
571 | ||
f2ed9b32 | 572 | RX_MATCH_UTF8_set(rx,utf8_target); |
cad2e5aa | 573 | |
3c8556c3 | 574 | if (RX_UTF8(rx)) { |
b8d68ded JH |
575 | PL_reg_flags |= RF_utf8; |
576 | } | |
ab3bbdeb | 577 | DEBUG_EXECUTE_r( |
f2ed9b32 | 578 | debug_start_match(rx, utf8_target, strpos, strend, |
1de06328 YO |
579 | sv ? "Guessing start of match in sv for" |
580 | : "Guessing start of match in string for"); | |
2a782b5b | 581 | ); |
cad2e5aa | 582 | |
c344f387 JH |
583 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
584 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 585 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 586 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 587 | goto fail; |
2c2d71f5 | 588 | } |
1de06328 | 589 | |
a1933d95 | 590 | strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; |
1aa99e6b | 591 | PL_regeol = strend; |
f2ed9b32 | 592 | if (utf8_target) { |
33b8afdf JH |
593 | if (!prog->check_utf8 && prog->check_substr) |
594 | to_utf8_substr(prog); | |
595 | check = prog->check_utf8; | |
596 | } else { | |
597 | if (!prog->check_substr && prog->check_utf8) | |
598 | to_byte_substr(prog); | |
599 | check = prog->check_substr; | |
600 | } | |
1de06328 | 601 | if (check == &PL_sv_undef) { |
a3621e74 | 602 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
1de06328 | 603 | "Non-utf8 string cannot match utf8 check string\n")); |
33b8afdf JH |
604 | goto fail; |
605 | } | |
bbe252da YO |
606 | if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */ |
607 | ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) | |
608 | || ( (prog->extflags & RXf_ANCH_BOL) | |
7fba1cd6 | 609 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 610 | |
7e25d62c | 611 | if (!ml_anch) { |
bbe252da YO |
612 | if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */ |
613 | && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ | |
3f7c398e | 614 | /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ |
7e25d62c JH |
615 | && sv && !SvROK(sv) |
616 | && (strpos != strbeg)) { | |
a3621e74 | 617 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
618 | goto fail; |
619 | } | |
620 | if (prog->check_offset_min == prog->check_offset_max && | |
bbe252da | 621 | !(prog->extflags & RXf_CANY_SEEN)) { |
2c2d71f5 | 622 | /* Substring at constant offset from beg-of-str... */ |
cad2e5aa JH |
623 | I32 slen; |
624 | ||
1aa99e6b | 625 | s = HOP3c(strpos, prog->check_offset_min, strend); |
1de06328 | 626 | |
653099ff GS |
627 | if (SvTAIL(check)) { |
628 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 629 | |
9041c2e3 | 630 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 631 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 632 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 633 | goto fail_finish; |
cad2e5aa JH |
634 | } |
635 | /* Now should match s[0..slen-2] */ | |
636 | slen--; | |
3f7c398e | 637 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 638 | || (slen > 1 |
3f7c398e | 639 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 640 | report_neq: |
a3621e74 | 641 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
642 | goto fail_finish; |
643 | } | |
cad2e5aa | 644 | } |
3f7c398e | 645 | else if (*SvPVX_const(check) != *s |
653099ff | 646 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 647 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 648 | goto report_neq; |
c315bfe8 | 649 | check_at = s; |
2c2d71f5 | 650 | goto success_at_start; |
7e25d62c | 651 | } |
cad2e5aa | 652 | } |
2c2d71f5 | 653 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 654 | s = strpos; |
2c2d71f5 | 655 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
1de06328 YO |
656 | end_shift = prog->check_end_shift; |
657 | ||
2c2d71f5 | 658 | if (!ml_anch) { |
a3b680e6 | 659 | const I32 end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 660 | - (SvTAIL(check) != 0); |
a3b680e6 | 661 | const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
662 | |
663 | if (end_shift < eshift) | |
664 | end_shift = eshift; | |
665 | } | |
cad2e5aa | 666 | } |
2c2d71f5 | 667 | else { /* Can match at random position */ |
cad2e5aa JH |
668 | ml_anch = 0; |
669 | s = strpos; | |
1de06328 YO |
670 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
671 | end_shift = prog->check_end_shift; | |
672 | ||
673 | /* end shift should be non negative here */ | |
cad2e5aa JH |
674 | } |
675 | ||
bcdf7404 | 676 | #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 677 | if (end_shift < 0) |
1de06328 | 678 | Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", |
220fc49f | 679 | (IV)end_shift, RX_PRECOMP(prog)); |
2c2d71f5 JH |
680 | #endif |
681 | ||
2c2d71f5 JH |
682 | restart: |
683 | /* Find a possible match in the region s..strend by looking for | |
684 | the "check" substring in the region corrected by start/end_shift. */ | |
1de06328 YO |
685 | |
686 | { | |
687 | I32 srch_start_shift = start_shift; | |
688 | I32 srch_end_shift = end_shift; | |
689 | if (srch_start_shift < 0 && strbeg - s > srch_start_shift) { | |
690 | srch_end_shift -= ((strbeg - s) - srch_start_shift); | |
691 | srch_start_shift = strbeg - s; | |
692 | } | |
6bda09f9 | 693 | DEBUG_OPTIMISE_MORE_r({ |
1de06328 YO |
694 | PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n", |
695 | (IV)prog->check_offset_min, | |
696 | (IV)srch_start_shift, | |
697 | (IV)srch_end_shift, | |
698 | (IV)prog->check_end_shift); | |
699 | }); | |
700 | ||
f1905e1b | 701 | if ((flags & REXEC_SCREAM) && SvSCREAM(sv)) { |
cad2e5aa | 702 | I32 p = -1; /* Internal iterator of scream. */ |
a3b680e6 | 703 | I32 * const pp = data ? data->scream_pos : &p; |
4185c919 | 704 | const MAGIC *mg; |
72de20cd | 705 | bool found = FALSE; |
cad2e5aa | 706 | |
4185c919 NC |
707 | assert(SvMAGICAL(sv)); |
708 | mg = mg_find(sv, PERL_MAGIC_study); | |
709 | assert(mg); | |
710 | ||
72de20cd NC |
711 | if (mg->mg_private == 1) { |
712 | found = ((U8 *)mg->mg_ptr)[BmRARE(check)] != (U8)~0; | |
713 | } else if (mg->mg_private == 2) { | |
714 | found = ((U16 *)mg->mg_ptr)[BmRARE(check)] != (U16)~0; | |
715 | } else { | |
716 | assert (mg->mg_private == 4); | |
717 | found = ((U32 *)mg->mg_ptr)[BmRARE(check)] != (U32)~0; | |
718 | } | |
719 | ||
720 | if (found | |
2c2d71f5 | 721 | || ( BmRARE(check) == '\n' |
85c508c3 | 722 | && (BmPREVIOUS(check) == SvCUR(check) - 1) |
2c2d71f5 | 723 | && SvTAIL(check) )) |
9041c2e3 | 724 | s = screaminstr(sv, check, |
1de06328 | 725 | srch_start_shift + (s - strbeg), srch_end_shift, pp, 0); |
cad2e5aa | 726 | else |
2c2d71f5 | 727 | goto fail_finish; |
4addbd3b | 728 | /* we may be pointing at the wrong string */ |
07bc277f | 729 | if (s && RXp_MATCH_COPIED(prog)) |
3f7c398e | 730 | s = strbeg + (s - SvPVX_const(sv)); |
cad2e5aa JH |
731 | if (data) |
732 | *data->scream_olds = s; | |
733 | } | |
1de06328 YO |
734 | else { |
735 | U8* start_point; | |
736 | U8* end_point; | |
bbe252da | 737 | if (prog->extflags & RXf_CANY_SEEN) { |
1de06328 YO |
738 | start_point= (U8*)(s + srch_start_shift); |
739 | end_point= (U8*)(strend - srch_end_shift); | |
740 | } else { | |
741 | start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); | |
742 | end_point= HOP3(strend, -srch_end_shift, strbeg); | |
743 | } | |
6bda09f9 | 744 | DEBUG_OPTIMISE_MORE_r({ |
56570a2c | 745 | PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", |
1de06328 | 746 | (int)(end_point - start_point), |
fc8cd66c | 747 | (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), |
1de06328 YO |
748 | start_point); |
749 | }); | |
750 | ||
751 | s = fbm_instr( start_point, end_point, | |
7fba1cd6 | 752 | check, multiline ? FBMrf_MULTILINE : 0); |
1de06328 YO |
753 | } |
754 | } | |
cad2e5aa JH |
755 | /* Update the count-of-usability, remove useless subpatterns, |
756 | unshift s. */ | |
2c2d71f5 | 757 | |
ab3bbdeb | 758 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 759 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
760 | SvPVX_const(check), RE_SV_DUMPLEN(check), 30); |
761 | PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", | |
2c2d71f5 | 762 | (s ? "Found" : "Did not find"), |
f2ed9b32 | 763 | (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) |
ab3bbdeb YO |
764 | ? "anchored" : "floating"), |
765 | quoted, | |
766 | RE_SV_TAIL(check), | |
767 | (s ? " at offset " : "...\n") ); | |
768 | }); | |
2c2d71f5 JH |
769 | |
770 | if (!s) | |
771 | goto fail_finish; | |
2c2d71f5 | 772 | /* Finish the diagnostic message */ |
a3621e74 | 773 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 | 774 | |
1de06328 YO |
775 | /* XXX dmq: first branch is for positive lookbehind... |
776 | Our check string is offset from the beginning of the pattern. | |
777 | So we need to do any stclass tests offset forward from that | |
778 | point. I think. :-( | |
779 | */ | |
780 | ||
781 | ||
782 | ||
783 | check_at=s; | |
784 | ||
785 | ||
2c2d71f5 JH |
786 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. |
787 | Start with the other substr. | |
788 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 789 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
790 | *always* match. Probably should be marked during compile... |
791 | Probably it is right to do no SCREAM here... | |
792 | */ | |
793 | ||
f2ed9b32 | 794 | if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8) |
1de06328 YO |
795 | : (prog->float_substr && prog->anchored_substr)) |
796 | { | |
30944b6d | 797 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
798 | /* XXXX May be hopelessly wrong for UTF... */ |
799 | if (!other_last) | |
6eb5f6b9 | 800 | other_last = strpos; |
f2ed9b32 | 801 | if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
802 | do_other_anchored: |
803 | { | |
890ce7af AL |
804 | char * const last = HOP3c(s, -start_shift, strbeg); |
805 | char *last1, *last2; | |
be8e71aa | 806 | char * const saved_s = s; |
33b8afdf | 807 | SV* must; |
2c2d71f5 | 808 | |
2c2d71f5 JH |
809 | t = s - prog->check_offset_max; |
810 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
f2ed9b32 | 811 | && (!utf8_target |
0ce71af7 | 812 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 813 | && t > strpos))) |
6f207bd3 | 814 | NOOP; |
2c2d71f5 JH |
815 | else |
816 | t = strpos; | |
1aa99e6b | 817 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
818 | if (t < other_last) /* These positions already checked */ |
819 | t = other_last; | |
1aa99e6b | 820 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
821 | if (last < last1) |
822 | last1 = last; | |
1de06328 YO |
823 | /* XXXX It is not documented what units *_offsets are in. |
824 | We assume bytes, but this is clearly wrong. | |
825 | Meaning this code needs to be carefully reviewed for errors. | |
826 | dmq. | |
827 | */ | |
828 | ||
2c2d71f5 | 829 | /* On end-of-str: see comment below. */ |
f2ed9b32 | 830 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; |
33b8afdf JH |
831 | if (must == &PL_sv_undef) { |
832 | s = (char*)NULL; | |
1de06328 | 833 | DEBUG_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
834 | } |
835 | else | |
836 | s = fbm_instr( | |
837 | (unsigned char*)t, | |
838 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
839 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
840 | must, | |
7fba1cd6 | 841 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 842 | ); |
ab3bbdeb | 843 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 844 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
845 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
846 | PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", | |
2c2d71f5 | 847 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
848 | quoted, RE_SV_TAIL(must)); |
849 | }); | |
850 | ||
851 | ||
2c2d71f5 JH |
852 | if (!s) { |
853 | if (last1 >= last2) { | |
a3621e74 | 854 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
855 | ", giving up...\n")); |
856 | goto fail_finish; | |
857 | } | |
a3621e74 | 858 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 859 | ", trying floating at offset %ld...\n", |
be8e71aa | 860 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
861 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
862 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
863 | goto restart; |
864 | } | |
865 | else { | |
a3621e74 | 866 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 867 | (long)(s - i_strpos))); |
1aa99e6b IH |
868 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
869 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 870 | s = saved_s; |
2c2d71f5 JH |
871 | if (t == strpos) |
872 | goto try_at_start; | |
2c2d71f5 JH |
873 | goto try_at_offset; |
874 | } | |
30944b6d | 875 | } |
2c2d71f5 JH |
876 | } |
877 | else { /* Take into account the floating substring. */ | |
33b8afdf | 878 | char *last, *last1; |
be8e71aa | 879 | char * const saved_s = s; |
33b8afdf JH |
880 | SV* must; |
881 | ||
882 | t = HOP3c(s, -start_shift, strbeg); | |
883 | last1 = last = | |
884 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
885 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
886 | last = HOP3c(t, prog->float_max_offset, strend); | |
887 | s = HOP3c(t, prog->float_min_offset, strend); | |
888 | if (s < other_last) | |
889 | s = other_last; | |
2c2d71f5 | 890 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
f2ed9b32 | 891 | must = utf8_target ? prog->float_utf8 : prog->float_substr; |
33b8afdf JH |
892 | /* fbm_instr() takes into account exact value of end-of-str |
893 | if the check is SvTAIL(ed). Since false positives are OK, | |
894 | and end-of-str is not later than strend we are OK. */ | |
895 | if (must == &PL_sv_undef) { | |
896 | s = (char*)NULL; | |
1de06328 | 897 | DEBUG_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
898 | } |
899 | else | |
2c2d71f5 | 900 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
901 | (unsigned char*)last + SvCUR(must) |
902 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 903 | must, multiline ? FBMrf_MULTILINE : 0); |
ab3bbdeb | 904 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 905 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
906 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
907 | PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", | |
33b8afdf | 908 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
909 | quoted, RE_SV_TAIL(must)); |
910 | }); | |
33b8afdf JH |
911 | if (!s) { |
912 | if (last1 == last) { | |
a3621e74 | 913 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
914 | ", giving up...\n")); |
915 | goto fail_finish; | |
2c2d71f5 | 916 | } |
a3621e74 | 917 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 918 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 919 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
920 | other_last = last; |
921 | s = HOP3c(t, 1, strend); | |
922 | goto restart; | |
923 | } | |
924 | else { | |
a3621e74 | 925 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
926 | (long)(s - i_strpos))); |
927 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 928 | s = saved_s; |
33b8afdf JH |
929 | if (t == strpos) |
930 | goto try_at_start; | |
931 | goto try_at_offset; | |
932 | } | |
2c2d71f5 | 933 | } |
cad2e5aa | 934 | } |
2c2d71f5 | 935 | |
1de06328 | 936 | |
9ef43ace | 937 | t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); |
1de06328 | 938 | |
6bda09f9 | 939 | DEBUG_OPTIMISE_MORE_r( |
1de06328 YO |
940 | PerlIO_printf(Perl_debug_log, |
941 | "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", | |
942 | (IV)prog->check_offset_min, | |
943 | (IV)prog->check_offset_max, | |
944 | (IV)(s-strpos), | |
945 | (IV)(t-strpos), | |
946 | (IV)(t-s), | |
947 | (IV)(strend-strpos) | |
948 | ) | |
949 | ); | |
950 | ||
2c2d71f5 | 951 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
f2ed9b32 | 952 | && (!utf8_target |
9ef43ace | 953 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) |
1de06328 YO |
954 | && t > strpos))) |
955 | { | |
2c2d71f5 JH |
956 | /* Fixed substring is found far enough so that the match |
957 | cannot start at strpos. */ | |
958 | try_at_offset: | |
cad2e5aa | 959 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
960 | /* Eventually fbm_*() should handle this, but often |
961 | anchored_offset is not 0, so this check will not be wasted. */ | |
962 | /* XXXX In the code below we prefer to look for "^" even in | |
963 | presence of anchored substrings. And we search even | |
964 | beyond the found float position. These pessimizations | |
965 | are historical artefacts only. */ | |
966 | find_anchor: | |
2c2d71f5 | 967 | while (t < strend - prog->minlen) { |
cad2e5aa | 968 | if (*t == '\n') { |
4ee3650e | 969 | if (t < check_at - prog->check_offset_min) { |
f2ed9b32 | 970 | if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
971 | /* Since we moved from the found position, |
972 | we definitely contradict the found anchored | |
30944b6d IZ |
973 | substr. Due to the above check we do not |
974 | contradict "check" substr. | |
975 | Thus we can arrive here only if check substr | |
976 | is float. Redo checking for "other"=="fixed". | |
977 | */ | |
9041c2e3 | 978 | strpos = t + 1; |
a3621e74 | 979 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 980 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
981 | goto do_other_anchored; |
982 | } | |
4ee3650e GS |
983 | /* We don't contradict the found floating substring. */ |
984 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 985 | s = t + 1; |
a3621e74 | 986 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 987 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
988 | goto set_useful; |
989 | } | |
4ee3650e GS |
990 | /* Position contradicts check-string */ |
991 | /* XXXX probably better to look for check-string | |
992 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 993 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 994 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 995 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
996 | goto restart; |
997 | } | |
998 | t++; | |
999 | } | |
a3621e74 | 1000 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 1001 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 1002 | goto fail_finish; |
cad2e5aa | 1003 | } |
f5952150 | 1004 | else { |
a3621e74 | 1005 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 1006 | PL_colors[0], PL_colors[1])); |
f5952150 | 1007 | } |
cad2e5aa JH |
1008 | s = t; |
1009 | set_useful: | |
f2ed9b32 | 1010 | ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
1011 | } |
1012 | else { | |
f5952150 | 1013 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 1014 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
1015 | unless it was an MBOL and we are not after MBOL, |
1016 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
1017 | try_at_start: |
1018 | /* Even in this situation we may use MBOL flag if strpos is offset | |
1019 | wrt the start of the string. */ | |
05b4157f | 1020 | if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ |
a1933d95 | 1021 | && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d | 1022 | /* May be due to an implicit anchor of m{.*foo} */ |
bbe252da | 1023 | && !(prog->intflags & PREGf_IMPLICIT)) |
d506a20d | 1024 | { |
cad2e5aa JH |
1025 | t = strpos; |
1026 | goto find_anchor; | |
1027 | } | |
a3621e74 | 1028 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 1029 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
70685ca0 | 1030 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 1031 | ); |
2c2d71f5 | 1032 | success_at_start: |
bbe252da | 1033 | if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ |
f2ed9b32 | 1034 | && (utf8_target ? ( |
33b8afdf JH |
1035 | prog->check_utf8 /* Could be deleted already */ |
1036 | && --BmUSEFUL(prog->check_utf8) < 0 | |
1037 | && (prog->check_utf8 == prog->float_utf8) | |
1038 | ) : ( | |
1039 | prog->check_substr /* Could be deleted already */ | |
1040 | && --BmUSEFUL(prog->check_substr) < 0 | |
1041 | && (prog->check_substr == prog->float_substr) | |
1042 | ))) | |
66e933ab | 1043 | { |
cad2e5aa | 1044 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 1045 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
f2ed9b32 KW |
1046 | /* XXX Does the destruction order has to change with utf8_target? */ |
1047 | SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr); | |
1048 | SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
1049 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
1050 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
1051 | check = NULL; /* abort */ | |
cad2e5aa | 1052 | s = strpos; |
486ec47a | 1053 | /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag |
c9415951 YO |
1054 | see http://bugs.activestate.com/show_bug.cgi?id=87173 */ |
1055 | if (prog->intflags & PREGf_IMPLICIT) | |
1056 | prog->extflags &= ~RXf_ANCH_MBOL; | |
3cf5c195 IZ |
1057 | /* XXXX This is a remnant of the old implementation. It |
1058 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 1059 | other heuristics. */ |
bbe252da | 1060 | prog->extflags &= ~RXf_USE_INTUIT; |
c9415951 | 1061 | /* XXXX What other flags might need to be cleared in this branch? */ |
cad2e5aa JH |
1062 | } |
1063 | else | |
1064 | s = strpos; | |
1065 | } | |
1066 | ||
6eb5f6b9 JH |
1067 | /* Last resort... */ |
1068 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
1de06328 YO |
1069 | /* trie stclasses are too expensive to use here, we are better off to |
1070 | leave it to regmatch itself */ | |
f8fc2ecf | 1071 | if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { |
6eb5f6b9 JH |
1072 | /* minlen == 0 is possible if regstclass is \b or \B, |
1073 | and the fixed substr is ''$. | |
1074 | Since minlen is already taken into account, s+1 is before strend; | |
1075 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
1076 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
1077 | regstclass does not come from lookahead... */ | |
1078 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
af944926 | 1079 | This leaves EXACTF-ish only, which are dealt with in find_byclass(). */ |
f8fc2ecf YO |
1080 | const U8* const str = (U8*)STRING(progi->regstclass); |
1081 | const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT | |
1082 | ? CHR_DIST(str+STR_LEN(progi->regstclass), str) | |
66e933ab | 1083 | : 1); |
1de06328 YO |
1084 | char * endpos; |
1085 | if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) | |
1086 | endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); | |
1087 | else if (prog->float_substr || prog->float_utf8) | |
1088 | endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); | |
1089 | else | |
1090 | endpos= strend; | |
1091 | ||
70685ca0 JH |
1092 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n", |
1093 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg))); | |
1de06328 | 1094 | |
6eb5f6b9 | 1095 | t = s; |
f8fc2ecf | 1096 | s = find_byclass(prog, progi->regstclass, s, endpos, NULL); |
6eb5f6b9 JH |
1097 | if (!s) { |
1098 | #ifdef DEBUGGING | |
cbbf8932 | 1099 | const char *what = NULL; |
6eb5f6b9 JH |
1100 | #endif |
1101 | if (endpos == strend) { | |
a3621e74 | 1102 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
1103 | "Could not match STCLASS...\n") ); |
1104 | goto fail; | |
1105 | } | |
a3621e74 | 1106 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 1107 | "This position contradicts STCLASS...\n") ); |
bbe252da | 1108 | if ((prog->extflags & RXf_ANCH) && !ml_anch) |
653099ff | 1109 | goto fail; |
6eb5f6b9 | 1110 | /* Contradict one of substrings */ |
33b8afdf | 1111 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 | 1112 | if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) { |
a3621e74 | 1113 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 1114 | hop_and_restart: |
1aa99e6b | 1115 | s = HOP3c(t, 1, strend); |
66e933ab GS |
1116 | if (s + start_shift + end_shift > strend) { |
1117 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 1118 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
1119 | "Could not match STCLASS...\n") ); |
1120 | goto fail; | |
1121 | } | |
5e39e1e5 HS |
1122 | if (!check) |
1123 | goto giveup; | |
a3621e74 | 1124 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1125 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1126 | what, (long)(s + start_shift - i_strpos)) ); |
1127 | goto restart; | |
1128 | } | |
66e933ab | 1129 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
1130 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
1131 | goto retry_floating_check; | |
1132 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 1133 | s = check_at; |
5e39e1e5 HS |
1134 | if (!check) |
1135 | goto giveup; | |
a3621e74 | 1136 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1137 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1138 | (long)(other_last - i_strpos)) ); |
1139 | goto do_other_anchored; | |
1140 | } | |
60e71179 GS |
1141 | /* Another way we could have checked stclass at the |
1142 | current position only: */ | |
1143 | if (ml_anch) { | |
1144 | s = t = t + 1; | |
5e39e1e5 HS |
1145 | if (!check) |
1146 | goto giveup; | |
a3621e74 | 1147 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1148 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 1149 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 1150 | goto try_at_offset; |
66e933ab | 1151 | } |
f2ed9b32 | 1152 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 1153 | goto fail; |
486ec47a | 1154 | /* Check is floating substring. */ |
6eb5f6b9 JH |
1155 | retry_floating_check: |
1156 | t = check_at - start_shift; | |
a3621e74 | 1157 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
1158 | goto hop_and_restart; |
1159 | } | |
b7953727 | 1160 | if (t != s) { |
a3621e74 | 1161 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 1162 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
1163 | (long)(t - i_strpos), (long)(s - i_strpos)) |
1164 | ); | |
1165 | } | |
1166 | else { | |
a3621e74 | 1167 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1168 | "Does not contradict STCLASS...\n"); |
1169 | ); | |
1170 | } | |
6eb5f6b9 | 1171 | } |
5e39e1e5 | 1172 | giveup: |
a3621e74 | 1173 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
1174 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
1175 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 1176 | return s; |
2c2d71f5 JH |
1177 | |
1178 | fail_finish: /* Substring not found */ | |
33b8afdf | 1179 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
f2ed9b32 | 1180 | BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ |
cad2e5aa | 1181 | fail: |
a3621e74 | 1182 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 1183 | PL_colors[4], PL_colors[5])); |
bd61b366 | 1184 | return NULL; |
cad2e5aa | 1185 | } |
9661b544 | 1186 | |
a0a388a1 YO |
1187 | #define DECL_TRIE_TYPE(scan) \ |
1188 | const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \ | |
1189 | trie_type = (scan->flags != EXACT) \ | |
f2ed9b32 KW |
1190 | ? (utf8_target ? trie_utf8_fold : (UTF_PATTERN ? trie_latin_utf8_fold : trie_plain)) \ |
1191 | : (utf8_target ? trie_utf8 : trie_plain) | |
3b0527fe | 1192 | |
55eed653 NC |
1193 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ |
1194 | uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ | |
4cadc6a9 YO |
1195 | switch (trie_type) { \ |
1196 | case trie_utf8_fold: \ | |
1197 | if ( foldlen>0 ) { \ | |
0abd0d78 | 1198 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ |
4cadc6a9 YO |
1199 | foldlen -= len; \ |
1200 | uscan += len; \ | |
1201 | len=0; \ | |
1202 | } else { \ | |
ad0bff0e KW |
1203 | uvc = to_utf8_fold( (U8 *) uc, foldbuf, &foldlen ); \ |
1204 | len = UTF8SKIP(uc); \ | |
4cadc6a9 YO |
1205 | foldlen -= UNISKIP( uvc ); \ |
1206 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1207 | } \ | |
1208 | break; \ | |
a0a388a1 YO |
1209 | case trie_latin_utf8_fold: \ |
1210 | if ( foldlen>0 ) { \ | |
1211 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1212 | foldlen -= len; \ | |
1213 | uscan += len; \ | |
1214 | len=0; \ | |
1215 | } else { \ | |
1216 | len = 1; \ | |
1217 | uvc = to_uni_fold( *(U8*)uc, foldbuf, &foldlen ); \ | |
1218 | foldlen -= UNISKIP( uvc ); \ | |
1219 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1220 | } \ | |
1221 | break; \ | |
4cadc6a9 YO |
1222 | case trie_utf8: \ |
1223 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ | |
1224 | break; \ | |
1225 | case trie_plain: \ | |
1226 | uvc = (UV)*uc; \ | |
1227 | len = 1; \ | |
1228 | } \ | |
4cadc6a9 YO |
1229 | if (uvc < 256) { \ |
1230 | charid = trie->charmap[ uvc ]; \ | |
1231 | } \ | |
1232 | else { \ | |
1233 | charid = 0; \ | |
55eed653 NC |
1234 | if (widecharmap) { \ |
1235 | SV** const svpp = hv_fetch(widecharmap, \ | |
4cadc6a9 YO |
1236 | (char*)&uvc, sizeof(UV), 0); \ |
1237 | if (svpp) \ | |
1238 | charid = (U16)SvIV(*svpp); \ | |
1239 | } \ | |
1240 | } \ | |
1241 | } STMT_END | |
1242 | ||
4cadc6a9 YO |
1243 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ |
1244 | STMT_START { \ | |
1245 | while (s <= e) { \ | |
1246 | if ( (CoNd) \ | |
fac1af77 | 1247 | && (ln == 1 || folder(s, pat_string, ln)) \ |
9a5a5549 | 1248 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1249 | goto got_it; \ |
1250 | s++; \ | |
1251 | } \ | |
1252 | } STMT_END | |
1253 | ||
1254 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
1255 | STMT_START { \ | |
1256 | while (s + (uskip = UTF8SKIP(s)) <= strend) { \ | |
1257 | CoDe \ | |
1258 | s += uskip; \ | |
1259 | } \ | |
1260 | } STMT_END | |
1261 | ||
1262 | #define REXEC_FBC_SCAN(CoDe) \ | |
1263 | STMT_START { \ | |
1264 | while (s < strend) { \ | |
1265 | CoDe \ | |
1266 | s++; \ | |
1267 | } \ | |
1268 | } STMT_END | |
1269 | ||
1270 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1271 | REXEC_FBC_UTF8_SCAN( \ | |
1272 | if (CoNd) { \ | |
24b23f37 | 1273 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1274 | goto got_it; \ |
1275 | else \ | |
1276 | tmp = doevery; \ | |
1277 | } \ | |
1278 | else \ | |
1279 | tmp = 1; \ | |
1280 | ) | |
1281 | ||
1282 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1283 | REXEC_FBC_SCAN( \ | |
1284 | if (CoNd) { \ | |
24b23f37 | 1285 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1286 | goto got_it; \ |
1287 | else \ | |
1288 | tmp = doevery; \ | |
1289 | } \ | |
1290 | else \ | |
1291 | tmp = 1; \ | |
1292 | ) | |
1293 | ||
1294 | #define REXEC_FBC_TRYIT \ | |
24b23f37 | 1295 | if ((!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1296 | goto got_it |
1297 | ||
e1d1eefb | 1298 | #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \ |
f2ed9b32 | 1299 | if (utf8_target) { \ |
e1d1eefb YO |
1300 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1301 | } \ | |
1302 | else { \ | |
1303 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1304 | } |
e1d1eefb | 1305 | |
4cadc6a9 | 1306 | #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ |
f2ed9b32 | 1307 | if (utf8_target) { \ |
4cadc6a9 YO |
1308 | UtFpReLoAd; \ |
1309 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1310 | } \ | |
1311 | else { \ | |
1312 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1313 | } |
4cadc6a9 YO |
1314 | |
1315 | #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ | |
1316 | PL_reg_flags |= RF_tainted; \ | |
f2ed9b32 | 1317 | if (utf8_target) { \ |
4cadc6a9 YO |
1318 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1319 | } \ | |
1320 | else { \ | |
1321 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1322 | } |
4cadc6a9 | 1323 | |
786e8c11 YO |
1324 | #define DUMP_EXEC_POS(li,s,doutf8) \ |
1325 | dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8) | |
1326 | ||
cfaf538b KW |
1327 | |
1328 | #define UTF8_NOLOAD(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ | |
1329 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1330 | tmp = TEST_NON_UTF8(tmp); \ | |
1331 | REXEC_FBC_UTF8_SCAN( \ | |
1332 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1333 | tmp = !tmp; \ | |
1334 | IF_SUCCESS; \ | |
1335 | } \ | |
1336 | else { \ | |
1337 | IF_FAIL; \ | |
1338 | } \ | |
1339 | ); \ | |
1340 | ||
1341 | #define UTF8_LOAD(TeSt1_UtF8, TeSt2_UtF8, IF_SUCCESS, IF_FAIL) \ | |
1342 | if (s == PL_bostr) { \ | |
1343 | tmp = '\n'; \ | |
1344 | } \ | |
1345 | else { \ | |
1346 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); \ | |
1347 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); \ | |
1348 | } \ | |
