Commit | Line | Data |
---|---|---|
a0d0e21e LW |
1 | /* regexec.c |
2 | */ | |
3 | ||
4 | /* | |
4ac71550 TC |
5 | * One Ring to rule them all, One Ring to find them |
6 | & | |
7 | * [p.v of _The Lord of the Rings_, opening poem] | |
8 | * [p.50 of _The Lord of the Rings_, I/iii: "The Shadow of the Past"] | |
9 | * [p.254 of _The Lord of the Rings_, II/ii: "The Council of Elrond"] | |
a0d0e21e LW |
10 | */ |
11 | ||
61296642 DM |
12 | /* This file contains functions for executing a regular expression. See |
13 | * also regcomp.c which funnily enough, contains functions for compiling | |
166f8a29 | 14 | * a regular expression. |
e4a054ea DM |
15 | * |
16 | * This file is also copied at build time to ext/re/re_exec.c, where | |
17 | * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT. | |
18 | * This causes the main functions to be compiled under new names and with | |
19 | * debugging support added, which makes "use re 'debug'" work. | |
166f8a29 DM |
20 | */ |
21 | ||
a687059c LW |
22 | /* NOTE: this is derived from Henry Spencer's regexp code, and should not |
23 | * confused with the original package (see point 3 below). Thanks, Henry! | |
24 | */ | |
25 | ||
26 | /* Additional note: this code is very heavily munged from Henry's version | |
27 | * in places. In some spots I've traded clarity for efficiency, so don't | |
28 | * blame Henry for some of the lack of readability. | |
29 | */ | |
30 | ||
e50aee73 AD |
31 | /* The names of the functions have been changed from regcomp and |
32 | * regexec to pregcomp and pregexec in order to avoid conflicts | |
33 | * with the POSIX routines of the same names. | |
34 | */ | |
35 | ||
b9d5759e | 36 | #ifdef PERL_EXT_RE_BUILD |
54df2634 | 37 | #include "re_top.h" |
9041c2e3 | 38 | #endif |
56953603 | 39 | |
a687059c | 40 | /* |
e50aee73 | 41 | * pregcomp and pregexec -- regsub and regerror are not used in perl |
a687059c LW |
42 | * |
43 | * Copyright (c) 1986 by University of Toronto. | |
44 | * Written by Henry Spencer. Not derived from licensed software. | |
45 | * | |
46 | * Permission is granted to anyone to use this software for any | |
47 | * purpose on any computer system, and to redistribute it freely, | |
48 | * subject to the following restrictions: | |
49 | * | |
50 | * 1. The author is not responsible for the consequences of use of | |
51 | * this software, no matter how awful, even if they arise | |
52 | * from defects in it. | |
53 | * | |
54 | * 2. The origin of this software must not be misrepresented, either | |
55 | * by explicit claim or by omission. | |
56 | * | |
57 | * 3. Altered versions must be plainly marked as such, and must not | |
58 | * be misrepresented as being the original software. | |
59 | * | |
60 | **** Alterations to Henry's code are... | |
61 | **** | |
4bb101f2 | 62 | **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, |
1129b882 NC |
63 | **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 |
64 | **** by Larry Wall and others | |
a687059c | 65 | **** |
9ef589d8 LW |
66 | **** You may distribute under the terms of either the GNU General Public |
67 | **** License or the Artistic License, as specified in the README file. | |
a687059c LW |
68 | * |
69 | * Beware that some of this code is subtly aware of the way operator | |
70 | * precedence is structured in regular expressions. Serious changes in | |
71 | * regular-expression syntax might require a total rethink. | |
72 | */ | |
73 | #include "EXTERN.h" | |
864dbfa3 | 74 | #define PERL_IN_REGEXEC_C |
a687059c | 75 | #include "perl.h" |
0f5d15d6 | 76 | |
54df2634 NC |
77 | #ifdef PERL_IN_XSUB_RE |
78 | # include "re_comp.h" | |
79 | #else | |
80 | # include "regcomp.h" | |
81 | #endif | |
a687059c | 82 | |
ef07e810 | 83 | #define RF_tainted 1 /* tainted information used? e.g. locale */ |
c277df42 | 84 | #define RF_warned 2 /* warned about big count? */ |
faec1544 | 85 | |
ab3bbdeb | 86 | #define RF_utf8 8 /* Pattern contains multibyte chars? */ |
a0ed51b3 | 87 | |
f2ed9b32 | 88 | #define UTF_PATTERN ((PL_reg_flags & RF_utf8) != 0) |
ce862d02 | 89 | |
a687059c LW |
90 | #ifndef STATIC |
91 | #define STATIC static | |
92 | #endif | |
93 | ||
7e2509c1 KW |
94 | /* Valid for non-utf8 strings, non-ANYOFV nodes only: avoids the reginclass |
95 | * call if there are no complications: i.e., if everything matchable is | |
96 | * straight forward in the bitmap */ | |
af364d03 KW |
97 | #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) \ |
98 | : ANYOF_BITMAP_TEST(p,*(c))) | |
7d3e948e | 99 | |
c277df42 IZ |
100 | /* |
101 | * Forwards. | |
102 | */ | |
103 | ||
f2ed9b32 | 104 | #define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv)) |
53c4c00c | 105 | #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b) |
a0ed51b3 | 106 | |
3dab1dad YO |
107 | #define HOPc(pos,off) \ |
108 | (char *)(PL_reg_match_utf8 \ | |
52657f30 | 109 | ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \ |
3dab1dad YO |
110 | : (U8*)(pos + off)) |
111 | #define HOPBACKc(pos, off) \ | |
07be1b83 YO |
112 | (char*)(PL_reg_match_utf8\ |
113 | ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \ | |
114 | : (pos - off >= PL_bostr) \ | |
8e11feef | 115 | ? (U8*)pos - off \ |
3dab1dad | 116 | : NULL) |
efb30f32 | 117 | |
e7409c1b | 118 | #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off)) |
1aa99e6b | 119 | #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim)) |
1aa99e6b | 120 | |
20d0b1e9 | 121 | /* these are unrolled below in the CCC_TRY_XXX defined */ |
a9a337f8 KW |
122 | #ifdef EBCDIC |
123 | /* Often 'str' is a hard-coded utf8 string instead of utfebcdic. so just | |
124 | * skip the check on EBCDIC platforms */ | |
125 | # define LOAD_UTF8_CHARCLASS(class,str) LOAD_UTF8_CHARCLASS_NO_CHECK(class) | |
126 | #else | |
127 | # define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ | |
9c4fdda1 RB |
128 | if (!CAT2(PL_utf8_,class)) { \ |
129 | bool ok; \ | |
130 | ENTER; save_re_context(); \ | |
131 | ok=CAT2(is_utf8_,class)((const U8*)str); \ | |
9ac792f4 | 132 | assert(ok); assert(CAT2(PL_utf8_,class)); LEAVE; } } STMT_END |
a9a337f8 | 133 | #endif |
37e2e78e KW |
134 | |
135 | /* Doesn't do an assert to verify that is correct */ | |
136 | #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \ | |
9c4fdda1 | 137 | if (!CAT2(PL_utf8_,class)) { \ |
9d63fa07 | 138 | bool throw_away PERL_UNUSED_DECL; \ |
9c4fdda1 RB |
139 | ENTER; save_re_context(); \ |
140 | throw_away = CAT2(is_utf8_,class)((const U8*)" "); \ | |
141 | LEAVE; } } STMT_END | |
37e2e78e | 142 | |
1a4fad37 AL |
143 | #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a") |
144 | #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0") | |
145 | #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ") | |
51371543 | 146 | |
37e2e78e | 147 | #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \ |
d275fa5e CB |
148 | LOAD_UTF8_CHARCLASS(X_begin, " "); \ |
149 | LOAD_UTF8_CHARCLASS(X_non_hangul, "A"); \ | |
37e2e78e KW |
150 | /* These are utf8 constants, and not utf-ebcdic constants, so the \ |
151 | * assert should likely and hopefully fail on an EBCDIC machine */ \ | |
d275fa5e | 152 | LOAD_UTF8_CHARCLASS(X_extend, "\xcc\x80"); /* U+0300 */ \ |
37e2e78e KW |
153 | \ |
154 | /* No asserts are done for these, in case called on an early \ | |
155 | * Unicode version in which they map to nothing */ \ | |
d275fa5e CB |
156 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_prepend);/* U+0E40 "\xe0\xb9\x80" */ \ |
157 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_L); /* U+1100 "\xe1\x84\x80" */ \ | |
158 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV); /* U+AC00 "\xea\xb0\x80" */ \ | |
159 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LVT); /* U+AC01 "\xea\xb0\x81" */ \ | |
160 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV_LVT_V);/* U+AC01 "\xea\xb0\x81" */\ | |
161 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_T); /* U+11A8 "\xe1\x86\xa8" */ \ | |
37e2e78e | 162 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_V) /* U+1160 "\xe1\x85\xa0" */ |
20d0b1e9 | 163 | |
1dcf4a1b | 164 | #define PLACEHOLDER /* Something for the preprocessor to grab onto */ |
d1eb3177 | 165 | |
ee9a90b8 KW |
166 | /* The actual code for CCC_TRY, which uses several variables from the routine |
167 | * it's callable from. It is designed to be the bulk of a case statement. | |
168 | * FUNC is the macro or function to call on non-utf8 targets that indicate if | |
169 | * nextchr matches the class. | |
170 | * UTF8_TEST is the whole test string to use for utf8 targets | |
171 | * LOAD is what to use to test, and if not present to load in the swash for the | |
172 | * class | |
173 | * POS_OR_NEG is either empty or ! to complement the results of FUNC or | |
174 | * UTF8_TEST test. | |
175 | * The logic is: Fail if we're at the end-of-string; otherwise if the target is | |
176 | * utf8 and a variant, load the swash if necessary and test using the utf8 | |
177 | * test. Advance to the next character if test is ok, otherwise fail; If not | |
178 | * utf8 or an invariant under utf8, use the non-utf8 test, and fail if it | |
179 | * fails, or advance to the next character */ | |
180 | ||
181 | #define _CCC_TRY_CODE(POS_OR_NEG, FUNC, UTF8_TEST, CLASS, STR) \ | |
182 | if (locinput >= PL_regeol) { \ | |
183 | sayNO; \ | |
184 | } \ | |
185 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ | |
186 | LOAD_UTF8_CHARCLASS(CLASS, STR); \ | |
187 | if (POS_OR_NEG (UTF8_TEST)) { \ | |
188 | sayNO; \ | |
189 | } \ | |
190 | locinput += PL_utf8skip[nextchr]; \ | |
191 | nextchr = UCHARAT(locinput); \ | |
192 | break; \ | |
193 | } \ | |
194 | if (POS_OR_NEG (FUNC(nextchr))) { \ | |
195 | sayNO; \ | |
196 | } \ | |
197 | nextchr = UCHARAT(++locinput); \ | |
980866de KW |
198 | break; |
199 | ||
ee9a90b8 KW |
200 | /* Handle the non-locale cases for a character class and its complement. It |
201 | * calls _CCC_TRY_CODE with a ! to complement the test for the character class. | |
202 | * This is because that code fails when the test succeeds, so we want to have | |
203 | * the test fail so that the code succeeds. The swash is stored in a | |
204 | * predictable PL_ place */ | |
cfaf538b KW |
205 | #define _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, \ |
206 | CLASS, STR) \ | |
ee9a90b8 KW |
207 | case NAME: \ |
208 | _CCC_TRY_CODE( !, FUNC, \ | |
209 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ | |
210 | (U8*)locinput, TRUE)), \ | |
211 | CLASS, STR) \ | |
212 | case NNAME: \ | |
1dcf4a1b | 213 | _CCC_TRY_CODE( PLACEHOLDER , FUNC, \ |
ee9a90b8 KW |
214 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ |
215 | (U8*)locinput, TRUE)), \ | |
216 | CLASS, STR) \ | |
217 | ||
218 | /* Generate the case statements for both locale and non-locale character | |
219 | * classes in regmatch for classes that don't have special unicode semantics. | |
220 | * Locales don't use an immediate swash, but an intermediary special locale | |
221 | * function that is called on the pointer to the current place in the input | |
222 | * string. That function will resolve to needing the same swash. One might | |
223 | * think that because we don't know what the locale will match, we shouldn't | |
224 | * check with the swash loading function that it loaded properly; ie, that we | |
225 | * should use LOAD_UTF8_CHARCLASS_NO_CHECK for those, but what is passed to the | |
226 | * regular LOAD_UTF8_CHARCLASS is in non-locale terms, and so locale is | |
227 | * irrelevant here */ | |
228 | #define CCC_TRY(NAME, NNAME, FUNC, \ | |
229 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
cfaf538b | 230 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 KW |
231 | CLASS, STR) \ |
232 | case NAMEL: \ | |
233 | PL_reg_flags |= RF_tainted; \ | |
234 | _CCC_TRY_CODE( !, LCFUNC, LCFUNC_utf8((U8*)locinput), CLASS, STR) \ | |
235 | case NNAMEL: \ | |
236 | PL_reg_flags |= RF_tainted; \ | |
1dcf4a1b KW |
237 | _CCC_TRY_CODE( PLACEHOLDER, LCFUNC, LCFUNC_utf8((U8*)locinput), \ |
238 | CLASS, STR) \ | |
cfaf538b KW |
239 | case NAMEA: \ |
240 | if (locinput >= PL_regeol || ! FUNCA(nextchr)) { \ | |
241 | sayNO; \ | |
242 | } \ | |
243 | /* Matched a utf8-invariant, so don't have to worry about utf8 */ \ | |
244 | nextchr = UCHARAT(++locinput); \ | |
245 | break; \ | |
246 | case NNAMEA: \ | |
247 | if (locinput >= PL_regeol || FUNCA(nextchr)) { \ | |
248 | sayNO; \ | |
249 | } \ | |
250 | if (utf8_target) { \ | |
251 | locinput += PL_utf8skip[nextchr]; \ | |
252 | nextchr = UCHARAT(locinput); \ | |
253 | } \ | |
254 | else { \ | |
255 | nextchr = UCHARAT(++locinput); \ | |
256 | } \ | |
257 | break; \ | |
ee9a90b8 KW |
258 | /* Generate the non-locale cases */ \ |
259 | _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, CLASS, STR) | |
260 | ||
261 | /* This is like CCC_TRY, but has an extra set of parameters for generating case | |
262 | * statements to handle separate Unicode semantics nodes */ | |
263 | #define CCC_TRY_U(NAME, NNAME, FUNC, \ | |
264 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
265 | NAMEU, NNAMEU, FUNCU, \ | |
cfaf538b | 266 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 | 267 | CLASS, STR) \ |
cfaf538b KW |
268 | CCC_TRY(NAME, NNAME, FUNC, \ |
269 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
270 | NAMEA, NNAMEA, FUNCA, \ | |
271 | CLASS, STR) \ | |
ee9a90b8 | 272 | _CCC_TRY_NONLOCALE(NAMEU, NNAMEU, FUNCU, CLASS, STR) |
d1eb3177 | 273 | |
3dab1dad YO |
274 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
275 | ||
5f80c4cf | 276 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
c35dcbe2 YO |
277 | /* it would be nice to rework regcomp.sym to generate this stuff. sigh |
278 | * | |
279 | * NOTE that *nothing* that affects backtracking should be in here, specifically | |
280 | * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a | |
281 | * node that is in between two EXACT like nodes when ascertaining what the required | |
282 | * "follow" character is. This should probably be moved to regex compile time | |
283 | * although it may be done at run time beause of the REF possibility - more | |
284 | * investigation required. -- demerphq | |
285 | */ | |
3e901dc0 YO |
286 | #define JUMPABLE(rn) ( \ |
287 | OP(rn) == OPEN || \ | |
288 | (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \ | |
289 | OP(rn) == EVAL || \ | |
cca55fe3 JP |
290 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
291 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
d1c771f5 | 292 | OP(rn) == KEEPS || \ |
3dab1dad | 293 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 | 294 | ) |
ee9b8eae | 295 | #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT) |
e2d8ce26 | 296 | |
ee9b8eae YO |
297 | #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF ) |
298 | ||
299 | #if 0 | |
300 | /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so | |
301 | we don't need this definition. */ | |
302 | #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) | |
fab2782b | 303 | #define IS_TEXTF(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFU_TRICKYFOLD || OP(rn)==EXACTFA || OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF ) |
ee9b8eae YO |
304 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) |
305 | ||
306 | #else | |
307 | /* ... so we use this as its faster. */ | |
308 | #define IS_TEXT(rn) ( OP(rn)==EXACT ) | |
fab2782b | 309 | #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFU_TRICKYFOLD || OP(rn) == EXACTFA) |
ee9b8eae YO |
310 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) |
311 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) | |
312 | ||
313 | #endif | |
e2d8ce26 | 314 | |
a84d97b6 HS |
315 | /* |
316 | Search for mandatory following text node; for lookahead, the text must | |
317 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
318 | */ | |
cca55fe3 | 319 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
320 | while (JUMPABLE(rn)) { \ |
321 | const OPCODE type = OP(rn); \ | |
322 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 323 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 324 | else if (type == PLUS) \ |
cca55fe3 | 325 | rn = NEXTOPER(rn); \ |
3dab1dad | 326 | else if (type == IFMATCH) \ |
a84d97b6 | 327 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 328 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 329 | } \ |
5f80c4cf | 330 | } STMT_END |
74750237 | 331 | |
c476f425 | 332 | |
acfe0abc | 333 | static void restore_pos(pTHX_ void *arg); |
51371543 | 334 | |
87c0511b | 335 | #define REGCP_PAREN_ELEMS 3 |
f067efbf | 336 | #define REGCP_OTHER_ELEMS 3 |
e0fa7e2b | 337 | #define REGCP_FRAME_ELEMS 1 |
620d5b66 NC |
338 | /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and |
339 | * are needed for the regexp context stack bookkeeping. */ | |
340 | ||
76e3520e | 341 | STATIC CHECKPOINT |
b93070ed | 342 | S_regcppush(pTHX_ const regexp *rex, I32 parenfloor) |
a0d0e21e | 343 | { |
97aff369 | 344 | dVAR; |
a3b680e6 | 345 | const int retval = PL_savestack_ix; |
a3b680e6 | 346 | const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; |
e0fa7e2b NC |
347 | const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS; |
348 | const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT; | |
87c0511b | 349 | I32 p; |
40a82448 | 350 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 351 | |
b93070ed DM |
352 | PERL_ARGS_ASSERT_REGCPPUSH; |
353 | ||
e49a9654 | 354 | if (paren_elems_to_push < 0) |
5637ef5b NC |
355 | Perl_croak(aTHX_ "panic: paren_elems_to_push, %i < 0", |
356 | paren_elems_to_push); | |
e49a9654 | 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 | |
495f47a5 DM |
365 | DEBUG_BUFFERS_r( |
366 | if ((int)PL_regsize > (int)parenfloor) | |
367 | PerlIO_printf(Perl_debug_log, | |
368 | "rex=0x%"UVxf" offs=0x%"UVxf": saving capture indices:\n", | |
369 | PTR2UV(rex), | |
370 | PTR2UV(rex->offs) | |
371 | ); | |
372 | ); | |
87c0511b | 373 | for (p = parenfloor+1; p <= (I32)PL_regsize; p++) { |
b1ce53c5 | 374 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
b93070ed DM |
375 | SSPUSHINT(rex->offs[p].end); |
376 | SSPUSHINT(rex->offs[p].start); | |
1ca2007e | 377 | SSPUSHINT(rex->offs[p].start_tmp); |
e7707071 | 378 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
495f47a5 DM |
