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 | ||
32fc9b6a | 97 | #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c))) |
7d3e948e | 98 | |
c277df42 IZ |
99 | /* |
100 | * Forwards. | |
101 | */ | |
102 | ||
f2ed9b32 | 103 | #define CHR_SVLEN(sv) (utf8_target ? sv_len_utf8(sv) : SvCUR(sv)) |
53c4c00c | 104 | #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b) |
a0ed51b3 | 105 | |
3dab1dad YO |
106 | #define HOPc(pos,off) \ |
107 | (char *)(PL_reg_match_utf8 \ | |
52657f30 | 108 | ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \ |
3dab1dad YO |
109 | : (U8*)(pos + off)) |
110 | #define HOPBACKc(pos, off) \ | |
07be1b83 YO |
111 | (char*)(PL_reg_match_utf8\ |
112 | ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \ | |
113 | : (pos - off >= PL_bostr) \ | |
8e11feef | 114 | ? (U8*)pos - off \ |
3dab1dad | 115 | : NULL) |
efb30f32 | 116 | |
e7409c1b | 117 | #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off)) |
1aa99e6b | 118 | #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim)) |
1aa99e6b | 119 | |
20d0b1e9 | 120 | /* these are unrolled below in the CCC_TRY_XXX defined */ |
1a4fad37 AL |
121 | #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ |
122 | 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 |
123 | |
124 | /* Doesn't do an assert to verify that is correct */ | |
125 | #define LOAD_UTF8_CHARCLASS_NO_CHECK(class) STMT_START { \ | |
126 | if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)" "); LEAVE; } } STMT_END | |
127 | ||
1a4fad37 AL |
128 | #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a") |
129 | #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0") | |
130 | #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ") | |
51371543 | 131 | |
37e2e78e | 132 | #define LOAD_UTF8_CHARCLASS_GCB() /* Grapheme cluster boundaries */ \ |
d275fa5e CB |
133 | LOAD_UTF8_CHARCLASS(X_begin, " "); \ |
134 | LOAD_UTF8_CHARCLASS(X_non_hangul, "A"); \ | |
37e2e78e KW |
135 | /* These are utf8 constants, and not utf-ebcdic constants, so the \ |
136 | * assert should likely and hopefully fail on an EBCDIC machine */ \ | |
d275fa5e | 137 | LOAD_UTF8_CHARCLASS(X_extend, "\xcc\x80"); /* U+0300 */ \ |
37e2e78e KW |
138 | \ |
139 | /* No asserts are done for these, in case called on an early \ | |
140 | * Unicode version in which they map to nothing */ \ | |
d275fa5e CB |
141 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_prepend);/* U+0E40 "\xe0\xb9\x80" */ \ |
142 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_L); /* U+1100 "\xe1\x84\x80" */ \ | |
143 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV); /* U+AC00 "\xea\xb0\x80" */ \ | |
144 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LVT); /* U+AC01 "\xea\xb0\x81" */ \ | |
145 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_LV_LVT_V);/* U+AC01 "\xea\xb0\x81" */\ | |
146 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_T); /* U+11A8 "\xe1\x86\xa8" */ \ | |
37e2e78e | 147 | LOAD_UTF8_CHARCLASS_NO_CHECK(X_V) /* U+1160 "\xe1\x85\xa0" */ |
20d0b1e9 | 148 | |
d1eb3177 YO |
149 | /* |
150 | We dont use PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS as the direct test | |
151 | so that it is possible to override the option here without having to | |
152 | rebuild the entire core. as we are required to do if we change regcomp.h | |
153 | which is where PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS is defined. | |
154 | */ | |
155 | #if PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS | |
156 | #define BROKEN_UNICODE_CHARCLASS_MAPPINGS | |
157 | #endif | |
158 | ||
159 | #ifdef BROKEN_UNICODE_CHARCLASS_MAPPINGS | |
160 | #define LOAD_UTF8_CHARCLASS_PERL_WORD() LOAD_UTF8_CHARCLASS_ALNUM() | |
161 | #define LOAD_UTF8_CHARCLASS_PERL_SPACE() LOAD_UTF8_CHARCLASS_SPACE() | |
162 | #define LOAD_UTF8_CHARCLASS_POSIX_DIGIT() LOAD_UTF8_CHARCLASS_DIGIT() | |
163 | #define RE_utf8_perl_word PL_utf8_alnum | |
164 | #define RE_utf8_perl_space PL_utf8_space | |
165 | #define RE_utf8_posix_digit PL_utf8_digit | |
166 | #define perl_word alnum | |
167 | #define perl_space space | |
168 | #define posix_digit digit | |
169 | #else | |
170 | #define LOAD_UTF8_CHARCLASS_PERL_WORD() LOAD_UTF8_CHARCLASS(perl_word,"a") | |
171 | #define LOAD_UTF8_CHARCLASS_PERL_SPACE() LOAD_UTF8_CHARCLASS(perl_space," ") | |
172 | #define LOAD_UTF8_CHARCLASS_POSIX_DIGIT() LOAD_UTF8_CHARCLASS(posix_digit,"0") | |
173 | #define RE_utf8_perl_word PL_utf8_perl_word | |
174 | #define RE_utf8_perl_space PL_utf8_perl_space | |
175 | #define RE_utf8_posix_digit PL_utf8_posix_digit | |
176 | #endif | |
177 | ||
178 | ||
20d0b1e9 YO |
179 | #define CCC_TRY_AFF(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC,LCFUNC) \ |
180 | case NAMEL: \ | |
181 | PL_reg_flags |= RF_tainted; \ | |
182 | /* FALL THROUGH */ \ | |
183 | case NAME: \ | |
184 | if (!nextchr) \ | |
185 | sayNO; \ | |
f2ed9b32 | 186 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ |
20d0b1e9 YO |
187 | if (!CAT2(PL_utf8_,CLASS)) { \ |
188 | bool ok; \ | |
189 | ENTER; \ | |
190 | save_re_context(); \ | |
191 | ok=CAT2(is_utf8_,CLASS)((const U8*)STR); \ | |
192 | assert(ok); \ | |
193 | LEAVE; \ | |
194 | } \ | |
195 | if (!(OP(scan) == NAME \ | |
f2ed9b32 | 196 | ? cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, utf8_target)) \ |
20d0b1e9 YO |
197 | : LCFUNC_utf8((U8*)locinput))) \ |
198 | { \ | |
199 | sayNO; \ | |
200 | } \ | |
201 | locinput += PL_utf8skip[nextchr]; \ | |
202 | nextchr = UCHARAT(locinput); \ | |
203 | break; \ | |
204 | } \ | |
205 | if (!(OP(scan) == NAME ? FUNC(nextchr) : LCFUNC(nextchr))) \ | |
206 | sayNO; \ | |
207 | nextchr = UCHARAT(++locinput); \ | |
208 | break | |
209 | ||
210 | #define CCC_TRY_NEG(NAME,NAMEL,CLASS,STR,LCFUNC_utf8,FUNC,LCFUNC) \ | |
211 | case NAMEL: \ | |
212 | PL_reg_flags |= RF_tainted; \ | |
213 | /* FALL THROUGH */ \ | |
214 | case NAME : \ | |
215 | if (!nextchr && locinput >= PL_regeol) \ | |
216 | sayNO; \ | |
f2ed9b32 | 217 | if (utf8_target && UTF8_IS_CONTINUED(nextchr)) { \ |
20d0b1e9 YO |
218 | if (!CAT2(PL_utf8_,CLASS)) { \ |
219 | bool ok; \ | |
220 | ENTER; \ | |
221 | save_re_context(); \ | |
222 | ok=CAT2(is_utf8_,CLASS)((const U8*)STR); \ | |
223 | assert(ok); \ | |
224 | LEAVE; \ | |
225 | } \ | |
226 | if ((OP(scan) == NAME \ | |
f2ed9b32 | 227 | ? cBOOL(swash_fetch(CAT2(PL_utf8_,CLASS), (U8*)locinput, utf8_target)) \ |
20d0b1e9 YO |
228 | : LCFUNC_utf8((U8*)locinput))) \ |
229 | { \ | |
230 | sayNO; \ | |
231 | } \ | |
232 | locinput += PL_utf8skip[nextchr]; \ | |
233 | nextchr = UCHARAT(locinput); \ | |
234 | break; \ | |
235 | } \ | |
236 | if ((OP(scan) == NAME ? FUNC(nextchr) : LCFUNC(nextchr))) \ | |
237 | sayNO; \ | |
238 | nextchr = UCHARAT(++locinput); \ | |
239 | break | |
240 | ||
241 | ||
d1eb3177 YO |
242 | |
243 | ||
244 | ||
3dab1dad YO |
245 | /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */ |
246 | ||
5f80c4cf | 247 | /* for use after a quantifier and before an EXACT-like node -- japhy */ |
3e901dc0 YO |
248 | /* it would be nice to rework regcomp.sym to generate this stuff. sigh */ |
249 | #define JUMPABLE(rn) ( \ | |
250 | OP(rn) == OPEN || \ | |
251 | (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \ | |
252 | OP(rn) == EVAL || \ | |
cca55fe3 JP |
253 | OP(rn) == SUSPEND || OP(rn) == IFMATCH || \ |
254 | OP(rn) == PLUS || OP(rn) == MINMOD || \ | |
ee9b8eae | 255 | OP(rn) == KEEPS || (PL_regkind[OP(rn)] == VERB) || \ |
3dab1dad | 256 | (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \ |
e2d8ce26 | 257 | ) |
ee9b8eae | 258 | #define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT) |
e2d8ce26 | 259 | |
ee9b8eae YO |
260 | #define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF ) |
261 | ||
262 | #if 0 | |
263 | /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so | |
264 | we don't need this definition. */ | |
265 | #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) | |
266 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF ) | |
267 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) | |
268 | ||
269 | #else | |
270 | /* ... so we use this as its faster. */ | |
271 | #define IS_TEXT(rn) ( OP(rn)==EXACT ) | |
272 | #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) | |
273 | #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) | |
274 | ||
275 | #endif | |
e2d8ce26 | 276 | |
a84d97b6 HS |
277 | /* |
278 | Search for mandatory following text node; for lookahead, the text must | |
279 | follow but for lookbehind (rn->flags != 0) we skip to the next step. | |
280 | */ | |
cca55fe3 | 281 | #define FIND_NEXT_IMPT(rn) STMT_START { \ |
3dab1dad YO |
282 | while (JUMPABLE(rn)) { \ |
283 | const OPCODE type = OP(rn); \ | |
284 | if (type == SUSPEND || PL_regkind[type] == CURLY) \ | |
e2d8ce26 | 285 | rn = NEXTOPER(NEXTOPER(rn)); \ |
3dab1dad | 286 | else if (type == PLUS) \ |
cca55fe3 | 287 | rn = NEXTOPER(rn); \ |
3dab1dad | 288 | else if (type == IFMATCH) \ |
a84d97b6 | 289 | rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \ |
e2d8ce26 | 290 | else rn += NEXT_OFF(rn); \ |
3dab1dad | 291 | } \ |
5f80c4cf | 292 | } STMT_END |
74750237 | 293 | |
c476f425 | 294 | |
acfe0abc | 295 | static void restore_pos(pTHX_ void *arg); |
51371543 | 296 | |
620d5b66 NC |
297 | #define REGCP_PAREN_ELEMS 4 |
298 | #define REGCP_OTHER_ELEMS 5 | |
e0fa7e2b | 299 | #define REGCP_FRAME_ELEMS 1 |
620d5b66 NC |
300 | /* REGCP_FRAME_ELEMS are not part of the REGCP_OTHER_ELEMS and |
301 | * are needed for the regexp context stack bookkeeping. */ | |
302 | ||
76e3520e | 303 | STATIC CHECKPOINT |
cea2e8a9 | 304 | S_regcppush(pTHX_ I32 parenfloor) |
a0d0e21e | 305 | { |
97aff369 | 306 | dVAR; |
a3b680e6 | 307 | const int retval = PL_savestack_ix; |
a3b680e6 | 308 | const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; |
e0fa7e2b NC |
309 | const UV total_elems = paren_elems_to_push + REGCP_OTHER_ELEMS; |
310 | const UV elems_shifted = total_elems << SAVE_TIGHT_SHIFT; | |
a0d0e21e | 311 | int p; |
40a82448 | 312 | GET_RE_DEBUG_FLAGS_DECL; |
a0d0e21e | 313 | |
e49a9654 IH |
314 | if (paren_elems_to_push < 0) |
315 | Perl_croak(aTHX_ "panic: paren_elems_to_push < 0"); | |
316 | ||
e0fa7e2b NC |
317 | if ((elems_shifted >> SAVE_TIGHT_SHIFT) != total_elems) |
318 | Perl_croak(aTHX_ "panic: paren_elems_to_push offset %"UVuf | |
5df417d0 JH |
319 | " out of range (%lu-%ld)", |
320 | total_elems, (unsigned long)PL_regsize, (long)parenfloor); | |
e0fa7e2b | 321 | |
620d5b66 | 322 | SSGROW(total_elems + REGCP_FRAME_ELEMS); |
7f69552c | 323 | |
3280af22 | 324 | for (p = PL_regsize; p > parenfloor; p--) { |
b1ce53c5 | 325 | /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ |
f0ab9afb NC |
326 | SSPUSHINT(PL_regoffs[p].end); |
327 | SSPUSHINT(PL_regoffs[p].start); | |
3280af22 | 328 | SSPUSHPTR(PL_reg_start_tmp[p]); |
a0d0e21e | 329 | SSPUSHINT(p); |
e7707071 | 330 | DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log, |
40a82448 | 331 | " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n", |
f0ab9afb | 332 | (UV)p, (IV)PL_regoffs[p].start, |
40a82448 | 333 | (IV)(PL_reg_start_tmp[p] - PL_bostr), |
f0ab9afb | 334 | (IV)PL_regoffs[p].end |
40a82448 | 335 | )); |
a0d0e21e | 336 | } |
b1ce53c5 | 337 | /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ |
f0ab9afb | 338 | SSPUSHPTR(PL_regoffs); |
3280af22 NIS |
339 | SSPUSHINT(PL_regsize); |
340 | SSPUSHINT(*PL_reglastparen); | |
a01268b5 | 341 | SSPUSHINT(*PL_reglastcloseparen); |
3280af22 | 342 | SSPUSHPTR(PL_reginput); |
e0fa7e2b | 343 | SSPUSHUV(SAVEt_REGCONTEXT | elems_shifted); /* Magic cookie. */ |
41123dfd | 344 | |
a0d0e21e LW |
345 | return retval; |
346 | } | |
347 | ||
c277df42 | 348 | /* These are needed since we do not localize EVAL nodes: */ |
ab3bbdeb YO |
349 | #define REGCP_SET(cp) \ |
350 | DEBUG_STATE_r( \ | |
ab3bbdeb | 351 | PerlIO_printf(Perl_debug_log, \ |
e4f74956 | 352 | " Setting an EVAL scope, savestack=%"IVdf"\n", \ |
ab3bbdeb YO |
353 | (IV)PL_savestack_ix)); \ |
354 | cp = PL_savestack_ix | |
c3464db5 | 355 | |
ab3bbdeb | 356 | #define REGCP_UNWIND(cp) \ |
e4f74956 | 357 | DEBUG_STATE_r( \ |
ab3bbdeb | 358 | if (cp != PL_savestack_ix) \ |
e4f74956 YO |
359 | PerlIO_printf(Perl_debug_log, \ |
360 | " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ | |
ab3bbdeb YO |
361 | (IV)(cp), (IV)PL_savestack_ix)); \ |
362 | regcpblow(cp) | |
c277df42 | 363 | |
76e3520e | 364 | STATIC char * |
097eb12c | 365 | S_regcppop(pTHX_ const regexp *rex) |
a0d0e21e | 366 | { |
97aff369 | 367 | dVAR; |
e0fa7e2b | 368 | UV i; |
a0d0e21e | 369 | char *input; |
a3621e74 YO |
370 | GET_RE_DEBUG_FLAGS_DECL; |
371 | ||
7918f24d NC |
372 | PERL_ARGS_ASSERT_REGCPPOP; |
373 | ||
b1ce53c5 | 374 | /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ |
c6bf6a65 | 375 | i = SSPOPUV; |
e0fa7e2b NC |
376 | assert((i & SAVE_MASK) == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ |
377 | i >>= SAVE_TIGHT_SHIFT; /* Parentheses elements to pop. */ | |
a0d0e21e | 378 | input = (char *) SSPOPPTR; |
a01268b5 | 379 | *PL_reglastcloseparen = SSPOPINT; |
3280af22 NIS |
380 | *PL_reglastparen = SSPOPINT; |
381 | PL_regsize = SSPOPINT; | |
f0ab9afb | 382 | PL_regoffs=(regexp_paren_pair *) SSPOPPTR; |
b1ce53c5 | 383 | |
620d5b66 | 384 | i -= REGCP_OTHER_ELEMS; |
b1ce53c5 | 385 | /* Now restore the parentheses context. */ |
620d5b66 | 386 | for ( ; i > 0; i -= REGCP_PAREN_ELEMS) { |
1df70142 | 387 | I32 tmps; |
097eb12c | 388 | U32 paren = (U32)SSPOPINT; |
3280af22 | 389 | PL_reg_start_tmp[paren] = (char *) SSPOPPTR; |
f0ab9afb | 390 | PL_regoffs[paren].start = SSPOPINT; |
cf93c79d | 391 | tmps = SSPOPINT; |
3280af22 | 392 | if (paren <= *PL_reglastparen) |
f0ab9afb | 393 | PL_regoffs[paren].end = tmps; |
e7707071 | 394 | DEBUG_BUFFERS_r( |
c3464db5 | 395 | PerlIO_printf(Perl_debug_log, |
b900a521 | 396 | " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n", |
f0ab9afb | 397 | (UV)paren, (IV)PL_regoffs[paren].start, |
b900a521 | 398 | (IV)(PL_reg_start_tmp[paren] - PL_bostr), |
f0ab9afb | 399 | (IV)PL_regoffs[paren].end, |
3280af22 | 400 | (paren > *PL_reglastparen ? "(no)" : "")); |
c277df42 | 401 | ); |
a0d0e21e | 402 | } |
e7707071 | 403 | DEBUG_BUFFERS_r( |
bb7a0f54 | 404 | if (*PL_reglastparen + 1 <= rex->nparens) { |
c3464db5 | 405 | PerlIO_printf(Perl_debug_log, |
faccc32b | 406 | " restoring \\%"IVdf"..\\%"IVdf" to undef\n", |
4f639d21 | 407 | (IV)(*PL_reglastparen + 1), (IV)rex->nparens); |
c277df42 IZ |
408 | } |
409 | ); | |
daf18116 | 410 | #if 1 |
dafc8851 JH |
411 | /* It would seem that the similar code in regtry() |
412 | * already takes care of this, and in fact it is in | |
413 | * a better location to since this code can #if 0-ed out | |
414 | * but the code in regtry() is needed or otherwise tests | |
415 | * requiring null fields (pat.t#187 and split.t#{13,14} | |
daf18116 JH |
416 | * (as of patchlevel 7877) will fail. Then again, |
417 | * this code seems to be necessary or otherwise | |
225593e1 DM |
418 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ |
419 | * --jhi updated by dapm */ | |
3b6647e0 | 420 | for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) { |
097eb12c | 421 | if (i > PL_regsize) |
f0ab9afb NC |
422 | PL_regoffs[i].start = -1; |
423 | PL_regoffs[i].end = -1; | |
a0d0e21e | 424 | } |
dafc8851 | 425 | #endif |
a0d0e21e LW |
426 | return input; |
427 | } | |
428 | ||
02db2b7b | 429 | #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ |
a0d0e21e | 430 | |
a687059c | 431 | /* |
e50aee73 | 432 | * pregexec and friends |
a687059c LW |
433 | */ |
434 | ||
76234dfb | 435 | #ifndef PERL_IN_XSUB_RE |
a687059c | 436 | /* |
c277df42 | 437 | - pregexec - match a regexp against a string |
a687059c | 438 | */ |
c277df42 | 439 | I32 |
49d7dfbc | 440 | Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend, |
c3464db5 | 441 | char *strbeg, I32 minend, SV *screamer, U32 nosave) |
c277df42 IZ |
442 | /* strend: pointer to null at end of string */ |
443 | /* strbeg: real beginning of string */ | |
444 | /* minend: end of match must be >=minend after stringarg. */ | |
445 | /* nosave: For optimizations. */ | |
446 | { | |
7918f24d NC |
447 | PERL_ARGS_ASSERT_PREGEXEC; |
448 | ||
c277df42 | 449 | return |
9041c2e3 | 450 | regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, |
c277df42 IZ |
451 | nosave ? 0 : REXEC_COPY_STR); |
452 | } | |
76234dfb | 453 | #endif |
22e551b9 | 454 | |
9041c2e3 | 455 | /* |
cad2e5aa JH |
456 | * Need to implement the following flags for reg_anch: |
457 | * | |
458 | * USE_INTUIT_NOML - Useful to call re_intuit_start() first | |
459 | * USE_INTUIT_ML | |
460 | * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer | |
461 | * INTUIT_AUTORITATIVE_ML | |
462 | * INTUIT_ONCE_NOML - Intuit can match in one location only. | |
463 | * INTUIT_ONCE_ML | |
464 | * | |
465 | * Another flag for this function: SECOND_TIME (so that float substrs | |
466 | * with giant delta may be not rechecked). | |
467 | */ | |
468 | ||
469 | /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */ | |
470 | ||
3f7c398e | 471 | /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend. |
cad2e5aa JH |
472 | Otherwise, only SvCUR(sv) is used to get strbeg. */ |
473 | ||
474 | /* XXXX We assume that strpos is strbeg unless sv. */ | |
475 | ||
6eb5f6b9 JH |
476 | /* XXXX Some places assume that there is a fixed substring. |
477 | An update may be needed if optimizer marks as "INTUITable" | |
478 | RExen without fixed substrings. Similarly, it is assumed that | |
479 | lengths of all the strings are no more than minlen, thus they | |
480 | cannot come from lookahead. | |
40d049e4 YO |
481 | (Or minlen should take into account lookahead.) |
482 | NOTE: Some of this comment is not correct. minlen does now take account | |
483 | of lookahead/behind. Further research is required. -- demerphq | |
484 | ||
485 | */ | |
6eb5f6b9 | 486 | |
2c2d71f5 JH |
487 | /* A failure to find a constant substring means that there is no need to make |
488 | an expensive call to REx engine, thus we celebrate a failure. Similarly, | |
489 | finding a substring too deep into the string means that less calls to | |
30944b6d IZ |
490 | regtry() should be needed. |
491 | ||
492 | REx compiler's optimizer found 4 possible hints: | |
493 | a) Anchored substring; | |
494 | b) Fixed substring; | |
495 | c) Whether we are anchored (beginning-of-line or \G); | |
496 | d) First node (of those at offset 0) which may distingush positions; | |
6eb5f6b9 | 497 | We use a)b)d) and multiline-part of c), and try to find a position in the |
30944b6d IZ |
498 | string which does not contradict any of them. |
499 | */ | |
2c2d71f5 | 500 | |
6eb5f6b9 JH |
501 | /* Most of decisions we do here should have been done at compile time. |
502 | The nodes of the REx which we used for the search should have been | |
503 | deleted from the finite automaton. */ | |
504 | ||
cad2e5aa | 505 | char * |
288b8c02 | 506 | Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, |
9f61653a | 507 | char *strend, const U32 flags, re_scream_pos_data *data) |
cad2e5aa | 508 | { |
97aff369 | 509 | dVAR; |
288b8c02 | 510 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
b7953727 | 511 | register I32 start_shift = 0; |
cad2e5aa | 512 | /* Should be nonnegative! */ |
b7953727 | 513 | register I32 end_shift = 0; |
2c2d71f5 JH |
514 | register char *s; |
515 | register SV *check; | |
a1933d95 | 516 | char *strbeg; |
cad2e5aa | 517 | char *t; |
f2ed9b32 | 518 | const bool utf8_target = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */ |
cad2e5aa | 519 | I32 ml_anch; |
bd61b366 SS |
520 | register char *other_last = NULL; /* other substr checked before this */ |
521 | char *check_at = NULL; /* check substr found at this pos */ | |
bbe252da | 522 | const I32 multiline = prog->extflags & RXf_PMf_MULTILINE; |
f8fc2ecf | 523 | RXi_GET_DECL(prog,progi); |
30944b6d | 524 | #ifdef DEBUGGING |
890ce7af | 525 | const char * const i_strpos = strpos; |
30944b6d | 526 | #endif |
a3621e74 YO |
527 | GET_RE_DEBUG_FLAGS_DECL; |
528 | ||
7918f24d NC |
529 | PERL_ARGS_ASSERT_RE_INTUIT_START; |
530 | ||
f2ed9b32 | 531 | RX_MATCH_UTF8_set(rx,utf8_target); |
cad2e5aa | 532 | |
3c8556c3 | 533 | if (RX_UTF8(rx)) { |
b8d68ded JH |
534 | PL_reg_flags |= RF_utf8; |
535 | } | |
ab3bbdeb | 536 | DEBUG_EXECUTE_r( |
f2ed9b32 | 537 | debug_start_match(rx, utf8_target, strpos, strend, |
1de06328 YO |
538 | sv ? "Guessing start of match in sv for" |
539 | : "Guessing start of match in string for"); | |
2a782b5b | 540 | ); |
cad2e5aa | 541 | |
c344f387 JH |
542 | /* CHR_DIST() would be more correct here but it makes things slow. */ |
543 | if (prog->minlen > strend - strpos) { | |
a3621e74 | 544 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 | 545 | "String too short... [re_intuit_start]\n")); |
cad2e5aa | 546 | goto fail; |
2c2d71f5 | 547 | } |
1de06328 | 548 | |
a1933d95 | 549 | strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; |
1aa99e6b | 550 | PL_regeol = strend; |
f2ed9b32 | 551 | if (utf8_target) { |
33b8afdf JH |
552 | if (!prog->check_utf8 && prog->check_substr) |
553 | to_utf8_substr(prog); | |
554 | check = prog->check_utf8; | |
555 | } else { | |
556 | if (!prog->check_substr && prog->check_utf8) | |
557 | to_byte_substr(prog); | |
558 | check = prog->check_substr; | |
559 | } | |
1de06328 | 560 | if (check == &PL_sv_undef) { |
a3621e74 | 561 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
1de06328 | 562 | "Non-utf8 string cannot match utf8 check string\n")); |
33b8afdf JH |
563 | goto fail; |
564 | } | |
bbe252da YO |
565 | if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */ |
566 | ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE) | |
567 | || ( (prog->extflags & RXf_ANCH_BOL) | |
7fba1cd6 | 568 | && !multiline ) ); /* Check after \n? */ |
cad2e5aa | 569 | |
7e25d62c | 570 | if (!ml_anch) { |
bbe252da YO |
571 | if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */ |
572 | && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */ | |
3f7c398e | 573 | /* SvCUR is not set on references: SvRV and SvPVX_const overlap */ |
7e25d62c JH |
574 | && sv && !SvROK(sv) |
575 | && (strpos != strbeg)) { | |
a3621e74 | 576 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); |
7e25d62c JH |
577 | goto fail; |
578 | } | |
579 | if (prog->check_offset_min == prog->check_offset_max && | |
bbe252da | 580 | !(prog->extflags & RXf_CANY_SEEN)) { |
2c2d71f5 | 581 | /* Substring at constant offset from beg-of-str... */ |
cad2e5aa JH |
582 | I32 slen; |
583 | ||
1aa99e6b | 584 | s = HOP3c(strpos, prog->check_offset_min, strend); |
1de06328 | 585 | |
653099ff GS |
586 | if (SvTAIL(check)) { |
587 | slen = SvCUR(check); /* >= 1 */ | |
cad2e5aa | 588 | |
9041c2e3 | 589 | if ( strend - s > slen || strend - s < slen - 1 |
2c2d71f5 | 590 | || (strend - s == slen && strend[-1] != '\n')) { |
a3621e74 | 591 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n")); |
2c2d71f5 | 592 | goto fail_finish; |
cad2e5aa JH |
593 | } |
594 | /* Now should match s[0..slen-2] */ | |
595 | slen--; | |
3f7c398e | 596 | if (slen && (*SvPVX_const(check) != *s |
cad2e5aa | 597 | || (slen > 1 |
3f7c398e | 598 | && memNE(SvPVX_const(check), s, slen)))) { |
2c2d71f5 | 599 | report_neq: |
a3621e74 | 600 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); |
2c2d71f5 JH |
601 | goto fail_finish; |
602 | } | |
cad2e5aa | 603 | } |
3f7c398e | 604 | else if (*SvPVX_const(check) != *s |
653099ff | 605 | || ((slen = SvCUR(check)) > 1 |
3f7c398e | 606 | && memNE(SvPVX_const(check), s, slen))) |
2c2d71f5 | 607 | goto report_neq; |
c315bfe8 | 608 | check_at = s; |
2c2d71f5 | 609 | goto success_at_start; |
7e25d62c | 610 | } |
cad2e5aa | 611 | } |
2c2d71f5 | 612 | /* Match is anchored, but substr is not anchored wrt beg-of-str. */ |
cad2e5aa | 613 | s = strpos; |
2c2d71f5 | 614 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
1de06328 YO |
615 | end_shift = prog->check_end_shift; |
616 | ||
2c2d71f5 | 617 | if (!ml_anch) { |
a3b680e6 | 618 | const I32 end = prog->check_offset_max + CHR_SVLEN(check) |
653099ff | 619 | - (SvTAIL(check) != 0); |
a3b680e6 | 620 | const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; |
2c2d71f5 JH |
621 | |
622 | if (end_shift < eshift) | |
623 | end_shift = eshift; | |
624 | } | |
cad2e5aa | 625 | } |
2c2d71f5 | 626 | else { /* Can match at random position */ |
cad2e5aa JH |
627 | ml_anch = 0; |
628 | s = strpos; | |
1de06328 YO |
629 | start_shift = prog->check_offset_min; /* okay to underestimate on CC */ |
630 | end_shift = prog->check_end_shift; | |
631 | ||
632 | /* end shift should be non negative here */ | |
cad2e5aa JH |
633 | } |
634 | ||
bcdf7404 | 635 | #ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */ |
0033605d | 636 | if (end_shift < 0) |
1de06328 | 637 | Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ", |
220fc49f | 638 | (IV)end_shift, RX_PRECOMP(prog)); |
2c2d71f5 JH |
639 | #endif |
640 | ||
2c2d71f5 JH |
641 | restart: |
642 | /* Find a possible match in the region s..strend by looking for | |
643 | the "check" substring in the region corrected by start/end_shift. */ | |
1de06328 YO |
644 | |
645 | { | |
646 | I32 srch_start_shift = start_shift; | |
647 | I32 srch_end_shift = end_shift; | |
648 | if (srch_start_shift < 0 && strbeg - s > srch_start_shift) { | |
649 | srch_end_shift -= ((strbeg - s) - srch_start_shift); | |
650 | srch_start_shift = strbeg - s; | |
651 | } | |
6bda09f9 | 652 | DEBUG_OPTIMISE_MORE_r({ |
1de06328 YO |
653 | PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n", |
654 | (IV)prog->check_offset_min, | |
655 | (IV)srch_start_shift, | |
656 | (IV)srch_end_shift, | |
657 | (IV)prog->check_end_shift); | |
658 | }); | |
659 | ||
cad2e5aa | 660 | if (flags & REXEC_SCREAM) { |
cad2e5aa | 661 | I32 p = -1; /* Internal iterator of scream. */ |
a3b680e6 | 662 | I32 * const pp = data ? data->scream_pos : &p; |
cad2e5aa | 663 | |
2c2d71f5 JH |
664 | if (PL_screamfirst[BmRARE(check)] >= 0 |
665 | || ( BmRARE(check) == '\n' | |
85c508c3 | 666 | && (BmPREVIOUS(check) == SvCUR(check) - 1) |
2c2d71f5 | 667 | && SvTAIL(check) )) |
9041c2e3 | 668 | s = screaminstr(sv, check, |
1de06328 | 669 | srch_start_shift + (s - strbeg), srch_end_shift, pp, 0); |
cad2e5aa | 670 | else |
2c2d71f5 | 671 | goto fail_finish; |
4addbd3b | 672 | /* we may be pointing at the wrong string */ |
07bc277f | 673 | if (s && RXp_MATCH_COPIED(prog)) |
3f7c398e | 674 | s = strbeg + (s - SvPVX_const(sv)); |
cad2e5aa JH |
675 | if (data) |
676 | *data->scream_olds = s; | |
677 | } | |
1de06328 YO |
678 | else { |
679 | U8* start_point; | |
680 | U8* end_point; | |
bbe252da | 681 | if (prog->extflags & RXf_CANY_SEEN) { |
1de06328 YO |
682 | start_point= (U8*)(s + srch_start_shift); |
683 | end_point= (U8*)(strend - srch_end_shift); | |
684 | } else { | |
685 | start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend); | |
686 | end_point= HOP3(strend, -srch_end_shift, strbeg); | |
687 | } | |
6bda09f9 | 688 | DEBUG_OPTIMISE_MORE_r({ |
56570a2c | 689 | PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n", |
1de06328 | 690 | (int)(end_point - start_point), |
fc8cd66c | 691 | (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point), |
1de06328 YO |
692 | start_point); |
693 | }); | |
694 | ||
695 | s = fbm_instr( start_point, end_point, | |
7fba1cd6 | 696 | check, multiline ? FBMrf_MULTILINE : 0); |
1de06328 YO |
697 | } |
698 | } | |
cad2e5aa JH |
699 | /* Update the count-of-usability, remove useless subpatterns, |
700 | unshift s. */ | |
2c2d71f5 | 701 | |
ab3bbdeb | 702 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 703 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
704 | SvPVX_const(check), RE_SV_DUMPLEN(check), 30); |
705 | PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s", | |
2c2d71f5 | 706 | (s ? "Found" : "Did not find"), |
f2ed9b32 | 707 | (check == (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) |
ab3bbdeb YO |
708 | ? "anchored" : "floating"), |
709 | quoted, | |
710 | RE_SV_TAIL(check), | |
711 | (s ? " at offset " : "...\n") ); | |
712 | }); | |
2c2d71f5 JH |
713 | |
714 | if (!s) | |
715 | goto fail_finish; | |
2c2d71f5 | 716 | /* Finish the diagnostic message */ |
a3621e74 | 717 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); |
2c2d71f5 | 718 | |
1de06328 YO |
719 | /* XXX dmq: first branch is for positive lookbehind... |
720 | Our check string is offset from the beginning of the pattern. | |
721 | So we need to do any stclass tests offset forward from that | |
722 | point. I think. :-( | |
723 | */ | |
724 | ||
725 | ||
726 | ||
727 | check_at=s; | |
728 | ||
729 | ||
2c2d71f5 JH |
730 | /* Got a candidate. Check MBOL anchoring, and the *other* substr. |
731 | Start with the other substr. | |
732 | XXXX no SCREAM optimization yet - and a very coarse implementation | |
a0288114 | 733 | XXXX /ttx+/ results in anchored="ttx", floating="x". floating will |
2c2d71f5 JH |
734 | *always* match. Probably should be marked during compile... |
735 | Probably it is right to do no SCREAM here... | |
736 | */ | |
737 | ||
f2ed9b32 | 738 | if (utf8_target ? (prog->float_utf8 && prog->anchored_utf8) |
1de06328 YO |
739 | : (prog->float_substr && prog->anchored_substr)) |
740 | { | |
30944b6d | 741 | /* Take into account the "other" substring. */ |
2c2d71f5 JH |
742 | /* XXXX May be hopelessly wrong for UTF... */ |
743 | if (!other_last) | |
6eb5f6b9 | 744 | other_last = strpos; |
f2ed9b32 | 745 | if (check == (utf8_target ? prog->float_utf8 : prog->float_substr)) { |
30944b6d IZ |
746 | do_other_anchored: |
747 | { | |
890ce7af AL |
748 | char * const last = HOP3c(s, -start_shift, strbeg); |
749 | char *last1, *last2; | |
be8e71aa | 750 | char * const saved_s = s; |
33b8afdf | 751 | SV* must; |
2c2d71f5 | 752 | |
2c2d71f5 JH |
753 | t = s - prog->check_offset_max; |
754 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ | |
f2ed9b32 | 755 | && (!utf8_target |
0ce71af7 | 756 | || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos)) |
2c2d71f5 | 757 | && t > strpos))) |
6f207bd3 | 758 | NOOP; |
2c2d71f5 JH |
759 | else |
760 | t = strpos; | |
1aa99e6b | 761 | t = HOP3c(t, prog->anchored_offset, strend); |
6eb5f6b9 JH |
762 | if (t < other_last) /* These positions already checked */ |
763 | t = other_last; | |
1aa99e6b | 764 | last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); |
2c2d71f5 JH |
765 | if (last < last1) |
766 | last1 = last; | |
1de06328 YO |
767 | /* XXXX It is not documented what units *_offsets are in. |
768 | We assume bytes, but this is clearly wrong. | |
769 | Meaning this code needs to be carefully reviewed for errors. | |
770 | dmq. | |
771 | */ | |
772 | ||
2c2d71f5 | 773 | /* On end-of-str: see comment below. */ |
f2ed9b32 | 774 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; |
33b8afdf JH |
775 | if (must == &PL_sv_undef) { |
776 | s = (char*)NULL; | |
1de06328 | 777 | DEBUG_r(must = prog->anchored_utf8); /* for debug */ |
33b8afdf JH |
778 | } |
779 | else | |
780 | s = fbm_instr( | |
781 | (unsigned char*)t, | |
782 | HOP3(HOP3(last1, prog->anchored_offset, strend) | |
783 | + SvCUR(must), -(SvTAIL(must)!=0), strbeg), | |
784 | must, | |
7fba1cd6 | 785 | multiline ? FBMrf_MULTILINE : 0 |
33b8afdf | 786 | ); |
ab3bbdeb | 787 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 788 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
789 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
790 | PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s", | |
2c2d71f5 | 791 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
792 | quoted, RE_SV_TAIL(must)); |
793 | }); | |
794 | ||
795 | ||
2c2d71f5 JH |
796 | if (!s) { |
797 | if (last1 >= last2) { | |
a3621e74 | 798 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 JH |
799 | ", giving up...\n")); |
800 | goto fail_finish; | |
