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 | |
c277df42 IZ |
83 | #define RF_tainted 1 /* tainted information used? */ |
84 | #define RF_warned 2 /* warned about big count? */ | |
faec1544 | 85 | |
ab3bbdeb | 86 | #define RF_utf8 8 /* Pattern contains multibyte chars? */ |
a0ed51b3 | 87 | |
f2ed9b32 | 88 | #define UTF_PATTERN ((PL_reg_flags & RF_utf8) != 0) |
ce862d02 IZ |
89 | |
90 | #define RS_init 1 /* eval environment created */ | |
91 | #define RS_set 2 /* replsv value is set */ | |
c277df42 | 92 | |
a687059c LW |
93 | #ifndef STATIC |
94 | #define STATIC static | |
95 | #endif | |
96 | ||
af364d03 KW |
97 | /* Valid for non-utf8 strings only: avoids the reginclass call if there are no |
98 | * complications: i.e., if everything matchable is straight forward in the | |
99 | * bitmap */ | |
100 | #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) \ | |
101 | : ANYOF_BITMAP_TEST(p,*(c))) | |
7d3e948e | 102 | |
c277df42 IZ |
103 | /* |
104 | * Forwards. | |
105 | */ | |
106 | ||
f2ed9b32 | 107 | #define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv)) |
53c4c00c | 108 | #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b) |
a0ed51b3 | 109 | |
3dab1dad YO |
110 | #define HOPc(pos,off) \ |
111 | (char *)(PL_reg_match_utf8 \ | |
52657f30 | 112 | ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \ |
3dab1dad YO |
113 | : (U8*)(pos + off)) |
114 | #define HOPBACKc(pos, off) \ | |
07be1b83 YO |
115 | (char*)(PL_reg_match_utf8\ |
116 | ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \ | |
117 | : (pos - off >= PL_bostr) \ | |
8e11feef | 118 | ? (U8*)pos - off \ |
3dab1dad | 119 | : NULL) |
efb30f32 | 120 | |
e7409c1b | 121 | #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off)) |
1aa99e6b | 122 | #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim)) |
1aa99e6b | 123 | |
20d0b1e9 | 124 | /* these are unrolled below in the CCC_TRY_XXX defined */ |
1a4fad37 AL |
125 | #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ |
126 | if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END | |
37e2e78e KW |
127 | |
128 | /* Doesn't do an assert to verify that is correct */ | |
129 | #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \ | |
c269f403 | 130 | if (!CAT2(PL_utf8_,class)) { bool throw_away; ENTER; save_re_context(); throw_away = CAT2(is_utf8_,class)((const U8*)" "); LEAVE; } } STMT_END |
37e2e78e | 131 | |
1a4fad37 AL |
132 | #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a") |
133 | #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0") | |
134 | #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ") | |
51371543 | 135 | |
37e2e78e | 136 | #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \ |
d275fa5e CB |
137 | LOAD_UTF8_CHARCLASS(X_begin, " "); \ |
138 | LOAD_UTF8_CHARCLASS(X_non_hangul, "A"); \ | |
37e2e78e KW |
139 | /* These are utf8 constants, and not utf-ebcdic constants, so the \ |
140 | * assert should likely and hopefully fail on an EBCDIC machine */ \ | |
d275fa5e | 141 | LOAD_UTF8_CHARCLASS(X_extend, "\xcc\x80"); /* U+0300 */ \ |
37e2e78e KW |
142 | \ |
143 | /* No asserts are done for these, in case called on an early \ | |
144 | * Unicode version in which they map to nothing */ \ | |
d275fa5e CB |
145 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_prepend);/* U+0E40 "\xe0\xb9\x80" */ \ |
146 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_L); /* U+1100 "\xe1\x84\x80" */ \ | |
147 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV); /* U+AC00 "\xea\xb0\x80" */ \ | |
148 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LVT); /* U+AC01 "\xea\xb0\x81" */ \ | |
149 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV_LVT_V);/* U+AC01 "\xea\xb0\x81" */\ | |
150 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_T); /* U+11A8 "\xe1\x86\xa8" */ \ | |
37e2e78e | 151 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_V) /* U+1160 "\xe1\x85\xa0" */ |
20d0b1e9 | 152 | |
d1eb3177 YO |
153 | /* |
154 | We dont use PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS as the direct test | |
155 | so that it is possible to override the option here without having to | |
156 | rebuild the entire core. as we are required to do if we change regcomp.h | |
157 | which is where PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS is defined. | |
158 | */ | |
159 | #if PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS | |
160 | #define BROKEN_UNICODE_CHARCLASS_MAPPINGS | |
161 | #endif | |
162 | ||
163 | #ifdef BROKEN_UNICODE_CHARCLASS_MAPPINGS | |
164 | #define LOAD_UTF8_CHARCLASS_PERL_WORD() LOAD_UTF8_CHARCLASS_ALNUM() | |
165 | #define LOAD_UTF8_CHARCLASS_PERL_SPACE() LOAD_UTF8_CHARCLASS_SPACE() | |
166 | #define LOAD_UTF8_CHARCLASS_POSIX_DIGIT() LOAD_UTF8_CHARCLASS_DIGIT() | |
167 | #define RE_utf8_perl_word PL_utf8_alnum | |
168 | #define RE_utf8_perl_space PL_utf8_space | |
169 | #define RE_utf8_posix_digit PL_utf8_digit | |
170 | #define perl_word alnum | |
171 | #define perl_space space | |
172 | #define posix_digit digit | |
173 | #else | |
174 | #define LOAD_UTF8_CHARCLASS_PERL_WORD() LOAD_UTF8_CHARCLASS(perl_word,"a") | |
175 | #define LOAD_UTF8_CHARCLASS_PERL_SPACE() LOAD_UTF8_CHARCLASS(perl_space," ") | |
176 | #define LOAD_UTF8_CHARCLASS_POSIX_DIGIT() LOAD_UTF8_CHARCLASS(posix_digit,"0") | |
177 | #define RE_utf8_perl_word PL_utf8_perl_word | |
178 | #define RE_utf8_perl_space PL_utf8_perl_space | |
179 | #define RE_utf8_posix_digit PL_utf8_posix_digit | |
180 | #endif | |
181 | ||
1dcf4a1b | 182 | #define PLACEHOLDER /* Something for the preprocessor to grab onto */ |
d1eb3177 | 183 | |
ee9a90b8 KW |
184 | /* The actual code for CCC_TRY, which uses several variables from the routine |
185 | * it's callable from. It is designed to be the bulk of a case statement. | |
186 | * FUNC is the macro or function to call on non-utf8 targets that indicate if | |
187 | * nextchr matches the class. | |
188 | * UTF8_TEST is the whole test string to use for utf8 targets | |
189 | * LOAD is what to use to test, and if not present to load in the swash for the | |
190 | * class | |
191 | * POS_OR_NEG is either empty or ! to complement the results of FUNC or | |
192 | * UTF8_TEST test. | |
193 | * The logic is: Fail if we're at the end-of-string; otherwise if the target is | |
194 | * utf8 and a variant, load the swash if necessary and test using the utf8 | |
195 | * test. Advance to the next character if test is ok, otherwise fail; If not | |
196 | * utf8 or an invariant under utf8, use the non-utf8 test, and fail if it | |
197 | * fails, or advance to the next character */ | |
198 | ||
199 | #define _CCC_TRY_CODE(POS_OR_NEG, FUNC, UTF8_TEST, CLASS, STR) \ | |
200 | if (locinput >= PL_regeol) { \ | |
201 | sayNO; \ | |
202 | } \ | |
203 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ | |
204 | LOAD_UTF8_CHARCLASS(CLASS, STR); \ | |
205 | if (POS_OR_NEG (UTF8_TEST)) { \ | |
206 | sayNO; \ | |
207 | } \ | |
208 | locinput += PL_utf8skip[nextchr]; \ | |
209 | nextchr = UCHARAT(locinput); \ | |
210 | break; \ | |
211 | } \ | |
212 | if (POS_OR_NEG (FUNC(nextchr))) { \ | |
213 | sayNO; \ | |
214 | } \ | |
215 | nextchr = UCHARAT(++locinput); \ | |
980866de KW |
216 | break; |
217 | ||
ee9a90b8 KW |
218 | /* Handle the non-locale cases for a character class and its complement. It |
219 | * calls _CCC_TRY_CODE with a ! to complement the test for the character class. | |
220 | * This is because that code fails when the test succeeds, so we want to have | |
221 | * the test fail so that the code succeeds. The swash is stored in a | |
222 | * predictable PL_ place */ | |
cfaf538b KW |
223 | #define _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, \ |
224 | CLASS, STR) \ | |
ee9a90b8 KW |
225 | case NAME: \ |
226 | _CCC_TRY_CODE( !, FUNC, \ | |
227 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ | |
228 | (U8*)locinput, TRUE)), \ | |
229 | CLASS, STR) \ | |
230 | case NNAME: \ | |
1dcf4a1b | 231 | _CCC_TRY_CODE( PLACEHOLDER , FUNC, \ |
ee9a90b8 KW |
232 | cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), \ |
233 | (U8*)locinput, TRUE)), \ | |
234 | CLASS, STR) \ | |
235 | ||
236 | /* Generate the case statements for both locale and non-locale character | |
237 | * classes in regmatch for classes that don't have special unicode semantics. | |
238 | * Locales don't use an immediate swash, but an intermediary special locale | |
239 | * function that is called on the pointer to the current place in the input | |
240 | * string. That function will resolve to needing the same swash. One might | |
241 | * think that because we don't know what the locale will match, we shouldn't | |
242 | * check with the swash loading function that it loaded properly; ie, that we | |
243 | * should use LOAD_UTF8_CHARCLASS_NO_CHECK for those, but what is passed to the | |
244 | * regular LOAD_UTF8_CHARCLASS is in non-locale terms, and so locale is | |
245 | * irrelevant here */ | |
246 | #define CCC_TRY(NAME, NNAME, FUNC, \ | |
247 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
cfaf538b | 248 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 KW |
249 | CLASS, STR) \ |
250 | case NAMEL: \ | |
251 | PL_reg_flags |= RF_tainted; \ | |
252 | _CCC_TRY_CODE( !, LCFUNC, LCFUNC_utf8((U8*)locinput), CLASS, STR) \ | |
253 | case NNAMEL: \ | |
254 | PL_reg_flags |= RF_tainted; \ | |
1dcf4a1b KW |
255 | _CCC_TRY_CODE( PLACEHOLDER, LCFUNC, LCFUNC_utf8((U8*)locinput), \ |
256 | CLASS, STR) \ | |
cfaf538b KW |
257 | case NAMEA: \ |
258 | if (locinput >= PL_regeol || ! FUNCA(nextchr)) { \ | |
259 | sayNO; \ | |
260 | } \ | |
261 | /* Matched a utf8-invariant, so don't have to worry about utf8 */ \ | |
262 | nextchr = UCHARAT(++locinput); \ | |
263 | break; \ | |
264 | case NNAMEA: \ | |
265 | if (locinput >= PL_regeol || FUNCA(nextchr)) { \ | |
266 | sayNO; \ | |
267 | } \ | |
268 | if (utf8_target) { \ | |
269 | locinput += PL_utf8skip[nextchr]; \ | |
270 | nextchr = UCHARAT(locinput); \ | |
271 | } \ | |
272 | else { \ | |
273 | nextchr = UCHARAT(++locinput); \ | |
274 | } \ | |
275 | break; \ | |
ee9a90b8 KW |
276 | /* Generate the non-locale cases */ \ |
277 | _CCC_TRY_NONLOCALE(NAME, NNAME, FUNC, CLASS, STR) | |
278 | ||
279 | /* This is like CCC_TRY, but has an extra set of parameters for generating case | |
280 | * statements to handle separate Unicode semantics nodes */ | |
281 | #define CCC_TRY_U(NAME, NNAME, FUNC, \ | |
282 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
283 | NAMEU, NNAMEU, FUNCU, \ | |
cfaf538b | 284 | NAMEA, NNAMEA, FUNCA, \ |
ee9a90b8 | 285 | CLASS, STR) \ |
cfaf538b KW |
286 | CCC_TRY(NAME, NNAME, FUNC, \ |
287 | NAMEL, NNAMEL, LCFUNC, LCFUNC_utf8, \ | |
288 | NAMEA, NNAMEA, FUNCA, \ | |
289 | CLASS, STR) \ | |
ee9a90b8 | 290 | _CCC_TRY_NONLOCALE(NAMEU, NNAMEU, FUNCU, CLASS, STR) |
d1eb3177 | 291 | |
3dab1dad YO |
292 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
293 | ||
5f80c4cf | 294 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
c35dcbe2 YO |
295 | /* it would be nice to rework regcomp.sym to generate this stuff. sigh |
296 | * | |
297 | * NOTE that *nothing* that affects backtracking should be in here, specifically | |
298 | * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a | |
299 | * node that is in between two EXACT like nodes when ascertaining what the required | |
300 | * "follow" character is. This should probably be moved to regex compile time | |
301 | * although it may be done at run time beause of the REF possibility - more | |
302 | * investigation required. -- demerphq | |
303 | */ | |
3e901dc0 YO |
304 | #define JUMPABLE(rn) ( \ |
305 | OP(rn) == OPEN || \ | |
306 | (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \ | |
307 | OP(rn) == EVAL || \ | |
cca55fe3 JP |
308 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
309 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
d1c771f5 | 310 | OP(rn) == KEEPS || \ |
3dab1dad | 311 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 | 312 | ) |
ee9b8eae | 313 | #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT) |
e2d8ce26 | 314 | |
ee9b8eae YO |
315 | #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF ) |
316 | ||
317 | #if 0 | |
318 | /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so | |
319 | we don't need this definition. */ | |
320 | #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) | |
9a5a5549 | 321 | #define IS_TEXTF(rn) ( (OP(rn)==EXACTFU || OP(rn)==EXACTF) || OP(rn)==REFF || OP(rn)==NREFF ) |
ee9b8eae YO |
322 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) |
323 | ||
324 | #else | |
325 | /* ... so we use this as its faster. */ | |
326 | #define IS_TEXT(rn) ( OP(rn)==EXACT ) | |
9a5a5549 | 327 | #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU ) |
ee9b8eae YO |
328 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) |
329 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) | |
330 | ||
331 | #endif | |
e2d8ce26 | 332 | |
a84d97b6 HS |
333 | /* |
334 | Search for mandatory following text node; for lookahead, the text must | |
335 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
336 | */ | |
cca55fe3 | 337 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
338 | while (JUMPABLE(rn)) { \ |
339 | const OPCODE type = OP(rn); \ | |
340 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 341 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 342 | else if (type == PLUS) \ |
cca55fe3 | 343 | rn = NEXTOPER(rn); \ |
3dab1dad | 344 | else if (type == IFMATCH) \ |
a84d97b6 | 345 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 346 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 347 | } \ |
5f80c4cf | 348 | } STMT_END |
74750237 | 349 | |
c476f425 | 350 | |
acfe0abc | 351 | static void restore_pos(pTHX_ void *arg); |
51371543 | 352 | |
620d5b66 NC |
353 | #define REGCP_PAREN_ELEMS 4 |
354 | #define REGCP_OTHER_ELEMS 5 | |
e0fa7e2b | 355 | #define REGCP_FRAME_ELEMS 1 |
620d5b66 NC |
356 | /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and |
357 | * are needed for the regexp context stack bookkeeping. */ | |
358 | ||
76e3520e | 359 | STATIC CHECKPOINT |
cea2e8a9 | 360 | S_regcppush(pTHX_ I32 parenfloor) |
a0d0e21e | 361 | { |
97aff369 | 362 | dVAR; |
a3b680e6 | 363 | const int retval = PL_savestack_ix; |
a3b680e6 | 364 | const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; |
e0fa7e2b NC |
365 | const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS; |
366 | const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT; | |
a0d0e21e | 367 | int p; |
40a82448 | 368 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 369 | |
e49a9654 IH |
370 | if (paren_elems_to_push < 0) |
371 | Perl_croak(aTHX_ "panic: paren_elems_to_push < 0"); | |
372 | ||
e0fa7e2b NC |
373 | if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems) |
374 | Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf | |
5df417d0 JH |
375 | " out of range (%lu-%ld)", |
376 | total_elems, (unsigned long)PL_regsize, (long)parenfloor); | |
e0fa7e2b | 377 | |
620d5b66 | 378 | SSGROW(total_elems + REGCP_FRAME_ELEMS); |
7f69552c | 379 | |
3280af22 | 380 | for (p = PL_regsize; p > parenfloor; p--) { |
b1ce53c5 | 381 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
f0ab9afb NC |
382 | SSPUSHINT(PL_regoffs[p].end); |
383 | SSPUSHINT(PL_regoffs[p].start); | |
3280af22 | 384 | SSPUSHPTR(PL_reg_start_tmp[p]); |
a0d0e21e | 385 | SSPUSHINT(p); |
e7707071 | 386 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
40a82448 | 387 | " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n", |
f0ab9afb | 388 | (UV)p, (IV)PL_regoffs[p].start, |
40a82448 | 389 | (IV)(PL_reg_start_tmp[p] - PL_bostr), |
f0ab9afb | 390 | (IV)PL_regoffs[p].end |
40a82448 | 391 | )); |
a0d0e21e | 392 | } |
b1ce53c5 | 393 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
f0ab9afb | 394 | SSPUSHPTR(PL_regoffs); |
3280af22 NIS |
395 | SSPUSHINT(PL_regsize); |
396 | SSPUSHINT(*PL_reglastparen); | |
a01268b5 | 397 | SSPUSHINT(*PL_reglastcloseparen); |
3280af22 | 398 | SSPUSHPTR(PL_reginput); |
e0fa7e2b | 399 | SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */ |
41123dfd | 400 | |
a0d0e21e LW |
401 | return retval; |
402 | } | |
403 | ||
c277df42 | 404 | /* These are needed since we do not localize EVAL nodes: */ |
ab3bbdeb YO |
405 | #define REGCP_SET(cp) \ |
406 | DEBUG_STATE_r( \ | |
ab3bbdeb | 407 | PerlIO_printf(Perl_debug_log, \ |
e4f74956 | 408 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
ab3bbdeb YO |
409 | (IV)PL_savestack_ix)); \ |
410 | cp = PL_savestack_ix | |
c3464db5 | 411 | |
ab3bbdeb | 412 | #define REGCP_UNWIND(cp) \ |
e4f74956 | 413 | DEBUG_STATE_r( \ |
ab3bbdeb | 414 | if (cp != PL_savestack_ix) \ |
e4f74956 YO |
415 | PerlIO_printf(Perl_debug_log, \ |
416 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ | |
ab3bbdeb YO |
417 | (IV)(cp), (IV)PL_savestack_ix)); \ |
418 | regcpblow(cp) | |
c277df42 | 419 | |
76e3520e | 420 | STATIC char * |
097eb12c | 421 | S_regcppop(pTHX_ const regexp *rex) |
a0d0e21e | 422 | { |
97aff369 | 423 | dVAR; |
e0fa7e2b | 424 | UV i; |
a0d0e21e | 425 | char *input; |
a3621e74 YO |
426 | GET_RE_DEBUG_FLAGS_DECL; |
427 | ||
7918f24d NC |
428 | PERL_ARGS_ASSERT_REGCPPOP; |
429 | ||
b1ce53c5 | 430 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
c6bf6a65 | 431 | i = SSPOPUV; |
e0fa7e2b NC |
432 | assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
433 | i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */ | |
a0d0e21e | 434 | input = (char *) SSPOPPTR; |