1349 | tmp = TeSt1_UtF8; \ | |
1350 | LOAD_UTF8_CHARCLASS_ALNUM(); \ | |
1351 | REXEC_FBC_UTF8_SCAN( \ | |
1352 | if (tmp == ! (TeSt2_UtF8)) { \ | |
1353 | tmp = !tmp; \ | |
1354 | IF_SUCCESS; \ | |
1355 | } \ | |
1356 | else { \ | |
1357 | IF_FAIL; \ | |
1358 | } \ | |
1359 | ); \ | |
1360 | ||
63ac0dad KW |
1361 | /* The only difference between the BOUND and NBOUND cases is that |
1362 | * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in | |
1363 | * NBOUND. This is accomplished by passing it in either the if or else clause, | |
1364 | * with the other one being empty */ | |
1365 | #define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1366 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
cfaf538b KW |
1367 | |
1368 | #define FBC_BOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1369 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
63ac0dad KW |
1370 | |
1371 | #define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1372 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b KW |
1373 | |
1374 | #define FBC_NBOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1375 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b | 1376 | |
63ac0dad KW |
1377 | |
1378 | /* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to | |
1379 | * be passed in completely with the variable name being tested, which isn't | |
1380 | * such a clean interface, but this is easier to read than it was before. We | |
1381 | * are looking for the boundary (or non-boundary between a word and non-word | |
1382 | * character. The utf8 and non-utf8 cases have the same logic, but the details | |
1383 | * must be different. Find the "wordness" of the character just prior to this | |
1384 | * one, and compare it with the wordness of this one. If they differ, we have | |
1385 | * a boundary. At the beginning of the string, pretend that the previous | |
1386 | * character was a new-line */ | |
cfaf538b | 1387 | #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ |
63ac0dad | 1388 | if (utf8_target) { \ |
cfaf538b | 1389 | UTF8_CODE \ |
63ac0dad KW |
1390 | } \ |
1391 | else { /* Not utf8 */ \ | |
1392 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1393 | tmp = TEST_NON_UTF8(tmp); \ | |
1394 | REXEC_FBC_SCAN( \ | |
1395 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1396 | tmp = !tmp; \ | |
1397 | IF_SUCCESS; \ | |
1398 | } \ | |
1399 | else { \ | |
1400 | IF_FAIL; \ | |
1401 | } \ | |
1402 | ); \ | |
1403 | } \ | |
1404 | if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) \ | |
1405 | goto got_it; | |
1406 | ||
786e8c11 YO |
1407 | /* We know what class REx starts with. Try to find this position... */ |
1408 | /* if reginfo is NULL, its a dryrun */ | |
1409 | /* annoyingly all the vars in this routine have different names from their counterparts | |
1410 | in regmatch. /grrr */ | |
1411 | ||
3c3eec57 | 1412 | STATIC char * |
07be1b83 | 1413 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
24b23f37 | 1414 | const char *strend, regmatch_info *reginfo) |
a687059c | 1415 | { |
27da23d5 | 1416 | dVAR; |
bbe252da | 1417 | const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; |
fac1af77 KW |
1418 | char *pat_string; /* The pattern's exactish string */ |
1419 | char *pat_end; /* ptr to end char of pat_string */ | |
1420 | re_fold_t folder; /* Function for computing non-utf8 folds */ | |
1421 | const U8 *fold_array; /* array for folding ords < 256 */ | |
d8093b23 | 1422 | STRLEN ln; |
5dab1207 | 1423 | STRLEN lnc; |
078c425b | 1424 | register STRLEN uskip; |
fac1af77 KW |
1425 | U8 c1; |
1426 | U8 c2; | |
6eb5f6b9 JH |
1427 | char *e; |
1428 | register I32 tmp = 1; /* Scratch variable? */ | |
f2ed9b32 | 1429 | register const bool utf8_target = PL_reg_match_utf8; |
453bfd44 | 1430 | UV utf8_fold_flags = 0; |
f8fc2ecf | 1431 | RXi_GET_DECL(prog,progi); |
7918f24d NC |
1432 | |
1433 | PERL_ARGS_ASSERT_FIND_BYCLASS; | |
f8fc2ecf | 1434 | |
6eb5f6b9 JH |
1435 | /* We know what class it must start with. */ |
1436 | switch (OP(c)) { | |
f56b6394 | 1437 | case ANYOFV: |
6eb5f6b9 | 1438 | case ANYOF: |
f56b6394 | 1439 | if (utf8_target || OP(c) == ANYOFV) { |
b1e3e569 KW |
1440 | STRLEN inclasslen = strend - s; |
1441 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1442 | reginclass(prog, c, (U8*)s, &inclasslen, utf8_target)); | |
388cc4de HS |
1443 | } |
1444 | else { | |
6ef69d56 | 1445 | REXEC_FBC_CLASS_SCAN(REGINCLASS(prog, c, (U8*)s)); |
a0d0e21e | 1446 | } |
6eb5f6b9 | 1447 | break; |
f33976b4 | 1448 | case CANY: |
4cadc6a9 | 1449 | REXEC_FBC_SCAN( |
24b23f37 | 1450 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
f33976b4 DB |
1451 | goto got_it; |
1452 | else | |
1453 | tmp = doevery; | |
4cadc6a9 | 1454 | ); |
f33976b4 | 1455 | break; |
2f7f8cb1 KW |
1456 | |
1457 | case EXACTFA: | |
1458 | if (UTF_PATTERN || utf8_target) { | |
1459 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
1460 | goto do_exactf_utf8; | |
1461 | } | |
1462 | fold_array = PL_fold_latin1; /* Latin1 folds are not affected by */ | |
1463 | folder = foldEQ_latin1; /* /a, except the sharp s one which */ | |
1464 | goto do_exactf_non_utf8; /* isn't dealt with by these */ | |
1465 | ||
6eb5f6b9 | 1466 | case EXACTF: |
fac1af77 | 1467 | if (UTF_PATTERN || utf8_target) { |
77a6d856 KW |
1468 | |
1469 | /* regcomp.c already folded this if pattern is in UTF-8 */ | |
1470 | utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0; | |
fac1af77 KW |
1471 | goto do_exactf_utf8; |
1472 | } | |
1473 | fold_array = PL_fold; | |
1474 | folder = foldEQ; | |
1475 | goto do_exactf_non_utf8; | |
1476 | ||
1477 | case EXACTFL: | |
1478 | if (UTF_PATTERN || utf8_target) { | |
17580e7a | 1479 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
fac1af77 KW |
1480 | goto do_exactf_utf8; |
1481 | } | |
1482 | fold_array = PL_fold_locale; | |
1483 | folder = foldEQ_locale; | |
16d951b7 KW |
1484 | goto do_exactf_non_utf8; |
1485 | ||
1486 | case EXACTFU: | |
1487 | if (UTF_PATTERN || utf8_target) { | |
77a6d856 | 1488 | utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0; |
16d951b7 KW |
1489 | goto do_exactf_utf8; |
1490 | } | |
1491 | ||
1492 | /* Any 'ss' in the pattern should have been replaced by regcomp, | |
1493 | * so we don't have to worry here about this single special case | |
1494 | * in the Latin1 range */ | |
1495 | fold_array = PL_fold_latin1; | |
1496 | folder = foldEQ_latin1; | |
fac1af77 KW |
1497 | |
1498 | /* FALL THROUGH */ | |
1499 | ||
1500 | do_exactf_non_utf8: /* Neither pattern nor string are UTF8 */ | |
1501 | ||
1502 | /* The idea in the non-utf8 EXACTF* cases is to first find the | |
1503 | * first character of the EXACTF* node and then, if necessary, | |
1504 | * case-insensitively compare the full text of the node. c1 is the | |
1505 | * first character. c2 is its fold. This logic will not work for | |
1506 | * Unicode semantics and the german sharp ss, which hence should | |
1507 | * not be compiled into a node that gets here. */ | |
1508 | pat_string = STRING(c); | |
1509 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1510 | ||
8a90a8fe KW |
1511 | /* We know that we have to match at least 'ln' bytes (which is the |
1512 | * same as characters, since not utf8). If we have to match 3 | |
1513 | * characters, and there are only 2 availabe, we know without | |
1514 | * trying that it will fail; so don't start a match past the | |
1515 | * required minimum number from the far end */ | |
fac1af77 KW |
1516 | e = HOP3c(strend, -((I32)ln), s); |
1517 | ||
1518 | if (!reginfo && e < s) { | |
1519 | e = s; /* Due to minlen logic of intuit() */ | |
1520 | } | |
1521 | ||
1522 | c1 = *pat_string; | |
1523 | c2 = fold_array[c1]; | |
1524 | if (c1 == c2) { /* If char and fold are the same */ | |
1525 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); | |
1526 | } | |
1527 | else { | |
1528 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); | |
1529 | } | |
1530 | break; | |
1531 | ||
1532 | do_exactf_utf8: | |
e067297c KW |
1533 | { |
1534 | unsigned expansion; | |
1535 | ||
fac1af77 KW |
1536 | |
1537 | /* If one of the operands is in utf8, we can't use the simpler | |
1538 | * folding above, due to the fact that many different characters | |
1539 | * can have the same fold, or portion of a fold, or different- | |
1540 | * length fold */ | |
1541 | pat_string = STRING(c); | |
1542 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1543 | pat_end = pat_string + ln; | |
1544 | lnc = (UTF_PATTERN) /* length to match in characters */ | |
1545 | ? utf8_length((U8 *) pat_string, (U8 *) pat_end) | |
1546 | : ln; | |
1547 | ||
e067297c KW |
1548 | /* We have 'lnc' characters to match in the pattern, but because of |
1549 | * multi-character folding, each character in the target can match | |
1550 | * up to 3 characters (Unicode guarantees it will never exceed | |
1551 | * this) if it is utf8-encoded; and up to 2 if not (based on the | |
1552 | * fact that the Latin 1 folds are already determined, and the | |
1553 | * only multi-char fold in that range is the sharp-s folding to | |
1554 | * 'ss'. Thus, a pattern character can match as little as 1/3 of a | |
89378d8a KW |
1555 | * string character. Adjust lnc accordingly, rounding up, so that |
1556 | * if we need to match at least 4+1/3 chars, that really is 5. */ | |
e067297c | 1557 | expansion = (utf8_target) ? UTF8_MAX_FOLD_CHAR_EXPAND : 2; |
89378d8a | 1558 | lnc = (lnc + expansion - 1) / expansion; |
e067297c KW |
1559 | |
1560 | /* As in the non-UTF8 case, if we have to match 3 characters, and | |
1561 | * only 2 are left, it's guaranteed to fail, so don't start a | |
1562 | * match that would require us to go beyond the end of the string | |
1563 | */ | |
1564 | e = HOP3c(strend, -((I32)lnc), s); | |
fac1af77 KW |
1565 | |
1566 | if (!reginfo && e < s) { | |
1567 | e = s; /* Due to minlen logic of intuit() */ | |
1568 | } | |
1569 | ||
b33105db KW |
1570 | /* XXX Note that we could recalculate e to stop the loop earlier, |
1571 | * as the worst case expansion above will rarely be met, and as we | |
1572 | * go along we would usually find that e moves further to the left. | |
1573 | * This would happen only after we reached the point in the loop | |
1574 | * where if there were no expansion we should fail. Unclear if | |
1575 | * worth the expense */ | |
e067297c | 1576 | |
fac1af77 KW |
1577 | while (s <= e) { |
1578 | char *my_strend= (char *)strend; | |
1579 | if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target, | |
1580 | pat_string, NULL, ln, cBOOL(UTF_PATTERN), utf8_fold_flags) | |
1581 | && (!reginfo || regtry(reginfo, &s)) ) | |
1582 | { | |
1583 | goto got_it; | |
1584 | } | |
bbdd8bad | 1585 | s += (utf8_target) ? UTF8SKIP(s) : 1; |
fac1af77 KW |
1586 | } |
1587 | break; | |
e067297c | 1588 | } |
bbce6d69 | 1589 | case BOUNDL: |
3280af22 | 1590 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1591 | FBC_BOUND(isALNUM_LC, |
1592 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1593 | isALNUM_LC_utf8((U8*)s)); | |
a0ed51b3 | 1594 | break; |
bbce6d69 | 1595 | case NBOUNDL: |
3280af22 | 1596 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1597 | FBC_NBOUND(isALNUM_LC, |
1598 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1599 | isALNUM_LC_utf8((U8*)s)); | |
1600 | break; | |
1601 | case BOUND: | |
1602 | FBC_BOUND(isWORDCHAR, | |
1603 | isALNUM_uni(tmp), | |
1604 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1605 | break; | |
cfaf538b KW |
1606 | case BOUNDA: |
1607 | FBC_BOUND_NOLOAD(isWORDCHAR_A, | |
1608 | isWORDCHAR_A(tmp), | |
1609 | isWORDCHAR_A((U8*)s)); | |
1610 | break; | |
a0d0e21e | 1611 | case NBOUND: |
63ac0dad KW |
1612 | FBC_NBOUND(isWORDCHAR, |
1613 | isALNUM_uni(tmp), | |
1614 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1615 | break; | |
cfaf538b KW |
1616 | case NBOUNDA: |
1617 | FBC_NBOUND_NOLOAD(isWORDCHAR_A, | |
1618 | isWORDCHAR_A(tmp), | |
1619 | isWORDCHAR_A((U8*)s)); | |
1620 | break; | |
63ac0dad KW |
1621 | case BOUNDU: |
1622 | FBC_BOUND(isWORDCHAR_L1, | |
1623 | isALNUM_uni(tmp), | |
1624 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1625 | break; | |
1626 | case NBOUNDU: | |
1627 | FBC_NBOUND(isWORDCHAR_L1, | |
1628 | isALNUM_uni(tmp), | |
1629 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
a0ed51b3 | 1630 | break; |
bbce6d69 | 1631 | case ALNUML: |
4cadc6a9 YO |
1632 | REXEC_FBC_CSCAN_TAINT( |
1633 | isALNUM_LC_utf8((U8*)s), | |
1634 | isALNUM_LC(*s) | |
1635 | ); | |
6895a8aa | 1636 | break; |
980866de KW |
1637 | case ALNUMU: |
1638 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1639 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1640 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1641 | isWORDCHAR_L1((U8) *s) |
1642 | ); | |
6895a8aa | 1643 | break; |
980866de KW |
1644 | case ALNUM: |
1645 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1646 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1647 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1648 | isWORDCHAR((U8) *s) |
1649 | ); | |
6895a8aa | 1650 | break; |
cfaf538b | 1651 | case ALNUMA: |
8e9da4d4 KW |
1652 | /* Don't need to worry about utf8, as it can match only a single |
1653 | * byte invariant character */ | |
cfaf538b | 1654 | REXEC_FBC_CLASS_SCAN( isWORDCHAR_A(*s)); |
6895a8aa | 1655 | break; |
980866de KW |
1656 | case NALNUMU: |
1657 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 | 1658 | LOAD_UTF8_CHARCLASS_ALNUM(), |
359960d4 | 1659 | !swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), |
980866de KW |
1660 | ! isWORDCHAR_L1((U8) *s) |
1661 | ); | |
6895a8aa | 1662 | break; |
a0d0e21e | 1663 | case NALNUM: |
4cadc6a9 | 1664 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1665 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1666 | !swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target), | |
980866de | 1667 | ! isALNUM(*s) |
4cadc6a9 | 1668 | ); |
6895a8aa | 1669 | break; |
cfaf538b | 1670 | case NALNUMA: |
8e9da4d4 KW |
1671 | REXEC_FBC_CSCAN( |
1672 | !isWORDCHAR_A(*s), | |
1673 | !isWORDCHAR_A(*s) | |
1674 | ); | |
1675 | break; | |
bbce6d69 | 1676 | case NALNUML: |
4cadc6a9 YO |
1677 | REXEC_FBC_CSCAN_TAINT( |
1678 | !isALNUM_LC_utf8((U8*)s), | |
1679 | !isALNUM_LC(*s) | |
1680 | ); | |
6895a8aa | 1681 | break; |
980866de KW |
1682 | case SPACEU: |
1683 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1684 | LOAD_UTF8_CHARCLASS_SPACE(), |
1685 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de KW |
1686 | isSPACE_L1((U8) *s) |
1687 | ); | |
6895a8aa | 1688 | break; |
a0d0e21e | 1689 | case SPACE: |
4cadc6a9 | 1690 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1691 | LOAD_UTF8_CHARCLASS_SPACE(), |
1692 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de | 1693 | isSPACE((U8) *s) |
4cadc6a9 | 1694 | ); |
6895a8aa | 1695 | break; |
cfaf538b | 1696 | case SPACEA: |
8e9da4d4 KW |
1697 | /* Don't need to worry about utf8, as it can match only a single |
1698 | * byte invariant character */ | |
cfaf538b | 1699 | REXEC_FBC_CLASS_SCAN( isSPACE_A(*s)); |
6895a8aa | 1700 | break; |
bbce6d69 | 1701 | case SPACEL: |
4cadc6a9 | 1702 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1703 | isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1704 | isSPACE_LC(*s) |
1705 | ); | |
6895a8aa | 1706 | break; |
980866de KW |
1707 | case NSPACEU: |
1708 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1709 | LOAD_UTF8_CHARCLASS_SPACE(), |
1710 | !( *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de KW |
1711 | ! isSPACE_L1((U8) *s) |
1712 | ); | |
6895a8aa | 1713 | break; |
a0d0e21e | 1714 | case NSPACE: |
4cadc6a9 | 1715 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1716 | LOAD_UTF8_CHARCLASS_SPACE(), |
1717 | !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de | 1718 | ! isSPACE((U8) *s) |
4cadc6a9 | 1719 | ); |
6895a8aa | 1720 | break; |
cfaf538b | 1721 | case NSPACEA: |
8e9da4d4 KW |
1722 | REXEC_FBC_CSCAN( |
1723 | !isSPACE_A(*s), | |
1724 | !isSPACE_A(*s) | |
1725 | ); | |
1726 | break; | |
bbce6d69 | 1727 | case NSPACEL: |
4cadc6a9 | 1728 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1729 | !isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1730 | !isSPACE_LC(*s) |
1731 | ); | |
6895a8aa | 1732 | break; |
a0d0e21e | 1733 | case DIGIT: |
4cadc6a9 | 1734 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1735 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1736 | swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1737 | isDIGIT(*s) |
1738 | ); | |
6895a8aa | 1739 | break; |
cfaf538b | 1740 | case DIGITA: |
8e9da4d4 KW |
1741 | /* Don't need to worry about utf8, as it can match only a single |
1742 | * byte invariant character */ | |
cfaf538b | 1743 | REXEC_FBC_CLASS_SCAN( isDIGIT_A(*s)); |
6895a8aa | 1744 | break; |
b8c5462f | 1745 | case DIGITL: |
4cadc6a9 YO |
1746 | REXEC_FBC_CSCAN_TAINT( |
1747 | isDIGIT_LC_utf8((U8*)s), | |
1748 | isDIGIT_LC(*s) | |
1749 | ); | |
6895a8aa | 1750 | break; |
a0d0e21e | 1751 | case NDIGIT: |
4cadc6a9 | 1752 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1753 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1754 | !swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1755 | !isDIGIT(*s) |
1756 | ); | |
6895a8aa | 1757 | break; |
cfaf538b | 1758 | case NDIGITA: |
8e9da4d4 KW |
1759 | REXEC_FBC_CSCAN( |
1760 | !isDIGIT_A(*s), | |
1761 | !isDIGIT_A(*s) | |
1762 | ); | |
1763 | break; | |
b8c5462f | 1764 | case NDIGITL: |
4cadc6a9 YO |
1765 | REXEC_FBC_CSCAN_TAINT( |
1766 | !isDIGIT_LC_utf8((U8*)s), | |
1767 | !isDIGIT_LC(*s) | |
1768 | ); | |
6895a8aa | 1769 | break; |
e1d1eefb YO |
1770 | case LNBREAK: |
1771 | REXEC_FBC_CSCAN( | |
1772 | is_LNBREAK_utf8(s), | |
1773 | is_LNBREAK_latin1(s) | |
1774 | ); | |
6895a8aa | 1775 | break; |
e1d1eefb YO |
1776 | case VERTWS: |
1777 | REXEC_FBC_CSCAN( | |
1778 | is_VERTWS_utf8(s), | |
1779 | is_VERTWS_latin1(s) | |
1780 | ); | |
6895a8aa | 1781 | break; |
e1d1eefb YO |
1782 | case NVERTWS: |
1783 | REXEC_FBC_CSCAN( | |
1784 | !is_VERTWS_utf8(s), | |
1785 | !is_VERTWS_latin1(s) | |
1786 | ); | |
6895a8aa | 1787 | break; |
e1d1eefb YO |
1788 | case HORIZWS: |
1789 | REXEC_FBC_CSCAN( | |
1790 | is_HORIZWS_utf8(s), | |
1791 | is_HORIZWS_latin1(s) | |
1792 | ); | |
6895a8aa | 1793 | break; |
e1d1eefb YO |
1794 | case NHORIZWS: |
1795 | REXEC_FBC_CSCAN( | |
1796 | !is_HORIZWS_utf8(s), | |
1797 | !is_HORIZWS_latin1(s) | |
1798 | ); | |
6895a8aa | 1799 | break; |
1de06328 YO |
1800 | case AHOCORASICKC: |
1801 | case AHOCORASICK: | |
07be1b83 | 1802 | { |
a0a388a1 | 1803 | DECL_TRIE_TYPE(c); |
07be1b83 YO |
1804 | /* what trie are we using right now */ |
1805 | reg_ac_data *aho | |
f8fc2ecf | 1806 | = (reg_ac_data*)progi->data->data[ ARG( c ) ]; |
3251b653 NC |
1807 | reg_trie_data *trie |
1808 | = (reg_trie_data*)progi->data->data[ aho->trie ]; | |
85fbaab2 | 1809 | HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); |
07be1b83 YO |
1810 | |
1811 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1812 | #ifdef DEBUGGING |
07be1b83 | 1813 | const char *real_start = s; |
6148ee25 | 1814 | #endif |
07be1b83 | 1815 | STRLEN maxlen = trie->maxlen; |
be8e71aa YO |
1816 | SV *sv_points; |
1817 | U8 **points; /* map of where we were in the input string | |
786e8c11 | 1818 | when reading a given char. For ASCII this |
be8e71aa | 1819 | is unnecessary overhead as the relationship |
38a44b82 NC |
1820 | is always 1:1, but for Unicode, especially |
1821 | case folded Unicode this is not true. */ | |
f9e705e8 | 1822 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
786e8c11 YO |
1823 | U8 *bitmap=NULL; |
1824 | ||
07be1b83 YO |
1825 | |
1826 | GET_RE_DEBUG_FLAGS_DECL; | |
1827 | ||
be8e71aa YO |
1828 | /* We can't just allocate points here. We need to wrap it in |
1829 | * an SV so it gets freed properly if there is a croak while | |
1830 | * running the match */ | |
1831 | ENTER; | |
1832 | SAVETMPS; | |
1833 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1834 | SvCUR_set(sv_points, | |
1835 | maxlen * sizeof(U8 *)); | |
1836 | SvPOK_on(sv_points); | |
1837 | sv_2mortal(sv_points); | |
1838 | points=(U8**)SvPV_nolen(sv_points ); | |
1de06328 YO |
1839 | if ( trie_type != trie_utf8_fold |
1840 | && (trie->bitmap || OP(c)==AHOCORASICKC) ) | |
1841 | { | |
786e8c11 YO |
1842 | if (trie->bitmap) |
1843 | bitmap=(U8*)trie->bitmap; | |
1844 | else | |
1845 | bitmap=(U8*)ANYOF_BITMAP(c); | |
07be1b83 | 1846 | } |
786e8c11 YO |
1847 | /* this is the Aho-Corasick algorithm modified a touch |
1848 | to include special handling for long "unknown char" | |
1849 | sequences. The basic idea being that we use AC as long | |
1850 | as we are dealing with a possible matching char, when | |
1851 | we encounter an unknown char (and we have not encountered | |
1852 | an accepting state) we scan forward until we find a legal | |
1853 | starting char. | |
1854 | AC matching is basically that of trie matching, except | |
1855 | that when we encounter a failing transition, we fall back | |
1856 | to the current states "fail state", and try the current char | |
1857 | again, a process we repeat until we reach the root state, | |
1858 | state 1, or a legal transition. If we fail on the root state | |
1859 | then we can either terminate if we have reached an accepting | |
1860 | state previously, or restart the entire process from the beginning | |
1861 | if we have not. | |
1862 | ||
1863 | */ | |
07be1b83 YO |
1864 | while (s <= last_start) { |
1865 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1866 | U8 *uc = (U8*)s; | |
1867 | U16 charid = 0; | |
1868 | U32 base = 1; | |
1869 | U32 state = 1; | |
1870 | UV uvc = 0; | |
1871 | STRLEN len = 0; | |
1872 | STRLEN foldlen = 0; | |
1873 | U8 *uscan = (U8*)NULL; | |
1874 | U8 *leftmost = NULL; | |
786e8c11 YO |
1875 | #ifdef DEBUGGING |
1876 | U32 accepted_word= 0; | |
1877 | #endif | |
07be1b83 YO |
1878 | U32 pointpos = 0; |
1879 | ||
1880 | while ( state && uc <= (U8*)strend ) { | |
1881 | int failed=0; | |
786e8c11 YO |
1882 | U32 word = aho->states[ state ].wordnum; |
1883 | ||
1de06328 YO |
1884 | if( state==1 ) { |
1885 | if ( bitmap ) { | |
1886 | DEBUG_TRIE_EXECUTE_r( | |
1887 | if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1888 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1889 | (char *)uc, utf8_target ); |
1de06328 YO |
1890 | PerlIO_printf( Perl_debug_log, |
1891 | " Scanning for legal start char...\n"); | |
1892 | } | |
d085b490 YO |
1893 | ); |
1894 | if (utf8_target) { | |
1895 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1896 | uc += UTF8SKIP(uc); | |
1897 | } | |
1898 | } else { | |
1899 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1900 | uc++; | |
1901 | } | |
1902 | } | |
1de06328 | 1903 | s= (char *)uc; |
786e8c11 | 1904 | } |
786e8c11 YO |
1905 | if (uc >(U8*)last_start) break; |
1906 | } | |
1907 | ||
1908 | if ( word ) { | |
2e64971a | 1909 | U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ]; |
786e8c11 YO |
1910 | if (!leftmost || lpos < leftmost) { |
1911 | DEBUG_r(accepted_word=word); | |
07be1b83 | 1912 | leftmost= lpos; |
786e8c11 | 1913 | } |
07be1b83 | 1914 | if (base==0) break; |
786e8c11 | 1915 | |
07be1b83 YO |
1916 | } |
1917 | points[pointpos++ % maxlen]= uc; | |
55eed653 NC |
1918 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
1919 | uscan, len, uvc, charid, foldlen, | |
1920 | foldbuf, uniflags); | |
786e8c11 YO |
1921 | DEBUG_TRIE_EXECUTE_r({ |
1922 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1923 | s, utf8_target ); |
07be1b83 | 1924 | PerlIO_printf(Perl_debug_log, |
786e8c11 YO |
1925 | " Charid:%3u CP:%4"UVxf" ", |
1926 | charid, uvc); | |
1927 | }); | |
07be1b83 YO |
1928 | |
1929 | do { | |
6148ee25 | 1930 | #ifdef DEBUGGING |
786e8c11 | 1931 | word = aho->states[ state ].wordnum; |
6148ee25 | 1932 | #endif |
07be1b83 YO |
1933 | base = aho->states[ state ].trans.base; |
1934 | ||
786e8c11 YO |
1935 | DEBUG_TRIE_EXECUTE_r({ |
1936 | if (failed) | |
1937 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1938 | s, utf8_target ); |
07be1b83 | 1939 | PerlIO_printf( Perl_debug_log, |
786e8c11 YO |
1940 | "%sState: %4"UVxf", word=%"UVxf, |
1941 | failed ? " Fail transition to " : "", | |
1942 | (UV)state, (UV)word); | |
1943 | }); | |
07be1b83 YO |
1944 | if ( base ) { |
1945 | U32 tmp; | |
6dd2be57 | 1946 | I32 offset; |
07be1b83 | 1947 | if (charid && |
6dd2be57 DM |
1948 | ( ((offset = base + charid |
1949 | - 1 - trie->uniquecharcount)) >= 0) | |
1950 | && ((U32)offset < trie->lasttrans) | |
1951 | && trie->trans[offset].check == state | |
1952 | && (tmp=trie->trans[offset].next)) | |
07be1b83 | 1953 | { |
786e8c11 YO |
1954 | DEBUG_TRIE_EXECUTE_r( |
1955 | PerlIO_printf( Perl_debug_log," - legal\n")); | |
07be1b83 YO |
1956 | state = tmp; |
1957 | break; | |
1958 | } | |
1959 | else { | |
786e8c11 YO |
1960 | DEBUG_TRIE_EXECUTE_r( |
1961 | PerlIO_printf( Perl_debug_log," - fail\n")); | |
1962 | failed = 1; | |
1963 | state = aho->fail[state]; | |
07be1b83 YO |
1964 | } |
1965 | } | |
1966 | else { | |
1967 | /* we must be accepting here */ | |
786e8c11 YO |
1968 | DEBUG_TRIE_EXECUTE_r( |
1969 | PerlIO_printf( Perl_debug_log," - accepting\n")); | |
1970 | failed = 1; | |
07be1b83 YO |
1971 | break; |
1972 | } | |
1973 | } while(state); | |
786e8c11 | 1974 | uc += len; |
07be1b83 YO |
1975 | if (failed) { |
1976 | if (leftmost) | |
1977 | break; | |
786e8c11 | 1978 | if (!state) state = 1; |
07be1b83 YO |
1979 | } |
1980 | } | |
1981 | if ( aho->states[ state ].wordnum ) { | |
2e64971a | 1982 | U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ]; |
786e8c11 YO |
1983 | if (!leftmost || lpos < leftmost) { |
1984 | DEBUG_r(accepted_word=aho->states[ state ].wordnum); | |
07be1b83 | 1985 | leftmost = lpos; |
786e8c11 | 1986 | } |
07be1b83 | 1987 | } |
07be1b83 YO |
1988 | if (leftmost) { |
1989 | s = (char*)leftmost; | |
786e8c11 YO |
1990 | DEBUG_TRIE_EXECUTE_r({ |
1991 | PerlIO_printf( | |
70685ca0 JH |
1992 | Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", |
1993 | (UV)accepted_word, (IV)(s - real_start) | |
786e8c11 YO |
1994 | ); |
1995 | }); | |
24b23f37 | 1996 | if (!reginfo || regtry(reginfo, &s)) { |
be8e71aa YO |
1997 | FREETMPS; |
1998 | LEAVE; | |
07be1b83 | 1999 | goto got_it; |
be8e71aa | 2000 | } |
07be1b83 | 2001 | s = HOPc(s,1); |
786e8c11 YO |
2002 | DEBUG_TRIE_EXECUTE_r({ |
2003 | PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); | |
2004 | }); | |
07be1b83 | 2005 | } else { |
786e8c11 YO |
2006 | DEBUG_TRIE_EXECUTE_r( |
2007 | PerlIO_printf( Perl_debug_log,"No match.\n")); | |
07be1b83 YO |
2008 | break; |
2009 | } | |
2010 | } | |
be8e71aa YO |
2011 | FREETMPS; |
2012 | LEAVE; | |
07be1b83 YO |
2013 | } |
2014 | break; | |
b3c9acc1 | 2015 | default: |
3c3eec57 GS |
2016 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); |
2017 | break; | |
d6a28714 | 2018 | } |
6eb5f6b9 JH |
2019 | return 0; |
2020 | got_it: | |
2021 | return s; | |
2022 | } | |
2023 | ||
fae667d5 | 2024 | |
6eb5f6b9 JH |
2025 | /* |
2026 | - regexec_flags - match a regexp against a string | |
2027 | */ | |
2028 | I32 | |
288b8c02 | 2029 | Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend, |
6eb5f6b9 JH |
2030 | char *strbeg, I32 minend, SV *sv, void *data, U32 flags) |
2031 | /* strend: pointer to null at end of string */ | |
2032 | /* strbeg: real beginning of string */ | |
2033 | /* minend: end of match must be >=minend after stringarg. */ | |
58e23c8d YO |
2034 | /* data: May be used for some additional optimizations. |
2035 | Currently its only used, with a U32 cast, for transmitting | |
2036 | the ganch offset when doing a /g match. This will change */ | |
6eb5f6b9 JH |
2037 | /* nosave: For optimizations. */ |
2038 | { | |
97aff369 | 2039 | dVAR; |
288b8c02 | 2040 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
24b23f37 | 2041 | /*register*/ char *s; |
6eb5f6b9 | 2042 | register regnode *c; |
24b23f37 | 2043 | /*register*/ char *startpos = stringarg; |
6eb5f6b9 JH |
2044 | I32 minlen; /* must match at least this many chars */ |
2045 | I32 dontbother = 0; /* how many characters not to try at end */ | |
6eb5f6b9 JH |
2046 | I32 end_shift = 0; /* Same for the end. */ /* CC */ |
2047 | I32 scream_pos = -1; /* Internal iterator of scream. */ | |
ccac19ea | 2048 | char *scream_olds = NULL; |
f2ed9b32 | 2049 | const bool utf8_target = cBOOL(DO_UTF8(sv)); |
2757e526 | 2050 | I32 multiline; |
f8fc2ecf | 2051 | RXi_GET_DECL(prog,progi); |
3b0527fe | 2052 | regmatch_info reginfo; /* create some info to pass to regtry etc */ |
e9105d30 | 2053 | regexp_paren_pair *swap = NULL; |
a3621e74 YO |
2054 | GET_RE_DEBUG_FLAGS_DECL; |
2055 | ||
7918f24d | 2056 | PERL_ARGS_ASSERT_REGEXEC_FLAGS; |
9d4ba2ae | 2057 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
2058 | |
2059 | /* Be paranoid... */ | |
2060 | if (prog == NULL || startpos == NULL) { | |
2061 | Perl_croak(aTHX_ "NULL regexp parameter"); | |
2062 | return 0; | |
2063 | } | |
2064 | ||
bbe252da | 2065 | multiline = prog->extflags & RXf_PMf_MULTILINE; |
288b8c02 | 2066 | reginfo.prog = rx; /* Yes, sorry that this is confusing. */ |
2757e526 | 2067 | |
f2ed9b32 | 2068 | RX_MATCH_UTF8_set(rx, utf8_target); |
1de06328 | 2069 | DEBUG_EXECUTE_r( |
f2ed9b32 | 2070 | debug_start_match(rx, utf8_target, startpos, strend, |
1de06328 YO |
2071 | "Matching"); |
2072 | ); | |
bac06658 | 2073 | |
6eb5f6b9 | 2074 | minlen = prog->minlen; |
1de06328 YO |
2075 | |
2076 | if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { | |
a3621e74 | 2077 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
2078 | "String too short [regexec_flags]...\n")); |
2079 | goto phooey; | |
1aa99e6b | 2080 | } |
6eb5f6b9 | 2081 | |
1de06328 | 2082 | |
6eb5f6b9 | 2083 | /* Check validity of program. */ |
f8fc2ecf | 2084 | if (UCHARAT(progi->program) != REG_MAGIC) { |
6eb5f6b9 JH |
2085 | Perl_croak(aTHX_ "corrupted regexp program"); |
2086 | } | |
2087 | ||
2088 | PL_reg_flags = 0; | |
2089 | PL_reg_eval_set = 0; | |
2090 | PL_reg_maxiter = 0; | |
2091 | ||
3c8556c3 | 2092 | if (RX_UTF8(rx)) |
6eb5f6b9 JH |
2093 | PL_reg_flags |= RF_utf8; |
2094 | ||
2095 | /* Mark beginning of line for ^ and lookbehind. */ | |
3b0527fe | 2096 | reginfo.bol = startpos; /* XXX not used ??? */ |
6eb5f6b9 | 2097 | PL_bostr = strbeg; |
3b0527fe | 2098 | reginfo.sv = sv; |
6eb5f6b9 JH |
2099 | |
2100 | /* Mark end of line for $ (and such) */ | |
2101 | PL_regeol = strend; | |
2102 | ||
2103 | /* see how far we have to get to not match where we matched before */ | |
3b0527fe | 2104 | reginfo.till = startpos+minend; |
6eb5f6b9 | 2105 | |
6eb5f6b9 JH |
2106 | /* If there is a "must appear" string, look for it. */ |
2107 | s = startpos; | |
2108 | ||
bbe252da | 2109 | if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ |
6eb5f6b9 | 2110 | MAGIC *mg; |
2c296965 | 2111 | if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */ |
58e23c8d | 2112 | reginfo.ganch = startpos + prog->gofs; |
2c296965 | 2113 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2114 | "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs)); |
2c296965 | 2115 | } else if (sv && SvTYPE(sv) >= SVt_PVMG |
6eb5f6b9 | 2116 | && SvMAGIC(sv) |
14befaf4 DM |
2117 | && (mg = mg_find(sv, PERL_MAGIC_regex_global)) |
2118 | && mg->mg_len >= 0) { | |
3b0527fe | 2119 | reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ |
2c296965 | 2120 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2121 | "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len)); |
2c296965 | 2122 | |
bbe252da | 2123 | if (prog->extflags & RXf_ANCH_GPOS) { |
3b0527fe | 2124 | if (s > reginfo.ganch) |
6eb5f6b9 | 2125 | goto phooey; |
58e23c8d | 2126 | s = reginfo.ganch - prog->gofs; |
2c296965 | 2127 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2128 | "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs)); |
c584a96e YO |
2129 | if (s < strbeg) |
2130 | goto phooey; | |
6eb5f6b9 JH |
2131 | } |
2132 | } | |
58e23c8d | 2133 | else if (data) { |
70685ca0 | 2134 | reginfo.ganch = strbeg + PTR2UV(data); |
2c296965 YO |
2135 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2136 | "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data))); | |
2137 | ||
2138 | } else { /* pos() not defined */ | |
3b0527fe | 2139 | reginfo.ganch = strbeg; |
2c296965 YO |
2140 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2141 | "GPOS: reginfo.ganch = strbeg\n")); | |