801 | } | |
a3621e74 | 802 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
2c2d71f5 | 803 | ", trying floating at offset %ld...\n", |
be8e71aa | 804 | (long)(HOP3c(saved_s, 1, strend) - i_strpos))); |
1aa99e6b IH |
805 | other_last = HOP3c(last1, prog->anchored_offset+1, strend); |
806 | s = HOP3c(last, 1, strend); | |
2c2d71f5 JH |
807 | goto restart; |
808 | } | |
809 | else { | |
a3621e74 | 810 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
30944b6d | 811 | (long)(s - i_strpos))); |
1aa99e6b IH |
812 | t = HOP3c(s, -prog->anchored_offset, strbeg); |
813 | other_last = HOP3c(s, 1, strend); | |
be8e71aa | 814 | s = saved_s; |
2c2d71f5 JH |
815 | if (t == strpos) |
816 | goto try_at_start; | |
2c2d71f5 JH |
817 | goto try_at_offset; |
818 | } | |
30944b6d | 819 | } |
2c2d71f5 JH |
820 | } |
821 | else { /* Take into account the floating substring. */ | |
33b8afdf | 822 | char *last, *last1; |
be8e71aa | 823 | char * const saved_s = s; |
33b8afdf JH |
824 | SV* must; |
825 | ||
826 | t = HOP3c(s, -start_shift, strbeg); | |
827 | last1 = last = | |
828 | HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); | |
829 | if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) | |
830 | last = HOP3c(t, prog->float_max_offset, strend); | |
831 | s = HOP3c(t, prog->float_min_offset, strend); | |
832 | if (s < other_last) | |
833 | s = other_last; | |
2c2d71f5 | 834 | /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ |
f2ed9b32 | 835 | must = utf8_target ? prog->float_utf8 : prog->float_substr; |
33b8afdf JH |
836 | /* fbm_instr() takes into account exact value of end-of-str |
837 | if the check is SvTAIL(ed). Since false positives are OK, | |
838 | and end-of-str is not later than strend we are OK. */ | |
839 | if (must == &PL_sv_undef) { | |
840 | s = (char*)NULL; | |
1de06328 | 841 | DEBUG_r(must = prog->float_utf8); /* for debug message */ |
33b8afdf JH |
842 | } |
843 | else | |
2c2d71f5 | 844 | s = fbm_instr((unsigned char*)s, |
33b8afdf JH |
845 | (unsigned char*)last + SvCUR(must) |
846 | - (SvTAIL(must)!=0), | |
7fba1cd6 | 847 | must, multiline ? FBMrf_MULTILINE : 0); |
ab3bbdeb | 848 | DEBUG_EXECUTE_r({ |
f2ed9b32 | 849 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
850 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
851 | PerlIO_printf(Perl_debug_log, "%s floating substr %s%s", | |
33b8afdf | 852 | (s ? "Found" : "Contradicts"), |
ab3bbdeb YO |
853 | quoted, RE_SV_TAIL(must)); |
854 | }); | |
33b8afdf JH |
855 | if (!s) { |
856 | if (last1 == last) { | |
a3621e74 | 857 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf JH |
858 | ", giving up...\n")); |
859 | goto fail_finish; | |
2c2d71f5 | 860 | } |
a3621e74 | 861 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
33b8afdf | 862 | ", trying anchored starting at offset %ld...\n", |
be8e71aa | 863 | (long)(saved_s + 1 - i_strpos))); |
33b8afdf JH |
864 | other_last = last; |
865 | s = HOP3c(t, 1, strend); | |
866 | goto restart; | |
867 | } | |
868 | else { | |
a3621e74 | 869 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", |
33b8afdf JH |
870 | (long)(s - i_strpos))); |
871 | other_last = s; /* Fix this later. --Hugo */ | |
be8e71aa | 872 | s = saved_s; |
33b8afdf JH |
873 | if (t == strpos) |
874 | goto try_at_start; | |
875 | goto try_at_offset; | |
876 | } | |
2c2d71f5 | 877 | } |
cad2e5aa | 878 | } |
2c2d71f5 | 879 | |
1de06328 | 880 | |
9ef43ace | 881 | t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos); |
1de06328 | 882 | |
6bda09f9 | 883 | DEBUG_OPTIMISE_MORE_r( |
1de06328 YO |
884 | PerlIO_printf(Perl_debug_log, |
885 | "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n", | |
886 | (IV)prog->check_offset_min, | |
887 | (IV)prog->check_offset_max, | |
888 | (IV)(s-strpos), | |
889 | (IV)(t-strpos), | |
890 | (IV)(t-s), | |
891 | (IV)(strend-strpos) | |
892 | ) | |
893 | ); | |
894 | ||
2c2d71f5 | 895 | if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ |
f2ed9b32 | 896 | && (!utf8_target |
9ef43ace | 897 | || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos))) |
1de06328 YO |
898 | && t > strpos))) |
899 | { | |
2c2d71f5 JH |
900 | /* Fixed substring is found far enough so that the match |
901 | cannot start at strpos. */ | |
902 | try_at_offset: | |
cad2e5aa | 903 | if (ml_anch && t[-1] != '\n') { |
30944b6d IZ |
904 | /* Eventually fbm_*() should handle this, but often |
905 | anchored_offset is not 0, so this check will not be wasted. */ | |
906 | /* XXXX In the code below we prefer to look for "^" even in | |
907 | presence of anchored substrings. And we search even | |
908 | beyond the found float position. These pessimizations | |
909 | are historical artefacts only. */ | |
910 | find_anchor: | |
2c2d71f5 | 911 | while (t < strend - prog->minlen) { |
cad2e5aa | 912 | if (*t == '\n') { |
4ee3650e | 913 | if (t < check_at - prog->check_offset_min) { |
f2ed9b32 | 914 | if (utf8_target ? prog->anchored_utf8 : prog->anchored_substr) { |
4ee3650e GS |
915 | /* Since we moved from the found position, |
916 | we definitely contradict the found anchored | |
30944b6d IZ |
917 | substr. Due to the above check we do not |
918 | contradict "check" substr. | |
919 | Thus we can arrive here only if check substr | |
920 | is float. Redo checking for "other"=="fixed". | |
921 | */ | |
9041c2e3 | 922 | strpos = t + 1; |
a3621e74 | 923 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n", |
e4584336 | 924 | PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); |
30944b6d IZ |
925 | goto do_other_anchored; |
926 | } | |
4ee3650e GS |
927 | /* We don't contradict the found floating substring. */ |
928 | /* XXXX Why not check for STCLASS? */ | |
cad2e5aa | 929 | s = t + 1; |
a3621e74 | 930 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", |
e4584336 | 931 | PL_colors[0], PL_colors[1], (long)(s - i_strpos))); |
cad2e5aa JH |
932 | goto set_useful; |
933 | } | |
4ee3650e GS |
934 | /* Position contradicts check-string */ |
935 | /* XXXX probably better to look for check-string | |
936 | than for "\n", so one should lower the limit for t? */ | |
a3621e74 | 937 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", |
e4584336 | 938 | PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos))); |
0e41cd87 | 939 | other_last = strpos = s = t + 1; |
cad2e5aa JH |
940 | goto restart; |
941 | } | |
942 | t++; | |
943 | } | |
a3621e74 | 944 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n", |
e4584336 | 945 | PL_colors[0], PL_colors[1])); |
2c2d71f5 | 946 | goto fail_finish; |
cad2e5aa | 947 | } |
f5952150 | 948 | else { |
a3621e74 | 949 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", |
e4584336 | 950 | PL_colors[0], PL_colors[1])); |
f5952150 | 951 | } |
cad2e5aa JH |
952 | s = t; |
953 | set_useful: | |
f2ed9b32 | 954 | ++BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr); /* hooray/5 */ |
cad2e5aa JH |
955 | } |
956 | else { | |
f5952150 | 957 | /* The found string does not prohibit matching at strpos, |
2c2d71f5 | 958 | - no optimization of calling REx engine can be performed, |
f5952150 GS |
959 | unless it was an MBOL and we are not after MBOL, |
960 | or a future STCLASS check will fail this. */ | |
2c2d71f5 JH |
961 | try_at_start: |
962 | /* Even in this situation we may use MBOL flag if strpos is offset | |
963 | wrt the start of the string. */ | |
05b4157f | 964 | if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ |
a1933d95 | 965 | && (strpos != strbeg) && strpos[-1] != '\n' |
d506a20d | 966 | /* May be due to an implicit anchor of m{.*foo} */ |
bbe252da | 967 | && !(prog->intflags & PREGf_IMPLICIT)) |
d506a20d | 968 | { |
cad2e5aa JH |
969 | t = strpos; |
970 | goto find_anchor; | |
971 | } | |
a3621e74 | 972 | DEBUG_EXECUTE_r( if (ml_anch) |
f5952150 | 973 | PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", |
70685ca0 | 974 | (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]); |
30944b6d | 975 | ); |
2c2d71f5 | 976 | success_at_start: |
bbe252da | 977 | if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */ |
f2ed9b32 | 978 | && (utf8_target ? ( |
33b8afdf JH |
979 | prog->check_utf8 /* Could be deleted already */ |
980 | && --BmUSEFUL(prog->check_utf8) < 0 | |
981 | && (prog->check_utf8 == prog->float_utf8) | |
982 | ) : ( | |
983 | prog->check_substr /* Could be deleted already */ | |
984 | && --BmUSEFUL(prog->check_substr) < 0 | |
985 | && (prog->check_substr == prog->float_substr) | |
986 | ))) | |
66e933ab | 987 | { |
cad2e5aa | 988 | /* If flags & SOMETHING - do not do it many times on the same match */ |
a3621e74 | 989 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); |
f2ed9b32 KW |
990 | /* XXX Does the destruction order has to change with utf8_target? */ |
991 | SvREFCNT_dec(utf8_target ? prog->check_utf8 : prog->check_substr); | |
992 | SvREFCNT_dec(utf8_target ? prog->check_substr : prog->check_utf8); | |
a0714e2c SS |
993 | prog->check_substr = prog->check_utf8 = NULL; /* disable */ |
994 | prog->float_substr = prog->float_utf8 = NULL; /* clear */ | |
995 | check = NULL; /* abort */ | |
cad2e5aa | 996 | s = strpos; |
c9415951 YO |
997 | /* XXXX If the check string was an implicit check MBOL, then we need to unset the relevent flag |
998 | see http://bugs.activestate.com/show_bug.cgi?id=87173 */ | |
999 | if (prog->intflags & PREGf_IMPLICIT) | |
1000 | prog->extflags &= ~RXf_ANCH_MBOL; | |
3cf5c195 IZ |
1001 | /* XXXX This is a remnant of the old implementation. It |
1002 | looks wasteful, since now INTUIT can use many | |
6eb5f6b9 | 1003 | other heuristics. */ |
bbe252da | 1004 | prog->extflags &= ~RXf_USE_INTUIT; |
c9415951 | 1005 | /* XXXX What other flags might need to be cleared in this branch? */ |
cad2e5aa JH |
1006 | } |
1007 | else | |
1008 | s = strpos; | |
1009 | } | |
1010 | ||
6eb5f6b9 JH |
1011 | /* Last resort... */ |
1012 | /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ | |
1de06328 YO |
1013 | /* trie stclasses are too expensive to use here, we are better off to |
1014 | leave it to regmatch itself */ | |
f8fc2ecf | 1015 | if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) { |
6eb5f6b9 JH |
1016 | /* minlen == 0 is possible if regstclass is \b or \B, |
1017 | and the fixed substr is ''$. | |
1018 | Since minlen is already taken into account, s+1 is before strend; | |
1019 | accidentally, minlen >= 1 guaranties no false positives at s + 1 | |
1020 | even for \b or \B. But (minlen? 1 : 0) below assumes that | |
1021 | regstclass does not come from lookahead... */ | |
1022 | /* If regstclass takes bytelength more than 1: If charlength==1, OK. | |
1023 | This leaves EXACTF only, which is dealt with in find_byclass(). */ | |
f8fc2ecf YO |
1024 | const U8* const str = (U8*)STRING(progi->regstclass); |
1025 | const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT | |
1026 | ? CHR_DIST(str+STR_LEN(progi->regstclass), str) | |
66e933ab | 1027 | : 1); |
1de06328 YO |
1028 | char * endpos; |
1029 | if (prog->anchored_substr || prog->anchored_utf8 || ml_anch) | |
1030 | endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend); | |
1031 | else if (prog->float_substr || prog->float_utf8) | |
1032 | endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend); | |
1033 | else | |
1034 | endpos= strend; | |
1035 | ||
70685ca0 JH |
1036 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n", |
1037 | (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg))); | |
1de06328 | 1038 | |
6eb5f6b9 | 1039 | t = s; |
f8fc2ecf | 1040 | s = find_byclass(prog, progi->regstclass, s, endpos, NULL); |
6eb5f6b9 JH |
1041 | if (!s) { |
1042 | #ifdef DEBUGGING | |
cbbf8932 | 1043 | const char *what = NULL; |
6eb5f6b9 JH |
1044 | #endif |
1045 | if (endpos == strend) { | |
a3621e74 | 1046 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
6eb5f6b9 JH |
1047 | "Could not match STCLASS...\n") ); |
1048 | goto fail; | |
1049 | } | |
a3621e74 | 1050 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab | 1051 | "This position contradicts STCLASS...\n") ); |
bbe252da | 1052 | if ((prog->extflags & RXf_ANCH) && !ml_anch) |
653099ff | 1053 | goto fail; |
6eb5f6b9 | 1054 | /* Contradict one of substrings */ |
33b8afdf | 1055 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 | 1056 | if ((utf8_target ? prog->anchored_utf8 : prog->anchored_substr) == check) { |
a3621e74 | 1057 | DEBUG_EXECUTE_r( what = "anchored" ); |
6eb5f6b9 | 1058 | hop_and_restart: |
1aa99e6b | 1059 | s = HOP3c(t, 1, strend); |
66e933ab GS |
1060 | if (s + start_shift + end_shift > strend) { |
1061 | /* XXXX Should be taken into account earlier? */ | |
a3621e74 | 1062 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
66e933ab GS |
1063 | "Could not match STCLASS...\n") ); |
1064 | goto fail; | |
1065 | } | |
5e39e1e5 HS |
1066 | if (!check) |
1067 | goto giveup; | |
a3621e74 | 1068 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1069 | "Looking for %s substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1070 | what, (long)(s + start_shift - i_strpos)) ); |
1071 | goto restart; | |
1072 | } | |
66e933ab | 1073 | /* Have both, check_string is floating */ |
6eb5f6b9 JH |
1074 | if (t + start_shift >= check_at) /* Contradicts floating=check */ |
1075 | goto retry_floating_check; | |
1076 | /* Recheck anchored substring, but not floating... */ | |
9041c2e3 | 1077 | s = check_at; |
5e39e1e5 HS |
1078 | if (!check) |
1079 | goto giveup; | |
a3621e74 | 1080 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1081 | "Looking for anchored substr starting at offset %ld...\n", |
6eb5f6b9 JH |
1082 | (long)(other_last - i_strpos)) ); |
1083 | goto do_other_anchored; | |
1084 | } | |
60e71179 GS |
1085 | /* Another way we could have checked stclass at the |
1086 | current position only: */ | |
1087 | if (ml_anch) { | |
1088 | s = t = t + 1; | |
5e39e1e5 HS |
1089 | if (!check) |
1090 | goto giveup; | |
a3621e74 | 1091 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
f5952150 | 1092 | "Looking for /%s^%s/m starting at offset %ld...\n", |
e4584336 | 1093 | PL_colors[0], PL_colors[1], (long)(t - i_strpos)) ); |
60e71179 | 1094 | goto try_at_offset; |
66e933ab | 1095 | } |
f2ed9b32 | 1096 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */ |
60e71179 | 1097 | goto fail; |
6eb5f6b9 JH |
1098 | /* Check is floating subtring. */ |
1099 | retry_floating_check: | |
1100 | t = check_at - start_shift; | |
a3621e74 | 1101 | DEBUG_EXECUTE_r( what = "floating" ); |
6eb5f6b9 JH |
1102 | goto hop_and_restart; |
1103 | } | |
b7953727 | 1104 | if (t != s) { |
a3621e74 | 1105 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
6eb5f6b9 | 1106 | "By STCLASS: moving %ld --> %ld\n", |
b7953727 JH |
1107 | (long)(t - i_strpos), (long)(s - i_strpos)) |
1108 | ); | |
1109 | } | |
1110 | else { | |
a3621e74 | 1111 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
b7953727 JH |
1112 | "Does not contradict STCLASS...\n"); |
1113 | ); | |
1114 | } | |
6eb5f6b9 | 1115 | } |
5e39e1e5 | 1116 | giveup: |
a3621e74 | 1117 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", |
5e39e1e5 HS |
1118 | PL_colors[4], (check ? "Guessed" : "Giving up"), |
1119 | PL_colors[5], (long)(s - i_strpos)) ); | |
cad2e5aa | 1120 | return s; |
2c2d71f5 JH |
1121 | |
1122 | fail_finish: /* Substring not found */ | |
33b8afdf | 1123 | if (prog->check_substr || prog->check_utf8) /* could be removed already */ |
f2ed9b32 | 1124 | BmUSEFUL(utf8_target ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */ |
cad2e5aa | 1125 | fail: |
a3621e74 | 1126 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", |
e4584336 | 1127 | PL_colors[4], PL_colors[5])); |
bd61b366 | 1128 | return NULL; |
cad2e5aa | 1129 | } |
9661b544 | 1130 | |
a0a388a1 YO |
1131 | #define DECL_TRIE_TYPE(scan) \ |
1132 | const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \ | |
1133 | trie_type = (scan->flags != EXACT) \ | |
f2ed9b32 KW |
1134 | ? (utf8_target ? trie_utf8_fold : (UTF_PATTERN ? trie_latin_utf8_fold : trie_plain)) \ |
1135 | : (utf8_target ? trie_utf8 : trie_plain) | |
3b0527fe | 1136 | |
55eed653 NC |
1137 | #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ |
1138 | uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ | |
4cadc6a9 YO |
1139 | switch (trie_type) { \ |
1140 | case trie_utf8_fold: \ | |
1141 | if ( foldlen>0 ) { \ | |
0abd0d78 | 1142 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ |
4cadc6a9 YO |
1143 | foldlen -= len; \ |
1144 | uscan += len; \ | |
1145 | len=0; \ | |
1146 | } else { \ | |
0abd0d78 | 1147 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ |
4cadc6a9 YO |
1148 | uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \ |
1149 | foldlen -= UNISKIP( uvc ); \ | |
1150 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1151 | } \ | |
1152 | break; \ | |
a0a388a1 YO |
1153 | case trie_latin_utf8_fold: \ |
1154 | if ( foldlen>0 ) { \ | |
1155 | uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ | |
1156 | foldlen -= len; \ | |
1157 | uscan += len; \ | |
1158 | len=0; \ | |
1159 | } else { \ | |
1160 | len = 1; \ | |
1161 | uvc = to_uni_fold( *(U8*)uc, foldbuf, &foldlen ); \ | |
1162 | foldlen -= UNISKIP( uvc ); \ | |
1163 | uscan = foldbuf + UNISKIP( uvc ); \ | |
1164 | } \ | |
1165 | break; \ | |
4cadc6a9 YO |
1166 | case trie_utf8: \ |
1167 | uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ | |
1168 | break; \ | |
1169 | case trie_plain: \ | |
1170 | uvc = (UV)*uc; \ | |
1171 | len = 1; \ | |
1172 | } \ | |
4cadc6a9 YO |
1173 | if (uvc < 256) { \ |
1174 | charid = trie->charmap[ uvc ]; \ | |
1175 | } \ | |
1176 | else { \ | |
1177 | charid = 0; \ | |
55eed653 NC |
1178 | if (widecharmap) { \ |
1179 | SV** const svpp = hv_fetch(widecharmap, \ | |
4cadc6a9 YO |
1180 | (char*)&uvc, sizeof(UV), 0); \ |
1181 | if (svpp) \ | |
1182 | charid = (U16)SvIV(*svpp); \ | |
1183 | } \ | |
1184 | } \ | |
1185 | } STMT_END | |
1186 | ||
a0a388a1 YO |
1187 | #define REXEC_FBC_EXACTISH_CHECK(CoNd) \ |
1188 | { \ | |
1189 | char *my_strend= (char *)strend; \ | |
4cadc6a9 YO |
1190 | if ( (CoNd) \ |
1191 | && (ln == len || \ | |
f2ed9b32 KW |
1192 | foldEQ_utf8(s, &my_strend, 0, utf8_target, \ |
1193 | m, NULL, ln, cBOOL(UTF_PATTERN))) \ | |
a0a388a1 | 1194 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1195 | goto got_it; \ |
1196 | else { \ | |
1197 | U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \ | |
1198 | uvchr_to_utf8(tmpbuf, c); \ | |
1199 | f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \ | |
1200 | if ( f != c \ | |
1201 | && (f == c1 || f == c2) \ | |
a0a388a1 | 1202 | && (ln == len || \ |
f2ed9b32 KW |
1203 | foldEQ_utf8(s, &my_strend, 0, utf8_target,\ |
1204 | m, NULL, ln, cBOOL(UTF_PATTERN)))\ | |
a0a388a1 | 1205 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1206 | goto got_it; \ |
1207 | } \ | |
a0a388a1 YO |
1208 | } \ |
1209 | s += len | |
4cadc6a9 YO |
1210 | |
1211 | #define REXEC_FBC_EXACTISH_SCAN(CoNd) \ | |
1212 | STMT_START { \ | |
1213 | while (s <= e) { \ | |
1214 | if ( (CoNd) \ | |
4c1b470c KW |
1215 | && (ln == 1 || (OP(c) == EXACTF \ |
1216 | ? foldEQ(s, m, ln) \ | |
1217 | : foldEQ_locale(s, m, ln))) \ | |
24b23f37 | 1218 | && (!reginfo || regtry(reginfo, &s)) ) \ |
4cadc6a9 YO |
1219 | goto got_it; \ |
1220 | s++; \ | |
1221 | } \ | |
1222 | } STMT_END | |
1223 | ||
1224 | #define REXEC_FBC_UTF8_SCAN(CoDe) \ | |
1225 | STMT_START { \ | |
1226 | while (s + (uskip = UTF8SKIP(s)) <= strend) { \ | |
1227 | CoDe \ | |
1228 | s += uskip; \ | |
1229 | } \ | |
1230 | } STMT_END | |
1231 | ||
1232 | #define REXEC_FBC_SCAN(CoDe) \ | |
1233 | STMT_START { \ | |
1234 | while (s < strend) { \ | |
1235 | CoDe \ | |
1236 | s++; \ | |
1237 | } \ | |
1238 | } STMT_END | |
1239 | ||
1240 | #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \ | |
1241 | REXEC_FBC_UTF8_SCAN( \ | |
1242 | if (CoNd) { \ | |
24b23f37 | 1243 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1244 | goto got_it; \ |
1245 | else \ | |
1246 | tmp = doevery; \ | |
1247 | } \ | |
1248 | else \ | |
1249 | tmp = 1; \ | |
1250 | ) | |
1251 | ||
1252 | #define REXEC_FBC_CLASS_SCAN(CoNd) \ | |
1253 | REXEC_FBC_SCAN( \ | |
1254 | if (CoNd) { \ | |
24b23f37 | 1255 | if (tmp && (!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1256 | goto got_it; \ |
1257 | else \ | |
1258 | tmp = doevery; \ | |
1259 | } \ | |
1260 | else \ | |
1261 | tmp = 1; \ | |
1262 | ) | |
1263 | ||
1264 | #define REXEC_FBC_TRYIT \ | |
24b23f37 | 1265 | if ((!reginfo || regtry(reginfo, &s))) \ |
4cadc6a9 YO |
1266 | goto got_it |
1267 | ||
e1d1eefb | 1268 | #define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \ |
f2ed9b32 | 1269 | if (utf8_target) { \ |
e1d1eefb YO |
1270 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1271 | } \ | |
1272 | else { \ | |
1273 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
1274 | } \ | |
1275 | break | |
1276 | ||
4cadc6a9 | 1277 | #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \ |
f2ed9b32 | 1278 | if (utf8_target) { \ |
4cadc6a9 YO |
1279 | UtFpReLoAd; \ |
1280 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ | |
1281 | } \ | |
1282 | else { \ | |
1283 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
1284 | } \ | |
1285 | break | |
1286 | ||
1287 | #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \ | |
1288 | PL_reg_flags |= RF_tainted; \ | |
f2ed9b32 | 1289 | if (utf8_target) { \ |
4cadc6a9 YO |
1290 | REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \ |
1291 | } \ | |
1292 | else { \ | |
1293 | REXEC_FBC_CLASS_SCAN(CoNd); \ | |
1294 | } \ | |
1295 | break | |
1296 | ||
786e8c11 YO |
1297 | #define DUMP_EXEC_POS(li,s,doutf8) \ |
1298 | dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8) | |
1299 | ||
1300 | /* We know what class REx starts with. Try to find this position... */ | |
1301 | /* if reginfo is NULL, its a dryrun */ | |
1302 | /* annoyingly all the vars in this routine have different names from their counterparts | |
1303 | in regmatch. /grrr */ | |
1304 | ||
3c3eec57 | 1305 | STATIC char * |
07be1b83 | 1306 | S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, |
24b23f37 | 1307 | const char *strend, regmatch_info *reginfo) |
a687059c | 1308 | { |
27da23d5 | 1309 | dVAR; |
bbe252da | 1310 | const I32 doevery = (prog->intflags & PREGf_SKIP) == 0; |
6eb5f6b9 | 1311 | char *m; |
d8093b23 | 1312 | STRLEN ln; |
5dab1207 | 1313 | STRLEN lnc; |
078c425b | 1314 | register STRLEN uskip; |
d8093b23 G |
1315 | unsigned int c1; |
1316 | unsigned int c2; | |
6eb5f6b9 JH |
1317 | char *e; |
1318 | register I32 tmp = 1; /* Scratch variable? */ | |
f2ed9b32 | 1319 | register const bool utf8_target = PL_reg_match_utf8; |
f8fc2ecf | 1320 | RXi_GET_DECL(prog,progi); |
7918f24d NC |
1321 | |
1322 | PERL_ARGS_ASSERT_FIND_BYCLASS; | |
f8fc2ecf | 1323 | |
6eb5f6b9 JH |
1324 | /* We know what class it must start with. */ |
1325 | switch (OP(c)) { | |
6eb5f6b9 | 1326 | case ANYOF: |
f2ed9b32 | 1327 | if (utf8_target) { |
4cadc6a9 | 1328 | REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_UNICODE) || |
388cc4de | 1329 | !UTF8_IS_INVARIANT((U8)s[0]) ? |
f2ed9b32 | 1330 | reginclass(prog, c, (U8*)s, 0, utf8_target) : |
4cadc6a9 | 1331 | REGINCLASS(prog, c, (U8*)s)); |
388cc4de HS |
1332 | } |
1333 | else { | |
1334 | while (s < strend) { | |
1335 | STRLEN skip = 1; | |
1336 | ||
32fc9b6a | 1337 | if (REGINCLASS(prog, c, (U8*)s) || |
388cc4de HS |
1338 | (ANYOF_FOLD_SHARP_S(c, s, strend) && |
1339 | /* The assignment of 2 is intentional: | |
1340 | * for the folded sharp s, the skip is 2. */ | |
1341 | (skip = SHARP_S_SKIP))) { | |
24b23f37 | 1342 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
388cc4de HS |
1343 | goto got_it; |
1344 | else | |
1345 | tmp = doevery; | |
1346 | } | |
1347 | else | |
1348 | tmp = 1; | |
1349 | s += skip; | |
1350 | } | |
a0d0e21e | 1351 | } |
6eb5f6b9 | 1352 | break; |
f33976b4 | 1353 | case CANY: |
4cadc6a9 | 1354 | REXEC_FBC_SCAN( |
24b23f37 | 1355 | if (tmp && (!reginfo || regtry(reginfo, &s))) |
f33976b4 DB |
1356 | goto got_it; |
1357 | else | |
1358 | tmp = doevery; | |
4cadc6a9 | 1359 | ); |
f33976b4 | 1360 | break; |
6eb5f6b9 | 1361 | case EXACTF: |
5dab1207 NIS |
1362 | m = STRING(c); |
1363 | ln = STR_LEN(c); /* length to match in octets/bytes */ | |
1364 | lnc = (I32) ln; /* length to match in characters */ | |
f2ed9b32 | 1365 | if (UTF_PATTERN) { |
a2a2844f | 1366 | STRLEN ulen1, ulen2; |
5dab1207 | 1367 | U8 *sm = (U8 *) m; |
89ebb4a3 JH |
1368 | U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; |
1369 | U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; | |
97dc7d3e RGS |
1370 | /* used by commented-out code below */ |
1371 | /*const U32 uniflags = UTF8_ALLOW_DEFAULT;*/ | |
a0a388a1 YO |
1372 | |
1373 | /* XXX: Since the node will be case folded at compile | |
1374 | time this logic is a little odd, although im not | |
1375 | sure that its actually wrong. --dmq */ | |
1376 | ||
1377 | c1 = to_utf8_lower((U8*)m, tmpbuf1, &ulen1); | |
1378 | c2 = to_utf8_upper((U8*)m, tmpbuf2, &ulen2); | |
1379 | ||
1380 | /* XXX: This is kinda strange. to_utf8_XYZ returns the | |
1381 | codepoint of the first character in the converted | |
1382 | form, yet originally we did the extra step. | |
1383 | No tests fail by commenting this code out however | |
1384 | so Ive left it out. -- dmq. | |
1385 | ||
89ebb4a3 | 1386 | c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE, |
041457d9 | 1387 | 0, uniflags); |
89ebb4a3 | 1388 | c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE, |
041457d9 | 1389 | 0, uniflags); |
a0a388a1 YO |
1390 | */ |
1391 | ||
5dab1207 NIS |
1392 | lnc = 0; |
1393 | while (sm < ((U8 *) m + ln)) { | |
1394 | lnc++; | |
1395 | sm += UTF8SKIP(sm); | |
1396 | } | |
1aa99e6b IH |
1397 | } |
1398 | else { | |
1399 | c1 = *(U8*)m; | |
1400 | c2 = PL_fold[c1]; | |
1401 | } | |
6eb5f6b9 JH |
1402 | goto do_exactf; |
1403 | case EXACTFL: | |
5dab1207 NIS |
1404 | m = STRING(c); |
1405 | ln = STR_LEN(c); | |
1406 | lnc = (I32) ln; | |
d8093b23 | 1407 | c1 = *(U8*)m; |
6eb5f6b9 JH |
1408 | c2 = PL_fold_locale[c1]; |
1409 | do_exactf: | |
db12adc6 | 1410 | e = HOP3c(strend, -((I32)lnc), s); |
b3c9acc1 | 1411 | |
3b0527fe | 1412 | if (!reginfo && e < s) |
6eb5f6b9 | 1413 | e = s; /* Due to minlen logic of intuit() */ |
1aa99e6b | 1414 | |
60a8b682 JH |
1415 | /* The idea in the EXACTF* cases is to first find the |
1416 | * first character of the EXACTF* node and then, if | |
1417 | * necessary, case-insensitively compare the full | |
1418 | * text of the node. The c1 and c2 are the first | |
1419 | * characters (though in Unicode it gets a bit | |
1420 | * more complicated because there are more cases | |
7f16dd3d JH |
1421 | * than just upper and lower: one needs to use |
1422 | * the so-called folding case for case-insensitive | |
1423 | * matching (called "loose matching" in Unicode). | |
4c1b470c | 1424 | * foldEQ_utf8() will do just that. */ |
60a8b682 | 1425 | |
f2ed9b32 | 1426 | if (utf8_target || UTF_PATTERN) { |
575cac57 | 1427 | UV c, f; |
89ebb4a3 | 1428 | U8 tmpbuf [UTF8_MAXBYTES+1]; |
a0a388a1 YO |
1429 | STRLEN len = 1; |
1430 | STRLEN foldlen; | |
4ad0818d | 1431 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
09091399 | 1432 | if (c1 == c2) { |
5dab1207 NIS |
1433 | /* Upper and lower of 1st char are equal - |
1434 | * probably not a "letter". */ | |
1aa99e6b | 1435 | while (s <= e) { |
f2ed9b32 | 1436 | if (utf8_target) { |
a0a388a1 | 1437 | c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len, |
041457d9 | 1438 | uniflags); |
a0a388a1 YO |
1439 | } else { |
1440 | c = *((U8*)s); | |
1441 | } | |
4cadc6a9 | 1442 | REXEC_FBC_EXACTISH_CHECK(c == c1); |
1aa99e6b | 1443 | } |
09091399 JH |
1444 | } |
1445 | else { | |
1aa99e6b | 1446 | while (s <= e) { |
f2ed9b32 | 1447 | if (utf8_target) { |
a0a388a1 | 1448 | c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len, |
041457d9 | 1449 | uniflags); |
a0a388a1 YO |
1450 | } else { |
1451 | c = *((U8*)s); | |
1452 | } | |
80aecb99 | 1453 | |
60a8b682 | 1454 | /* Handle some of the three Greek sigmas cases. |
8c01da3c JH |
1455 | * Note that not all the possible combinations |
1456 | * are handled here: some of them are handled | |
1457 | * by the standard folding rules, and some of | |
1458 | * them (the character class or ANYOF cases) | |
1459 | * are handled during compiletime in | |
1460 | * regexec.c:S_regclass(). */ | |
880bd946 JH |
1461 | if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA || |
1462 | c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA) | |
1463 | c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA; | |
80aecb99 | 1464 | |
4cadc6a9 | 1465 | REXEC_FBC_EXACTISH_CHECK(c == c1 || c == c2); |
1aa99e6b | 1466 | } |
09091399 | 1467 | } |
1aa99e6b IH |
1468 | } |
1469 | else { | |
a0a388a1 | 1470 | /* Neither pattern nor string are UTF8 */ |
1aa99e6b | 1471 | if (c1 == c2) |
4cadc6a9 | 1472 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1); |
1aa99e6b | 1473 | else |
4cadc6a9 | 1474 | REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2); |
b3c9acc1 IZ |
1475 | } |
1476 | break; | |
bbce6d69 | 1477 | case BOUNDL: |
3280af22 | 1478 | PL_reg_flags |= RF_tainted; |
bbce6d69 | 1479 | /* FALL THROUGH */ |
a0d0e21e | 1480 | case BOUND: |
f2ed9b32 | 1481 | if (utf8_target) { |
12d33761 | 1482 | if (s == PL_bostr) |
ffc61ed2 JH |
1483 | tmp = '\n'; |
1484 | else { | |
6136c704 | 1485 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); |
4ad0818d | 1486 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); |
ffc61ed2 JH |
1487 | } |
1488 | tmp = ((OP(c) == BOUND ? | |
9041c2e3 | 1489 | isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0); |
1a4fad37 | 1490 | LOAD_UTF8_CHARCLASS_ALNUM(); |
4cadc6a9 | 1491 | REXEC_FBC_UTF8_SCAN( |
ffc61ed2 | 1492 | if (tmp == !(OP(c) == BOUND ? |
f2ed9b32 | 1493 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)) : |
ffc61ed2 JH |
1494 | isALNUM_LC_utf8((U8*)s))) |
1495 | { | |
1496 | tmp = !tmp; | |
4cadc6a9 | 1497 | REXEC_FBC_TRYIT; |
a687059c | 1498 | } |
4cadc6a9 | 1499 | ); |
a0d0e21e | 1500 | } |
667bb95a | 1501 | else { |
12d33761 | 1502 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; |
ffc61ed2 | 1503 | tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0); |
4cadc6a9 | 1504 | REXEC_FBC_SCAN( |
ffc61ed2 JH |
1505 | if (tmp == |
1506 | !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) { | |
1507 | tmp = !tmp; | |
4cadc6a9 | 1508 | REXEC_FBC_TRYIT; |
a0ed51b3 | 1509 | } |
4cadc6a9 | 1510 | ); |
a0ed51b3 | 1511 | } |
24b23f37 | 1512 | if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s))) |
a0ed51b3 LW |
1513 | goto got_it; |
1514 | break; | |
bbce6d69 | 1515 | case NBOUNDL: |
3280af22 | 1516 | PL_reg_flags |= RF_tainted; |
bbce6d69 | 1517 | /* FALL THROUGH */ |
a0d0e21e | 1518 | case NBOUND: |
f2ed9b32 | 1519 | if (utf8_target) { |
12d33761 | 1520 | if (s == PL_bostr) |
ffc61ed2 JH |
1521 | tmp = '\n'; |
1522 | else { | |
6136c704 | 1523 | U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); |
4ad0818d | 1524 | tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); |
ffc61ed2 JH |
1525 | } |
1526 | tmp = ((OP(c) == NBOUND ? | |
9041c2e3 | 1527 | isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0); |
1a4fad37 | 1528 | LOAD_UTF8_CHARCLASS_ALNUM(); |
4cadc6a9 | 1529 | REXEC_FBC_UTF8_SCAN( |
ffc61ed2 | 1530 | if (tmp == !(OP(c) == NBOUND ? |
f2ed9b32 | 1531 | cBOOL(swash_fetch(PL_utf8_alnum, (U8*)s, utf8_target)) : |
ffc61ed2 JH |
1532 | isALNUM_LC_utf8((U8*)s))) |
1533 | tmp = !tmp; | |
4cadc6a9 YO |
1534 | else REXEC_FBC_TRYIT; |
1535 | ); | |
a0d0e21e | 1536 | } |
667bb95a | 1537 | else { |
12d33761 | 1538 | tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n'; |
ffc61ed2 JH |
1539 | tmp = ((OP(c) == NBOUND ? |
1540 | isALNUM(tmp) : isALNUM_LC(tmp)) != 0); | |
4cadc6a9 | 1541 | REXEC_FBC_SCAN( |
ffc61ed2 JH |
1542 | if (tmp == |