379 | " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"\n", |
380 | (UV)p, | |
381 | (IV)rex->offs[p].start, | |
382 | (IV)rex->offs[p].start_tmp, | |
383 | (IV)rex->offs[p].end | |
40a82448 | 384 | )); |
a0d0e21e | 385 | } |
b1ce53c5 | 386 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
3280af22 | 387 | SSPUSHINT(PL_regsize); |
b93070ed DM |
388 | SSPUSHINT(rex->lastparen); |
389 | SSPUSHINT(rex->lastcloseparen); | |
e0fa7e2b | 390 | SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */ |
41123dfd | 391 | |
a0d0e21e LW |
392 | return retval; |
393 | } | |
394 | ||
c277df42 | 395 | /* These are needed since we do not localize EVAL nodes: */ |
ab3bbdeb YO |
396 | #define REGCP_SET(cp) \ |
397 | DEBUG_STATE_r( \ | |
ab3bbdeb | 398 | PerlIO_printf(Perl_debug_log, \ |
e4f74956 | 399 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
ab3bbdeb YO |
400 | (IV)PL_savestack_ix)); \ |
401 | cp = PL_savestack_ix | |
c3464db5 | 402 | |
ab3bbdeb | 403 | #define REGCP_UNWIND(cp) \ |
e4f74956 | 404 | DEBUG_STATE_r( \ |
ab3bbdeb | 405 | if (cp != PL_savestack_ix) \ |
e4f74956 YO |
406 | PerlIO_printf(Perl_debug_log, \ |
407 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ | |
ab3bbdeb YO |
408 | (IV)(cp), (IV)PL_savestack_ix)); \ |
409 | regcpblow(cp) | |
c277df42 | 410 | |
f067efbf | 411 | STATIC void |
b93070ed | 412 | S_regcppop(pTHX_ regexp *rex) |
a0d0e21e | 413 | { |
97aff369 | 414 | dVAR; |
e0fa7e2b | 415 | UV i; |
87c0511b | 416 | U32 paren; |
a3621e74 YO |
417 | GET_RE_DEBUG_FLAGS_DECL; |
418 | ||
7918f24d NC |
419 | PERL_ARGS_ASSERT_REGCPPOP; |
420 | ||
b1ce53c5 | 421 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
c6bf6a65 | 422 | i = SSPOPUV; |
e0fa7e2b NC |
423 | assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
424 | i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */ | |
b93070ed DM |
425 | rex->lastcloseparen = SSPOPINT; |
426 | rex->lastparen = SSPOPINT; | |
3280af22 | 427 | PL_regsize = SSPOPINT; |
b1ce53c5 | 428 | |
620d5b66 | 429 | i -= REGCP_OTHER_ELEMS; |
b1ce53c5 | 430 | /* Now restore the parentheses context. */ |
495f47a5 DM |
431 | DEBUG_BUFFERS_r( |
432 | if (i || rex->lastparen + 1 <= rex->nparens) | |
433 | PerlIO_printf(Perl_debug_log, | |
434 | "rex=0x%"UVxf" offs=0x%"UVxf": restoring capture indices to:\n", | |
435 | PTR2UV(rex), | |
436 | PTR2UV(rex->offs) | |
437 | ); | |
438 | ); | |
87c0511b | 439 | paren = PL_regsize; |
620d5b66 | 440 | for ( ; i > 0; i -= REGCP_PAREN_ELEMS) { |
1df70142 | 441 | I32 tmps; |
1ca2007e | 442 | rex->offs[paren].start_tmp = SSPOPINT; |
b93070ed | 443 | rex->offs[paren].start = SSPOPINT; |
cf93c79d | 444 | tmps = SSPOPINT; |
b93070ed DM |
445 | if (paren <= rex->lastparen) |
446 | rex->offs[paren].end = tmps; | |
495f47a5 DM |
447 | DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log, |
448 | " \\%"UVuf": %"IVdf"(%"IVdf")..%"IVdf"%s\n", | |
449 | (UV)paren, | |
450 | (IV)rex->offs[paren].start, | |
451 | (IV)rex->offs[paren].start_tmp, | |
452 | (IV)rex->offs[paren].end, | |
453 | (paren > rex->lastparen ? "(skipped)" : "")); | |
c277df42 | 454 | ); |
87c0511b | 455 | paren--; |
a0d0e21e | 456 | } |
daf18116 | 457 | #if 1 |
dafc8851 JH |
458 | /* It would seem that the similar code in regtry() |
459 | * already takes care of this, and in fact it is in | |
460 | * a better location to since this code can #if 0-ed out | |
461 | * but the code in regtry() is needed or otherwise tests | |
462 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
463 | * (as of patchlevel 7877) will fail. Then again, |
464 | * this code seems to be necessary or otherwise | |
225593e1 DM |
465 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ |
466 | * --jhi updated by dapm */ | |
b93070ed | 467 | for (i = rex->lastparen + 1; i <= rex->nparens; i++) { |
097eb12c | 468 | if (i > PL_regsize) |
b93070ed DM |
469 | rex->offs[i].start = -1; |
470 | rex->offs[i].end = -1; | |
495f47a5 DM |
471 | DEBUG_BUFFERS_r( PerlIO_printf(Perl_debug_log, |
472 | " \\%"UVuf": %s ..-1 undeffing\n", | |
473 | (UV)i, | |
474 | (i > PL_regsize) ? "-1" : " " | |
475 | )); | |
a0d0e21e | 476 | } |
dafc8851 | 477 | #endif |
a0d0e21e LW |
478 | } |
479 | ||
02db2b7b | 480 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 481 | |
a687059c | 482 | /* |
e50aee73 | 483 | * pregexec and friends |
a687059c LW |
484 | */ |
485 | ||
76234dfb | 486 | #ifndef PERL_IN_XSUB_RE |
a687059c | 487 | /* |
c277df42 | 488 | - pregexec - match a regexp against a string |
a687059c | 489 | */ |
c277df42 | 490 | I32 |
49d7dfbc | 491 | Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend, |
c3464db5 | 492 | char *strbeg, I32 minend, SV *screamer, U32 nosave) |
c277df42 IZ |
493 | /* strend: pointer to null at end of string */ |
494 | /* strbeg: real beginning of string */ | |
495 | /* minend: end of match must be >=minend after stringarg. */ | |
496 | /* nosave: For optimizations. */ | |
497 | { | |
7918f24d NC |
498 | PERL_ARGS_ASSERT_PREGEXEC; |
499 | ||
c277df42 | 500 | return |
9041c2e3 | 501 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
502 | nosave ? 0 : REXEC_COPY_STR); |
503 | } | |
76234dfb | 504 | #endif |
22e551b9 | 505 | |
9041c2e3 | 506 | /* |
cad2e5aa JH |
507 | * Need to implement the following flags for reg_anch: |
508 | * | |
509 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
510 | * USE_INTUIT_ML | |
511 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
512 | * INTUIT_AUTORITATIVE_ML | |
513 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
514 | * INTUIT_ONCE_ML | |
515 | * | |
516 | * Another flag for this function: SECOND_TIME (so that float substrs | |
517 | * with giant delta may be not rechecked). | |
518 | */ | |
519 | ||
520 | /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ | |
521 | ||
3f7c398e | 522 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
523 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
524 | ||
525 | /* XXXX We assume that strpos is strbeg unless sv. */ | |
526 | ||
6eb5f6b9 JH |
527 | /* XXXX Some places assume that there is a fixed substring. |
528 | An update may be needed if optimizer marks as "INTUITable" | |
529 | RExen without fixed substrings. Similarly, it is assumed that | |
530 | lengths of all the strings are no more than minlen, thus they | |
531 | cannot come from lookahead. | |
40d049e4 YO |
532 | (Or minlen should take into account lookahead.) |
533 | NOTE: Some of this comment is not correct. minlen does now take account | |
534 | of lookahead/behind. Further research is required. -- demerphq | |
535 | ||
536 | */ | |
6eb5f6b9 | 537 | |
2c2d71f5 JH |
538 | /* A failure to find a constant substring means that there is no need to make |
539 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
540 | finding a substring too deep into the string means that less calls to | |
30944b6d IZ |
541 | regtry() should be needed. |
542 | ||
543 | REx compiler's optimizer found 4 possible hints: | |
544 | a) Anchored substring; | |
545 | b) Fixed substring; | |
546 | c) Whether we are anchored (beginning-of-line or \G); | |
486ec47a | 547 | d) First node (of those at offset 0) which may distinguish positions; |
6eb5f6b9 | 548 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
549 | string which does not contradict any of them. |
550 | */ | |
2c2d71f5 | 551 | |
6eb5f6b9 JH |
552 | /* Most of decisions we do here should have been done at compile time. |
553 | The nodes of the REx which we used for the search should have been | |
554 | deleted from the finite automaton. */ | |
555 | ||
cad2e5aa | 556 | char * |
288b8c02 | 557 | Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, |
9f61653a | 558 | char *strend, const U32 flags, re_scream_pos_data *data) |
cad2e5aa | 559 | { |
97aff369 | 560 | dVAR; |
288b8c02 | 561 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
b7953727 | 562 | register I32 start_shift = 0; |
cad2e5aa | 563 | /* Should be nonnegative! */ |
b7953727 | 564 | register I32 end_shift = 0; |
2c2d71f5 JH |
565 | register char *s; |
566 | register SV *check; | |
a1933d95 | 567 | char *strbeg; |
cad2e5aa | 568 | char *t; |
f2ed9b32 | 569 | const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 570 | I32 ml_anch; |
bd61b366 SS |
571 | register char *other_last = NULL; /* other substr checked before this */ |
572 | char *check_at = NULL; /* check substr found at this pos */ | |
d8080198 | 573 | char *checked_upto = NULL; /* how far into the string we have already checked using find_byclass*/ |
bbe252da | 574 | const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; |
f8fc2ecf | 575 | RXi_GET_DECL(prog,progi); |
30944b6d | 576 | #ifdef DEBUGGING |
890ce7af | 577 | const char * const i_strpos = strpos; |
30944b6d | 578 | #endif |
a3621e74 YO |
579 | GET_RE_DEBUG_FLAGS_DECL; |
580 | ||
7918f24d | 581 | PERL_ARGS_ASSERT_RE_INTUIT_START; |
c33e64f0 FC |
582 | PERL_UNUSED_ARG(flags); |
583 | PERL_UNUSED_ARG(data); | |
7918f24d | 584 | |
f2ed9b32 | 585 | RX_MATCH_UTF8_set(rx,utf8_target); |
cad2e5aa | 586 | |
3c8556c3 | 587 | if (RX_UTF8(rx)) { |
b8d68ded JH |
588 | PL_reg_flags |= RF_utf8; |
589 | } | |
ab3bbdeb | 590 | DEBUG_EXECUTE_r( |
f2ed9b32 | 591 | debug_start_match(rx, utf8_target, strpos, strend, |
1de06328 YO |
592 | sv ? "Guessing start of match in sv for" |
593 | : "Guessing start of match in string for"); | |
2a782b5b | 594 | ); |
cad2e5aa | 595 | |
c344f387 JH |
596 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
597 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 598 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 599 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 600 | goto fail; |
2c2d71f5 | 601 | } |
1de06328 | 602 | |
a1933d95 | 603 | strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; |
1aa99e6b | 604 | PL_regeol = strend; |
f2ed9b32 | 605 | if (utf8_target) { |
33b8afdf JH |
606 | if (!prog->check_utf8 && prog->check_substr) |
607 | to_utf8_substr(prog); | |
608 | check = prog->check_utf8; | |
609 | } else { | |
610 | if (!prog->check_substr && prog->check_utf8) | |
611 | to_byte_substr(prog); | |
612 | check = prog->check_substr; | |
613 | } | |
1de06328 | 614 | if (check == &PL_sv_undef) { |
a3621e74 | 615 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
1de06328 | 616 | "Non-utf8 string cannot match utf8 check string\n")); |
33b8afdf JH |
617 | goto fail; |
618 | } | |
bbe252da YO |
619 | if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */ |
620 | ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) | |
621 | || ( (prog->extflags & RXf_ANCH_BOL) | |
7fba1cd6 | 622 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 623 | |
7e25d62c | 624 | if (!ml_anch) { |
bbe252da YO |
625 | if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */ |
626 | && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ | |
3f7c398e | 627 | /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ |
7e25d62c JH |
628 | && sv && !SvROK(sv) |
629 | && (strpos != strbeg)) { | |
a3621e74 | 630 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
631 | goto fail; |
632 | } | |
633 | if (prog->check_offset_min == prog->check_offset_max && | |
bbe252da | 634 | !(prog->extflags & RXf_CANY_SEEN)) { |
2c2d71f5 | 635 | /* Substring at constant offset from beg-of-str... */ |
cad2e5aa JH |
636 | I32 slen; |
637 | ||
1aa99e6b | 638 | s = HOP3c(strpos, prog->check_offset_min, strend); |
1de06328 | 639 | |
653099ff GS |
640 | if (SvTAIL(check)) { |
641 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 642 | |
9041c2e3 | 643 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 644 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 645 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 646 | goto fail_finish; |
cad2e5aa JH |
647 | } |
648 | /* Now should match s[0..slen-2] */ | |
649 | slen--; | |
3f7c398e | 650 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 651 | || (slen > 1 |
3f7c398e | 652 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 653 | report_neq: |
a3621e74 | 654 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
655 | goto fail_finish; |
656 | } | |
cad2e5aa | 657 | } |
3f7c398e | 658 | else if (*SvPVX_const(check) != *s |
653099ff | 659 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 660 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 661 | goto report_neq; |
c315bfe8 | 662 | check_at = s; |
2c2d71f5 | 663 | goto success_at_start; |
7e25d62c | 664 | } |
cad2e5aa | 665 | } |
2c2d71f5 | 666 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 667 | s = strpos; |
2c2d71f5 | 668 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
1de06328 YO |
669 | end_shift = prog->check_end_shift; |
670 | ||
2c2d71f5 | 671 | if (!ml_anch) { |
a3b680e6 | 672 | const I32 end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 673 | - (SvTAIL(check) != 0); |
a3b680e6 | 674 | const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
675 | |
676 | if (end_shift < eshift) | |
677 | end_shift = eshift; | |
678 | } | |
cad2e5aa | 679 | } |
2c2d71f5 | 680 | else { /* Can match at random position */ |
cad2e5aa JH |
681 | ml_anch = 0; |
682 | s = strpos; | |
1de06328 YO |
683 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
684 | end_shift = prog->check_end_shift; | |
685 | ||
686 | /* end shift should be non negative here */ | |
cad2e5aa JH |
687 | } |
688 | ||
bcdf7404 | 689 | #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 690 | if (end_shift < 0) |
1de06328 | 691 | Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", |
220fc49f | 692 | (IV)end_shift, RX_PRECOMP(prog)); |
2c2d71f5 JH |
693 | #endif |
694 | ||
2c2d71f5 JH |
695 | restart: |
696 | /* Find a possible match in the region s..strend by looking for | |
697 | the "check" substring in the region corrected by start/end_shift. */ | |
1de06328 YO |
698 | |
699 | { | |
700 | I32 srch_start_shift = start_shift; | |
701 | I32 srch_end_shift = end_shift; | |
c33e64f0 FC |
702 | U8* start_point; |
703 | U8* end_point; | |
1de06328 YO |
704 | if (srch_start_shift < 0 && strbeg - s > srch_start_shift) { |
705 | srch_end_shift -= ((strbeg - s) - srch_start_shift); | |
706 | srch_start_shift = strbeg - s; | |
707 | } | |
6bda09f9 | 708 | DEBUG_OPTIMISE_MORE_r({ |
1de06328 YO |
709 | PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n", |
710 | (IV)prog->check_offset_min, | |
711 | (IV)srch_start_shift, | |
712 | (IV)srch_end_shift, | |
713 | (IV)prog->check_end_shift); | |
714 | }); | |
715 | ||
bbe252da | 716 | if (prog->extflags & RXf_CANY_SEEN) { |
1de06328 YO |
717 | start_point= (U8*)(s + srch_start_shift); |
718 | end_point= (U8*)(strend - srch_end_shift); | |
719 | } else { | |
720 | start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); | |
721 | end_point= HOP3(strend, -srch_end_shift, strbeg); | |
722 | } | |
6bda09f9 | 723 | DEBUG_OPTIMISE_MORE_r({ |
56570a2c | 724 | PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", |
1de06328 | 725 | (int)(end_point - start_point), |
fc8cd66c | 726 | (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), |
1de06328 YO |
727 | start_point); |
728 | }); | |
729 | ||
730 | s = fbm_instr( start_point, end_point, | |
7fba1cd6 | 731 | check, multiline ? FBMrf_MULTILINE : 0); |
1de06328 | 732 | } |
cad2e5aa JH |
733 | /* Update the count-of-usability, remove useless subpatterns, |
734 | unshift s. */ | |
2c2d71f5 | 735 | |
ab3bbdeb | 736 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 737 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
738 | SvPVX_const(check), RE_SV_DUMPLEN(check), 30); |
739 | PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", | |
2c2d71f5 | 740 | (s ? "Found" : "Did not find"), |
f2ed9b32 | 741 | (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) |
ab3bbdeb YO |
742 | ? "anchored" : "floating"), |
743 | quoted, | |
744 | RE_SV_TAIL(check), | |
745 | (s ? " at offset " : "...\n") ); | |
746 | }); | |
2c2d71f5 JH |
747 | |
748 | if (!s) | |
749 | goto fail_finish; | |
2c2d71f5 | 750 | /* Finish the diagnostic message */ |
a3621e74 | 751 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 | 752 | |
1de06328 YO |
753 | /* XXX dmq: first branch is for positive lookbehind... |
754 | Our check string is offset from the beginning of the pattern. | |
755 | So we need to do any stclass tests offset forward from that | |
756 | point. I think. :-( | |
757 | */ | |
758 | ||
759 | ||
760 | ||
761 | check_at=s; | |
762 | ||
763 | ||
2c2d71f5 JH |
764 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. |
765 | Start with the other substr. | |
766 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 767 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
768 | *always* match. Probably should be marked during compile... |
769 | Probably it is right to do no SCREAM here... | |
770 | */ | |
771 | ||
f2ed9b32 | 772 | if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8) |
1de06328 YO |
773 | : (prog->float_substr && prog->anchored_substr)) |
774 | { | |
30944b6d | 775 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
776 | /* XXXX May be hopelessly wrong for UTF... */ |
777 | if (!other_last) | |
6eb5f6b9 | 778 | other_last = strpos; |
f2ed9b32 | 779 | if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
780 | do_other_anchored: |
781 | { | |
890ce7af AL |
782 | char * const last = HOP3c(s, -start_shift, strbeg); |
783 | char *last1, *last2; | |
be8e71aa | 784 | char * const saved_s = s; |
33b8afdf | 785 | SV* must; |
2c2d71f5 | 786 | |
2c2d71f5 JH |
787 | t = s - prog->check_offset_max; |
788 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
f2ed9b32 | 789 | && (!utf8_target |
0ce71af7 | 790 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 791 | && t > strpos))) |
6f207bd3 | 792 | NOOP; |
2c2d71f5 JH |
793 | else |
794 | t = strpos; | |
1aa99e6b | 795 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
796 | if (t < other_last) /* These positions already checked */ |
797 | t = other_last; | |
1aa99e6b | 798 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
799 | if (last < last1) |
800 | last1 = last; | |
1de06328 YO |
801 | /* XXXX It is not documented what units *_offsets are in. |
802 | We assume bytes, but this is clearly wrong. | |
803 | Meaning this code needs to be carefully reviewed for errors. | |
804 | dmq. | |
805 | */ | |
806 | ||
2c2d71f5 | 807 | /* On end-of-str: see comment below. */ |