2142 | } | |
6eb5f6b9 | 2143 | } |
288b8c02 | 2144 | if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { |
e9105d30 GG |
2145 | /* We have to be careful. If the previous successful match |
2146 | was from this regex we don't want a subsequent partially | |
2147 | successful match to clobber the old results. | |
2148 | So when we detect this possibility we add a swap buffer | |
2149 | to the re, and switch the buffer each match. If we fail | |
2150 | we switch it back, otherwise we leave it swapped. | |
2151 | */ | |
2152 | swap = prog->offs; | |
2153 | /* do we need a save destructor here for eval dies? */ | |
2154 | Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); | |
c74340f9 | 2155 | } |
a0714e2c | 2156 | if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { |
6eb5f6b9 JH |
2157 | re_scream_pos_data d; |
2158 | ||
2159 | d.scream_olds = &scream_olds; | |
2160 | d.scream_pos = &scream_pos; | |
288b8c02 | 2161 | s = re_intuit_start(rx, sv, s, strend, flags, &d); |
3fa9c3d7 | 2162 | if (!s) { |
a3621e74 | 2163 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n")); |
6eb5f6b9 | 2164 | goto phooey; /* not present */ |
3fa9c3d7 | 2165 | } |
6eb5f6b9 JH |
2166 | } |
2167 | ||
1de06328 | 2168 | |
6eb5f6b9 JH |
2169 | |
2170 | /* Simplest case: anchored match need be tried only once. */ | |
2171 | /* [unless only anchor is BOL and multiline is set] */ | |
bbe252da | 2172 | if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { |
24b23f37 | 2173 | if (s == startpos && regtry(®info, &startpos)) |
6eb5f6b9 | 2174 | goto got_it; |
bbe252da YO |
2175 | else if (multiline || (prog->intflags & PREGf_IMPLICIT) |
2176 | || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ | |
6eb5f6b9 JH |
2177 | { |
2178 | char *end; | |
2179 | ||
2180 | if (minlen) | |
2181 | dontbother = minlen - 1; | |
1aa99e6b | 2182 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 2183 | /* for multiline we only have to try after newlines */ |
33b8afdf | 2184 | if (prog->check_substr || prog->check_utf8) { |
92f3d482 YO |
2185 | /* because of the goto we can not easily reuse the macros for bifurcating the |
2186 | unicode/non-unicode match modes here like we do elsewhere - demerphq */ | |
2187 | if (utf8_target) { | |
2188 | if (s == startpos) | |
2189 | goto after_try_utf8; | |
2190 | while (1) { | |
2191 | if (regtry(®info, &s)) { | |
2192 | goto got_it; | |
2193 | } | |
2194 | after_try_utf8: | |
2195 | if (s > end) { | |
2196 | goto phooey; | |
2197 | } | |
2198 | if (prog->extflags & RXf_USE_INTUIT) { | |
2199 | s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL); | |
2200 | if (!s) { | |
2201 | goto phooey; | |
2202 | } | |
2203 | } | |
2204 | else { | |
2205 | s += UTF8SKIP(s); | |
2206 | } | |
2207 | } | |
2208 | } /* end search for check string in unicode */ | |
2209 | else { | |
2210 | if (s == startpos) { | |
2211 | goto after_try_latin; | |
2212 | } | |
2213 | while (1) { | |
2214 | if (regtry(®info, &s)) { | |
2215 | goto got_it; | |
2216 | } | |
2217 | after_try_latin: | |
2218 | if (s > end) { | |
2219 | goto phooey; | |
2220 | } | |
2221 | if (prog->extflags & RXf_USE_INTUIT) { | |
2222 | s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL); | |
2223 | if (!s) { | |
2224 | goto phooey; | |
2225 | } | |
2226 | } | |
2227 | else { | |
2228 | s++; | |
2229 | } | |
2230 | } | |
2231 | } /* end search for check string in latin*/ | |
2232 | } /* end search for check string */ | |
2233 | else { /* search for newline */ | |
2234 | if (s > startpos) { | |
2235 | /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/ | |
6eb5f6b9 | 2236 | s--; |
92f3d482 YO |
2237 | } |
2238 | /* We can use a more efficient search as newlines are the same in unicode as they are in latin */ | |
6eb5f6b9 JH |
2239 | while (s < end) { |
2240 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ | |
24b23f37 | 2241 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2242 | goto got_it; |
2243 | } | |
92f3d482 YO |
2244 | } |
2245 | } /* end search for newline */ | |
2246 | } /* end anchored/multiline check string search */ | |
6eb5f6b9 | 2247 | goto phooey; |
bbe252da | 2248 | } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) |
f9f4320a | 2249 | { |
486ec47a | 2250 | /* the warning about reginfo.ganch being used without initialization |
bbe252da | 2251 | is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN |
f9f4320a | 2252 | and we only enter this block when the same bit is set. */ |
58e23c8d | 2253 | char *tmp_s = reginfo.ganch - prog->gofs; |
c584a96e YO |
2254 | |
2255 | if (tmp_s >= strbeg && regtry(®info, &tmp_s)) | |
6eb5f6b9 JH |
2256 | goto got_it; |
2257 | goto phooey; | |
2258 | } | |
2259 | ||
2260 | /* Messy cases: unanchored match. */ | |
bbe252da | 2261 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { |
6eb5f6b9 | 2262 | /* we have /x+whatever/ */ |
f2ed9b32 | 2263 | /* it must be a one character string (XXXX Except UTF_PATTERN?) */ |
33b8afdf | 2264 | char ch; |
bf93d4cc GS |
2265 | #ifdef DEBUGGING |
2266 | int did_match = 0; | |
2267 | #endif | |
f2ed9b32 KW |
2268 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2269 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2270 | ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0]; | |
bf93d4cc | 2271 | |
f2ed9b32 | 2272 | if (utf8_target) { |
4cadc6a9 | 2273 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2274 | if (*s == ch) { |
a3621e74 | 2275 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2276 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2277 | s += UTF8SKIP(s); |
2278 | while (s < strend && *s == ch) | |
2279 | s += UTF8SKIP(s); | |
2280 | } | |
4cadc6a9 | 2281 | ); |
6eb5f6b9 JH |
2282 | } |
2283 | else { | |
4cadc6a9 | 2284 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2285 | if (*s == ch) { |
a3621e74 | 2286 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2287 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2288 | s++; |
2289 | while (s < strend && *s == ch) | |
2290 | s++; | |
2291 | } | |
4cadc6a9 | 2292 | ); |
6eb5f6b9 | 2293 | } |
a3621e74 | 2294 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 2295 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
2296 | "Did not find anchored character...\n") |
2297 | ); | |
6eb5f6b9 | 2298 | } |
a0714e2c SS |
2299 | else if (prog->anchored_substr != NULL |
2300 | || prog->anchored_utf8 != NULL | |
2301 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
2302 | && prog->float_max_offset < strend - s)) { |
2303 | SV *must; | |
2304 | I32 back_max; | |
2305 | I32 back_min; | |
2306 | char *last; | |
6eb5f6b9 | 2307 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
2308 | #ifdef DEBUGGING |
2309 | int did_match = 0; | |
2310 | #endif | |
33b8afdf | 2311 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 KW |
2312 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2313 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2314 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; | |
33b8afdf JH |
2315 | back_max = back_min = prog->anchored_offset; |
2316 | } else { | |
f2ed9b32 KW |
2317 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2318 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2319 | must = utf8_target ? prog->float_utf8 : prog->float_substr; | |
33b8afdf JH |
2320 | back_max = prog->float_max_offset; |
2321 | back_min = prog->float_min_offset; | |
2322 | } | |
1de06328 YO |
2323 | |
2324 | ||
33b8afdf JH |
2325 | if (must == &PL_sv_undef) |
2326 | /* could not downgrade utf8 check substring, so must fail */ | |
2327 | goto phooey; | |
2328 | ||
1de06328 YO |
2329 | if (back_min<0) { |
2330 | last = strend; | |
2331 | } else { | |
2332 | last = HOP3c(strend, /* Cannot start after this */ | |
2333 | -(I32)(CHR_SVLEN(must) | |
2334 | - (SvTAIL(must) != 0) + back_min), strbeg); | |
2335 | } | |
6eb5f6b9 JH |
2336 | if (s > PL_bostr) |
2337 | last1 = HOPc(s, -1); | |
2338 | else | |
2339 | last1 = s - 1; /* bogus */ | |
2340 | ||
a0288114 | 2341 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 JH |
2342 | check_substr==must. */ |
2343 | scream_pos = -1; | |
2344 | dontbother = end_shift; | |
2345 | strend = HOPc(strend, -dontbother); | |
2346 | while ( (s <= last) && | |
f1905e1b | 2347 | ((flags & REXEC_SCREAM) && SvSCREAM(sv) |
1de06328 | 2348 | ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg, |
6eb5f6b9 | 2349 | end_shift, &scream_pos, 0)) |
1de06328 | 2350 | : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), |
9041c2e3 | 2351 | (unsigned char*)strend, must, |
7fba1cd6 | 2352 | multiline ? FBMrf_MULTILINE : 0))) ) { |
4addbd3b | 2353 | /* we may be pointing at the wrong string */ |
07bc277f | 2354 | if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog)) |
3f7c398e | 2355 | s = strbeg + (s - SvPVX_const(sv)); |
a3621e74 | 2356 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
2357 | if (HOPc(s, -back_max) > last1) { |
2358 | last1 = HOPc(s, -back_min); | |
2359 | s = HOPc(s, -back_max); | |
2360 | } | |
2361 | else { | |
52657f30 | 2362 | char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; |
6eb5f6b9 JH |
2363 | |
2364 | last1 = HOPc(s, -back_min); | |
52657f30 | 2365 | s = t; |
6eb5f6b9 | 2366 | } |
f2ed9b32 | 2367 | if (utf8_target) { |
6eb5f6b9 | 2368 | while (s <= last1) { |
24b23f37 | 2369 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2370 | goto got_it; |
2371 | s += UTF8SKIP(s); | |
2372 | } | |
2373 | } | |
2374 | else { | |
2375 | while (s <= last1) { | |
24b23f37 | 2376 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2377 | goto got_it; |
2378 | s++; | |
2379 | } | |
2380 | } | |
2381 | } | |
ab3bbdeb | 2382 | DEBUG_EXECUTE_r(if (!did_match) { |
f2ed9b32 | 2383 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
2384 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
2385 | PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", | |
33b8afdf | 2386 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc | 2387 | ? "anchored" : "floating"), |
ab3bbdeb YO |
2388 | quoted, RE_SV_TAIL(must)); |
2389 | }); | |
6eb5f6b9 JH |
2390 | goto phooey; |
2391 | } | |
f8fc2ecf | 2392 | else if ( (c = progi->regstclass) ) { |
f14c76ed | 2393 | if (minlen) { |
f8fc2ecf | 2394 | const OPCODE op = OP(progi->regstclass); |
66e933ab | 2395 | /* don't bother with what can't match */ |
786e8c11 | 2396 | if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) |
f14c76ed RGS |
2397 | strend = HOPc(strend, -(minlen - 1)); |
2398 | } | |
a3621e74 | 2399 | DEBUG_EXECUTE_r({ |
be8e71aa | 2400 | SV * const prop = sv_newmortal(); |
32fc9b6a | 2401 | regprop(prog, prop, c); |
0df25f3d | 2402 | { |
f2ed9b32 | 2403 | RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb | 2404 | s,strend-s,60); |
0df25f3d | 2405 | PerlIO_printf(Perl_debug_log, |
1c8f8eb1 | 2406 | "Matching stclass %.*s against %s (%d bytes)\n", |
e4f74956 | 2407 | (int)SvCUR(prop), SvPVX_const(prop), |
ab3bbdeb | 2408 | quoted, (int)(strend - s)); |
0df25f3d | 2409 | } |
ffc61ed2 | 2410 | }); |
3b0527fe | 2411 | if (find_byclass(prog, c, s, strend, ®info)) |
6eb5f6b9 | 2412 | goto got_it; |
07be1b83 | 2413 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
2414 | } |
2415 | else { | |
2416 | dontbother = 0; | |
a0714e2c | 2417 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 2418 | /* Trim the end. */ |
d6a28714 | 2419 | char *last; |
33b8afdf JH |
2420 | SV* float_real; |
2421 | ||
f2ed9b32 KW |
2422 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2423 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2424 | float_real = utf8_target ? prog->float_utf8 : prog->float_substr; | |
d6a28714 | 2425 | |
f1905e1b | 2426 | if ((flags & REXEC_SCREAM) && SvSCREAM(sv)) { |
33b8afdf | 2427 | last = screaminstr(sv, float_real, s - strbeg, |
d6a28714 JH |
2428 | end_shift, &scream_pos, 1); /* last one */ |
2429 | if (!last) | |
ffc61ed2 | 2430 | last = scream_olds; /* Only one occurrence. */ |
4addbd3b | 2431 | /* we may be pointing at the wrong string */ |
07bc277f | 2432 | else if (RXp_MATCH_COPIED(prog)) |
3f7c398e | 2433 | s = strbeg + (s - SvPVX_const(sv)); |
b8c5462f | 2434 | } |
d6a28714 JH |
2435 | else { |
2436 | STRLEN len; | |
cfd0369c | 2437 | const char * const little = SvPV_const(float_real, len); |
d6a28714 | 2438 | |
33b8afdf | 2439 | if (SvTAIL(float_real)) { |
d6a28714 JH |
2440 | if (memEQ(strend - len + 1, little, len - 1)) |
2441 | last = strend - len + 1; | |
7fba1cd6 | 2442 | else if (!multiline) |
9041c2e3 | 2443 | last = memEQ(strend - len, little, len) |
bd61b366 | 2444 | ? strend - len : NULL; |
b8c5462f | 2445 | else |
d6a28714 JH |
2446 | goto find_last; |
2447 | } else { | |
2448 | find_last: | |
9041c2e3 | 2449 | if (len) |
d6a28714 | 2450 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 2451 | else |
a0288114 | 2452 | last = strend; /* matching "$" */ |
b8c5462f | 2453 | } |
b8c5462f | 2454 | } |
bf93d4cc | 2455 | if (last == NULL) { |
6bda09f9 YO |
2456 | DEBUG_EXECUTE_r( |
2457 | PerlIO_printf(Perl_debug_log, | |
2458 | "%sCan't trim the tail, match fails (should not happen)%s\n", | |
2459 | PL_colors[4], PL_colors[5])); | |
bf93d4cc GS |
2460 | goto phooey; /* Should not happen! */ |
2461 | } | |
d6a28714 JH |
2462 | dontbother = strend - last + prog->float_min_offset; |
2463 | } | |
2464 | if (minlen && (dontbother < minlen)) | |
2465 | dontbother = minlen - 1; | |
2466 | strend -= dontbother; /* this one's always in bytes! */ | |
2467 | /* We don't know much -- general case. */ | |
f2ed9b32 | 2468 | if (utf8_target) { |
d6a28714 | 2469 | for (;;) { |
24b23f37 | 2470 | if (regtry(®info, &s)) |
d6a28714 JH |
2471 | goto got_it; |
2472 | if (s >= strend) | |
2473 | break; | |
b8c5462f | 2474 | s += UTF8SKIP(s); |
d6a28714 JH |
2475 | }; |
2476 | } | |
2477 | else { | |
2478 | do { | |
24b23f37 | 2479 | if (regtry(®info, &s)) |
d6a28714 JH |
2480 | goto got_it; |
2481 | } while (s++ < strend); | |
2482 | } | |
2483 | } | |
2484 | ||
2485 | /* Failure. */ | |
2486 | goto phooey; | |
2487 | ||
2488 | got_it: | |
e9105d30 | 2489 | Safefree(swap); |
288b8c02 | 2490 | RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted); |
d6a28714 | 2491 | |
19b95bf0 | 2492 | if (PL_reg_eval_set) |
4f639d21 | 2493 | restore_pos(aTHX_ prog); |
5daac39c NC |
2494 | if (RXp_PAREN_NAMES(prog)) |
2495 | (void)hv_iterinit(RXp_PAREN_NAMES(prog)); | |
d6a28714 JH |
2496 | |
2497 | /* make sure $`, $&, $', and $digit will work later */ | |
2498 | if ( !(flags & REXEC_NOT_FIRST) ) { | |
288b8c02 | 2499 | RX_MATCH_COPY_FREE(rx); |
d6a28714 | 2500 | if (flags & REXEC_COPY_STR) { |
be8e71aa | 2501 | const I32 i = PL_regeol - startpos + (stringarg - strbeg); |
f8c7b90f | 2502 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2503 | if ((SvIsCOW(sv) |
2504 | || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { | |
2505 | if (DEBUG_C_TEST) { | |
2506 | PerlIO_printf(Perl_debug_log, | |
2507 | "Copy on write: regexp capture, type %d\n", | |
2508 | (int) SvTYPE(sv)); | |
2509 | } | |
2510 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); | |
d5263905 | 2511 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
ed252734 NC |
2512 | assert (SvPOKp(prog->saved_copy)); |
2513 | } else | |
2514 | #endif | |
2515 | { | |
288b8c02 | 2516 | RX_MATCH_COPIED_on(rx); |
ed252734 NC |
2517 | s = savepvn(strbeg, i); |
2518 | prog->subbeg = s; | |
2519 | } | |
d6a28714 | 2520 | prog->sublen = i; |
d6a28714 JH |
2521 | } |
2522 | else { | |
2523 | prog->subbeg = strbeg; | |
2524 | prog->sublen = PL_regeol - strbeg; /* strend may have been modified */ | |
2525 | } | |
2526 | } | |
9041c2e3 | 2527 | |
d6a28714 JH |
2528 | return 1; |
2529 | ||
2530 | phooey: | |
a3621e74 | 2531 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 2532 | PL_colors[4], PL_colors[5])); |
d6a28714 | 2533 | if (PL_reg_eval_set) |
4f639d21 | 2534 | restore_pos(aTHX_ prog); |
e9105d30 | 2535 | if (swap) { |
c74340f9 | 2536 | /* we failed :-( roll it back */ |
e9105d30 GG |
2537 | Safefree(prog->offs); |
2538 | prog->offs = swap; | |
2539 | } | |
2540 | ||
d6a28714 JH |
2541 | return 0; |
2542 | } | |
2543 | ||
6bda09f9 | 2544 | |
d6a28714 JH |
2545 | /* |
2546 | - regtry - try match at specific point | |
2547 | */ | |
2548 | STATIC I32 /* 0 failure, 1 success */ | |
24b23f37 | 2549 | S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) |
d6a28714 | 2550 | { |
97aff369 | 2551 | dVAR; |
d6a28714 | 2552 | CHECKPOINT lastcp; |
288b8c02 NC |
2553 | REGEXP *const rx = reginfo->prog; |
2554 | regexp *const prog = (struct regexp *)SvANY(rx); | |
f8fc2ecf | 2555 | RXi_GET_DECL(prog,progi); |
a3621e74 | 2556 | GET_RE_DEBUG_FLAGS_DECL; |
7918f24d NC |
2557 | |
2558 | PERL_ARGS_ASSERT_REGTRY; | |
2559 | ||
24b23f37 | 2560 | reginfo->cutpoint=NULL; |
d6a28714 | 2561 | |
bbe252da | 2562 | if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) { |
d6a28714 JH |
2563 | MAGIC *mg; |
2564 | ||
2565 | PL_reg_eval_set = RS_init; | |
a3621e74 | 2566 | DEBUG_EXECUTE_r(DEBUG_s( |
b900a521 JH |
2567 | PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n", |
2568 | (IV)(PL_stack_sp - PL_stack_base)); | |
d6a28714 | 2569 | )); |
ea8d6ae1 | 2570 | SAVESTACK_CXPOS(); |
d6a28714 JH |
2571 | cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base; |
2572 | /* Otherwise OP_NEXTSTATE will free whatever on stack now. */ | |
2573 | SAVETMPS; | |
2574 | /* Apparently this is not needed, judging by wantarray. */ | |
e8347627 | 2575 | /* SAVEI8(cxstack[cxstack_ix].blk_gimme); |
d6a28714 JH |
2576 | cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ |
2577 | ||
3b0527fe | 2578 | if (reginfo->sv) { |
d6a28714 | 2579 | /* Make $_ available to executed code. */ |
3b0527fe | 2580 | if (reginfo->sv != DEFSV) { |
59f00321 | 2581 | SAVE_DEFSV; |
414bf5ae | 2582 | DEFSV_set(reginfo->sv); |
b8c5462f | 2583 | } |
d6a28714 | 2584 | |
3b0527fe DM |
2585 | if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) |
2586 | && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { | |
d6a28714 | 2587 | /* prepare for quick setting of pos */ |
d300d9fa | 2588 | #ifdef PERL_OLD_COPY_ON_WRITE |
51a9ea20 NC |
2589 | if (SvIsCOW(reginfo->sv)) |
2590 | sv_force_normal_flags(reginfo->sv, 0); | |
d300d9fa | 2591 | #endif |
3dab1dad | 2592 | mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, |
d300d9fa | 2593 | &PL_vtbl_mglob, NULL, 0); |
d6a28714 | 2594 | mg->mg_len = -1; |
b8c5462f | 2595 | } |
d6a28714 JH |
2596 | PL_reg_magic = mg; |
2597 | PL_reg_oldpos = mg->mg_len; | |
4f639d21 | 2598 | SAVEDESTRUCTOR_X(restore_pos, prog); |
d6a28714 | 2599 | } |
09687e5a | 2600 | if (!PL_reg_curpm) { |
a02a5408 | 2601 | Newxz(PL_reg_curpm, 1, PMOP); |
09687e5a AB |
2602 | #ifdef USE_ITHREADS |
2603 | { | |
14a49a24 | 2604 | SV* const repointer = &PL_sv_undef; |
92313705 NC |
2605 | /* this regexp is also owned by the new PL_reg_curpm, which |
2606 | will try to free it. */ | |
d2ece331 | 2607 | av_push(PL_regex_padav, repointer); |
09687e5a AB |
2608 | PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); |
2609 | PL_regex_pad = AvARRAY(PL_regex_padav); | |
2610 | } | |
2611 | #endif | |
2612 | } | |
86c29d75 NC |
2613 | #ifdef USE_ITHREADS |
2614 | /* It seems that non-ithreads works both with and without this code. | |
2615 | So for efficiency reasons it seems best not to have the code | |
2616 | compiled when it is not needed. */ | |
92313705 NC |
2617 | /* This is safe against NULLs: */ |
2618 | ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); | |
2619 | /* PM_reg_curpm owns a reference to this regexp. */ | |
cd733337 | 2620 | (void)ReREFCNT_inc(rx); |
86c29d75 | 2621 | #endif |
288b8c02 | 2622 | PM_SETRE(PL_reg_curpm, rx); |
d6a28714 JH |
2623 | PL_reg_oldcurpm = PL_curpm; |
2624 | PL_curpm = PL_reg_curpm; | |
07bc277f | 2625 | if (RXp_MATCH_COPIED(prog)) { |
d6a28714 JH |
2626 | /* Here is a serious problem: we cannot rewrite subbeg, |
2627 | since it may be needed if this match fails. Thus | |
2628 | $` inside (?{}) could fail... */ | |
2629 | PL_reg_oldsaved = prog->subbeg; | |
2630 | PL_reg_oldsavedlen = prog->sublen; | |
f8c7b90f | 2631 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2632 | PL_nrs = prog->saved_copy; |
2633 | #endif | |
07bc277f | 2634 | RXp_MATCH_COPIED_off(prog); |
d6a28714 JH |
2635 | } |
2636 | else | |
bd61b366 | 2637 | PL_reg_oldsaved = NULL; |
d6a28714 JH |
2638 | prog->subbeg = PL_bostr; |
2639 | prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ | |
2640 | } | |
24b23f37 | 2641 | DEBUG_EXECUTE_r(PL_reg_starttry = *startpos); |
f0ab9afb | 2642 | prog->offs[0].start = *startpos - PL_bostr; |
24b23f37 | 2643 | PL_reginput = *startpos; |
d6a28714 | 2644 | PL_reglastparen = &prog->lastparen; |
a01268b5 | 2645 | PL_reglastcloseparen = &prog->lastcloseparen; |
d6a28714 | 2646 | prog->lastparen = 0; |
03994de8 | 2647 | prog->lastcloseparen = 0; |
d6a28714 | 2648 | PL_regsize = 0; |
f0ab9afb | 2649 | PL_regoffs = prog->offs; |
d6a28714 JH |
2650 | if (PL_reg_start_tmpl <= prog->nparens) { |
2651 | PL_reg_start_tmpl = prog->nparens*3/2 + 3; | |
2652 | if(PL_reg_start_tmp) | |
2653 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
2654 | else | |
a02a5408 | 2655 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); |
d6a28714 JH |
2656 | } |
2657 | ||
2658 | /* XXXX What this code is doing here?!!! There should be no need | |
2659 | to do this again and again, PL_reglastparen should take care of | |
3dd2943c | 2660 | this! --ilya*/ |
dafc8851 JH |
2661 | |
2662 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2663 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
daf18116 | 2664 | * PL_reglastparen), is not needed at all by the test suite |
225593e1 DM |
2665 | * (op/regexp, op/pat, op/split), but that code is needed otherwise |
2666 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ | |
2667 | * Meanwhile, this code *is* needed for the | |
daf18116 JH |
2668 | * above-mentioned test suite tests to succeed. The common theme |
2669 | * on those tests seems to be returning null fields from matches. | |
225593e1 | 2670 | * --jhi updated by dapm */ |
dafc8851 | 2671 | #if 1 |
d6a28714 | 2672 | if (prog->nparens) { |
f0ab9afb | 2673 | regexp_paren_pair *pp = PL_regoffs; |
097eb12c | 2674 | register I32 i; |
eb160463 | 2675 | for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) { |
f0ab9afb NC |
2676 | ++pp; |
2677 | pp->start = -1; | |
2678 | pp->end = -1; | |
d6a28714 JH |
2679 | } |
2680 | } | |
dafc8851 | 2681 | #endif |
02db2b7b | 2682 | REGCP_SET(lastcp); |
f8fc2ecf | 2683 | if (regmatch(reginfo, progi->program + 1)) { |
f0ab9afb | 2684 | PL_regoffs[0].end = PL_reginput - PL_bostr; |
d6a28714 JH |
2685 | return 1; |
2686 | } | |
24b23f37 YO |
2687 | if (reginfo->cutpoint) |
2688 | *startpos= reginfo->cutpoint; | |
02db2b7b | 2689 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
2690 | return 0; |
2691 | } | |
2692 | ||
02db2b7b | 2693 | |
8ba1375e MJD |
2694 | #define sayYES goto yes |
2695 | #define sayNO goto no | |
262b90c4 | 2696 | #define sayNO_SILENT goto no_silent |
8ba1375e | 2697 | |
f9f4320a YO |
2698 | /* we dont use STMT_START/END here because it leads to |
2699 | "unreachable code" warnings, which are bogus, but distracting. */ | |
2700 | #define CACHEsayNO \ | |
c476f425 DM |
2701 | if (ST.cache_mask) \ |
2702 | PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \ | |
f9f4320a | 2703 | sayNO |
3298f257 | 2704 | |
a3621e74 | 2705 | /* this is used to determine how far from the left messages like |
265c4333 YO |
2706 | 'failed...' are printed. It should be set such that messages |
2707 | are inline with the regop output that created them. | |
a3621e74 | 2708 | */ |
265c4333 | 2709 | #define REPORT_CODE_OFF 32 |
a3621e74 YO |
2710 | |
2711 | ||
40a82448 DM |
2712 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
2713 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
9e137952 | 2714 | |
86545054 DM |
2715 | #define SLAB_FIRST(s) (&(s)->states[0]) |
2716 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
2717 | ||
5d9a96ca DM |
2718 | /* grab a new slab and return the first slot in it */ |
2719 | ||
2720 | STATIC regmatch_state * | |
2721 | S_push_slab(pTHX) | |
2722 | { | |
a35a87e7 | 2723 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2724 | dMY_CXT; |
2725 | #endif | |
5d9a96ca DM |
2726 | regmatch_slab *s = PL_regmatch_slab->next; |
2727 | if (!s) { | |
2728 | Newx(s, 1, regmatch_slab); | |
2729 | s->prev = PL_regmatch_slab; | |
2730 | s->next = NULL; | |
2731 | PL_regmatch_slab->next = s; | |
2732 | } | |
2733 | PL_regmatch_slab = s; | |
86545054 | 2734 | return SLAB_FIRST(s); |
5d9a96ca | 2735 | } |
5b47454d | 2736 | |
95b24440 | 2737 | |
40a82448 DM |
2738 | /* push a new state then goto it */ |
2739 | ||
2740 | #define PUSH_STATE_GOTO(state, node) \ | |
2741 | scan = node; \ | |
2742 | st->resume_state = state; \ | |
2743 | goto push_state; | |
2744 | ||
2745 | /* push a new state with success backtracking, then goto it */ | |
2746 | ||
2747 | #define PUSH_YES_STATE_GOTO(state, node) \ | |
2748 | scan = node; \ | |
2749 | st->resume_state = state; \ | |
2750 | goto push_yes_state; | |
2751 | ||
aa283a38 | 2752 | |
aa283a38 | 2753 | |
d6a28714 | 2754 | /* |
95b24440 | 2755 | |
bf1f174e DM |
2756 | regmatch() - main matching routine |
2757 | ||
2758 | This is basically one big switch statement in a loop. We execute an op, | |
2759 | set 'next' to point the next op, and continue. If we come to a point which | |
2760 | we may need to backtrack to on failure such as (A|B|C), we push a | |
2761 | backtrack state onto the backtrack stack. On failure, we pop the top | |
2762 | state, and re-enter the loop at the state indicated. If there are no more | |
2763 | states to pop, we return failure. | |
2764 | ||
2765 | Sometimes we also need to backtrack on success; for example /A+/, where | |
2766 | after successfully matching one A, we need to go back and try to | |
2767 | match another one; similarly for lookahead assertions: if the assertion | |
2768 | completes successfully, we backtrack to the state just before the assertion | |
2769 | and then carry on. In these cases, the pushed state is marked as | |
2770 | 'backtrack on success too'. This marking is in fact done by a chain of | |
2771 | pointers, each pointing to the previous 'yes' state. On success, we pop to | |
2772 | the nearest yes state, discarding any intermediate failure-only states. | |
2773 | Sometimes a yes state is pushed just to force some cleanup code to be | |
2774 | called at the end of a successful match or submatch; e.g. (??{$re}) uses | |
2775 | it to free the inner regex. | |
2776 | ||
2777 | Note that failure backtracking rewinds the cursor position, while | |
2778 | success backtracking leaves it alone. | |
2779 | ||
2780 | A pattern is complete when the END op is executed, while a subpattern | |
2781 | such as (?=foo) is complete when the SUCCESS op is executed. Both of these | |
2782 | ops trigger the "pop to last yes state if any, otherwise return true" | |
2783 | behaviour. | |
2784 | ||
2785 | A common convention in this function is to use A and B to refer to the two | |
2786 | subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is | |
2787 | the subpattern to be matched possibly multiple times, while B is the entire | |
2788 | rest of the pattern. Variable and state names reflect this convention. | |
2789 | ||
2790 | The states in the main switch are the union of ops and failure/success of | |
2791 | substates associated with with that op. For example, IFMATCH is the op | |
2792 | that does lookahead assertions /(?=A)B/ and so the IFMATCH state means | |
2793 | 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just | |
2794 | successfully matched A and IFMATCH_A_fail is a state saying that we have | |
2795 | just failed to match A. Resume states always come in pairs. The backtrack | |
2796 | state we push is marked as 'IFMATCH_A', but when that is popped, we resume | |
2797 | at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking | |
2798 | on success or failure. | |
2799 | ||
2800 | The struct that holds a backtracking state is actually a big union, with | |
2801 | one variant for each major type of op. The variable st points to the | |
2802 | top-most backtrack struct. To make the code clearer, within each | |
2803 | block of code we #define ST to alias the relevant union. | |
2804 | ||
2805 | Here's a concrete example of a (vastly oversimplified) IFMATCH | |
2806 | implementation: | |
2807 | ||
2808 | switch (state) { | |
2809 | .... | |
2810 | ||
2811 | #define ST st->u.ifmatch | |
2812 | ||
2813 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B | |
2814 | ST.foo = ...; // some state we wish to save | |
95b24440 | 2815 | ... |
bf1f174e DM |
2816 | // push a yes backtrack state with a resume value of |
2817 | // IFMATCH_A/IFMATCH_A_fail, then continue execution at the | |
2818 | // first node of A: | |
2819 | PUSH_YES_STATE_GOTO(IFMATCH_A, A); | |
2820 | // NOTREACHED | |
2821 | ||
2822 | case IFMATCH_A: // we have successfully executed A; now continue with B | |
2823 | next = B; | |
2824 | bar = ST.foo; // do something with the preserved value | |
2825 | break; | |
2826 | ||
2827 | case IFMATCH_A_fail: // A failed, so the assertion failed | |
2828 | ...; // do some housekeeping, then ... | |
2829 | sayNO; // propagate the failure | |
2830 | ||
2831 | #undef ST | |
95b24440 | 2832 | |
bf1f174e DM |
2833 | ... |
2834 | } | |
95b24440 | 2835 | |
bf1f174e DM |
2836 | For any old-timers reading this who are familiar with the old recursive |
2837 | approach, the code above is equivalent to: | |
95b24440 | 2838 | |
bf1f174e DM |
2839 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B |
2840 | { | |
2841 | int foo = ... | |
95b24440 | 2842 | ... |
bf1f174e DM |
2843 | if (regmatch(A)) { |
2844 | next = B; | |
2845 | bar = foo; | |
2846 | break; | |
95b24440 | 2847 | } |
bf1f174e DM |
2848 | ...; // do some housekeeping, then ... |
2849 | sayNO; // propagate the failure | |
95b24440 | 2850 | } |
bf1f174e DM |
2851 | |
2852 | The topmost backtrack state, pointed to by st, is usually free. If you | |
2853 | want to claim it, populate any ST.foo fields in it with values you wish to | |
2854 | save, then do one of | |
2855 | ||
2856 | PUSH_STATE_GOTO(resume_state, node); | |
2857 | PUSH_YES_STATE_GOTO(resume_state, node); | |
2858 | ||
2859 | which sets that backtrack state's resume value to 'resume_state', pushes a | |
2860 | new free entry to the top of the backtrack stack, then goes to 'node'. | |
2861 | On backtracking, the free slot is popped, and the saved state becomes the | |
2862 | new free state. An ST.foo field in this new top state can be temporarily | |
2863 | accessed to retrieve values, but once the main loop is re-entered, it | |
2864 | becomes available for reuse. | |
2865 | ||
2866 | Note that the depth of the backtrack stack constantly increases during the | |
2867 | left-to-right execution of the pattern, rather than going up and down with | |
2868 | the pattern nesting. For example the stack is at its maximum at Z at the | |
2869 | end of the pattern, rather than at X in the following: | |
2870 | ||
2871 | /(((X)+)+)+....(Y)+....Z/ | |
2872 | ||
2873 | The only exceptions to this are lookahead/behind assertions and the cut, | |
2874 | (?>A), which pop all the backtrack states associated with A before | |
2875 | continuing. | |
2876 | ||
486ec47a | 2877 | Backtrack state structs are allocated in slabs of about 4K in size. |
bf1f174e DM |
2878 | PL_regmatch_state and st always point to the currently active state, |
2879 | and PL_regmatch_slab points to the slab currently containing | |
2880 | PL_regmatch_state. The first time regmatch() is called, the first slab is | |
2881 | allocated, and is never freed until interpreter destruction. When the slab | |
2882 | is full, a new one is allocated and chained to the end. At exit from | |
2883 | regmatch(), slabs allocated since entry are freed. | |
2884 | ||
2885 | */ | |
95b24440 | 2886 | |
40a82448 | 2887 | |
5bc10b2c | 2888 | #define DEBUG_STATE_pp(pp) \ |
265c4333 | 2889 | DEBUG_STATE_r({ \ |
f2ed9b32 | 2890 | DUMP_EXEC_POS(locinput, scan, utf8_target); \ |
5bc10b2c | 2891 | PerlIO_printf(Perl_debug_log, \ |
5d458dd8 | 2892 | " %*s"pp" %s%s%s%s%s\n", \ |
5bc10b2c | 2893 | depth*2, "", \ |
13d6edb4 | 2894 | PL_reg_name[st->resume_state], \ |
5d458dd8 YO |
2895 | ((st==yes_state||st==mark_state) ? "[" : ""), \ |