a01268b5 | 435 | *PL_reglastcloseparen = SSPOPINT; |
3280af22 NIS |
436 | *PL_reglastparen = SSPOPINT; |
437 | PL_regsize = SSPOPINT; | |
f0ab9afb | 438 | PL_regoffs=(regexp_paren_pair *) SSPOPPTR; |
b1ce53c5 | 439 | |
620d5b66 | 440 | i -= REGCP_OTHER_ELEMS; |
b1ce53c5 | 441 | /* Now restore the parentheses context. */ |
620d5b66 | 442 | for ( ; i > 0; i -= REGCP_PAREN_ELEMS) { |
1df70142 | 443 | I32 tmps; |
097eb12c | 444 | U32 paren = (U32)SSPOPINT; |
3280af22 | 445 | PL_reg_start_tmp[paren] = (char *) SSPOPPTR; |
f0ab9afb | 446 | PL_regoffs[paren].start = SSPOPINT; |
cf93c79d | 447 | tmps = SSPOPINT; |
3280af22 | 448 | if (paren <= *PL_reglastparen) |
f0ab9afb | 449 | PL_regoffs[paren].end = tmps; |
e7707071 | 450 | DEBUG_BUFFERS_r( |
c3464db5 | 451 | PerlIO_printf(Perl_debug_log, |
b900a521 | 452 | " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n", |
f0ab9afb | 453 | (UV)paren, (IV)PL_regoffs[paren].start, |
b900a521 | 454 | (IV)(PL_reg_start_tmp[paren] - PL_bostr), |
f0ab9afb | 455 | (IV)PL_regoffs[paren].end, |
3280af22 | 456 | (paren > *PL_reglastparen ? "(no)" : "")); |
c277df42 | 457 | ); |
a0d0e21e | 458 | } |
e7707071 | 459 | DEBUG_BUFFERS_r( |
bb7a0f54 | 460 | if (*PL_reglastparen + 1 <= rex->nparens) { |
c3464db5 | 461 | PerlIO_printf(Perl_debug_log, |
faccc32b | 462 | " restoring \\%"IVdf"..\\%"IVdf" to undef\n", |
4f639d21 | 463 | (IV)(*PL_reglastparen + 1), (IV)rex->nparens); |
c277df42 IZ |
464 | } |
465 | ); | |
daf18116 | 466 | #if 1 |
dafc8851 JH |
467 | /* It would seem that the similar code in regtry() |
468 | * already takes care of this, and in fact it is in | |
469 | * a better location to since this code can #if 0-ed out | |
470 | * but the code in regtry() is needed or otherwise tests | |
471 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
472 | * (as of patchlevel 7877) will fail. Then again, |
473 | * this code seems to be necessary or otherwise | |
225593e1 DM |
474 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ |
475 | * --jhi updated by dapm */ | |
3b6647e0 | 476 | for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) { |
097eb12c | 477 | if (i > PL_regsize) |
f0ab9afb NC |
478 | PL_regoffs[i].start = -1; |
479 | PL_regoffs[i].end = -1; | |
a0d0e21e | 480 | } |
dafc8851 | 481 | #endif |
a0d0e21e LW |
482 | return input; |
483 | } | |
484 | ||
02db2b7b | 485 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 486 | |
a687059c | 487 | /* |
e50aee73 | 488 | * pregexec and friends |
a687059c LW |
489 | */ |
490 | ||
76234dfb | 491 | #ifndef PERL_IN_XSUB_RE |
a687059c | 492 | /* |
c277df42 | 493 | - pregexec - match a regexp against a string |
a687059c | 494 | */ |
c277df42 | 495 | I32 |
49d7dfbc | 496 | Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend, |
c3464db5 | 497 | char *strbeg, I32 minend, SV *screamer, U32 nosave) |
c277df42 IZ |
498 | /* strend: pointer to null at end of string */ |
499 | /* strbeg: real beginning of string */ | |
500 | /* minend: end of match must be >=minend after stringarg. */ | |
501 | /* nosave: For optimizations. */ | |
502 | { | |
7918f24d NC |
503 | PERL_ARGS_ASSERT_PREGEXEC; |
504 | ||
c277df42 | 505 | return |
9041c2e3 | 506 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
507 | nosave ? 0 : REXEC_COPY_STR); |
508 | } | |
76234dfb | 509 | #endif |
22e551b9 | 510 | |
9041c2e3 | 511 | /* |
cad2e5aa JH |
512 | * Need to implement the following flags for reg_anch: |
513 | * | |
514 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
515 | * USE_INTUIT_ML | |
516 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
517 | * INTUIT_AUTORITATIVE_ML | |
518 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
519 | * INTUIT_ONCE_ML | |
520 | * | |
521 | * Another flag for this function: SECOND_TIME (so that float substrs | |
522 | * with giant delta may be not rechecked). | |
523 | */ | |
524 | ||
525 | /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ | |
526 | ||
3f7c398e | 527 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
528 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
529 | ||
530 | /* XXXX We assume that strpos is strbeg unless sv. */ | |
531 | ||
6eb5f6b9 JH |
532 | /* XXXX Some places assume that there is a fixed substring. |
533 | An update may be needed if optimizer marks as "INTUITable" | |
534 | RExen without fixed substrings. Similarly, it is assumed that | |
535 | lengths of all the strings are no more than minlen, thus they | |
536 | cannot come from lookahead. | |
40d049e4 YO |
537 | (Or minlen should take into account lookahead.) |
538 | NOTE: Some of this comment is not correct. minlen does now take account | |
539 | of lookahead/behind. Further research is required. -- demerphq | |
540 | ||
541 | */ | |
6eb5f6b9 | 542 | |
2c2d71f5 JH |
543 | /* A failure to find a constant substring means that there is no need to make |
544 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
545 | finding a substring too deep into the string means that less calls to | |
30944b6d IZ |
546 | regtry() should be needed. |
547 | ||
548 | REx compiler's optimizer found 4 possible hints: | |
549 | a) Anchored substring; | |
550 | b) Fixed substring; | |
551 | c) Whether we are anchored (beginning-of-line or \G); | |
486ec47a | 552 | d) First node (of those at offset 0) which may distinguish positions; |
6eb5f6b9 | 553 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
554 | string which does not contradict any of them. |
555 | */ | |
2c2d71f5 | 556 | |
6eb5f6b9 JH |
557 | /* Most of decisions we do here should have been done at compile time. |
558 | The nodes of the REx which we used for the search should have been | |
559 | deleted from the finite automaton. */ | |
560 | ||
cad2e5aa | 561 | char * |
288b8c02 | 562 | Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, |
9f61653a | 563 | char *strend, const U32 flags, re_scream_pos_data *data) |
cad2e5aa | 564 | { |
97aff369 | 565 | dVAR; |
288b8c02 | 566 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
b7953727 | 567 | register I32 start_shift = 0; |
cad2e5aa | 568 | /* Should be nonnegative! */ |
b7953727 | 569 | register I32 end_shift = 0; |
2c2d71f5 JH |
570 | register char *s; |
571 | register SV *check; | |
a1933d95 | 572 | char *strbeg; |
cad2e5aa | 573 | char *t; |
f2ed9b32 | 574 | const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 575 | I32 ml_anch; |
bd61b366 SS |
576 | register char *other_last = NULL; /* other substr checked before this */ |
577 | char *check_at = NULL; /* check substr found at this pos */ | |
bbe252da | 578 | const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; |
f8fc2ecf | 579 | RXi_GET_DECL(prog,progi); |
30944b6d | 580 | #ifdef DEBUGGING |
890ce7af | 581 | const char * const i_strpos = strpos; |
30944b6d | 582 | #endif |
a3621e74 YO |
583 | GET_RE_DEBUG_FLAGS_DECL; |
584 | ||
7918f24d NC |
585 | PERL_ARGS_ASSERT_RE_INTUIT_START; |
586 | ||
f2ed9b32 | 587 | RX_MATCH_UTF8_set(rx,utf8_target); |
cad2e5aa | 588 | |
3c8556c3 | 589 | if (RX_UTF8(rx)) { |
b8d68ded JH |
590 | PL_reg_flags |= RF_utf8; |
591 | } | |
ab3bbdeb | 592 | DEBUG_EXECUTE_r( |
f2ed9b32 | 593 | debug_start_match(rx, utf8_target, strpos, strend, |
1de06328 YO |
594 | sv ? "Guessing start of match in sv for" |
595 | : "Guessing start of match in string for"); | |
2a782b5b | 596 | ); |
cad2e5aa | 597 | |
c344f387 JH |
598 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
599 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 600 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 601 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 602 | goto fail; |
2c2d71f5 | 603 | } |
1de06328 | 604 | |
a1933d95 | 605 | strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; |
1aa99e6b | 606 | PL_regeol = strend; |
f2ed9b32 | 607 | if (utf8_target) { |
33b8afdf JH |
608 | if (!prog->check_utf8 && prog->check_substr) |
609 | to_utf8_substr(prog); | |
610 | check = prog->check_utf8; | |
611 | } else { | |
612 | if (!prog->check_substr && prog->check_utf8) | |
613 | to_byte_substr(prog); | |
614 | check = prog->check_substr; | |
615 | } | |
1de06328 | 616 | if (check == &PL_sv_undef) { |
a3621e74 | 617 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
1de06328 | 618 | "Non-utf8 string cannot match utf8 check string\n")); |
33b8afdf JH |
619 | goto fail; |
620 | } | |
bbe252da YO |
621 | if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */ |
622 | ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) | |
623 | || ( (prog->extflags & RXf_ANCH_BOL) | |
7fba1cd6 | 624 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 625 | |
7e25d62c | 626 | if (!ml_anch) { |
bbe252da YO |
627 | if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */ |
628 | && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ | |
3f7c398e | 629 | /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ |
7e25d62c JH |
630 | && sv && !SvROK(sv) |
631 | && (strpos != strbeg)) { | |
a3621e74 | 632 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
633 | goto fail; |
634 | } | |
635 | if (prog->check_offset_min == prog->check_offset_max && | |
bbe252da | 636 | !(prog->extflags & RXf_CANY_SEEN)) { |
2c2d71f5 | 637 | /* Substring at constant offset from beg-of-str... */ |
cad2e5aa JH |
638 | I32 slen; |
639 | ||
1aa99e6b | 640 | s = HOP3c(strpos, prog->check_offset_min, strend); |
1de06328 | 641 | |
653099ff GS |
642 | if (SvTAIL(check)) { |
643 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 644 | |
9041c2e3 | 645 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 646 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 647 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 648 | goto fail_finish; |
cad2e5aa JH |
649 | } |
650 | /* Now should match s[0..slen-2] */ | |
651 | slen--; | |
3f7c398e | 652 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 653 | || (slen > 1 |
3f7c398e | 654 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 655 | report_neq: |
a3621e74 | 656 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
657 | goto fail_finish; |
658 | } | |
cad2e5aa | 659 | } |
3f7c398e | 660 | else if (*SvPVX_const(check) != *s |
653099ff | 661 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 662 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 663 | goto report_neq; |
c315bfe8 | 664 | check_at = s; |
2c2d71f5 | 665 | goto success_at_start; |
7e25d62c | 666 | } |
cad2e5aa | 667 | } |
2c2d71f5 | 668 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 669 | s = strpos; |
2c2d71f5 | 670 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
1de06328 YO |
671 | end_shift = prog->check_end_shift; |
672 | ||
2c2d71f5 | 673 | if (!ml_anch) { |
a3b680e6 | 674 | const I32 end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 675 | - (SvTAIL(check) != 0); |
a3b680e6 | 676 | const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
677 | |
678 | if (end_shift < eshift) | |
679 | end_shift = eshift; | |
680 | } | |
cad2e5aa | 681 | } |
2c2d71f5 | 682 | else { /* Can match at random position */ |
cad2e5aa JH |
683 | ml_anch = 0; |
684 | s = strpos; | |
1de06328 YO |
685 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
686 | end_shift = prog->check_end_shift; | |
687 | ||
688 | /* end shift should be non negative here */ | |
cad2e5aa JH |
689 | } |
690 | ||
bcdf7404 | 691 | #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 692 | if (end_shift < 0) |
1de06328 | 693 | Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", |
220fc49f | 694 | (IV)end_shift, RX_PRECOMP(prog)); |
2c2d71f5 JH |
695 | #endif |
696 | ||
2c2d71f5 JH |
697 | restart: |
698 | /* Find a possible match in the region s..strend by looking for | |
699 | the "check" substring in the region corrected by start/end_shift. */ | |
1de06328 YO |
700 | |
701 | { | |
702 | I32 srch_start_shift = start_shift; | |
703 | I32 srch_end_shift = end_shift; | |
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 | ||
cad2e5aa | 716 | if (flags & REXEC_SCREAM) { |
cad2e5aa | 717 | I32 p = -1; /* Internal iterator of scream. */ |
a3b680e6 | 718 | I32 * const pp = data ? data->scream_pos : &p; |
cad2e5aa | 719 | |
2c2d71f5 JH |
720 | if (PL_screamfirst[BmRARE(check)] >= 0 |
721 | || ( BmRARE(check) == '\n' | |
85c508c3 | 722 | && (BmPREVIOUS(check) == SvCUR(check) - 1) |
2c2d71f5 | 723 | && SvTAIL(check) )) |
9041c2e3 | 724 | s = screaminstr(sv, check, |
1de06328 | 725 | srch_start_shift + (s - strbeg), srch_end_shift, pp, 0); |
cad2e5aa | 726 | else |
2c2d71f5 | 727 | goto fail_finish; |
4addbd3b | 728 | /* we may be pointing at the wrong string */ |
07bc277f | 729 | if (s && RXp_MATCH_COPIED(prog)) |
3f7c398e | 730 | s = strbeg + (s - SvPVX_const(sv)); |
cad2e5aa JH |
731 | if (data) |
732 | *data->scream_olds = s; | |
733 | } | |
1de06328 YO |
734 | else { |
735 | U8* start_point; | |
736 | U8* end_point; | |
bbe252da | 737 | if (prog->extflags & RXf_CANY_SEEN) { |
1de06328 YO |
738 | start_point= (U8*)(s + srch_start_shift); |
739 | end_point= (U8*)(strend - srch_end_shift); | |
740 | } else { | |
741 | start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); | |
742 | end_point= HOP3(strend, -srch_end_shift, strbeg); | |
743 | } | |
6bda09f9 | 744 | DEBUG_OPTIMISE_MORE_r({ |
56570a2c | 745 | PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", |
1de06328 | 746 | (int)(end_point - start_point), |
fc8cd66c | 747 | (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), |
1de06328 YO |
748 | start_point); |
749 | }); | |
750 | ||
751 | s = fbm_instr( start_point, end_point, | |
7fba1cd6 | 752 | check, multiline ? FBMrf_MULTILINE : 0); |
1de06328 YO |
753 | } |
754 | } | |
cad2e5aa JH |
755 | /* Update the count-of-usability, remove useless subpatterns, |
756 | unshift s. */ | |
2c2d71f5 | 757 | |
ab3bbdeb | 758 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 759 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
760 | SvPVX_const(check), RE_SV_DUMPLEN(check), 30); |
761 | PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", | |
2c2d71f5 | 762 | (s ? "Found" : "Did not find"), |
f2ed9b32 | 763 | (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) |
ab3bbdeb YO |
764 | ? "anchored" : "floating"), |
765 | quoted, | |
766 | RE_SV_TAIL(check), | |
767 | (s ? " at offset " : "...\n") ); | |
768 | }); | |
2c2d71f5 JH |
769 | |
770 | if (!s) | |
771 | goto fail_finish; | |
2c2d71f5 | 772 | /* Finish the diagnostic message */ |
a3621e74 | 773 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 | 774 | |
1de06328 YO |
775 | /* XXX dmq: first branch is for positive lookbehind... |
776 | Our check string is offset from the beginning of the pattern. | |
777 | So we need to do any stclass tests offset forward from that | |
778 | point. I think. :-( | |
779 | */ | |
780 | ||
781 | ||
782 | ||
783 | check_at=s; | |
784 | ||
785 | ||
2c2d71f5 JH |
786 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. |
787 | Start with the other substr. | |
788 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 789 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
790 | *always* match. Probably should be marked during compile... |
791 | Probably it is right to do no SCREAM here... | |
792 | */ | |
793 | ||
f2ed9b32 | 794 | if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8) |
1de06328 YO |
795 | : (prog->float_substr && prog->anchored_substr)) |
796 | { | |
30944b6d | 797 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
798 | /* XXXX May be hopelessly wrong for UTF... */ |
799 | if (!other_last) | |
6eb5f6b9 | 800 | other_last = strpos; |
f2ed9b32 | 801 | if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
802 | do_other_anchored: |
803 | { | |
890ce7af AL |
804 | char * const last = HOP3c(s, -start_shift, strbeg); |
805 | char *last1, *last2; | |
be8e71aa | 806 | char * const saved_s = s; |
33b8afdf | 807 | SV* must; |
2c2d71f5 | 808 | |
2c2d71f5 JH |
809 | t = s - prog->check_offset_max; |
810 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
f2ed9b32 | 811 | && (!utf8_target |
0ce71af7 | 812 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 813 | && t > strpos))) |
6f207bd3 | 814 | NOOP; |
2c2d71f5 JH |
815 | else |
816 | t = strpos; | |
1aa99e6b | 817 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
818 | if (t < other_last) /* These positions already checked */ |
819 | t = other_last; | |
1aa99e6b | 820 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
821 | if (last < last1) |
822 | last1 = last; | |
1de06328 YO |
823 | /* XXXX It is not documented what units *_offsets are in. |
824 | We assume bytes, but this is clearly wrong. | |
825 | Meaning this code needs to be carefully reviewed for errors. | |
826 | dmq. | |
827 | */ | |
828 | ||
2c2d71f5 | 829 | /* On end-of-str: see comment below. */ |
f2ed9b32 | 830 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; |
33b8afdf JH |
831 | if (must == &PL_sv_undef) { |
832 | s = (char*)NULL; | |
1de06328 | 833 | DEBUG_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
834 | } |
835 | else | |
836 | s = fbm_instr( | |
837 | (unsigned char*)t, | |
838 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
839 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
840 | must, | |
7fba1cd6 | 841 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 842 | ); |
ab3bbdeb | 843 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 844 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
845 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
846 | PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", | |