f2ed9b32 | 808 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; |
33b8afdf JH |
809 | if (must == &PL_sv_undef) { |
810 | s = (char*)NULL; | |
1de06328 | 811 | DEBUG_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
812 | } |
813 | else | |
814 | s = fbm_instr( | |
815 | (unsigned char*)t, | |
816 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
817 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
818 | must, | |
7fba1cd6 | 819 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 820 | ); |
ab3bbdeb | 821 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 822 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
823 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
824 | PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", | |
2c2d71f5 | 825 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
826 | quoted, RE_SV_TAIL(must)); |
827 | }); | |
828 | ||
829 | ||
2c2d71f5 JH |
830 | if (!s) { |
831 | if (last1 >= last2) { | |
a3621e74 | 832 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
833 | ", giving up...\n")); |
834 | goto fail_finish; | |
835 | } | |
a3621e74 | 836 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 837 | ", trying floating at offset %ld...\n", |
be8e71aa | 838 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
839 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
840 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
841 | goto restart; |
842 | } | |
843 | else { | |
a3621e74 | 844 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 845 | (long)(s - i_strpos))); |
1aa99e6b IH |
846 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
847 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 848 | s = saved_s; |
2c2d71f5 JH |
849 | if (t == strpos) |
850 | goto try_at_start; | |
2c2d71f5 JH |
851 | goto try_at_offset; |
852 | } | |
30944b6d | 853 | } |
2c2d71f5 JH |
854 | } |
855 | else { /* Take into account the floating substring. */ | |
33b8afdf | 856 | char *last, *last1; |
be8e71aa | 857 | char * const saved_s = s; |
33b8afdf JH |
858 | SV* must; |
859 | ||
860 | t = HOP3c(s, -start_shift, strbeg); | |
861 | last1 = last = | |
862 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
863 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
864 | last = HOP3c(t, prog->float_max_offset, strend); | |
865 | s = HOP3c(t, prog->float_min_offset, strend); | |
866 | if (s < other_last) | |
867 | s = other_last; | |
2c2d71f5 | 868 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
f2ed9b32 | 869 | must = utf8_target ? prog->float_utf8 : prog->float_substr; |
33b8afdf JH |
870 | /* fbm_instr() takes into account exact value of end-of-str |
871 | if the check is SvTAIL(ed). Since false positives are OK, | |
872 | and end-of-str is not later than strend we are OK. */ | |
873 | if (must == &PL_sv_undef) { | |
874 | s = (char*)NULL; | |
1de06328 | 875 | DEBUG_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
876 | } |
877 | else | |
2c2d71f5 | 878 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
879 | (unsigned char*)last + SvCUR(must) |
880 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 881 | must, multiline ? FBMrf_MULTILINE : 0); |
ab3bbdeb | 882 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 883 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
884 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
885 | PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", | |
33b8afdf | 886 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
887 | quoted, RE_SV_TAIL(must)); |
888 | }); | |
33b8afdf JH |
889 | if (!s) { |
890 | if (last1 == last) { | |
a3621e74 | 891 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
892 | ", giving up...\n")); |
893 | goto fail_finish; | |
2c2d71f5 | 894 | } |
a3621e74 | 895 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 896 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 897 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
898 | other_last = last; |
899 | s = HOP3c(t, 1, strend); | |
900 | goto restart; | |
901 | } | |
902 | else { | |
a3621e74 | 903 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
904 | (long)(s - i_strpos))); |
905 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 906 | s = saved_s; |
33b8afdf JH |
907 | if (t == strpos) |
908 | goto try_at_start; | |
909 | goto try_at_offset; | |
910 | } | |
2c2d71f5 | 911 | } |
cad2e5aa | 912 | } |
2c2d71f5 | 913 | |
1de06328 | 914 | |
9ef43ace | 915 | t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); |
1de06328 | 916 | |
6bda09f9 | 917 | DEBUG_OPTIMISE_MORE_r( |
1de06328 YO |
918 | PerlIO_printf(Perl_debug_log, |
919 | "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", | |
920 | (IV)prog->check_offset_min, | |
921 | (IV)prog->check_offset_max, | |
922 | (IV)(s-strpos), | |
923 | (IV)(t-strpos), | |
924 | (IV)(t-s), | |
925 | (IV)(strend-strpos) | |
926 | ) | |
927 | ); | |
928 | ||
2c2d71f5 | 929 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
f2ed9b32 | 930 | && (!utf8_target |
9ef43ace | 931 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) |
1de06328 YO |
932 | && t > strpos))) |
933 | { | |
2c2d71f5 JH |
934 | /* Fixed substring is found far enough so that the match |
935 | cannot start at strpos. */ | |
936 | try_at_offset: | |
cad2e5aa | 937 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
938 | /* Eventually fbm_*() should handle this, but often |
939 | anchored_offset is not 0, so this check will not be wasted. */ | |
940 | /* XXXX In the code below we prefer to look for "^" even in | |
941 | presence of anchored substrings. And we search even | |
942 | beyond the found float position. These pessimizations | |
943 | are historical artefacts only. */ | |
944 | find_anchor: | |
2c2d71f5 | 945 | while (t < strend - prog->minlen) { |
cad2e5aa | 946 | if (*t == '\n') { |
4ee3650e | 947 | if (t < check_at - prog->check_offset_min) { |
f2ed9b32 | 948 | if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
949 | /* Since we moved from the found position, |
950 | we definitely contradict the found anchored | |
30944b6d IZ |
951 | substr. Due to the above check we do not |
952 | contradict "check" substr. | |
953 | Thus we can arrive here only if check substr | |
954 | is float. Redo checking for "other"=="fixed". | |
955 | */ | |
9041c2e3 | 956 | strpos = t + 1; |
a3621e74 | 957 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 958 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
959 | goto do_other_anchored; |
960 | } | |
4ee3650e GS |
961 | /* We don't contradict the found floating substring. */ |
962 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 963 | s = t + 1; |
a3621e74 | 964 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 965 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
966 | goto set_useful; |
967 | } | |
4ee3650e GS |
968 | /* Position contradicts check-string */ |
969 | /* XXXX probably better to look for check-string | |
970 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 971 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 972 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 973 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
974 | goto restart; |
975 | } | |
976 | t++; | |
977 | } | |
a3621e74 | 978 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 979 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 980 | goto fail_finish; |
cad2e5aa | 981 | } |
f5952150 | 982 | else { |
a3621e74 | 983 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 984 | PL_colors[0], PL_colors[1])); |
f5952150 | 985 | } |
cad2e5aa JH |
986 | s = t; |
987 | set_useful: | |
f2ed9b32 | 988 | ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
989 | } |
990 | else { | |
f5952150 | 991 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 992 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
993 | unless it was an MBOL and we are not after MBOL, |
994 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
995 | try_at_start: |
996 | /* Even in this situation we may use MBOL flag if strpos is offset | |
997 | wrt the start of the string. */ | |
05b4157f | 998 | if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ |
a1933d95 | 999 | && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d | 1000 | /* May be due to an implicit anchor of m{.*foo} */ |
bbe252da | 1001 | && !(prog->intflags & PREGf_IMPLICIT)) |
d506a20d | 1002 | { |
cad2e5aa JH |
1003 | t = strpos; |
1004 | goto find_anchor; | |
1005 | } | |
a3621e74 | 1006 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 1007 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
70685ca0 | 1008 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 1009 | ); |
2c2d71f5 | 1010 | success_at_start: |
bbe252da | 1011 | if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ |
f2ed9b32 | 1012 | && (utf8_target ? ( |
33b8afdf JH |
1013 | prog->check_utf8 /* Could be deleted already */ |
1014 | && --BmUSEFUL(prog->check_utf8) < 0 | |
1015 | && (prog->check_utf8 == prog->float_utf8) | |
1016 | ) : ( | |
1017 | prog->check_substr /* Could be deleted already */ | |
1018 | && --BmUSEFUL(prog->check_substr) < 0 | |
1019 | && (prog->check_substr == prog->float_substr) | |
1020 | ))) | |
66e933ab | 1021 | { |
cad2e5aa | 1022 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 1023 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
f2ed9b32 KW |
1024 | /* XXX Does the destruction order has to change with utf8_target? */ |
1025 | SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr); | |
1026 | SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
1027 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
1028 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
1029 | check = NULL; /* abort */ | |
cad2e5aa | 1030 | s = strpos; |
486ec47a | 1031 | /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag |
c9415951 YO |
1032 | see http://bugs.activestate.com/show_bug.cgi?id=87173 */ |
1033 | if (prog->intflags & PREGf_IMPLICIT) | |
1034 | prog->extflags &= ~RXf_ANCH_MBOL; | |
3cf5c195 IZ |
1035 | /* XXXX This is a remnant of the old implementation. It |
1036 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 1037 | other heuristics. */ |
bbe252da | 1038 | prog->extflags &= ~RXf_USE_INTUIT; |
c9415951 | 1039 | /* XXXX What other flags might need to be cleared in this branch? */ |
cad2e5aa JH |
1040 | } |
1041 | else | |
1042 | s = strpos; | |
1043 | } | |
1044 | ||
6eb5f6b9 JH |
1045 | /* Last resort... */ |
1046 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
1de06328 YO |
1047 | /* trie stclasses are too expensive to use here, we are better off to |
1048 | leave it to regmatch itself */ | |
f8fc2ecf | 1049 | if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { |
6eb5f6b9 JH |
1050 | /* minlen == 0 is possible if regstclass is \b or \B, |
1051 | and the fixed substr is ''$. | |
1052 | Since minlen is already taken into account, s+1 is before strend; | |
1053 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
1054 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
1055 | regstclass does not come from lookahead... */ | |
1056 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
af944926 | 1057 | This leaves EXACTF-ish only, which are dealt with in find_byclass(). */ |
f8fc2ecf YO |
1058 | const U8* const str = (U8*)STRING(progi->regstclass); |
1059 | const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT | |
1060 | ? CHR_DIST(str+STR_LEN(progi->regstclass), str) | |
66e933ab | 1061 | : 1); |
1de06328 YO |
1062 | char * endpos; |
1063 | if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) | |
1064 | endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); | |
1065 | else if (prog->float_substr || prog->float_utf8) | |
1066 | endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); | |
1067 | else | |
1068 | endpos= strend; | |
1069 | ||
d8080198 YO |
1070 | if (checked_upto < s) |
1071 | checked_upto = s; | |
1072 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n", | |
1073 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg))); | |
1074 | ||
6eb5f6b9 | 1075 | t = s; |
d8080198 YO |
1076 | s = find_byclass(prog, progi->regstclass, checked_upto, endpos, NULL); |
1077 | if (s) { | |
1078 | checked_upto = s; | |
1079 | } else { | |
6eb5f6b9 | 1080 | #ifdef DEBUGGING |
cbbf8932 | 1081 | const char *what = NULL; |
6eb5f6b9 JH |
1082 | #endif |
1083 | if (endpos == strend) { | |
a3621e74 | 1084 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
1085 | "Could not match STCLASS...\n") ); |
1086 | goto fail; | |
1087 | } | |
a3621e74 | 1088 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 1089 | "This position contradicts STCLASS...\n") ); |
bbe252da | 1090 | if ((prog->extflags & RXf_ANCH) && !ml_anch) |
653099ff | 1091 | goto fail; |
d8080198 YO |
1092 | checked_upto = HOPBACKc(endpos, start_shift); |
1093 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" endpos: %"IVdf" checked_upto: %"IVdf"\n", | |
1094 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(endpos - strbeg), (IV)(checked_upto- strbeg))); | |
6eb5f6b9 | 1095 | /* Contradict one of substrings */ |
33b8afdf | 1096 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 | 1097 | if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) { |
a3621e74 | 1098 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 1099 | hop_and_restart: |
1aa99e6b | 1100 | s = HOP3c(t, 1, strend); |
66e933ab GS |
1101 | if (s + start_shift + end_shift > strend) { |
1102 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 1103 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
1104 | "Could not match STCLASS...\n") ); |
1105 | goto fail; | |
1106 | } | |
5e39e1e5 HS |
1107 | if (!check) |
1108 | goto giveup; | |
a3621e74 | 1109 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1110 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1111 | what, (long)(s + start_shift - i_strpos)) ); |
1112 | goto restart; | |
1113 | } | |
66e933ab | 1114 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
1115 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
1116 | goto retry_floating_check; | |
1117 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 1118 | s = check_at; |
5e39e1e5 HS |
1119 | if (!check) |
1120 | goto giveup; | |
a3621e74 | 1121 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1122 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1123 | (long)(other_last - i_strpos)) ); |
1124 | goto do_other_anchored; | |
1125 | } | |
60e71179 GS |
1126 | /* Another way we could have checked stclass at the |
1127 | current position only: */ | |
1128 | if (ml_anch) { | |
1129 | s = t = t + 1; | |
5e39e1e5 HS |
1130 | if (!check) |
1131 | goto giveup; | |
a3621e74 | 1132 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1133 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 1134 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 1135 | goto try_at_offset; |
66e933ab | 1136 | } |
f2ed9b32 | 1137 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 1138 | goto fail; |
486ec47a | 1139 | /* Check is floating substring. */ |
6eb5f6b9 JH |
1140 | retry_floating_check: |
1141 | t = check_at - start_shift; | |
a3621e74 | 1142 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
1143 | goto hop_and_restart; |
1144 | } | |
b7953727 | 1145 | if (t != s) { |
a3621e74 | 1146 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 1147 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
1148 | (long)(t - i_strpos), (long)(s - i_strpos)) |
1149 | ); | |
1150 | } | |
1151 | else { | |
a3621e74 | 1152 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1153 | "Does not contradict STCLASS...\n"); |
1154 | ); | |
1155 | } | |
6eb5f6b9 | 1156 | } |
5e39e1e5 | 1157 | giveup: |
a3621e74 | 1158 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
1159 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
1160 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 1161 | return s; |
2c2d71f5 JH |
1162 | |
1163 | fail_finish: /* Substring not found */ | |
33b8afdf | 1164 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
f2ed9b32 | 1165 | BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ |
cad2e5aa | 1166 | fail: |
a3621e74 | 1167 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 1168 | PL_colors[4], PL_colors[5])); |
bd61b366 | 1169 | return NULL; |
cad2e5aa | 1170 | } |
9661b544 | 1171 | |
a0a388a1 YO |
1172 | #define DECL_TRIE_TYPE(scan) \ |
1173 | const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \ | |
fab2782b YO |
1174 | trie_type = ((scan->flags == EXACT) \ |
1175 | ? (utf8_target ? trie_utf8 : trie_plain) \ | |
1176 | : (utf8_target ? trie_utf8_fold : trie_latin_utf8_fold)) | |
1177 | ||
1178 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ | |
1179 | uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ | |
1180 | STRLEN skiplen; \ | |
1181 | switch (trie_type) { \ | |
1182 | case trie_utf8_fold: \ | |
1183 | if ( foldlen>0 ) { \ | |
1184 | uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1185 | foldlen -= len; \ | |
1186 | uscan += len; \ | |
1187 | len=0; \ | |
1188 | } else { \ | |
1189 | uvc = to_utf8_fold( (const U8*) uc, foldbuf, &foldlen ); \ | |
1190 | len = UTF8SKIP(uc); \ | |
1191 | skiplen = UNISKIP( uvc ); \ | |
1192 | foldlen -= skiplen; \ | |
1193 | uscan = foldbuf + skiplen; \ | |
1194 | } \ | |
1195 | break; \ | |
1196 | case trie_latin_utf8_fold: \ | |
1197 | if ( foldlen>0 ) { \ | |
1198 | uvc = utf8n_to_uvuni( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1199 | foldlen -= len; \ | |
1200 | uscan += len; \ | |
1201 | len=0; \ | |
1202 | } else { \ | |
1203 | len = 1; \ | |
1204 | uvc = _to_fold_latin1( (U8) *uc, foldbuf, &foldlen, 1); \ | |
1205 | skiplen = UNISKIP( uvc ); \ | |
1206 | foldlen -= skiplen; \ | |
1207 | uscan = foldbuf + skiplen; \ | |
1208 | } \ | |
1209 | break; \ | |
1210 | case trie_utf8: \ | |
1211 | uvc = utf8n_to_uvuni( (const U8*) uc, UTF8_MAXLEN, &len, uniflags ); \ | |
1212 | break; \ | |
1213 | case trie_plain: \ | |
1214 | uvc = (UV)*uc; \ | |
1215 | len = 1; \ | |
1216 | } \ | |
1217 | if (uvc < 256) { \ | |
1218 | charid = trie->charmap[ uvc ]; \ | |
1219 | } \ | |
1220 | else { \ | |
1221 | charid = 0; \ | |
1222 | if (widecharmap) { \ | |
1223 | SV** const svpp = hv_fetch(widecharmap, \ | |
1224 | (char*)&uvc, sizeof(UV), 0); \ | |
