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 { \ | |
130 | if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)" "); LEAVE; } } STMT_END | |
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 | ||
182 | ||
c02a0702 KW |
183 | #define _CCC_TRY_AFF_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC) \ |
184 | case NAMEL: \ | |
185 | PL_reg_flags |= RF_tainted; \ | |
186 | /* FALL THROUGH */ \ | |
187 | case NAME: \ | |
188 | if (!nextchr) \ | |
189 | sayNO; \ | |
190 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ | |
191 | if (!CAT2(PL_utf8_,CLASS)) { \ | |
192 | bool ok; \ | |
193 | ENTER; \ | |
194 | save_re_context(); \ | |
195 | ok=CAT2(is_utf8_,CLASS)((const U8*)STR); \ | |
196 | assert(ok); \ | |
197 | LEAVE; \ | |
198 | } \ | |
199 | if (!(OP(scan) == NAME \ | |
f2ed9b32 | 200 | ? cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, utf8_target)) \ |
c02a0702 KW |
201 | : LCFUNC_utf8((U8*)locinput))) \ |
202 | { \ | |
203 | sayNO; \ | |
204 | } \ | |
205 | locinput += PL_utf8skip[nextchr]; \ | |
206 | nextchr = UCHARAT(locinput); \ | |
207 | break; \ | |
208 | } \ | |
a12cf05f | 209 | /* Drops through to the macro that calls this one */ |
c02a0702 KW |
210 | |
211 | #define CCC_TRY_AFF(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC,LCFUNC) \ | |
212 | _CCC_TRY_AFF_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC) \ | |
213 | if (!(OP(scan) == NAME ? FUNC(nextchr) : LCFUNC(nextchr))) \ | |
214 | sayNO; \ | |
215 | nextchr = UCHARAT(++locinput); \ | |
20d0b1e9 YO |
216 | break |
217 | ||
b10ac0d8 KW |
218 | /* Almost identical to the above, but has a case for a node that matches chars |
219 | * between 128 and 255 using Unicode (latin1) semantics. */ | |
220 | #define CCC_TRY_AFF_U(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNCU,LCFUNC) \ | |
c02a0702 | 221 | _CCC_TRY_AFF_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC) \ |
b10ac0d8 | 222 | if (!(OP(scan) == NAMEL ? LCFUNC(nextchr) : (FUNCU(nextchr) && (isASCII(nextchr) || (FLAGS(scan) & USE_UNI))))) \ |
c02a0702 KW |
223 | sayNO; \ |
224 | nextchr = UCHARAT(++locinput); \ | |
b10ac0d8 KW |
225 | break |
226 | ||
c02a0702 KW |
227 | #define _CCC_TRY_NEG_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC) \ |
228 | case NAMEL: \ | |
229 | PL_reg_flags |= RF_tainted; \ | |
230 | /* FALL THROUGH */ \ | |
231 | case NAME : \ | |
232 | if (!nextchr && locinput >= PL_regeol) \ | |
233 | sayNO; \ | |
234 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ | |
235 | if (!CAT2(PL_utf8_,CLASS)) { \ | |
236 | bool ok; \ | |
237 | ENTER; \ | |
238 | save_re_context(); \ | |
239 | ok=CAT2(is_utf8_,CLASS)((const U8*)STR); \ | |
240 | assert(ok); \ | |
241 | LEAVE; \ | |
242 | } \ | |
243 | if ((OP(scan) == NAME \ | |
f2ed9b32 | 244 | ? cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, utf8_target)) \ |
c02a0702 KW |
245 | : LCFUNC_utf8((U8*)locinput))) \ |
246 | { \ | |
247 | sayNO; \ | |
248 | } \ | |
249 | locinput += PL_utf8skip[nextchr]; \ | |
250 | nextchr = UCHARAT(locinput); \ | |
251 | break; \ | |
b10ac0d8 KW |
252 | } |
253 | ||
c02a0702 KW |
254 | #define CCC_TRY_NEG(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC,LCFUNC) \ |
255 | _CCC_TRY_NEG_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC) \ | |
256 | if ((OP(scan) == NAME ? FUNC(nextchr) : LCFUNC(nextchr))) \ | |
257 | sayNO; \ | |
258 | nextchr = UCHARAT(++locinput); \ | |
20d0b1e9 YO |
259 | break |
260 | ||
261 | ||
c02a0702 KW |
262 | #define CCC_TRY_NEG_U(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNCU,LCFUNC) \ |
263 | _CCC_TRY_NEG_COMMON(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNCU) \ | |
b10ac0d8 | 264 | if ((OP(scan) == NAMEL ? LCFUNC(nextchr) : (FUNCU(nextchr) && (isASCII(nextchr) || (FLAGS(scan) & USE_UNI))))) \ |
c02a0702 KW |
265 | sayNO; \ |
266 | nextchr = UCHARAT(++locinput); \ | |
b10ac0d8 | 267 | break |
d1eb3177 YO |
268 | |
269 | ||
270 | ||
3dab1dad YO |
271 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
272 | ||
5f80c4cf | 273 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
c35dcbe2 YO |
274 | /* it would be nice to rework regcomp.sym to generate this stuff. sigh |
275 | * | |
276 | * NOTE that *nothing* that affects backtracking should be in here, specifically | |
277 | * VERBS must NOT be included. JUMPABLE is used to determine if we can ignore a | |
278 | * node that is in between two EXACT like nodes when ascertaining what the required | |
279 | * "follow" character is. This should probably be moved to regex compile time | |
280 | * although it may be done at run time beause of the REF possibility - more | |
281 | * investigation required. -- demerphq | |
282 | */ | |
3e901dc0 YO |
283 | #define JUMPABLE(rn) ( \ |
284 | OP(rn) == OPEN || \ | |
285 | (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \ | |
286 | OP(rn) == EVAL || \ | |
cca55fe3 JP |
287 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
288 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
d1c771f5 | 289 | OP(rn) == KEEPS || \ |
3dab1dad | 290 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 | 291 | ) |
ee9b8eae | 292 | #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT) |
e2d8ce26 | 293 | |
ee9b8eae YO |
294 | #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF ) |
295 | ||
296 | #if 0 | |
297 | /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so | |
298 | we don't need this definition. */ | |
299 | #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) | |
9a5a5549 | 300 | #define IS_TEXTF(rn) ( (OP(rn)==EXACTFU || OP(rn)==EXACTF) || OP(rn)==REFF || OP(rn)==NREFF ) |
ee9b8eae YO |
301 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) |
302 | ||
303 | #else | |
304 | /* ... so we use this as its faster. */ | |
305 | #define IS_TEXT(rn) ( OP(rn)==EXACT ) | |
9a5a5549 | 306 | #define IS_TEXTFU(rn) ( OP(rn)==EXACTFU ) |
ee9b8eae YO |
307 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) |
308 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) | |
309 | ||
310 | #endif | |
e2d8ce26 | 311 | |
a84d97b6 HS |
312 | /* |
313 | Search for mandatory following text node; for lookahead, the text must | |
314 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
315 | */ | |
cca55fe3 | 316 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
317 | while (JUMPABLE(rn)) { \ |
318 | const OPCODE type = OP(rn); \ | |
319 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 320 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 321 | else if (type == PLUS) \ |
cca55fe3 | 322 | rn = NEXTOPER(rn); \ |
3dab1dad | 323 | else if (type == IFMATCH) \ |
a84d97b6 | 324 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 325 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 326 | } \ |
5f80c4cf | 327 | } STMT_END |
74750237 | 328 | |
c476f425 | 329 | |
acfe0abc | 330 | static void restore_pos(pTHX_ void *arg); |
51371543 | 331 | |
620d5b66 NC |
332 | #define REGCP_PAREN_ELEMS 4 |
333 | #define REGCP_OTHER_ELEMS 5 | |
e0fa7e2b | 334 | #define REGCP_FRAME_ELEMS 1 |
620d5b66 NC |
335 | /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and |
336 | * are needed for the regexp context stack bookkeeping. */ | |
337 | ||
76e3520e | 338 | STATIC CHECKPOINT |
cea2e8a9 | 339 | S_regcppush(pTHX_ I32 parenfloor) |
a0d0e21e | 340 | { |
97aff369 | 341 | dVAR; |
a3b680e6 | 342 | const int retval = PL_savestack_ix; |
a3b680e6 | 343 | const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; |
e0fa7e2b NC |
344 | const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS; |
345 | const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT; | |
a0d0e21e | 346 | int p; |
40a82448 | 347 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 348 | |
e49a9654 IH |
349 | if (paren_elems_to_push < 0) |
350 | Perl_croak(aTHX_ "panic: paren_elems_to_push < 0"); | |
351 | ||
e0fa7e2b NC |
352 | if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems) |
353 | Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf | |
5df417d0 JH |
354 | " out of range (%lu-%ld)", |
355 | total_elems, (unsigned long)PL_regsize, (long)parenfloor); | |
e0fa7e2b | 356 | |
620d5b66 | 357 | SSGROW(total_elems + REGCP_FRAME_ELEMS); |
7f69552c | 358 | |
3280af22 | 359 | for (p = PL_regsize; p > parenfloor; p--) { |
b1ce53c5 | 360 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
f0ab9afb NC |
361 | SSPUSHINT(PL_regoffs[p].end); |
362 | SSPUSHINT(PL_regoffs[p].start); | |
3280af22 | 363 | SSPUSHPTR(PL_reg_start_tmp[p]); |
a0d0e21e | 364 | SSPUSHINT(p); |
e7707071 | 365 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
40a82448 | 366 | " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n", |
f0ab9afb | 367 | (UV)p, (IV)PL_regoffs[p].start, |
40a82448 | 368 | (IV)(PL_reg_start_tmp[p] - PL_bostr), |
f0ab9afb | 369 | (IV)PL_regoffs[p].end |
40a82448 | 370 | )); |
a0d0e21e | 371 | } |
b1ce53c5 | 372 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
f0ab9afb | 373 | SSPUSHPTR(PL_regoffs); |
3280af22 NIS |
374 | SSPUSHINT(PL_regsize); |
375 | SSPUSHINT(*PL_reglastparen); | |
a01268b5 | 376 | SSPUSHINT(*PL_reglastcloseparen); |
3280af22 | 377 | SSPUSHPTR(PL_reginput); |
e0fa7e2b | 378 | SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */ |
41123dfd | 379 | |
a0d0e21e LW |
380 | return retval; |
381 | } | |
382 | ||
c277df42 | 383 | /* These are needed since we do not localize EVAL nodes: */ |
ab3bbdeb YO |
384 | #define REGCP_SET(cp) \ |
385 | DEBUG_STATE_r( \ | |
ab3bbdeb | 386 | PerlIO_printf(Perl_debug_log, \ |
e4f74956 | 387 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
ab3bbdeb YO |
388 | (IV)PL_savestack_ix)); \ |
389 | cp = PL_savestack_ix | |
c3464db5 | 390 | |
ab3bbdeb | 391 | #define REGCP_UNWIND(cp) \ |
e4f74956 | 392 | DEBUG_STATE_r( \ |
ab3bbdeb | 393 | if (cp != PL_savestack_ix) \ |
e4f74956 YO |
394 | PerlIO_printf(Perl_debug_log, \ |
395 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ | |
ab3bbdeb YO |
396 | (IV)(cp), (IV)PL_savestack_ix)); \ |
397 | regcpblow(cp) | |
c277df42 | 398 | |
76e3520e | 399 | STATIC char * |
097eb12c | 400 | S_regcppop(pTHX_ const regexp *rex) |
a0d0e21e | 401 | { |
97aff369 | 402 | dVAR; |
e0fa7e2b | 403 | UV i; |
a0d0e21e | 404 | char *input; |
a3621e74 YO |
405 | GET_RE_DEBUG_FLAGS_DECL; |
406 | ||
7918f24d NC |
407 | PERL_ARGS_ASSERT_REGCPPOP; |
408 | ||
b1ce53c5 | 409 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
c6bf6a65 | 410 | i = SSPOPUV; |
e0fa7e2b NC |
411 | assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
412 | i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */ | |
a0d0e21e | 413 | input = (char *) SSPOPPTR; |
a01268b5 | 414 | *PL_reglastcloseparen = SSPOPINT; |
3280af22 NIS |
415 | *PL_reglastparen = SSPOPINT; |
416 | PL_regsize = SSPOPINT; | |
f0ab9afb | 417 | PL_regoffs=(regexp_paren_pair *) SSPOPPTR; |
b1ce53c5 | 418 | |
620d5b66 | 419 | i -= REGCP_OTHER_ELEMS; |
b1ce53c5 | 420 | /* Now restore the parentheses context. */ |
620d5b66 | 421 | for ( ; i > 0; i -= REGCP_PAREN_ELEMS) { |
1df70142 | 422 | I32 tmps; |
097eb12c | 423 | U32 paren = (U32)SSPOPINT; |
3280af22 | 424 | PL_reg_start_tmp[paren] = (char *) SSPOPPTR; |
f0ab9afb | 425 | PL_regoffs[paren].start = SSPOPINT; |
cf93c79d | 426 | tmps = SSPOPINT; |
3280af22 | 427 | if (paren <= *PL_reglastparen) |
f0ab9afb | 428 | PL_regoffs[paren].end = tmps; |
e7707071 | 429 | DEBUG_BUFFERS_r( |
c3464db5 | 430 | PerlIO_printf(Perl_debug_log, |
b900a521 | 431 | " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n", |
f0ab9afb | 432 | (UV)paren, (IV)PL_regoffs[paren].start, |
b900a521 | 433 | (IV)(PL_reg_start_tmp[paren] - PL_bostr), |
f0ab9afb | 434 | (IV)PL_regoffs[paren].end, |
3280af22 | 435 | (paren > *PL_reglastparen ? "(no)" : "")); |
c277df42 | 436 | ); |
a0d0e21e | 437 | } |
e7707071 | 438 | DEBUG_BUFFERS_r( |
bb7a0f54 | 439 | if (*PL_reglastparen + 1 <= rex->nparens) { |
c3464db5 | 440 | PerlIO_printf(Perl_debug_log, |
faccc32b | 441 | " restoring \\%"IVdf"..\\%"IVdf" to undef\n", |
4f639d21 | 442 | (IV)(*PL_reglastparen + 1), (IV)rex->nparens); |
c277df42 IZ |
443 | } |
444 | ); | |
daf18116 | 445 | #if 1 |
dafc8851 JH |
446 | /* It would seem that the similar code in regtry() |
447 | * already takes care of this, and in fact it is in | |
448 | * a better location to since this code can #if 0-ed out | |
449 | * but the code in regtry() is needed or otherwise tests | |
450 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
451 | * (as of patchlevel 7877) will fail. Then again, |
452 | * this code seems to be necessary or otherwise | |
225593e1 DM |
453 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ |
454 | * --jhi updated by dapm */ | |
3b6647e0 | 455 | for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) { |
097eb12c | 456 | if (i > PL_regsize) |
f0ab9afb NC |
457 | PL_regoffs[i].start = -1; |
458 | PL_regoffs[i].end = -1; | |
a0d0e21e | 459 | } |
dafc8851 | 460 | #endif |
a0d0e21e LW |
461 | return input; |
462 | } | |
463 | ||
02db2b7b | 464 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 465 | |
a687059c | 466 | /* |
e50aee73 | 467 | * pregexec and friends |
a687059c LW |
468 | */ |
469 | ||
76234dfb | 470 | #ifndef PERL_IN_XSUB_RE |
a687059c | 471 | /* |
c277df42 | 472 | - pregexec - match a regexp against a string |
a687059c | 473 | */ |
c277df42 | 474 | I32 |
49d7dfbc | 475 | Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend, |
c3464db5 | 476 | char *strbeg, I32 minend, SV *screamer, U32 nosave) |
c277df42 IZ |
477 | /* strend: pointer to null at end of string */ |
478 | /* strbeg: real beginning of string */ | |
479 | /* minend: end of match must be >=minend after stringarg. */ | |
480 | /* nosave: For optimizations. */ | |
481 | { | |
7918f24d NC |
482 | PERL_ARGS_ASSERT_PREGEXEC; |
483 | ||
c277df42 | 484 | return |
9041c2e3 | 485 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
486 | nosave ? 0 : REXEC_COPY_STR); |
487 | } | |
76234dfb | 488 | #endif |
22e551b9 | 489 | |
9041c2e3 | 490 | /* |
cad2e5aa JH |
491 | * Need to implement the following flags for reg_anch: |
492 | * | |
493 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
494 | * USE_INTUIT_ML | |
495 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
496 | * INTUIT_AUTORITATIVE_ML | |
497 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
498 | * INTUIT_ONCE_ML | |
499 | * | |
500 | * Another flag for this function: SECOND_TIME (so that float substrs | |
501 | * with giant delta may be not rechecked). | |
502 | */ | |
503 | ||
504 | /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ | |
505 | ||
3f7c398e | 506 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
507 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
508 | ||
509 | /* XXXX We assume that strpos is strbeg unless sv. */ | |
510 | ||
6eb5f6b9 JH |
511 | /* XXXX Some places assume that there is a fixed substring. |
512 | An update may be needed if optimizer marks as "INTUITable" | |
513 | RExen without fixed substrings. Similarly, it is assumed that | |
514 | lengths of all the strings are no more than minlen, thus they | |
515 | cannot come from lookahead. | |
40d049e4 YO |
516 | (Or minlen should take into account lookahead.) |
517 | NOTE: Some of this comment is not correct. minlen does now take account | |
518 | of lookahead/behind. Further research is required. -- demerphq | |
519 | ||
520 | */ | |
6eb5f6b9 | 521 | |
2c2d71f5 JH |
522 | /* A failure to find a constant substring means that there is no need to make |
523 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
524 | finding a substring too deep into the string means that less calls to | |
30944b6d IZ |
525 | regtry() should be needed. |
526 | ||
527 | REx compiler's optimizer found 4 possible hints: | |
528 | a) Anchored substring; | |
529 | b) Fixed substring; | |
530 | c) Whether we are anchored (beginning-of-line or \G); | |
531 | d) First node (of those at offset 0) which may distingush positions; | |
6eb5f6b9 | 532 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
533 | string which does not contradict any of them. |
534 | */ | |
2c2d71f5 | 535 | |
6eb5f6b9 JH |
536 | /* Most of decisions we do here should have been done at compile time. |
537 | The nodes of the REx which we used for the search should have been | |
538 | deleted from the finite automaton. */ | |
539 | ||
cad2e5aa | 540 | char * |
288b8c02 | 541 | Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, |
9f61653a | 542 | char *strend, const U32 flags, re_scream_pos_data *data) |
cad2e5aa | 543 | { |
97aff369 | 544 | dVAR; |
288b8c02 | 545 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
b7953727 | 546 | register I32 start_shift = 0; |
cad2e5aa | 547 | /* Should be nonnegative! */ |
b7953727 | 548 | register I32 end_shift = 0; |
2c2d71f5 JH |
549 | register char *s; |
550 | register SV *check; | |
a1933d95 | 551 | char *strbeg; |
cad2e5aa | 552 | char *t; |
f2ed9b32 | 553 | const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 554 | I32 ml_anch; |
bd61b366 SS |
555 | register char *other_last = NULL; /* other substr checked before this */ |
556 | char *check_at = NULL; /* check substr found at this pos */ | |
bbe252da | 557 | const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; |
f8fc2ecf | 558 | RXi_GET_DECL(prog,progi); |
30944b6d | 559 | #ifdef DEBUGGING |
890ce7af | 560 | const char * const i_strpos = strpos; |
30944b6d | 561 | #endif |
a3621e74 YO |
562 | GET_RE_DEBUG_FLAGS_DECL; |
563 | ||
7918f24d NC |
564 | PERL_ARGS_ASSERT_RE_INTUIT_START; |
565 | ||
f2ed9b32 | 566 | RX_MATCH_UTF8_set(rx,utf8_target); |
cad2e5aa | 567 | |
3c8556c3 | 568 | if (RX_UTF8(rx)) { |
b8d68ded JH |
569 | PL_reg_flags |= RF_utf8; |
570 | } | |
ab3bbdeb | 571 | DEBUG_EXECUTE_r( |
f2ed9b32 | 572 | debug_start_match(rx, utf8_target, strpos, strend, |
1de06328 YO |
573 | sv ? "Guessing start of match in sv for" |
574 | : "Guessing start of match in string for"); | |
2a782b5b | 575 | ); |
cad2e5aa | 576 | |
c344f387 JH |
577 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
578 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 579 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 580 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 581 | goto fail; |
2c2d71f5 | 582 | } |
1de06328 | 583 | |
a1933d95 | 584 | strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; |
1aa99e6b | 585 | PL_regeol = strend; |
f2ed9b32 | 586 | if (utf8_target) { |
33b8afdf JH |
587 | if (!prog->check_utf8 && prog->check_substr) |
588 | to_utf8_substr(prog); | |
589 | check = prog->check_utf8; | |
590 | } else { | |
591 | if (!prog->check_substr && prog->check_utf8) | |
592 | to_byte_substr(prog); | |
593 | check = prog->check_substr; | |
594 | } | |
1de06328 | 595 | if (check == &PL_sv_undef) { |
a3621e74 | 596 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
1de06328 | 597 | "Non-utf8 string cannot match utf8 check string\n")); |
33b8afdf JH |
598 | goto fail; |
599 | } | |
bbe252da YO |
600 | if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */ |
601 | ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) | |
602 | || ( (prog->extflags & RXf_ANCH_BOL) | |
7fba1cd6 | 603 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 604 | |
7e25d62c | 605 | if (!ml_anch) { |
bbe252da YO |
606 | if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */ |
607 | && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ | |
3f7c398e | 608 | /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ |
7e25d62c JH |
609 | && sv && !SvROK(sv) |
610 | && (strpos != strbeg)) { | |
a3621e74 | 611 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
612 | goto fail; |
613 | } | |
614 | if (prog->check_offset_min == prog->check_offset_max && | |
bbe252da | 615 | !(prog->extflags & RXf_CANY_SEEN)) { |
2c2d71f5 | 616 | /* Substring at constant offset from beg-of-str... */ |
cad2e5aa JH |
617 | I32 slen; |
618 | ||
1aa99e6b | 619 | s = HOP3c(strpos, prog->check_offset_min, strend); |
1de06328 | 620 | |
653099ff GS |
621 | if (SvTAIL(check)) { |
622 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 623 | |
9041c2e3 | 624 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 625 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 626 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 627 | goto fail_finish; |
cad2e5aa JH |
628 | } |
629 | /* Now should match s[0..slen-2] */ | |
630 | slen--; | |
3f7c398e | 631 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 632 | || (slen > 1 |
3f7c398e | 633 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 634 | report_neq: |
a3621e74 | 635 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
636 | goto fail_finish; |
637 | } | |
cad2e5aa | 638 | } |
3f7c398e | 639 | else if (*SvPVX_const(check) != *s |
653099ff | 640 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 641 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 642 | goto report_neq; |
c315bfe8 | 643 | check_at = s; |
2c2d71f5 | 644 | goto success_at_start; |
7e25d62c | 645 | } |
cad2e5aa | 646 | } |
2c2d71f5 | 647 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 648 | s = strpos; |
2c2d71f5 | 649 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
1de06328 YO |
650 | end_shift = prog->check_end_shift; |
651 | ||
2c2d71f5 | 652 | if (!ml_anch) { |
a3b680e6 | 653 | const I32 end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 654 | - (SvTAIL(check) != 0); |
a3b680e6 | 655 | const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
656 | |
657 | if (end_shift < eshift) | |
658 | end_shift = eshift; | |
659 | } | |
cad2e5aa | 660 | } |
2c2d71f5 | 661 | else { /* Can match at random position */ |
cad2e5aa JH |
662 | ml_anch = 0; |
663 | s = strpos; | |
1de06328 YO |
664 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
665 | end_shift = prog->check_end_shift; | |
666 | ||
667 | /* end shift should be non negative here */ | |
cad2e5aa JH |
668 | } |
669 | ||
bcdf7404 | 670 | #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 671 | if (end_shift < 0) |
1de06328 | 672 | Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", |
220fc49f | 673 | (IV)end_shift, RX_PRECOMP(prog)); |
2c2d71f5 JH |
674 | #endif |
675 | ||
2c2d71f5 JH |
676 | restart: |
677 | /* Find a possible match in the region s..strend by looking for | |
678 | the "check" substring in the region corrected by start/end_shift. */ | |
1de06328 YO |
679 | |
680 | { | |
681 | I32 srch_start_shift = start_shift; | |
682 | I32 srch_end_shift = end_shift; | |
683 | if (srch_start_shift < 0 && strbeg - s > srch_start_shift) { | |
684 | srch_end_shift -= ((strbeg - s) - srch_start_shift); | |
685 | srch_start_shift = strbeg - s; | |
686 | } | |
6bda09f9 | 687 | DEBUG_OPTIMISE_MORE_r({ |
1de06328 YO |
688 | PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n", |
689 | (IV)prog->check_offset_min, | |
690 | (IV)srch_start_shift, | |
691 | (IV)srch_end_shift, | |
692 | (IV)prog->check_end_shift); | |
693 | }); | |
694 | ||
cad2e5aa | 695 | if (flags & REXEC_SCREAM) { |
cad2e5aa | 696 | I32 p = -1; /* Internal iterator of scream. */ |
a3b680e6 | 697 | I32 * const pp = data ? data->scream_pos : &p; |
cad2e5aa | 698 | |
2c2d71f5 JH |
699 | if (PL_screamfirst[BmRARE(check)] >= 0 |
700 | || ( BmRARE(check) == '\n' | |
85c508c3 | 701 | && (BmPREVIOUS(check) == SvCUR(check) - 1) |
2c2d71f5 | 702 | && SvTAIL(check) )) |
9041c2e3 | 703 | s = screaminstr(sv, check, |
1de06328 | 704 | srch_start_shift + (s - strbeg), srch_end_shift, pp, 0); |
cad2e5aa | 705 | else |
2c2d71f5 | 706 | goto fail_finish; |
4addbd3b | 707 | /* we may be pointing at the wrong string */ |
07bc277f | 708 | if (s && RXp_MATCH_COPIED(prog)) |
3f7c398e | 709 | s = strbeg + (s - SvPVX_const(sv)); |
cad2e5aa JH |
710 | if (data) |
711 | *data->scream_olds = s; | |
712 | } | |
1de06328 YO |
713 | else { |
714 | U8* start_point; | |
715 | U8* end_point; | |
bbe252da | 716 | if (prog->extflags & RXf_CANY_SEEN) { |
1de06328 YO |
717 | start_point= (U8*)(s + srch_start_shift); |
718 | end_point= (U8*)(strend - srch_end_shift); | |
719 | } else { | |
720 | start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); | |
721 | end_point= HOP3(strend, -srch_end_shift, strbeg); | |
722 | } | |
6bda09f9 | 723 | DEBUG_OPTIMISE_MORE_r({ |
56570a2c | 724 | PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", |
1de06328 | 725 | (int)(end_point - start_point), |
fc8cd66c | 726 | (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), |
1de06328 YO |
727 | start_point); |
728 | }); | |
729 | ||
730 | s = fbm_instr( start_point, end_point, | |
7fba1cd6 | 731 | check, multiline ? FBMrf_MULTILINE : 0); |
1de06328 YO |
732 | } |
733 | } | |
cad2e5aa JH |
734 | /* Update the count-of-usability, remove useless subpatterns, |
735 | unshift s. */ | |
2c2d71f5 | 736 | |
ab3bbdeb | 737 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 738 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
739 | SvPVX_const(check), RE_SV_DUMPLEN(check), 30); |
740 | PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", | |
2c2d71f5 | 741 | (s ? "Found" : "Did not find"), |
f2ed9b32 | 742 | (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) |
ab3bbdeb YO |
743 | ? "anchored" : "floating"), |
744 | quoted, | |
745 | RE_SV_TAIL(check), | |
746 | (s ? " at offset " : "...\n") ); | |
747 | }); | |
2c2d71f5 JH |
748 | |
749 | if (!s) | |
750 | goto fail_finish; | |
2c2d71f5 | 751 | /* Finish the diagnostic message */ |
a3621e74 | 752 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 | 753 | |
1de06328 YO |
754 | /* XXX dmq: first branch is for positive lookbehind... |
755 | Our check string is offset from the beginning of the pattern. | |
756 | So we need to do any stclass tests offset forward from that | |
757 | point. I think. :-( | |
758 | */ | |
759 | ||
760 | ||
761 | ||
762 | check_at=s; | |
763 | ||
764 | ||
2c2d71f5 JH |
765 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. |
766 | Start with the other substr. | |
767 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 768 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
769 | *always* match. Probably should be marked during compile... |
770 | Probably it is right to do no SCREAM here... | |
771 | */ | |
772 | ||
f2ed9b32 | 773 | if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8) |
1de06328 YO |
774 | : (prog->float_substr && prog->anchored_substr)) |
775 | { | |
30944b6d | 776 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
777 | /* XXXX May be hopelessly wrong for UTF... */ |
778 | if (!other_last) | |
6eb5f6b9 | 779 | other_last = strpos; |
f2ed9b32 | 780 | if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
781 | do_other_anchored: |
782 | { | |
890ce7af AL |
783 | char * const last = HOP3c(s, -start_shift, strbeg); |
784 | char *last1, *last2; | |
be8e71aa | 785 | char * const saved_s = s; |
33b8afdf | 786 | SV* must; |
2c2d71f5 | 787 | |
2c2d71f5 JH |
788 | t = s - prog->check_offset_max; |
789 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
f2ed9b32 | 790 | && (!utf8_target |
0ce71af7 | 791 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 792 | && t > strpos))) |
6f207bd3 | 793 | NOOP; |
2c2d71f5 JH |
794 | else |
795 | t = strpos; | |
1aa99e6b | 796 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
797 | if (t < other_last) /* These positions already checked */ |
798 | t = other_last; | |
1aa99e6b | 799 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
800 | if (last < last1) |
801 | last1 = last; | |
1de06328 YO |
802 | /* XXXX It is not documented what units *_offsets are in. |
803 | We assume bytes, but this is clearly wrong. | |
804 | Meaning this code needs to be carefully reviewed for errors. | |
805 | dmq. | |
806 | */ | |
807 | ||
2c2d71f5 | 808 | /* On end-of-str: see comment below. */ |
f2ed9b32 | 809 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; |
33b8afdf JH |
810 | if (must == &PL_sv_undef) { |
811 | s = (char*)NULL; | |
1de06328 | 812 | DEBUG_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
813 | } |
814 | else | |
815 | s = fbm_instr( | |
816 | (unsigned char*)t, | |
817 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
818 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
819 | must, | |
7fba1cd6 | 820 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 821 | ); |
ab3bbdeb | 822 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 823 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
824 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
825 | PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", | |
2c2d71f5 | 826 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
827 | quoted, RE_SV_TAIL(must)); |
828 | }); | |
829 | ||
830 | ||
2c2d71f5 JH |
831 | if (!s) { |
832 | if (last1 >= last2) { | |
a3621e74 | 833 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
834 | ", giving up...\n")); |
835 | goto fail_finish; | |
836 | } | |
a3621e74 | 837 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 838 | ", trying floating at offset %ld...\n", |
be8e71aa | 839 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
840 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
841 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
842 | goto restart; |
843 | } | |
844 | else { | |
a3621e74 | 845 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 846 | (long)(s - i_strpos))); |
1aa99e6b IH |
847 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
848 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 849 | s = saved_s; |
2c2d71f5 JH |
850 | if (t == strpos) |
851 | goto try_at_start; | |
2c2d71f5 JH |
852 | goto try_at_offset; |
853 | } | |
30944b6d | 854 | } |
2c2d71f5 JH |
855 | } |
856 | else { /* Take into account the floating substring. */ | |
33b8afdf | 857 | char *last, *last1; |
be8e71aa | 858 | char * const saved_s = s; |
33b8afdf JH |
859 | SV* must; |
860 | ||
861 | t = HOP3c(s, -start_shift, strbeg); | |
862 | last1 = last = | |
863 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
864 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
865 | last = HOP3c(t, prog->float_max_offset, strend); | |
866 | s = HOP3c(t, prog->float_min_offset, strend); | |
867 | if (s < other_last) | |
868 | s = other_last; | |
2c2d71f5 | 869 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
f2ed9b32 | 870 | must = utf8_target ? prog->float_utf8 : prog->float_substr; |
33b8afdf JH |
871 | /* fbm_instr() takes into account exact value of end-of-str |
872 | if the check is SvTAIL(ed). Since false positives are OK, | |
873 | and end-of-str is not later than strend we are OK. */ | |
874 | if (must == &PL_sv_undef) { | |
875 | s = (char*)NULL; | |
1de06328 | 876 | DEBUG_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
877 | } |
878 | else | |
2c2d71f5 | 879 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
880 | (unsigned char*)last + SvCUR(must) |
881 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 882 | must, multiline ? FBMrf_MULTILINE : 0); |
ab3bbdeb | 883 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 884 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
885 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
886 | PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", | |
33b8afdf | 887 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
888 | quoted, RE_SV_TAIL(must)); |
889 | }); | |
33b8afdf JH |
890 | if (!s) { |
891 | if (last1 == last) { | |
a3621e74 | 892 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
893 | ", giving up...\n")); |
894 | goto fail_finish; | |
2c2d71f5 | 895 | } |
a3621e74 | 896 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 897 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 898 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
899 | other_last = last; |
900 | s = HOP3c(t, 1, strend); | |
901 | goto restart; | |
902 | } | |
903 | else { | |
a3621e74 | 904 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
905 | (long)(s - i_strpos))); |
906 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 907 | s = saved_s; |
33b8afdf JH |
908 | if (t == strpos) |
909 | goto try_at_start; | |
910 | goto try_at_offset; | |
911 | } | |
2c2d71f5 | 912 | } |
cad2e5aa | 913 | } |
2c2d71f5 | 914 | |
1de06328 | 915 | |
9ef43ace | 916 | t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); |
1de06328 | 917 | |
6bda09f9 | 918 | DEBUG_OPTIMISE_MORE_r( |
1de06328 YO |
919 | PerlIO_printf(Perl_debug_log, |
920 | "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", | |
921 | (IV)prog->check_offset_min, | |
922 | (IV)prog->check_offset_max, | |
923 | (IV)(s-strpos), | |
924 | (IV)(t-strpos), | |
925 | (IV)(t-s), | |
926 | (IV)(strend-strpos) | |
927 | ) | |
928 | ); | |
929 | ||
2c2d71f5 | 930 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
f2ed9b32 | 931 | && (!utf8_target |
9ef43ace | 932 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) |
1de06328 YO |
933 | && t > strpos))) |
934 | { | |
2c2d71f5 JH |