1543 | !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s))) | |
1544 | tmp = !tmp; | |
4cadc6a9 YO |
1545 | else REXEC_FBC_TRYIT; |
1546 | ); | |
a0ed51b3 | 1547 | } |
24b23f37 | 1548 | if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, &s))) |
a0ed51b3 LW |
1549 | goto got_it; |
1550 | break; | |
a0d0e21e | 1551 | case ALNUM: |
4cadc6a9 | 1552 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1553 | LOAD_UTF8_CHARCLASS_PERL_WORD(), |
f2ed9b32 | 1554 | swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target), |
4cadc6a9 YO |
1555 | isALNUM(*s) |
1556 | ); | |
bbce6d69 | 1557 | case ALNUML: |
4cadc6a9 YO |
1558 | REXEC_FBC_CSCAN_TAINT( |
1559 | isALNUM_LC_utf8((U8*)s), | |
1560 | isALNUM_LC(*s) | |
1561 | ); | |
a0d0e21e | 1562 | case NALNUM: |
4cadc6a9 | 1563 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1564 | LOAD_UTF8_CHARCLASS_PERL_WORD(), |
f2ed9b32 | 1565 | !swash_fetch(RE_utf8_perl_word, (U8*)s, utf8_target), |
4cadc6a9 YO |
1566 | !isALNUM(*s) |
1567 | ); | |
bbce6d69 | 1568 | case NALNUML: |
4cadc6a9 YO |
1569 | REXEC_FBC_CSCAN_TAINT( |
1570 | !isALNUM_LC_utf8((U8*)s), | |
1571 | !isALNUM_LC(*s) | |
1572 | ); | |
a0d0e21e | 1573 | case SPACE: |
4cadc6a9 | 1574 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1575 | LOAD_UTF8_CHARCLASS_PERL_SPACE(), |
f2ed9b32 | 1576 | *s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target), |
4cadc6a9 YO |
1577 | isSPACE(*s) |
1578 | ); | |
bbce6d69 | 1579 | case SPACEL: |
4cadc6a9 YO |
1580 | REXEC_FBC_CSCAN_TAINT( |
1581 | *s == ' ' || isSPACE_LC_utf8((U8*)s), | |
1582 | isSPACE_LC(*s) | |
1583 | ); | |
a0d0e21e | 1584 | case NSPACE: |
4cadc6a9 | 1585 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1586 | LOAD_UTF8_CHARCLASS_PERL_SPACE(), |
f2ed9b32 | 1587 | !(*s == ' ' || swash_fetch(RE_utf8_perl_space,(U8*)s, utf8_target)), |
4cadc6a9 YO |
1588 | !isSPACE(*s) |
1589 | ); | |
bbce6d69 | 1590 | case NSPACEL: |
4cadc6a9 YO |
1591 | REXEC_FBC_CSCAN_TAINT( |
1592 | !(*s == ' ' || isSPACE_LC_utf8((U8*)s)), | |
1593 | !isSPACE_LC(*s) | |
1594 | ); | |
a0d0e21e | 1595 | case DIGIT: |
4cadc6a9 | 1596 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1597 | LOAD_UTF8_CHARCLASS_POSIX_DIGIT(), |
f2ed9b32 | 1598 | swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target), |
4cadc6a9 YO |
1599 | isDIGIT(*s) |
1600 | ); | |
b8c5462f | 1601 | case DIGITL: |
4cadc6a9 YO |
1602 | REXEC_FBC_CSCAN_TAINT( |
1603 | isDIGIT_LC_utf8((U8*)s), | |
1604 | isDIGIT_LC(*s) | |
1605 | ); | |
a0d0e21e | 1606 | case NDIGIT: |
4cadc6a9 | 1607 | REXEC_FBC_CSCAN_PRELOAD( |
d1eb3177 | 1608 | LOAD_UTF8_CHARCLASS_POSIX_DIGIT(), |
f2ed9b32 | 1609 | !swash_fetch(RE_utf8_posix_digit,(U8*)s, utf8_target), |
4cadc6a9 YO |
1610 | !isDIGIT(*s) |
1611 | ); | |
b8c5462f | 1612 | case NDIGITL: |
4cadc6a9 YO |
1613 | REXEC_FBC_CSCAN_TAINT( |
1614 | !isDIGIT_LC_utf8((U8*)s), | |
1615 | !isDIGIT_LC(*s) | |
1616 | ); | |
e1d1eefb YO |
1617 | case LNBREAK: |
1618 | REXEC_FBC_CSCAN( | |
1619 | is_LNBREAK_utf8(s), | |
1620 | is_LNBREAK_latin1(s) | |
1621 | ); | |
1622 | case VERTWS: | |
1623 | REXEC_FBC_CSCAN( | |
1624 | is_VERTWS_utf8(s), | |
1625 | is_VERTWS_latin1(s) | |
1626 | ); | |
1627 | case NVERTWS: | |
1628 | REXEC_FBC_CSCAN( | |
1629 | !is_VERTWS_utf8(s), | |
1630 | !is_VERTWS_latin1(s) | |
1631 | ); | |
1632 | case HORIZWS: | |
1633 | REXEC_FBC_CSCAN( | |
1634 | is_HORIZWS_utf8(s), | |
1635 | is_HORIZWS_latin1(s) | |
1636 | ); | |
1637 | case NHORIZWS: | |
1638 | REXEC_FBC_CSCAN( | |
1639 | !is_HORIZWS_utf8(s), | |
1640 | !is_HORIZWS_latin1(s) | |
1641 | ); | |
1de06328 YO |
1642 | case AHOCORASICKC: |
1643 | case AHOCORASICK: | |
07be1b83 | 1644 | { |
a0a388a1 | 1645 | DECL_TRIE_TYPE(c); |
07be1b83 YO |
1646 | /* what trie are we using right now */ |
1647 | reg_ac_data *aho | |
f8fc2ecf | 1648 | = (reg_ac_data*)progi->data->data[ ARG( c ) ]; |
3251b653 NC |
1649 | reg_trie_data *trie |
1650 | = (reg_trie_data*)progi->data->data[ aho->trie ]; | |
85fbaab2 | 1651 | HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); |
07be1b83 YO |
1652 | |
1653 | const char *last_start = strend - trie->minlen; | |
6148ee25 | 1654 | #ifdef DEBUGGING |
07be1b83 | 1655 | const char *real_start = s; |
6148ee25 | 1656 | #endif |
07be1b83 | 1657 | STRLEN maxlen = trie->maxlen; |
be8e71aa YO |
1658 | SV *sv_points; |
1659 | U8 **points; /* map of where we were in the input string | |
786e8c11 | 1660 | when reading a given char. For ASCII this |
be8e71aa | 1661 | is unnecessary overhead as the relationship |
38a44b82 NC |
1662 | is always 1:1, but for Unicode, especially |
1663 | case folded Unicode this is not true. */ | |
f9e705e8 | 1664 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
786e8c11 YO |
1665 | U8 *bitmap=NULL; |
1666 | ||
07be1b83 YO |
1667 | |
1668 | GET_RE_DEBUG_FLAGS_DECL; | |
1669 | ||
be8e71aa YO |
1670 | /* We can't just allocate points here. We need to wrap it in |
1671 | * an SV so it gets freed properly if there is a croak while | |
1672 | * running the match */ | |
1673 | ENTER; | |
1674 | SAVETMPS; | |
1675 | sv_points=newSV(maxlen * sizeof(U8 *)); | |
1676 | SvCUR_set(sv_points, | |
1677 | maxlen * sizeof(U8 *)); | |
1678 | SvPOK_on(sv_points); | |
1679 | sv_2mortal(sv_points); | |
1680 | points=(U8**)SvPV_nolen(sv_points ); | |
1de06328 YO |
1681 | if ( trie_type != trie_utf8_fold |
1682 | && (trie->bitmap || OP(c)==AHOCORASICKC) ) | |
1683 | { | |
786e8c11 YO |
1684 | if (trie->bitmap) |
1685 | bitmap=(U8*)trie->bitmap; | |
1686 | else | |
1687 | bitmap=(U8*)ANYOF_BITMAP(c); | |
07be1b83 | 1688 | } |
786e8c11 YO |
1689 | /* this is the Aho-Corasick algorithm modified a touch |
1690 | to include special handling for long "unknown char" | |
1691 | sequences. The basic idea being that we use AC as long | |
1692 | as we are dealing with a possible matching char, when | |
1693 | we encounter an unknown char (and we have not encountered | |
1694 | an accepting state) we scan forward until we find a legal | |
1695 | starting char. | |
1696 | AC matching is basically that of trie matching, except | |
1697 | that when we encounter a failing transition, we fall back | |
1698 | to the current states "fail state", and try the current char | |
1699 | again, a process we repeat until we reach the root state, | |
1700 | state 1, or a legal transition. If we fail on the root state | |
1701 | then we can either terminate if we have reached an accepting | |
1702 | state previously, or restart the entire process from the beginning | |
1703 | if we have not. | |
1704 | ||
1705 | */ | |
07be1b83 YO |
1706 | while (s <= last_start) { |
1707 | const U32 uniflags = UTF8_ALLOW_DEFAULT; | |
1708 | U8 *uc = (U8*)s; | |
1709 | U16 charid = 0; | |
1710 | U32 base = 1; | |
1711 | U32 state = 1; | |
1712 | UV uvc = 0; | |
1713 | STRLEN len = 0; | |
1714 | STRLEN foldlen = 0; | |
1715 | U8 *uscan = (U8*)NULL; | |
1716 | U8 *leftmost = NULL; | |
786e8c11 YO |
1717 | #ifdef DEBUGGING |
1718 | U32 accepted_word= 0; | |
1719 | #endif | |
07be1b83 YO |
1720 | U32 pointpos = 0; |
1721 | ||
1722 | while ( state && uc <= (U8*)strend ) { | |
1723 | int failed=0; | |
786e8c11 YO |
1724 | U32 word = aho->states[ state ].wordnum; |
1725 | ||
1de06328 YO |
1726 | if( state==1 ) { |
1727 | if ( bitmap ) { | |
1728 | DEBUG_TRIE_EXECUTE_r( | |
1729 | if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1730 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1731 | (char *)uc, utf8_target ); |
1de06328 YO |
1732 | PerlIO_printf( Perl_debug_log, |
1733 | " Scanning for legal start char...\n"); | |
1734 | } | |
1735 | ); | |
1736 | while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) { | |
1737 | uc++; | |
786e8c11 | 1738 | } |
1de06328 | 1739 | s= (char *)uc; |
786e8c11 | 1740 | } |
786e8c11 YO |
1741 | if (uc >(U8*)last_start) break; |
1742 | } | |
1743 | ||
1744 | if ( word ) { | |
2e64971a | 1745 | U8 *lpos= points[ (pointpos - trie->wordinfo[word].len) % maxlen ]; |
786e8c11 YO |
1746 | if (!leftmost || lpos < leftmost) { |
1747 | DEBUG_r(accepted_word=word); | |
07be1b83 | 1748 | leftmost= lpos; |
786e8c11 | 1749 | } |
07be1b83 | 1750 | if (base==0) break; |
786e8c11 | 1751 | |
07be1b83 YO |
1752 | } |
1753 | points[pointpos++ % maxlen]= uc; | |
55eed653 NC |
1754 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
1755 | uscan, len, uvc, charid, foldlen, | |
1756 | foldbuf, uniflags); | |
786e8c11 YO |
1757 | DEBUG_TRIE_EXECUTE_r({ |
1758 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1759 | s, utf8_target ); |
07be1b83 | 1760 | PerlIO_printf(Perl_debug_log, |
786e8c11 YO |
1761 | " Charid:%3u CP:%4"UVxf" ", |
1762 | charid, uvc); | |
1763 | }); | |
07be1b83 YO |
1764 | |
1765 | do { | |
6148ee25 | 1766 | #ifdef DEBUGGING |
786e8c11 | 1767 | word = aho->states[ state ].wordnum; |
6148ee25 | 1768 | #endif |
07be1b83 YO |
1769 | base = aho->states[ state ].trans.base; |
1770 | ||
786e8c11 YO |
1771 | DEBUG_TRIE_EXECUTE_r({ |
1772 | if (failed) | |
1773 | dump_exec_pos( (char *)uc, c, strend, real_start, | |
f2ed9b32 | 1774 | s, utf8_target ); |
07be1b83 | 1775 | PerlIO_printf( Perl_debug_log, |
786e8c11 YO |
1776 | "%sState: %4"UVxf", word=%"UVxf, |
1777 | failed ? " Fail transition to " : "", | |
1778 | (UV)state, (UV)word); | |
1779 | }); | |
07be1b83 YO |
1780 | if ( base ) { |
1781 | U32 tmp; | |
6dd2be57 | 1782 | I32 offset; |
07be1b83 | 1783 | if (charid && |
6dd2be57 DM |
1784 | ( ((offset = base + charid |
1785 | - 1 - trie->uniquecharcount)) >= 0) | |
1786 | && ((U32)offset < trie->lasttrans) | |
1787 | && trie->trans[offset].check == state | |
1788 | && (tmp=trie->trans[offset].next)) | |
07be1b83 | 1789 | { |
786e8c11 YO |
1790 | DEBUG_TRIE_EXECUTE_r( |
1791 | PerlIO_printf( Perl_debug_log," - legal\n")); | |
07be1b83 YO |
1792 | state = tmp; |
1793 | break; | |
1794 | } | |
1795 | else { | |
786e8c11 YO |
1796 | DEBUG_TRIE_EXECUTE_r( |
1797 | PerlIO_printf( Perl_debug_log," - fail\n")); | |
1798 | failed = 1; | |
1799 | state = aho->fail[state]; | |
07be1b83 YO |
1800 | } |
1801 | } | |
1802 | else { | |
1803 | /* we must be accepting here */ | |
786e8c11 YO |
1804 | DEBUG_TRIE_EXECUTE_r( |
1805 | PerlIO_printf( Perl_debug_log," - accepting\n")); | |
1806 | failed = 1; | |
07be1b83 YO |
1807 | break; |
1808 | } | |
1809 | } while(state); | |
786e8c11 | 1810 | uc += len; |
07be1b83 YO |
1811 | if (failed) { |
1812 | if (leftmost) | |
1813 | break; | |
786e8c11 | 1814 | if (!state) state = 1; |
07be1b83 YO |
1815 | } |
1816 | } | |
1817 | if ( aho->states[ state ].wordnum ) { | |
2e64971a | 1818 | U8 *lpos = points[ (pointpos - trie->wordinfo[aho->states[ state ].wordnum].len) % maxlen ]; |
786e8c11 YO |
1819 | if (!leftmost || lpos < leftmost) { |
1820 | DEBUG_r(accepted_word=aho->states[ state ].wordnum); | |
07be1b83 | 1821 | leftmost = lpos; |
786e8c11 | 1822 | } |
07be1b83 | 1823 | } |
07be1b83 YO |
1824 | if (leftmost) { |
1825 | s = (char*)leftmost; | |
786e8c11 YO |
1826 | DEBUG_TRIE_EXECUTE_r({ |
1827 | PerlIO_printf( | |
70685ca0 JH |
1828 | Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n", |
1829 | (UV)accepted_word, (IV)(s - real_start) | |
786e8c11 YO |
1830 | ); |
1831 | }); | |
24b23f37 | 1832 | if (!reginfo || regtry(reginfo, &s)) { |
be8e71aa YO |
1833 | FREETMPS; |
1834 | LEAVE; | |
07be1b83 | 1835 | goto got_it; |
be8e71aa | 1836 | } |
07be1b83 | 1837 | s = HOPc(s,1); |
786e8c11 YO |
1838 | DEBUG_TRIE_EXECUTE_r({ |
1839 | PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n"); | |
1840 | }); | |
07be1b83 | 1841 | } else { |
786e8c11 YO |
1842 | DEBUG_TRIE_EXECUTE_r( |
1843 | PerlIO_printf( Perl_debug_log,"No match.\n")); | |
07be1b83 YO |
1844 | break; |
1845 | } | |
1846 | } | |
be8e71aa YO |
1847 | FREETMPS; |
1848 | LEAVE; | |
07be1b83 YO |
1849 | } |
1850 | break; | |
b3c9acc1 | 1851 | default: |
3c3eec57 GS |
1852 | Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); |
1853 | break; | |
d6a28714 | 1854 | } |
6eb5f6b9 JH |
1855 | return 0; |
1856 | got_it: | |
1857 | return s; | |
1858 | } | |
1859 | ||
fae667d5 | 1860 | |
6eb5f6b9 JH |
1861 | /* |
1862 | - regexec_flags - match a regexp against a string | |
1863 | */ | |
1864 | I32 | |
288b8c02 | 1865 | Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend, |
6eb5f6b9 JH |
1866 | char *strbeg, I32 minend, SV *sv, void *data, U32 flags) |
1867 | /* strend: pointer to null at end of string */ | |
1868 | /* strbeg: real beginning of string */ | |
1869 | /* minend: end of match must be >=minend after stringarg. */ | |
58e23c8d YO |
1870 | /* data: May be used for some additional optimizations. |
1871 | Currently its only used, with a U32 cast, for transmitting | |
1872 | the ganch offset when doing a /g match. This will change */ | |
6eb5f6b9 JH |
1873 | /* nosave: For optimizations. */ |
1874 | { | |
97aff369 | 1875 | dVAR; |
288b8c02 | 1876 | struct regexp *const prog = (struct regexp *)SvANY(rx); |
24b23f37 | 1877 | /*register*/ char *s; |
6eb5f6b9 | 1878 | register regnode *c; |
24b23f37 | 1879 | /*register*/ char *startpos = stringarg; |
6eb5f6b9 JH |
1880 | I32 minlen; /* must match at least this many chars */ |
1881 | I32 dontbother = 0; /* how many characters not to try at end */ | |
6eb5f6b9 JH |
1882 | I32 end_shift = 0; /* Same for the end. */ /* CC */ |
1883 | I32 scream_pos = -1; /* Internal iterator of scream. */ | |
ccac19ea | 1884 | char *scream_olds = NULL; |
f2ed9b32 | 1885 | const bool utf8_target = cBOOL(DO_UTF8(sv)); |
2757e526 | 1886 | I32 multiline; |
f8fc2ecf | 1887 | RXi_GET_DECL(prog,progi); |
3b0527fe | 1888 | regmatch_info reginfo; /* create some info to pass to regtry etc */ |
e9105d30 | 1889 | regexp_paren_pair *swap = NULL; |
a3621e74 YO |
1890 | GET_RE_DEBUG_FLAGS_DECL; |
1891 | ||
7918f24d | 1892 | PERL_ARGS_ASSERT_REGEXEC_FLAGS; |
9d4ba2ae | 1893 | PERL_UNUSED_ARG(data); |
6eb5f6b9 JH |
1894 | |
1895 | /* Be paranoid... */ | |
1896 | if (prog == NULL || startpos == NULL) { | |
1897 | Perl_croak(aTHX_ "NULL regexp parameter"); | |
1898 | return 0; | |
1899 | } | |
1900 | ||
bbe252da | 1901 | multiline = prog->extflags & RXf_PMf_MULTILINE; |
288b8c02 | 1902 | reginfo.prog = rx; /* Yes, sorry that this is confusing. */ |
2757e526 | 1903 | |
f2ed9b32 | 1904 | RX_MATCH_UTF8_set(rx, utf8_target); |
1de06328 | 1905 | DEBUG_EXECUTE_r( |
f2ed9b32 | 1906 | debug_start_match(rx, utf8_target, startpos, strend, |
1de06328 YO |
1907 | "Matching"); |
1908 | ); | |
bac06658 | 1909 | |
6eb5f6b9 | 1910 | minlen = prog->minlen; |
1de06328 YO |
1911 | |
1912 | if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) { | |
a3621e74 | 1913 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
a72c7584 JH |
1914 | "String too short [regexec_flags]...\n")); |
1915 | goto phooey; | |
1aa99e6b | 1916 | } |
6eb5f6b9 | 1917 | |
1de06328 | 1918 | |
6eb5f6b9 | 1919 | /* Check validity of program. */ |
f8fc2ecf | 1920 | if (UCHARAT(progi->program) != REG_MAGIC) { |
6eb5f6b9 JH |
1921 | Perl_croak(aTHX_ "corrupted regexp program"); |
1922 | } | |
1923 | ||
1924 | PL_reg_flags = 0; | |
1925 | PL_reg_eval_set = 0; | |
1926 | PL_reg_maxiter = 0; | |
1927 | ||
3c8556c3 | 1928 | if (RX_UTF8(rx)) |
6eb5f6b9 JH |
1929 | PL_reg_flags |= RF_utf8; |
1930 | ||
1931 | /* Mark beginning of line for ^ and lookbehind. */ | |
3b0527fe | 1932 | reginfo.bol = startpos; /* XXX not used ??? */ |
6eb5f6b9 | 1933 | PL_bostr = strbeg; |
3b0527fe | 1934 | reginfo.sv = sv; |
6eb5f6b9 JH |
1935 | |
1936 | /* Mark end of line for $ (and such) */ | |
1937 | PL_regeol = strend; | |
1938 | ||
1939 | /* see how far we have to get to not match where we matched before */ | |
3b0527fe | 1940 | reginfo.till = startpos+minend; |
6eb5f6b9 | 1941 | |
6eb5f6b9 JH |
1942 | /* If there is a "must appear" string, look for it. */ |
1943 | s = startpos; | |
1944 | ||
bbe252da | 1945 | if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ |
6eb5f6b9 | 1946 | MAGIC *mg; |
2c296965 | 1947 | if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */ |
58e23c8d | 1948 | reginfo.ganch = startpos + prog->gofs; |
2c296965 | 1949 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 1950 | "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs)); |
2c296965 | 1951 | } else if (sv && SvTYPE(sv) >= SVt_PVMG |
6eb5f6b9 | 1952 | && SvMAGIC(sv) |
14befaf4 DM |
1953 | && (mg = mg_find(sv, PERL_MAGIC_regex_global)) |
1954 | && mg->mg_len >= 0) { | |
3b0527fe | 1955 | reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ |
2c296965 | 1956 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 1957 | "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len)); |
2c296965 | 1958 | |
bbe252da | 1959 | if (prog->extflags & RXf_ANCH_GPOS) { |
3b0527fe | 1960 | if (s > reginfo.ganch) |
6eb5f6b9 | 1961 | goto phooey; |
58e23c8d | 1962 | s = reginfo.ganch - prog->gofs; |
2c296965 | 1963 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
ed549f2e | 1964 | "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs)); |
c584a96e YO |
1965 | if (s < strbeg) |
1966 | goto phooey; | |
6eb5f6b9 JH |
1967 | } |
1968 | } | |
58e23c8d | 1969 | else if (data) { |
70685ca0 | 1970 | reginfo.ganch = strbeg + PTR2UV(data); |
2c296965 YO |
1971 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
1972 | "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data))); | |
1973 | ||
1974 | } else { /* pos() not defined */ | |
3b0527fe | 1975 | reginfo.ganch = strbeg; |
2c296965 YO |
1976 | DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, |
1977 | "GPOS: reginfo.ganch = strbeg\n")); | |
1978 | } | |
6eb5f6b9 | 1979 | } |
288b8c02 | 1980 | if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { |
e9105d30 GG |
1981 | /* We have to be careful. If the previous successful match |
1982 | was from this regex we don't want a subsequent partially | |
1983 | successful match to clobber the old results. | |
1984 | So when we detect this possibility we add a swap buffer | |
1985 | to the re, and switch the buffer each match. If we fail | |
1986 | we switch it back, otherwise we leave it swapped. | |
1987 | */ | |
1988 | swap = prog->offs; | |
1989 | /* do we need a save destructor here for eval dies? */ | |
1990 | Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); | |
c74340f9 | 1991 | } |
a0714e2c | 1992 | if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { |
6eb5f6b9 JH |
1993 | re_scream_pos_data d; |
1994 | ||
1995 | d.scream_olds = &scream_olds; | |
1996 | d.scream_pos = &scream_pos; | |
288b8c02 | 1997 | s = re_intuit_start(rx, sv, s, strend, flags, &d); |
3fa9c3d7 | 1998 | if (!s) { |
a3621e74 | 1999 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n")); |
6eb5f6b9 | 2000 | goto phooey; /* not present */ |
3fa9c3d7 | 2001 | } |
6eb5f6b9 JH |
2002 | } |
2003 | ||
1de06328 | 2004 | |
6eb5f6b9 JH |
2005 | |
2006 | /* Simplest case: anchored match need be tried only once. */ | |
2007 | /* [unless only anchor is BOL and multiline is set] */ | |
bbe252da | 2008 | if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) { |
24b23f37 | 2009 | if (s == startpos && regtry(®info, &startpos)) |
6eb5f6b9 | 2010 | goto got_it; |
bbe252da YO |
2011 | else if (multiline || (prog->intflags & PREGf_IMPLICIT) |
2012 | || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */ | |
6eb5f6b9 JH |
2013 | { |
2014 | char *end; | |
2015 | ||
2016 | if (minlen) | |
2017 | dontbother = minlen - 1; | |
1aa99e6b | 2018 | end = HOP3c(strend, -dontbother, strbeg) - 1; |
6eb5f6b9 | 2019 | /* for multiline we only have to try after newlines */ |
33b8afdf | 2020 | if (prog->check_substr || prog->check_utf8) { |
92f3d482 YO |
2021 | /* because of the goto we can not easily reuse the macros for bifurcating the |
2022 | unicode/non-unicode match modes here like we do elsewhere - demerphq */ | |
2023 | if (utf8_target) { | |
2024 | if (s == startpos) | |
2025 | goto after_try_utf8; | |
2026 | while (1) { | |
2027 | if (regtry(®info, &s)) { | |
2028 | goto got_it; | |
2029 | } | |
2030 | after_try_utf8: | |
2031 | if (s > end) { | |
2032 | goto phooey; | |
2033 | } | |
2034 | if (prog->extflags & RXf_USE_INTUIT) { | |
2035 | s = re_intuit_start(rx, sv, s + UTF8SKIP(s), strend, flags, NULL); | |
2036 | if (!s) { | |
2037 | goto phooey; | |
2038 | } | |
2039 | } | |
2040 | else { | |
2041 | s += UTF8SKIP(s); | |
2042 | } | |
2043 | } | |
2044 | } /* end search for check string in unicode */ | |
2045 | else { | |
2046 | if (s == startpos) { | |
2047 | goto after_try_latin; | |
2048 | } | |
2049 | while (1) { | |
2050 | if (regtry(®info, &s)) { | |
2051 | goto got_it; | |
2052 | } | |
2053 | after_try_latin: | |
2054 | if (s > end) { | |
2055 | goto phooey; | |
2056 | } | |
2057 | if (prog->extflags & RXf_USE_INTUIT) { | |
2058 | s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL); | |
2059 | if (!s) { | |
2060 | goto phooey; | |
2061 | } | |
2062 | } | |
2063 | else { | |
2064 | s++; | |
2065 | } | |
2066 | } | |
2067 | } /* end search for check string in latin*/ | |
2068 | } /* end search for check string */ | |
2069 | else { /* search for newline */ | |
2070 | if (s > startpos) { | |
2071 | /*XXX: The s-- is almost definitely wrong here under unicode - demeprhq*/ | |
6eb5f6b9 | 2072 | s--; |
92f3d482 YO |
2073 | } |
2074 | /* We can use a more efficient search as newlines are the same in unicode as they are in latin */ | |
6eb5f6b9 JH |
2075 | while (s < end) { |
2076 | if (*s++ == '\n') { /* don't need PL_utf8skip here */ | |
24b23f37 | 2077 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2078 | goto got_it; |
2079 | } | |
92f3d482 YO |
2080 | } |
2081 | } /* end search for newline */ | |
2082 | } /* end anchored/multiline check string search */ | |
6eb5f6b9 | 2083 | goto phooey; |
bbe252da | 2084 | } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK)) |
f9f4320a YO |
2085 | { |
2086 | /* the warning about reginfo.ganch being used without intialization | |
bbe252da | 2087 | is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN |
f9f4320a | 2088 | and we only enter this block when the same bit is set. */ |
58e23c8d | 2089 | char *tmp_s = reginfo.ganch - prog->gofs; |
c584a96e YO |
2090 | |
2091 | if (tmp_s >= strbeg && regtry(®info, &tmp_s)) | |
6eb5f6b9 JH |
2092 | goto got_it; |
2093 | goto phooey; | |
2094 | } | |
2095 | ||
2096 | /* Messy cases: unanchored match. */ | |
bbe252da | 2097 | if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) { |
6eb5f6b9 | 2098 | /* we have /x+whatever/ */ |
f2ed9b32 | 2099 | /* it must be a one character string (XXXX Except UTF_PATTERN?) */ |
33b8afdf | 2100 | char ch; |
bf93d4cc GS |
2101 | #ifdef DEBUGGING |
2102 | int did_match = 0; | |
2103 | #endif | |
f2ed9b32 KW |
2104 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2105 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2106 | ch = SvPVX_const(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)[0]; | |
bf93d4cc | 2107 | |
f2ed9b32 | 2108 | if (utf8_target) { |
4cadc6a9 | 2109 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2110 | if (*s == ch) { |
a3621e74 | 2111 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2112 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2113 | s += UTF8SKIP(s); |
2114 | while (s < strend && *s == ch) | |
2115 | s += UTF8SKIP(s); | |
2116 | } | |
4cadc6a9 | 2117 | ); |
6eb5f6b9 JH |
2118 | } |
2119 | else { | |
4cadc6a9 | 2120 | REXEC_FBC_SCAN( |
6eb5f6b9 | 2121 | if (*s == ch) { |
a3621e74 | 2122 | DEBUG_EXECUTE_r( did_match = 1 ); |
24b23f37 | 2123 | if (regtry(®info, &s)) goto got_it; |
6eb5f6b9 JH |
2124 | s++; |
2125 | while (s < strend && *s == ch) | |
2126 | s++; | |
2127 | } | |
4cadc6a9 | 2128 | ); |
6eb5f6b9 | 2129 | } |
a3621e74 | 2130 | DEBUG_EXECUTE_r(if (!did_match) |
bf93d4cc | 2131 | PerlIO_printf(Perl_debug_log, |
b7953727 JH |
2132 | "Did not find anchored character...\n") |
2133 | ); | |
6eb5f6b9 | 2134 | } |
a0714e2c SS |
2135 | else if (prog->anchored_substr != NULL |
2136 | || prog->anchored_utf8 != NULL | |
2137 | || ((prog->float_substr != NULL || prog->float_utf8 != NULL) | |
33b8afdf JH |
2138 | && prog->float_max_offset < strend - s)) { |
2139 | SV *must; | |
2140 | I32 back_max; | |
2141 | I32 back_min; | |
2142 | char *last; | |
6eb5f6b9 | 2143 | char *last1; /* Last position checked before */ |
bf93d4cc GS |
2144 | #ifdef DEBUGGING |
2145 | int did_match = 0; | |
2146 | #endif | |
33b8afdf | 2147 | if (prog->anchored_substr || prog->anchored_utf8) { |
f2ed9b32 KW |
2148 | if (!(utf8_target ? prog->anchored_utf8 : prog->anchored_substr)) |
2149 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2150 | must = utf8_target ? prog->anchored_utf8 : prog->anchored_substr; | |
33b8afdf JH |
2151 | back_max = back_min = prog->anchored_offset; |
2152 | } else { | |
f2ed9b32 KW |
2153 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2154 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2155 | must = utf8_target ? prog->float_utf8 : prog->float_substr; | |
33b8afdf JH |
2156 | back_max = prog->float_max_offset; |
2157 | back_min = prog->float_min_offset; | |
2158 | } | |
1de06328 YO |
2159 | |
2160 | ||
33b8afdf JH |
2161 | if (must == &PL_sv_undef) |
2162 | /* could not downgrade utf8 check substring, so must fail */ | |
2163 | goto phooey; | |
2164 | ||
1de06328 YO |
2165 | if (back_min<0) { |
2166 | last = strend; | |
2167 | } else { | |
2168 | last = HOP3c(strend, /* Cannot start after this */ | |
2169 | -(I32)(CHR_SVLEN(must) | |
2170 | - (SvTAIL(must) != 0) + back_min), strbeg); | |
2171 | } | |
6eb5f6b9 JH |
2172 | if (s > PL_bostr) |
2173 | last1 = HOPc(s, -1); | |
2174 | else | |
2175 | last1 = s - 1; /* bogus */ | |
2176 | ||
a0288114 | 2177 | /* XXXX check_substr already used to find "s", can optimize if |
6eb5f6b9 JH |
2178 | check_substr==must. */ |
2179 | scream_pos = -1; | |
2180 | dontbother = end_shift; | |
2181 | strend = HOPc(strend, -dontbother); | |
2182 | while ( (s <= last) && | |
9041c2e3 | 2183 | ((flags & REXEC_SCREAM) |
1de06328 | 2184 | ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg, |
6eb5f6b9 | 2185 | end_shift, &scream_pos, 0)) |
1de06328 | 2186 | : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)), |
9041c2e3 | 2187 | (unsigned char*)strend, must, |
7fba1cd6 | 2188 | multiline ? FBMrf_MULTILINE : 0))) ) { |
4addbd3b | 2189 | /* we may be pointing at the wrong string */ |
07bc277f | 2190 | if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog)) |
3f7c398e | 2191 | s = strbeg + (s - SvPVX_const(sv)); |
a3621e74 | 2192 | DEBUG_EXECUTE_r( did_match = 1 ); |
6eb5f6b9 JH |
2193 | if (HOPc(s, -back_max) > last1) { |
2194 | last1 = HOPc(s, -back_min); | |
2195 | s = HOPc(s, -back_max); | |
2196 | } | |
2197 | else { | |
52657f30 | 2198 | char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; |
6eb5f6b9 JH |
2199 | |
2200 | last1 = HOPc(s, -back_min); | |
52657f30 | 2201 | s = t; |
6eb5f6b9 | 2202 | } |
f2ed9b32 | 2203 | if (utf8_target) { |
6eb5f6b9 | 2204 | while (s <= last1) { |
24b23f37 | 2205 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2206 | goto got_it; |
2207 | s += UTF8SKIP(s); | |
2208 | } | |
2209 | } | |
2210 | else { | |
2211 | while (s <= last1) { | |
24b23f37 | 2212 | if (regtry(®info, &s)) |
6eb5f6b9 JH |
2213 | goto got_it; |
2214 | s++; | |
2215 | } | |
2216 | } | |
2217 | } | |
ab3bbdeb | 2218 | DEBUG_EXECUTE_r(if (!did_match) { |
f2ed9b32 | 2219 | RE_PV_QUOTED_DECL(quoted, utf8_target, PERL_DEBUG_PAD_ZERO(0), |
ab3bbdeb YO |
2220 | SvPVX_const(must), RE_SV_DUMPLEN(must), 30); |
2221 | PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n", | |
33b8afdf | 2222 | ((must == prog->anchored_substr || must == prog->anchored_utf8) |
bf93d4cc | 2223 | ? "anchored" : "floating"), |
ab3bbdeb YO |
2224 | quoted, RE_SV_TAIL(must)); |
2225 | }); | |
6eb5f6b9 JH |
2226 | goto phooey; |
2227 | } | |
f8fc2ecf | 2228 | else if ( (c = progi->regstclass) ) { |
f14c76ed | 2229 | if (minlen) { |
f8fc2ecf | 2230 | const OPCODE op = OP(progi->regstclass); |
66e933ab | 2231 | /* don't bother with what can't match */ |
786e8c11 | 2232 | if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE) |
f14c76ed RGS |
2233 | strend = HOPc(strend, -(minlen - 1)); |
2234 | } | |
a3621e74 | 2235 | DEBUG_EXECUTE_r({ |
be8e71aa | 2236 | SV * const prop = sv_newmortal(); |
32fc9b6a | 2237 | regprop(prog, prop, c); |
0df25f3d | 2238 | { |
f2ed9b32 | 2239 | RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb | 2240 | s,strend-s,60); |
0df25f3d | 2241 | PerlIO_printf(Perl_debug_log, |
1c8f8eb1 | 2242 | "Matching stclass %.*s against %s (%d bytes)\n", |
e4f74956 | 2243 | (int)SvCUR(prop), SvPVX_const(prop), |
ab3bbdeb | 2244 | quoted, (int)(strend - s)); |
0df25f3d | 2245 | } |
ffc61ed2 | 2246 | }); |
3b0527fe | 2247 | if (find_byclass(prog, c, s, strend, ®info)) |
6eb5f6b9 | 2248 | goto got_it; |
07be1b83 | 2249 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n")); |
d6a28714 JH |
2250 | } |
2251 | else { | |
2252 | dontbother = 0; | |
a0714e2c | 2253 | if (prog->float_substr != NULL || prog->float_utf8 != NULL) { |
33b8afdf | 2254 | /* Trim the end. */ |
d6a28714 | 2255 | char *last; |
33b8afdf JH |
2256 | SV* float_real; |
2257 | ||
f2ed9b32 KW |
2258 | if (!(utf8_target ? prog->float_utf8 : prog->float_substr)) |
2259 | utf8_target ? to_utf8_substr(prog) : to_byte_substr(prog); | |
2260 | float_real = utf8_target ? prog->float_utf8 : prog->float_substr; | |
d6a28714 JH |
2261 | |
2262 | if (flags & REXEC_SCREAM) { | |
33b8afdf | 2263 | last = screaminstr(sv, float_real, s - strbeg, |
d6a28714 JH |
2264 | end_shift, &scream_pos, 1); /* last one */ |
2265 | if (!last) | |
ffc61ed2 | 2266 | last = scream_olds; /* Only one occurrence. */ |
4addbd3b | 2267 | /* we may be pointing at the wrong string */ |
07bc277f | 2268 | else if (RXp_MATCH_COPIED(prog)) |
3f7c398e | 2269 | s = strbeg + (s - SvPVX_const(sv)); |
b8c5462f | 2270 | } |
d6a28714 JH |
2271 | else { |
2272 | STRLEN len; | |
cfd0369c | 2273 | const char * const little = SvPV_const(float_real, len); |
d6a28714 | 2274 | |
33b8afdf | 2275 | if (SvTAIL(float_real)) { |
d6a28714 JH |
2276 | if (memEQ(strend - len + 1, little, len - 1)) |
2277 | last = strend - len + 1; | |
7fba1cd6 | 2278 | else if (!multiline) |
9041c2e3 | 2279 | last = memEQ(strend - len, little, len) |
bd61b366 | 2280 | ? strend - len : NULL; |
b8c5462f | 2281 | else |
d6a28714 JH |
2282 | goto find_last; |
2283 | } else { | |
2284 | find_last: | |
9041c2e3 | 2285 | if (len) |
d6a28714 | 2286 | last = rninstr(s, strend, little, little + len); |
b8c5462f | 2287 | else |
a0288114 | 2288 | last = strend; /* matching "$" */ |
b8c5462f | 2289 | } |
b8c5462f | 2290 | } |
bf93d4cc | 2291 | if (last == NULL) { |
6bda09f9 YO |
2292 | DEBUG_EXECUTE_r( |
2293 | PerlIO_printf(Perl_debug_log, | |
2294 | "%sCan't trim the tail, match fails (should not happen)%s\n", | |
2295 | PL_colors[4], PL_colors[5])); | |
bf93d4cc GS |
2296 | goto phooey; /* Should not happen! */ |
2297 | } | |
d6a28714 JH |
2298 | dontbother = strend - last + prog->float_min_offset; |
2299 | } | |
2300 | if (minlen && (dontbother < minlen)) | |
2301 | dontbother = minlen - 1; | |
2302 | strend -= dontbother; /* this one's always in bytes! */ | |
2303 | /* We don't know much -- general case. */ | |
f2ed9b32 | 2304 | if (utf8_target) { |
d6a28714 | 2305 | for (;;) { |
24b23f37 | 2306 | if (regtry(®info, &s)) |
d6a28714 JH |
2307 | goto got_it; |
2308 | if (s >= strend) | |
2309 | break; | |
b8c5462f | 2310 | s += UTF8SKIP(s); |
d6a28714 JH |
2311 | }; |
2312 | } | |
2313 | else { | |
2314 | do { | |
24b23f37 | 2315 | if (regtry(®info, &s)) |
d6a28714 JH |
2316 | goto got_it; |
2317 | } while (s++ < strend); | |
2318 | } | |
2319 | } | |
2320 | ||
2321 | /* Failure. */ | |
2322 | goto phooey; | |
2323 | ||
2324 | got_it: | |
e9105d30 | 2325 | Safefree(swap); |
288b8c02 | 2326 | RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted); |
d6a28714 | 2327 | |