2c2d71f5 | 847 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
848 | quoted, RE_SV_TAIL(must)); |
849 | }); | |
850 | ||
851 | ||
2c2d71f5 JH |
852 | if (!s) { |
853 | if (last1 >= last2) { | |
a3621e74 | 854 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
855 | ", giving up...\n")); |
856 | goto fail_finish; | |
857 | } | |
a3621e74 | 858 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 859 | ", trying floating at offset %ld...\n", |
be8e71aa | 860 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
861 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
862 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
863 | goto restart; |
864 | } | |
865 | else { | |
a3621e74 | 866 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 867 | (long)(s - i_strpos))); |
1aa99e6b IH |
868 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
869 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 870 | s = saved_s; |
2c2d71f5 JH |
871 | if (t == strpos) |
872 | goto try_at_start; | |
2c2d71f5 JH |
873 | goto try_at_offset; |
874 | } | |
30944b6d | 875 | } |
2c2d71f5 JH |
876 | } |
877 | else { /* Take into account the floating substring. */ | |
33b8afdf | 878 | char *last, *last1; |
be8e71aa | 879 | char * const saved_s = s; |
33b8afdf JH |
880 | SV* must; |
881 | ||
882 | t = HOP3c(s, -start_shift, strbeg); | |
883 | last1 = last = | |
884 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
885 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
886 | last = HOP3c(t, prog->float_max_offset, strend); | |
887 | s = HOP3c(t, prog->float_min_offset, strend); | |
888 | if (s < other_last) | |
889 | s = other_last; | |
2c2d71f5 | 890 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
f2ed9b32 | 891 | must = utf8_target ? prog->float_utf8 : prog->float_substr; |
33b8afdf JH |
892 | /* fbm_instr() takes into account exact value of end-of-str |
893 | if the check is SvTAIL(ed). Since false positives are OK, | |
894 | and end-of-str is not later than strend we are OK. */ | |
895 | if (must == &PL_sv_undef) { | |
896 | s = (char*)NULL; | |
1de06328 | 897 | DEBUG_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
898 | } |
899 | else | |
2c2d71f5 | 900 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
901 | (unsigned char*)last + SvCUR(must) |
902 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 903 | must, multiline ? FBMrf_MULTILINE : 0); |
ab3bbdeb | 904 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 905 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
906 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
907 | PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", | |
33b8afdf | 908 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
909 | quoted, RE_SV_TAIL(must)); |
910 | }); | |
33b8afdf JH |
911 | if (!s) { |
912 | if (last1 == last) { | |
a3621e74 | 913 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
914 | ", giving up...\n")); |
915 | goto fail_finish; | |
2c2d71f5 | 916 | } |
a3621e74 | 917 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 918 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 919 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
920 | other_last = last; |
921 | s = HOP3c(t, 1, strend); | |
922 | goto restart; | |
923 | } | |
924 | else { | |
a3621e74 | 925 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
926 | (long)(s - i_strpos))); |
927 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 928 | s = saved_s; |
33b8afdf JH |
929 | if (t == strpos) |
930 | goto try_at_start; | |
931 | goto try_at_offset; | |
932 | } | |
2c2d71f5 | 933 | } |
cad2e5aa | 934 | } |
2c2d71f5 | 935 | |
1de06328 | 936 | |
9ef43ace | 937 | t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); |
1de06328 | 938 | |
6bda09f9 | 939 | DEBUG_OPTIMISE_MORE_r( |
1de06328 YO |
940 | PerlIO_printf(Perl_debug_log, |
941 | "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", | |
942 | (IV)prog->check_offset_min, | |
943 | (IV)prog->check_offset_max, | |
944 | (IV)(s-strpos), | |
945 | (IV)(t-strpos), | |
946 | (IV)(t-s), | |
947 | (IV)(strend-strpos) | |
948 | ) | |
949 | ); | |
950 | ||
2c2d71f5 | 951 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
f2ed9b32 | 952 | && (!utf8_target |
9ef43ace | 953 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) |
1de06328 YO |
954 | && t > strpos))) |
955 | { | |
2c2d71f5 JH |
956 | /* Fixed substring is found far enough so that the match |
957 | cannot start at strpos. */ | |
958 | try_at_offset: | |
cad2e5aa | 959 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
960 | /* Eventually fbm_*() should handle this, but often |
961 | anchored_offset is not 0, so this check will not be wasted. */ | |
962 | /* XXXX In the code below we prefer to look for "^" even in | |
963 | presence of anchored substrings. And we search even | |
964 | beyond the found float position. These pessimizations | |
965 | are historical artefacts only. */ | |
966 | find_anchor: | |
2c2d71f5 | 967 | while (t < strend - prog->minlen) { |
cad2e5aa | 968 | if (*t == '\n') { |
4ee3650e | 969 | if (t < check_at - prog->check_offset_min) { |
f2ed9b32 | 970 | if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
971 | /* Since we moved from the found position, |
972 | we definitely contradict the found anchored | |
30944b6d IZ |
973 | substr. Due to the above check we do not |
974 | contradict "check" substr. | |
975 | Thus we can arrive here only if check substr | |
976 | is float. Redo checking for "other"=="fixed". | |
977 | */ | |
9041c2e3 | 978 | strpos = t + 1; |
a3621e74 | 979 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 980 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
981 | goto do_other_anchored; |
982 | } | |
4ee3650e GS |
983 | /* We don't contradict the found floating substring. */ |
984 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 985 | s = t + 1; |
a3621e74 | 986 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 987 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
988 | goto set_useful; |
989 | } | |
4ee3650e GS |
990 | /* Position contradicts check-string */ |
991 | /* XXXX probably better to look for check-string | |
992 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 993 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 994 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 995 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
996 | goto restart; |
997 | } | |
998 | t++; | |
999 | } | |
a3621e74 | 1000 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 1001 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 1002 | goto fail_finish; |
cad2e5aa | 1003 | } |
f5952150 | 1004 | else { |
a3621e74 | 1005 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 1006 | PL_colors[0], PL_colors[1])); |
f5952150 | 1007 | } |
cad2e5aa JH |
1008 | s = t; |
1009 | set_useful: | |
f2ed9b32 | 1010 | ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
1011 | } |
1012 | else { | |
f5952150 | 1013 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 1014 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
1015 | unless it was an MBOL and we are not after MBOL, |
1016 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
1017 | try_at_start: |
1018 | /* Even in this situation we may use MBOL flag if strpos is offset | |
1019 | wrt the start of the string. */ | |
05b4157f | 1020 | if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ |
a1933d95 | 1021 | && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d | 1022 | /* May be due to an implicit anchor of m{.*foo} */ |
bbe252da | 1023 | && !(prog->intflags & PREGf_IMPLICIT)) |
d506a20d | 1024 | { |
cad2e5aa JH |
1025 | t = strpos; |
1026 | goto find_anchor; | |
1027 | } | |
a3621e74 | 1028 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 1029 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
70685ca0 | 1030 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 1031 | ); |
2c2d71f5 | 1032 | success_at_start: |
bbe252da | 1033 | if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ |
f2ed9b32 | 1034 | && (utf8_target ? ( |
33b8afdf JH |
1035 | prog->check_utf8 /* Could be deleted already */ |
1036 | && --BmUSEFUL(prog->check_utf8) < 0 | |
1037 | && (prog->check_utf8 == prog->float_utf8) | |
1038 | ) : ( | |
1039 | prog->check_substr /* Could be deleted already */ | |
1040 | && --BmUSEFUL(prog->check_substr) < 0 | |
1041 | && (prog->check_substr == prog->float_substr) | |
1042 | ))) | |
66e933ab | 1043 | { |
cad2e5aa | 1044 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 1045 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
f2ed9b32 KW |
1046 | /* XXX Does the destruction order has to change with utf8_target? */ |
1047 | SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr); | |
1048 | SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
1049 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
1050 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
1051 | check = NULL; /* abort */ | |
cad2e5aa | 1052 | s = strpos; |
486ec47a | 1053 | /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevant flag |
c9415951 YO |
1054 | see http://bugs.activestate.com/show_bug.cgi?id=87173 */ |
1055 | if (prog->intflags & PREGf_IMPLICIT) | |
1056 | prog->extflags &= ~RXf_ANCH_MBOL; | |
3cf5c195 IZ |
1057 | /* XXXX This is a remnant of the old implementation. It |
1058 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 1059 | other heuristics. */ |
bbe252da | 1060 | prog->extflags &= ~RXf_USE_INTUIT; |
c9415951 | 1061 | /* XXXX What other flags might need to be cleared in this branch? */ |
cad2e5aa JH |
1062 | } |
1063 | else | |
1064 | s = strpos; | |
1065 | } | |
1066 | ||
6eb5f6b9 JH |
1067 | /* Last resort... */ |
1068 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
1de06328 YO |
1069 | /* trie stclasses are too expensive to use here, we are better off to |
1070 | leave it to regmatch itself */ | |
f8fc2ecf | 1071 | if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { |
6eb5f6b9 JH |
1072 | /* minlen == 0 is possible if regstclass is \b or \B, |
1073 | and the fixed substr is ''$. | |
1074 | Since minlen is already taken into account, s+1 is before strend; | |
1075 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
1076 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
1077 | regstclass does not come from lookahead... */ | |
1078 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
9a5a5549 | 1079 | This leaves EXACTF, EXACTFU only, which are dealt with in find_byclass(). */ |
f8fc2ecf YO |
1080 | const U8* const str = (U8*)STRING(progi->regstclass); |
1081 | const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT | |
1082 | ? CHR_DIST(str+STR_LEN(progi->regstclass), str) | |
66e933ab | 1083 | : 1); |
1de06328 YO |
1084 | char * endpos; |
1085 | if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) | |
1086 | endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); | |
1087 | else if (prog->float_substr || prog->float_utf8) | |
1088 | endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); | |
1089 | else | |
1090 | endpos= strend; | |
1091 | ||
70685ca0 JH |
1092 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n", |
1093 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg))); | |
1de06328 | 1094 | |
6eb5f6b9 | 1095 | t = s; |
f8fc2ecf | 1096 | s = find_byclass(prog, progi->regstclass, s, endpos, NULL); |
6eb5f6b9 JH |
1097 | if (!s) { |
1098 | #ifdef DEBUGGING | |
cbbf8932 | 1099 | const char *what = NULL; |
6eb5f6b9 JH |
1100 | #endif |
1101 | if (endpos == strend) { | |
a3621e74 | 1102 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
1103 | "Could not match STCLASS...\n") ); |
1104 | goto fail; | |
1105 | } | |
a3621e74 | 1106 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 1107 | "This position contradicts STCLASS...\n") ); |
bbe252da | 1108 | if ((prog->extflags & RXf_ANCH) && !ml_anch) |
653099ff | 1109 | goto fail; |
6eb5f6b9 | 1110 | /* Contradict one of substrings */ |
33b8afdf | 1111 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 | 1112 | if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) { |
a3621e74 | 1113 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 1114 | hop_and_restart: |
1aa99e6b | 1115 | s = HOP3c(t, 1, strend); |
66e933ab GS |
1116 | if (s + start_shift + end_shift > strend) { |
1117 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 1118 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
1119 | "Could not match STCLASS...\n") ); |
1120 | goto fail; | |
1121 | } | |
5e39e1e5 HS |
1122 | if (!check) |
1123 | goto giveup; | |
a3621e74 | 1124 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1125 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1126 | what, (long)(s + start_shift - i_strpos)) ); |
1127 | goto restart; | |
1128 | } | |
66e933ab | 1129 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
1130 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
1131 | goto retry_floating_check; | |
1132 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 1133 | s = check_at; |
5e39e1e5 HS |
1134 | if (!check) |
1135 | goto giveup; | |
a3621e74 | 1136 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1137 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1138 | (long)(other_last - i_strpos)) ); |
1139 | goto do_other_anchored; | |
1140 | } | |
60e71179 GS |
1141 | /* Another way we could have checked stclass at the |
1142 | current position only: */ | |
1143 | if (ml_anch) { | |
1144 | s = t = t + 1; | |
5e39e1e5 HS |
1145 | if (!check) |
1146 | goto giveup; | |
a3621e74 | 1147 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1148 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 1149 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 1150 | goto try_at_offset; |
66e933ab | 1151 | } |
f2ed9b32 | 1152 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 1153 | goto fail; |
486ec47a | 1154 | /* Check is floating substring. */ |
6eb5f6b9 JH |
1155 | retry_floating_check: |
1156 | t = check_at - start_shift; | |
a3621e74 | 1157 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
1158 | goto hop_and_restart; |
1159 | } | |
b7953727 | 1160 | if (t != s) { |
a3621e74 | 1161 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 1162 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
1163 | (long)(t - i_strpos), (long)(s - i_strpos)) |
1164 | ); | |
1165 | } | |
1166 | else { | |
a3621e74 | 1167 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1168 | "Does not contradict STCLASS...\n"); |
1169 | ); | |
1170 | } | |
6eb5f6b9 | 1171 | } |
5e39e1e5 | 1172 | giveup: |
a3621e74 | 1173 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
1174 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
1175 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 1176 | return s; |
2c2d71f5 JH |
1177 | |
1178 | fail_finish: /* Substring not found */ | |
33b8afdf | 1179 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
f2ed9b32 | 1180 | BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ |
cad2e5aa | 1181 | fail: |
a3621e74 | 1182 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 1183 | PL_colors[4], PL_colors[5])); |
bd61b366 | 1184 | return NULL; |
cad2e5aa | 1185 | } |
9661b544 | 1186 | |
a0a388a1 YO |
1187 | #define DECL_TRIE_TYPE(scan) \ |
1188 | const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \ | |
1189 | trie_type = (scan->flags != EXACT) \ | |
f2ed9b32 KW |
1190 | ? (utf8_target ? trie_utf8_fold : (UTF_PATTERN ? trie_latin_utf8_fold : trie_plain)) \ |
1191 | : (utf8_target ? trie_utf8 : trie_plain) | |
3b0527fe | 1192 | |
55eed653 NC |
1193 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ |
1194 | uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ | |
4cadc6a9 YO |
1195 | switch (trie_type) { \ |
1196 | case trie_utf8_fold: \ | |
1197 | if ( foldlen>0 ) { \ | |
0abd0d78 | 1198 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ |
4cadc6a9 YO |
1199 | foldlen -= len; \ |
1200 | uscan += len; \ | |
1201 | len=0; \ | |
1202 | } else { \ | |
0abd0d78 | 1203 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ |
4cadc6a9 YO |
1204 | uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \ |
1205 | foldlen -= UNISKIP( uvc ); \ | |
1206 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1207 | } \ | |
1208 | break; \ | |
a0a388a1 YO |
1209 | case trie_latin_utf8_fold: \ |
1210 | if ( foldlen>0 ) { \ | |
1211 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1212 | foldlen -= len; \ | |
1213 | uscan += len; \ | |
1214 | len=0; \ | |
1215 | } else { \ | |
1216 | len = 1; \ | |
1217 | uvc = to_uni_fold( *(U8*)uc, foldbuf, &foldlen ); \ | |
1218 | foldlen -= UNISKIP( uvc ); \ | |
1219 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1220 | } \ | |
1221 | break; \ | |
4cadc6a9 YO |
1222 | case trie_utf8: \ |
1223 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ | |
1224 | break; \ | |
1225 | case trie_plain: \ | |
1226 | uvc = (UV)*uc; \ | |
1227 | len = 1; \ | |
1228 | } \ | |
4cadc6a9 YO |
1229 | if (uvc < 256) { \ |
1230 | charid = trie->charmap[ uvc ]; \ | |
1231 | } \ | |
1232 | else { \ | |
1233 | charid = 0; \ | |
55eed653 NC |
1234 | if (widecharmap) { \ |
1235 | SV** const svpp = hv_fetch(widecharmap, \ | |
4cadc6a9 YO |
1236 | (char*)&uvc, sizeof(UV), 0); \ |
1237 | if (svpp) \ | |
1238 | charid = (U16)SvIV(*svpp); \ | |
1239 | } \ | |
1240 | } \ | |
1241 | } STMT_END | |
1242 | ||
a0a388a1 YO |
1243 | #define REXEC_FBC_EXACTISH_CHECK(CoNd) \ |
1244 | { \ | |
1245 | char *my_strend= (char *)strend; \ | |