935 | /* Fixed substring is found far enough so that the match |
936 | cannot start at strpos. */ | |
937 | try_at_offset: | |
cad2e5aa | 938 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
939 | /* Eventually fbm_*() should handle this, but often |
940 | anchored_offset is not 0, so this check will not be wasted. */ | |
941 | /* XXXX In the code below we prefer to look for "^" even in | |
942 | presence of anchored substrings. And we search even | |
943 | beyond the found float position. These pessimizations | |
944 | are historical artefacts only. */ | |
945 | find_anchor: | |
2c2d71f5 | 946 | while (t < strend - prog->minlen) { |
cad2e5aa | 947 | if (*t == '\n') { |
4ee3650e | 948 | if (t < check_at - prog->check_offset_min) { |
f2ed9b32 | 949 | if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
950 | /* Since we moved from the found position, |
951 | we definitely contradict the found anchored | |
30944b6d IZ |
952 | substr. Due to the above check we do not |
953 | contradict "check" substr. | |
954 | Thus we can arrive here only if check substr | |
955 | is float. Redo checking for "other"=="fixed". | |
956 | */ | |
9041c2e3 | 957 | strpos = t + 1; |
a3621e74 | 958 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 959 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
960 | goto do_other_anchored; |
961 | } | |
4ee3650e GS |
962 | /* We don't contradict the found floating substring. */ |
963 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 964 | s = t + 1; |
a3621e74 | 965 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 966 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
967 | goto set_useful; |
968 | } | |
4ee3650e GS |
969 | /* Position contradicts check-string */ |
970 | /* XXXX probably better to look for check-string | |
971 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 972 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 973 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 974 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
975 | goto restart; |
976 | } | |
977 | t++; | |
978 | } | |
a3621e74 | 979 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 980 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 981 | goto fail_finish; |
cad2e5aa | 982 | } |
f5952150 | 983 | else { |
a3621e74 | 984 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 985 | PL_colors[0], PL_colors[1])); |
f5952150 | 986 | } |
cad2e5aa JH |
987 | s = t; |
988 | set_useful: | |
f2ed9b32 | 989 | ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
990 | } |
991 | else { | |
f5952150 | 992 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 993 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
994 | unless it was an MBOL and we are not after MBOL, |
995 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
996 | try_at_start: |
997 | /* Even in this situation we may use MBOL flag if strpos is offset | |
998 | wrt the start of the string. */ | |
05b4157f | 999 | if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ |
a1933d95 | 1000 | && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d | 1001 | /* May be due to an implicit anchor of m{.*foo} */ |
bbe252da | 1002 | && !(prog->intflags & PREGf_IMPLICIT)) |
d506a20d | 1003 | { |
cad2e5aa JH |
1004 | t = strpos; |
1005 | goto find_anchor; | |
1006 | } | |
a3621e74 | 1007 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 1008 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
70685ca0 | 1009 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 1010 | ); |
2c2d71f5 | 1011 | success_at_start: |
bbe252da | 1012 | if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ |
f2ed9b32 | 1013 | && (utf8_target ? ( |
33b8afdf JH |
1014 | prog->check_utf8 /* Could be deleted already */ |
1015 | && --BmUSEFUL(prog->check_utf8) < 0 | |
1016 | && (prog->check_utf8 == prog->float_utf8) | |
1017 | ) : ( | |
1018 | prog->check_substr /* Could be deleted already */ | |
1019 | && --BmUSEFUL(prog->check_substr) < 0 | |
1020 | && (prog->check_substr == prog->float_substr) | |
1021 | ))) | |
66e933ab | 1022 | { |
cad2e5aa | 1023 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 1024 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
f2ed9b32 KW |
1025 | /* XXX Does the destruction order has to change with utf8_target? */ |
1026 | SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr); | |
1027 | SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
1028 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
1029 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
1030 | check = NULL; /* abort */ | |
cad2e5aa | 1031 | s = strpos; |
c9415951 YO |
1032 | /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevent flag |
1033 | see http://bugs.activestate.com/show_bug.cgi?id=87173 */ | |
1034 | if (prog->intflags & PREGf_IMPLICIT) | |
1035 | prog->extflags &= ~RXf_ANCH_MBOL; | |
3cf5c195 IZ |
1036 | /* XXXX This is a remnant of the old implementation. It |
1037 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 1038 | other heuristics. */ |
bbe252da | 1039 | prog->extflags &= ~RXf_USE_INTUIT; |
c9415951 | 1040 | /* XXXX What other flags might need to be cleared in this branch? */ |
cad2e5aa JH |
1041 | } |
1042 | else | |
1043 | s = strpos; | |
1044 | } | |
1045 | ||
6eb5f6b9 JH |
1046 | /* Last resort... */ |
1047 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
1de06328 YO |
1048 | /* trie stclasses are too expensive to use here, we are better off to |
1049 | leave it to regmatch itself */ | |
f8fc2ecf | 1050 | if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { |
6eb5f6b9 JH |
1051 | /* minlen == 0 is possible if regstclass is \b or \B, |
1052 | and the fixed substr is ''$. | |
1053 | Since minlen is already taken into account, s+1 is before strend; | |
1054 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
1055 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
1056 | regstclass does not come from lookahead... */ | |
1057 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
9a5a5549 | 1058 | This leaves EXACTF, EXACTFU only, which are dealt with in find_byclass(). */ |
f8fc2ecf YO |
1059 | const U8* const str = (U8*)STRING(progi->regstclass); |
1060 | const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT | |
1061 | ? CHR_DIST(str+STR_LEN(progi->regstclass), str) | |
66e933ab | 1062 | : 1); |
1de06328 YO |
1063 | char * endpos; |
1064 | if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) | |
1065 | endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); | |
1066 | else if (prog->float_substr || prog->float_utf8) | |
1067 | endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); | |
1068 | else | |
1069 | endpos= strend; | |
1070 | ||
70685ca0 JH |
1071 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n", |
1072 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg))); | |
1de06328 | 1073 | |
6eb5f6b9 | 1074 | t = s; |
f8fc2ecf | 1075 | s = find_byclass(prog, progi->regstclass, s, endpos, NULL); |
6eb5f6b9 JH |
1076 | if (!s) { |
1077 | #ifdef DEBUGGING | |
cbbf8932 | 1078 | const char *what = NULL; |
6eb5f6b9 JH |
1079 | #endif |
1080 | if (endpos == strend) { | |
a3621e74 | 1081 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
1082 | "Could not match STCLASS...\n") ); |
1083 | goto fail; | |
1084 | } | |
a3621e74 | 1085 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 1086 | "This position contradicts STCLASS...\n") ); |
bbe252da | 1087 | if ((prog->extflags & RXf_ANCH) && !ml_anch) |
653099ff | 1088 | goto fail; |
6eb5f6b9 | 1089 | /* Contradict one of substrings */ |
33b8afdf | 1090 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 | 1091 | if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) { |
a3621e74 | 1092 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 1093 | hop_and_restart: |
1aa99e6b | 1094 | s = HOP3c(t, 1, strend); |
66e933ab GS |
1095 | if (s + start_shift + end_shift > strend) { |
1096 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 1097 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
1098 | "Could not match STCLASS...\n") ); |
1099 | goto fail; | |
1100 | } | |
5e39e1e5 HS |
1101 | if (!check) |
1102 | goto giveup; | |
a3621e74 | 1103 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1104 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1105 | what, (long)(s + start_shift - i_strpos)) ); |
1106 | goto restart; | |
1107 | } | |
66e933ab | 1108 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
1109 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
1110 | goto retry_floating_check; | |
1111 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 1112 | s = check_at; |
5e39e1e5 HS |
1113 | if (!check) |
1114 | goto giveup; | |
a3621e74 | 1115 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1116 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1117 | (long)(other_last - i_strpos)) ); |
1118 | goto do_other_anchored; | |
1119 | } | |
60e71179 GS |
1120 | /* Another way we could have checked stclass at the |
1121 | current position only: */ | |
1122 | if (ml_anch) { | |
1123 | s = t = t + 1; | |
5e39e1e5 HS |
1124 | if (!check) |
1125 | goto giveup; | |
a3621e74 | 1126 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1127 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 1128 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 1129 | goto try_at_offset; |
66e933ab | 1130 | } |
f2ed9b32 | 1131 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 1132 | goto fail; |
6eb5f6b9 JH |
1133 | /* Check is floating subtring. */ |
1134 | retry_floating_check: | |
1135 | t = check_at - start_shift; | |
a3621e74 | 1136 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
1137 | goto hop_and_restart; |
1138 | } | |
b7953727 | 1139 | if (t != s) { |
a3621e74 | 1140 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 1141 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
1142 | (long)(t - i_strpos), (long)(s - i_strpos)) |
1143 | ); | |
1144 | } | |
1145 | else { | |
a3621e74 | 1146 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1147 | "Does not contradict STCLASS...\n"); |
1148 | ); | |
1149 | } | |
6eb5f6b9 | 1150 | } |
5e39e1e5 | 1151 | giveup: |
a3621e74 | 1152 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
1153 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
1154 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 1155 | return s; |
2c2d71f5 JH |
1156 | |
1157 | fail_finish: /* Substring not found */ | |
33b8afdf | 1158 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
f2ed9b32 | 1159 | BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ |
cad2e5aa | 1160 | fail: |
a3621e74 | 1161 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 1162 | PL_colors[4], PL_colors[5])); |
bd61b366 | 1163 | return NULL; |
cad2e5aa | 1164 | } |
9661b544 | 1165 | |
a0a388a1 YO |
1166 | #define DECL_TRIE_TYPE(scan) \ |
1167 | const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \ | |
1168 | trie_type = (scan->flags != EXACT) \ | |
f2ed9b32 KW |
1169 | ? (utf8_target ? trie_utf8_fold : (UTF_PATTERN ? trie_latin_utf8_fold : trie_plain)) \ |
1170 | : (utf8_target ? trie_utf8 : trie_plain) | |
3b0527fe | 1171 | |
55eed653 NC |
1172 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ |
1173 | uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ | |
4cadc6a9 YO |
1174 | switch (trie_type) { \ |
1175 | case trie_utf8_fold: \ | |
1176 | if ( foldlen>0 ) { \ | |
0abd0d78 | 1177 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ |
4cadc6a9 YO |
1178 | foldlen -= len; \ |
1179 | uscan += len; \ | |
1180 | len=0; \ | |
1181 | } else { \ | |
0abd0d78 | 1182 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ |
4cadc6a9 YO |
1183 | uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \ |
1184 | foldlen -= UNISKIP( uvc ); \ | |
1185 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1186 | } \ | |
1187 | break; \ | |
a0a388a1 YO |
1188 | case trie_latin_utf8_fold: \ |
1189 | if ( foldlen>0 ) { \ | |
1190 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1191 | foldlen -= len; \ | |
1192 | uscan += len; \ | |
1193 | len=0; \ | |
1194 | } else { \ | |
1195 | len = 1; \ | |
1196 | uvc = to_uni_fold( *(U8*)uc, foldbuf, &foldlen ); \ | |
1197 | foldlen -= UNISKIP( uvc ); \ | |
1198 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1199 | } \ | |
1200 | break; \ | |
4cadc6a9 YO |
1201 | case trie_utf8: \ |
1202 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ | |
1203 | break; \ | |
1204 | case trie_plain: \ | |
1205 | uvc = (UV)*uc; \ | |
1206 | len = 1; \ | |
1207 | } \ | |
4cadc6a9 YO |
1208 | if (uvc < 256) { \ |
1209 | charid = trie->charmap[ uvc ]; \ | |
1210 | } \ | |
1211 | else { \ | |
1212 | charid = 0; \ | |
55eed653 NC |
1213 | if (widecharmap) { \ |
1214 | SV** const svpp = hv_fetch(widecharmap, \ | |
4cadc6a9 YO |
1215 | (char*)&uvc, sizeof(UV), 0); \ |
1216 | if (svpp) \ | |
1217 | charid = (U16)SvIV(*svpp); \ | |
1218 | } \ | |
1219 | } \ | |
1220 | } STMT_END | |
1221 | ||
a0a388a1 YO |
1222 | #define REXEC_FBC_EXACTISH_CHECK(CoNd) \ |
1223 | { \ | |
1224 | char *my_strend= (char *)strend; \ | |
4cadc6a9 YO |
1225 | if ( (CoNd) \ |
1226 | && (ln == len || \ | |
f2ed9b32 KW |
1227 | foldEQ_utf8(s, &my_strend, 0, utf8_target, \ |
1228 | m, NULL, ln, cBOOL(UTF_PATTERN))) \ | |
a0a388a1 | 1229 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1230 | goto got_it; \ |
1231 | else { \ | |
1232 | U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \ | |
1233 | uvchr_to_utf8(tmpbuf, c); \ | |
1234 | f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \ | |
1235 | if ( f != c \ | |
1236 | && (f == c1 || f == c2) \ | |
a0a388a1 | 1237 | && (ln == len || \ |
f2ed9b32 KW |
1238 | foldEQ_utf8(s, &my_strend, 0, utf8_target,\ |
1239 | m, NULL, ln, cBOOL(UTF_PATTERN)))\ | |
a0a388a1 | 1240 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1241 | goto got_it; \ |
1242 | } \ | |
a0a388a1 YO |
1243 | } \ |
1244 | s += len | |
4cadc6a9 YO |
1245 | |
1246 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ | |
1247 | STMT_START { \ | |
a932d541 | 1248 | re_fold_t folder; \ |
9a5a5549 KW |
1249 | switch (OP(c)) { \ |
1250 | case EXACTFU: folder = foldEQ_latin1; break; \ | |
1251 | case EXACTFL: folder = foldEQ_locale; break; \ | |
1252 | case EXACTF: folder = foldEQ; break; \ | |
1253 | default: \ | |
1254 | Perl_croak(aTHX_ "panic: Unexpected op %u", OP(c)); \ | |
1255 | } \ | |
4cadc6a9 YO |
1256 | while (s <= e) { \ |
1257 | if ( (CoNd) \ | |
9a5a5549 KW |
1258 | && (ln == 1 || folder(s, m, ln)) \ |
1259 | && (!reginfo || regtry(reginfo, &s)) ) \ | |
4cadc6a9 YO |
1260 | goto got_it; \ |
1261 | s++; \ | |
1262 | } \ | |
1263 | } STMT_END | |
1264 | ||
1265 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
1266 | STMT_START { \ | |
1267 | while (s + (uskip = UTF8SKIP(s)) <= strend) { \ | |
1268 | CoDe \ | |
1269 | s += uskip; \ | |
1270 | } \ | |
1271 | } STMT_END | |
1272 | ||
1273 | #define REXEC_FBC_SCAN(CoDe) \ | |
1274 | STMT_START { \ | |
1275 | while (s < strend) { \ | |
1276 | CoDe \ | |
1277 | s++; \ | |
1278 | } \ | |
1279 | } STMT_END | |
1280 | ||
1281 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1282 | REXEC_FBC_UTF8_SCAN( \ | |
1283 | if (CoNd) { \ | |
24b23f37 | 1284 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1285 | goto got_it; \ |
1286 | else \ | |
1287 | tmp = doevery; \ | |
1288 | } \ | |
1289 | else \ | |
1290 | tmp = 1; \ | |
1291 | ) | |
1292 | ||
1293 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1294 | REXEC_FBC_SCAN( \ | |
1295 | if (CoNd) { \ | |
24b23f37 | 1296 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1297 | goto got_it; \ |
1298 | else \ | |
1299 | tmp = doevery; \ | |
1300 | } \ | |
1301 | else \ | |
1302 | tmp = 1; \ | |
1303 | ) | |
1304 | ||
1305 | #define REXEC_FBC_TRYIT \ | |
24b23f37 | 1306 | if ((!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1307 | goto got_it |
1308 | ||
e1d1eefb | 1309 | #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \ |
f2ed9b32 | 1310 | if (utf8_target) { \ |
e1d1eefb YO |
1311 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1312 | } \ | |
1313 | else { \ | |
1314 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
1315 | } \ | |
1316 | break | |
1317 | ||
4cadc6a9 | 1318 | #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ |
f2ed9b32 | 1319 | if (utf8_target) { \ |
4cadc6a9 YO |
1320 | UtFpReLoAd; \ |
1321 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1322 | } \ | |
1323 | else { \ | |
1324 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
1325 | } \ | |
1326 | break | |
1327 | ||
1328 | #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ | |
1329 | PL_reg_flags |= RF_tainted; \ | |
f2ed9b32 | 1330 | if (utf8_target) { \ |
4cadc6a9 YO |
1331 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1332 | } \ | |
1333 | else { \ | |
1334 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
1335 | } \ | |
1336 | break | |
1337 | ||
786e8c11 YO |
1338 | #define DUMP_EXEC_POS(li,s,doutf8) \ |
1339 | dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8) | |
1340 | ||
1341 | /* We know what class REx starts with. Try to find this position... */ | |
1342 | /* if reginfo is NULL, its a dryrun */ | |
1343 | /* annoyingly all the vars in this routine have different names from their counterparts | |
1344 | in regmatch. /grrr */ | |
1345 | ||
3c3eec57 | 1346 | STATIC char * |
07be1b83 | 1347 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
24b23f37 | 1348 | const char *strend, regmatch_info *reginfo) |
a687059c | 1349 | { |
27da23d5 | 1350 | dVAR; |
bbe252da | 1351 | const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; |
6eb5f6b9 | 1352 | char *m; |
d8093b23 | 1353 | STRLEN ln; |
5dab1207 | 1354 | STRLEN lnc; |
078c425b | 1355 | register STRLEN uskip; |
d8093b23 G |
1356 | unsigned int c1; |
1357 | unsigned int c2; | |
6eb5f6b9 JH |
1358 | char *e; |
1359 | register I32 tmp = 1; /* Scratch variable? */ | |
f2ed9b32 | 1360 | register const bool utf8_target = PL_reg_match_utf8; |
f8fc2ecf | 1361 | RXi_GET_DECL(prog,progi); |
7918f24d NC |
1362 | |
1363 | PERL_ARGS_ASSERT_FIND_BYCLASS; | |
f8fc2ecf | 1364 | |
6eb5f6b9 JH |
1365 | /* We know what class it must start with. */ |
1366 | switch (OP(c)) { | |
6eb5f6b9 | 1367 | case ANYOF: |
f2ed9b32 | 1368 | if (utf8_target) { |
3ff7ceb3 | 1369 | REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_NONBITMAP) || |
388cc4de | 1370 | !UTF8_IS_INVARIANT((U8)s[0]) ? |
f2ed9b32 | 1371 | reginclass(prog, c, (U8*)s, 0, utf8_target) : |
4cadc6a9 | 1372 | REGINCLASS(prog, c, (U8*)s)); |
388cc4de HS |
1373 | } |
1374 | else { | |
1375 | while (s < strend) { | |
1376 | STRLEN skip = 1; | |
1377 | ||
32fc9b6a | 1378 | if (REGINCLASS(prog, c, (U8*)s) || |
388cc4de HS |
1379 | (ANYOF_FOLD_SHARP_S(c, s, strend) && |
1380 | /* The assignment of 2 is intentional: | |
1381 | * for the folded sharp s, the skip is 2. */ | |
1382 | (skip = SHARP_S_SKIP))) { | |
24b23f37 | 1383 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
388cc4de HS |
1384 | goto got_it; |
1385 | else | |
1386 | tmp = doevery; | |
1387 | } | |
1388 | else | |
1389 | tmp = 1; | |
1390 | s += skip; | |
1391 | } | |
a0d0e21e | 1392 | } |
6eb5f6b9 | 1393 | break; |
f33976b4 | 1394 | case CANY: |
4cadc6a9 | 1395 | REXEC_FBC_SCAN( |
24b23f37 | 1396 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
f33976b4 DB |
1397 | goto got_it; |
1398 | else | |
1399 | tmp = doevery; | |
4cadc6a9 | 1400 | ); |
f33976b4 | 1401 | break; |
9a5a5549 | 1402 | case EXACTFU: |
6eb5f6b9 | 1403 | case EXACTF: |
5dab1207 NIS |
1404 | m = STRING(c); |
1405 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1406 | lnc = (I32) ln; /* length to match in characters */ | |
f2ed9b32 | 1407 | if (UTF_PATTERN) { |
a2a2844f | 1408 | STRLEN ulen1, ulen2; |
5dab1207 | 1409 | U8 *sm = (U8 *) m; |
89ebb4a3 JH |
1410 | U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; |
1411 | U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; | |
97dc7d3e RGS |
1412 | /* used by commented-out code below */ |
1413 | /*const U32 uniflags = UTF8_ALLOW_DEFAULT;*/ | |
a0a388a1 YO |
1414 | |
1415 | /* XXX: Since the node will be case folded at compile | |
1416 | time this logic is a little odd, although im not | |
1417 | sure that its actually wrong. --dmq */ | |
1418 | ||
1419 | c1 = to_utf8_lower((U8*)m, tmpbuf1, &ulen1); | |
1420 | c2 = to_utf8_upper((U8*)m, tmpbuf2, &ulen2); | |
1421 | ||
1422 | /* XXX: This is kinda strange. to_utf8_XYZ returns the | |
1423 | codepoint of the first character in the converted | |
1424 | form, yet originally we did the extra step. | |
1425 | No tests fail by commenting this code out however | |
1426 | so Ive left it out. -- dmq. | |
1427 | ||
89ebb4a3 | 1428 | c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE, |
041457d9 | 1429 | 0, uniflags); |
89ebb4a3 | 1430 | c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE, |
041457d9 | 1431 | 0, uniflags); |
a0a388a1 YO |
1432 | */ |
1433 | ||
5dab1207 NIS |
1434 | lnc = 0; |
1435 | while (sm < ((U8 *) m + ln)) { | |
1436 | lnc++; | |
1437 | sm += UTF8SKIP(sm); | |
1438 | } | |
1aa99e6b IH |
1439 | } |
1440 | else { | |
1441 | c1 = *(U8*)m; | |
9a5a5549 KW |
1442 | if (utf8_target || OP(c) == EXACTFU) { |
1443 | ||
1444 | /* Micro sign folds to GREEK SMALL LETTER MU; | |
1445 | LATIN_SMALL_LETTER_SHARP_S folds to 'ss', and this sets | |
1446 | c2 to the first 's' of the pair, and the code below will | |
1447 | look for others */ | |
1448 | c2 = (c1 == MICRO_SIGN) | |
1449 | ? GREEK_SMALL_LETTER_MU | |
1450 | : (c1 == LATIN_SMALL_LETTER_SHARP_S) | |
1451 | ? 's' | |
1452 | : PL_fold_latin1[c1]; | |
1453 | } else c2 = PL_fold[c1]; | |
1aa99e6b | 1454 | } |
6eb5f6b9 JH |
1455 | goto do_exactf; |
1456 | case EXACTFL: | |
5dab1207 NIS |
1457 | m = STRING(c); |
1458 | ln = STR_LEN(c); | |
1459 | lnc = (I32) ln; | |
d8093b23 | 1460 | c1 = *(U8*)m; |
6eb5f6b9 JH |
1461 | c2 = PL_fold_locale[c1]; |
1462 | do_exactf: | |
db12adc6 | 1463 | e = HOP3c(strend, -((I32)lnc), s); |
b3c9acc1 | 1464 | |
3b0527fe | 1465 | if (!reginfo && e < s) |
6eb5f6b9 | 1466 | e = s; /* Due to minlen logic of intuit() */ |
1aa99e6b | 1467 | |
60a8b682 JH |
1468 | /* The idea in the EXACTF* cases is to first find the |
1469 | * first character of the EXACTF* node and then, if | |
1470 | * necessary, case-insensitively compare the full | |
1471 | * text of the node. The c1 and c2 are the first | |
1472 | * characters (though in Unicode it gets a bit | |
1473 | * more complicated because there are more cases | |
7f16dd3d JH |
1474 | * than just upper and lower: one needs to use |
1475 | * the so-called folding case for case-insensitive | |
1476 | * matching (called "loose matching" in Unicode). | |
4c1b470c | 1477 | * foldEQ_utf8() will do just that. */ |
60a8b682 | 1478 | |
f2ed9b32 | 1479 | if (utf8_target || UTF_PATTERN) { |
575cac57 | 1480 | UV c, f; |
89ebb4a3 | 1481 | U8 tmpbuf [UTF8_MAXBYTES+1]; |
a0a388a1 YO |
1482 | STRLEN len = 1; |
1483 | STRLEN foldlen; | |
4ad0818d | 1484 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
09091399 | 1485 | if (c1 == c2) { |
5dab1207 NIS |
1486 | /* Upper and lower of 1st char are equal - |
1487 | * probably not a "letter". */ | |
1aa99e6b | 1488 | while (s <= e) { |
f2ed9b32 | 1489 | if (utf8_target) { |
a0a388a1 | 1490 | c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len, |
041457d9 | 1491 | uniflags); |
a0a388a1 YO |
1492 | } else { |
1493 | c = *((U8*)s); | |
1494 | } | |
4cadc6a9 | 1495 | REXEC_FBC_EXACTISH_CHECK(c == c1); |
1aa99e6b | 1496 | } |
09091399 JH |
1497 | } |
1498 | else { | |
1aa99e6b | 1499 | while (s <= e) { |
f2ed9b32 | 1500 | if (utf8_target) { |
a0a388a1 | 1501 | c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len, |
041457d9 | 1502 | uniflags); |
a0a388a1 YO |
1503 | } else { |
1504 | c = *((U8*)s); | |
1505 | } | |
80aecb99 | 1506 | |
60a8b682 | 1507 | /* Handle some of the three Greek sigmas cases. |
8c01da3c JH |
1508 | * Note that not all the possible combinations |
1509 | * are handled here: some of them are handled | |
1510 | * by the standard folding rules, and some of | |
1511 | * them (the character class or ANYOF cases) | |
1512 | * are handled during compiletime in | |
1513 | * regexec.c:S_regclass(). */ | |
880bd946 JH |
1514 | if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA || |
1515 | c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA) | |
1516 | c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA; | |
80aecb99 | 1517 | |
4cadc6a9 | 1518 | REXEC_FBC_EXACTISH_CHECK(c == c1 || c == c2); |
1aa99e6b | 1519 | } |
09091399 | 1520 | } |
1aa99e6b IH |
1521 | } |
1522 | else { | |
a0a388a1 | 1523 | /* Neither pattern nor string are UTF8 */ |
1aa99e6b | 1524 | if (c1 == c2) |
4cadc6a9 | 1525 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); |
1aa99e6b | 1526 | else |
4cadc6a9 | 1527 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); |
b3c9acc1 IZ |
1528 | } |
1529 | break; | |
bbce6d69 | 1530 | case BOUNDL: |
3280af22 | 1531 | PL_reg_flags |= RF_tainted; |
bbce6d69 | 1532 | /* FALL THROUGH */ |
a0d0e21e | 1533 | case BOUND: |
f2ed9b32 | 1534 | if (utf8_target) { |
12d33761 | 1535 | if (s == PL_bostr) |
ffc61ed2 JH |
1536 | tmp = '\n'; |
1537 | else { | |
6136c704 | 1538 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); |
4ad0818d | 1539 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); |
ffc61ed2 JH |
1540 | } |
1541 | tmp = ((OP(c) == BOUND ? | |
9041c2e3 | 1542 | isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0); |
1a4fad37 | 1543 | LOAD_UTF8_CHARCLASS_ALNUM(); |
4cadc6a9 | 1544 | REXEC_FBC_UTF8_SCAN( |
ffc61ed2 | 1545 | if (tmp == !(OP(c) == BOUND ? |
f2ed9b32 | 1546 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)) : |
ffc61ed2 JH |
1547 | isALNUM_LC_utf8((U8*)s))) |
1548 | { | |
1549 | tmp = !tmp; | |
4cadc6a9 | 1550 | REXEC_FBC_TRYIT; |
a687059c | 1551 | } |
4cadc6a9 | 1552 | ); |
a0d0e21e | 1553 | } |
a12cf05f | 1554 | else { /* Not utf8 */ |
12d33761 | 1555 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; |
a12cf05f KW |
1556 | tmp = cBOOL((OP(c) == BOUNDL) |
1557 | ? isALNUM_LC(tmp) | |
1558 | : (isWORDCHAR_L1(tmp) | |
1559 | && (isASCII(tmp) || (FLAGS(c) & USE_UNI)))); | |
4cadc6a9 | 1560 | REXEC_FBC_SCAN( |
ffc61ed2 | 1561 | if (tmp == |
a12cf05f KW |
1562 | !((OP(c) == BOUNDL) |
1563 | ? isALNUM_LC(*s) | |
1564 | : (isWORDCHAR_L1((U8) *s) | |
1565 | && (isASCII((U8) *s) || (FLAGS(c) & USE_UNI))))) | |
1566 | { | |
ffc61ed2 | 1567 | tmp = !tmp; |
4cadc6a9 | 1568 | REXEC_FBC_TRYIT; |
a0ed51b3 | 1569 | } |
4cadc6a9 | 1570 | ); |
a0ed51b3 | 1571 | } |
24b23f37 | 1572 | if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) |
a0ed51b3 LW |
1573 | goto got_it; |
1574 | break; | |
bbce6d69 | 1575 | case NBOUNDL: |
3280af22 | 1576 | PL_reg_flags |= RF_tainted; |
bbce6d69 | 1577 | /* FALL THROUGH */ |
a0d0e21e | 1578 | case NBOUND: |
f2ed9b32 | 1579 | if (utf8_target) { |
12d33761 | 1580 | if (s == PL_bostr) |
ffc61ed2 JH |
1581 | tmp = '\n'; |
1582 | else { | |
6136c704 | 1583 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); |
4ad0818d | 1584 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); |
ffc61ed2 JH |
1585 | } |
1586 | tmp = ((OP(c) == NBOUND ? | |
9041c2e3 | 1587 | isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0); |
1a4fad37 | 1588 | LOAD_UTF8_CHARCLASS_ALNUM(); |
4cadc6a9 | 1589 | REXEC_FBC_UTF8_SCAN( |
ffc61ed2 | 1590 | if (tmp == !(OP(c) == NBOUND ? |
f2ed9b32 | 1591 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)) : |
ffc61ed2 JH |
1592 | isALNUM_LC_utf8((U8*)s))) |
1593 | tmp = !tmp; | |
4cadc6a9 YO |
1594 | else REXEC_FBC_TRYIT; |
1595 | ); | |
a0d0e21e | 1596 | } |
667bb95a | 1597 | else { |
12d33761 | 1598 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; |
a12cf05f KW |
1599 | tmp = cBOOL((OP(c) == NBOUNDL) |
1600 | ? isALNUM_LC(tmp) | |
1601 | : (isWORDCHAR_L1(tmp) | |
1602 | && (isASCII(tmp) || (FLAGS(c) & USE_UNI)))); | |
4cadc6a9 | 1603 | REXEC_FBC_SCAN( |
a12cf05f KW |
1604 | if (tmp == ! cBOOL( |
1605 | (OP(c) == NBOUNDL) | |
1606 | ? isALNUM_LC(*s) | |
1607 | : (isWORDCHAR_L1((U8) *s) | |
1608 | && (isASCII((U8) *s) || (FLAGS(c) & USE_UNI))))) | |
1609 | { | |
ffc61ed2 | 1610 | tmp = !tmp; |
a12cf05f | 1611 | } |
4cadc6a9 YO |
1612 | else REXEC_FBC_TRYIT; |
1613 | ); | |
a0ed51b3 | 1614 | } |
24b23f37 | 1615 | if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, &s))) |
a0ed51b3 LW |
1616 | goto got_it; |
1617 | break; | |
a0d0e21e | 1618 | case ALNUM: |
4cadc6a9 | 1619 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1620 | LOAD_UTF8_CHARCLASS_PERL_WORD(), |
f2ed9b32 | 1621 | swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target), |
a12cf05f | 1622 | (FLAGS(c) & USE_UNI) ? isWORDCHAR_L1((U8) *s) : isALNUM(*s) |
4cadc6a9 | 1623 | ); |
bbce6d69 | 1624 | case ALNUML: |
4cadc6a9 YO |
1625 | REXEC_FBC_CSCAN_TAINT( |
1626 | isALNUM_LC_utf8((U8*)s), | |
1627 | isALNUM_LC(*s) | |
1628 | ); | |
a0d0e21e | 1629 | case NALNUM: |
4cadc6a9 | 1630 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1631 | LOAD_UTF8_CHARCLASS_PERL_WORD(), |
f2ed9b32 | 1632 | !swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target), |
a12cf05f | 1633 | ! ((FLAGS(c) & USE_UNI) ? isWORDCHAR_L1((U8) *s) : isALNUM(*s)) |
4cadc6a9 | 1634 | ); |
bbce6d69 | 1635 | case NALNUML: |
4cadc6a9 YO |
1636 | REXEC_FBC_CSCAN_TAINT( |
1637 | !isALNUM_LC_utf8((U8*)s), | |
1638 | !isALNUM_LC(*s) | |
1639 | ); | |
a0d0e21e | 1640 | case SPACE: |
4cadc6a9 | 1641 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1642 | LOAD_UTF8_CHARCLASS_PERL_SPACE(), |
f2ed9b32 | 1643 | *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target), |
a12cf05f | 1644 | isSPACE_L1((U8) *s) && (isASCII((U8) *s) || (FLAGS(c) & USE_UNI)) |
4cadc6a9 | 1645 | ); |
bbce6d69 | 1646 | case SPACEL: |
4cadc6a9 | 1647 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1648 | isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1649 | isSPACE_LC(*s) |
1650 | ); | |
a0d0e21e | 1651 | case NSPACE: |
4cadc6a9 | 1652 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1653 | LOAD_UTF8_CHARCLASS_PERL_SPACE(), |
f2ed9b32 | 1654 | !(*s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target)), |
a12cf05f | 1655 | !(isSPACE_L1((U8) *s) && (isASCII((U8) *s) || (FLAGS(c) & USE_UNI))) |
4cadc6a9 | 1656 | ); |
bbce6d69 | 1657 | case NSPACEL: |
4cadc6a9 | 1658 | REXEC_FBC_CSCAN_TAINT( |
6bbba904 | 1659 | !isSPACE_LC_utf8((U8*)s), |
4cadc6a9 YO |
1660 | !isSPACE_LC(*s) |
1661 | ); | |
a0d0e21e | 1662 | case DIGIT: |
4cadc6a9 | 1663 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1664 | LOAD_UTF8_CHARCLASS_POSIX_DIGIT(), |
f2ed9b32 | 1665 | swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target), |
4cadc6a9 YO |
1666 | isDIGIT(*s) |
1667 | ); | |
b8c5462f | 1668 | case DIGITL: |
4cadc6a9 YO |
1669 | REXEC_FBC_CSCAN_TAINT( |
1670 | isDIGIT_LC_utf8((U8*)s), | |
1671 | isDIGIT_LC(*s) | |
1672 | ); | |
a0d0e21e | 1673 | case NDIGIT: |
4cadc6a9 | 1674 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1675 | LOAD_UTF8_CHARCLASS_POSIX_DIGIT(), |
f2ed9b32 | 1676 | !swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target), |
4cadc6a9 YO |
1677 | !isDIGIT(*s) |
1678 | ); | |
b8c5462f | 1679 | case NDIGITL: |
4cadc6a9 YO |
1680 | REXEC_FBC_CSCAN_TAINT( |
1681 | !isDIGIT_LC_utf8((U8*)s), | |
1682 | !isDIGIT_LC(*s) | |
1683 | ); | |
e1d1eefb YO |
1684 | case LNBREAK: |
1685 | REXEC_FBC_CSCAN( | |
1686 | is_LNBREAK_utf8(s), | |
1687 | is_LNBREAK_latin1(s) | |
1688 | ); | |
1689 | case VERTWS: | |
1690 | REXEC_FBC_CSCAN( | |
1691 | is_VERTWS_utf8(s), | |
1692 | is_VERTWS_latin1(s) | |
1693 | ); | |
1694 | case NVERTWS: | |
1695 | REXEC_FBC_CSCAN( | |
1696 | !is_VERTWS_utf8(s), | |
1697 | !is_VERTWS_latin1(s) | |
1698 | ); | |
1699 | case HORIZWS: | |
1700 | REXEC_FBC_CSCAN( | |
1701 | is_HORIZWS_utf8(s), | |
1702 | is_HORIZWS_latin1(s) | |
1703 | ); | |
1704 | case NHORIZWS: | |
1705 | REXEC_FBC_CSCAN( | |
1706 | !is_HORIZWS_utf8(s), | |
1707 | !is_HORIZWS_latin1(s) | |
1708 | ); | |
1de06328 YO |
1709 | case AHOCORASICKC: |
1710 | case AHOCORASICK: | |
07be1b83 | 1711 | { |
a0a388a1 | 1712 | DECL_TRIE_TYPE(c); |
07be1b83 YO |
1713 | /* what trie are we using right now */ |
1714 | reg_ac_data *aho | |
f8fc2ecf | 1715 | = (reg_ac_data*)progi->data->data[ ARG( c ) ]; |
3251b653 NC |
1716 | reg_trie_data *trie |
1717 | = (reg_trie_data*)progi->data->data[ aho->trie ]; | |
85fbaab2 | 1718 | HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); |
07be1b83 YO |
1719 | |
1720 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1721 | #ifdef DEBUGGING |
07be1b83 | 1722 | const char *real_start = s; |
6148ee25 | 1723 | #endif |
07be1b83 | 1724 | STRLEN maxlen = trie->maxlen; |
be8e71aa YO |
1725 | SV *sv_points; |
1726 | U8 **points; /* map of where we were in the input string | |
786e8c11 | 1727 | when reading a given char. For ASCII this |