19b95bf0 | 2328 | if (PL_reg_eval_set) |
4f639d21 | 2329 | restore_pos(aTHX_ prog); |
5daac39c NC |
2330 | if (RXp_PAREN_NAMES(prog)) |
2331 | (void)hv_iterinit(RXp_PAREN_NAMES(prog)); | |
d6a28714 JH |
2332 | |
2333 | /* make sure $`, $&, $', and $digit will work later */ | |
2334 | if ( !(flags & REXEC_NOT_FIRST) ) { | |
288b8c02 | 2335 | RX_MATCH_COPY_FREE(rx); |
d6a28714 | 2336 | if (flags & REXEC_COPY_STR) { |
be8e71aa | 2337 | const I32 i = PL_regeol - startpos + (stringarg - strbeg); |
f8c7b90f | 2338 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2339 | if ((SvIsCOW(sv) |
2340 | || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { | |
2341 | if (DEBUG_C_TEST) { | |
2342 | PerlIO_printf(Perl_debug_log, | |
2343 | "Copy on write: regexp capture, type %d\n", | |
2344 | (int) SvTYPE(sv)); | |
2345 | } | |
2346 | prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv); | |
d5263905 | 2347 | prog->subbeg = (char *)SvPVX_const(prog->saved_copy); |
ed252734 NC |
2348 | assert (SvPOKp(prog->saved_copy)); |
2349 | } else | |
2350 | #endif | |
2351 | { | |
288b8c02 | 2352 | RX_MATCH_COPIED_on(rx); |
ed252734 NC |
2353 | s = savepvn(strbeg, i); |
2354 | prog->subbeg = s; | |
2355 | } | |
d6a28714 | 2356 | prog->sublen = i; |
d6a28714 JH |
2357 | } |
2358 | else { | |
2359 | prog->subbeg = strbeg; | |
2360 | prog->sublen = PL_regeol - strbeg; /* strend may have been modified */ | |
2361 | } | |
2362 | } | |
9041c2e3 | 2363 | |
d6a28714 JH |
2364 | return 1; |
2365 | ||
2366 | phooey: | |
a3621e74 | 2367 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", |
e4584336 | 2368 | PL_colors[4], PL_colors[5])); |
d6a28714 | 2369 | if (PL_reg_eval_set) |
4f639d21 | 2370 | restore_pos(aTHX_ prog); |
e9105d30 | 2371 | if (swap) { |
c74340f9 | 2372 | /* we failed :-( roll it back */ |
e9105d30 GG |
2373 | Safefree(prog->offs); |
2374 | prog->offs = swap; | |
2375 | } | |
2376 | ||
d6a28714 JH |
2377 | return 0; |
2378 | } | |
2379 | ||
6bda09f9 | 2380 | |
d6a28714 JH |
2381 | /* |
2382 | - regtry - try match at specific point | |
2383 | */ | |
2384 | STATIC I32 /* 0 failure, 1 success */ | |
24b23f37 | 2385 | S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) |
d6a28714 | 2386 | { |
97aff369 | 2387 | dVAR; |
d6a28714 | 2388 | CHECKPOINT lastcp; |
288b8c02 NC |
2389 | REGEXP *const rx = reginfo->prog; |
2390 | regexp *const prog = (struct regexp *)SvANY(rx); | |
f8fc2ecf | 2391 | RXi_GET_DECL(prog,progi); |
a3621e74 | 2392 | GET_RE_DEBUG_FLAGS_DECL; |
7918f24d NC |
2393 | |
2394 | PERL_ARGS_ASSERT_REGTRY; | |
2395 | ||
24b23f37 | 2396 | reginfo->cutpoint=NULL; |
d6a28714 | 2397 | |
bbe252da | 2398 | if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) { |
d6a28714 JH |
2399 | MAGIC *mg; |
2400 | ||
2401 | PL_reg_eval_set = RS_init; | |
a3621e74 | 2402 | DEBUG_EXECUTE_r(DEBUG_s( |
b900a521 JH |
2403 | PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n", |
2404 | (IV)(PL_stack_sp - PL_stack_base)); | |
d6a28714 | 2405 | )); |
ea8d6ae1 | 2406 | SAVESTACK_CXPOS(); |
d6a28714 JH |
2407 | cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base; |
2408 | /* Otherwise OP_NEXTSTATE will free whatever on stack now. */ | |
2409 | SAVETMPS; | |
2410 | /* Apparently this is not needed, judging by wantarray. */ | |
e8347627 | 2411 | /* SAVEI8(cxstack[cxstack_ix].blk_gimme); |
d6a28714 JH |
2412 | cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ |
2413 | ||
3b0527fe | 2414 | if (reginfo->sv) { |
d6a28714 | 2415 | /* Make $_ available to executed code. */ |
3b0527fe | 2416 | if (reginfo->sv != DEFSV) { |
59f00321 | 2417 | SAVE_DEFSV; |
414bf5ae | 2418 | DEFSV_set(reginfo->sv); |
b8c5462f | 2419 | } |
d6a28714 | 2420 | |
3b0527fe DM |
2421 | if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) |
2422 | && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { | |
d6a28714 | 2423 | /* prepare for quick setting of pos */ |
d300d9fa | 2424 | #ifdef PERL_OLD_COPY_ON_WRITE |
51a9ea20 NC |
2425 | if (SvIsCOW(reginfo->sv)) |
2426 | sv_force_normal_flags(reginfo->sv, 0); | |
d300d9fa | 2427 | #endif |
3dab1dad | 2428 | mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global, |
d300d9fa | 2429 | &PL_vtbl_mglob, NULL, 0); |
d6a28714 | 2430 | mg->mg_len = -1; |
b8c5462f | 2431 | } |
d6a28714 JH |
2432 | PL_reg_magic = mg; |
2433 | PL_reg_oldpos = mg->mg_len; | |
4f639d21 | 2434 | SAVEDESTRUCTOR_X(restore_pos, prog); |
d6a28714 | 2435 | } |
09687e5a | 2436 | if (!PL_reg_curpm) { |
a02a5408 | 2437 | Newxz(PL_reg_curpm, 1, PMOP); |
09687e5a AB |
2438 | #ifdef USE_ITHREADS |
2439 | { | |
14a49a24 | 2440 | SV* const repointer = &PL_sv_undef; |
92313705 NC |
2441 | /* this regexp is also owned by the new PL_reg_curpm, which |
2442 | will try to free it. */ | |
d2ece331 | 2443 | av_push(PL_regex_padav, repointer); |
09687e5a AB |
2444 | PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); |
2445 | PL_regex_pad = AvARRAY(PL_regex_padav); | |
2446 | } | |
2447 | #endif | |
2448 | } | |
86c29d75 NC |
2449 | #ifdef USE_ITHREADS |
2450 | /* It seems that non-ithreads works both with and without this code. | |
2451 | So for efficiency reasons it seems best not to have the code | |
2452 | compiled when it is not needed. */ | |
92313705 NC |
2453 | /* This is safe against NULLs: */ |
2454 | ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); | |
2455 | /* PM_reg_curpm owns a reference to this regexp. */ | |
2456 | ReREFCNT_inc(rx); | |
86c29d75 | 2457 | #endif |
288b8c02 | 2458 | PM_SETRE(PL_reg_curpm, rx); |
d6a28714 JH |
2459 | PL_reg_oldcurpm = PL_curpm; |
2460 | PL_curpm = PL_reg_curpm; | |
07bc277f | 2461 | if (RXp_MATCH_COPIED(prog)) { |
d6a28714 JH |
2462 | /* Here is a serious problem: we cannot rewrite subbeg, |
2463 | since it may be needed if this match fails. Thus | |
2464 | $` inside (?{}) could fail... */ | |
2465 | PL_reg_oldsaved = prog->subbeg; | |
2466 | PL_reg_oldsavedlen = prog->sublen; | |
f8c7b90f | 2467 | #ifdef PERL_OLD_COPY_ON_WRITE |
ed252734 NC |
2468 | PL_nrs = prog->saved_copy; |
2469 | #endif | |
07bc277f | 2470 | RXp_MATCH_COPIED_off(prog); |
d6a28714 JH |
2471 | } |
2472 | else | |
bd61b366 | 2473 | PL_reg_oldsaved = NULL; |
d6a28714 JH |
2474 | prog->subbeg = PL_bostr; |
2475 | prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ | |
2476 | } | |
24b23f37 | 2477 | DEBUG_EXECUTE_r(PL_reg_starttry = *startpos); |
f0ab9afb | 2478 | prog->offs[0].start = *startpos - PL_bostr; |
24b23f37 | 2479 | PL_reginput = *startpos; |
d6a28714 | 2480 | PL_reglastparen = &prog->lastparen; |
a01268b5 | 2481 | PL_reglastcloseparen = &prog->lastcloseparen; |
d6a28714 | 2482 | prog->lastparen = 0; |
03994de8 | 2483 | prog->lastcloseparen = 0; |
d6a28714 | 2484 | PL_regsize = 0; |
f0ab9afb | 2485 | PL_regoffs = prog->offs; |
d6a28714 JH |
2486 | if (PL_reg_start_tmpl <= prog->nparens) { |
2487 | PL_reg_start_tmpl = prog->nparens*3/2 + 3; | |
2488 | if(PL_reg_start_tmp) | |
2489 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
2490 | else | |
a02a5408 | 2491 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); |
d6a28714 JH |
2492 | } |
2493 | ||
2494 | /* XXXX What this code is doing here?!!! There should be no need | |
2495 | to do this again and again, PL_reglastparen should take care of | |
3dd2943c | 2496 | this! --ilya*/ |
dafc8851 JH |
2497 | |
2498 | /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. | |
2499 | * Actually, the code in regcppop() (which Ilya may be meaning by | |
daf18116 | 2500 | * PL_reglastparen), is not needed at all by the test suite |
225593e1 DM |
2501 | * (op/regexp, op/pat, op/split), but that code is needed otherwise |
2502 | * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/ | |
2503 | * Meanwhile, this code *is* needed for the | |
daf18116 JH |
2504 | * above-mentioned test suite tests to succeed. The common theme |
2505 | * on those tests seems to be returning null fields from matches. | |
225593e1 | 2506 | * --jhi updated by dapm */ |
dafc8851 | 2507 | #if 1 |
d6a28714 | 2508 | if (prog->nparens) { |
f0ab9afb | 2509 | regexp_paren_pair *pp = PL_regoffs; |
097eb12c | 2510 | register I32 i; |
eb160463 | 2511 | for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) { |
f0ab9afb NC |
2512 | ++pp; |
2513 | pp->start = -1; | |
2514 | pp->end = -1; | |
d6a28714 JH |
2515 | } |
2516 | } | |
dafc8851 | 2517 | #endif |
02db2b7b | 2518 | REGCP_SET(lastcp); |
f8fc2ecf | 2519 | if (regmatch(reginfo, progi->program + 1)) { |
f0ab9afb | 2520 | PL_regoffs[0].end = PL_reginput - PL_bostr; |
d6a28714 JH |
2521 | return 1; |
2522 | } | |
24b23f37 YO |
2523 | if (reginfo->cutpoint) |
2524 | *startpos= reginfo->cutpoint; | |
02db2b7b | 2525 | REGCP_UNWIND(lastcp); |
d6a28714 JH |
2526 | return 0; |
2527 | } | |
2528 | ||
02db2b7b | 2529 | |
8ba1375e MJD |
2530 | #define sayYES goto yes |
2531 | #define sayNO goto no | |
262b90c4 | 2532 | #define sayNO_SILENT goto no_silent |
8ba1375e | 2533 | |
f9f4320a YO |
2534 | /* we dont use STMT_START/END here because it leads to |
2535 | "unreachable code" warnings, which are bogus, but distracting. */ | |
2536 | #define CACHEsayNO \ | |
c476f425 DM |
2537 | if (ST.cache_mask) \ |
2538 | PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \ | |
f9f4320a | 2539 | sayNO |
3298f257 | 2540 | |
a3621e74 | 2541 | /* this is used to determine how far from the left messages like |
265c4333 YO |
2542 | 'failed...' are printed. It should be set such that messages |
2543 | are inline with the regop output that created them. | |
a3621e74 | 2544 | */ |
265c4333 | 2545 | #define REPORT_CODE_OFF 32 |
a3621e74 YO |
2546 | |
2547 | ||
40a82448 DM |
2548 | #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */ |
2549 | #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */ | |
9e137952 | 2550 | |
86545054 DM |
2551 | #define SLAB_FIRST(s) (&(s)->states[0]) |
2552 | #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) | |
2553 | ||
5d9a96ca DM |
2554 | /* grab a new slab and return the first slot in it */ |
2555 | ||
2556 | STATIC regmatch_state * | |
2557 | S_push_slab(pTHX) | |
2558 | { | |
a35a87e7 | 2559 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2560 | dMY_CXT; |
2561 | #endif | |
5d9a96ca DM |
2562 | regmatch_slab *s = PL_regmatch_slab->next; |
2563 | if (!s) { | |
2564 | Newx(s, 1, regmatch_slab); | |
2565 | s->prev = PL_regmatch_slab; | |
2566 | s->next = NULL; | |
2567 | PL_regmatch_slab->next = s; | |
2568 | } | |
2569 | PL_regmatch_slab = s; | |
86545054 | 2570 | return SLAB_FIRST(s); |
5d9a96ca | 2571 | } |
5b47454d | 2572 | |
95b24440 | 2573 | |
40a82448 DM |
2574 | /* push a new state then goto it */ |
2575 | ||
2576 | #define PUSH_STATE_GOTO(state, node) \ | |
2577 | scan = node; \ | |
2578 | st->resume_state = state; \ | |
2579 | goto push_state; | |
2580 | ||
2581 | /* push a new state with success backtracking, then goto it */ | |
2582 | ||
2583 | #define PUSH_YES_STATE_GOTO(state, node) \ | |
2584 | scan = node; \ | |
2585 | st->resume_state = state; \ | |
2586 | goto push_yes_state; | |
2587 | ||
aa283a38 | 2588 | |
aa283a38 | 2589 | |
d6a28714 | 2590 | /* |
95b24440 | 2591 | |
bf1f174e DM |
2592 | regmatch() - main matching routine |
2593 | ||
2594 | This is basically one big switch statement in a loop. We execute an op, | |
2595 | set 'next' to point the next op, and continue. If we come to a point which | |
2596 | we may need to backtrack to on failure such as (A|B|C), we push a | |
2597 | backtrack state onto the backtrack stack. On failure, we pop the top | |
2598 | state, and re-enter the loop at the state indicated. If there are no more | |
2599 | states to pop, we return failure. | |
2600 | ||
2601 | Sometimes we also need to backtrack on success; for example /A+/, where | |
2602 | after successfully matching one A, we need to go back and try to | |
2603 | match another one; similarly for lookahead assertions: if the assertion | |
2604 | completes successfully, we backtrack to the state just before the assertion | |
2605 | and then carry on. In these cases, the pushed state is marked as | |
2606 | 'backtrack on success too'. This marking is in fact done by a chain of | |
2607 | pointers, each pointing to the previous 'yes' state. On success, we pop to | |
2608 | the nearest yes state, discarding any intermediate failure-only states. | |
2609 | Sometimes a yes state is pushed just to force some cleanup code to be | |
2610 | called at the end of a successful match or submatch; e.g. (??{$re}) uses | |
2611 | it to free the inner regex. | |
2612 | ||
2613 | Note that failure backtracking rewinds the cursor position, while | |
2614 | success backtracking leaves it alone. | |
2615 | ||
2616 | A pattern is complete when the END op is executed, while a subpattern | |
2617 | such as (?=foo) is complete when the SUCCESS op is executed. Both of these | |
2618 | ops trigger the "pop to last yes state if any, otherwise return true" | |
2619 | behaviour. | |
2620 | ||
2621 | A common convention in this function is to use A and B to refer to the two | |
2622 | subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is | |
2623 | the subpattern to be matched possibly multiple times, while B is the entire | |
2624 | rest of the pattern. Variable and state names reflect this convention. | |
2625 | ||
2626 | The states in the main switch are the union of ops and failure/success of | |
2627 | substates associated with with that op. For example, IFMATCH is the op | |
2628 | that does lookahead assertions /(?=A)B/ and so the IFMATCH state means | |
2629 | 'execute IFMATCH'; while IFMATCH_A is a state saying that we have just | |
2630 | successfully matched A and IFMATCH_A_fail is a state saying that we have | |
2631 | just failed to match A. Resume states always come in pairs. The backtrack | |
2632 | state we push is marked as 'IFMATCH_A', but when that is popped, we resume | |
2633 | at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking | |
2634 | on success or failure. | |
2635 | ||
2636 | The struct that holds a backtracking state is actually a big union, with | |
2637 | one variant for each major type of op. The variable st points to the | |
2638 | top-most backtrack struct. To make the code clearer, within each | |
2639 | block of code we #define ST to alias the relevant union. | |
2640 | ||
2641 | Here's a concrete example of a (vastly oversimplified) IFMATCH | |
2642 | implementation: | |
2643 | ||
2644 | switch (state) { | |
2645 | .... | |
2646 | ||
2647 | #define ST st->u.ifmatch | |
2648 | ||
2649 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B | |
2650 | ST.foo = ...; // some state we wish to save | |
95b24440 | 2651 | ... |
bf1f174e DM |
2652 | // push a yes backtrack state with a resume value of |
2653 | // IFMATCH_A/IFMATCH_A_fail, then continue execution at the | |
2654 | // first node of A: | |
2655 | PUSH_YES_STATE_GOTO(IFMATCH_A, A); | |
2656 | // NOTREACHED | |
2657 | ||
2658 | case IFMATCH_A: // we have successfully executed A; now continue with B | |
2659 | next = B; | |
2660 | bar = ST.foo; // do something with the preserved value | |
2661 | break; | |
2662 | ||
2663 | case IFMATCH_A_fail: // A failed, so the assertion failed | |
2664 | ...; // do some housekeeping, then ... | |
2665 | sayNO; // propagate the failure | |
2666 | ||
2667 | #undef ST | |
95b24440 | 2668 | |
bf1f174e DM |
2669 | ... |
2670 | } | |
95b24440 | 2671 | |
bf1f174e DM |
2672 | For any old-timers reading this who are familiar with the old recursive |
2673 | approach, the code above is equivalent to: | |
95b24440 | 2674 | |
bf1f174e DM |
2675 | case IFMATCH: // we are executing the IFMATCH op, (?=A)B |
2676 | { | |
2677 | int foo = ... | |
95b24440 | 2678 | ... |
bf1f174e DM |
2679 | if (regmatch(A)) { |
2680 | next = B; | |
2681 | bar = foo; | |
2682 | break; | |
95b24440 | 2683 | } |
bf1f174e DM |
2684 | ...; // do some housekeeping, then ... |
2685 | sayNO; // propagate the failure | |
95b24440 | 2686 | } |
bf1f174e DM |
2687 | |
2688 | The topmost backtrack state, pointed to by st, is usually free. If you | |
2689 | want to claim it, populate any ST.foo fields in it with values you wish to | |
2690 | save, then do one of | |
2691 | ||
2692 | PUSH_STATE_GOTO(resume_state, node); | |
2693 | PUSH_YES_STATE_GOTO(resume_state, node); | |
2694 | ||
2695 | which sets that backtrack state's resume value to 'resume_state', pushes a | |
2696 | new free entry to the top of the backtrack stack, then goes to 'node'. | |
2697 | On backtracking, the free slot is popped, and the saved state becomes the | |
2698 | new free state. An ST.foo field in this new top state can be temporarily | |
2699 | accessed to retrieve values, but once the main loop is re-entered, it | |
2700 | becomes available for reuse. | |
2701 | ||
2702 | Note that the depth of the backtrack stack constantly increases during the | |
2703 | left-to-right execution of the pattern, rather than going up and down with | |
2704 | the pattern nesting. For example the stack is at its maximum at Z at the | |
2705 | end of the pattern, rather than at X in the following: | |
2706 | ||
2707 | /(((X)+)+)+....(Y)+....Z/ | |
2708 | ||
2709 | The only exceptions to this are lookahead/behind assertions and the cut, | |
2710 | (?>A), which pop all the backtrack states associated with A before | |
2711 | continuing. | |
2712 | ||
2713 | Bascktrack state structs are allocated in slabs of about 4K in size. | |
2714 | PL_regmatch_state and st always point to the currently active state, | |
2715 | and PL_regmatch_slab points to the slab currently containing | |
2716 | PL_regmatch_state. The first time regmatch() is called, the first slab is | |
2717 | allocated, and is never freed until interpreter destruction. When the slab | |
2718 | is full, a new one is allocated and chained to the end. At exit from | |
2719 | regmatch(), slabs allocated since entry are freed. | |
2720 | ||
2721 | */ | |
95b24440 | 2722 | |
40a82448 | 2723 | |
5bc10b2c | 2724 | #define DEBUG_STATE_pp(pp) \ |
265c4333 | 2725 | DEBUG_STATE_r({ \ |
f2ed9b32 | 2726 | DUMP_EXEC_POS(locinput, scan, utf8_target); \ |
5bc10b2c | 2727 | PerlIO_printf(Perl_debug_log, \ |
5d458dd8 | 2728 | " %*s"pp" %s%s%s%s%s\n", \ |
5bc10b2c | 2729 | depth*2, "", \ |
13d6edb4 | 2730 | PL_reg_name[st->resume_state], \ |
5d458dd8 YO |
2731 | ((st==yes_state||st==mark_state) ? "[" : ""), \ |
2732 | ((st==yes_state) ? "Y" : ""), \ | |
2733 | ((st==mark_state) ? "M" : ""), \ | |
2734 | ((st==yes_state||st==mark_state) ? "]" : "") \ | |
2735 | ); \ | |
265c4333 | 2736 | }); |
5bc10b2c | 2737 | |
40a82448 | 2738 | |
3dab1dad | 2739 | #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1) |
95b24440 | 2740 | |
3df15adc | 2741 | #ifdef DEBUGGING |
5bc10b2c | 2742 | |
ab3bbdeb | 2743 | STATIC void |
f2ed9b32 | 2744 | S_debug_start_match(pTHX_ const REGEXP *prog, const bool utf8_target, |
ab3bbdeb YO |
2745 | const char *start, const char *end, const char *blurb) |
2746 | { | |
efd26800 | 2747 | const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; |
7918f24d NC |
2748 | |
2749 | PERL_ARGS_ASSERT_DEBUG_START_MATCH; | |
2750 | ||
ab3bbdeb YO |
2751 | if (!PL_colorset) |
2752 | reginitcolors(); | |
2753 | { | |
2754 | RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), | |
d2c6dc5e | 2755 | RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); |
ab3bbdeb | 2756 | |
f2ed9b32 | 2757 | RE_PV_QUOTED_DECL(s1, utf8_target, PERL_DEBUG_PAD_ZERO(1), |
ab3bbdeb YO |
2758 | start, end - start, 60); |
2759 | ||
2760 | PerlIO_printf(Perl_debug_log, | |
2761 | "%s%s REx%s %s against %s\n", | |
2762 | PL_colors[4], blurb, PL_colors[5], s0, s1); | |
2763 | ||
f2ed9b32 | 2764 | if (utf8_target||utf8_pat) |
1de06328 YO |
2765 | PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n", |
2766 | utf8_pat ? "pattern" : "", | |
f2ed9b32 KW |
2767 | utf8_pat && utf8_target ? " and " : "", |
2768 | utf8_target ? "string" : "" | |
ab3bbdeb YO |
2769 | ); |
2770 | } | |
2771 | } | |
3df15adc YO |
2772 | |
2773 | STATIC void | |
786e8c11 YO |
2774 | S_dump_exec_pos(pTHX_ const char *locinput, |
2775 | const regnode *scan, | |
2776 | const char *loc_regeol, | |
2777 | const char *loc_bostr, | |
2778 | const char *loc_reg_starttry, | |
f2ed9b32 | 2779 | const bool utf8_target) |
07be1b83 | 2780 | { |
786e8c11 | 2781 | const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4]; |
07be1b83 | 2782 | const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ |
786e8c11 | 2783 | int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput); |
07be1b83 YO |
2784 | /* The part of the string before starttry has one color |
2785 | (pref0_len chars), between starttry and current | |
2786 | position another one (pref_len - pref0_len chars), | |
2787 | after the current position the third one. | |
2788 | We assume that pref0_len <= pref_len, otherwise we | |
2789 | decrease pref0_len. */ | |
786e8c11 YO |
2790 | int pref_len = (locinput - loc_bostr) > (5 + taill) - l |
2791 | ? (5 + taill) - l : locinput - loc_bostr; | |
07be1b83 YO |
2792 | int pref0_len; |
2793 | ||
7918f24d NC |
2794 | PERL_ARGS_ASSERT_DUMP_EXEC_POS; |
2795 | ||
f2ed9b32 | 2796 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) |
07be1b83 | 2797 | pref_len++; |
786e8c11 YO |
2798 | pref0_len = pref_len - (locinput - loc_reg_starttry); |
2799 | if (l + pref_len < (5 + taill) && l < loc_regeol - locinput) | |
2800 | l = ( loc_regeol - locinput > (5 + taill) - pref_len | |
2801 | ? (5 + taill) - pref_len : loc_regeol - locinput); | |
f2ed9b32 | 2802 | while (utf8_target && UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) |
07be1b83 YO |
2803 | l--; |
2804 | if (pref0_len < 0) | |
2805 | pref0_len = 0; | |
2806 | if (pref0_len > pref_len) | |
2807 | pref0_len = pref_len; | |
2808 | { | |
f2ed9b32 | 2809 | const int is_uni = (utf8_target && OP(scan) != CANY) ? 1 : 0; |
0df25f3d | 2810 | |
ab3bbdeb | 2811 | RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0), |
1de06328 | 2812 | (locinput - pref_len),pref0_len, 60, 4, 5); |
0df25f3d | 2813 | |
ab3bbdeb | 2814 | RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1), |
3df15adc | 2815 | (locinput - pref_len + pref0_len), |
1de06328 | 2816 | pref_len - pref0_len, 60, 2, 3); |
0df25f3d | 2817 | |
ab3bbdeb | 2818 | RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2), |
1de06328 | 2819 | locinput, loc_regeol - locinput, 10, 0, 1); |
0df25f3d | 2820 | |
1de06328 | 2821 | const STRLEN tlen=len0+len1+len2; |
3df15adc | 2822 | PerlIO_printf(Perl_debug_log, |
ab3bbdeb | 2823 | "%4"IVdf" <%.*s%.*s%s%.*s>%*s|", |
786e8c11 | 2824 | (IV)(locinput - loc_bostr), |
07be1b83 | 2825 | len0, s0, |
07be1b83 | 2826 | len1, s1, |
07be1b83 | 2827 | (docolor ? "" : "> <"), |
07be1b83 | 2828 | len2, s2, |
f9f4320a | 2829 | (int)(tlen > 19 ? 0 : 19 - tlen), |
07be1b83 YO |
2830 | ""); |
2831 | } | |
2832 | } | |
3df15adc | 2833 | |
07be1b83 YO |
2834 | #endif |
2835 | ||
0a4db386 YO |
2836 | /* reg_check_named_buff_matched() |
2837 | * Checks to see if a named buffer has matched. The data array of | |
2838 | * buffer numbers corresponding to the buffer is expected to reside | |
2839 | * in the regexp->data->data array in the slot stored in the ARG() of | |
2840 | * node involved. Note that this routine doesn't actually care about the | |
2841 | * name, that information is not preserved from compilation to execution. | |
2842 | * Returns the index of the leftmost defined buffer with the given name | |
2843 | * or 0 if non of the buffers matched. | |
2844 | */ | |
2845 | STATIC I32 | |
7918f24d NC |
2846 | S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) |
2847 | { | |
0a4db386 | 2848 | I32 n; |
f8fc2ecf | 2849 | RXi_GET_DECL(rex,rexi); |
ad64d0ec | 2850 | SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
0a4db386 | 2851 | I32 *nums=(I32*)SvPVX(sv_dat); |
7918f24d NC |
2852 | |
2853 | PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; | |
2854 | ||
0a4db386 YO |
2855 | for ( n=0; n<SvIVX(sv_dat); n++ ) { |
2856 | if ((I32)*PL_reglastparen >= nums[n] && | |
f0ab9afb | 2857 | PL_regoffs[nums[n]].end != -1) |
0a4db386 YO |
2858 | { |
2859 | return nums[n]; | |
2860 | } | |
2861 | } | |
2862 | return 0; | |
2863 | } | |
2864 | ||
2f554ef7 DM |
2865 | |
2866 | /* free all slabs above current one - called during LEAVE_SCOPE */ | |
2867 | ||
2868 | STATIC void | |
2869 | S_clear_backtrack_stack(pTHX_ void *p) | |
2870 | { | |
2871 | regmatch_slab *s = PL_regmatch_slab->next; | |
2872 | PERL_UNUSED_ARG(p); | |
2873 | ||
2874 | if (!s) | |
2875 | return; | |
2876 | PL_regmatch_slab->next = NULL; | |
2877 | while (s) { | |
2878 | regmatch_slab * const osl = s; | |
2879 | s = s->next; | |
2880 | Safefree(osl); | |
2881 | } | |
2882 | } | |
2883 | ||
2884 | ||
28d8d7f4 YO |
2885 | #define SETREX(Re1,Re2) \ |
2886 | if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \ | |
2887 | Re1 = (Re2) | |
2888 | ||
d6a28714 | 2889 | STATIC I32 /* 0 failure, 1 success */ |
24b23f37 | 2890 | S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) |
d6a28714 | 2891 | { |
a35a87e7 | 2892 | #if PERL_VERSION < 9 && !defined(PERL_CORE) |
54df2634 NC |
2893 | dMY_CXT; |
2894 | #endif | |
27da23d5 | 2895 | dVAR; |
f2ed9b32 | 2896 | register const bool utf8_target = PL_reg_match_utf8; |
4ad0818d | 2897 | const U32 uniflags = UTF8_ALLOW_DEFAULT; |
288b8c02 NC |
2898 | REGEXP *rex_sv = reginfo->prog; |
2899 | regexp *rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 2900 | RXi_GET_DECL(rex,rexi); |
2f554ef7 | 2901 | I32 oldsave; |
5d9a96ca DM |
2902 | /* the current state. This is a cached copy of PL_regmatch_state */ |
2903 | register regmatch_state *st; | |
5d9a96ca DM |
2904 | /* cache heavy used fields of st in registers */ |
2905 | register regnode *scan; | |
2906 | register regnode *next; | |
438e9bae | 2907 | register U32 n = 0; /* general value; init to avoid compiler warning */ |
24d3c4a9 | 2908 | register I32 ln = 0; /* len or last; init to avoid compiler warning */ |
5d9a96ca | 2909 | register char *locinput = PL_reginput; |
5d9a96ca | 2910 | register I32 nextchr; /* is always set to UCHARAT(locinput) */ |
24d3c4a9 | 2911 | |
b69b0499 | 2912 | bool result = 0; /* return value of S_regmatch */ |
24d3c4a9 | 2913 | int depth = 0; /* depth of backtrack stack */ |
4b196cd4 YO |
2914 | U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */ |
2915 | const U32 max_nochange_depth = | |
2916 | (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? | |
2917 | 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; | |
77cb431f DM |
2918 | regmatch_state *yes_state = NULL; /* state to pop to on success of |
2919 | subpattern */ | |
e2e6a0f1 YO |
2920 | /* mark_state piggy backs on the yes_state logic so that when we unwind |
2921 | the stack on success we can update the mark_state as we go */ | |
2922 | regmatch_state *mark_state = NULL; /* last mark state we have seen */ | |
faec1544 | 2923 | regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ |
b8591aee | 2924 | struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ |
40a82448 | 2925 | U32 state_num; |
5d458dd8 YO |
2926 | bool no_final = 0; /* prevent failure from backtracking? */ |
2927 | bool do_cutgroup = 0; /* no_final only until next branch/trie entry */ | |
e2e6a0f1 | 2928 | char *startpoint = PL_reginput; |
5d458dd8 YO |
2929 | SV *popmark = NULL; /* are we looking for a mark? */ |
2930 | SV *sv_commit = NULL; /* last mark name seen in failure */ | |
2931 | SV *sv_yes_mark = NULL; /* last mark name we have seen | |
2932 | during a successfull match */ | |
2933 | U32 lastopen = 0; /* last open we saw */ | |
2934 | bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; | |
19b95bf0 | 2935 | SV* const oreplsv = GvSV(PL_replgv); |
24d3c4a9 DM |
2936 | /* these three flags are set by various ops to signal information to |
2937 | * the very next op. They have a useful lifetime of exactly one loop | |
2938 | * iteration, and are not preserved or restored by state pushes/pops | |
2939 | */ | |
2940 | bool sw = 0; /* the condition value in (?(cond)a|b) */ | |
2941 | bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */ | |
2942 | int logical = 0; /* the following EVAL is: | |
2943 | 0: (?{...}) | |
2944 | 1: (?(?{...})X|Y) | |
2945 | 2: (??{...}) | |
2946 | or the following IFMATCH/UNLESSM is: | |
2947 | false: plain (?=foo) | |
2948 | true: used as a condition: (?(?=foo)) | |
2949 | */ | |
95b24440 | 2950 | #ifdef DEBUGGING |
e68ec53f | 2951 | GET_RE_DEBUG_FLAGS_DECL; |
d6a28714 JH |
2952 | #endif |
2953 | ||
7918f24d NC |
2954 | PERL_ARGS_ASSERT_REGMATCH; |
2955 | ||
3b57cd43 | 2956 | DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ |
24b23f37 | 2957 | PerlIO_printf(Perl_debug_log,"regmatch start\n"); |
3b57cd43 | 2958 | })); |
5d9a96ca DM |
2959 | /* on first ever call to regmatch, allocate first slab */ |
2960 | if (!PL_regmatch_slab) { | |
2961 | Newx(PL_regmatch_slab, 1, regmatch_slab); | |
2962 | PL_regmatch_slab->prev = NULL; | |
2963 | PL_regmatch_slab->next = NULL; | |
86545054 | 2964 | PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); |
5d9a96ca DM |
2965 | } |
2966 | ||
2f554ef7 DM |
2967 | oldsave = PL_savestack_ix; |
2968 | SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL); | |
2969 | SAVEVPTR(PL_regmatch_slab); | |
2970 | SAVEVPTR(PL_regmatch_state); | |
5d9a96ca DM |
2971 | |
2972 | /* grab next free state slot */ | |
2973 | st = ++PL_regmatch_state; | |
86545054 | 2974 | if (st > SLAB_LAST(PL_regmatch_slab)) |
5d9a96ca DM |
2975 | st = PL_regmatch_state = S_push_slab(aTHX); |
2976 | ||
d6a28714 JH |
2977 | /* Note that nextchr is a byte even in UTF */ |
2978 | nextchr = UCHARAT(locinput); | |
2979 | scan = prog; | |
2980 | while (scan != NULL) { | |
8ba1375e | 2981 | |
a3621e74 | 2982 | DEBUG_EXECUTE_r( { |
6136c704 | 2983 | SV * const prop = sv_newmortal(); |
1de06328 | 2984 | regnode *rnext=regnext(scan); |
f2ed9b32 | 2985 | DUMP_EXEC_POS( locinput, scan, utf8_target ); |
32fc9b6a | 2986 | regprop(rex, prop, scan); |
07be1b83 YO |
2987 | |
2988 | PerlIO_printf(Perl_debug_log, | |
2989 | "%3"IVdf":%*s%s(%"IVdf")\n", | |
f8fc2ecf | 2990 | (IV)(scan - rexi->program), depth*2, "", |
07be1b83 | 2991 | SvPVX_const(prop), |
1de06328 | 2992 | (PL_regkind[OP(scan)] == END || !rnext) ? |
f8fc2ecf | 2993 | 0 : (IV)(rnext - rexi->program)); |
2a782b5b | 2994 | }); |
d6a28714 JH |
2995 | |
2996 | next = scan + NEXT_OFF(scan); | |
2997 | if (next == scan) | |
2998 | next = NULL; | |
40a82448 | 2999 | state_num = OP(scan); |
d6a28714 | 3000 | |
40a82448 | 3001 | reenter_switch: |
34a81e2b B |
3002 | |
3003 | assert(PL_reglastparen == &rex->lastparen); | |
3004 | assert(PL_reglastcloseparen == &rex->lastcloseparen); | |
3005 | assert(PL_regoffs == rex->offs); | |
3006 | ||
40a82448 | 3007 | switch (state_num) { |
d6a28714 | 3008 | case BOL: |
7fba1cd6 | 3009 | if (locinput == PL_bostr) |
d6a28714 | 3010 | { |
3b0527fe | 3011 | /* reginfo->till = reginfo->bol; */ |
b8c5462f JH |
3012 | break; |
3013 | } | |
d6a28714 JH |
3014 | sayNO; |
3015 | case MBOL: | |
12d33761 HS |
3016 | if (locinput == PL_bostr || |
3017 | ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n')) | |
d6a28714 | 3018 | { |
b8c5462f JH |
3019 | break; |
3020 | } | |
d6a28714 JH |
3021 | sayNO; |
3022 | case SBOL: | |
c2a73568 | 3023 | if (locinput == PL_bostr) |
b8c5462f | 3024 | break; |
d6a28714 JH |
3025 | sayNO; |
3026 | case GPOS: | |
3b0527fe | 3027 | if (locinput == reginfo->ganch) |
d6a28714 JH |
3028 | break; |
3029 | sayNO; | |
ee9b8eae YO |
3030 | |
3031 | case KEEPS: | |
3032 | /* update the startpoint */ | |
f0ab9afb | 3033 | st->u.keeper.val = PL_regoffs[0].start; |
ee9b8eae | 3034 | PL_reginput = locinput; |
f0ab9afb | 3035 | PL_regoffs[0].start = locinput - PL_bostr; |
ee9b8eae YO |
3036 | PUSH_STATE_GOTO(KEEPS_next, next); |
3037 | /*NOT-REACHED*/ | |
3038 | case KEEPS_next_fail: | |