2896 | ((st==yes_state) ? "Y" : ""), \ | |
2897 | ((st==mark_state) ? "M" : ""), \ | |
2898 | ((st==yes_state||st==mark_state) ? "]" : "") \ | |
2899 | ); \ | |
265c4333 | 2900 | }); |
5bc10b2c | 2901 | |
40a82448 | 2902 | |
3dab1dad | 2903 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 2904 | |
3df15adc | 2905 | #ifdef DEBUGGING |
5bc10b2c | 2906 | |
ab3bbdeb | 2907 | STATIC void |
f2ed9b32 | 2908 | S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target, |
ab3bbdeb YO |
2909 | const char *start, const char *end, const char *blurb) |
2910 | { | |
efd26800 | 2911 | const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; |
7918f24d NC |
2912 | |
2913 | PERL_ARGS_ASSERT_DEBUG_START_MATCH; | |
2914 | ||
ab3bbdeb YO |
2915 | if (!PL_colorset) |
2916 | reginitcolors(); | |
2917 | { | |
2918 | RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), | |
d2c6dc5e | 2919 | RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); |
ab3bbdeb | 2920 | |
f2ed9b32 | 2921 | RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb YO |
2922 | start, end - start, 60); |
2923 | ||
2924 | PerlIO_printf(Perl_debug_log, | |
2925 | "%s%s REx%s %s against %s\n", | |
2926 | PL_colors[4], blurb, PL_colors[5], s0, s1); | |
2927 | ||
f2ed9b32 | 2928 | if (utf8_target||utf8_pat) |
1de06328 YO |
2929 | PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", |
2930 | utf8_pat ? "pattern" : "", | |
f2ed9b32 KW |
2931 | utf8_pat && utf8_target ? " and " : "", |
2932 | utf8_target ? "string" : "" | |
ab3bbdeb YO |
2933 | ); |
2934 | } | |
2935 | } | |
3df15adc YO |
2936 | |
2937 | STATIC void | |
786e8c11 YO |
2938 | S_dump_exec_pos(pTHX_ const char *locinput, |
2939 | const regnode *scan, | |
2940 | const char *loc_regeol, | |
2941 | const char *loc_bostr, | |
2942 | const char *loc_reg_starttry, | |
f2ed9b32 | 2943 | const bool utf8_target) |
07be1b83 | 2944 | { |
786e8c11 | 2945 | const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; |
07be1b83 | 2946 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ |
786e8c11 | 2947 | int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); |
07be1b83 YO |
2948 | /* The part of the string before starttry has one color |
2949 | (pref0_len chars), between starttry and current | |
2950 | position another one (pref_len - pref0_len chars), | |
2951 | after the current position the third one. | |
2952 | We assume that pref0_len <= pref_len, otherwise we | |
2953 | decrease pref0_len. */ | |
786e8c11 YO |
2954 | int pref_len = (locinput - loc_bostr) > (5 + taill) - l |
2955 | ? (5 + taill) - l : locinput - loc_bostr; | |
07be1b83 YO |
2956 | int pref0_len; |
2957 | ||
7918f24d NC |
2958 | PERL_ARGS_ASSERT_DUMP_EXEC_POS; |
2959 | ||
f2ed9b32 | 2960 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) |
07be1b83 | 2961 | pref_len++; |
786e8c11 YO |
2962 | pref0_len = pref_len - (locinput - loc_reg_starttry); |
2963 | if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) | |
2964 | l = ( loc_regeol - locinput > (5 + taill) - pref_len | |
2965 | ? (5 + taill) - pref_len : loc_regeol - locinput); | |
f2ed9b32 | 2966 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) |
07be1b83 YO |
2967 | l--; |
2968 | if (pref0_len < 0) | |
2969 | pref0_len = 0; | |
2970 | if (pref0_len > pref_len) | |
2971 | pref0_len = pref_len; | |
2972 | { | |
f2ed9b32 | 2973 | const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0; |
0df25f3d | 2974 | |
ab3bbdeb | 2975 | RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), |
1de06328 | 2976 | (locinput - pref_len),pref0_len, 60, 4, 5); |
0df25f3d | 2977 | |
ab3bbdeb | 2978 | RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), |
3df15adc | 2979 | (locinput - pref_len + pref0_len), |
1de06328 | 2980 | pref_len - pref0_len, 60, 2, 3); |
0df25f3d | 2981 | |
ab3bbdeb | 2982 | RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), |
1de06328 | 2983 | locinput, loc_regeol - locinput, 10, 0, 1); |
0df25f3d | 2984 | |
1de06328 | 2985 | const STRLEN tlen=len0+len1+len2; |
3df15adc | 2986 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb | 2987 | "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", |
786e8c11 | 2988 | (IV)(locinput - loc_bostr), |
07be1b83 | 2989 | len0, s0, |
07be1b83 | 2990 | len1, s1, |
07be1b83 | 2991 | (docolor ? "" : "> <"), |
07be1b83 | 2992 | len2, s2, |
f9f4320a | 2993 | (int)(tlen > 19 ? 0 : 19 - tlen), |
07be1b83 YO |
2994 | ""); |
2995 | } | |
2996 | } | |
3df15adc | 2997 | |
07be1b83 YO |
2998 | #endif |
2999 | ||
0a4db386 YO |
3000 | /* reg_check_named_buff_matched() |
3001 | * Checks to see if a named buffer has matched. The data array of | |
3002 | * buffer numbers corresponding to the buffer is expected to reside | |
3003 | * in the regexp->data->data array in the slot stored in the ARG() of | |
3004 | * node involved. Note that this routine doesn't actually care about the | |
3005 | * name, that information is not preserved from compilation to execution. | |
3006 | * Returns the index of the leftmost defined buffer with the given name | |
3007 | * or 0 if non of the buffers matched. | |
3008 | */ | |
3009 | STATIC I32 | |
7918f24d NC |
3010 | S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) |
3011 | { | |
0a4db386 | 3012 | I32 n; |
f8fc2ecf | 3013 | RXi_GET_DECL(rex,rexi); |
ad64d0ec | 3014 | SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
0a4db386 | 3015 | I32 *nums=(I32*)SvPVX(sv_dat); |
7918f24d NC |
3016 | |
3017 | PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; | |
3018 | ||
0a4db386 YO |
3019 | for ( n=0; n<SvIVX(sv_dat); n++ ) { |
3020 | if ((I32)*PL_reglastparen >= nums[n] && | |
f0ab9afb | 3021 | PL_regoffs[nums[n]].end != -1) |
0a4db386 YO |
3022 | { |
3023 | return nums[n]; | |
3024 | } | |
3025 | } | |
3026 | return 0; | |
3027 | } | |
3028 | ||
2f554ef7 DM |
3029 | |
3030 | /* free all slabs above current one - called during LEAVE_SCOPE */ | |
3031 | ||
3032 | STATIC void | |
3033 | S_clear_backtrack_stack(pTHX_ void *p) | |
3034 | { | |
3035 | regmatch_slab *s = PL_regmatch_slab->next; | |
3036 | PERL_UNUSED_ARG(p); | |
3037 | ||
3038 | if (!s) | |
3039 | return; | |
3040 | PL_regmatch_slab->next = NULL; | |
3041 | while (s) { | |
3042 | regmatch_slab * const osl = s; | |
3043 | s = s->next; | |
3044 | Safefree(osl); | |
3045 | } | |
3046 | } | |
3047 | ||
3048 | ||
28d8d7f4 YO |
3049 | #define SETREX(Re1,Re2) \ |
3050 | if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \ | |
3051 | Re1 = (Re2) | |
3052 | ||
d6a28714 | 3053 | STATIC I32 /* 0 failure, 1 success */ |
24b23f37 | 3054 | S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) |
d6a28714 | 3055 | { |
a35a87e7 | 3056 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
3057 | dMY_CXT; |
3058 | #endif | |
27da23d5 | 3059 | dVAR; |
f2ed9b32 | 3060 | register const bool utf8_target = PL_reg_match_utf8; |
4ad0818d | 3061 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
288b8c02 NC |
3062 | REGEXP *rex_sv = reginfo->prog; |
3063 | regexp *rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 3064 | RXi_GET_DECL(rex,rexi); |
2f554ef7 | 3065 | I32 oldsave; |
5d9a96ca DM |
3066 | /* the current state. This is a cached copy of PL_regmatch_state */ |
3067 | register regmatch_state *st; | |
5d9a96ca DM |
3068 | /* cache heavy used fields of st in registers */ |
3069 | register regnode *scan; | |
3070 | register regnode *next; | |
438e9bae | 3071 | register U32 n = 0; /* general value; init to avoid compiler warning */ |
24d3c4a9 | 3072 | register I32 ln = 0; /* len or last; init to avoid compiler warning */ |
5d9a96ca | 3073 | register char *locinput = PL_reginput; |
5d9a96ca | 3074 | register I32 nextchr; /* is always set to UCHARAT(locinput) */ |
24d3c4a9 | 3075 | |
b69b0499 | 3076 | bool result = 0; /* return value of S_regmatch */ |
24d3c4a9 | 3077 | int depth = 0; /* depth of backtrack stack */ |
4b196cd4 YO |
3078 | U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */ |
3079 | const U32 max_nochange_depth = | |
3080 | (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? | |
3081 | 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; | |
77cb431f DM |
3082 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
3083 | subpattern */ | |
e2e6a0f1 YO |
3084 | /* mark_state piggy backs on the yes_state logic so that when we unwind |
3085 | the stack on success we can update the mark_state as we go */ | |
3086 | regmatch_state *mark_state = NULL; /* last mark state we have seen */ | |
faec1544 | 3087 | regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ |
b8591aee | 3088 | struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ |
40a82448 | 3089 | U32 state_num; |
5d458dd8 YO |
3090 | bool no_final = 0; /* prevent failure from backtracking? */ |
3091 | bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ | |
e2e6a0f1 | 3092 | char *startpoint = PL_reginput; |
5d458dd8 YO |
3093 | SV *popmark = NULL; /* are we looking for a mark? */ |
3094 | SV *sv_commit = NULL; /* last mark name seen in failure */ | |
3095 | SV *sv_yes_mark = NULL; /* last mark name we have seen | |
486ec47a | 3096 | during a successful match */ |
5d458dd8 YO |
3097 | U32 lastopen = 0; /* last open we saw */ |
3098 | bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; | |
19b95bf0 | 3099 | SV* const oreplsv = GvSV(PL_replgv); |
24d3c4a9 DM |
3100 | /* these three flags are set by various ops to signal information to |
3101 | * the very next op. They have a useful lifetime of exactly one loop | |
3102 | * iteration, and are not preserved or restored by state pushes/pops | |
3103 | */ | |
3104 | bool sw = 0; /* the condition value in (?(cond)a|b) */ | |
3105 | bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ | |
3106 | int logical = 0; /* the following EVAL is: | |
3107 | 0: (?{...}) | |
3108 | 1: (?(?{...})X|Y) | |
3109 | 2: (??{...}) | |
3110 | or the following IFMATCH/UNLESSM is: | |
3111 | false: plain (?=foo) | |
3112 | true: used as a condition: (?(?=foo)) | |
3113 | */ | |
95b24440 | 3114 | #ifdef DEBUGGING |
e68ec53f | 3115 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
3116 | #endif |
3117 | ||
7918f24d NC |
3118 | PERL_ARGS_ASSERT_REGMATCH; |
3119 | ||
3b57cd43 | 3120 | DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ |
24b23f37 | 3121 | PerlIO_printf(Perl_debug_log,"regmatch start\n"); |
3b57cd43 | 3122 | })); |
5d9a96ca DM |
3123 | /* on first ever call to regmatch, allocate first slab */ |
3124 | if (!PL_regmatch_slab) { | |
3125 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
3126 | PL_regmatch_slab->prev = NULL; | |
3127 | PL_regmatch_slab->next = NULL; | |
86545054 | 3128 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); |
5d9a96ca DM |
3129 | } |
3130 | ||
2f554ef7 DM |
3131 | oldsave = PL_savestack_ix; |
3132 | SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL); | |
3133 | SAVEVPTR(PL_regmatch_slab); | |
3134 | SAVEVPTR(PL_regmatch_state); | |
5d9a96ca DM |
3135 | |
3136 | /* grab next free state slot */ | |
3137 | st = ++PL_regmatch_state; | |
86545054 | 3138 | if (st > SLAB_LAST(PL_regmatch_slab)) |
5d9a96ca DM |
3139 | st = PL_regmatch_state = S_push_slab(aTHX); |
3140 | ||
d6a28714 JH |
3141 | /* Note that nextchr is a byte even in UTF */ |
3142 | nextchr = UCHARAT(locinput); | |
3143 | scan = prog; | |
3144 | while (scan != NULL) { | |
8ba1375e | 3145 | |
a3621e74 | 3146 | DEBUG_EXECUTE_r( { |
6136c704 | 3147 | SV * const prop = sv_newmortal(); |
1de06328 | 3148 | regnode *rnext=regnext(scan); |
f2ed9b32 | 3149 | DUMP_EXEC_POS( locinput, scan, utf8_target ); |
32fc9b6a | 3150 | regprop(rex, prop, scan); |
07be1b83 YO |
3151 | |
3152 | PerlIO_printf(Perl_debug_log, | |
3153 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
f8fc2ecf | 3154 | (IV)(scan - rexi->program), depth*2, "", |
07be1b83 | 3155 | SvPVX_const(prop), |
1de06328 | 3156 | (PL_regkind[OP(scan)] == END || !rnext) ? |
f8fc2ecf | 3157 | 0 : (IV)(rnext - rexi->program)); |
2a782b5b | 3158 | }); |
d6a28714 JH |
3159 | |
3160 | next = scan + NEXT_OFF(scan); | |
3161 | if (next == scan) | |
3162 | next = NULL; | |
40a82448 | 3163 | state_num = OP(scan); |
d6a28714 | 3164 | |
40a82448 | 3165 | reenter_switch: |
34a81e2b B |
3166 | |
3167 | assert(PL_reglastparen == &rex->lastparen); | |
3168 | assert(PL_reglastcloseparen == &rex->lastcloseparen); | |
3169 | assert(PL_regoffs == rex->offs); | |
3170 | ||
40a82448 | 3171 | switch (state_num) { |
d6a28714 | 3172 | case BOL: |
7fba1cd6 | 3173 | if (locinput == PL_bostr) |
d6a28714 | 3174 | { |
3b0527fe | 3175 | /* reginfo->till = reginfo->bol; */ |
b8c5462f JH |
3176 | break; |
3177 | } | |
d6a28714 JH |
3178 | sayNO; |
3179 | case MBOL: | |
12d33761 HS |
3180 | if (locinput == PL_bostr || |
3181 | ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n')) | |
d6a28714 | 3182 | { |
b8c5462f JH |
3183 | break; |
3184 | } | |
d6a28714 JH |
3185 | sayNO; |
3186 | case SBOL: | |
c2a73568 | 3187 | if (locinput == PL_bostr) |
b8c5462f | 3188 | break; |
d6a28714 JH |
3189 | sayNO; |
3190 | case GPOS: | |
3b0527fe | 3191 | if (locinput == reginfo->ganch) |
d6a28714 JH |
3192 | break; |
3193 | sayNO; | |
ee9b8eae YO |
3194 | |
3195 | case KEEPS: | |
3196 | /* update the startpoint */ | |
f0ab9afb | 3197 | st->u.keeper.val = PL_regoffs[0].start; |
ee9b8eae | 3198 | PL_reginput = locinput; |
f0ab9afb | 3199 | PL_regoffs[0].start = locinput - PL_bostr; |
ee9b8eae YO |
3200 | PUSH_STATE_GOTO(KEEPS_next, next); |
3201 | /*NOT-REACHED*/ | |
3202 | case KEEPS_next_fail: | |
3203 | /* rollback the start point change */ | |
f0ab9afb | 3204 | PL_regoffs[0].start = st->u.keeper.val; |
ee9b8eae YO |
3205 | sayNO_SILENT; |
3206 | /*NOT-REACHED*/ | |
d6a28714 | 3207 | case EOL: |
d6a28714 JH |
3208 | goto seol; |
3209 | case MEOL: | |
d6a28714 | 3210 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') |
b8c5462f | 3211 | sayNO; |
b8c5462f | 3212 | break; |
d6a28714 JH |
3213 | case SEOL: |
3214 | seol: | |
3215 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') | |
b8c5462f | 3216 | sayNO; |
d6a28714 | 3217 | if (PL_regeol - locinput > 1) |
b8c5462f | 3218 | sayNO; |
b8c5462f | 3219 | break; |
d6a28714 JH |
3220 | case EOS: |
3221 | if (PL_regeol != locinput) | |
b8c5462f | 3222 | sayNO; |
d6a28714 | 3223 | break; |
ffc61ed2 | 3224 | case SANY: |
d6a28714 | 3225 | if (!nextchr && locinput >= PL_regeol) |
4633a7c4 | 3226 | sayNO; |
f2ed9b32 | 3227 | if (utf8_target) { |
f33976b4 DB |
3228 | locinput += PL_utf8skip[nextchr]; |
3229 | if (locinput > PL_regeol) | |
3230 | sayNO; | |
3231 | nextchr = UCHARAT(locinput); | |
3232 | } | |
3233 | else | |
3234 | nextchr = UCHARAT(++locinput); | |
3235 | break; | |
3236 | case CANY: | |
3237 | if (!nextchr && locinput >= PL_regeol) | |
3238 | sayNO; | |
b8c5462f | 3239 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 3240 | break; |
ffc61ed2 | 3241 | case REG_ANY: |
1aa99e6b IH |
3242 | if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n') |
3243 | sayNO; | |
f2ed9b32 | 3244 | if (utf8_target) { |
b8c5462f | 3245 | locinput += PL_utf8skip[nextchr]; |
d6a28714 JH |
3246 | if (locinput > PL_regeol) |
3247 | sayNO; | |
a0ed51b3 | 3248 | nextchr = UCHARAT(locinput); |
a0ed51b3 | 3249 | } |
1aa99e6b IH |
3250 | else |
3251 | nextchr = UCHARAT(++locinput); | |
a0ed51b3 | 3252 | break; |
166ba7cd DM |
3253 | |
3254 | #undef ST | |
3255 | #define ST st->u.trie | |
786e8c11 YO |
3256 | case TRIEC: |
3257 | /* In this case the charclass data is available inline so | |
3258 | we can fail fast without a lot of extra overhead. | |
3259 | */ | |
f2ed9b32 | 3260 | if (scan->flags == EXACT || !utf8_target) { |
786e8c11 YO |
3261 | if(!ANYOF_BITMAP_TEST(scan, *locinput)) { |
3262 | DEBUG_EXECUTE_r( | |
3263 | PerlIO_printf(Perl_debug_log, | |
3264 | "%*s %sfailed to match trie start class...%s\n", | |
5bc10b2c | 3265 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
786e8c11 YO |
3266 | ); |
3267 | sayNO_SILENT; | |
3268 | /* NOTREACHED */ | |
3269 | } | |
3270 | } | |
3271 | /* FALL THROUGH */ | |
5b47454d | 3272 | case TRIE: |
2e64971a DM |
3273 | /* the basic plan of execution of the trie is: |
3274 | * At the beginning, run though all the states, and | |
3275 | * find the longest-matching word. Also remember the position | |
3276 | * of the shortest matching word. For example, this pattern: | |
3277 | * 1 2 3 4 5 | |
3278 | * ab|a|x|abcd|abc | |
3279 | * when matched against the string "abcde", will generate | |
3280 | * accept states for all words except 3, with the longest | |
3281 | * matching word being 4, and the shortest being 1 (with | |
3282 | * the position being after char 1 of the string). | |
3283 | * | |
3284 | * Then for each matching word, in word order (i.e. 1,2,4,5), | |
3285 | * we run the remainder of the pattern; on each try setting | |
3286 | * the current position to the character following the word, | |
3287 | * returning to try the next word on failure. | |
3288 | * | |
3289 | * We avoid having to build a list of words at runtime by | |
3290 | * using a compile-time structure, wordinfo[].prev, which | |
3291 | * gives, for each word, the previous accepting word (if any). | |
3292 | * In the case above it would contain the mappings 1->2, 2->0, | |
3293 | * 3->0, 4->5, 5->1. We can use this table to generate, from | |
3294 | * the longest word (4 above), a list of all words, by | |
3295 | * following the list of prev pointers; this gives us the | |
3296 | * unordered list 4,5,1,2. Then given the current word we have | |
3297 | * just tried, we can go through the list and find the | |
3298 | * next-biggest word to try (so if we just failed on word 2, | |
3299 | * the next in the list is 4). | |
3300 | * | |
3301 | * Since at runtime we don't record the matching position in | |
3302 | * the string for each word, we have to work that out for | |
3303 | * each word we're about to process. The wordinfo table holds | |
3304 | * the character length of each word; given that we recorded | |
3305 | * at the start: the position of the shortest word and its | |
3306 | * length in chars, we just need to move the pointer the | |
3307 | * difference between the two char lengths. Depending on | |
3308 | * Unicode status and folding, that's cheap or expensive. | |
3309 | * | |
3310 | * This algorithm is optimised for the case where are only a | |
3311 | * small number of accept states, i.e. 0,1, or maybe 2. | |
3312 | * With lots of accepts states, and having to try all of them, | |
3313 | * it becomes quadratic on number of accept states to find all | |
3314 | * the next words. | |
3315 | */ | |
3316 | ||
3dab1dad | 3317 | { |
07be1b83 | 3318 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
a0a388a1 | 3319 | DECL_TRIE_TYPE(scan); |
3dab1dad YO |
3320 | |
3321 | /* what trie are we using right now */ | |
be8e71aa | 3322 | reg_trie_data * const trie |
f8fc2ecf | 3323 | = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; |
85fbaab2 | 3324 | HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); |
3dab1dad | 3325 | U32 state = trie->startstate; |
166ba7cd | 3326 | |
3dab1dad YO |
3327 | if (trie->bitmap && trie_type != trie_utf8_fold && |
3328 | !TRIE_BITMAP_TEST(trie,*locinput) | |
3329 | ) { | |
3330 | if (trie->states[ state ].wordnum) { | |
3331 | DEBUG_EXECUTE_r( | |
3332 | PerlIO_printf(Perl_debug_log, | |
3333 | "%*s %smatched empty string...%s\n", | |
5bc10b2c | 3334 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad | 3335 | ); |
20dbff7c YO |
3336 | if (!trie->jump) |
3337 | break; | |
3dab1dad YO |
3338 | } else { |
3339 | DEBUG_EXECUTE_r( | |
3340 | PerlIO_printf(Perl_debug_log, | |
786e8c11 | 3341 | "%*s %sfailed to match trie start class...%s\n", |
5bc10b2c | 3342 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad YO |
3343 | ); |
3344 | sayNO_SILENT; | |
3345 | } | |
3346 | } | |
166ba7cd | 3347 | |
786e8c11 YO |
3348 | { |
3349 | U8 *uc = ( U8* )locinput; | |
3350 | ||
3351 | STRLEN len = 0; | |
3352 | STRLEN foldlen = 0; | |
3353 | U8 *uscan = (U8*)NULL; | |
786e8c11 | 3354 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
2e64971a DM |
3355 | U32 charcount = 0; /* how many input chars we have matched */ |
3356 | U32 accepted = 0; /* have we seen any accepting states? */ | |
786e8c11 | 3357 | |
786e8c11 YO |
3358 | ST.B = next; |
3359 | ST.jump = trie->jump; | |
786e8c11 | 3360 | ST.me = scan; |
2e64971a DM |
3361 | ST.firstpos = NULL; |
3362 | ST.longfold = FALSE; /* char longer if folded => it's harder */ | |
3363 | ST.nextword = 0; | |
3364 | ||
3365 | /* fully traverse the TRIE; note the position of the | |
3366 | shortest accept state and the wordnum of the longest | |
3367 | accept state */ | |
07be1b83 | 3368 | |
a3621e74 | 3369 | while ( state && uc <= (U8*)PL_regeol ) { |
786e8c11 | 3370 | U32 base = trie->states[ state ].trans.base; |
f9f4320a | 3371 | UV uvc = 0; |
acb909b4 | 3372 | U16 charid = 0; |
2e64971a DM |
3373 | U16 wordnum; |
3374 | wordnum = trie->states[ state ].wordnum; | |
3375 | ||
3376 | if (wordnum) { /* it's an accept state */ | |
3377 | if (!accepted) { | |
3378 | accepted = 1; | |
3379 | /* record first match position */ | |
3380 | if (ST.longfold) { | |
3381 | ST.firstpos = (U8*)locinput; | |
3382 | ST.firstchars = 0; | |
5b47454d | 3383 | } |
2e64971a DM |
3384 | else { |
3385 | ST.firstpos = uc; | |
3386 | ST.firstchars = charcount; | |
3387 | } | |
3388 | } | |
3389 | if (!ST.nextword || wordnum < ST.nextword) | |
3390 | ST.nextword = wordnum; | |
3391 | ST.topword = wordnum; | |
786e8c11 | 3392 | } |
a3621e74 | 3393 | |
07be1b83 | 3394 | DEBUG_TRIE_EXECUTE_r({ |
f2ed9b32 | 3395 | DUMP_EXEC_POS( (char *)uc, scan, utf8_target ); |
a3621e74 | 3396 | PerlIO_printf( Perl_debug_log, |
2e64971a | 3397 | "%*s %sState: %4"UVxf" Accepted: %c ", |
5bc10b2c | 3398 | 2+depth * 2, "", PL_colors[4], |
2e64971a | 3399 | (UV)state, (accepted ? 'Y' : 'N')); |
07be1b83 | 3400 | }); |
a3621e74 | 3401 | |
2e64971a | 3402 | /* read a char and goto next state */ |
a3621e74 | 3403 | if ( base ) { |
6dd2be57 | 3404 | I32 offset; |
55eed653 NC |
3405 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
3406 | uscan, len, uvc, charid, foldlen, | |
3407 | foldbuf, uniflags); | |
2e64971a DM |
3408 | charcount++; |
3409 | if (foldlen>0) | |
3410 | ST.longfold = TRUE; | |
5b47454d | 3411 | if (charid && |
6dd2be57 DM |
3412 | ( ((offset = |
3413 | base + charid - 1 - trie->uniquecharcount)) >= 0) | |
3414 | ||
3415 | && ((U32)offset < trie->lasttrans) | |
3416 | && trie->trans[offset].check == state) | |
5b47454d | 3417 | { |
6dd2be57 | 3418 | state = trie->trans[offset].next; |
5b47454d DM |
3419 | } |
3420 | else { | |
3421 | state = 0; | |
3422 | } | |
3423 | uc += len; | |
3424 | ||
3425 | } | |
3426 | else { | |
a3621e74 YO |
3427 | state = 0; |
3428 | } | |
3429 | DEBUG_TRIE_EXECUTE_r( | |
e4584336 | 3430 | PerlIO_printf( Perl_debug_log, |
786e8c11 | 3431 | "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n", |
e4584336 | 3432 | charid, uvc, (UV)state, PL_colors[5] ); |
a3621e74 YO |
3433 | ); |
3434 | } | |
2e64971a | 3435 | if (!accepted) |
a3621e74 | 3436 | sayNO; |
a3621e74 | 3437 | |
2e64971a DM |
3438 | /* calculate total number of accept states */ |
3439 | { | |
3440 | U16 w = ST.topword; | |
3441 | accepted = 0; | |
3442 | while (w) { | |
3443 | w = trie->wordinfo[w].prev; | |
3444 | accepted++; | |
3445 | } | |
3446 | ST.accepted = accepted; | |
3447 | } | |
3448 | ||
166ba7cd DM |
3449 | DEBUG_EXECUTE_r( |
3450 | PerlIO_printf( Perl_debug_log, | |
3451 | "%*s %sgot %"IVdf" possible matches%s\n", | |
5bc10b2c | 3452 | REPORT_CODE_OFF + depth * 2, "", |
166ba7cd DM |
3453 | PL_colors[4], (IV)ST.accepted, PL_colors[5] ); |
3454 | ); | |
2e64971a | 3455 | goto trie_first_try; /* jump into the fail handler */ |
786e8c11 | 3456 | }} |
fae667d5 | 3457 | /* NOTREACHED */ |
2e64971a DM |
3458 | |
3459 | case TRIE_next_fail: /* we failed - try next alternative */ | |
fae667d5 YO |
3460 | if ( ST.jump) { |
3461 | REGCP_UNWIND(ST.cp); | |
3462 | for (n = *PL_reglastparen; n > ST.lastparen; n--) | |
f0ab9afb | 3463 | PL_regoffs[n].end = -1; |
fae667d5 YO |
3464 | *PL_reglastparen = n; |
3465 | } | |
2e64971a DM |
3466 | if (!--ST.accepted) { |
3467 | DEBUG_EXECUTE_r({ | |
3468 | PerlIO_printf( Perl_debug_log, | |
3469 | "%*s %sTRIE failed...%s\n", | |
3470 | REPORT_CODE_OFF+depth*2, "", | |
3471 | PL_colors[4], | |
3472 | PL_colors[5] ); | |
3473 | }); | |
3474 | sayNO_SILENT; | |
3475 | } | |
3476 | { | |
3477 | /* Find next-highest word to process. Note that this code | |
3478 | * is O(N^2) per trie run (O(N) per branch), so keep tight */ | |
d9a396a3 DM |
3479 | register U16 min = 0; |
3480 | register U16 word; | |
2e64971a DM |
3481 | register U16 const nextword = ST.nextword; |
3482 | register reg_trie_wordinfo * const wordinfo | |
3483 | = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo; | |
3484 | for (word=ST.topword; word; word=wordinfo[word].prev) { | |
3485 | if (word > nextword && (!min || word < min)) | |
3486 | min = word; | |
3487 | } | |
3488 | ST.nextword = min; | |
3489 | } | |
3490 | ||
fae667d5 | 3491 | trie_first_try: |
5d458dd8 YO |
3492 | if (do_cutgroup) { |
3493 | do_cutgroup = 0; | |
3494 | no_final = 0; | |
3495 | } | |
fae667d5 YO |
3496 | |
3497 | if ( ST.jump) { | |
3498 | ST.lastparen = *PL_reglastparen; | |
3499 | REGCP_SET(ST.cp); | |
2e64971a | 3500 | } |
a3621e74 | 3501 | |
2e64971a | 3502 | /* find start char of end of current word */ |
166ba7cd | 3503 | { |
2e64971a DM |
3504 | U32 chars; /* how many chars to skip */ |
3505 | U8 *uc = ST.firstpos; | |
3506 | reg_trie_data * const trie | |
3507 | = (reg_trie_data*)rexi->data->data[ARG(ST.me)]; | |
3508 | ||
3509 | assert((trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3510 | >= ST.firstchars); | |
3511 | chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3512 | - ST.firstchars; | |
3513 | ||
3514 | if (ST.longfold) { | |
3515 | /* the hard option - fold each char in turn and find | |
3516 | * its folded length (which may be different */ | |
3517 | U8 foldbuf[UTF8_MAXBYTES_CASE + 1]; | |
3518 | STRLEN foldlen; | |
3519 | STRLEN len; | |
d9a396a3 | 3520 | UV uvc; |
2e64971a DM |
3521 | U8 *uscan; |
3522 | ||
3523 | while (chars) { | |
f2ed9b32 | 3524 | if (utf8_target) { |
2e64971a DM |
3525 | uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len, |
3526 | uniflags); | |
3527 | uc += len; | |
3528 | } | |
3529 | else { | |
3530 | uvc = *uc; | |
3531 | uc++; | |
3532 | } | |
3533 | uvc = to_uni_fold(uvc, foldbuf, &foldlen); | |
3534 | uscan = foldbuf; | |
3535 | while (foldlen) { | |
3536 | if (!--chars) | |
3537 | break; | |
3538 | uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len, | |
3539 | uniflags); | |
3540 | uscan += len; | |
3541 | foldlen -= len; | |
3542 | } | |
3543 | } | |
a3621e74 | 3544 | } |
2e64971a | 3545 | else { |
f2ed9b32 | 3546 | if (utf8_target) |
2e64971a DM |
3547 | while (chars--) |
3548 | uc += UTF8SKIP(uc); | |
3549 | else | |
3550 | uc += chars; | |
3551 | } | |
3552 | PL_reginput = (char *)uc; | |
3553 | } | |
166ba7cd | 3554 | |
2e64971a DM |
3555 | scan = (ST.jump && ST.jump[ST.nextword]) |
3556 | ? ST.me + ST.jump[ST.nextword] | |
3557 | : ST.B; | |
166ba7cd | 3558 | |
2e64971a DM |
3559 | DEBUG_EXECUTE_r({ |
3560 | PerlIO_printf( Perl_debug_log, | |
3561 | "%*s %sTRIE matched word #%d, continuing%s\n", | |
3562 | REPORT_CODE_OFF+depth*2, "", | |
3563 | PL_colors[4], | |
3564 | ST.nextword, | |
3565 | PL_colors[5] | |
3566 | ); | |
3567 | }); | |
3568 | ||
3569 | if (ST.accepted > 1 || has_cutgroup) { | |
3570 | PUSH_STATE_GOTO(TRIE_next, scan); | |
3571 | /* NOTREACHED */ | |
166ba7cd | 3572 | } |
2e64971a DM |
3573 | /* only one choice left - just continue */ |
3574 | DEBUG_EXECUTE_r({ | |
3575 | AV *const trie_words | |
3576 | = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]); | |
3577 | SV ** const tmp = av_fetch( trie_words, | |
3578 | ST.nextword-1, 0 ); | |
3579 | SV *sv= tmp ? sv_newmortal() : NULL; | |
3580 | ||
3581 | PerlIO_printf( Perl_debug_log, | |
3582 | "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n", | |
3583 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], | |
3584 | ST.nextword, | |
3585 | tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0, | |
3586 | PL_colors[0], PL_colors[1], | |
c89df6cf | 3587 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII |
2e64971a DM |
3588 | ) |
3589 | : "not compiled under -Dr", | |
3590 | PL_colors[5] ); | |
3591 | }); | |
3592 | ||
3593 | locinput = PL_reginput; | |
3594 | nextchr = UCHARAT(locinput); | |
3595 | continue; /* execute rest of RE */ | |
166ba7cd | 3596 | /* NOTREACHED */ |
166ba7cd DM |
3597 | #undef ST |
3598 | ||
95b24440 DM |
3599 | case EXACT: { |
3600 | char *s = STRING(scan); | |
24d3c4a9 | 3601 | ln = STR_LEN(scan); |
f2ed9b32 | 3602 | if (utf8_target != UTF_PATTERN) { |
bc517b45 | 3603 | /* The target and the pattern have differing utf8ness. */ |
1aa99e6b | 3604 | char *l = locinput; |
24d3c4a9 | 3605 | const char * const e = s + ln; |
a72c7584 | 3606 | |
f2ed9b32 | 3607 | if (utf8_target) { |
5ff6fc6d | 3608 | /* The target is utf8, the pattern is not utf8. */ |
1aa99e6b | 3609 | while (s < e) { |
a3b680e6 | 3610 | STRLEN ulen; |
1aa99e6b | 3611 | if (l >= PL_regeol) |
5ff6fc6d JH |
3612 | sayNO; |
3613 | if (NATIVE_TO_UNI(*(U8*)s) != | |
89ebb4a3 | 3614 | utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen, |
041457d9 | 3615 | uniflags)) |
5ff6fc6d | 3616 | sayNO; |
bc517b45 | 3617 | l += ulen; |
5ff6fc6d | 3618 | s ++; |
1aa99e6b | 3619 | } |
5ff6fc6d JH |
3620 | } |
3621 | else { | |
3622 | /* The target is not utf8, the pattern is utf8. */ | |
1aa99e6b | 3623 | while (s < e) { |
a3b680e6 | 3624 | STRLEN ulen; |
1aa99e6b IH |
3625 | if (l >= PL_regeol) |
3626 | sayNO; | |
5ff6fc6d | 3627 | if (NATIVE_TO_UNI(*((U8*)l)) != |
89ebb4a3 | 3628 | utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen, |
041457d9 | 3629 | uniflags)) |
1aa99e6b | 3630 | sayNO; |
bc517b45 | 3631 | s += ulen; |
a72c7584 | 3632 | l ++; |
1aa99e6b | 3633 | } |
5ff6fc6d | 3634 | } |
1aa99e6b IH |
3635 | locinput = l; |
3636 | nextchr = UCHARAT(locinput); | |
3637 | break; | |
3638 | } | |
bc517b45 | 3639 | /* The target and the pattern have the same utf8ness. */ |
d6a28714 JH |
3640 | /* Inline the first character, for speed. */ |
3641 | if (UCHARAT(s) != nextchr) | |
3642 | sayNO; | |
24d3c4a9 | 3643 | if (PL_regeol - locinput < ln) |
d6a28714 | 3644 | sayNO; |
24d3c4a9 | 3645 | if (ln > 1 && memNE(s, locinput, ln)) |
d6a28714 | 3646 | sayNO; |
24d3c4a9 | 3647 | locinput += ln; |
d6a28714 JH |
3648 | nextchr = UCHARAT(locinput); |
3649 | break; | |
95b24440 | 3650 | } |
9a5a5549 | 3651 | case EXACTFL: { |
a932d541 | 3652 | re_fold_t folder; |
9a5a5549 KW |
3653 | const U8 * fold_array; |
3654 | const char * s; | |
d513472c | 3655 | U32 fold_utf8_flags; |
9a5a5549 | 3656 | |
b8c5462f | 3657 | PL_reg_flags |= RF_tainted; |
9a5a5549 KW |
3658 | folder = foldEQ_locale; |
3659 | fold_array = PL_fold_locale; | |
17580e7a | 3660 | fold_utf8_flags = FOLDEQ_UTF8_LOCALE; |
9a5a5549 KW |
3661 | goto do_exactf; |
3662 | ||
3663 | case EXACTFU: | |
3664 | folder = foldEQ_latin1; | |
3665 | fold_array = PL_fold_latin1; | |
2daa8fee | 3666 | fold_utf8_flags = (UTF_PATTERN) ? FOLDEQ_S1_ALREADY_FOLDED : 0; |
9a5a5549 KW |
3667 | goto do_exactf; |
3668 | ||
2f7f8cb1 KW |
3669 | case EXACTFA: |
3670 | folder = foldEQ_latin1; | |
3671 | fold_array = PL_fold_latin1; | |
57014d77 | 3672 | fold_utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; |
2f7f8cb1 KW |
3673 | goto do_exactf; |
3674 | ||
9a5a5549 KW |
3675 | case EXACTF: |
3676 | folder = foldEQ; | |
3677 | fold_array = PL_fold; | |
2daa8fee | 3678 | fold_utf8_flags = (UTF_PATTERN) ? FOLDEQ_S1_ALREADY_FOLDED : 0; |
9a5a5549 KW |
3679 | |
3680 | do_exactf: | |
3681 | s = STRING(scan); | |
24d3c4a9 | 3682 | ln = STR_LEN(scan); |
d6a28714 | 3683 | |
f2ed9b32 | 3684 | if (utf8_target || UTF_PATTERN) { |
d07ddd77 | 3685 | /* Either target or the pattern are utf8. */ |
be8e71aa | 3686 | const char * const l = locinput; |
d07ddd77 | 3687 | char *e = PL_regeol; |
bc517b45 | 3688 | |
d513472c | 3689 | if (! foldEQ_utf8_flags(s, 0, ln, cBOOL(UTF_PATTERN), |
c3e1d013 KW |
3690 | l, &e, 0, utf8_target, fold_utf8_flags)) |
3691 | { | |
3692 | sayNO; | |
5486206c | 3693 | } |
d07ddd77 JH |
3694 | locinput = e; |
3695 | nextchr = UCHARAT(locinput); | |