be8e71aa | 1728 | is unnecessary overhead as the relationship |
38a44b82 NC |
1729 | is always 1:1, but for Unicode, especially |
1730 | case folded Unicode this is not true. */ | |
f9e705e8 | 1731 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
786e8c11 YO |
1732 | U8 *bitmap=NULL; |
1733 | ||
07be1b83 YO |
1734 | |
1735 | GET_RE_DEBUG_FLAGS_DECL; | |
1736 | ||
be8e71aa YO |
1737 | /* We can't just allocate points here. We need to wrap it in |
1738 | * an SV so it gets freed properly if there is a croak while | |
1739 | * running the match */ | |
1740 | ENTER; | |
1741 | SAVETMPS; | |
1742 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1743 | SvCUR_set(sv_points, | |
1744 | maxlen * sizeof(U8 *)); | |
1745 | SvPOK_on(sv_points); | |
1746 | sv_2mortal(sv_points); | |
1747 | points=(U8**)SvPV_nolen(sv_points ); | |
1de06328 YO |
1748 | if ( trie_type != trie_utf8_fold |
1749 | && (trie->bitmap || OP(c)==AHOCORASICKC) ) | |
1750 | { | |
786e8c11 YO |
1751 | if (trie->bitmap) |
1752 | bitmap=(U8*)trie->bitmap; | |
1753 | else | |
1754 | bitmap=(U8*)ANYOF_BITMAP(c); | |
07be1b83 | 1755 | } |
786e8c11 YO |
1756 | /* this is the Aho-Corasick algorithm modified a touch |
1757 | to include special handling for long "unknown char" | |
1758 | sequences. The basic idea being that we use AC as long | |
1759 | as we are dealing with a possible matching char, when | |
1760 | we encounter an unknown char (and we have not encountered | |
1761 | an accepting state) we scan forward until we find a legal | |
1762 | starting char. | |
1763 | AC matching is basically that of trie matching, except | |
1764 | that when we encounter a failing transition, we fall back | |
1765 | to the current states "fail state", and try the current char | |
1766 | again, a process we repeat until we reach the root state, | |
1767 | state 1, or a legal transition. If we fail on the root state | |
1768 | then we can either terminate if we have reached an accepting | |
1769 | state previously, or restart the entire process from the beginning | |
1770 | if we have not. | |
1771 | ||
1772 | */ | |
07be1b83 YO |
1773 | while (s <= last_start) { |
1774 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1775 | U8 *uc = (U8*)s; | |
1776 | U16 charid = 0; | |
1777 | U32 base = 1; | |
1778 | U32 state = 1; | |
1779 | UV uvc = 0; | |
1780 | STRLEN len = 0; | |
1781 | STRLEN foldlen = 0; | |
1782 | U8 *uscan = (U8*)NULL; | |
1783 | U8 *leftmost = NULL; | |
786e8c11 YO |
1784 | #ifdef DEBUGGING |
1785 | U32 accepted_word= 0; | |
1786 | #endif | |
07be1b83 YO |
1787 | U32 pointpos = 0; |
1788 | ||
1789 | while ( state && uc <= (U8*)strend ) { | |
1790 | int failed=0; | |
786e8c11 YO |
1791 | U32 word = aho->states[ state ].wordnum; |
1792 | ||
1de06328 YO |
1793 | if( state==1 ) { |
1794 | if ( bitmap ) { | |
1795 | DEBUG_TRIE_EXECUTE_r( | |
1796 | if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1797 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1798 | (char *)uc, utf8_target ); |
1de06328 YO |
1799 | PerlIO_printf( Perl_debug_log, |
1800 | " Scanning for legal start char...\n"); | |
1801 | } | |
d085b490 YO |
1802 | ); |
1803 | if (utf8_target) { | |
1804 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1805 | uc += UTF8SKIP(uc); | |
1806 | } | |
1807 | } else { | |
1808 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1809 | uc++; | |
1810 | } | |
1811 | } | |
1de06328 | 1812 | s= (char *)uc; |
786e8c11 | 1813 | } |
786e8c11 YO |
1814 | if (uc >(U8*)last_start) break; |
1815 | } | |
1816 | ||
1817 | if ( word ) { | |
2e64971a | 1818 | U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ]; |
786e8c11 YO |
1819 | if (!leftmost || lpos < leftmost) { |
1820 | DEBUG_r(accepted_word=word); | |
07be1b83 | 1821 | leftmost= lpos; |
786e8c11 | 1822 | } |
07be1b83 | 1823 | if (base==0) break; |
786e8c11 | 1824 | |
07be1b83 YO |
1825 | } |
1826 | points[pointpos++ % maxlen]= uc; | |
55eed653 NC |
1827 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
1828 | uscan, len, uvc, charid, foldlen, | |
1829 | foldbuf, uniflags); | |
786e8c11 YO |
1830 | DEBUG_TRIE_EXECUTE_r({ |
1831 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1832 | s, utf8_target ); |
07be1b83 | 1833 | PerlIO_printf(Perl_debug_log, |
786e8c11 YO |
1834 | " Charid:%3u CP:%4"UVxf" ", |
1835 | charid, uvc); | |
1836 | }); | |
07be1b83 YO |
1837 | |
1838 | do { | |
6148ee25 | 1839 | #ifdef DEBUGGING |
786e8c11 | 1840 | word = aho->states[ state ].wordnum; |
6148ee25 | 1841 | #endif |
07be1b83 YO |
1842 | base = aho->states[ state ].trans.base; |
1843 | ||
786e8c11 YO |
1844 | DEBUG_TRIE_EXECUTE_r({ |
1845 | if (failed) | |
1846 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1847 | s, utf8_target ); |
07be1b83 | 1848 | PerlIO_printf( Perl_debug_log, |
786e8c11 YO |
1849 | "%sState: %4"UVxf", word=%"UVxf, |
1850 | failed ? " Fail transition to " : "", | |
1851 | (UV)state, (UV)word); | |
1852 | }); | |
07be1b83 YO |
1853 | if ( base ) { |
1854 | U32 tmp; | |
6dd2be57 | 1855 | I32 offset; |
07be1b83 | 1856 | if (charid && |
6dd2be57 DM |
1857 | ( ((offset = base + charid |
1858 | - 1 - trie->uniquecharcount)) >= 0) | |
1859 | && ((U32)offset < trie->lasttrans) | |
1860 | && trie->trans[offset].check == state | |
1861 | && (tmp=trie->trans[offset].next)) | |
07be1b83 | 1862 | { |
786e8c11 YO |
1863 | DEBUG_TRIE_EXECUTE_r( |
1864 | PerlIO_printf( Perl_debug_log," - legal\n")); | |
07be1b83 YO |
1865 | state = tmp; |
1866 | break; | |
1867 | } | |
1868 | else { | |
786e8c11 YO |
1869 | DEBUG_TRIE_EXECUTE_r( |
1870 | PerlIO_printf( Perl_debug_log," - fail\n")); | |
1871 | failed = 1; | |
1872 | state = aho->fail[state]; | |
07be1b83 YO |
1873 | } |
1874 | } | |
1875 | else { | |
1876 | /* we must be accepting here */ | |
786e8c11 YO |
1877 | DEBUG_TRIE_EXECUTE_r( |
1878 | PerlIO_printf( Perl_debug_log," - accepting\n")); | |
1879 | failed = 1; | |
07be1b83 YO |
1880 | break; |
1881 | } | |
1882 | } while(state); | |
786e8c11 | 1883 | uc += len; |
07be1b83 YO |
1884 | if (failed) { |
1885 | if (leftmost) | |
1886 | break; | |
786e8c11 | 1887 | if (!state) state = 1; |
07be1b83 YO |
1888 | } |
1889 | } | |
1890 | if ( aho->states[ state ].wordnum ) { | |
2e64971a | 1891 | U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ]; |
786e8c11 YO |
1892 | if (!leftmost || lpos < leftmost) { |
1893 | DEBUG_r(accepted_word=aho->states[ state ].wordnum); | |
07be1b83 | 1894 | leftmost = lpos; |
786e8c11 | 1895 | } |
07be1b83 | 1896 | } |
07be1b83 YO |
1897 | if (leftmost) { |
1898 | s = (char*)leftmost; | |
786e8c11 YO |
1899 | DEBUG_TRIE_EXECUTE_r({ |
1900 | PerlIO_printf( | |
70685ca0 JH |
1901 | Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", |
1902 | (UV)accepted_word, (IV)(s - real_start) | |
786e8c11 YO |
1903 | ); |
1904 | }); | |
24b23f37 | 1905 | if (!reginfo || regtry(reginfo, &s)) { |
be8e71aa YO |
1906 | FREETMPS; |
1907 | LEAVE; | |
07be1b83 | 1908 | goto got_it; |
be8e71aa | 1909 | } |
07be1b83 | 1910 | s = HOPc(s,1); |
786e8c11 YO |
1911 | DEBUG_TRIE_EXECUTE_r({ |
1912 | PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); | |
1913 | }); | |
07be1b83 | 1914 | } else { |
786e8c11 YO |
1915 | DEBUG_TRIE_EXECUTE_r( |
1916 | PerlIO_printf( Perl_debug_log,"No match.\n")); | |
07be1b83 YO |
1917 | break; |
1918 | } | |
1919 | } | |
be8e71aa YO |
1920 | FREETMPS; |
1921 | LEAVE; | |
07be1b83 YO |
1922 | } |
1923 | break; | |
b3c9acc1 | 1924 | default: |
3c3eec57 GS |
1925 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); |
1926 | break; | |
d6a28714 | 1927 | } |
6eb5f6b9 JH |
1928 | return 0; |
1929 | got_it: | |
1930 | return s; | |
1931 | } | |
1932 | ||
fae667d5 | 1933 | |
6eb5f6b9 JH |
1934 | /* |
1935 | - regexec_flags - match a regexp against a string | |
1936 | */ | |
1937 | I32 | |
288b8c02 | 1938 | Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend, |
6eb5f6b9 JH |
1939 | char *strbeg, I32 minend, SV *sv, void *data, U32 flags) |
1940 | /* strend: pointer to null at end of string */ | |
1941 | /* strbeg: real beginning of string */ | |
1942 | /* minend: end of match must be >=minend after stringarg. */ | |
58e23c8d YO |
1943 | /* data: May be used for some additional optimizations. |
1944 | Currently its only used, with a U32 cast, for transmitting | |
1945 | the ganch offset when doing a /g match. This will change */ | |
6eb5f6b9 JH |
1946 | /* nosave: For optimizations. */ |
1947 | { | |
97aff369 | 1948 | dVAR; |
288b8c02 | 1949 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
24b23f37 | 1950 | /*register*/ char *s; |
6eb5f6b9 | 1951 | register regnode *c; |
24b23f37 | 1952 | /*register*/ char *startpos = stringarg; |
6eb5f6b9 JH |
1953 | I32 minlen; /* must match at least this many chars */ |
1954 | I32 dontbother = 0; /* how many characters not to try at end */ | |
6eb5f6b9 JH |
1955 | I32 end_shift = 0; /* Same for the end. */ /* CC */ |
1956 | I32 scream_pos = -1; /* Internal iterator of scream. */ | |
ccac19ea | 1957 | char *scream_olds = NULL; |
f2ed9b32 | 1958 | const bool utf8_target = cBOOL(DO_UTF8(sv)); |
2757e526 | 1959 | I32 multiline; |
f8fc2ecf | 1960 | RXi_GET_DECL(prog,progi); |
3b0527fe | 1961 | regmatch_info reginfo; /* create some info to pass to regtry etc */ |
e9105d30 | 1962 | regexp_paren_pair *swap = NULL; |
a3621e74 YO |
1963 | GET_RE_DEBUG_FLAGS_DECL; |
1964 | ||
7918f24d | 1965 | PERL_ARGS_ASSERT_REGEXEC_FLAGS; |
9d4ba2ae | 1966 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
1967 | |
1968 | /* Be paranoid... */ | |
1969 | if (prog == NULL || startpos == NULL) { | |
1970 | Perl_croak(aTHX_ "NULL regexp parameter"); | |
1971 | return 0; | |
1972 | } | |
1973 | ||
bbe252da | 1974 | multiline = prog->extflags & RXf_PMf_MULTILINE; |
288b8c02 | 1975 | reginfo.prog = rx; /* Yes, sorry that this is confusing. */ |
2757e526 | 1976 | |
f2ed9b32 | 1977 | RX_MATCH_UTF8_set(rx, utf8_target); |
1de06328 | 1978 | DEBUG_EXECUTE_r( |
f2ed9b32 | 1979 | debug_start_match(rx, utf8_target, startpos, strend, |
1de06328 YO |
1980 | "Matching"); |
1981 | ); | |
bac06658 | 1982 | |
6eb5f6b9 | 1983 | minlen = prog->minlen; |
1de06328 YO |
1984 | |
1985 | if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { | |
a3621e74 | 1986 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
1987 | "String too short [regexec_flags]...\n")); |
1988 | goto phooey; | |
1aa99e6b | 1989 | } |
6eb5f6b9 | 1990 | |
1de06328 | 1991 | |
6eb5f6b9 | 1992 | /* Check validity of program. */ |
f8fc2ecf | 1993 | if (UCHARAT(progi->program) != REG_MAGIC) { |
6eb5f6b9 JH |
1994 | Perl_croak(aTHX_ "corrupted regexp program"); |
1995 | } | |
1996 | ||
1997 | PL_reg_flags = 0; | |
1998 | PL_reg_eval_set = 0; | |
1999 | PL_reg_maxiter = 0; | |
2000 | ||
3c8556c3 | 2001 | if (RX_UTF8(rx)) |
6eb5f6b9 JH |
2002 | PL_reg_flags |= RF_utf8; |
2003 | ||
2004 | /* Mark beginning of line for ^ and lookbehind. */ | |
3b0527fe | 2005 | reginfo.bol = startpos; /* XXX not used ??? */ |
6eb5f6b9 | 2006 | PL_bostr = strbeg; |
3b0527fe | 2007 | reginfo.sv = sv; |
6eb5f6b9 JH |
2008 | |
2009 | /* Mark end of line for $ (and such) */ | |
2010 | PL_regeol = strend; | |
2011 | ||
2012 | /* see how far we have to get to not match where we matched before */ | |
3b0527fe | 2013 | reginfo.till = startpos+minend; |
6eb5f6b9 | 2014 | |
6eb5f6b9 JH |
2015 | /* If there is a "must appear" string, look for it. */ |
2016 | s = startpos; | |
2017 | ||
bbe252da | 2018 | if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ |
6eb5f6b9 | 2019 | MAGIC *mg; |
2c296965 | 2020 | if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */ |
58e23c8d | 2021 | reginfo.ganch = startpos + prog->gofs; |
2c296965 | 2022 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2023 | "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs)); |
2c296965 | 2024 | } else if (sv && SvTYPE(sv) >= SVt_PVMG |
6eb5f6b9 | 2025 | && SvMAGIC(sv) |
14befaf4 DM |
2026 | && (mg = mg_find(sv, PERL_MAGIC_regex_global)) |
2027 | && mg->mg_len >= 0) { | |
3b0527fe | 2028 | reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ |
2c296965 | 2029 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2030 | "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len)); |
2c296965 | 2031 | |
bbe252da | 2032 | if (prog->extflags & RXf_ANCH_GPOS) { |
3b0527fe | 2033 | if (s > reginfo.ganch) |
6eb5f6b9 | 2034 | goto phooey; |
58e23c8d | 2035 | s = reginfo.ganch - prog->gofs; |
2c296965 | 2036 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 2037 | "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs)); |
c584a96e YO |
2038 | if (s < strbeg) |
2039 | goto phooey; | |
6eb5f6b9 JH |
2040 | } |
2041 | } | |
58e23c8d | 2042 | else if (data) { |
70685ca0 | 2043 | reginfo.ganch = strbeg + PTR2UV(data); |
2c296965 YO |
2044 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2045 | "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data))); | |
2046 | ||
2047 | } else { /* pos() not defined */ | |
3b0527fe | 2048 | reginfo.ganch = strbeg; |
2c296965 YO |
2049 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
2050 | "GPOS: reginfo.ganch = strbeg\n")); | |
2051 | } | |
6eb5f6b9 | 2052 | } |
288b8c02 | 2053 | if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { |
e9105d30 GG |
2054 | /* We have to be careful. If the previous successful match |
2055 | was from this regex we don't want a subsequent partially | |
2056 | successful match to clobber the old results. | |
2057 | So when we detect this possibility we add a swap buffer | |
2058 | to the re, and switch the buffer each match. If we fail | |
2059 | we switch it back, otherwise we leave it swapped. | |
2060 | */ | |
2061 | swap = prog->offs; | |
2062 | /* do we need a save destructor here for eval dies? */ | |
2063 | Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); | |
c74340f9 | 2064 | } |
a0714e2c | 2065 | if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { |
6eb5f6b9 JH |
2066 | re_scream_pos_data d; |
2067 | ||
2068 | d.scream_olds = &scream_olds; | |
2069 | d.scream_pos = &scream_pos; | |
288b8c02 | 2070 | s = re_intuit_start(rx, sv, s, strend, flags, &d); |
3fa9c3d7 | 2071 | if (!s) { |
a3621e74 | 2072 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n")); |
6eb5f6b9 | 2073 | goto phooey; /* not present */ |
3fa9c3d7 | 2074 | } |
6eb5f6b9 JH |
2075 | } |
2076 | ||
1de06328 | 2077 | |
6eb5f6b9 JH |
2078 | |
2079 | /* Simplest case: anchored match need be tried only once. */ | |
2080 | /* [unless only anchor is BOL and multiline is set] */ | |
bbe252da | 2081 | if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { |
24b23f37 | 2082 | if (s == startpos && regtry(®info, &startpos)) |
6eb5f6b9 | 2083 | goto got_it; |
bbe252da YO |
2084 | else if (multiline || (prog->intflags & PREGf_IMPLICIT) |
2085 | || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ | |
6eb5f6b9 JH |
2086 | { |
2087 | char *end; | |
2088 | ||
2089 | if (minlen) | |
2090 | dontbother = minlen - 1; | |
1aa99e6b | 2091 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 2092 | /* for multiline we only have to try after newlines */ |
33b8afdf | 2093 | if (prog->check_substr || prog->check_utf8) { |
92f3d482 YO |
2094 | /* because of the goto we can not easily reuse the macros for bifurcating the |
2095 | unicode/non-unicode match modes here like we do elsewhere - demerphq */ | |
2096 | if (utf8_target) { | |
2097 | if (s == startpos) | |
2098 | goto after_try_utf8; | |
2099 | while (1) { | |
2100 | if (regtry(®info, &s)) { | |
2101 | goto got_it; | |
2102 | } | |
2103 | after_try_utf8: | |
2104 | if (s > end) { | |
2105 | goto phooey; | |
2106 | } | |
2107 | if (prog->extflags & RXf_USE_INTUIT) { | |
2108 | s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL); | |
2109 | if (!s) { | |
2110 | goto phooey; | |
2111 | } | |
2112 | } | |
2113 | else { | |
2114 | s += UTF8SKIP(s); | |
2115 | } | |
2116 | } | |
2117 | } /* end search for check string in unicode */ | |
2118 | else { | |
2119 | if (s == startpos) { | |
2120 | goto after_try_latin; | |
2121 | } | |
2122 | while (1) { | |
2123 | if (regtry(®info, &s)) { | |
2124 | goto got_it; | |
2125 | } | |
2126 | after_try_latin: | |
2127 | if (s > end) { | |
2128 | goto phooey; | |
2129 | } | |
2130 | if (prog->extflags & RXf_USE_INTUIT) { | |
2131 | s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL); | |
2132 | if (!s) { | |
2133 | goto phooey; | |
2134 | } | |
2135 | } | |
2136 | else { | |
2137 | s++; | |
2138 | } | |
2139 | } | |
2140 | } /* end search for check string in latin*/ | |
2141 | } /* end search for check string */ | |
2142 | else { /* search for newline */ | |
2143 | if (s > startpos) { | |
2144 | /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/ | |
6eb5f6b9 | 2145 | s--; |
92f3d482 YO |
2146 | } |
2147 | /* We can use a more efficient search as newlines are the same in unicode as they are in latin */ | |
6eb5f6b9 JH |
2148 | while (s < end) { |
2149 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ | |
24b23f37 | 2150 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2151 | goto got_it; |
2152 | } | |
92f3d482 YO |
2153 | } |
2154 | } /* end search for newline */ | |
2155 | } /* end anchored/multiline check string search */ | |
6eb5f6b9 | 2156 | goto phooey; |
bbe252da | 2157 | } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) |
f9f4320a YO |
2158 | { |
2159 | /* the warning about reginfo.ganch being used without intialization | |
bbe252da | 2160 | is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN |
f9f4320a | 2161 | and we only enter this block when the same bit is set. */ |
58e23c8d | 2162 | char *tmp_s = reginfo.ganch - prog->gofs; |
c584a96e YO |
2163 | |
2164 | if (tmp_s >= strbeg && regtry(®info, &tmp_s)) | |
6eb5f6b9 JH |
2165 | goto got_it; |
2166 | goto phooey; | |
2167 | } | |
2168 | ||
2169 | /* Messy cases: unanchored match. */ | |
bbe252da | 2170 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { |
6eb5f6b9 | 2171 | /* we have /x+whatever/ */ |
f2ed9b32 | 2172 | /* it must be a one character string (XXXX Except UTF_PATTERN?) */ |
33b8afdf | 2173 | char ch; |
bf93d4cc GS |
2174 | #ifdef DEBUGGING |
2175 | int did_match = 0; | |
2176 | #endif | |
f2ed9b32 KW |
2177 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2178 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2179 | ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0]; | |
bf93d4cc | 2180 | |
f2ed9b32 | 2181 | if (utf8_target) { |
4cadc6a9 | 2182 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2183 | if (*s == ch) { |
a3621e74 | 2184 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2185 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2186 | s += UTF8SKIP(s); |
2187 | while (s < strend && *s == ch) | |
2188 | s += UTF8SKIP(s); | |
2189 | } | |
4cadc6a9 | 2190 | ); |
6eb5f6b9 JH |
2191 | } |
2192 | else { | |
4cadc6a9 | 2193 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2194 | if (*s == ch) { |
a3621e74 | 2195 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2196 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2197 | s++; |
2198 | while (s < strend && *s == ch) | |
2199 | s++; | |
2200 | } | |
4cadc6a9 | 2201 | ); |
6eb5f6b9 | 2202 | } |
a3621e74 | 2203 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 2204 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
2205 | "Did not find anchored character...\n") |
2206 | ); | |
6eb5f6b9 | 2207 | } |
a0714e2c SS |
2208 | else if (prog->anchored_substr != NULL |
2209 | || prog->anchored_utf8 != NULL | |
2210 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
2211 | && prog->float_max_offset < strend - s)) { |
2212 | SV *must; | |
2213 | I32 back_max; | |
2214 | I32 back_min; | |
2215 | char *last; | |
6eb5f6b9 | 2216 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
2217 | #ifdef DEBUGGING |
2218 | int did_match = 0; | |
2219 | #endif | |
33b8afdf | 2220 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 KW |
2221 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2222 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2223 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; | |
33b8afdf JH |
2224 | back_max = back_min = prog->anchored_offset; |
2225 | } else { | |
f2ed9b32 KW |
2226 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2227 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2228 | must = utf8_target ? prog->float_utf8 : prog->float_substr; | |
33b8afdf JH |
2229 | back_max = prog->float_max_offset; |
2230 | back_min = prog->float_min_offset; | |
2231 | } | |
1de06328 YO |
2232 | |
2233 | ||
33b8afdf JH |
2234 | if (must == &PL_sv_undef) |
2235 | /* could not downgrade utf8 check substring, so must fail */ | |
2236 | goto phooey; | |
2237 | ||
1de06328 YO |
2238 | if (back_min<0) { |
2239 | last = strend; | |
2240 | } else { | |
2241 | last = HOP3c(strend, /* Cannot start after this */ | |
2242 | -(I32)(CHR_SVLEN(must) | |
2243 | - (SvTAIL(must) != 0) + back_min), strbeg); | |
2244 | } | |
6eb5f6b9 JH |
2245 | if (s > PL_bostr) |
2246 | last1 = HOPc(s, -1); | |
2247 | else | |
2248 | last1 = s - 1; /* bogus */ | |
2249 | ||
a0288114 | 2250 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 JH |
2251 | check_substr==must. */ |
2252 | scream_pos = -1; | |
2253 | dontbother = end_shift; | |
2254 | strend = HOPc(strend, -dontbother); | |
2255 | while ( (s <= last) && | |
9041c2e3 | 2256 | ((flags & REXEC_SCREAM) |
1de06328 | 2257 | ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg, |
6eb5f6b9 | 2258 | end_shift, &scream_pos, 0)) |
1de06328 | 2259 | : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), |
9041c2e3 | 2260 | (unsigned char*)strend, must, |
7fba1cd6 | 2261 | multiline ? FBMrf_MULTILINE : 0))) ) { |
4addbd3b | 2262 | /* we may be pointing at the wrong string */ |
07bc277f | 2263 | if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog)) |
3f7c398e | 2264 | s = strbeg + (s - SvPVX_const(sv)); |
a3621e74 | 2265 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
2266 | if (HOPc(s, -back_max) > last1) { |
2267 | last1 = HOPc(s, -back_min); | |
2268 | s = HOPc(s, -back_max); | |
2269 | } | |
2270 | else { | |
52657f30 | 2271 | char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; |
6eb5f6b9 JH |
2272 | |
2273 | last1 = HOPc(s, -back_min); | |
52657f30 | 2274 | s = t; |
6eb5f6b9 | 2275 | } |
f2ed9b32 | 2276 | if (utf8_target) { |
6eb5f6b9 | 2277 | while (s <= last1) { |
24b23f37 | 2278 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2279 | goto got_it; |
2280 | s += UTF8SKIP(s); | |
2281 | } | |
2282 | } | |
2283 | else { | |
2284 | while (s <= last1) { | |
24b23f37 | 2285 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2286 | goto got_it; |
2287 | s++; | |
2288 | } | |
2289 | } | |
2290 | } | |
ab3bbdeb | 2291 | DEBUG_EXECUTE_r(if (!did_match) { |
f2ed9b32 | 2292 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
2293 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
2294 | PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", | |
33b8afdf | 2295 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc | 2296 | ? "anchored" : "floating"), |
ab3bbdeb YO |
2297 | quoted, RE_SV_TAIL(must)); |
2298 | }); | |
6eb5f6b9 JH |
2299 | goto phooey; |
2300 | } | |
f8fc2ecf | 2301 | else if ( (c = progi->regstclass) ) { |
f14c76ed | 2302 | if (minlen) { |
f8fc2ecf | 2303 | const OPCODE op = OP(progi->regstclass); |
66e933ab | 2304 | /* don't bother with what can't match */ |
786e8c11 | 2305 | if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) |
f14c76ed RGS |
2306 | strend = HOPc(strend, -(minlen - 1)); |
2307 | } | |
a3621e74 | 2308 | DEBUG_EXECUTE_r({ |
be8e71aa | 2309 | SV * const prop = sv_newmortal(); |
32fc9b6a | 2310 | regprop(prog, prop, c); |
0df25f3d | 2311 | { |
f2ed9b32 | 2312 | RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb | 2313 | s,strend-s,60); |
0df25f3d | 2314 | PerlIO_printf(Perl_debug_log, |
1c8f8eb1 | 2315 | "Matching stclass %.*s against %s (%d bytes)\n", |
e4f74956 | 2316 | (int)SvCUR(prop), SvPVX_const(prop), |
ab3bbdeb | 2317 | quoted, (int)(strend - s)); |
0df25f3d | 2318 | } |
ffc61ed2 | 2319 | }); |
3b0527fe | 2320 | if (find_byclass(prog, c, s, strend, ®info)) |
6eb5f6b9 | 2321 | goto got_it; |
07be1b83 | 2322 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
2323 | } |
2324 | else { | |
2325 | dontbother = 0; | |
a0714e2c | 2326 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 2327 | /* Trim the end. */ |
d6a28714 | 2328 | char *last; |
33b8afdf JH |
2329 | SV* float_real; |
2330 | ||
f2ed9b32 KW |
2331 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2332 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2333 | float_real = utf8_target ? prog->float_utf8 : prog->float_substr; | |
d6a28714 JH |
2334 | |
2335 | if (flags & REXEC_SCREAM) { | |
33b8afdf | 2336 | last = screaminstr(sv, float_real, s - strbeg, |
d6a28714 JH |
2337 | end_shift, &scream_pos, 1); /* last one */ |
2338 | if (!last) | |
ffc61ed2 | 2339 | last = scream_olds; /* Only one occurrence. */ |
4addbd3b | 2340 | /* we may be pointing at the wrong string */ |
07bc277f | 2341 | else if (RXp_MATCH_COPIED(prog)) |
3f7c398e | 2342 | s = strbeg + (s - SvPVX_const(sv)); |
b8c5462f | 2343 | } |
d6a28714 JH |
2344 | else { |
2345 | STRLEN len; | |
cfd0369c | 2346 | const char * const little = SvPV_const(float_real, len); |
d6a28714 | 2347 | |
33b8afdf | 2348 | if (SvTAIL(float_real)) { |
d6a28714 JH |
2349 | if (memEQ(strend - len + 1, little, len - 1)) |
2350 | last = strend - len + 1; | |
7fba1cd6 | 2351 | else if (!multiline) |
9041c2e3 | 2352 | last = memEQ(strend - len, little, len) |
bd61b366 | 2353 | ? strend - len : NULL; |
b8c5462f | 2354 | else |
d6a28714 JH |
2355 | goto find_last; |
2356 | } else { | |
2357 | find_last: | |
9041c2e3 | 2358 | if (len) |
d6a28714 | 2359 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 2360 | else |
a0288114 | 2361 | last = strend; /* matching "$" */ |
b8c5462f | 2362 | } |
b8c5462f | 2363 | } |
bf93d4cc | 2364 | if (last == NULL) { |
6bda09f9 YO |
2365 | DEBUG_EXECUTE_r( |
2366 | PerlIO_printf(Perl_debug_log, | |
2367 | "%sCan't trim the tail, match fails (should not happen)%s\n", | |
2368 | PL_colors[4], PL_colors[5])); | |
bf93d4cc GS |
2369 | goto phooey; /* Should not happen! */ |
2370 | } | |
d6a28714 JH |
2371 | dontbother = strend - last + prog->float_min_offset; |
2372 | } | |
2373 | if (minlen && (dontbother < minlen)) | |
2374 | dontbother = minlen - 1; | |
2375 | strend -= dontbother; /* this one's always in bytes! */ | |
2376 | /* We don't know much -- general case. */ | |
f2ed9b32 | 2377 | if (utf8_target) { |
d6a28714 | 2378 | for (;;) { |
24b23f37 | 2379 | if (regtry(®info, &s)) |
d6a28714 JH |
2380 | goto got_it; |
2381 | if (s >= strend) | |
2382 | break; | |
b8c5462f | 2383 | s += UTF8SKIP(s); |
d6a28714 JH |
2384 | }; |
2385 | } | |
2386 | else { | |
2387 | do { | |
24b23f37 | 2388 | if (regtry(®info, &s)) |
d6a28714 JH |
2389 | goto got_it; |
2390 | } while (s++ < strend); | |
2391 | } | |
2392 | } | |
2393 | ||
2394 | /* Failure. */ | |
2395 | goto phooey; | |
2396 | ||
2397 | got_it: | |
e9105d30 | 2398 | Safefree(swap); |
288b8c02 | 2399 | RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted); |
d6a28714 | 2400 | |
19b95bf0 | 2401 | if (PL_reg_eval_set) |
4f639d21 | 2402 | restore_pos(aTHX_ prog); |
5daac39c NC |
2403 | if (RXp_PAREN_NAMES(prog)) |
2404 | (void)hv_iterinit(RXp_PAREN_NAMES(prog)); | |
d6a28714 JH |
2405 | |
2406 | /* make sure $`, $&, $', and $digit will work later */ | |
2407 | if ( !(flags & REXEC_NOT_FIRST) ) { | |
288b8c02 | 2408 | RX_MATCH_COPY_FREE(rx); |
d6a28714 | 2409 | if (flags & REXEC_COPY_STR) { |
be8e71aa | 2410 | const I32 i = PL_regeol - startpos + (stringarg - strbeg); |
f8c7b90f | 2411 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2412 | if ((SvIsCOW(sv) |
2413 | || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { | |
2414 | if (DEBUG_C_TEST) { | |
2415 | PerlIO_printf(Perl_debug_log, | |
2416 | "Copy on write: regexp capture, type %d\n", | |
2417 | (int) SvTYPE(sv)); | |
2418 | } | |
2419 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); | |
d5263905 | 2420 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
ed252734 NC |
2421 | assert (SvPOKp(prog->saved_copy)); |
2422 | } else | |
2423 | #endif | |
2424 | { | |
288b8c02 | 2425 | RX_MATCH_COPIED_on(rx); |
ed252734 NC |
2426 | s = savepvn(strbeg, i); |
2427 | prog->subbeg = s; | |
2428 | } | |
d6a28714 | 2429 | prog->sublen = i; |
d6a28714 JH |
2430 | } |
2431 | else { | |
2432 | prog->subbeg = strbeg; | |
2433 | prog->sublen = PL_regeol - strbeg; /* strend may have been modified */ | |
2434 | } | |
2435 | } | |
9041c2e3 | 2436 | |
d6a28714 JH |
2437 | return 1; |
2438 | ||
2439 | phooey: | |
a3621e74 | 2440 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 2441 | PL_colors[4], PL_colors[5])); |
d6a28714 | 2442 | if (PL_reg_eval_set) |
4f639d21 | 2443 | restore_pos(aTHX_ prog); |
e9105d30 | 2444 | if (swap) { |
c74340f9 | 2445 | /* we failed :-( roll it back */ |
e9105d30 GG |
2446 | Safefree(prog->offs); |
2447 | prog->offs = swap; | |
2448 | } | |
2449 | ||
d6a28714 JH |
2450 | return 0; |
2451 | } | |
2452 | ||
6bda09f9 | 2453 | |
d6a28714 JH |
2454 | /* |
2455 | - regtry - try match at specific point | |
2456 | */ | |
2457 | STATIC I32 /* 0 failure, 1 success */ | |
24b23f37 | 2458 | S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) |
d6a28714 | 2459 | { |
97aff369 | 2460 | dVAR; |
d6a28714 | 2461 | CHECKPOINT lastcp; |
288b8c02 NC |
2462 | REGEXP *const rx = reginfo->prog; |
2463 | regexp *const prog = (struct regexp *)SvANY(rx); | |
f8fc2ecf | 2464 | RXi_GET_DECL(prog,progi); |
a3621e74 | 2465 | GET_RE_DEBUG_FLAGS_DECL; |
7918f24d NC |
2466 | |
2467 | PERL_ARGS_ASSERT_REGTRY; | |
2468 | ||
24b23f37 | 2469 | reginfo->cutpoint=NULL; |
d6a28714 | 2470 | |
bbe252da | 2471 | if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) { |
d6a28714 JH |
2472 | MAGIC *mg; |
2473 | ||
2474 | PL_reg_eval_set = RS_init; | |
a3621e74 | 2475 | DEBUG_EXECUTE_r(DEBUG_s( |
b900a521 JH |
2476 | PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n", |
2477 | (IV)(PL_stack_sp - PL_stack_base)); | |
d6a28714 | 2478 | )); |
ea8d6ae1 | 2479 | SAVESTACK_CXPOS(); |
d6a28714 JH |
2480 | cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base; |
2481 | /* Otherwise OP_NEXTSTATE will free whatever on stack now. */ | |
2482 | SAVETMPS; | |
2483 | /* Apparently this is not needed, judging by wantarray. */ | |
e8347627 | 2484 | /* SAVEI8(cxstack[cxstack_ix].blk_gimme); |
d6a28714 JH |
2485 | cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ |
2486 | ||
3b0527fe | 2487 | if (reginfo->sv) { |
d6a28714 | 2488 | /* Make $_ available to executed code. */ |
3b0527fe | 2489 | if (reginfo->sv != DEFSV) { |
59f00321 | 2490 | SAVE_DEFSV; |
414bf5ae | 2491 | DEFSV_set(reginfo->sv); |
b8c5462f | 2492 | } |
d6a28714 | 2493 | |
3b0527fe DM |
2494 | if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) |
2495 | && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { | |
d6a28714 | 2496 | /* prepare for quick setting of pos */ |
d300d9fa | 2497 | #ifdef PERL_OLD_COPY_ON_WRITE |
51a9ea20 NC |
2498 | if (SvIsCOW(reginfo->sv)) |
2499 | sv_force_normal_flags(reginfo->sv, 0); | |
d300d9fa | 2500 | #endif |
3dab1dad | 2501 | mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, |
d300d9fa | 2502 | &PL_vtbl_mglob, NULL, 0); |
d6a28714 | 2503 | mg->mg_len = -1; |
b8c5462f | 2504 | } |
d6a28714 JH |
2505 | PL_reg_magic = mg; |
2506 | PL_reg_oldpos = mg->mg_len; | |
4f639d21 | 2507 | SAVEDESTRUCTOR_X(restore_pos, prog); |
d6a28714 | 2508 | } |
09687e5a | 2509 | if (!PL_reg_curpm) { |
a02a5408 | 2510 | Newxz(PL_reg_curpm, 1, PMOP); |
09687e5a AB |
2511 | #ifdef USE_ITHREADS |
2512 | { | |
14a49a24 | 2513 | SV* const repointer = &PL_sv_undef; |
92313705 NC |
2514 | /* this regexp is also owned by the new PL_reg_curpm, which |
2515 | will try to free it. */ | |
d2ece331 | 2516 | av_push(PL_regex_padav, repointer); |
09687e5a AB |
2517 | PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); |
2518 | PL_regex_pad = AvARRAY(PL_regex_padav); | |
2519 | } | |
2520 | #endif | |
2521 | } | |
86c29d75 NC |
2522 | #ifdef USE_ITHREADS |
2523 | /* It seems that non-ithreads works both with and without this code. | |
2524 | So for efficiency reasons it seems best not to have the code | |
2525 | compiled when it is not needed. */ | |
92313705 NC |
2526 | /* This is safe against NULLs: */ |
2527 | ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); | |
2528 | /* PM_reg_curpm owns a reference to this regexp. */ | |
2529 | ReREFCNT_inc(rx); | |
86c29d75 | 2530 | #endif |
288b8c02 | 2531 | PM_SETRE(PL_reg_curpm, rx); |
d6a28714 JH |
2532 | PL_reg_oldcurpm = PL_curpm; |
2533 | PL_curpm = PL_reg_curpm; | |
07bc277f | 2534 | if (RXp_MATCH_COPIED(prog)) { |
d6a28714 JH |
2535 | /* Here is a serious problem: we cannot rewrite subbeg, |