3039 | /* rollback the start point change */ | |
f0ab9afb | 3040 | PL_regoffs[0].start = st->u.keeper.val; |
ee9b8eae YO |
3041 | sayNO_SILENT; |
3042 | /*NOT-REACHED*/ | |
d6a28714 | 3043 | case EOL: |
d6a28714 JH |
3044 | goto seol; |
3045 | case MEOL: | |
d6a28714 | 3046 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') |
b8c5462f | 3047 | sayNO; |
b8c5462f | 3048 | break; |
d6a28714 JH |
3049 | case SEOL: |
3050 | seol: | |
3051 | if ((nextchr || locinput < PL_regeol) && nextchr != '\n') | |
b8c5462f | 3052 | sayNO; |
d6a28714 | 3053 | if (PL_regeol - locinput > 1) |
b8c5462f | 3054 | sayNO; |
b8c5462f | 3055 | break; |
d6a28714 JH |
3056 | case EOS: |
3057 | if (PL_regeol != locinput) | |
b8c5462f | 3058 | sayNO; |
d6a28714 | 3059 | break; |
ffc61ed2 | 3060 | case SANY: |
d6a28714 | 3061 | if (!nextchr && locinput >= PL_regeol) |
4633a7c4 | 3062 | sayNO; |
f2ed9b32 | 3063 | if (utf8_target) { |
f33976b4 DB |
3064 | locinput += PL_utf8skip[nextchr]; |
3065 | if (locinput > PL_regeol) | |
3066 | sayNO; | |
3067 | nextchr = UCHARAT(locinput); | |
3068 | } | |
3069 | else | |
3070 | nextchr = UCHARAT(++locinput); | |
3071 | break; | |
3072 | case CANY: | |
3073 | if (!nextchr && locinput >= PL_regeol) | |
3074 | sayNO; | |
b8c5462f | 3075 | nextchr = UCHARAT(++locinput); |
a0d0e21e | 3076 | break; |
ffc61ed2 | 3077 | case REG_ANY: |
1aa99e6b IH |
3078 | if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n') |
3079 | sayNO; | |
f2ed9b32 | 3080 | if (utf8_target) { |
b8c5462f | 3081 | locinput += PL_utf8skip[nextchr]; |
d6a28714 JH |
3082 | if (locinput > PL_regeol) |
3083 | sayNO; | |
a0ed51b3 | 3084 | nextchr = UCHARAT(locinput); |
a0ed51b3 | 3085 | } |
1aa99e6b IH |
3086 | else |
3087 | nextchr = UCHARAT(++locinput); | |
a0ed51b3 | 3088 | break; |
166ba7cd DM |
3089 | |
3090 | #undef ST | |
3091 | #define ST st->u.trie | |
786e8c11 YO |
3092 | case TRIEC: |
3093 | /* In this case the charclass data is available inline so | |
3094 | we can fail fast without a lot of extra overhead. | |
3095 | */ | |
f2ed9b32 | 3096 | if (scan->flags == EXACT || !utf8_target) { |
786e8c11 YO |
3097 | if(!ANYOF_BITMAP_TEST(scan, *locinput)) { |
3098 | DEBUG_EXECUTE_r( | |
3099 | PerlIO_printf(Perl_debug_log, | |
3100 | "%*s %sfailed to match trie start class...%s\n", | |
5bc10b2c | 3101 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
786e8c11 YO |
3102 | ); |
3103 | sayNO_SILENT; | |
3104 | /* NOTREACHED */ | |
3105 | } | |
3106 | } | |
3107 | /* FALL THROUGH */ | |
5b47454d | 3108 | case TRIE: |
2e64971a DM |
3109 | /* the basic plan of execution of the trie is: |
3110 | * At the beginning, run though all the states, and | |
3111 | * find the longest-matching word. Also remember the position | |
3112 | * of the shortest matching word. For example, this pattern: | |
3113 | * 1 2 3 4 5 | |
3114 | * ab|a|x|abcd|abc | |
3115 | * when matched against the string "abcde", will generate | |
3116 | * accept states for all words except 3, with the longest | |
3117 | * matching word being 4, and the shortest being 1 (with | |
3118 | * the position being after char 1 of the string). | |
3119 | * | |
3120 | * Then for each matching word, in word order (i.e. 1,2,4,5), | |
3121 | * we run the remainder of the pattern; on each try setting | |
3122 | * the current position to the character following the word, | |
3123 | * returning to try the next word on failure. | |
3124 | * | |
3125 | * We avoid having to build a list of words at runtime by | |
3126 | * using a compile-time structure, wordinfo[].prev, which | |
3127 | * gives, for each word, the previous accepting word (if any). | |
3128 | * In the case above it would contain the mappings 1->2, 2->0, | |
3129 | * 3->0, 4->5, 5->1. We can use this table to generate, from | |
3130 | * the longest word (4 above), a list of all words, by | |
3131 | * following the list of prev pointers; this gives us the | |
3132 | * unordered list 4,5,1,2. Then given the current word we have | |
3133 | * just tried, we can go through the list and find the | |
3134 | * next-biggest word to try (so if we just failed on word 2, | |
3135 | * the next in the list is 4). | |
3136 | * | |
3137 | * Since at runtime we don't record the matching position in | |
3138 | * the string for each word, we have to work that out for | |
3139 | * each word we're about to process. The wordinfo table holds | |
3140 | * the character length of each word; given that we recorded | |
3141 | * at the start: the position of the shortest word and its | |
3142 | * length in chars, we just need to move the pointer the | |
3143 | * difference between the two char lengths. Depending on | |
3144 | * Unicode status and folding, that's cheap or expensive. | |
3145 | * | |
3146 | * This algorithm is optimised for the case where are only a | |
3147 | * small number of accept states, i.e. 0,1, or maybe 2. | |
3148 | * With lots of accepts states, and having to try all of them, | |
3149 | * it becomes quadratic on number of accept states to find all | |
3150 | * the next words. | |
3151 | */ | |
3152 | ||
3dab1dad | 3153 | { |
07be1b83 | 3154 | /* what type of TRIE am I? (utf8 makes this contextual) */ |
a0a388a1 | 3155 | DECL_TRIE_TYPE(scan); |
3dab1dad YO |
3156 | |
3157 | /* what trie are we using right now */ | |
be8e71aa | 3158 | reg_trie_data * const trie |
f8fc2ecf | 3159 | = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; |
85fbaab2 | 3160 | HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); |
3dab1dad | 3161 | U32 state = trie->startstate; |
166ba7cd | 3162 | |
3dab1dad YO |
3163 | if (trie->bitmap && trie_type != trie_utf8_fold && |
3164 | !TRIE_BITMAP_TEST(trie,*locinput) | |
3165 | ) { | |
3166 | if (trie->states[ state ].wordnum) { | |
3167 | DEBUG_EXECUTE_r( | |
3168 | PerlIO_printf(Perl_debug_log, | |
3169 | "%*s %smatched empty string...%s\n", | |
5bc10b2c | 3170 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad YO |
3171 | ); |
3172 | break; | |
3173 | } else { | |
3174 | DEBUG_EXECUTE_r( | |
3175 | PerlIO_printf(Perl_debug_log, | |
786e8c11 | 3176 | "%*s %sfailed to match trie start class...%s\n", |
5bc10b2c | 3177 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) |
3dab1dad YO |
3178 | ); |
3179 | sayNO_SILENT; | |
3180 | } | |
3181 | } | |
166ba7cd | 3182 | |
786e8c11 YO |
3183 | { |
3184 | U8 *uc = ( U8* )locinput; | |
3185 | ||
3186 | STRLEN len = 0; | |
3187 | STRLEN foldlen = 0; | |
3188 | U8 *uscan = (U8*)NULL; | |
786e8c11 | 3189 | U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ]; |
2e64971a DM |
3190 | U32 charcount = 0; /* how many input chars we have matched */ |
3191 | U32 accepted = 0; /* have we seen any accepting states? */ | |
786e8c11 | 3192 | |
786e8c11 YO |
3193 | ST.B = next; |
3194 | ST.jump = trie->jump; | |
786e8c11 | 3195 | ST.me = scan; |
2e64971a DM |
3196 | ST.firstpos = NULL; |
3197 | ST.longfold = FALSE; /* char longer if folded => it's harder */ | |
3198 | ST.nextword = 0; | |
3199 | ||
3200 | /* fully traverse the TRIE; note the position of the | |
3201 | shortest accept state and the wordnum of the longest | |
3202 | accept state */ | |
07be1b83 | 3203 | |
a3621e74 | 3204 | while ( state && uc <= (U8*)PL_regeol ) { |
786e8c11 | 3205 | U32 base = trie->states[ state ].trans.base; |
f9f4320a | 3206 | UV uvc = 0; |
acb909b4 | 3207 | U16 charid = 0; |
2e64971a DM |
3208 | U16 wordnum; |
3209 | wordnum = trie->states[ state ].wordnum; | |
3210 | ||
3211 | if (wordnum) { /* it's an accept state */ | |
3212 | if (!accepted) { | |
3213 | accepted = 1; | |
3214 | /* record first match position */ | |
3215 | if (ST.longfold) { | |
3216 | ST.firstpos = (U8*)locinput; | |
3217 | ST.firstchars = 0; | |
5b47454d | 3218 | } |
2e64971a DM |
3219 | else { |
3220 | ST.firstpos = uc; | |
3221 | ST.firstchars = charcount; | |
3222 | } | |
3223 | } | |
3224 | if (!ST.nextword || wordnum < ST.nextword) | |
3225 | ST.nextword = wordnum; | |
3226 | ST.topword = wordnum; | |
786e8c11 | 3227 | } |
a3621e74 | 3228 | |
07be1b83 | 3229 | DEBUG_TRIE_EXECUTE_r({ |
f2ed9b32 | 3230 | DUMP_EXEC_POS( (char *)uc, scan, utf8_target ); |
a3621e74 | 3231 | PerlIO_printf( Perl_debug_log, |
2e64971a | 3232 | "%*s %sState: %4"UVxf" Accepted: %c ", |
5bc10b2c | 3233 | 2+depth * 2, "", PL_colors[4], |
2e64971a | 3234 | (UV)state, (accepted ? 'Y' : 'N')); |
07be1b83 | 3235 | }); |
a3621e74 | 3236 | |
2e64971a | 3237 | /* read a char and goto next state */ |
a3621e74 | 3238 | if ( base ) { |
6dd2be57 | 3239 | I32 offset; |
55eed653 NC |
3240 | REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, |
3241 | uscan, len, uvc, charid, foldlen, | |
3242 | foldbuf, uniflags); | |
2e64971a DM |
3243 | charcount++; |
3244 | if (foldlen>0) | |
3245 | ST.longfold = TRUE; | |
5b47454d | 3246 | if (charid && |
6dd2be57 DM |
3247 | ( ((offset = |
3248 | base + charid - 1 - trie->uniquecharcount)) >= 0) | |
3249 | ||
3250 | && ((U32)offset < trie->lasttrans) | |
3251 | && trie->trans[offset].check == state) | |
5b47454d | 3252 | { |
6dd2be57 | 3253 | state = trie->trans[offset].next; |
5b47454d DM |
3254 | } |
3255 | else { | |
3256 | state = 0; | |
3257 | } | |
3258 | uc += len; | |
3259 | ||
3260 | } | |
3261 | else { | |
a3621e74 YO |
3262 | state = 0; |
3263 | } | |
3264 | DEBUG_TRIE_EXECUTE_r( | |
e4584336 | 3265 | PerlIO_printf( Perl_debug_log, |
786e8c11 | 3266 | "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n", |
e4584336 | 3267 | charid, uvc, (UV)state, PL_colors[5] ); |
a3621e74 YO |
3268 | ); |
3269 | } | |
2e64971a | 3270 | if (!accepted) |
a3621e74 | 3271 | sayNO; |
a3621e74 | 3272 | |
2e64971a DM |
3273 | /* calculate total number of accept states */ |
3274 | { | |
3275 | U16 w = ST.topword; | |
3276 | accepted = 0; | |
3277 | while (w) { | |
3278 | w = trie->wordinfo[w].prev; | |
3279 | accepted++; | |
3280 | } | |
3281 | ST.accepted = accepted; | |
3282 | } | |
3283 | ||
166ba7cd DM |
3284 | DEBUG_EXECUTE_r( |
3285 | PerlIO_printf( Perl_debug_log, | |
3286 | "%*s %sgot %"IVdf" possible matches%s\n", | |
5bc10b2c | 3287 | REPORT_CODE_OFF + depth * 2, "", |
166ba7cd DM |
3288 | PL_colors[4], (IV)ST.accepted, PL_colors[5] ); |
3289 | ); | |
2e64971a | 3290 | goto trie_first_try; /* jump into the fail handler */ |
786e8c11 | 3291 | }} |
fae667d5 | 3292 | /* NOTREACHED */ |
2e64971a DM |
3293 | |
3294 | case TRIE_next_fail: /* we failed - try next alternative */ | |
fae667d5 YO |
3295 | if ( ST.jump) { |
3296 | REGCP_UNWIND(ST.cp); | |
3297 | for (n = *PL_reglastparen; n > ST.lastparen; n--) | |
f0ab9afb | 3298 | PL_regoffs[n].end = -1; |
fae667d5 YO |
3299 | *PL_reglastparen = n; |
3300 | } | |
2e64971a DM |
3301 | if (!--ST.accepted) { |
3302 | DEBUG_EXECUTE_r({ | |
3303 | PerlIO_printf( Perl_debug_log, | |
3304 | "%*s %sTRIE failed...%s\n", | |
3305 | REPORT_CODE_OFF+depth*2, "", | |
3306 | PL_colors[4], | |
3307 | PL_colors[5] ); | |
3308 | }); | |
3309 | sayNO_SILENT; | |
3310 | } | |
3311 | { | |
3312 | /* Find next-highest word to process. Note that this code | |
3313 | * is O(N^2) per trie run (O(N) per branch), so keep tight */ | |
d9a396a3 DM |
3314 | register U16 min = 0; |
3315 | register U16 word; | |
2e64971a DM |
3316 | register U16 const nextword = ST.nextword; |
3317 | register reg_trie_wordinfo * const wordinfo | |
3318 | = ((reg_trie_data*)rexi->data->data[ARG(ST.me)])->wordinfo; | |
3319 | for (word=ST.topword; word; word=wordinfo[word].prev) { | |
3320 | if (word > nextword && (!min || word < min)) | |
3321 | min = word; | |
3322 | } | |
3323 | ST.nextword = min; | |
3324 | } | |
3325 | ||
fae667d5 | 3326 | trie_first_try: |
5d458dd8 YO |
3327 | if (do_cutgroup) { |
3328 | do_cutgroup = 0; | |
3329 | no_final = 0; | |
3330 | } | |
fae667d5 YO |
3331 | |
3332 | if ( ST.jump) { | |
3333 | ST.lastparen = *PL_reglastparen; | |
3334 | REGCP_SET(ST.cp); | |
2e64971a | 3335 | } |
a3621e74 | 3336 | |
2e64971a | 3337 | /* find start char of end of current word */ |
166ba7cd | 3338 | { |
2e64971a DM |
3339 | U32 chars; /* how many chars to skip */ |
3340 | U8 *uc = ST.firstpos; | |
3341 | reg_trie_data * const trie | |
3342 | = (reg_trie_data*)rexi->data->data[ARG(ST.me)]; | |
3343 | ||
3344 | assert((trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3345 | >= ST.firstchars); | |
3346 | chars = (trie->wordinfo[ST.nextword].len - trie->prefixlen) | |
3347 | - ST.firstchars; | |
3348 | ||
3349 | if (ST.longfold) { | |
3350 | /* the hard option - fold each char in turn and find | |
3351 | * its folded length (which may be different */ | |
3352 | U8 foldbuf[UTF8_MAXBYTES_CASE + 1]; | |
3353 | STRLEN foldlen; | |
3354 | STRLEN len; | |
d9a396a3 | 3355 | UV uvc; |
2e64971a DM |
3356 | U8 *uscan; |
3357 | ||
3358 | while (chars) { | |
f2ed9b32 | 3359 | if (utf8_target) { |
2e64971a DM |
3360 | uvc = utf8n_to_uvuni((U8*)uc, UTF8_MAXLEN, &len, |
3361 | uniflags); | |
3362 | uc += len; | |
3363 | } | |
3364 | else { | |
3365 | uvc = *uc; | |
3366 | uc++; | |
3367 | } | |
3368 | uvc = to_uni_fold(uvc, foldbuf, &foldlen); | |
3369 | uscan = foldbuf; | |
3370 | while (foldlen) { | |
3371 | if (!--chars) | |
3372 | break; | |
3373 | uvc = utf8n_to_uvuni(uscan, UTF8_MAXLEN, &len, | |
3374 | uniflags); | |
3375 | uscan += len; | |
3376 | foldlen -= len; | |
3377 | } | |
3378 | } | |
a3621e74 | 3379 | } |
2e64971a | 3380 | else { |
f2ed9b32 | 3381 | if (utf8_target) |
2e64971a DM |
3382 | while (chars--) |
3383 | uc += UTF8SKIP(uc); | |
3384 | else | |
3385 | uc += chars; | |
3386 | } | |
3387 | PL_reginput = (char *)uc; | |
3388 | } | |
166ba7cd | 3389 | |
2e64971a DM |
3390 | scan = (ST.jump && ST.jump[ST.nextword]) |
3391 | ? ST.me + ST.jump[ST.nextword] | |
3392 | : ST.B; | |
166ba7cd | 3393 | |
2e64971a DM |
3394 | DEBUG_EXECUTE_r({ |
3395 | PerlIO_printf( Perl_debug_log, | |
3396 | "%*s %sTRIE matched word #%d, continuing%s\n", | |
3397 | REPORT_CODE_OFF+depth*2, "", | |
3398 | PL_colors[4], | |
3399 | ST.nextword, | |
3400 | PL_colors[5] | |
3401 | ); | |
3402 | }); | |
3403 | ||
3404 | if (ST.accepted > 1 || has_cutgroup) { | |
3405 | PUSH_STATE_GOTO(TRIE_next, scan); | |
3406 | /* NOTREACHED */ | |
166ba7cd | 3407 | } |
2e64971a DM |
3408 | /* only one choice left - just continue */ |
3409 | DEBUG_EXECUTE_r({ | |
3410 | AV *const trie_words | |
3411 | = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]); | |
3412 | SV ** const tmp = av_fetch( trie_words, | |
3413 | ST.nextword-1, 0 ); | |
3414 | SV *sv= tmp ? sv_newmortal() : NULL; | |
3415 | ||
3416 | PerlIO_printf( Perl_debug_log, | |
3417 | "%*s %sonly one match left, short-circuiting: #%d <%s>%s\n", | |
3418 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], | |
3419 | ST.nextword, | |
3420 | tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0, | |
3421 | PL_colors[0], PL_colors[1], | |
3422 | (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) | |
3423 | ) | |
3424 | : "not compiled under -Dr", | |
3425 | PL_colors[5] ); | |
3426 | }); | |
3427 | ||
3428 | locinput = PL_reginput; | |
3429 | nextchr = UCHARAT(locinput); | |
3430 | continue; /* execute rest of RE */ | |
166ba7cd | 3431 | /* NOTREACHED */ |
166ba7cd DM |
3432 | #undef ST |
3433 | ||
95b24440 DM |
3434 | case EXACT: { |
3435 | char *s = STRING(scan); | |
24d3c4a9 | 3436 | ln = STR_LEN(scan); |
f2ed9b32 | 3437 | if (utf8_target != UTF_PATTERN) { |
bc517b45 | 3438 | /* The target and the pattern have differing utf8ness. */ |
1aa99e6b | 3439 | char *l = locinput; |
24d3c4a9 | 3440 | const char * const e = s + ln; |
a72c7584 | 3441 | |
f2ed9b32 | 3442 | if (utf8_target) { |
5ff6fc6d | 3443 | /* The target is utf8, the pattern is not utf8. */ |
1aa99e6b | 3444 | while (s < e) { |
a3b680e6 | 3445 | STRLEN ulen; |
1aa99e6b | 3446 | if (l >= PL_regeol) |
5ff6fc6d JH |
3447 | sayNO; |
3448 | if (NATIVE_TO_UNI(*(U8*)s) != | |
89ebb4a3 | 3449 | utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen, |
041457d9 | 3450 | uniflags)) |
5ff6fc6d | 3451 | sayNO; |
bc517b45 | 3452 | l += ulen; |
5ff6fc6d | 3453 | s ++; |
1aa99e6b | 3454 | } |
5ff6fc6d JH |
3455 | } |
3456 | else { | |
3457 | /* The target is not utf8, the pattern is utf8. */ | |
1aa99e6b | 3458 | while (s < e) { |
a3b680e6 | 3459 | STRLEN ulen; |
1aa99e6b IH |
3460 | if (l >= PL_regeol) |
3461 | sayNO; | |
5ff6fc6d | 3462 | if (NATIVE_TO_UNI(*((U8*)l)) != |
89ebb4a3 | 3463 | utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen, |
041457d9 | 3464 | uniflags)) |
1aa99e6b | 3465 | sayNO; |
bc517b45 | 3466 | s += ulen; |
a72c7584 | 3467 | l ++; |
1aa99e6b | 3468 | } |
5ff6fc6d | 3469 | } |
1aa99e6b IH |
3470 | locinput = l; |
3471 | nextchr = UCHARAT(locinput); | |
3472 | break; | |
3473 | } | |
bc517b45 | 3474 | /* The target and the pattern have the same utf8ness. */ |
d6a28714 JH |
3475 | /* Inline the first character, for speed. */ |
3476 | if (UCHARAT(s) != nextchr) | |
3477 | sayNO; | |
24d3c4a9 | 3478 | if (PL_regeol - locinput < ln) |
d6a28714 | 3479 | sayNO; |
24d3c4a9 | 3480 | if (ln > 1 && memNE(s, locinput, ln)) |
d6a28714 | 3481 | sayNO; |
24d3c4a9 | 3482 | locinput += ln; |
d6a28714 JH |
3483 | nextchr = UCHARAT(locinput); |
3484 | break; | |
95b24440 | 3485 | } |
d6a28714 | 3486 | case EXACTFL: |
b8c5462f JH |
3487 | PL_reg_flags |= RF_tainted; |
3488 | /* FALL THROUGH */ | |
95b24440 | 3489 | case EXACTF: { |
be8e71aa | 3490 | char * const s = STRING(scan); |
24d3c4a9 | 3491 | ln = STR_LEN(scan); |
d6a28714 | 3492 | |
f2ed9b32 | 3493 | if (utf8_target || UTF_PATTERN) { |
d07ddd77 | 3494 | /* Either target or the pattern are utf8. */ |
be8e71aa | 3495 | const char * const l = locinput; |
d07ddd77 | 3496 | char *e = PL_regeol; |
bc517b45 | 3497 | |
f2ed9b32 KW |
3498 | if (! foldEQ_utf8(s, 0, ln, cBOOL(UTF_PATTERN), |
3499 | l, &e, 0, utf8_target)) { | |
5486206c JH |
3500 | /* One more case for the sharp s: |
3501 | * pack("U0U*", 0xDF) =~ /ss/i, | |
3502 | * the 0xC3 0x9F are the UTF-8 | |
3503 | * byte sequence for the U+00DF. */ | |
e1d1eefb | 3504 | |
f2ed9b32 | 3505 | if (!(utf8_target && |
e1d1eefb | 3506 | toLOWER(s[0]) == 's' && |
24d3c4a9 | 3507 | ln >= 2 && |
5486206c JH |
3508 | toLOWER(s[1]) == 's' && |
3509 | (U8)l[0] == 0xC3 && | |
3510 | e - l >= 2 && | |
3511 | (U8)l[1] == 0x9F)) | |
3512 | sayNO; | |
3513 | } | |
d07ddd77 JH |
3514 | locinput = e; |
3515 | nextchr = UCHARAT(locinput); | |
3516 | break; | |
a0ed51b3 | 3517 | } |
d6a28714 | 3518 | |
bc517b45 JH |
3519 | /* Neither the target and the pattern are utf8. */ |
3520 | ||
d6a28714 JH |
3521 | /* Inline the first character, for speed. */ |
3522 | if (UCHARAT(s) != nextchr && | |
3523 | UCHARAT(s) != ((OP(scan) == EXACTF) | |
3524 | ? PL_fold : PL_fold_locale)[nextchr]) | |
a0ed51b3 | 3525 | sayNO; |
24d3c4a9 | 3526 | if (PL_regeol - locinput < ln) |
b8c5462f | 3527 | sayNO; |
24d3c4a9 | 3528 | if (ln > 1 && (OP(scan) == EXACTF |
4c1b470c KW |
3529 | ? ! foldEQ(s, locinput, ln) |
3530 | : ! foldEQ_locale(s, locinput, ln))) | |
4633a7c4 | 3531 | sayNO; |
24d3c4a9 | 3532 | locinput += ln; |
d6a28714 | 3533 | nextchr = UCHARAT(locinput); |
a0d0e21e | 3534 | break; |
95b24440 | 3535 | } |
b2680017 YO |
3536 | case BOUNDL: |
3537 | case NBOUNDL: | |
3538 | PL_reg_flags |= RF_tainted; | |
3539 | /* FALL THROUGH */ | |
3540 | case BOUND: | |
3541 | case NBOUND: | |
3542 | /* was last char in word? */ | |
f2ed9b32 | 3543 | if (utf8_target) { |
b2680017 YO |
3544 | if (locinput == PL_bostr) |
3545 | ln = '\n'; | |
3546 | else { | |
3547 | const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); | |
3548 | ||
3549 | ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); | |
3550 | } | |
3551 | if (OP(scan) == BOUND || OP(scan) == NBOUND) { | |
3552 | ln = isALNUM_uni(ln); | |
3553 | LOAD_UTF8_CHARCLASS_ALNUM(); | |
f2ed9b32 | 3554 | n = swash_fetch(PL_utf8_alnum, (U8*)locinput, utf8_target); |
b2680017 YO |
3555 | } |
3556 | else { | |
3557 | ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); | |
3558 | n = isALNUM_LC_utf8((U8*)locinput); | |
3559 | } | |
3560 | } | |
3561 | else { | |
3562 | ln = (locinput != PL_bostr) ? | |
3563 | UCHARAT(locinput - 1) : '\n'; | |
3564 | if (OP(scan) == BOUND || OP(scan) == NBOUND) { | |
3565 | ln = isALNUM(ln); | |
3566 | n = isALNUM(nextchr); | |
3567 | } | |
3568 | else { | |
3569 | ln = isALNUM_LC(ln); | |
3570 | n = isALNUM_LC(nextchr); | |
3571 | } | |
3572 | } | |
3573 | if (((!ln) == (!n)) == (OP(scan) == BOUND || | |
3574 | OP(scan) == BOUNDL)) | |
3575 | sayNO; | |
3576 | break; | |
d6a28714 | 3577 | case ANYOF: |
f2ed9b32 | 3578 | if (utf8_target) { |
9e55ce06 JH |
3579 | STRLEN inclasslen = PL_regeol - locinput; |
3580 | ||
f2ed9b32 | 3581 | if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, utf8_target)) |
262b90c4 | 3582 | goto anyof_fail; |
ffc61ed2 JH |
3583 | if (locinput >= PL_regeol) |
3584 | sayNO; | |
0f0076b4 | 3585 | locinput += inclasslen ? inclasslen : UTF8SKIP(locinput); |
b8c5462f | 3586 | nextchr = UCHARAT(locinput); |
e0f9d4a8 | 3587 | break; |
ffc61ed2 JH |
3588 | } |
3589 | else { | |
3590 | if (nextchr < 0) | |
3591 | nextchr = UCHARAT(locinput); | |
32fc9b6a | 3592 | if (!REGINCLASS(rex, scan, (U8*)locinput)) |
262b90c4 | 3593 | goto anyof_fail; |
ffc61ed2 JH |
3594 | if (!nextchr && locinput >= PL_regeol) |
3595 | sayNO; | |
3596 | nextchr = UCHARAT(++locinput); | |
e0f9d4a8 JH |
3597 | break; |
3598 | } | |
262b90c4 | 3599 | anyof_fail: |
e0f9d4a8 JH |
3600 | /* If we might have the case of the German sharp s |
3601 | * in a casefolding Unicode character class. */ | |
3602 | ||
ebc501f0 JH |
3603 | if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) { |
3604 | locinput += SHARP_S_SKIP; | |
e0f9d4a8 | 3605 | nextchr = UCHARAT(locinput); |
ffc61ed2 | 3606 | } |
e0f9d4a8 JH |
3607 | else |
3608 | sayNO; | |
b8c5462f | 3609 | break; |
20d0b1e9 | 3610 | /* Special char classes - The defines start on line 129 or so */ |
d1eb3177 YO |
3611 | CCC_TRY_AFF( ALNUM, ALNUML, perl_word, "a", isALNUM_LC_utf8, isALNUM, isALNUM_LC); |
3612 | CCC_TRY_NEG(NALNUM, NALNUML, perl_word, "a", isALNUM_LC_utf8, isALNUM, isALNUM_LC); | |
20d0b1e9 | 3613 | |
d1eb3177 YO |
3614 | CCC_TRY_AFF( SPACE, SPACEL, perl_space, " ", isSPACE_LC_utf8, isSPACE, isSPACE_LC); |
3615 | CCC_TRY_NEG(NSPACE, NSPACEL, perl_space, " ", isSPACE_LC_utf8, isSPACE, isSPACE_LC); | |
20d0b1e9 | 3616 | |
d1eb3177 YO |
3617 | CCC_TRY_AFF( DIGIT, DIGITL, posix_digit, "0", isDIGIT_LC_utf8, isDIGIT, isDIGIT_LC); |
3618 | CCC_TRY_NEG(NDIGIT, NDIGITL, posix_digit, "0", isDIGIT_LC_utf8, isDIGIT, isDIGIT_LC); | |
20d0b1e9 | 3619 | |
37e2e78e KW |
3620 | case CLUMP: /* Match \X: logical Unicode character. This is defined as |
3621 | a Unicode extended Grapheme Cluster */ | |
3622 | /* From http://www.unicode.org/reports/tr29 (5.2 version). An | |
3623 | extended Grapheme Cluster is: | |
3624 | ||
3625 | CR LF | |
3626 | | Prepend* Begin Extend* | |
3627 | | . | |
3628 | ||
3629 | Begin is (Hangul-syllable | ! Control) | |
3630 | Extend is (Grapheme_Extend | Spacing_Mark) | |
3631 | Control is [ GCB_Control CR LF ] | |
3632 | ||
3633 | The discussion below shows how the code for CLUMP is derived | |
3634 | from this regex. Note that most of these concepts are from | |
3635 | property values of the Grapheme Cluster Boundary (GCB) property. | |
3636 | No code point can have multiple property values for a given | |
3637 | property. Thus a code point in Prepend can't be in Control, but | |
3638 | it must be in !Control. This is why Control above includes | |
3639 | GCB_Control plus CR plus LF. The latter two are used in the GCB | |
3640 | property separately, and so can't be in GCB_Control, even though | |
3641 | they logically are controls. Control is not the same as gc=cc, | |
3642 | but includes format and other characters as well. | |
3643 | ||
3644 | The Unicode definition of Hangul-syllable is: | |
3645 | L+ | |
3646 | | (L* ( ( V | LV ) V* | LVT ) T*) | |
3647 | | T+ | |
3648 | ) | |
3649 | Each of these is a value for the GCB property, and hence must be | |
3650 | disjoint, so the order they are tested is immaterial, so the | |
3651 | above can safely be changed to | |
3652 | T+ | |
3653 | | L+ | |
3654 | | (L* ( LVT | ( V | LV ) V*) T*) | |
3655 | ||
3656 | The last two terms can be combined like this: | |
3657 | L* ( L | |
3658 | | (( LVT | ( V | LV ) V*) T*)) | |
3659 | ||
3660 | And refactored into this: | |
3661 | L* (L | LVT T* | V V* T* | LV V* T*) | |
3662 | ||
3663 | That means that if we have seen any L's at all we can quit | |
3664 | there, but if the next character is a LVT, a V or and LV we | |
3665 | should keep going. | |
3666 | ||
3667 | There is a subtlety with Prepend* which showed up in testing. | |
3668 | Note that the Begin, and only the Begin is required in: | |
3669 | | Prepend* Begin Extend* | |
3670 | Also, Begin contains '! Control'. A Prepend must be a '! | |
3671 | Control', which means it must be a Begin. What it comes down to | |
3672 | is that if we match Prepend* and then find no suitable Begin | |
3673 | afterwards, that if we backtrack the last Prepend, that one will | |
3674 | be a suitable Begin. | |
3675 | */ | |
3676 | ||
b7c83a7e | 3677 | if (locinput >= PL_regeol) |
a0ed51b3 | 3678 | sayNO; |
f2ed9b32 | 3679 | if (! utf8_target) { |
37e2e78e KW |
3680 | |
3681 | /* Match either CR LF or '.', as all the other possibilities | |
3682 | * require utf8 */ | |
3683 | locinput++; /* Match the . or CR */ | |
3684 | if (nextchr == '\r' | |
3685 | && locinput < PL_regeol | |
3686 | && UCHARAT(locinput) == '\n') locinput++; | |
3687 | } | |
3688 | else { | |
3689 | ||
3690 | /* Utf8: See if is ( CR LF ); already know that locinput < | |
3691 | * PL_regeol, so locinput+1 is in bounds */ | |
3692 | if (nextchr == '\r' && UCHARAT(locinput + 1) == '\n') { | |
3693 | locinput += 2; | |
3694 | } | |
3695 | else { | |
3696 | /* In case have to backtrack to beginning, then match '.' */ | |
3697 | char *starting = locinput; | |
3698 | ||
3699 | /* In case have to backtrack the last prepend */ | |
3700 | char *previous_prepend = 0; | |
3701 | ||
3702 | LOAD_UTF8_CHARCLASS_GCB(); | |
3703 | ||
3704 | /* Match (prepend)* */ | |
3705 | while (locinput < PL_regeol | |
3706 | && swash_fetch(PL_utf8_X_prepend, | |
f2ed9b32 | 3707 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3708 | { |
3709 | previous_prepend = locinput; | |
3710 | locinput += UTF8SKIP(locinput); | |
3711 | } | |
3712 | ||
3713 | /* As noted above, if we matched a prepend character, but | |
3714 | * the next thing won't match, back off the last prepend we | |
3715 | * matched, as it is guaranteed to match the begin */ | |
3716 | if (previous_prepend | |
3717 | && (locinput >= PL_regeol | |
3718 | || ! swash_fetch(PL_utf8_X_begin, | |
f2ed9b32 | 3719 | (U8*)locinput, utf8_target))) |
37e2e78e KW |
3720 | { |
3721 | locinput = previous_prepend; | |
3722 | } | |
3723 | ||
3724 | /* Note that here we know PL_regeol > locinput, as we | |
3725 | * tested that upon input to this switch case, and if we | |
3726 | * moved locinput forward, we tested the result just above | |
3727 | * and it either passed, or we backed off so that it will | |
3728 | * now pass */ | |
f2ed9b32 | 3729 | if (! swash_fetch(PL_utf8_X_begin, (U8*)locinput, utf8_target)) { |
37e2e78e KW |
3730 | |
3731 | /* Here did not match the required 'Begin' in the | |
3732 | * second term. So just match the very first | |
3733 | * character, the '.' of the final term of the regex */ | |
3734 | locinput = starting + UTF8SKIP(starting); | |
3735 | } else { | |
3736 | ||
3737 | /* Here is the beginning of a character that can have | |
3738 | * an extender. It is either a hangul syllable, or a | |
3739 | * non-control */ | |
3740 | if (swash_fetch(PL_utf8_X_non_hangul, | |
f2ed9b32 | 3741 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3742 | { |
3743 | ||
3744 | /* Here not a Hangul syllable, must be a | |
3745 | * ('! * Control') */ | |
3746 | locinput += UTF8SKIP(locinput); | |
3747 | } else { | |
3748 | ||
3749 | /* Here is a Hangul syllable. It can be composed | |
3750 | * of several individual characters. One | |
3751 | * possibility is T+ */ | |
3752 | if (swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 3753 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3754 | { |
3755 | while (locinput < PL_regeol | |
3756 | && swash_fetch(PL_utf8_X_T, | |
f2ed9b32 | 3757 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3758 | { |
3759 | locinput += UTF8SKIP(locinput); | |
3760 | } | |
3761 | } else { | |
3762 | ||
3763 | /* Here, not T+, but is a Hangul. That means | |
3764 | * it is one of the others: L, LV, LVT or V, | |
3765 | * and matches: | |
3766 | * L* (L | LVT T* | V V* T* | LV V* T*) */ | |
3767 | ||
3768 | /* Match L* */ | |
3769 | while (locinput < PL_regeol | |
3770 | && swash_fetch(PL_utf8_X_L, | |
f2ed9b32 | 3771 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3772 | { |
3773 | locinput += UTF8SKIP(locinput); | |
3774 | } | |
3775 | ||
3776 | /* Here, have exhausted L*. If the next | |
3777 | * character is not an LV, LVT nor V, it means | |
3778 | * we had to have at least one L, so matches L+ | |
3779 | * in the original equation, we have a complete | |
3780 | * hangul syllable. Are done. */ | |
3781 | ||
3782 | if (locinput < PL_regeol | |
3783 | && swash_fetch(PL_utf8_X_LV_LVT_V, | |
f2ed9b32 | 3784 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3785 | { |
3786 | ||
3787 | /* Otherwise keep going. Must be LV, LVT | |
3788 | * or V. See if LVT */ | |
3789 | if (swash_fetch(PL_utf8_X_LVT, | |
f2ed9b32 | 3790 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3791 | { |
3792 | locinput += UTF8SKIP(locinput); | |
3793 | } else { | |
3794 | ||
3795 | /* Must be V or LV. Take it, then | |
3796 | * match V* */ | |
3797 | locinput += UTF8SKIP(locinput); | |
3798 | while (locinput < PL_regeol | |
3799 | && swash_fetch(PL_utf8_X_V, | |
f2ed9b32 | 3800 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3801 | { |
3802 | locinput += UTF8SKIP(locinput); | |
3803 | } | |
3804 | } | |
3805 | ||
3806 | /* And any of LV, LVT, or V can be followed | |
3807 | * by T* */ | |
3808 | while (locinput < PL_regeol | |
3809 | && swash_fetch(PL_utf8_X_T, | |
3810 | (U8*)locinput, | |
f2ed9b32 | 3811 | utf8_target)) |
37e2e78e KW |
3812 | { |
3813 | locinput += UTF8SKIP(locinput); | |
3814 | } | |
3815 | } | |
3816 | } | |
3817 | } | |
3818 | ||
3819 | /* Match any extender */ | |
3820 | while (locinput < PL_regeol | |
3821 | && swash_fetch(PL_utf8_X_extend, | |
f2ed9b32 | 3822 | (U8*)locinput, utf8_target)) |
37e2e78e KW |
3823 | { |
3824 | locinput += UTF8SKIP(locinput); | |
3825 | } | |
3826 | } | |
3827 | } | |
3828 | if (locinput > PL_regeol) sayNO; | |
3829 | } | |
a0ed51b3 LW |
3830 | nextchr = UCHARAT(locinput); |
3831 | break; | |
81714fb9 YO |
3832 | |
3833 | case NREFFL: | |
3834 | { | |
3835 | char *s; | |
ff1157ca | 3836 | char type; |
81714fb9 YO |
3837 | PL_reg_flags |= RF_tainted; |
3838 | /* FALL THROUGH */ | |
3839 | case NREF: | |
3840 | case NREFF: | |
ff1157ca | 3841 | type = OP(scan); |
0a4db386 YO |
3842 | n = reg_check_named_buff_matched(rex,scan); |
3843 | ||
3844 | if ( n ) { | |
3845 | type = REF + ( type - NREF ); | |
3846 | goto do_ref; | |
3847 | } else { | |
81714fb9 | 3848 | sayNO; |
0a4db386 YO |
3849 | } |
3850 | /* unreached */ | |
c8756f30 | 3851 | case REFFL: |
3280af22 | 3852 | PL_reg_flags |= RF_tainted; |
c8756f30 | 3853 | /* FALL THROUGH */ |
c277df42 | 3854 | case REF: |
81714fb9 | 3855 | case REFF: |
c277df42 | 3856 | n = ARG(scan); /* which paren pair */ |
81714fb9 YO |
3857 | type = OP(scan); |
3858 | do_ref: | |
f0ab9afb | 3859 | ln = PL_regoffs[n].start; |
2c2d71f5 | 3860 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
3b6647e0 | 3861 | if (*PL_reglastparen < n || ln == -1) |
af3f8c16 | 3862 | sayNO; /* Do not match unless seen CLOSEn. */ |
f0ab9afb | 3863 | if (ln == PL_regoffs[n].end) |