1225 | if (svpp) \ | |
1226 | charid = (U16)SvIV(*svpp); \ | |
1227 | } \ | |
1228 | } \ | |
4cadc6a9 YO |
1229 | } STMT_END |
1230 | ||
4cadc6a9 YO |
1231 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ |
1232 | STMT_START { \ | |
1233 | while (s <= e) { \ | |
1234 | if ( (CoNd) \ | |
fac1af77 | 1235 | && (ln == 1 || folder(s, pat_string, ln)) \ |
9a5a5549 | 1236 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1237 | goto got_it; \ |
1238 | s++; \ | |
1239 | } \ | |
1240 | } STMT_END | |
1241 | ||
1242 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
1243 | STMT_START { \ | |
1244 | while (s + (uskip = UTF8SKIP(s)) <= strend) { \ | |
1245 | CoDe \ | |
1246 | s += uskip; \ | |
1247 | } \ | |
1248 | } STMT_END | |
1249 | ||
1250 | #define REXEC_FBC_SCAN(CoDe) \ | |
1251 | STMT_START { \ | |
1252 | while (s < strend) { \ | |
1253 | CoDe \ | |
1254 | s++; \ | |
1255 | } \ | |
1256 | } STMT_END | |
1257 | ||
1258 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1259 | REXEC_FBC_UTF8_SCAN( \ | |
1260 | if (CoNd) { \ | |
24b23f37 | 1261 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1262 | goto got_it; \ |
1263 | else \ | |
1264 | tmp = doevery; \ | |
1265 | } \ | |
1266 | else \ | |
1267 | tmp = 1; \ | |
1268 | ) | |
1269 | ||
1270 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1271 | REXEC_FBC_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_TRYIT \ | |
24b23f37 | 1283 | if ((!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1284 | goto got_it |
1285 | ||
e1d1eefb | 1286 | #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \ |
f2ed9b32 | 1287 | if (utf8_target) { \ |
e1d1eefb YO |
1288 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1289 | } \ | |
1290 | else { \ | |
1291 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1292 | } |
e1d1eefb | 1293 | |
4cadc6a9 | 1294 | #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ |
f2ed9b32 | 1295 | if (utf8_target) { \ |
4cadc6a9 YO |
1296 | UtFpReLoAd; \ |
1297 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1298 | } \ | |
1299 | else { \ | |
1300 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1301 | } |
4cadc6a9 YO |
1302 | |
1303 | #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ | |
1304 | PL_reg_flags |= RF_tainted; \ | |
f2ed9b32 | 1305 | if (utf8_target) { \ |
4cadc6a9 YO |
1306 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1307 | } \ | |
1308 | else { \ | |
1309 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
d981ef24 | 1310 | } |
4cadc6a9 | 1311 | |
786e8c11 YO |
1312 | #define DUMP_EXEC_POS(li,s,doutf8) \ |
1313 | dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8) | |
1314 | ||
cfaf538b KW |
1315 | |
1316 | #define UTF8_NOLOAD(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ | |
1317 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1318 | tmp = TEST_NON_UTF8(tmp); \ | |
1319 | REXEC_FBC_UTF8_SCAN( \ | |
1320 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1321 | tmp = !tmp; \ | |
1322 | IF_SUCCESS; \ | |
1323 | } \ | |
1324 | else { \ | |
1325 | IF_FAIL; \ | |
1326 | } \ | |
1327 | ); \ | |
1328 | ||
1329 | #define UTF8_LOAD(TeSt1_UtF8, TeSt2_UtF8, IF_SUCCESS, IF_FAIL) \ | |
1330 | if (s == PL_bostr) { \ | |
1331 | tmp = '\n'; \ | |
1332 | } \ | |
1333 | else { \ | |
1334 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); \ | |
1335 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); \ | |
1336 | } \ | |
1337 | tmp = TeSt1_UtF8; \ | |
1338 | LOAD_UTF8_CHARCLASS_ALNUM(); \ | |
1339 | REXEC_FBC_UTF8_SCAN( \ | |
1340 | if (tmp == ! (TeSt2_UtF8)) { \ | |
1341 | tmp = !tmp; \ | |
1342 | IF_SUCCESS; \ | |
1343 | } \ | |
1344 | else { \ | |
1345 | IF_FAIL; \ | |
1346 | } \ | |
1347 | ); \ | |
1348 | ||
63ac0dad KW |
1349 | /* The only difference between the BOUND and NBOUND cases is that |
1350 | * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in | |
1351 | * NBOUND. This is accomplished by passing it in either the if or else clause, | |
1352 | * with the other one being empty */ | |
1353 | #define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1354 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
cfaf538b KW |
1355 | |
1356 | #define FBC_BOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1357 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
63ac0dad KW |
1358 | |
1359 | #define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1360 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b KW |
1361 | |
1362 | #define FBC_NBOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1363 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b | 1364 | |
63ac0dad KW |
1365 | |
1366 | /* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to | |
1367 | * be passed in completely with the variable name being tested, which isn't | |
1368 | * such a clean interface, but this is easier to read than it was before. We | |
1369 | * are looking for the boundary (or non-boundary between a word and non-word | |
1370 | * character. The utf8 and non-utf8 cases have the same logic, but the details | |
1371 | * must be different. Find the "wordness" of the character just prior to this | |
1372 | * one, and compare it with the wordness of this one. If they differ, we have | |
1373 | * a boundary. At the beginning of the string, pretend that the previous | |
1374 | * character was a new-line */ | |
cfaf538b | 1375 | #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ |
63ac0dad | 1376 | if (utf8_target) { \ |
cfaf538b | 1377 | UTF8_CODE \ |
63ac0dad KW |
1378 | } \ |
1379 | else { /* Not utf8 */ \ | |
1380 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1381 | tmp = TEST_NON_UTF8(tmp); \ | |
1382 | REXEC_FBC_SCAN( \ | |
1383 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1384 | tmp = !tmp; \ | |
1385 | IF_SUCCESS; \ | |
1386 | } \ | |
1387 | else { \ | |
1388 | IF_FAIL; \ | |
1389 | } \ | |
1390 | ); \ | |
1391 | } \ | |
1392 | if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) \ | |
1393 | goto got_it; | |
1394 | ||
786e8c11 YO |
1395 | /* We know what class REx starts with. Try to find this position... */ |
1396 | /* if reginfo is NULL, its a dryrun */ | |
1397 | /* annoyingly all the vars in this routine have different names from their counterparts | |
1398 | in regmatch. /grrr */ | |
1399 | ||
3c3eec57 | 1400 | STATIC char * |
07be1b83 | 1401 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
24b23f37 | 1402 | const char *strend, regmatch_info *reginfo) |
a687059c | 1403 | { |
27da23d5 | 1404 | dVAR; |
bbe252da | 1405 | const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; |
fac1af77 KW |
1406 | char *pat_string; /* The pattern's exactish string */ |
1407 | char *pat_end; /* ptr to end char of pat_string */ | |
1408 | re_fold_t folder; /* Function for computing non-utf8 folds */ | |
1409 | const U8 *fold_array; /* array for folding ords < 256 */ | |
d8093b23 | 1410 | STRLEN ln; |
5dab1207 | 1411 | STRLEN lnc; |
078c425b | 1412 | register STRLEN uskip; |
fac1af77 KW |
1413 | U8 c1; |
1414 | U8 c2; | |
6eb5f6b9 JH |
1415 | char *e; |
1416 | register I32 tmp = 1; /* Scratch variable? */ | |
f2ed9b32 | 1417 | register const bool utf8_target = PL_reg_match_utf8; |
453bfd44 | 1418 | UV utf8_fold_flags = 0; |
f8fc2ecf | 1419 | RXi_GET_DECL(prog,progi); |
7918f24d NC |
1420 | |
1421 | PERL_ARGS_ASSERT_FIND_BYCLASS; | |
f8fc2ecf | 1422 | |
6eb5f6b9 JH |
1423 | /* We know what class it must start with. */ |
1424 | switch (OP(c)) { | |
f56b6394 | 1425 | case ANYOFV: |
6eb5f6b9 | 1426 | case ANYOF: |
f56b6394 | 1427 | if (utf8_target || OP(c) == ANYOFV) { |
b1e3e569 KW |
1428 | STRLEN inclasslen = strend - s; |
1429 | REXEC_FBC_UTF8_CLASS_SCAN( | |
1430 | reginclass(prog, c, (U8*)s, &inclasslen, utf8_target)); | |
388cc4de HS |
1431 | } |
1432 | else { | |
6ef69d56 | 1433 | REXEC_FBC_CLASS_SCAN(REGINCLASS(prog, c, (U8*)s)); |
a0d0e21e | 1434 | } |
6eb5f6b9 | 1435 | break; |
f33976b4 | 1436 | case CANY: |
4cadc6a9 | 1437 | REXEC_FBC_SCAN( |
24b23f37 | 1438 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
f33976b4 DB |
1439 | goto got_it; |
1440 | else | |
1441 | tmp = doevery; | |
4cadc6a9 | 1442 | ); |
f33976b4 | 1443 | break; |
2f7f8cb1 KW |
1444 | |
1445 | case EXACTFA: | |
1446 | if (UTF_PATTERN || utf8_target) { | |
1447 | utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; | |
1448 | goto do_exactf_utf8; | |
1449 | } | |
1450 | fold_array = PL_fold_latin1; /* Latin1 folds are not affected by */ | |
1451 | folder = foldEQ_latin1; /* /a, except the sharp s one which */ | |
1452 | goto do_exactf_non_utf8; /* isn't dealt with by these */ | |
1453 | ||
6eb5f6b9 | 1454 | case EXACTF: |
62bf7766 | 1455 | if (utf8_target) { |
77a6d856 KW |
1456 | |
1457 | /* regcomp.c already folded this if pattern is in UTF-8 */ | |
62bf7766 | 1458 | utf8_fold_flags = 0; |
fac1af77 KW |
1459 | goto do_exactf_utf8; |
1460 | } | |
1461 | fold_array = PL_fold; | |
1462 | folder = foldEQ; | |
1463 | goto do_exactf_non_utf8; | |
1464 | ||
1465 | case EXACTFL: | |
1466 | if (UTF_PATTERN || utf8_target) { | |
17580e7a | 1467 | utf8_fold_flags = FOLDEQ_UTF8_LOCALE; |
fac1af77 KW |
1468 | goto do_exactf_utf8; |
1469 | } | |
1470 | fold_array = PL_fold_locale; | |
1471 | folder = foldEQ_locale; | |
16d951b7 KW |
1472 | goto do_exactf_non_utf8; |
1473 | ||
3c760661 KW |
1474 | case EXACTFU_SS: |
1475 | if (UTF_PATTERN) { | |
1476 | utf8_fold_flags = FOLDEQ_S2_ALREADY_FOLDED; | |
1477 | } | |
1478 | goto do_exactf_utf8; | |
1479 | ||
fab2782b | 1480 | case EXACTFU_TRICKYFOLD: |
16d951b7 KW |
1481 | case EXACTFU: |
1482 | if (UTF_PATTERN || utf8_target) { | |
77a6d856 | 1483 | utf8_fold_flags = (UTF_PATTERN) ? FOLDEQ_S2_ALREADY_FOLDED : 0; |
16d951b7 KW |
1484 | goto do_exactf_utf8; |
1485 | } | |
1486 | ||
1487 | /* Any 'ss' in the pattern should have been replaced by regcomp, | |
1488 | * so we don't have to worry here about this single special case | |
1489 | * in the Latin1 range */ | |
1490 | fold_array = PL_fold_latin1; | |
1491 | folder = foldEQ_latin1; | |
fac1af77 KW |
1492 | |
1493 | /* FALL THROUGH */ | |
1494 | ||
62bf7766 KW |
1495 | do_exactf_non_utf8: /* Neither pattern nor string are UTF8, and there |
1496 | are no glitches with fold-length differences | |
1497 | between the target string and pattern */ | |
fac1af77 KW |
1498 | |
1499 | /* The idea in the non-utf8 EXACTF* cases is to first find the | |
1500 | * first character of the EXACTF* node and then, if necessary, | |
1501 | * case-insensitively compare the full text of the node. c1 is the | |
1502 | * first character. c2 is its fold. This logic will not work for | |
1503 | * Unicode semantics and the german sharp ss, which hence should | |
1504 | * not be compiled into a node that gets here. */ | |
1505 | pat_string = STRING(c); | |
1506 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1507 | ||
8a90a8fe KW |
1508 | /* We know that we have to match at least 'ln' bytes (which is the |
1509 | * same as characters, since not utf8). If we have to match 3 | |
1510 | * characters, and there are only 2 availabe, we know without | |
1511 | * trying that it will fail; so don't start a match past the | |
1512 | * required minimum number from the far end */ | |
fac1af77 KW |
1513 | e = HOP3c(strend, -((I32)ln), s); |
1514 | ||
1515 | if (!reginfo && e < s) { | |
1516 | e = s; /* Due to minlen logic of intuit() */ | |
1517 | } | |
1518 | ||
1519 | c1 = *pat_string; | |
1520 | c2 = fold_array[c1]; | |
1521 | if (c1 == c2) { /* If char and fold are the same */ | |
1522 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); | |
1523 | } | |
1524 | else { | |
1525 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); | |
1526 | } | |
1527 | break; | |
1528 | ||
1529 | do_exactf_utf8: | |
e067297c KW |
1530 | { |
1531 | unsigned expansion; | |
1532 | ||
fac1af77 KW |
1533 | |
1534 | /* If one of the operands is in utf8, we can't use the simpler | |
1535 | * folding above, due to the fact that many different characters | |
1536 | * can have the same fold, or portion of a fold, or different- | |
1537 | * length fold */ | |
1538 | pat_string = STRING(c); | |
1539 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1540 | pat_end = pat_string + ln; | |
1541 | lnc = (UTF_PATTERN) /* length to match in characters */ | |
1542 | ? utf8_length((U8 *) pat_string, (U8 *) pat_end) | |
1543 | : ln; | |
1544 | ||
e067297c KW |
1545 | /* We have 'lnc' characters to match in the pattern, but because of |
1546 | * multi-character folding, each character in the target can match | |
1547 | * up to 3 characters (Unicode guarantees it will never exceed | |
1548 | * this) if it is utf8-encoded; and up to 2 if not (based on the | |
1549 | * fact that the Latin 1 folds are already determined, and the | |
1550 | * only multi-char fold in that range is the sharp-s folding to | |
1551 | * 'ss'. Thus, a pattern character can match as little as 1/3 of a | |
89378d8a KW |
1552 | * string character. Adjust lnc accordingly, rounding up, so that |
1553 | * if we need to match at least 4+1/3 chars, that really is 5. */ | |
e067297c | 1554 | expansion = (utf8_target) ? UTF8_MAX_FOLD_CHAR_EXPAND : 2; |
89378d8a | 1555 | lnc = (lnc + expansion - 1) / expansion; |
e067297c KW |
1556 | |
1557 | /* As in the non-UTF8 case, if we have to match 3 characters, and | |
1558 | * only 2 are left, it's guaranteed to fail, so don't start a | |
1559 | * match that would require us to go beyond the end of the string | |
1560 | */ | |
1561 | e = HOP3c(strend, -((I32)lnc), s); | |
fac1af77 KW |
1562 | |
1563 | if (!reginfo && e < s) { | |
1564 | e = s; /* Due to minlen logic of intuit() */ | |
1565 | } | |
1566 | ||
b33105db KW |
1567 | /* XXX Note that we could recalculate e to stop the loop earlier, |
1568 | * as the worst case expansion above will rarely be met, and as we | |
1569 | * go along we would usually find that e moves further to the left. | |
1570 | * This would happen only after we reached the point in the loop | |
1571 | * where if there were no expansion we should fail. Unclear if | |
1572 | * worth the expense */ | |
e067297c | 1573 | |
fac1af77 KW |
1574 | while (s <= e) { |
1575 | char *my_strend= (char *)strend; | |
1576 | if (foldEQ_utf8_flags(s, &my_strend, 0, utf8_target, | |
1577 | pat_string, NULL, ln, cBOOL(UTF_PATTERN), utf8_fold_flags) | |
1578 | && (!reginfo || regtry(reginfo, &s)) ) | |
1579 | { | |
1580 | goto got_it; | |
1581 | } | |
bbdd8bad | 1582 | s += (utf8_target) ? UTF8SKIP(s) : 1; |
fac1af77 KW |
1583 | } |
1584 | break; | |
e067297c | 1585 | } |
bbce6d69 | 1586 | case BOUNDL: |
3280af22 | 1587 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1588 | FBC_BOUND(isALNUM_LC, |
1589 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1590 | isALNUM_LC_utf8((U8*)s)); | |
a0ed51b3 | 1591 | break; |
bbce6d69 | 1592 | case NBOUNDL: |
3280af22 | 1593 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1594 | FBC_NBOUND(isALNUM_LC, |
1595 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1596 | isALNUM_LC_utf8((U8*)s)); | |
1597 | break; | |
1598 | case BOUND: | |
1599 | FBC_BOUND(isWORDCHAR, | |
1600 | isALNUM_uni(tmp), | |
1601 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1602 | break; | |
cfaf538b KW |
1603 | case BOUNDA: |
1604 | FBC_BOUND_NOLOAD(isWORDCHAR_A, | |
1605 | isWORDCHAR_A(tmp), | |
1606 | isWORDCHAR_A((U8*)s)); | |
1607 | break; | |
a0d0e21e | 1608 | case NBOUND: |
63ac0dad KW |
1609 | FBC_NBOUND(isWORDCHAR, |
1610 | isALNUM_uni(tmp), | |
1611 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1612 | break; | |
cfaf538b KW |
1613 | case NBOUNDA: |
1614 | FBC_NBOUND_NOLOAD(isWORDCHAR_A, | |
1615 | isWORDCHAR_A(tmp), | |
1616 | isWORDCHAR_A((U8*)s)); | |
1617 | break; | |
63ac0dad KW |
1618 | case BOUNDU: |
1619 | FBC_BOUND(isWORDCHAR_L1, | |
1620 | isALNUM_uni(tmp), | |
1621 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1622 | break; | |
1623 | case NBOUNDU: | |
1624 | FBC_NBOUND(isWORDCHAR_L1, | |
1625 | isALNUM_uni(tmp), | |
1626 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
a0ed51b3 | 1627 | break; |
bbce6d69 | 1628 | case ALNUML: |
4cadc6a9 YO |
1629 | REXEC_FBC_CSCAN_TAINT( |
1630 | isALNUM_LC_utf8((U8*)s), | |
1631 | isALNUM_LC(*s) | |
1632 | ); | |
6895a8aa | 1633 | break; |
980866de KW |
1634 | case ALNUMU: |
1635 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1636 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1637 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1638 | isWORDCHAR_L1((U8) *s) |
1639 | ); | |
6895a8aa | 1640 | break; |
980866de KW |
1641 | case ALNUM: |
1642 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1643 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1644 | swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), | |
980866de KW |
1645 | isWORDCHAR((U8) *s) |
1646 | ); | |
6895a8aa | 1647 | break; |
cfaf538b | 1648 | case ALNUMA: |
8e9da4d4 KW |
1649 | /* Don't need to worry about utf8, as it can match only a single |
1650 | * byte invariant character */ | |
cfaf538b | 1651 | REXEC_FBC_CLASS_SCAN( isWORDCHAR_A(*s)); |
6895a8aa | 1652 | break; |
980866de KW |
1653 | case NALNUMU: |
1654 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 | 1655 | LOAD_UTF8_CHARCLASS_ALNUM(), |
359960d4 | 1656 | !swash_fetch(PL_utf8_alnum,(U8*)s, utf8_target), |
980866de KW |
1657 | ! isWORDCHAR_L1((U8) *s) |
1658 | ); | |
6895a8aa | 1659 | break; |
a0d0e21e | 1660 | case NALNUM: |
4cadc6a9 | 1661 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1662 | LOAD_UTF8_CHARCLASS_ALNUM(), |
1663 | !swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target), | |
980866de | 1664 | ! isALNUM(*s) |
4cadc6a9 | 1665 | ); |
6895a8aa | 1666 | break; |
cfaf538b | 1667 | case NALNUMA: |
8e9da4d4 KW |
1668 | REXEC_FBC_CSCAN( |
1669 | !isWORDCHAR_A(*s), | |
1670 | !isWORDCHAR_A(*s) | |
1671 | ); | |
1672 | break; | |
bbce6d69 | 1673 | case NALNUML: |
4cadc6a9 YO |
1674 | REXEC_FBC_CSCAN_TAINT( |
1675 | !isALNUM_LC_utf8((U8*)s), | |
1676 | !isALNUM_LC(*s) | |
1677 | ); | |
6895a8aa | 1678 | break; |
980866de KW |
1679 | case SPACEU: |
1680 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1681 | LOAD_UTF8_CHARCLASS_SPACE(), |
1682 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de KW |
1683 | isSPACE_L1((U8) *s) |
1684 | ); | |
6895a8aa | 1685 | break; |