4cadc6a9 YO |
1246 | if ( (CoNd) \ |
1247 | && (ln == len || \ | |
f2ed9b32 KW |
1248 | foldEQ_utf8(s, &my_strend, 0, utf8_target, \ |
1249 | m, NULL, ln, cBOOL(UTF_PATTERN))) \ | |
a0a388a1 | 1250 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1251 | goto got_it; \ |
1252 | else { \ | |
1253 | U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \ | |
1254 | uvchr_to_utf8(tmpbuf, c); \ | |
1255 | f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \ | |
1256 | if ( f != c \ | |
1257 | && (f == c1 || f == c2) \ | |
a0a388a1 | 1258 | && (ln == len || \ |
f2ed9b32 KW |
1259 | foldEQ_utf8(s, &my_strend, 0, utf8_target,\ |
1260 | m, NULL, ln, cBOOL(UTF_PATTERN)))\ | |
a0a388a1 | 1261 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1262 | goto got_it; \ |
1263 | } \ | |
a0a388a1 YO |
1264 | } \ |
1265 | s += len | |
4cadc6a9 YO |
1266 | |
1267 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ | |
1268 | STMT_START { \ | |
a932d541 | 1269 | re_fold_t folder; \ |
9a5a5549 KW |
1270 | switch (OP(c)) { \ |
1271 | case EXACTFU: folder = foldEQ_latin1; break; \ | |
1272 | case EXACTFL: folder = foldEQ_locale; break; \ | |
1273 | case EXACTF: folder = foldEQ; break; \ | |
1274 | default: \ | |
1275 | Perl_croak(aTHX_ "panic: Unexpected op %u", OP(c)); \ | |
1276 | } \ | |
4cadc6a9 YO |
1277 | while (s <= e) { \ |
1278 | if ( (CoNd) \ | |
9a5a5549 KW |
1279 | && (ln == 1 || folder(s, m, ln)) \ |
1280 | && (!reginfo || regtry(reginfo, &s)) ) \ | |
4cadc6a9 YO |
1281 | goto got_it; \ |
1282 | s++; \ | |
1283 | } \ | |
1284 | } STMT_END | |
1285 | ||
1286 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
1287 | STMT_START { \ | |
1288 | while (s + (uskip = UTF8SKIP(s)) <= strend) { \ | |
1289 | CoDe \ | |
1290 | s += uskip; \ | |
1291 | } \ | |
1292 | } STMT_END | |
1293 | ||
1294 | #define REXEC_FBC_SCAN(CoDe) \ | |
1295 | STMT_START { \ | |
1296 | while (s < strend) { \ | |
1297 | CoDe \ | |
1298 | s++; \ | |
1299 | } \ | |
1300 | } STMT_END | |
1301 | ||
1302 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1303 | REXEC_FBC_UTF8_SCAN( \ | |
1304 | if (CoNd) { \ | |
24b23f37 | 1305 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1306 | goto got_it; \ |
1307 | else \ | |
1308 | tmp = doevery; \ | |
1309 | } \ | |
1310 | else \ | |
1311 | tmp = 1; \ | |
1312 | ) | |
1313 | ||
1314 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1315 | REXEC_FBC_SCAN( \ | |
1316 | if (CoNd) { \ | |
24b23f37 | 1317 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1318 | goto got_it; \ |
1319 | else \ | |
1320 | tmp = doevery; \ | |
1321 | } \ | |
1322 | else \ | |
1323 | tmp = 1; \ | |
1324 | ) | |
1325 | ||
1326 | #define REXEC_FBC_TRYIT \ | |
24b23f37 | 1327 | if ((!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1328 | goto got_it |
1329 | ||
e1d1eefb | 1330 | #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \ |
f2ed9b32 | 1331 | if (utf8_target) { \ |
e1d1eefb YO |
1332 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1333 | } \ | |
1334 | else { \ | |
1335 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
1336 | } \ | |
1337 | break | |
1338 | ||
4cadc6a9 | 1339 | #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ |
f2ed9b32 | 1340 | if (utf8_target) { \ |
4cadc6a9 YO |
1341 | UtFpReLoAd; \ |
1342 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1343 | } \ | |
1344 | else { \ | |
1345 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
1346 | } \ | |
1347 | break | |
1348 | ||
1349 | #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ | |
1350 | PL_reg_flags |= RF_tainted; \ | |
f2ed9b32 | 1351 | if (utf8_target) { \ |
4cadc6a9 YO |
1352 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1353 | } \ | |
1354 | else { \ | |
1355 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
1356 | } \ | |
1357 | break | |
1358 | ||
786e8c11 YO |
1359 | #define DUMP_EXEC_POS(li,s,doutf8) \ |
1360 | dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8) | |
1361 | ||
cfaf538b KW |
1362 | |
1363 | #define UTF8_NOLOAD(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ | |
1364 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1365 | tmp = TEST_NON_UTF8(tmp); \ | |
1366 | REXEC_FBC_UTF8_SCAN( \ | |
1367 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1368 | tmp = !tmp; \ | |
1369 | IF_SUCCESS; \ | |
1370 | } \ | |
1371 | else { \ | |
1372 | IF_FAIL; \ | |
1373 | } \ | |
1374 | ); \ | |
1375 | ||
1376 | #define UTF8_LOAD(TeSt1_UtF8, TeSt2_UtF8, IF_SUCCESS, IF_FAIL) \ | |
1377 | if (s == PL_bostr) { \ | |
1378 | tmp = '\n'; \ | |
1379 | } \ | |
1380 | else { \ | |
1381 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); \ | |
1382 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); \ | |
1383 | } \ | |
1384 | tmp = TeSt1_UtF8; \ | |
1385 | LOAD_UTF8_CHARCLASS_ALNUM(); \ | |
1386 | REXEC_FBC_UTF8_SCAN( \ | |
1387 | if (tmp == ! (TeSt2_UtF8)) { \ | |
1388 | tmp = !tmp; \ | |
1389 | IF_SUCCESS; \ | |
1390 | } \ | |
1391 | else { \ | |
1392 | IF_FAIL; \ | |
1393 | } \ | |
1394 | ); \ | |
1395 | ||
63ac0dad KW |
1396 | /* The only difference between the BOUND and NBOUND cases is that |
1397 | * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in | |
1398 | * NBOUND. This is accomplished by passing it in either the if or else clause, | |
1399 | * with the other one being empty */ | |
1400 | #define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1401 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
cfaf538b KW |
1402 | |
1403 | #define FBC_BOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1404 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) |
63ac0dad KW |
1405 | |
1406 | #define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1407 | FBC_BOUND_COMMON(UTF8_LOAD(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b KW |
1408 | |
1409 | #define FBC_NBOUND_NOLOAD(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ | |
1dcf4a1b | 1410 | FBC_BOUND_COMMON(UTF8_NOLOAD(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) |
cfaf538b | 1411 | |
63ac0dad KW |
1412 | |
1413 | /* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to | |
1414 | * be passed in completely with the variable name being tested, which isn't | |
1415 | * such a clean interface, but this is easier to read than it was before. We | |
1416 | * are looking for the boundary (or non-boundary between a word and non-word | |
1417 | * character. The utf8 and non-utf8 cases have the same logic, but the details | |
1418 | * must be different. Find the "wordness" of the character just prior to this | |
1419 | * one, and compare it with the wordness of this one. If they differ, we have | |
1420 | * a boundary. At the beginning of the string, pretend that the previous | |
1421 | * character was a new-line */ | |
cfaf538b | 1422 | #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ |
63ac0dad | 1423 | if (utf8_target) { \ |
cfaf538b | 1424 | UTF8_CODE \ |
63ac0dad KW |
1425 | } \ |
1426 | else { /* Not utf8 */ \ | |
1427 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; \ | |
1428 | tmp = TEST_NON_UTF8(tmp); \ | |
1429 | REXEC_FBC_SCAN( \ | |
1430 | if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ | |
1431 | tmp = !tmp; \ | |
1432 | IF_SUCCESS; \ | |
1433 | } \ | |
1434 | else { \ | |
1435 | IF_FAIL; \ | |
1436 | } \ | |
1437 | ); \ | |
1438 | } \ | |
1439 | if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) \ | |
1440 | goto got_it; | |
1441 | ||
786e8c11 YO |
1442 | /* We know what class REx starts with. Try to find this position... */ |
1443 | /* if reginfo is NULL, its a dryrun */ | |
1444 | /* annoyingly all the vars in this routine have different names from their counterparts | |
1445 | in regmatch. /grrr */ | |
1446 | ||
3c3eec57 | 1447 | STATIC char * |
07be1b83 | 1448 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
24b23f37 | 1449 | const char *strend, regmatch_info *reginfo) |
a687059c | 1450 | { |
27da23d5 | 1451 | dVAR; |
bbe252da | 1452 | const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; |
6eb5f6b9 | 1453 | char *m; |
d8093b23 | 1454 | STRLEN ln; |
5dab1207 | 1455 | STRLEN lnc; |
078c425b | 1456 | register STRLEN uskip; |
d8093b23 G |
1457 | unsigned int c1; |
1458 | unsigned int c2; | |
6eb5f6b9 JH |
1459 | char *e; |
1460 | register I32 tmp = 1; /* Scratch variable? */ | |
f2ed9b32 | 1461 | register const bool utf8_target = PL_reg_match_utf8; |
f8fc2ecf | 1462 | RXi_GET_DECL(prog,progi); |
7918f24d NC |
1463 | |
1464 | PERL_ARGS_ASSERT_FIND_BYCLASS; | |
f8fc2ecf | 1465 | |
6eb5f6b9 JH |
1466 | /* We know what class it must start with. */ |
1467 | switch (OP(c)) { | |
f56b6394 | 1468 | case ANYOFV: |
6eb5f6b9 | 1469 | case ANYOF: |
f56b6394 | 1470 | if (utf8_target || OP(c) == ANYOFV) { |
3ff7ceb3 | 1471 | REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_NONBITMAP) || |
388cc4de | 1472 | !UTF8_IS_INVARIANT((U8)s[0]) ? |
f2ed9b32 | 1473 | reginclass(prog, c, (U8*)s, 0, utf8_target) : |
4cadc6a9 | 1474 | REGINCLASS(prog, c, (U8*)s)); |
388cc4de HS |
1475 | } |
1476 | else { | |
1477 | while (s < strend) { | |
1478 | STRLEN skip = 1; | |
1479 | ||
32fc9b6a | 1480 | if (REGINCLASS(prog, c, (U8*)s) || |
388cc4de HS |
1481 | (ANYOF_FOLD_SHARP_S(c, s, strend) && |
1482 | /* The assignment of 2 is intentional: | |
1483 | * for the folded sharp s, the skip is 2. */ | |
1484 | (skip = SHARP_S_SKIP))) { | |
24b23f37 | 1485 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
388cc4de HS |
1486 | goto got_it; |
1487 | else | |
1488 | tmp = doevery; | |
1489 | } | |
1490 | else | |
1491 | tmp = 1; | |
1492 | s += skip; | |
1493 | } | |
a0d0e21e | 1494 | } |
6eb5f6b9 | 1495 | break; |
f33976b4 | 1496 | case CANY: |
4cadc6a9 | 1497 | REXEC_FBC_SCAN( |
24b23f37 | 1498 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
f33976b4 DB |
1499 | goto got_it; |
1500 | else | |
1501 | tmp = doevery; | |
4cadc6a9 | 1502 | ); |
f33976b4 | 1503 | break; |
9a5a5549 | 1504 | case EXACTFU: |
6eb5f6b9 | 1505 | case EXACTF: |
5dab1207 NIS |
1506 | m = STRING(c); |
1507 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1508 | lnc = (I32) ln; /* length to match in characters */ | |
f2ed9b32 | 1509 | if (UTF_PATTERN) { |
a2a2844f | 1510 | STRLEN ulen1, ulen2; |
5dab1207 | 1511 | U8 *sm = (U8 *) m; |
89ebb4a3 JH |
1512 | U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; |
1513 | U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; | |
97dc7d3e RGS |
1514 | /* used by commented-out code below */ |
1515 | /*const U32 uniflags = UTF8_ALLOW_DEFAULT;*/ | |
a0a388a1 YO |
1516 | |
1517 | /* XXX: Since the node will be case folded at compile | |
1518 | time this logic is a little odd, although im not | |
1519 | sure that its actually wrong. --dmq */ | |
1520 | ||
1521 | c1 = to_utf8_lower((U8*)m, tmpbuf1, &ulen1); | |
1522 | c2 = to_utf8_upper((U8*)m, tmpbuf2, &ulen2); | |
1523 | ||
1524 | /* XXX: This is kinda strange. to_utf8_XYZ returns the | |
1525 | codepoint of the first character in the converted | |
1526 | form, yet originally we did the extra step. | |
1527 | No tests fail by commenting this code out however | |
1528 | so Ive left it out. -- dmq. | |
1529 | ||
89ebb4a3 | 1530 | c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE, |
041457d9 | 1531 | 0, uniflags); |
89ebb4a3 | 1532 | c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE, |
041457d9 | 1533 | 0, uniflags); |
a0a388a1 YO |
1534 | */ |
1535 | ||
5dab1207 NIS |
1536 | lnc = 0; |
1537 | while (sm < ((U8 *) m + ln)) { | |
1538 | lnc++; | |
1539 | sm += UTF8SKIP(sm); | |
1540 | } | |
1aa99e6b IH |
1541 | } |
1542 | else { | |
1543 | c1 = *(U8*)m; | |
9a5a5549 KW |
1544 | if (utf8_target || OP(c) == EXACTFU) { |
1545 | ||
1546 | /* Micro sign folds to GREEK SMALL LETTER MU; | |
1547 | LATIN_SMALL_LETTER_SHARP_S folds to 'ss', and this sets | |
1548 | c2 to the first 's' of the pair, and the code below will | |
1549 | look for others */ | |
1550 | c2 = (c1 == MICRO_SIGN) | |
1551 | ? GREEK_SMALL_LETTER_MU | |
1552 | : (c1 == LATIN_SMALL_LETTER_SHARP_S) | |
1553 | ? 's' | |
1554 | : PL_fold_latin1[c1]; | |
1555 | } else c2 = PL_fold[c1]; | |
1aa99e6b | 1556 | } |
6eb5f6b9 JH |
1557 | goto do_exactf; |
1558 | case EXACTFL: | |
5dab1207 NIS |
1559 | m = STRING(c); |
1560 | ln = STR_LEN(c); | |
1561 | lnc = (I32) ln; | |
d8093b23 | 1562 | c1 = *(U8*)m; |
6eb5f6b9 JH |
1563 | c2 = PL_fold_locale[c1]; |
1564 | do_exactf: | |
db12adc6 | 1565 | e = HOP3c(strend, -((I32)lnc), s); |
b3c9acc1 | 1566 | |
3b0527fe | 1567 | if (!reginfo && e < s) |
6eb5f6b9 | 1568 | e = s; /* Due to minlen logic of intuit() */ |
1aa99e6b | 1569 | |
60a8b682 JH |
1570 | /* The idea in the EXACTF* cases is to first find the |
1571 | * first character of the EXACTF* node and then, if | |
1572 | * necessary, case-insensitively compare the full | |
1573 | * text of the node. The c1 and c2 are the first | |
1574 | * characters (though in Unicode it gets a bit | |
1575 | * more complicated because there are more cases | |
7f16dd3d JH |
1576 | * than just upper and lower: one needs to use |
1577 | * the so-called folding case for case-insensitive | |
1578 | * matching (called "loose matching" in Unicode). | |
4c1b470c | 1579 | * foldEQ_utf8() will do just that. */ |
60a8b682 | 1580 | |
f2ed9b32 | 1581 | if (utf8_target || UTF_PATTERN) { |
575cac57 | 1582 | UV c, f; |
89ebb4a3 | 1583 | U8 tmpbuf [UTF8_MAXBYTES+1]; |
a0a388a1 YO |
1584 | STRLEN len = 1; |
1585 | STRLEN foldlen; | |
4ad0818d | 1586 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
09091399 | 1587 | if (c1 == c2) { |
5dab1207 NIS |
1588 | /* Upper and lower of 1st char are equal - |
1589 | * probably not a "letter". */ | |
1aa99e6b | 1590 | while (s <= e) { |
f2ed9b32 | 1591 | if (utf8_target) { |
a0a388a1 | 1592 | c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len, |
041457d9 | 1593 | uniflags); |
a0a388a1 YO |
1594 | } else { |
1595 | c = *((U8*)s); | |
1596 | } | |
4cadc6a9 | 1597 | REXEC_FBC_EXACTISH_CHECK(c == c1); |
1aa99e6b | 1598 | } |
09091399 JH |
1599 | } |
1600 | else { | |
1aa99e6b | 1601 | while (s <= e) { |
f2ed9b32 | 1602 | if (utf8_target) { |
a0a388a1 | 1603 | c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len, |
041457d9 | 1604 | uniflags); |
a0a388a1 YO |
1605 | } else { |
1606 | c = *((U8*)s); | |
1607 | } | |
80aecb99 | 1608 | |
60a8b682 | 1609 | /* Handle some of the three Greek sigmas cases. |
8c01da3c JH |
1610 | * Note that not all the possible combinations |
1611 | * are handled here: some of them are handled | |
1612 | * by the standard folding rules, and some of | |
1613 | * them (the character class or ANYOF cases) | |
1614 | * are handled during compiletime in | |
1615 | * regexec.c:S_regclass(). */ | |
880bd946 JH |
1616 | if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA || |
1617 | c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA) | |
1618 | c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA; | |
80aecb99 | 1619 | |
4cadc6a9 | 1620 | REXEC_FBC_EXACTISH_CHECK(c == c1 || c == c2); |
1aa99e6b | 1621 | } |
09091399 | 1622 | } |
1aa99e6b IH |
1623 | } |
1624 | else { | |
a0a388a1 | 1625 | /* Neither pattern nor string are UTF8 */ |
1aa99e6b | 1626 | if (c1 == c2) |
4cadc6a9 | 1627 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); |
1aa99e6b | 1628 | else |
4cadc6a9 | 1629 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); |
b3c9acc1 IZ |
1630 | } |
1631 | break; | |
bbce6d69 | 1632 | case BOUNDL: |
3280af22 | 1633 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1634 | FBC_BOUND(isALNUM_LC, |
1635 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1636 | isALNUM_LC_utf8((U8*)s)); | |
a0ed51b3 | 1637 | break; |
bbce6d69 | 1638 | case NBOUNDL: |
3280af22 | 1639 | PL_reg_flags |= RF_tainted; |
63ac0dad KW |
1640 | FBC_NBOUND(isALNUM_LC, |
1641 | isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp)), | |
1642 | isALNUM_LC_utf8((U8*)s)); | |
1643 | break; | |
1644 | case BOUND: | |
1645 | FBC_BOUND(isWORDCHAR, | |
1646 | isALNUM_uni(tmp), | |
1647 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1648 | break; | |
cfaf538b KW |
1649 | case BOUNDA: |
1650 | FBC_BOUND_NOLOAD(isWORDCHAR_A, | |
1651 | isWORDCHAR_A(tmp), | |
1652 | isWORDCHAR_A((U8*)s)); | |
1653 | break; | |
a0d0e21e | 1654 | case NBOUND: |
63ac0dad KW |
1655 | FBC_NBOUND(isWORDCHAR, |
1656 | isALNUM_uni(tmp), | |
1657 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1658 | break; | |
cfaf538b KW |
1659 | case NBOUNDA: |
1660 | FBC_NBOUND_NOLOAD(isWORDCHAR_A, | |
1661 | isWORDCHAR_A(tmp), | |
1662 | isWORDCHAR_A((U8*)s)); | |
1663 | break; | |
63ac0dad KW |
1664 | case BOUNDU: |
1665 | FBC_BOUND(isWORDCHAR_L1, | |
1666 | isALNUM_uni(tmp), | |
1667 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
1668 | break; | |
1669 | case NBOUNDU: | |
1670 | FBC_NBOUND(isWORDCHAR_L1, | |
1671 | isALNUM_uni(tmp), | |
1672 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target))); | |
a0ed51b3 | 1673 | break; |
bbce6d69 | 1674 | case ALNUML: |
4cadc6a9 YO |
1675 | REXEC_FBC_CSCAN_TAINT( |
1676 | isALNUM_LC_utf8((U8*)s), | |
1677 | isALNUM_LC(*s) | |
1678 | ); | |
6895a8aa | 1679 | break; |
980866de KW |
1680 | case ALNUMU: |
1681 | REXEC_FBC_CSCAN_PRELOAD( | |
1682 | LOAD_UTF8_CHARCLASS_PERL_WORD(), | |
1683 | swash_fetch(RE_utf8_perl_word,(U8*)s, utf8_target), | |
1684 | isWORDCHAR_L1((U8) *s) | |
1685 | ); | |
6895a8aa | 1686 | break; |
980866de KW |
1687 | case ALNUM: |
1688 | REXEC_FBC_CSCAN_PRELOAD( | |