a0d0e21e | 3864 | break; |
a0ed51b3 | 3865 | |
24d3c4a9 | 3866 | s = PL_bostr + ln; |
f2ed9b32 | 3867 | if (utf8_target && type != REF) { /* REF can do byte comparison */ |
a0ed51b3 | 3868 | char *l = locinput; |
f0ab9afb | 3869 | const char *e = PL_bostr + PL_regoffs[n].end; |
a0ed51b3 LW |
3870 | /* |
3871 | * Note that we can't do the "other character" lookup trick as | |
3872 | * in the 8-bit case (no pun intended) because in Unicode we | |
3873 | * have to map both upper and title case to lower case. | |
3874 | */ | |
81714fb9 | 3875 | if (type == REFF) { |
a0ed51b3 | 3876 | while (s < e) { |
a3b680e6 AL |
3877 | STRLEN ulen1, ulen2; |
3878 | U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; | |
3879 | U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; | |
3880 | ||
a0ed51b3 LW |
3881 | if (l >= PL_regeol) |
3882 | sayNO; | |
a2a2844f JH |
3883 | toLOWER_utf8((U8*)s, tmpbuf1, &ulen1); |
3884 | toLOWER_utf8((U8*)l, tmpbuf2, &ulen2); | |
7114a2d2 | 3885 | if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1)) |
a0ed51b3 | 3886 | sayNO; |
a2a2844f JH |
3887 | s += ulen1; |
3888 | l += ulen2; | |
a0ed51b3 LW |
3889 | } |
3890 | } | |
3891 | locinput = l; | |
3892 | nextchr = UCHARAT(locinput); | |
3893 | break; | |
3894 | } | |
3895 | ||
a0d0e21e | 3896 | /* Inline the first character, for speed. */ |
76e3520e | 3897 | if (UCHARAT(s) != nextchr && |
81714fb9 YO |
3898 | (type == REF || |
3899 | (UCHARAT(s) != (type == REFF | |
3900 | ? PL_fold : PL_fold_locale)[nextchr]))) | |
4633a7c4 | 3901 | sayNO; |
f0ab9afb | 3902 | ln = PL_regoffs[n].end - ln; |
24d3c4a9 | 3903 | if (locinput + ln > PL_regeol) |
4633a7c4 | 3904 | sayNO; |
81714fb9 | 3905 | if (ln > 1 && (type == REF |
24d3c4a9 | 3906 | ? memNE(s, locinput, ln) |
81714fb9 | 3907 | : (type == REFF |
4c1b470c KW |
3908 | ? ! foldEQ(s, locinput, ln) |
3909 | : ! foldEQ_locale(s, locinput, ln)))) | |
4633a7c4 | 3910 | sayNO; |
24d3c4a9 | 3911 | locinput += ln; |
76e3520e | 3912 | nextchr = UCHARAT(locinput); |
a0d0e21e | 3913 | break; |
81714fb9 | 3914 | } |
a0d0e21e | 3915 | case NOTHING: |
c277df42 | 3916 | case TAIL: |
a0d0e21e LW |
3917 | break; |
3918 | case BACK: | |
3919 | break; | |
40a82448 DM |
3920 | |
3921 | #undef ST | |
3922 | #define ST st->u.eval | |
c277df42 | 3923 | { |
c277df42 | 3924 | SV *ret; |
d2f13c59 | 3925 | REGEXP *re_sv; |
6bda09f9 | 3926 | regexp *re; |
f8fc2ecf | 3927 | regexp_internal *rei; |
1a147d38 YO |
3928 | regnode *startpoint; |
3929 | ||
3930 | case GOSTART: | |
e7707071 YO |
3931 | case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */ |
3932 | if (cur_eval && cur_eval->locinput==locinput) { | |
24b23f37 | 3933 | if (cur_eval->u.eval.close_paren == (U32)ARG(scan)) |
1a147d38 | 3934 | Perl_croak(aTHX_ "Infinite recursion in regex"); |
4b196cd4 | 3935 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 YO |
3936 | Perl_croak(aTHX_ |
3937 | "Pattern subroutine nesting without pos change" | |
3938 | " exceeded limit in regex"); | |
6bda09f9 YO |
3939 | } else { |
3940 | nochange_depth = 0; | |
1a147d38 | 3941 | } |
288b8c02 | 3942 | re_sv = rex_sv; |
6bda09f9 | 3943 | re = rex; |
f8fc2ecf | 3944 | rei = rexi; |
288b8c02 | 3945 | (void)ReREFCNT_inc(rex_sv); |
1a147d38 | 3946 | if (OP(scan)==GOSUB) { |
6bda09f9 YO |
3947 | startpoint = scan + ARG2L(scan); |
3948 | ST.close_paren = ARG(scan); | |
3949 | } else { | |
f8fc2ecf | 3950 | startpoint = rei->program+1; |
6bda09f9 YO |
3951 | ST.close_paren = 0; |
3952 | } | |
3953 | goto eval_recurse_doit; | |
3954 | /* NOTREACHED */ | |
3955 | case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ | |
3956 | if (cur_eval && cur_eval->locinput==locinput) { | |
4b196cd4 | 3957 | if ( ++nochange_depth > max_nochange_depth ) |
1a147d38 | 3958 | Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex"); |
6bda09f9 YO |
3959 | } else { |
3960 | nochange_depth = 0; | |
3961 | } | |
8e5e9ebe | 3962 | { |
4aabdb9b DM |
3963 | /* execute the code in the {...} */ |
3964 | dSP; | |
6136c704 | 3965 | SV ** const before = SP; |
4aabdb9b DM |
3966 | OP_4tree * const oop = PL_op; |
3967 | COP * const ocurcop = PL_curcop; | |
3968 | PAD *old_comppad; | |
d80618d2 | 3969 | char *saved_regeol = PL_regeol; |
4aabdb9b DM |
3970 | |
3971 | n = ARG(scan); | |
f8fc2ecf | 3972 | PL_op = (OP_4tree*)rexi->data->data[n]; |
24b23f37 YO |
3973 | DEBUG_STATE_r( PerlIO_printf(Perl_debug_log, |
3974 | " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); | |
f8fc2ecf | 3975 | PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]); |
f0ab9afb | 3976 | PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr; |
4aabdb9b | 3977 | |
2bf803e2 YO |
3978 | if (sv_yes_mark) { |
3979 | SV *sv_mrk = get_sv("REGMARK", 1); | |
3980 | sv_setsv(sv_mrk, sv_yes_mark); | |
3981 | } | |
3982 | ||
8e5e9ebe RGS |
3983 | CALLRUNOPS(aTHX); /* Scalar context. */ |
3984 | SPAGAIN; | |
3985 | if (SP == before) | |
075aa684 | 3986 | ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */ |
8e5e9ebe RGS |
3987 | else { |
3988 | ret = POPs; | |
3989 | PUTBACK; | |
3990 | } | |
4aabdb9b DM |
3991 | |
3992 | PL_op = oop; | |
3993 | PAD_RESTORE_LOCAL(old_comppad); | |
3994 | PL_curcop = ocurcop; | |
d80618d2 | 3995 | PL_regeol = saved_regeol; |
24d3c4a9 | 3996 | if (!logical) { |
4aabdb9b DM |
3997 | /* /(?{...})/ */ |
3998 | sv_setsv(save_scalar(PL_replgv), ret); | |
4aabdb9b DM |
3999 | break; |
4000 | } | |
8e5e9ebe | 4001 | } |
24d3c4a9 DM |
4002 | if (logical == 2) { /* Postponed subexpression: /(??{...})/ */ |
4003 | logical = 0; | |
4aabdb9b | 4004 | { |
4f639d21 DM |
4005 | /* extract RE object from returned value; compiling if |
4006 | * necessary */ | |
6136c704 | 4007 | MAGIC *mg = NULL; |
288b8c02 | 4008 | REGEXP *rx = NULL; |
5c35adbb NC |
4009 | |
4010 | if (SvROK(ret)) { | |
288b8c02 | 4011 | SV *const sv = SvRV(ret); |
5c35adbb NC |
4012 | |
4013 | if (SvTYPE(sv) == SVt_REGEXP) { | |
d2f13c59 | 4014 | rx = (REGEXP*) sv; |
5c35adbb NC |
4015 | } else if (SvSMAGICAL(sv)) { |
4016 | mg = mg_find(sv, PERL_MAGIC_qr); | |
4017 | assert(mg); | |
4018 | } | |
4019 | } else if (SvTYPE(ret) == SVt_REGEXP) { | |
d2f13c59 | 4020 | rx = (REGEXP*) ret; |
5c35adbb | 4021 | } else if (SvSMAGICAL(ret)) { |
124ee91a NC |
4022 | if (SvGMAGICAL(ret)) { |
4023 | /* I don't believe that there is ever qr magic | |
4024 | here. */ | |
4025 | assert(!mg_find(ret, PERL_MAGIC_qr)); | |
faf82a0b | 4026 | sv_unmagic(ret, PERL_MAGIC_qr); |
124ee91a NC |
4027 | } |
4028 | else { | |
faf82a0b | 4029 | mg = mg_find(ret, PERL_MAGIC_qr); |
124ee91a NC |
4030 | /* testing suggests mg only ends up non-NULL for |
4031 | scalars who were upgraded and compiled in the | |
4032 | else block below. In turn, this is only | |
4033 | triggered in the "postponed utf8 string" tests | |
4034 | in t/op/pat.t */ | |
4035 | } | |
0f5d15d6 | 4036 | } |
faf82a0b | 4037 | |
0f5d15d6 | 4038 | if (mg) { |
d2f13c59 | 4039 | rx = (REGEXP *) mg->mg_obj; /*XXX:dmq*/ |
e2560c33 | 4040 | assert(rx); |
0f5d15d6 | 4041 | } |
288b8c02 | 4042 | if (rx) { |
f0826785 | 4043 | rx = reg_temp_copy(NULL, rx); |
288b8c02 | 4044 | } |
0f5d15d6 | 4045 | else { |
c737faaf | 4046 | U32 pm_flags = 0; |
a3b680e6 | 4047 | const I32 osize = PL_regsize; |
0f5d15d6 | 4048 | |
b9ad30b4 NC |
4049 | if (DO_UTF8(ret)) { |
4050 | assert (SvUTF8(ret)); | |
4051 | } else if (SvUTF8(ret)) { | |
4052 | /* Not doing UTF-8, despite what the SV says. Is | |
4053 | this only if we're trapped in use 'bytes'? */ | |
4054 | /* Make a copy of the octet sequence, but without | |
4055 | the flag on, as the compiler now honours the | |
4056 | SvUTF8 flag on ret. */ | |
4057 | STRLEN len; | |
4058 | const char *const p = SvPV(ret, len); | |
4059 | ret = newSVpvn_flags(p, len, SVs_TEMP); | |
4060 | } | |
288b8c02 | 4061 | rx = CALLREGCOMP(ret, pm_flags); |
9041c2e3 | 4062 | if (!(SvFLAGS(ret) |
faf82a0b | 4063 | & (SVs_TEMP | SVs_PADTMP | SVf_READONLY |
3ce3ed55 | 4064 | | SVs_GMG))) { |
a2794585 NC |
4065 | /* This isn't a first class regexp. Instead, it's |
4066 | caching a regexp onto an existing, Perl visible | |
4067 | scalar. */ | |
ad64d0ec | 4068 | sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0); |
3ce3ed55 | 4069 | } |
0f5d15d6 | 4070 | PL_regsize = osize; |
0f5d15d6 | 4071 | } |
288b8c02 NC |
4072 | re_sv = rx; |
4073 | re = (struct regexp *)SvANY(rx); | |
4aabdb9b | 4074 | } |
07bc277f | 4075 | RXp_MATCH_COPIED_off(re); |
28d8d7f4 YO |
4076 | re->subbeg = rex->subbeg; |
4077 | re->sublen = rex->sublen; | |
f8fc2ecf | 4078 | rei = RXi_GET(re); |
6bda09f9 | 4079 | DEBUG_EXECUTE_r( |
f2ed9b32 | 4080 | debug_start_match(re_sv, utf8_target, locinput, PL_regeol, |
6bda09f9 YO |
4081 | "Matching embedded"); |
4082 | ); | |
f8fc2ecf | 4083 | startpoint = rei->program + 1; |
1a147d38 | 4084 | ST.close_paren = 0; /* only used for GOSUB */ |
6bda09f9 YO |
4085 | /* borrowed from regtry */ |
4086 | if (PL_reg_start_tmpl <= re->nparens) { | |
4087 | PL_reg_start_tmpl = re->nparens*3/2 + 3; | |
4088 | if(PL_reg_start_tmp) | |
4089 | Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
4090 | else | |
4091 | Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*); | |
dd5def09 | 4092 | } |
aa283a38 | 4093 | |
1a147d38 | 4094 | eval_recurse_doit: /* Share code with GOSUB below this line */ |
aa283a38 | 4095 | /* run the pattern returned from (??{...}) */ |
40a82448 DM |
4096 | ST.cp = regcppush(0); /* Save *all* the positions. */ |
4097 | REGCP_SET(ST.lastcp); | |
6bda09f9 | 4098 | |
f0ab9afb | 4099 | PL_regoffs = re->offs; /* essentially NOOP on GOSUB */ |
6bda09f9 | 4100 | |
0357f1fd ML |
4101 | /* see regtry, specifically PL_reglast(?:close)?paren is a pointer! (i dont know why) :dmq */ |
4102 | PL_reglastparen = &re->lastparen; | |
4103 | PL_reglastcloseparen = &re->lastcloseparen; | |
4104 | re->lastparen = 0; | |
4105 | re->lastcloseparen = 0; | |
4106 | ||
4aabdb9b | 4107 | PL_reginput = locinput; |
ae0beba1 | 4108 | PL_regsize = 0; |
4aabdb9b DM |
4109 | |
4110 | /* XXXX This is too dramatic a measure... */ | |
4111 | PL_reg_maxiter = 0; | |
4112 | ||
faec1544 | 4113 | ST.toggle_reg_flags = PL_reg_flags; |
3c8556c3 | 4114 | if (RX_UTF8(re_sv)) |
faec1544 DM |
4115 | PL_reg_flags |= RF_utf8; |
4116 | else | |
4117 | PL_reg_flags &= ~RF_utf8; | |
4118 | ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */ | |
4119 | ||
288b8c02 | 4120 | ST.prev_rex = rex_sv; |
faec1544 | 4121 | ST.prev_curlyx = cur_curlyx; |
288b8c02 NC |
4122 | SETREX(rex_sv,re_sv); |
4123 | rex = re; | |
f8fc2ecf | 4124 | rexi = rei; |
faec1544 | 4125 | cur_curlyx = NULL; |
40a82448 | 4126 | ST.B = next; |
faec1544 DM |
4127 | ST.prev_eval = cur_eval; |
4128 | cur_eval = st; | |
faec1544 | 4129 | /* now continue from first node in postoned RE */ |
6bda09f9 | 4130 | PUSH_YES_STATE_GOTO(EVAL_AB, startpoint); |
4aabdb9b | 4131 | /* NOTREACHED */ |
a0ed51b3 | 4132 | } |
24d3c4a9 | 4133 | /* logical is 1, /(?(?{...})X|Y)/ */ |
f2338a2e | 4134 | sw = cBOOL(SvTRUE(ret)); |
24d3c4a9 | 4135 | logical = 0; |
c277df42 IZ |
4136 | break; |
4137 | } | |
40a82448 | 4138 | |
faec1544 DM |
4139 | case EVAL_AB: /* cleanup after a successful (??{A})B */ |
4140 | /* note: this is called twice; first after popping B, then A */ | |
4141 | PL_reg_flags ^= ST.toggle_reg_flags; | |
288b8c02 NC |
4142 | ReREFCNT_dec(rex_sv); |
4143 | SETREX(rex_sv,ST.prev_rex); | |
4144 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 4145 | rexi = RXi_GET(rex); |
faec1544 DM |
4146 | regcpblow(ST.cp); |
4147 | cur_eval = ST.prev_eval; | |
4148 | cur_curlyx = ST.prev_curlyx; | |
34a81e2b B |
4149 | |
4150 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ | |
0357f1fd ML |
4151 | PL_reglastparen = &rex->lastparen; |
4152 | PL_reglastcloseparen = &rex->lastcloseparen; | |
34a81e2b B |
4153 | /* also update PL_regoffs */ |
4154 | PL_regoffs = rex->offs; | |
0357f1fd | 4155 | |
40a82448 DM |
4156 | /* XXXX This is too dramatic a measure... */ |
4157 | PL_reg_maxiter = 0; | |
e7707071 | 4158 | if ( nochange_depth ) |
4b196cd4 | 4159 | nochange_depth--; |
262b90c4 | 4160 | sayYES; |
40a82448 | 4161 | |
40a82448 | 4162 | |
faec1544 DM |
4163 | case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */ |
4164 | /* note: this is called twice; first after popping B, then A */ | |
4165 | PL_reg_flags ^= ST.toggle_reg_flags; | |
288b8c02 NC |
4166 | ReREFCNT_dec(rex_sv); |
4167 | SETREX(rex_sv,ST.prev_rex); | |
4168 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 4169 | rexi = RXi_GET(rex); |
34a81e2b | 4170 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ |
0357f1fd ML |
4171 | PL_reglastparen = &rex->lastparen; |
4172 | PL_reglastcloseparen = &rex->lastcloseparen; | |
4173 | ||
40a82448 DM |
4174 | PL_reginput = locinput; |
4175 | REGCP_UNWIND(ST.lastcp); | |
4176 | regcppop(rex); | |
faec1544 DM |
4177 | cur_eval = ST.prev_eval; |
4178 | cur_curlyx = ST.prev_curlyx; | |
4179 | /* XXXX This is too dramatic a measure... */ | |
4180 | PL_reg_maxiter = 0; | |
e7707071 | 4181 | if ( nochange_depth ) |
4b196cd4 | 4182 | nochange_depth--; |
40a82448 | 4183 | sayNO_SILENT; |
40a82448 DM |
4184 | #undef ST |
4185 | ||
a0d0e21e | 4186 | case OPEN: |
c277df42 | 4187 | n = ARG(scan); /* which paren pair */ |
3280af22 NIS |
4188 | PL_reg_start_tmp[n] = locinput; |
4189 | if (n > PL_regsize) | |
4190 | PL_regsize = n; | |
e2e6a0f1 | 4191 | lastopen = n; |
a0d0e21e LW |
4192 | break; |
4193 | case CLOSE: | |
c277df42 | 4194 | n = ARG(scan); /* which paren pair */ |
f0ab9afb NC |
4195 | PL_regoffs[n].start = PL_reg_start_tmp[n] - PL_bostr; |
4196 | PL_regoffs[n].end = locinput - PL_bostr; | |
7f69552c YO |
4197 | /*if (n > PL_regsize) |
4198 | PL_regsize = n;*/ | |
3b6647e0 | 4199 | if (n > *PL_reglastparen) |
3280af22 | 4200 | *PL_reglastparen = n; |
a01268b5 | 4201 | *PL_reglastcloseparen = n; |
3b6647e0 | 4202 | if (cur_eval && cur_eval->u.eval.close_paren == n) { |
6bda09f9 YO |
4203 | goto fake_end; |
4204 | } | |
a0d0e21e | 4205 | break; |
e2e6a0f1 YO |
4206 | case ACCEPT: |
4207 | if (ARG(scan)){ | |
4208 | regnode *cursor; | |
4209 | for (cursor=scan; | |
4210 | cursor && OP(cursor)!=END; | |
4211 | cursor=regnext(cursor)) | |
4212 | { | |
4213 | if ( OP(cursor)==CLOSE ){ | |
4214 | n = ARG(cursor); | |
4215 | if ( n <= lastopen ) { | |
f0ab9afb NC |
4216 | PL_regoffs[n].start |
4217 | = PL_reg_start_tmp[n] - PL_bostr; | |
4218 | PL_regoffs[n].end = locinput - PL_bostr; | |
e2e6a0f1 YO |
4219 | /*if (n > PL_regsize) |
4220 | PL_regsize = n;*/ | |
3b6647e0 | 4221 | if (n > *PL_reglastparen) |
e2e6a0f1 YO |
4222 | *PL_reglastparen = n; |
4223 | *PL_reglastcloseparen = n; | |
3b6647e0 RB |
4224 | if ( n == ARG(scan) || (cur_eval && |
4225 | cur_eval->u.eval.close_paren == n)) | |
e2e6a0f1 YO |
4226 | break; |
4227 | } | |
4228 | } | |
4229 | } | |
4230 | } | |
4231 | goto fake_end; | |
4232 | /*NOTREACHED*/ | |
c277df42 IZ |
4233 | case GROUPP: |
4234 | n = ARG(scan); /* which paren pair */ | |
f2338a2e | 4235 | sw = cBOOL(*PL_reglastparen >= n && PL_regoffs[n].end != -1); |
c277df42 | 4236 | break; |
0a4db386 YO |
4237 | case NGROUPP: |
4238 | /* reg_check_named_buff_matched returns 0 for no match */ | |
f2338a2e | 4239 | sw = cBOOL(0 < reg_check_named_buff_matched(rex,scan)); |
0a4db386 | 4240 | break; |
1a147d38 | 4241 | case INSUBP: |
0a4db386 | 4242 | n = ARG(scan); |
3b6647e0 | 4243 | sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n)); |
0a4db386 YO |
4244 | break; |
4245 | case DEFINEP: | |
4246 | sw = 0; | |
4247 | break; | |
c277df42 | 4248 | case IFTHEN: |
2c2d71f5 | 4249 | PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ |
24d3c4a9 | 4250 | if (sw) |
c277df42 IZ |
4251 | next = NEXTOPER(NEXTOPER(scan)); |
4252 | else { | |
4253 | next = scan + ARG(scan); | |
4254 | if (OP(next) == IFTHEN) /* Fake one. */ | |
4255 | next = NEXTOPER(NEXTOPER(next)); | |
4256 | } | |
4257 | break; | |
4258 | case LOGICAL: | |
24d3c4a9 | 4259 | logical = scan->flags; |
c277df42 | 4260 | break; |
c476f425 | 4261 | |
2ab05381 | 4262 | /******************************************************************* |
2ab05381 | 4263 | |
c476f425 DM |
4264 | The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/ |
4265 | pattern, where A and B are subpatterns. (For simple A, CURLYM or | |
4266 | STAR/PLUS/CURLY/CURLYN are used instead.) | |
2ab05381 | 4267 | |
c476f425 | 4268 | A*B is compiled as <CURLYX><A><WHILEM><B> |
2ab05381 | 4269 | |
c476f425 DM |
4270 | On entry to the subpattern, CURLYX is called. This pushes a CURLYX |
4271 | state, which contains the current count, initialised to -1. It also sets | |
4272 | cur_curlyx to point to this state, with any previous value saved in the | |
4273 | state block. | |
2ab05381 | 4274 | |
c476f425 DM |
4275 | CURLYX then jumps straight to the WHILEM op, rather than executing A, |
4276 | since the pattern may possibly match zero times (i.e. it's a while {} loop | |
4277 | rather than a do {} while loop). | |
2ab05381 | 4278 | |
c476f425 DM |
4279 | Each entry to WHILEM represents a successful match of A. The count in the |
4280 | CURLYX block is incremented, another WHILEM state is pushed, and execution | |
4281 | passes to A or B depending on greediness and the current count. | |
2ab05381 | 4282 | |
c476f425 DM |
4283 | For example, if matching against the string a1a2a3b (where the aN are |
4284 | substrings that match /A/), then the match progresses as follows: (the | |
4285 | pushed states are interspersed with the bits of strings matched so far): | |
2ab05381 | 4286 | |
c476f425 DM |
4287 | <CURLYX cnt=-1> |
4288 | <CURLYX cnt=0><WHILEM> | |
4289 | <CURLYX cnt=1><WHILEM> a1 <WHILEM> | |
4290 | <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM> | |
4291 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> | |
4292 | <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b | |
2ab05381 | 4293 | |
c476f425 DM |
4294 | (Contrast this with something like CURLYM, which maintains only a single |
4295 | backtrack state: | |
2ab05381 | 4296 | |
c476f425 DM |
4297 | <CURLYM cnt=0> a1 |
4298 | a1 <CURLYM cnt=1> a2 | |
4299 | a1 a2 <CURLYM cnt=2> a3 | |
4300 | a1 a2 a3 <CURLYM cnt=3> b | |
4301 | ) | |
2ab05381 | 4302 | |
c476f425 DM |
4303 | Each WHILEM state block marks a point to backtrack to upon partial failure |
4304 | of A or B, and also contains some minor state data related to that | |
4305 | iteration. The CURLYX block, pointed to by cur_curlyx, contains the | |
4306 | overall state, such as the count, and pointers to the A and B ops. | |
2ab05381 | 4307 | |
c476f425 DM |
4308 | This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx |
4309 | must always point to the *current* CURLYX block, the rules are: | |
2ab05381 | 4310 | |
c476f425 DM |
4311 | When executing CURLYX, save the old cur_curlyx in the CURLYX state block, |
4312 | and set cur_curlyx to point the new block. | |
2ab05381 | 4313 | |
c476f425 DM |
4314 | When popping the CURLYX block after a successful or unsuccessful match, |
4315 | restore the previous cur_curlyx. | |
2ab05381 | 4316 | |
c476f425 DM |
4317 | When WHILEM is about to execute B, save the current cur_curlyx, and set it |
4318 | to the outer one saved in the CURLYX block. | |
2ab05381 | 4319 | |
c476f425 DM |
4320 | When popping the WHILEM block after a successful or unsuccessful B match, |
4321 | restore the previous cur_curlyx. | |
2ab05381 | 4322 | |
c476f425 DM |
4323 | Here's an example for the pattern (AI* BI)*BO |
4324 | I and O refer to inner and outer, C and W refer to CURLYX and WHILEM: | |
2ab05381 | 4325 | |
c476f425 DM |
4326 | cur_ |
4327 | curlyx backtrack stack | |
4328 | ------ --------------- | |
4329 | NULL | |
4330 | CO <CO prev=NULL> <WO> | |
4331 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4332 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi | |
4333 | NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo | |
2ab05381 | 4334 | |
c476f425 DM |
4335 | At this point the pattern succeeds, and we work back down the stack to |
4336 | clean up, restoring as we go: | |
95b24440 | 4337 | |
c476f425 DM |
4338 | CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi |
4339 | CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai | |
4340 | CO <CO prev=NULL> <WO> | |
4341 | NULL | |
a0374537 | 4342 | |
c476f425 DM |
4343 | *******************************************************************/ |
4344 | ||
4345 | #define ST st->u.curlyx | |
4346 | ||
4347 | case CURLYX: /* start of /A*B/ (for complex A) */ | |
4348 | { | |
4349 | /* No need to save/restore up to this paren */ | |
4350 | I32 parenfloor = scan->flags; | |
4351 | ||
4352 | assert(next); /* keep Coverity happy */ | |
4353 | if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ | |
4354 | next += ARG(next); | |
4355 | ||
4356 | /* XXXX Probably it is better to teach regpush to support | |
4357 | parenfloor > PL_regsize... */ | |
4358 | if (parenfloor > (I32)*PL_reglastparen) | |
4359 | parenfloor = *PL_reglastparen; /* Pessimization... */ | |
4360 | ||
4361 | ST.prev_curlyx= cur_curlyx; | |
4362 | cur_curlyx = st; | |
4363 | ST.cp = PL_savestack_ix; | |
4364 | ||
4365 | /* these fields contain the state of the current curly. | |
4366 | * they are accessed by subsequent WHILEMs */ | |
4367 | ST.parenfloor = parenfloor; | |
d02d6d97 | 4368 | ST.me = scan; |
c476f425 | 4369 | ST.B = next; |
24d3c4a9 DM |
4370 | ST.minmod = minmod; |
4371 | minmod = 0; | |
c476f425 DM |
4372 | ST.count = -1; /* this will be updated by WHILEM */ |
4373 | ST.lastloc = NULL; /* this will be updated by WHILEM */ | |
4374 | ||
4375 | PL_reginput = locinput; | |
4376 | PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next)); | |
5f66b61c | 4377 | /* NOTREACHED */ |
c476f425 | 4378 | } |
a0d0e21e | 4379 | |
c476f425 | 4380 | case CURLYX_end: /* just finished matching all of A*B */ |
c476f425 DM |
4381 | cur_curlyx = ST.prev_curlyx; |
4382 | sayYES; | |
4383 | /* NOTREACHED */ | |
a0d0e21e | 4384 | |
c476f425 DM |
4385 | case CURLYX_end_fail: /* just failed to match all of A*B */ |
4386 | regcpblow(ST.cp); | |
4387 | cur_curlyx = ST.prev_curlyx; | |
4388 | sayNO; | |
4389 | /* NOTREACHED */ | |
4633a7c4 | 4390 | |
a0d0e21e | 4391 | |
c476f425 DM |
4392 | #undef ST |
4393 | #define ST st->u.whilem | |
4394 | ||
4395 | case WHILEM: /* just matched an A in /A*B/ (for complex A) */ | |
4396 | { | |
4397 | /* see the discussion above about CURLYX/WHILEM */ | |
c476f425 | 4398 | I32 n; |
d02d6d97 DM |
4399 | int min = ARG1(cur_curlyx->u.curlyx.me); |
4400 | int max = ARG2(cur_curlyx->u.curlyx.me); | |
4401 | regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS; | |
4402 | ||
c476f425 DM |
4403 | assert(cur_curlyx); /* keep Coverity happy */ |
4404 | n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */ | |
4405 | ST.save_lastloc = cur_curlyx->u.curlyx.lastloc; | |
4406 | ST.cache_offset = 0; | |
4407 | ST.cache_mask = 0; | |
4408 | ||
4409 | PL_reginput = locinput; | |
4410 | ||
4411 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
d02d6d97 DM |
4412 | "%*s whilem: matched %ld out of %d..%d\n", |
4413 | REPORT_CODE_OFF+depth*2, "", (long)n, min, max) | |
c476f425 | 4414 | ); |
a0d0e21e | 4415 | |
c476f425 | 4416 | /* First just match a string of min A's. */ |
a0d0e21e | 4417 | |
d02d6d97 | 4418 | if (n < min) { |
c476f425 | 4419 | cur_curlyx->u.curlyx.lastloc = locinput; |
d02d6d97 | 4420 | PUSH_STATE_GOTO(WHILEM_A_pre, A); |
c476f425 DM |
4421 | /* NOTREACHED */ |
4422 | } | |
4423 | ||
4424 | /* If degenerate A matches "", assume A done. */ | |
4425 | ||
4426 | if (locinput == cur_curlyx->u.curlyx.lastloc) { | |
4427 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4428 | "%*s whilem: empty match detected, trying continuation...\n", | |
4429 | REPORT_CODE_OFF+depth*2, "") | |
4430 | ); | |
4431 | goto do_whilem_B_max; | |
4432 | } | |
4433 | ||
4434 | /* super-linear cache processing */ | |
4435 | ||
4436 | if (scan->flags) { | |
a0d0e21e | 4437 | |
2c2d71f5 | 4438 | if (!PL_reg_maxiter) { |
c476f425 DM |
4439 | /* start the countdown: Postpone detection until we |
4440 | * know the match is not *that* much linear. */ | |
2c2d71f5 | 4441 | PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4); |
66bf836d DM |
4442 | /* possible overflow for long strings and many CURLYX's */ |
4443 | if (PL_reg_maxiter < 0) | |
4444 | PL_reg_maxiter = I32_MAX; | |
2c2d71f5 JH |
4445 | PL_reg_leftiter = PL_reg_maxiter; |
4446 | } | |
c476f425 | 4447 | |
2c2d71f5 | 4448 | if (PL_reg_leftiter-- == 0) { |
c476f425 | 4449 | /* initialise cache */ |
3298f257 | 4450 | const I32 size = (PL_reg_maxiter + 7)/8; |
2c2d71f5 | 4451 | if (PL_reg_poscache) { |
eb160463 | 4452 | if ((I32)PL_reg_poscache_size < size) { |
2c2d71f5 JH |
4453 | Renew(PL_reg_poscache, size, char); |
4454 | PL_reg_poscache_size = size; | |
4455 | } | |
4456 | Zero(PL_reg_poscache, size, char); | |
4457 | } | |
4458 | else { | |
4459 | PL_reg_poscache_size = size; | |
a02a5408 | 4460 | Newxz(PL_reg_poscache, size, char); |
2c2d71f5 | 4461 | } |
c476f425 DM |
4462 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, |
4463 | "%swhilem: Detected a super-linear match, switching on caching%s...\n", | |
4464 | PL_colors[4], PL_colors[5]) | |
4465 | ); | |
2c2d71f5 | 4466 | } |
c476f425 | 4467 | |
2c2d71f5 | 4468 | if (PL_reg_leftiter < 0) { |
c476f425 DM |
4469 | /* have we already failed at this position? */ |
4470 | I32 offset, mask; | |
4471 | offset = (scan->flags & 0xf) - 1 | |
4472 | + (locinput - PL_bostr) * (scan->flags>>4); | |
4473 | mask = 1 << (offset % 8); | |
4474 | offset /= 8; | |
4475 | if (PL_reg_poscache[offset] & mask) { | |
4476 | DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, | |
4477 | "%*s whilem: (cache) already tried at this position...\n", | |
4478 | REPORT_CODE_OFF+depth*2, "") | |
2c2d71f5 | 4479 | ); |
3298f257 | 4480 | sayNO; /* cache records failure */ |
2c2d71f5 | 4481 | } |
c476f425 DM |
4482 | ST.cache_offset = offset; |
4483 | ST.cache_mask = mask; | |
2c2d71f5 | 4484 | } |
c476f425 | 4485 | } |
2c2d71f5 | 4486 | |
c476f425 | 4487 | /* Prefer B over A for minimal matching. */ |
a687059c | 4488 | |
c476f425 DM |
4489 | if (cur_curlyx->u.curlyx.minmod) { |
4490 | ST.save_curlyx = cur_curlyx; | |
4491 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4492 | ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor); | |
4493 | REGCP_SET(ST.lastcp); | |
4494 | PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B); | |
4495 | /* NOTREACHED */ | |
4496 | } | |
a0d0e21e | 4497 | |
c476f425 DM |
4498 | /* Prefer A over B for maximal matching. */ |
4499 | ||
d02d6d97 | 4500 | if (n < max) { /* More greed allowed? */ |
c476f425 DM |
4501 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); |
4502 | cur_curlyx->u.curlyx.lastloc = locinput; | |
4503 | REGCP_SET(ST.lastcp); | |
d02d6d97 | 4504 | PUSH_STATE_GOTO(WHILEM_A_max, A); |
c476f425 DM |
4505 | /* NOTREACHED */ |
4506 | } | |
4507 | goto do_whilem_B_max; | |
4508 | } | |
4509 | /* NOTREACHED */ | |
4510 | ||
4511 | case WHILEM_B_min: /* just matched B in a minimal match */ | |
4512 | case WHILEM_B_max: /* just matched B in a maximal match */ | |
4513 | cur_curlyx = ST.save_curlyx; | |
4514 | sayYES; | |
4515 | /* NOTREACHED */ | |
4516 | ||
4517 | case WHILEM_B_max_fail: /* just failed to match B in a maximal match */ | |
4518 | cur_curlyx = ST.save_curlyx; | |
4519 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
4520 | cur_curlyx->u.curlyx.count--; | |
4521 | CACHEsayNO; | |
4522 | /* NOTREACHED */ | |
4523 | ||
4524 | case WHILEM_A_min_fail: /* just failed to match A in a minimal match */ | |
4525 | REGCP_UNWIND(ST.lastcp); | |
4526 | regcppop(rex); | |
4527 | /* FALL THROUGH */ | |
4528 | case WHILEM_A_pre_fail: /* just failed to match even minimal A */ | |
4529 | cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; | |
4530 | cur_curlyx->u.curlyx.count--; | |
4531 | CACHEsayNO; | |
4532 | /* NOTREACHED */ | |
4533 | ||
4534 | case WHILEM_A_max_fail: /* just failed to match A in a maximal match */ | |
4535 | REGCP_UNWIND(ST.lastcp); | |
4536 | regcppop(rex); /* Restore some previous $<digit>s? */ | |
4537 | PL_reginput = locinput; | |
4538 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
4539 | "%*s whilem: failed, trying continuation...\n", | |
4540 | REPORT_CODE_OFF+depth*2, "") | |
4541 | ); | |
4542 | do_whilem_B_max: | |
4543 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
4544 | && ckWARN(WARN_REGEXP) | |
4545 | && !(PL_reg_flags & RF_warned)) | |
4546 | { | |
4547 | PL_reg_flags |= RF_warned; | |
4548 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded", | |
4549 | "Complex regular subexpression recursion", | |
4550 | REG_INFTY - 1); | |
4551 | } | |
4552 | ||
4553 | /* now try B */ | |
4554 | ST.save_curlyx = cur_curlyx; | |
4555 | cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; | |
4556 | PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B); | |
4557 | /* NOTREACHED */ | |
4558 | ||
4559 | case WHILEM_B_min_fail: /* just failed to match B in a minimal match */ | |
4560 | cur_curlyx = ST.save_curlyx; | |
4561 | REGCP_UNWIND(ST.lastcp); | |
4562 | regcppop(rex); | |
4563 | ||
d02d6d97 | 4564 | if (cur_curlyx->u.curlyx.count >= /*max*/ARG2(cur_curlyx->u.curlyx.me)) { |
c476f425 DM |
4565 | /* Maximum greed exceeded */ |
4566 | if (cur_curlyx->u.curlyx.count >= REG_INFTY | |
4567 | && ckWARN(WARN_REGEXP) | |
4568 | && !(PL_reg_flags & RF_warned)) | |
4569 | { | |
3280af22 | 4570 | PL_reg_flags |= RF_warned; |
c476f425 DM |
4571 | Perl_warner(aTHX_ packWARN(WARN_REGEXP), |
4572 | "%s limit (%d) exceeded", | |
4573 | "Complex regular subexpression recursion", | |
4574 | REG_INFTY - 1); | |
a0d0e21e | 4575 | } |
c476f425 | 4576 | cur_curlyx->u.curlyx.count--; |
3ab3c9b4 | 4577 | CACHEsayNO; |
a0d0e21e | 4578 | } |
c476f425 DM |
4579 | |
4580 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, | |
4581 | "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "") | |
4582 | ); | |
4583 | /* Try grabbing another A and see if it helps. */ | |
4584 | PL_reginput = locinput; | |
4585 | cur_curlyx->u.curlyx.lastloc = locinput; | |
4586 | ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor); | |
4587 | REGCP_SET(ST.lastcp); | |
d02d6d97 DM |
4588 | PUSH_STATE_GOTO(WHILEM_A_min, |
4589 | /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS); | |
5f66b61c | 4590 | /* NOTREACHED */ |
40a82448 DM |
4591 | |
4592 | #undef ST | |
4593 | #define ST st->u.branch | |
4594 | ||
4595 | case BRANCHJ: /* /(...|A|...)/ with long next pointer */ | |
c277df42 IZ |
4596 | next = scan + ARG(scan); |
4597 | if (next == scan) | |
4598 | next = NULL; | |
40a82448 DM |
4599 | scan = NEXTOPER(scan); |
4600 | /* FALL THROUGH */ | |
c277df42 | 4601 | |
40a82448 DM |
4602 | case BRANCH: /* /(...|A|...)/ */ |
4603 | scan = NEXTOPER(scan); /* scan now points to inner node */ | |
40a82448 DM |
4604 | ST.lastparen = *PL_reglastparen; |
4605 | ST.next_branch = next; | |
4606 | REGCP_SET(ST.cp); | |
4607 | PL_reginput = locinput; | |
02db2b7b | 4608 | |
40a82448 | 4609 | /* Now go into the branch */ |
5d458dd8 YO |
4610 | if (has_cutgroup) { |
4611 | PUSH_YES_STATE_GOTO(BRANCH_next, scan); | |
4612 | } else { | |
4613 | PUSH_STATE_GOTO(BRANCH_next, scan); | |
4614 | } | |
40a82448 | 4615 | /* NOTREACHED */ |
5d458dd8 YO |
4616 | case CUTGROUP: |
4617 | PL_reginput = locinput; | |
4618 | sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : | |
ad64d0ec | 4619 | MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
4620 | PUSH_STATE_GOTO(CUTGROUP_next,next); |
4621 | /* NOTREACHED */ | |
4622 | case CUTGROUP_next_fail: | |
4623 | do_cutgroup = 1; | |
4624 | no_final = 1; | |
4625 | if (st->u.mark.mark_name) | |
4626 | sv_commit = st->u.mark.mark_name; | |
4627 | sayNO; | |
4628 | /* NOTREACHED */ | |
4629 | case BRANCH_next: | |