2536 | since it may be needed if this match fails. Thus | |
2537 | $` inside (?{}) could fail... */ | |
2538 | PL_reg_oldsaved = prog->subbeg; | |
2539 | PL_reg_oldsavedlen = prog->sublen; | |
f8c7b90f | 2540 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2541 | PL_nrs = prog->saved_copy; |
2542 | #endif | |
07bc277f | 2543 | RXp_MATCH_COPIED_off(prog); |
d6a28714 JH |
2544 | } |
2545 | else | |
bd61b366 | 2546 | PL_reg_oldsaved = NULL; |
d6a28714 JH |
2547 | prog->subbeg = PL_bostr; |
2548 | prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ | |
2549 | } | |
24b23f37 | 2550 | DEBUG_EXECUTE_r(PL_reg_starttry = *startpos); |
f0ab9afb | 2551 | prog->offs[0].start = *startpos - PL_bostr; |
24b23f37 | 2552 | PL_reginput = *startpos; |
d6a28714 | 2553 | PL_reglastparen = &prog->lastparen; |
a01268b5 | 2554 | PL_reglastcloseparen = &prog->lastcloseparen; |
d6a28714 | 2555 | prog->lastparen = 0; |
03994de8 | 2556 | prog->lastcloseparen = 0; |
d6a28714 | 2557 | PL_regsize = 0; |
f0ab9afb | 2558 | PL_regoffs = prog->offs; |
d6a28714 JH |
2559 | if (PL_reg_start_tmpl <= prog->nparens) { |
2560 | PL_reg_start_tmpl = prog->nparens*3/2 + 3; | |
2561 | if(PL_reg_start_tmp) | |
2562 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
2563 | else | |
a02a5408 | 2564 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); |
d6a28714 JH |
2565 | } |
2566 | ||
2567 | /* XXXX What this code is doing here?!!! There should be no need | |
2568 | to do this again and again, PL_reglastparen should take care of | |
3dd2943c | 2569 | this! --ilya*/ |
dafc8851 JH |
2570 | |
2571 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2572 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
daf18116 | 2573 | * PL_reglastparen), is not needed at all by the test suite |
225593e1 DM |
2574 | * (op/regexp, op/pat, op/split), but that code is needed otherwise |
2575 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ | |
2576 | * Meanwhile, this code *is* needed for the | |
daf18116 JH |
2577 | * above-mentioned test suite tests to succeed. The common theme |
2578 | * on those tests seems to be returning null fields from matches. | |
225593e1 | 2579 | * --jhi updated by dapm */ |
dafc8851 | 2580 | #if 1 |
d6a28714 | 2581 | if (prog->nparens) { |
f0ab9afb | 2582 | regexp_paren_pair *pp = PL_regoffs; |
097eb12c | 2583 | register I32 i; |
eb160463 | 2584 | for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) { |
f0ab9afb NC |
2585 | ++pp; |
2586 | pp->start = -1; | |
2587 | pp->end = -1; | |
d6a28714 JH |
2588 | } |
2589 | } | |
dafc8851 | 2590 | #endif |
02db2b7b | 2591 | REGCP_SET(lastcp); |
f8fc2ecf | 2592 | if (regmatch(reginfo, progi->program + 1)) { |
f0ab9afb | 2593 | PL_regoffs[0].end = PL_reginput - PL_bostr; |
d6a28714 JH |
2594 | return 1; |
2595 | } | |
24b23f37 YO |
2596 | if (reginfo->cutpoint) |
2597 | *startpos= reginfo->cutpoint; | |
02db2b7b | 2598 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
2599 | return 0; |
2600 | } | |
2601 | ||
02db2b7b | 2602 | |
8ba1375e MJD |
2603 | #define sayYES goto yes |
2604 | #define sayNO goto no | |
262b90c4 | 2605 | #define sayNO_SILENT goto no_silent |
8ba1375e | 2606 | |
f9f4320a YO |
2607 | /* we dont use STMT_START/END here because it leads to |
2608 | "unreachable code" warnings, which are bogus, but distracting. */ | |
2609 | #define CACHEsayNO \ | |
c476f425 DM |
2610 | if (ST.cache_mask) \ |
2611 | PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \ | |
f9f4320a | 2612 | sayNO |
3298f257 | 2613 | |
a3621e74 | 2614 | /* this is used to determine how far from the left messages like |
265c4333 YO |
2615 | 'failed...' are printed. It should be set such that messages |
2616 | are inline with the regop output that created them. | |
a3621e74 | 2617 | */ |
265c4333 | 2618 | #define REPORT_CODE_OFF 32 |
a3621e74 YO |
2619 | |
2620 | ||
40a82448 DM |
2621 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
2622 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
9e137952 | 2623 | |
86545054 DM |
2624 | #define SLAB_FIRST(s) (&(s)->states[0]) |
2625 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
2626 | ||
5d9a96ca DM |
2627 | /* grab a new slab and return the first slot in it */ |
2628 | ||
2629 | STATIC regmatch_state * | |
2630 | S_push_slab(pTHX) | |
2631 | { | |
a35a87e7 | 2632 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2633 | dMY_CXT; |
2634 | #endif | |
5d9a96ca DM |
2635 | regmatch_slab *s = PL_regmatch_slab->next; |
2636 | if (!s) { | |
2637 | Newx(s, 1, regmatch_slab); | |
2638 | s->prev = PL_regmatch_slab; | |
2639 | s->next = NULL; | |
2640 | PL_regmatch_slab->next = s; | |
2641 | } | |
2642 | PL_regmatch_slab = s; | |
86545054 | 2643 | return SLAB_FIRST(s); |
5d9a96ca | 2644 | } |
5b47454d | 2645 | |
95b24440 | 2646 | |
40a82448 DM |
2647 | /* push a new state then goto it */ |
2648 | ||
2649 | #define PUSH_STATE_GOTO(state, node) \ | |
2650 | scan = node; \ | |
2651 | st->resume_state = state; \ | |
2652 | goto push_state; | |
2653 | ||
2654 | /* push a new state with success backtracking, then goto it */ | |
2655 | ||
2656 | #define PUSH_YES_STATE_GOTO(state, node) \ | |
2657 | scan = node; \ | |
2658 | st->resume_state = state; \ | |
2659 | goto push_yes_state; | |
2660 | ||
aa283a38 | 2661 | |
aa283a38 | 2662 | |
d6a28714 | 2663 | /* |
95b24440 | 2664 | |
bf1f174e DM |
2665 | regmatch() - main matching routine |
2666 | ||
2667 | This is basically one big switch statement in a loop. We execute an op, | |
2668 | set 'next' to point the next op, and continue. If we come to a point which | |
2669 | we may need to backtrack to on failure such as (A|B|C), we push a | |
2670 | backtrack state onto the backtrack stack. On failure, we pop the top | |
2671 | state, and re-enter the loop at the state indicated. If there are no more | |
2672 | states to pop, we return failure. | |
2673 | ||
2674 | Sometimes we also need to backtrack on success; for example /A+/, where | |
2675 | after successfully matching one A, we need to go back and try to | |
2676 | match another one; similarly for lookahead assertions: if the assertion | |
2677 | completes successfully, we backtrack to the state just before the assertion | |
2678 | and then carry on. In these cases, the pushed state is marked as | |
2679 | 'backtrack on success too'. This marking is in fact done by a chain of | |
2680 | pointers, each pointing to the previous 'yes' state. On success, we pop to | |
2681 | the nearest yes state, discarding any intermediate failure-only states. | |
2682 | Sometimes a yes state is pushed just to force some cleanup code to be | |
2683 | called at the end of a successful match or submatch; e.g. (??{$re}) uses | |
2684 | it to free the inner regex. | |
2685 | ||
2686 | Note that failure backtracking rewinds the cursor position, while | |
2687 | success backtracking leaves it alone. | |
2688 | ||
2689 | A pattern is complete when the END op is executed, while a subpattern | |
2690 | such as (?=foo) is complete when the SUCCESS op is executed. Both of these | |
2691 | ops trigger the "pop to last yes state if any, otherwise return true" | |
2692 | behaviour. | |
2693 | ||
2694 | A common convention in this function is to use A and B to refer to the two | |
2695 | subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is | |
2696 | the subpattern to be matched possibly multiple times, while B is the entire | |
2697 | rest of the pattern. Variable and state names reflect this convention. | |
2698 | ||
2699 | The states in the main switch are the union of ops and failure/success of | |
2700 | substates associated with with that op. For example, IFMATCH is the op | |
2701 | that does lookahead assertions /(?=A)B/ and so the IFMATCH state means | |
2702 | 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just | |
2703 | successfully matched A and IFMATCH_A_fail is a state saying that we have | |
2704 | just failed to match A. Resume states always come in pairs. The backtrack | |
2705 | state we push is marked as 'IFMATCH_A', but when that is popped, we resume | |
2706 | at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking | |
2707 | on success or failure. | |
2708 | ||
2709 | The struct that holds a backtracking state is actually a big union, with | |
2710 | one variant for each major type of op. The variable st points to the | |
2711 | top-most backtrack struct. To make the code clearer, within each | |
2712 | block of code we #define ST to alias the relevant union. | |
2713 | ||
2714 | Here's a concrete example of a (vastly oversimplified) IFMATCH | |
2715 | implementation: | |
2716 | ||
2717 | switch (state) { | |
2718 | .... | |
2719 | ||
2720 | #define ST st->u.ifmatch | |
2721 | ||
2722 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B | |
2723 | ST.foo = ...; // some state we wish to save | |
95b24440 | 2724 | ... |
bf1f174e DM |
2725 | // push a yes backtrack state with a resume value of |
2726 | // IFMATCH_A/IFMATCH_A_fail, then continue execution at the | |
2727 | // first node of A: | |
2728 | PUSH_YES_STATE_GOTO(IFMATCH_A, A); | |
2729 | // NOTREACHED | |
2730 | ||
2731 | case IFMATCH_A: // we have successfully executed A; now continue with B | |
2732 | next = B; | |
2733 | bar = ST.foo; // do something with the preserved value | |
2734 | break; | |
2735 | ||
2736 | case IFMATCH_A_fail: // A failed, so the assertion failed | |
2737 | ...; // do some housekeeping, then ... | |
2738 | sayNO; // propagate the failure | |
2739 | ||
2740 | #undef ST | |
95b24440 | 2741 | |
bf1f174e DM |
2742 | ... |
2743 | } | |
95b24440 | 2744 | |
bf1f174e DM |
2745 | For any old-timers reading this who are familiar with the old recursive |
2746 | approach, the code above is equivalent to: | |
95b24440 | 2747 | |
bf1f174e DM |
2748 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B |
2749 | { | |
2750 | int foo = ... | |
95b24440 | 2751 | ... |
bf1f174e DM |
2752 | if (regmatch(A)) { |
2753 | next = B; | |
2754 | bar = foo; | |
2755 | break; | |
95b24440 | 2756 | } |
bf1f174e DM |
2757 | ...; // do some housekeeping, then ... |
2758 | sayNO; // propagate the failure | |
95b24440 | 2759 | } |
bf1f174e DM |
2760 | |
2761 | The topmost backtrack state, pointed to by st, is usually free. If you | |
2762 | want to claim it, populate any ST.foo fields in it with values you wish to | |
2763 | save, then do one of | |
2764 | ||
2765 | PUSH_STATE_GOTO(resume_state, node); | |
2766 | PUSH_YES_STATE_GOTO(resume_state, node); | |
2767 | ||
2768 | which sets that backtrack state's resume value to 'resume_state', pushes a | |
2769 | new free entry to the top of the backtrack stack, then goes to 'node'. | |
2770 | On backtracking, the free slot is popped, and the saved state becomes the | |
2771 | new free state. An ST.foo field in this new top state can be temporarily | |
2772 | accessed to retrieve values, but once the main loop is re-entered, it | |
2773 | becomes available for reuse. | |
2774 | ||
2775 | Note that the depth of the backtrack stack constantly increases during the | |
2776 | left-to-right execution of the pattern, rather than going up and down with | |
2777 | the pattern nesting. For example the stack is at its maximum at Z at the | |
2778 | end of the pattern, rather than at X in the following: | |
2779 | ||
2780 | /(((X)+)+)+....(Y)+....Z/ | |
2781 | ||
2782 | The only exceptions to this are lookahead/behind assertions and the cut, | |
2783 | (?>A), which pop all the backtrack states associated with A before | |
2784 | continuing. | |
2785 | ||
2786 | Bascktrack state structs are allocated in slabs of about 4K in size. | |
2787 | PL_regmatch_state and st always point to the currently active state, | |
2788 | and PL_regmatch_slab points to the slab currently containing | |
2789 | PL_regmatch_state. The first time regmatch() is called, the first slab is | |
2790 | allocated, and is never freed until interpreter destruction. When the slab | |
2791 | is full, a new one is allocated and chained to the end. At exit from | |
2792 | regmatch(), slabs allocated since entry are freed. | |
2793 | ||
2794 | */ | |
95b24440 | 2795 | |
40a82448 | 2796 | |
5bc10b2c | 2797 | #define DEBUG_STATE_pp(pp) \ |
265c4333 | 2798 | DEBUG_STATE_r({ \ |
f2ed9b32 | 2799 | DUMP_EXEC_POS(locinput, scan, utf8_target); \ |
5bc10b2c | 2800 | PerlIO_printf(Perl_debug_log, \ |
5d458dd8 | 2801 | " %*s"pp" %s%s%s%s%s\n", \ |
5bc10b2c | 2802 | depth*2, "", \ |
13d6edb4 | 2803 | PL_reg_name[st->resume_state], \ |
5d458dd8 YO |
2804 | ((st==yes_state||st==mark_state) ? "[" : ""), \ |
2805 | ((st==yes_state) ? "Y" : ""), \ | |
2806 | ((st==mark_state) ? "M" : ""), \ | |
2807 | ((st==yes_state||st==mark_state) ? "]" : "") \ | |
2808 | ); \ | |
265c4333 | 2809 | }); |
5bc10b2c | 2810 | |
40a82448 | 2811 | |
3dab1dad | 2812 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 2813 | |
3df15adc | 2814 | #ifdef DEBUGGING |
5bc10b2c | 2815 | |
ab3bbdeb | 2816 | STATIC void |
f2ed9b32 | 2817 | S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target, |
ab3bbdeb YO |
2818 | const char *start, const char *end, const char *blurb) |
2819 | { | |
efd26800 | 2820 | const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; |
7918f24d NC |
2821 | |
2822 | PERL_ARGS_ASSERT_DEBUG_START_MATCH; | |
2823 | ||
ab3bbdeb YO |
2824 | if (!PL_colorset) |
2825 | reginitcolors(); | |
2826 | { | |
2827 | RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), | |
d2c6dc5e | 2828 | RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); |
ab3bbdeb | 2829 | |
f2ed9b32 | 2830 | RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb YO |
2831 | start, end - start, 60); |
2832 | ||
2833 | PerlIO_printf(Perl_debug_log, | |
2834 | "%s%s REx%s %s against %s\n", | |
2835 | PL_colors[4], blurb, PL_colors[5], s0, s1); | |
2836 | ||
f2ed9b32 | 2837 | if (utf8_target||utf8_pat) |
1de06328 YO |
2838 | PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", |
2839 | utf8_pat ? "pattern" : "", | |
f2ed9b32 KW |
2840 | utf8_pat && utf8_target ? " and " : "", |
2841 | utf8_target ? "string" : "" | |
ab3bbdeb YO |
2842 | ); |
2843 | } | |
2844 | } | |
3df15adc YO |
2845 | |
2846 | STATIC void | |
786e8c11 YO |
2847 | S_dump_exec_pos(pTHX_ const char *locinput, |
2848 | const regnode *scan, | |
2849 | const char *loc_regeol, | |
2850 | const char *loc_bostr, | |
2851 | const char *loc_reg_starttry, | |
f2ed9b32 | 2852 | const bool utf8_target) |
07be1b83 | 2853 | { |
786e8c11 | 2854 | const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; |
07be1b83 | 2855 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ |
786e8c11 | 2856 | int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); |
07be1b83 YO |
2857 | /* The part of the string before starttry has one color |
2858 | (pref0_len chars), between starttry and current | |
2859 | position another one (pref_len - pref0_len chars), | |
2860 | after the current position the third one. | |
2861 | We assume that pref0_len <= pref_len, otherwise we | |
2862 | decrease pref0_len. */ | |
786e8c11 YO |
2863 | int pref_len = (locinput - loc_bostr) > (5 + taill) - l |
2864 | ? (5 + taill) - l : locinput - loc_bostr; | |
07be1b83 YO |
2865 | int pref0_len; |
2866 | ||
7918f24d NC |
2867 | PERL_ARGS_ASSERT_DUMP_EXEC_POS; |
2868 | ||
f2ed9b32 | 2869 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) |
07be1b83 | 2870 | pref_len++; |
786e8c11 YO |
2871 | pref0_len = pref_len - (locinput - loc_reg_starttry); |
2872 | if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) | |
2873 | l = ( loc_regeol - locinput > (5 + taill) - pref_len | |
2874 | ? (5 + taill) - pref_len : loc_regeol - locinput); | |
f2ed9b32 | 2875 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) |
07be1b83 YO |
2876 | l--; |
2877 | if (pref0_len < 0) | |
2878 | pref0_len = 0; | |
2879 | if (pref0_len > pref_len) | |
2880 | pref0_len = pref_len; | |
2881 | { | |
f2ed9b32 | 2882 | const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0; |
0df25f3d | 2883 | |
ab3bbdeb | 2884 | RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), |
1de06328 | 2885 | (locinput - pref_len),pref0_len, 60, 4, 5); |
0df25f3d | 2886 | |
ab3bbdeb | 2887 | RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), |
3df15adc | 2888 | (locinput - pref_len + pref0_len), |
1de06328 | 2889 | pref_len - pref0_len, 60, 2, 3); |
0df25f3d | 2890 | |
ab3bbdeb | 2891 | RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), |
1de06328 | 2892 | locinput, loc_regeol - locinput, 10, 0, 1); |
0df25f3d | 2893 | |
1de06328 | 2894 | const STRLEN tlen=len0+len1+len2; |
3df15adc | 2895 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb | 2896 | "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", |
786e8c11 | 2897 | (IV)(locinput - loc_bostr), |
07be1b83 | 2898 | len0, s0, |
07be1b83 | 2899 | len1, s1, |
07be1b83 | 2900 | (docolor ? "" : "> <"), |
07be1b83 | 2901 | len2, s2, |
f9f4320a | 2902 | (int)(tlen > 19 ? 0 : 19 - tlen), |
07be1b83 YO |
2903 | ""); |
2904 | } | |
2905 | } | |
3df15adc | 2906 | |
07be1b83 YO |
2907 | #endif |
2908 | ||
0a4db386 YO |
2909 | /* reg_check_named_buff_matched() |
2910 | * Checks to see if a named buffer has matched. The data array of | |
2911 | * buffer numbers corresponding to the buffer is expected to reside | |
2912 | * in the regexp->data->data array in the slot stored in the ARG() of | |
2913 | * node involved. Note that this routine doesn't actually care about the | |
2914 | * name, that information is not preserved from compilation to execution. | |
2915 | * Returns the index of the leftmost defined buffer with the given name | |
2916 | * or 0 if non of the buffers matched. | |
2917 | */ | |
2918 | STATIC I32 | |
7918f24d NC |
2919 | S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) |
2920 | { | |
0a4db386 | 2921 | I32 n; |
f8fc2ecf | 2922 | RXi_GET_DECL(rex,rexi); |
ad64d0ec | 2923 | SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
0a4db386 | 2924 | I32 *nums=(I32*)SvPVX(sv_dat); |
7918f24d NC |
2925 | |
2926 | PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; | |
2927 | ||
0a4db386 YO |
2928 | for ( n=0; n<SvIVX(sv_dat); n++ ) { |
2929 | if ((I32)*PL_reglastparen >= nums[n] && | |
f0ab9afb | 2930 | PL_regoffs[nums[n]].end != -1) |
0a4db386 YO |
2931 | { |
2932 | return nums[n]; | |
2933 | } | |
2934 | } | |
2935 | return 0; | |
2936 | } | |
2937 | ||
2f554ef7 DM |
2938 | |
2939 | /* free all slabs above current one - called during LEAVE_SCOPE */ | |
2940 | ||
2941 | STATIC void | |
2942 | S_clear_backtrack_stack(pTHX_ void *p) | |
2943 | { | |
2944 | regmatch_slab *s = PL_regmatch_slab->next; | |
2945 | PERL_UNUSED_ARG(p); | |
2946 | ||
2947 | if (!s) | |
2948 | return; | |
2949 | PL_regmatch_slab->next = NULL; | |
2950 | while (s) { | |
2951 | regmatch_slab * const osl = s; | |
2952 | s = s->next; | |
2953 | Safefree(osl); | |
2954 | } | |
2955 | } | |
2956 | ||
2957 | ||
28d8d7f4 YO |
2958 | #define SETREX(Re1,Re2) \ |
2959 | if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \ | |
2960 | Re1 = (Re2) | |
2961 | ||
d6a28714 | 2962 | STATIC I32 /* 0 failure, 1 success */ |
24b23f37 | 2963 | S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) |
d6a28714 | 2964 | { |
a35a87e7 | 2965 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2966 | dMY_CXT; |
2967 | #endif | |
27da23d5 | 2968 | dVAR; |
f2ed9b32 | 2969 | register const bool utf8_target = PL_reg_match_utf8; |
4ad0818d | 2970 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
288b8c02 NC |
2971 | REGEXP *rex_sv = reginfo->prog; |
2972 | regexp *rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 2973 | RXi_GET_DECL(rex,rexi); |
2f554ef7 | 2974 | I32 oldsave; |
5d9a96ca DM |
2975 | /* the current state. This is a cached copy of PL_regmatch_state */ |
2976 | register regmatch_state *st; | |
5d9a96ca DM |
2977 | /* cache heavy used fields of st in registers */ |
2978 | register regnode *scan; | |
2979 | register regnode *next; | |
438e9bae | 2980 | register U32 n = 0; /* general value; init to avoid compiler warning */ |
24d3c4a9 | 2981 | register I32 ln = 0; /* len or last; init to avoid compiler warning */ |
5d9a96ca | 2982 | register char *locinput = PL_reginput; |
5d9a96ca | 2983 | register I32 nextchr; /* is always set to UCHARAT(locinput) */ |
24d3c4a9 | 2984 | |
b69b0499 | 2985 | bool result = 0; /* return value of S_regmatch */ |
24d3c4a9 | 2986 | int depth = 0; /* depth of backtrack stack */ |
4b196cd4 YO |
2987 | U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */ |
2988 | const U32 max_nochange_depth = | |
2989 | (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? | |
2990 | 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; | |
77cb431f DM |
2991 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
2992 | subpattern */ | |
e2e6a0f1 YO |
2993 | /* mark_state piggy backs on the yes_state logic so that when we unwind |
2994 | the stack on success we can update the mark_state as we go */ | |
2995 | regmatch_state *mark_state = NULL; /* last mark state we have seen */ | |
faec1544 | 2996 | regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ |
b8591aee | 2997 | struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ |
40a82448 | 2998 | U32 state_num; |
5d458dd8 YO |
2999 | bool no_final = 0; /* prevent failure from backtracking? */ |
3000 | bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ | |
e2e6a0f1 | 3001 | char *startpoint = PL_reginput; |
5d458dd8 YO |
3002 | SV *popmark = NULL; /* are we looking for a mark? */ |
3003 | SV *sv_commit = NULL; /* last mark name seen in failure */ | |
3004 | SV *sv_yes_mark = NULL; /* last mark name we have seen | |
3005 | during a successfull match */ | |
3006 | U32 lastopen = 0; /* last open we saw */ | |
3007 | bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; | |
19b95bf0 | 3008 | SV* const oreplsv = GvSV(PL_replgv); |
24d3c4a9 DM |
3009 | /* these three flags are set by various ops to signal information to |
3010 | * the very next op. They have a useful lifetime of exactly one loop | |
3011 | * iteration, and are not preserved or restored by state pushes/pops | |
3012 | */ | |
3013 | bool sw = 0; /* the condition value in (?(cond)a|b) */ | |
3014 | bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ | |
3015 | int logical = 0; /* the following EVAL is: | |
3016 | 0: (?{...}) | |
3017 | 1: (?(?{...})X|Y) | |
3018 | 2: (??{...}) | |
3019 | or the following IFMATCH/UNLESSM is: | |
3020 | false: plain (?=foo) | |
3021 | true: used as a condition: (?(?=foo)) | |
3022 | */ | |
95b24440 | 3023 | #ifdef DEBUGGING |
e68ec53f | 3024 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
3025 | #endif |
3026 | ||
7918f24d NC |
3027 | PERL_ARGS_ASSERT_REGMATCH; |
3028 | ||
3b57cd43 | 3029 | DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ |
24b23f37 | 3030 | PerlIO_printf(Perl_debug_log,"regmatch start\n"); |
3b57cd43 | 3031 | })); |
5d9a96ca DM |
3032 | /* on first ever call to regmatch, allocate first slab */ |
3033 | if (!PL_regmatch_slab) { | |
3034 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
3035 | PL_regmatch_slab->prev = NULL; | |
3036 | PL_regmatch_slab->next = NULL; | |
86545054 | 3037 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); |
5d9a96ca DM |
3038 | } |
3039 | ||
2f554ef7 DM |
3040 | oldsave = PL_savestack_ix; |
3041 | SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL); | |
3042 | SAVEVPTR(PL_regmatch_slab); | |
3043 | SAVEVPTR(PL_regmatch_state); | |
5d9a96ca DM |
3044 | |
3045 | /* grab next free state slot */ | |
3046 | st = ++PL_regmatch_state; | |
86545054 | 3047 | if (st > SLAB_LAST(PL_regmatch_slab)) |
5d9a96ca DM |
3048 | st = PL_regmatch_state = S_push_slab(aTHX); |
3049 | ||
d6a28714 JH |
3050 | /* Note that nextchr is a byte even in UTF */ |
3051 | nextchr = UCHARAT(locinput); | |
3052 | scan = prog; | |
3053 | while (scan != NULL) { | |
8ba1375e | 3054 | |
a3621e74 | 3055 | DEBUG_EXECUTE_r( { |
6136c704 | 3056 | SV * const prop = sv_newmortal(); |
1de06328 | 3057 | regnode *rnext=regnext(scan); |
f2ed9b32 | 3058 | DUMP_EXEC_POS( locinput, scan, utf8_target ); |
32fc9b6a | 3059 | regprop(rex, prop, scan); |
07be1b83 YO |
3060 | |
3061 | PerlIO_printf(Perl_debug_log, | |
3062 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
f8fc2ecf | 3063 | (IV)(scan - rexi->program), depth*2, "", |
07be1b83 | 3064 | SvPVX_const(prop), |
1de06328 | 3065 | (PL_regkind[OP(scan)] == END || !rnext) ? |
f8fc2ecf | 3066 | 0 : (IV)(rnext - rexi->program)); |
2a782b5b | 3067 | }); |
d6a28714 JH |
3068 | |
3069 | next = scan + NEXT_OFF(scan); | |
3070 | if (next == scan) | |
3071 | next = NULL; | |
40a82448 | 3072 | state_num = OP(scan); |
d6a28714 | 3073 | |
40a82448 | 3074 | reenter_switch: |
34a81e2b B |
3075 | |
3076 | assert(PL_reglastparen == &rex->lastparen); | |
3077 | assert(PL_reglastcloseparen == &rex->lastcloseparen); | |
3078 | assert(PL_regoffs == rex->offs); | |
3079 | ||
40a82448 | 3080 | switch (state_num) { |
d6a28714 | 3081 | case BOL: |
7fba1cd6 | 3082 | if (locinput == PL_bostr) |
d6a28714 | 3083 | { |
3b0527fe | 3084 | /* reginfo->till = reginfo->bol; */ |
b8c5462f JH |
3085 | break; |
3086 | } | |
d6a28714 JH |
3087 | sayNO; |
3088 | case MBOL: | |
12d33761 HS |
3089 | if (locinput == PL_bostr || |
3090 | ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n')) | |
d6a28714 | 3091 | { |
b8c5462f JH |
3092 | break; |
3093 | } | |
d6a28714 JH |
3094 | sayNO; |
3095 | case SBOL: | |
c2a73568 | 3096 | if (locinput == PL_bostr) |
b8c5462f | 3097 | break; |
d6a28714 JH |
3098 | sayNO; |
3099 | case GPOS: | |
3b0527fe | 3100 | if (locinput == reginfo->ganch) |
d6a28714 JH |
3101 | break; |
3102 | sayNO; | |
ee9b8eae YO |
3103 | |
3104 | case KEEPS: | |
3105 | /* update the startpoint */ | |
f0ab9afb | 3106 | st->u.keeper.val = PL_regoffs[0].start; |
ee9b8eae | 3107 | PL_reginput = locinput; |
f0ab9afb | 3108 | PL_regoffs[0].start = locinput - PL_bostr; |
ee9b8eae YO |
3109 | PUSH_STATE_GOTO(KEEPS_next, next); |
3110 | /*NOT-REACHED*/ | |
3111 | case KEEPS_next_fail: | |
3112 | /* rollback the start point change */ | |
f0ab9afb | 3113 | PL_regoffs[0].start = st->u.keeper.val; |
ee9b8eae YO |
3114 | sayNO_SILENT; |
3115 | /*NOT-REACHED*/ | |
d6a28714 | 3116 | case EOL: |
d6a28714 JH |
3117 | goto seol; |
3118 | case MEOL: | |
d6a28714 | 3119 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') |
b8c5462f | 3120 | sayNO; |
b8c5462f | 3121 | break; |
d6a28714 JH |
3122 | case SEOL: |
3123 | seol: | |
3124 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') | |
b8c5462f | 3125 | sayNO; |
d6a28714 | 3126 | if (PL_regeol - locinput > 1) |
b8c5462f | 3127 | sayNO; |
b8c5462f | 3128 | break; |
d6a28714 JH |
3129 | case EOS: |
3130 | if (PL_regeol != locinput) | |
b8c5462f | 3131 | sayNO; |
d6a28714 | 3132 | break; |
ffc61ed2 | 3133 | case SANY: |
d6a28714 | 3134 | if (!nextchr && locinput >= PL_regeol) |
4633a7c4 | 3135 | sayNO; |
f2ed9b32 | 3136 | if (utf8_target) { |
f33976b4 DB |
3137 | locinput += PL_utf8skip[nextchr]; |
3138 | if (locinput > PL_regeol) | |
3139 | sayNO; | |
3140 | nextchr = UCHARAT(locinput); | |
3141 | } | |
3142 | else | |
3143 | nextchr = UCHARAT(++locinput); | |
3144 | break; | |
3145 | case CANY: | |
3146 | if (!nextchr && locinput >= PL_regeol) | |
3147 | sayNO; | |
b8c5462f | 3148 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 3149 | break; |
ffc61ed2 | 3150 | case REG_ANY: |
1aa99e6b IH |
3151 | if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n') |
3152 | sayNO; | |
f2ed9b32 | 3153 | if (utf8_target) { |
b8c5462f | 3154 | locinput += PL_utf8skip[nextchr]; |
d6a28714 JH |
3155 | if (locinput > PL_regeol) |
3156 | sayNO; | |
a0ed51b3 | 3157 | nextchr = UCHARAT(locinput); |
a0ed51b3 | 3158 | } |
1aa99e6b IH |
3159 | else |
3160 | nextchr = UCHARAT(++locinput); | |
a0ed51b3 | 3161 | break; |
166ba7cd DM |
3162 | |
3163 | #undef ST | |
3164 | #define ST st->u.trie | |
786e8c11 YO |
3165 | case TRIEC: |
3166 | /* In this case the charclass data is available inline so | |
3167 | we can fail fast without a lot of extra overhead. | |
3168 | */ | |
f2ed9b32 | 3169 | if (scan->flags == EXACT || !utf8_target) { |
786e8c11 YO |
3170 | if(!ANYOF_BITMAP_TEST(scan, *locinput)) { |
3171 | DEBUG_EXECUTE_r( | |
3172 | PerlIO_printf(Perl_debug_log, | |
3173 | "%*s %sfailed to match trie start class...%s\n", | |
5bc10b2c | 3174 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
786e8c11 YO |
3175 | ); |
3176 | sayNO_SILENT; | |
3177 | /* NOTREACHED */ | |
3178 | } | |
3179 | } | |
3180 | /* FALL THROUGH */ | |
5b47454d | 3181 | case TRIE: |
2e64971a DM |
3182 | /* the basic plan of execution of the trie is: |
3183 | * At the beginning, run though all the states, and | |
3184 | * find the longest-matching word. Also remember the position | |
3185 | * of the shortest matching word. For example, this pattern: | |
3186 | * 1 2 3 4 5 | |
3187 | * ab|a|x|abcd|abc | |
3188 | * when matched against the string "abcde", will generate | |
3189 | * accept states for all words except 3, with the longest | |
3190 | * matching word being 4, and the shortest being 1 (with | |
3191 | * the position being after char 1 of the string). | |
3192 | * | |
3193 | * Then for each matching word, in word order (i.e. 1,2,4,5), | |
3194 | * we run the remainder of the pattern; on each try setting | |
3195 | * the current position to the character following the word, | |
3196 | * returning to try the next word on failure. | |
3197 | * | |
3198 | * We avoid having to build a list of words at runtime by | |
3199 | * using a compile-time structure, wordinfo[].prev, which | |
3200 | * gives, for each word, the previous accepting word (if any). | |
3201 | * In the case above it would contain the mappings 1->2, 2->0, | |
3202 | * 3->0, 4->5, 5->1. We can use this table to generate, from | |
3203 | * the longest word (4 above), a list of all words, by | |
3204 | * following the list of prev pointers; this gives us the | |
3205 | * unordered list 4,5,1,2. Then given the current word we have | |
3206 | * just tried, we can go through the list and find the | |
3207 | * next-biggest word to try (so if we just failed on word 2, | |
3208 | * the next in the list is 4). | |
3209 | * | |
3210 | * Since at runtime we don't record the matching position in | |
3211 | * the string for each word, we have to work that out for | |
3212 | * each word we're about to process. The wordinfo table holds | |
3213 | * the character length of each word; given that we recorded | |
3214 | * at the start: the position of the shortest word and its | |
3215 | * length in chars, we just need to move the pointer the | |
3216 | * difference between the two char lengths. Depending on | |
3217 | * Unicode status and folding, that's cheap or expensive. | |
3218 | * | |
3219 | * This algorithm is optimised for the case where are only a | |
3220 | * small number of accept states, i.e. 0,1, or maybe 2. | |
3221 | * With lots of accepts states, and having to try all of them, | |
3222 | * it becomes quadratic on number of accept states to find all | |
3223 | * the next words. | |
3224 | */ | |
3225 | ||
3dab1dad | 3226 | { |
07be1b83 | 3227 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
a0a388a1 | 3228 | DECL_TRIE_TYPE(scan); |
3dab1dad YO |
3229 | |
3230 | /* what trie are we using right now */ | |
be8e71aa | 3231 | reg_trie_data * const trie |
f8fc2ecf | 3232 | = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; |
85fbaab2 | 3233 | HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); |
3dab1dad | 3234 | U32 state = trie->startstate; |
166ba7cd | 3235 | |
3dab1dad YO |
3236 | if (trie->bitmap && trie_type != trie_utf8_fold && |
3237 | !TRIE_BITMAP_TEST(trie,*locinput) | |
3238 | ) { | |
3239 | if (trie->states[ state ].wordnum) { | |
3240 | DEBUG_EXECUTE_r( | |
3241 | PerlIO_printf(Perl_debug_log, | |
3242 | "%*s %smatched empty string...%s\n", | |
5bc10b2c | 3243 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad | 3244 | ); |
20dbff7c YO |
3245 | if (!trie->jump) |
3246 | break; | |
3dab1dad YO |
3247 | } else { |
3248 | DEBUG_EXECUTE_r( | |
3249 | PerlIO_printf(Perl_debug_log, | |
786e8c11 | 3250 | "%*s %sfailed to match trie start class...%s\n", |
5bc10b2c | 3251 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad YO |
3252 | ); |
3253 | sayNO_SILENT; | |
3254 | } | |
3255 | } | |
166ba7cd | 3256 | |
786e8c11 YO |
3257 | { |
3258 | U8 *uc = ( U8* )locinput; | |
3259 | ||
3260 | STRLEN len = 0; | |
3261 | STRLEN foldlen = 0; | |
3262 | U8 *uscan = (U8*)NULL; | |
786e8c11 | 3263 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
2e64971a DM |
3264 | U32 charcount = 0; /* how many input chars we have matched */ |
3265 | U32 accepted = 0; /* have we seen any accepting states? */ | |
786e8c11 | 3266 | |
786e8c11 YO |
3267 | ST.B = next; |
3268 | ST.jump = trie->jump; | |
786e8c11 | 3269 | ST.me = scan; |
2e64971a DM |
3270 | ST.firstpos = NULL; |