a0d0e21e | 1686 | case SPACE: |
4cadc6a9 | 1687 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1688 | LOAD_UTF8_CHARCLASS_SPACE(), |
1689 | *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target), | |
980866de | 1690 | isSPACE((U8) *s) |
4cadc6a9 | 1691 | ); |
6895a8aa | 1692 | break; |
cfaf538b | 1693 | case SPACEA: |
8e9da4d4 KW |
1694 | /* Don't need to worry about utf8, as it can match only a single |
1695 | * byte invariant character */ | |
cfaf538b | 1696 | REXEC_FBC_CLASS_SCAN( isSPACE_A(*s)); |
6895a8aa | 1697 | break; |
bbce6d69 | 1698 | case SPACEL: |
4cadc6a9 | 1699 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1700 | isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1701 | isSPACE_LC(*s) |
1702 | ); | |
6895a8aa | 1703 | break; |
980866de KW |
1704 | case NSPACEU: |
1705 | REXEC_FBC_CSCAN_PRELOAD( | |
779d7b58 KW |
1706 | LOAD_UTF8_CHARCLASS_SPACE(), |
1707 | !( *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de KW |
1708 | ! isSPACE_L1((U8) *s) |
1709 | ); | |
6895a8aa | 1710 | break; |
a0d0e21e | 1711 | case NSPACE: |
4cadc6a9 | 1712 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1713 | LOAD_UTF8_CHARCLASS_SPACE(), |
1714 | !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, utf8_target)), | |
980866de | 1715 | ! isSPACE((U8) *s) |
4cadc6a9 | 1716 | ); |
6895a8aa | 1717 | break; |
cfaf538b | 1718 | case NSPACEA: |
8e9da4d4 KW |
1719 | REXEC_FBC_CSCAN( |
1720 | !isSPACE_A(*s), | |
1721 | !isSPACE_A(*s) | |
1722 | ); | |
1723 | break; | |
bbce6d69 | 1724 | case NSPACEL: |
4cadc6a9 | 1725 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1726 | !isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1727 | !isSPACE_LC(*s) |
1728 | ); | |
6895a8aa | 1729 | break; |
a0d0e21e | 1730 | case DIGIT: |
4cadc6a9 | 1731 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1732 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1733 | swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1734 | isDIGIT(*s) |
1735 | ); | |
6895a8aa | 1736 | break; |
cfaf538b | 1737 | case DIGITA: |
8e9da4d4 KW |
1738 | /* Don't need to worry about utf8, as it can match only a single |
1739 | * byte invariant character */ | |
cfaf538b | 1740 | REXEC_FBC_CLASS_SCAN( isDIGIT_A(*s)); |
6895a8aa | 1741 | break; |
b8c5462f | 1742 | case DIGITL: |
4cadc6a9 YO |
1743 | REXEC_FBC_CSCAN_TAINT( |
1744 | isDIGIT_LC_utf8((U8*)s), | |
1745 | isDIGIT_LC(*s) | |
1746 | ); | |
6895a8aa | 1747 | break; |
a0d0e21e | 1748 | case NDIGIT: |
4cadc6a9 | 1749 | REXEC_FBC_CSCAN_PRELOAD( |
779d7b58 KW |
1750 | LOAD_UTF8_CHARCLASS_DIGIT(), |
1751 | !swash_fetch(PL_utf8_digit,(U8*)s, utf8_target), | |
4cadc6a9 YO |
1752 | !isDIGIT(*s) |
1753 | ); | |
6895a8aa | 1754 | break; |
cfaf538b | 1755 | case NDIGITA: |
8e9da4d4 KW |
1756 | REXEC_FBC_CSCAN( |
1757 | !isDIGIT_A(*s), | |
1758 | !isDIGIT_A(*s) | |
1759 | ); | |
1760 | break; | |
b8c5462f | 1761 | case NDIGITL: |
4cadc6a9 YO |
1762 | REXEC_FBC_CSCAN_TAINT( |
1763 | !isDIGIT_LC_utf8((U8*)s), | |
1764 | !isDIGIT_LC(*s) | |
1765 | ); | |
6895a8aa | 1766 | break; |
e1d1eefb YO |
1767 | case LNBREAK: |
1768 | REXEC_FBC_CSCAN( | |
1769 | is_LNBREAK_utf8(s), | |
1770 | is_LNBREAK_latin1(s) | |
1771 | ); | |
6895a8aa | 1772 | break; |
e1d1eefb YO |
1773 | case VERTWS: |
1774 | REXEC_FBC_CSCAN( | |
1775 | is_VERTWS_utf8(s), | |
1776 | is_VERTWS_latin1(s) | |
1777 | ); | |
6895a8aa | 1778 | break; |
e1d1eefb YO |
1779 | case NVERTWS: |
1780 | REXEC_FBC_CSCAN( | |
1781 | !is_VERTWS_utf8(s), | |
1782 | !is_VERTWS_latin1(s) | |
1783 | ); | |
6895a8aa | 1784 | break; |
e1d1eefb YO |
1785 | case HORIZWS: |
1786 | REXEC_FBC_CSCAN( | |
1787 | is_HORIZWS_utf8(s), | |
1788 | is_HORIZWS_latin1(s) | |
1789 | ); | |
6895a8aa | 1790 | break; |
e1d1eefb YO |
1791 | case NHORIZWS: |
1792 | REXEC_FBC_CSCAN( | |
1793 | !is_HORIZWS_utf8(s), | |
1794 | !is_HORIZWS_latin1(s) | |
1795 | ); | |
6895a8aa | 1796 | break; |
1de06328 YO |
1797 | case AHOCORASICKC: |
1798 | case AHOCORASICK: | |
07be1b83 | 1799 | { |
a0a388a1 | 1800 | DECL_TRIE_TYPE(c); |
07be1b83 YO |
1801 | /* what trie are we using right now */ |
1802 | reg_ac_data *aho | |
f8fc2ecf | 1803 | = (reg_ac_data*)progi->data->data[ ARG( c ) ]; |
3251b653 NC |
1804 | reg_trie_data *trie |
1805 | = (reg_trie_data*)progi->data->data[ aho->trie ]; | |
85fbaab2 | 1806 | HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); |
07be1b83 YO |
1807 | |
1808 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1809 | #ifdef DEBUGGING |
07be1b83 | 1810 | const char *real_start = s; |
6148ee25 | 1811 | #endif |
07be1b83 | 1812 | STRLEN maxlen = trie->maxlen; |
be8e71aa YO |
1813 | SV *sv_points; |
1814 | U8 **points; /* map of where we were in the input string | |
786e8c11 | 1815 | when reading a given char. For ASCII this |
be8e71aa | 1816 | is unnecessary overhead as the relationship |
38a44b82 NC |
1817 | is always 1:1, but for Unicode, especially |
1818 | case folded Unicode this is not true. */ | |
f9e705e8 | 1819 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
786e8c11 YO |
1820 | U8 *bitmap=NULL; |
1821 | ||
07be1b83 YO |
1822 | |
1823 | GET_RE_DEBUG_FLAGS_DECL; | |
1824 | ||
be8e71aa YO |
1825 | /* We can't just allocate points here. We need to wrap it in |
1826 | * an SV so it gets freed properly if there is a croak while | |
1827 | * running the match */ | |
1828 | ENTER; | |
1829 | SAVETMPS; | |
1830 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1831 | SvCUR_set(sv_points, | |
1832 | maxlen * sizeof(U8 *)); | |
1833 | SvPOK_on(sv_points); | |
1834 | sv_2mortal(sv_points); | |
1835 | points=(U8**)SvPV_nolen(sv_points ); | |
1de06328 YO |
1836 | if ( trie_type != trie_utf8_fold |
1837 | && (trie->bitmap || OP(c)==AHOCORASICKC) ) | |
1838 | { | |
786e8c11 YO |
1839 | if (trie->bitmap) |
1840 | bitmap=(U8*)trie->bitmap; | |
1841 | else | |
1842 | bitmap=(U8*)ANYOF_BITMAP(c); | |
07be1b83 | 1843 | } |
786e8c11 YO |
1844 | /* this is the Aho-Corasick algorithm modified a touch |
1845 | to include special handling for long "unknown char" | |
1846 | sequences. The basic idea being that we use AC as long | |
1847 | as we are dealing with a possible matching char, when | |
1848 | we encounter an unknown char (and we have not encountered | |
1849 | an accepting state) we scan forward until we find a legal | |
1850 | starting char. | |
1851 | AC matching is basically that of trie matching, except | |
1852 | that when we encounter a failing transition, we fall back | |
1853 | to the current states "fail state", and try the current char | |
1854 | again, a process we repeat until we reach the root state, | |
1855 | state 1, or a legal transition. If we fail on the root state | |
1856 | then we can either terminate if we have reached an accepting | |
1857 | state previously, or restart the entire process from the beginning | |
1858 | if we have not. | |
1859 | ||
1860 | */ | |
07be1b83 YO |
1861 | while (s <= last_start) { |
1862 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1863 | U8 *uc = (U8*)s; | |
1864 | U16 charid = 0; | |
1865 | U32 base = 1; | |
1866 | U32 state = 1; | |
1867 | UV uvc = 0; | |
1868 | STRLEN len = 0; | |
1869 | STRLEN foldlen = 0; | |
1870 | U8 *uscan = (U8*)NULL; | |
1871 | U8 *leftmost = NULL; | |
786e8c11 YO |
1872 | #ifdef DEBUGGING |
1873 | U32 accepted_word= 0; | |
1874 | #endif | |
07be1b83 YO |
1875 | U32 pointpos = 0; |
1876 | ||
1877 | while ( state && uc <= (U8*)strend ) { | |
1878 | int failed=0; | |
786e8c11 YO |
1879 | U32 word = aho->states[ state ].wordnum; |
1880 | ||
1de06328 YO |
1881 | if( state==1 ) { |
1882 | if ( bitmap ) { | |
1883 | DEBUG_TRIE_EXECUTE_r( | |
1884 | if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1885 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1886 | (char *)uc, utf8_target ); |
1de06328 YO |
1887 | PerlIO_printf( Perl_debug_log, |
1888 | " Scanning for legal start char...\n"); | |
1889 | } | |
d085b490 YO |
1890 | ); |
1891 | if (utf8_target) { | |
1892 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1893 | uc += UTF8SKIP(uc); | |
1894 | } | |
1895 | } else { | |
1896 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1897 | uc++; | |
1898 | } | |
1899 | } | |
1de06328 | 1900 | s= (char *)uc; |
786e8c11 | 1901 | } |
786e8c11 YO |
1902 | if (uc >(U8*)last_start) break; |
1903 | } | |
1904 | ||
1905 | if ( word ) { | |
2e64971a | 1906 | U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ]; |
786e8c11 YO |
1907 | if (!leftmost || lpos < leftmost) { |
1908 | DEBUG_r(accepted_word=word); | |
07be1b83 | 1909 | leftmost= lpos; |
786e8c11 | 1910 | } |
07be1b83 | 1911 | if (base==0) break; |
786e8c11 | 1912 | |
07be1b83 YO |
1913 | } |
1914 | points[pointpos++ % maxlen]= uc; | |
55eed653 NC |
1915 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
1916 | uscan, len, uvc, charid, foldlen, | |
1917 | foldbuf, uniflags); | |
786e8c11 YO |
1918 | DEBUG_TRIE_EXECUTE_r({ |
1919 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1920 | s, utf8_target ); |
07be1b83 | 1921 | PerlIO_printf(Perl_debug_log, |
786e8c11 YO |
1922 | " Charid:%3u CP:%4"UVxf" ", |
1923 | charid, uvc); | |
1924 | }); | |
07be1b83 YO |
1925 | |
1926 | do { | |
6148ee25 | 1927 | #ifdef DEBUGGING |
786e8c11 | 1928 | word = aho->states[ state ].wordnum; |
6148ee25 | 1929 | #endif |
07be1b83 YO |
1930 | base = aho->states[ state ].trans.base; |
1931 | ||
786e8c11 YO |
1932 | DEBUG_TRIE_EXECUTE_r({ |
1933 | if (failed) | |
1934 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1935 | s, utf8_target ); |
07be1b83 | 1936 | PerlIO_printf( Perl_debug_log, |
786e8c11 YO |
1937 | "%sState: %4"UVxf", word=%"UVxf, |
1938 | failed ? " Fail transition to " : "", | |
1939 | (UV)state, (UV)word); | |
1940 | }); | |
07be1b83 YO |
1941 | if ( base ) { |
1942 | U32 tmp; | |
6dd2be57 | 1943 | I32 offset; |
07be1b83 | 1944 | if (charid && |
6dd2be57 DM |
1945 | ( ((offset = base + charid |
1946 | - 1 - trie->uniquecharcount)) >= 0) | |
1947 | && ((U32)offset < trie->lasttrans) | |
1948 | && trie->trans[offset].check == state | |
1949 | && (tmp=trie->trans[offset].next)) | |
07be1b83 | 1950 | { |
786e8c11 YO |
1951 | DEBUG_TRIE_EXECUTE_r( |
1952 | PerlIO_printf( Perl_debug_log," - legal\n")); | |
07be1b83 YO |
1953 | state = tmp; |
1954 | break; | |
1955 | } | |
1956 | else { | |
786e8c11 YO |
1957 | DEBUG_TRIE_EXECUTE_r( |
1958 | PerlIO_printf( Perl_debug_log," - fail\n")); | |
1959 | failed = 1; | |
1960 | state = aho->fail[state]; | |
07be1b83 YO |
1961 | } |
1962 | } | |
1963 | else { | |
1964 | /* we must be accepting here */ | |
786e8c11 YO |
1965 | DEBUG_TRIE_EXECUTE_r( |
1966 | PerlIO_printf( Perl_debug_log," - accepting\n")); | |
1967 | failed = 1; | |
07be1b83 YO |
1968 | break; |
1969 | } | |
1970 | } while(state); | |
786e8c11 | 1971 | uc += len; |
07be1b83 YO |
1972 | if (failed) { |
1973 | if (leftmost) | |
1974 | break; | |
786e8c11 | 1975 | if (!state) state = 1; |
07be1b83 YO |
1976 | } |
1977 | } | |
1978 | if ( aho->states[ state ].wordnum ) { | |
2e64971a | 1979 | U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ]; |
786e8c11 YO |
1980 | if (!leftmost || lpos < leftmost) { |
1981 | DEBUG_r(accepted_word=aho->states[ state ].wordnum); | |
07be1b83 | 1982 | leftmost = lpos; |
786e8c11 | 1983 | } |
07be1b83 | 1984 | } |
07be1b83 YO |
1985 | if (leftmost) { |
1986 | s = (char*)leftmost; | |
786e8c11 YO |
1987 | DEBUG_TRIE_EXECUTE_r({ |
1988 | PerlIO_printf( | |
70685ca0 JH |
1989 | Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", |
1990 | (UV)accepted_word, (IV)(s - real_start) | |
786e8c11 YO |
1991 | ); |
1992 | }); | |
24b23f37 | 1993 | if (!reginfo || regtry(reginfo, &s)) { |
be8e71aa YO |
1994 | FREETMPS; |
1995 | LEAVE; | |
07be1b83 | 1996 | goto got_it; |
be8e71aa | 1997 | } |
07be1b83 | 1998 | s = HOPc(s,1); |
786e8c11 YO |
1999 | DEBUG_TRIE_EXECUTE_r({ |
2000 | PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); | |
2001 | }); | |
07be1b83 | 2002 | } else { |
786e8c11 YO |
2003 | DEBUG_TRIE_EXECUTE_r( |
2004 | PerlIO_printf( Perl_debug_log,"No match.\n")); | |
07be1b83 YO |
2005 | break; |
2006 | } | |
2007 | } | |
be8e71aa YO |
2008 | FREETMPS; |
2009 | LEAVE; | |
07be1b83 YO |
2010 | } |
2011 | break; | |
b3c9acc1 | 2012 | default: |
3c3eec57 GS |
2013 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); |
2014 | break; | |
d6a28714 | 2015 | } |
6eb5f6b9 JH |
2016 | return 0; |
2017 | got_it: | |
2018 | return s; | |
2019 | } | |
2020 | ||
fae667d5 | 2021 | |
6eb5f6b9 JH |
2022 | /* |
2023 | - regexec_flags - match a regexp against a string | |
2024 | */ | |
2025 | I32 | |
288b8c02 | 2026 | Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend, |
6eb5f6b9 JH |
2027 | char *strbeg, I32 minend, SV *sv, void *data, U32 flags) |
2028 | /* strend: pointer to null at end of string */ | |
2029 | /* strbeg: real beginning of string */ | |
2030 | /* minend: end of match must be >=minend after stringarg. */ | |
58e23c8d YO |
2031 | /* data: May be used for some additional optimizations. |
2032 | Currently its only used, with a U32 cast, for transmitting | |
2033 | the ganch offset when doing a /g match. This will change */ | |
6eb5f6b9 JH |
2034 | /* nosave: For optimizations. */ |
2035 | { | |
97aff369 | 2036 | dVAR; |
288b8c02 | 2037 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
24b23f37 | 2038 | /*register*/ char *s; |
6eb5f6b9 | 2039 | register regnode *c; |
24b23f37 | 2040 | /*register*/ char *startpos = stringarg; |
6eb5f6b9 JH |
2041 | I32 minlen; /* must match at least this many chars */ |
2042 | I32 dontbother = 0; /* how many characters not to try at end */ | |
6eb5f6b9 JH |
2043 | I32 end_shift = 0; /* Same for the end. */ /* CC */ |
2044 | I32 scream_pos = -1; /* Internal iterator of scream. */ | |
ccac19ea | 2045 | char *scream_olds = NULL; |
f2ed9b32 | 2046 | const bool utf8_target = cBOOL(DO_UTF8(sv)); |
2757e526 | 2047 | I32 multiline; |
f8fc2ecf | 2048 | RXi_GET_DECL(prog,progi); |
3b0527fe | 2049 | regmatch_info reginfo; /* create some info to pass to regtry etc */ |
e9105d30 | 2050 | regexp_paren_pair *swap = NULL; |
a3621e74 YO |
2051 | GET_RE_DEBUG_FLAGS_DECL; |
2052 | ||
7918f24d | 2053 | PERL_ARGS_ASSERT_REGEXEC_FLAGS; |
9d4ba2ae | 2054 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
2055 | |
2056 | /* Be paranoid... */ | |
2057 | if (prog == NULL || startpos == NULL) { | |
2058 | Perl_croak(aTHX_ "NULL regexp parameter"); | |
2059 | return 0; | |
2060 | } | |
2061 | ||
bbe252da | 2062 | multiline = prog->extflags & RXf_PMf_MULTILINE; |
288b8c02 | 2063 | reginfo.prog = rx; /* Yes, sorry that this is confusing. */ |
2757e526 | 2064 | |
f2ed9b32 | 2065 | RX_MATCH_UTF8_set(rx, utf8_target); |
1de06328 | 2066 | DEBUG_EXECUTE_r( |
f2ed9b32 | 2067 | debug_start_match(rx, utf8_target, startpos, strend, |
1de06328 YO |
2068 | "Matching"); |
2069 | ); | |
bac06658 | 2070 | |
6eb5f6b9 | 2071 | minlen = prog->minlen; |
1de06328 YO |
2072 | |
2073 | if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { | |
a3621e74 | 2074 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
2075 | "String too short [regexec_flags]...\n")); |
2076 | goto phooey; | |
1aa99e6b | 2077 | } |
6eb5f6b9 | 2078 | |
1de06328 | 2079 | |
6eb5f6b9 | 2080 | /* Check validity of program. */ |
f8fc2ecf | 2081 | if (UCHARAT(progi->program) != REG_MAGIC) { |
6eb5f6b9 JH |
2082 | Perl_croak(aTHX_ "corrupted regexp program"); |
2083 | } | |
2084 | ||
2085 | PL_reg_flags = 0; | |
ed301438 | 2086 | PL_reg_state.re_state_eval_setup_done = FALSE; |
6eb5f6b9 JH |
2087 | PL_reg_maxiter = 0; |
2088 | ||
3c8556c3 | 2089 | if (RX_UTF8(rx)) |
6eb5f6b9 JH |
2090 | PL_reg_flags |= RF_utf8; |
2091 | ||
2092 | /* Mark beginning of line for ^ and lookbehind. */ | |
3b0527fe | 2093 | reginfo.bol = startpos; /* XXX not used ??? */ |
6eb5f6b9 | 2094 | PL_bostr = strbeg; |
3b0527fe | 2095 | reginfo.sv = sv; |
6eb5f6b9 JH |
2096 | |
2097 | /* Mark end of line for $ (and such) */ | |
2098 | PL_regeol = strend; | |
2099 | ||
2100 | /* see how far we have to get to not match where we matched before */ | |
3b0527fe | 2101 | reginfo.till = startpos+minend; |
6eb5f6b9 | 2102 | |
6eb5f6b9 JH |
2103 | /* If there is a "must appear" string, look for it. */ |
2104 | s = startpos; | |
2105 | ||
bbe252da | 2106 | if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ |
6eb5f6b9 | 2107 | MAGIC *mg; |
2c296965 | 2108 | if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */ |
58e23c8d | 2109 | reginfo.ganch = startpos + prog->gofs; |
2c296965 | 2110 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2111 | "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs)); |
2c296965 | 2112 | } else if (sv && SvTYPE(sv) >= SVt_PVMG |
6eb5f6b9 | 2113 | && SvMAGIC(sv) |
14befaf4 DM |
2114 | && (mg = mg_find(sv, PERL_MAGIC_regex_global)) |
2115 | && mg->mg_len >= 0) { | |
3b0527fe | 2116 | reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ |
2c296965 | 2117 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2118 | "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len)); |
2c296965 | 2119 | |
bbe252da | 2120 | if (prog->extflags & RXf_ANCH_GPOS) { |
3b0527fe | 2121 | if (s > reginfo.ganch) |
6eb5f6b9 | 2122 | goto phooey; |
58e23c8d | 2123 | s = reginfo.ganch - prog->gofs; |
2c296965 | 2124 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2125 | "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs)); |
c584a96e YO |
2126 | if (s < strbeg) |
2127 | goto phooey; | |
6eb5f6b9 JH |
2128 | } |
2129 | } | |
58e23c8d | 2130 | else if (data) { |
70685ca0 | 2131 | reginfo.ganch = strbeg + PTR2UV(data); |
2c296965 YO |
2132 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2133 | "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data))); | |
2134 | ||
2135 | } else { /* pos() not defined */ | |
3b0527fe | 2136 | reginfo.ganch = strbeg; |
2c296965 YO |
2137 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2138 | "GPOS: reginfo.ganch = strbeg\n")); | |
2139 | } | |
6eb5f6b9 | 2140 | } |
288b8c02 | 2141 | if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { |
e9105d30 GG |
2142 | /* We have to be careful. If the previous successful match |
2143 | was from this regex we don't want a subsequent partially | |
2144 | successful match to clobber the old results. | |