4630 | sayYES; | |
4631 | /* NOTREACHED */ | |
40a82448 | 4632 | case BRANCH_next_fail: /* that branch failed; try the next, if any */ |
5d458dd8 YO |
4633 | if (do_cutgroup) { |
4634 | do_cutgroup = 0; | |
4635 | no_final = 0; | |
4636 | } | |
40a82448 DM |
4637 | REGCP_UNWIND(ST.cp); |
4638 | for (n = *PL_reglastparen; n > ST.lastparen; n--) | |
f0ab9afb | 4639 | PL_regoffs[n].end = -1; |
40a82448 | 4640 | *PL_reglastparen = n; |
0a4db386 | 4641 | /*dmq: *PL_reglastcloseparen = n; */ |
40a82448 DM |
4642 | scan = ST.next_branch; |
4643 | /* no more branches? */ | |
5d458dd8 YO |
4644 | if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) { |
4645 | DEBUG_EXECUTE_r({ | |
4646 | PerlIO_printf( Perl_debug_log, | |
4647 | "%*s %sBRANCH failed...%s\n", | |
4648 | REPORT_CODE_OFF+depth*2, "", | |
4649 | PL_colors[4], | |
4650 | PL_colors[5] ); | |
4651 | }); | |
4652 | sayNO_SILENT; | |
4653 | } | |
40a82448 DM |
4654 | continue; /* execute next BRANCH[J] op */ |
4655 | /* NOTREACHED */ | |
4656 | ||
a0d0e21e | 4657 | case MINMOD: |
24d3c4a9 | 4658 | minmod = 1; |
a0d0e21e | 4659 | break; |
40a82448 DM |
4660 | |
4661 | #undef ST | |
4662 | #define ST st->u.curlym | |
4663 | ||
4664 | case CURLYM: /* /A{m,n}B/ where A is fixed-length */ | |
4665 | ||
4666 | /* This is an optimisation of CURLYX that enables us to push | |
84d2fa14 | 4667 | * only a single backtracking state, no matter how many matches |
40a82448 DM |
4668 | * there are in {m,n}. It relies on the pattern being constant |
4669 | * length, with no parens to influence future backrefs | |
4670 | */ | |
4671 | ||
4672 | ST.me = scan; | |
dc45a647 | 4673 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; |
40a82448 DM |
4674 | |
4675 | /* if paren positive, emulate an OPEN/CLOSE around A */ | |
4676 | if (ST.me->flags) { | |
3b6647e0 | 4677 | U32 paren = ST.me->flags; |
40a82448 DM |
4678 | if (paren > PL_regsize) |
4679 | PL_regsize = paren; | |
3b6647e0 | 4680 | if (paren > *PL_reglastparen) |
40a82448 | 4681 | *PL_reglastparen = paren; |
c277df42 | 4682 | scan += NEXT_OFF(scan); /* Skip former OPEN. */ |
6407bf3b | 4683 | } |
40a82448 DM |
4684 | ST.A = scan; |
4685 | ST.B = next; | |
4686 | ST.alen = 0; | |
4687 | ST.count = 0; | |
24d3c4a9 DM |
4688 | ST.minmod = minmod; |
4689 | minmod = 0; | |
40a82448 DM |
4690 | ST.c1 = CHRTEST_UNINIT; |
4691 | REGCP_SET(ST.cp); | |
6407bf3b | 4692 | |
40a82448 DM |
4693 | if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */ |
4694 | goto curlym_do_B; | |
4695 | ||
4696 | curlym_do_A: /* execute the A in /A{m,n}B/ */ | |
6407bf3b | 4697 | PL_reginput = locinput; |
40a82448 DM |
4698 | PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */ |
4699 | /* NOTREACHED */ | |
5f80c4cf | 4700 | |
40a82448 DM |
4701 | case CURLYM_A: /* we've just matched an A */ |
4702 | locinput = st->locinput; | |
4703 | nextchr = UCHARAT(locinput); | |
4704 | ||
4705 | ST.count++; | |
4706 | /* after first match, determine A's length: u.curlym.alen */ | |
4707 | if (ST.count == 1) { | |
4708 | if (PL_reg_match_utf8) { | |
4709 | char *s = locinput; | |
4710 | while (s < PL_reginput) { | |
4711 | ST.alen++; | |
4712 | s += UTF8SKIP(s); | |
4713 | } | |
4714 | } | |
4715 | else { | |
4716 | ST.alen = PL_reginput - locinput; | |
4717 | } | |
4718 | if (ST.alen == 0) | |
4719 | ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me); | |
4720 | } | |
0cadcf80 DM |
4721 | DEBUG_EXECUTE_r( |
4722 | PerlIO_printf(Perl_debug_log, | |
40a82448 | 4723 | "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n", |
5bc10b2c | 4724 | (int)(REPORT_CODE_OFF+(depth*2)), "", |
40a82448 | 4725 | (IV) ST.count, (IV)ST.alen) |
0cadcf80 DM |
4726 | ); |
4727 | ||
40a82448 | 4728 | locinput = PL_reginput; |
0a4db386 YO |
4729 | |
4730 | if (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 4731 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
4732 | goto fake_end; |
4733 | ||
c966426a DM |
4734 | { |
4735 | I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)); | |
4736 | if ( max == REG_INFTY || ST.count < max ) | |
4737 | goto curlym_do_A; /* try to match another A */ | |
4738 | } | |
40a82448 | 4739 | goto curlym_do_B; /* try to match B */ |
5f80c4cf | 4740 | |
40a82448 DM |
4741 | case CURLYM_A_fail: /* just failed to match an A */ |
4742 | REGCP_UNWIND(ST.cp); | |
0a4db386 YO |
4743 | |
4744 | if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ | |
4745 | || (cur_eval && cur_eval->u.eval.close_paren && | |
24b23f37 | 4746 | cur_eval->u.eval.close_paren == (U32)ST.me->flags)) |
40a82448 | 4747 | sayNO; |
0cadcf80 | 4748 | |
40a82448 DM |
4749 | curlym_do_B: /* execute the B in /A{m,n}B/ */ |
4750 | PL_reginput = locinput; | |
4751 | if (ST.c1 == CHRTEST_UNINIT) { | |
4752 | /* calculate c1 and c2 for possible match of 1st char | |
4753 | * following curly */ | |
4754 | ST.c1 = ST.c2 = CHRTEST_VOID; | |
4755 | if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) { | |
4756 | regnode *text_node = ST.B; | |
4757 | if (! HAS_TEXT(text_node)) | |
4758 | FIND_NEXT_IMPT(text_node); | |
ee9b8eae YO |
4759 | /* this used to be |
4760 | ||
4761 | (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT) | |
4762 | ||
4763 | But the former is redundant in light of the latter. | |
4764 | ||
4765 | if this changes back then the macro for | |
4766 | IS_TEXT and friends need to change. | |
4767 | */ | |
4768 | if (PL_regkind[OP(text_node)] == EXACT) | |
40a82448 | 4769 | { |
ee9b8eae | 4770 | |
40a82448 DM |
4771 | ST.c1 = (U8)*STRING(text_node); |
4772 | ST.c2 = | |
ee9b8eae | 4773 | (IS_TEXTF(text_node)) |
40a82448 | 4774 | ? PL_fold[ST.c1] |
ee9b8eae | 4775 | : (IS_TEXTFL(text_node)) |
40a82448 DM |
4776 | ? PL_fold_locale[ST.c1] |
4777 | : ST.c1; | |
c277df42 | 4778 | } |
c277df42 | 4779 | } |
40a82448 DM |
4780 | } |
4781 | ||
4782 | DEBUG_EXECUTE_r( | |
4783 | PerlIO_printf(Perl_debug_log, | |
4784 | "%*s CURLYM trying tail with matches=%"IVdf"...\n", | |
5bc10b2c | 4785 | (int)(REPORT_CODE_OFF+(depth*2)), |
40a82448 DM |
4786 | "", (IV)ST.count) |
4787 | ); | |
4788 | if (ST.c1 != CHRTEST_VOID | |
4789 | && UCHARAT(PL_reginput) != ST.c1 | |
4790 | && UCHARAT(PL_reginput) != ST.c2) | |
4791 | { | |
4792 | /* simulate B failing */ | |
3e901dc0 YO |
4793 | DEBUG_OPTIMISE_r( |
4794 | PerlIO_printf(Perl_debug_log, | |
4795 | "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n", | |
4796 | (int)(REPORT_CODE_OFF+(depth*2)),"", | |
4797 | (IV)ST.c1,(IV)ST.c2 | |
4798 | )); | |
40a82448 DM |
4799 | state_num = CURLYM_B_fail; |
4800 | goto reenter_switch; | |
4801 | } | |
4802 | ||
4803 | if (ST.me->flags) { | |
4804 | /* mark current A as captured */ | |
4805 | I32 paren = ST.me->flags; | |
4806 | if (ST.count) { | |
f0ab9afb | 4807 | PL_regoffs[paren].start |
40a82448 | 4808 | = HOPc(PL_reginput, -ST.alen) - PL_bostr; |
f0ab9afb | 4809 | PL_regoffs[paren].end = PL_reginput - PL_bostr; |
0a4db386 | 4810 | /*dmq: *PL_reglastcloseparen = paren; */ |
c277df42 | 4811 | } |
40a82448 | 4812 | else |
f0ab9afb | 4813 | PL_regoffs[paren].end = -1; |
0a4db386 | 4814 | if (cur_eval && cur_eval->u.eval.close_paren && |
24b23f37 | 4815 | cur_eval->u.eval.close_paren == (U32)ST.me->flags) |
0a4db386 YO |
4816 | { |
4817 | if (ST.count) | |
4818 | goto fake_end; | |
4819 | else | |
4820 | sayNO; | |
4821 | } | |
c277df42 | 4822 | } |
0a4db386 | 4823 | |
40a82448 | 4824 | PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */ |
5f66b61c | 4825 | /* NOTREACHED */ |
40a82448 DM |
4826 | |
4827 | case CURLYM_B_fail: /* just failed to match a B */ | |
4828 | REGCP_UNWIND(ST.cp); | |
4829 | if (ST.minmod) { | |
84d2fa14 HS |
4830 | I32 max = ARG2(ST.me); |
4831 | if (max != REG_INFTY && ST.count == max) | |
40a82448 DM |
4832 | sayNO; |
4833 | goto curlym_do_A; /* try to match a further A */ | |
4834 | } | |
4835 | /* backtrack one A */ | |
4836 | if (ST.count == ARG1(ST.me) /* min */) | |
4837 | sayNO; | |
4838 | ST.count--; | |
4839 | locinput = HOPc(locinput, -ST.alen); | |
4840 | goto curlym_do_B; /* try to match B */ | |
4841 | ||
c255a977 DM |
4842 | #undef ST |
4843 | #define ST st->u.curly | |
40a82448 | 4844 | |
c255a977 DM |
4845 | #define CURLY_SETPAREN(paren, success) \ |
4846 | if (paren) { \ | |
4847 | if (success) { \ | |
f0ab9afb NC |
4848 | PL_regoffs[paren].start = HOPc(locinput, -1) - PL_bostr; \ |
4849 | PL_regoffs[paren].end = locinput - PL_bostr; \ | |
0a4db386 | 4850 | *PL_reglastcloseparen = paren; \ |
c255a977 DM |
4851 | } \ |
4852 | else \ | |
f0ab9afb | 4853 | PL_regoffs[paren].end = -1; \ |
c255a977 DM |
4854 | } |
4855 | ||
4856 | case STAR: /* /A*B/ where A is width 1 */ | |
4857 | ST.paren = 0; | |
4858 | ST.min = 0; | |
4859 | ST.max = REG_INFTY; | |
a0d0e21e LW |
4860 | scan = NEXTOPER(scan); |
4861 | goto repeat; | |
c255a977 DM |
4862 | case PLUS: /* /A+B/ where A is width 1 */ |
4863 | ST.paren = 0; | |
4864 | ST.min = 1; | |
4865 | ST.max = REG_INFTY; | |
c277df42 | 4866 | scan = NEXTOPER(scan); |
c255a977 DM |
4867 | goto repeat; |
4868 | case CURLYN: /* /(A){m,n}B/ where A is width 1 */ | |
4869 | ST.paren = scan->flags; /* Which paren to set */ | |
4870 | if (ST.paren > PL_regsize) | |
4871 | PL_regsize = ST.paren; | |
3b6647e0 | 4872 | if (ST.paren > *PL_reglastparen) |
c255a977 DM |
4873 | *PL_reglastparen = ST.paren; |
4874 | ST.min = ARG1(scan); /* min to match */ | |
4875 | ST.max = ARG2(scan); /* max to match */ | |
0a4db386 | 4876 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 4877 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
4878 | ST.min=1; |
4879 | ST.max=1; | |
4880 | } | |
c255a977 DM |
4881 | scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); |
4882 | goto repeat; | |
4883 | case CURLY: /* /A{m,n}B/ where A is width 1 */ | |
4884 | ST.paren = 0; | |
4885 | ST.min = ARG1(scan); /* min to match */ | |
4886 | ST.max = ARG2(scan); /* max to match */ | |
4887 | scan = NEXTOPER(scan) + NODE_STEP_REGNODE; | |
c277df42 | 4888 | repeat: |
a0d0e21e LW |
4889 | /* |
4890 | * Lookahead to avoid useless match attempts | |
4891 | * when we know what character comes next. | |
c255a977 | 4892 | * |
5f80c4cf JP |
4893 | * Used to only do .*x and .*?x, but now it allows |
4894 | * for )'s, ('s and (?{ ... })'s to be in the way | |
4895 | * of the quantifier and the EXACT-like node. -- japhy | |
4896 | */ | |
4897 | ||
c255a977 DM |
4898 | if (ST.min > ST.max) /* XXX make this a compile-time check? */ |
4899 | sayNO; | |
cca55fe3 | 4900 | if (HAS_TEXT(next) || JUMPABLE(next)) { |
5f80c4cf JP |
4901 | U8 *s; |
4902 | regnode *text_node = next; | |
4903 | ||
3dab1dad YO |
4904 | if (! HAS_TEXT(text_node)) |
4905 | FIND_NEXT_IMPT(text_node); | |
5f80c4cf | 4906 | |
9e137952 | 4907 | if (! HAS_TEXT(text_node)) |
c255a977 | 4908 | ST.c1 = ST.c2 = CHRTEST_VOID; |
5f80c4cf | 4909 | else { |
ee9b8eae | 4910 | if ( PL_regkind[OP(text_node)] != EXACT ) { |
c255a977 | 4911 | ST.c1 = ST.c2 = CHRTEST_VOID; |
44a68960 | 4912 | goto assume_ok_easy; |
cca55fe3 | 4913 | } |
be8e71aa YO |
4914 | else |
4915 | s = (U8*)STRING(text_node); | |
ee9b8eae YO |
4916 | |
4917 | /* Currently we only get here when | |
4918 | ||
4919 | PL_rekind[OP(text_node)] == EXACT | |
4920 | ||
4921 | if this changes back then the macro for IS_TEXT and | |
4922 | friends need to change. */ | |
f2ed9b32 | 4923 | if (!UTF_PATTERN) { |
c255a977 | 4924 | ST.c2 = ST.c1 = *s; |
ee9b8eae | 4925 | if (IS_TEXTF(text_node)) |
c255a977 | 4926 | ST.c2 = PL_fold[ST.c1]; |
ee9b8eae | 4927 | else if (IS_TEXTFL(text_node)) |
c255a977 | 4928 | ST.c2 = PL_fold_locale[ST.c1]; |
1aa99e6b | 4929 | } |
f2ed9b32 | 4930 | else { /* UTF_PATTERN */ |
ee9b8eae | 4931 | if (IS_TEXTF(text_node)) { |
a2a2844f | 4932 | STRLEN ulen1, ulen2; |
89ebb4a3 JH |
4933 | U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; |
4934 | U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; | |
a2a2844f JH |
4935 | |
4936 | to_utf8_lower((U8*)s, tmpbuf1, &ulen1); | |
4937 | to_utf8_upper((U8*)s, tmpbuf2, &ulen2); | |
e294cc5d JH |
4938 | #ifdef EBCDIC |
4939 | ST.c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN, 0, | |
4940 | ckWARN(WARN_UTF8) ? | |
4941 | 0 : UTF8_ALLOW_ANY); | |
4942 | ST.c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN, 0, | |
4943 | ckWARN(WARN_UTF8) ? | |
4944 | 0 : UTF8_ALLOW_ANY); | |
4945 | #else | |
c255a977 | 4946 | ST.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0, |
e294cc5d | 4947 | uniflags); |
c255a977 | 4948 | ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0, |
e294cc5d JH |
4949 | uniflags); |
4950 | #endif | |
5f80c4cf JP |
4951 | } |
4952 | else { | |
c255a977 | 4953 | ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, |
041457d9 | 4954 | uniflags); |
5f80c4cf | 4955 | } |
1aa99e6b IH |
4956 | } |
4957 | } | |
bbce6d69 | 4958 | } |
a0d0e21e | 4959 | else |
c255a977 | 4960 | ST.c1 = ST.c2 = CHRTEST_VOID; |
cca55fe3 | 4961 | assume_ok_easy: |
c255a977 DM |
4962 | |
4963 | ST.A = scan; | |
4964 | ST.B = next; | |
3280af22 | 4965 | PL_reginput = locinput; |
24d3c4a9 DM |
4966 | if (minmod) { |
4967 | minmod = 0; | |
e2e6a0f1 | 4968 | if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min) |
4633a7c4 | 4969 | sayNO; |
c255a977 | 4970 | ST.count = ST.min; |
a0ed51b3 | 4971 | locinput = PL_reginput; |
c255a977 DM |
4972 | REGCP_SET(ST.cp); |
4973 | if (ST.c1 == CHRTEST_VOID) | |
4974 | goto curly_try_B_min; | |
4975 | ||
4976 | ST.oldloc = locinput; | |
4977 | ||
4978 | /* set ST.maxpos to the furthest point along the | |
4979 | * string that could possibly match */ | |
4980 | if (ST.max == REG_INFTY) { | |
4981 | ST.maxpos = PL_regeol - 1; | |
f2ed9b32 | 4982 | if (utf8_target) |
c255a977 DM |
4983 | while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos)) |
4984 | ST.maxpos--; | |
4985 | } | |
f2ed9b32 | 4986 | else if (utf8_target) { |
c255a977 DM |
4987 | int m = ST.max - ST.min; |
4988 | for (ST.maxpos = locinput; | |
4989 | m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--) | |
4990 | ST.maxpos += UTF8SKIP(ST.maxpos); | |
4991 | } | |
4992 | else { | |
4993 | ST.maxpos = locinput + ST.max - ST.min; | |
4994 | if (ST.maxpos >= PL_regeol) | |
4995 | ST.maxpos = PL_regeol - 1; | |
4996 | } | |
4997 | goto curly_try_B_min_known; | |
4998 | ||
4999 | } | |
5000 | else { | |
e2e6a0f1 | 5001 | ST.count = regrepeat(rex, ST.A, ST.max, depth); |
c255a977 DM |
5002 | locinput = PL_reginput; |
5003 | if (ST.count < ST.min) | |
5004 | sayNO; | |
5005 | if ((ST.count > ST.min) | |
5006 | && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL)) | |
5007 | { | |
5008 | /* A{m,n} must come at the end of the string, there's | |
5009 | * no point in backing off ... */ | |
5010 | ST.min = ST.count; | |
5011 | /* ...except that $ and \Z can match before *and* after | |
5012 | newline at the end. Consider "\n\n" =~ /\n+\Z\n/. | |
5013 | We may back off by one in this case. */ | |
5014 | if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS) | |
5015 | ST.min--; | |
5016 | } | |
5017 | REGCP_SET(ST.cp); | |
5018 | goto curly_try_B_max; | |
5019 | } | |
5020 | /* NOTREACHED */ | |
5021 | ||
5022 | ||
5023 | case CURLY_B_min_known_fail: | |
5024 | /* failed to find B in a non-greedy match where c1,c2 valid */ | |
5025 | if (ST.paren && ST.count) | |
f0ab9afb | 5026 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5027 | |
5028 | PL_reginput = locinput; /* Could be reset... */ | |
5029 | REGCP_UNWIND(ST.cp); | |
5030 | /* Couldn't or didn't -- move forward. */ | |
5031 | ST.oldloc = locinput; | |
f2ed9b32 | 5032 | if (utf8_target) |
c255a977 DM |
5033 | locinput += UTF8SKIP(locinput); |
5034 | else | |
5035 | locinput++; | |
5036 | ST.count++; | |
5037 | curly_try_B_min_known: | |
5038 | /* find the next place where 'B' could work, then call B */ | |
5039 | { | |
5040 | int n; | |
f2ed9b32 | 5041 | if (utf8_target) { |
c255a977 DM |
5042 | n = (ST.oldloc == locinput) ? 0 : 1; |
5043 | if (ST.c1 == ST.c2) { | |
5044 | STRLEN len; | |
5045 | /* set n to utf8_distance(oldloc, locinput) */ | |
5046 | while (locinput <= ST.maxpos && | |
5047 | utf8n_to_uvchr((U8*)locinput, | |
5048 | UTF8_MAXBYTES, &len, | |
5049 | uniflags) != (UV)ST.c1) { | |
5050 | locinput += len; | |
5051 | n++; | |
5052 | } | |
1aa99e6b IH |
5053 | } |
5054 | else { | |
c255a977 DM |
5055 | /* set n to utf8_distance(oldloc, locinput) */ |
5056 | while (locinput <= ST.maxpos) { | |
5057 | STRLEN len; | |
5058 | const UV c = utf8n_to_uvchr((U8*)locinput, | |
5059 | UTF8_MAXBYTES, &len, | |
5060 | uniflags); | |
5061 | if (c == (UV)ST.c1 || c == (UV)ST.c2) | |
5062 | break; | |
5063 | locinput += len; | |
5064 | n++; | |
1aa99e6b | 5065 | } |
0fe9bf95 IZ |
5066 | } |
5067 | } | |
c255a977 DM |
5068 | else { |
5069 | if (ST.c1 == ST.c2) { | |
5070 | while (locinput <= ST.maxpos && | |
5071 | UCHARAT(locinput) != ST.c1) | |
5072 | locinput++; | |
bbce6d69 | 5073 | } |
c255a977 DM |
5074 | else { |
5075 | while (locinput <= ST.maxpos | |
5076 | && UCHARAT(locinput) != ST.c1 | |
5077 | && UCHARAT(locinput) != ST.c2) | |
5078 | locinput++; | |
a0ed51b3 | 5079 | } |
c255a977 DM |
5080 | n = locinput - ST.oldloc; |
5081 | } | |
5082 | if (locinput > ST.maxpos) | |
5083 | sayNO; | |
5084 | /* PL_reginput == oldloc now */ | |
5085 | if (n) { | |
5086 | ST.count += n; | |
e2e6a0f1 | 5087 | if (regrepeat(rex, ST.A, n, depth) < n) |
4633a7c4 | 5088 | sayNO; |
a0d0e21e | 5089 | } |
c255a977 DM |
5090 | PL_reginput = locinput; |
5091 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5092 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5093 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5094 | goto fake_end; |
5095 | } | |
c255a977 | 5096 | PUSH_STATE_GOTO(CURLY_B_min_known, ST.B); |
a0d0e21e | 5097 | } |
c255a977 DM |
5098 | /* NOTREACHED */ |
5099 | ||
5100 | ||
5101 | case CURLY_B_min_fail: | |
5102 | /* failed to find B in a non-greedy match where c1,c2 invalid */ | |
5103 | if (ST.paren && ST.count) | |
f0ab9afb | 5104 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5105 | |
5106 | REGCP_UNWIND(ST.cp); | |
5107 | /* failed -- move forward one */ | |
5108 | PL_reginput = locinput; | |
e2e6a0f1 | 5109 | if (regrepeat(rex, ST.A, 1, depth)) { |
c255a977 | 5110 | ST.count++; |
a0ed51b3 | 5111 | locinput = PL_reginput; |
c255a977 DM |
5112 | if (ST.count <= ST.max || (ST.max == REG_INFTY && |
5113 | ST.count > 0)) /* count overflow ? */ | |
15272685 | 5114 | { |
c255a977 DM |
5115 | curly_try_B_min: |
5116 | CURLY_SETPAREN(ST.paren, ST.count); | |
0a4db386 | 5117 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5118 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
0a4db386 YO |
5119 | goto fake_end; |
5120 | } | |
c255a977 | 5121 | PUSH_STATE_GOTO(CURLY_B_min, ST.B); |
a0d0e21e LW |
5122 | } |
5123 | } | |
4633a7c4 | 5124 | sayNO; |
c255a977 DM |
5125 | /* NOTREACHED */ |
5126 | ||
5127 | ||
5128 | curly_try_B_max: | |
5129 | /* a successful greedy match: now try to match B */ | |
40d049e4 | 5130 | if (cur_eval && cur_eval->u.eval.close_paren && |
86413ec0 | 5131 | cur_eval->u.eval.close_paren == (U32)ST.paren) { |
40d049e4 YO |
5132 | goto fake_end; |
5133 | } | |
c255a977 DM |
5134 | { |
5135 | UV c = 0; | |
5136 | if (ST.c1 != CHRTEST_VOID) | |
f2ed9b32 | 5137 | c = utf8_target ? utf8n_to_uvchr((U8*)PL_reginput, |
c255a977 | 5138 | UTF8_MAXBYTES, 0, uniflags) |
466787eb | 5139 | : (UV) UCHARAT(PL_reginput); |
c255a977 DM |
5140 | /* If it could work, try it. */ |
5141 | if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) { | |
5142 | CURLY_SETPAREN(ST.paren, ST.count); | |
5143 | PUSH_STATE_GOTO(CURLY_B_max, ST.B); | |
5144 | /* NOTREACHED */ | |
5145 | } | |
5146 | } | |
5147 | /* FALL THROUGH */ | |
5148 | case CURLY_B_max_fail: | |
5149 | /* failed to find B in a greedy match */ | |
5150 | if (ST.paren && ST.count) | |
f0ab9afb | 5151 | PL_regoffs[ST.paren].end = -1; |
c255a977 DM |
5152 | |
5153 | REGCP_UNWIND(ST.cp); | |
5154 | /* back up. */ | |
5155 | if (--ST.count < ST.min) | |
5156 | sayNO; | |
5157 | PL_reginput = locinput = HOPc(locinput, -1); | |
5158 | goto curly_try_B_max; | |
5159 | ||
5160 | #undef ST | |
5161 | ||
a0d0e21e | 5162 | case END: |
6bda09f9 | 5163 | fake_end: |
faec1544 DM |
5164 | if (cur_eval) { |
5165 | /* we've just finished A in /(??{A})B/; now continue with B */ | |
5166 | I32 tmpix; | |
faec1544 DM |
5167 | st->u.eval.toggle_reg_flags |
5168 | = cur_eval->u.eval.toggle_reg_flags; | |
5169 | PL_reg_flags ^= st->u.eval.toggle_reg_flags; | |
5170 | ||
288b8c02 NC |
5171 | st->u.eval.prev_rex = rex_sv; /* inner */ |
5172 | SETREX(rex_sv,cur_eval->u.eval.prev_rex); | |
5173 | rex = (struct regexp *)SvANY(rex_sv); | |
f8fc2ecf | 5174 | rexi = RXi_GET(rex); |
faec1544 | 5175 | cur_curlyx = cur_eval->u.eval.prev_curlyx; |
288b8c02 | 5176 | ReREFCNT_inc(rex_sv); |
faec1544 | 5177 | st->u.eval.cp = regcppush(0); /* Save *all* the positions. */ |
34a81e2b B |
5178 | |
5179 | /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ | |
5180 | PL_reglastparen = &rex->lastparen; | |
5181 | PL_reglastcloseparen = &rex->lastcloseparen; | |
5182 | ||
faec1544 DM |
5183 | REGCP_SET(st->u.eval.lastcp); |
5184 | PL_reginput = locinput; | |
5185 | ||
5186 | /* Restore parens of the outer rex without popping the | |
5187 | * savestack */ | |
5188 | tmpix = PL_savestack_ix; | |
5189 | PL_savestack_ix = cur_eval->u.eval.lastcp; | |
5190 | regcppop(rex); | |
5191 | PL_savestack_ix = tmpix; | |
5192 | ||
5193 | st->u.eval.prev_eval = cur_eval; | |
5194 | cur_eval = cur_eval->u.eval.prev_eval; | |
5195 | DEBUG_EXECUTE_r( | |
2a49f0f5 JH |
5196 | PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n", |
5197 | REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval));); | |
e7707071 YO |
5198 | if ( nochange_depth ) |
5199 | nochange_depth--; | |
5200 | ||
5201 | PUSH_YES_STATE_GOTO(EVAL_AB, | |
faec1544 DM |
5202 | st->u.eval.prev_eval->u.eval.B); /* match B */ |
5203 | } | |
5204 | ||
3b0527fe | 5205 | if (locinput < reginfo->till) { |
a3621e74 | 5206 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, |
7821416a IZ |
5207 | "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n", |
5208 | PL_colors[4], | |
5209 | (long)(locinput - PL_reg_starttry), | |
3b0527fe | 5210 | (long)(reginfo->till - PL_reg_starttry), |
7821416a | 5211 | PL_colors[5])); |
58e23c8d | 5212 | |
262b90c4 | 5213 | sayNO_SILENT; /* Cannot match: too short. */ |
7821416a IZ |
5214 | } |
5215 | PL_reginput = locinput; /* put where regtry can find it */ | |
262b90c4 | 5216 | sayYES; /* Success! */ |
dad79028 DM |
5217 | |
5218 | case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */ | |
5219 | DEBUG_EXECUTE_r( | |
5220 | PerlIO_printf(Perl_debug_log, | |
5221 | "%*s %ssubpattern success...%s\n", | |
5bc10b2c | 5222 | REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])); |
3280af22 | 5223 | PL_reginput = locinput; /* put where regtry can find it */ |
262b90c4 | 5224 | sayYES; /* Success! */ |
dad79028 | 5225 | |
40a82448 DM |
5226 | #undef ST |
5227 | #define ST st->u.ifmatch | |
5228 | ||
5229 | case SUSPEND: /* (?>A) */ | |
5230 | ST.wanted = 1; | |
9fe1d20c | 5231 | PL_reginput = locinput; |
9041c2e3 | 5232 | goto do_ifmatch; |
dad79028 | 5233 | |
40a82448 DM |
5234 | case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */ |
5235 | ST.wanted = 0; | |
dad79028 DM |
5236 | goto ifmatch_trivial_fail_test; |
5237 | ||
40a82448 DM |
5238 | case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */ |
5239 | ST.wanted = 1; | |
dad79028 | 5240 | ifmatch_trivial_fail_test: |
a0ed51b3 | 5241 | if (scan->flags) { |
52657f30 | 5242 | char * const s = HOPBACKc(locinput, scan->flags); |
dad79028 DM |
5243 | if (!s) { |
5244 | /* trivial fail */ | |
24d3c4a9 DM |
5245 | if (logical) { |
5246 | logical = 0; | |
f2338a2e | 5247 | sw = 1 - cBOOL(ST.wanted); |
dad79028 | 5248 | } |
40a82448 | 5249 | else if (ST.wanted) |
dad79028 DM |
5250 | sayNO; |
5251 | next = scan + ARG(scan); | |
5252 | if (next == scan) | |
5253 | next = NULL; | |
5254 | break; | |
5255 | } | |
efb30f32 | 5256 | PL_reginput = s; |
a0ed51b3 LW |
5257 | } |
5258 | else | |
5259 | PL_reginput = locinput; | |
5260 | ||
c277df42 | 5261 | do_ifmatch: |
40a82448 | 5262 | ST.me = scan; |
24d3c4a9 | 5263 | ST.logical = logical; |
24d786f4 YO |
5264 | logical = 0; /* XXX: reset state of logical once it has been saved into ST */ |
5265 | ||
40a82448 DM |
5266 | /* execute body of (?...A) */ |
5267 | PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan))); | |
5268 | /* NOTREACHED */ | |
5269 | ||
5270 | case IFMATCH_A_fail: /* body of (?...A) failed */ | |
5271 | ST.wanted = !ST.wanted; | |
5272 | /* FALL THROUGH */ | |
5273 | ||
5274 | case IFMATCH_A: /* body of (?...A) succeeded */ | |
24d3c4a9 | 5275 | if (ST.logical) { |
f2338a2e | 5276 | sw = cBOOL(ST.wanted); |
40a82448 DM |
5277 | } |
5278 | else if (!ST.wanted) | |
5279 | sayNO; | |
5280 | ||
5281 | if (OP(ST.me) == SUSPEND) | |
5282 | locinput = PL_reginput; | |
5283 | else { | |
5284 | locinput = PL_reginput = st->locinput; | |
5285 | nextchr = UCHARAT(locinput); | |
5286 | } | |
5287 | scan = ST.me + ARG(ST.me); | |
5288 | if (scan == ST.me) | |
5289 | scan = NULL; | |
5290 | continue; /* execute B */ | |
5291 | ||
5292 | #undef ST | |
dad79028 | 5293 | |
c277df42 | 5294 | case LONGJMP: |
c277df42 IZ |
5295 | next = scan + ARG(scan); |
5296 | if (next == scan) | |
5297 | next = NULL; | |
a0d0e21e | 5298 | break; |
54612592 | 5299 | case COMMIT: |
e2e6a0f1 YO |
5300 | reginfo->cutpoint = PL_regeol; |
5301 | /* FALLTHROUGH */ | |
5d458dd8 | 5302 | case PRUNE: |
24b23f37 | 5303 | PL_reginput = locinput; |
e2e6a0f1 | 5304 | if (!scan->flags) |
ad64d0ec | 5305 | sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
54612592 YO |
5306 | PUSH_STATE_GOTO(COMMIT_next,next); |
5307 | /* NOTREACHED */ | |
5308 | case COMMIT_next_fail: | |
5309 | no_final = 1; | |
5310 | /* FALLTHROUGH */ | |
7f69552c YO |
5311 | case OPFAIL: |
5312 | sayNO; | |
e2e6a0f1 YO |
5313 | /* NOTREACHED */ |
5314 | ||
5315 | #define ST st->u.mark | |
5316 | case MARKPOINT: | |
5317 | ST.prev_mark = mark_state; | |
5d458dd8 | 5318 | ST.mark_name = sv_commit = sv_yes_mark |
ad64d0ec | 5319 | = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
e2e6a0f1 YO |
5320 | mark_state = st; |
5321 | ST.mark_loc = PL_reginput = locinput; | |
5322 | PUSH_YES_STATE_GOTO(MARKPOINT_next,next); | |
5323 | /* NOTREACHED */ | |
5324 | case MARKPOINT_next: | |
5325 | mark_state = ST.prev_mark; | |
5326 | sayYES; | |
5327 | /* NOTREACHED */ | |
5328 | case MARKPOINT_next_fail: | |
5d458dd8 | 5329 | if (popmark && sv_eq(ST.mark_name,popmark)) |
e2e6a0f1 YO |
5330 | { |
5331 | if (ST.mark_loc > startpoint) | |
5332 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5333 | popmark = NULL; /* we found our mark */ | |
5334 | sv_commit = ST.mark_name; | |
5335 | ||
5336 | DEBUG_EXECUTE_r({ | |
5d458dd8 | 5337 | PerlIO_printf(Perl_debug_log, |
e2e6a0f1 YO |
5338 | "%*s %ssetting cutpoint to mark:%"SVf"...%s\n", |
5339 | REPORT_CODE_OFF+depth*2, "", | |
be2597df | 5340 | PL_colors[4], SVfARG(sv_commit), PL_colors[5]); |
e2e6a0f1 YO |
5341 | }); |
5342 | } | |
5343 | mark_state = ST.prev_mark; | |
5d458dd8 YO |
5344 | sv_yes_mark = mark_state ? |
5345 | mark_state->u.mark.mark_name : NULL; | |
e2e6a0f1 YO |
5346 | sayNO; |
5347 | /* NOTREACHED */ | |
5d458dd8 YO |
5348 | case SKIP: |
5349 | PL_reginput = locinput; | |
5350 | if (scan->flags) { | |
2bf803e2 | 5351 | /* (*SKIP) : if we fail we cut here*/ |
5d458dd8 | 5352 | ST.mark_name = NULL; |
e2e6a0f1 | 5353 | ST.mark_loc = locinput; |
5d458dd8 YO |
5354 | PUSH_STATE_GOTO(SKIP_next,next); |
5355 | } else { | |
2bf803e2 | 5356 | /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was, |
5d458dd8 YO |
5357 | otherwise do nothing. Meaning we need to scan |
5358 | */ | |
5359 | regmatch_state *cur = mark_state; | |
ad64d0ec | 5360 | SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); |
5d458dd8 YO |
5361 | |
5362 | while (cur) { | |
5363 | if ( sv_eq( cur->u.mark.mark_name, | |
5364 | find ) ) | |
5365 | { | |
5366 | ST.mark_name = find; | |
5367 | PUSH_STATE_GOTO( SKIP_next, next ); | |
5368 | } | |
5369 | cur = cur->u.mark.prev_mark; | |
5370 | } | |
e2e6a0f1 | 5371 | } |
2bf803e2 | 5372 | /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */ |
5d458dd8 YO |
5373 | break; |
5374 | case SKIP_next_fail: | |
5375 | if (ST.mark_name) { | |
5376 | /* (*CUT:NAME) - Set up to search for the name as we | |
5377 | collapse the stack*/ | |
5378 | popmark = ST.mark_name; | |
5379 | } else { | |
5380 | /* (*CUT) - No name, we cut here.*/ | |
e2e6a0f1 YO |
5381 | if (ST.mark_loc > startpoint) |
5382 | reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1); | |
5d458dd8 YO |
5383 | /* but we set sv_commit to latest mark_name if there |
5384 | is one so they can test to see how things lead to this | |
5385 | cut */ | |
5386 | if (mark_state) | |
5387 | sv_commit=mark_state->u.mark.mark_name; | |
5388 | } | |
e2e6a0f1 YO |
5389 | no_final = 1; |
5390 | sayNO; | |
5391 | /* NOTREACHED */ | |
5392 | #undef ST | |
32e6a07c YO |
5393 | case FOLDCHAR: |
5394 | n = ARG(scan); | |
f2ed9b32 | 5395 | if ( n == (U32)what_len_TRICKYFOLD(locinput,utf8_target,ln) ) { |
e64b1bd1 | 5396 | locinput += ln; |
f2ed9b32 | 5397 | } else if ( 0xDF == n && !utf8_target && !UTF_PATTERN ) { |
e64b1bd1 YO |
5398 | sayNO; |
5399 | } else { | |
5400 | U8 folded[UTF8_MAXBYTES_CASE+1]; | |
5401 | STRLEN foldlen; | |
5402 | const char * const l = locinput; | |
5403 | char *e = PL_regeol; | |
5404 | to_uni_fold(n, folded, &foldlen); | |
5405 | ||
4c1b470c | 5406 | if (! foldEQ_utf8((const char*) folded, 0, foldlen, 1, |
f2ed9b32 | 5407 | l, &e, 0, utf8_target)) { |
32e6a07c | 5408 | sayNO; |
e64b1bd1 YO |
5409 | } |
5410 | locinput = e; | |
32e6a07c YO |
5411 | } |
5412 | nextchr = UCHARAT(locinput); | |
5413 | break; | |
e1d1eefb | 5414 | case LNBREAK: |
f2ed9b32 | 5415 | if ((n=is_LNBREAK(locinput,utf8_target))) { |
e1d1eefb YO |
5416 | locinput += n; |
5417 | nextchr = UCHARAT(locinput); | |
5418 | } else | |
5419 | sayNO; | |
5420 | break; | |
5421 | ||
5422 | #define CASE_CLASS(nAmE) \ | |
5423 | case nAmE: \ | |
f2ed9b32 | 5424 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5425 | locinput += n; \ |
5426 | nextchr = UCHARAT(locinput); \ | |
5427 | } else \ | |
5428 | sayNO; \ | |
5429 | break; \ | |
5430 | case N##nAmE: \ | |
f2ed9b32 | 5431 | if ((n=is_##nAmE(locinput,utf8_target))) { \ |
e1d1eefb YO |
5432 | sayNO; \ |
5433 | } else { \ | |
5434 | locinput += UTF8SKIP(locinput); \ | |
5435 | nextchr = UCHARAT(locinput); \ | |
5436 | } \ | |
5437 | break | |
5438 | ||
5439 | CASE_CLASS(VERTWS); | |
5440 | CASE_CLASS(HORIZWS); | |
5441 | #undef CASE_CLASS | |
5442 | ||
a0d0e21e | 5443 | default: |
b900a521 | 5444 | PerlIO_printf(Perl_error_log, "%"UVxf" %d\n", |
d7d93a81 | 5445 | PTR2UV(scan), OP(scan)); |
cea2e8a9 | 5446 | Perl_croak(aTHX_ "regexp memory corruption"); |
5d458dd8 YO |
5447 | |
5448 | } /* end switch */ | |
95b24440 | 5449 | |
5d458dd8 YO |
5450 | /* switch break jumps here */ |
5451 | scan = next; /* prepare to execute the next op and ... */ | |
5452 | continue; /* ... jump back to the top, reusing st */ | |
95b24440 DM |
5453 | /* NOTREACHED */ |
5454 | ||
40a82448 DM |
5455 | push_yes_state: |
5456 | /* push a state that backtracks on success */ | |
5457 | st->u.yes.prev_yes_state = yes_state; | |
5458 | yes_state = st; | |
5459 | /* FALL THROUGH */ | |
5460 | push_state: | |
5461 | /* push a new regex state, then continue at scan */ | |
5462 | { | |
5463 | regmatch_state *newst; | |
5464 | ||
24b23f37 YO |
5465 | DEBUG_STACK_r({ |
5466 | regmatch_state *cur = st; | |
5467 | regmatch_state *curyes = yes_state; | |
5468 | int curd = depth; | |
5469 | regmatch_slab *slab = PL_regmatch_slab; | |
5470 | for (;curd > -1;cur--,curd--) { | |