3696 | break; | |
a0ed51b3 | 3697 | } |
d6a28714 | 3698 | |
0a138b74 | 3699 | /* Neither the target nor the pattern are utf8 */ |
d6a28714 | 3700 | if (UCHARAT(s) != nextchr && |
9a5a5549 KW |
3701 | UCHARAT(s) != fold_array[nextchr]) |
3702 | { | |
a0ed51b3 | 3703 | sayNO; |
9a5a5549 | 3704 | } |
24d3c4a9 | 3705 | if (PL_regeol - locinput < ln) |
b8c5462f | 3706 | sayNO; |
9a5a5549 | 3707 | if (ln > 1 && ! folder(s, locinput, ln)) |
4633a7c4 | 3708 | sayNO; |
24d3c4a9 | 3709 | locinput += ln; |
d6a28714 | 3710 | nextchr = UCHARAT(locinput); |
a0d0e21e | 3711 | break; |
9a5a5549 | 3712 | } |
63ac0dad KW |
3713 | |
3714 | /* XXX Could improve efficiency by separating these all out using a | |
3715 | * macro or in-line function. At that point regcomp.c would no longer | |
3716 | * have to set the FLAGS fields of these */ | |
b2680017 YO |
3717 | case BOUNDL: |
3718 | case NBOUNDL: | |
3719 | PL_reg_flags |= RF_tainted; | |
3720 | /* FALL THROUGH */ | |
3721 | case BOUND: | |
63ac0dad | 3722 | case BOUNDU: |
cfaf538b | 3723 | case BOUNDA: |
b2680017 | 3724 | case NBOUND: |
63ac0dad | 3725 | case NBOUNDU: |
cfaf538b | 3726 | case NBOUNDA: |
b2680017 | 3727 | /* was last char in word? */ |
f2e96b5d KW |
3728 | if (utf8_target |
3729 | && FLAGS(scan) != REGEX_ASCII_RESTRICTED_CHARSET | |
3730 | && FLAGS(scan) != REGEX_ASCII_MORE_RESTRICTED_CHARSET) | |
3731 | { | |
b2680017 YO |
3732 | if (locinput == PL_bostr) |
3733 | ln = '\n'; | |
3734 | else { | |
3735 | const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); | |
3736 | ||
3737 | ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); | |
3738 | } | |
63ac0dad | 3739 | if (FLAGS(scan) != REGEX_LOCALE_CHARSET) { |
b2680017 YO |
3740 | ln = isALNUM_uni(ln); |
3741 | LOAD_UTF8_CHARCLASS_ALNUM(); | |
f2ed9b32 | 3742 | n = swash_fetch(PL_utf8_alnum, (U8*)locinput, utf8_target); |
b2680017 YO |
3743 | } |
3744 | else { | |
3745 | ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); | |
3746 | n = isALNUM_LC_utf8((U8*)locinput); | |
3747 | } | |
3748 | } | |
3749 | else { | |
cfaf538b KW |
3750 | |
3751 | /* Here the string isn't utf8, or is utf8 and only ascii | |
3752 | * characters are to match \w. In the latter case looking at | |
3753 | * the byte just prior to the current one may be just the final | |
3754 | * byte of a multi-byte character. This is ok. There are two | |
3755 | * cases: | |
3756 | * 1) it is a single byte character, and then the test is doing | |
3757 | * just what it's supposed to. | |
3758 | * 2) it is a multi-byte character, in which case the final | |
3759 | * byte is never mistakable for ASCII, and so the test | |
3760 | * will say it is not a word character, which is the | |
3761 | * correct answer. */ | |
b2680017 YO |
3762 | ln = (locinput != PL_bostr) ? |
3763 | UCHARAT(locinput - 1) : '\n'; | |
63ac0dad KW |
3764 | switch (FLAGS(scan)) { |
3765 | case REGEX_UNICODE_CHARSET: | |
3766 | ln = isWORDCHAR_L1(ln); | |
3767 | n = isWORDCHAR_L1(nextchr); | |
3768 | break; | |
3769 | case REGEX_LOCALE_CHARSET: | |
3770 | ln = isALNUM_LC(ln); | |
3771 | n = isALNUM_LC(nextchr); | |
3772 | break; | |
3773 | case REGEX_DEPENDS_CHARSET: | |
3774 | ln = isALNUM(ln); | |
3775 | n = isALNUM(nextchr); | |
3776 | break; | |
cfaf538b | 3777 | case REGEX_ASCII_RESTRICTED_CHARSET: |
c973bd4f | 3778 | case REGEX_ASCII_MORE_RESTRICTED_CHARSET: |
cfaf538b KW |
3779 | ln = isWORDCHAR_A(ln); |
3780 | n = isWORDCHAR_A(nextchr); | |
3781 | break; | |
63ac0dad KW |
3782 | default: |
3783 | Perl_croak(aTHX_ "panic: Unexpected FLAGS %u in op %u", FLAGS(scan), OP(scan)); | |
3784 | break; | |
b2680017 YO |
3785 | } |
3786 | } | |
63ac0dad KW |
3787 | /* Note requires that all BOUNDs be lower than all NBOUNDs in |
3788 | * regcomp.sym */ | |
3789 | if (((!ln) == (!n)) == (OP(scan) < NBOUND)) | |
b2680017 YO |
3790 | sayNO; |
3791 | break; | |
f56b6394 | 3792 | case ANYOFV: |
d6a28714 | 3793 | case ANYOF: |
f56b6394 | 3794 | if (utf8_target || state_num == ANYOFV) { |
9e55ce06 | 3795 | STRLEN inclasslen = PL_regeol - locinput; |
20ed0b26 KW |
3796 | if (locinput >= PL_regeol) |
3797 | sayNO; | |
9e55ce06 | 3798 | |
f2ed9b32 | 3799 | if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target)) |
09b08e9b | 3800 | sayNO; |
b32d7d3e | 3801 | locinput += inclasslen; |
b8c5462f | 3802 | nextchr = UCHARAT(locinput); |
e0f9d4a8 | 3803 | break; |
ffc61ed2 JH |
3804 | } |
3805 | else { | |
3806 | if (nextchr < 0) | |
3807 | nextchr = UCHARAT(locinput); | |
ffc61ed2 JH |
3808 | if (!nextchr && locinput >= PL_regeol) |
3809 | sayNO; | |
20ed0b26 | 3810 | if (!REGINCLASS(rex, scan, (U8*)locinput)) |
09b08e9b | 3811 | sayNO; |
ffc61ed2 | 3812 | nextchr = UCHARAT(++locinput); |
e0f9d4a8 JH |
3813 | break; |
3814 | } | |
b8c5462f | 3815 | break; |
20d0b1e9 | 3816 | /* Special char classes - The defines start on line 129 or so */ |
ee9a90b8 KW |
3817 | CCC_TRY_U(ALNUM, NALNUM, isWORDCHAR, |
3818 | ALNUML, NALNUML, isALNUM_LC, isALNUM_LC_utf8, | |
3819 | ALNUMU, NALNUMU, isWORDCHAR_L1, | |
cfaf538b | 3820 | ALNUMA, NALNUMA, isWORDCHAR_A, |
779d7b58 | 3821 | alnum, "a"); |
ee9a90b8 KW |
3822 | |
3823 | CCC_TRY_U(SPACE, NSPACE, isSPACE, | |
3824 | SPACEL, NSPACEL, isSPACE_LC, isSPACE_LC_utf8, | |
3825 | SPACEU, NSPACEU, isSPACE_L1, | |
cfaf538b | 3826 | SPACEA, NSPACEA, isSPACE_A, |
779d7b58 | 3827 | space, " "); |
ee9a90b8 KW |
3828 | |
3829 | CCC_TRY(DIGIT, NDIGIT, isDIGIT, | |
3830 | DIGITL, NDIGITL, isDIGIT_LC, isDIGIT_LC_utf8, | |
cfaf538b | 3831 | DIGITA, NDIGITA, isDIGIT_A, |
779d7b58 | 3832 | digit, "0"); |
20d0b1e9 | 3833 | |
37e2e78e KW |
3834 | case CLUMP: /* Match \X: logical Unicode character. This is defined as |
3835 | a Unicode extended Grapheme Cluster */ | |
3836 | /* From http://www.unicode.org/reports/tr29 (5.2 version). An | |
3837 | extended Grapheme Cluster is: | |
3838 | ||
3839 | CR LF | |
3840 | | Prepend* Begin Extend* | |
3841 | | . | |
3842 | ||
3843 | Begin is (Hangul-syllable | ! Control) | |
3844 | Extend is (Grapheme_Extend | Spacing_Mark) | |
3845 | Control is [ GCB_Control CR LF ] | |
3846 | ||
3847 | The discussion below shows how the code for CLUMP is derived | |
3848 | from this regex. Note that most of these concepts are from | |
3849 | property values of the Grapheme Cluster Boundary (GCB) property. | |
3850 | No code point can have multiple property values for a given | |
3851 | property. Thus a code point in Prepend can't be in Control, but | |
3852 | it must be in !Control. This is why Control above includes | |
3853 | GCB_Control plus CR plus LF. The latter two are used in the GCB | |
3854 | property separately, and so can't be in GCB_Control, even though | |
3855 | they logically are controls. Control is not the same as gc=cc, | |
3856 | but includes format and other characters as well. | |
3857 | ||
3858 | The Unicode definition of Hangul-syllable is: | |
3859 | L+ | |
3860 | | (L* ( ( V | LV ) V* | LVT ) T*) | |
3861 | | T+ | |
3862 | ) | |
3863 | Each of these is a value for the GCB property, and hence must be | |
3864 | disjoint, so the order they are tested is immaterial, so the | |
3865 | above can safely be changed to | |
3866 | T+ | |
3867 | | L+ | |
3868 | | (L* ( LVT | ( V | LV ) V*) T*) | |
3869 | ||
3870 | The last two terms can be combined like this: | |
3871 | L* ( L | |
3872 | | (( LVT | ( V | LV ) V*) T*)) | |
3873 | ||
3874 | And refactored into this: | |
3875 | L* (L | LVT T* | V V* T* | LV V* T*) | |
3876 | ||
3877 | That means that if we have seen any L's at all we can quit | |
cc3b396d | 3878 | there, but if the next character is an LVT, a V, or an LV we |
37e2e78e KW |
3879 | should keep going. |
3880 | ||
3881 | There is a subtlety with Prepend* which showed up in testing. | |
3882 | Note that the Begin, and only the Begin is required in: | |
3883 | | Prepend* Begin Extend* | |
cc3b396d KW |
3884 | Also, Begin contains '! Control'. A Prepend must be a |
3885 | '! Control', which means it must also be a Begin. What it | |
3886 | comes down to is that if we match Prepend* and then find no | |
3887 | suitable Begin afterwards, that if we backtrack the last | |
3888 | Prepend, that one will be a suitable Begin. | |
37e2e78e KW |
3889 | */ |
3890 | ||
b7c83a7e | 3891 | if (locinput >= PL_regeol) |
a0ed51b3 | 3892 | sayNO; |
f2ed9b32 | 3893 | if (! utf8_target) { |
37e2e78e KW |
3894 | |
3895 | /* Match either CR LF or '.', as all the other possibilities | |
3896 | * require utf8 */ | |
3897 | locinput++; /* Match the . or CR */ | |
cc3b396d KW |
3898 | if (nextchr == '\r' /* And if it was CR, and the next is LF, |
3899 | match the LF */ | |
37e2e78e KW |
3900 | && locinput < PL_regeol |
3901 | && UCHARAT(locinput) == '\n') locinput++; | |
3902 | } | |
3903 | else { | |
3904 | ||
3905 | /* Utf8: See if is ( CR LF ); already know that locinput < | |
3906 | * PL_regeol, so locinput+1 is in bounds */ | |
3907 | if (nextchr == '\r' && UCHARAT(locinput + 1) == '\n') { | |
3908 | locinput += 2; | |
3909 | } | |
3910 | else { | |
3911 | /* In case have to backtrack to beginning, then match '.' */ | |
3912 | char *starting = locinput; | |
3913 | ||
3914 | /* In case have to backtrack the last prepend */ | |
3915 | char *previous_prepend = 0; | |
3916 | ||
3917 | LOAD_UTF8_CHARCLASS_GCB(); | |
3918 | ||
3919 | /* Match (prepend)* */ | |
3920 | while (locinput < PL_regeol | |
3921 | && swash_fetch(PL_utf8_X_prepend, | |
f2ed9b32 | 3922 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3923 | { |
3924 | previous_prepend = locinput; | |
3925 | locinput += UTF8SKIP(locinput); | |
3926 | } | |
3927 | ||
3928 | /* As noted above, if we matched a prepend character, but | |
3929 | * the next thing won't match, back off the last prepend we | |
3930 | * matched, as it is guaranteed to match the begin */ | |
3931 | if (previous_prepend | |
3932 | && (locinput >= PL_regeol | |
3933 | || ! swash_fetch(PL_utf8_X_begin, | |
f2ed9b32 | 3934 | (U8*)locinput, utf8_target))) |
37e2e78e KW |
3935 | { |
3936 | locinput = previous_prepend; | |
3937 | } | |
3938 | ||
3939 | /* Note that here we know PL_regeol > locinput, as we | |
3940 | * tested that upon input to this switch case, and if we | |
3941 | * moved locinput forward, we tested the result just above | |
3942 | * and it either passed, or we backed off so that it will | |
3943 | * now pass */ | |
f2ed9b32 | 3944 | if (! swash_fetch(PL_utf8_X_begin, (U8*)locinput, utf8_target)) { |
37e2e78e KW |
3945 | |
3946 | /* Here did not match the required 'Begin' in the | |
3947 | * second term. So just match the very first | |
3948 | * character, the '.' of the final term of the regex */ | |
3949 | locinput = starting + UTF8SKIP(starting); | |
3950 | } else { | |
3951 | ||
3952 | /* Here is the beginning of a character that can have | |
3953 | * an extender. It is either a hangul syllable, or a | |
3954 | * non-control */ | |
3955 | if (swash_fetch(PL_utf8_X_non_hangul, | |
f2ed9b32 | 3956 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3957 | { |
3958 | ||
3959 | /* Here not a Hangul syllable, must be a | |
3960 | * ('! * Control') */ | |
3961 | locinput += UTF8SKIP(locinput); | |
3962 | } else { | |
3963 | ||
3964 | /* Here is a Hangul syllable. It can be composed | |
3965 | * of several individual characters. One | |
3966 | * possibility is T+ */ | |
3967 | if (swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 3968 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3969 | { |
3970 | while (locinput < PL_regeol | |
3971 | && swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 3972 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3973 | { |
3974 | locinput += UTF8SKIP(locinput); | |
3975 | } | |
3976 | } else { | |
3977 | ||
3978 | /* Here, not T+, but is a Hangul. That means | |
3979 | * it is one of the others: L, LV, LVT or V, | |
3980 | * and matches: | |
3981 | * L* (L | LVT T* | V V* T* | LV V* T*) */ | |
3982 | ||
3983 | /* Match L* */ | |
3984 | while (locinput < PL_regeol | |
3985 | && swash_fetch(PL_utf8_X_L, | |
f2ed9b32 | 3986 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3987 | { |
3988 | locinput += UTF8SKIP(locinput); | |
3989 | } | |
3990 | ||
3991 | /* Here, have exhausted L*. If the next | |
3992 | * character is not an LV, LVT nor V, it means | |
3993 | * we had to have at least one L, so matches L+ | |
3994 | * in the original equation, we have a complete | |
3995 | * hangul syllable. Are done. */ | |
3996 | ||
3997 | if (locinput < PL_regeol | |
3998 | && swash_fetch(PL_utf8_X_LV_LVT_V, | |
f2ed9b32 | 3999 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4000 | { |
4001 | ||
4002 | /* Otherwise keep going. Must be LV, LVT | |
4003 | * or V. See if LVT */ | |
4004 | if (swash_fetch(PL_utf8_X_LVT, | |
f2ed9b32 | 4005 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4006 | { |
4007 | locinput += UTF8SKIP(locinput); | |
4008 | } else { | |
4009 | ||
4010 | /* Must be V or LV. Take it, then | |
4011 | * match V* */ | |
4012 | locinput += UTF8SKIP(locinput); | |
4013 | while (locinput < PL_regeol | |
4014 | && swash_fetch(PL_utf8_X_V, | |
f2ed9b32 | 4015 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4016 | { |
4017 | locinput += UTF8SKIP(locinput); | |
4018 | } | |
4019 | } | |
4020 | ||
4021 | /* And any of LV, LVT, or V can be followed | |
4022 | * by T* */ | |
4023 | while (locinput < PL_regeol | |
4024 | && swash_fetch(PL_utf8_X_T, | |
4025 | (U8*)locinput, | |
f2ed9b32 | 4026 | utf8_target)) |
37e2e78e KW |
4027 | { |
4028 | locinput += UTF8SKIP(locinput); | |
4029 | } | |
4030 | } | |
4031 | } | |
4032 | } | |
4033 | ||
4034 | /* Match any extender */ | |
4035 | while (locinput < PL_regeol | |
4036 | && swash_fetch(PL_utf8_X_extend, | |
f2ed9b32 | 4037 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
4038 | { |
4039 | locinput += UTF8SKIP(locinput); | |
4040 | } | |
4041 | } | |
4042 | } | |
4043 | if (locinput > PL_regeol) sayNO; | |
4044 | } | |
a0ed51b3 LW |
4045 | nextchr = UCHARAT(locinput); |
4046 | break; | |
81714fb9 YO |
4047 | |
4048 | case NREFFL: | |
d7ef4b73 KW |
4049 | { /* The capture buffer cases. The ones beginning with N for the |
4050 | named buffers just convert to the equivalent numbered and | |
4051 | pretend they were called as the corresponding numbered buffer | |
4052 | op. */ | |
26ecd678 TC |
4053 | /* don't initialize these in the declaration, it makes C++ |
4054 | unhappy */ | |
81714fb9 | 4055 | char *s; |
ff1157ca | 4056 | char type; |
8368298a TC |
4057 | re_fold_t folder; |
4058 | const U8 *fold_array; | |
26ecd678 | 4059 | UV utf8_fold_flags; |
8368298a | 4060 | |
81714fb9 | 4061 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4062 | folder = foldEQ_locale; |
4063 | fold_array = PL_fold_locale; | |
4064 | type = REFFL; | |
17580e7a | 4065 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4066 | goto do_nref; |
4067 | ||
2f7f8cb1 KW |
4068 | case NREFFA: |
4069 | folder = foldEQ_latin1; | |
4070 | fold_array = PL_fold_latin1; | |
4071 | type = REFFA; | |
4072 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4073 | goto do_nref; | |
4074 | ||
d7ef4b73 KW |
4075 | case NREFFU: |
4076 | folder = foldEQ_latin1; | |
4077 | fold_array = PL_fold_latin1; | |
4078 | type = REFFU; | |
d513472c | 4079 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4080 | goto do_nref; |
4081 | ||
81714fb9 | 4082 | case NREFF: |
d7ef4b73 KW |
4083 | folder = foldEQ; |
4084 | fold_array = PL_fold; | |
4085 | type = REFF; | |
d513472c | 4086 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4087 | goto do_nref; |
4088 | ||
4089 | case NREF: | |
4090 | type = REF; | |
83d7b90b KW |
4091 | folder = NULL; |
4092 | fold_array = NULL; | |
d513472c | 4093 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4094 | do_nref: |
4095 | ||
4096 | /* For the named back references, find the corresponding buffer | |
4097 | * number */ | |
0a4db386 YO |
4098 | n = reg_check_named_buff_matched(rex,scan); |
4099 | ||
d7ef4b73 | 4100 | if ( ! n ) { |
81714fb9 | 4101 | sayNO; |
d7ef4b73 KW |
4102 | } |
4103 | goto do_nref_ref_common; | |
4104 | ||
c8756f30 | 4105 | case REFFL: |
3280af22 | 4106 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
4107 | folder = foldEQ_locale; |
4108 | fold_array = PL_fold_locale; | |
17580e7a | 4109 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
d7ef4b73 KW |
4110 | goto do_ref; |
4111 | ||
2f7f8cb1 KW |
4112 | case REFFA: |
4113 | folder = foldEQ_latin1; | |
4114 | fold_array = PL_fold_latin1; | |
4115 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
4116 | goto do_ref; | |
4117 | ||
d7ef4b73 KW |
4118 | case REFFU: |
4119 | folder = foldEQ_latin1; | |
4120 | fold_array = PL_fold_latin1; | |
d513472c | 4121 | utf8_fold_flags = 0; |
d7ef4b73 KW |
4122 | goto do_ref; |
4123 | ||
4124 | case REFF: | |
4125 | folder = foldEQ; | |
4126 | fold_array = PL_fold; | |
d513472c | 4127 | utf8_fold_flags = 0; |
83d7b90b | 4128 | goto do_ref; |
d7ef4b73 | 4129 | |
c277df42 | 4130 | case REF: |
83d7b90b KW |
4131 | folder = NULL; |
4132 | fold_array = NULL; | |
d513472c | 4133 | utf8_fold_flags = 0; |
83d7b90b | 4134 | |
d7ef4b73 | 4135 | do_ref: |
81714fb9 | 4136 | type = OP(scan); |
d7ef4b73 KW |
4137 | n = ARG(scan); /* which paren pair */ |
4138 | ||
4139 | do_nref_ref_common: | |
f0ab9afb | 4140 | ln = PL_regoffs[n].start; |
2c2d71f5 | 4141 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
3b6647e0 | 4142 | if (*PL_reglastparen < n || ln == -1) |
af3f8c16 | 4143 | sayNO; /* Do not match unless seen CLOSEn. */ |
f0ab9afb | 4144 | if (ln == PL_regoffs[n].end) |
a0d0e21e | 4145 | break; |
a0ed51b3 | 4146 | |
24d3c4a9 | 4147 | s = PL_bostr + ln; |
d7ef4b73 | 4148 | if (type != REF /* REF can do byte comparison */ |
2f65c56d | 4149 | && (utf8_target || type == REFFU)) |
d7ef4b73 KW |
4150 | { /* XXX handle REFFL better */ |
4151 | char * limit = PL_regeol; | |
4152 | ||
4153 | /* This call case insensitively compares the entire buffer | |
4154 | * at s, with the current input starting at locinput, but | |
4155 | * not going off the end given by PL_regeol, and returns in | |
4156 | * limit upon success, how much of the current input was | |
4157 | * matched */ | |
d513472c KW |
4158 | if (! foldEQ_utf8_flags(s, NULL, PL_regoffs[n].end - ln, utf8_target, |
4159 | locinput, &limit, 0, utf8_target, utf8_fold_flags)) | |
d7ef4b73 KW |
4160 | { |
4161 | sayNO; | |
a0ed51b3 | 4162 | } |
d7ef4b73 | 4163 | locinput = limit; |
a0ed51b3 LW |
4164 | nextchr = UCHARAT(locinput); |
4165 | break; | |
4166 | } | |
4167 | ||
d7ef4b73 | 4168 | /* Not utf8: Inline the first character, for speed. */ |
76e3520e | 4169 | if (UCHARAT(s) != nextchr && |
81714fb9 | 4170 | (type == REF || |
d7ef4b73 | 4171 | UCHARAT(s) != fold_array[nextchr])) |
4633a7c4 | 4172 | sayNO; |
f0ab9afb | 4173 | ln = PL_regoffs[n].end - ln; |
24d3c4a9 | 4174 | if (locinput + ln > PL_regeol) |
4633a7c4 | 4175 | sayNO; |
81714fb9 | 4176 | if (ln > 1 && (type == REF |
24d3c4a9 | 4177 | ? memNE(s, locinput, ln) |
d7ef4b73 | 4178 | : ! folder(s, locinput, ln))) |
4633a7c4 | 4179 | sayNO; |
24d3c4a9 | 4180 | locinput += ln; |
76e3520e | 4181 | nextchr = UCHARAT(locinput); |
a0d0e21e | 4182 | break; |
81714fb9 | 4183 | } |
a0d0e21e | 4184 | case NOTHING: |
c277df42 | 4185 | case TAIL: |
a0d0e21e LW |
4186 | break; |
4187 | case BACK: | |
4188 | break; | |
40a82448 DM |
4189 | |
4190 | #undef ST | |
4191 | #define ST st->u.eval | |
c277df42 | 4192 | { |
c277df42 | 4193 | SV *ret; |
d2f13c59 | 4194 | REGEXP *re_sv; |
6bda09f9 | 4195 | regexp *re; |
f8fc2ecf | 4196 | regexp_internal *rei; |
1a147d38 YO |
4197 | regnode *startpoint; |
4198 | ||
4199 | case GOSTART: | |
e7707071 YO |
4200 | case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */ |
4201 | if (cur_eval && cur_eval->locinput==locinput) { | |
24b23f37 | 4202 | if (cur_eval->u.eval.close_paren == (U32)ARG(scan)) |
1a147d38 | 4203 | Perl_croak(aTHX_ "Infinite recursion in regex"); |
4b196cd4 | 4204 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 YO |
4205 | Perl_croak(aTHX_ |
4206 | "Pattern subroutine nesting without pos change" | |
4207 | " exceeded limit in regex"); | |
6bda09f9 YO |
4208 | } else { |
4209 | nochange_depth = 0; | |
1a147d38 | 4210 | } |
288b8c02 | 4211 | re_sv = rex_sv; |
6bda09f9 | 4212 | re = rex; |
f8fc2ecf | 4213 | rei = rexi; |
288b8c02 | 4214 | (void)ReREFCNT_inc(rex_sv); |
1a147d38 | 4215 | if (OP(scan)==GOSUB) { |
6bda09f9 YO |
4216 | startpoint = scan + ARG2L(scan); |
4217 | ST.close_paren = ARG(scan); | |
4218 | } else { | |
f8fc2ecf | 4219 | startpoint = rei->program+1; |
6bda09f9 YO |
4220 | ST.close_paren = 0; |
4221 | } | |
4222 | goto eval_recurse_doit; | |
4223 | /* NOTREACHED */ | |
4224 | case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ | |
4225 | if (cur_eval && cur_eval->locinput==locinput) { | |
4b196cd4 | 4226 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 | 4227 | Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex"); |
6bda09f9 YO |
4228 | } else { |
4229 | nochange_depth = 0; | |
4230 | } | |
8e5e9ebe | 4231 | { |
4aabdb9b DM |
4232 | /* execute the code in the {...} */ |
4233 | dSP; | |
6136c704 | 4234 | SV ** const before = SP; |
4aabdb9b DM |
4235 | OP_4tree * const oop = PL_op; |
4236 | COP * const ocurcop = PL_curcop; | |
4237 | PAD *old_comppad; | |
d80618d2 | 4238 | char *saved_regeol = PL_regeol; |
91332126 FR |
4239 | struct re_save_state saved_state; |
4240 | ||
6562f1c4 | 4241 | /* To not corrupt the existing regex state while executing the |
b7f4cd04 FR |
4242 | * eval we would normally put it on the save stack, like with |
4243 | * save_re_context. However, re-evals have a weird scoping so we | |
4244 | * can't just add ENTER/LEAVE here. With that, things like | |
4245 | * | |
4246 | * (?{$a=2})(a(?{local$a=$a+1}))*aak*c(?{$b=$a}) | |
4247 | * | |
4248 | * would break, as they expect the localisation to be unwound | |
4249 | * only when the re-engine backtracks through the bit that | |
4250 | * localised it. | |
4251 | * | |
4252 | * What we do instead is just saving the state in a local c | |
4253 | * variable. | |
4254 | */ | |
91332126 FR |
4255 | Copy(&PL_reg_state, &saved_state, 1, struct re_save_state); |
4256 | ||
4aabdb9b | 4257 | n = ARG(scan); |
f8fc2ecf | 4258 | PL_op = (OP_4tree*)rexi->data->data[n]; |
24b23f37 YO |
4259 | DEBUG_STATE_r( PerlIO_printf(Perl_debug_log, |
4260 | " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); | |
9e103e26 DM |
4261 | /* wrap the call in two SAVECOMPPADs. This ensures that |
4262 | * when the save stack is eventually unwound, all the | |
4263 | * accumulated SAVEt_CLEARSV's will be processed with | |
4264 | * interspersed SAVEt_COMPPAD's to ensure that lexicals | |
4265 | * are cleared in the right pad */ | |
4266 | SAVECOMPPAD(); | |
f8fc2ecf | 4267 | PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]); |
f0ab9afb | 4268 | PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr; |
4aabdb9b | 4269 | |
2bf803e2 YO |
4270 | if (sv_yes_mark) { |
4271 | SV *sv_mrk = get_sv("REGMARK", 1); | |
4272 | sv_setsv(sv_mrk, sv_yes_mark); | |
4273 | } | |
4274 | ||
8e5e9ebe RGS |
4275 | CALLRUNOPS(aTHX); /* Scalar context. */ |
4276 | SPAGAIN; | |
4277 | if (SP == before) | |
075aa684 | 4278 | ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */ |
8e5e9ebe RGS |
4279 | else { |
4280 | ret = POPs; | |
4281 | PUTBACK; | |
4282 | } | |
4aabdb9b | 4283 | |
91332126 FR |
4284 | Copy(&saved_state, &PL_reg_state, 1, struct re_save_state); |
4285 | ||
4aabdb9b | 4286 | PL_op = oop; |
9e103e26 | 4287 | SAVECOMPPAD(); |
4aabdb9b DM |
4288 | PAD_RESTORE_LOCAL(old_comppad); |
4289 | PL_curcop = ocurcop; | |
d80618d2 | 4290 | PL_regeol = saved_regeol; |
24d3c4a9 | 4291 | if (!logical) { |
4aabdb9b DM |
4292 | /* /(?{...})/ */ |
4293 | sv_setsv(save_scalar(PL_replgv), ret); | |
4aabdb9b DM |
4294 | break; |
4295 | } | |
8e5e9ebe | 4296 | } |
24d3c4a9 DM |
4297 | if (logical == 2) { /* Postponed subexpression: /(??{...})/ */ |
4298 | logical = 0; | |
4aabdb9b | 4299 | { |
4f639d21 DM |
4300 | /* extract RE object from returned value; compiling if |
4301 | * necessary */ | |
6136c704 | 4302 | MAGIC *mg = NULL; |
288b8c02 | 4303 | REGEXP *rx = NULL; |
5c35adbb NC |
4304 | |
4305 | if (SvROK(ret)) { | |
288b8c02 | 4306 | SV *const sv = SvRV(ret); |
5c35adbb NC |
4307 | |
4308 | if (SvTYPE(sv) == SVt_REGEXP) { | |
d2f13c59 | 4309 | rx = (REGEXP*) sv; |
5c35adbb NC |
4310 | } else if (SvSMAGICAL(sv)) { |
4311 | mg = mg_find(sv, PERL_MAGIC_qr); | |
4312 | assert(mg); | |
4313 | } | |
4314 | } else if (SvTYPE(ret) == SVt_REGEXP) { | |
d2f13c59 | 4315 | rx = (REGEXP*) ret; |
5c35adbb | 4316 | } else if (SvSMAGICAL(ret)) { |
124ee91a NC |
4317 | if (SvGMAGICAL(ret)) { |
4318 | /* I don't believe that there is ever qr magic | |
4319 | here. */ | |
4320 | assert(!mg_find(ret, PERL_MAGIC_qr)); | |
faf82a0b | 4321 | sv_unmagic(ret, PERL_MAGIC_qr); |
124ee91a NC |
4322 | } |
4323 | else { | |
faf82a0b | 4324 | mg = mg_find(ret, PERL_MAGIC_qr); |
124ee91a NC |
4325 | /* testing suggests mg only ends up non-NULL for |
4326 | scalars who were upgraded and compiled in the | |
4327 | else block below. In turn, this is only | |
4328 | triggered in the "postponed utf8 string" tests | |
4329 | in t/op/pat.t */ | |
4330 | } | |
0f5d15d6 | 4331 | } |
faf82a0b | 4332 | |
0f5d15d6 | 4333 | if (mg) { |
d2f13c59 | 4334 | rx = (REGEXP *) mg->mg_obj; /*XXX:dmq*/ |
e2560c33 | 4335 | assert(rx); |
0f5d15d6 | 4336 | } |
288b8c02 | 4337 | if (rx) { |
f0826785 | 4338 | rx = reg_temp_copy(NULL, rx); |
288b8c02 | 4339 | } |
0f5d15d6 | 4340 | else { |
c737faaf | 4341 | U32 pm_flags = 0; |
a3b680e6 | 4342 | const I32 osize = PL_regsize; |
0f5d15d6 | 4343 | |
b9ad30b4 NC |
4344 | if (DO_UTF8(ret)) { |
4345 | assert (SvUTF8(ret)); | |
4346 | } else if (SvUTF8(ret)) { | |
4347 | /* Not doing UTF-8, despite what the SV says. Is | |
4348 | this only if we're trapped in use 'bytes'? */ | |
4349 | /* Make a copy of the octet sequence, but without | |
4350 | the flag on, as the compiler now honours the | |
4351 | SvUTF8 flag on ret. */ | |
4352 | STRLEN len; | |
4353 | const char *const p = SvPV(ret, len); | |
4354 | ret = newSVpvn_flags(p, len, SVs_TEMP); | |
4355 | } | |
288b8c02 | 4356 | rx = CALLREGCOMP(ret, pm_flags); |
9041c2e3 | 4357 | if (!(SvFLAGS(ret) |
faf82a0b | 4358 | & (SVs_TEMP | SVs_PADTMP | SVf_READONLY |
3ce3ed55 | 4359 | | SVs_GMG))) { |
a2794585 NC |
4360 | /* This isn't a first class regexp. Instead, it's |
4361 | caching a regexp onto an existing, Perl visible | |
4362 | scalar. */ | |
ad64d0ec | 4363 | sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0); |
3ce3ed55 | 4364 | } |
0f5d15d6 | 4365 | PL_regsize = osize; |
0f5d15d6 | 4366 | } |
288b8c02 NC |
4367 | re_sv = rx; |
4368 | re = (struct regexp *)SvANY(rx); | |
4aabdb9b | 4369 | } |
07bc277f | 4370 | RXp_MATCH_COPIED_off(re); |
28d8d7f4 YO |
4371 | re->subbeg = rex->subbeg; |
4372 | re->sublen = rex->sublen; | |
f8fc2ecf | 4373 | rei = RXi_GET(re); |
6bda09f9 | 4374 | DEBUG_EXECUTE_r( |
f2ed9b32 | 4375 | debug_start_match(re_sv, utf8_target, locinput, PL_regeol, |
6bda09f9 YO |
4376 | "Matching embedded"); |
4377 | ); | |
f8fc2ecf | 4378 | startpoint = rei->program + 1; |
1a147d38 | 4379 | ST.close_paren = 0; /* only used for GOSUB */ |
6bda09f9 YO |
4380 | /* borrowed from regtry */ |
4381 | if (PL_reg_start_tmpl <= re->nparens) { | |
4382 | PL_reg_start_tmpl = re->nparens*3/2 + 3; | |
4383 | if(PL_reg_start_tmp) | |
4384 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
4385 | else | |
4386 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
dd5def09 | 4387 | } |
aa283a38 | 4388 | |
1a147d38 | 4389 | eval_recurse_doit: /* Share code with GOSUB below this line */ |
aa283a38 | 4390 | /* run the pattern returned from (??{...}) */ |
40a82448 DM |
4391 | ST.cp = regcppush(0); /* Save *all* the positions. */ |
4392 | REGCP_SET(ST.lastcp); | |
6bda09f9 | 4393 | |
f0ab9afb | 4394 | PL_regoffs = re->offs; /* essentially NOOP on GOSUB */ |
6bda09f9 | 4395 | |
0357f1fd ML |
4396 | /* see regtry, specifically PL_reglast(?:close)?paren is a pointer! (i dont know why) :dmq */ |
4397 | PL_reglastparen = &re->lastparen; | |
4398 | PL_reglastcloseparen = &re->lastcloseparen; | |
4399 | re->lastparen = 0; | |
4400 | re->lastcloseparen = 0; | |
4401 | ||
4aabdb9b | 4402 | PL_reginput = locinput; |
ae0beba1 | 4403 | PL_regsize = 0; |
4aabdb9b DM |
4404 | |
4405 | /* XXXX This is too dramatic a measure... */ | |
4406 | PL_reg_maxiter = 0; | |
4407 | ||
faec1544 | 4408 | ST.toggle_reg_flags = PL_reg_flags; |
3c8556c3 | 4409 | if (RX_UTF8(re_sv)) |
faec1544 DM |
4410 | PL_reg_flags |= RF_utf8; |
4411 | else | |
4412 | PL_reg_flags &= ~RF_utf8; | |
4413 | ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */ | |
4414 | ||
288b8c02 | 4415 | ST.prev_rex = rex_sv; |
faec1544 | 4416 | ST.prev_curlyx = cur_curlyx; |
288b8c02 NC |
4417 | SETREX(rex_sv,re_sv); |
4418 | rex = re; | |
f8fc2ecf | 4419 | rexi = rei; |
faec1544 | 4420 | cur_curlyx = NULL; |
40a82448 | 4421 | ST.B = next; |
faec1544 DM |
4422 | ST.prev_eval = cur_eval; |
4423 | cur_eval = st; | |
faec1544 | 4424 | /* now continue from first node in postoned RE */ |
6bda09f9 | 4425 | PUSH_YES_STATE_GOTO(EVAL_AB, startpoint); |
4aabdb9b | 4426 | /* NOTREACHED */ |
a0ed51b3 | 4427 | } |
24d3c4a9 | 4428 | /* logical is 1, /(?(?{...})X|Y)/ */ |
f2338a2e | 4429 | sw = cBOOL(SvTRUE(ret)); |
24d3c4a9 | 4430 | logical = 0; |
c277df42 IZ |
4431 | break; |
4432 | } | |
40a82448 | 4433 | |
faec1544 DM |
4434 | case EVAL_AB: /* cleanup after a successful (??{A})B */ |
4435 | /* note: this is called twice; first after popping B, then A */ | |
4436 | PL_reg_flags ^= ST.toggle_reg_flags; | |
288b8c02 NC |
4437 | ReREFCNT_dec(rex_sv); |
4438 | SETREX(rex_sv,ST.prev_rex); | |
4439 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 4440 | rexi = RXi_GET(rex); |
faec1544 DM |
4441 | regcpblow(ST.cp); |
4442 | cur_eval = ST.prev_eval; | |
4443 | cur_curlyx = ST.prev_curlyx; | |
34a81e2b B |
4444 | |
4445 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ | |
0357f1fd ML |
4446 | PL_reglastparen = &rex->lastparen; |
4447 | PL_reglastcloseparen = &rex->lastcloseparen; | |
34a81e2b B |
4448 | /* also update PL_regoffs */ |
4449 | PL_regoffs = rex->offs; | |
0357f1fd | 4450 | |
40a82448 DM |
4451 | /* XXXX This is too dramatic a measure... */ |
4452 | PL_reg_maxiter = 0; | |
e7707071 | 4453 | if ( nochange_depth ) |
4b196cd4 | 4454 | nochange_depth--; |
262b90c4 | 4455 | sayYES; |
40a82448 | 4456 | |
40a82448 | 4457 | |
faec1544 DM |
4458 | case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */ |
4459 | /* note: this is called twice; first after popping B, then A */ | |
4460 | PL_reg_flags ^= ST.toggle_reg_flags; | |
288b8c02 NC |
4461 | ReREFCNT_dec(rex_sv); |
4462 | SETREX(rex_sv,ST.prev_rex); | |
4463 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 4464 | rexi = RXi_GET(rex); |
34a81e2b | 4465 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ |
0357f1fd ML |
4466 | PL_reglastparen = &rex->lastparen; |
4467 | PL_reglastcloseparen = &rex->lastcloseparen; | |
4468 | ||
40a82448 DM |
4469 | PL_reginput = locinput; |
4470 | REGCP_UNWIND(ST.lastcp); | |
4471 | regcppop(rex); | |
faec1544 DM |
4472 | cur_eval = ST.prev_eval; |
4473 | cur_curlyx = ST.prev_curlyx; | |
4474 | /* XXXX This is too dramatic a measure... */ | |
4475 | PL_reg_maxiter = 0; | |
e7707071 | 4476 | if ( nochange_depth ) |
4b196cd4 | 4477 | nochange_depth--; |
40a82448 | 4478 | sayNO_SILENT; |
40a82448 DM |
4479 | #undef ST |
4480 | ||
a0d0e21e | 4481 | case OPEN: |
c277df42 | 4482 | n = ARG(scan); /* which paren pair */ |
3280af22 NIS |
4483 | PL_reg_start_tmp[n] = locinput; |
4484 | if (n > PL_regsize) | |
4485 | PL_regsize = n; | |
e2e6a0f1 | 4486 | lastopen = n; |
a0d0e21e LW |
4487 | break; |
4488 | case CLOSE: | |
c277df42 | 4489 | n = ARG(scan); /* which paren pair */ |
f0ab9afb NC |
4490 | PL_regoffs[n].start = PL_reg_start_tmp[n] - PL_bostr; |
4491 | PL_regoffs[n].end = locinput - PL_bostr; | |
7f69552c YO |
4492 | /*if (n > PL_regsize) |
4493 | PL_regsize = n;*/ | |
3b6647e0 | 4494 | if (n > *PL_reglastparen) |
3280af22 | 4495 | *PL_reglastparen = n; |
a01268b5 | 4496 | *PL_reglastcloseparen = n; |
3b6647e0 | 4497 | if (cur_eval && cur_eval->u.eval.close_paren == n) { |
6bda09f9 YO |
4498 | goto fake_end; |
4499 | } | |
a0d0e21e | 4500 | break; |
e2e6a0f1 YO |
4501 | case ACCEPT: |