2145 | So when we detect this possibility we add a swap buffer | |
2146 | to the re, and switch the buffer each match. If we fail | |
2147 | we switch it back, otherwise we leave it swapped. | |
2148 | */ | |
2149 | swap = prog->offs; | |
2150 | /* do we need a save destructor here for eval dies? */ | |
2151 | Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); | |
495f47a5 DM |
2152 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
2153 | "rex=0x%"UVxf" saving offs: orig=0x%"UVxf" new=0x%"UVxf"\n", | |
2154 | PTR2UV(prog), | |
2155 | PTR2UV(swap), | |
2156 | PTR2UV(prog->offs) | |
2157 | )); | |
c74340f9 | 2158 | } |
a0714e2c | 2159 | if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { |
6eb5f6b9 JH |
2160 | re_scream_pos_data d; |
2161 | ||
2162 | d.scream_olds = &scream_olds; | |
2163 | d.scream_pos = &scream_pos; | |
288b8c02 | 2164 | s = re_intuit_start(rx, sv, s, strend, flags, &d); |
3fa9c3d7 | 2165 | if (!s) { |
a3621e74 | 2166 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n")); |
6eb5f6b9 | 2167 | goto phooey; /* not present */ |
3fa9c3d7 | 2168 | } |
6eb5f6b9 JH |
2169 | } |
2170 | ||
1de06328 | 2171 | |
6eb5f6b9 JH |
2172 | |
2173 | /* Simplest case: anchored match need be tried only once. */ | |
2174 | /* [unless only anchor is BOL and multiline is set] */ | |
bbe252da | 2175 | if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { |
24b23f37 | 2176 | if (s == startpos && regtry(®info, &startpos)) |
6eb5f6b9 | 2177 | goto got_it; |
bbe252da YO |
2178 | else if (multiline || (prog->intflags & PREGf_IMPLICIT) |
2179 | || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ | |
6eb5f6b9 JH |
2180 | { |
2181 | char *end; | |
2182 | ||
2183 | if (minlen) | |
2184 | dontbother = minlen - 1; | |
1aa99e6b | 2185 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 2186 | /* for multiline we only have to try after newlines */ |
33b8afdf | 2187 | if (prog->check_substr || prog->check_utf8) { |
92f3d482 YO |
2188 | /* because of the goto we can not easily reuse the macros for bifurcating the |
2189 | unicode/non-unicode match modes here like we do elsewhere - demerphq */ | |
2190 | if (utf8_target) { | |
2191 | if (s == startpos) | |
2192 | goto after_try_utf8; | |
2193 | while (1) { | |
2194 | if (regtry(®info, &s)) { | |
2195 | goto got_it; | |
2196 | } | |
2197 | after_try_utf8: | |
2198 | if (s > end) { | |
2199 | goto phooey; | |
2200 | } | |
2201 | if (prog->extflags & RXf_USE_INTUIT) { | |
2202 | s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL); | |
2203 | if (!s) { | |
2204 | goto phooey; | |
2205 | } | |
2206 | } | |
2207 | else { | |
2208 | s += UTF8SKIP(s); | |
2209 | } | |
2210 | } | |
2211 | } /* end search for check string in unicode */ | |
2212 | else { | |
2213 | if (s == startpos) { | |
2214 | goto after_try_latin; | |
2215 | } | |
2216 | while (1) { | |
2217 | if (regtry(®info, &s)) { | |
2218 | goto got_it; | |
2219 | } | |
2220 | after_try_latin: | |
2221 | if (s > end) { | |
2222 | goto phooey; | |
2223 | } | |
2224 | if (prog->extflags & RXf_USE_INTUIT) { | |
2225 | s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL); | |
2226 | if (!s) { | |
2227 | goto phooey; | |
2228 | } | |
2229 | } | |
2230 | else { | |
2231 | s++; | |
2232 | } | |
2233 | } | |
2234 | } /* end search for check string in latin*/ | |
2235 | } /* end search for check string */ | |
2236 | else { /* search for newline */ | |
2237 | if (s > startpos) { | |
2238 | /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/ | |
6eb5f6b9 | 2239 | s--; |
92f3d482 | 2240 | } |
21eede78 YO |
2241 | /* We can use a more efficient search as newlines are the same in unicode as they are in latin */ |
2242 | while (s <= end) { /* note it could be possible to match at the end of the string */ | |
6eb5f6b9 | 2243 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ |
24b23f37 | 2244 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2245 | goto got_it; |
2246 | } | |
92f3d482 YO |
2247 | } |
2248 | } /* end search for newline */ | |
2249 | } /* end anchored/multiline check string search */ | |
6eb5f6b9 | 2250 | goto phooey; |
bbe252da | 2251 | } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) |
f9f4320a | 2252 | { |
486ec47a | 2253 | /* the warning about reginfo.ganch being used without initialization |
bbe252da | 2254 | is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN |
f9f4320a | 2255 | and we only enter this block when the same bit is set. */ |
58e23c8d | 2256 | char *tmp_s = reginfo.ganch - prog->gofs; |
c584a96e YO |
2257 | |
2258 | if (tmp_s >= strbeg && regtry(®info, &tmp_s)) | |
6eb5f6b9 JH |
2259 | goto got_it; |
2260 | goto phooey; | |
2261 | } | |
2262 | ||
2263 | /* Messy cases: unanchored match. */ | |
bbe252da | 2264 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { |
6eb5f6b9 | 2265 | /* we have /x+whatever/ */ |
f2ed9b32 | 2266 | /* it must be a one character string (XXXX Except UTF_PATTERN?) */ |
33b8afdf | 2267 | char ch; |
bf93d4cc GS |
2268 | #ifdef DEBUGGING |
2269 | int did_match = 0; | |
2270 | #endif | |
f2ed9b32 KW |
2271 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2272 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2273 | ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0]; | |
bf93d4cc | 2274 | |
f2ed9b32 | 2275 | if (utf8_target) { |
4cadc6a9 | 2276 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2277 | if (*s == ch) { |
a3621e74 | 2278 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2279 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2280 | s += UTF8SKIP(s); |
2281 | while (s < strend && *s == ch) | |
2282 | s += UTF8SKIP(s); | |
2283 | } | |
4cadc6a9 | 2284 | ); |
6eb5f6b9 JH |
2285 | } |
2286 | else { | |
4cadc6a9 | 2287 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2288 | if (*s == ch) { |
a3621e74 | 2289 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2290 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2291 | s++; |
2292 | while (s < strend && *s == ch) | |
2293 | s++; | |
2294 | } | |
4cadc6a9 | 2295 | ); |
6eb5f6b9 | 2296 | } |
a3621e74 | 2297 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 2298 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
2299 | "Did not find anchored character...\n") |
2300 | ); | |
6eb5f6b9 | 2301 | } |
a0714e2c SS |
2302 | else if (prog->anchored_substr != NULL |
2303 | || prog->anchored_utf8 != NULL | |
2304 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
2305 | && prog->float_max_offset < strend - s)) { |
2306 | SV *must; | |
2307 | I32 back_max; | |
2308 | I32 back_min; | |
2309 | char *last; | |
6eb5f6b9 | 2310 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
2311 | #ifdef DEBUGGING |
2312 | int did_match = 0; | |
2313 | #endif | |
33b8afdf | 2314 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 KW |
2315 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2316 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2317 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; | |
33b8afdf JH |
2318 | back_max = back_min = prog->anchored_offset; |
2319 | } else { | |
f2ed9b32 KW |
2320 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2321 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2322 | must = utf8_target ? prog->float_utf8 : prog->float_substr; | |
33b8afdf JH |
2323 | back_max = prog->float_max_offset; |
2324 | back_min = prog->float_min_offset; | |
2325 | } | |
1de06328 YO |
2326 | |
2327 | ||
33b8afdf JH |
2328 | if (must == &PL_sv_undef) |
2329 | /* could not downgrade utf8 check substring, so must fail */ | |
2330 | goto phooey; | |
2331 | ||
1de06328 YO |
2332 | if (back_min<0) { |
2333 | last = strend; | |
2334 | } else { | |
2335 | last = HOP3c(strend, /* Cannot start after this */ | |
2336 | -(I32)(CHR_SVLEN(must) | |
2337 | - (SvTAIL(must) != 0) + back_min), strbeg); | |
2338 | } | |
6eb5f6b9 JH |
2339 | if (s > PL_bostr) |
2340 | last1 = HOPc(s, -1); | |
2341 | else | |
2342 | last1 = s - 1; /* bogus */ | |
2343 | ||
a0288114 | 2344 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 JH |
2345 | check_substr==must. */ |
2346 | scream_pos = -1; | |
2347 | dontbother = end_shift; | |
2348 | strend = HOPc(strend, -dontbother); | |
2349 | while ( (s <= last) && | |
c33e64f0 | 2350 | (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), |
9041c2e3 | 2351 | (unsigned char*)strend, must, |
c33e64f0 | 2352 | multiline ? FBMrf_MULTILINE : 0)) ) { |
a3621e74 | 2353 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
2354 | if (HOPc(s, -back_max) > last1) { |
2355 | last1 = HOPc(s, -back_min); | |
2356 | s = HOPc(s, -back_max); | |
2357 | } | |
2358 | else { | |
52657f30 | 2359 | char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; |
6eb5f6b9 JH |
2360 | |
2361 | last1 = HOPc(s, -back_min); | |
52657f30 | 2362 | s = t; |
6eb5f6b9 | 2363 | } |
f2ed9b32 | 2364 | if (utf8_target) { |
6eb5f6b9 | 2365 | while (s <= last1) { |
24b23f37 | 2366 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2367 | goto got_it; |
2368 | s += UTF8SKIP(s); | |
2369 | } | |
2370 | } | |
2371 | else { | |
2372 | while (s <= last1) { | |
24b23f37 | 2373 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2374 | goto got_it; |
2375 | s++; | |
2376 | } | |
2377 | } | |
2378 | } | |
ab3bbdeb | 2379 | DEBUG_EXECUTE_r(if (!did_match) { |
f2ed9b32 | 2380 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
2381 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
2382 | PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", | |
33b8afdf | 2383 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc | 2384 | ? "anchored" : "floating"), |
ab3bbdeb YO |
2385 | quoted, RE_SV_TAIL(must)); |
2386 | }); | |
6eb5f6b9 JH |
2387 | goto phooey; |
2388 | } | |
f8fc2ecf | 2389 | else if ( (c = progi->regstclass) ) { |
f14c76ed | 2390 | if (minlen) { |
f8fc2ecf | 2391 | const OPCODE op = OP(progi->regstclass); |
66e933ab | 2392 | /* don't bother with what can't match */ |
786e8c11 | 2393 | if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) |
f14c76ed RGS |
2394 | strend = HOPc(strend, -(minlen - 1)); |
2395 | } | |
a3621e74 | 2396 | DEBUG_EXECUTE_r({ |
be8e71aa | 2397 | SV * const prop = sv_newmortal(); |
32fc9b6a | 2398 | regprop(prog, prop, c); |
0df25f3d | 2399 | { |
f2ed9b32 | 2400 | RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb | 2401 | s,strend-s,60); |
0df25f3d | 2402 | PerlIO_printf(Perl_debug_log, |
1c8f8eb1 | 2403 | "Matching stclass %.*s against %s (%d bytes)\n", |
e4f74956 | 2404 | (int)SvCUR(prop), SvPVX_const(prop), |
ab3bbdeb | 2405 | quoted, (int)(strend - s)); |
0df25f3d | 2406 | } |
ffc61ed2 | 2407 | }); |
3b0527fe | 2408 | if (find_byclass(prog, c, s, strend, ®info)) |
6eb5f6b9 | 2409 | goto got_it; |
07be1b83 | 2410 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
2411 | } |
2412 | else { | |
2413 | dontbother = 0; | |
a0714e2c | 2414 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 2415 | /* Trim the end. */ |
6af40bd7 | 2416 | char *last= NULL; |
33b8afdf | 2417 | SV* float_real; |
c33e64f0 FC |
2418 | STRLEN len; |
2419 | const char *little; | |
33b8afdf | 2420 | |
f2ed9b32 KW |
2421 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2422 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2423 | float_real = utf8_target ? prog->float_utf8 : prog->float_substr; | |
d6a28714 | 2424 | |
c33e64f0 FC |
2425 | little = SvPV_const(float_real, len); |
2426 | if (SvTAIL(float_real)) { | |
1a13b075 YO |
2427 | /* This means that float_real contains an artificial \n on the end |
2428 | * due to the presence of something like this: /foo$/ | |
2429 | * where we can match both "foo" and "foo\n" at the end of the string. | |
2430 | * So we have to compare the end of the string first against the float_real | |
2431 | * without the \n and then against the full float_real with the string. | |
2432 | * We have to watch out for cases where the string might be smaller | |
2433 | * than the float_real or the float_real without the \n. | |
2434 | */ | |
2435 | char *checkpos= strend - len; | |
2436 | DEBUG_OPTIMISE_r( | |
2437 | PerlIO_printf(Perl_debug_log, | |
2438 | "%sChecking for float_real.%s\n", | |
2439 | PL_colors[4], PL_colors[5])); | |
2440 | if (checkpos + 1 < strbeg) { | |
2441 | /* can't match, even if we remove the trailing \n string is too short to match */ | |
2442 | DEBUG_EXECUTE_r( | |
2443 | PerlIO_printf(Perl_debug_log, | |
2444 | "%sString shorter than required trailing substring, cannot match.%s\n", | |
2445 | PL_colors[4], PL_colors[5])); | |
2446 | goto phooey; | |
2447 | } else if (memEQ(checkpos + 1, little, len - 1)) { | |
2448 | /* can match, the end of the string matches without the "\n" */ | |
2449 | last = checkpos + 1; | |
2450 | } else if (checkpos < strbeg) { | |
2451 | /* cant match, string is too short when the "\n" is included */ | |
2452 | DEBUG_EXECUTE_r( | |
2453 | PerlIO_printf(Perl_debug_log, | |
2454 | "%sString does not contain required trailing substring, cannot match.%s\n", | |
2455 | PL_colors[4], PL_colors[5])); | |
2456 | goto phooey; | |
2457 | } else if (!multiline) { | |
2458 | /* non multiline match, so compare with the "\n" at the end of the string */ | |
2459 | if (memEQ(checkpos, little, len)) { | |
2460 | last= checkpos; | |
2461 | } else { | |
2462 | DEBUG_EXECUTE_r( | |
2463 | PerlIO_printf(Perl_debug_log, | |
2464 | "%sString does not contain required trailing substring, cannot match.%s\n", | |
2465 | PL_colors[4], PL_colors[5])); | |
2466 | goto phooey; | |
2467 | } | |
2468 | } else { | |
2469 | /* multiline match, so we have to search for a place where the full string is located */ | |
d6a28714 | 2470 | goto find_last; |
1a13b075 | 2471 | } |
c33e64f0 | 2472 | } else { |
d6a28714 | 2473 | find_last: |
9041c2e3 | 2474 | if (len) |
d6a28714 | 2475 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 2476 | else |
a0288114 | 2477 | last = strend; /* matching "$" */ |
b8c5462f | 2478 | } |
6af40bd7 YO |
2479 | if (!last) { |
2480 | /* at one point this block contained a comment which was probably | |
2481 | * incorrect, which said that this was a "should not happen" case. | |
2482 | * Even if it was true when it was written I am pretty sure it is | |
2483 | * not anymore, so I have removed the comment and replaced it with | |
2484 | * this one. Yves */ | |
6bda09f9 YO |
2485 | DEBUG_EXECUTE_r( |
2486 | PerlIO_printf(Perl_debug_log, | |
6af40bd7 YO |
2487 | "String does not contain required substring, cannot match.\n" |
2488 | )); | |
2489 | goto phooey; | |
bf93d4cc | 2490 | } |
d6a28714 JH |
2491 | dontbother = strend - last + prog->float_min_offset; |
2492 | } | |
2493 | if (minlen && (dontbother < minlen)) | |
2494 | dontbother = minlen - 1; | |
2495 | strend -= dontbother; /* this one's always in bytes! */ | |
2496 | /* We don't know much -- general case. */ | |
f2ed9b32 | 2497 | if (utf8_target) { |
d6a28714 | 2498 | for (;;) { |
24b23f37 | 2499 | if (regtry(®info, &s)) |
d6a28714 JH |
2500 | goto got_it; |
2501 | if (s >= strend) | |
2502 | break; | |
b8c5462f | 2503 | s += UTF8SKIP(s); |
d6a28714 JH |
2504 | }; |
2505 | } | |
2506 | else { | |
2507 | do { | |
24b23f37 | 2508 | if (regtry(®info, &s)) |
d6a28714 JH |
2509 | goto got_it; |
2510 | } while (s++ < strend); | |
2511 | } | |
2512 | } | |
2513 | ||
2514 | /* Failure. */ | |
2515 | goto phooey; | |
2516 | ||
2517 | got_it: | |
495f47a5 DM |
2518 | DEBUG_BUFFERS_r( |
2519 | if (swap) | |
2520 | PerlIO_printf(Perl_debug_log, | |
2521 | "rex=0x%"UVxf" freeing offs: 0x%"UVxf"\n", | |
2522 | PTR2UV(prog), | |
2523 | PTR2UV(swap) | |
2524 | ); | |
2525 | ); | |
e9105d30 | 2526 | Safefree(swap); |
288b8c02 | 2527 | RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted); |
d6a28714 | 2528 | |
ed301438 | 2529 | if (PL_reg_state.re_state_eval_setup_done) |
4f639d21 | 2530 | restore_pos(aTHX_ prog); |
5daac39c NC |
2531 | if (RXp_PAREN_NAMES(prog)) |
2532 | (void)hv_iterinit(RXp_PAREN_NAMES(prog)); | |
d6a28714 JH |
2533 | |
2534 | /* make sure $`, $&, $', and $digit will work later */ | |
2535 | if ( !(flags & REXEC_NOT_FIRST) ) { | |
288b8c02 | 2536 | RX_MATCH_COPY_FREE(rx); |
d6a28714 | 2537 | if (flags & REXEC_COPY_STR) { |
be8e71aa | 2538 | const I32 i = PL_regeol - startpos + (stringarg - strbeg); |
f8c7b90f | 2539 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2540 | if ((SvIsCOW(sv) |
2541 | || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { | |
2542 | if (DEBUG_C_TEST) { | |
2543 | PerlIO_printf(Perl_debug_log, | |
2544 | "Copy on write: regexp capture, type %d\n", | |
2545 | (int) SvTYPE(sv)); | |
2546 | } | |
2547 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); | |
d5263905 | 2548 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
ed252734 NC |
2549 | assert (SvPOKp(prog->saved_copy)); |
2550 | } else | |
2551 | #endif | |
2552 | { | |
288b8c02 | 2553 | RX_MATCH_COPIED_on(rx); |
ed252734 NC |
2554 | s = savepvn(strbeg, i); |
2555 | prog->subbeg = s; | |
2556 | } | |
d6a28714 | 2557 | prog->sublen = i; |
d6a28714 JH |
2558 | } |
2559 | else { | |
2560 | prog->subbeg = strbeg; | |
2561 | prog->sublen = PL_regeol - strbeg; /* strend may have been modified */ | |
2562 | } | |
2563 | } | |
9041c2e3 | 2564 | |
d6a28714 JH |
2565 | return 1; |
2566 | ||
2567 | phooey: | |
a3621e74 | 2568 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 2569 | PL_colors[4], PL_colors[5])); |
ed301438 | 2570 | if (PL_reg_state.re_state_eval_setup_done) |
4f639d21 | 2571 | restore_pos(aTHX_ prog); |
e9105d30 | 2572 | if (swap) { |
c74340f9 | 2573 | /* we failed :-( roll it back */ |
495f47a5 DM |
2574 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
2575 | "rex=0x%"UVxf" rolling back offs: freeing=0x%"UVxf" restoring=0x%"UVxf"\n", | |
2576 | PTR2UV(prog), | |
2577 | PTR2UV(prog->offs), | |
2578 | PTR2UV(swap) | |
2579 | )); | |
e9105d30 GG |
2580 | Safefree(prog->offs); |
2581 | prog->offs = swap; | |
2582 | } | |
2583 | ||
d6a28714 JH |
2584 | return 0; |
2585 | } | |
2586 | ||
6bda09f9 | 2587 | |
ec43f78b DM |
2588 | /* Set which rex is pointed to by PL_reg_state, handling ref counting. |
2589 | * Do inc before dec, in case old and new rex are the same */ | |
2590 | #define SET_reg_curpm(Re2) \ | |
2591 | if (PL_reg_state.re_state_eval_setup_done) { \ | |
2592 | (void)ReREFCNT_inc(Re2); \ | |
2593 | ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); \ | |