3271 | ST.longfold = FALSE; /* char longer if folded => it's harder */ | |
3272 | ST.nextword = 0; | |
3273 | ||
3274 | /* fully traverse the TRIE; note the position of the | |
3275 | shortest accept state and the wordnum of the longest | |
3276 | accept state */ | |
07be1b83 | 3277 | |
a3621e74 | 3278 | while ( state && uc <= (U8*)PL_regeol ) { |
786e8c11 | 3279 | U32 base = trie->states[ state ].trans.base; |
f9f4320a | 3280 | UV uvc = 0; |
acb909b4 | 3281 | U16 charid = 0; |
2e64971a DM |
3282 | U16 wordnum; |
3283 | wordnum = trie->states[ state ].wordnum; | |
3284 | ||
3285 | if (wordnum) { /* it's an accept state */ | |
3286 | if (!accepted) { | |
3287 | accepted = 1; | |
3288 | /* record first match position */ | |
3289 | if (ST.longfold) { | |
3290 | ST.firstpos = (U8*)locinput; | |
3291 | ST.firstchars = 0; | |
5b47454d | 3292 | } |
2e64971a DM |
3293 | else { |
3294 | ST.firstpos = uc; | |
3295 | ST.firstchars = charcount; | |
3296 | } | |
3297 | } | |
3298 | if (!ST.nextword || wordnum < ST.nextword) | |
3299 | ST.nextword = wordnum; | |
3300 | ST.topword = wordnum; | |
786e8c11 | 3301 | } |
a3621e74 | 3302 | |
07be1b83 | 3303 | DEBUG_TRIE_EXECUTE_r({ |
f2ed9b32 | 3304 | DUMP_EXEC_POS( (char *)uc, scan, utf8_target ); |
a3621e74 | 3305 | PerlIO_printf( Perl_debug_log, |
2e64971a | 3306 | "%*s %sState: %4"UVxf" Accepted: %c ", |
5bc10b2c | 3307 | 2+depth * 2, "", PL_colors[4], |
2e64971a | 3308 | (UV)state, (accepted ? 'Y' : 'N')); |
07be1b83 | 3309 | }); |
a3621e74 | 3310 | |
2e64971a | 3311 | /* read a char and goto next state */ |
a3621e74 | 3312 | if ( base ) { |
6dd2be57 | 3313 | I32 offset; |
55eed653 NC |
3314 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
3315 | uscan, len, uvc, charid, foldlen, | |
3316 | foldbuf, uniflags); | |
2e64971a DM |
3317 | charcount++; |
3318 | if (foldlen>0) | |
3319 | ST.longfold = TRUE; | |
5b47454d | 3320 | if (charid && |
6dd2be57 DM |
3321 | ( ((offset = |
3322 | base + charid - 1 - trie->uniquecharcount)) >= 0) | |
3323 | ||
3324 | && ((U32)offset < trie->lasttrans) | |
3325 | && trie->trans[offset].check == state) | |
5b47454d | 3326 | { |
6dd2be57 | 3327 | state = trie->trans[offset].next; |
5b47454d DM |
3328 | } |
3329 | else { | |
3330 | state = 0; | |
3331 | } | |
3332 | uc += len; | |
3333 | ||
3334 | } | |
3335 | else { | |
a3621e74 YO |
3336 | state = 0; |
3337 | } | |
3338 | DEBUG_TRIE_EXECUTE_r( | |
e4584336 | 3339 | PerlIO_printf( Perl_debug_log, |
786e8c11 | 3340 | "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n", |
e4584336 | 3341 | charid, uvc, (UV)state, PL_colors[5] ); |
a3621e74 YO |
3342 | ); |
3343 | } | |
2e64971a | 3344 | if (!accepted) |
a3621e74 | 3345 | sayNO; |
a3621e74 | 3346 | |
2e64971a DM |
3347 | /* calculate total number of accept states */ |
3348 | { | |
3349 | U16 w = ST.topword; | |
3350 | accepted = 0; | |
3351 | while (w) { | |
3352 | w = trie->wordinfo[w].prev; | |
3353 | accepted++; | |
3354 | } | |
3355 | ST.accepted = accepted; | |
3356 | } | |
3357 | ||
166ba7cd DM |
3358 | DEBUG_EXECUTE_r( |
3359 | PerlIO_printf( Perl_debug_log, | |
3360 | "%*s %sgot %"IVdf" possible matches%s\n", | |
5bc10b2c | 3361 | REPORT_CODE_OFF + depth * 2, "", |
166ba7cd DM |
3362 | PL_colors[4], (IV)ST.accepted, PL_colors[5] ); |
3363 | ); | |
2e64971a | 3364 | goto trie_first_try; /* jump into the fail handler */ |
786e8c11 | 3365 | }} |
fae667d5 | 3366 | /* NOTREACHED */ |
2e64971a DM |
3367 | |
3368 | case TRIE_next_fail: /* we failed - try next alternative */ | |
fae667d5 YO |
3369 | if ( ST.jump) { |
3370 | REGCP_UNWIND(ST.cp); | |
3371 | for (n = *PL_reglastparen; n > ST.lastparen; n--) | |
f0ab9afb | 3372 | PL_regoffs[n].end = -1; |
fae667d5 YO |
3373 | *PL_reglastparen = n; |
3374 | } | |
2e64971a DM |
3375 | if (!--ST.accepted) { |
3376 | DEBUG_EXECUTE_r({ | |
3377 | PerlIO_printf( Perl_debug_log, | |
3378 | "%*s %sTRIE failed...%s\n", | |
3379 | REPORT_CODE_OFF+depth*2, "", | |
3380 | PL_colors[4], | |
3381 | PL_colors[5] ); | |
3382 | }); | |
3383 | sayNO_SILENT; | |
3384 | } | |
3385 | { | |
3386 | /* Find next-highest word to process. Note that this code | |
3387 | * is O(N^2) per trie run (O(N) per branch), so keep tight */ | |
d9a396a3 DM |
3388 | register U16 min = 0; |
3389 | register U16 word; | |
2e64971a DM |
3390 | register U16 const nextword = ST.nextword; |
3391 | register reg_trie_wordinfo * const wordinfo | |
3392 | = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo; | |
3393 | for (word=ST.topword; word; word=wordinfo[word].prev) { | |
3394 | if (word > nextword && (!min || word < min)) | |
3395 | min = word; | |
3396 | } | |
3397 | ST.nextword = min; | |
3398 | } | |
3399 | ||
fae667d5 | 3400 | trie_first_try: |
5d458dd8 YO |
3401 | if (do_cutgroup) { |
3402 | do_cutgroup = 0; | |
3403 | no_final = 0; | |
3404 | } | |
fae667d5 YO |
3405 | |
3406 | if ( ST.jump) { | |
3407 | ST.lastparen = *PL_reglastparen; | |
3408 | REGCP_SET(ST.cp); | |
2e64971a | 3409 | } |
a3621e74 | 3410 | |
2e64971a | 3411 | /* find start char of end of current word */ |
166ba7cd | 3412 | { |
2e64971a DM |
3413 | U32 chars; /* how many chars to skip */ |
3414 | U8 *uc = ST.firstpos; | |
3415 | reg_trie_data * const trie | |
3416 | = (reg_trie_data*)rexi->data->data[ARG(ST.me)]; | |
3417 | ||
3418 | assert((trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3419 | >= ST.firstchars); | |
3420 | chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3421 | - ST.firstchars; | |
3422 | ||
3423 | if (ST.longfold) { | |
3424 | /* the hard option - fold each char in turn and find | |
3425 | * its folded length (which may be different */ | |
3426 | U8 foldbuf[UTF8_MAXBYTES_CASE + 1]; | |
3427 | STRLEN foldlen; | |
3428 | STRLEN len; | |
d9a396a3 | 3429 | UV uvc; |
2e64971a DM |
3430 | U8 *uscan; |
3431 | ||
3432 | while (chars) { | |
f2ed9b32 | 3433 | if (utf8_target) { |
2e64971a DM |
3434 | uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len, |
3435 | uniflags); | |
3436 | uc += len; | |
3437 | } | |
3438 | else { | |
3439 | uvc = *uc; | |
3440 | uc++; | |
3441 | } | |
3442 | uvc = to_uni_fold(uvc, foldbuf, &foldlen); | |
3443 | uscan = foldbuf; | |
3444 | while (foldlen) { | |
3445 | if (!--chars) | |
3446 | break; | |
3447 | uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len, | |
3448 | uniflags); | |
3449 | uscan += len; | |
3450 | foldlen -= len; | |
3451 | } | |
3452 | } | |
a3621e74 | 3453 | } |
2e64971a | 3454 | else { |
f2ed9b32 | 3455 | if (utf8_target) |
2e64971a DM |
3456 | while (chars--) |
3457 | uc += UTF8SKIP(uc); | |
3458 | else | |
3459 | uc += chars; | |
3460 | } | |
3461 | PL_reginput = (char *)uc; | |
3462 | } | |
166ba7cd | 3463 | |
2e64971a DM |
3464 | scan = (ST.jump && ST.jump[ST.nextword]) |
3465 | ? ST.me + ST.jump[ST.nextword] | |
3466 | : ST.B; | |
166ba7cd | 3467 | |
2e64971a DM |
3468 | DEBUG_EXECUTE_r({ |
3469 | PerlIO_printf( Perl_debug_log, | |
3470 | "%*s %sTRIE matched word #%d, continuing%s\n", | |
3471 | REPORT_CODE_OFF+depth*2, "", | |
3472 | PL_colors[4], | |
3473 | ST.nextword, | |
3474 | PL_colors[5] | |
3475 | ); | |
3476 | }); | |
3477 | ||
3478 | if (ST.accepted > 1 || has_cutgroup) { | |
3479 | PUSH_STATE_GOTO(TRIE_next, scan); | |
3480 | /* NOTREACHED */ | |
166ba7cd | 3481 | } |
2e64971a DM |
3482 | /* only one choice left - just continue */ |
3483 | DEBUG_EXECUTE_r({ | |
3484 | AV *const trie_words | |
3485 | = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]); | |
3486 | SV ** const tmp = av_fetch( trie_words, | |
3487 | ST.nextword-1, 0 ); | |
3488 | SV *sv= tmp ? sv_newmortal() : NULL; | |
3489 | ||
3490 | PerlIO_printf( Perl_debug_log, | |
3491 | "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n", | |
3492 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], | |
3493 | ST.nextword, | |
3494 | tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0, | |
3495 | PL_colors[0], PL_colors[1], | |
c89df6cf | 3496 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)|PERL_PV_ESCAPE_NONASCII |
2e64971a DM |
3497 | ) |
3498 | : "not compiled under -Dr", | |
3499 | PL_colors[5] ); | |
3500 | }); | |
3501 | ||
3502 | locinput = PL_reginput; | |
3503 | nextchr = UCHARAT(locinput); | |
3504 | continue; /* execute rest of RE */ | |
166ba7cd | 3505 | /* NOTREACHED */ |
166ba7cd DM |
3506 | #undef ST |
3507 | ||
95b24440 DM |
3508 | case EXACT: { |
3509 | char *s = STRING(scan); | |
24d3c4a9 | 3510 | ln = STR_LEN(scan); |
f2ed9b32 | 3511 | if (utf8_target != UTF_PATTERN) { |
bc517b45 | 3512 | /* The target and the pattern have differing utf8ness. */ |
1aa99e6b | 3513 | char *l = locinput; |
24d3c4a9 | 3514 | const char * const e = s + ln; |
a72c7584 | 3515 | |
f2ed9b32 | 3516 | if (utf8_target) { |
5ff6fc6d | 3517 | /* The target is utf8, the pattern is not utf8. */ |
1aa99e6b | 3518 | while (s < e) { |
a3b680e6 | 3519 | STRLEN ulen; |
1aa99e6b | 3520 | if (l >= PL_regeol) |
5ff6fc6d JH |
3521 | sayNO; |
3522 | if (NATIVE_TO_UNI(*(U8*)s) != | |
89ebb4a3 | 3523 | utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen, |
041457d9 | 3524 | uniflags)) |
5ff6fc6d | 3525 | sayNO; |
bc517b45 | 3526 | l += ulen; |
5ff6fc6d | 3527 | s ++; |
1aa99e6b | 3528 | } |
5ff6fc6d JH |
3529 | } |
3530 | else { | |
3531 | /* The target is not utf8, the pattern is utf8. */ | |
1aa99e6b | 3532 | while (s < e) { |
a3b680e6 | 3533 | STRLEN ulen; |
1aa99e6b IH |
3534 | if (l >= PL_regeol) |
3535 | sayNO; | |
5ff6fc6d | 3536 | if (NATIVE_TO_UNI(*((U8*)l)) != |
89ebb4a3 | 3537 | utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen, |
041457d9 | 3538 | uniflags)) |
1aa99e6b | 3539 | sayNO; |
bc517b45 | 3540 | s += ulen; |
a72c7584 | 3541 | l ++; |
1aa99e6b | 3542 | } |
5ff6fc6d | 3543 | } |
1aa99e6b IH |
3544 | locinput = l; |
3545 | nextchr = UCHARAT(locinput); | |
3546 | break; | |
3547 | } | |
bc517b45 | 3548 | /* The target and the pattern have the same utf8ness. */ |
d6a28714 JH |
3549 | /* Inline the first character, for speed. */ |
3550 | if (UCHARAT(s) != nextchr) | |
3551 | sayNO; | |
24d3c4a9 | 3552 | if (PL_regeol - locinput < ln) |
d6a28714 | 3553 | sayNO; |
24d3c4a9 | 3554 | if (ln > 1 && memNE(s, locinput, ln)) |
d6a28714 | 3555 | sayNO; |
24d3c4a9 | 3556 | locinput += ln; |
d6a28714 JH |
3557 | nextchr = UCHARAT(locinput); |
3558 | break; | |
95b24440 | 3559 | } |
9a5a5549 | 3560 | case EXACTFL: { |
a932d541 | 3561 | re_fold_t folder; |
9a5a5549 KW |
3562 | const U8 * fold_array; |
3563 | const char * s; | |
3564 | ||
b8c5462f | 3565 | PL_reg_flags |= RF_tainted; |
9a5a5549 KW |
3566 | folder = foldEQ_locale; |
3567 | fold_array = PL_fold_locale; | |
3568 | goto do_exactf; | |
3569 | ||
3570 | case EXACTFU: | |
3571 | folder = foldEQ_latin1; | |
3572 | fold_array = PL_fold_latin1; | |
3573 | goto do_exactf; | |
3574 | ||
3575 | case EXACTF: | |
3576 | folder = foldEQ; | |
3577 | fold_array = PL_fold; | |
3578 | ||
3579 | do_exactf: | |
3580 | s = STRING(scan); | |
24d3c4a9 | 3581 | ln = STR_LEN(scan); |
d6a28714 | 3582 | |
f2ed9b32 | 3583 | if (utf8_target || UTF_PATTERN) { |
d07ddd77 | 3584 | /* Either target or the pattern are utf8. */ |
be8e71aa | 3585 | const char * const l = locinput; |
d07ddd77 | 3586 | char *e = PL_regeol; |
bc517b45 | 3587 | |
f2ed9b32 KW |
3588 | if (! foldEQ_utf8(s, 0, ln, cBOOL(UTF_PATTERN), |
3589 | l, &e, 0, utf8_target)) { | |
5486206c JH |
3590 | /* One more case for the sharp s: |
3591 | * pack("U0U*", 0xDF) =~ /ss/i, | |
3592 | * the 0xC3 0x9F are the UTF-8 | |
3593 | * byte sequence for the U+00DF. */ | |
e1d1eefb | 3594 | |
f2ed9b32 | 3595 | if (!(utf8_target && |
e1d1eefb | 3596 | toLOWER(s[0]) == 's' && |
24d3c4a9 | 3597 | ln >= 2 && |
5486206c JH |
3598 | toLOWER(s[1]) == 's' && |
3599 | (U8)l[0] == 0xC3 && | |
3600 | e - l >= 2 && | |
3601 | (U8)l[1] == 0x9F)) | |
3602 | sayNO; | |
3603 | } | |
d07ddd77 JH |
3604 | locinput = e; |
3605 | nextchr = UCHARAT(locinput); | |
3606 | break; | |
a0ed51b3 | 3607 | } |
d6a28714 | 3608 | |
bc517b45 JH |
3609 | /* Neither the target and the pattern are utf8. */ |
3610 | ||
d6a28714 JH |
3611 | /* Inline the first character, for speed. */ |
3612 | if (UCHARAT(s) != nextchr && | |
9a5a5549 KW |
3613 | UCHARAT(s) != fold_array[nextchr]) |
3614 | { | |
a0ed51b3 | 3615 | sayNO; |
9a5a5549 | 3616 | } |
24d3c4a9 | 3617 | if (PL_regeol - locinput < ln) |
b8c5462f | 3618 | sayNO; |
9a5a5549 | 3619 | if (ln > 1 && ! folder(s, locinput, ln)) |
4633a7c4 | 3620 | sayNO; |
24d3c4a9 | 3621 | locinput += ln; |
d6a28714 | 3622 | nextchr = UCHARAT(locinput); |
a0d0e21e | 3623 | break; |
9a5a5549 | 3624 | } |
b2680017 YO |
3625 | case BOUNDL: |
3626 | case NBOUNDL: | |
3627 | PL_reg_flags |= RF_tainted; | |
3628 | /* FALL THROUGH */ | |
3629 | case BOUND: | |
3630 | case NBOUND: | |
3631 | /* was last char in word? */ | |
f2ed9b32 | 3632 | if (utf8_target) { |
b2680017 YO |
3633 | if (locinput == PL_bostr) |
3634 | ln = '\n'; | |
3635 | else { | |
3636 | const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); | |
3637 | ||
3638 | ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); | |
3639 | } | |
3640 | if (OP(scan) == BOUND || OP(scan) == NBOUND) { | |
3641 | ln = isALNUM_uni(ln); | |
3642 | LOAD_UTF8_CHARCLASS_ALNUM(); | |
f2ed9b32 | 3643 | n = swash_fetch(PL_utf8_alnum, (U8*)locinput, utf8_target); |
b2680017 YO |
3644 | } |
3645 | else { | |
3646 | ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); | |
3647 | n = isALNUM_LC_utf8((U8*)locinput); | |
3648 | } | |
3649 | } | |
3650 | else { | |
3651 | ln = (locinput != PL_bostr) ? | |
3652 | UCHARAT(locinput - 1) : '\n'; | |
a12cf05f KW |
3653 | if (FLAGS(scan) & USE_UNI) { |
3654 | ||
3655 | /* Here, can't be BOUNDL or NBOUNDL because they never set | |
3656 | * the flags to USE_UNI */ | |
3657 | ln = isWORDCHAR_L1(ln); | |
3658 | n = isWORDCHAR_L1(nextchr); | |
3659 | } | |
3660 | else if (OP(scan) == BOUND || OP(scan) == NBOUND) { | |
b2680017 YO |
3661 | ln = isALNUM(ln); |
3662 | n = isALNUM(nextchr); | |
3663 | } | |
3664 | else { | |
3665 | ln = isALNUM_LC(ln); | |
3666 | n = isALNUM_LC(nextchr); | |
3667 | } | |
3668 | } | |
3669 | if (((!ln) == (!n)) == (OP(scan) == BOUND || | |
3670 | OP(scan) == BOUNDL)) | |
3671 | sayNO; | |
3672 | break; | |
d6a28714 | 3673 | case ANYOF: |
f2ed9b32 | 3674 | if (utf8_target) { |
9e55ce06 | 3675 | STRLEN inclasslen = PL_regeol - locinput; |
20ed0b26 KW |
3676 | if (locinput >= PL_regeol) |
3677 | sayNO; | |
9e55ce06 | 3678 | |
f2ed9b32 | 3679 | if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target)) |
262b90c4 | 3680 | goto anyof_fail; |
b32d7d3e | 3681 | locinput += inclasslen; |
b8c5462f | 3682 | nextchr = UCHARAT(locinput); |
e0f9d4a8 | 3683 | break; |
ffc61ed2 JH |
3684 | } |
3685 | else { | |
3686 | if (nextchr < 0) | |
3687 | nextchr = UCHARAT(locinput); | |
ffc61ed2 JH |
3688 | if (!nextchr && locinput >= PL_regeol) |
3689 | sayNO; | |
20ed0b26 KW |
3690 | if (!REGINCLASS(rex, scan, (U8*)locinput)) |
3691 | goto anyof_fail; | |
ffc61ed2 | 3692 | nextchr = UCHARAT(++locinput); |
e0f9d4a8 JH |
3693 | break; |
3694 | } | |
262b90c4 | 3695 | anyof_fail: |
e0f9d4a8 JH |
3696 | /* If we might have the case of the German sharp s |
3697 | * in a casefolding Unicode character class. */ | |
3698 | ||
ebc501f0 JH |
3699 | if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) { |
3700 | locinput += SHARP_S_SKIP; | |
e0f9d4a8 | 3701 | nextchr = UCHARAT(locinput); |
ffc61ed2 | 3702 | } |
e0f9d4a8 JH |
3703 | else |
3704 | sayNO; | |
b8c5462f | 3705 | break; |
20d0b1e9 | 3706 | /* Special char classes - The defines start on line 129 or so */ |
a12cf05f KW |
3707 | CCC_TRY_AFF_U( ALNUM, ALNUML, perl_word, "a", isALNUM_LC_utf8, isWORDCHAR_L1, isALNUM_LC); |
3708 | CCC_TRY_NEG_U(NALNUM, NALNUML, perl_word, "a", isALNUM_LC_utf8, isWORDCHAR_L1, isALNUM_LC); | |
20d0b1e9 | 3709 | |
a12cf05f KW |
3710 | CCC_TRY_AFF_U( SPACE, SPACEL, perl_space, " ", isSPACE_LC_utf8, isSPACE_L1, isSPACE_LC); |
3711 | CCC_TRY_NEG_U(NSPACE, NSPACEL, perl_space, " ", isSPACE_LC_utf8, isSPACE_L1, isSPACE_LC); | |
20d0b1e9 | 3712 | |
d1eb3177 YO |
3713 | CCC_TRY_AFF( DIGIT, DIGITL, posix_digit, "0", isDIGIT_LC_utf8, isDIGIT, isDIGIT_LC); |
3714 | CCC_TRY_NEG(NDIGIT, NDIGITL, posix_digit, "0", isDIGIT_LC_utf8, isDIGIT, isDIGIT_LC); | |
20d0b1e9 | 3715 | |
37e2e78e KW |
3716 | case CLUMP: /* Match \X: logical Unicode character. This is defined as |
3717 | a Unicode extended Grapheme Cluster */ | |
3718 | /* From http://www.unicode.org/reports/tr29 (5.2 version). An | |
3719 | extended Grapheme Cluster is: | |
3720 | ||
3721 | CR LF | |
3722 | | Prepend* Begin Extend* | |
3723 | | . | |
3724 | ||
3725 | Begin is (Hangul-syllable | ! Control) | |
3726 | Extend is (Grapheme_Extend | Spacing_Mark) | |
3727 | Control is [ GCB_Control CR LF ] | |
3728 | ||
3729 | The discussion below shows how the code for CLUMP is derived | |
3730 | from this regex. Note that most of these concepts are from | |
3731 | property values of the Grapheme Cluster Boundary (GCB) property. | |
3732 | No code point can have multiple property values for a given | |
3733 | property. Thus a code point in Prepend can't be in Control, but | |
3734 | it must be in !Control. This is why Control above includes | |
3735 | GCB_Control plus CR plus LF. The latter two are used in the GCB | |
3736 | property separately, and so can't be in GCB_Control, even though | |
3737 | they logically are controls. Control is not the same as gc=cc, | |
3738 | but includes format and other characters as well. | |
3739 | ||
3740 | The Unicode definition of Hangul-syllable is: | |
3741 | L+ | |
3742 | | (L* ( ( V | LV ) V* | LVT ) T*) | |
3743 | | T+ | |
3744 | ) | |
3745 | Each of these is a value for the GCB property, and hence must be | |
3746 | disjoint, so the order they are tested is immaterial, so the | |
3747 | above can safely be changed to | |
3748 | T+ | |
3749 | | L+ | |
3750 | | (L* ( LVT | ( V | LV ) V*) T*) | |
3751 | ||
3752 | The last two terms can be combined like this: | |
3753 | L* ( L | |
3754 | | (( LVT | ( V | LV ) V*) T*)) | |
3755 | ||
3756 | And refactored into this: | |
3757 | L* (L | LVT T* | V V* T* | LV V* T*) | |
3758 | ||
3759 | That means that if we have seen any L's at all we can quit | |
3760 | there, but if the next character is a LVT, a V or and LV we | |
3761 | should keep going. | |
3762 | ||
3763 | There is a subtlety with Prepend* which showed up in testing. | |
3764 | Note that the Begin, and only the Begin is required in: | |
3765 | | Prepend* Begin Extend* | |
3766 | Also, Begin contains '! Control'. A Prepend must be a '! | |
3767 | Control', which means it must be a Begin. What it comes down to | |
3768 | is that if we match Prepend* and then find no suitable Begin | |
3769 | afterwards, that if we backtrack the last Prepend, that one will | |
3770 | be a suitable Begin. | |
3771 | */ | |
3772 | ||
b7c83a7e | 3773 | if (locinput >= PL_regeol) |
a0ed51b3 | 3774 | sayNO; |
f2ed9b32 | 3775 | if (! utf8_target) { |
37e2e78e KW |
3776 | |
3777 | /* Match either CR LF or '.', as all the other possibilities | |
3778 | * require utf8 */ | |
3779 | locinput++; /* Match the . or CR */ | |
3780 | if (nextchr == '\r' | |
3781 | && locinput < PL_regeol | |
3782 | && UCHARAT(locinput) == '\n') locinput++; | |
3783 | } | |
3784 | else { | |
3785 | ||
3786 | /* Utf8: See if is ( CR LF ); already know that locinput < | |
3787 | * PL_regeol, so locinput+1 is in bounds */ | |
3788 | if (nextchr == '\r' && UCHARAT(locinput + 1) == '\n') { | |
3789 | locinput += 2; | |
3790 | } | |
3791 | else { | |
3792 | /* In case have to backtrack to beginning, then match '.' */ | |
3793 | char *starting = locinput; | |
3794 | ||
3795 | /* In case have to backtrack the last prepend */ | |
3796 | char *previous_prepend = 0; | |
3797 | ||
3798 | LOAD_UTF8_CHARCLASS_GCB(); | |
3799 | ||
3800 | /* Match (prepend)* */ | |
3801 | while (locinput < PL_regeol | |
3802 | && swash_fetch(PL_utf8_X_prepend, | |
f2ed9b32 | 3803 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3804 | { |
3805 | previous_prepend = locinput; | |
3806 | locinput += UTF8SKIP(locinput); | |
3807 | } | |
3808 | ||
3809 | /* As noted above, if we matched a prepend character, but | |
3810 | * the next thing won't match, back off the last prepend we | |
3811 | * matched, as it is guaranteed to match the begin */ | |
3812 | if (previous_prepend | |
3813 | && (locinput >= PL_regeol | |
3814 | || ! swash_fetch(PL_utf8_X_begin, | |
f2ed9b32 | 3815 | (U8*)locinput, utf8_target))) |
37e2e78e KW |
3816 | { |
3817 | locinput = previous_prepend; | |
3818 | } | |
3819 | ||
3820 | /* Note that here we know PL_regeol > locinput, as we | |
3821 | * tested that upon input to this switch case, and if we | |
3822 | * moved locinput forward, we tested the result just above | |
3823 | * and it either passed, or we backed off so that it will | |
3824 | * now pass */ | |
f2ed9b32 | 3825 | if (! swash_fetch(PL_utf8_X_begin, (U8*)locinput, utf8_target)) { |
37e2e78e KW |
3826 | |
3827 | /* Here did not match the required 'Begin' in the | |
3828 | * second term. So just match the very first | |
3829 | * character, the '.' of the final term of the regex */ | |
3830 | locinput = starting + UTF8SKIP(starting); | |
3831 | } else { | |
3832 | ||
3833 | /* Here is the beginning of a character that can have | |
3834 | * an extender. It is either a hangul syllable, or a | |
3835 | * non-control */ | |
3836 | if (swash_fetch(PL_utf8_X_non_hangul, | |
f2ed9b32 | 3837 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3838 | { |
3839 | ||
3840 | /* Here not a Hangul syllable, must be a | |
3841 | * ('! * Control') */ | |
3842 | locinput += UTF8SKIP(locinput); | |
3843 | } else { | |
3844 | ||
3845 | /* Here is a Hangul syllable. It can be composed | |
3846 | * of several individual characters. One | |
3847 | * possibility is T+ */ | |
3848 | if (swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 3849 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3850 | { |
3851 | while (locinput < PL_regeol | |
3852 | && swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 3853 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3854 | { |
3855 | locinput += UTF8SKIP(locinput); | |
3856 | } | |
3857 | } else { | |
3858 | ||
3859 | /* Here, not T+, but is a Hangul. That means | |
3860 | * it is one of the others: L, LV, LVT or V, | |
3861 | * and matches: | |
3862 | * L* (L | LVT T* | V V* T* | LV V* T*) */ | |
3863 | ||
3864 | /* Match L* */ | |
3865 | while (locinput < PL_regeol | |
3866 | && swash_fetch(PL_utf8_X_L, | |
f2ed9b32 | 3867 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3868 | { |
3869 | locinput += UTF8SKIP(locinput); | |
3870 | } | |
3871 | ||
3872 | /* Here, have exhausted L*. If the next | |
3873 | * character is not an LV, LVT nor V, it means | |
3874 | * we had to have at least one L, so matches L+ | |
3875 | * in the original equation, we have a complete | |
3876 | * hangul syllable. Are done. */ | |
3877 | ||
3878 | if (locinput < PL_regeol | |
3879 | && swash_fetch(PL_utf8_X_LV_LVT_V, | |
f2ed9b32 | 3880 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3881 | { |
3882 | ||
3883 | /* Otherwise keep going. Must be LV, LVT | |
3884 | * or V. See if LVT */ | |
3885 | if (swash_fetch(PL_utf8_X_LVT, | |
f2ed9b32 | 3886 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3887 | { |
3888 | locinput += UTF8SKIP(locinput); | |
3889 | } else { | |
3890 | ||
3891 | /* Must be V or LV. Take it, then | |
3892 | * match V* */ | |
3893 | locinput += UTF8SKIP(locinput); | |
3894 | while (locinput < PL_regeol | |
3895 | && swash_fetch(PL_utf8_X_V, | |
f2ed9b32 | 3896 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3897 | { |
3898 | locinput += UTF8SKIP(locinput); | |
3899 | } | |
3900 | } | |
3901 | ||
3902 | /* And any of LV, LVT, or V can be followed | |
3903 | * by T* */ | |
3904 | while (locinput < PL_regeol | |
3905 | && swash_fetch(PL_utf8_X_T, | |
3906 | (U8*)locinput, | |
f2ed9b32 | 3907 | utf8_target)) |
37e2e78e KW |
3908 | { |
3909 | locinput += UTF8SKIP(locinput); | |
3910 | } | |
3911 | } | |
3912 | } | |
3913 | } | |
3914 | ||
3915 | /* Match any extender */ | |
3916 | while (locinput < PL_regeol | |
3917 | && swash_fetch(PL_utf8_X_extend, | |
f2ed9b32 | 3918 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3919 | { |
3920 | locinput += UTF8SKIP(locinput); | |
3921 | } | |
3922 | } | |
3923 | } | |
3924 | if (locinput > PL_regeol) sayNO; | |
3925 | } | |
a0ed51b3 LW |
3926 | nextchr = UCHARAT(locinput); |
3927 | break; | |
81714fb9 YO |
3928 | |
3929 | case NREFFL: | |
d7ef4b73 KW |
3930 | { /* The capture buffer cases. The ones beginning with N for the |
3931 | named buffers just convert to the equivalent numbered and | |
3932 | pretend they were called as the corresponding numbered buffer | |
3933 | op. */ | |
8368298a | 3934 | /* don't initialize these, it makes C++ unhappy */ |
81714fb9 | 3935 | char *s; |
ff1157ca | 3936 | char type; |
8368298a TC |
3937 | re_fold_t folder; |
3938 | const U8 *fold_array; | |
3939 | ||
81714fb9 | 3940 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
3941 | folder = foldEQ_locale; |
3942 | fold_array = PL_fold_locale; | |
3943 | type = REFFL; | |
3944 | goto do_nref; | |
3945 | ||
3946 | case NREFFU: | |
3947 | folder = foldEQ_latin1; | |
3948 | fold_array = PL_fold_latin1; | |
3949 | type = REFFU; | |
3950 | goto do_nref; | |
3951 | ||
81714fb9 | 3952 | case NREFF: |
d7ef4b73 KW |
3953 | folder = foldEQ; |
3954 | fold_array = PL_fold; | |
3955 | type = REFF; | |
3956 | goto do_nref; | |
3957 | ||
3958 | case NREF: | |
3959 | type = REF; | |
83d7b90b KW |
3960 | folder = NULL; |
3961 | fold_array = NULL; | |
d7ef4b73 KW |
3962 | do_nref: |
3963 | ||
3964 | /* For the named back references, find the corresponding buffer | |
3965 | * number */ | |
0a4db386 YO |
3966 | n = reg_check_named_buff_matched(rex,scan); |
3967 | ||
d7ef4b73 | 3968 | if ( ! n ) { |
81714fb9 | 3969 | sayNO; |
d7ef4b73 KW |
3970 | } |
3971 | goto do_nref_ref_common; | |
3972 | ||
c8756f30 | 3973 | case REFFL: |
3280af22 | 3974 | PL_reg_flags |= RF_tainted; |
d7ef4b73 KW |
3975 | folder = foldEQ_locale; |
3976 | fold_array = PL_fold_locale; | |
3977 | goto do_ref; | |
3978 | ||
3979 | case REFFU: | |
3980 | folder = foldEQ_latin1; | |
3981 | fold_array = PL_fold_latin1; | |
3982 | goto do_ref; | |
3983 | ||
3984 | case REFF: | |
3985 | folder = foldEQ; | |
3986 | fold_array = PL_fold; | |
83d7b90b | 3987 | goto do_ref; |
d7ef4b73 | 3988 | |
c277df42 | 3989 | case REF: |
83d7b90b KW |
3990 | folder = NULL; |
3991 | fold_array = NULL; | |
3992 | ||
d7ef4b73 | 3993 | do_ref: |
81714fb9 | 3994 | type = OP(scan); |
d7ef4b73 KW |
3995 | n = ARG(scan); /* which paren pair */ |
3996 | ||
3997 | do_nref_ref_common: | |
f0ab9afb | 3998 | ln = PL_regoffs[n].start; |
2c2d71f5 | 3999 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
3b6647e0 | 4000 | if (*PL_reglastparen < n || ln == -1) |
af3f8c16 | 4001 | sayNO; /* Do not match unless seen CLOSEn. */ |
f0ab9afb | 4002 | if (ln == PL_regoffs[n].end) |
a0d0e21e | 4003 | break; |
a0ed51b3 | 4004 | |
24d3c4a9 | 4005 | s = PL_bostr + ln; |
d7ef4b73 KW |
4006 | if (type != REF /* REF can do byte comparison */ |
4007 | && (utf8_target | |
4008 | || (type == REFFU | |
4009 | && (*s == (char) LATIN_SMALL_LETTER_SHARP_S | |
4010 | || *locinput == (char) LATIN_SMALL_LETTER_SHARP_S)))) | |
4011 | { /* XXX handle REFFL better */ | |
4012 | char * limit = PL_regeol; | |
4013 | ||
4014 | /* This call case insensitively compares the entire buffer | |
4015 | * at s, with the current input starting at locinput, but | |
4016 | * not going off the end given by PL_regeol, and returns in | |
4017 | * limit upon success, how much of the current input was | |
4018 | * matched */ | |
4019 | if (! foldEQ_utf8(s, NULL, PL_regoffs[n].end - ln, utf8_target, | |
4020 | locinput, &limit, 0, utf8_target)) | |
4021 | { | |
4022 | sayNO; | |
a0ed51b3 | 4023 | } |
d7ef4b73 | 4024 | locinput = limit; |
a0ed51b3 LW |
4025 | nextchr = UCHARAT(locinput); |
4026 | break; | |
4027 | } | |
4028 | ||
d7ef4b73 | 4029 | /* Not utf8: Inline the first character, for speed. */ |
76e3520e | 4030 | if (UCHARAT(s) != nextchr && |
81714fb9 | 4031 | (type == REF || |
d7ef4b73 | 4032 | UCHARAT(s) != fold_array[nextchr])) |
4633a7c4 | 4033 | sayNO; |
f0ab9afb | 4034 | ln = PL_regoffs[n].end - ln; |
24d3c4a9 | 4035 | if (locinput + ln > PL_regeol) |
4633a7c4 | 4036 | sayNO; |
81714fb9 | 4037 | if (ln > 1 && (type == REF |
24d3c4a9 | 4038 | ? memNE(s, locinput, ln) |
d7ef4b73 | 4039 | : ! folder(s, locinput, ln))) |
4633a7c4 | 4040 | sayNO; |
24d3c4a9 | 4041 | locinput += ln; |
76e3520e | 4042 | nextchr = UCHARAT(locinput); |
a0d0e21e | 4043 | break; |
81714fb9 | 4044 | } |
a0d0e21e | 4045 | case NOTHING: |
c277df42 | 4046 | case TAIL: |
a0d0e21e LW |
4047 | break; |
4048 | case BACK: | |
4049 | break; | |
40a82448 DM |
4050 | |
4051 | #undef ST | |
4052 | #define ST st->u.eval | |
c277df42 | 4053 | { |
c277df42 | 4054 | SV *ret; |
d2f13c59 | 4055 | REGEXP *re_sv; |
6bda09f9 | 4056 | regexp *re; |
f8fc2ecf | 4057 | regexp_internal *rei; |
1a147d38 YO |
4058 | regnode *startpoint; |
4059 | ||
4060 | case GOSTART: | |
e7707071 YO |
4061 | case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */ |
4062 | if (cur_eval && cur_eval->locinput==locinput) { | |
24b23f37 | 4063 | if (cur_eval->u.eval.close_paren == (U32)ARG(scan)) |
1a147d38 | 4064 | Perl_croak(aTHX_ "Infinite recursion in regex"); |
4b196cd4 | 4065 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 YO |
4066 | Perl_croak(aTHX_ |
4067 | "Pattern subroutine nesting without pos change" | |
4068 | " exceeded limit in regex"); | |
6bda09f9 YO |
4069 | } else { |
4070 | nochange_depth = 0; | |
1a147d38 | 4071 | } |
288b8c02 | 4072 | re_sv = rex_sv; |
6bda09f9 | 4073 | re = rex; |
f8fc2ecf | 4074 | rei = rexi; |
288b8c02 | 4075 | (void)ReREFCNT_inc(rex_sv); |
1a147d38 | 4076 | if (OP(scan)==GOSUB) { |
6bda09f9 YO |
4077 | startpoint = scan + ARG2L(scan); |
4078 | ST.close_paren = ARG(scan); | |
4079 | } else { | |
f8fc2ecf | 4080 | startpoint = rei->program+1; |
6bda09f9 YO |
4081 | ST.close_paren = 0; |
4082 | } | |
4083 | goto eval_recurse_doit; | |
4084 | /* NOTREACHED */ | |
4085 | case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ | |
4086 | if (cur_eval && cur_eval->locinput==locinput) { | |
4b196cd4 | 4087 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 | 4088 | Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex"); |
6bda09f9 YO |
4089 | } else { |
4090 | nochange_depth = 0; | |
4091 | } | |
8e5e9ebe | 4092 | { |
4aabdb9b DM |
4093 | /* execute the code in the {...} */ |
4094 | dSP; | |
6136c704 | 4095 | SV ** const before = SP; |
4aabdb9b DM |
4096 | OP_4tree * const oop = PL_op; |
4097 | COP * const ocurcop = PL_curcop; | |
4098 | PAD *old_comppad; | |
d80618d2 | 4099 | char *saved_regeol = PL_regeol; |
91332126 FR |
4100 | struct re_save_state saved_state; |
4101 | ||
6562f1c4 | 4102 | /* To not corrupt the existing regex state while executing the |
b7f4cd04 FR |
4103 | * eval we would normally put it on the save stack, like with |
4104 | * save_re_context. However, re-evals have a weird scoping so we | |
4105 | * can't just add ENTER/LEAVE here. With that, things like | |
4106 | * | |
4107 | * (?{$a=2})(a(?{local$a=$a+1}))*aak*c(?{$b=$a}) | |
4108 | * | |
4109 | * would break, as they expect the localisation to be unwound | |
4110 | * only when the re-engine backtracks through the bit that | |
4111 | * localised it. | |
4112 | * | |
4113 | * What we do instead is just saving the state in a local c | |
4114 | * variable. | |
4115 | */ | |
91332126 FR |
4116 | Copy(&PL_reg_state, &saved_state, 1, struct re_save_state); |
4117 | ||
4aabdb9b | 4118 | n = ARG(scan); |
f8fc2ecf | 4119 | PL_op = (OP_4tree*)rexi->data->data[n]; |
24b23f37 YO |
4120 | DEBUG_STATE_r( PerlIO_printf(Perl_debug_log, |
4121 | " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); | |
f8fc2ecf | 4122 | PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]); |
f0ab9afb | 4123 | PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr; |
4aabdb9b | 4124 | |
2bf803e2 YO |
4125 | if (sv_yes_mark) { |
4126 | SV *sv_mrk = get_sv("REGMARK", 1); | |
4127 | sv_setsv(sv_mrk, sv_yes_mark); | |
4128 | } | |
4129 | ||
8e5e9ebe RGS |
4130 | CALLRUNOPS(aTHX); /* Scalar context. */ |
4131 | SPAGAIN; | |
4132 | if (SP == before) | |
075aa684 | 4133 | ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */ |
8e5e9ebe RGS |
4134 | else { |
4135 | ret = POPs; | |
4136 | PUTBACK; | |
4137 | } | |
4aabdb9b | 4138 | |
91332126 FR |
4139 | Copy(&saved_state, &PL_reg_state, 1, struct re_save_state); |
4140 | ||
4aabdb9b DM |
4141 | PL_op = oop; |
4142 | PAD_RESTORE_LOCAL(old_comppad); | |
4143 | PL_curcop = ocurcop; | |
d80618d2 | 4144 | PL_regeol = saved_regeol; |
24d3c4a9 | 4145 | if (!logical) { |