4502 | if (ARG(scan)){ | |
4503 | regnode *cursor; | |
4504 | for (cursor=scan; | |
4505 | cursor && OP(cursor)!=END; | |
4506 | cursor=regnext(cursor)) | |
4507 | { | |
4508 | if ( OP(cursor)==CLOSE ){ | |
4509 | n = ARG(cursor); | |
4510 | if ( n <= lastopen ) { | |
f0ab9afb NC |
4511 | PL_regoffs[n].start |
4512 | = PL_reg_start_tmp[n] - PL_bostr; | |
4513 | PL_regoffs[n].end = locinput - PL_bostr; | |
e2e6a0f1 YO |
4514 | /*if (n > PL_regsize) |
4515 | PL_regsize = n;*/ | |
3b6647e0 | 4516 | if (n > *PL_reglastparen) |
e2e6a0f1 YO |
4517 | *PL_reglastparen = n; |
4518 | *PL_reglastcloseparen = n; | |
3b6647e0 RB |
4519 | if ( n == ARG(scan) || (cur_eval && |
4520 | cur_eval->u.eval.close_paren == n)) | |
e2e6a0f1 YO |
4521 | break; |
4522 | } | |
4523 | } | |
4524 | } | |
4525 | } | |
4526 | goto fake_end; | |
4527 | /*NOTREACHED*/ | |
c277df42 IZ |
4528 | case GROUPP: |
4529 | n = ARG(scan); /* which paren pair */ | |
f2338a2e | 4530 | sw = cBOOL(*PL_reglastparen >= n && PL_regoffs[n].end != -1); |
c277df42 | 4531 | break; |
0a4db386 YO |
4532 | case NGROUPP: |
4533 | /* reg_check_named_buff_matched returns 0 for no match */ | |
f2338a2e | 4534 | sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan)); |
0a4db386 | 4535 | break; |
1a147d38 | 4536 | case INSUBP: |
0a4db386 | 4537 | n = ARG(scan); |
3b6647e0 | 4538 | sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n)); |
0a4db386 YO |
4539 | break; |
4540 | case DEFINEP: | |
4541 | sw = 0; | |
4542 | break; | |
c277df42 | 4543 | case IFTHEN: |
2c2d71f5 | 4544 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
24d3c4a9 | 4545 | if (sw) |
c277df42 IZ |
4546 | next = NEXTOPER(NEXTOPER(scan)); |
4547 | else { | |
4548 | next = scan + ARG(scan); | |
4549 | if (OP(next) == IFTHEN) /* Fake one. */ | |
4550 | next = NEXTOPER(NEXTOPER(next)); | |
4551 | } | |
4552 | break; | |
4553 | case LOGICAL: | |
24d3c4a9 | 4554 | logical = scan->flags; |
c277df42 | 4555 | break; |
c476f425 | 4556 | |
2ab05381 | 4557 | /******************************************************************* |
2ab05381 | 4558 | |
c476f425 DM |
4559 | The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/ |
4560 | pattern, where A and B are subpatterns. (For simple A, CURLYM or | |
4561 | STAR/PLUS/CURLY/CURLYN are used instead.) | |
2ab05381 | 4562 | |
c476f425 | 4563 | A*B is compiled as <CURLYX><A><WHILEM><B> |
2ab05381 | 4564 | |
c476f425 DM |
4565 | On entry to the subpattern, CURLYX is called. This pushes a CURLYX |
4566 | state, which contains the current count, initialised to -1. It also sets | |
4567 | cur_curlyx to point to this state, with any previous value saved in the | |
4568 | state block. | |
2ab05381 | 4569 | |
c476f425 DM |
4570 | CURLYX then jumps straight to the WHILEM op, rather than executing A, |
4571 | since the pattern may possibly match zero times (i.e. it's a while {} loop | |
4572 | rather than a do {} while loop). | |
2ab05381 | 4573 | |
c476f425 DM |
4574 | Each entry to WHILEM represents a successful match of A. The count in the |
4575 | CURLYX block is incremented, another WHILEM state is pushed, and execution | |
4576 | passes to A or B depending on greediness and the current count. | |
2ab05381 | 4577 | |
c476f425 DM |
4578 | For example, if matching against the string a1a2a3b (where the aN are |
4579 | substrings that match /A/), then the match progresses as follows: (the | |
4580 | pushed states are interspersed with the bits of strings matched so far): | |
2ab05381 | 4581 | |
c476f425 DM |
4582 | <CURLYX cnt=-1> |
4583 | <CURLYX cnt=0><WHILEM> | |
4584 | <CURLYX cnt=1><WHILEM> a1 <WHILEM> | |
4585 | <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM> | |
4586 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> | |
4587 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b | |
2ab05381 | 4588 | |
c476f425 DM |
4589 | (Contrast this with something like CURLYM, which maintains only a single |
4590 | backtrack state: | |
2ab05381 | 4591 | |
c476f425 DM |
4592 | <CURLYM cnt=0> a1 |
4593 | a1 <CURLYM cnt=1> a2 | |
4594 | a1 a2 <CURLYM cnt=2> a3 | |
4595 | a1 a2 a3 <CURLYM cnt=3> b | |
4596 | ) | |
2ab05381 | 4597 | |
c476f425 DM |
4598 | Each WHILEM state block marks a point to backtrack to upon partial failure |
4599 | of A or B, and also contains some minor state data related to that | |
4600 | iteration. The CURLYX block, pointed to by cur_curlyx, contains the | |
4601 | overall state, such as the count, and pointers to the A and B ops. | |
2ab05381 | 4602 | |
c476f425 DM |
4603 | This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx |
4604 | must always point to the *current* CURLYX block, the rules are: | |
2ab05381 | 4605 | |
c476f425 DM |
4606 | When executing CURLYX, save the old cur_curlyx in the CURLYX state block, |
4607 | and set cur_curlyx to point the new block. | |
2ab05381 | 4608 | |
c476f425 DM |
4609 | When popping the CURLYX block after a successful or unsuccessful match, |
4610 | restore the previous cur_curlyx. | |
2ab05381 | 4611 | |
c476f425 DM |
4612 | When WHILEM is about to execute B, save the current cur_curlyx, and set it |
4613 | to the outer one saved in the CURLYX block. | |
2ab05381 | 4614 | |
c476f425 DM |
4615 | When popping the WHILEM block after a successful or unsuccessful B match, |
4616 | restore the previous cur_curlyx. | |
2ab05381 | 4617 | |
c476f425 DM |
4618 | Here's an example for the pattern (AI* BI)*BO |
4619 | I and O refer to inner and outer, C and W refer to CURLYX and WHILEM: | |
2ab05381 | 4620 | |
c476f425 DM |
4621 | cur_ |
4622 | curlyx backtrack stack | |
4623 | ------ --------------- | |
4624 | NULL | |
4625 | CO <CO prev=NULL> <WO> | |
4626 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4627 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi | |
4628 | NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo | |
2ab05381 | 4629 | |
c476f425 DM |
4630 | At this point the pattern succeeds, and we work back down the stack to |
4631 | clean up, restoring as we go: | |
95b24440 | 4632 | |
c476f425 DM |
4633 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi |
4634 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4635 | CO <CO prev=NULL> <WO> | |
4636 | NULL | |
a0374537 | 4637 | |
c476f425 DM |
4638 | *******************************************************************/ |
4639 | ||
4640 | #define ST st->u.curlyx | |
4641 | ||
4642 | case CURLYX: /* start of /A*B/ (for complex A) */ | |
4643 | { | |
4644 | /* No need to save/restore up to this paren */ | |
4645 | I32 parenfloor = scan->flags; | |
4646 | ||
4647 | assert(next); /* keep Coverity happy */ | |
4648 | if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ | |
4649 | next += ARG(next); | |
4650 | ||
4651 | /* XXXX Probably it is better to teach regpush to support | |
4652 | parenfloor > PL_regsize... */ | |
4653 | if (parenfloor > (I32)*PL_reglastparen) | |
4654 | parenfloor = *PL_reglastparen; /* Pessimization... */ | |
4655 | ||
4656 | ST.prev_curlyx= cur_curlyx; | |
4657 | cur_curlyx = st; | |
4658 | ST.cp = PL_savestack_ix; | |
4659 | ||
4660 | /* these fields contain the state of the current curly. | |
4661 | * they are accessed by subsequent WHILEMs */ | |
4662 | ST.parenfloor = parenfloor; | |
d02d6d97 | 4663 | ST.me = scan; |
c476f425 | 4664 | ST.B = next; |
24d3c4a9 DM |
4665 | ST.minmod = minmod; |
4666 | minmod = 0; | |
c476f425 DM |
4667 | ST.count = -1; /* this will be updated by WHILEM */ |
4668 | ST.lastloc = NULL; /* this will be updated by WHILEM */ | |
4669 | ||
4670 | PL_reginput = locinput; | |
4671 | PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next)); | |
5f66b61c | 4672 | /* NOTREACHED */ |
c476f425 | 4673 | } |
a0d0e21e | 4674 | |
c476f425 | 4675 | case CURLYX_end: /* just finished matching all of A*B */ |
c476f425 DM |
4676 | cur_curlyx = ST.prev_curlyx; |
4677 | sayYES; | |
4678 | /* NOTREACHED */ | |
a0d0e21e | 4679 | |
c476f425 DM |
4680 | case CURLYX_end_fail: /* just failed to match all of A*B */ |
4681 | regcpblow(ST.cp); | |
4682 | cur_curlyx = ST.prev_curlyx; | |
4683 | sayNO; | |
4684 | /* NOTREACHED */ | |
4633a7c4 | 4685 | |
a0d0e21e | 4686 | |
c476f425 DM |
4687 | #undef ST |
4688 | #define ST st->u.whilem | |
4689 | ||
4690 | case WHILEM: /* just matched an A in /A*B/ (for complex A) */ | |
4691 | { | |
4692 | /* see the discussion above about CURLYX/WHILEM */ | |
c476f425 | 4693 | I32 n; |
d02d6d97 DM |
4694 | int min = ARG1(cur_curlyx->u.curlyx.me); |
4695 | int max = ARG2(cur_curlyx->u.curlyx.me); | |
4696 | regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS; | |
4697 | ||
c476f425 DM |
4698 | assert(cur_curlyx); /* keep Coverity happy */ |
4699 | n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */ | |
4700 | ST.save_lastloc = cur_curlyx->u.curlyx.lastloc; | |
4701 | ST.cache_offset = 0; | |
4702 | ST.cache_mask = 0; | |
4703 | ||
4704 | PL_reginput = locinput; | |
4705 | ||
4706 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
d02d6d97 DM |
4707 | "%*s whilem: matched %ld out of %d..%d\n", |
4708 | REPORT_CODE_OFF+depth*2, "", (long)n, min, max) | |
c476f425 | 4709 | ); |
a0d0e21e | 4710 | |
c476f425 | 4711 | /* First just match a string of min A's. */ |
a0d0e21e | 4712 | |
d02d6d97 | 4713 | if (n < min) { |
92e82afa | 4714 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); |
c476f425 | 4715 | cur_curlyx->u.curlyx.lastloc = locinput; |
92e82afa YO |
4716 | REGCP_SET(ST.lastcp); |
4717 | ||
d02d6d97 | 4718 | PUSH_STATE_GOTO(WHILEM_A_pre, A); |
c476f425 DM |
4719 | /* NOTREACHED */ |
4720 | } | |
4721 | ||
4722 | /* If degenerate A matches "", assume A done. */ | |
4723 | ||
4724 | if (locinput == cur_curlyx->u.curlyx.lastloc) { | |
4725 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4726 | "%*s whilem: empty match detected, trying continuation...\n", | |
4727 | REPORT_CODE_OFF+depth*2, "") | |
4728 | ); | |
4729 | goto do_whilem_B_max; | |
4730 | } | |
4731 | ||
4732 | /* super-linear cache processing */ | |
4733 | ||
4734 | if (scan->flags) { | |
a0d0e21e | 4735 | |
2c2d71f5 | 4736 | if (!PL_reg_maxiter) { |
c476f425 DM |
4737 | /* start the countdown: Postpone detection until we |
4738 | * know the match is not *that* much linear. */ | |
2c2d71f5 | 4739 | PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4); |
66bf836d DM |
4740 | /* possible overflow for long strings and many CURLYX's */ |
4741 | if (PL_reg_maxiter < 0) | |
4742 | PL_reg_maxiter = I32_MAX; | |
2c2d71f5 JH |
4743 | PL_reg_leftiter = PL_reg_maxiter; |
4744 | } | |
c476f425 | 4745 | |
2c2d71f5 | 4746 | if (PL_reg_leftiter-- == 0) { |
c476f425 | 4747 | /* initialise cache */ |
3298f257 | 4748 | const I32 size = (PL_reg_maxiter + 7)/8; |
2c2d71f5 | 4749 | if (PL_reg_poscache) { |
eb160463 | 4750 | if ((I32)PL_reg_poscache_size < size) { |
2c2d71f5 JH |
4751 | Renew(PL_reg_poscache, size, char); |
4752 | PL_reg_poscache_size = size; | |
4753 | } | |
4754 | Zero(PL_reg_poscache, size, char); | |
4755 | } | |
4756 | else { | |
4757 | PL_reg_poscache_size = size; | |
a02a5408 | 4758 | Newxz(PL_reg_poscache, size, char); |
2c2d71f5 | 4759 | } |
c476f425 DM |
4760 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
4761 | "%swhilem: Detected a super-linear match, switching on caching%s...\n", | |
4762 | PL_colors[4], PL_colors[5]) | |
4763 | ); | |
2c2d71f5 | 4764 | } |
c476f425 | 4765 | |
2c2d71f5 | 4766 | if (PL_reg_leftiter < 0) { |
c476f425 DM |
4767 | /* have we already failed at this position? */ |
4768 | I32 offset, mask; | |
4769 | offset = (scan->flags & 0xf) - 1 | |
4770 | + (locinput - PL_bostr) * (scan->flags>>4); | |
4771 | mask = 1 << (offset % 8); | |
4772 | offset /= 8; | |
4773 | if (PL_reg_poscache[offset] & mask) { | |
4774 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4775 | "%*s whilem: (cache) already tried at this position...\n", | |
4776 | REPORT_CODE_OFF+depth*2, "") | |
2c2d71f5 | 4777 | ); |
3298f257 | 4778 | sayNO; /* cache records failure */ |
2c2d71f5 | 4779 | } |
c476f425 DM |
4780 | ST.cache_offset = offset; |
4781 | ST.cache_mask = mask; | |
2c2d71f5 | 4782 | } |
c476f425 | 4783 | } |
2c2d71f5 | 4784 | |
c476f425 | 4785 | /* Prefer B over A for minimal matching. */ |
a687059c | 4786 | |
c476f425 DM |
4787 | if (cur_curlyx->u.curlyx.minmod) { |
4788 | ST.save_curlyx = cur_curlyx; | |
4789 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4790 | ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor); | |
4791 | REGCP_SET(ST.lastcp); | |
4792 | PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B); | |
4793 | /* NOTREACHED */ | |
4794 | } | |
a0d0e21e | 4795 | |
c476f425 DM |
4796 | /* Prefer A over B for maximal matching. */ |
4797 | ||
d02d6d97 | 4798 | if (n < max) { /* More greed allowed? */ |
c476f425 DM |
4799 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); |
4800 | cur_curlyx->u.curlyx.lastloc = locinput; | |
4801 | REGCP_SET(ST.lastcp); | |
d02d6d97 | 4802 | PUSH_STATE_GOTO(WHILEM_A_max, A); |
c476f425 DM |
4803 | /* NOTREACHED */ |
4804 | } | |
4805 | goto do_whilem_B_max; | |
4806 | } | |
4807 | /* NOTREACHED */ | |
4808 | ||
4809 | case WHILEM_B_min: /* just matched B in a minimal match */ | |
4810 | case WHILEM_B_max: /* just matched B in a maximal match */ | |
4811 | cur_curlyx = ST.save_curlyx; | |
4812 | sayYES; | |
4813 | /* NOTREACHED */ | |
4814 | ||
4815 | case WHILEM_B_max_fail: /* just failed to match B in a maximal match */ | |
4816 | cur_curlyx = ST.save_curlyx; | |
4817 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
4818 | cur_curlyx->u.curlyx.count--; | |
4819 | CACHEsayNO; | |
4820 | /* NOTREACHED */ | |
4821 | ||
4822 | case WHILEM_A_min_fail: /* just failed to match A in a minimal match */ | |
c476f425 DM |
4823 | /* FALL THROUGH */ |
4824 | case WHILEM_A_pre_fail: /* just failed to match even minimal A */ | |
92e82afa YO |
4825 | REGCP_UNWIND(ST.lastcp); |
4826 | regcppop(rex); | |
c476f425 DM |
4827 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; |
4828 | cur_curlyx->u.curlyx.count--; | |
4829 | CACHEsayNO; | |
4830 | /* NOTREACHED */ | |
4831 | ||
4832 | case WHILEM_A_max_fail: /* just failed to match A in a maximal match */ | |
4833 | REGCP_UNWIND(ST.lastcp); | |
4834 | regcppop(rex); /* Restore some previous $<digit>s? */ | |
4835 | PL_reginput = locinput; | |
4836 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
4837 | "%*s whilem: failed, trying continuation...\n", | |
4838 | REPORT_CODE_OFF+depth*2, "") | |
4839 | ); | |
4840 | do_whilem_B_max: | |
4841 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
4842 | && ckWARN(WARN_REGEXP) | |
4843 | && !(PL_reg_flags & RF_warned)) | |
4844 | { | |
4845 | PL_reg_flags |= RF_warned; | |
4846 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded", | |
4847 | "Complex regular subexpression recursion", | |
4848 | REG_INFTY - 1); | |
4849 | } | |
4850 | ||
4851 | /* now try B */ | |
4852 | ST.save_curlyx = cur_curlyx; | |
4853 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4854 | PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B); | |
4855 | /* NOTREACHED */ | |
4856 | ||
4857 | case WHILEM_B_min_fail: /* just failed to match B in a minimal match */ | |
4858 | cur_curlyx = ST.save_curlyx; | |
4859 | REGCP_UNWIND(ST.lastcp); | |
4860 | regcppop(rex); | |
4861 | ||
d02d6d97 | 4862 | if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) { |
c476f425 DM |
4863 | /* Maximum greed exceeded */ |
4864 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
4865 | && ckWARN(WARN_REGEXP) | |
4866 | && !(PL_reg_flags & RF_warned)) | |
4867 | { | |
3280af22 | 4868 | PL_reg_flags |= RF_warned; |
c476f425 DM |
4869 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
4870 | "%s limit (%d) exceeded", | |
4871 | "Complex regular subexpression recursion", | |
4872 | REG_INFTY - 1); | |
a0d0e21e | 4873 | } |
c476f425 | 4874 | cur_curlyx->u.curlyx.count--; |
3ab3c9b4 | 4875 | CACHEsayNO; |
a0d0e21e | 4876 | } |
c476f425 DM |
4877 | |
4878 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
4879 | "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "") | |
4880 | ); | |
4881 | /* Try grabbing another A and see if it helps. */ | |
4882 | PL_reginput = locinput; | |
4883 | cur_curlyx->u.curlyx.lastloc = locinput; | |
4884 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); | |
4885 | REGCP_SET(ST.lastcp); | |
d02d6d97 DM |
4886 | PUSH_STATE_GOTO(WHILEM_A_min, |
4887 | /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS); | |
5f66b61c | 4888 | /* NOTREACHED */ |
40a82448 DM |
4889 | |
4890 | #undef ST | |
4891 | #define ST st->u.branch | |
4892 | ||
4893 | case BRANCHJ: /* /(...|A|...)/ with long next pointer */ | |
c277df42 IZ |
4894 | next = scan + ARG(scan); |
4895 | if (next == scan) | |
4896 | next = NULL; | |
40a82448 DM |
4897 | scan = NEXTOPER(scan); |
4898 | /* FALL THROUGH */ | |
c277df42 | 4899 | |
40a82448 DM |
4900 | case BRANCH: /* /(...|A|...)/ */ |
4901 | scan = NEXTOPER(scan); /* scan now points to inner node */ | |
40a82448 DM |
4902 | ST.lastparen = *PL_reglastparen; |
4903 | ST.next_branch = next; | |
4904 | REGCP_SET(ST.cp); | |
4905 | PL_reginput = locinput; | |
02db2b7b | 4906 | |
40a82448 | 4907 | /* Now go into the branch */ |
5d458dd8 YO |
4908 | if (has_cutgroup) { |
4909 | PUSH_YES_STATE_GOTO(BRANCH_next, scan); | |
4910 | } else { | |
4911 | PUSH_STATE_GOTO(BRANCH_next, scan); | |
4912 | } | |
40a82448 | 4913 | /* NOTREACHED */ |
5d458dd8 YO |
4914 | case CUTGROUP: |
4915 | PL_reginput = locinput; | |
4916 | sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : | |
ad64d0ec | 4917 | MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
4918 | PUSH_STATE_GOTO(CUTGROUP_next,next); |
4919 | /* NOTREACHED */ | |
4920 | case CUTGROUP_next_fail: | |
4921 | do_cutgroup = 1; | |
4922 | no_final = 1; | |
4923 | if (st->u.mark.mark_name) | |
4924 | sv_commit = st->u.mark.mark_name; | |
4925 | sayNO; | |
4926 | /* NOTREACHED */ | |
4927 | case BRANCH_next: | |
4928 | sayYES; | |
4929 | /* NOTREACHED */ | |
40a82448 | 4930 | case BRANCH_next_fail: /* that branch failed; try the next, if any */ |
5d458dd8 YO |
4931 | if (do_cutgroup) { |
4932 | do_cutgroup = 0; | |
4933 | no_final = 0; | |
4934 | } | |
40a82448 DM |
4935 | REGCP_UNWIND(ST.cp); |
4936 | for (n = *PL_reglastparen; n > ST.lastparen; n--) | |
f0ab9afb | 4937 | PL_regoffs[n].end = -1; |
40a82448 | 4938 | *PL_reglastparen = n; |
0a4db386 | 4939 | /*dmq: *PL_reglastcloseparen = n; */ |
40a82448 DM |
4940 | scan = ST.next_branch; |
4941 | /* no more branches? */ | |
5d458dd8 YO |
4942 | if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) { |
4943 | DEBUG_EXECUTE_r({ | |
4944 | PerlIO_printf( Perl_debug_log, | |
4945 | "%*s %sBRANCH failed...%s\n", | |
4946 | REPORT_CODE_OFF+depth*2, "", | |
4947 | PL_colors[4], | |
4948 | PL_colors[5] ); | |
4949 | }); | |
4950 | sayNO_SILENT; | |
4951 | } | |
40a82448 DM |
4952 | continue; /* execute next BRANCH[J] op */ |
4953 | /* NOTREACHED */ | |
4954 | ||
a0d0e21e | 4955 | case MINMOD: |
24d3c4a9 | 4956 | minmod = 1; |
a0d0e21e | 4957 | break; |
40a82448 DM |
4958 | |
4959 | #undef ST | |
4960 | #define ST st->u.curlym | |
4961 | ||
4962 | case CURLYM: /* /A{m,n}B/ where A is fixed-length */ | |
4963 | ||
4964 | /* This is an optimisation of CURLYX that enables us to push | |
84d2fa14 | 4965 | * only a single backtracking state, no matter how many matches |
40a82448 DM |
4966 | * there are in {m,n}. It relies on the pattern being constant |
4967 | * length, with no parens to influence future backrefs | |
4968 | */ | |
4969 | ||
4970 | ST.me = scan; | |
dc45a647 | 4971 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; |
40a82448 DM |
4972 | |
4973 | /* if paren positive, emulate an OPEN/CLOSE around A */ | |
4974 | if (ST.me->flags) { | |
3b6647e0 | 4975 | U32 paren = ST.me->flags; |
40a82448 DM |
4976 | if (paren > PL_regsize) |
4977 | PL_regsize = paren; | |
3b6647e0 | 4978 | if (paren > *PL_reglastparen) |
40a82448 | 4979 | *PL_reglastparen = paren; |
c277df42 | 4980 | scan += NEXT_OFF(scan); /* Skip former OPEN. */ |
6407bf3b | 4981 | } |
40a82448 DM |
4982 | ST.A = scan; |
4983 | ST.B = next; | |
4984 | ST.alen = 0; | |
4985 | ST.count = 0; | |
24d3c4a9 DM |
4986 | ST.minmod = minmod; |
4987 | minmod = 0; | |
40a82448 DM |
4988 | ST.c1 = CHRTEST_UNINIT; |
4989 | REGCP_SET(ST.cp); | |
6407bf3b | 4990 | |
40a82448 DM |
4991 | if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */ |
4992 | goto curlym_do_B; | |
4993 | ||
4994 | curlym_do_A: /* execute the A in /A{m,n}B/ */ | |
6407bf3b | 4995 | PL_reginput = locinput; |
40a82448 DM |
4996 | PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */ |
4997 | /* NOTREACHED */ | |
5f80c4cf | 4998 | |
40a82448 DM |
4999 | case CURLYM_A: /* we've just matched an A */ |
5000 | locinput = st->locinput; | |
5001 | nextchr = UCHARAT(locinput); | |
5002 | ||
5003 | ST.count++; | |
5004 | /* after first match, determine A's length: u.curlym.alen */ | |
5005 | if (ST.count == 1) { | |
5006 | if (PL_reg_match_utf8) { | |
5007 | char *s = locinput; | |
5008 | while (s < PL_reginput) { | |
5009 | ST.alen++; | |
5010 | s += UTF8SKIP(s); | |
5011 | } | |
5012 | } | |
5013 | else { | |
5014 | ST.alen = PL_reginput - locinput; | |
5015 | } | |
5016 | if (ST.alen == 0) | |
5017 | ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me); | |
5018 | } | |
0cadcf80 DM |
5019 | DEBUG_EXECUTE_r( |
5020 | PerlIO_printf(Perl_debug_log, | |
40a82448 | 5021 | "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n", |
5bc10b2c | 5022 | (int)(REPORT_CODE_OFF+(depth*2)), "", |
40a82448 | 5023 | (IV) ST.count, (IV)ST.alen) |
0cadcf80 DM |
5024 | ); |
5025 | ||
40a82448 | 5026 | locinput = PL_reginput; |
0a4db386 YO |
5027 | |
5028 | if (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 5029 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5030 | goto fake_end; |
5031 | ||
c966426a DM |
5032 | { |
5033 | I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)); | |
5034 | if ( max == REG_INFTY || ST.count < max ) | |
5035 | goto curlym_do_A; /* try to match another A */ | |
5036 | } | |
40a82448 | 5037 | goto curlym_do_B; /* try to match B */ |
5f80c4cf | 5038 | |
40a82448 DM |
5039 | case CURLYM_A_fail: /* just failed to match an A */ |
5040 | REGCP_UNWIND(ST.cp); | |
0a4db386 YO |
5041 | |
5042 | if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ | |
5043 | || (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 5044 | cur_eval->u.eval.close_paren == (U32)ST.me->flags)) |
40a82448 | 5045 | sayNO; |
0cadcf80 | 5046 | |
40a82448 DM |
5047 | curlym_do_B: /* execute the B in /A{m,n}B/ */ |
5048 | PL_reginput = locinput; | |
5049 | if (ST.c1 == CHRTEST_UNINIT) { | |
5050 | /* calculate c1 and c2 for possible match of 1st char | |
5051 | * following curly */ | |
5052 | ST.c1 = ST.c2 = CHRTEST_VOID; | |
5053 | if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) { | |
5054 | regnode *text_node = ST.B; | |
5055 | if (! HAS_TEXT(text_node)) | |
5056 | FIND_NEXT_IMPT(text_node); | |
ee9b8eae YO |
5057 | /* this used to be |
5058 | ||
5059 | (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT) | |
5060 | ||
5061 | But the former is redundant in light of the latter. | |
5062 | ||
5063 | if this changes back then the macro for | |
5064 | IS_TEXT and friends need to change. | |
5065 | */ | |
5066 | if (PL_regkind[OP(text_node)] == EXACT) | |
40a82448 | 5067 | { |
ee9b8eae | 5068 | |
40a82448 | 5069 | ST.c1 = (U8)*STRING(text_node); |
9a5a5549 KW |
5070 | switch (OP(text_node)) { |
5071 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
2f7f8cb1 | 5072 | case EXACTFA: |
9a5a5549 KW |
5073 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; |
5074 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
5075 | default: ST.c2 = ST.c1; | |
5076 | } | |
c277df42 | 5077 | } |
c277df42 | 5078 | } |
40a82448 DM |
5079 | } |
5080 | ||
5081 | DEBUG_EXECUTE_r( | |
5082 | PerlIO_printf(Perl_debug_log, | |
5083 | "%*s CURLYM trying tail with matches=%"IVdf"...\n", | |
5bc10b2c | 5084 | (int)(REPORT_CODE_OFF+(depth*2)), |
40a82448 DM |
5085 | "", (IV)ST.count) |
5086 | ); | |
5087 | if (ST.c1 != CHRTEST_VOID | |
5088 | && UCHARAT(PL_reginput) != ST.c1 | |
5089 | && UCHARAT(PL_reginput) != ST.c2) | |
5090 | { | |
5091 | /* simulate B failing */ | |
3e901dc0 YO |
5092 | DEBUG_OPTIMISE_r( |
5093 | PerlIO_printf(Perl_debug_log, | |
5094 | "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n", | |
5095 | (int)(REPORT_CODE_OFF+(depth*2)),"", | |
5096 | (IV)ST.c1,(IV)ST.c2 | |
5097 | )); | |
40a82448 DM |
5098 | state_num = CURLYM_B_fail; |
5099 | goto reenter_switch; | |
5100 | } | |
5101 | ||
5102 | if (ST.me->flags) { | |
5103 | /* mark current A as captured */ | |
5104 | I32 paren = ST.me->flags; | |
5105 | if (ST.count) { | |
f0ab9afb | 5106 | PL_regoffs[paren].start |
40a82448 | 5107 | = HOPc(PL_reginput, -ST.alen) - PL_bostr; |
f0ab9afb | 5108 | PL_regoffs[paren].end = PL_reginput - PL_bostr; |
0a4db386 | 5109 | /*dmq: *PL_reglastcloseparen = paren; */ |
c277df42 | 5110 | } |
40a82448 | 5111 | else |
f0ab9afb | 5112 | PL_regoffs[paren].end = -1; |
0a4db386 | 5113 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 5114 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
5115 | { |
5116 | if (ST.count) | |
5117 | goto fake_end; | |
5118 | else | |
5119 | sayNO; | |
5120 | } | |
c277df42 | 5121 | } |
0a4db386 | 5122 | |
40a82448 | 5123 | PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */ |
5f66b61c | 5124 | /* NOTREACHED */ |
40a82448 DM |
5125 | |
5126 | case CURLYM_B_fail: /* just failed to match a B */ | |
5127 | REGCP_UNWIND(ST.cp); | |
5128 | if (ST.minmod) { | |
84d2fa14 HS |
5129 | I32 max = ARG2(ST.me); |
5130 | if (max != REG_INFTY && ST.count == max) | |
40a82448 DM |
5131 | sayNO; |
5132 | goto curlym_do_A; /* try to match a further A */ | |
5133 | } | |
5134 | /* backtrack one A */ | |
5135 | if (ST.count == ARG1(ST.me) /* min */) | |
5136 | sayNO; | |
5137 | ST.count--; | |
5138 | locinput = HOPc(locinput, -ST.alen); | |
5139 | goto curlym_do_B; /* try to match B */ | |
5140 | ||
c255a977 DM |
5141 | #undef ST |
5142 | #define ST st->u.curly | |
40a82448 | 5143 | |
c255a977 DM |
5144 | #define CURLY_SETPAREN(paren, success) \ |
5145 | if (paren) { \ | |
5146 | if (success) { \ | |
f0ab9afb NC |
5147 | PL_regoffs[paren].start = HOPc(locinput, -1) - PL_bostr; \ |
5148 | PL_regoffs[paren].end = locinput - PL_bostr; \ | |
0a4db386 | 5149 | *PL_reglastcloseparen = paren; \ |
c255a977 DM |
5150 | } \ |
5151 | else \ | |
f0ab9afb | 5152 | PL_regoffs[paren].end = -1; \ |
c255a977 DM |
5153 | } |
5154 | ||
5155 | case STAR: /* /A*B/ where A is width 1 */ | |
5156 | ST.paren = 0; | |
5157 | ST.min = 0; | |
5158 | ST.max = REG_INFTY; | |
a0d0e21e LW |
5159 | scan = NEXTOPER(scan); |
5160 | goto repeat; | |
c255a977 DM |
5161 | case PLUS: /* /A+B/ where A is width 1 */ |
5162 | ST.paren = 0; | |
5163 | ST.min = 1; | |
5164 | ST.max = REG_INFTY; | |
c277df42 | 5165 | scan = NEXTOPER(scan); |
c255a977 DM |
5166 | goto repeat; |
5167 | case CURLYN: /* /(A){m,n}B/ where A is width 1 */ | |
5168 | ST.paren = scan->flags; /* Which paren to set */ | |
5169 | if (ST.paren > PL_regsize) | |
5170 | PL_regsize = ST.paren; | |
3b6647e0 | 5171 | if (ST.paren > *PL_reglastparen) |
c255a977 DM |
5172 | *PL_reglastparen = ST.paren; |
5173 | ST.min = ARG1(scan); /* min to match */ | |
5174 | ST.max = ARG2(scan); /* max to match */ | |
0a4db386 | 5175 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5176 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5177 | ST.min=1; |
5178 | ST.max=1; | |
5179 | } | |
c255a977 DM |
5180 | scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); |
5181 | goto repeat; | |
5182 | case CURLY: /* /A{m,n}B/ where A is width 1 */ | |
5183 | ST.paren = 0; | |
5184 | ST.min = ARG1(scan); /* min to match */ | |
5185 | ST.max = ARG2(scan); /* max to match */ | |
5186 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; | |
c277df42 | 5187 | repeat: |
a0d0e21e LW |
5188 | /* |
5189 | * Lookahead to avoid useless match attempts | |
5190 | * when we know what character comes next. | |
c255a977 | 5191 | * |
5f80c4cf JP |
5192 | * Used to only do .*x and .*?x, but now it allows |
5193 | * for )'s, ('s and (?{ ... })'s to be in the way | |
5194 | * of the quantifier and the EXACT-like node. -- japhy | |
5195 | */ | |
5196 | ||
c255a977 DM |
5197 | if (ST.min > ST.max) /* XXX make this a compile-time check? */ |
5198 | sayNO; | |
cca55fe3 | 5199 | if (HAS_TEXT(next) || JUMPABLE(next)) { |
5f80c4cf JP |
5200 | U8 *s; |
5201 | regnode *text_node = next; | |
5202 | ||
3dab1dad YO |
5203 | if (! HAS_TEXT(text_node)) |
5204 | FIND_NEXT_IMPT(text_node); | |
5f80c4cf | 5205 | |
9e137952 | 5206 | if (! HAS_TEXT(text_node)) |
c255a977 | 5207 | ST.c1 = ST.c2 = CHRTEST_VOID; |
5f80c4cf | 5208 | else { |
ee9b8eae | 5209 | if ( PL_regkind[OP(text_node)] != EXACT ) { |
c255a977 | 5210 | ST.c1 = ST.c2 = CHRTEST_VOID; |
44a68960 | 5211 | goto assume_ok_easy; |
cca55fe3 | 5212 | } |
be8e71aa YO |
5213 | else |
5214 | s = (U8*)STRING(text_node); | |
ee9b8eae YO |
5215 | |
5216 | /* Currently we only get here when | |
5217 | ||
5218 | PL_rekind[OP(text_node)] == EXACT | |
5219 | ||
5220 | if this changes back then the macro for IS_TEXT and | |
5221 | friends need to change. */ | |
f2ed9b32 | 5222 | if (!UTF_PATTERN) { |
9a5a5549 KW |
5223 | ST.c1 = *s; |
5224 | switch (OP(text_node)) { | |
5225 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
2f7f8cb1 | 5226 | case EXACTFA: |
9a5a5549 KW |
5227 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; |
5228 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
5229 | default: ST.c2 = ST.c1; break; | |
5230 | } | |
1aa99e6b | 5231 | } |
f2ed9b32 | 5232 | else { /* UTF_PATTERN */ |
9a5a5549 | 5233 | if (IS_TEXTFU(text_node) || IS_TEXTF(text_node)) { |
a2a2844f | 5234 | STRLEN ulen1, ulen2; |
89ebb4a3 JH |
5235 | U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; |
5236 | U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; | |
a2a2844f JH |
5237 | |
5238 | to_utf8_lower((U8*)s, tmpbuf1, &ulen1); | |
5239 | to_utf8_upper((U8*)s, tmpbuf2, &ulen2); | |
e294cc5d JH |
5240 | #ifdef EBCDIC |
5241 | ST.c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN, 0, | |
5242 | ckWARN(WARN_UTF8) ? | |
5243 | 0 : UTF8_ALLOW_ANY); | |
5244 | ST.c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN, 0, | |
5245 | ckWARN(WARN_UTF8) ? | |
5246 | 0 : UTF8_ALLOW_ANY); | |
5247 | #else | |
c255a977 | 5248 | ST.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0, |
e294cc5d | 5249 | uniflags); |
c255a977 | 5250 | ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0, |
e294cc5d JH |
5251 | uniflags); |
5252 | #endif | |
5f80c4cf JP |
5253 | } |
5254 | else { | |
c255a977 | 5255 | ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, |
041457d9 | 5256 | uniflags); |
5f80c4cf | 5257 | } |
1aa99e6b IH |
5258 | } |
5259 | } | |
bbce6d69 | 5260 | } |
a0d0e21e | 5261 | else |
c255a977 | 5262 | ST.c1 = ST.c2 = CHRTEST_VOID; |
cca55fe3 | 5263 | assume_ok_easy: |
c255a977 DM |
5264 | |
5265 | ST.A = scan; | |
5266 | ST.B = next; | |
3280af22 | 5267 | PL_reginput = locinput; |
24d3c4a9 DM |
5268 | if (minmod) { |
5269 | minmod = 0; | |
e2e6a0f1 | 5270 | if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min) |
4633a7c4 | 5271 | sayNO; |
c255a977 | 5272 | ST.count = ST.min; |
a0ed51b3 | 5273 | locinput = PL_reginput; |
c255a977 DM |
5274 | REGCP_SET(ST.cp); |
5275 | if (ST.c1 == CHRTEST_VOID) | |
5276 | goto curly_try_B_min; | |
5277 | ||
5278 | ST.oldloc = locinput; | |
5279 | ||
5280 | /* set ST.maxpos to the furthest point along the | |
5281 | * string that could possibly match */ | |
5282 | if (ST.max == REG_INFTY) { | |
5283 | ST.maxpos = PL_regeol - 1; | |
f2ed9b32 | 5284 | if (utf8_target) |
c255a977 DM |
5285 | while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos)) |
5286 | ST.maxpos--; | |
5287 | } | |
f2ed9b32 | 5288 | else if (utf8_target) { |
c255a977 DM |
5289 | int m = ST.max - ST.min; |
5290 | for (ST.maxpos = locinput; | |
5291 | m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--) | |
5292 | ST.maxpos += UTF8SKIP(ST.maxpos); | |
5293 | } | |
5294 | else { | |
5295 | ST.maxpos = locinput + ST.max - ST.min; | |
5296 | if (ST.maxpos >= PL_regeol) | |
5297 | ST.maxpos = PL_regeol - 1; | |
5298 | } | |
5299 | goto curly_try_B_min_known; | |
5300 | ||
5301 | } | |