2594 | PM_SETRE((PL_reg_curpm), (Re2)); \ | |
2595 | } | |
2596 | ||
2597 | ||
d6a28714 JH |
2598 | /* |
2599 | - regtry - try match at specific point | |
2600 | */ | |
2601 | STATIC I32 /* 0 failure, 1 success */ | |
24b23f37 | 2602 | S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) |
d6a28714 | 2603 | { |
97aff369 | 2604 | dVAR; |
d6a28714 | 2605 | CHECKPOINT lastcp; |
288b8c02 NC |
2606 | REGEXP *const rx = reginfo->prog; |
2607 | regexp *const prog = (struct regexp *)SvANY(rx); | |
f8fc2ecf | 2608 | RXi_GET_DECL(prog,progi); |
a3621e74 | 2609 | GET_RE_DEBUG_FLAGS_DECL; |
7918f24d NC |
2610 | |
2611 | PERL_ARGS_ASSERT_REGTRY; | |
2612 | ||
24b23f37 | 2613 | reginfo->cutpoint=NULL; |
d6a28714 | 2614 | |
ed301438 DM |
2615 | if ((prog->extflags & RXf_EVAL_SEEN) |
2616 | && !PL_reg_state.re_state_eval_setup_done) | |
2617 | { | |
d6a28714 JH |
2618 | MAGIC *mg; |
2619 | ||
ed301438 | 2620 | PL_reg_state.re_state_eval_setup_done = TRUE; |
3b0527fe | 2621 | if (reginfo->sv) { |
d6a28714 | 2622 | /* Make $_ available to executed code. */ |
3b0527fe | 2623 | if (reginfo->sv != DEFSV) { |
59f00321 | 2624 | SAVE_DEFSV; |
414bf5ae | 2625 | DEFSV_set(reginfo->sv); |
b8c5462f | 2626 | } |
d6a28714 | 2627 | |
3b0527fe DM |
2628 | if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) |
2629 | && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { | |
d6a28714 | 2630 | /* prepare for quick setting of pos */ |
d300d9fa | 2631 | #ifdef PERL_OLD_COPY_ON_WRITE |
51a9ea20 NC |
2632 | if (SvIsCOW(reginfo->sv)) |
2633 | sv_force_normal_flags(reginfo->sv, 0); | |
d300d9fa | 2634 | #endif |
3dab1dad | 2635 | mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, |
d300d9fa | 2636 | &PL_vtbl_mglob, NULL, 0); |
d6a28714 | 2637 | mg->mg_len = -1; |
b8c5462f | 2638 | } |
d6a28714 JH |
2639 | PL_reg_magic = mg; |
2640 | PL_reg_oldpos = mg->mg_len; | |
4f639d21 | 2641 | SAVEDESTRUCTOR_X(restore_pos, prog); |
d6a28714 | 2642 | } |
09687e5a | 2643 | if (!PL_reg_curpm) { |
a02a5408 | 2644 | Newxz(PL_reg_curpm, 1, PMOP); |
09687e5a AB |
2645 | #ifdef USE_ITHREADS |
2646 | { | |
14a49a24 | 2647 | SV* const repointer = &PL_sv_undef; |
92313705 NC |
2648 | /* this regexp is also owned by the new PL_reg_curpm, which |
2649 | will try to free it. */ | |
d2ece331 | 2650 | av_push(PL_regex_padav, repointer); |
09687e5a AB |
2651 | PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); |
2652 | PL_regex_pad = AvARRAY(PL_regex_padav); | |
2653 | } | |
2654 | #endif | |
2655 | } | |
ec43f78b | 2656 | SET_reg_curpm(rx); |
d6a28714 JH |
2657 | PL_reg_oldcurpm = PL_curpm; |
2658 | PL_curpm = PL_reg_curpm; | |
07bc277f | 2659 | if (RXp_MATCH_COPIED(prog)) { |
d6a28714 JH |
2660 | /* Here is a serious problem: we cannot rewrite subbeg, |
2661 | since it may be needed if this match fails. Thus | |
2662 | $` inside (?{}) could fail... */ | |
2663 | PL_reg_oldsaved = prog->subbeg; | |
2664 | PL_reg_oldsavedlen = prog->sublen; | |
f8c7b90f | 2665 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2666 | PL_nrs = prog->saved_copy; |
2667 | #endif | |
07bc277f | 2668 | RXp_MATCH_COPIED_off(prog); |
d6a28714 JH |
2669 | } |
2670 | else | |
bd61b366 | 2671 | PL_reg_oldsaved = NULL; |
d6a28714 JH |
2672 | prog->subbeg = PL_bostr; |
2673 | prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ | |
2674 | } | |
97ca13b7 DM |
2675 | #ifdef DEBUGGING |
2676 | PL_reg_starttry = *startpos; | |
2677 | #endif | |
f0ab9afb | 2678 | prog->offs[0].start = *startpos - PL_bostr; |
24b23f37 | 2679 | PL_reginput = *startpos; |
d6a28714 | 2680 | prog->lastparen = 0; |
03994de8 | 2681 | prog->lastcloseparen = 0; |
d6a28714 | 2682 | PL_regsize = 0; |
d6a28714 JH |
2683 | |
2684 | /* XXXX What this code is doing here?!!! There should be no need | |
b93070ed | 2685 | to do this again and again, prog->lastparen should take care of |
3dd2943c | 2686 | this! --ilya*/ |
dafc8851 JH |
2687 | |
2688 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2689 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
b93070ed | 2690 | * prog->lastparen), is not needed at all by the test suite |
225593e1 DM |
2691 | * (op/regexp, op/pat, op/split), but that code is needed otherwise |
2692 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ | |
2693 | * Meanwhile, this code *is* needed for the | |
daf18116 JH |
2694 | * above-mentioned test suite tests to succeed. The common theme |
2695 | * on those tests seems to be returning null fields from matches. | |
225593e1 | 2696 | * --jhi updated by dapm */ |
dafc8851 | 2697 | #if 1 |
d6a28714 | 2698 | if (prog->nparens) { |
b93070ed | 2699 | regexp_paren_pair *pp = prog->offs; |
097eb12c | 2700 | register I32 i; |
b93070ed | 2701 | for (i = prog->nparens; i > (I32)prog->lastparen; i--) { |
f0ab9afb NC |
2702 | ++pp; |
2703 | pp->start = -1; | |
2704 | pp->end = -1; | |
d6a28714 JH |
2705 | } |
2706 | } | |
dafc8851 | 2707 | #endif |
02db2b7b | 2708 | REGCP_SET(lastcp); |
f8fc2ecf | 2709 | if (regmatch(reginfo, progi->program + 1)) { |
b93070ed | 2710 | prog->offs[0].end = PL_reginput - PL_bostr; |
d6a28714 JH |
2711 | return 1; |
2712 | } | |
24b23f37 YO |
2713 | if (reginfo->cutpoint) |
2714 | *startpos= reginfo->cutpoint; | |
02db2b7b | 2715 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
2716 | return 0; |
2717 | } | |
2718 | ||
02db2b7b | 2719 | |
8ba1375e MJD |
2720 | #define sayYES goto yes |
2721 | #define sayNO goto no | |
262b90c4 | 2722 | #define sayNO_SILENT goto no_silent |
8ba1375e | 2723 | |
f9f4320a YO |
2724 | /* we dont use STMT_START/END here because it leads to |
2725 | "unreachable code" warnings, which are bogus, but distracting. */ | |
2726 | #define CACHEsayNO \ | |
c476f425 DM |
2727 | if (ST.cache_mask) \ |
2728 | PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \ | |
f9f4320a | 2729 | sayNO |
3298f257 | 2730 | |
a3621e74 | 2731 | /* this is used to determine how far from the left messages like |
265c4333 YO |
2732 | 'failed...' are printed. It should be set such that messages |
2733 | are inline with the regop output that created them. | |
a3621e74 | 2734 | */ |
265c4333 | 2735 | #define REPORT_CODE_OFF 32 |
a3621e74 YO |
2736 | |
2737 | ||
40a82448 DM |
2738 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
2739 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
9e137952 | 2740 | |
86545054 DM |
2741 | #define SLAB_FIRST(s) (&(s)->states[0]) |
2742 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
2743 | ||
5d9a96ca DM |
2744 | /* grab a new slab and return the first slot in it */ |
2745 | ||
2746 | STATIC regmatch_state * | |
2747 | S_push_slab(pTHX) | |
2748 | { | |
a35a87e7 | 2749 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2750 | dMY_CXT; |
2751 | #endif | |
5d9a96ca DM |
2752 | regmatch_slab *s = PL_regmatch_slab->next; |
2753 | if (!s) { | |
2754 | Newx(s, 1, regmatch_slab); | |
2755 | s->prev = PL_regmatch_slab; | |
2756 | s->next = NULL; | |
2757 | PL_regmatch_slab->next = s; | |
2758 | } | |
2759 | PL_regmatch_slab = s; | |
86545054 | 2760 | return SLAB_FIRST(s); |
5d9a96ca | 2761 | } |
5b47454d | 2762 | |
95b24440 | 2763 | |
40a82448 DM |
2764 | /* push a new state then goto it */ |
2765 | ||
2766 | #define PUSH_STATE_GOTO(state, node) \ | |
2767 | scan = node; \ | |
2768 | st->resume_state = state; \ | |
2769 | goto push_state; | |
2770 | ||
2771 | /* push a new state with success backtracking, then goto it */ | |
2772 | ||
2773 | #define PUSH_YES_STATE_GOTO(state, node) \ | |
2774 | scan = node; \ | |
2775 | st->resume_state = state; \ | |
2776 | goto push_yes_state; | |
2777 | ||
aa283a38 | 2778 | |
aa283a38 | 2779 | |
d6a28714 | 2780 | /* |
95b24440 | 2781 | |
bf1f174e DM |
2782 | regmatch() - main matching routine |
2783 | ||
2784 | This is basically one big switch statement in a loop. We execute an op, | |
2785 | set 'next' to point the next op, and continue. If we come to a point which | |
2786 | we may need to backtrack to on failure such as (A|B|C), we push a | |
2787 | backtrack state onto the backtrack stack. On failure, we pop the top | |
2788 | state, and re-enter the loop at the state indicated. If there are no more | |
2789 | states to pop, we return failure. | |
2790 | ||
2791 | Sometimes we also need to backtrack on success; for example /A+/, where | |
2792 | after successfully matching one A, we need to go back and try to | |
2793 | match another one; similarly for lookahead assertions: if the assertion | |
2794 | completes successfully, we backtrack to the state just before the assertion | |
2795 | and then carry on. In these cases, the pushed state is marked as | |
2796 | 'backtrack on success too'. This marking is in fact done by a chain of | |
2797 | pointers, each pointing to the previous 'yes' state. On success, we pop to | |
2798 | the nearest yes state, discarding any intermediate failure-only states. | |
2799 | Sometimes a yes state is pushed just to force some cleanup code to be | |
2800 | called at the end of a successful match or submatch; e.g. (??{$re}) uses | |
2801 | it to free the inner regex. | |
2802 | ||
2803 | Note that failure backtracking rewinds the cursor position, while | |
2804 | success backtracking leaves it alone. | |
2805 | ||
2806 | A pattern is complete when the END op is executed, while a subpattern | |
2807 | such as (?=foo) is complete when the SUCCESS op is executed. Both of these | |
2808 | ops trigger the "pop to last yes state if any, otherwise return true" | |
2809 | behaviour. | |
2810 | ||
2811 | A common convention in this function is to use A and B to refer to the two | |
2812 | subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is | |
2813 | the subpattern to be matched possibly multiple times, while B is the entire | |
2814 | rest of the pattern. Variable and state names reflect this convention. | |
2815 | ||
2816 | The states in the main switch are the union of ops and failure/success of | |
2817 | substates associated with with that op. For example, IFMATCH is the op | |
2818 | that does lookahead assertions /(?=A)B/ and so the IFMATCH state means | |
2819 | 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just | |
2820 | successfully matched A and IFMATCH_A_fail is a state saying that we have | |
2821 | just failed to match A. Resume states always come in pairs. The backtrack | |
2822 | state we push is marked as 'IFMATCH_A', but when that is popped, we resume | |
2823 | at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking | |
2824 | on success or failure. | |
2825 | ||
2826 | The struct that holds a backtracking state is actually a big union, with | |
2827 | one variant for each major type of op. The variable st points to the | |
2828 | top-most backtrack struct. To make the code clearer, within each | |
2829 | block of code we #define ST to alias the relevant union. | |
2830 | ||
2831 | Here's a concrete example of a (vastly oversimplified) IFMATCH | |
2832 | implementation: | |
2833 | ||
2834 | switch (state) { | |
2835 | .... | |
2836 | ||
2837 | #define ST st->u.ifmatch | |
2838 | ||
2839 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B | |
2840 | ST.foo = ...; // some state we wish to save | |
95b24440 | 2841 | ... |
bf1f174e DM |
2842 | // push a yes backtrack state with a resume value of |
2843 | // IFMATCH_A/IFMATCH_A_fail, then continue execution at the | |
2844 | // first node of A: | |
2845 | PUSH_YES_STATE_GOTO(IFMATCH_A, A); | |
2846 | // NOTREACHED | |
2847 | ||
2848 | case IFMATCH_A: // we have successfully executed A; now continue with B | |
2849 | next = B; | |
2850 | bar = ST.foo; // do something with the preserved value | |
2851 | break; | |
2852 | ||
2853 | case IFMATCH_A_fail: // A failed, so the assertion failed | |
2854 | ...; // do some housekeeping, then ... | |
2855 | sayNO; // propagate the failure | |
2856 | ||
2857 | #undef ST | |
95b24440 | 2858 | |
bf1f174e DM |
2859 | ... |
2860 | } | |
95b24440 | 2861 | |
bf1f174e DM |
2862 | For any old-timers reading this who are familiar with the old recursive |
2863 | approach, the code above is equivalent to: | |
95b24440 | 2864 | |
bf1f174e DM |
2865 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B |
2866 | { | |
2867 | int foo = ... | |
95b24440 | 2868 | ... |
bf1f174e DM |
2869 | if (regmatch(A)) { |
2870 | next = B; | |
2871 | bar = foo; | |
2872 | break; | |
95b24440 | 2873 | } |
bf1f174e DM |
2874 | ...; // do some housekeeping, then ... |
2875 | sayNO; // propagate the failure | |
95b24440 | 2876 | } |
bf1f174e DM |
2877 | |
2878 | The topmost backtrack state, pointed to by st, is usually free. If you | |
2879 | want to claim it, populate any ST.foo fields in it with values you wish to | |
2880 | save, then do one of | |
2881 | ||
2882 | PUSH_STATE_GOTO(resume_state, node); | |
2883 | PUSH_YES_STATE_GOTO(resume_state, node); | |
2884 | ||
2885 | which sets that backtrack state's resume value to 'resume_state', pushes a | |
2886 | new free entry to the top of the backtrack stack, then goes to 'node'. | |
2887 | On backtracking, the free slot is popped, and the saved state becomes the | |
2888 | new free state. An ST.foo field in this new top state can be temporarily | |
2889 | accessed to retrieve values, but once the main loop is re-entered, it | |
2890 | becomes available for reuse. | |
2891 | ||
2892 | Note that the depth of the backtrack stack constantly increases during the | |
2893 | left-to-right execution of the pattern, rather than going up and down with | |
2894 | the pattern nesting. For example the stack is at its maximum at Z at the | |
2895 | end of the pattern, rather than at X in the following: | |
2896 | ||
2897 | /(((X)+)+)+....(Y)+....Z/ | |
2898 | ||
2899 | The only exceptions to this are lookahead/behind assertions and the cut, | |
2900 | (?>A), which pop all the backtrack states associated with A before | |
2901 | continuing. | |
2902 | ||
486ec47a | 2903 | Backtrack state structs are allocated in slabs of about 4K in size. |
bf1f174e DM |
2904 | PL_regmatch_state and st always point to the currently active state, |
2905 | and PL_regmatch_slab points to the slab currently containing | |
2906 | PL_regmatch_state. The first time regmatch() is called, the first slab is | |
2907 | allocated, and is never freed until interpreter destruction. When the slab | |
2908 | is full, a new one is allocated and chained to the end. At exit from | |
2909 | regmatch(), slabs allocated since entry are freed. | |
2910 | ||
2911 | */ | |
95b24440 | 2912 | |
40a82448 | 2913 | |
5bc10b2c | 2914 | #define DEBUG_STATE_pp(pp) \ |
265c4333 | 2915 | DEBUG_STATE_r({ \ |
f2ed9b32 | 2916 | DUMP_EXEC_POS(locinput, scan, utf8_target); \ |
5bc10b2c | 2917 | PerlIO_printf(Perl_debug_log, \ |
5d458dd8 | 2918 | " %*s"pp" %s%s%s%s%s\n", \ |
5bc10b2c | 2919 | depth*2, "", \ |
13d6edb4 | 2920 | PL_reg_name[st->resume_state], \ |
5d458dd8 YO |
2921 | ((st==yes_state||st==mark_state) ? "[" : ""), \ |
2922 | ((st==yes_state) ? "Y" : ""), \ | |
2923 | ((st==mark_state) ? "M" : ""), \ | |
2924 | ((st==yes_state||st==mark_state) ? "]" : "") \ | |
2925 | ); \ | |
265c4333 | 2926 | }); |
5bc10b2c | 2927 | |
40a82448 | 2928 | |
3dab1dad | 2929 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 2930 | |
3df15adc | 2931 | #ifdef DEBUGGING |
5bc10b2c | 2932 | |
ab3bbdeb | 2933 | STATIC void |
f2ed9b32 | 2934 | S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target, |
ab3bbdeb YO |
2935 | const char *start, const char *end, const char *blurb) |
2936 | { | |
efd26800 | 2937 | const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; |
7918f24d NC |
2938 | |
2939 | PERL_ARGS_ASSERT_DEBUG_START_MATCH; | |
2940 | ||
ab3bbdeb YO |
2941 | if (!PL_colorset) |
2942 | reginitcolors(); | |
2943 | { | |
2944 | RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), | |
d2c6dc5e | 2945 | RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); |
ab3bbdeb | 2946 | |
f2ed9b32 | 2947 | RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb YO |
2948 | start, end - start, 60); |
2949 | ||
2950 | PerlIO_printf(Perl_debug_log, | |
2951 | "%s%s REx%s %s against %s\n", | |
2952 | PL_colors[4], blurb, PL_colors[5], s0, s1); | |
2953 | ||
f2ed9b32 | 2954 | if (utf8_target||utf8_pat) |
1de06328 YO |
2955 | PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", |
2956 | utf8_pat ? "pattern" : "", | |
f2ed9b32 KW |
2957 | utf8_pat && utf8_target ? " and " : "", |
2958 | utf8_target ? "string" : "" | |
ab3bbdeb YO |
2959 | ); |
2960 | } | |
2961 | } | |
3df15adc YO |
2962 | |
2963 | STATIC void | |
786e8c11 YO |
2964 | S_dump_exec_pos(pTHX_ const char *locinput, |
2965 | const regnode *scan, | |
2966 | const char *loc_regeol, | |
2967 | const char *loc_bostr, | |
2968 | const char *loc_reg_starttry, | |
f2ed9b32 | 2969 | const bool utf8_target) |
07be1b83 | 2970 | { |
786e8c11 | 2971 | const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; |
07be1b83 | 2972 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ |
786e8c11 | 2973 | int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); |
07be1b83 YO |
2974 | /* The part of the string before starttry has one color |
2975 | (pref0_len chars), between starttry and current | |
2976 | position another one (pref_len - pref0_len chars), | |
2977 | after the current position the third one. | |
2978 | We assume that pref0_len <= pref_len, otherwise we | |
2979 | decrease pref0_len. */ | |
786e8c11 YO |
2980 | int pref_len = (locinput - loc_bostr) > (5 + taill) - l |
2981 | ? (5 + taill) - l : locinput - loc_bostr; | |
07be1b83 YO |
2982 | int pref0_len; |
2983 | ||
7918f24d NC |
2984 | PERL_ARGS_ASSERT_DUMP_EXEC_POS; |
2985 | ||
f2ed9b32 | 2986 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) |
07be1b83 | 2987 | pref_len++; |
786e8c11 YO |
2988 | pref0_len = pref_len - (locinput - loc_reg_starttry); |
2989 | if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) | |
2990 | l = ( loc_regeol - locinput > (5 + taill) - pref_len | |
2991 | ? (5 + taill) - pref_len : loc_regeol - locinput); | |
f2ed9b32 | 2992 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) |
07be1b83 YO |
2993 | l--; |
2994 | if (pref0_len < 0) | |
2995 | pref0_len = 0; | |
2996 | if (pref0_len > pref_len) | |
2997 | pref0_len = pref_len; | |
2998 | { | |
f2ed9b32 | 2999 | const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0; |