1689 | LOAD_UTF8_CHARCLASS_PERL_WORD(), | |
1690 | swash_fetch(RE_utf8_perl_word,(U8*)s, utf8_target), | |
1691 | isWORDCHAR((U8) *s) | |
1692 | ); | |
6895a8aa | 1693 | break; |
cfaf538b | 1694 | case ALNUMA: |
8e9da4d4 KW |
1695 | /* Don't need to worry about utf8, as it can match only a single |
1696 | * byte invariant character */ | |
cfaf538b | 1697 | REXEC_FBC_CLASS_SCAN( isWORDCHAR_A(*s)); |
6895a8aa | 1698 | break; |
980866de KW |
1699 | case NALNUMU: |
1700 | REXEC_FBC_CSCAN_PRELOAD( | |
1701 | LOAD_UTF8_CHARCLASS_PERL_WORD(), | |
1702 | swash_fetch(RE_utf8_perl_word,(U8*)s, utf8_target), | |
1703 | ! isWORDCHAR_L1((U8) *s) | |
1704 | ); | |
6895a8aa | 1705 | break; |
a0d0e21e | 1706 | case NALNUM: |
4cadc6a9 | 1707 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1708 | LOAD_UTF8_CHARCLASS_PERL_WORD(), |
f2ed9b32 | 1709 | !swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target), |
980866de | 1710 | ! isALNUM(*s) |
4cadc6a9 | 1711 | ); |
6895a8aa | 1712 | break; |
cfaf538b | 1713 | case NALNUMA: |
8e9da4d4 KW |
1714 | REXEC_FBC_CSCAN( |
1715 | !isWORDCHAR_A(*s), | |
1716 | !isWORDCHAR_A(*s) | |
1717 | ); | |
1718 | break; | |
bbce6d69 | 1719 | case NALNUML: |
4cadc6a9 YO |
1720 | REXEC_FBC_CSCAN_TAINT( |
1721 | !isALNUM_LC_utf8((U8*)s), | |
1722 | !isALNUM_LC(*s) | |
1723 | ); | |
6895a8aa | 1724 | break; |
980866de KW |
1725 | case SPACEU: |
1726 | REXEC_FBC_CSCAN_PRELOAD( | |
1727 | LOAD_UTF8_CHARCLASS_PERL_SPACE(), | |
1728 | *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target), | |
1729 | isSPACE_L1((U8) *s) | |
1730 | ); | |
6895a8aa | 1731 | break; |
a0d0e21e | 1732 | case SPACE: |
4cadc6a9 | 1733 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1734 | LOAD_UTF8_CHARCLASS_PERL_SPACE(), |
f2ed9b32 | 1735 | *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target), |
980866de | 1736 | isSPACE((U8) *s) |
4cadc6a9 | 1737 | ); |
6895a8aa | 1738 | break; |
cfaf538b | 1739 | case SPACEA: |
8e9da4d4 KW |
1740 | /* Don't need to worry about utf8, as it can match only a single |
1741 | * byte invariant character */ | |
cfaf538b | 1742 | REXEC_FBC_CLASS_SCAN( isSPACE_A(*s)); |
6895a8aa | 1743 | break; |
bbce6d69 | 1744 | case SPACEL: |
4cadc6a9 | 1745 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1746 | isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1747 | isSPACE_LC(*s) |
1748 | ); | |
6895a8aa | 1749 | break; |
980866de KW |
1750 | case NSPACEU: |
1751 | REXEC_FBC_CSCAN_PRELOAD( | |
1752 | LOAD_UTF8_CHARCLASS_PERL_SPACE(), | |
1753 | !( *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target)), | |
1754 | ! isSPACE_L1((U8) *s) | |
1755 | ); | |
6895a8aa | 1756 | break; |
a0d0e21e | 1757 | case NSPACE: |
4cadc6a9 | 1758 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1759 | LOAD_UTF8_CHARCLASS_PERL_SPACE(), |
f2ed9b32 | 1760 | !(*s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target)), |
980866de | 1761 | ! isSPACE((U8) *s) |
4cadc6a9 | 1762 | ); |
6895a8aa | 1763 | break; |
cfaf538b | 1764 | case NSPACEA: |
8e9da4d4 KW |
1765 | REXEC_FBC_CSCAN( |
1766 | !isSPACE_A(*s), | |
1767 | !isSPACE_A(*s) | |
1768 | ); | |
1769 | break; | |
bbce6d69 | 1770 | case NSPACEL: |
4cadc6a9 | 1771 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1772 | !isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1773 | !isSPACE_LC(*s) |
1774 | ); | |
6895a8aa | 1775 | break; |
a0d0e21e | 1776 | case DIGIT: |
4cadc6a9 | 1777 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1778 | LOAD_UTF8_CHARCLASS_POSIX_DIGIT(), |
f2ed9b32 | 1779 | swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target), |
4cadc6a9 YO |
1780 | isDIGIT(*s) |
1781 | ); | |
6895a8aa | 1782 | break; |
cfaf538b | 1783 | case DIGITA: |
8e9da4d4 KW |
1784 | /* Don't need to worry about utf8, as it can match only a single |
1785 | * byte invariant character */ | |
cfaf538b | 1786 | REXEC_FBC_CLASS_SCAN( isDIGIT_A(*s)); |
6895a8aa | 1787 | break; |
b8c5462f | 1788 | case DIGITL: |
4cadc6a9 YO |
1789 | REXEC_FBC_CSCAN_TAINT( |
1790 | isDIGIT_LC_utf8((U8*)s), | |
1791 | isDIGIT_LC(*s) | |
1792 | ); | |
6895a8aa | 1793 | break; |
a0d0e21e | 1794 | case NDIGIT: |
4cadc6a9 | 1795 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1796 | LOAD_UTF8_CHARCLASS_POSIX_DIGIT(), |
f2ed9b32 | 1797 | !swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target), |
4cadc6a9 YO |
1798 | !isDIGIT(*s) |
1799 | ); | |
6895a8aa | 1800 | break; |
cfaf538b | 1801 | case NDIGITA: |
8e9da4d4 KW |
1802 | REXEC_FBC_CSCAN( |
1803 | !isDIGIT_A(*s), | |
1804 | !isDIGIT_A(*s) | |
1805 | ); | |
1806 | break; | |
b8c5462f | 1807 | case NDIGITL: |
4cadc6a9 YO |
1808 | REXEC_FBC_CSCAN_TAINT( |
1809 | !isDIGIT_LC_utf8((U8*)s), | |
1810 | !isDIGIT_LC(*s) | |
1811 | ); | |
6895a8aa | 1812 | break; |
e1d1eefb YO |
1813 | case LNBREAK: |
1814 | REXEC_FBC_CSCAN( | |
1815 | is_LNBREAK_utf8(s), | |
1816 | is_LNBREAK_latin1(s) | |
1817 | ); | |
6895a8aa | 1818 | break; |
e1d1eefb YO |
1819 | case VERTWS: |
1820 | REXEC_FBC_CSCAN( | |
1821 | is_VERTWS_utf8(s), | |
1822 | is_VERTWS_latin1(s) | |
1823 | ); | |
6895a8aa | 1824 | break; |
e1d1eefb YO |
1825 | case NVERTWS: |
1826 | REXEC_FBC_CSCAN( | |
1827 | !is_VERTWS_utf8(s), | |
1828 | !is_VERTWS_latin1(s) | |
1829 | ); | |
6895a8aa | 1830 | break; |
e1d1eefb YO |
1831 | case HORIZWS: |
1832 | REXEC_FBC_CSCAN( | |
1833 | is_HORIZWS_utf8(s), | |
1834 | is_HORIZWS_latin1(s) | |
1835 | ); | |
6895a8aa | 1836 | break; |
e1d1eefb YO |
1837 | case NHORIZWS: |
1838 | REXEC_FBC_CSCAN( | |
1839 | !is_HORIZWS_utf8(s), | |
1840 | !is_HORIZWS_latin1(s) | |
1841 | ); | |
6895a8aa | 1842 | break; |
1de06328 YO |
1843 | case AHOCORASICKC: |
1844 | case AHOCORASICK: | |
07be1b83 | 1845 | { |
a0a388a1 | 1846 | DECL_TRIE_TYPE(c); |
07be1b83 YO |
1847 | /* what trie are we using right now */ |
1848 | reg_ac_data *aho | |
f8fc2ecf | 1849 | = (reg_ac_data*)progi->data->data[ ARG( c ) ]; |
3251b653 NC |
1850 | reg_trie_data *trie |
1851 | = (reg_trie_data*)progi->data->data[ aho->trie ]; | |
85fbaab2 | 1852 | HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); |
07be1b83 YO |
1853 | |
1854 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1855 | #ifdef DEBUGGING |
07be1b83 | 1856 | const char *real_start = s; |
6148ee25 | 1857 | #endif |
07be1b83 | 1858 | STRLEN maxlen = trie->maxlen; |
be8e71aa YO |
1859 | SV *sv_points; |
1860 | U8 **points; /* map of where we were in the input string | |
786e8c11 | 1861 | when reading a given char. For ASCII this |
be8e71aa | 1862 | is unnecessary overhead as the relationship |
38a44b82 NC |
1863 | is always 1:1, but for Unicode, especially |
1864 | case folded Unicode this is not true. */ | |
f9e705e8 | 1865 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
786e8c11 YO |
1866 | U8 *bitmap=NULL; |
1867 | ||
07be1b83 YO |
1868 | |
1869 | GET_RE_DEBUG_FLAGS_DECL; | |
1870 | ||
be8e71aa YO |
1871 | /* We can't just allocate points here. We need to wrap it in |
1872 | * an SV so it gets freed properly if there is a croak while | |
1873 | * running the match */ | |
1874 | ENTER; | |
1875 | SAVETMPS; | |
1876 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1877 | SvCUR_set(sv_points, | |
1878 | maxlen * sizeof(U8 *)); | |
1879 | SvPOK_on(sv_points); | |
1880 | sv_2mortal(sv_points); | |
1881 | points=(U8**)SvPV_nolen(sv_points ); | |
1de06328 YO |
1882 | if ( trie_type != trie_utf8_fold |
1883 | && (trie->bitmap || OP(c)==AHOCORASICKC) ) | |
1884 | { | |
786e8c11 YO |
1885 | if (trie->bitmap) |
1886 | bitmap=(U8*)trie->bitmap; | |
1887 | else | |
1888 | bitmap=(U8*)ANYOF_BITMAP(c); | |
07be1b83 | 1889 | } |
786e8c11 YO |
1890 | /* this is the Aho-Corasick algorithm modified a touch |
1891 | to include special handling for long "unknown char" | |
1892 | sequences. The basic idea being that we use AC as long | |
1893 | as we are dealing with a possible matching char, when | |
1894 | we encounter an unknown char (and we have not encountered | |
1895 | an accepting state) we scan forward until we find a legal | |
1896 | starting char. | |
1897 | AC matching is basically that of trie matching, except | |
1898 | that when we encounter a failing transition, we fall back | |
1899 | to the current states "fail state", and try the current char | |
1900 | again, a process we repeat until we reach the root state, | |
1901 | state 1, or a legal transition. If we fail on the root state | |
1902 | then we can either terminate if we have reached an accepting | |
1903 | state previously, or restart the entire process from the beginning | |
1904 | if we have not. | |
1905 | ||
1906 | */ | |
07be1b83 YO |
1907 | while (s <= last_start) { |
1908 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1909 | U8 *uc = (U8*)s; | |
1910 | U16 charid = 0; | |
1911 | U32 base = 1; | |
1912 | U32 state = 1; | |
1913 | UV uvc = 0; | |
1914 | STRLEN len = 0; | |
1915 | STRLEN foldlen = 0; | |
1916 | U8 *uscan = (U8*)NULL; | |
1917 | U8 *leftmost = NULL; | |
786e8c11 YO |
1918 | #ifdef DEBUGGING |
1919 | U32 accepted_word= 0; | |
1920 | #endif | |
07be1b83 YO |
1921 | U32 pointpos = 0; |
1922 | ||
1923 | while ( state && uc <= (U8*)strend ) { | |
1924 | int failed=0; | |
786e8c11 YO |
1925 | U32 word = aho->states[ state ].wordnum; |
1926 | ||
1de06328 YO |
1927 | if( state==1 ) { |
1928 | if ( bitmap ) { | |
1929 | DEBUG_TRIE_EXECUTE_r( | |
1930 | if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1931 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1932 | (char *)uc, utf8_target ); |
1de06328 YO |
1933 | PerlIO_printf( Perl_debug_log, |
1934 | " Scanning for legal start char...\n"); | |
1935 | } | |
d085b490 YO |
1936 | ); |
1937 | if (utf8_target) { | |
1938 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1939 | uc += UTF8SKIP(uc); | |
1940 | } | |
1941 | } else { | |
1942 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1943 | uc++; | |
1944 | } | |
1945 | } | |
1de06328 | 1946 | s= (char *)uc; |
786e8c11 | 1947 | } |
786e8c11 YO |
1948 | if (uc >(U8*)last_start) break; |
1949 | } | |
1950 | ||
1951 | if ( word ) { | |
2e64971a | 1952 | U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ]; |
786e8c11 YO |
1953 | if (!leftmost || lpos < leftmost) { |
1954 | DEBUG_r(accepted_word=word); | |
07be1b83 | 1955 | leftmost= lpos; |
786e8c11 | 1956 | } |
07be1b83 | 1957 | if (base==0) break; |
786e8c11 | 1958 | |
07be1b83 YO |
1959 | } |
1960 | points[pointpos++ % maxlen]= uc; | |
55eed653 NC |
1961 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
1962 | uscan, len, uvc, charid, foldlen, | |
1963 | foldbuf, uniflags); | |
786e8c11 YO |
1964 | DEBUG_TRIE_EXECUTE_r({ |
1965 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1966 | s, utf8_target ); |
07be1b83 | 1967 | PerlIO_printf(Perl_debug_log, |
786e8c11 YO |
1968 | " Charid:%3u CP:%4"UVxf" ", |
1969 | charid, uvc); | |
1970 | }); | |
07be1b83 YO |
1971 | |
1972 | do { | |
6148ee25 | 1973 | #ifdef DEBUGGING |
786e8c11 | 1974 | word = aho->states[ state ].wordnum; |
6148ee25 | 1975 | #endif |
07be1b83 YO |
1976 | base = aho->states[ state ].trans.base; |
1977 | ||
786e8c11 YO |
1978 | DEBUG_TRIE_EXECUTE_r({ |
1979 | if (failed) | |
1980 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1981 | s, utf8_target ); |
07be1b83 | 1982 | PerlIO_printf( Perl_debug_log, |
786e8c11 YO |
1983 | "%sState: %4"UVxf", word=%"UVxf, |
1984 | failed ? " Fail transition to " : "", | |
1985 | (UV)state, (UV)word); | |
1986 | }); | |
07be1b83 YO |
1987 | if ( base ) { |
1988 | U32 tmp; | |
6dd2be57 | 1989 | I32 offset; |
07be1b83 | 1990 | if (charid && |
6dd2be57 DM |
1991 | ( ((offset = base + charid |
1992 | - 1 - trie->uniquecharcount)) >= 0) | |
1993 | && ((U32)offset < trie->lasttrans) | |
1994 | && trie->trans[offset].check == state | |
1995 | && (tmp=trie->trans[offset].next)) | |
07be1b83 | 1996 | { |
786e8c11 YO |
1997 | DEBUG_TRIE_EXECUTE_r( |
1998 | PerlIO_printf( Perl_debug_log," - legal\n")); | |
07be1b83 YO |
1999 | state = tmp; |
2000 | break; | |
2001 | } | |
2002 | else { | |
786e8c11 YO |
2003 | DEBUG_TRIE_EXECUTE_r( |
2004 | PerlIO_printf( Perl_debug_log," - fail\n")); | |
2005 | failed = 1; | |
2006 | state = aho->fail[state]; | |
07be1b83 YO |
2007 | } |
2008 | } | |
2009 | else { | |
2010 | /* we must be accepting here */ | |
786e8c11 YO |
2011 | DEBUG_TRIE_EXECUTE_r( |
2012 | PerlIO_printf( Perl_debug_log," - accepting\n")); | |
2013 | failed = 1; | |
07be1b83 YO |
2014 | break; |
2015 | } | |
2016 | } while(state); | |
786e8c11 | 2017 | uc += len; |
07be1b83 YO |
2018 | if (failed) { |
2019 | if (leftmost) | |
2020 | break; | |
786e8c11 | 2021 | if (!state) state = 1; |
07be1b83 YO |
2022 | } |
2023 | } | |
2024 | if ( aho->states[ state ].wordnum ) { | |
2e64971a | 2025 | U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ]; |
786e8c11 YO |
2026 | if (!leftmost || lpos < leftmost) { |
2027 | DEBUG_r(accepted_word=aho->states[ state ].wordnum); | |
07be1b83 | 2028 | leftmost = lpos; |
786e8c11 | 2029 | } |
07be1b83 | 2030 | } |
07be1b83 YO |
2031 | if (leftmost) { |
2032 | s = (char*)leftmost; | |
786e8c11 YO |
2033 | DEBUG_TRIE_EXECUTE_r({ |
2034 | PerlIO_printf( | |
70685ca0 JH |
2035 | Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", |
2036 | (UV)accepted_word, (IV)(s - real_start) | |
786e8c11 YO |
2037 | ); |
2038 | }); | |
24b23f37 | 2039 | if (!reginfo || regtry(reginfo, &s)) { |
be8e71aa YO |
2040 | FREETMPS; |
2041 | LEAVE; | |
07be1b83 | 2042 | goto got_it; |
be8e71aa | 2043 | } |
07be1b83 | 2044 | s = HOPc(s,1); |
786e8c11 YO |
2045 | DEBUG_TRIE_EXECUTE_r({ |
2046 | PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); | |
2047 | }); | |
07be1b83 | 2048 | } else { |
786e8c11 YO |
2049 | DEBUG_TRIE_EXECUTE_r( |
2050 | PerlIO_printf( Perl_debug_log,"No match.\n")); | |
07be1b83 YO |
2051 | break; |
2052 | } | |
2053 | } | |
be8e71aa YO |
2054 | FREETMPS; |
2055 | LEAVE; | |
07be1b83 YO |
2056 | } |
2057 | break; | |
b3c9acc1 | 2058 | default: |
3c3eec57 GS |
2059 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); |
2060 | break; | |
d6a28714 | 2061 | } |
6eb5f6b9 JH |
2062 | return 0; |
2063 | got_it: | |
2064 | return s; | |
2065 | } | |
2066 | ||
fae667d5 | 2067 | |
6eb5f6b9 JH |
2068 | /* |
2069 | - regexec_flags - match a regexp against a string | |
2070 | */ | |
2071 | I32 | |
288b8c02 | 2072 | Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend, |
6eb5f6b9 JH |
2073 | char *strbeg, I32 minend, SV *sv, void *data, U32 flags) |
2074 | /* strend: pointer to null at end of string */ | |
2075 | /* strbeg: real beginning of string */ | |
2076 | /* minend: end of match must be >=minend after stringarg. */ | |
58e23c8d YO |
2077 | /* data: May be used for some additional optimizations. |
2078 | Currently its only used, with a U32 cast, for transmitting | |
2079 | the ganch offset when doing a /g match. This will change */ | |
6eb5f6b9 JH |
2080 | /* nosave: For optimizations. */ |
2081 | { | |
97aff369 | 2082 | dVAR; |
288b8c02 | 2083 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
24b23f37 | 2084 | /*register*/ char *s; |
6eb5f6b9 | 2085 | register regnode *c; |
24b23f37 | 2086 | /*register*/ char *startpos = stringarg; |
6eb5f6b9 JH |
2087 | I32 minlen; /* must match at least this many chars */ |
2088 | I32 dontbother = 0; /* how many characters not to try at end */ | |
6eb5f6b9 JH |
2089 | I32 end_shift = 0; /* Same for the end. */ /* CC */ |
2090 | I32 scream_pos = -1; /* Internal iterator of scream. */ | |
ccac19ea | 2091 | char *scream_olds = NULL; |
f2ed9b32 | 2092 | const bool utf8_target = cBOOL(DO_UTF8(sv)); |
2757e526 | 2093 | I32 multiline; |
f8fc2ecf | 2094 | RXi_GET_DECL(prog,progi); |
3b0527fe | 2095 | regmatch_info reginfo; /* create some info to pass to regtry etc */ |
e9105d30 | 2096 | regexp_paren_pair *swap = NULL; |
a3621e74 YO |
2097 | GET_RE_DEBUG_FLAGS_DECL; |
2098 | ||
7918f24d | 2099 | PERL_ARGS_ASSERT_REGEXEC_FLAGS; |
9d4ba2ae | 2100 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
2101 | |
2102 | /* Be paranoid... */ | |
2103 | if (prog == NULL || startpos == NULL) { | |
2104 | Perl_croak(aTHX_ "NULL regexp parameter"); | |
2105 | return 0; | |
2106 | } | |
2107 | ||
bbe252da | 2108 | multiline = prog->extflags & RXf_PMf_MULTILINE; |
288b8c02 | 2109 | reginfo.prog = rx; /* Yes, sorry that this is confusing. */ |
2757e526 | 2110 | |
f2ed9b32 | 2111 | RX_MATCH_UTF8_set(rx, utf8_target); |
1de06328 | 2112 | DEBUG_EXECUTE_r( |
f2ed9b32 | 2113 | debug_start_match(rx, utf8_target, startpos, strend, |
1de06328 YO |
2114 | "Matching"); |
2115 | ); | |
bac06658 | 2116 | |
6eb5f6b9 | 2117 | minlen = prog->minlen; |
1de06328 YO |
2118 | |
2119 | if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { | |
a3621e74 | 2120 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
2121 | "String too short [regexec_flags]...\n")); |
2122 | goto phooey; | |
1aa99e6b | 2123 | } |
6eb5f6b9 | 2124 | |
1de06328 | 2125 | |
6eb5f6b9 | 2126 | /* Check validity of program. */ |
f8fc2ecf | 2127 | if (UCHARAT(progi->program) != REG_MAGIC) { |
6eb5f6b9 JH |
2128 | Perl_croak(aTHX_ "corrupted regexp program"); |
2129 | } | |
2130 | ||
2131 | PL_reg_flags = 0; | |
2132 | PL_reg_eval_set = 0; | |
2133 | PL_reg_maxiter = 0; | |
2134 | ||
3c8556c3 | 2135 | if (RX_UTF8(rx)) |
6eb5f6b9 JH |
2136 | PL_reg_flags |= RF_utf8; |
2137 | ||
2138 | /* Mark beginning of line for ^ and lookbehind. */ | |
3b0527fe | 2139 | reginfo.bol = startpos; /* XXX not used ??? */ |
6eb5f6b9 | 2140 | PL_bostr = strbeg; |
3b0527fe | 2141 | reginfo.sv = sv; |