4aabdb9b DM |
4146 | /* /(?{...})/ */ |
4147 | sv_setsv(save_scalar(PL_replgv), ret); | |
4aabdb9b DM |
4148 | break; |
4149 | } | |
8e5e9ebe | 4150 | } |
24d3c4a9 DM |
4151 | if (logical == 2) { /* Postponed subexpression: /(??{...})/ */ |
4152 | logical = 0; | |
4aabdb9b | 4153 | { |
4f639d21 DM |
4154 | /* extract RE object from returned value; compiling if |
4155 | * necessary */ | |
6136c704 | 4156 | MAGIC *mg = NULL; |
288b8c02 | 4157 | REGEXP *rx = NULL; |
5c35adbb NC |
4158 | |
4159 | if (SvROK(ret)) { | |
288b8c02 | 4160 | SV *const sv = SvRV(ret); |
5c35adbb NC |
4161 | |
4162 | if (SvTYPE(sv) == SVt_REGEXP) { | |
d2f13c59 | 4163 | rx = (REGEXP*) sv; |
5c35adbb NC |
4164 | } else if (SvSMAGICAL(sv)) { |
4165 | mg = mg_find(sv, PERL_MAGIC_qr); | |
4166 | assert(mg); | |
4167 | } | |
4168 | } else if (SvTYPE(ret) == SVt_REGEXP) { | |
d2f13c59 | 4169 | rx = (REGEXP*) ret; |
5c35adbb | 4170 | } else if (SvSMAGICAL(ret)) { |
124ee91a NC |
4171 | if (SvGMAGICAL(ret)) { |
4172 | /* I don't believe that there is ever qr magic | |
4173 | here. */ | |
4174 | assert(!mg_find(ret, PERL_MAGIC_qr)); | |
faf82a0b | 4175 | sv_unmagic(ret, PERL_MAGIC_qr); |
124ee91a NC |
4176 | } |
4177 | else { | |
faf82a0b | 4178 | mg = mg_find(ret, PERL_MAGIC_qr); |
124ee91a NC |
4179 | /* testing suggests mg only ends up non-NULL for |
4180 | scalars who were upgraded and compiled in the | |
4181 | else block below. In turn, this is only | |
4182 | triggered in the "postponed utf8 string" tests | |
4183 | in t/op/pat.t */ | |
4184 | } | |
0f5d15d6 | 4185 | } |
faf82a0b | 4186 | |
0f5d15d6 | 4187 | if (mg) { |
d2f13c59 | 4188 | rx = (REGEXP *) mg->mg_obj; /*XXX:dmq*/ |
e2560c33 | 4189 | assert(rx); |
0f5d15d6 | 4190 | } |
288b8c02 | 4191 | if (rx) { |
f0826785 | 4192 | rx = reg_temp_copy(NULL, rx); |
288b8c02 | 4193 | } |
0f5d15d6 | 4194 | else { |
c737faaf | 4195 | U32 pm_flags = 0; |
a3b680e6 | 4196 | const I32 osize = PL_regsize; |
0f5d15d6 | 4197 | |
b9ad30b4 NC |
4198 | if (DO_UTF8(ret)) { |
4199 | assert (SvUTF8(ret)); | |
4200 | } else if (SvUTF8(ret)) { | |
4201 | /* Not doing UTF-8, despite what the SV says. Is | |
4202 | this only if we're trapped in use 'bytes'? */ | |
4203 | /* Make a copy of the octet sequence, but without | |
4204 | the flag on, as the compiler now honours the | |
4205 | SvUTF8 flag on ret. */ | |
4206 | STRLEN len; | |
4207 | const char *const p = SvPV(ret, len); | |
4208 | ret = newSVpvn_flags(p, len, SVs_TEMP); | |
4209 | } | |
288b8c02 | 4210 | rx = CALLREGCOMP(ret, pm_flags); |
9041c2e3 | 4211 | if (!(SvFLAGS(ret) |
faf82a0b | 4212 | & (SVs_TEMP | SVs_PADTMP | SVf_READONLY |
3ce3ed55 | 4213 | | SVs_GMG))) { |
a2794585 NC |
4214 | /* This isn't a first class regexp. Instead, it's |
4215 | caching a regexp onto an existing, Perl visible | |
4216 | scalar. */ | |
ad64d0ec | 4217 | sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0); |
3ce3ed55 | 4218 | } |
0f5d15d6 | 4219 | PL_regsize = osize; |
0f5d15d6 | 4220 | } |
288b8c02 NC |
4221 | re_sv = rx; |
4222 | re = (struct regexp *)SvANY(rx); | |
4aabdb9b | 4223 | } |
07bc277f | 4224 | RXp_MATCH_COPIED_off(re); |
28d8d7f4 YO |
4225 | re->subbeg = rex->subbeg; |
4226 | re->sublen = rex->sublen; | |
f8fc2ecf | 4227 | rei = RXi_GET(re); |
6bda09f9 | 4228 | DEBUG_EXECUTE_r( |
f2ed9b32 | 4229 | debug_start_match(re_sv, utf8_target, locinput, PL_regeol, |
6bda09f9 YO |
4230 | "Matching embedded"); |
4231 | ); | |
f8fc2ecf | 4232 | startpoint = rei->program + 1; |
1a147d38 | 4233 | ST.close_paren = 0; /* only used for GOSUB */ |
6bda09f9 YO |
4234 | /* borrowed from regtry */ |
4235 | if (PL_reg_start_tmpl <= re->nparens) { | |
4236 | PL_reg_start_tmpl = re->nparens*3/2 + 3; | |
4237 | if(PL_reg_start_tmp) | |
4238 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
4239 | else | |
4240 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
dd5def09 | 4241 | } |
aa283a38 | 4242 | |
1a147d38 | 4243 | eval_recurse_doit: /* Share code with GOSUB below this line */ |
aa283a38 | 4244 | /* run the pattern returned from (??{...}) */ |
40a82448 DM |
4245 | ST.cp = regcppush(0); /* Save *all* the positions. */ |
4246 | REGCP_SET(ST.lastcp); | |
6bda09f9 | 4247 | |
f0ab9afb | 4248 | PL_regoffs = re->offs; /* essentially NOOP on GOSUB */ |
6bda09f9 | 4249 | |
0357f1fd ML |
4250 | /* see regtry, specifically PL_reglast(?:close)?paren is a pointer! (i dont know why) :dmq */ |
4251 | PL_reglastparen = &re->lastparen; | |
4252 | PL_reglastcloseparen = &re->lastcloseparen; | |
4253 | re->lastparen = 0; | |
4254 | re->lastcloseparen = 0; | |
4255 | ||
4aabdb9b | 4256 | PL_reginput = locinput; |
ae0beba1 | 4257 | PL_regsize = 0; |
4aabdb9b DM |
4258 | |
4259 | /* XXXX This is too dramatic a measure... */ | |
4260 | PL_reg_maxiter = 0; | |
4261 | ||
faec1544 | 4262 | ST.toggle_reg_flags = PL_reg_flags; |
3c8556c3 | 4263 | if (RX_UTF8(re_sv)) |
faec1544 DM |
4264 | PL_reg_flags |= RF_utf8; |
4265 | else | |
4266 | PL_reg_flags &= ~RF_utf8; | |
4267 | ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */ | |
4268 | ||
288b8c02 | 4269 | ST.prev_rex = rex_sv; |
faec1544 | 4270 | ST.prev_curlyx = cur_curlyx; |
288b8c02 NC |
4271 | SETREX(rex_sv,re_sv); |
4272 | rex = re; | |
f8fc2ecf | 4273 | rexi = rei; |
faec1544 | 4274 | cur_curlyx = NULL; |
40a82448 | 4275 | ST.B = next; |
faec1544 DM |
4276 | ST.prev_eval = cur_eval; |
4277 | cur_eval = st; | |
faec1544 | 4278 | /* now continue from first node in postoned RE */ |
6bda09f9 | 4279 | PUSH_YES_STATE_GOTO(EVAL_AB, startpoint); |
4aabdb9b | 4280 | /* NOTREACHED */ |
a0ed51b3 | 4281 | } |
24d3c4a9 | 4282 | /* logical is 1, /(?(?{...})X|Y)/ */ |
f2338a2e | 4283 | sw = cBOOL(SvTRUE(ret)); |
24d3c4a9 | 4284 | logical = 0; |
c277df42 IZ |
4285 | break; |
4286 | } | |
40a82448 | 4287 | |
faec1544 DM |
4288 | case EVAL_AB: /* cleanup after a successful (??{A})B */ |
4289 | /* note: this is called twice; first after popping B, then A */ | |
4290 | PL_reg_flags ^= ST.toggle_reg_flags; | |
288b8c02 NC |
4291 | ReREFCNT_dec(rex_sv); |
4292 | SETREX(rex_sv,ST.prev_rex); | |
4293 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 4294 | rexi = RXi_GET(rex); |
faec1544 DM |
4295 | regcpblow(ST.cp); |
4296 | cur_eval = ST.prev_eval; | |
4297 | cur_curlyx = ST.prev_curlyx; | |
34a81e2b B |
4298 | |
4299 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ | |
0357f1fd ML |
4300 | PL_reglastparen = &rex->lastparen; |
4301 | PL_reglastcloseparen = &rex->lastcloseparen; | |
34a81e2b B |
4302 | /* also update PL_regoffs */ |
4303 | PL_regoffs = rex->offs; | |
0357f1fd | 4304 | |
40a82448 DM |
4305 | /* XXXX This is too dramatic a measure... */ |
4306 | PL_reg_maxiter = 0; | |
e7707071 | 4307 | if ( nochange_depth ) |
4b196cd4 | 4308 | nochange_depth--; |
262b90c4 | 4309 | sayYES; |
40a82448 | 4310 | |
40a82448 | 4311 | |
faec1544 DM |
4312 | case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */ |
4313 | /* note: this is called twice; first after popping B, then A */ | |
4314 | PL_reg_flags ^= ST.toggle_reg_flags; | |
288b8c02 NC |
4315 | ReREFCNT_dec(rex_sv); |
4316 | SETREX(rex_sv,ST.prev_rex); | |
4317 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 4318 | rexi = RXi_GET(rex); |
34a81e2b | 4319 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ |
0357f1fd ML |
4320 | PL_reglastparen = &rex->lastparen; |
4321 | PL_reglastcloseparen = &rex->lastcloseparen; | |
4322 | ||
40a82448 DM |
4323 | PL_reginput = locinput; |
4324 | REGCP_UNWIND(ST.lastcp); | |
4325 | regcppop(rex); | |
faec1544 DM |
4326 | cur_eval = ST.prev_eval; |
4327 | cur_curlyx = ST.prev_curlyx; | |
4328 | /* XXXX This is too dramatic a measure... */ | |
4329 | PL_reg_maxiter = 0; | |
e7707071 | 4330 | if ( nochange_depth ) |
4b196cd4 | 4331 | nochange_depth--; |
40a82448 | 4332 | sayNO_SILENT; |
40a82448 DM |
4333 | #undef ST |
4334 | ||
a0d0e21e | 4335 | case OPEN: |
c277df42 | 4336 | n = ARG(scan); /* which paren pair */ |
3280af22 NIS |
4337 | PL_reg_start_tmp[n] = locinput; |
4338 | if (n > PL_regsize) | |
4339 | PL_regsize = n; | |
e2e6a0f1 | 4340 | lastopen = n; |
a0d0e21e LW |
4341 | break; |
4342 | case CLOSE: | |
c277df42 | 4343 | n = ARG(scan); /* which paren pair */ |
f0ab9afb NC |
4344 | PL_regoffs[n].start = PL_reg_start_tmp[n] - PL_bostr; |
4345 | PL_regoffs[n].end = locinput - PL_bostr; | |
7f69552c YO |
4346 | /*if (n > PL_regsize) |
4347 | PL_regsize = n;*/ | |
3b6647e0 | 4348 | if (n > *PL_reglastparen) |
3280af22 | 4349 | *PL_reglastparen = n; |
a01268b5 | 4350 | *PL_reglastcloseparen = n; |
3b6647e0 | 4351 | if (cur_eval && cur_eval->u.eval.close_paren == n) { |
6bda09f9 YO |
4352 | goto fake_end; |
4353 | } | |
a0d0e21e | 4354 | break; |
e2e6a0f1 YO |
4355 | case ACCEPT: |
4356 | if (ARG(scan)){ | |
4357 | regnode *cursor; | |
4358 | for (cursor=scan; | |
4359 | cursor && OP(cursor)!=END; | |
4360 | cursor=regnext(cursor)) | |
4361 | { | |
4362 | if ( OP(cursor)==CLOSE ){ | |
4363 | n = ARG(cursor); | |
4364 | if ( n <= lastopen ) { | |
f0ab9afb NC |
4365 | PL_regoffs[n].start |
4366 | = PL_reg_start_tmp[n] - PL_bostr; | |
4367 | PL_regoffs[n].end = locinput - PL_bostr; | |
e2e6a0f1 YO |
4368 | /*if (n > PL_regsize) |
4369 | PL_regsize = n;*/ | |
3b6647e0 | 4370 | if (n > *PL_reglastparen) |
e2e6a0f1 YO |
4371 | *PL_reglastparen = n; |
4372 | *PL_reglastcloseparen = n; | |
3b6647e0 RB |
4373 | if ( n == ARG(scan) || (cur_eval && |
4374 | cur_eval->u.eval.close_paren == n)) | |
e2e6a0f1 YO |
4375 | break; |
4376 | } | |
4377 | } | |
4378 | } | |
4379 | } | |
4380 | goto fake_end; | |
4381 | /*NOTREACHED*/ | |
c277df42 IZ |
4382 | case GROUPP: |
4383 | n = ARG(scan); /* which paren pair */ | |
f2338a2e | 4384 | sw = cBOOL(*PL_reglastparen >= n && PL_regoffs[n].end != -1); |
c277df42 | 4385 | break; |
0a4db386 YO |
4386 | case NGROUPP: |
4387 | /* reg_check_named_buff_matched returns 0 for no match */ | |
f2338a2e | 4388 | sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan)); |
0a4db386 | 4389 | break; |
1a147d38 | 4390 | case INSUBP: |
0a4db386 | 4391 | n = ARG(scan); |
3b6647e0 | 4392 | sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n)); |
0a4db386 YO |
4393 | break; |
4394 | case DEFINEP: | |
4395 | sw = 0; | |
4396 | break; | |
c277df42 | 4397 | case IFTHEN: |
2c2d71f5 | 4398 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
24d3c4a9 | 4399 | if (sw) |
c277df42 IZ |
4400 | next = NEXTOPER(NEXTOPER(scan)); |
4401 | else { | |
4402 | next = scan + ARG(scan); | |
4403 | if (OP(next) == IFTHEN) /* Fake one. */ | |
4404 | next = NEXTOPER(NEXTOPER(next)); | |
4405 | } | |
4406 | break; | |
4407 | case LOGICAL: | |
24d3c4a9 | 4408 | logical = scan->flags; |
c277df42 | 4409 | break; |
c476f425 | 4410 | |
2ab05381 | 4411 | /******************************************************************* |
2ab05381 | 4412 | |
c476f425 DM |
4413 | The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/ |
4414 | pattern, where A and B are subpatterns. (For simple A, CURLYM or | |
4415 | STAR/PLUS/CURLY/CURLYN are used instead.) | |
2ab05381 | 4416 | |
c476f425 | 4417 | A*B is compiled as <CURLYX><A><WHILEM><B> |
2ab05381 | 4418 | |
c476f425 DM |
4419 | On entry to the subpattern, CURLYX is called. This pushes a CURLYX |
4420 | state, which contains the current count, initialised to -1. It also sets | |
4421 | cur_curlyx to point to this state, with any previous value saved in the | |
4422 | state block. | |
2ab05381 | 4423 | |
c476f425 DM |
4424 | CURLYX then jumps straight to the WHILEM op, rather than executing A, |
4425 | since the pattern may possibly match zero times (i.e. it's a while {} loop | |
4426 | rather than a do {} while loop). | |
2ab05381 | 4427 | |
c476f425 DM |
4428 | Each entry to WHILEM represents a successful match of A. The count in the |
4429 | CURLYX block is incremented, another WHILEM state is pushed, and execution | |
4430 | passes to A or B depending on greediness and the current count. | |
2ab05381 | 4431 | |
c476f425 DM |
4432 | For example, if matching against the string a1a2a3b (where the aN are |
4433 | substrings that match /A/), then the match progresses as follows: (the | |
4434 | pushed states are interspersed with the bits of strings matched so far): | |
2ab05381 | 4435 | |
c476f425 DM |
4436 | <CURLYX cnt=-1> |
4437 | <CURLYX cnt=0><WHILEM> | |
4438 | <CURLYX cnt=1><WHILEM> a1 <WHILEM> | |
4439 | <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM> | |
4440 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> | |
4441 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b | |
2ab05381 | 4442 | |
c476f425 DM |
4443 | (Contrast this with something like CURLYM, which maintains only a single |
4444 | backtrack state: | |
2ab05381 | 4445 | |
c476f425 DM |
4446 | <CURLYM cnt=0> a1 |
4447 | a1 <CURLYM cnt=1> a2 | |
4448 | a1 a2 <CURLYM cnt=2> a3 | |
4449 | a1 a2 a3 <CURLYM cnt=3> b | |
4450 | ) | |
2ab05381 | 4451 | |
c476f425 DM |
4452 | Each WHILEM state block marks a point to backtrack to upon partial failure |
4453 | of A or B, and also contains some minor state data related to that | |
4454 | iteration. The CURLYX block, pointed to by cur_curlyx, contains the | |
4455 | overall state, such as the count, and pointers to the A and B ops. | |
2ab05381 | 4456 | |
c476f425 DM |
4457 | This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx |
4458 | must always point to the *current* CURLYX block, the rules are: | |
2ab05381 | 4459 | |
c476f425 DM |
4460 | When executing CURLYX, save the old cur_curlyx in the CURLYX state block, |
4461 | and set cur_curlyx to point the new block. | |
2ab05381 | 4462 | |
c476f425 DM |
4463 | When popping the CURLYX block after a successful or unsuccessful match, |
4464 | restore the previous cur_curlyx. | |
2ab05381 | 4465 | |
c476f425 DM |
4466 | When WHILEM is about to execute B, save the current cur_curlyx, and set it |
4467 | to the outer one saved in the CURLYX block. | |
2ab05381 | 4468 | |
c476f425 DM |
4469 | When popping the WHILEM block after a successful or unsuccessful B match, |
4470 | restore the previous cur_curlyx. | |
2ab05381 | 4471 | |
c476f425 DM |
4472 | Here's an example for the pattern (AI* BI)*BO |
4473 | I and O refer to inner and outer, C and W refer to CURLYX and WHILEM: | |
2ab05381 | 4474 | |
c476f425 DM |
4475 | cur_ |
4476 | curlyx backtrack stack | |
4477 | ------ --------------- | |
4478 | NULL | |
4479 | CO <CO prev=NULL> <WO> | |
4480 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4481 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi | |
4482 | NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo | |
2ab05381 | 4483 | |
c476f425 DM |
4484 | At this point the pattern succeeds, and we work back down the stack to |
4485 | clean up, restoring as we go: | |
95b24440 | 4486 | |
c476f425 DM |
4487 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi |
4488 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4489 | CO <CO prev=NULL> <WO> | |
4490 | NULL | |
a0374537 | 4491 | |
c476f425 DM |
4492 | *******************************************************************/ |
4493 | ||
4494 | #define ST st->u.curlyx | |
4495 | ||
4496 | case CURLYX: /* start of /A*B/ (for complex A) */ | |
4497 | { | |
4498 | /* No need to save/restore up to this paren */ | |
4499 | I32 parenfloor = scan->flags; | |
4500 | ||
4501 | assert(next); /* keep Coverity happy */ | |
4502 | if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ | |
4503 | next += ARG(next); | |
4504 | ||
4505 | /* XXXX Probably it is better to teach regpush to support | |
4506 | parenfloor > PL_regsize... */ | |
4507 | if (parenfloor > (I32)*PL_reglastparen) | |
4508 | parenfloor = *PL_reglastparen; /* Pessimization... */ | |
4509 | ||
4510 | ST.prev_curlyx= cur_curlyx; | |
4511 | cur_curlyx = st; | |
4512 | ST.cp = PL_savestack_ix; | |
4513 | ||
4514 | /* these fields contain the state of the current curly. | |
4515 | * they are accessed by subsequent WHILEMs */ | |
4516 | ST.parenfloor = parenfloor; | |
d02d6d97 | 4517 | ST.me = scan; |
c476f425 | 4518 | ST.B = next; |
24d3c4a9 DM |
4519 | ST.minmod = minmod; |
4520 | minmod = 0; | |
c476f425 DM |
4521 | ST.count = -1; /* this will be updated by WHILEM */ |
4522 | ST.lastloc = NULL; /* this will be updated by WHILEM */ | |
4523 | ||
4524 | PL_reginput = locinput; | |
4525 | PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next)); | |
5f66b61c | 4526 | /* NOTREACHED */ |
c476f425 | 4527 | } |
a0d0e21e | 4528 | |
c476f425 | 4529 | case CURLYX_end: /* just finished matching all of A*B */ |
c476f425 DM |
4530 | cur_curlyx = ST.prev_curlyx; |
4531 | sayYES; | |
4532 | /* NOTREACHED */ | |
a0d0e21e | 4533 | |
c476f425 DM |
4534 | case CURLYX_end_fail: /* just failed to match all of A*B */ |
4535 | regcpblow(ST.cp); | |
4536 | cur_curlyx = ST.prev_curlyx; | |
4537 | sayNO; | |
4538 | /* NOTREACHED */ | |
4633a7c4 | 4539 | |
a0d0e21e | 4540 | |
c476f425 DM |
4541 | #undef ST |
4542 | #define ST st->u.whilem | |
4543 | ||
4544 | case WHILEM: /* just matched an A in /A*B/ (for complex A) */ | |
4545 | { | |
4546 | /* see the discussion above about CURLYX/WHILEM */ | |
c476f425 | 4547 | I32 n; |
d02d6d97 DM |
4548 | int min = ARG1(cur_curlyx->u.curlyx.me); |
4549 | int max = ARG2(cur_curlyx->u.curlyx.me); | |
4550 | regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS; | |
4551 | ||
c476f425 DM |
4552 | assert(cur_curlyx); /* keep Coverity happy */ |
4553 | n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */ | |
4554 | ST.save_lastloc = cur_curlyx->u.curlyx.lastloc; | |
4555 | ST.cache_offset = 0; | |
4556 | ST.cache_mask = 0; | |
4557 | ||
4558 | PL_reginput = locinput; | |
4559 | ||
4560 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
d02d6d97 DM |
4561 | "%*s whilem: matched %ld out of %d..%d\n", |
4562 | REPORT_CODE_OFF+depth*2, "", (long)n, min, max) | |
c476f425 | 4563 | ); |
a0d0e21e | 4564 | |
c476f425 | 4565 | /* First just match a string of min A's. */ |
a0d0e21e | 4566 | |
d02d6d97 | 4567 | if (n < min) { |
c476f425 | 4568 | cur_curlyx->u.curlyx.lastloc = locinput; |
d02d6d97 | 4569 | PUSH_STATE_GOTO(WHILEM_A_pre, A); |
c476f425 DM |
4570 | /* NOTREACHED */ |
4571 | } | |
4572 | ||
4573 | /* If degenerate A matches "", assume A done. */ | |
4574 | ||
4575 | if (locinput == cur_curlyx->u.curlyx.lastloc) { | |
4576 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4577 | "%*s whilem: empty match detected, trying continuation...\n", | |
4578 | REPORT_CODE_OFF+depth*2, "") | |
4579 | ); | |
4580 | goto do_whilem_B_max; | |
4581 | } | |
4582 | ||
4583 | /* super-linear cache processing */ | |
4584 | ||
4585 | if (scan->flags) { | |
a0d0e21e | 4586 | |
2c2d71f5 | 4587 | if (!PL_reg_maxiter) { |
c476f425 DM |
4588 | /* start the countdown: Postpone detection until we |
4589 | * know the match is not *that* much linear. */ | |
2c2d71f5 | 4590 | PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4); |
66bf836d DM |
4591 | /* possible overflow for long strings and many CURLYX's */ |
4592 | if (PL_reg_maxiter < 0) | |
4593 | PL_reg_maxiter = I32_MAX; | |
2c2d71f5 JH |
4594 | PL_reg_leftiter = PL_reg_maxiter; |
4595 | } | |
c476f425 | 4596 | |
2c2d71f5 | 4597 | if (PL_reg_leftiter-- == 0) { |
c476f425 | 4598 | /* initialise cache */ |
3298f257 | 4599 | const I32 size = (PL_reg_maxiter + 7)/8; |
2c2d71f5 | 4600 | if (PL_reg_poscache) { |
eb160463 | 4601 | if ((I32)PL_reg_poscache_size < size) { |
2c2d71f5 JH |
4602 | Renew(PL_reg_poscache, size, char); |
4603 | PL_reg_poscache_size = size; | |
4604 | } | |
4605 | Zero(PL_reg_poscache, size, char); | |
4606 | } | |
4607 | else { | |
4608 | PL_reg_poscache_size = size; | |
a02a5408 | 4609 | Newxz(PL_reg_poscache, size, char); |
2c2d71f5 | 4610 | } |
c476f425 DM |
4611 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
4612 | "%swhilem: Detected a super-linear match, switching on caching%s...\n", | |
4613 | PL_colors[4], PL_colors[5]) | |
4614 | ); | |
2c2d71f5 | 4615 | } |
c476f425 | 4616 | |
2c2d71f5 | 4617 | if (PL_reg_leftiter < 0) { |
c476f425 DM |
4618 | /* have we already failed at this position? */ |
4619 | I32 offset, mask; | |
4620 | offset = (scan->flags & 0xf) - 1 | |
4621 | + (locinput - PL_bostr) * (scan->flags>>4); | |
4622 | mask = 1 << (offset % 8); | |
4623 | offset /= 8; | |
4624 | if (PL_reg_poscache[offset] & mask) { | |
4625 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4626 | "%*s whilem: (cache) already tried at this position...\n", | |
4627 | REPORT_CODE_OFF+depth*2, "") | |
2c2d71f5 | 4628 | ); |
3298f257 | 4629 | sayNO; /* cache records failure */ |
2c2d71f5 | 4630 | } |
c476f425 DM |
4631 | ST.cache_offset = offset; |
4632 | ST.cache_mask = mask; | |
2c2d71f5 | 4633 | } |
c476f425 | 4634 | } |
2c2d71f5 | 4635 | |
c476f425 | 4636 | /* Prefer B over A for minimal matching. */ |
a687059c | 4637 | |
c476f425 DM |
4638 | if (cur_curlyx->u.curlyx.minmod) { |
4639 | ST.save_curlyx = cur_curlyx; | |
4640 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4641 | ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor); | |
4642 | REGCP_SET(ST.lastcp); | |
4643 | PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B); | |
4644 | /* NOTREACHED */ | |
4645 | } | |
a0d0e21e | 4646 | |
c476f425 DM |
4647 | /* Prefer A over B for maximal matching. */ |
4648 | ||
d02d6d97 | 4649 | if (n < max) { /* More greed allowed? */ |
c476f425 DM |
4650 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); |
4651 | cur_curlyx->u.curlyx.lastloc = locinput; | |
4652 | REGCP_SET(ST.lastcp); | |
d02d6d97 | 4653 | PUSH_STATE_GOTO(WHILEM_A_max, A); |
c476f425 DM |
4654 | /* NOTREACHED */ |
4655 | } | |
4656 | goto do_whilem_B_max; | |
4657 | } | |
4658 | /* NOTREACHED */ | |
4659 | ||
4660 | case WHILEM_B_min: /* just matched B in a minimal match */ | |
4661 | case WHILEM_B_max: /* just matched B in a maximal match */ | |
4662 | cur_curlyx = ST.save_curlyx; | |
4663 | sayYES; | |
4664 | /* NOTREACHED */ | |
4665 | ||
4666 | case WHILEM_B_max_fail: /* just failed to match B in a maximal match */ | |
4667 | cur_curlyx = ST.save_curlyx; | |
4668 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
4669 | cur_curlyx->u.curlyx.count--; | |
4670 | CACHEsayNO; | |
4671 | /* NOTREACHED */ | |
4672 | ||
4673 | case WHILEM_A_min_fail: /* just failed to match A in a minimal match */ | |
4674 | REGCP_UNWIND(ST.lastcp); | |
4675 | regcppop(rex); | |
4676 | /* FALL THROUGH */ | |
4677 | case WHILEM_A_pre_fail: /* just failed to match even minimal A */ | |
4678 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
4679 | cur_curlyx->u.curlyx.count--; | |
4680 | CACHEsayNO; | |
4681 | /* NOTREACHED */ | |
4682 | ||
4683 | case WHILEM_A_max_fail: /* just failed to match A in a maximal match */ | |
4684 | REGCP_UNWIND(ST.lastcp); | |
4685 | regcppop(rex); /* Restore some previous $<digit>s? */ | |
4686 | PL_reginput = locinput; | |
4687 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
4688 | "%*s whilem: failed, trying continuation...\n", | |
4689 | REPORT_CODE_OFF+depth*2, "") | |
4690 | ); | |
4691 | do_whilem_B_max: | |
4692 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
4693 | && ckWARN(WARN_REGEXP) | |
4694 | && !(PL_reg_flags & RF_warned)) | |
4695 | { | |
4696 | PL_reg_flags |= RF_warned; | |
4697 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded", | |
4698 | "Complex regular subexpression recursion", | |
4699 | REG_INFTY - 1); | |
4700 | } | |
4701 | ||
4702 | /* now try B */ | |
4703 | ST.save_curlyx = cur_curlyx; | |
4704 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4705 | PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B); | |
4706 | /* NOTREACHED */ | |
4707 | ||
4708 | case WHILEM_B_min_fail: /* just failed to match B in a minimal match */ | |
4709 | cur_curlyx = ST.save_curlyx; | |
4710 | REGCP_UNWIND(ST.lastcp); | |
4711 | regcppop(rex); | |
4712 | ||
d02d6d97 | 4713 | if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) { |
c476f425 DM |
4714 | /* Maximum greed exceeded */ |
4715 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
4716 | && ckWARN(WARN_REGEXP) | |
4717 | && !(PL_reg_flags & RF_warned)) | |
4718 | { | |
3280af22 | 4719 | PL_reg_flags |= RF_warned; |
c476f425 DM |
4720 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
4721 | "%s limit (%d) exceeded", | |
4722 | "Complex regular subexpression recursion", | |
4723 | REG_INFTY - 1); | |
a0d0e21e | 4724 | } |
c476f425 | 4725 | cur_curlyx->u.curlyx.count--; |
3ab3c9b4 | 4726 | CACHEsayNO; |
a0d0e21e | 4727 | } |
c476f425 DM |
4728 | |
4729 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
4730 | "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "") | |
4731 | ); | |
4732 | /* Try grabbing another A and see if it helps. */ | |
4733 | PL_reginput = locinput; | |
4734 | cur_curlyx->u.curlyx.lastloc = locinput; | |
4735 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); | |
4736 | REGCP_SET(ST.lastcp); | |
d02d6d97 DM |
4737 | PUSH_STATE_GOTO(WHILEM_A_min, |
4738 | /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS); | |
5f66b61c | 4739 | /* NOTREACHED */ |
40a82448 DM |
4740 | |
4741 | #undef ST | |
4742 | #define ST st->u.branch | |
4743 | ||
4744 | case BRANCHJ: /* /(...|A|...)/ with long next pointer */ | |
c277df42 IZ |
4745 | next = scan + ARG(scan); |
4746 | if (next == scan) | |
4747 | next = NULL; | |
40a82448 DM |
4748 | scan = NEXTOPER(scan); |
4749 | /* FALL THROUGH */ | |
c277df42 | 4750 | |
40a82448 DM |
4751 | case BRANCH: /* /(...|A|...)/ */ |
4752 | scan = NEXTOPER(scan); /* scan now points to inner node */ | |
40a82448 DM |
4753 | ST.lastparen = *PL_reglastparen; |
4754 | ST.next_branch = next; | |
4755 | REGCP_SET(ST.cp); | |
4756 | PL_reginput = locinput; | |
02db2b7b | 4757 | |
40a82448 | 4758 | /* Now go into the branch */ |
5d458dd8 YO |
4759 | if (has_cutgroup) { |
4760 | PUSH_YES_STATE_GOTO(BRANCH_next, scan); | |
4761 | } else { | |
4762 | PUSH_STATE_GOTO(BRANCH_next, scan); | |
4763 | } | |
40a82448 | 4764 | /* NOTREACHED */ |
5d458dd8 YO |
4765 | case CUTGROUP: |
4766 | PL_reginput = locinput; | |
4767 | sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : | |
ad64d0ec | 4768 | MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
4769 | PUSH_STATE_GOTO(CUTGROUP_next,next); |
4770 | /* NOTREACHED */ | |
4771 | case CUTGROUP_next_fail: | |
4772 | do_cutgroup = 1; | |
4773 | no_final = 1; | |
4774 | if (st->u.mark.mark_name) | |
4775 | sv_commit = st->u.mark.mark_name; | |
4776 | sayNO; | |
4777 | /* NOTREACHED */ | |
4778 | case BRANCH_next: | |
4779 | sayYES; | |
4780 | /* NOTREACHED */ | |
40a82448 | 4781 | case BRANCH_next_fail: /* that branch failed; try the next, if any */ |
5d458dd8 YO |
4782 | if (do_cutgroup) { |
4783 | do_cutgroup = 0; | |
4784 | no_final = 0; | |
4785 | } | |
40a82448 DM |
4786 | REGCP_UNWIND(ST.cp); |
4787 | for (n = *PL_reglastparen; n > ST.lastparen; n--) | |
f0ab9afb | 4788 | PL_regoffs[n].end = -1; |
40a82448 | 4789 | *PL_reglastparen = n; |
0a4db386 | 4790 | /*dmq: *PL_reglastcloseparen = n; */ |
40a82448 DM |
4791 | scan = ST.next_branch; |
4792 | /* no more branches? */ | |
5d458dd8 YO |
4793 | if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) { |
4794 | DEBUG_EXECUTE_r({ | |
4795 | PerlIO_printf( Perl_debug_log, | |
4796 | "%*s %sBRANCH failed...%s\n", | |
4797 | REPORT_CODE_OFF+depth*2, "", | |
4798 | PL_colors[4], | |
4799 | PL_colors[5] ); | |
4800 | }); | |
4801 | sayNO_SILENT; | |
4802 | } | |
40a82448 DM |
4803 | continue; /* execute next BRANCH[J] op */ |
4804 | /* NOTREACHED */ | |
4805 | ||
a0d0e21e | 4806 | case MINMOD: |
24d3c4a9 | 4807 | minmod = 1; |
a0d0e21e | 4808 | break; |
40a82448 DM |
4809 | |
4810 | #undef ST | |
4811 | #define ST st->u.curlym | |
4812 | ||
4813 | case CURLYM: /* /A{m,n}B/ where A is fixed-length */ | |
4814 | ||
4815 | /* This is an optimisation of CURLYX that enables us to push | |
84d2fa14 | 4816 | * only a single backtracking state, no matter how many matches |
40a82448 DM |
4817 | * there are in {m,n}. It relies on the pattern being constant |
4818 | * length, with no parens to influence future backrefs | |
4819 | */ | |
4820 | ||
4821 | ST.me = scan; | |
dc45a647 | 4822 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; |
40a82448 DM |
4823 | |
4824 | /* if paren positive, emulate an OPEN/CLOSE around A */ | |
4825 | if (ST.me->flags) { | |
3b6647e0 | 4826 | U32 paren = ST.me->flags; |
40a82448 DM |
4827 | if (paren > PL_regsize) |
4828 | PL_regsize = paren; | |
3b6647e0 | 4829 | if (paren > *PL_reglastparen) |
40a82448 | 4830 | *PL_reglastparen = paren; |
c277df42 | 4831 | scan += NEXT_OFF(scan); /* Skip former OPEN. */ |
6407bf3b | 4832 | } |
40a82448 DM |
4833 | ST.A = scan; |
4834 | ST.B = next; | |
4835 | ST.alen = 0; | |
4836 | ST.count = 0; | |
24d3c4a9 DM |
4837 | ST.minmod = minmod; |
4838 | minmod = 0; | |
40a82448 DM |
4839 | ST.c1 = CHRTEST_UNINIT; |
4840 | REGCP_SET(ST.cp); | |
6407bf3b | 4841 | |
40a82448 DM |
4842 | if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */ |
4843 | goto curlym_do_B; | |
4844 | ||
4845 | curlym_do_A: /* execute the A in /A{m,n}B/ */ | |
6407bf3b | 4846 | PL_reginput = locinput; |
40a82448 DM |
4847 | PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */ |
4848 | /* NOTREACHED */ | |
5f80c4cf | 4849 | |
40a82448 DM |
4850 | case CURLYM_A: /* we've just matched an A */ |
4851 | locinput = st->locinput; | |
4852 | nextchr = UCHARAT(locinput); | |
4853 | ||
4854 | ST.count++; | |
4855 | /* after first match, determine A's length: u.curlym.alen */ | |
4856 | if (ST.count == 1) { | |
4857 | if (PL_reg_match_utf8) { | |
4858 | char *s = locinput; | |
4859 | while (s < PL_reginput) { | |
4860 | ST.alen++; | |
4861 | s += UTF8SKIP(s); | |
4862 | } | |
4863 | } | |
4864 | else { | |
4865 | ST.alen = PL_reginput - locinput; | |
4866 | } | |
4867 | if (ST.alen == 0) | |
4868 | ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me); | |
4869 | } | |
0cadcf80 DM |
4870 | DEBUG_EXECUTE_r( |
4871 | PerlIO_printf(Perl_debug_log, | |
40a82448 | 4872 | "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n", |
5bc10b2c | 4873 | (int)(REPORT_CODE_OFF+(depth*2)), "", |
40a82448 | 4874 | (IV) ST.count, (IV)ST.alen) |
0cadcf80 DM |
4875 | ); |
4876 | ||
40a82448 | 4877 | locinput = PL_reginput; |
0a4db386 YO |
4878 | |
4879 | if (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 4880 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
4881 | goto fake_end; |
4882 | ||
c966426a DM |
4883 | { |
4884 | I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)); | |
4885 | if ( max == REG_INFTY || ST.count < max ) | |
4886 | goto curlym_do_A; /* try to match another A */ | |
4887 | } | |
40a82448 | 4888 | goto curlym_do_B; /* try to match B */ |
5f80c4cf | 4889 | |
40a82448 DM |
4890 | case CURLYM_A_fail: /* just failed to match an A */ |
4891 | REGCP_UNWIND(ST.cp); | |
0a4db386 YO |
4892 | |
4893 | if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ | |
4894 | || (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 4895 | cur_eval->u.eval.close_paren == (U32)ST.me->flags)) |
40a82448 | 4896 | sayNO; |
0cadcf80 | 4897 | |
40a82448 DM |
4898 | curlym_do_B: /* execute the B in /A{m,n}B/ */ |
4899 | PL_reginput = locinput; | |
4900 | if (ST.c1 == CHRTEST_UNINIT) { | |
4901 | /* calculate c1 and c2 for possible match of 1st char | |
4902 | * following curly */ | |
4903 | ST.c1 = ST.c2 = CHRTEST_VOID; | |
4904 | if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) { | |
4905 | regnode *text_node = ST.B; | |
4906 | if (! HAS_TEXT(text_node)) | |
4907 | FIND_NEXT_IMPT(text_node); | |
ee9b8eae YO |
4908 | /* this used to be |
4909 | ||
4910 | (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT) | |
4911 | ||
4912 | But the former is redundant in light of the latter. | |
4913 | ||
4914 | if this changes back then the macro for | |
4915 | IS_TEXT and friends need to change. | |
4916 | */ | |
4917 | if (PL_regkind[OP(text_node)] == EXACT) | |
40a82448 | 4918 | { |
ee9b8eae | 4919 | |
40a82448 | 4920 | ST.c1 = (U8)*STRING(text_node); |
9a5a5549 KW |
4921 | switch (OP(text_node)) { |
4922 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
4923 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; | |
4924 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
4925 | default: ST.c2 = ST.c1; | |
4926 | } | |
c277df42 | 4927 | } |
c277df42 | 4928 | } |
40a82448 DM |
4929 | } |
4930 | ||
4931 | DEBUG_EXECUTE_r( | |
4932 | PerlIO_printf(Perl_debug_log, | |
4933 | "%*s CURLYM trying tail with matches=%"IVdf"...\n", | |
5bc10b2c | 4934 | (int)(REPORT_CODE_OFF+(depth*2)), |
40a82448 DM |
4935 | "", (IV)ST.count) |
4936 | ); | |
4937 | if (ST.c1 != CHRTEST_VOID | |
4938 | && UCHARAT(PL_reginput) != ST.c1 | |
4939 | && UCHARAT(PL_reginput) != ST.c2) | |
4940 | { | |
4941 | /* simulate B failing */ | |
3e901dc0 YO |
4942 | DEBUG_OPTIMISE_r( |
4943 | PerlIO_printf(Perl_debug_log, | |
4944 | "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n", | |