0df25f3d | 3000 | |
ab3bbdeb | 3001 | RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), |
1de06328 | 3002 | (locinput - pref_len),pref0_len, 60, 4, 5); |
0df25f3d | 3003 | |
ab3bbdeb | 3004 | RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), |
3df15adc | 3005 | (locinput - pref_len + pref0_len), |
1de06328 | 3006 | pref_len - pref0_len, 60, 2, 3); |
0df25f3d | 3007 | |
ab3bbdeb | 3008 | RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), |
1de06328 | 3009 | locinput, loc_regeol - locinput, 10, 0, 1); |
0df25f3d | 3010 | |
1de06328 | 3011 | const STRLEN tlen=len0+len1+len2; |
3df15adc | 3012 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb | 3013 | "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", |
786e8c11 | 3014 | (IV)(locinput - loc_bostr), |
07be1b83 | 3015 | len0, s0, |
07be1b83 | 3016 | len1, s1, |
07be1b83 | 3017 | (docolor ? "" : "> <"), |
07be1b83 | 3018 | len2, s2, |
f9f4320a | 3019 | (int)(tlen > 19 ? 0 : 19 - tlen), |
07be1b83 YO |
3020 | ""); |
3021 | } | |
3022 | } | |
3df15adc | 3023 | |
07be1b83 YO |
3024 | #endif |
3025 | ||
0a4db386 YO |
3026 | /* reg_check_named_buff_matched() |
3027 | * Checks to see if a named buffer has matched. The data array of | |
3028 | * buffer numbers corresponding to the buffer is expected to reside | |
3029 | * in the regexp->data->data array in the slot stored in the ARG() of | |
3030 | * node involved. Note that this routine doesn't actually care about the | |
3031 | * name, that information is not preserved from compilation to execution. | |
3032 | * Returns the index of the leftmost defined buffer with the given name | |
3033 | * or 0 if non of the buffers matched. | |
3034 | */ | |
3035 | STATIC I32 | |
7918f24d NC |
3036 | S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) |
3037 | { | |
0a4db386 | 3038 | I32 n; |
f8fc2ecf | 3039 | RXi_GET_DECL(rex,rexi); |
ad64d0ec | 3040 | SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
0a4db386 | 3041 | I32 *nums=(I32*)SvPVX(sv_dat); |
7918f24d NC |
3042 | |
3043 | PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; | |
3044 | ||
0a4db386 | 3045 | for ( n=0; n<SvIVX(sv_dat); n++ ) { |
b93070ed DM |
3046 | if ((I32)rex->lastparen >= nums[n] && |
3047 | rex->offs[nums[n]].end != -1) | |
0a4db386 YO |
3048 | { |
3049 | return nums[n]; | |
3050 | } | |
3051 | } | |
3052 | return 0; | |
3053 | } | |
3054 | ||
2f554ef7 DM |
3055 | |
3056 | /* free all slabs above current one - called during LEAVE_SCOPE */ | |
3057 | ||
3058 | STATIC void | |
3059 | S_clear_backtrack_stack(pTHX_ void *p) | |
3060 | { | |
3061 | regmatch_slab *s = PL_regmatch_slab->next; | |
3062 | PERL_UNUSED_ARG(p); | |
3063 | ||
3064 | if (!s) | |
3065 | return; | |
3066 | PL_regmatch_slab->next = NULL; | |
3067 | while (s) { | |
3068 | regmatch_slab * const osl = s; | |
3069 | s = s->next; | |
3070 | Safefree(osl); | |
3071 | } | |
3072 | } | |
3073 | ||
3074 | ||
d6a28714 | 3075 | STATIC I32 /* 0 failure, 1 success */ |
24b23f37 | 3076 | S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) |
d6a28714 | 3077 | { |
a35a87e7 | 3078 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
3079 | dMY_CXT; |
3080 | #endif | |
27da23d5 | 3081 | dVAR; |
f2ed9b32 | 3082 | register const bool utf8_target = PL_reg_match_utf8; |
4ad0818d | 3083 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
288b8c02 NC |
3084 | REGEXP *rex_sv = reginfo->prog; |
3085 | regexp *rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 3086 | RXi_GET_DECL(rex,rexi); |
2f554ef7 | 3087 | I32 oldsave; |
5d9a96ca DM |
3088 | /* the current state. This is a cached copy of PL_regmatch_state */ |
3089 | register regmatch_state *st; | |
5d9a96ca DM |
3090 | /* cache heavy used fields of st in registers */ |
3091 | register regnode *scan; | |
3092 | register regnode *next; | |
438e9bae | 3093 | register U32 n = 0; /* general value; init to avoid compiler warning */ |
24d3c4a9 | 3094 | register I32 ln = 0; /* len or last; init to avoid compiler warning */ |
5d9a96ca | 3095 | register char *locinput = PL_reginput; |
5d9a96ca | 3096 | register I32 nextchr; /* is always set to UCHARAT(locinput) */ |
24d3c4a9 | 3097 | |
b69b0499 | 3098 | bool result = 0; /* return value of S_regmatch */ |
24d3c4a9 | 3099 | int depth = 0; /* depth of backtrack stack */ |
4b196cd4 YO |
3100 | U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */ |
3101 | const U32 max_nochange_depth = | |
3102 | (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? | |
3103 | 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; | |
77cb431f DM |
3104 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
3105 | subpattern */ | |
e2e6a0f1 YO |
3106 | /* mark_state piggy backs on the yes_state logic so that when we unwind |
3107 | the stack on success we can update the mark_state as we go */ | |
3108 | regmatch_state *mark_state = NULL; /* last mark state we have seen */ | |
faec1544 | 3109 | regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ |
b8591aee | 3110 | struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ |
40a82448 | 3111 | U32 state_num; |
5d458dd8 YO |
3112 | bool no_final = 0; /* prevent failure from backtracking? */ |
3113 | bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ | |
e2e6a0f1 | 3114 | char *startpoint = PL_reginput; |
5d458dd8 YO |
3115 | SV *popmark = NULL; /* are we looking for a mark? */ |
3116 | SV *sv_commit = NULL; /* last mark name seen in failure */ | |
3117 | SV *sv_yes_mark = NULL; /* last mark name we have seen | |
486ec47a | 3118 | during a successful match */ |
5d458dd8 YO |
3119 | U32 lastopen = 0; /* last open we saw */ |
3120 | bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; | |
19b95bf0 | 3121 | SV* const oreplsv = GvSV(PL_replgv); |
24d3c4a9 DM |
3122 | /* these three flags are set by various ops to signal information to |
3123 | * the very next op. They have a useful lifetime of exactly one loop | |
3124 | * iteration, and are not preserved or restored by state pushes/pops | |
3125 | */ | |
3126 | bool sw = 0; /* the condition value in (?(cond)a|b) */ | |
3127 | bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ | |
3128 | int logical = 0; /* the following EVAL is: | |
3129 | 0: (?{...}) | |
3130 | 1: (?(?{...})X|Y) | |
3131 | 2: (??{...}) | |
3132 | or the following IFMATCH/UNLESSM is: | |
3133 | false: plain (?=foo) | |
3134 | true: used as a condition: (?(?=foo)) | |
3135 | */ | |
81ed78b2 DM |
3136 | PAD* last_pad = NULL; |
3137 | dMULTICALL; | |
3138 | I32 gimme = G_SCALAR; | |
3139 | CV *caller_cv = NULL; /* who called us */ | |
3140 | CV *last_pushed_cv = NULL; /* most recently called (?{}) CV */ | |
3141 | ||
95b24440 | 3142 | #ifdef DEBUGGING |
e68ec53f | 3143 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
3144 | #endif |
3145 | ||
81ed78b2 DM |
3146 | /* shut up 'may be used uninitialized' compiler warnings for dMULTICALL */ |
3147 | multicall_oldcatch = 0; | |
3148 | multicall_cv = NULL; | |
3149 | cx = NULL; | |
3150 | ||
3151 | ||
7918f24d NC |
3152 | PERL_ARGS_ASSERT_REGMATCH; |
3153 | ||
3b57cd43 | 3154 | DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ |
24b23f37 | 3155 | PerlIO_printf(Perl_debug_log,"regmatch start\n"); |
3b57cd43 | 3156 | })); |
5d9a96ca DM |
3157 | /* on first ever call to regmatch, allocate first slab */ |
3158 | if (!PL_regmatch_slab) { | |
3159 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
3160 | PL_regmatch_slab->prev = NULL; | |
3161 | PL_regmatch_slab->next = NULL; | |
86545054 | 3162 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); |
5d9a96ca DM |
3163 | } |
3164 | ||
2f554ef7 DM |
3165 | oldsave = PL_savestack_ix; |
3166 | SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL); | |
3167 | SAVEVPTR(PL_regmatch_slab); | |
3168 | SAVEVPTR(PL_regmatch_state); | |
5d9a96ca DM |
3169 | |
3170 | /* grab next free state slot */ | |
3171 | st = ++PL_regmatch_state; | |
86545054 | 3172 | if (st > SLAB_LAST(PL_regmatch_slab)) |
5d9a96ca DM |
3173 | st = PL_regmatch_state = S_push_slab(aTHX); |
3174 | ||
d6a28714 JH |
3175 | /* Note that nextchr is a byte even in UTF */ |
3176 | nextchr = UCHARAT(locinput); | |
3177 | scan = prog; | |
3178 | while (scan != NULL) { | |
8ba1375e | 3179 | |
a3621e74 | 3180 | DEBUG_EXECUTE_r( { |
6136c704 | 3181 | SV * const prop = sv_newmortal(); |
1de06328 | 3182 | regnode *rnext=regnext(scan); |
f2ed9b32 | 3183 | DUMP_EXEC_POS( locinput, scan, utf8_target ); |
32fc9b6a | 3184 | regprop(rex, prop, scan); |
07be1b83 YO |
3185 | |
3186 | PerlIO_printf(Perl_debug_log, | |
3187 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
f8fc2ecf | 3188 | (IV)(scan - rexi->program), depth*2, "", |
07be1b83 | 3189 | SvPVX_const(prop), |
1de06328 | 3190 | (PL_regkind[OP(scan)] == END || !rnext) ? |
f8fc2ecf | 3191 | 0 : (IV)(rnext - rexi->program)); |
2a782b5b | 3192 | }); |
d6a28714 JH |
3193 | |
3194 | next = scan + NEXT_OFF(scan); | |
3195 | if (next == scan) | |
3196 | next = NULL; | |
40a82448 | 3197 | state_num = OP(scan); |
d6a28714 | 3198 | |
40a82448 | 3199 | reenter_switch: |
34a81e2b | 3200 | |
40a82448 | 3201 | switch (state_num) { |
d6a28714 | 3202 | case BOL: |
7fba1cd6 | 3203 | if (locinput == PL_bostr) |
d6a28714 | 3204 | { |
3b0527fe | 3205 | /* reginfo->till = reginfo->bol; */ |
b8c5462f JH |
3206 | break; |
3207 | } | |
d6a28714 JH |
3208 | sayNO; |
3209 | case MBOL: | |
12d33761 HS |
3210 | if (locinput == PL_bostr || |
3211 | ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n')) | |
d6a28714 | 3212 | { |
b8c5462f JH |
3213 | break; |
3214 | } | |
d6a28714 JH |
3215 | sayNO; |
3216 | case SBOL: | |
c2a73568 | 3217 | if (locinput == PL_bostr) |
b8c5462f | 3218 | break; |
d6a28714 JH |
3219 | sayNO; |
3220 | case GPOS: | |
3b0527fe | 3221 | if (locinput == reginfo->ganch) |
d6a28714 JH |
3222 | break; |
3223 | sayNO; | |
ee9b8eae YO |
3224 | |
3225 | case KEEPS: | |
3226 | /* update the startpoint */ | |
b93070ed | 3227 | st->u.keeper.val = rex->offs[0].start; |
ee9b8eae | 3228 | PL_reginput = locinput; |
b93070ed | 3229 | rex->offs[0].start = locinput - PL_bostr; |
ee9b8eae YO |
3230 | PUSH_STATE_GOTO(KEEPS_next, next); |
3231 | /*NOT-REACHED*/ | |
3232 | case KEEPS_next_fail: | |
3233 | /* rollback the start point change */ | |
b93070ed | 3234 | rex->offs[0].start = st->u.keeper.val; |
ee9b8eae YO |
3235 | sayNO_SILENT; |
3236 | /*NOT-REACHED*/ | |
d6a28714 | 3237 | case EOL: |
d6a28714 JH |
3238 | goto seol; |
3239 | case MEOL: | |
d6a28714 | 3240 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') |
b8c5462f | 3241 | sayNO; |
b8c5462f | 3242 | break; |
d6a28714 JH |
3243 | case SEOL: |
3244 | seol: | |
3245 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') | |
b8c5462f | 3246 | sayNO; |
d6a28714 | 3247 | if (PL_regeol - locinput > 1) |
b8c5462f | 3248 | sayNO; |
b8c5462f | 3249 | break; |
d6a28714 JH |
3250 | case EOS: |
3251 | if (PL_regeol != locinput) | |
b8c5462f | 3252 | sayNO; |
d6a28714 | 3253 | break; |
ffc61ed2 | 3254 | case SANY: |
d6a28714 | 3255 | if (!nextchr && locinput >= PL_regeol) |
4633a7c4 | 3256 | sayNO; |
f2ed9b32 | 3257 | if (utf8_target) { |
f33976b4 DB |
3258 | locinput += PL_utf8skip[nextchr]; |
3259 | if (locinput > PL_regeol) | |
3260 | sayNO; | |
3261 | nextchr = UCHARAT(locinput); | |
3262 | } | |
3263 | else | |
3264 | nextchr = UCHARAT(++locinput); | |
3265 | break; | |
3266 | case CANY: | |
3267 | if (!nextchr && locinput >= PL_regeol) | |
3268 | sayNO; | |
b8c5462f | 3269 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 3270 | break; |
ffc61ed2 | 3271 | case REG_ANY: |
1aa99e6b IH |
3272 | if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n') |
3273 | sayNO; | |
f2ed9b32 | 3274 | if (utf8_target) { |
b8c5462f | 3275 | locinput += PL_utf8skip[nextchr]; |
d6a28714 JH |
3276 | if (locinput > PL_regeol) |
3277 | sayNO; | |
a0ed51b3 | 3278 | nextchr = UCHARAT(locinput); |
a0ed51b3 | 3279 | } |
1aa99e6b IH |
3280 | else |
3281 | nextchr = UCHARAT(++locinput); | |
a0ed51b3 | 3282 | break; |
166ba7cd DM |
3283 | |
3284 | #undef ST | |
3285 | #define ST st->u.trie | |
786e8c11 YO |
3286 | case TRIEC: |
3287 | /* In this case the charclass data is available inline so | |
3288 | we can fail fast without a lot of extra overhead. | |
3289 | */ | |
fab2782b YO |
3290 | if(!ANYOF_BITMAP_TEST(scan, *locinput)) { |
3291 | DEBUG_EXECUTE_r( | |
3292 | PerlIO_printf(Perl_debug_log, | |
3293 | "%*s %sfailed to match trie start class...%s\n", | |
3294 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) | |
3295 | ); | |
3296 | sayNO_SILENT; | |
3297 | /* NOTREACHED */ | |
786e8c11 YO |
3298 | } |
3299 | /* FALL THROUGH */ | |
5b47454d | 3300 | case TRIE: |
2e64971a DM |
3301 | /* the basic plan of execution of the trie is: |
3302 | * At the beginning, run though all the states, and | |
3303 | * find the longest-matching word. Also remember the position | |
3304 | * of the shortest matching word. For example, this pattern: | |
3305 | * 1 2 3 4 5 | |
3306 | * ab|a|x|abcd|abc | |
3307 | * when matched against the string "abcde", will generate | |
3308 | * accept states for all words except 3, with the longest | |
3309 | * matching word being 4, and the shortest being 1 (with | |
3310 | * the position being after char 1 of the string). | |
3311 | * | |
3312 | * Then for each matching word, in word order (i.e. 1,2,4,5), | |
3313 | * we run the remainder of the pattern; on each try setting | |
3314 | * the current position to the character following the word, | |
3315 | * returning to try the next word on failure. | |
3316 | * | |
3317 | * We avoid having to build a list of words at runtime by | |
3318 | * using a compile-time structure, wordinfo[].prev, which | |
3319 | * gives, for each word, the previous accepting word (if any). | |
3320 | * In the case above it would contain the mappings 1->2, 2->0, | |
3321 | * 3->0, 4->5, 5->1. We can use this table to generate, from | |
3322 | * the longest word (4 above), a list of all words, by | |
3323 | * following the list of prev pointers; this gives us the | |
3324 | * unordered list 4,5,1,2. Then given the current word we have | |
3325 | * just tried, we can go through the list and find the | |
3326 | * next-biggest word to try (so if we just failed on word 2, | |
3327 | * the next in the list is 4). | |
3328 | * | |
3329 | * Since at runtime we don't record the matching position in | |
3330 | * the string for each word, we have to work that out for | |
3331 | * each word we're about to process. The wordinfo table holds | |
3332 | * the character length of each word; given that we recorded | |
3333 | * at the start: the position of the shortest word and its | |
3334 | * length in chars, we just need to move the pointer the | |
3335 | * difference between the two char lengths. Depending on | |
3336 | * Unicode status and folding, that's cheap or expensive. | |
3337 | * | |
3338 | * This algorithm is optimised for the case where are only a | |
3339 | * small number of accept states, i.e. 0,1, or maybe 2. | |
3340 | * With lots of accepts states, and having to try all of them, | |
3341 | * it becomes quadratic on number of accept states to find all | |
3342 | * the next words. | |
3343 | */ | |
3344 | ||
3dab1dad | 3345 | { |
07be1b83 | 3346 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
a0a388a1 | 3347 | DECL_TRIE_TYPE(scan); |
3dab1dad YO |
3348 | |
3349 | /* what trie are we using right now */ | |
be8e71aa | 3350 | reg_trie_data * const trie |
f8fc2ecf | 3351 | = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; |
85fbaab2 | 3352 | HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); |
3dab1dad | 3353 | U32 state = trie->startstate; |
166ba7cd | 3354 | |
fab2782b | 3355 | if (trie->bitmap && !TRIE_BITMAP_TEST(trie,*locinput) ) { |
3dab1dad YO |
3356 | if (trie->states[ state ].wordnum) { |
3357 | DEBUG_EXECUTE_r( | |
3358 | PerlIO_printf(Perl_debug_log, | |
3359 | "%*s %smatched empty string...%s\n", | |
5bc10b2c | 3360 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad | 3361 | ); |
20dbff7c YO |
3362 | if (!trie->jump) |
3363 | break; | |
3dab1dad YO |
3364 | } else { |
3365 | DEBUG_EXECUTE_r( | |
3366 | PerlIO_printf(Perl_debug_log, | |
786e8c11 | 3367 | "%*s %sfailed to match trie start class...%s\n", |
5bc10b2c | 3368 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad YO |
3369 | ); |
3370 | sayNO_SILENT; | |
3371 | } | |
3372 | } | |
166ba7cd | 3373 | |
786e8c11 YO |
3374 | { |
3375 | U8 *uc = ( U8* )locinput; | |
3376 | ||
3377 | STRLEN len = 0; | |
3378 | STRLEN foldlen = 0; | |
3379 | U8 *uscan = (U8*)NULL; | |
786e8c11 | 3380 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
2e64971a DM |
3381 | U32 charcount = 0; /* how many input chars we have matched */ |
3382 | U32 accepted = 0; /* have we seen any accepting states? */ | |
786e8c11 | 3383 | |
786e8c11 | 3384 | ST.jump = trie->jump; |
786e8c11 | 3385 | ST.me = scan; |
2e64971a DM |
3386 | ST.firstpos = NULL; |
3387 | ST.longfold = FALSE; /* char longer if folded => it's harder */ | |
3388 | ST.nextword = 0; | |
3389 | ||
3390 | /* fully traverse the TRIE; note the position of the | |
3391 | shortest accept state and the wordnum of the longest | |
3392 | accept state */ | |
07be1b83 | 3393 | |
a3621e74 | 3394 | while ( state && uc <= (U8*)PL_regeol ) { |
786e8c11 | 3395 | U32 base = trie->states[ state ].trans.base; |
f9f4320a | 3396 | UV uvc = 0; |
acb909b4 | 3397 | U16 charid = 0; |
2e64971a DM |
3398 | U16 wordnum; |
3399 | wordnum = trie->states[ state ].wordnum; | |
3400 | ||
3401 | if (wordnum) { /* it's an accept state */ | |
3402 | if (!accepted) { | |
3403 | accepted = 1; | |
3404 | /* record first match position */ | |
3405 | if (ST.longfold) { | |
3406 | ST.firstpos = (U8*)locinput; | |
3407 | ST.firstchars = 0; | |
5b47454d | 3408 | } |
2e64971a DM |
3409 | else { |
3410 | ST.firstpos = uc; | |
3411 | ST.firstchars = charcount; | |
3412 | } | |
3413 | } | |
3414 | if (!ST.nextword || wordnum < ST.nextword) | |
3415 | ST.nextword = wordnum; | |
3416 | ST.topword = wordnum; | |
786e8c11 | 3417 | } |
a3621e74 | 3418 | |
07be1b83 | 3419 | DEBUG_TRIE_EXECUTE_r({ |
f2ed9b32 | 3420 | DUMP_EXEC_POS( (char *)uc, scan, utf8_target ); |
a3621e74 | 3421 | PerlIO_printf( Perl_debug_log, |
2e64971a | 3422 | "%*s %sState: %4"UVxf" Accepted: %c ", |