5302 | else { | |
e2e6a0f1 | 5303 | ST.count = regrepeat(rex, ST.A, ST.max, depth); |
c255a977 DM |
5304 | locinput = PL_reginput; |
5305 | if (ST.count < ST.min) | |
5306 | sayNO; | |
5307 | if ((ST.count > ST.min) | |
5308 | && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL)) | |
5309 | { | |
5310 | /* A{m,n} must come at the end of the string, there's | |
5311 | * no point in backing off ... */ | |
5312 | ST.min = ST.count; | |
5313 | /* ...except that $ and \Z can match before *and* after | |
5314 | newline at the end. Consider "\n\n" =~ /\n+\Z\n/. | |
5315 | We may back off by one in this case. */ | |
5316 | if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS) | |
5317 | ST.min--; | |
5318 | } | |
5319 | REGCP_SET(ST.cp); | |
5320 | goto curly_try_B_max; | |
5321 | } | |
5322 | /* NOTREACHED */ | |
5323 | ||
5324 | ||
5325 | case CURLY_B_min_known_fail: | |
5326 | /* failed to find B in a non-greedy match where c1,c2 valid */ | |
5327 | if (ST.paren && ST.count) | |
f0ab9afb | 5328 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5329 | |
5330 | PL_reginput = locinput; /* Could be reset... */ | |
5331 | REGCP_UNWIND(ST.cp); | |
5332 | /* Couldn't or didn't -- move forward. */ | |
5333 | ST.oldloc = locinput; | |
f2ed9b32 | 5334 | if (utf8_target) |
c255a977 DM |
5335 | locinput += UTF8SKIP(locinput); |
5336 | else | |
5337 | locinput++; | |
5338 | ST.count++; | |
5339 | curly_try_B_min_known: | |
5340 | /* find the next place where 'B' could work, then call B */ | |
5341 | { | |
5342 | int n; | |
f2ed9b32 | 5343 | if (utf8_target) { |
c255a977 DM |
5344 | n = (ST.oldloc == locinput) ? 0 : 1; |
5345 | if (ST.c1 == ST.c2) { | |
5346 | STRLEN len; | |
5347 | /* set n to utf8_distance(oldloc, locinput) */ | |
5348 | while (locinput <= ST.maxpos && | |
5349 | utf8n_to_uvchr((U8*)locinput, | |
5350 | UTF8_MAXBYTES, &len, | |
5351 | uniflags) != (UV)ST.c1) { | |
5352 | locinput += len; | |
5353 | n++; | |
5354 | } | |
1aa99e6b IH |
5355 | } |
5356 | else { | |
c255a977 DM |
5357 | /* set n to utf8_distance(oldloc, locinput) */ |
5358 | while (locinput <= ST.maxpos) { | |
5359 | STRLEN len; | |
5360 | const UV c = utf8n_to_uvchr((U8*)locinput, | |
5361 | UTF8_MAXBYTES, &len, | |
5362 | uniflags); | |
5363 | if (c == (UV)ST.c1 || c == (UV)ST.c2) | |
5364 | break; | |
5365 | locinput += len; | |
5366 | n++; | |
1aa99e6b | 5367 | } |
0fe9bf95 IZ |
5368 | } |
5369 | } | |
c255a977 DM |
5370 | else { |
5371 | if (ST.c1 == ST.c2) { | |
5372 | while (locinput <= ST.maxpos && | |
5373 | UCHARAT(locinput) != ST.c1) | |
5374 | locinput++; | |
bbce6d69 | 5375 | } |
c255a977 DM |
5376 | else { |
5377 | while (locinput <= ST.maxpos | |
5378 | && UCHARAT(locinput) != ST.c1 | |
5379 | && UCHARAT(locinput) != ST.c2) | |
5380 | locinput++; | |
a0ed51b3 | 5381 | } |
c255a977 DM |
5382 | n = locinput - ST.oldloc; |
5383 | } | |
5384 | if (locinput > ST.maxpos) | |
5385 | sayNO; | |
5386 | /* PL_reginput == oldloc now */ | |
5387 | if (n) { | |
5388 | ST.count += n; | |
e2e6a0f1 | 5389 | if (regrepeat(rex, ST.A, n, depth) < n) |
4633a7c4 | 5390 | sayNO; |
a0d0e21e | 5391 | } |
c255a977 DM |
5392 | PL_reginput = locinput; |
5393 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5394 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5395 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5396 | goto fake_end; |
5397 | } | |
c255a977 | 5398 | PUSH_STATE_GOTO(CURLY_B_min_known, ST.B); |
a0d0e21e | 5399 | } |
c255a977 DM |
5400 | /* NOTREACHED */ |
5401 | ||
5402 | ||
5403 | case CURLY_B_min_fail: | |
5404 | /* failed to find B in a non-greedy match where c1,c2 invalid */ | |
5405 | if (ST.paren && ST.count) | |
f0ab9afb | 5406 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5407 | |
5408 | REGCP_UNWIND(ST.cp); | |
5409 | /* failed -- move forward one */ | |
5410 | PL_reginput = locinput; | |
e2e6a0f1 | 5411 | if (regrepeat(rex, ST.A, 1, depth)) { |
c255a977 | 5412 | ST.count++; |
a0ed51b3 | 5413 | locinput = PL_reginput; |
c255a977 DM |
5414 | if (ST.count <= ST.max || (ST.max == REG_INFTY && |
5415 | ST.count > 0)) /* count overflow ? */ | |
15272685 | 5416 | { |
c255a977 DM |
5417 | curly_try_B_min: |
5418 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5419 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5420 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5421 | goto fake_end; |
5422 | } | |
c255a977 | 5423 | PUSH_STATE_GOTO(CURLY_B_min, ST.B); |
a0d0e21e LW |
5424 | } |
5425 | } | |
4633a7c4 | 5426 | sayNO; |
c255a977 DM |
5427 | /* NOTREACHED */ |
5428 | ||
5429 | ||
5430 | curly_try_B_max: | |
5431 | /* a successful greedy match: now try to match B */ | |
40d049e4 | 5432 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5433 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
40d049e4 YO |
5434 | goto fake_end; |
5435 | } | |
c255a977 DM |
5436 | { |
5437 | UV c = 0; | |
5438 | if (ST.c1 != CHRTEST_VOID) | |
f2ed9b32 | 5439 | c = utf8_target ? utf8n_to_uvchr((U8*)PL_reginput, |
c255a977 | 5440 | UTF8_MAXBYTES, 0, uniflags) |
466787eb | 5441 | : (UV) UCHARAT(PL_reginput); |
c255a977 DM |
5442 | /* If it could work, try it. */ |
5443 | if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) { | |
5444 | CURLY_SETPAREN(ST.paren, ST.count); | |
5445 | PUSH_STATE_GOTO(CURLY_B_max, ST.B); | |
5446 | /* NOTREACHED */ | |
5447 | } | |
5448 | } | |
5449 | /* FALL THROUGH */ | |
5450 | case CURLY_B_max_fail: | |
5451 | /* failed to find B in a greedy match */ | |
5452 | if (ST.paren && ST.count) | |
f0ab9afb | 5453 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5454 | |
5455 | REGCP_UNWIND(ST.cp); | |
5456 | /* back up. */ | |
5457 | if (--ST.count < ST.min) | |
5458 | sayNO; | |
5459 | PL_reginput = locinput = HOPc(locinput, -1); | |
5460 | goto curly_try_B_max; | |
5461 | ||
5462 | #undef ST | |
5463 | ||
a0d0e21e | 5464 | case END: |
6bda09f9 | 5465 | fake_end: |
faec1544 DM |
5466 | if (cur_eval) { |
5467 | /* we've just finished A in /(??{A})B/; now continue with B */ | |
5468 | I32 tmpix; | |
faec1544 DM |
5469 | st->u.eval.toggle_reg_flags |
5470 | = cur_eval->u.eval.toggle_reg_flags; | |
5471 | PL_reg_flags ^= st->u.eval.toggle_reg_flags; | |
5472 | ||
288b8c02 NC |
5473 | st->u.eval.prev_rex = rex_sv; /* inner */ |
5474 | SETREX(rex_sv,cur_eval->u.eval.prev_rex); | |
5475 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 5476 | rexi = RXi_GET(rex); |
faec1544 | 5477 | cur_curlyx = cur_eval->u.eval.prev_curlyx; |
cd733337 | 5478 | (void)ReREFCNT_inc(rex_sv); |
faec1544 | 5479 | st->u.eval.cp = regcppush(0); /* Save *all* the positions. */ |
34a81e2b B |
5480 | |
5481 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ | |
5482 | PL_reglastparen = &rex->lastparen; | |
5483 | PL_reglastcloseparen = &rex->lastcloseparen; | |
5484 | ||
faec1544 DM |
5485 | REGCP_SET(st->u.eval.lastcp); |
5486 | PL_reginput = locinput; | |
5487 | ||
5488 | /* Restore parens of the outer rex without popping the | |
5489 | * savestack */ | |
5490 | tmpix = PL_savestack_ix; | |
5491 | PL_savestack_ix = cur_eval->u.eval.lastcp; | |
5492 | regcppop(rex); | |
5493 | PL_savestack_ix = tmpix; | |
5494 | ||
5495 | st->u.eval.prev_eval = cur_eval; | |
5496 | cur_eval = cur_eval->u.eval.prev_eval; | |
5497 | DEBUG_EXECUTE_r( | |
2a49f0f5 JH |
5498 | PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n", |
5499 | REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval));); | |
e7707071 YO |
5500 | if ( nochange_depth ) |
5501 | nochange_depth--; | |
5502 | ||
5503 | PUSH_YES_STATE_GOTO(EVAL_AB, | |
faec1544 DM |
5504 | st->u.eval.prev_eval->u.eval.B); /* match B */ |
5505 | } | |
5506 | ||
3b0527fe | 5507 | if (locinput < reginfo->till) { |
a3621e74 | 5508 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
7821416a IZ |
5509 | "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n", |
5510 | PL_colors[4], | |
5511 | (long)(locinput - PL_reg_starttry), | |
3b0527fe | 5512 | (long)(reginfo->till - PL_reg_starttry), |
7821416a | 5513 | PL_colors[5])); |
58e23c8d | 5514 | |
262b90c4 | 5515 | sayNO_SILENT; /* Cannot match: too short. */ |
7821416a IZ |
5516 | } |
5517 | PL_reginput = locinput; /* put where regtry can find it */ | |
262b90c4 | 5518 | sayYES; /* Success! */ |
dad79028 DM |
5519 | |
5520 | case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */ | |
5521 | DEBUG_EXECUTE_r( | |
5522 | PerlIO_printf(Perl_debug_log, | |
5523 | "%*s %ssubpattern success...%s\n", | |
5bc10b2c | 5524 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])); |
3280af22 | 5525 | PL_reginput = locinput; /* put where regtry can find it */ |
262b90c4 | 5526 | sayYES; /* Success! */ |
dad79028 | 5527 | |
40a82448 DM |
5528 | #undef ST |
5529 | #define ST st->u.ifmatch | |
5530 | ||
5531 | case SUSPEND: /* (?>A) */ | |
5532 | ST.wanted = 1; | |
9fe1d20c | 5533 | PL_reginput = locinput; |
9041c2e3 | 5534 | goto do_ifmatch; |
dad79028 | 5535 | |
40a82448 DM |
5536 | case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */ |
5537 | ST.wanted = 0; | |
dad79028 DM |
5538 | goto ifmatch_trivial_fail_test; |
5539 | ||
40a82448 DM |
5540 | case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */ |
5541 | ST.wanted = 1; | |
dad79028 | 5542 | ifmatch_trivial_fail_test: |
a0ed51b3 | 5543 | if (scan->flags) { |
52657f30 | 5544 | char * const s = HOPBACKc(locinput, scan->flags); |
dad79028 DM |
5545 | if (!s) { |
5546 | /* trivial fail */ | |
24d3c4a9 DM |
5547 | if (logical) { |
5548 | logical = 0; | |
f2338a2e | 5549 | sw = 1 - cBOOL(ST.wanted); |
dad79028 | 5550 | } |
40a82448 | 5551 | else if (ST.wanted) |
dad79028 DM |
5552 | sayNO; |
5553 | next = scan + ARG(scan); | |
5554 | if (next == scan) | |
5555 | next = NULL; | |
5556 | break; | |
5557 | } | |
efb30f32 | 5558 | PL_reginput = s; |
a0ed51b3 LW |
5559 | } |
5560 | else | |
5561 | PL_reginput = locinput; | |
5562 | ||
c277df42 | 5563 | do_ifmatch: |
40a82448 | 5564 | ST.me = scan; |
24d3c4a9 | 5565 | ST.logical = logical; |
24d786f4 YO |
5566 | logical = 0; /* XXX: reset state of logical once it has been saved into ST */ |
5567 | ||
40a82448 DM |
5568 | /* execute body of (?...A) */ |
5569 | PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan))); | |
5570 | /* NOTREACHED */ | |
5571 | ||
5572 | case IFMATCH_A_fail: /* body of (?...A) failed */ | |
5573 | ST.wanted = !ST.wanted; | |
5574 | /* FALL THROUGH */ | |
5575 | ||
5576 | case IFMATCH_A: /* body of (?...A) succeeded */ | |
24d3c4a9 | 5577 | if (ST.logical) { |
f2338a2e | 5578 | sw = cBOOL(ST.wanted); |
40a82448 DM |
5579 | } |
5580 | else if (!ST.wanted) | |
5581 | sayNO; | |
5582 | ||
5583 | if (OP(ST.me) == SUSPEND) | |
5584 | locinput = PL_reginput; | |
5585 | else { | |
5586 | locinput = PL_reginput = st->locinput; | |
5587 | nextchr = UCHARAT(locinput); | |
5588 | } | |
5589 | scan = ST.me + ARG(ST.me); | |
5590 | if (scan == ST.me) | |
5591 | scan = NULL; | |
5592 | continue; /* execute B */ | |
5593 | ||
5594 | #undef ST | |
dad79028 | 5595 | |
c277df42 | 5596 | case LONGJMP: |
c277df42 IZ |
5597 | next = scan + ARG(scan); |
5598 | if (next == scan) | |
5599 | next = NULL; | |
a0d0e21e | 5600 | break; |
54612592 | 5601 | case COMMIT: |
e2e6a0f1 YO |
5602 | reginfo->cutpoint = PL_regeol; |
5603 | /* FALLTHROUGH */ | |
5d458dd8 | 5604 | case PRUNE: |
24b23f37 | 5605 | PL_reginput = locinput; |
e2e6a0f1 | 5606 | if (!scan->flags) |
ad64d0ec | 5607 | sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
54612592 YO |
5608 | PUSH_STATE_GOTO(COMMIT_next,next); |
5609 | /* NOTREACHED */ | |
5610 | case COMMIT_next_fail: | |
5611 | no_final = 1; | |
5612 | /* FALLTHROUGH */ | |
7f69552c YO |
5613 | case OPFAIL: |
5614 | sayNO; | |
e2e6a0f1 YO |
5615 | /* NOTREACHED */ |
5616 | ||
5617 | #define ST st->u.mark | |
5618 | case MARKPOINT: | |
5619 | ST.prev_mark = mark_state; | |
5d458dd8 | 5620 | ST.mark_name = sv_commit = sv_yes_mark |
ad64d0ec | 5621 | = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
e2e6a0f1 YO |
5622 | mark_state = st; |
5623 | ST.mark_loc = PL_reginput = locinput; | |
5624 | PUSH_YES_STATE_GOTO(MARKPOINT_next,next); | |
5625 | /* NOTREACHED */ | |
5626 | case MARKPOINT_next: | |
5627 | mark_state = ST.prev_mark; | |
5628 | sayYES; | |
5629 | /* NOTREACHED */ | |
5630 | case MARKPOINT_next_fail: | |
5d458dd8 | 5631 | if (popmark && sv_eq(ST.mark_name,popmark)) |
e2e6a0f1 YO |
5632 | { |
5633 | if (ST.mark_loc > startpoint) | |
5634 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5635 | popmark = NULL; /* we found our mark */ | |
5636 | sv_commit = ST.mark_name; | |
5637 | ||
5638 | DEBUG_EXECUTE_r({ | |
5d458dd8 | 5639 | PerlIO_printf(Perl_debug_log, |
e2e6a0f1 YO |
5640 | "%*s %ssetting cutpoint to mark:%"SVf"...%s\n", |
5641 | REPORT_CODE_OFF+depth*2, "", | |
be2597df | 5642 | PL_colors[4], SVfARG(sv_commit), PL_colors[5]); |
e2e6a0f1 YO |
5643 | }); |
5644 | } | |
5645 | mark_state = ST.prev_mark; | |
5d458dd8 YO |
5646 | sv_yes_mark = mark_state ? |
5647 | mark_state->u.mark.mark_name : NULL; | |
e2e6a0f1 YO |
5648 | sayNO; |
5649 | /* NOTREACHED */ | |
5d458dd8 YO |
5650 | case SKIP: |
5651 | PL_reginput = locinput; | |
5652 | if (scan->flags) { | |
2bf803e2 | 5653 | /* (*SKIP) : if we fail we cut here*/ |
5d458dd8 | 5654 | ST.mark_name = NULL; |
e2e6a0f1 | 5655 | ST.mark_loc = locinput; |
5d458dd8 YO |
5656 | PUSH_STATE_GOTO(SKIP_next,next); |
5657 | } else { | |
2bf803e2 | 5658 | /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was, |
5d458dd8 YO |
5659 | otherwise do nothing. Meaning we need to scan |
5660 | */ | |
5661 | regmatch_state *cur = mark_state; | |
ad64d0ec | 5662 | SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
5663 | |
5664 | while (cur) { | |
5665 | if ( sv_eq( cur->u.mark.mark_name, | |
5666 | find ) ) | |
5667 | { | |
5668 | ST.mark_name = find; | |
5669 | PUSH_STATE_GOTO( SKIP_next, next ); | |
5670 | } | |
5671 | cur = cur->u.mark.prev_mark; | |
5672 | } | |
e2e6a0f1 | 5673 | } |
2bf803e2 | 5674 | /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */ |
5d458dd8 YO |
5675 | break; |
5676 | case SKIP_next_fail: | |
5677 | if (ST.mark_name) { | |
5678 | /* (*CUT:NAME) - Set up to search for the name as we | |
5679 | collapse the stack*/ | |
5680 | popmark = ST.mark_name; | |
5681 | } else { | |
5682 | /* (*CUT) - No name, we cut here.*/ | |
e2e6a0f1 YO |
5683 | if (ST.mark_loc > startpoint) |
5684 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5d458dd8 YO |
5685 | /* but we set sv_commit to latest mark_name if there |
5686 | is one so they can test to see how things lead to this | |
5687 | cut */ | |
5688 | if (mark_state) | |
5689 | sv_commit=mark_state->u.mark.mark_name; | |
5690 | } | |
e2e6a0f1 YO |
5691 | no_final = 1; |
5692 | sayNO; | |
5693 | /* NOTREACHED */ | |
5694 | #undef ST | |
32e6a07c YO |
5695 | case FOLDCHAR: |
5696 | n = ARG(scan); | |
f2ed9b32 | 5697 | if ( n == (U32)what_len_TRICKYFOLD(locinput,utf8_target,ln) ) { |
e64b1bd1 | 5698 | locinput += ln; |
ced7f090 | 5699 | } else if ( LATIN_SMALL_LETTER_SHARP_S == n && !utf8_target && !UTF_PATTERN ) { |
e64b1bd1 YO |
5700 | sayNO; |
5701 | } else { | |
5702 | U8 folded[UTF8_MAXBYTES_CASE+1]; | |
5703 | STRLEN foldlen; | |
5704 | const char * const l = locinput; | |
5705 | char *e = PL_regeol; | |
5706 | to_uni_fold(n, folded, &foldlen); | |
5707 | ||
4c1b470c | 5708 | if (! foldEQ_utf8((const char*) folded, 0, foldlen, 1, |
f2ed9b32 | 5709 | l, &e, 0, utf8_target)) { |
32e6a07c | 5710 | sayNO; |
e64b1bd1 YO |
5711 | } |
5712 | locinput = e; | |
32e6a07c YO |
5713 | } |
5714 | nextchr = UCHARAT(locinput); | |
5715 | break; | |
e1d1eefb | 5716 | case LNBREAK: |
f2ed9b32 | 5717 | if ((n=is_LNBREAK(locinput,utf8_target))) { |
e1d1eefb YO |
5718 | locinput += n; |
5719 | nextchr = UCHARAT(locinput); | |
5720 | } else | |
5721 | sayNO; | |
5722 | break; | |
5723 | ||
5724 | #define CASE_CLASS(nAmE) \ | |
5725 | case nAmE: \ | |
1380b51d DM |
5726 | if (locinput >= PL_regeol) \ |
5727 | sayNO; \ | |
f2ed9b32 | 5728 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5729 | locinput += n; \ |
5730 | nextchr = UCHARAT(locinput); \ | |
5731 | } else \ | |
5732 | sayNO; \ | |
5733 | break; \ | |
5734 | case N##nAmE: \ | |
1380b51d DM |
5735 | if (locinput >= PL_regeol) \ |
5736 | sayNO; \ | |
f2ed9b32 | 5737 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5738 | sayNO; \ |
5739 | } else { \ | |
5740 | locinput += UTF8SKIP(locinput); \ | |
5741 | nextchr = UCHARAT(locinput); \ | |
5742 | } \ | |
5743 | break | |
5744 | ||
5745 | CASE_CLASS(VERTWS); | |
5746 | CASE_CLASS(HORIZWS); | |
5747 | #undef CASE_CLASS | |
5748 | ||
a0d0e21e | 5749 | default: |
b900a521 | 5750 | PerlIO_printf(Perl_error_log, "%"UVxf" %d\n", |
d7d93a81 | 5751 | PTR2UV(scan), OP(scan)); |
cea2e8a9 | 5752 | Perl_croak(aTHX_ "regexp memory corruption"); |
5d458dd8 YO |
5753 | |
5754 | } /* end switch */ | |
95b24440 | 5755 | |
5d458dd8 YO |
5756 | /* switch break jumps here */ |
5757 | scan = next; /* prepare to execute the next op and ... */ | |
5758 | continue; /* ... jump back to the top, reusing st */ | |
95b24440 DM |
5759 | /* NOTREACHED */ |
5760 | ||
40a82448 DM |
5761 | push_yes_state: |
5762 | /* push a state that backtracks on success */ | |
5763 | st->u.yes.prev_yes_state = yes_state; | |
5764 | yes_state = st; | |
5765 | /* FALL THROUGH */ | |
5766 | push_state: | |
5767 | /* push a new regex state, then continue at scan */ | |
5768 | { | |
5769 | regmatch_state *newst; | |
5770 | ||
24b23f37 YO |
5771 | DEBUG_STACK_r({ |
5772 | regmatch_state *cur = st; | |
5773 | regmatch_state *curyes = yes_state; | |
5774 | int curd = depth; | |
5775 | regmatch_slab *slab = PL_regmatch_slab; | |
5776 | for (;curd > -1;cur--,curd--) { | |
5777 | if (cur < SLAB_FIRST(slab)) { | |
5778 | slab = slab->prev; | |
5779 | cur = SLAB_LAST(slab); | |
5780 | } | |
5781 | PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n", | |
5782 | REPORT_CODE_OFF + 2 + depth * 2,"", | |
13d6edb4 | 5783 | curd, PL_reg_name[cur->resume_state], |
24b23f37 YO |
5784 | (curyes == cur) ? "yes" : "" |
5785 | ); | |
5786 | if (curyes == cur) | |
5787 | curyes = cur->u.yes.prev_yes_state; | |
5788 | } | |
5789 | } else | |
5790 | DEBUG_STATE_pp("push") | |
5791 | ); | |
40a82448 | 5792 | depth++; |
40a82448 DM |
5793 | st->locinput = locinput; |
5794 | newst = st+1; | |
5795 | if (newst > SLAB_LAST(PL_regmatch_slab)) | |
5796 | newst = S_push_slab(aTHX); | |
5797 | PL_regmatch_state = newst; | |
786e8c11 | 5798 | |
40a82448 DM |
5799 | locinput = PL_reginput; |
5800 | nextchr = UCHARAT(locinput); | |
5801 | st = newst; | |
5802 | continue; | |
5803 | /* NOTREACHED */ | |
5804 | } | |
a0d0e21e | 5805 | } |
a687059c | 5806 | |
a0d0e21e LW |
5807 | /* |
5808 | * We get here only if there's trouble -- normally "case END" is | |
5809 | * the terminating point. | |
5810 | */ | |
cea2e8a9 | 5811 | Perl_croak(aTHX_ "corrupted regexp pointers"); |
a0d0e21e | 5812 | /*NOTREACHED*/ |
4633a7c4 LW |
5813 | sayNO; |
5814 | ||
262b90c4 | 5815 | yes: |
77cb431f DM |
5816 | if (yes_state) { |
5817 | /* we have successfully completed a subexpression, but we must now | |
5818 | * pop to the state marked by yes_state and continue from there */ | |
77cb431f | 5819 | assert(st != yes_state); |
5bc10b2c DM |
5820 | #ifdef DEBUGGING |
5821 | while (st != yes_state) { | |
5822 | st--; | |
5823 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
5824 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5825 | st = SLAB_LAST(PL_regmatch_slab); | |
5826 | } | |
e2e6a0f1 | 5827 | DEBUG_STATE_r({ |
54612592 YO |
5828 | if (no_final) { |
5829 | DEBUG_STATE_pp("pop (no final)"); | |
5830 | } else { | |
5831 | DEBUG_STATE_pp("pop (yes)"); | |
5832 | } | |
e2e6a0f1 | 5833 | }); |
5bc10b2c DM |
5834 | depth--; |
5835 | } | |
5836 | #else | |
77cb431f DM |
5837 | while (yes_state < SLAB_FIRST(PL_regmatch_slab) |
5838 | || yes_state > SLAB_LAST(PL_regmatch_slab)) | |
5839 | { | |
5840 | /* not in this slab, pop slab */ | |
5841 | depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1); | |
5842 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5843 | st = SLAB_LAST(PL_regmatch_slab); | |
5844 | } | |
5845 | depth -= (st - yes_state); | |
5bc10b2c | 5846 | #endif |
77cb431f DM |
5847 | st = yes_state; |
5848 | yes_state = st->u.yes.prev_yes_state; | |
5849 | PL_regmatch_state = st; | |
24b23f37 | 5850 | |
5d458dd8 YO |
5851 | if (no_final) { |
5852 | locinput= st->locinput; | |
5853 | nextchr = UCHARAT(locinput); | |
5854 | } | |
54612592 | 5855 | state_num = st->resume_state + no_final; |
24d3c4a9 | 5856 | goto reenter_switch; |
77cb431f DM |
5857 | } |
5858 | ||
a3621e74 | 5859 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", |
e4584336 | 5860 | PL_colors[4], PL_colors[5])); |
02db2b7b | 5861 | |
19b95bf0 DM |
5862 | if (PL_reg_eval_set) { |
5863 | /* each successfully executed (?{...}) block does the equivalent of | |
5864 | * local $^R = do {...} | |
5865 | * When popping the save stack, all these locals would be undone; | |
5866 | * bypass this by setting the outermost saved $^R to the latest | |
5867 | * value */ | |
5868 | if (oreplsv != GvSV(PL_replgv)) | |
5869 | sv_setsv(oreplsv, GvSV(PL_replgv)); | |
5870 | } | |
95b24440 | 5871 | result = 1; |
aa283a38 | 5872 | goto final_exit; |
4633a7c4 LW |
5873 | |
5874 | no: | |
a3621e74 | 5875 | DEBUG_EXECUTE_r( |
7821416a | 5876 | PerlIO_printf(Perl_debug_log, |
786e8c11 | 5877 | "%*s %sfailed...%s\n", |
5bc10b2c | 5878 | REPORT_CODE_OFF+depth*2, "", |
786e8c11 | 5879 | PL_colors[4], PL_colors[5]) |
7821416a | 5880 | ); |
aa283a38 | 5881 | |
262b90c4 | 5882 | no_silent: |
54612592 YO |
5883 | if (no_final) { |
5884 | if (yes_state) { | |
5885 | goto yes; | |
5886 | } else { | |
5887 | goto final_exit; | |
5888 | } | |
5889 | } | |
aa283a38 DM |
5890 | if (depth) { |
5891 | /* there's a previous state to backtrack to */ | |
40a82448 DM |
5892 | st--; |
5893 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
5894 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5895 | st = SLAB_LAST(PL_regmatch_slab); | |
5896 | } | |
5897 | PL_regmatch_state = st; | |
40a82448 DM |
5898 | locinput= st->locinput; |
5899 | nextchr = UCHARAT(locinput); | |
5900 | ||
5bc10b2c DM |
5901 | DEBUG_STATE_pp("pop"); |
5902 | depth--; | |
262b90c4 DM |
5903 | if (yes_state == st) |
5904 | yes_state = st->u.yes.prev_yes_state; | |
5bc10b2c | 5905 | |
24d3c4a9 DM |
5906 | state_num = st->resume_state + 1; /* failure = success + 1 */ |
5907 | goto reenter_switch; | |
95b24440 | 5908 | } |
24d3c4a9 | 5909 | result = 0; |
aa283a38 | 5910 | |
262b90c4 | 5911 | final_exit: |
bbe252da | 5912 | if (rex->intflags & PREGf_VERBARG_SEEN) { |
5d458dd8 YO |
5913 | SV *sv_err = get_sv("REGERROR", 1); |
5914 | SV *sv_mrk = get_sv("REGMARK", 1); | |
5915 | if (result) { | |
e2e6a0f1 | 5916 | sv_commit = &PL_sv_no; |
5d458dd8 YO |
5917 | if (!sv_yes_mark) |
5918 | sv_yes_mark = &PL_sv_yes; | |
5919 | } else { | |
5920 | if (!sv_commit) | |
5921 | sv_commit = &PL_sv_yes; | |
5922 | sv_yes_mark = &PL_sv_no; | |
5923 | } | |
5924 | sv_setsv(sv_err, sv_commit); | |
5925 | sv_setsv(sv_mrk, sv_yes_mark); | |
e2e6a0f1 | 5926 | } |
19b95bf0 | 5927 | |
2f554ef7 DM |
5928 | /* clean up; in particular, free all slabs above current one */ |
5929 | LEAVE_SCOPE(oldsave); | |
5d9a96ca | 5930 | |
95b24440 | 5931 | return result; |
a687059c LW |
5932 | } |
5933 | ||
5934 | /* | |
5935 | - regrepeat - repeatedly match something simple, report how many | |
5936 | */ | |
5937 | /* | |
5938 | * [This routine now assumes that it will only match on things of length 1. | |
5939 | * That was true before, but now we assume scan - reginput is the count, | |
a0ed51b3 | 5940 | * rather than incrementing count on every character. [Er, except utf8.]] |
a687059c | 5941 | */ |
76e3520e | 5942 | STATIC I32 |
e2e6a0f1 | 5943 | S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth) |
a687059c | 5944 | { |
27da23d5 | 5945 | dVAR; |
a0d0e21e | 5946 | register char *scan; |
a0d0e21e | 5947 | register I32 c; |
3280af22 | 5948 | register char *loceol = PL_regeol; |
a0ed51b3 | 5949 | register I32 hardcount = 0; |
f2ed9b32 | 5950 | register bool utf8_target = PL_reg_match_utf8; |
d513472c | 5951 | UV utf8_flags; |
4f55667c SP |
5952 | #ifndef DEBUGGING |
5953 | PERL_UNUSED_ARG(depth); | |
5954 | #endif | |
a0d0e21e | 5955 | |
7918f24d NC |
5956 | PERL_ARGS_ASSERT_REGREPEAT; |
5957 | ||
3280af22 | 5958 | scan = PL_reginput; |
faf11cac HS |
5959 | if (max == REG_INFTY) |
5960 | max = I32_MAX; | |
5961 | else if (max < loceol - scan) | |
7f596f4c | 5962 | loceol = scan + max; |
a0d0e21e | 5963 | switch (OP(p)) { |
22c35a8c | 5964 | case REG_ANY: |
f2ed9b32 | 5965 | if (utf8_target) { |
ffc61ed2 | 5966 | loceol = PL_regeol; |
1aa99e6b | 5967 | while (scan < loceol && hardcount < max && *scan != '\n') { |
ffc61ed2 JH |
5968 | scan += UTF8SKIP(scan); |
5969 | hardcount++; | |
5970 | } | |
5971 | } else { | |
5972 | while (scan < loceol && *scan != '\n') | |
5973 | scan++; | |
a0ed51b3 LW |
5974 | } |
5975 | break; | |
ffc61ed2 | 5976 | case SANY: |
f2ed9b32 | 5977 | if (utf8_target) { |
def8e4ea | 5978 | loceol = PL_regeol; |
a0804c9e | 5979 | while (scan < loceol && hardcount < max) { |
def8e4ea JH |
5980 | scan += UTF8SKIP(scan); |
5981 | hardcount++; | |
5982 | } | |
5983 | } | |
5984 | else | |
5985 | scan = loceol; | |
a0ed51b3 | 5986 | break; |
f33976b4 DB |
5987 | case CANY: |
5988 | scan = loceol; | |
5989 | break; | |
59d32103 KW |
5990 | case EXACT: |
5991 | /* To get here, EXACTish nodes must have *byte* length == 1. That | |
5992 | * means they match only characters in the string that can be expressed | |
5993 | * as a single byte. For non-utf8 strings, that means a simple match. | |
5994 | * For utf8 strings, the character matched must be an invariant, or | |
5995 | * downgradable to a single byte. The pattern's utf8ness is | |
5996 | * irrelevant, as since it's a single byte, it either isn't utf8, or if | |
5997 | * it is, it's an invariant */ | |
5998 | ||
5999 | c = (U8)*STRING(p); | |
6000 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); | |
6001 | ||
6002 | if (! utf8_target || UNI_IS_INVARIANT(c)) { | |
6003 | while (scan < loceol && UCHARAT(scan) == c) { | |
6004 | scan++; | |
6005 | } | |
6006 | } | |
6007 | else { | |
6008 | ||
6009 | /* Here, the string is utf8, and the pattern char is different | |
6010 | * in utf8 than not, so can't compare them directly. Outside the | |
5908807f | 6011 | * loop, find the two utf8 bytes that represent c, and then |
59d32103 KW |
6012 | * look for those in sequence in the utf8 string */ |
6013 | U8 high = UTF8_TWO_BYTE_HI(c); | |
6014 | U8 low = UTF8_TWO_BYTE_LO(c); | |
6015 | loceol = PL_regeol; | |
6016 | ||
6017 | while (hardcount < max | |
6018 | && scan + 1 < loceol | |
6019 | && UCHARAT(scan) == high | |
6020 | && UCHARAT(scan + 1) == low) | |
6021 | { | |
6022 | scan += 2; | |
6023 | hardcount++; | |
6024 | } | |
6025 | } | |
6026 | break; | |
2f7f8cb1 KW |
6027 | case EXACTFA: |
6028 | utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
6029 | goto do_exactf; | |
6030 | ||
d4e0b827 KW |
6031 | case EXACTFL: |
6032 | PL_reg_flags |= RF_tainted; | |
17580e7a KW |
6033 | utf8_flags = FOLDEQ_UTF8_LOCALE; |
6034 | goto do_exactf; | |
6035 | ||
d4e0b827 | 6036 | case EXACTF: |
9a5a5549 | 6037 | case EXACTFU: |
05f861a2 | 6038 | utf8_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0; |
59d32103 | 6039 | |
2be3e190 KW |
6040 | /* The comments for the EXACT case above apply as well to these fold |
6041 | * ones */ | |
634c83a2 | 6042 | |
2f7f8cb1 | 6043 | do_exactf: |
090f7165 | 6044 | c = (U8)*STRING(p); |
634c83a2 | 6045 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); |
d4e0b827 | 6046 | |
59d32103 | 6047 | if (utf8_target) { /* Use full Unicode fold matching */ |
59d32103 KW |
6048 | char *tmpeol = loceol; |
6049 | while (hardcount < max | |
d513472c KW |
6050 | && foldEQ_utf8_flags(scan, &tmpeol, 0, utf8_target, |
6051 | STRING(p), NULL, 1, cBOOL(UTF_PATTERN), utf8_flags)) | |
59d32103 KW |
6052 | { |
6053 | scan = tmpeol; | |
6054 | tmpeol = loceol; | |
6055 | hardcount++; | |
6056 | } | |
634c83a2 | 6057 | |
59d32103 KW |
6058 | /* XXX Note that the above handles properly the German sharp s in |
6059 | * the pattern matching ss in the string. But it doesn't handle | |
6060 | * properly cases where the string contains say 'LIGATURE ff' and | |
6061 | * the pattern is 'f+'. This would require, say, a new function or | |
6062 | * revised interface to foldEQ_utf8(), in which the maximum number | |
6063 | * of characters to match could be passed and it would return how | |
6064 | * many actually did. This is just one of many cases where | |
6065 | * multi-char folds don't work properly, and so the fix is being | |
6066 | * deferred */ | |
6067 | } | |
6068 | else { | |
87381386 | 6069 | U8 folded; |
59d32103 | 6070 | |
2be3e190 KW |
6071 | /* Here, the string isn't utf8 and c is a single byte; and either |
6072 | * the pattern isn't utf8 or c is an invariant, so its utf8ness | |
6073 | * doesn't affect c. Can just do simple comparisons for exact or | |
6074 | * fold matching. */ | |
d4e0b827 | 6075 | switch (OP(p)) { |
87381386 | 6076 | case EXACTF: folded = PL_fold[c]; break; |
2f7f8cb1 | 6077 | case EXACTFA: |
9a5a5549 | 6078 | case EXACTFU: folded = PL_fold_latin1[c]; break; |
87381386 KW |
6079 | case EXACTFL: folded = PL_fold_locale[c]; break; |
6080 | default: Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p)); | |
6081 | } | |
6082 | while (scan < loceol && | |
6083 | (UCHARAT(scan) == c || UCHARAT(scan) == folded)) | |
6084 | { | |
6085 | scan++; | |
634c83a2 KW |
6086 | } |
6087 | } | |
bbce6d69 | 6088 | break; |
f56b6394 | 6089 | case ANYOFV: |
a0d0e21e | 6090 | case ANYOF: |
4e8910e0 KW |
6091 | if (utf8_target || OP(p) == ANYOFV) { |
6092 | STRLEN inclasslen; | |
ffc61ed2 | 6093 | loceol = PL_regeol; |
4e8910e0 KW |
6094 | inclasslen = loceol - scan; |
6095 | while (hardcount < max | |
6096 | && ((inclasslen = loceol - scan) > 0) | |
6097 | && reginclass(prog, p, (U8*)scan, &inclasslen, utf8_target)) | |
6098 | { | |
6099 | scan += inclasslen; | |
ffc61ed2 JH |
6100 | hardcount++; |
6101 | } | |
6102 | } else { | |
32fc9b6a | 6103 | while (scan < loceol && REGINCLASS(prog, p, (U8*)scan)) |
ffc61ed2 JH |
6104 | scan++; |
6105 | } | |
a0d0e21e | 6106 | break; |
980866de | 6107 | case ALNUMU: |
f2ed9b32 | 6108 | if (utf8_target) { |
980866de | 6109 | utf8_wordchar: |
ffc61ed2 | 6110 | loceol = PL_regeol; |
1a4fad37 | 6111 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6112 | while (hardcount < max && scan < loceol && |
a12cf05f KW |
6113 | swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
6114 | { | |
ffc61ed2 JH |
6115 | scan += UTF8SKIP(scan); |
6116 | hardcount++; | |
6117 | } | |
980866de | 6118 | } else { |
a12cf05f KW |
6119 | while (scan < loceol && isWORDCHAR_L1((U8) *scan)) { |
6120 | scan++; | |
6121 | } | |
980866de KW |
6122 | } |
6123 | break; | |
6124 | case ALNUM: | |
6125 | if (utf8_target) | |
6126 | goto utf8_wordchar; | |
6127 | while (scan < loceol && isALNUM((U8) *scan)) { | |
6128 | scan++; | |
a0ed51b3 LW |
6129 | } |
6130 | break; | |
cfaf538b KW |
6131 | case ALNUMA: |
6132 | while (scan < loceol && isWORDCHAR_A((U8) *scan)) { | |
6133 | scan++; | |
6134 | } | |
6135 | break; | |
bbce6d69 | 6136 | case ALNUML: |
3280af22 | 6137 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6138 | if (utf8_target) { |
ffc61ed2 | 6139 | loceol = PL_regeol; |
1aa99e6b IH |
6140 | while (hardcount < max && scan < loceol && |
6141 | isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6142 | scan += UTF8SKIP(scan); |
6143 | hardcount++; | |
6144 | } | |
6145 | } else { | |
6146 | while (scan < loceol && isALNUM_LC(*scan)) | |
6147 | scan++; | |
a0ed51b3 LW |
6148 | } |
6149 | break; | |
980866de | 6150 | case NALNUMU: |
f2ed9b32 | 6151 | if (utf8_target) { |
980866de KW |
6152 | |
6153 | utf8_Nwordchar: | |
6154 | ||
ffc61ed2 | 6155 | loceol = PL_regeol; |
1a4fad37 | 6156 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 6157 | while (hardcount < max && scan < loceol && |
980866de | 6158 | ! swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
a12cf05f | 6159 | { |
ffc61ed2 JH |
6160 | scan += UTF8SKIP(scan); |
6161 | hardcount++; | |
6162 | } | |
980866de | 6163 | } else { |
a12cf05f KW |