4945 | (int)(REPORT_CODE_OFF+(depth*2)),"", | |
4946 | (IV)ST.c1,(IV)ST.c2 | |
4947 | )); | |
40a82448 DM |
4948 | state_num = CURLYM_B_fail; |
4949 | goto reenter_switch; | |
4950 | } | |
4951 | ||
4952 | if (ST.me->flags) { | |
4953 | /* mark current A as captured */ | |
4954 | I32 paren = ST.me->flags; | |
4955 | if (ST.count) { | |
f0ab9afb | 4956 | PL_regoffs[paren].start |
40a82448 | 4957 | = HOPc(PL_reginput, -ST.alen) - PL_bostr; |
f0ab9afb | 4958 | PL_regoffs[paren].end = PL_reginput - PL_bostr; |
0a4db386 | 4959 | /*dmq: *PL_reglastcloseparen = paren; */ |
c277df42 | 4960 | } |
40a82448 | 4961 | else |
f0ab9afb | 4962 | PL_regoffs[paren].end = -1; |
0a4db386 | 4963 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 4964 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
4965 | { |
4966 | if (ST.count) | |
4967 | goto fake_end; | |
4968 | else | |
4969 | sayNO; | |
4970 | } | |
c277df42 | 4971 | } |
0a4db386 | 4972 | |
40a82448 | 4973 | PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */ |
5f66b61c | 4974 | /* NOTREACHED */ |
40a82448 DM |
4975 | |
4976 | case CURLYM_B_fail: /* just failed to match a B */ | |
4977 | REGCP_UNWIND(ST.cp); | |
4978 | if (ST.minmod) { | |
84d2fa14 HS |
4979 | I32 max = ARG2(ST.me); |
4980 | if (max != REG_INFTY && ST.count == max) | |
40a82448 DM |
4981 | sayNO; |
4982 | goto curlym_do_A; /* try to match a further A */ | |
4983 | } | |
4984 | /* backtrack one A */ | |
4985 | if (ST.count == ARG1(ST.me) /* min */) | |
4986 | sayNO; | |
4987 | ST.count--; | |
4988 | locinput = HOPc(locinput, -ST.alen); | |
4989 | goto curlym_do_B; /* try to match B */ | |
4990 | ||
c255a977 DM |
4991 | #undef ST |
4992 | #define ST st->u.curly | |
40a82448 | 4993 | |
c255a977 DM |
4994 | #define CURLY_SETPAREN(paren, success) \ |
4995 | if (paren) { \ | |
4996 | if (success) { \ | |
f0ab9afb NC |
4997 | PL_regoffs[paren].start = HOPc(locinput, -1) - PL_bostr; \ |
4998 | PL_regoffs[paren].end = locinput - PL_bostr; \ | |
0a4db386 | 4999 | *PL_reglastcloseparen = paren; \ |
c255a977 DM |
5000 | } \ |
5001 | else \ | |
f0ab9afb | 5002 | PL_regoffs[paren].end = -1; \ |
c255a977 DM |
5003 | } |
5004 | ||
5005 | case STAR: /* /A*B/ where A is width 1 */ | |
5006 | ST.paren = 0; | |
5007 | ST.min = 0; | |
5008 | ST.max = REG_INFTY; | |
a0d0e21e LW |
5009 | scan = NEXTOPER(scan); |
5010 | goto repeat; | |
c255a977 DM |
5011 | case PLUS: /* /A+B/ where A is width 1 */ |
5012 | ST.paren = 0; | |
5013 | ST.min = 1; | |
5014 | ST.max = REG_INFTY; | |
c277df42 | 5015 | scan = NEXTOPER(scan); |
c255a977 DM |
5016 | goto repeat; |
5017 | case CURLYN: /* /(A){m,n}B/ where A is width 1 */ | |
5018 | ST.paren = scan->flags; /* Which paren to set */ | |
5019 | if (ST.paren > PL_regsize) | |
5020 | PL_regsize = ST.paren; | |
3b6647e0 | 5021 | if (ST.paren > *PL_reglastparen) |
c255a977 DM |
5022 | *PL_reglastparen = ST.paren; |
5023 | ST.min = ARG1(scan); /* min to match */ | |
5024 | ST.max = ARG2(scan); /* max to match */ | |
0a4db386 | 5025 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5026 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5027 | ST.min=1; |
5028 | ST.max=1; | |
5029 | } | |
c255a977 DM |
5030 | scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); |
5031 | goto repeat; | |
5032 | case CURLY: /* /A{m,n}B/ where A is width 1 */ | |
5033 | ST.paren = 0; | |
5034 | ST.min = ARG1(scan); /* min to match */ | |
5035 | ST.max = ARG2(scan); /* max to match */ | |
5036 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; | |
c277df42 | 5037 | repeat: |
a0d0e21e LW |
5038 | /* |
5039 | * Lookahead to avoid useless match attempts | |
5040 | * when we know what character comes next. | |
c255a977 | 5041 | * |
5f80c4cf JP |
5042 | * Used to only do .*x and .*?x, but now it allows |
5043 | * for )'s, ('s and (?{ ... })'s to be in the way | |
5044 | * of the quantifier and the EXACT-like node. -- japhy | |
5045 | */ | |
5046 | ||
c255a977 DM |
5047 | if (ST.min > ST.max) /* XXX make this a compile-time check? */ |
5048 | sayNO; | |
cca55fe3 | 5049 | if (HAS_TEXT(next) || JUMPABLE(next)) { |
5f80c4cf JP |
5050 | U8 *s; |
5051 | regnode *text_node = next; | |
5052 | ||
3dab1dad YO |
5053 | if (! HAS_TEXT(text_node)) |
5054 | FIND_NEXT_IMPT(text_node); | |
5f80c4cf | 5055 | |
9e137952 | 5056 | if (! HAS_TEXT(text_node)) |
c255a977 | 5057 | ST.c1 = ST.c2 = CHRTEST_VOID; |
5f80c4cf | 5058 | else { |
ee9b8eae | 5059 | if ( PL_regkind[OP(text_node)] != EXACT ) { |
c255a977 | 5060 | ST.c1 = ST.c2 = CHRTEST_VOID; |
44a68960 | 5061 | goto assume_ok_easy; |
cca55fe3 | 5062 | } |
be8e71aa YO |
5063 | else |
5064 | s = (U8*)STRING(text_node); | |
ee9b8eae YO |
5065 | |
5066 | /* Currently we only get here when | |
5067 | ||
5068 | PL_rekind[OP(text_node)] == EXACT | |
5069 | ||
5070 | if this changes back then the macro for IS_TEXT and | |
5071 | friends need to change. */ | |
f2ed9b32 | 5072 | if (!UTF_PATTERN) { |
9a5a5549 KW |
5073 | ST.c1 = *s; |
5074 | switch (OP(text_node)) { | |
5075 | case EXACTF: ST.c2 = PL_fold[ST.c1]; break; | |
5076 | case EXACTFU: ST.c2 = PL_fold_latin1[ST.c1]; break; | |
5077 | case EXACTFL: ST.c2 = PL_fold_locale[ST.c1]; break; | |
5078 | default: ST.c2 = ST.c1; break; | |
5079 | } | |
1aa99e6b | 5080 | } |
f2ed9b32 | 5081 | else { /* UTF_PATTERN */ |
9a5a5549 | 5082 | if (IS_TEXTFU(text_node) || IS_TEXTF(text_node)) { |
a2a2844f | 5083 | STRLEN ulen1, ulen2; |
89ebb4a3 JH |
5084 | U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; |
5085 | U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; | |
a2a2844f JH |
5086 | |
5087 | to_utf8_lower((U8*)s, tmpbuf1, &ulen1); | |
5088 | to_utf8_upper((U8*)s, tmpbuf2, &ulen2); | |
e294cc5d JH |
5089 | #ifdef EBCDIC |
5090 | ST.c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN, 0, | |
5091 | ckWARN(WARN_UTF8) ? | |
5092 | 0 : UTF8_ALLOW_ANY); | |
5093 | ST.c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN, 0, | |
5094 | ckWARN(WARN_UTF8) ? | |
5095 | 0 : UTF8_ALLOW_ANY); | |
5096 | #else | |
c255a977 | 5097 | ST.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0, |
e294cc5d | 5098 | uniflags); |
c255a977 | 5099 | ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0, |
e294cc5d JH |
5100 | uniflags); |
5101 | #endif | |
5f80c4cf JP |
5102 | } |
5103 | else { | |
c255a977 | 5104 | ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, |
041457d9 | 5105 | uniflags); |
5f80c4cf | 5106 | } |
1aa99e6b IH |
5107 | } |
5108 | } | |
bbce6d69 | 5109 | } |
a0d0e21e | 5110 | else |
c255a977 | 5111 | ST.c1 = ST.c2 = CHRTEST_VOID; |
cca55fe3 | 5112 | assume_ok_easy: |
c255a977 DM |
5113 | |
5114 | ST.A = scan; | |
5115 | ST.B = next; | |
3280af22 | 5116 | PL_reginput = locinput; |
24d3c4a9 DM |
5117 | if (minmod) { |
5118 | minmod = 0; | |
e2e6a0f1 | 5119 | if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min) |
4633a7c4 | 5120 | sayNO; |
c255a977 | 5121 | ST.count = ST.min; |
a0ed51b3 | 5122 | locinput = PL_reginput; |
c255a977 DM |
5123 | REGCP_SET(ST.cp); |
5124 | if (ST.c1 == CHRTEST_VOID) | |
5125 | goto curly_try_B_min; | |
5126 | ||
5127 | ST.oldloc = locinput; | |
5128 | ||
5129 | /* set ST.maxpos to the furthest point along the | |
5130 | * string that could possibly match */ | |
5131 | if (ST.max == REG_INFTY) { | |
5132 | ST.maxpos = PL_regeol - 1; | |
f2ed9b32 | 5133 | if (utf8_target) |
c255a977 DM |
5134 | while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos)) |
5135 | ST.maxpos--; | |
5136 | } | |
f2ed9b32 | 5137 | else if (utf8_target) { |
c255a977 DM |
5138 | int m = ST.max - ST.min; |
5139 | for (ST.maxpos = locinput; | |
5140 | m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--) | |
5141 | ST.maxpos += UTF8SKIP(ST.maxpos); | |
5142 | } | |
5143 | else { | |
5144 | ST.maxpos = locinput + ST.max - ST.min; | |
5145 | if (ST.maxpos >= PL_regeol) | |
5146 | ST.maxpos = PL_regeol - 1; | |
5147 | } | |
5148 | goto curly_try_B_min_known; | |
5149 | ||
5150 | } | |
5151 | else { | |
e2e6a0f1 | 5152 | ST.count = regrepeat(rex, ST.A, ST.max, depth); |
c255a977 DM |
5153 | locinput = PL_reginput; |
5154 | if (ST.count < ST.min) | |
5155 | sayNO; | |
5156 | if ((ST.count > ST.min) | |
5157 | && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL)) | |
5158 | { | |
5159 | /* A{m,n} must come at the end of the string, there's | |
5160 | * no point in backing off ... */ | |
5161 | ST.min = ST.count; | |
5162 | /* ...except that $ and \Z can match before *and* after | |
5163 | newline at the end. Consider "\n\n" =~ /\n+\Z\n/. | |
5164 | We may back off by one in this case. */ | |
5165 | if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS) | |
5166 | ST.min--; | |
5167 | } | |
5168 | REGCP_SET(ST.cp); | |
5169 | goto curly_try_B_max; | |
5170 | } | |
5171 | /* NOTREACHED */ | |
5172 | ||
5173 | ||
5174 | case CURLY_B_min_known_fail: | |
5175 | /* failed to find B in a non-greedy match where c1,c2 valid */ | |
5176 | if (ST.paren && ST.count) | |
f0ab9afb | 5177 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5178 | |
5179 | PL_reginput = locinput; /* Could be reset... */ | |
5180 | REGCP_UNWIND(ST.cp); | |
5181 | /* Couldn't or didn't -- move forward. */ | |
5182 | ST.oldloc = locinput; | |
f2ed9b32 | 5183 | if (utf8_target) |
c255a977 DM |
5184 | locinput += UTF8SKIP(locinput); |
5185 | else | |
5186 | locinput++; | |
5187 | ST.count++; | |
5188 | curly_try_B_min_known: | |
5189 | /* find the next place where 'B' could work, then call B */ | |
5190 | { | |
5191 | int n; | |
f2ed9b32 | 5192 | if (utf8_target) { |
c255a977 DM |
5193 | n = (ST.oldloc == locinput) ? 0 : 1; |
5194 | if (ST.c1 == ST.c2) { | |
5195 | STRLEN len; | |
5196 | /* set n to utf8_distance(oldloc, locinput) */ | |
5197 | while (locinput <= ST.maxpos && | |
5198 | utf8n_to_uvchr((U8*)locinput, | |
5199 | UTF8_MAXBYTES, &len, | |
5200 | uniflags) != (UV)ST.c1) { | |
5201 | locinput += len; | |
5202 | n++; | |
5203 | } | |
1aa99e6b IH |
5204 | } |
5205 | else { | |
c255a977 DM |
5206 | /* set n to utf8_distance(oldloc, locinput) */ |
5207 | while (locinput <= ST.maxpos) { | |
5208 | STRLEN len; | |
5209 | const UV c = utf8n_to_uvchr((U8*)locinput, | |
5210 | UTF8_MAXBYTES, &len, | |
5211 | uniflags); | |
5212 | if (c == (UV)ST.c1 || c == (UV)ST.c2) | |
5213 | break; | |
5214 | locinput += len; | |
5215 | n++; | |
1aa99e6b | 5216 | } |
0fe9bf95 IZ |
5217 | } |
5218 | } | |
c255a977 DM |
5219 | else { |
5220 | if (ST.c1 == ST.c2) { | |
5221 | while (locinput <= ST.maxpos && | |
5222 | UCHARAT(locinput) != ST.c1) | |
5223 | locinput++; | |
bbce6d69 | 5224 | } |
c255a977 DM |
5225 | else { |
5226 | while (locinput <= ST.maxpos | |
5227 | && UCHARAT(locinput) != ST.c1 | |
5228 | && UCHARAT(locinput) != ST.c2) | |
5229 | locinput++; | |
a0ed51b3 | 5230 | } |
c255a977 DM |
5231 | n = locinput - ST.oldloc; |
5232 | } | |
5233 | if (locinput > ST.maxpos) | |
5234 | sayNO; | |
5235 | /* PL_reginput == oldloc now */ | |
5236 | if (n) { | |
5237 | ST.count += n; | |
e2e6a0f1 | 5238 | if (regrepeat(rex, ST.A, n, depth) < n) |
4633a7c4 | 5239 | sayNO; |
a0d0e21e | 5240 | } |
c255a977 DM |
5241 | PL_reginput = locinput; |
5242 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5243 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5244 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5245 | goto fake_end; |
5246 | } | |
c255a977 | 5247 | PUSH_STATE_GOTO(CURLY_B_min_known, ST.B); |
a0d0e21e | 5248 | } |
c255a977 DM |
5249 | /* NOTREACHED */ |
5250 | ||
5251 | ||
5252 | case CURLY_B_min_fail: | |
5253 | /* failed to find B in a non-greedy match where c1,c2 invalid */ | |
5254 | if (ST.paren && ST.count) | |
f0ab9afb | 5255 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5256 | |
5257 | REGCP_UNWIND(ST.cp); | |
5258 | /* failed -- move forward one */ | |
5259 | PL_reginput = locinput; | |
e2e6a0f1 | 5260 | if (regrepeat(rex, ST.A, 1, depth)) { |
c255a977 | 5261 | ST.count++; |
a0ed51b3 | 5262 | locinput = PL_reginput; |
c255a977 DM |
5263 | if (ST.count <= ST.max || (ST.max == REG_INFTY && |
5264 | ST.count > 0)) /* count overflow ? */ | |
15272685 | 5265 | { |
c255a977 DM |
5266 | curly_try_B_min: |
5267 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5268 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5269 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5270 | goto fake_end; |
5271 | } | |
c255a977 | 5272 | PUSH_STATE_GOTO(CURLY_B_min, ST.B); |
a0d0e21e LW |
5273 | } |
5274 | } | |
4633a7c4 | 5275 | sayNO; |
c255a977 DM |
5276 | /* NOTREACHED */ |
5277 | ||
5278 | ||
5279 | curly_try_B_max: | |
5280 | /* a successful greedy match: now try to match B */ | |
40d049e4 | 5281 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5282 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
40d049e4 YO |
5283 | goto fake_end; |
5284 | } | |
c255a977 DM |
5285 | { |
5286 | UV c = 0; | |
5287 | if (ST.c1 != CHRTEST_VOID) | |
f2ed9b32 | 5288 | c = utf8_target ? utf8n_to_uvchr((U8*)PL_reginput, |
c255a977 | 5289 | UTF8_MAXBYTES, 0, uniflags) |
466787eb | 5290 | : (UV) UCHARAT(PL_reginput); |
c255a977 DM |
5291 | /* If it could work, try it. */ |
5292 | if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) { | |
5293 | CURLY_SETPAREN(ST.paren, ST.count); | |
5294 | PUSH_STATE_GOTO(CURLY_B_max, ST.B); | |
5295 | /* NOTREACHED */ | |
5296 | } | |
5297 | } | |
5298 | /* FALL THROUGH */ | |
5299 | case CURLY_B_max_fail: | |
5300 | /* failed to find B in a greedy match */ | |
5301 | if (ST.paren && ST.count) | |
f0ab9afb | 5302 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5303 | |
5304 | REGCP_UNWIND(ST.cp); | |
5305 | /* back up. */ | |
5306 | if (--ST.count < ST.min) | |
5307 | sayNO; | |
5308 | PL_reginput = locinput = HOPc(locinput, -1); | |
5309 | goto curly_try_B_max; | |
5310 | ||
5311 | #undef ST | |
5312 | ||
a0d0e21e | 5313 | case END: |
6bda09f9 | 5314 | fake_end: |
faec1544 DM |
5315 | if (cur_eval) { |
5316 | /* we've just finished A in /(??{A})B/; now continue with B */ | |
5317 | I32 tmpix; | |
faec1544 DM |
5318 | st->u.eval.toggle_reg_flags |
5319 | = cur_eval->u.eval.toggle_reg_flags; | |
5320 | PL_reg_flags ^= st->u.eval.toggle_reg_flags; | |
5321 | ||
288b8c02 NC |
5322 | st->u.eval.prev_rex = rex_sv; /* inner */ |
5323 | SETREX(rex_sv,cur_eval->u.eval.prev_rex); | |
5324 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 5325 | rexi = RXi_GET(rex); |
faec1544 | 5326 | cur_curlyx = cur_eval->u.eval.prev_curlyx; |
288b8c02 | 5327 | ReREFCNT_inc(rex_sv); |
faec1544 | 5328 | st->u.eval.cp = regcppush(0); /* Save *all* the positions. */ |
34a81e2b B |
5329 | |
5330 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ | |
5331 | PL_reglastparen = &rex->lastparen; | |
5332 | PL_reglastcloseparen = &rex->lastcloseparen; | |
5333 | ||
faec1544 DM |
5334 | REGCP_SET(st->u.eval.lastcp); |
5335 | PL_reginput = locinput; | |
5336 | ||
5337 | /* Restore parens of the outer rex without popping the | |
5338 | * savestack */ | |
5339 | tmpix = PL_savestack_ix; | |
5340 | PL_savestack_ix = cur_eval->u.eval.lastcp; | |
5341 | regcppop(rex); | |
5342 | PL_savestack_ix = tmpix; | |
5343 | ||
5344 | st->u.eval.prev_eval = cur_eval; | |
5345 | cur_eval = cur_eval->u.eval.prev_eval; | |
5346 | DEBUG_EXECUTE_r( | |
2a49f0f5 JH |
5347 | PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n", |
5348 | REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval));); | |
e7707071 YO |
5349 | if ( nochange_depth ) |
5350 | nochange_depth--; | |
5351 | ||
5352 | PUSH_YES_STATE_GOTO(EVAL_AB, | |
faec1544 DM |
5353 | st->u.eval.prev_eval->u.eval.B); /* match B */ |
5354 | } | |
5355 | ||
3b0527fe | 5356 | if (locinput < reginfo->till) { |
a3621e74 | 5357 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
7821416a IZ |
5358 | "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n", |
5359 | PL_colors[4], | |
5360 | (long)(locinput - PL_reg_starttry), | |
3b0527fe | 5361 | (long)(reginfo->till - PL_reg_starttry), |
7821416a | 5362 | PL_colors[5])); |
58e23c8d | 5363 | |
262b90c4 | 5364 | sayNO_SILENT; /* Cannot match: too short. */ |
7821416a IZ |
5365 | } |
5366 | PL_reginput = locinput; /* put where regtry can find it */ | |
262b90c4 | 5367 | sayYES; /* Success! */ |
dad79028 DM |
5368 | |
5369 | case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */ | |
5370 | DEBUG_EXECUTE_r( | |
5371 | PerlIO_printf(Perl_debug_log, | |
5372 | "%*s %ssubpattern success...%s\n", | |
5bc10b2c | 5373 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])); |
3280af22 | 5374 | PL_reginput = locinput; /* put where regtry can find it */ |
262b90c4 | 5375 | sayYES; /* Success! */ |
dad79028 | 5376 | |
40a82448 DM |
5377 | #undef ST |
5378 | #define ST st->u.ifmatch | |
5379 | ||
5380 | case SUSPEND: /* (?>A) */ | |
5381 | ST.wanted = 1; | |
9fe1d20c | 5382 | PL_reginput = locinput; |
9041c2e3 | 5383 | goto do_ifmatch; |
dad79028 | 5384 | |
40a82448 DM |
5385 | case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */ |
5386 | ST.wanted = 0; | |
dad79028 DM |
5387 | goto ifmatch_trivial_fail_test; |
5388 | ||
40a82448 DM |
5389 | case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */ |
5390 | ST.wanted = 1; | |
dad79028 | 5391 | ifmatch_trivial_fail_test: |
a0ed51b3 | 5392 | if (scan->flags) { |
52657f30 | 5393 | char * const s = HOPBACKc(locinput, scan->flags); |
dad79028 DM |
5394 | if (!s) { |
5395 | /* trivial fail */ | |
24d3c4a9 DM |
5396 | if (logical) { |
5397 | logical = 0; | |
f2338a2e | 5398 | sw = 1 - cBOOL(ST.wanted); |
dad79028 | 5399 | } |
40a82448 | 5400 | else if (ST.wanted) |
dad79028 DM |
5401 | sayNO; |
5402 | next = scan + ARG(scan); | |
5403 | if (next == scan) | |
5404 | next = NULL; | |
5405 | break; | |
5406 | } | |
efb30f32 | 5407 | PL_reginput = s; |
a0ed51b3 LW |
5408 | } |
5409 | else | |
5410 | PL_reginput = locinput; | |
5411 | ||
c277df42 | 5412 | do_ifmatch: |
40a82448 | 5413 | ST.me = scan; |
24d3c4a9 | 5414 | ST.logical = logical; |
24d786f4 YO |
5415 | logical = 0; /* XXX: reset state of logical once it has been saved into ST */ |
5416 | ||
40a82448 DM |
5417 | /* execute body of (?...A) */ |
5418 | PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan))); | |
5419 | /* NOTREACHED */ | |
5420 | ||
5421 | case IFMATCH_A_fail: /* body of (?...A) failed */ | |
5422 | ST.wanted = !ST.wanted; | |
5423 | /* FALL THROUGH */ | |
5424 | ||
5425 | case IFMATCH_A: /* body of (?...A) succeeded */ | |
24d3c4a9 | 5426 | if (ST.logical) { |
f2338a2e | 5427 | sw = cBOOL(ST.wanted); |
40a82448 DM |
5428 | } |
5429 | else if (!ST.wanted) | |
5430 | sayNO; | |
5431 | ||
5432 | if (OP(ST.me) == SUSPEND) | |
5433 | locinput = PL_reginput; | |
5434 | else { | |
5435 | locinput = PL_reginput = st->locinput; | |
5436 | nextchr = UCHARAT(locinput); | |
5437 | } | |
5438 | scan = ST.me + ARG(ST.me); | |
5439 | if (scan == ST.me) | |
5440 | scan = NULL; | |
5441 | continue; /* execute B */ | |
5442 | ||
5443 | #undef ST | |
dad79028 | 5444 | |
c277df42 | 5445 | case LONGJMP: |
c277df42 IZ |
5446 | next = scan + ARG(scan); |
5447 | if (next == scan) | |
5448 | next = NULL; | |
a0d0e21e | 5449 | break; |
54612592 | 5450 | case COMMIT: |
e2e6a0f1 YO |
5451 | reginfo->cutpoint = PL_regeol; |
5452 | /* FALLTHROUGH */ | |
5d458dd8 | 5453 | case PRUNE: |
24b23f37 | 5454 | PL_reginput = locinput; |
e2e6a0f1 | 5455 | if (!scan->flags) |
ad64d0ec | 5456 | sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
54612592 YO |
5457 | PUSH_STATE_GOTO(COMMIT_next,next); |
5458 | /* NOTREACHED */ | |
5459 | case COMMIT_next_fail: | |
5460 | no_final = 1; | |
5461 | /* FALLTHROUGH */ | |
7f69552c YO |
5462 | case OPFAIL: |
5463 | sayNO; | |
e2e6a0f1 YO |
5464 | /* NOTREACHED */ |
5465 | ||
5466 | #define ST st->u.mark | |
5467 | case MARKPOINT: | |
5468 | ST.prev_mark = mark_state; | |
5d458dd8 | 5469 | ST.mark_name = sv_commit = sv_yes_mark |
ad64d0ec | 5470 | = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
e2e6a0f1 YO |
5471 | mark_state = st; |
5472 | ST.mark_loc = PL_reginput = locinput; | |
5473 | PUSH_YES_STATE_GOTO(MARKPOINT_next,next); | |
5474 | /* NOTREACHED */ | |
5475 | case MARKPOINT_next: | |
5476 | mark_state = ST.prev_mark; | |
5477 | sayYES; | |
5478 | /* NOTREACHED */ | |
5479 | case MARKPOINT_next_fail: | |
5d458dd8 | 5480 | if (popmark && sv_eq(ST.mark_name,popmark)) |
e2e6a0f1 YO |
5481 | { |
5482 | if (ST.mark_loc > startpoint) | |
5483 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5484 | popmark = NULL; /* we found our mark */ | |
5485 | sv_commit = ST.mark_name; | |
5486 | ||
5487 | DEBUG_EXECUTE_r({ | |
5d458dd8 | 5488 | PerlIO_printf(Perl_debug_log, |
e2e6a0f1 YO |
5489 | "%*s %ssetting cutpoint to mark:%"SVf"...%s\n", |
5490 | REPORT_CODE_OFF+depth*2, "", | |
be2597df | 5491 | PL_colors[4], SVfARG(sv_commit), PL_colors[5]); |
e2e6a0f1 YO |
5492 | }); |
5493 | } | |
5494 | mark_state = ST.prev_mark; | |
5d458dd8 YO |
5495 | sv_yes_mark = mark_state ? |
5496 | mark_state->u.mark.mark_name : NULL; | |
e2e6a0f1 YO |
5497 | sayNO; |
5498 | /* NOTREACHED */ | |
5d458dd8 YO |
5499 | case SKIP: |
5500 | PL_reginput = locinput; | |
5501 | if (scan->flags) { | |
2bf803e2 | 5502 | /* (*SKIP) : if we fail we cut here*/ |
5d458dd8 | 5503 | ST.mark_name = NULL; |
e2e6a0f1 | 5504 | ST.mark_loc = locinput; |
5d458dd8 YO |
5505 | PUSH_STATE_GOTO(SKIP_next,next); |
5506 | } else { | |
2bf803e2 | 5507 | /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was, |
5d458dd8 YO |
5508 | otherwise do nothing. Meaning we need to scan |
5509 | */ | |
5510 | regmatch_state *cur = mark_state; | |
ad64d0ec | 5511 | SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
5512 | |
5513 | while (cur) { | |
5514 | if ( sv_eq( cur->u.mark.mark_name, | |
5515 | find ) ) | |
5516 | { | |
5517 | ST.mark_name = find; | |
5518 | PUSH_STATE_GOTO( SKIP_next, next ); | |
5519 | } | |
5520 | cur = cur->u.mark.prev_mark; | |
5521 | } | |
e2e6a0f1 | 5522 | } |
2bf803e2 | 5523 | /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */ |
5d458dd8 YO |
5524 | break; |
5525 | case SKIP_next_fail: | |
5526 | if (ST.mark_name) { | |
5527 | /* (*CUT:NAME) - Set up to search for the name as we | |
5528 | collapse the stack*/ | |
5529 | popmark = ST.mark_name; | |
5530 | } else { | |
5531 | /* (*CUT) - No name, we cut here.*/ | |
e2e6a0f1 YO |
5532 | if (ST.mark_loc > startpoint) |
5533 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5d458dd8 YO |
5534 | /* but we set sv_commit to latest mark_name if there |
5535 | is one so they can test to see how things lead to this | |
5536 | cut */ | |
5537 | if (mark_state) | |
5538 | sv_commit=mark_state->u.mark.mark_name; | |
5539 | } | |
e2e6a0f1 YO |
5540 | no_final = 1; |
5541 | sayNO; | |
5542 | /* NOTREACHED */ | |
5543 | #undef ST | |
32e6a07c YO |
5544 | case FOLDCHAR: |
5545 | n = ARG(scan); | |
f2ed9b32 | 5546 | if ( n == (U32)what_len_TRICKYFOLD(locinput,utf8_target,ln) ) { |
e64b1bd1 | 5547 | locinput += ln; |
ced7f090 | 5548 | } else if ( LATIN_SMALL_LETTER_SHARP_S == n && !utf8_target && !UTF_PATTERN ) { |
e64b1bd1 YO |
5549 | sayNO; |
5550 | } else { | |
5551 | U8 folded[UTF8_MAXBYTES_CASE+1]; | |
5552 | STRLEN foldlen; | |
5553 | const char * const l = locinput; | |
5554 | char *e = PL_regeol; | |
5555 | to_uni_fold(n, folded, &foldlen); | |
5556 | ||
4c1b470c | 5557 | if (! foldEQ_utf8((const char*) folded, 0, foldlen, 1, |
f2ed9b32 | 5558 | l, &e, 0, utf8_target)) { |
32e6a07c | 5559 | sayNO; |
e64b1bd1 YO |
5560 | } |
5561 | locinput = e; | |
32e6a07c YO |
5562 | } |
5563 | nextchr = UCHARAT(locinput); | |
5564 | break; | |
e1d1eefb | 5565 | case LNBREAK: |
f2ed9b32 | 5566 | if ((n=is_LNBREAK(locinput,utf8_target))) { |
e1d1eefb YO |
5567 | locinput += n; |
5568 | nextchr = UCHARAT(locinput); | |
5569 | } else | |
5570 | sayNO; | |
5571 | break; | |
5572 | ||
5573 | #define CASE_CLASS(nAmE) \ | |
5574 | case nAmE: \ | |
f2ed9b32 | 5575 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5576 | locinput += n; \ |
5577 | nextchr = UCHARAT(locinput); \ | |
5578 | } else \ | |
5579 | sayNO; \ | |
5580 | break; \ | |
5581 | case N##nAmE: \ | |
f2ed9b32 | 5582 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5583 | sayNO; \ |
5584 | } else { \ | |
5585 | locinput += UTF8SKIP(locinput); \ | |
5586 | nextchr = UCHARAT(locinput); \ | |
5587 | } \ | |
5588 | break | |
5589 | ||
5590 | CASE_CLASS(VERTWS); | |
5591 | CASE_CLASS(HORIZWS); | |
5592 | #undef CASE_CLASS | |
5593 | ||
a0d0e21e | 5594 | default: |
b900a521 | 5595 | PerlIO_printf(Perl_error_log, "%"UVxf" %d\n", |
d7d93a81 | 5596 | PTR2UV(scan), OP(scan)); |
cea2e8a9 | 5597 | Perl_croak(aTHX_ "regexp memory corruption"); |
5d458dd8 YO |
5598 | |
5599 | } /* end switch */ | |
95b24440 | 5600 | |
5d458dd8 YO |
5601 | /* switch break jumps here */ |
5602 | scan = next; /* prepare to execute the next op and ... */ | |
5603 | continue; /* ... jump back to the top, reusing st */ | |
95b24440 DM |
5604 | /* NOTREACHED */ |
5605 | ||
40a82448 DM |
5606 | push_yes_state: |
5607 | /* push a state that backtracks on success */ | |
5608 | st->u.yes.prev_yes_state = yes_state; | |
5609 | yes_state = st; | |
5610 | /* FALL THROUGH */ | |
5611 | push_state: | |
5612 | /* push a new regex state, then continue at scan */ | |
5613 | { | |
5614 | regmatch_state *newst; | |
5615 | ||
24b23f37 YO |
5616 | DEBUG_STACK_r({ |
5617 | regmatch_state *cur = st; | |
5618 | regmatch_state *curyes = yes_state; | |
5619 | int curd = depth; | |
5620 | regmatch_slab *slab = PL_regmatch_slab; | |
5621 | for (;curd > -1;cur--,curd--) { | |
5622 | if (cur < SLAB_FIRST(slab)) { | |
5623 | slab = slab->prev; | |
5624 | cur = SLAB_LAST(slab); | |
5625 | } | |
5626 | PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n", | |
5627 | REPORT_CODE_OFF + 2 + depth * 2,"", | |
13d6edb4 | 5628 | curd, PL_reg_name[cur->resume_state], |
24b23f37 YO |
5629 | (curyes == cur) ? "yes" : "" |
5630 | ); | |
5631 | if (curyes == cur) | |
5632 | curyes = cur->u.yes.prev_yes_state; | |
5633 | } | |
5634 | } else | |
5635 | DEBUG_STATE_pp("push") | |
5636 | ); | |
40a82448 | 5637 | depth++; |
40a82448 DM |
5638 | st->locinput = locinput; |
5639 | newst = st+1; | |
5640 | if (newst > SLAB_LAST(PL_regmatch_slab)) | |
5641 | newst = S_push_slab(aTHX); | |
5642 | PL_regmatch_state = newst; | |
786e8c11 | 5643 | |
40a82448 DM |
5644 | locinput = PL_reginput; |
5645 | nextchr = UCHARAT(locinput); | |
5646 | st = newst; | |
5647 | continue; | |
5648 | /* NOTREACHED */ | |
5649 | } | |
a0d0e21e | 5650 | } |
a687059c | 5651 | |
a0d0e21e LW |
5652 | /* |
5653 | * We get here only if there's trouble -- normally "case END" is | |
5654 | * the terminating point. | |
5655 | */ | |
cea2e8a9 | 5656 | Perl_croak(aTHX_ "corrupted regexp pointers"); |
a0d0e21e | 5657 | /*NOTREACHED*/ |
4633a7c4 LW |
5658 | sayNO; |
5659 | ||
262b90c4 | 5660 | yes: |
77cb431f DM |
5661 | if (yes_state) { |
5662 | /* we have successfully completed a subexpression, but we must now | |
5663 | * pop to the state marked by yes_state and continue from there */ | |
77cb431f | 5664 | assert(st != yes_state); |
5bc10b2c DM |
5665 | #ifdef DEBUGGING |
5666 | while (st != yes_state) { | |
5667 | st--; | |
5668 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
5669 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5670 | st = SLAB_LAST(PL_regmatch_slab); | |
5671 | } | |
e2e6a0f1 | 5672 | DEBUG_STATE_r({ |
54612592 YO |
5673 | if (no_final) { |
5674 | DEBUG_STATE_pp("pop (no final)"); | |
5675 | } else { | |
5676 | DEBUG_STATE_pp("pop (yes)"); | |
5677 | } | |
e2e6a0f1 | 5678 | }); |
5bc10b2c DM |
5679 | depth--; |
5680 | } | |
5681 | #else | |
77cb431f DM |
5682 | while (yes_state < SLAB_FIRST(PL_regmatch_slab) |
5683 | || yes_state > SLAB_LAST(PL_regmatch_slab)) | |
5684 | { | |
5685 | /* not in this slab, pop slab */ | |
5686 | depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1); | |
5687 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5688 | st = SLAB_LAST(PL_regmatch_slab); | |
5689 | } | |
5690 | depth -= (st - yes_state); | |
5bc10b2c | 5691 | #endif |
77cb431f DM |
5692 | st = yes_state; |
5693 | yes_state = st->u.yes.prev_yes_state; | |
5694 | PL_regmatch_state = st; | |
24b23f37 | 5695 | |
5d458dd8 YO |
5696 | if (no_final) { |
5697 | locinput= st->locinput; | |
5698 | nextchr = UCHARAT(locinput); | |
5699 | } | |
54612592 | 5700 | state_num = st->resume_state + no_final; |
24d3c4a9 | 5701 | goto reenter_switch; |
77cb431f DM |
5702 | } |
5703 | ||
a3621e74 | 5704 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", |
e4584336 | 5705 | PL_colors[4], PL_colors[5])); |
02db2b7b | 5706 | |
19b95bf0 DM |
5707 | if (PL_reg_eval_set) { |
5708 | /* each successfully executed (?{...}) block does the equivalent of | |
5709 | * local $^R = do {...} | |
5710 | * When popping the save stack, all these locals would be undone; | |
5711 | * bypass this by setting the outermost saved $^R to the latest | |
5712 | * value */ | |
5713 | if (oreplsv != GvSV(PL_replgv)) | |
5714 | sv_setsv(oreplsv, GvSV(PL_replgv)); | |
5715 | } | |
95b24440 | 5716 | result = 1; |
aa283a38 | 5717 | goto final_exit; |
4633a7c4 LW |
5718 | |
5719 | no: | |
a3621e74 | 5720 | DEBUG_EXECUTE_r( |
7821416a | 5721 | PerlIO_printf(Perl_debug_log, |
786e8c11 | 5722 | "%*s %sfailed...%s\n", |
5bc10b2c | 5723 | REPORT_CODE_OFF+depth*2, "", |
786e8c11 | 5724 | PL_colors[4], PL_colors[5]) |
7821416a | 5725 | ); |
aa283a38 | 5726 | |
262b90c4 | 5727 | no_silent: |
54612592 YO |
5728 | if (no_final) { |
5729 | if (yes_state) { | |
5730 | goto yes; | |
5731 | } else { | |
5732 | goto final_exit; | |
5733 | } | |
5734 | } | |
aa283a38 DM |
5735 | if (depth) { |
5736 | /* there's a previous state to backtrack to */ | |
40a82448 DM |
5737 | st--; |
5738 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
5739 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5740 | st = SLAB_LAST(PL_regmatch_slab); | |
5741 | } | |
5742 | PL_regmatch_state = st; | |
40a82448 DM |
5743 | locinput= st->locinput; |
5744 | nextchr = UCHARAT(locinput); | |
5745 | ||
5bc10b2c DM |
5746 | DEBUG_STATE_pp("pop"); |
5747 | depth--; | |
262b90c4 DM |
5748 | if (yes_state == st) |
5749 | yes_state = st->u.yes.prev_yes_state; | |
5bc10b2c | 5750 | |
24d3c4a9 DM |
5751 | state_num = st->resume_state + 1; /* failure = success + 1 */ |
5752 | goto reenter_switch; | |
95b24440 | 5753 | } |
24d3c4a9 | 5754 | result = 0; |
aa283a38 | 5755 | |
262b90c4 | 5756 | final_exit: |
bbe252da | 5757 | if (rex->intflags & PREGf_VERBARG_SEEN) { |
5d458dd8 YO |
5758 | SV *sv_err = get_sv("REGERROR", 1); |
5759 | SV *sv_mrk = get_sv("REGMARK", 1); | |
5760 | if (result) { | |
e2e6a0f1 | 5761 | sv_commit = &PL_sv_no; |
5d458dd8 YO |
5762 | if (!sv_yes_mark) |
5763 | sv_yes_mark = &PL_sv_yes; | |
5764 | } else { | |
5765 | if (!sv_commit) | |
5766 | sv_commit = &PL_sv_yes; | |
5767 | sv_yes_mark = &PL_sv_no; | |
5768 | } | |
5769 | sv_setsv(sv_err, sv_commit); | |
5770 | sv_setsv(sv_mrk, sv_yes_mark); | |
e2e6a0f1 | 5771 | } |
19b95bf0 | 5772 | |
2f554ef7 DM |
5773 | /* clean up; in particular, free all slabs above current one */ |
5774 | LEAVE_SCOPE(oldsave); | |
5d9a96ca | 5775 | |
95b24440 | 5776 | return result; |
a687059c LW |
5777 | } |
5778 | ||
5779 | /* | |
5780 | - regrepeat - repeatedly match something simple, report how many | |
5781 | */ | |
5782 | /* | |
5783 | * [This routine now assumes that it will only match on things of length 1. | |
5784 | * That was true before, but now we assume scan - reginput is the count, | |
a0ed51b3 | 5785 | * rather than incrementing count on every character. [Er, except utf8.]] |
a687059c | 5786 | */ |
76e3520e | 5787 | STATIC I32 |
e2e6a0f1 | 5788 | S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth) |
a687059c | 5789 | { |
27da23d5 | 5790 | dVAR; |
a0d0e21e | 5791 | register char *scan; |
a0d0e21e | 5792 | register I32 c; |
3280af22 | 5793 | register char *loceol = PL_regeol; |
a0ed51b3 | 5794 | register I32 hardcount = 0; |
f2ed9b32 | 5795 | register bool utf8_target = PL_reg_match_utf8; |
4f55667c SP |
5796 | #ifndef DEBUGGING |
5797 | PERL_UNUSED_ARG(depth); | |
5798 | #endif | |
a0d0e21e | 5799 | |
7918f24d NC |
5800 | PERL_ARGS_ASSERT_REGREPEAT; |
5801 | ||
3280af22 | 5802 | scan = PL_reginput; |
faf11cac HS |
5803 | if (max == REG_INFTY) |
5804 | max = I32_MAX; | |
5805 | else if (max < loceol - scan) | |
7f596f4c | 5806 | loceol = scan + max; |
a0d0e21e | 5807 | switch (OP(p)) { |
22c35a8c | 5808 | case REG_ANY: |
f2ed9b32 | 5809 | if (utf8_target) { |
ffc61ed2 | 5810 | loceol = PL_regeol; |
1aa99e6b | 5811 | while (scan < loceol && hardcount < max && *scan != '\n') { |
ffc61ed2 JH |
5812 | scan += UTF8SKIP(scan); |
5813 | hardcount++; | |
5814 | } | |
5815 | } else { | |
5816 | while (scan < loceol && *scan != '\n') | |
5817 | scan++; | |
a0ed51b3 LW |
5818 | } |
5819 | break; | |
ffc61ed2 | 5820 | case SANY: |
f2ed9b32 | 5821 | if (utf8_target) { |