6eb5f6b9 JH |
2142 | |
2143 | /* Mark end of line for $ (and such) */ | |
2144 | PL_regeol = strend; | |
2145 | ||
2146 | /* see how far we have to get to not match where we matched before */ | |
3b0527fe | 2147 | reginfo.till = startpos+minend; |
6eb5f6b9 | 2148 | |
6eb5f6b9 JH |
2149 | /* If there is a "must appear" string, look for it. */ |
2150 | s = startpos; | |
2151 | ||
bbe252da | 2152 | if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ |
6eb5f6b9 | 2153 | MAGIC *mg; |
2c296965 | 2154 | if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */ |
58e23c8d | 2155 | reginfo.ganch = startpos + prog->gofs; |
2c296965 | 2156 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2157 | "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs)); |
2c296965 | 2158 | } else if (sv && SvTYPE(sv) >= SVt_PVMG |
6eb5f6b9 | 2159 | && SvMAGIC(sv) |
14befaf4 DM |
2160 | && (mg = mg_find(sv, PERL_MAGIC_regex_global)) |
2161 | && mg->mg_len >= 0) { | |
3b0527fe | 2162 | reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ |
2c296965 | 2163 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2164 | "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len)); |
2c296965 | 2165 | |
bbe252da | 2166 | if (prog->extflags & RXf_ANCH_GPOS) { |
3b0527fe | 2167 | if (s > reginfo.ganch) |
6eb5f6b9 | 2168 | goto phooey; |
58e23c8d | 2169 | s = reginfo.ganch - prog->gofs; |
2c296965 | 2170 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2171 | "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs)); |
c584a96e YO |
2172 | if (s < strbeg) |
2173 | goto phooey; | |
6eb5f6b9 JH |
2174 | } |
2175 | } | |
58e23c8d | 2176 | else if (data) { |
70685ca0 | 2177 | reginfo.ganch = strbeg + PTR2UV(data); |
2c296965 YO |
2178 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2179 | "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data))); | |
2180 | ||
2181 | } else { /* pos() not defined */ | |
3b0527fe | 2182 | reginfo.ganch = strbeg; |
2c296965 YO |
2183 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2184 | "GPOS: reginfo.ganch = strbeg\n")); | |
2185 | } | |
6eb5f6b9 | 2186 | } |
288b8c02 | 2187 | if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { |
e9105d30 GG |
2188 | /* We have to be careful. If the previous successful match |
2189 | was from this regex we don't want a subsequent partially | |
2190 | successful match to clobber the old results. | |
2191 | So when we detect this possibility we add a swap buffer | |
2192 | to the re, and switch the buffer each match. If we fail | |
2193 | we switch it back, otherwise we leave it swapped. | |
2194 | */ | |
2195 | swap = prog->offs; | |
2196 | /* do we need a save destructor here for eval dies? */ | |
2197 | Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); | |
c74340f9 | 2198 | } |
a0714e2c | 2199 | if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { |
6eb5f6b9 JH |
2200 | re_scream_pos_data d; |
2201 | ||
2202 | d.scream_olds = &scream_olds; | |
2203 | d.scream_pos = &scream_pos; | |
288b8c02 | 2204 | s = re_intuit_start(rx, sv, s, strend, flags, &d); |
3fa9c3d7 | 2205 | if (!s) { |
a3621e74 | 2206 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n")); |
6eb5f6b9 | 2207 | goto phooey; /* not present */ |
3fa9c3d7 | 2208 | } |
6eb5f6b9 JH |
2209 | } |
2210 | ||
1de06328 | 2211 | |
6eb5f6b9 JH |
2212 | |
2213 | /* Simplest case: anchored match need be tried only once. */ | |
2214 | /* [unless only anchor is BOL and multiline is set] */ | |
bbe252da | 2215 | if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { |
24b23f37 | 2216 | if (s == startpos && regtry(®info, &startpos)) |
6eb5f6b9 | 2217 | goto got_it; |
bbe252da YO |
2218 | else if (multiline || (prog->intflags & PREGf_IMPLICIT) |
2219 | || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ | |
6eb5f6b9 JH |
2220 | { |
2221 | char *end; | |
2222 | ||
2223 | if (minlen) | |
2224 | dontbother = minlen - 1; | |
1aa99e6b | 2225 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 2226 | /* for multiline we only have to try after newlines */ |
33b8afdf | 2227 | if (prog->check_substr || prog->check_utf8) { |
92f3d482 YO |
2228 | /* because of the goto we can not easily reuse the macros for bifurcating the |
2229 | unicode/non-unicode match modes here like we do elsewhere - demerphq */ | |
2230 | if (utf8_target) { | |
2231 | if (s == startpos) | |
2232 | goto after_try_utf8; | |
2233 | while (1) { | |
2234 | if (regtry(®info, &s)) { | |
2235 | goto got_it; | |
2236 | } | |
2237 | after_try_utf8: | |
2238 | if (s > end) { | |
2239 | goto phooey; | |
2240 | } | |
2241 | if (prog->extflags & RXf_USE_INTUIT) { | |
2242 | s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL); | |
2243 | if (!s) { | |
2244 | goto phooey; | |
2245 | } | |
2246 | } | |
2247 | else { | |
2248 | s += UTF8SKIP(s); | |
2249 | } | |
2250 | } | |
2251 | } /* end search for check string in unicode */ | |
2252 | else { | |
2253 | if (s == startpos) { | |
2254 | goto after_try_latin; | |
2255 | } | |
2256 | while (1) { | |
2257 | if (regtry(®info, &s)) { | |
2258 | goto got_it; | |
2259 | } | |
2260 | after_try_latin: | |
2261 | if (s > end) { | |
2262 | goto phooey; | |
2263 | } | |
2264 | if (prog->extflags & RXf_USE_INTUIT) { | |
2265 | s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL); | |
2266 | if (!s) { | |
2267 | goto phooey; | |
2268 | } | |
2269 | } | |
2270 | else { | |
2271 | s++; | |
2272 | } | |
2273 | } | |
2274 | } /* end search for check string in latin*/ | |
2275 | } /* end search for check string */ | |
2276 | else { /* search for newline */ | |
2277 | if (s > startpos) { | |
2278 | /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/ | |
6eb5f6b9 | 2279 | s--; |
92f3d482 YO |
2280 | } |
2281 | /* We can use a more efficient search as newlines are the same in unicode as they are in latin */ | |
6eb5f6b9 JH |
2282 | while (s < end) { |
2283 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ | |
24b23f37 | 2284 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2285 | goto got_it; |
2286 | } | |
92f3d482 YO |
2287 | } |
2288 | } /* end search for newline */ | |
2289 | } /* end anchored/multiline check string search */ | |
6eb5f6b9 | 2290 | goto phooey; |
bbe252da | 2291 | } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) |
f9f4320a | 2292 | { |
486ec47a | 2293 | /* the warning about reginfo.ganch being used without initialization |
bbe252da | 2294 | is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN |
f9f4320a | 2295 | and we only enter this block when the same bit is set. */ |
58e23c8d | 2296 | char *tmp_s = reginfo.ganch - prog->gofs; |
c584a96e YO |
2297 | |
2298 | if (tmp_s >= strbeg && regtry(®info, &tmp_s)) | |
6eb5f6b9 JH |
2299 | goto got_it; |
2300 | goto phooey; | |
2301 | } | |
2302 | ||
2303 | /* Messy cases: unanchored match. */ | |
bbe252da | 2304 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { |
6eb5f6b9 | 2305 | /* we have /x+whatever/ */ |
f2ed9b32 | 2306 | /* it must be a one character string (XXXX Except UTF_PATTERN?) */ |
33b8afdf | 2307 | char ch; |
bf93d4cc GS |
2308 | #ifdef DEBUGGING |
2309 | int did_match = 0; | |
2310 | #endif | |
f2ed9b32 KW |
2311 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2312 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2313 | ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0]; | |
bf93d4cc | 2314 | |
f2ed9b32 | 2315 | if (utf8_target) { |
4cadc6a9 | 2316 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2317 | if (*s == ch) { |
a3621e74 | 2318 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2319 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2320 | s += UTF8SKIP(s); |
2321 | while (s < strend && *s == ch) | |
2322 | s += UTF8SKIP(s); | |
2323 | } | |
4cadc6a9 | 2324 | ); |
6eb5f6b9 JH |
2325 | } |
2326 | else { | |
4cadc6a9 | 2327 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2328 | if (*s == ch) { |
a3621e74 | 2329 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2330 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2331 | s++; |
2332 | while (s < strend && *s == ch) | |
2333 | s++; | |
2334 | } | |
4cadc6a9 | 2335 | ); |
6eb5f6b9 | 2336 | } |
a3621e74 | 2337 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 2338 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
2339 | "Did not find anchored character...\n") |
2340 | ); | |
6eb5f6b9 | 2341 | } |
a0714e2c SS |
2342 | else if (prog->anchored_substr != NULL |
2343 | || prog->anchored_utf8 != NULL | |
2344 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
2345 | && prog->float_max_offset < strend - s)) { |
2346 | SV *must; | |
2347 | I32 back_max; | |
2348 | I32 back_min; | |
2349 | char *last; | |
6eb5f6b9 | 2350 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
2351 | #ifdef DEBUGGING |
2352 | int did_match = 0; | |
2353 | #endif | |
33b8afdf | 2354 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 KW |
2355 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2356 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2357 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; | |
33b8afdf JH |
2358 | back_max = back_min = prog->anchored_offset; |
2359 | } else { | |
f2ed9b32 KW |
2360 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2361 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2362 | must = utf8_target ? prog->float_utf8 : prog->float_substr; | |
33b8afdf JH |
2363 | back_max = prog->float_max_offset; |
2364 | back_min = prog->float_min_offset; | |
2365 | } | |
1de06328 YO |
2366 | |
2367 | ||
33b8afdf JH |
2368 | if (must == &PL_sv_undef) |
2369 | /* could not downgrade utf8 check substring, so must fail */ | |
2370 | goto phooey; | |
2371 | ||
1de06328 YO |
2372 | if (back_min<0) { |
2373 | last = strend; | |
2374 | } else { | |
2375 | last = HOP3c(strend, /* Cannot start after this */ | |
2376 | -(I32)(CHR_SVLEN(must) | |
2377 | - (SvTAIL(must) != 0) + back_min), strbeg); | |
2378 | } | |
6eb5f6b9 JH |
2379 | if (s > PL_bostr) |
2380 | last1 = HOPc(s, -1); | |
2381 | else | |
2382 | last1 = s - 1; /* bogus */ | |
2383 | ||
a0288114 | 2384 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 JH |
2385 | check_substr==must. */ |
2386 | scream_pos = -1; | |
2387 | dontbother = end_shift; | |
2388 | strend = HOPc(strend, -dontbother); | |
2389 | while ( (s <= last) && | |
9041c2e3 | 2390 | ((flags & REXEC_SCREAM) |
1de06328 | 2391 | ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg, |
6eb5f6b9 | 2392 | end_shift, &scream_pos, 0)) |
1de06328 | 2393 | : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), |
9041c2e3 | 2394 | (unsigned char*)strend, must, |
7fba1cd6 | 2395 | multiline ? FBMrf_MULTILINE : 0))) ) { |
4addbd3b | 2396 | /* we may be pointing at the wrong string */ |
07bc277f | 2397 | if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog)) |
3f7c398e | 2398 | s = strbeg + (s - SvPVX_const(sv)); |
a3621e74 | 2399 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
2400 | if (HOPc(s, -back_max) > last1) { |
2401 | last1 = HOPc(s, -back_min); | |
2402 | s = HOPc(s, -back_max); | |
2403 | } | |
2404 | else { | |
52657f30 | 2405 | char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; |
6eb5f6b9 JH |
2406 | |
2407 | last1 = HOPc(s, -back_min); | |
52657f30 | 2408 | s = t; |
6eb5f6b9 | 2409 | } |
f2ed9b32 | 2410 | if (utf8_target) { |
6eb5f6b9 | 2411 | while (s <= last1) { |
24b23f37 | 2412 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2413 | goto got_it; |
2414 | s += UTF8SKIP(s); | |
2415 | } | |
2416 | } | |
2417 | else { | |
2418 | while (s <= last1) { | |
24b23f37 | 2419 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2420 | goto got_it; |
2421 | s++; | |
2422 | } | |
2423 | } | |
2424 | } | |
ab3bbdeb | 2425 | DEBUG_EXECUTE_r(if (!did_match) { |
f2ed9b32 | 2426 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
2427 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
2428 | PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", | |
33b8afdf | 2429 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc | 2430 | ? "anchored" : "floating"), |
ab3bbdeb YO |
2431 | quoted, RE_SV_TAIL(must)); |
2432 | }); | |
6eb5f6b9 JH |
2433 | goto phooey; |
2434 | } | |
f8fc2ecf | 2435 | else if ( (c = progi->regstclass) ) { |
f14c76ed | 2436 | if (minlen) { |
f8fc2ecf | 2437 | const OPCODE op = OP(progi->regstclass); |
66e933ab | 2438 | /* don't bother with what can't match */ |
786e8c11 | 2439 | if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) |
f14c76ed RGS |
2440 | strend = HOPc(strend, -(minlen - 1)); |
2441 | } | |
a3621e74 | 2442 | DEBUG_EXECUTE_r({ |
be8e71aa | 2443 | SV * const prop = sv_newmortal(); |
32fc9b6a | 2444 | regprop(prog, prop, c); |
0df25f3d | 2445 | { |
f2ed9b32 | 2446 | RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb | 2447 | s,strend-s,60); |
0df25f3d | 2448 | PerlIO_printf(Perl_debug_log, |
1c8f8eb1 | 2449 | "Matching stclass %.*s against %s (%d bytes)\n", |
e4f74956 | 2450 | (int)SvCUR(prop), SvPVX_const(prop), |
ab3bbdeb | 2451 | quoted, (int)(strend - s)); |
0df25f3d | 2452 | } |
ffc61ed2 | 2453 | }); |
3b0527fe | 2454 | if (find_byclass(prog, c, s, strend, ®info)) |
6eb5f6b9 | 2455 | goto got_it; |
07be1b83 | 2456 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
2457 | } |
2458 | else { | |
2459 | dontbother = 0; | |
a0714e2c | 2460 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 2461 | /* Trim the end. */ |
d6a28714 | 2462 | char *last; |
33b8afdf JH |
2463 | SV* float_real; |
2464 | ||
f2ed9b32 KW |
2465 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2466 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2467 | float_real = utf8_target ? prog->float_utf8 : prog->float_substr; | |
d6a28714 JH |
2468 | |
2469 | if (flags & REXEC_SCREAM) { | |
33b8afdf | 2470 | last = screaminstr(sv, float_real, s - strbeg, |
d6a28714 JH |
2471 | end_shift, &scream_pos, 1); /* last one */ |
2472 | if (!last) | |
ffc61ed2 | 2473 | last = scream_olds; /* Only one occurrence. */ |
4addbd3b | 2474 | /* we may be pointing at the wrong string */ |
07bc277f | 2475 | else if (RXp_MATCH_COPIED(prog)) |
3f7c398e | 2476 | s = strbeg + (s - SvPVX_const(sv)); |
b8c5462f | 2477 | } |
d6a28714 JH |
2478 | else { |
2479 | STRLEN len; | |
cfd0369c | 2480 | const char * const little = SvPV_const(float_real, len); |
d6a28714 | 2481 | |
33b8afdf | 2482 | if (SvTAIL(float_real)) { |
d6a28714 JH |
2483 | if (memEQ(strend - len + 1, little, len - 1)) |
2484 | last = strend - len + 1; | |
7fba1cd6 | 2485 | else if (!multiline) |
9041c2e3 | 2486 | last = memEQ(strend - len, little, len) |
bd61b366 | 2487 | ? strend - len : NULL; |
b8c5462f | 2488 | else |
d6a28714 JH |
2489 | goto find_last; |
2490 | } else { | |
2491 | find_last: | |
9041c2e3 | 2492 | if (len) |
d6a28714 | 2493 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 2494 | else |
a0288114 | 2495 | last = strend; /* matching "$" */ |
b8c5462f | 2496 | } |
b8c5462f | 2497 | } |
bf93d4cc | 2498 | if (last == NULL) { |
6bda09f9 YO |
2499 | DEBUG_EXECUTE_r( |
2500 | PerlIO_printf(Perl_debug_log, | |
2501 | "%sCan't trim the tail, match fails (should not happen)%s\n", | |
2502 | PL_colors[4], PL_colors[5])); | |
bf93d4cc GS |
2503 | goto phooey; /* Should not happen! */ |
2504 | } | |
d6a28714 JH |
2505 | dontbother = strend - last + prog->float_min_offset; |
2506 | } | |
2507 | if (minlen && (dontbother < minlen)) | |
2508 | dontbother = minlen - 1; | |
2509 | strend -= dontbother; /* this one's always in bytes! */ | |
2510 | /* We don't know much -- general case. */ | |
f2ed9b32 | 2511 | if (utf8_target) { |
d6a28714 | 2512 | for (;;) { |
24b23f37 | 2513 | if (regtry(®info, &s)) |
d6a28714 JH |
2514 | goto got_it; |
2515 | if (s >= strend) | |
2516 | break; | |
b8c5462f | 2517 | s += UTF8SKIP(s); |
d6a28714 JH |
2518 | }; |
2519 | } | |
2520 | else { | |
2521 | do { | |
24b23f37 | 2522 | if (regtry(®info, &s)) |
d6a28714 JH |
2523 | goto got_it; |
2524 | } while (s++ < strend); | |
2525 | } | |
2526 | } | |
2527 | ||
2528 | /* Failure. */ | |
2529 | goto phooey; | |
2530 | ||
2531 | got_it: | |
e9105d30 | 2532 | Safefree(swap); |
288b8c02 | 2533 | RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted); |
d6a28714 | 2534 | |
19b95bf0 | 2535 | if (PL_reg_eval_set) |
4f639d21 | 2536 | restore_pos(aTHX_ prog); |
5daac39c NC |
2537 | if (RXp_PAREN_NAMES(prog)) |
2538 | (void)hv_iterinit(RXp_PAREN_NAMES(prog)); | |
d6a28714 JH |
2539 | |
2540 | /* make sure $`, $&, $', and $digit will work later */ | |
2541 | if ( !(flags & REXEC_NOT_FIRST) ) { | |
288b8c02 | 2542 | RX_MATCH_COPY_FREE(rx); |
d6a28714 | 2543 | if (flags & REXEC_COPY_STR) { |
be8e71aa | 2544 | const I32 i = PL_regeol - startpos + (stringarg - strbeg); |
f8c7b90f | 2545 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2546 | if ((SvIsCOW(sv) |
2547 | || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { | |
2548 | if (DEBUG_C_TEST) { | |
2549 | PerlIO_printf(Perl_debug_log, | |
2550 | "Copy on write: regexp capture, type %d\n", | |
2551 | (int) SvTYPE(sv)); | |
2552 | } | |
2553 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); | |
d5263905 | 2554 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
ed252734 NC |
2555 | assert (SvPOKp(prog->saved_copy)); |
2556 | } else | |
2557 | #endif | |
2558 | { | |
288b8c02 | 2559 | RX_MATCH_COPIED_on(rx); |
ed252734 NC |
2560 | s = savepvn(strbeg, i); |
2561 | prog->subbeg = s; | |
2562 | } | |
d6a28714 | 2563 | prog->sublen = i; |
d6a28714 JH |
2564 | } |
2565 | else { | |