6164 | while (scan < loceol && ! isWORDCHAR_L1((U8) *scan)) { |
6165 | scan++; | |
6166 | } | |
980866de KW |
6167 | } |
6168 | break; | |
6169 | case NALNUM: | |
6170 | if (utf8_target) | |
6171 | goto utf8_Nwordchar; | |
6172 | while (scan < loceol && ! isALNUM((U8) *scan)) { | |
6173 | scan++; | |
a0ed51b3 LW |
6174 | } |
6175 | break; | |
cfaf538b KW |
6176 | case NALNUMA: |
6177 | if (utf8_target) { | |
6178 | while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) { | |
6179 | scan += UTF8SKIP(scan); | |
6180 | } | |
6181 | } | |
6182 | else { | |
6183 | while (scan < loceol && ! isWORDCHAR_A((U8) *scan)) { | |
6184 | scan++; | |
6185 | } | |
6186 | } | |
6187 | break; | |
bbce6d69 | 6188 | case NALNUML: |
3280af22 | 6189 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6190 | if (utf8_target) { |
ffc61ed2 | 6191 | loceol = PL_regeol; |
1aa99e6b IH |
6192 | while (hardcount < max && scan < loceol && |
6193 | !isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6194 | scan += UTF8SKIP(scan); |
6195 | hardcount++; | |
6196 | } | |
6197 | } else { | |
6198 | while (scan < loceol && !isALNUM_LC(*scan)) | |
6199 | scan++; | |
a0ed51b3 LW |
6200 | } |
6201 | break; | |
980866de | 6202 | case SPACEU: |
f2ed9b32 | 6203 | if (utf8_target) { |
980866de KW |
6204 | |
6205 | utf8_space: | |
6206 | ||
ffc61ed2 | 6207 | loceol = PL_regeol; |
1a4fad37 | 6208 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6209 | while (hardcount < max && scan < loceol && |
3568d838 | 6210 | (*scan == ' ' || |
a12cf05f KW |
6211 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) |
6212 | { | |
ffc61ed2 JH |
6213 | scan += UTF8SKIP(scan); |
6214 | hardcount++; | |
6215 | } | |
980866de KW |
6216 | break; |
6217 | } | |
6218 | else { | |
a12cf05f KW |
6219 | while (scan < loceol && isSPACE_L1((U8) *scan)) { |
6220 | scan++; | |
6221 | } | |
980866de KW |
6222 | break; |
6223 | } | |
6224 | case SPACE: | |
6225 | if (utf8_target) | |
6226 | goto utf8_space; | |
6227 | ||
6228 | while (scan < loceol && isSPACE((U8) *scan)) { | |
6229 | scan++; | |
a0ed51b3 LW |
6230 | } |
6231 | break; | |
cfaf538b KW |
6232 | case SPACEA: |
6233 | while (scan < loceol && isSPACE_A((U8) *scan)) { | |
6234 | scan++; | |
6235 | } | |
6236 | break; | |
bbce6d69 | 6237 | case SPACEL: |
3280af22 | 6238 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6239 | if (utf8_target) { |
ffc61ed2 | 6240 | loceol = PL_regeol; |
1aa99e6b | 6241 | while (hardcount < max && scan < loceol && |
6bbba904 | 6242 | isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6243 | scan += UTF8SKIP(scan); |
6244 | hardcount++; | |
6245 | } | |
6246 | } else { | |
6247 | while (scan < loceol && isSPACE_LC(*scan)) | |
6248 | scan++; | |
a0ed51b3 LW |
6249 | } |
6250 | break; | |
980866de | 6251 | case NSPACEU: |
f2ed9b32 | 6252 | if (utf8_target) { |
980866de KW |
6253 | |
6254 | utf8_Nspace: | |
6255 | ||
ffc61ed2 | 6256 | loceol = PL_regeol; |
1a4fad37 | 6257 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6258 | while (hardcount < max && scan < loceol && |
980866de KW |
6259 | ! (*scan == ' ' || |
6260 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) | |
a12cf05f | 6261 | { |
ffc61ed2 JH |
6262 | scan += UTF8SKIP(scan); |
6263 | hardcount++; | |
6264 | } | |
980866de KW |
6265 | break; |
6266 | } | |
6267 | else { | |
a12cf05f KW |
6268 | while (scan < loceol && ! isSPACE_L1((U8) *scan)) { |
6269 | scan++; | |
6270 | } | |
980866de KW |
6271 | } |
6272 | break; | |
6273 | case NSPACE: | |
6274 | if (utf8_target) | |
6275 | goto utf8_Nspace; | |
6276 | ||
6277 | while (scan < loceol && ! isSPACE((U8) *scan)) { | |
6278 | scan++; | |
a0ed51b3 | 6279 | } |
0008a298 | 6280 | break; |
cfaf538b KW |
6281 | case NSPACEA: |
6282 | if (utf8_target) { | |
6283 | while (scan < loceol && ! isSPACE_A((U8) *scan)) { | |
6284 | scan += UTF8SKIP(scan); | |
6285 | } | |
6286 | } | |
6287 | else { | |
6288 | while (scan < loceol && ! isSPACE_A((U8) *scan)) { | |
6289 | scan++; | |
6290 | } | |
6291 | } | |
6292 | break; | |
bbce6d69 | 6293 | case NSPACEL: |
3280af22 | 6294 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6295 | if (utf8_target) { |
ffc61ed2 | 6296 | loceol = PL_regeol; |
1aa99e6b | 6297 | while (hardcount < max && scan < loceol && |
6bbba904 | 6298 | !isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6299 | scan += UTF8SKIP(scan); |
6300 | hardcount++; | |
6301 | } | |
6302 | } else { | |
6303 | while (scan < loceol && !isSPACE_LC(*scan)) | |
6304 | scan++; | |
a0ed51b3 LW |
6305 | } |
6306 | break; | |
a0d0e21e | 6307 | case DIGIT: |
f2ed9b32 | 6308 | if (utf8_target) { |
ffc61ed2 | 6309 | loceol = PL_regeol; |
1a4fad37 | 6310 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6311 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6312 | swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6313 | scan += UTF8SKIP(scan); |
6314 | hardcount++; | |
6315 | } | |
6316 | } else { | |
6317 | while (scan < loceol && isDIGIT(*scan)) | |
6318 | scan++; | |
a0ed51b3 LW |
6319 | } |
6320 | break; | |
cfaf538b KW |
6321 | case DIGITA: |
6322 | while (scan < loceol && isDIGIT_A((U8) *scan)) { | |
6323 | scan++; | |
6324 | } | |
6325 | break; | |
b77393f6 KW |
6326 | case DIGITL: |
6327 | PL_reg_flags |= RF_tainted; | |
6328 | if (utf8_target) { | |
6329 | loceol = PL_regeol; | |
6330 | while (hardcount < max && scan < loceol && | |
6331 | isDIGIT_LC_utf8((U8*)scan)) { | |
6332 | scan += UTF8SKIP(scan); | |
6333 | hardcount++; | |
6334 | } | |
6335 | } else { | |
6336 | while (scan < loceol && isDIGIT_LC(*scan)) | |
6337 | scan++; | |
6338 | } | |
6339 | break; | |
a0d0e21e | 6340 | case NDIGIT: |
f2ed9b32 | 6341 | if (utf8_target) { |
ffc61ed2 | 6342 | loceol = PL_regeol; |
1a4fad37 | 6343 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6344 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6345 | !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6346 | scan += UTF8SKIP(scan); |
6347 | hardcount++; | |
6348 | } | |
6349 | } else { | |
6350 | while (scan < loceol && !isDIGIT(*scan)) | |
6351 | scan++; | |
a0ed51b3 | 6352 | } |
cfaf538b KW |
6353 | break; |
6354 | case NDIGITA: | |
6355 | if (utf8_target) { | |
6356 | while (scan < loceol && ! isDIGIT_A((U8) *scan)) { | |
6357 | scan += UTF8SKIP(scan); | |
6358 | } | |
6359 | } | |
6360 | else { | |
6361 | while (scan < loceol && ! isDIGIT_A((U8) *scan)) { | |
6362 | scan++; | |
6363 | } | |
6364 | } | |
6365 | break; | |
b77393f6 KW |
6366 | case NDIGITL: |
6367 | PL_reg_flags |= RF_tainted; | |
6368 | if (utf8_target) { | |
6369 | loceol = PL_regeol; | |
6370 | while (hardcount < max && scan < loceol && | |
6371 | !isDIGIT_LC_utf8((U8*)scan)) { | |
6372 | scan += UTF8SKIP(scan); | |
6373 | hardcount++; | |
6374 | } | |
6375 | } else { | |
6376 | while (scan < loceol && !isDIGIT_LC(*scan)) | |
6377 | scan++; | |
6378 | } | |
6379 | break; | |
e1d1eefb | 6380 | case LNBREAK: |
f2ed9b32 | 6381 | if (utf8_target) { |
e1d1eefb YO |
6382 | loceol = PL_regeol; |
6383 | while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) { | |
6384 | scan += c; | |
6385 | hardcount++; | |
6386 | } | |
6387 | } else { | |
6388 | /* | |
6389 | LNBREAK can match two latin chars, which is ok, | |
6390 | because we have a null terminated string, but we | |
6391 | have to use hardcount in this situation | |
6392 | */ | |
6393 | while (scan < loceol && (c=is_LNBREAK_latin1(scan))) { | |
6394 | scan+=c; | |
6395 | hardcount++; | |
6396 | } | |
6397 | } | |
6398 | break; | |
6399 | case HORIZWS: | |
f2ed9b32 | 6400 | if (utf8_target) { |
e1d1eefb YO |
6401 | loceol = PL_regeol; |
6402 | while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) { | |
6403 | scan += c; | |
6404 | hardcount++; | |
6405 | } | |
6406 | } else { | |
6407 | while (scan < loceol && is_HORIZWS_latin1(scan)) | |
6408 | scan++; | |
6409 | } | |
a0ed51b3 | 6410 | break; |
e1d1eefb | 6411 | case NHORIZWS: |
f2ed9b32 | 6412 | if (utf8_target) { |
e1d1eefb YO |
6413 | loceol = PL_regeol; |
6414 | while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) { | |
6415 | scan += UTF8SKIP(scan); | |
6416 | hardcount++; | |
6417 | } | |
6418 | } else { | |
6419 | while (scan < loceol && !is_HORIZWS_latin1(scan)) | |
6420 | scan++; | |
6421 | ||
6422 | } | |
6423 | break; | |
6424 | case VERTWS: | |
f2ed9b32 | 6425 | if (utf8_target) { |
e1d1eefb YO |
6426 | loceol = PL_regeol; |
6427 | while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) { | |
6428 | scan += c; | |
6429 | hardcount++; | |
6430 | } | |
6431 | } else { | |
6432 | while (scan < loceol && is_VERTWS_latin1(scan)) | |
6433 | scan++; | |
6434 | ||
6435 | } | |
6436 | break; | |
6437 | case NVERTWS: | |
f2ed9b32 | 6438 | if (utf8_target) { |
e1d1eefb YO |
6439 | loceol = PL_regeol; |
6440 | while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) { | |
6441 | scan += UTF8SKIP(scan); | |
6442 | hardcount++; | |
6443 | } | |
6444 | } else { | |
6445 | while (scan < loceol && !is_VERTWS_latin1(scan)) | |
6446 | scan++; | |
6447 | ||
6448 | } | |
6449 | break; | |
6450 | ||
a0d0e21e LW |
6451 | default: /* Called on something of 0 width. */ |
6452 | break; /* So match right here or not at all. */ | |
6453 | } | |
a687059c | 6454 | |
a0ed51b3 LW |
6455 | if (hardcount) |
6456 | c = hardcount; | |
6457 | else | |
6458 | c = scan - PL_reginput; | |
3280af22 | 6459 | PL_reginput = scan; |
a687059c | 6460 | |
a3621e74 | 6461 | DEBUG_r({ |
e68ec53f | 6462 | GET_RE_DEBUG_FLAGS_DECL; |
be8e71aa | 6463 | DEBUG_EXECUTE_r({ |
e68ec53f YO |
6464 | SV * const prop = sv_newmortal(); |
6465 | regprop(prog, prop, p); | |
6466 | PerlIO_printf(Perl_debug_log, | |
be8e71aa | 6467 | "%*s %s can match %"IVdf" times out of %"IVdf"...\n", |
e2e6a0f1 | 6468 | REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max); |
a3621e74 | 6469 | }); |
be8e71aa | 6470 | }); |
9041c2e3 | 6471 | |
a0d0e21e | 6472 | return(c); |
a687059c LW |
6473 | } |
6474 | ||
c277df42 | 6475 | |
be8e71aa | 6476 | #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) |
c277df42 | 6477 | /* |
ffc61ed2 JH |
6478 | - regclass_swash - prepare the utf8 swash |
6479 | */ | |
6480 | ||
6481 | SV * | |
32fc9b6a | 6482 | Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) |
ffc61ed2 | 6483 | { |
97aff369 | 6484 | dVAR; |
9e55ce06 JH |
6485 | SV *sw = NULL; |
6486 | SV *si = NULL; | |
6487 | SV *alt = NULL; | |
f8fc2ecf YO |
6488 | RXi_GET_DECL(prog,progi); |
6489 | const struct reg_data * const data = prog ? progi->data : NULL; | |
ffc61ed2 | 6490 | |
7918f24d NC |
6491 | PERL_ARGS_ASSERT_REGCLASS_SWASH; |
6492 | ||
ccb2541c KW |
6493 | assert(ANYOF_NONBITMAP(node)); |
6494 | ||
4f639d21 | 6495 | if (data && data->count) { |
a3b680e6 | 6496 | const U32 n = ARG(node); |
ffc61ed2 | 6497 | |
4f639d21 | 6498 | if (data->what[n] == 's') { |
ad64d0ec NC |
6499 | SV * const rv = MUTABLE_SV(data->data[n]); |
6500 | AV * const av = MUTABLE_AV(SvRV(rv)); | |
2d03de9c | 6501 | SV **const ary = AvARRAY(av); |
9e55ce06 | 6502 | SV **a, **b; |
9041c2e3 | 6503 | |
711a919c | 6504 | /* See the end of regcomp.c:S_regclass() for |
9e55ce06 JH |
6505 | * documentation of these array elements. */ |
6506 | ||
b11f357e | 6507 | si = *ary; |
fe5bfecd JH |
6508 | a = SvROK(ary[1]) ? &ary[1] : NULL; |
6509 | b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : NULL; | |
b11f357e | 6510 | |
ffc61ed2 JH |
6511 | if (a) |
6512 | sw = *a; | |
6513 | else if (si && doinit) { | |
6514 | sw = swash_init("utf8", "", si, 1, 0); | |
6515 | (void)av_store(av, 1, sw); | |
6516 | } | |
9e55ce06 JH |
6517 | if (b) |
6518 | alt = *b; | |
ffc61ed2 JH |
6519 | } |
6520 | } | |
6521 | ||
9e55ce06 JH |
6522 | if (listsvp) |
6523 | *listsvp = si; | |
6524 | if (altsvp) | |
6525 | *altsvp = alt; | |
ffc61ed2 JH |
6526 | |
6527 | return sw; | |
6528 | } | |
76234dfb | 6529 | #endif |
ffc61ed2 JH |
6530 | |
6531 | /* | |
ba7b4546 | 6532 | - reginclass - determine if a character falls into a character class |
832705d4 | 6533 | |
6698fab5 KW |
6534 | n is the ANYOF regnode |
6535 | p is the target string | |
6536 | lenp is pointer to the maximum number of bytes of how far to go in p | |
6537 | (This is assumed wthout checking to always be at least the current | |
6538 | character's size) | |
6539 | utf8_target tells whether p is in UTF-8. | |
832705d4 | 6540 | |
4b3cda86 KW |
6541 | Returns true if matched; false otherwise. If lenp is not NULL, on return |
6542 | from a successful match, the value it points to will be updated to how many | |
6543 | bytes in p were matched. If there was no match, the value is undefined, | |
6544 | possibly changed from the input. | |
eba1359e | 6545 | |
d5788240 KW |
6546 | Note that this can be a synthetic start class, a combination of various |
6547 | nodes, so things you think might be mutually exclusive, such as locale, | |
6548 | aren't. It can match both locale and non-locale | |
6549 | ||
bbce6d69 | 6550 | */ |
6551 | ||
76e3520e | 6552 | STATIC bool |
f6ad78d8 | 6553 | 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 | 6554 | { |
27da23d5 | 6555 | dVAR; |
a3b680e6 | 6556 | const char flags = ANYOF_FLAGS(n); |
bbce6d69 | 6557 | bool match = FALSE; |
cc07378b | 6558 | UV c = *p; |
f7ab54c6 | 6559 | STRLEN c_len = 0; |
6698fab5 | 6560 | STRLEN maxlen; |
1aa99e6b | 6561 | |
7918f24d NC |
6562 | PERL_ARGS_ASSERT_REGINCLASS; |
6563 | ||
4b3cda86 | 6564 | /* If c is not already the code point, get it */ |
f2ed9b32 | 6565 | if (utf8_target && !UTF8_IS_INVARIANT(c)) { |
f7ab54c6 | 6566 | c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &c_len, |
6182169b KW |
6567 | (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) |
6568 | | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY); | |
6569 | /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for | |
6570 | * UTF8_ALLOW_FFFF */ | |
f7ab54c6 | 6571 | if (c_len == (STRLEN)-1) |
e8a70c6f | 6572 | Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); |
19f67299 | 6573 | } |
a5a291f5 KW |
6574 | else { |
6575 | c_len = 1; | |
6576 | } | |
bbce6d69 | 6577 | |
a5a291f5 KW |
6578 | /* Use passed in max length, or one character if none passed in or less |
6579 | * than one character. And assume will match just one character. This is | |
6580 | * overwritten later if matched more. */ | |
4b3cda86 | 6581 | if (lenp) { |
a5a291f5 KW |
6582 | maxlen = (*lenp > c_len) ? *lenp : c_len; |
6583 | *lenp = c_len; | |
4b3cda86 KW |
6584 | |
6585 | } | |
6586 | else { | |
a5a291f5 | 6587 | maxlen = c_len; |
4b3cda86 KW |
6588 | } |
6589 | ||
7cdde544 KW |
6590 | /* If this character is potentially in the bitmap, check it */ |
6591 | if (c < 256) { | |
ffc61ed2 JH |
6592 | if (ANYOF_BITMAP_TEST(n, c)) |
6593 | match = TRUE; | |
11454c59 KW |
6594 | else if (flags & ANYOF_NON_UTF8_LATIN1_ALL |
6595 | && ! utf8_target | |
6596 | && ! isASCII(c)) | |
6597 | { | |
6598 | match = TRUE; | |
6599 | } | |
a0ed51b3 | 6600 | |
78969a98 KW |
6601 | else if (flags & ANYOF_LOCALE) { |
6602 | PL_reg_flags |= RF_tainted; | |
6603 | ||
6604 | if ((flags & ANYOF_LOC_NONBITMAP_FOLD) | |
6605 | && ANYOF_BITMAP_TEST(n, PL_fold_locale[c])) | |
6606 | { | |
ffc61ed2 | 6607 | match = TRUE; |
78969a98 | 6608 | } |
040aea3a KW |
6609 | else if (ANYOF_CLASS_TEST_ANY_SET(n) && |
6610 | ((ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) || | |
6611 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) || | |
6612 | (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) || | |
6613 | (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) || | |
6614 | (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) || | |
6615 | (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) || | |
6616 | (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) || | |
6617 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) || | |
6618 | (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) || | |
6619 | (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) || | |
6620 | (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) || | |
6621 | (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) || | |
6622 | (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) || | |
6623 | (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) || | |
6624 | (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) || | |
6625 | (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) || | |
6626 | (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) || | |
6627 | (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) || | |
6628 | (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) || | |
6629 | (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) || | |
6630 | (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) || | |
6631 | (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) || | |
6632 | (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) || | |
6633 | (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) || | |
6634 | (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) || | |
6635 | (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) || | |
6636 | (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) || | |
6637 | (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) || | |
6638 | (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) || | |
6639 | (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c)) | |
6640 | ) /* How's that for a conditional? */ | |
78969a98 | 6641 | ) { |
ffc61ed2 JH |
6642 | match = TRUE; |
6643 | } | |
a0ed51b3 | 6644 | } |
a0ed51b3 LW |
6645 | } |
6646 | ||
7cdde544 | 6647 | /* If the bitmap didn't (or couldn't) match, and something outside the |
de87c4fe KW |
6648 | * bitmap could match, try that. Locale nodes specifiy completely the |
6649 | * behavior of code points in the bit map (otherwise, a utf8 target would | |
c613755a | 6650 | * cause them to be treated as Unicode and not locale), except in |
de87c4fe | 6651 | * the very unlikely event when this node is a synthetic start class, which |
c613755a KW |
6652 | * could be a combination of locale and non-locale nodes. So allow locale |
6653 | * to match for the synthetic start class, which will give a false | |
6654 | * positive that will be resolved when the match is done again as not part | |
6655 | * of the synthetic start class */ | |
ef87b810 | 6656 | if (!match) { |
10ee90d2 KW |
6657 | if (utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256) { |
6658 | match = TRUE; /* Everything above 255 matches */ | |
e051a21d | 6659 | } |
6f8d7d0d KW |
6660 | else if (ANYOF_NONBITMAP(n) |
6661 | && ((flags & ANYOF_NONBITMAP_NON_UTF8) | |
6662 | || (utf8_target | |
6663 | && (c >=256 | |
6664 | || (! (flags & ANYOF_LOCALE)) | |
6665 | || (flags & ANYOF_IS_SYNTHETIC))))) | |
ef87b810 | 6666 | { |
7cdde544 KW |
6667 | AV *av; |
6668 | SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av); | |
6669 | ||
6670 | if (sw) { | |
6671 | U8 * utf8_p; | |
6672 | if (utf8_target) { | |
6673 | utf8_p = (U8 *) p; | |
6674 | } else { | |
f56b6394 KW |
6675 | |
6676 | /* Not utf8. Convert as much of the string as available up | |
6677 | * to the limit of how far the (single) character in the | |
6678 | * pattern can possibly match (no need to go further). If | |
6679 | * the node is a straight ANYOF or not folding, it can't | |
6680 | * match more than one. Otherwise, It can match up to how | |
6681 | * far a single char can fold to. Since not utf8, each | |
6682 | * character is a single byte, so the max it can be in | |
6683 | * bytes is the same as the max it can be in characters */ | |
6684 | STRLEN len = (OP(n) == ANYOF | |
6685 | || ! (flags & ANYOF_LOC_NONBITMAP_FOLD)) | |
6686 | ? 1 | |
6687 | : (maxlen < UTF8_MAX_FOLD_CHAR_EXPAND) | |
6688 | ? maxlen | |
6689 | : UTF8_MAX_FOLD_CHAR_EXPAND; | |
7cdde544 KW |
6690 | utf8_p = bytes_to_utf8(p, &len); |
6691 | } | |
f56b6394 | 6692 | |
f7eb7452 | 6693 | if (swash_fetch(sw, utf8_p, TRUE)) |
7cdde544 | 6694 | match = TRUE; |
39065660 | 6695 | else if (flags & ANYOF_LOC_NONBITMAP_FOLD) { |
f56b6394 KW |
6696 | |
6697 | /* Here, we need to test if the fold of the target string | |
8d5d17fa KW |
6698 | * matches. The non-multi char folds have all been moved to |
6699 | * the compilation phase, and the multi-char folds have | |
6700 | * been stored by regcomp into 'av'; we linearly check to | |
6701 | * see if any match the target string (folded). We know | |
6702 | * that the originals were each one character, but we don't | |
6703 | * currently know how many characters/bytes each folded to, | |
6704 | * except we do know that there are small limits imposed by | |
6705 | * Unicode. XXX A performance enhancement would be to have | |
6706 | * regcomp.c store the max number of chars/bytes that are | |
6707 | * in an av entry, as, say the 0th element. Even better | |
6708 | * would be to have a hash of the few characters that can | |
6709 | * start a multi-char fold to the max number of chars of | |
6710 | * those folds. | |
f56b6394 KW |
6711 | * |
6712 | * If there is a match, we will need to advance (if lenp is | |
6713 | * specified) the match pointer in the target string. But | |
6714 | * what we are comparing here isn't that string directly, | |
6715 | * but its fold, whose length may differ from the original. | |
6716 | * As we go along in constructing the fold, therefore, we | |
6717 | * create a map so that we know how many bytes in the | |
6718 | * source to advance given that we have matched a certain | |
6719 | * number of bytes in the fold. This map is stored in | |
6438af90 KW |
6720 | * 'map_fold_len_back'. Let n mean the number of bytes in |
6721 | * the fold of the first character that we are folding. | |
6722 | * Then map_fold_len_back[n] is set to the number of bytes | |
6723 | * in that first character. Similarly let m be the | |
6724 | * corresponding number for the second character to be | |
6725 | * folded. Then map_fold_len_back[n+m] is set to the | |
6726 | * number of bytes occupied by the first two source | |
6727 | * characters. ... */ | |
6728 | U8 map_fold_len_back[UTF8_MAXBYTES_CASE+1] = { 0 }; | |
f56b6394 KW |
6729 | U8 folded[UTF8_MAXBYTES_CASE+1]; |
6730 | STRLEN foldlen = 0; /* num bytes in fold of 1st char */ | |
deba3d96 | 6731 | STRLEN total_foldlen = 0; /* num bytes in fold of all |
6438af90 | 6732 | chars */ |
f56b6394 KW |
6733 | |
6734 | if (OP(n) == ANYOF || maxlen == 1 || ! lenp || ! av) { | |
6735 | ||
6736 | /* Here, only need to fold the first char of the target | |
6438af90 | 6737 | * string. It the source wasn't utf8, is 1 byte long */ |
f56b6394 | 6738 | to_utf8_fold(utf8_p, folded, &foldlen); |
deba3d96 | 6739 | total_foldlen = foldlen; |
6438af90 KW |
6740 | map_fold_len_back[foldlen] = (utf8_target) |
6741 | ? UTF8SKIP(utf8_p) | |
6742 | : 1; | |
f56b6394 KW |
6743 | } |
6744 | else { | |
6745 | ||
6746 | /* Here, need to fold more than the first char. Do so | |
6747 | * up to the limits */ | |
f56b6394 KW |
6748 | U8* source_ptr = utf8_p; /* The source for the fold |
6749 | is the regex target | |
6750 | string */ | |
6751 | U8* folded_ptr = folded; | |
6752 | U8* e = utf8_p + maxlen; /* Can't go beyond last | |
6753 | available byte in the | |
6754 | target string */ | |
6438af90 KW |
6755 | U8 i; |
6756 | for (i = 0; | |
6757 | i < UTF8_MAX_FOLD_CHAR_EXPAND && source_ptr < e; | |
6758 | i++) | |
f56b6394 KW |
6759 | { |
6760 | ||
6761 | /* Fold the next character */ | |
6762 | U8 this_char_folded[UTF8_MAXBYTES_CASE+1]; | |
6763 | STRLEN this_char_foldlen; | |
6764 | to_utf8_fold(source_ptr, | |
6765 | this_char_folded, | |
6766 | &this_char_foldlen); | |
6767 | ||
6768 | /* Bail if it would exceed the byte limit for | |
6769 | * folding a single char. */ | |
6770 | if (this_char_foldlen + folded_ptr - folded > | |
6771 | UTF8_MAXBYTES_CASE) | |
6772 | { | |
6773 | break; | |
6774 | } | |
6775 | ||
6438af90 | 6776 | /* Add the fold of this character */ |
f56b6394 KW |
6777 | Copy(this_char_folded, |
6778 | folded_ptr, | |
6779 | this_char_foldlen, | |
6780 | U8); | |
f56b6394 | 6781 | source_ptr += UTF8SKIP(source_ptr); |
6438af90 | 6782 | folded_ptr += this_char_foldlen; |
deba3d96 | 6783 | total_foldlen = folded_ptr - folded; |
6438af90 KW |
6784 | |
6785 | /* Create map from the number of bytes in the fold | |
6786 | * back to the number of bytes in the source. If | |
6787 | * the source isn't utf8, the byte count is just | |
6788 | * the number of characters so far */ | |
deba3d96 | 6789 | map_fold_len_back[total_foldlen] |
6438af90 KW |
6790 | = (utf8_target) |
6791 | ? source_ptr - utf8_p | |
6792 | : i + 1; | |
f56b6394 KW |
6793 | } |
6794 | *folded_ptr = '\0'; | |
f56b6394 KW |
6795 | } |
6796 | ||
6797 | ||
6798 | /* Do the linear search to see if the fold is in the list | |
8f2655f7 KW |
6799 | * of multi-char folds. */ |
6800 | if (av) { | |
7cdde544 KW |
6801 | I32 i; |
6802 | for (i = 0; i <= av_len(av); i++) { | |
6803 | SV* const sv = *av_fetch(av, i, FALSE); | |
6804 | STRLEN len; | |
6805 | const char * const s = SvPV_const(sv, len); | |
6438af90 | 6806 | |
f9126265 KW |
6807 | if (len <= total_foldlen |
6808 | && memEQ(s, (char*)folded, len) | |
6809 | ||
6810 | /* If 0, means matched a partial char. See | |
6811 | * [perl #90536] */ | |
6812 | && map_fold_len_back[len]) | |
f56b6394 KW |
6813 | { |
6814 | ||
6815 | /* Advance the target string ptr to account for | |
6816 | * this fold, but have to translate from the | |
6817 | * folded length to the corresponding source | |
6438af90 | 6818 | * length. */ |
8f2655f7 KW |
6819 | if (lenp) { |
6820 | *lenp = map_fold_len_back[len]; | |
8f2655f7 | 6821 | } |
7cdde544 KW |
6822 | match = TRUE; |
6823 | break; | |
6824 | } | |
6825 | } | |
6826 | } | |
7cdde544 KW |
6827 | } |
6828 | ||
6829 | /* If we allocated a string above, free it */ | |
6830 | if (! utf8_target) Safefree(utf8_p); | |
6831 | } | |
6832 | } | |
6833 | } | |
6834 | ||
a0ed51b3 LW |
6835 | return (flags & ANYOF_INVERT) ? !match : match; |
6836 | } | |
161b471a | 6837 | |
dfe13c55 | 6838 | STATIC U8 * |
0ce71af7 | 6839 | S_reghop3(U8 *s, I32 off, const U8* lim) |
9041c2e3 | 6840 | { |
6af86488 KW |
6841 | /* return the position 'off' UTF-8 characters away from 's', forward if |
6842 | * 'off' >= 0, backwards if negative. But don't go outside of position | |
6843 | * 'lim', which better be < s if off < 0 */ | |
6844 | ||
97aff369 | 6845 | dVAR; |
7918f24d NC |
6846 | |
6847 | PERL_ARGS_ASSERT_REGHOP3; | |
6848 | ||
a0ed51b3 | 6849 | if (off >= 0) { |
1aa99e6b | 6850 | while (off-- && s < lim) { |
ffc61ed2 | 6851 | /* XXX could check well-formedness here */ |
a0ed51b3 | 6852 | s += UTF8SKIP(s); |
ffc61ed2 | 6853 | } |
a0ed51b3 LW |
6854 | } |
6855 | else { | |
1de06328 YO |
6856 | while (off++ && s > lim) { |
6857 | s--; | |
6858 | if (UTF8_IS_CONTINUED(*s)) { | |
6859 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
6860 | s--; | |
a0ed51b3 | 6861 | } |
1de06328 | 6862 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
6863 | } |
6864 | } | |
6865 | return s; | |
6866 | } | |
161b471a | 6867 | |
f9f4320a YO |
6868 | #ifdef XXX_dmq |
6869 | /* there are a bunch of places where we use two reghop3's that should | |
6870 | be replaced with this routine. but since thats not done yet | |
6871 | we ifdef it out - dmq | |
6872 | */ | |
dfe13c55 | 6873 | STATIC U8 * |
1de06328 YO |
6874 | S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim) |
6875 | { | |
6876 | dVAR; | |
7918f24d NC |
6877 | |
6878 | PERL_ARGS_ASSERT_REGHOP4; | |
6879 | ||
1de06328 YO |
6880 | if (off >= 0) { |
6881 | while (off-- && s < rlim) { | |
6882 | /* XXX could check well-formedness here */ | |
6883 | s += UTF8SKIP(s); | |
6884 | } | |
6885 | } | |
6886 | else { | |
6887 | while (off++ && s > llim) { | |
6888 | s--; | |
6889 | if (UTF8_IS_CONTINUED(*s)) { | |
6890 | while (s > llim && UTF8_IS_CONTINUATION(*s)) | |
6891 | s--; | |
6892 | } | |
6893 | /* XXX could check well-formedness here */ | |
6894 | } | |
6895 | } | |
6896 | return s; | |
6897 | } | |
f9f4320a | 6898 | #endif |
1de06328 YO |
6899 | |
6900 | STATIC U8 * | |
0ce71af7 | 6901 | S_reghopmaybe3(U8* s, I32 off, const U8* lim) |
a0ed51b3 | 6902 | { |
97aff369 | 6903 | dVAR; |
7918f24d NC |
6904 | |
6905 | PERL_ARGS_ASSERT_REGHOPMAYBE3; | |
6906 | ||
a0ed51b3 | 6907 | if (off >= 0) { |
1aa99e6b | 6908 | while (off-- && s < lim) { |
ffc61ed2 | 6909 | /* XXX could check well-formedness here */ |
a0ed51b3 | 6910 | s += UTF8SKIP(s); |
ffc61ed2 | 6911 | } |
a0ed51b3 | 6912 | if (off >= 0) |
3dab1dad | 6913 | return NULL; |
a0ed51b3 LW |
6914 | } |
6915 | else { | |
1de06328 YO |
6916 | while (off++ && s > lim) { |
6917 | s--; | |
6918 | if (UTF8_IS_CONTINUED(*s)) { | |
6919 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
6920 | s--; | |
a0ed51b3 | 6921 | } |
1de06328 | 6922 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
6923 | } |
6924 | if (off <= 0) | |
3dab1dad | 6925 | return NULL; |
a0ed51b3 LW |
6926 | } |
6927 | return s; | |
6928 | } | |
51371543 | 6929 | |
51371543 | 6930 | static void |
acfe0abc | 6931 | restore_pos(pTHX_ void *arg) |
51371543 | 6932 | { |
97aff369 | 6933 | dVAR; |
097eb12c | 6934 | regexp * const rex = (regexp *)arg; |
51371543 GS |
6935 | if (PL_reg_eval_set) { |
6936 | if (PL_reg_oldsaved) { | |
4f639d21 DM |
6937 | rex->subbeg = PL_reg_oldsaved; |
6938 | rex->sublen = PL_reg_oldsavedlen; | |
f8c7b90f | 6939 | #ifdef PERL_OLD_COPY_ON_WRITE |
4f639d21 | 6940 | rex->saved_copy = PL_nrs; |
ed252734 | 6941 | #endif |
07bc277f | 6942 | RXp_MATCH_COPIED_on(rex); |
51371543 GS |
6943 | } |
6944 | PL_reg_magic->mg_len = PL_reg_oldpos; | |
6945 | PL_reg_eval_set = 0; | |
6946 | PL_curpm = PL_reg_oldcurpm; | |
6947 | } | |
6948 | } | |
33b8afdf JH |
6949 | |
6950 | STATIC void | |
6951 | S_to_utf8_substr(pTHX_ register regexp *prog) | |
6952 | { | |
a1cac82e | 6953 | int i = 1; |
7918f24d NC |
6954 | |
6955 | PERL_ARGS_ASSERT_TO_UTF8_SUBSTR; | |
6956 | ||
a1cac82e NC |
6957 | do { |
6958 | if (prog->substrs->data[i].substr | |
6959 | && !prog->substrs->data[i].utf8_substr) { | |
6960 | SV* const sv = newSVsv(prog->substrs->data[i].substr); | |
6961 | prog->substrs->data[i].utf8_substr = sv; | |
6962 | sv_utf8_upgrade(sv); | |
610460f9 | 6963 | if (SvVALID(prog->substrs->data[i].substr)) { |
cffe132d | 6964 | if (SvTAIL(prog->substrs->data[i].substr)) { |
610460f9 NC |
6965 | /* Trim the trailing \n that fbm_compile added last |
6966 | time. */ | |
6967 | SvCUR_set(sv, SvCUR(sv) - 1); | |
6968 | /* Whilst this makes the SV technically "invalid" (as its | |
6969 | buffer is no longer followed by "\0") when fbm_compile() | |
6970 | adds the "\n" back, a "\0" is restored. */ | |
cffe132d NC |
6971 | fbm_compile(sv, FBMcf_TAIL); |
6972 | } else | |
6973 | fbm_compile(sv, 0); | |
610460f9 | 6974 | } |
a1cac82e NC |
6975 | if (prog->substrs->data[i].substr == prog->check_substr) |
6976 | prog->check_utf8 = sv; | |
6977 | } | |
6978 | } while (i--); | |
33b8afdf JH |
6979 | } |
6980 | ||
6981 | STATIC void | |
6982 | S_to_byte_substr(pTHX_ register regexp *prog) | |
6983 | { | |
97aff369 | 6984 | dVAR; |
a1cac82e | 6985 | int i = 1; |
7918f24d NC |
6986 | |
6987 | PERL_ARGS_ASSERT_TO_BYTE_SUBSTR; | |
6988 | ||
a1cac82e NC |
6989 | do { |
6990 | if (prog->substrs->data[i].utf8_substr | |
6991 | && !prog->substrs->data[i].substr) { | |
6992 | SV* sv = newSVsv(prog->substrs->data[i].utf8_substr); | |
6993 | if (sv_utf8_downgrade(sv, TRUE)) { | |
610460f9 | 6994 | if (SvVALID(prog->substrs->data[i].utf8_substr)) { |
cffe132d | 6995 | if (SvTAIL(prog->substrs->data[i].utf8_substr)) { |
610460f9 NC |
6996 | /* Trim the trailing \n that fbm_compile added last |
6997 | time. */ | |
6998 | SvCUR_set(sv, SvCUR(sv) - 1); | |
cffe132d NC |
6999 | fbm_compile(sv, FBMcf_TAIL); |
7000 | } else | |
7001 | fbm_compile(sv, 0); | |
7002 | } | |
a1cac82e NC |
7003 | } else { |
7004 | SvREFCNT_dec(sv); | |
7005 | sv = &PL_sv_undef; | |
7006 | } | |
7007 | prog->substrs->data[i].substr = sv; | |
7008 | if (prog->substrs->data[i].utf8_substr == prog->check_utf8) | |
7009 | prog->check_substr = sv; | |
33b8afdf | 7010 | } |
a1cac82e | 7011 | } while (i--); |
33b8afdf | 7012 | } |
66610fdd RGS |
7013 | |
7014 | /* | |
7015 | * Local variables: | |
7016 | * c-indentation-style: bsd | |
7017 | * c-basic-offset: 4 | |
7018 | * indent-tabs-mode: t | |
7019 | * End: | |
7020 | * | |
37442d52 RGS |
7021 | * ex: set ts=8 sts=4 sw=4 noet: |
7022 | */ |