5471 | if (cur < SLAB_FIRST(slab)) { | |
5472 | slab = slab->prev; | |
5473 | cur = SLAB_LAST(slab); | |
5474 | } | |
5475 | PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n", | |
5476 | REPORT_CODE_OFF + 2 + depth * 2,"", | |
13d6edb4 | 5477 | curd, PL_reg_name[cur->resume_state], |
24b23f37 YO |
5478 | (curyes == cur) ? "yes" : "" |
5479 | ); | |
5480 | if (curyes == cur) | |
5481 | curyes = cur->u.yes.prev_yes_state; | |
5482 | } | |
5483 | } else | |
5484 | DEBUG_STATE_pp("push") | |
5485 | ); | |
40a82448 | 5486 | depth++; |
40a82448 DM |
5487 | st->locinput = locinput; |
5488 | newst = st+1; | |
5489 | if (newst > SLAB_LAST(PL_regmatch_slab)) | |
5490 | newst = S_push_slab(aTHX); | |
5491 | PL_regmatch_state = newst; | |
786e8c11 | 5492 | |
40a82448 DM |
5493 | locinput = PL_reginput; |
5494 | nextchr = UCHARAT(locinput); | |
5495 | st = newst; | |
5496 | continue; | |
5497 | /* NOTREACHED */ | |
5498 | } | |
a0d0e21e | 5499 | } |
a687059c | 5500 | |
a0d0e21e LW |
5501 | /* |
5502 | * We get here only if there's trouble -- normally "case END" is | |
5503 | * the terminating point. | |
5504 | */ | |
cea2e8a9 | 5505 | Perl_croak(aTHX_ "corrupted regexp pointers"); |
a0d0e21e | 5506 | /*NOTREACHED*/ |
4633a7c4 LW |
5507 | sayNO; |
5508 | ||
262b90c4 | 5509 | yes: |
77cb431f DM |
5510 | if (yes_state) { |
5511 | /* we have successfully completed a subexpression, but we must now | |
5512 | * pop to the state marked by yes_state and continue from there */ | |
77cb431f | 5513 | assert(st != yes_state); |
5bc10b2c DM |
5514 | #ifdef DEBUGGING |
5515 | while (st != yes_state) { | |
5516 | st--; | |
5517 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
5518 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5519 | st = SLAB_LAST(PL_regmatch_slab); | |
5520 | } | |
e2e6a0f1 | 5521 | DEBUG_STATE_r({ |
54612592 YO |
5522 | if (no_final) { |
5523 | DEBUG_STATE_pp("pop (no final)"); | |
5524 | } else { | |
5525 | DEBUG_STATE_pp("pop (yes)"); | |
5526 | } | |
e2e6a0f1 | 5527 | }); |
5bc10b2c DM |
5528 | depth--; |
5529 | } | |
5530 | #else | |
77cb431f DM |
5531 | while (yes_state < SLAB_FIRST(PL_regmatch_slab) |
5532 | || yes_state > SLAB_LAST(PL_regmatch_slab)) | |
5533 | { | |
5534 | /* not in this slab, pop slab */ | |
5535 | depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1); | |
5536 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5537 | st = SLAB_LAST(PL_regmatch_slab); | |
5538 | } | |
5539 | depth -= (st - yes_state); | |
5bc10b2c | 5540 | #endif |
77cb431f DM |
5541 | st = yes_state; |
5542 | yes_state = st->u.yes.prev_yes_state; | |
5543 | PL_regmatch_state = st; | |
24b23f37 | 5544 | |
5d458dd8 YO |
5545 | if (no_final) { |
5546 | locinput= st->locinput; | |
5547 | nextchr = UCHARAT(locinput); | |
5548 | } | |
54612592 | 5549 | state_num = st->resume_state + no_final; |
24d3c4a9 | 5550 | goto reenter_switch; |
77cb431f DM |
5551 | } |
5552 | ||
a3621e74 | 5553 | DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", |
e4584336 | 5554 | PL_colors[4], PL_colors[5])); |
02db2b7b | 5555 | |
19b95bf0 DM |
5556 | if (PL_reg_eval_set) { |
5557 | /* each successfully executed (?{...}) block does the equivalent of | |
5558 | * local $^R = do {...} | |
5559 | * When popping the save stack, all these locals would be undone; | |
5560 | * bypass this by setting the outermost saved $^R to the latest | |
5561 | * value */ | |
5562 | if (oreplsv != GvSV(PL_replgv)) | |
5563 | sv_setsv(oreplsv, GvSV(PL_replgv)); | |
5564 | } | |
95b24440 | 5565 | result = 1; |
aa283a38 | 5566 | goto final_exit; |
4633a7c4 LW |
5567 | |
5568 | no: | |
a3621e74 | 5569 | DEBUG_EXECUTE_r( |
7821416a | 5570 | PerlIO_printf(Perl_debug_log, |
786e8c11 | 5571 | "%*s %sfailed...%s\n", |
5bc10b2c | 5572 | REPORT_CODE_OFF+depth*2, "", |
786e8c11 | 5573 | PL_colors[4], PL_colors[5]) |
7821416a | 5574 | ); |
aa283a38 | 5575 | |
262b90c4 | 5576 | no_silent: |
54612592 YO |
5577 | if (no_final) { |
5578 | if (yes_state) { | |
5579 | goto yes; | |
5580 | } else { | |
5581 | goto final_exit; | |
5582 | } | |
5583 | } | |
aa283a38 DM |
5584 | if (depth) { |
5585 | /* there's a previous state to backtrack to */ | |
40a82448 DM |
5586 | st--; |
5587 | if (st < SLAB_FIRST(PL_regmatch_slab)) { | |
5588 | PL_regmatch_slab = PL_regmatch_slab->prev; | |
5589 | st = SLAB_LAST(PL_regmatch_slab); | |
5590 | } | |
5591 | PL_regmatch_state = st; | |
40a82448 DM |
5592 | locinput= st->locinput; |
5593 | nextchr = UCHARAT(locinput); | |
5594 | ||
5bc10b2c DM |
5595 | DEBUG_STATE_pp("pop"); |
5596 | depth--; | |
262b90c4 DM |
5597 | if (yes_state == st) |
5598 | yes_state = st->u.yes.prev_yes_state; | |
5bc10b2c | 5599 | |
24d3c4a9 DM |
5600 | state_num = st->resume_state + 1; /* failure = success + 1 */ |
5601 | goto reenter_switch; | |
95b24440 | 5602 | } |
24d3c4a9 | 5603 | result = 0; |
aa283a38 | 5604 | |
262b90c4 | 5605 | final_exit: |
bbe252da | 5606 | if (rex->intflags & PREGf_VERBARG_SEEN) { |
5d458dd8 YO |
5607 | SV *sv_err = get_sv("REGERROR", 1); |
5608 | SV *sv_mrk = get_sv("REGMARK", 1); | |
5609 | if (result) { | |
e2e6a0f1 | 5610 | sv_commit = &PL_sv_no; |
5d458dd8 YO |
5611 | if (!sv_yes_mark) |
5612 | sv_yes_mark = &PL_sv_yes; | |
5613 | } else { | |
5614 | if (!sv_commit) | |
5615 | sv_commit = &PL_sv_yes; | |
5616 | sv_yes_mark = &PL_sv_no; | |
5617 | } | |
5618 | sv_setsv(sv_err, sv_commit); | |
5619 | sv_setsv(sv_mrk, sv_yes_mark); | |
e2e6a0f1 | 5620 | } |
19b95bf0 | 5621 | |
2f554ef7 DM |
5622 | /* clean up; in particular, free all slabs above current one */ |
5623 | LEAVE_SCOPE(oldsave); | |
5d9a96ca | 5624 | |
95b24440 | 5625 | return result; |
a687059c LW |
5626 | } |
5627 | ||
5628 | /* | |
5629 | - regrepeat - repeatedly match something simple, report how many | |
5630 | */ | |
5631 | /* | |
5632 | * [This routine now assumes that it will only match on things of length 1. | |
5633 | * That was true before, but now we assume scan - reginput is the count, | |
a0ed51b3 | 5634 | * rather than incrementing count on every character. [Er, except utf8.]] |
a687059c | 5635 | */ |
76e3520e | 5636 | STATIC I32 |
e2e6a0f1 | 5637 | S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth) |
a687059c | 5638 | { |
27da23d5 | 5639 | dVAR; |
a0d0e21e | 5640 | register char *scan; |
a0d0e21e | 5641 | register I32 c; |
3280af22 | 5642 | register char *loceol = PL_regeol; |
a0ed51b3 | 5643 | register I32 hardcount = 0; |
f2ed9b32 | 5644 | register bool utf8_target = PL_reg_match_utf8; |
4f55667c SP |
5645 | #ifndef DEBUGGING |
5646 | PERL_UNUSED_ARG(depth); | |
5647 | #endif | |
a0d0e21e | 5648 | |
7918f24d NC |
5649 | PERL_ARGS_ASSERT_REGREPEAT; |
5650 | ||
3280af22 | 5651 | scan = PL_reginput; |
faf11cac HS |
5652 | if (max == REG_INFTY) |
5653 | max = I32_MAX; | |
5654 | else if (max < loceol - scan) | |
7f596f4c | 5655 | loceol = scan + max; |
a0d0e21e | 5656 | switch (OP(p)) { |
22c35a8c | 5657 | case REG_ANY: |
f2ed9b32 | 5658 | if (utf8_target) { |
ffc61ed2 | 5659 | loceol = PL_regeol; |
1aa99e6b | 5660 | while (scan < loceol && hardcount < max && *scan != '\n') { |
ffc61ed2 JH |
5661 | scan += UTF8SKIP(scan); |
5662 | hardcount++; | |
5663 | } | |
5664 | } else { | |
5665 | while (scan < loceol && *scan != '\n') | |
5666 | scan++; | |
a0ed51b3 LW |
5667 | } |
5668 | break; | |
ffc61ed2 | 5669 | case SANY: |
f2ed9b32 | 5670 | if (utf8_target) { |
def8e4ea | 5671 | loceol = PL_regeol; |
a0804c9e | 5672 | while (scan < loceol && hardcount < max) { |
def8e4ea JH |
5673 | scan += UTF8SKIP(scan); |
5674 | hardcount++; | |
5675 | } | |
5676 | } | |
5677 | else | |
5678 | scan = loceol; | |
a0ed51b3 | 5679 | break; |
f33976b4 DB |
5680 | case CANY: |
5681 | scan = loceol; | |
5682 | break; | |
090f7165 JH |
5683 | case EXACT: /* length of string is 1 */ |
5684 | c = (U8)*STRING(p); | |
5685 | while (scan < loceol && UCHARAT(scan) == c) | |
5686 | scan++; | |
bbce6d69 | 5687 | break; |
5688 | case EXACTF: /* length of string is 1 */ | |
cd439c50 | 5689 | c = (U8)*STRING(p); |
bbce6d69 | 5690 | while (scan < loceol && |
22c35a8c | 5691 | (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c])) |
bbce6d69 | 5692 | scan++; |
5693 | break; | |
5694 | case EXACTFL: /* length of string is 1 */ | |
3280af22 | 5695 | PL_reg_flags |= RF_tainted; |
cd439c50 | 5696 | c = (U8)*STRING(p); |
bbce6d69 | 5697 | while (scan < loceol && |
22c35a8c | 5698 | (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c])) |
a0d0e21e LW |
5699 | scan++; |
5700 | break; | |
5701 | case ANYOF: | |
f2ed9b32 | 5702 | if (utf8_target) { |
ffc61ed2 | 5703 | loceol = PL_regeol; |
cfc92286 | 5704 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 5705 | reginclass(prog, p, (U8*)scan, 0, utf8_target)) { |
ffc61ed2 JH |
5706 | scan += UTF8SKIP(scan); |
5707 | hardcount++; | |
5708 | } | |
5709 | } else { | |
32fc9b6a | 5710 | while (scan < loceol && REGINCLASS(prog, p, (U8*)scan)) |
ffc61ed2 JH |
5711 | scan++; |
5712 | } | |
a0d0e21e LW |
5713 | break; |
5714 | case ALNUM: | |
f2ed9b32 | 5715 | if (utf8_target) { |
ffc61ed2 | 5716 | loceol = PL_regeol; |
1a4fad37 | 5717 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 5718 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 5719 | swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
5720 | scan += UTF8SKIP(scan); |
5721 | hardcount++; | |
5722 | } | |
5723 | } else { | |
5724 | while (scan < loceol && isALNUM(*scan)) | |
5725 | scan++; | |
a0ed51b3 LW |
5726 | } |
5727 | break; | |
bbce6d69 | 5728 | case ALNUML: |
3280af22 | 5729 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 5730 | if (utf8_target) { |
ffc61ed2 | 5731 | loceol = PL_regeol; |
1aa99e6b IH |
5732 | while (hardcount < max && scan < loceol && |
5733 | isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
5734 | scan += UTF8SKIP(scan); |
5735 | hardcount++; | |
5736 | } | |
5737 | } else { | |
5738 | while (scan < loceol && isALNUM_LC(*scan)) | |
5739 | scan++; | |
a0ed51b3 LW |
5740 | } |
5741 | break; | |
a0d0e21e | 5742 | case NALNUM: |
f2ed9b32 | 5743 | if (utf8_target) { |
ffc61ed2 | 5744 | loceol = PL_regeol; |
1a4fad37 | 5745 | LOAD_UTF8_CHARCLASS_ALNUM(); |
1aa99e6b | 5746 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 5747 | !swash_fetch(PL_utf8_alnum, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
5748 | scan += UTF8SKIP(scan); |
5749 | hardcount++; | |
5750 | } | |
5751 | } else { | |
5752 | while (scan < loceol && !isALNUM(*scan)) | |
5753 | scan++; | |
a0ed51b3 LW |
5754 | } |
5755 | break; | |
bbce6d69 | 5756 | case NALNUML: |
3280af22 | 5757 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 5758 | if (utf8_target) { |
ffc61ed2 | 5759 | loceol = PL_regeol; |
1aa99e6b IH |
5760 | while (hardcount < max && scan < loceol && |
5761 | !isALNUM_LC_utf8((U8*)scan)) { | |
ffc61ed2 JH |
5762 | scan += UTF8SKIP(scan); |
5763 | hardcount++; | |
5764 | } | |
5765 | } else { | |
5766 | while (scan < loceol && !isALNUM_LC(*scan)) | |
5767 | scan++; | |
a0ed51b3 LW |
5768 | } |
5769 | break; | |
a0d0e21e | 5770 | case SPACE: |
f2ed9b32 | 5771 | if (utf8_target) { |
ffc61ed2 | 5772 | loceol = PL_regeol; |
1a4fad37 | 5773 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 5774 | while (hardcount < max && scan < loceol && |
3568d838 | 5775 | (*scan == ' ' || |
f2ed9b32 | 5776 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) { |
ffc61ed2 JH |
5777 | scan += UTF8SKIP(scan); |
5778 | hardcount++; | |
5779 | } | |
5780 | } else { | |
5781 | while (scan < loceol && isSPACE(*scan)) | |
5782 | scan++; | |
a0ed51b3 LW |
5783 | } |
5784 | break; | |
bbce6d69 | 5785 | case SPACEL: |
3280af22 | 5786 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 5787 | if (utf8_target) { |
ffc61ed2 | 5788 | loceol = PL_regeol; |
1aa99e6b | 5789 | while (hardcount < max && scan < loceol && |
ffc61ed2 JH |
5790 | (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) { |
5791 | scan += UTF8SKIP(scan); | |
5792 | hardcount++; | |
5793 | } | |
5794 | } else { | |
5795 | while (scan < loceol && isSPACE_LC(*scan)) | |
5796 | scan++; | |
a0ed51b3 LW |
5797 | } |
5798 | break; | |
a0d0e21e | 5799 | case NSPACE: |
f2ed9b32 | 5800 | if (utf8_target) { |
ffc61ed2 | 5801 | loceol = PL_regeol; |
1a4fad37 | 5802 | LOAD_UTF8_CHARCLASS_SPACE(); |
1aa99e6b | 5803 | while (hardcount < max && scan < loceol && |
3568d838 | 5804 | !(*scan == ' ' || |
f2ed9b32 | 5805 | swash_fetch(PL_utf8_space,(U8*)scan, utf8_target))) { |
ffc61ed2 JH |
5806 | scan += UTF8SKIP(scan); |
5807 | hardcount++; | |
5808 | } | |
5809 | } else { | |
5810 | while (scan < loceol && !isSPACE(*scan)) | |
5811 | scan++; | |
a0ed51b3 | 5812 | } |
0008a298 | 5813 | break; |
bbce6d69 | 5814 | case NSPACEL: |
3280af22 | 5815 | PL_reg_flags |= RF_tainted; |
f2ed9b32 | 5816 | if (utf8_target) { |
ffc61ed2 | 5817 | loceol = PL_regeol; |
1aa99e6b | 5818 | while (hardcount < max && scan < loceol && |
ffc61ed2 JH |
5819 | !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) { |
5820 | scan += UTF8SKIP(scan); | |
5821 | hardcount++; | |
5822 | } | |
5823 | } else { | |
5824 | while (scan < loceol && !isSPACE_LC(*scan)) | |
5825 | scan++; | |
a0ed51b3 LW |
5826 | } |
5827 | break; | |
a0d0e21e | 5828 | case DIGIT: |
f2ed9b32 | 5829 | if (utf8_target) { |
ffc61ed2 | 5830 | loceol = PL_regeol; |
1a4fad37 | 5831 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 5832 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 5833 | swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
5834 | scan += UTF8SKIP(scan); |
5835 | hardcount++; | |
5836 | } | |
5837 | } else { | |
5838 | while (scan < loceol && isDIGIT(*scan)) | |
5839 | scan++; | |
a0ed51b3 LW |
5840 | } |
5841 | break; | |
a0d0e21e | 5842 | case NDIGIT: |
f2ed9b32 | 5843 | if (utf8_target) { |
ffc61ed2 | 5844 | loceol = PL_regeol; |
1a4fad37 | 5845 | LOAD_UTF8_CHARCLASS_DIGIT(); |
1aa99e6b | 5846 | while (hardcount < max && scan < loceol && |
f2ed9b32 | 5847 | !swash_fetch(PL_utf8_digit, (U8*)scan, utf8_target)) { |
ffc61ed2 JH |
5848 | scan += UTF8SKIP(scan); |
5849 | hardcount++; | |
5850 | } | |
5851 | } else { | |
5852 | while (scan < loceol && !isDIGIT(*scan)) | |
5853 | scan++; | |
a0ed51b3 | 5854 | } |
e1d1eefb | 5855 | case LNBREAK: |
f2ed9b32 | 5856 | if (utf8_target) { |
e1d1eefb YO |
5857 | loceol = PL_regeol; |
5858 | while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) { | |
5859 | scan += c; | |
5860 | hardcount++; | |
5861 | } | |
5862 | } else { | |
5863 | /* | |
5864 | LNBREAK can match two latin chars, which is ok, | |
5865 | because we have a null terminated string, but we | |
5866 | have to use hardcount in this situation | |
5867 | */ | |
5868 | while (scan < loceol && (c=is_LNBREAK_latin1(scan))) { | |
5869 | scan+=c; | |
5870 | hardcount++; | |
5871 | } | |
5872 | } | |
5873 | break; | |
5874 | case HORIZWS: | |
f2ed9b32 | 5875 | if (utf8_target) { |
e1d1eefb YO |
5876 | loceol = PL_regeol; |
5877 | while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) { | |
5878 | scan += c; | |
5879 | hardcount++; | |
5880 | } | |
5881 | } else { | |
5882 | while (scan < loceol && is_HORIZWS_latin1(scan)) | |
5883 | scan++; | |
5884 | } | |
a0ed51b3 | 5885 | break; |
e1d1eefb | 5886 | case NHORIZWS: |
f2ed9b32 | 5887 | if (utf8_target) { |
e1d1eefb YO |
5888 | loceol = PL_regeol; |
5889 | while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) { | |
5890 | scan += UTF8SKIP(scan); | |
5891 | hardcount++; | |
5892 | } | |
5893 | } else { | |
5894 | while (scan < loceol && !is_HORIZWS_latin1(scan)) | |
5895 | scan++; | |
5896 | ||
5897 | } | |
5898 | break; | |
5899 | case VERTWS: | |
f2ed9b32 | 5900 | if (utf8_target) { |
e1d1eefb YO |
5901 | loceol = PL_regeol; |
5902 | while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) { | |
5903 | scan += c; | |
5904 | hardcount++; | |
5905 | } | |
5906 | } else { | |
5907 | while (scan < loceol && is_VERTWS_latin1(scan)) | |
5908 | scan++; | |
5909 | ||
5910 | } | |
5911 | break; | |
5912 | case NVERTWS: | |
f2ed9b32 | 5913 | if (utf8_target) { |
e1d1eefb YO |
5914 | loceol = PL_regeol; |
5915 | while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) { | |
5916 | scan += UTF8SKIP(scan); | |
5917 | hardcount++; | |
5918 | } | |
5919 | } else { | |
5920 | while (scan < loceol && !is_VERTWS_latin1(scan)) | |
5921 | scan++; | |
5922 | ||
5923 | } | |
5924 | break; | |
5925 | ||
a0d0e21e LW |
5926 | default: /* Called on something of 0 width. */ |
5927 | break; /* So match right here or not at all. */ | |
5928 | } | |
a687059c | 5929 | |
a0ed51b3 LW |
5930 | if (hardcount) |
5931 | c = hardcount; | |
5932 | else | |
5933 | c = scan - PL_reginput; | |
3280af22 | 5934 | PL_reginput = scan; |
a687059c | 5935 | |
a3621e74 | 5936 | DEBUG_r({ |
e68ec53f | 5937 | GET_RE_DEBUG_FLAGS_DECL; |
be8e71aa | 5938 | DEBUG_EXECUTE_r({ |
e68ec53f YO |
5939 | SV * const prop = sv_newmortal(); |
5940 | regprop(prog, prop, p); | |
5941 | PerlIO_printf(Perl_debug_log, | |
be8e71aa | 5942 | "%*s %s can match %"IVdf" times out of %"IVdf"...\n", |
e2e6a0f1 | 5943 | REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max); |
a3621e74 | 5944 | }); |
be8e71aa | 5945 | }); |
9041c2e3 | 5946 | |
a0d0e21e | 5947 | return(c); |
a687059c LW |
5948 | } |
5949 | ||
c277df42 | 5950 | |
be8e71aa | 5951 | #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) |
c277df42 | 5952 | /* |
ffc61ed2 JH |
5953 | - regclass_swash - prepare the utf8 swash |
5954 | */ | |
5955 | ||
5956 | SV * | |
32fc9b6a | 5957 | Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) |
ffc61ed2 | 5958 | { |
97aff369 | 5959 | dVAR; |
9e55ce06 JH |
5960 | SV *sw = NULL; |
5961 | SV *si = NULL; | |
5962 | SV *alt = NULL; | |
f8fc2ecf YO |
5963 | RXi_GET_DECL(prog,progi); |
5964 | const struct reg_data * const data = prog ? progi->data : NULL; | |
ffc61ed2 | 5965 | |
7918f24d NC |
5966 | PERL_ARGS_ASSERT_REGCLASS_SWASH; |
5967 | ||
4f639d21 | 5968 | if (data && data->count) { |
a3b680e6 | 5969 | const U32 n = ARG(node); |
ffc61ed2 | 5970 | |
4f639d21 | 5971 | if (data->what[n] == 's') { |
ad64d0ec NC |
5972 | SV * const rv = MUTABLE_SV(data->data[n]); |
5973 | AV * const av = MUTABLE_AV(SvRV(rv)); | |
2d03de9c | 5974 | SV **const ary = AvARRAY(av); |
9e55ce06 | 5975 | SV **a, **b; |
9041c2e3 | 5976 | |
711a919c | 5977 | /* See the end of regcomp.c:S_regclass() for |
9e55ce06 JH |
5978 | * documentation of these array elements. */ |
5979 | ||
b11f357e | 5980 | si = *ary; |
fe5bfecd JH |
5981 | a = SvROK(ary[1]) ? &ary[1] : NULL; |
5982 | b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : NULL; | |
b11f357e | 5983 | |
ffc61ed2 JH |
5984 | if (a) |
5985 | sw = *a; | |
5986 | else if (si && doinit) { | |
5987 | sw = swash_init("utf8", "", si, 1, 0); | |
5988 | (void)av_store(av, 1, sw); | |
5989 | } | |
9e55ce06 JH |
5990 | if (b) |
5991 | alt = *b; | |
ffc61ed2 JH |
5992 | } |
5993 | } | |
5994 | ||
9e55ce06 JH |
5995 | if (listsvp) |
5996 | *listsvp = si; | |
5997 | if (altsvp) | |
5998 | *altsvp = alt; | |
ffc61ed2 JH |
5999 | |
6000 | return sw; | |
6001 | } | |
76234dfb | 6002 | #endif |
ffc61ed2 JH |
6003 | |
6004 | /* | |
ba7b4546 | 6005 | - reginclass - determine if a character falls into a character class |
832705d4 JH |
6006 | |
6007 | The n is the ANYOF regnode, the p is the target string, lenp | |
6008 | is pointer to the maximum length of how far to go in the p | |
6009 | (if the lenp is zero, UTF8SKIP(p) is used), | |
f2ed9b32 | 6010 | utf8_target tells whether the target string is in UTF-8. |
832705d4 | 6011 | |
bbce6d69 | 6012 | */ |
6013 | ||
76e3520e | 6014 | STATIC bool |
f2ed9b32 | 6015 | S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const U8* p, STRLEN* lenp, register bool utf8_target) |
bbce6d69 | 6016 | { |
27da23d5 | 6017 | dVAR; |
a3b680e6 | 6018 | const char flags = ANYOF_FLAGS(n); |
bbce6d69 | 6019 | bool match = FALSE; |
cc07378b | 6020 | UV c = *p; |
ae9ddab8 | 6021 | STRLEN len = 0; |
9e55ce06 | 6022 | STRLEN plen; |
1aa99e6b | 6023 | |
7918f24d NC |
6024 | PERL_ARGS_ASSERT_REGINCLASS; |
6025 | ||
f2ed9b32 | 6026 | if (utf8_target && !UTF8_IS_INVARIANT(c)) { |
19f67299 | 6027 | c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len, |
6182169b KW |
6028 | (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) |
6029 | | UTF8_ALLOW_FFFF | UTF8_CHECK_ONLY); | |
6030 | /* see [perl #37836] for UTF8_ALLOW_ANYUV; [perl #38293] for | |
6031 | * UTF8_ALLOW_FFFF */ | |
e8a70c6f SP |
6032 | if (len == (STRLEN)-1) |
6033 | Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); | |
19f67299 | 6034 | } |
bbce6d69 | 6035 | |
0f0076b4 | 6036 | plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c)); |
f2ed9b32 | 6037 | if (utf8_target || (flags & ANYOF_UNICODE)) { |
9e55ce06 JH |
6038 | if (lenp) |
6039 | *lenp = 0; | |
f2ed9b32 | 6040 | if (utf8_target && !ANYOF_RUNTIME(n)) { |
ffc61ed2 JH |
6041 | if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c)) |
6042 | match = TRUE; | |
bbce6d69 | 6043 | } |
f2ed9b32 | 6044 | if (!match && utf8_target && (flags & ANYOF_UNICODE_ALL) && c >= 256) |
1aa99e6b | 6045 | match = TRUE; |
ffc61ed2 | 6046 | if (!match) { |
9e55ce06 | 6047 | AV *av; |
32fc9b6a | 6048 | SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av); |
ffc61ed2 JH |
6049 | |
6050 | if (sw) { | |
3f0c5693 | 6051 | U8 * utf8_p; |
f2ed9b32 | 6052 | if (utf8_target) { |
3f0c5693 KW |
6053 | utf8_p = (U8 *) p; |
6054 | } else { | |
6055 | STRLEN len = 1; | |
6056 | utf8_p = bytes_to_utf8(p, &len); | |
6057 | } | |
6058 | if (swash_fetch(sw, utf8_p, 1)) | |
ffc61ed2 JH |
6059 | match = TRUE; |
6060 | else if (flags & ANYOF_FOLD) { | |
9e55ce06 JH |
6061 | if (!match && lenp && av) { |
6062 | I32 i; | |
9e55ce06 | 6063 | for (i = 0; i <= av_len(av); i++) { |
890ce7af | 6064 | SV* const sv = *av_fetch(av, i, FALSE); |
9e55ce06 | 6065 | STRLEN len; |
890ce7af | 6066 | const char * const s = SvPV_const(sv, len); |
3f0c5693 | 6067 | if (len <= plen && memEQ(s, (char*)utf8_p, len)) { |
9e55ce06 JH |
6068 | *lenp = len; |
6069 | match = TRUE; | |
6070 | break; | |
6071 | } | |
6072 | } | |
6073 | } | |
6074 | if (!match) { | |
89ebb4a3 | 6075 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; |
4a623e43 | 6076 | |
3f0c5693 KW |
6077 | STRLEN tmplen; |
6078 | to_utf8_fold(utf8_p, tmpbuf, &tmplen); | |
6079 | if (swash_fetch(sw, tmpbuf, 1)) | |
9e55ce06 JH |
6080 | match = TRUE; |
6081 | } | |
ffc61ed2 | 6082 | } |
b3a04dd3 KW |
6083 | |
6084 | /* If we allocated a string above, free it */ | |
f2ed9b32 | 6085 | if (! utf8_target) Safefree(utf8_p); |
ffc61ed2 | 6086 | } |
bbce6d69 | 6087 | } |
9e55ce06 | 6088 | if (match && lenp && *lenp == 0) |
0f0076b4 | 6089 | *lenp = UNISKIP(NATIVE_TO_UNI(c)); |
bbce6d69 | 6090 | } |
1aa99e6b | 6091 | if (!match && c < 256) { |
ffc61ed2 JH |
6092 | if (ANYOF_BITMAP_TEST(n, c)) |
6093 | match = TRUE; | |
6094 | else if (flags & ANYOF_FOLD) { | |
eb160463 | 6095 | U8 f; |
a0ed51b3 | 6096 | |
ffc61ed2 JH |
6097 | if (flags & ANYOF_LOCALE) { |
6098 | PL_reg_flags |= RF_tainted; | |
6099 | f = PL_fold_locale[c]; | |
6100 | } | |
6101 | else | |
6102 | f = PL_fold[c]; | |
6103 | if (f != c && ANYOF_BITMAP_TEST(n, f)) | |
6104 | match = TRUE; | |
6105 | } | |
6106 | ||
6107 | if (!match && (flags & ANYOF_CLASS)) { | |
a0ed51b3 | 6108 | PL_reg_flags |= RF_tainted; |
ffc61ed2 JH |
6109 | if ( |
6110 | (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) || | |
6111 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) || | |
6112 | (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) || | |
6113 | (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) || | |
6114 | (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) || | |
6115 | (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) || | |
6116 | (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) || | |
6117 | (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) || | |
6118 | (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) || | |
6119 | (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) || | |
6120 | (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) || | |
6121 | (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) || | |
6122 | (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) || | |
6123 | (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) || | |
6124 | (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) || | |
6125 | (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) || | |
6126 | (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) || | |
6127 | (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) || | |
6128 | (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) || | |
6129 | (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) || | |
6130 | (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) || | |
6131 | (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) || | |
6132 | (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) || | |
6133 | (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) || | |
6134 | (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) || | |
6135 | (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) || | |
6136 | (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) || | |
6137 | (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) || | |
6138 | (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) || | |
6139 | (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c)) | |
6140 | ) /* How's that for a conditional? */ | |
6141 | { | |
6142 | match = TRUE; | |
6143 | } | |
a0ed51b3 | 6144 | } |
a0ed51b3 LW |
6145 | } |
6146 | ||
a0ed51b3 LW |
6147 | return (flags & ANYOF_INVERT) ? !match : match; |
6148 | } | |
161b471a | 6149 | |
dfe13c55 | 6150 | STATIC U8 * |
0ce71af7 | 6151 | S_reghop3(U8 *s, I32 off, const U8* lim) |
9041c2e3 | 6152 | { |
97aff369 | 6153 | dVAR; |
7918f24d NC |
6154 | |
6155 | PERL_ARGS_ASSERT_REGHOP3; | |
6156 | ||
a0ed51b3 | 6157 | if (off >= 0) { |
1aa99e6b | 6158 | while (off-- && s < lim) { |
ffc61ed2 | 6159 | /* XXX could check well-formedness here */ |
a0ed51b3 | 6160 | s += UTF8SKIP(s); |
ffc61ed2 | 6161 | } |
a0ed51b3 LW |
6162 | } |
6163 | else { | |
1de06328 YO |
6164 | while (off++ && s > lim) { |
6165 | s--; | |
6166 | if (UTF8_IS_CONTINUED(*s)) { | |
6167 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
6168 | s--; | |
a0ed51b3 | 6169 | } |
1de06328 | 6170 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
6171 | } |
6172 | } | |
6173 | return s; | |
6174 | } | |
161b471a | 6175 | |
f9f4320a YO |
6176 | #ifdef XXX_dmq |
6177 | /* there are a bunch of places where we use two reghop3's that should | |
6178 | be replaced with this routine. but since thats not done yet | |
6179 | we ifdef it out - dmq | |
6180 | */ | |
dfe13c55 | 6181 | STATIC U8 * |
1de06328 YO |
6182 | S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim) |
6183 | { | |
6184 | dVAR; | |
7918f24d NC |
6185 | |
6186 | PERL_ARGS_ASSERT_REGHOP4; | |
6187 | ||
1de06328 YO |
6188 | if (off >= 0) { |
6189 | while (off-- && s < rlim) { | |
6190 | /* XXX could check well-formedness here */ | |
6191 | s += UTF8SKIP(s); | |
6192 | } | |
6193 | } | |
6194 | else { | |
6195 | while (off++ && s > llim) { | |
6196 | s--; | |
6197 | if (UTF8_IS_CONTINUED(*s)) { | |
6198 | while (s > llim && UTF8_IS_CONTINUATION(*s)) | |
6199 | s--; | |
6200 | } | |
6201 | /* XXX could check well-formedness here */ | |
6202 | } | |
6203 | } | |
6204 | return s; | |
6205 | } | |
f9f4320a | 6206 | #endif |
1de06328 YO |
6207 | |
6208 | STATIC U8 * | |
0ce71af7 | 6209 | S_reghopmaybe3(U8* s, I32 off, const U8* lim) |
a0ed51b3 | 6210 | { |
97aff369 | 6211 | dVAR; |
7918f24d NC |
6212 | |
6213 | PERL_ARGS_ASSERT_REGHOPMAYBE3; | |
6214 | ||
a0ed51b3 | 6215 | if (off >= 0) { |
1aa99e6b | 6216 | while (off-- && s < lim) { |
ffc61ed2 | 6217 | /* XXX could check well-formedness here */ |
a0ed51b3 | 6218 | s += UTF8SKIP(s); |
ffc61ed2 | 6219 | } |
a0ed51b3 | 6220 | if (off >= 0) |
3dab1dad | 6221 | return NULL; |
a0ed51b3 LW |
6222 | } |
6223 | else { | |
1de06328 YO |
6224 | while (off++ && s > lim) { |
6225 | s--; | |
6226 | if (UTF8_IS_CONTINUED(*s)) { | |
6227 | while (s > lim && UTF8_IS_CONTINUATION(*s)) | |
6228 | s--; | |
a0ed51b3 | 6229 | } |
1de06328 | 6230 | /* XXX could check well-formedness here */ |
a0ed51b3 LW |
6231 | } |
6232 | if (off <= 0) | |
3dab1dad | 6233 | return NULL; |
a0ed51b3 LW |
6234 | } |
6235 | return s; | |
6236 | } | |
51371543 | 6237 | |
51371543 | 6238 | static void |
acfe0abc | 6239 | restore_pos(pTHX_ void *arg) |
51371543 | 6240 | { |
97aff369 | 6241 | dVAR; |
097eb12c | 6242 | regexp * const rex = (regexp *)arg; |
51371543 GS |
6243 | if (PL_reg_eval_set) { |
6244 | if (PL_reg_oldsaved) { | |
4f639d21 DM |
6245 | rex->subbeg = PL_reg_oldsaved; |
6246 | rex->sublen = PL_reg_oldsavedlen; | |
f8c7b90f | 6247 | #ifdef PERL_OLD_COPY_ON_WRITE |
4f639d21 | 6248 | rex->saved_copy = PL_nrs; |
ed252734 | 6249 | #endif |
07bc277f | 6250 | RXp_MATCH_COPIED_on(rex); |
51371543 GS |
6251 | } |
6252 | PL_reg_magic->mg_len = PL_reg_oldpos; | |
6253 | PL_reg_eval_set = 0; | |
6254 | PL_curpm = PL_reg_oldcurpm; | |
6255 | } | |
6256 | } | |
33b8afdf JH |
6257 | |
6258 | STATIC void | |
6259 | S_to_utf8_substr(pTHX_ register regexp *prog) | |
6260 | { | |
a1cac82e | 6261 | int i = 1; |
7918f24d NC |
6262 | |
6263 | PERL_ARGS_ASSERT_TO_UTF8_SUBSTR; | |
6264 | ||
a1cac82e NC |
6265 | do { |
6266 | if (prog->substrs->data[i].substr | |
6267 | && !prog->substrs->data[i].utf8_substr) { | |
6268 | SV* const sv = newSVsv(prog->substrs->data[i].substr); | |
6269 | prog->substrs->data[i].utf8_substr = sv; | |
6270 | sv_utf8_upgrade(sv); | |
610460f9 NC |
6271 | if (SvVALID(prog->substrs->data[i].substr)) { |
6272 | const U8 flags = BmFLAGS(prog->substrs->data[i].substr); | |
6273 | if (flags & FBMcf_TAIL) { | |
6274 | /* Trim the trailing \n that fbm_compile added last | |
6275 | time. */ | |
6276 | SvCUR_set(sv, SvCUR(sv) - 1); | |
6277 | /* Whilst this makes the SV technically "invalid" (as its | |
6278 | buffer is no longer followed by "\0") when fbm_compile() | |
6279 | adds the "\n" back, a "\0" is restored. */ | |
6280 | } | |
6281 | fbm_compile(sv, flags); | |
6282 | } | |
a1cac82e NC |
6283 | if (prog->substrs->data[i].substr == prog->check_substr) |
6284 | prog->check_utf8 = sv; | |
6285 | } | |
6286 | } while (i--); | |
33b8afdf JH |
6287 | } |
6288 | ||
6289 | STATIC void | |
6290 | S_to_byte_substr(pTHX_ register regexp *prog) | |
6291 | { | |
97aff369 | 6292 | dVAR; |
a1cac82e | 6293 | int i = 1; |
7918f24d NC |
6294 | |
6295 | PERL_ARGS_ASSERT_TO_BYTE_SUBSTR; | |
6296 | ||
a1cac82e NC |
6297 | do { |
6298 | if (prog->substrs->data[i].utf8_substr | |
6299 | && !prog->substrs->data[i].substr) { | |
6300 | SV* sv = newSVsv(prog->substrs->data[i].utf8_substr); | |
6301 | if (sv_utf8_downgrade(sv, TRUE)) { | |
610460f9 NC |
6302 | if (SvVALID(prog->substrs->data[i].utf8_substr)) { |
6303 | const U8 flags | |
6304 | = BmFLAGS(prog->substrs->data[i].utf8_substr); | |
6305 | if (flags & FBMcf_TAIL) { | |
6306 | /* Trim the trailing \n that fbm_compile added last | |
6307 | time. */ | |
6308 | SvCUR_set(sv, SvCUR(sv) - 1); | |
6309 | } | |
6310 | fbm_compile(sv, flags); | |
6311 | } | |
a1cac82e NC |
6312 | } else { |
6313 | SvREFCNT_dec(sv); | |
6314 | sv = &PL_sv_undef; | |
6315 | } | |
6316 | prog->substrs->data[i].substr = sv; | |
6317 | if (prog->substrs->data[i].utf8_substr == prog->check_utf8) | |
6318 | prog->check_substr = sv; | |
33b8afdf | 6319 | } |
a1cac82e | 6320 | } while (i--); |
33b8afdf | 6321 | } |
66610fdd RGS |
6322 | |
6323 | /* | |
6324 | * Local variables: | |
6325 | * c-indentation-style: bsd | |
6326 | * c-basic-offset: 4 | |
6327 | * indent-tabs-mode: t | |
6328 | * End: | |
6329 | * | |
37442d52 RGS |
6330 | * ex: set ts=8 sts=4 sw=4 noet: |
6331 | */ |