2566 | prog->subbeg = strbeg; | |
2567 | prog->sublen = PL_regeol - strbeg; /* strend may have been modified */ | |
2568 | } | |
2569 | } | |
9041c2e3 | 2570 | |
d6a28714 JH |
2571 | return 1; |
2572 | ||
2573 | phooey: | |
a3621e74 | 2574 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 2575 | PL_colors[4], PL_colors[5])); |
d6a28714 | 2576 | if (PL_reg_eval_set) |
4f639d21 | 2577 | restore_pos(aTHX_ prog); |
e9105d30 | 2578 | if (swap) { |
c74340f9 | 2579 | /* we failed :-( roll it back */ |
e9105d30 GG |
2580 | Safefree(prog->offs); |
2581 | prog->offs = swap; | |
2582 | } | |
2583 | ||
d6a28714 JH |
2584 | return 0; |
2585 | } | |
2586 | ||
6bda09f9 | 2587 | |
d6a28714 JH |
2588 | /* |
2589 | - regtry - try match at specific point | |
2590 | */ | |
2591 | STATIC I32 /* 0 failure, 1 success */ | |
24b23f37 | 2592 | S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) |
d6a28714 | 2593 | { |
97aff369 | 2594 | dVAR; |
d6a28714 | 2595 | CHECKPOINT lastcp; |
288b8c02 NC |
2596 | REGEXP *const rx = reginfo->prog; |
2597 | regexp *const prog = (struct regexp *)SvANY(rx); | |
f8fc2ecf | 2598 | RXi_GET_DECL(prog,progi); |
a3621e74 | 2599 | GET_RE_DEBUG_FLAGS_DECL; |
7918f24d NC |
2600 | |
2601 | PERL_ARGS_ASSERT_REGTRY; | |
2602 | ||
24b23f37 | 2603 | reginfo->cutpoint=NULL; |
d6a28714 | 2604 | |
bbe252da | 2605 | if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) { |
d6a28714 JH |
2606 | MAGIC *mg; |
2607 | ||
2608 | PL_reg_eval_set = RS_init; | |
a3621e74 | 2609 | DEBUG_EXECUTE_r(DEBUG_s( |
b900a521 JH |
2610 | PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n", |
2611 | (IV)(PL_stack_sp - PL_stack_base)); | |
d6a28714 | 2612 | )); |
ea8d6ae1 | 2613 | SAVESTACK_CXPOS(); |
d6a28714 JH |
2614 | cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base; |
2615 | /* Otherwise OP_NEXTSTATE will free whatever on stack now. */ | |
2616 | SAVETMPS; | |
2617 | /* Apparently this is not needed, judging by wantarray. */ | |
e8347627 | 2618 | /* SAVEI8(cxstack[cxstack_ix].blk_gimme); |
d6a28714 JH |
2619 | cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ |
2620 | ||
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 | } | |
86c29d75 NC |
2656 | #ifdef USE_ITHREADS |
2657 | /* It seems that non-ithreads works both with and without this code. | |
2658 | So for efficiency reasons it seems best not to have the code | |
2659 | compiled when it is not needed. */ | |
92313705 NC |
2660 | /* This is safe against NULLs: */ |
2661 | ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); | |
2662 | /* PM_reg_curpm owns a reference to this regexp. */ | |
2663 | ReREFCNT_inc(rx); | |
86c29d75 | 2664 | #endif |
288b8c02 | 2665 | PM_SETRE(PL_reg_curpm, rx); |
d6a28714 JH |
2666 | PL_reg_oldcurpm = PL_curpm; |
2667 | PL_curpm = PL_reg_curpm; | |
07bc277f | 2668 | if (RXp_MATCH_COPIED(prog)) { |
d6a28714 JH |
2669 | /* Here is a serious problem: we cannot rewrite subbeg, |
2670 | since it may be needed if this match fails. Thus | |
2671 | $` inside (?{}) could fail... */ | |
2672 | PL_reg_oldsaved = prog->subbeg; | |
2673 | PL_reg_oldsavedlen = prog->sublen; | |
f8c7b90f | 2674 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2675 | PL_nrs = prog->saved_copy; |
2676 | #endif | |
07bc277f | 2677 | RXp_MATCH_COPIED_off(prog); |
d6a28714 JH |
2678 | } |
2679 | else | |
bd61b366 | 2680 | PL_reg_oldsaved = NULL; |
d6a28714 JH |
2681 | prog->subbeg = PL_bostr; |
2682 | prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ | |
2683 | } | |
24b23f37 | 2684 | DEBUG_EXECUTE_r(PL_reg_starttry = *startpos); |
f0ab9afb | 2685 | prog->offs[0].start = *startpos - PL_bostr; |
24b23f37 | 2686 | PL_reginput = *startpos; |
d6a28714 | 2687 | PL_reglastparen = &prog->lastparen; |
a01268b5 | 2688 | PL_reglastcloseparen = &prog->lastcloseparen; |
d6a28714 | 2689 | prog->lastparen = 0; |
03994de8 | 2690 | prog->lastcloseparen = 0; |
d6a28714 | 2691 | PL_regsize = 0; |
f0ab9afb | 2692 | PL_regoffs = prog->offs; |
d6a28714 JH |
2693 | if (PL_reg_start_tmpl <= prog->nparens) { |
2694 | PL_reg_start_tmpl = prog->nparens*3/2 + 3; | |
2695 | if(PL_reg_start_tmp) | |
2696 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
2697 | else | |
a02a5408 | 2698 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); |
d6a28714 JH |
2699 | } |
2700 | ||
2701 | /* XXXX What this code is doing here?!!! There should be no need | |
2702 | to do this again and again, PL_reglastparen should take care of | |
3dd2943c | 2703 | this! --ilya*/ |
dafc8851 JH |
2704 | |
2705 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2706 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
daf18116 | 2707 | * PL_reglastparen), is not needed at all by the test suite |
225593e1 DM |
2708 | * (op/regexp, op/pat, op/split), but that code is needed otherwise |
2709 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ | |
2710 | * Meanwhile, this code *is* needed for the | |
daf18116 JH |
2711 | * above-mentioned test suite tests to succeed. The common theme |
2712 | * on those tests seems to be returning null fields from matches. | |
225593e1 | 2713 | * --jhi updated by dapm */ |
dafc8851 | 2714 | #if 1 |
d6a28714 | 2715 | if (prog->nparens) { |
f0ab9afb | 2716 | regexp_paren_pair *pp = PL_regoffs; |
097eb12c | 2717 | register I32 i; |
eb160463 | 2718 | for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) { |
f0ab9afb NC |
2719 | ++pp; |
2720 | pp->start = -1; | |
2721 | pp->end = -1; | |
d6a28714 JH |
2722 | } |
2723 | } | |
dafc8851 | 2724 | #endif |
02db2b7b | 2725 | REGCP_SET(lastcp); |
f8fc2ecf | 2726 | if (regmatch(reginfo, progi->program + 1)) { |
f0ab9afb | 2727 | PL_regoffs[0].end = PL_reginput - PL_bostr; |
d6a28714 JH |
2728 | return 1; |
2729 | } | |
24b23f37 YO |
2730 | if (reginfo->cutpoint) |
2731 | *startpos= reginfo->cutpoint; | |
02db2b7b | 2732 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
2733 | return 0; |
2734 | } | |
2735 | ||
02db2b7b | 2736 | |
8ba1375e MJD |
2737 | #define sayYES goto yes |
2738 | #define sayNO goto no | |
262b90c4 | 2739 | #define sayNO_SILENT goto no_silent |
8ba1375e | 2740 | |
f9f4320a YO |
2741 | /* we dont use STMT_START/END here because it leads to |
2742 | "unreachable code" warnings, which are bogus, but distracting. */ | |
2743 | #define CACHEsayNO \ | |
c476f425 DM |
2744 | if (ST.cache_mask) \ |
2745 | PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \ | |
f9f4320a | 2746 | sayNO |
3298f257 | 2747 | |
a3621e74 | 2748 | /* this is used to determine how far from the left messages like |
265c4333 YO |
2749 | 'failed...' are printed. It should be set such that messages |
2750 | are inline with the regop output that created them. | |
a3621e74 | 2751 | */ |
265c4333 | 2752 | #define REPORT_CODE_OFF 32 |
a3621e74 YO |
2753 | |
2754 | ||
40a82448 DM |
2755 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
2756 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
9e137952 | 2757 | |
86545054 DM |
2758 | #define SLAB_FIRST(s) (&(s)->states[0]) |
2759 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
2760 | ||
5d9a96ca DM |
2761 | /* grab a new slab and return the first slot in it */ |
2762 | ||
2763 | STATIC regmatch_state * | |
2764 | S_push_slab(pTHX) | |
2765 | { | |
a35a87e7 | 2766 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2767 | dMY_CXT; |
2768 | #endif | |
5d9a96ca DM |
2769 | regmatch_slab *s = PL_regmatch_slab->next; |
2770 | if (!s) { | |
2771 | Newx(s, 1, regmatch_slab); | |
2772 | s->prev = PL_regmatch_slab; | |
2773 | s->next = NULL; | |
2774 | PL_regmatch_slab->next = s; | |
2775 | } | |
2776 | PL_regmatch_slab = s; | |
86545054 | 2777 | return SLAB_FIRST(s); |
5d9a96ca | 2778 | } |
5b47454d | 2779 | |
95b24440 | 2780 | |
40a82448 DM |
2781 | /* push a new state then goto it */ |
2782 | ||
2783 | #define PUSH_STATE_GOTO(state, node) \ | |
2784 | scan = node; \ | |
2785 | st->resume_state = state; \ | |
2786 | goto push_state; | |
2787 | ||
2788 | /* push a new state with success backtracking, then goto it */ | |
2789 | ||
2790 | #define PUSH_YES_STATE_GOTO(state, node) \ | |
2791 | scan = node; \ | |
2792 | st->resume_state = state; \ | |
2793 | goto push_yes_state; | |
2794 | ||
aa283a38 | 2795 | |
aa283a38 | 2796 | |
d6a28714 | 2797 | /* |
95b24440 | 2798 | |
bf1f174e DM |
2799 | regmatch() - main matching routine |
2800 | ||
2801 | This is basically one big switch statement in a loop. We execute an op, | |
2802 | set 'next' to point the next op, and continue. If we come to a point which | |
2803 | we may need to backtrack to on failure such as (A|B|C), we push a | |
2804 | backtrack state onto the backtrack stack. On failure, we pop the top | |
2805 | state, and re-enter the loop at the state indicated. If there are no more | |
2806 | states to pop, we return failure. | |
2807 | ||
2808 | Sometimes we also need to backtrack on success; for example /A+/, where | |
2809 | after successfully matching one A, we need to go back and try to | |
2810 | match another one; similarly for lookahead assertions: if the assertion | |
2811 | completes successfully, we backtrack to the state just before the assertion | |
2812 | and then carry on. In these cases, the pushed state is marked as | |
2813 | 'backtrack on success too'. This marking is in fact done by a chain of | |
2814 | pointers, each pointing to the previous 'yes' state. On success, we pop to | |
2815 | the nearest yes state, discarding any intermediate failure-only states. | |
2816 | Sometimes a yes state is pushed just to force some cleanup code to be | |
2817 | called at the end of a successful match or submatch; e.g. (??{$re}) uses | |
2818 | it to free the inner regex. | |
2819 | ||
2820 | Note that failure backtracking rewinds the cursor position, while | |
2821 | success backtracking leaves it alone. | |
2822 | ||
2823 | A pattern is complete when the END op is executed, while a subpattern | |
2824 | such as (?=foo) is complete when the SUCCESS op is executed. Both of these | |
2825 | ops trigger the "pop to last yes state if any, otherwise return true" | |
2826 | behaviour. | |
2827 | ||
2828 | A common convention in this function is to use A and B to refer to the two | |
2829 | subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is | |
2830 | the subpattern to be matched possibly multiple times, while B is the entire | |
2831 | rest of the pattern. Variable and state names reflect this convention. | |
2832 | ||
2833 | The states in the main switch are the union of ops and failure/success of | |
2834 | substates associated with with that op. For example, IFMATCH is the op | |
2835 | that does lookahead assertions /(?=A)B/ and so the IFMATCH state means | |
2836 | 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just | |
2837 | successfully matched A and IFMATCH_A_fail is a state saying that we have | |
2838 | just failed to match A. Resume states always come in pairs. The backtrack | |
2839 | state we push is marked as 'IFMATCH_A', but when that is popped, we resume | |
2840 | at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking | |
2841 | on success or failure. | |
2842 | ||
2843 | The struct that holds a backtracking state is actually a big union, with | |
2844 | one variant for each major type of op. The variable st points to the | |
2845 | top-most backtrack struct. To make the code clearer, within each | |
2846 | block of code we #define ST to alias the relevant union. | |
2847 | ||
2848 | Here's a concrete example of a (vastly oversimplified) IFMATCH | |
2849 | implementation: | |
2850 | ||
2851 | switch (state) { | |
2852 | .... | |
2853 | ||
2854 | #define ST st->u.ifmatch | |
2855 | ||
2856 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B | |
2857 | ST.foo = ...; // some state we wish to save | |
95b24440 | 2858 | ... |
bf1f174e DM |
2859 | // push a yes backtrack state with a resume value of |
2860 | // IFMATCH_A/IFMATCH_A_fail, then continue execution at the | |
2861 | // first node of A: | |
2862 | PUSH_YES_STATE_GOTO(IFMATCH_A, A); | |
2863 | // NOTREACHED | |
2864 | ||
2865 | case IFMATCH_A: // we have successfully executed A; now continue with B | |
2866 | next = B; | |
2867 | bar = ST.foo; // do something with the preserved value | |
2868 | break; | |
2869 | ||
2870 | case IFMATCH_A_fail: // A failed, so the assertion failed | |
2871 | ...; // do some housekeeping, then ... | |
2872 | sayNO; // propagate the failure | |
2873 | ||
2874 | #undef ST | |
95b24440 | 2875 | |
bf1f174e DM |
2876 | ... |
2877 | } | |
95b24440 | 2878 | |
bf1f174e DM |
2879 | For any old-timers reading this who are familiar with the old recursive |
2880 | approach, the code above is equivalent to: | |
95b24440 | 2881 | |
bf1f174e DM |
2882 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B |
2883 | { | |
2884 | int foo = ... | |
95b24440 | 2885 | ... |
bf1f174e DM |
2886 | if (regmatch(A)) { |
2887 | next = B; | |
2888 | bar = foo; | |
2889 | break; | |
95b24440 | 2890 | } |
bf1f174e DM |
2891 | ...; // do some housekeeping, then ... |
2892 | sayNO; // propagate the failure | |
95b24440 | 2893 | } |
bf1f174e DM |
2894 | |
2895 | The topmost backtrack state, pointed to by st, is usually free. If you | |
2896 | want to claim it, populate any ST.foo fields in it with values you wish to | |
2897 | save, then do one of | |
2898 | ||
2899 | PUSH_STATE_GOTO(resume_state, node); | |
2900 | PUSH_YES_STATE_GOTO(resume_state, node); | |
2901 | ||
2902 | which sets that backtrack state's resume value to 'resume_state', pushes a | |
2903 | new free entry to the top of the backtrack stack, then goes to 'node'. | |
2904 | On backtracking, the free slot is popped, and the saved state becomes the | |
2905 | new free state. An ST.foo field in this new top state can be temporarily | |
2906 | accessed to retrieve values, but once the main loop is re-entered, it | |
2907 | becomes available for reuse. | |
2908 | ||
2909 | Note that the depth of the backtrack stack constantly increases during the | |
2910 | left-to-right execution of the pattern, rather than going up and down with | |
2911 | the pattern nesting. For example the stack is at its maximum at Z at the | |
2912 | end of the pattern, rather than at X in the following: | |
2913 | ||
2914 | /(((X)+)+)+....(Y)+....Z/ | |
2915 | ||
2916 | The only exceptions to this are lookahead/behind assertions and the cut, | |
2917 | (?>A), which pop all the backtrack states associated with A before | |
2918 | continuing. | |
2919 | ||
486ec47a | 2920 | Backtrack state structs are allocated in slabs of about 4K in size. |
bf1f174e DM |
2921 | PL_regmatch_state and st always point to the currently active state, |
2922 | and PL_regmatch_slab points to the slab currently containing | |
2923 | PL_regmatch_state. The first time regmatch() is called, the first slab is | |
2924 | allocated, and is never freed until interpreter destruction. When the slab | |
2925 | is full, a new one is allocated and chained to the end. At exit from | |
2926 | regmatch(), slabs allocated since entry are freed. | |
2927 | ||
2928 | */ | |
95b24440 | 2929 | |
40a82448 | 2930 | |
5bc10b2c | 2931 | #define DEBUG_STATE_pp(pp) \ |
265c4333 | 2932 | DEBUG_STATE_r({ \ |
f2ed9b32 | 2933 | DUMP_EXEC_POS(locinput, scan, utf8_target); \ |
5bc10b2c | 2934 | PerlIO_printf(Perl_debug_log, \ |
5d458dd8 | 2935 | " %*s"pp" %s%s%s%s%s\n", \ |
5bc10b2c | 2936 | depth*2, "", \ |
13d6edb4 | 2937 | PL_reg_name[st->resume_state], \ |
5d458dd8 YO |
2938 | ((st==yes_state||st==mark_state) ? "[" : ""), \ |
2939 | ((st==yes_state) ? "Y" : ""), \ | |
2940 | ((st==mark_state) ? "M" : ""), \ | |
2941 | ((st==yes_state||st==mark_state) ? "]" : "") \ | |
2942 | ); \ | |
265c4333 | 2943 | }); |
5bc10b2c | 2944 | |
40a82448 | 2945 | |
3dab1dad | 2946 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 2947 | |
3df15adc | 2948 | #ifdef DEBUGGING |
5bc10b2c | 2949 | |
ab3bbdeb | 2950 | STATIC void |
f2ed9b32 | 2951 | S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target, |
ab3bbdeb YO |
2952 | const char *start, const char *end, const char *blurb) |
2953 | { | |
efd26800 | 2954 | const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; |
7918f24d NC |
2955 | |
2956 | PERL_ARGS_ASSERT_DEBUG_START_MATCH; | |
2957 | ||
ab3bbdeb YO |
2958 | if (!PL_colorset) |
2959 | reginitcolors(); | |
2960 | { | |
2961 | RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), | |
d2c6dc5e | 2962 | RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); |
ab3bbdeb | 2963 | |
f2ed9b32 | 2964 | RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb YO |
2965 | start, end - start, 60); |
2966 | ||
2967 | PerlIO_printf(Perl_debug_log, | |
2968 | "%s%s REx%s %s against %s\n", | |
2969 | PL_colors[4], blurb, PL_colors[5], s0, s1); | |
2970 | ||
f2ed9b32 | 2971 | if (utf8_target||utf8_pat) |
1de06328 YO |
2972 | PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", |
2973 | utf8_pat ? "pattern" : "", | |
f2ed9b32 KW |
2974 | utf8_pat && utf8_target ? " and " : "", |
2975 | utf8_target ? "string" : "" | |
ab3bbdeb YO |
2976 | ); |
2977 | } | |
2978 | } | |
3df15adc YO |
2979 | |
2980 | STATIC void | |
786e8c11 YO |
2981 | S_dump_exec_pos(pTHX_ const char *locinput, |
2982 | const regnode *scan, | |
2983 | const char *loc_regeol, | |
2984 | const char *loc_bostr, | |
2985 | const char *loc_reg_starttry, | |
f2ed9b32 | 2986 | const bool utf8_target) |
07be1b83 | 2987 | { |
786e8c11 | 2988 | const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; |
07be1b83 | 2989 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ |
786e8c11 | 2990 | int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); |
07be1b83 YO |
2991 | /* The part of the string before starttry has one color |
2992 | (pref0_len chars), between starttry and current | |
2993 | position another one (pref_len - pref0_len chars), | |
2994 | after the current position the third one. | |
2995 | We assume that pref0_len <= pref_len, otherwise we | |
2996 | decrease pref0_len. */ | |
786e8c11 YO |
2997 | int pref_len = (locinput - loc_bostr) > (5 + taill) - l |
2998 | ? (5 + taill) - l : locinput - loc_bostr; | |
07be1b83 YO |
2999 | int pref0_len; |
3000 | ||
7918f24d NC |
3001 | PERL_ARGS_ASSERT_DUMP_EXEC_POS; |
3002 | ||
f2ed9b32 | 3003 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) |
07be1b83 | 3004 | pref_len++; |
786e8c11 YO |
3005 | pref0_len = pref_len - (locinput - loc_reg_starttry); |
3006 | if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) | |
3007 | l = ( loc_regeol - locinput > (5 + taill) - pref_len | |
3008 | ? (5 + taill) - pref_len : loc_regeol - locinput); | |
f2ed9b32 | 3009 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) |
07be1b83 YO |
3010 | l--; |
3011 | if (pref0_len < 0) | |
3012 | pref0_len = 0; | |
3013 | if (pref0_len > pref_len) | |
3014 | pref0_len = pref_len; | |
3015 | { | |
f2ed9b32 | 3016 | const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0; |
0df25f3d | 3017 | |
ab3bbdeb | 3018 | RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), |
1de06328 | 3019 | (locinput - pref_len),pref0_len, 60, 4, 5); |
0df25f3d | 3020 | |
ab3bbdeb | 3021 | RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), |
3df15adc | 3022 | (locinput - pref_len + pref0_len), |
1de06328 | 3023 | pref_len - pref0_len, 60, 2, 3); |
0df25f3d | 3024 | |
ab3bbdeb | 3025 | RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), |
1de06328 | 3026 | locinput, loc_regeol - locinput, 10, 0, 1); |
0df25f3d | 3027 | |
1de06328 | 3028 | const STRLEN tlen=len0+len1+len2; |
3df15adc | 3029 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb | 3030 | "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", |
786e8c11 | 3031 | (IV)(locinput - loc_bostr), |
07be1b83 | 3032 | len0, s0, |
07be1b83 | 3033 | len1, s1, |
07be1b83 | 3034 | (docolor ? "" : "> <"), |
07be1b83 | 3035 | len2, s2, |
f9f4320a | 3036 | (int)(tlen > 19 ? 0 : 19 - tlen), |
07be1b83 YO |
3037 | ""); |
3038 | } | |
3039 | } | |
3df15adc | 3040 | |
07be1b83 YO |
3041 | #endif |
3042 | ||
0a4db386 YO |
3043 | /* reg_check_named_buff_matched() |
3044 | * Checks to see if a named buffer has matched. The data array of | |
3045 | * buffer numbers corresponding to the buffer is expected to reside | |
3046 | * in the regexp->data->data array in the slot stored in the ARG() of | |
3047 | * node involved. Note that this routine doesn't actually care about the | |
3048 | * name, that information is not preserved from compilation to execution. | |
3049 | * Returns the index of the leftmost defined buffer with the given name | |
3050 | * or 0 if non of the buffers matched. | |
3051 | */ | |
3052 | STATIC I32 | |
7918f24d NC |
3053 | S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) |
3054 | { | |
0a4db386 | 3055 | I32 n; |
f8fc2ecf | 3056 | RXi_GET_DECL(rex,rexi); |
ad64d0ec | 3057 | SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
0a4db386 | 3058 | I32 *nums=(I32*)SvPVX(sv_dat); |
7918f24d NC |
3059 | |
3060 | PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; | |
3061 | ||
0a4db386 YO |
3062 | for ( n=0; n<SvIVX(sv_dat); n++ ) { |
3063 | if ((I32)*PL_reglastparen >= nums[n] && | |
f0ab9afb | 3064 | PL_regoffs[nums[n]].end != -1) |
0a4db386 YO |
3065 | { |
3066 | return nums[n]; | |
3067 | } | |
3068 | } | |
3069 | return 0; | |
3070 | } | |
3071 | ||
2f554ef7 DM |
3072 | |
3073 | /* free all slabs above current one - called during LEAVE_SCOPE */ | |
3074 | ||
3075 | STATIC void | |
3076 | S_clear_backtrack_stack(pTHX_ void *p) | |
3077 | { | |
3078 | regmatch_slab *s = PL_regmatch_slab->next; | |
3079 | PERL_UNUSED_ARG(p); | |
3080 | ||
3081 | if (!s) | |
3082 | return; | |
3083 | PL_regmatch_slab->next = NULL; | |
3084 | while (s) { | |
3085 | regmatch_slab * const osl = s; | |
3086 | s = s->next; | |
3087 | Safefree(osl); | |
3088 | } | |
3089 | } | |
3090 | ||
3091 | ||
28d8d7f4 YO |
3092 | #define SETREX(Re1,Re2) \ |
3093 | if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \ | |
3094 | Re1 = (Re2) | |
3095 | ||
d6a28714 | 3096 | STATIC I32 /* 0 failure, 1 success */ |
24b23f37 | 3097 | S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) |
d6a28714 | 3098 | { |
a35a87e7 | 3099 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
3100 | dMY_CXT; |
3101 | #endif | |
27da23d5 | 3102 | dVAR; |
f2ed9b32 | 3103 | register const bool utf8_target = PL_reg_match_utf8; |
4ad0818d | 3104 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
288b8c02 NC |
3105 | REGEXP *rex_sv = reginfo->prog; |
3106 | regexp *rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 3107 | RXi_GET_DECL(rex,rexi); |
2f554ef7 | 3108 | I32 oldsave; |
5d9a96ca DM |
3109 | /* the current state. This is a cached copy of PL_regmatch_state */ |
3110 | register regmatch_state *st; | |
5d9a96ca DM |
3111 | /* cache heavy used fields of st in registers */ |
3112 | register regnode *scan; | |
3113 | register regnode *next; | |
438e9bae | 3114 | register U32 n = 0; /* general value; init to avoid compiler warning */ |
24d3c4a9 | 3115 | register I32 ln = 0; /* len or last; init to avoid compiler warning */ |
5d9a96ca | 3116 | register char *locinput = PL_reginput; |
5d9a96ca | 3117 | register I32 nextchr; /* is always set to UCHARAT(locinput) */ |
24d3c4a9 | 3118 | |
b69b0499 | 3119 | bool result = 0; /* return value of S_regmatch */ |
24d3c4a9 | 3120 | int depth = 0; /* depth of backtrack stack */ |
4b196cd4 YO |
3121 | U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */ |
3122 | const U32 max_nochange_depth = | |
3123 | (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? | |
3124 | 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; | |
77cb431f DM |
3125 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
3126 | subpattern */ | |
e2e6a0f1 YO |
3127 | /* mark_state piggy backs on the yes_state logic so that when we unwind |
3128 | the stack on success we can update the mark_state as we go */ | |
3129 | regmatch_state *mark_state = NULL; /* last mark state we have seen */ | |
faec1544 | 3130 | regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ |
b8591aee | 3131 | struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ |
40a82448 | 3132 | U32 state_num; |
5d458dd8 YO |
3133 | bool no_final = 0; /* prevent failure from backtracking? */ |
3134 | bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ | |
e2e6a0f1 | 3135 | char *startpoint = PL_reginput; |
5d458dd8 YO |
3136 | SV *popmark = NULL; /* are we looking for a mark? */ |
3137 | SV *sv_commit = NULL; /* last mark name seen in failure */ | |
3138 | SV *sv_yes_mark = NULL; /* last mark name we have seen | |
486ec47a | 3139 | during a successful match */ |
5d458dd8 YO |
3140 | U32 lastopen = 0; /* last open we saw */ |
3141 | bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; | |
19b95bf0 | 3142 | SV* const oreplsv = GvSV(PL_replgv); |
24d3c4a9 DM |
3143 | /* these three flags are set by various ops to signal information to |
3144 | * the very next op. They have a useful lifetime of exactly one loop | |
3145 | * iteration, and are not preserved or restored by state pushes/pops | |
3146 | */ | |
3147 | bool sw = 0; /* the condition value in (?(cond)a|b) */ | |
3148 | bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ | |
3149 | int logical = 0; /* the following EVAL is: | |
3150 | 0: (?{...}) | |
3151 | 1: (?(?{...})X|Y) | |
3152 | 2: (??{...}) | |
3153 | or the following IFMATCH/UNLESSM is: | |
3154 | false: plain (?=foo) | |
3155 | true: used as a condition: (?(?=foo)) | |
3156 | */ | |
95b24440 | 3157 | #ifdef DEBUGGING |
e68ec53f | 3158 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
3159 | #endif |
3160 | ||
7918f24d NC |
3161 | PERL_ARGS_ASSERT_REGMATCH; |
3162 | ||
3b57cd43 | 3163 | DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ |
24b23f37 | 3164 | PerlIO_printf(Perl_debug_log,"regmatch start\n"); |
3b57cd43 | 3165 | })); |
5d9a96ca DM |
3166 | /* on first ever call to regmatch, allocate first slab */ |
3167 | if (!PL_regmatch_slab) { | |
3168 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
3169 | PL_regmatch_slab->prev = NULL; | |
3170 | PL_regmatch_slab->next = NULL; | |
86545054 | 3171 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); |
5d9a96ca DM |
3172 | } |
3173 | ||
2f554ef7 DM |
3174 | oldsave = PL_savestack_ix; |
3175 | SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL); | |
3176 | SAVEVPTR(PL_regmatch_slab); | |
3177 | SAVEVPTR(PL_regmatch_state); | |
5d9a96ca DM |
3178 | |
3179 | /* grab next free state slot */ | |
3180 | st = ++PL_regmatch_state; | |
86545054 | 3181 | if (st > SLAB_LAST(PL_regmatch_slab)) |
5d9a96ca DM |
3182 | st = PL_regmatch_state = S_push_slab(aTHX); |
3183 | ||
d6a28714 JH |
3184 | /* Note that nextchr is a byte even in UTF */ |
3185 | nextchr = UCHARAT(locinput); | |
3186 | scan = prog; | |
3187 | while (scan != NULL) { | |
8ba1375e | 3188 | |
a3621e74 | 3189 | DEBUG_EXECUTE_r( { |
6136c704 | 3190 | SV * const prop = sv_newmortal(); |
1de06328 | 3191 | regnode *rnext=regnext(scan); |
f2ed9b32 | 3192 | DUMP_EXEC_POS( locinput, scan, utf8_target ); |
32fc9b6a | 3193 | regprop(rex, prop, scan); |
07be1b83 YO |
3194 | |
3195 | PerlIO_printf(Perl_debug_log, | |
3196 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
f8fc2ecf | 3197 | (IV)(scan - rexi->program), depth*2, "", |
07be1b83 | 3198 | SvPVX_const(prop), |
1de06328 | 3199 | (PL_regkind[OP(scan)] == END || !rnext) ? |
f8fc2ecf | 3200 | 0 : (IV)(rnext - rexi->program)); |
2a782b5b | 3201 | }); |
d6a28714 JH |
3202 | |
3203 | next = scan + NEXT_OFF(scan); | |
3204 | if (next == scan) | |
3205 | next = NULL; | |
40a82448 | 3206 | state_num = OP(scan); |
d6a28714 | 3207 | |
40a82448 | 3208 | reenter_switch: |
34a81e2b B |
3209 | |
3210 | assert(PL_reglastparen == &rex->lastparen); | |
3211 | assert(PL_reglastcloseparen == &rex->lastcloseparen); | |
3212 | assert(PL_regoffs == rex->offs); | |
3213 | ||
40a82448 | 3214 | switch (state_num) { |
d6a28714 | 3215 | case BOL: |
7fba1cd6 | 3216 | if (locinput == PL_bostr) |
d6a28714 | 3217 | { |
3b0527fe | 3218 | /* reginfo->till = reginfo->bol; */ |
b8c5462f JH |
3219 | break; |
3220 | } | |
d6a28714 JH |
3221 | sayNO; |
3222 | case MBOL: | |
12d33761 HS |
3223 | if (locinput == PL_bostr || |
3224 | ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n')) | |
d6a28714 | 3225 | { |
b8c5462f JH |
3226 | break; |
3227 | } | |
d6a28714 JH |
3228 | sayNO; |
3229 | case SBOL: | |
c2a73568 | 3230 | if (locinput == PL_bostr) |
b8c5462f | 3231 | break; |
d6a28714 JH |
3232 | sayNO; |
3233 | case GPOS: | |
3b0527fe | 3234 | if (locinput == reginfo->ganch) |
d6a28714 JH |
3235 | break; |
3236 | sayNO; | |
ee9b8eae YO |
3237 | |
3238 | case KEEPS: | |
3239 | /* update the startpoint */ | |
f0ab9afb | 3240 | st->u.keeper.val = PL_regoffs[0].start; |
ee9b8eae | 3241 | PL_reginput = locinput; |
f0ab9afb | 3242 | PL_regoffs[0].start = locinput - PL_bostr; |
ee9b8eae YO |
3243 | PUSH_STATE_GOTO(KEEPS_next, next); |
3244 | /*NOT-REACHED*/ | |
3245 | case KEEPS_next_fail: | |
3246 | /* rollback the start point change */ | |
f0ab9afb | 3247 | PL_regoffs[0].start = st->u.keeper.val; |
ee9b8eae YO |
3248 | sayNO_SILENT; |
3249 | /*NOT-REACHED*/ | |
d6a28714 | 3250 | case EOL: |
d6a28714 JH |
3251 | goto seol; |
3252 | case MEOL: | |
d6a28714 | 3253 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') |
b8c5462f | 3254 | sayNO; |
b8c5462f | 3255 | break; |
d6a28714 JH |
3256 | case SEOL: |
3257 | seol: | |
3258 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') | |
b8c5462f | 3259 | sayNO; |
d6a28714 | 3260 | if (PL_regeol - locinput > 1) |
b8c5462f | 3261 | sayNO; |
b8c5462f | 3262 | break; |
d6a28714 JH |
3263 | case EOS: |
3264 | if (PL_regeol != locinput) | |
b8c5462f | 3265 | sayNO; |
d6a28714 | 3266 | break; |
ffc61ed2 | 3267 | case SANY: |
d6a28714 | 3268 | if (!nextchr && locinput >= PL_regeol) |
4633a7c4 | 3269 | sayNO; |
f2ed9b32 | 3270 | if (utf8_target) { |
f33976b4 DB |
3271 | locinput += PL_utf8skip[nextchr]; |
3272 | if (locinput > PL_regeol) | |
3273 | sayNO; | |
3274 | nextchr = UCHARAT(locinput); | |
3275 | } | |
3276 | else | |
3277 | nextchr = UCHARAT(++locinput); | |
3278 | break; | |
3279 | case CANY: | |
3280 | if (!nextchr && locinput >= PL_regeol) | |
3281 | sayNO; | |
b8c5462f | 3282 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 3283 | break; |
ffc61ed2 | 3284 | case REG_ANY: |
1aa99e6b IH |
3285 | if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n') |
3286 | sayNO; | |
f2ed9b32 | 3287 | if (utf8_target) { |
b8c5462f | 3288 | locinput += PL_utf8skip[nextchr]; |
d6a28714 JH |
3289 | if (locinput > PL_regeol) |
3290 | sayNO; | |
a0ed51b3 | 3291 | nextchr = UCHARAT(locinput); |
a0ed51b3 | 3292 | } |
1aa99e6b IH |
3293 | else |
3294 | nextchr = UCHARAT(++locinput); | |
a0ed51b3 | 3295 | break; |
166ba7cd DM |
3296 | |
3297 | #undef ST | |
3298 | #define ST st->u.trie | |
786e8c11 YO |
3299 | case TRIEC: |
3300 | /* In this case the charclass data is available inline so | |
3301 | we can fail fast without a lot of extra overhead. | |
3302 | */ | |
f2ed9b32 | 3303 | if (scan->flags == EXACT || !utf8_target) { |
786e8c11 YO |
3304 | if(!ANYOF_BITMAP_TEST(scan, *locinput)) { |
3305 | DEBUG_EXECUTE_r( | |
3306 | PerlIO_printf(Perl_debug_log, | |
3307 | "%*s %sfailed to match trie start class...%s\n", | |
5bc10b2c | 3308 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
786e8c11 YO |
3309 | ); |
3310 | sayNO_SILENT; | |
3311 | /* NOTREACHED */ | |
3312 | } | |
3313 | } | |
3314 | /* FALL THROUGH */ | |
5b47454d | 3315 | case TRIE: |
2e64971a DM |
3316 | /* the basic plan of execution of the trie is: |
3317 | * At the beginning, run though all the states, and | |
3318 | * find the longest-matching word. Also remember the position | |
3319 | * of the shortest matching word. For example, this pattern: | |
3320 | * 1 2 3 4 5 | |
3321 | * ab|a|x|abcd|abc | |
3322 | * when matched against the string "abcde", will generate | |
3323 | * accept states for all words except 3, with the longest | |
3324 | * matching word being 4, and the shortest being 1 (with | |
3325 | * the position being after char 1 of the string). | |
3326 | * | |
3327 | * Then for each matching word, in word order (i.e. 1,2,4,5), | |
3328 | * we run the remainder of the pattern; on each try setting | |
3329 | * the current position to the character following the word, | |
3330 | * returning to try the next word on failure. | |
3331 | * | |
3332 | * We avoid having to build a list of words at runtime by | |
3333 | * using a compile-time structure, wordinfo[].prev, which | |
3334 | * gives, for each word, the previous accepting word (if any). | |
3335 | * In the case above it would contain the mappings 1->2, 2->0, | |
3336 | * 3->0, 4->5, 5->1. We can use this table to generate, from | |
3337 | * the longest word (4 above), a list of all words, by | |
3338 | * following the list of prev pointers; this gives us the | |
3339 | * unordered list 4,5,1,2. Then given the current word we have | |
3340 | * just tried, we can go through the list and find the | |
3341 | * next-biggest word to try (so if we just failed on word 2, | |
3342 | * the next in the list is 4). | |
3343 | * | |
3344 | * Since at runtime we don't record the matching position in | |
3345 | * the string for each word, we have to work that out for | |
3346 | * each word we're about to process. The wordinfo table holds | |
3347 | * the character length of each word; given that we recorded | |
3348 | * at the start: the position of the shortest word and its | |
3349 | * length in chars, we just need to move the pointer the | |
3350 | * difference between the two char lengths. Depending on | |
3351 | * Unicode status and folding, that's cheap or expensive. | |
3352 | * | |
3353 | * This algorithm is optimised for the case where are only a | |
3354 | * small number of accept states, i.e. 0,1, or maybe 2. | |
3355 | * With lots of accepts states, and having to try all of them, | |
3356 | * it becomes quadratic on number of accept states to find all | |
3357 | * the next words. | |
3358 | */ | |
3359 | ||
3dab1dad | 3360 | { |
07be1b83 | 3361 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
a0a388a1 | 3362 | DECL_TRIE_TYPE(scan); |
3dab1dad YO |
3363 | |
3364 | /* what trie are we using right now */ | |
be8e71aa | 3365 | reg_trie_data * const trie |
f8fc2ecf | 3366 | = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; |
85fbaab2 | 3367 | HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); |
3dab1dad | 3368 | U32 state = trie->startstate; |
166ba7cd | 3369 | |
3dab1dad YO |
3370 | if (trie->bitmap && trie_type != trie_utf8_fold && |
3371 | !TRIE_BITMAP_TEST(trie,*locinput) | |
3372 | ) { | |
3373 | if (trie->states[ state ].wordnum) { | |
3374 | DEBUG_EXECUTE_r( | |
3375 | PerlIO_printf(Perl_debug_log, | |
3376 | "%*s %smatched empty string...%s\n", | |
5bc10b2c | 3377 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad | 3378 | ); |
20dbff7c YO |
3379 | if (!trie->jump) |
3380 | break; | |
3dab1dad YO |