def8e4ea | 5822 | loceol = PL_regeol; |
a0804c9e | 5823 | while (scan < loceol && hardcount < max) { |
def8e4ea JH |
5824 | scan += UTF8SKIP(scan); |
5825 | hardcount++; | |
5826 | } | |
5827 | } | |
5828 | else | |
5829 | scan = loceol; | |
a0ed51b3 | 5830 | break; |
f33976b4 DB |
5831 | case CANY: |
5832 | scan = loceol; | |
5833 | break; | |
59d32103 KW |
5834 | case EXACT: |
5835 | /* To get here, EXACTish nodes must have *byte* length == 1. That | |
5836 | * means they match only characters in the string that can be expressed | |
5837 | * as a single byte. For non-utf8 strings, that means a simple match. | |
5838 | * For utf8 strings, the character matched must be an invariant, or | |
5839 | * downgradable to a single byte. The pattern's utf8ness is | |
5840 | * irrelevant, as since it's a single byte, it either isn't utf8, or if | |
5841 | * it is, it's an invariant */ | |
5842 | ||
5843 | c = (U8)*STRING(p); | |
5844 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); | |
5845 | ||
5846 | if (! utf8_target || UNI_IS_INVARIANT(c)) { | |
5847 | while (scan < loceol && UCHARAT(scan) == c) { | |
5848 | scan++; | |
5849 | } | |
5850 | } | |
5851 | else { | |
5852 | ||
5853 | /* Here, the string is utf8, and the pattern char is different | |
5854 | * in utf8 than not, so can't compare them directly. Outside the | |
5855 | * loop, find find the two utf8 bytes that represent c, and then | |
5856 | * look for those in sequence in the utf8 string */ | |
5857 | U8 high = UTF8_TWO_BYTE_HI(c); | |
5858 | U8 low = UTF8_TWO_BYTE_LO(c); | |
5859 | loceol = PL_regeol; | |
5860 | ||
5861 | while (hardcount < max | |
5862 | && scan + 1 < loceol | |
5863 | && UCHARAT(scan) == high | |
5864 | && UCHARAT(scan + 1) == low) | |
5865 | { | |
5866 | scan += 2; | |
5867 | hardcount++; | |
5868 | } | |
5869 | } | |
5870 | break; | |
d4e0b827 KW |
5871 | case EXACTFL: |
5872 | PL_reg_flags |= RF_tainted; | |
5873 | /* FALL THROUGH */ | |
d4e0b827 | 5874 | case EXACTF: |
9a5a5549 | 5875 | case EXACTFU: |
59d32103 | 5876 | |
2be3e190 KW |
5877 | /* The comments for the EXACT case above apply as well to these fold |
5878 | * ones */ | |
634c83a2 | 5879 | |
090f7165 | 5880 | c = (U8)*STRING(p); |
634c83a2 | 5881 | assert(! UTF_PATTERN || UNI_IS_INVARIANT(c)); |
d4e0b827 | 5882 | |
59d32103 KW |
5883 | if (utf8_target) { /* Use full Unicode fold matching */ |
5884 | ||
5885 | /* For the EXACTFL case, It doesn't really make sense to compare | |
5886 | * locale and utf8, but it is best we can do. The documents warn | |
5887 | * against mixing them */ | |
5888 | ||
5889 | char *tmpeol = loceol; | |
5890 | while (hardcount < max | |
5891 | && foldEQ_utf8(scan, &tmpeol, 0, utf8_target, | |
2dcac756 | 5892 | STRING(p), NULL, 1, cBOOL(UTF_PATTERN))) |
59d32103 KW |
5893 | { |
5894 | scan = tmpeol; | |
5895 | tmpeol = loceol; | |
5896 | hardcount++; | |
5897 | } | |
634c83a2 | 5898 | |
59d32103 KW |
5899 | /* XXX Note that the above handles properly the German sharp s in |
5900 | * the pattern matching ss in the string. But it doesn't handle | |
5901 | * properly cases where the string contains say 'LIGATURE ff' and | |
5902 | * the pattern is 'f+'. This would require, say, a new function or | |
5903 | * revised interface to foldEQ_utf8(), in which the maximum number | |
5904 | * of characters to match could be passed and it would return how | |
5905 | * many actually did. This is just one of many cases where | |
5906 | * multi-char folds don't work properly, and so the fix is being | |
5907 | * deferred */ | |
5908 | } | |
5909 | else { | |
87381386 | 5910 | U8 folded; |
59d32103 | 5911 | |
2be3e190 KW |
5912 | /* Here, the string isn't utf8 and c is a single byte; and either |
5913 | * the pattern isn't utf8 or c is an invariant, so its utf8ness | |
5914 | * doesn't affect c. Can just do simple comparisons for exact or | |
5915 | * fold matching. */ | |
d4e0b827 | 5916 | switch (OP(p)) { |
87381386 | 5917 | case EXACTF: folded = PL_fold[c]; break; |
9a5a5549 | 5918 | case EXACTFU: folded = PL_fold_latin1[c]; break; |
87381386 KW |
5919 | case EXACTFL: folded = PL_fold_locale[c]; break; |
5920 | default: Perl_croak(aTHX_ "panic: Unexpected op %u", OP(p)); | |
5921 | } | |
5922 | while (scan < loceol && | |
5923 | (UCHARAT(scan) == c || UCHARAT(scan) == folded)) | |
5924 | { | |
5925 | scan++; | |
634c83a2 KW |
5926 | } |
5927 | } | |
bbce6d69 | 5928 | break; |
a0d0e21e | 5929 | case ANYOF: |
f2ed9b32 | 5930 | if (utf8_target) { |
ffc61ed2 | 5931 | loceol = PL_regeol; |
cfc92286 | 5932 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 5933 | reginclass(prog, p, (U8*)scan, 0, utf8_target)) { |
ffc61ed2 JH |
5934 | scan += UTF8SKIP(scan); |
5935 | hardcount++; | |
5936 | } | |
5937 | } else { | |
32fc9b6a | 5938 | while (scan < loceol && REGINCLASS(prog, p, (U8*)scan)) |
ffc61ed2 JH |
5939 | scan++; |
5940 | } | |
a0d0e21e LW |
5941 | break; |
5942 | case ALNUM: | |
f2ed9b32 | 5943 | if (utf8_target) { |
ffc61ed2 | 5944 | loceol = PL_regeol; |
1a4fad37 | 5945 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 5946 | while (hardcount < max && scan < loceol && |
a12cf05f KW |
5947 | swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
5948 | { | |
ffc61ed2 JH |
5949 | scan += UTF8SKIP(scan); |
5950 | hardcount++; | |
5951 | } | |
a12cf05f KW |
5952 | } else if (FLAGS(p) & USE_UNI) { |
5953 | while (scan < loceol && isWORDCHAR_L1((U8) *scan)) { | |
5954 | scan++; | |
5955 | } | |
ffc61ed2 | 5956 | } else { |
a12cf05f KW |
5957 | while (scan < loceol && isALNUM((U8) *scan)) { |
5958 | scan++; | |
5959 | } | |
a0ed51b3 LW |
5960 | } |
5961 | break; | |
bbce6d69 | 5962 | case ALNUML: |
3280af22 | 5963 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 5964 | if (utf8_target) { |
ffc61ed2 | 5965 | loceol = PL_regeol; |
1aa99e6b IH |
5966 | while (hardcount < max && scan < loceol && |
5967 | isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
5968 | scan += UTF8SKIP(scan); |
5969 | hardcount++; | |
5970 | } | |
5971 | } else { | |
5972 | while (scan < loceol && isALNUM_LC(*scan)) | |
5973 | scan++; | |
a0ed51b3 LW |
5974 | } |
5975 | break; | |
a0d0e21e | 5976 | case NALNUM: |
f2ed9b32 | 5977 | if (utf8_target) { |
ffc61ed2 | 5978 | loceol = PL_regeol; |
1a4fad37 | 5979 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 5980 | while (hardcount < max && scan < loceol && |
a12cf05f KW |
5981 | !swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) |
5982 | { | |
ffc61ed2 JH |
5983 | scan += UTF8SKIP(scan); |
5984 | hardcount++; | |
5985 | } | |
a12cf05f KW |
5986 | } else if (FLAGS(p) & USE_UNI) { |
5987 | while (scan < loceol && ! isWORDCHAR_L1((U8) *scan)) { | |
5988 | scan++; | |
5989 | } | |
ffc61ed2 | 5990 | } else { |
a12cf05f KW |
5991 | while (scan < loceol && ! isALNUM((U8) *scan)) { |
5992 | scan++; | |
5993 | } | |
a0ed51b3 LW |
5994 | } |
5995 | break; | |
bbce6d69 | 5996 | case NALNUML: |
3280af22 | 5997 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 5998 | if (utf8_target) { |
ffc61ed2 | 5999 | loceol = PL_regeol; |
1aa99e6b IH |
6000 | while (hardcount < max && scan < loceol && |
6001 | !isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
6002 | scan += UTF8SKIP(scan); |
6003 | hardcount++; | |
6004 | } | |
6005 | } else { | |
6006 | while (scan < loceol && !isALNUM_LC(*scan)) | |
6007 | scan++; | |
a0ed51b3 LW |
6008 | } |
6009 | break; | |
a0d0e21e | 6010 | case SPACE: |
f2ed9b32 | 6011 | if (utf8_target) { |
ffc61ed2 | 6012 | loceol = PL_regeol; |
1a4fad37 | 6013 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6014 | while (hardcount < max && scan < loceol && |
3568d838 | 6015 | (*scan == ' ' || |
a12cf05f KW |
6016 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) |
6017 | { | |
ffc61ed2 JH |
6018 | scan += UTF8SKIP(scan); |
6019 | hardcount++; | |
6020 | } | |
a12cf05f KW |
6021 | } else if (FLAGS(p) & USE_UNI) { |
6022 | while (scan < loceol && isSPACE_L1((U8) *scan)) { | |
6023 | scan++; | |
6024 | } | |
ffc61ed2 | 6025 | } else { |
a12cf05f KW |
6026 | while (scan < loceol && isSPACE((U8) *scan)) |
6027 | scan++; | |
a0ed51b3 LW |
6028 | } |
6029 | break; | |
bbce6d69 | 6030 | case SPACEL: |
3280af22 | 6031 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6032 | if (utf8_target) { |
ffc61ed2 | 6033 | loceol = PL_regeol; |
1aa99e6b | 6034 | while (hardcount < max && scan < loceol && |
6bbba904 | 6035 | isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6036 | scan += UTF8SKIP(scan); |
6037 | hardcount++; | |
6038 | } | |
6039 | } else { | |
6040 | while (scan < loceol && isSPACE_LC(*scan)) | |
6041 | scan++; | |
a0ed51b3 LW |
6042 | } |
6043 | break; | |
a0d0e21e | 6044 | case NSPACE: |
f2ed9b32 | 6045 | if (utf8_target) { |
ffc61ed2 | 6046 | loceol = PL_regeol; |
1a4fad37 | 6047 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 6048 | while (hardcount < max && scan < loceol && |
3568d838 | 6049 | !(*scan == ' ' || |
a12cf05f KW |
6050 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) |
6051 | { | |
ffc61ed2 JH |
6052 | scan += UTF8SKIP(scan); |
6053 | hardcount++; | |
6054 | } | |
a12cf05f KW |
6055 | } else if (FLAGS(p) & USE_UNI) { |
6056 | while (scan < loceol && ! isSPACE_L1((U8) *scan)) { | |
6057 | scan++; | |
6058 | } | |
ffc61ed2 | 6059 | } else { |
a12cf05f KW |
6060 | while (scan < loceol && ! isSPACE((U8) *scan)) { |
6061 | scan++; | |
6062 | } | |
a0ed51b3 | 6063 | } |
0008a298 | 6064 | break; |
bbce6d69 | 6065 | case NSPACEL: |
3280af22 | 6066 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 6067 | if (utf8_target) { |
ffc61ed2 | 6068 | loceol = PL_regeol; |
1aa99e6b | 6069 | while (hardcount < max && scan < loceol && |
6bbba904 | 6070 | !isSPACE_LC_utf8((U8*)scan)) { |
ffc61ed2 JH |
6071 | scan += UTF8SKIP(scan); |
6072 | hardcount++; | |
6073 | } | |
6074 | } else { | |
6075 | while (scan < loceol && !isSPACE_LC(*scan)) | |
6076 | scan++; | |
a0ed51b3 LW |
6077 | } |
6078 | break; | |
a0d0e21e | 6079 | case DIGIT: |
f2ed9b32 | 6080 | if (utf8_target) { |
ffc61ed2 | 6081 | loceol = PL_regeol; |
1a4fad37 | 6082 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6083 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6084 | swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6085 | scan += UTF8SKIP(scan); |
6086 | hardcount++; | |
6087 | } | |
6088 | } else { | |
6089 | while (scan < loceol && isDIGIT(*scan)) | |
6090 | scan++; | |
a0ed51b3 LW |
6091 | } |
6092 | break; | |
b77393f6 KW |
6093 | case DIGITL: |
6094 | PL_reg_flags |= RF_tainted; | |
6095 | if (utf8_target) { | |
6096 | loceol = PL_regeol; | |
6097 | while (hardcount < max && scan < loceol && | |
6098 | isDIGIT_LC_utf8((U8*)scan)) { | |
6099 | scan += UTF8SKIP(scan); | |
6100 | hardcount++; | |
6101 | } | |
6102 | } else { | |
6103 | while (scan < loceol && isDIGIT_LC(*scan)) | |
6104 | scan++; | |
6105 | } | |
6106 | break; | |
a0d0e21e | 6107 | case NDIGIT: |
f2ed9b32 | 6108 | if (utf8_target) { |
ffc61ed2 | 6109 | loceol = PL_regeol; |
1a4fad37 | 6110 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 6111 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 6112 | !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
6113 | scan += UTF8SKIP(scan); |
6114 | hardcount++; | |
6115 | } | |
6116 | } else { | |
6117 | while (scan < loceol && !isDIGIT(*scan)) | |
6118 | scan++; | |
a0ed51b3 | 6119 | } |
b77393f6 KW |
6120 | case NDIGITL: |
6121 | PL_reg_flags |= RF_tainted; | |
6122 | if (utf8_target) { | |
6123 | loceol = PL_regeol; | |
6124 | while (hardcount < max && scan < loceol && | |
6125 | !isDIGIT_LC_utf8((U8*)scan)) { | |
6126 | scan += UTF8SKIP(scan); | |
6127 | hardcount++; | |
6128 | } | |
6129 | } else { | |
6130 | while (scan < loceol && !isDIGIT_LC(*scan)) | |
6131 | scan++; | |
6132 | } | |
6133 | break; | |
e1d1eefb | 6134 | case LNBREAK: |
f2ed9b32 | 6135 | if (utf8_target) { |
e1d1eefb YO |
6136 | loceol = PL_regeol; |
6137 | while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) { | |
6138 | scan += c; | |
6139 | hardcount++; | |
6140 | } | |
6141 | } else { | |
6142 | /* | |
6143 | LNBREAK can match two latin chars, which is ok, | |
6144 | because we have a null terminated string, but we | |
6145 | have to use hardcount in this situation | |
6146 | */ | |
6147 | while (scan < loceol && (c=is_LNBREAK_latin1(scan))) { | |
6148 | scan+=c; | |
6149 | hardcount++; | |
6150 | } | |
6151 | } | |
6152 | break; | |
6153 | case HORIZWS: | |
f2ed9b32 | 6154 | if (utf8_target) { |
e1d1eefb YO |
6155 | loceol = PL_regeol; |
6156 | while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) { | |
6157 | scan += c; | |
6158 | hardcount++; | |
6159 | } | |
6160 | } else { | |
6161 | while (scan < loceol && is_HORIZWS_latin1(scan)) | |
6162 | scan++; | |
6163 | } | |
a0ed51b3 | 6164 | break; |
e1d1eefb | 6165 | case NHORIZWS: |
f2ed9b32 | 6166 | if (utf8_target) { |
e1d1eefb YO |
6167 | loceol = PL_regeol; |
6168 | while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) { | |
6169 | scan += UTF8SKIP(scan); | |
6170 | hardcount++; | |
6171 | } | |
6172 | } else { | |
6173 | while (scan < loceol && !is_HORIZWS_latin1(scan)) | |
6174 | scan++; | |
6175 | ||
6176 | } | |
6177 | break; | |
6178 | case VERTWS: | |
f2ed9b32 | 6179 | if (utf8_target) { |
e1d1eefb YO |
6180 | loceol = PL_regeol; |
6181 | while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) { | |
6182 | scan += c; | |
6183 | hardcount++; | |
6184 | } | |
6185 | } else { | |
6186 | while (scan < loceol && is_VERTWS_latin1(scan)) | |
6187 | scan++; | |
6188 | ||
6189 | } | |
6190 | break; | |
6191 | case NVERTWS: | |
f2ed9b32 | 6192 | if (utf8_target) { |
e1d1eefb YO |
6193 | loceol = PL_regeol; |
6194 | while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) { | |
6195 | scan += UTF8SKIP(scan); | |
6196 | hardcount++; | |
6197 | } | |
6198 | } else { | |
6199 | while (scan < loceol && !is_VERTWS_latin1(scan)) | |
6200 | scan++; | |
6201 | ||
6202 | } | |
6203 | break; | |
6204 | ||
a0d0e21e LW |
6205 | default: /* Called on something of 0 width. */ |
6206 | break; /* So match right here or not at all. */ | |
6207 | } | |
a687059c | 6208 | |
a0ed51b3 LW |
6209 | if (hardcount) |
6210 | c = hardcount; | |
6211 | else | |
6212 | c = scan - PL_reginput; | |
3280af22 | 6213 | PL_reginput = scan; |
a687059c | 6214 | |
a3621e74 | 6215 | DEBUG_r({ |
e68ec53f | 6216 | GET_RE_DEBUG_FLAGS_DECL; |
be8e71aa | 6217 | DEBUG_EXECUTE_r({ |
e68ec53f YO |
6218 | SV * const prop = sv_newmortal(); |
6219 | regprop(prog, prop, p); | |
6220 | PerlIO_printf(Perl_debug_log, | |
be8e71aa | 6221 | "%*s %s can match %"IVdf" times out of %"IVdf"...\n", |
e2e6a0f1 | 6222 | REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max); |
a3621e74 | 6223 | }); |
be8e71aa | 6224 | }); |
9041c2e3 | 6225 | |
a0d0e21e | 6226 | return(c); |
a687059c LW |
6227 | } |
6228 | ||
c277df42 | 6229 | |
be8e71aa | 6230 | #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) |
c277df42 | 6231 | /* |
ffc61ed2 JH |
6232 | - regclass_swash - prepare the utf8 swash |
6233 | */ | |
6234 | ||
6235 | SV * | |
32fc9b6a | 6236 | Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) |
ffc61ed2 | 6237 | { |
97aff369 | 6238 | dVAR; |
9e55ce06 JH |
6239 | SV *sw = NULL; |
6240 | SV *si = NULL; | |
6241 | SV *alt = NULL; | |
f8fc2ecf YO |
6242 | RXi_GET_DECL(prog,progi); |
6243 | const struct reg_data * const data = prog ? progi->data : NULL; | |
ffc61ed2 | 6244 | |
7918f24d NC |
6245 | PERL_ARGS_ASSERT_REGCLASS_SWASH; |
6246 | ||
4f639d21 | 6247 | if (data && data->count) { |
a3b680e6 | 6248 | const U32 n = ARG(node); |
ffc61ed2 | 6249 | |
4f639d21 | 6250 | if (data->what[n] == 's') { |
ad64d0ec NC |
6251 | SV * const rv = MUTABLE_SV(data->data[n]); |
6252 | AV * const av = MUTABLE_AV(SvRV(rv)); | |
2d03de9c | 6253 | SV **const ary = AvARRAY(av); |
9e55ce06 | 6254 | SV **a, **b; |
9041c2e3 | 6255 | |
711a919c | 6256 | /* See the end of regcomp.c:S_regclass() for |
9e55ce06 JH |
6257 | * documentation of these array elements. */ |
6258 | ||
b11f357e | 6259 | si = *ary; |
fe5bfecd JH |
6260 | a = SvROK(ary[1]) ? &ary[1] : NULL; |
6261 | b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : NULL; | |
b11f357e | 6262 | |
ffc61ed2 JH |
6263 | if (a) |
6264 | sw = *a; | |
6265 | else if (si && doinit) { | |
6266 | sw = swash_init("utf8", "", si, 1, 0); | |
6267 | (void)av_store(av, 1, sw); | |
6268 | } | |
9e55ce06 JH |
6269 | if (b) |
6270 | alt = *b; | |
ffc61ed2 JH |
6271 | } |
6272 | } | |
6273 | ||
9e55ce06 JH |
6274 | if (listsvp) |
6275 | *listsvp = si; | |
6276 | if (altsvp) | |
6277 | *altsvp = alt; | |
ffc61ed2 JH |
6278 | |
6279 | return sw; | |
6280 | } | |
76234dfb | 6281 | #endif |
ffc61ed2 JH |
6282 | |
6283 | /* | |
ba7b4546 | 6284 | - reginclass - determine if a character falls into a character class |
832705d4 | 6285 | |
6698fab5 KW |
6286 | n is the ANYOF regnode |
6287 | p is the target string | |
6288 | lenp is pointer to the maximum number of bytes of how far to go in p | |
6289 | (This is assumed wthout checking to always be at least the current | |
6290 | character's size) | |
6291 | utf8_target tells whether p is in UTF-8. | |
832705d4 | 6292 | |
4b3cda86 KW |
6293 | Returns true if matched; false otherwise. If lenp is not NULL, on return |
6294 | from a successful match, the value it points to will be updated to how many | |
6295 | bytes in p were matched. If there was no match, the value is undefined, | |
6296 | possibly changed from the input. | |
eba1359e | 6297 | |
bbce6d69 | 6298 | */ |
6299 | ||
76e3520e | 6300 | STATIC bool |
f6ad78d8 | 6301 | S_reginclass(pTHX_ const regexp * const prog, register const regnode * const n, register const U8* const p, STRLEN* lenp, register const bool utf8_target) |
bbce6d69 | 6302 | { |
27da23d5 | 6303 | dVAR; |
a3b680e6 | 6304 | const char flags = ANYOF_FLAGS(n); |
bbce6d69 | 6305 | bool match = FALSE; |
cc07378b | 6306 | UV c = *p; |
f7ab54c6 | 6307 | STRLEN c_len = 0; |
6698fab5 | 6308 | STRLEN maxlen; |
1aa99e6b | 6309 | |
7918f24d NC |
6310 | PERL_ARGS_ASSERT_REGINCLASS; |
6311 | ||
4b3cda86 | 6312 | /* If c is not already the code point, get it */ |
f2ed9b32 | 6313 | if (utf8_target && !UTF8_IS_INVARIANT(c)) { |
f7ab54c6 | 6314 | c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &c_len, |
6182169b KW |
6315 | (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) |
6316 | | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY); | |
6317 | /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for | |
6318 | * UTF8_ALLOW_FFFF */ | |
f7ab54c6 | 6319 | if (c_len == (STRLEN)-1) |
e8a70c6f | 6320 | Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); |
19f67299 | 6321 | } |
a5a291f5 KW |
6322 | else { |
6323 | c_len = 1; | |
6324 | } | |
bbce6d69 | 6325 | |
a5a291f5 KW |
6326 | /* Use passed in max length, or one character if none passed in or less |
6327 | * than one character. And assume will match just one character. This is | |
6328 | * overwritten later if matched more. */ | |
4b3cda86 | 6329 | if (lenp) { |
a5a291f5 KW |
6330 | maxlen = (*lenp > c_len) ? *lenp : c_len; |
6331 | *lenp = c_len; | |
4b3cda86 KW |
6332 | |
6333 | } | |
6334 | else { | |
a5a291f5 | 6335 | maxlen = c_len; |
4b3cda86 KW |
6336 | } |
6337 | ||
7cdde544 KW |
6338 | /* If this character is potentially in the bitmap, check it */ |
6339 | if (c < 256) { | |
ffc61ed2 JH |
6340 | if (ANYOF_BITMAP_TEST(n, c)) |
6341 | match = TRUE; | |
a0ed51b3 | 6342 | |
78969a98 KW |
6343 | else if (flags & ANYOF_LOCALE) { |
6344 | PL_reg_flags |= RF_tainted; | |
6345 | ||
6346 | if ((flags & ANYOF_LOC_NONBITMAP_FOLD) | |
6347 | && ANYOF_BITMAP_TEST(n, PL_fold_locale[c])) | |
6348 | { | |
ffc61ed2 | 6349 | match = TRUE; |
78969a98 | 6350 | } |
040aea3a KW |
6351 | else if (ANYOF_CLASS_TEST_ANY_SET(n) && |
6352 | ((ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) || | |
6353 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) || | |
6354 | (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) || | |
6355 | (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) || | |
6356 | (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) || | |
6357 | (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) || | |
6358 | (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) || | |
6359 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) || | |
6360 | (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) || | |
6361 | (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) || | |
6362 | (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) || | |
6363 | (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) || | |
6364 | (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) || | |
6365 | (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) || | |
6366 | (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) || | |
6367 | (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) || | |
6368 | (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) || | |
6369 | (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) || | |
6370 | (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) || | |
6371 | (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) || | |
6372 | (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) || | |
6373 | (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) || | |
6374 | (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) || | |
6375 | (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) || | |
6376 | (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) || | |
6377 | (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) || | |
6378 | (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) || | |
6379 | (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) || | |
6380 | (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) || | |
6381 | (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c)) | |
6382 | ) /* How's that for a conditional? */ | |
78969a98 | 6383 | ) { |
ffc61ed2 JH |
6384 | match = TRUE; |
6385 | } | |
a0ed51b3 | 6386 | } |
a0ed51b3 LW |
6387 | } |
6388 | ||
7cdde544 KW |
6389 | /* If the bitmap didn't (or couldn't) match, and something outside the |
6390 | * bitmap could match, try that */ | |
ef87b810 | 6391 | if (!match) { |
e2e75538 KW |
6392 | if (utf8_target && (flags & ANYOF_UNICODE_ALL)) { |
6393 | if (c >= 256 | |
78969a98 KW |
6394 | || ((flags & ANYOF_LOC_NONBITMAP_FOLD) /* Latin1 1 that has a |
6395 | non-Latin1 fold | |
6396 | should match */ | |
e2e75538 KW |
6397 | && _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c))) |
6398 | { | |
6399 | match = TRUE; | |
6400 | } | |
e051a21d | 6401 | } |
e2e75538 KW |
6402 | if (!match && ((flags & ANYOF_NONBITMAP_NON_UTF8) |
6403 | || (utf8_target && flags & ANYOF_UTF8))) | |
ef87b810 | 6404 | { |
7cdde544 KW |
6405 | AV *av; |
6406 | SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av); | |
6407 | ||
6408 | if (sw) { | |
6409 | U8 * utf8_p; | |
6410 | if (utf8_target) { | |
6411 | utf8_p = (U8 *) p; | |
6412 | } else { | |
6413 | STRLEN len = 1; | |
6414 | utf8_p = bytes_to_utf8(p, &len); | |
6415 | } | |
6416 | if (swash_fetch(sw, utf8_p, 1)) | |
6417 | match = TRUE; | |
39065660 | 6418 | else if (flags & ANYOF_LOC_NONBITMAP_FOLD) { |
7cdde544 KW |
6419 | if (!match && lenp && av) { |
6420 | I32 i; | |
6421 | for (i = 0; i <= av_len(av); i++) { | |
6422 | SV* const sv = *av_fetch(av, i, FALSE); | |
6423 | STRLEN len; | |
6424 | const char * const s = SvPV_const(sv, len); | |
6425 | if (len <= maxlen && memEQ(s, (char*)utf8_p, len)) { | |
6426 | *lenp = len; | |
6427 | match = TRUE; | |
6428 | break; | |
6429 | } | |
6430 | } | |
6431 | } | |
4ee329b3 | 6432 | if (!match) { /* See if the folded version matches */ |
a85c03da | 6433 | U8 folded[UTF8_MAXBYTES_CASE+1]; |
4ee329b3 | 6434 | SV** listp; |
a85c03da | 6435 | STRLEN foldlen; |
4ee329b3 | 6436 | |
a85c03da | 6437 | to_utf8_fold(utf8_p, folded, &foldlen); |
e34d62e9 | 6438 | |
4ee329b3 KW |
6439 | /* Consider "k" =~ /[K]/i. The line above would have |
6440 | * just folded the 'k' to itself, and that isn't going | |
6441 | * to match 'K'. So we look through the closure of | |
6442 | * everything that folds to 'k'. That will find the | |
6443 | * 'K'. Initialize the list, if necessary */ | |
6444 | if (! PL_utf8_foldclosures) { | |
6445 | ||
6446 | /* If the folds haven't been read in, call a fold | |
6447 | * function to force that */ | |
6448 | if (! PL_utf8_tofold) { | |
6449 | U8 dummy[UTF8_MAXBYTES+1]; | |
6450 | STRLEN dummy_len; | |
6451 | to_utf8_fold((U8*) "A", dummy, &dummy_len); | |
6452 | } | |
6453 | PL_utf8_foldclosures = | |
6454 | _swash_inversion_hash(PL_utf8_tofold); | |
6455 | } | |
2726813d | 6456 | |
4ee329b3 KW |
6457 | /* The data structure is a hash with the keys every |
6458 | * character that is folded to, like 'k', and the | |
6459 | * values each an array of everything that folds to its | |
6460 | * key. e.g. [ 'k', 'K', KELVIN_SIGN ] */ | |
6461 | if ((listp = hv_fetch(PL_utf8_foldclosures, | |
6462 | (char *) folded, foldlen, FALSE))) | |
6463 | { | |
6464 | AV* list = (AV*) *listp; | |
6465 | IV i; | |
6466 | for (i = 0; i <= av_len(list); i++) { | |
6467 | SV** try_p = av_fetch(list, i, FALSE); | |
6468 | char* try_c; | |
6469 | if (try_p == NULL) { | |
6470 | Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure"); | |
6471 | } | |
6472 | /* Don't have to worry about embeded nulls | |
6473 | * since NULL isn't folded or foldable */ | |
6474 | try_c = SvPVX(*try_p); | |
6475 | ||
6476 | /* The fold in a few cases of an above Latin1 | |
6477 | * char is in the Latin1 range, and hence may | |
6478 | * be in the bitmap */ | |
6479 | if (UTF8_IS_INVARIANT(*try_c) | |
6480 | && ANYOF_BITMAP_TEST(n, | |
6481 | UNI_TO_NATIVE(*try_c))) | |
a9ef5332 | 6482 | { |
4ee329b3 KW |
6483 | match = TRUE; |
6484 | break; | |
6485 | } | |
6486 | else if | |
6487 | (UTF8_IS_DOWNGRADEABLE_START(*try_c) | |
6488 | && ANYOF_BITMAP_TEST(n, UNI_TO_NATIVE( | |
1f933721 | 6489 | TWO_BYTE_UTF8_TO_UNI(try_c[0], |
4ee329b3 KW |
6490 | try_c[1])))) |
6491 | { | |
6492 | /* Since the fold comes from internally | |
6493 | * generated data, we can safely assume it | |
6494 | * is valid utf8 in the test above */ | |
6495 | match = TRUE; | |
6496 | break; | |
6497 | } else if (swash_fetch(sw, (U8*) try_c, 1)) { | |
6498 | match = TRUE; | |
6499 | break; | |
2726813d | 6500 | } |
4ee329b3 KW |
6501 | } |
6502 | } | |
7cdde544 KW |
6503 | } |
6504 | } | |
6505 | ||
6506 | /* If we allocated a string above, free it */ | |
6507 | if (! utf8_target) Safefree(utf8_p); | |
6508 | } | |
6509 | } | |
6510 | } | |
6511 | ||
a0ed51b3 LW |
6512 | return (flags & ANYOF_INVERT) ? !match : match; |
6513 | } | |
161b471a | 6514 | |
dfe13c55 | 6515 | STATIC U8 * |
0ce71af7 | 6516 | S_reghop3(U8 *s, I32 off, const U8* lim) |
9041c2e3 | 6517 | { |
97aff369 | 6518 | dVAR; |
7918f24d NC |
6519 | |
6520 | PERL_ARGS_ASSERT_REGHOP3; | |
6521 | ||
a0ed51b3 | 6522 | if (off >= 0) { |
1aa99e6b | 6523 | while (off-- && s < lim) { |
ffc61ed2 | 6524 | /* XXX could check well-formedness here */ |
a0ed51b3 | 6525 | s += UTF8SKIP(s); |
ffc61ed2 | 6526 | } |
a0ed51b3 LW |
6527 | } |
6528 | else { | |
1de06328 YO |
6529 | while (off++ && s > lim) { |
6530 | s--; | |
6531 | if (UTF8_IS_CONTINUED(*s)) { | |
6532 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
6533 | s--; | |
a0ed51b3 | 6534 | } |
1de06328 | 6535 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
6536 | } |
6537 | } | |
6538 | return s; | |
6539 | } | |
161b471a | 6540 | |
f9f4320a YO |
6541 | #ifdef XXX_dmq |
6542 | /* there are a bunch of places where we use two reghop3's that should | |
6543 | be replaced with this routine. but since thats not done yet | |
6544 | we ifdef it out - dmq | |
6545 | */ | |
dfe13c55 | 6546 | STATIC U8 * |
1de06328 YO |
6547 | S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim) |
6548 | { | |
6549 | dVAR; | |
7918f24d NC |
6550 | |
6551 | PERL_ARGS_ASSERT_REGHOP4; | |
6552 | ||
1de06328 YO |
6553 | if (off >= 0) { |
6554 | while (off-- && s < rlim) { | |
6555 | /* XXX could check well-formedness here */ | |
6556 | s += UTF8SKIP(s); | |
6557 | } | |
6558 | } | |
6559 | else { | |
6560 | while (off++ && s > llim) { | |
6561 | s--; | |
6562 | if (UTF8_IS_CONTINUED(*s)) { | |
6563 | while (s > llim && UTF8_IS_CONTINUATION(*s)) | |
6564 | s--; | |
6565 | } | |
6566 | /* XXX could check well-formedness here */ | |
6567 | } | |
6568 | } | |
6569 | return s; | |
6570 | } | |
f9f4320a | 6571 | #endif |
1de06328 YO |
6572 | |
6573 | STATIC U8 * | |
0ce71af7 | 6574 | S_reghopmaybe3(U8* s, I32 off, const U8* lim) |
a0ed51b3 | 6575 | { |
97aff369 | 6576 | dVAR; |
7918f24d NC |
6577 | |
6578 | PERL_ARGS_ASSERT_REGHOPMAYBE3; | |
6579 | ||
a0ed51b3 | 6580 | if (off >= 0) { |
1aa99e6b | 6581 | while (off-- && s < lim) { |
ffc61ed2 | 6582 | /* XXX could check well-formedness here */ |
a0ed51b3 | 6583 | s += UTF8SKIP(s); |
ffc61ed2 | 6584 | } |
a0ed51b3 | 6585 | if (off >= 0) |
3dab1dad | 6586 | return NULL; |
a0ed51b3 LW |
6587 | } |
6588 | else { | |
1de06328 YO |
6589 | while (off++ && s > lim) { |
6590 | s--; | |
6591 | if (UTF8_IS_CONTINUED(*s)) { | |
6592 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
6593 | s--; | |
a0ed51b3 | 6594 | } |
1de06328 | 6595 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
6596 | } |
6597 | if (off <= 0) | |
3dab1dad | 6598 | return NULL; |
a0ed51b3 LW |
6599 | } |
6600 | return s; | |
6601 | } | |
51371543 | 6602 | |
51371543 | 6603 | static void |
acfe0abc | 6604 | restore_pos(pTHX_ void *arg) |
51371543 | 6605 | { |
97aff369 | 6606 | dVAR; |
097eb12c | 6607 | regexp * const rex = (regexp *)arg; |
51371543 GS |
6608 | if (PL_reg_eval_set) { |
6609 | if (PL_reg_oldsaved) { | |
4f639d21 DM |
6610 | rex->subbeg = PL_reg_oldsaved; |
6611 | rex->sublen = PL_reg_oldsavedlen; | |
f8c7b90f | 6612 | #ifdef PERL_OLD_COPY_ON_WRITE |
4f639d21 | 6613 | rex->saved_copy = PL_nrs; |
ed252734 | 6614 | #endif |
07bc277f | 6615 | RXp_MATCH_COPIED_on(rex); |
51371543 GS |
6616 | } |
6617 | PL_reg_magic->mg_len = PL_reg_oldpos; | |
6618 | PL_reg_eval_set = 0; | |
6619 | PL_curpm = PL_reg_oldcurpm; | |
6620 | } | |
6621 | } | |
33b8afdf JH |
6622 | |
6623 | STATIC void | |
6624 | S_to_utf8_substr(pTHX_ register regexp *prog) | |
6625 | { | |
a1cac82e | 6626 | int i = 1; |
7918f24d NC |
6627 | |
6628 | PERL_ARGS_ASSERT_TO_UTF8_SUBSTR; | |
6629 | ||
a1cac82e NC |
6630 | do { |
6631 | if (prog->substrs->data[i].substr | |
6632 | && !prog->substrs->data[i].utf8_substr) { | |
6633 | SV* const sv = newSVsv(prog->substrs->data[i].substr); | |
6634 | prog->substrs->data[i].utf8_substr = sv; | |
6635 | sv_utf8_upgrade(sv); | |
610460f9 NC |
6636 | if (SvVALID(prog->substrs->data[i].substr)) { |
6637 | const U8 flags = BmFLAGS(prog->substrs->data[i].substr); | |
6638 | if (flags & FBMcf_TAIL) { | |
6639 | /* Trim the trailing \n that fbm_compile added last | |
6640 | time. */ | |
6641 | SvCUR_set(sv, SvCUR(sv) - 1); | |
6642 | /* Whilst this makes the SV technically "invalid" (as its | |
6643 | buffer is no longer followed by "\0") when fbm_compile() | |
6644 | adds the "\n" back, a "\0" is restored. */ | |
6645 | } | |
6646 | fbm_compile(sv, flags); | |
6647 | } | |
a1cac82e NC |
6648 | if (prog->substrs->data[i].substr == prog->check_substr) |
6649 | prog->check_utf8 = sv; | |
6650 | } | |
6651 | } while (i--); | |
33b8afdf JH |
6652 | } |
6653 | ||
6654 | STATIC void | |
6655 | S_to_byte_substr(pTHX_ register regexp *prog) | |
6656 | { | |
97aff369 | 6657 | dVAR; |
a1cac82e | 6658 | int i = 1; |
7918f24d NC |
6659 | |
6660 | PERL_ARGS_ASSERT_TO_BYTE_SUBSTR; | |
6661 | ||
a1cac82e NC |
6662 | do { |
6663 | if (prog->substrs->data[i].utf8_substr | |
6664 | && !prog->substrs->data[i].substr) { | |
6665 | SV* sv = newSVsv(prog->substrs->data[i].utf8_substr); | |
6666 | if (sv_utf8_downgrade(sv, TRUE)) { | |
610460f9 NC |
6667 | if (SvVALID(prog->substrs->data[i].utf8_substr)) { |
6668 | const U8 flags | |
6669 | = BmFLAGS(prog->substrs->data[i].utf8_substr); | |
6670 | if (flags & FBMcf_TAIL) { | |
6671 | /* Trim the trailing \n that fbm_compile added last | |
6672 | time. */ | |
6673 | SvCUR_set(sv, SvCUR(sv) - 1); | |
6674 | } | |
6675 | fbm_compile(sv, flags); | |
6676 | } | |
a1cac82e NC |
6677 | } else { |
6678 | SvREFCNT_dec(sv); | |
6679 | sv = &PL_sv_undef; | |
6680 | } | |
6681 | prog->substrs->data[i].substr = sv; | |
6682 | if (prog->substrs->data[i].utf8_substr == prog->check_utf8) | |
6683 | prog->check_substr = sv; | |
33b8afdf | 6684 | } |
a1cac82e | 6685 | } while (i--); |
33b8afdf | 6686 | } |
66610fdd RGS |
6687 | |
6688 | /* | |
6689 | * Local variables: | |
6690 | * c-indentation-style: bsd | |
6691 | * c-basic-offset: 4 | |
6692 | * indent-tabs-mode: t | |
6693 | * End: | |
6694 | * | |
37442d52 RGS |
6695 | * ex: set ts=8 sts=4 sw=4 noet: |
6696 | */ |